\
N
.1
i
A Generative Introduction
Second edition
Andrew
to
. Blackwell
fj Publishing
r n i
•
Syntax
Introducing Linguistics
This outstanding series is an indispensable resource for
students and teachers - a concise and engaging introduction to
the central subjects of contemporary linguistics. Presupposing
no prior knowledge on the part of the reader, each volume sets
out the fundamental skills and knowledge of the field,
and so provides the ideal educational platform for further study
in linguistics.
1 Andrew Spencer Phonology
2 John Saeed Semantics, Second Edition
3 Barbara Johnstone Discourse Analysis
4 Andrew Carnie Syntax, Second Edition
Syntax
A Generative Introduction
Second Edition
Andrew Carnie
4£k Blackwell
*Cr Publishing
© 2007 by Andrew Carnie
BLACKWELL PUBLISHING
350 Main Street, Maiden, MA 02148-5020, USA
9600 Garsington Road, Oxford OX4 2DQ, UK
550 Swanston Street, Carlton, Victoria 3053, Australia
The right of Andrew Carnie to be identified as the Author of this Work has been
asserted in accordance with the UKCopyright, 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 UKCopyright,
Designs, and Patents Act 1988,without the prior permission of the publisher.
First edition published 2002by BlackwellPublishers Ltd
Second edition published 2007by Blackwell Publishing Ltd
1 2007
Library ofCongress Cataloging-in-Publicatiott Data
Carnie, Andrew, 1969Syntax
: a generative introduction / Andrew Carnie. — 2nd ed.
p. cm.
Includes bibliographical references and index.
ISBN-13:978-1-4051-3384-5 (alk. paper)
ISBN-10:1-4051-3384-8 (alk. paper)
1. Grammar, Comparative and general—Syntax. 2. Generative
grammar. I. Title.
P295.C37 2006
415—dc22
2006008835
A catalogue record for this title is available from the British Library.
Layout by Andrew Carnie. Typeset in 10pt Palatino using ReadSmart® from
Language Technologies Inc.
Printed and bound in Singapore by COS Printers Pte Ltd
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 coverboard used have met acceptableenvironmental
accreditation standards.
For further information on Blackwell Publishing, visit our website:
www.blackwellpublishing.com
Dedicated with love to my parents, Robert and Jean
and in memory of my teacher and mentor, Ken Hale
The text formatting in this book has been adapted and modified by
ReadSmart®, an algorithm which imposes almost undetectable changes in
presentation. This formatting isbased on simple psycholinguisticprinciples,
and has been shown in studies to enhance readers' reading speed,
comprehension and enjoyment and to make the text more persuasive. The
ReadSmart® treatment of the text was provided by Language Technologies
Inc., which offers this service for all forms of publication.
www.lngtek.com
Contents
Preface and Acknowledgments xv
Part 1 Preliminaries 1
1 Generative Grammar 3
0. Preliminaries 3
1. Syntax as a Cognitive Science 4
2. Modeling Syntax 5
3. Syntax as Science - the Scientific Method 6
3.1 An Example of theScientific Method as Applied toSyntax 9
3.2 Sources of Data 12
4. Where do the Rules Come From? 14
4.1 Learning vs. Acquisition 14
4.2 Innateness: Language as an Instinct 15
4.3 The Logical Problem ofLanguage Acquisition 15
4.4 OtherArgumentsfor UG 19
4.5 Explaining Language Variation 22
5. Choosing among Theories about Syntax 24
6. The Scientific Method and the Structure of this Textbook 24
7. Summary 26
Ideas Introduced in this Chapter 26
Further Reading 28
General Problem Sets 29
Challenge Problem Sets 31
viii Contents
2 Parts of Speech 37
0. Words and Why They Matter to Syntax 37
1. Determining Part of Speech 38
1.1 The Problem of Traditional Definitions 38
1.2 Distributional Criteria 40
2. The Major Parts of Speech: N, V,Adj, and Adv 41
2.2 Nouns 42
2.2 Verbs 43
2.3 Adjectives 43
2.4 Adverbs 44
3. Open vs. Closed; Lexical vs. Functional 45
3.1 Open vs. Closed PartsofSpeech 45
3.2 Lexical vs. Functional 45
3.3 Some Functional (Closed) Categories of English 46
3.4 Summary 48
4. Subcategories and Features 48
4.1 Subcategories ofNouns 49
4.2 Subcategories of Verbs 50
5. Summary 54
Ideas Introduced in this Chapter 54
Further Reading 56
General Problem Sets 56
Challenge Problem Sets 60
3 Constituency, Trees, and Rules 63
0. Introduction 63
1. Rules and Trees 66
1.1 Noun Phrases (NPs) 66
1.2 Adjective Phrases (AdjPs) andAdverb Phrases (AdvPs) 68
2.3 Prepositional Phrases (PPs) 71
2.4 Verb Phrases (VPs) 72
1.5 Clauses 74
2.6 Summary 79
2. How to Draw a Tree 81
2.2 Bottom-up Trees 81
2.2 The Top-down Method ofDraiuing Trees 84
2.3 Bracketed Diagrams 86
3. Modification and Ambiguity 87
4. Constituency Tests 88
5. Summary and Conclusion 91
Contents ix
Appendix: How to Do Foreign Language PSR Problems 92
Al. Doing problems with word-by-word glosses 92
A2. Doing problems without word-by-zvord glosses 94
Ideas Introduced in this Chapter 95
Further Reading 96
General Problem Sets 96
Challenge Problem Sets 101
Structural Relations 103
0. Introduction 103
1. The Parts of a Tree 104
2. Domination 106
2.2 Domination 106
2.2 Exhaustive Domination 107
2.3 Immediate Domination 109
3. Precedence 110
4. C-command 113
5. Grammatical Relations 118
6. Summary and Conclusions 121
Ideas Introduced in this Chapter 121
Further Reading 124
General Problem Sets 124
Challenge Problem Sets 132
Binding Theory 135
0. Introduction 135
1. The Notions Comdex and Antecedent 137
2. Binding 138
3. Locality Conditions on the Bindingof Anaphors 141
4. The Distribution of Pronouns 142
5. The Distribution of R-expressions 143
6. Conclusion 144
Ideas Introduced in this Chapter 145
Further Reading 146
General Problem Sets 146
Challenge Problem Sets 147
x Contents
Part 2 The Base 151
X-bar Theory 153
0. Introduction 153
1. Bar-level Projections 155
2.2 V-bar 156
1.2 Adj-bar andAdv-bar 158
1.3 P-bar 160
2. Generalizing the Rules: The X-bar Schema 160
3. Complements, Adjuncts, and Specifiers 163
3.2 Complements and Adjuncts in NPs 165
3.2 Complements and Adjuncts in VPs, AdjPs, AdvPs, andPPs 170
3.3 The NotionSpecifier 172
4. Some Definitional Housekeeping 173
5. Parameters of Word Order 174
6. Drawing Trees in X-bar Notation 177
6.2 Important Considerations in Tree Drawing 177
6.2 A Sample Tree 180
7. X-barTheory: A Summary 187
Ideas Introduced in this Chapter 188
Further Reading 189
General Problem Sets 190
Challenge Problem Sets 194
Extending X-bar Theoryto FunctionalCategories 197
0. Introduction 197
1. Determiner Phrases (DPs) 198
2. A Descriptive Tangent into Clause Types 201
3. Complementizer Phrases (CPs) 206
4. Tense Phrases (TPs) 210
5. CP,TP,DPTree 212
Ideas Introduced in this Chapter 213
Further Reading 214
General Problem Sets 214
Challenge Problem Sets 216
Constraining X-bar Theory: The Lexicon 219
0. Introduction 219
1. Some Basic Terminology 220
2. Thematic Relations and Theta Roles 221
Contents xi
3. The Lexicon 226
4. Expletives and the Extended Projection Principle 228
5. Summary 230
Ideas Introduced in this Chapter 230
Further Reading 232
General Problem Sets 232
Challenge Problem Sets 236
Part 3 Movement 241
9 Head-to-Head Movement 243
0. Introduction 243
1. Verb Movement (V -> T) 246
2.2 French 246
2.2 Irish 254
2. T Movement (T -> C) 260
3. Do-support 263
4. MultipleAuxiliariesand Affix-hopping in English 265
4.2 Multiple Auxiliaries 265
4.2 Affix-hopping 268
5. Summary 271
Appendix:Testsfor Determiningif a Languagehas V —> T
or Affix Lowering 272
Ideas Introduced in this Chapter 272
Further Reading 273
General Problem Sets 274
Challenge Problem Sets 279
10 DP Movement 285
0. Introduction 285
1. A Puzzle for the Theory of Theta Roles 285
2. Passives 291
3. Case 295
4. Raising: Reprise 300
5. Passives: Reprise 302
6. Closing Up a Loose End 304
7. Conclusion 306
Ideas Introduced in this Chapter 306
Further Reading 307
General Problem Sets 307
Challenge Problem Sets 313
xii Contents
11 W7z-movement 317
0. Introduction 317
1. Movement in W/i-questions 318
2. Islands 334
3. The Minimal Link Condition 337
3.2 Wh-islands and theMinimal LinkCondition 337
3.2 The MLC in DP Movement and Head Movement 339
4. Echo Questions(Wfr-in-situ) in English 342
5. Conclusion 343
Ideas Introduced in this Chapter 344
Further Reading 345
General Problem Sets 346
Challenge Problem Sets 348
12 A Unified Theory of Movement 353
0. Introduction 353
1. Move 355
2. Explaining Cross-linguistic Differences 359
3. Scope, Covert Movement, and the MLC 364
3.2 MLC Effects in Wh-in-situ Languages 364
3.2 English Quantifiers andScope 365
4. Conclusion 368
Ideas Introduced in this Chapter 368
Further Reading 369
General Problem Sets 369
Challenge Problem Sets 370
Part 4 Advanced Topics 373
13 Expanded VPs 375
0. Introduction 375
1. The Problem of Ditransitive Verbs 375
2. Light Verbs 377
3. Object Shift 380
4. Ditransitives: Reprise 385
Ideas Introduced in this Chapter 389
Further Reading 389
General Problem Sets 390
Challenge Problem Sets 392
Contents xiii
14 Raising, Control, and Empty Categories 395
0. Introduction 395
1. Raising vs. Control 396
2.2 Two Kinds ofThetaGridsfor MainPredicates 396
2.2 Distinguishing Raisingfrom Control 403
2.3 What is PRO? 405
2. Two Kinds of Raising, Two Kinds of Control 405
2.2 Two Kinds ofRaising 405
2.2 Two Kinds ofControl 409
2.3 Summary ofPredicate Types 410
3. Control Theory 412
4. Another Kind of Null Subject: "Little" pro 416
5. Summary 417
Ideas Introduced in this Chapter 417
Further Reading 419
General Problem Sets 419
Challenge Problem Sets 422
15 Advanced Topics in Binding Theory 423
0. A Quick Review of Chapter 5 Binding Theory 423
1. Levels of Representation 424
2. The Definition of Binding Domain 426
2.2 A Miscellany ofDomain Violations 426
2.2 Anaphors 427
2.3 Pronouns 430
Ideas Introduced in this Chapter 432
Further Reading 433
General Problem Sets 433
Challenge Problem Sets 434
Part 5 Alternatives 435
16 Lexical-Functional Grammar 437
0. Alternative Theories 437
1. C-structure 438
2. Functions 439
3. The Lexicon 440
4. F-structure 441
4.2 Wliy F-structures? 444
5. Assorted Phenomena 445
xiv Contents
5.2 Head Mobility 446
5.2 Passives 447
5.3 Raisingand Control 447
5.4 Wh-movement: Long Distance Dependencies 450
6. Conclusion 450
Ideas Introduced in this Chapter 451
Further Reading 452
General Problem Set 453
Challenge Problem Sets 453
17 Head-Driven Phrase Structure Grammar 455
0. Introduction 455
1. Features 456
2. The Lexicon 459
3. Rules, Features, and Trees 461
4. Binding 465
5. Long Distance Dependencies 465
Ideas Introduced in this Chapter 467
Further Reading 470
General Problem Set 470
Challenge Problem Sets 471
Conclusions and Directions for Further Study 473
References 475
Index 482
A Glossary and other supplementary materials can be found on the book's
website at www.blackwellpublishing.com
Preface and
Acknowledgments
Almost every preface to every syntax textbook out there starts out by telling
the reader how different this book is from every other syntax textbook. On one hand,
this is often the truth: each author shows their own particular spin or emphasis. This
iscertainly true ofthis textbook. Forexample, you'llbehard pressed to find another
textbook on Principles and Parameters syntax that uses as many Irish examples
as this one does. Nor will you find another P&P textbook with a supplementary
discussion of alternative theoretical approaches like LFG or HPSG. On the other
hand, let's face facts. The basic material to be covered in an introductory textbook
doesn't really vary much. One linguist may prefer a little more on binding theory,
and a little less on control, etc. In this text, I've attempted to provide a relatively
balanced presentation of most of the major issues and I've tried to do this in
a student-friendly way. I've occasionally abstracted away from some of the thornier
controversies, where I felt theyweren't crucial toa student understanding the basics.
This may, to a certain extent, make the professional syntactician feel that I've cut
corners or laid out too rosy a picture. Idid this on purpose, however, to give students
a chance to absorb the fundamentals before challenging the issues. This is a textbook,
not a scholarly tome, so its aim is to reach as many students as possible. The style
is deliberately low key and friendly. This doesn't mean I don't want the students
to challenge the materialI've presented here. Throughout the book, you'll find gray
"textboxes" that contain issues for further discussion, or interesting tidbits. Many of
the problem sets alsoinvite thestudent tochallenge theblackand white presentation
I'vegiveninthetext. Iencourage instructors toassign these, and students todo them,
as they form an important part of the textbook. Instructors may note that if a favorite
topic is not dealt with in the body of the text, a problem set may very well treat
the question.
A quick word on the level of this textbook: This book is intended as an
introduction to syntactictheory. It takes the student through most of the major issues
in Principles and Parameters, from tree drawing to constraints on movement.
While this book is written as an introduction, some students have reported it to be
xvi Preface and Acknowledgments
challenging. I use this text in my upper division undergraduate introduction
to syntax with success, but I can certainly see it being used in more advanced classes.
I hope instructors will flesh out the book, and walk their students through some
of the thornier issues.
This textbook has grown out of my lecture notes for my own classes. Needless
to say, the form and shape of these notes have been influenced in terms of choice
of material and presentation by the textbooks my own students have used. While
the book you are reading is entirely my fault, it does owe a particular intellectual
debt to the following three textbooks, which I have used in teaching at various times:
Cowper, Elizabeth (1992) A Concise Introduction to Syntactic Theory: The Government
and Binding Approach. Chicago: Chicago University Press.
Haegeman, Liliane (1994) Introduction toGovernment and Binding Theory (2nd edition).
Oxford: Blackwell.
Radford, Andrew (1988) Transformational Grammar: A First Course. Cambridge:
Cambridge University Press.
I'd like to thank the authors of these books for breaking ground in presenting
a complicated and integrated theory to the beginner. Writing this book has given me
new appreciation for the difficulty of this task and their presentation of the material
has undoubtedly influenced mine.
Sadly, during the final stages of putting the first edition of this text together,
my dissertation director, teacher, mentor, and academic hero, Ken Hale passed
away after a long illness. Ken always pushed the idea that theoretical syntax
is best informed by cross-linguistic research; while at the same time, the accurate
documentation of languages requires a sophisticatedunderstanding of grammatical
theory. These were important lessons that I learned from Ken and I hopestudents
willglean thesignificance ofboth by reading this text. WhileI was writing thisbook
(and much other work) Ken gave me many comments and his unfettered support.
He was a great man and I will miss him terribly.
This, the second edition of this book, is considerably different from the first
edition. Here is a brief list of the major differences between the two editions. This list
is not comprehensive, many more minor differences can be found.
• The exercise sections of the chapters are now organized differently and are
greatly expanded. Exercises are presented in the order that the material appears
in the chapter. Ihave attempted to categorize each exerciseforlevel and type.
• There are two types of problem sets: General and Challenge. These two types
roughly correspond to the exercises that I assign to my regular students
and my honors students respectively. Challenge Problem Sets often challenge
the straightforward presentation of the material in the main body of the text.
• The former chapter 2 on structure and parts of speech has been split into two
chapters. The new chapter 2 contains new information on subcategorization
Preface and Acknowledgments xvii
that some instructors requested to better inform students about the role of part
of speech in phrase structure processes. Also Adjectives are now distinguished
from Adverbs.
• The phrase structure rules in the new chapter 3 have been completely revised.
In particular, I'm using non-X-bar versions of TP and CP here, and have added
embedded clauses to all the relevant rules.
• The definitions of precedence, exhaustive domination and c-command have all
been significantly revised in the chapter on structural relations. A limited
version of government is given for those instructors who wish to teach it
to their students.
• The chapters on X-bar theory have many more trees and examples.
• DPs are used consistently from chapter 7 forward.
• I have added categories to the theta grids in the chapter on the Lexicon in order
to tie them to the subcategories introduced in chapter 2.
• A new section on stacked VPs and affix-hopping has been added to the chapter
on head movement
• VP-internal subjects are used consistently from chapter 9 forward
• The treatment of passives in chapter 10is completelydifferent from the previous
edition. I have moved towards a Baker, Johnson and Roberts style approach
where the -en morphology is directly assigned the internal theta role
and accusative case by the verb in the syntax rather than in the lexicon.
• The treatment of locality conditions in the chapter on zo/i-movement is entirely
new. I've dropped subjacency in favor ofan MLC based approach. The chapter
now includes an inventory of the majorisland types; but theoretical coverage is
only given to ^//-islands. (Although the chapter also contains a brief discussion
of the Head-Movement Constraint and Super-raising in the context of the MLC).
• Chapter 12 now containsa more accurate discussion of zvh-in situ and develops
the ideas of feature checking, covert movement, and SPELLOUT.
• There is a brand new chapter on split VPs in a brand new section on "advanced
topics," including sections on object shift, ditransitives, a Lasnik style analysis
of Pseudogapping and a Hornstein style analysis of ACD.
• Thechapteron Raising and Control has beenmoved to the new part of the book
on advanced topics, and uses a split VP (vP-AgrOP-VP) structure
to avoid ternary branching.
• Thereis a new chapter on advanced topics in binding theory. This looks at issues
on level of representation, chains and the copytheory of movement. It also takes
a relativized view ofbinding domain consistent with Chomsky (1986).
• I've taken some of the more controversial "comparing theories" language out of
the chapters on LFG and HPSG.
I hope that instructors and students will find these revisions helpful.
I have attempted where possible to take into account all the many comments
and suggestions I received from people using the first edition. Although of course
in order to maintain consistency, I was unable to do them all.
xviii Preface and Acknowledgments
Acknowledgements:
I'd like to thank the many people who taught me syntax through the years:
Barb Brunson, Noam Chomsky, Elizabeth Cowper, Ken Hale, Alec Marantz,
Diane Massam, Jim McCloskey, Shigeru Miyagawa, and David Pesetsky. A number
of people have read through this book or the previous edition and have given me
extensive and helpful comments: WilliamAlexander, Ash Asudeh, Brett Baker, Andy
Barss, Mark Baltin, Luis Barragan, Emily Bender, Abbas Benmamoun, Joan Bresnan,
Dirk Bury, RoyChan, Danny Chen, Deborah Chen Pichler, Barbara Citko, Peter Cole,
Lorenzo Demery, Sheila Dooley, Yehuda Falk, Leslie Ford, Alexandra Galani,
Jila Ghomeshi, Erin Good, Paul Hagstrom, Ken Hale, Heidi Harley, Josh Harrison,
Rachel Hayes, Bernhard Heigl, One-Soon Her, Stephan Hurtubise, Alana Johns, Mark
Johnson, Simin Karimi, Andreas Kathol, Peter Lazar, Anne Lobeck, Leila Lomashivili,
Sarah Longstaff, Ahmad Reza Lotfi, Ricardo Mairal, Joan Maling, Jack Martin, Diane
Massam, Nathan McWhorter, Dave Medeiros, Martha McGinnis, Mirjana MiskovicLukovic,
Jon Nissenbaum, Peter Norquest, Kazutoshi Ohno, Heidi Orcutt, Hyeson
Park, David Pesetsky, Colin Phillips, Carl Pollard, Janet Randall, Norvin Richards,
Frank Richter, Betsy Ritter, Ed Rubin, Ivan Sag, Theresa Satterfield, Leslie Saxon,
Leah Shocket, Dan Siddiqi, Nick Sobin, Peggy Speas, Tania Strahan, Joshua Strauss,
Lisa deMena Travis, Robert Van Valin, Enwei Wang, Dainon Woudstra, Susi
Wurmbrand, Kim Youngroung, and several anonymous Blackwell reviewers. I'm
absolutely convinced I've left someone off this large list, if it's you many apologies,
I really did appreciate the help you gave me. The students in my Introduction to
Syntax classesin Michiganin 1997, and in Arizona in 1998-2006 have used allor parts
of this textbook as their reading. Thanks to them for their patience and suggestions.
A special thanks to Dave Medeiros who has been my TA during the writing
of this edition, his contributions and suggestions have been invaluable. A number
of problem sets in this book were graciously contributed by Sheila Dooley, Jila
Ghomeshi, Erin Good, Heidi Harley, Simin Karimi, Chris Kennedy, Amy LaCross,
Betsy Ritter and Leslie Saxon. Tom Bever, Chris Nicholas, and Roeland Hancock
at Language Technologies, Inc. have been incredibly helpful and responsive in their
implementation of the ReadSmart technology on this text. Chris Nicholas is to be
particularly thanked for his last-minute late-night help in formatting. I also owe a
great debt to all my colleagues here at the University of Arizona for their help and
support. In particular, DianaArchangeli, Andy Barss, SheilaDooley, Sandiway Fong,
Mike Hammond, Heidi Harley, Eloise Jelinek, Simin Karimi, Diane Ohala, Adam
Ussishkin, and Andy Wedel deserve special mention. Ada Brunstein, Sarah Coleman,
Lisa Eaton, Simon Eckley, Tami Kaplan, Becky Kennison, Anna Oxbury, Rhonda
Pearce, Beth Remmes, and Steve Smith of Blackwell all deserve many thanks for
encouraging me to write this and the previous edition up, and then smoothing its
way towards production. My family (Jean, Bob, Morag, Fiona, Pangur and Calvin)
were all incredible in their support and love. Go raibh maith agaibh!
The clip art in chapters 1 and 6 is taken from the Microsoft Clip Art Gallery and is used
withpermission. Thisusagedoes notimply Microsoft's endorsement ofthisbook.
+j
03
Q.
Preliminaries
>w#ww
Generative Grammar
0. Preliminaries
Although we use it every day, and although we all have strong opinions
about its proper form and appropriate use, we rarely stop to think about
the wonder of language. So-called language "experts" like William Safire tell
us about the misuse of hopefully or lecture us about the origins of the word
boondoggle, but surprisingly, they never get at the true wonder of language:
how it actually works. Think about it for a minute; you are reading
this and understanding it but you have no conscious knowledge of how you
are doing it. The study of this mystery is the science of linguistics. This book
is about one aspect of how language works - how sentences are structured:
syntax.
Language is a psychological or cognitive property of humans. That is,
there is some set of neurons in my head firing madly away that allows me to
sit here and produce this set of letters, and there is some other set of neurons
in your head firing away that allows you to translate these squiggles
into coherent ideas and thoughts. There are several subsystems at work here.
If you were listening to me speak, I would be producing sound waves with
my vocal cords and articulating particular speech sounds with my tongue,
lips, and vocal cords. On the other end of things you'd be hearing those
sound waves and translating them into speech sounds using your auditory
apparatus. The study of the acoustics and articulation of speech
is called phonetics. Once you've translated the waves of sound into mental
representations of speech sounds, you analyze them into syllables and
4 Preliminaries
pattern them appropriately. Forexample, speakers of English know that the
made-up word bluve isa possible word of English, but the word bnuck is not.
This is part of the science called phonology. Then you take these groups
of sounds and organize them into meaningful units (called morphemes)
and words. For example, the word dancer is made up of two meaningful bits:
dance and the suffix -er. The study of this level of Language is
called morphology. Next you organize the words into phrases and sentences.
Syntax is the cover term for studies of this level of Language. Finally, you
take the sentences and phrases you hear and translate them into thoughts
and ideas. This last step is what we refer to as the semantic level
of Language.
Syntax, then, studies the level of Language that lies between words and
the meaning of utterances: sentences. It is the level that mediates between
sounds that someone produces (organized into words) and what
they intended to say.
Perhaps one of the truly amazing aspects of the study of Language is not
the origins of the word demerit, or how to properly punctuate a quote inside
parentheses, or how kids have, like, destroyed the English language, eh?
Instead it's the question of how we subconsciously get from sounds
to meaning. This is the study of syntax.
Language vs. language
When I utter the term language, most people immediately think of some
particular language such as English, French, or KiSwahili. But this is not
the way linguists use the term; when linguists talk about Language (or ilanguage),
they are generally talking about the ability of humans to speak
any (particular) language. Some people (most notably Noam Chomsky)
also call this the Human Language Capacity. Language (written with a
capital L) is the part of the mind or brain that allows you to speak,
whereas language (with a lower case 1) (also known as e-language) is an
instantiation of this ability (like French or English). In this book we'll be
using language as our primary data, but we'll be trying to come up with a
model of Language.
1. Syntax as a Cognitive Science
Cognitive science is a cover term for a group of disciplines that all aim for
the same goal: describing and explaining human beings' ability to think (or
more particularly, to think about abstract notions like subatomic particles,
Chapter 1: Generative Grammar 5
the possibility of life on other planets or even how many angels can fit on
the head of a pin, etc.). One thing that distinguishes us from other animals,
even relatively smart ones like chimps and elephants, is our ability to use
productive, combinatory Language. Language plays an important role in
how we think about abstract notions, or, at the very least, Language appears
to be structured in such a way that it allows us to express abstractnotions.'
The discipline of linguistics, along with psychology, philosophy, and
computer science, thus forms an important subdiscipline within cognitive
science. Sentences are how we get at expressing abstract thought processes,
so the study of syntax is an important foundation stone for understanding
how we communicate and interact with each other as humans.
2. Modeling Syntax
The dominant theory of syntax is due to Noam Chomsky and his colleagues,
starting in the mid 1950s and continuing to this day. This theory, which
has had many different names through its development (Transformational
Grammar (TG), Transformational Generative Grammar, Standard Theory,
Extended Standard Theory, Government and Binding Theory (GB),
Principles and Parameters approach (P&P) and Minimalism (MP)), is often
given theblanket name Generative Grammar. A number of alternate theories
of syntax have also branched off of this research program; these include
Lexical-Functional Grammar (LFG) and Head-Driven Phrase Structure
Grammar (HPSG). These are also considered part of generative grammar;
but we won't cover them extensively in this book, except in chapters 16 and
17. The particular version of generative grammar that we will mostly look
at here is roughly the Principles andParameters approach, although we will
occasional stray from this into the more recent version called Minimalism.
The underlying thesis of generative grammar is that sentences
are generated by a subconscious set of procedures (like computer programs).
These procedures are part of our minds (or of our cognitive abilities
if you prefer). The goal of syntactic theory is to model these procedures.
In other words, we are trying to figure out what we subconsciously know
about the syntax of our language.
In generative grammar, the means for modeling these procedures is
through a set of formal grammatical rules. Note that these rules are nothing
1Whether language constrains what abstract things we can think about (this idea
is called the Sapir-Whorf hypothesis) is a matter of great debate and one that lies
outside the domain of syntax per se.
6 Preliminaries
like the rules of grammar you might have learned in school. These rules
don't tell you how to properly punctuate a sentence or not to split
an infinitive. Instead, they tell you the order in which to put your words (in
English, forexample, we put the subject of a sentence before its verb; this is
the kind of information encoded in generative rules). These rules are thought
to generate the sentences of a language, hence the name generative grammar.
You can think of these rules as being like the command lines in a computer
program. They tell you step by step how to put together words into
a sentence. We'll look at precise examples of these rules in the next chapter.
But before we can get into the nitty-gritty of sentence structure, let's look
at some of the underlying assumptions of generative grammar.
Noam Chomsky
Avram Noam Chomsky was born on the 7th of December 1928, in
Philadelphia. His father was a Hebrew grammarian and his mother a
teacher. Chomsky got his Ph.D. from the University of Pennsylvania,
where he studied linguistics under Zellig Harris. He took a position in
machine translation and language teaching at the Massachusetts Institute
of Technology. Eventually his ideas about the structure of language
transformed the field of linguistics. Reviled by some and admired by
others, Chomsky's ideas have laid the groundwork for the discipline of
linguistics, and have been very influential in computer science, and
philosophy.
Chomsky is also one of the leading intellectuals in the anarchist
socialist movement. His political writings about the media and political
injustice have profoundly influenced many. Chomsky is among the most
quoted authors in the world (among the top ten and the only living
person on the list).
3. Syntax as Science - the Scientific Method
To many people the study of language properly belongs in the domain of the
humanities. That is, the study of language is all about the beauty of its usage
in fine (and not so fine) literature. However, there is no particular reason,
other than our biases, that the study of language should be confined to a
humanistic approach. It is also possible to approach the study of language
from a scientific perspective; this is the domain of linguistics. People who
study literature often accuse linguists of abstracting away from the richness
of good prose and obscuring the beauty of language. Nothing
could be further from the truth. Most linguists, including the present author,
Chapter 1: Generative Grammar
enjoy nothing more than reading a finely crafted piece of fiction, and many
linguists often study, as a sideline, the more humanistic aspects of language.
This doesn't mean, however, that one can't appreciate and study the formal
properties (or rules) of language and do it from a scientific perspective.
The two approaches to language study are both valid; they complement
each other; and neither takes away from the other.
Science is perhaps one of the most poorly defined words of the English
language. We regularly talk of scientists as people who study bacteria,
particle physics, and the formation of chemical compounds, but ask
your average Joe or Jill on the street what science means, and you'll be hard
pressed to get a decent definition. Science refers to a particular methodology
for study: the scientific method. The scientific method dates backs to the
ancient Greeks, such as Aristotle, Euclid, and Archimedes. The method
involves observing some data, making some generalizations about patterns
in the data, developing hypotheses that account for these generalizations,
and testing the hypotheses against more data. Finally, the hypotheses
are revised to account for any new data and then tested again. A flow chart
showing the method is given in (1):
1) Gather and observe data
1 f
Make generalizations
' f
Develop hypotheses
In syntax, we apply this methodology to sentence structure. Syntacticians
start2 by observing data about the language they are studying, then
they make generalizations about patterns in the data (e.g., in simple English
declarative sentences, the subject precedes the verb). They then generate a
hypothesis - preferably one that makespredictions - and test the hypothesis
against more syntactic data, and if necessary go back and re-evaluate
their hypotheses.
2 This is a bit of an oversimplification. We really have a "chicken and the egg"
problem here. Youcan't know what data to study unless you have a hypothesis about
what is important, and you can't have a hypothesis unless you have some basic
understanding of the data. Fortunately, as working syntacticians this philosophical
conundrum is often irrelevant, as we can just jump feet-first into both the hypothesisforming
and data-analysis at the same time.
8 Preliminaries
Hypotheses are only useful to the extent that they make predictions.
A hypothesis that makes no predictions (or worse yet, predicts everything)
is useless from a scientific perspective. In particular, the hypothesis must be
falsifiable. That is we must, in principle, be able to look for some data, which
if true, show that the hypothesis is wrong. This means that we are often
looking for the cases where our hypotheses predict that a sentence will be
grammatical (and it is not), or the cases where they predict that the sentence
will be ungrammatical (but it is).
In syntax, hypotheses are called rules, and the group of hypotheses that
describe a language's syntax is called a grammar.
Do Rules Really Exist?
Generative grammar claims to be a theory of cognitive psychology, so the
natural question to ask at this point is whether formal rules really exist in
the brain/minds of speakers. After all, a brain is a mass of neurons firing
away, how can formal mathematical rules exist up there? Remember,
however, that we are attempting to model Language, we aren't trying to
describe Language exactly. This question confuses two disciplines:
psychology and neurology. Psychology is concerned with the mind,
which represents the output and the abstract organization of the brain.
Neurology is concerned with the actual firing of the neurons and the
physiology of the brain. Generative grammar doesn't try to be a theory of
neurology. Instead it is a model of the psychology of Language.
Obviously, the rules don't exist, per se in our brains, but they do represent
the external behavior of the mind. For more discussion of this issue, look
at the readings in the further reading section of this chapter.
The term grammar strikes terror into the hearts of many people. But you
should note that there are two ways to go about writing grammatical rules.
One is to tell people how they should speak (this is of course the domain
of English teachers and copy-editors); we call these kinds of rule prescriptive
rules (as they prescribe how people should speak according
to some standard). Some examples of prescriptive rules include "never end
a sentence with a preposition," "use whom not.who," "don't split infinitives."
These rules tell us how we are supposed to use our language. The other
approach is to write rules that describe how people actually speak, whether
or not they are speaking "correctly." These are called descriptive rules.
Consider for a moment the approach we're taking in this book; which of the
two types (descriptive or prescriptive) is more scientific? Which kind of rule
is more likely to give us insight into how the mind uses Language? For these
Chapter 1: Generative Grammar 9
reasons, we focus on descriptive rules. This doesn't mean that prescriptive
rules aren't important (in fact, in the problem sets section of this chapter you
are asked to critically examine the question of descriptive vs. prescriptive
rules), but for our purposes descriptive rules are more important. For
an interesting discussion of the prescriptive/descriptive debate, see Pinker's
1995 book: The Language Instinct.
You now have enough information to answer General Problem Set 1
3.1 An Example ofthe Scientific Method asApplied toSyntax
Let's turn now to a real world application of the scientific method to some
language data. The following data concern the form of a specific kind
of noun, called an anaphor (plural: anaphors, the phenomenon is called
anaphora). These are the nouns that end with -se//(e.g., himself, herself, itself,
etc.). Inchapter 5, we look at the distribution ofanaphora in detail; here we'll
only consider one superficial aspect of them. In the following sentences,
as is standard in the syntactic literature, a sentence that isn't well-formed
is marked with an asterisk (*) before it. For these sentences assume that Bill
is male and Sally is female.
2) a) Bill kissed himself.
b) *Bill kissed herself.
c) Sally kissed herself.
d) *Sally kissed himself.
e) *Kiss himself.
To the unskilled eye, the ill-formedsentences in (2band d) just look silly. It is
obvious that Bill can't kiss herself, because Bill is male. However, no matter
how matter-of-factly obvious this is, it is part of a bigger generalization
about the distribution of anaphors. In particular, the generalization we can
draw about the sentences in (2) is that an anaphor must agree ingender with
the noun it refers to (its antecedent). So in (2a and b) we see that the anaphor
must agree in gender with Bill, its antecedent. The anaphor must take the
masculine form himself. The situation in (2c and d) is the same; the anaphor
must take the form herself so that it agrees in gender with the feminine Sally.
Note further that a sentence like (2e) shows us that anaphors must have an
antecedent. An anaphor without an antecedent is unacceptable. A plausible
hypothesis (or rule) given the data in (2), then, is stated in (3):
3) An anaphor must (i) have an antecedent and (ii) agree in gender
(masculine, feminine, or neuter) with that antecedent.
10 Preliminaries
The next step in the scientific method is to test this hypothesis against more
data. Consider the additional data in (4):
4) a) The robot kissed itself.
b) She knocked herself on the head with a zucchini.
c) *She knocked himself on the head with a zucchini.
d) The snake flattened itselfagainst the rock.
e) ?The snake flattened himself/herself against the rock.
f) The Joneses think themselves the best family on the block.
g) *TheJoneses think himself the most wealthy guy on the block,
h) Gary and Kevin ran themselves into exhaustion.
i) *Gary and Kevin ran himself into exhaustion.
Sentences (4a, b, and c) are all consistent with our hypothesis that anaphors
must agree in gender with their antecedents, which at least confirms that the
hypothesis is on the right track. What about the data in (4d and e)? It appears
as if any gender is compatible with the antecedent the snake. This appears,
on the surface, to be a contradiction to our hypothesis. Think
about these examples a little more closely, however. Whether sentence (4e) is
well-formed or not depends upon your assumptions about the gender of the
snake. If you assume (or know) the snake to be male, then The snakeflattened
himself against the rock is perfectly well-formed. But under the same
assumption, the sentence The snake flattened herself against the rock seems
very odd indeed, although it is fine if you assume the snake is female. So it
appears as if this example also meets the generalization in (3); the vagueness
about its well-formedness has to do with the fact that we are rarely sure
what gender a snake is and not with the actual structure of the sentence.
Now, look at the sentences in (4f-i); note that the ill-formedness of (g)
and (i) is not predicted by our generalization. In fact, our generalization
predicts that sentence (4i) should be perfectly grammatical, since himself
agrees in gender (masculine) with its antecedents Gary and Kevin. Yetthere is
clearly something wrong with this sentence. The hypothesis needs revision.
It appears as if the anaphor must agree in gender and number
with the antecedent. Number refers to the quantity of individuals involved
in the sentence; English primarily distinguishes singular number
from plural number. (5) reflects our revised hypothesis.
5) An anaphor must agree in gender and number with its antecedent.
If there is more than one person or object mentioned in the antecedent, then
the anaphor must be plural (i.e., themselves).
Chapter 1: Generative Grammar 11
Testing this against more data, we can see that this partially makes
the right predictions (6a), but it doesn't properly predict the grammaticality
of sentences (6b-e):
6) a) People from Tucson think very highly of themselves.
b) *Igave yourself the bucket of ice cream.
c) I gave myself the bucket of ice cream.
d) *Shehit myself with a hammer.
e) She hit herself with a hammer.
Even more revision is in order. The phenomenon seen in (6b-e) revolves
around a grammatical distinction called person. Person refers to the
perspective of the speaker with respect to the other participants in the speech
act. First person refers to the speaker. Second person refers to the listener.
Third person refers to people being discussed that aren't participating in the
conversation. Here are the English pronouns associated with each person:
(Nominative refers to the case form the pronouns take when in subject
position like 7 in "I love peanut butter;" accusative refers to the form they
take when in object positions like me in "John loves me." We will look at case
in much more detail in chapter 9, so don't worry if you don't understand
it right now.)
7)
Nominative Accusative Anaphoric
Singular Plural Singular Plural Singular Plural
1 I we me us myself ourselves
2 you you you you yourself yourselves
3 masc he
they
him
them
himself
themselves3 fern she her herself
3 neut it it itself
As you can see from this chart, the form of the anaphor seems also to agree
in person with its antecedent. So once again we revise our hypothesis (rule):
8) An anaphor must agree in person, gender and number with its
antecedent.
With this hypothesis, we have a straightforward statement of the distribution
of this noun type, derived using the scientific method. In the problem sets
below, and in chapter 5, you'll have an opportunity to revise the rule in (8)
with even more data.
You nowhaveenough information to try Challenge Problem Sets 1 &2
12 Preliminaries
3.2 Sources of Data
If we are going to apply the scientific method to syntax, it is important
to consider the sources of data. One obvious source is in collections of either
spoken or written texts. Such data are called corpora (singular: corpus).
There are many corpora available, including some searchable through the
World Wide Web. For languages without a literary tradition or ones spoken
by a small minority, it is often necessary for the linguist to go and gather
data and compile a corpus in the field. In the early part of this century,
this was the primary occupation of linguists, and it is proudly carried
on today by many researchers.
While corpora are unquestionably invaluable sources of data, they are
only a partial representation of what goes on in the mind. More particularly,
corpora often contain instances of only grammatical (or more precisely
well-formed) sentences (sentences that sound "OK" to a native speaker).
For example, the online New York Times contains very few ungrammatical
sentences. Even corpora of naturalistic speech complete with the errors
every speaker makes don't necessarily contain the data we need to test
the falsifiable predictions of our hypotheses.
You might think that what's in a corpus would be enough for a linguist
to do her job. But corpora are just not enough: there is no way of knowing
whether a corpus has all possible forms of grammatical sentences. In fact,
as we will see in the next chapter, due to the productive nature of language,
a corpus could never contain all the grammatical forms of a language,
nor could it even contain a representative sample. To really get at what
we know about our languages (remember syntax is a cognitive science), we
have to know what sentences are not well-formed. That is, in order to know
the range of what are acceptable sentences of English, Italian or Igbo, wefirst
have to know what are not acceptable sentences in English, Italian or Igbo.
This kind of negative information is very rarely available in corpora,
which mostly provide grammatical, or well-formed, sentences.
Consider the following sentence:
9) *Who do you wonder what bought?
For most speakers of English, this sentence borders on word salad - it is not
a good sentence of English. How do you know that? Were you ever taught
in school that you can't say sentences like (9)? Has anyone ever uttered
this sentence in your presence before? I seriously doubt it. The fact
that a sentence like (9) sounds strange, but similar sentences like (10a and b)
dosound OK is not reflected anywhere in a corpus:
Chapter 1: Generative Grammar 13
10) a) Who do you think bought the bread machine?
b) I wonder what Fiona bought.
Instead we have to rely on our knowledge of our native language (or on the
knowledge of a native speaker consultant for languages that we don't speak
natively). Notice that this is notconscious knowledge. Idoubt there are many
native speakers of English that could tell you why sentence (9) is terrible, but
most can tell you that it is. This is subconscious knowledge. The trick is to get
at and describe this subconscious knowledge.
The psychological experiment used to get this subconscious kind
of knowledge is called the grammaticality judgment task. The judgment
task involves asking a native speaker to read a sentence, and judge whether
it is well-formed (grammatical), marginally well-formed, or ill-formed
(unacceptable or ungrammatical).
Judgments as Science?
Many linguists refer to the grammaticality judgment task as "drawing
upon our native speaker intuitions." The word "intuition" here is slightly
misleading. The last thing that pops into our heads when we hear the term
"intuition" is science. Generative grammar has been severely criticized by
many for relying on "unscientific" intuitions. But this is based primarily
on a misunderstanding of the term. To the lay person, the term "intuition"
brings to mind guesses and luck. This usage of the term is certainly
standard. When a generative grammarian refers to "intuition" however,
she is using the term to mean "tapping into our subconscious knowledge."
The term "intuition" may have been badly chosen, but in this
circumstance it refers to a real psychological effect. Intuition (as a
grammaticality judgment) has an entirely scientific basis. It is replicable
under strictly controlled experimental conditions (these conditions are
rarely applied, but the validity of the task is well established). Other
disciplines also use intuitions or judgment tasks. Forexample, within the
study of vision, it has been determined that people can accurately judge
differences in light intensity, drawing upon their subconscious knowledge
(Bard et al. 1996). To avoid the negative associations with the term
intuition, we will use the term judgment instead.
There are actually several different kinds of grammaticality judgments.
Both of the following sentences are ill-formed, but for different reasons:
11) a) #The toothbrush is pregnant.
b) Toothbrush the is blue.
14 Preliminaries
Sentence (Ha) sounds bizarre (cf. the toothbrush is blue) because we know
that toothbrushes (except in the world of fantasy/science fiction or poetry)
cannot be pregnant. The meaning of the sentence is strange, but the form is
OK. We call this semanticill-formedness and mark the sentence with a #. By
contrast, we can glean the meaning of sentence (lib); it seems semantically
reasonable (toothbrushes can be blue), but it is ill-formed from a structural
point of view. That is, the determiner the is in the wrong place
in the sentence. This is a syntactically ill-formed sentence. A native speaker
of English will judge both thesesentences as ill-formed, but for very different
reasons. In this text, we will be concerned primarily with syntactic
well-formedness.
You now have enough information to answer General Problem Set 2
4. Where do the Rules Come From?
In this chapter we've been talking about our subconscious knowledge of
syntactic rules, but we haven't dealt with how we get this knowledge. This
is sort of a side issue, but it may affect the shape of our theory. If we know
howchildren acquire theirrules, thenweare ina betterposition fora proper
formalization ofthem. The way by which children develop knowledge isan
important question in cognitive science. The theory of generative grammar
makes some very specific (and very surprising) claims about this.
4.1 Learning vs. Acquisition
One of the most common misconceptions about Language is the idea
that children and adults "learn" languages. Recall that the basic kind
of knowledge we are talking abouthere is subconscious knowledge. When
producing a sentence you don't consciously think about where to put the
subject, where toput the verb, etc. Your subconscious language faculty does
that for you. Cognitive scientists make a distinction in howwegetconscious
and subconscious knowledge. Conscious knowledge (like the rules
of algebra, syntactictheory, principles of organicchemistry, or how to take
apart a carburetor) is learned. Subconscious knowledge, like how to speak
or the ability to visually identify discrete objects, is acquired. In part,
this explains why classes in the formal grammar of a foreign language
often fail abysmally to train people to speak those languages. By contrast,
being immersed in an environment where you can subconsciously acquire
a language is much more effective. In this text we'll be primarily interested
Chapter 1: Generative Grammar 15
in how people acquire the rules oftheirlanguage. Notall rules of grammar
are acquired, however. Some facts about Language seem to be built
into our brains, or innate.
You now have enough information to answer General Problem Set 3
4.2 Innateness: Language as an Instinct
Ifyou think about the other types ofknowledgethat are subconscious, you'll
see that many3 of them (for example, the ability to walk) are built directly
into our brains - they are instincts. No one had to teach you to walk (despite
what your parents might think!). Kidsstart walking on their own. Walking is
an instinct. Probably the most controversial claim of Noam Chomsky's is that
Language is also an instinct. Many parts of Language are built in, or innate.
Much of Language is an ability hard-wired into our brains by our genes.
Obviously, particular languages are not innate. It isn't the case that
a child of Slovakparents growing up in North America who is never spoken
to in Slovak, grows up speaking Slovak. They'll speak English (or whatever
other language is spoken around them). So on the surface it seems crazy
to claim that Language is an instinct. There are very good reasons to believe,
however, that a human facility for Language (perhaps in the form
of a "Language organ" in the brain) is innate. We call this facility Universal
Grammar (or UG).
4.3 The Logical Problem ofLanguage Acquisition
What follows is a fairly technical proof of the idea that Language is at least
plausibly construed as an innate, in-built system. If you aren't interested
in this proof (and the problems with it), then you can reasonably skip
ahead to section 4.4.
Theargument in this section is that a productive system like the rules of
Language probably have not been learned or acquired. Infinite systems are
in principle, given certain assumptions, both unlearnable and unacquirable.
Since we all have such an infinite system in our heads, and we shouldn't
have been able to aquire it. So it follows that it is built in. The argument
presented here is based on an unpublished paper by Alec Marantz,
but is based on an argument dating back to at least Chomsky (1965).
First here's a sketch of the proof, which takes the classical form of an
argument by modus ponens:
3 but not all!
16 Preliminaries
Premise (i): Syntaxisa productive,recursive and infinitesystem
Premise (ii): Rulegoverned infinitesystemsare unlearnable.
Conclusion: Therefore syntax is an unlearnable system. Since we have it,
it follows that at least parts of syntax are innate.
There are parts ofthisargument thatare very controversial. In thechallenge
problem sets at theend ofthischapter youareinvited tothinkverycritically
about the form of this proof. Challenge Problem Set 3 considers
the possibility that premise (i) is false (but hopefully, you will conclude
that despite the argument given in the problemset, that the idea Language
is productive and infinite is correct). Premise (ii) is more dubious, and is
the topic of Challenge Problem Set 4. Here, in the main body of the text,
I willgive you the classic versions ofthesupport forthesepremises, without
criticizing them. You are invited to be skeptical and critical of them when
you do the Challenge Problem sets.
Let's start with premise (i). Language is a productive system. That is,
you can produce and understand sentencesyou have never heard before. For
example, I can practically guarantee that you have never heard the following
sentence:
12) The dancing chorus-line of elephants broke my television set.
The magic of syntax is that it can generate forms that have never
been produced before. Another example of the productive quality lies
in what is called recursion. It is possible to utter a sentence like (13):
13) Rosie loves magazine ads.
It is also possible to put this sentence inside another sentence, like (14):
14) I think [Rosie loves magazine ads].
Similarly you can put this larger sentence inside of another one:
15) Drew believes [I think [Rosieloves magazine ads]].
and of course you can put this bigger sentence inside of another one:
16) Dana doubts that [Drew believes [I think [Rosie loves magazine ads]]].
and so on, and so on ad infinitum. It is always possible to embed a sentence
inside of a larger one. This means that Language is a productive (probably
infinite) system. There are no limits on what we can talk about. Other
examples of the productivity of syntax can be seen in the fact that you can
infinitely repeat adjectives (17) and you can infinitely add coordinated nouns
to a noun phrase (18):
Chapter 1: Generative Grammar 17
17) a) a very big peanut
b) a very very big peanut
c) a very very very big peanut
d) a very very very very big peanut
etc.
18) a) Dave left
b) Dave and Alina left
c) Dave, Dan and Alina left
d) Dave, Dan, Erin and Alina left
e) Dave, Dan, Erin, Jaime and Alina left
etc.
It follows that for every grammatical sentence of English, you can find
a longer one (based on one of the rules of recursion, adjective repetition,
or coordination) that is longer. This means that language is at least
countably infinite. This premise is relatively uncontroversial (however,
see the discussion in Challenge Problem Set 3).
Let's now turn to premise (ii): The idea that infinite systems are
unlearnable. In order to make this more concrete, let's consider an algebraic
treatment of a linguistic example. Imagine that the task of a child
is to determine the rules by which her language is constructed. Further,
let's simplify the task, and say a child simply has to match up situations
in the real world with utterances she hears.4 So upon hearing the utterance
the cat spots Die kissing fishes, she identifies it with an appropriate situation
in the context around her (as represented by the picture).
19) "the cat spots the kissing fishes" =
1The task is actually several magnitudes more difficult than this, as the child has
to work out the phonology, etc., too, but for argument's sake, let's stick with
this simplified example.
18 Preliminaries
Herjob, then, is tocorrectly match up thesentences withthesituation.5 More
crucially she has tomakesure that shedoes not matchit up withall the other
possible alternatives, such as the things going on around her (like her older
brother kickingthe furniture, or her mother making her breakfast, etc.). This
matching of situations with expressions is a kind of mathematical relation
(or function) that maps sentences onto a particular situation. Another way of
putting it is that she has to figure out the rule(s) that decode(s) the meaning
of the sentences. It turns out that this task is, at least very difficult
if not impossible.
Let's make this even more abstract to get at the mathematics of the
situation. Assign each sentence some number. This number will represent
the input to the rule. Similarly we will assign each situation a number.
The function (or rule) modeling language acquisition maps from the set
of sentence numbers to the set of situation numbers. Now let's assume that
the child has the following set of inputs and correctly matched situations
(perhaps explicitly pointed out to her by her parents). The x value represents
the sentences she hears. The y is the number correctly associated
with the situation.
20) Sentence (input) Situation (output)
x v
1 i
2 2
3 3
4 4
5 5
Given this input, what do you suppose that the output where x = 6 will be?
6 ?
Most people will jump to the conclusion that the output will be 6 as well.
That is, they assume that the function (the rule) mapping between inputs
and outputs is x = y. But what if I were to tell you that in the hypothetical
situation I envision here, the correct answer is situation number 126. The rule
that generated the table in (20) is actually:
21) I(x-5)*(x-4)*(x-3)*(x-2)*(x-l)] + x = y
With this rule, all inputs equal to or less than 5 will give an output equal to
the input, but for all inputs greater than 5, will give some large number.
5Note that this is the job of the child who is using universal grammar, not the job of
UG itself.
Chapter 1: Generative Grammar 19
When you hypothesized the rule was x = y, you didn't have all the
crucial information; you only had part of the data. This seems to mean that if
you hear only the first five pieces of data in our table then you won't
get the rule, but if you learn the sixth you willfigure it out. Is this necessarily
the case? Unfortunately not: Even if you add a sixth line, you have no way
of being sure that you have the right function until you have heard all the
possible inputs. The important information might be in the sixth line, but it
might also be in the 7,902,821,123,765th sentence that you hear. Youhave no
way of knowing for sure if you have heard all the relevant data until you
have heard them all. In an infinite system you can't hear them all, even if you
were to hear 1 sentence every 10 seconds for your entire life. If we assume
the average person lives to be about 75 years old, if they heard one new
sentence every ten seconds, ignoring leap years and assuming they never
sleep, they'd have only heard about 39,420,000 sentences over their lifetime.
This is a much smaller number than infinity. Despite this poverty of input,
by the age of 5 most children are fairly confident with their use of
complicated syntax. Productive systems are (possibly) unlearnable, because
you never have enough input to be sure you have all the relevant facts.
This is called the logical problem of language acquisition.
Generative grammar gets around this logical puzzle by claiming that the
child acquiring English, Irish, or Yoruba has some help: a flexible blueprint
to use in constructing her knowledge of language called Universal Grammar.
Universal Grammar restricts the number of possible functions that
map between situations and utterances, thus making language learnable.
You now have enough information to try Challenge Problem Sets 3 &4
4.4 Other Argumentsfor UG
The evidence for UG doesn't rely on the logical problem alone, however.
There are many other arguments that support the hypothesis that at least
a certain amount of language is built in.
An argument that is directly related to the logical problem of language
acquisition discussed above has to do with the fact that we know things
about the grammar of our language that we couldn't possibly have learned.
Start with the data in (20). A child might plausibly have heard sentences
of these types (the underline represents the place where the question word
who plausibly starts out - that is either as the object or subject of the verb
will question):
20 Preliminaries
22) a) Who do you think that Ciaran will question first?
b) Who do you think Ciaran will question first?
c) Who do you think will question Seamus first?
The child has to draw a hypothesis about the distribution of the word that in
English sentences. Oneconclusion consistent with this observed data is that
the word that in English is optional. You can either have it or not.
Unfortunately this conclusion is not accurate. Consider the fourth sentence
in the paradigm in (22). Thissentence is the same as (22c) but with a that:
d) *Who do you think that will question Seamus first?
It appears as if that is only optional when the question word (who in this case)
starts in object position (as in 22a and b) It is obligatorily absent when the
question word starts in subject position (as in 22c and d) (don't worry about
the details of this generalization). What is important to note is that no one has
ever taught you that (22d) is ungrammatical. Nor could you have come to
that conclusion on the basis of the data you've heard. The logical hypothesis
on the basis of the data in (220a-c) predicts sentence (22d) to be grammatical.
There is nothing in the input a child hears that would lead them to
the conclusion that (22d) is ungrammatical, yet every English-speaking child
knows it is. One solution to this conundrum is that we are born
with the knowledge that sentences like (22d) are ungrammatical.6 This kind
of argument is often called the underdetermination of the data argument
for UG.
Most parents raising a toddler will swear up and down that they are
teaching their children to speak; that they actively engage in instructing their
child in the proper form of the language. That overt instruction by parents
plays any role in language development is easily falsified. The evidence
from the experimental language acquisition literature is very clear: parents,
despite their best intentions, do not, for the most part, correct ungrammatical
utterances by their children. More generally, they correct the content rather
6 The phenomenon in (22) is sometimes called the that-trace effect. There is no
disputing the fact that this phenomenon is not learnable. However, it is also a fact
that it is not a universal property of all languages. For example, French and Irish
don't seem to have the f/wMrace effect. Here is a challenge for those of you who like
to do logic puzzles: If the that-trace effect is not learnable and thus must be
biologically built in, how is it possible for a speaker of French or Irish to violate it?
Think carefully about what kind of input a child might have to have in order to learn
an "exception" to a built-in principle. This is a hard problem, but there is a solution.
It may become clearer below when we discuss parameters.
Chapter 1: Generative Grammar 21
than the form of their child's utterance (see for example the extensive
discussion in Holzman 1997).
23) (from Marcus et al. 1992)
Adult: Where is that big piece of paper I gave you yesterday?
Child: Remember? I writed on it.
Adult: Oh thafs right, don't you have any paper down here, buddy?
When a parent does try to correct a child's sentence structure, it is more often
than not ignored by the child:
24) (from Pinker 1995: 281 - attributed to Martin Braine)
Child: Want other one spoon, Daddy
Adult: You mean, you want the other spoon.
Child: Yes, I want other one spoon, please Daddy.
Adult: Can you say "the other spoon"?
Child: Other ... one ... spoon
Adult: Say "other".
Child: other
Adult: "spoon"
Child: spoon
Adult: "other ... spoon"
Child: other ... spoon. Now give me other one spoon.
This humorous example is typical of parental attempts to "instruct" their
children in language. When they do occur, they fail. However, children still
acquire language in the face of a complete lack of instruction. Perhaps one
of the most convincing explanations for this is UG. In the problem set part
of this chapter, you are asked to consider other possible explanations
and evaluate which are the most convincing.
There are also typological arguments for the existence of an innate
language faculty. All the languages of the world share certain properties
(for example they all have subjects and predicates - other examples will be
seen throughout the rest of this book). These properties are called universals
of Language. If we assume UG, then the explanation for these language
universals is straightforward - they exist because all speakers of human
languages share the same basicinnate materials forbuilding their language's
grammar. In addition to sharing many similarcharacteristics, recent research
into Language acquisition has begun to show that there is a certain amount
of consistency cross-linguistically in the way children acquire Language.
For example, children seem to go through the same stages and make
the same kinds of mistakes when acquiring their language, no matter
what their cultural background.
22 Preliminaries
Statistical Probability or UG?
In looking at the logical problem of language acquisition you might be
asking yourself "Ok, so maybe kids don't get all the data, but maybe they
get enough to draw conclusions about what is the most likely structure of
their grammar?" For example, we might conclude that a child learning
English would observe the total absence of any sentences that have that
followed by a trace (e.g., 22d), so after hearing some threshold of
sentences they conclude that this sentence type is ungrammatical. This is
a common objection to the hypothesis of UG. Unfortunately, this
hypothesis can't explain why many sentence types that are extremely rare
(to the point that they are probably never heard by children) are still
judged as grammatical by the children. For example, English speakers
rarely (if ever) produce sentences with seven embeddings (John said that
Mary thinks that Susan believes that Matt exclaimed that Marian claimed that
Art said that Andrew wondered if Gwen had lost her pen); yet speakers of
English routinely agree these are acceptable. The actual speech of adult
speakers is riddled with errors (due to all sorts of external factors:
memory, slips of the tongue, tiredness, distraction, etc.). But children do
not seem to assume that any of these errors, which they hear frequently,
are part of the data that determines their grammars.
Finally, there are a number of biological arguments in favor of UG.
As noted above, Language seems to be both human-specific and pervasive
across the species. All humans, unless they have some kind of physical
impairment, seem to have Language as we know it. This points towards
it being a genetically endowed instinct. Additionally, research
from neurolinguistics seems to point towards certain parts of the brain
being linked to specific linguistic functions.
With very few exceptions, most linguists believe that some Language
is innate. What is of controversy is how much is innate and whether
the innateness is specific to Language, or follows from more general
innate cognitive functions. We leave these questions unanswered here.
You nowhave enough information to try General Problem Set 4
4.5 Explaining Language Variation
The evidence for UG seems to be very strong. However, we are still left with
the annoying problem that languages differ from one another. This problem
is what makes the study of syntax so interesting. It is also not an unsolvable
Chapter 1: Generative Grammar 23
one. One way in which languages differ is in terms of the words used in the
language. These clearly have to be learned or memorized. Other differences
between languages (such as the fact that basic English word order is subjectverb-object
(SVO), but the order of an Irish sentence is verb-subject-object
(VSO) and the order of a Turkish sentence is subject-object-verb (SOV))
must also be acquired. The explanation for this kind of fact will be explored
in chapter 6. Foreshadowing slightly, we'll claim there that differencesin the
grammars of languages can be boiled down to the setting of certain innate
parameters (or switches) that select among possible variants. Language
variation thus reduces to learning the correct set of words and selecting
from a predetermined set of options.
Oversimplifying slightly, most languages put the order of elements in a
sentence in one of the following word orders:
25) a) Subject VerbObject (SVO) (e.g., English)
b) Subject Object Verb(SOV) (e.g., Turkish)
c) VerbSubjectObject (VSO) (e.g., Irish)
A few languages use:
d) VerbObjectSubject(VOS) (e.g., Malagasy)
No (or almost no)7 languages use
e) ObjectSubjectVerb (OSV)
f) ObjectVerbSubject(OVS)
Let usimagine thatpart ofUG isa parameter thatdetermines thebasic word
order. Four oftheoptions (SVO, SOV, VSO, andVOS) areinnately available
as possible settings. Two of the possible word orders are not part of UG. The
child who isacquiring English isinnately biased towards one ofthecommon
orders, when she hears a sentence like "Mommy loves Kirsten," if the child
knows the meaning ofeach of the words, then she might hypothesize two
possible word orders for English: SVO and OVS. None of the others are
consistent with the data. The child thus rejects all the other hypotheses. OVS
is not allowed, since it isn't one of the innately available forms. This leaves
SVO, which is the correct order for English. So children acquiring English
will choose to set the word order parameter at the innately available
SVO setting.
In his excellent book The Atoms ofLanguage, Mark Baker inventories a set
of possible parameters of a language variation within the UG hypothesis.
7This is a matter of some debate. Derbyshire (1985) has claimed that the language
Hixkaryana has objectinitial order.
24 Preliminaries
This is an excellent and highly accessible treatment of parameters. I strongly
recommend this book.
.You nowhave enough information totry General Problem Set5 and Challenge Set5
5. Choosing among Theories about Syntax
There is one last preliminary we have to touch on before actually doing
some real syntax. In this book we are going to posit many hypotheses. Some
of these we'll keep, others we'll revise, and still others we'll reject. How do
we know what is a good hypothesis and what is a bad? Chomsky
(1965) proposed that we can evaluate how good theories of syntax are,
using what are called the levelsofadequacy. Chomskyclaimed that there are
threestages that a grammar (thecollection ofdescriptive rules that constitute
your theory) can attain in terms ofadequacy.
If your theory only accounts for the data in a corpus (say a series of
printed texts) and nothing more it is said to be an observationally adequate
grammar. Needless to say, this isn't much use if we are trying to account
for the cognition of Language. As we discussed above, it doesn't tell
us the whole picture. We also need to know what kinds of sentences
are unacceptable, or ill-formed. A theory that accounts forboth corpora and
native speaker judgments about well-formedness is called a descriptively
adequate grammar. On the surface this may seem to be all we need.
Chomsky, however, has claimed that we can go one step better. He points
out that a theory that also accounts for how children acquire their language
is the best. He calls this an explanatorily adequate grammar. The simple
theory of parameters might get this label. Generative grammar strives
towards explanatorily adequate grammars.
You now have enough information totry General Problem Set6
6. The Scientific Method and the Structure of this
Textbook
Throughout this chapter I've emphasized the importance of the scientific
method tothe study ofsyntax. It's worth noting that we're notonly going to
apply this principle tosmall problems orspecific rules, butwe'll also apply
it in a more global way. This principle isin part a guide tothe wayin which
the rest of this book is structured.
In chapters 2-5 (the remainder of Part I of the book) we're
goingto developan initial hypothesis about the way in whichsyntactic rules
Chapter 1: Generative Grammar 25
are formed. These are the Phrase Structure Rules (PSRs). Chapters 2 and 3
examine the words these rules use, the form of the rules, and they structures
they generate. Chapters 4 and 5 look at ways we can detail the structure
of the trees formed by the PSRs.
In chapters 6-8 (Part 2 of the book), we examine some data that presents
problems for the simple grammar presented in Part 1. When faced with
more complicated data, we revise our hypotheses, and this is precisely what
we do. We develop a special refined kind of PSRknown as an X-bar rule. Xbar
rules are still phrase structure rules, but they offer a more sophisticated
way of looking at trees. This more sophisticated version also needs
an additional constraint known as the "theta criterion" which is the focus
of chapter 8.
In chapters 9-12 (Part 3) we consider even more data, and refine our
hypothesis again. This time adding a new rule type: the transformation
(we retain X-bar, but enrich it with transformations). Part 4 of the book
(chapters13-16) refines these proposals even further.
Witheachstep webuild upon our initial hypothesis, just as the scientific
method tells us to. I've been teaching with this proposal-revision method
theory construction for a couple ofyears now, andevery nowand thenIhear
the complaint froma student that we should just start with the final answer
(i.e. the revised hypothesis found in the later chapters in the book).
Why bother learning all this "other" "wrong" stuff? Why should we bother
learning Phrase Structure Rules? Why don't wejustjump straight intoX-bar
theory? Well, in principle, I could have constructed a book like that,
butthen you, the student, wouldn't understand why things aretheway they
are in the latter chapters. The theory would appear to be unmotivated, and
you wouldn't understand what the technology actually does. By proposing
a simple hypothesis early on in the initial chapters, and then refining
and revising it, building new ideas onto old ones, you not only get an
understanding ofthe motivations for and inner workings of our theoretical
premises, but you getpractice in working like a real linguist. Professional
linguists, like all scientists, work from a set ofsimple hypotheses and revise
them in light of predictions made by the hypotheses. The earlier versions
ofthe theoryaren't "wrong" so much as they need refinement and revision.
These early versions represent the foundations out of which the rest
of the theory has been built This is how science works.
26 Preliminaries
7. Summary
In this chapter, we've done very little syntax but talked a lot about
the assumptions underlying the approach we're going to take to the study
of sentence structure. Thebasic approach to syntax that we'll be using here is
generative grammar; we've seen that this approach is scientific in that it uses
the scientific method. It is descriptive and rule based. Further, it assumes
that a certain amount of grammar is built in and the rest is acquired.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Syntax: The level of linguistic organization that mediates between
sounds and meaning, where words are organized into phrases
and sentences.
ii) Language (capital L): Thepsychological ability ofhumans toproduce
and understand a particular language. Also called the Human
Language Capacity or i-Language. This is the object of study in
this book.
iii) language (lower-case I): A language like English or French. Theseare
the particular instances of the human Language. The data source
we use to examine Language is language. Also called e-language.
iv) Generative Grammar. A theory of linguistics in which grammar is
viewedas a cognitive faculty. Language isgeneratedby a set ofrules
or procedures. The version of generative grammar we are looking
at here is primarily the Principles and Parameters approach
(P&P) touching occasionally on Minimalism.
v) The Scientific Method: Observe some data, make generalizations
about that data, draw a hypothesis, test the hypothesis against
more data.
vi) Falsifiable Prediction: To prove that a hypothesis correct you have
to look for the data that would prove it wrong. The prediction
that might prove a hypothesiswrong is said to be falsifiable.
vii) Grammar. Not what you learned in school. This is the set of rules
that generate a language.
viii) Prescriptive Grammar. The grammar rules as taught by so called
"language experts." These rules, often inaccurate descriptively,
Chapter 1: Generative Grammar 27
prescribe how people should talk/ write, rather than describe
what they actually do.
ix) Descriptive Grammar: A scientific grammar that describes, rather
than prescribes, how people talk/write.
x) Anaphor: A word that ends in -selfor -selves (a better definition will
be given in chapter 5).
xi) Antecedent: The noun an anaphor refers to.
xii) Asterisk: * used to mark syntactically ill-formed (unacceptable
or ungrammatical) sentences. The hash mark, pound, or number
sign (#) is used to mark semantically strange, but syntactically
well-formed, sentences.
xiii) Gender (Grammatical): Masculine vs. Feminine vs. Neuter. Does not
have to be identical to the actual sex of the referent. For example, a
dog might be female, but we can refer to it with the neuter pronoun
it. Similarly, boats don't have a sex, but are grammatically feminine.
xiv) Number: The quantity of individuals or things described by a noun.
English distinguishes singular (e.g., a cat) from plural (e.g., the cats).
Other languages have more or less complicated number systems.
xv) Person: The perspective of the participants in the conversation.
The speaker or speakers (/, me, we, us) are called first person.
The listener(s) (you), are called the second person. Anyone
else (those not involved in the conversation) (he, him, she, her,
it, they, them) is referred to as the third person.
xvi) Case: The form a noun takes depending upon its position in the
sentence. We discuss this more in chapter 10.
xvii) Nominative: The form of a noun in subject position (/, you, he, she, it,
we, they).
xviii) Accusative: The form of a noun in object position (me, you, him, her,
it, us, them).
xix) Corpus (pi. Corpora): A collectionof real-world language data.
xx) Native Speaker Judgments (intuitions): Information about the
subconscious knowledge of a language. This information is tapped
by means of the grammaticality judgment task.
28 Preliminaries
xxi) Semantic Judgment: A judgment about the meaning of a sentence,
often relying on our knowledge of the context in which the sentence
was uttered.
xxii) Syntactic Judgment: A judgment about the form or structure of a
sentence.
xxiii) Learning: The gathering of conscious knowledge (like linguistics or
chemistry).
xxiv) Acquisition: The gathering of subconscious information (like
language).
xxv) Innate: Hard-wired or built in, an instinct.
xxvi) Recursion: The ability to embed structures iteratively inside one
another. Allows us to produce sentences we've never heard before.
xxvii) Universal Grammar (UG): The innate (or instinctual) part of each
language's grammar.
xxviii) The Logical Problem of Language Acquisition: The proof that an
infinite system like human language cannot be learned on the basis
of observed data - an argument for UG.
xxix) Underdetermination of the Data: The idea that we know things
about our language that we could not have possibly learned an
argument for UG.
xxx) Universal: A propertyfound in all the languages ofthe world.
xxxi) Observationally Adequate Grammar. A grammar that accounts for
observed real-world data (such as corpora).
xxxii) Descriptively Adequate Grammar. A grammar that accounts for
observed real-world data and native speaker judgments.
xxxiii) Explanatorily Adequate Grammar: A grammar that accounts for
observed real-world data and native speaker judgments and offers
an explanation for the facts of language acquisition.
Further Reading
Baker, Mark (2001) The Atoms of Language: The Mind's Hidden Rules of
Grammar. New York: Basic Books.
Chapter 1: Generative Grammar 29
Barsky, Robert (1997) Noam Chomsky: A Life ofDissent. Cambridge: MIT Press.
Chomsky, Noam (1965) Aspects ofthe Theory ofSyntax. Cambridge: MITPress.
Jackendoff, Ray (1993)Patterns in the Mind. London: Harvester-Wheatsheaf.
Pinker, Steven (1995) The Language Instinct. New York:Harper Perennial.
Sampson, Geoffrey (1997) Educating Eve: The Language Instinct Debate.
London: Cassell.
Uriagereka, Juan (1998) Rhyme and Reason: An Introduction to Minimalist
Syntax. Cambridge: MIT Press.
General Problem Sets
1. Prescriptive Rules
[Creative and Critical Thinking; Basic]
In the text above, we argued that descriptive rules are the primary focus
of syntactic theory. This doesn't mean that prescriptive rules don't have
their uses. What are these uses? Why do we maintain prescriptive rules
in our society?
2. Judgments
[Application of Skills; Intermediate]
All of the following sentences have been claimed to be ungrammatical
or unacceptable by someone at some time. For each sentence, indicate
whether this unacceptability is
i) a prescriptive or a descriptive judgment, and
ii) for all descriptive judgments indicate whether the ungrammaticality has
to do with syntax or semantics (or both).
One- or two-word answers are appropriate. If you are not a native speaker
of English, enlist the help of someone who is. If you are not familiar
with the prescriptive rules of English grammar, you may want to consult
a writing guide or Englishgrammar or look at Pinker's The LanguageInstinct.
a) Who did you see in Las Vegas?
b) You are taller than me.
c) My red is refrigerator.
d) Who do you think that saw Bill?
e) Hopefully, we'll make it through the winter without snow.
f) Myfriends wanted to quickly leave the party.
g) Bunnies carrots eat.
h) John's sister is not his sibling.
30 Preliminaries
3. Learning vs. Acquisition
[Creative and Critical Thinking; Basic]
We have distinguished between learning and acquiring knowledge. Learning
is conscious, acquisition is automatic and subconscious. (Note that acquired
things are not necessarily innate. They are just subconsciously obtained.)
Other than language are there other things we acquire? What other things do
we learn? What about walking? Or reading? Or sexual identity? An important
point in answering this question is to talk about what kind of evidence
is necessary to distinguish between learning and acquisition.
4. Universals
[Creative and Critical Thinking; Intermediate]
Pretend for a moment that you don't believe Chomsky and that you don't
believe in the innateness of syntax (but only pretend)). How might
you account for the existence of universals (see definition above) across
languages?
5. Innateness
[Creative and Critical Thinking; Intermediate]
We argued that some amount of syntax is innate (inborn). Can you think of
an argument that might be raised against innateness? (Itdoesn't have to be
an argument that works, just a plausible one.) Alternately, could you come up
with a hypothetical experiment that could disprove innateness? What would
such an experiment have to show? Remember that cross-linguistic variation
(differences between languages) is not an argument against innateness
or UG, because UG contains parameters that allow minute variations.
6. Levels of Adequacy
[Application of Skills; Basic]
Below, you'll find the description of several differentlinguists' work. Attribute
a level ofadequacy to them (state whether the grammars they developed are
observationally adequate, descriptively adequate, or explanatorily adequate).
Explain whyyou assigned the level of adequacy that you did.
a) Juan Martinez has been working with speakers of Chicano
English in the barrios of Los Angeles. He has been looking
both at corpora (rap music, recorded snatches of speech)
and working with adult native speakers.
b) Fredrike Schwarz has been looking at the structure of sentences
in eleventh-century Welsh poems. She has been working
at the national archives of Wales in Cardiff.
c) Boris Dimitrov has been working with adults and corpora
on the formation of questions in Rhodopian Bulgarian. He is also
conducting a longitudinal study of some two-year-old children
learning the language to test his hypotheses.
Chapter 1: Generative Grammar 31
Challenge Problem Sets
Challenge Problem Sets are special exercises that either challenge the presentation
ofthe main text oroffer significant enrichment. Students are encouraged tocomplete
the other problem sets before trying the Challenge Sets. Challenge Sets can vary in
level from interesting puzzles to dozvnright impossible conundrums. Tryyour best!
Challenge Problem Set 1: Anaphora
[Creative and Critical Thinking and Data Analysis; Challenge]
In this chapter, as an example of the scientific method, we looked
at the distribution of anaphora (nouns like himself, herself, etc.). We came
to the following conclusion about their distribution:
An anaphor must agree in person, gender, and number with its
antecedent.
However, there is much more to say about the distribution of these nouns (in
fact, chapter 5 of this book is entirely devoted to the question).
Part 1: Consider the data below. Can you make an addition to the above
statement that explains the distribution of anaphors and antecedents in the
very limited data below?
a) Geordi sang to himself.
b) *Himself sang to Geordi.
c) Betsy loves herself in blue leather.
d) *Blue leather shows herself that Betsy is pretty.
Part 2: Now consider thefollowing sentences:8
e) Everyone should be able to defend himself/herself/themselves.
f) I hope nobody will hurt themselves/himself/?herself.
Do these sentences obey your revised generalization? Why or why not?
Is there something special about the antecedents that forces an exception
here, or can you modify your generalization to fit these cases?
Challenge Problem Set 2: Yourself
[Creative and Critical Thinking; Challenge]
In the main body of the text we claimed that all anaphors need
an antecedent. Consider the following acceptable sentence. This kind
of sentence is called an "imperative" and is used to give orders.
a) Don't hit yourself!
8Thanks to Ahmad Lotfi for suggesting this part of the question.
32 Preliminaries
Part 1: Are all anaphors allowed in sentences like (a)? Which ones are
allowed there, and which ones aren't.
Part 2: Where is the antecedent for yourself? Is this a counter-example
to our rule? Why is this rule an exception? It is easy to add a stipulation
to our rule; but we'd rather have an explanatory rule. What is special
about the sentence in (a)?
Challenge Problem Set 3: Is Language Really Infinite?
[Creative and Critical Thinking; Extra Challenge]
[Note to instructors: this question requires some background either
in formal logic or mathematical proofs.]
In the text, it was claimed that because language is recursive, it follows that it
is infinite. (This was premise (i) of the discussion in section 4.3). The idea is
straightforward and at least intuitively correct: if you have some well-formed
sentence, and you have a rule that can embed it inside another structure;
then you can also take this new structure and embed it inside another and
so on and so on. Intuitively this leads to an infinitely large number of possible
sentences. Pullum and Scholz (2005) have shown that one formal version
of this intuitive idea is either circular or a contradiction.
Here is the structure of the traditional argument (paraphrased and
simplified from the version in Pullum and Scholz). This proof is cast in such
a way that the way we count the number of sentences is by comparing
the number of words in the sentence. If for any (extremely high) number
of words, we can find a longer sentence, then we know the set is infinite.
First some terminology:
> Terminology: call the set of well-formed sentences E. If a sentence x is
an element of this set we write E(x).
> Terminology: let us refer to the length of a sentence by counting the
number of words in it. The number of words in a sentence is expressed
by the variable n. There is a special measurement operation (function)
which counts the number of words, this is called ft. If the sentence
called x has 4 words in it then we say /<(x) = 4.
Next the formal argument:
Premise 1: There is at least one well-formed sentence that has more than
zero words in it.
3x[E(x) & p(x) > 0]
Premise 2: There is an operation in the PSRs such that any sentence may
be embedded in another with more words in it. That means for any
sentence in the language, there is another longer sentence. (If some
expression has the length n, then some other well-formed sentence
has a size greater than n).
Vn px[E(x) & p(x) = n]] -* [3y[E(y) &//(y) > n]]
Chapter 1: Generative Grammar 33
Conclusion: Therefore for every positive integer n, there are well-formed
sentences with a length longer than n (i.e., the set of well-formed
English expressions is at least countably infinite):
.-.Vfi|3y[E(K) &„(/)> fi]]
Pullum and Scholz claim that the problem with this argument lies with
the nature of the set E. Sets come of two kinds: there are finite sets which
have a fixed number of elements (e.g. the set {a, b, c, d} has 4 and exactly
4 members). There are also infinite sets, which have an endless possible
number of members (e.g., the set {a, b, c, ... } has an infinite number
of elements).
Question 1: Assume that E, the set of well-formed sentences, is finite. This is
a contradiction of one of the two premises given above. Which one? Why is it
a contradiction?
Question 2: Assume that E, the set of well-formed sentences, is infinite.
This leads to a circularity in the argument. What is the circularity (i.e., why is
the proof circular)?
Question 3: If the logical argument is either contradictory or circular
what does that make of our claim that the number of sentences possible
in a language is infinite? Is it totally wrong? What does the proof given
immediately above really prove?
Question 4: Given that E can be neither a finite nor an infinite set, is there
any way we might recast the premises, terminology, or conclusion in order
not to have a circular argument and capture the intuitive insight of the claim?
Explain how we might do this or why it's impossible. Try to be creative. There
is no "right" answer to this question. Hint: one might try a proof that proves
that a subset of the sentences of English is infinite (and by definition
the entire set of sentences in English is infinite) or one might try a proof
by contradiction.
Important notes:
1) Your answers can be given in English prose, you do not need to give a
formal mathematical answer.
2) Do not try to look up the answer in the papers cited above. That's just
cheating! Try to work out the answers for yourself.
Challenge Problem Set 4: Are Infinite Systems Really Unlearnable?
[Creative and Critical Thinking; Challenge]
In section 4.3, you saw the claim that iflanguage is an infinite system then it
must be unlearnable. In this problem set, you should aim a critical eye
at the premise that infinite systems can't be learned on the basis of the data
you hear.
34 Preliminaries
While given the extreme view in section 4.3 is logically true, consider the
following alternative possibilities:
a) We as humans have some kind of "cut off mechanism" that stops
considering new data after we've heard some threshold number of
examples. If we don't hear the crucial example after some period of time
we simply assume it doesn't exist. Rules simply can't exist that require
access to sentence types so rare that you don't hear them before the cut
off point.
b) We are purely statistical engines. Rare sentences types are simply
ignored as "statistical noise." We consider only those sentences
that are frequent in the input when constructing our rules.
c) Child-directed speech (motherese) is specially designed to give
you precisely the kinds of data you need to construct your rule system.
The child listens for very specific "triggers" or "cues" in the parental input
in order to determine the rules.
Question 1: To what extent are (a), (b) or (c) compatible with the hypothesis
of Universal Grammar. If(a), (b) or (c) turned out to be true, would this mean
that there was no innate grammar? Explain your answer.
Question 2: How might you experimentally or observationally distinguish
between (a), (b), (c) and the infinite input hypothesis of 4.3? What kinds
of evidence would you need to tell them apart?
Question 3: When people speak, they make errors. (They switch words
around, they mispronounce things, they use the wrong word, they stop midsentence
without completing what they are saying etc.) Nevertheless children
seem to be able to ignore these errors and still come up with the right set
of rules. Is this fact compatible with any of the infinite hypothesis, (a), (b),
or(c)?
Challenge Problem Set 5: Learning Parameters: Pro Drop
[Critical Thinking, Data Analysis; Challenge]
Background: Among the Indo-European languages there are two large
groups of languages that pattern differently with respect to whether
they require a pronoun (like he, she, it) in the subject position, or whether
such pronouns can be "dropped". For example, in both English and French,
pronouns are required. Sentences without them are usually ungrammatical:
a) He left
b) *Left
c) II est parti
he is gone
"he left"
d) *est parti
(French)
(French)
Chapter 1: Generative Grammar 35
In languages, such as Spanish and Italian, however, such pronouns are
routinely omitted (1s = first person, singular):
e) lo telephono (Italian)
I called.1s
"Icalled (phoned)"
f) telephono
called. 1s
"Called"
Question 1: Now imagine that you are a small child learning a language.
What kind of data would you need to know in order to tell if your language
was "pro drop" or not? (Hint. Does the English child hear sentences both with
and without subjects? Does the Italian child? Are they listening for sentences
with subjects or without them?
Question 2: Assume that one of the two possible settings for this parameter
(either your language is pro-drop or it is not) is the "default" setting.
This default setting is the version of the parameter one gets if one doesn't
hear the right kind of input. Which of the two possibilities is the default?
Question 3: English has imperative constructions such as:
g) Leave now!
Why doesn't the English child assume on the basis of such sentences that
English is pro-drop?
Parts of Speech
0. Words and Why They Matter to Syntax
It goes without saying that sentences are made up of words, so before we
get into the meat of this book, it's worth looking carefully at different kinds
of words.
What is most important to us here is the word's part of speech (also
known as syntactic category). The most common parts of speech are nouns,
verbs, adjectives, adverbs, and prepositions (we will also look at some other
less familiar parts of speech below). Parts of speech tell us how a word
is going to function in the sentence. Consider the sentences in (1).
Notice that we can substitute various words that are of the type noun for
the second word in the sentence:
1) a) The man loved peanut butter cookies.
b) The puppy loved peanut butter cookies.
c) The king loved peanut butter cookies.
However, we cannot substitute words that aren't nouns:1
2) a) *Thegreen loved peanut butter cookies.
b) *The in loved peanut butter cookies.
c) *Thesing loved peanut butter cookies.
Remember, the *symbol means that a sentence is syntactically ill-formed.
38 Preliminaries
The same holds true for larger groups of words (the square brackets [ ... ]
mark off the relevant groups of words).
3) a) IJohnJ went to the store.
b) [Theman] went to the store.
c) *IQuickly walksl went to the store.
4) a) [Norvell kissed the blarney stone.
b) *[To the washroom! kissed the blarney stone.
If we have categories for words that can appear in certain positions
and categories for those that don't we can make generalizations (scientific
ones) about the behavior of different word types. This is why we need parts
of speech in syntactic theory.
1. Determining Part of Speech
1.1 The Problem of Traditional Definitions
If you were taught any grammar in school, you may have been told that a
noun is a "person, place, or thing," or that a verb is "an action, state, or state
of being." Alas, this is a very over-simplistic way to characterize various
parts of speech. It also isn't terribly scientific or accurate. The first thing
to notice about definitions like this is that they are based on semantic criteria.
It doesn't take much effort to find counterexamples to these semantic
definitions. Consider the following:
5) The destruction of the city bothered the Mongols.
The meaning of destruction is not a "person, place, or thing." It is an action.
Bysemantic criteria, this word should be a verb. But in fact, native speakers
unanimously identify it as a noun. Similar cases are seen in (6):
6) a) Sincerity is an important quality.
b) The assassination of the president.
c) Tucson is a great place to live.
Sincerity is an attribute, a property normally associated with adjectives.
Yet in (6a), sincerity is a noun. Similarly in (6b) assassination, an action,
is functioning as a noun. (6c) is more subtle. The semantic property of
identifying a location is usually attributed to a preposition; in (6c) however,
the noun Tucson refers to a location, but isn't itself a preposition. It thus
seems difficult (if not impossible) to rigorously define the parts of speech
based solely on semantic criteria. This is made even clearer when we see
Chapter 2: Parts of Speech 39
that a word can change its part of speech depending upon where it appears
in a sentence:
7) a) Gabrielle's mother is an axe-murderer. (N)
b) Anteaters mother attractive offspring. (V)
c) Wendy's mother country is Iceland. (Adj)
The situation gets even muddier when we consider languages other than
English. Consider the following data from Warlpiri:
8) Wita-ngku ka maliki wajilipinyi.
small-SUBJ AUX dog chase.PRES
"The small (one) is chasing the dog."
In this sentence, we have a thing we'd normally call an adjective (the word
wita "small") functioning like a noun (e.g., taking subject marking). Is this
a noun or an adjective?
It's worth noting that some parts of speech don't lend themselves to
semantic definitions at all. Consider the sentence in (9). What is the meaning
of the word that?
9) Mikaela said that parts of speech intrigued her.
If parts of speech are based on the meaning of the word, how can we assign
a part ofspeech to word for which the meaningisn't clear.2
Perhaps the most striking evidence that we can't use semantic
definitions for parts of speech comes from the fact that you can know
the part of speech of a word without even knowing what it means:
10) The yinkish dripner blorked quastofically into the nindin with the
pidibs.
Every native speaker of English will tell you that yinkish is an adjective,
dripner a noun, blorked a verb, quastofically an adverb, and nindin and pidibs
both nouns, but they'd be very hard pressed to tell you what these words
actually mean. How then can you know the part of speech of a word without
knowing its meaning? The answer is simple: The definitions for the various
parts of speech are not semantically defined. Instead they depend on where
the words appear in the sentence and what kinds of affixes they take. Nouns
are things that that appear in "noun positions" and take "noun suffixes"
2Be careful here: the function of the word is clear (it is used to subordinate clauses
inside of sentences) but it doesn't have an obvious meaning with respect to the real
world.
40 Preliminaries
(endings). The same is true for verbs, adjectives, etc. Here are the criteria
that we used to determine the parts of speech in sentence (10):
a) yinkish between the and a noun
takes -ishadjective ending
b) dripner after an adjective (and the)
takes -er noun ending
subject of the sentence
c) blorked after subject noun
takes -edverb ending
d) quastofically after a verb
takes -ly adverb ending
e) nindin after the and after a preposition
0 pidibs after theand after a preposition
takes -s noun plural ending
The part of speech of a word is determined by its place in the sentence
and by its morphology, not by its meaning. In the next section, there is a list
of rules and distributional criteria that you can use to determine the part
of speech of a word.
1.2 Distributional Criteria
The criteria we use for determining part of speech then aren't based
on the meanings of the word, but on its distribution. We will use two kinds
of distributional tests for determining part of speech: morphological
distribution and syntactic distribution.
First we look at morphological distribution-, this refers to the kinds of
affixes (prefixes and suffixes) and other morphology that appear on a word.
Let's consider two different types of affixes. First, we have affixes that make
words out of other words. We call these affixes derivational morphemes.
These suffixes usually result in a particular part ofspeech. Forexample, if we
take the word distribute we can add the derivational suffix -(t)ion and we get
the noun distribution. The -(t)ion affix thus creates nouns. Any word ending
in -(t)ion is a noun. This is an example of a morphological distribution.
A similar example is found with the affix -al, which creates adjectives. If we
take distribution, and add -al to it, we get the adjective distributional. The -al
ending is a test for being an adjective. Derivational affixes make a word
a particular category; by contrast inflectional morphemes don't make a word
into a particular category, but instead only attach to certain categories.
Take for example the superlative suffix -est. This affix only attaches to words
that are already adjectives: big, biggest, (cf. dog, *doggest). Because they are
Chapter 2: Parts of Speech 41
sensitive to what category they attach to, inflectional suffixes can also serve
as a test for determining part of speech category.
A Warning: Homophony in English Affixes
Compared to other languages with much richer morphological systems,
English morphology is rather poor. In many cases, the same affix can be
used in very different ways. For example, the inflectional suffix -s is
found both as a marker of present tense in the third person, hewalks and
as the plural marker, peanuts. In fact, leaving aside the difference in
punctuation (the apostrophe), it is also used to mark possessors: ]olin's_
backpack; its cover. A similar effect is seen with many other suffixes. For
example -er is used both derivationally to form nouns: dancer, and as a
comparative inflectional marker on adjectives: bigger. Because so many
suffixes in Englishare homophonous (sound the same, but have different
usages), be very careful when using them for morphological distribution
tests.
The other kind of test we use for determining part of speech uses
syntactic distribution. Syntactic distribution refers to what other words
appear near the word. For example, in nouns typically appear after
determiners (articles) such as the, although they need not do so to be nouns.
We can thus take appearance after the to be a test for noun-hood.
Something to Think about: Circularity
In section 1 of this chapter, it was claimed that we needed parts of speech
to help us determine where in the sentence a word appeared. So for
example, we know that verbs and adjectives in English don't function as
thesubjects ofsentences. Above wehave given onetestfor part ofspeech
category in terms of the words distribution in the sentence. Here's
something to think about. Have we created a circular argument: category
determines position in the sentence and the position in the sentence
determines category? Is this really circular? Does it matter?
2. The Major Parts of Speech: N, V, Adj, and Adv
Having determined that we are going to use distributional criteria
for determining the part of speech of a word, we'll now turn to some tests
for particular lexical items. We'll limit ourselves to the majorclasses of noun
42 Preliminaries
(N), verb (V), adjective (Adj), and adverb (Adv). We'll look at other parts
of speech in later sections.
One thing that you'll notice is that these are specific to English.
Every language will have its own distributional criteria. For each language
linguists havetodevelop lists like theones below.3
A final word of qualification is in order, not every test will work in every
situation, so it is usually best to use multiple morphological and syntactic
tests for any given word if you can.
2.1 Nouns
Derivational Suffixes: In English, nouns often end in derivational endings
such as -ment (basement), -ness (friendliness), -ity (sincerity), -ty (certainty),
-(t)ion (devotion), -ation (expectation), -ist (specialist), -ant (attendant), -ery
(shrubbery), -ee(employee), -ship (hardship), -aire (billionaire), -acy(advocacy), -let
(piglet), -ling (underling), -hood (neighborhood), -ism(socialism), -ing (fencing).
Inflectional Suffixes: Nouns in English don't show much inflection,
but when pluralized can take suffixes such as -s (cats), -es (glasses), -en(oxen),
-ren (children), -i (cacti), -a (addenda).
Note that the following endings have homophonous usage
with other parts of speech: -ing, -s, 's, -er, -en.
Syntactic Distribution: Nouns often appear after determiners such as the,
those, these, (e.g., these peanuts) and can appear after adjectives (the big peanut).
Nouns can also follow prepositions (in school). All of these conditions can
happen together: in the big gymnasium). Nouns can appear as the subject of
the sentence (we will define subject rigorously in a later chapter): The syntax
paper was incomprehensible; or as the direct object: J read the syntax paper.
Nouns can be negated by no(as opposed to notor un-): Noapples were eaten.
One easy way to see if something is a noun is to see if you can replace
it with another word that is clearly a noun. So if we want to see if the word
people is a noun or not, we can substitute another word we know for sure
to be a noun, e.g., John (I saw people running all over the place vs.
/ saw John running all over theplace).
3The lists in this section are based on the discussions of English morphology found in
Katamba (2004) and Harley (2006).
Chapter 2: Parts ofSpeech 43
2.2 Verbs
Derivational Suffixes: Verbs often end in derivational endings such as -ate
(dissipate), and -izel-ise (regularize).
Inflectional Suffixes: In the past tense, verbs usually take an -ed or -t ending.
In the present tense, third person singular (he, she, it), they take the -s
ending. Verbs can also take an -ing ending in some aspectual constructions,
(she was walking) and take either an -en or an -ed suffix when they are
passivized (more on passivization in later chapters): the ice-cream was eaten.
Note that the following endings have homophonous usage with other
parts of speech: ate, -ing, -s, -er, -en, -ed.
Syntactic Distribution: Verbs can follow auxiliaries and modals such as will,
have, having, had, has, am, be, been, being, is, are, were, was, would, can, could,
shall, should, and the special infinitive marker to. Verbs follow subjects,
and can follow adverbs such as often and frequently. Verbs can be negated
with not (as opposed to no and un-4).
2.3 Adjectives
Derivational Suffixes: Adjectives often end in derivational endings
such as -ing (the dancing cat), -ive (indicative), -able (readable), -al (traditional),
-ate (intimate), -ish (childish), -some (tiresome), -G)an (reptilian), -ful (wishful),
-less (selfless), -ly (friendly).
Inflectional Suffixes: Adjectives can be inflected into a comparative form
using -er (alternately they follow the word more). They can also be inflected
into their superlative form using -est (alternately they follow the word most).
Adjectives are typically negated using the prefix un- (in its sense meaning
"not," not in its sense meaning "undo").
Note that the following affixes have homophonous usage with other
parts of speech: -ing, -er, -en, -ed, un-, -ly.
Syntactic Distribution: Adjectives can appear betweendeterminers such as
the, a, these etc. and nouns: (the big peanut). They also can follow the auxiliary
am/is/are/zvas/zoere/be/been/being (warning: this distribution overlaps with
verbs). Frequently, adjectives can be modified by the adverb very (warning:
this distribution overlaps with adverbs).
4There are verbs that begin with un-, but in these circumstances un-usually means
"reverse" not negation.
44 Preliminaries
You now have enough information to try ChallengeProblem Sets 1 & 2
Adjectives and Adverbs: Part of the Same Category?
Look carefully at the distributions of Adjectives and Adverbs. There is a
great deal of overlap between them. Adverbs typically take -ly; however,
there are also a number of clear adjectives that take this suffix too (e.g.,
thefriendly cub). Both Adj and Adv can be modified by the word very, and
they both have the same basic function in the grammar - to attribute
properties to the items they modify. In fact the only major distinction
between them is syntactic: Adjectives appear inside NPs, Adverbs appear
elsewhere. This kind of phenomenon is called Complementary
Distribution. (Where you get an adjective vs. an adverb is entirely
predictable.) When two elements are in complementary distribution in
linguistics, we normally think of them as variants of the same basic
category. For example, when two sounds in phonology are in
complementary distribution, we say they are allophones of the same
phoneme. We might extend this analysis to parts of speech: there is one
"supercategory" labeled "A" that has two subcategories in it (allo-partsof-speech
if you will): Adj and Adv. In this book we'll stick with the
traditional Adj and Adv categories, simply because they are familiar to
most people. Butyou should keep in mind that the category A (including
both Adjectives and Adverbs) might provide a better analysis and might
be better motivated scientifically.
2.4 Adverbs
Derivational Suffixes: Many adverbs end in -ly: quickly, frequently, etc.
Inflectional Suffixes: Adverbs generally don't take any inflectional suffixes.
However, on rare occasions they can be used comparatively and follow
the word more: She went more quickly than he did. Adverbstypically don't take
the prefix un- unless the adjective they are derived from does first
(e.g., unhelpfully from unhelpful, but *unquickly, *unquick).
Syntactic Distribution: The syntactic distribution of adverbs is most easily
described by stating where they can't appear. Adverbs can't appear between
a determiner and a noun (*the quicklyfox) or after the verb is and its variants.5
s In some prescriptive variants of English, there are a limited set of adverbs that can
appear after is. For example, well is prescriptively preferred over good, in such
constructions as / am well vs. / amgood(referring to your stale of being rather than the
Chapter 2: Parts of Speech 45
They can really appear pretty much anywhere else in the sentence, although
typically they either appear at the beginning or end of the clause/sentence.
Frequently, like adjectives, they can be modified by the adverb very.
You nowhaveenough information to answer General Problem Sets 1 & 2
3. Open vs. Closed; Lexical vs. Functional
3.1 Open vs. Closed Partsof Speech
Some parts of speech allow you to add neologisms (new words).
For example, imagine I invented a new tool especially for the purpose of
removing spines from cacti, and I called this tool a pulfice. This kind of word
is easily learned and adopted by speakers of English. In fact, we might even
predict that speakers would take pulfice and develop a verb pulficize, which
means to remove cacti spines using a pulfice. New words may be coined
at any time, if they are open class(e.g., fax, internet, grody). Bycontrast there
are some parts of speech that don't allow new forms. Suppose I wanted to
describe a situation where one arm is under the table and another is over the
table, and I called this new preposition uvder: My arms are uvder the table. It's
fairly unlikely that my new preposition, no matter how useful it is, will be
adopted into the language. Parts of speech that allow new members are said
to be open class. Those that don't (or where coinages are very rare) are
closed class. All of the cases that we've looked at so far have been open class
parts of speech.
3.2 Lexical vs. Functional
The open/closed distinction is similar to (but not identical to) another
useful distinction in parts of speech. This is the distinction between lexical
and functionalparts of speech. Lexical parts ofspeech provide the "content"
of the sentence. Nouns, verbs, adjectives and adverbs are all lexical parts
of speech. Functional parts of speech by contrast provide the grammatical
information. Functional items are the "glue" that holds a sentence together.
One way to tell if a lexical item is functional or lexical is to see if it is left
behind in "telegraphic speech" (that is, the way a telegram would be written;
e.g., Brian bring computer! Disaster looms!). Functional categories include
acceptability of your behavior). Most speakers of American English don't allow any
adverbs after is.
46 Preliminaries
Determiners, Prepositions, Complementizers, Conjunctions, Negation,
Auxiliaries and Modals. We will detail some of these below in section 3.3.
A Closed Lexical Subclass
You may have noticed that the open class and lexical class correspond to
exactly the same categories; similarly all of the cases of functional
categories I've mentioned are pretty clearly closed class items. So you
might be wondering why we have both the notions open and lexical and
both the notions closed and functional. There are two cases where we have
a mismatch between the terms: Pronouns and Anaphors. These are lexical
(they are a subtype of N), but they are closed classes.
You nowhave enough information toanswer Challenge Problem Set 3
3.3 Some Functional (Closed) Categories of English
We'll survey here some of the main functional categories of English. This list
is by no means complete. While it is possible to provide distributional
definitions for various functional parts of speech, because they are closed,
there are relatively few members of each class; so it's possible to simply list
most of them.
We'll start our categorization with Prepositions (abbreviated P).
Prepositions appear before nouns (or more precisely noun phrases). English
prepositions include the following:
12) Prepositions of English (P): to, from, under, over, with, by, at, above,
before, after, through, near, on, off, for, in, into, of, during, across,
without, since, until.
The class of determiners (D) is a little broader. It contains a number
of subcategories including articles, quantifiers, numerals, deictics, and
possessive pronouns. Determiners appear at the very beginning of English
noun phrases.
13) Determiners of English (D)
a) Articles: the, a, an
b) Deictic articles: this, that, these, those, yon
Chapter 2: Parts ofSpeech 47
c) Quantifiers6: every, some, many, most, few, all, each, any,less, fewer,
no
d) (Cardinal) numerals: one, two, three, four, etc.
e) Possessive pronouns7: my, your, his, her, its, our, their
f) Some wh-question words: which, whose
Conjunctions (Conj) are words that connect two or more phrases
together on an equal level:
14) Conjunctions ofEnglish (Conj):and, or, nor, neither ... nor, either ... or
The class of complementizers (C) also connectsstructures together, but they
embed one clause inside of another instead of keeping them on an equal
level:
15) Complementizers ofEnglish (C): that, for, if, whether
One of the most important categories that we'll use is the category
of Tense (T). For the moment we will not include tense suffixes such as -ed
and -s in this class, and treat those as parts of verbs (we will revisit this issue
in chapter 8). Instead the category T consists of auxiliaries, modals and
the non-finite clause marker. In the older syntactic literature, the category T
is sometimes called Infl (inflection) or Aux (Auxiliary). We'll use the
more modern T.
16) Tense categories ofEnglish (T)
Auxiliaries: have/has/had, am/is/are/was/were, do
Modals: will, would, shall, should, can, could
Non-finite Tense marker: to
There is one special category containing only one word: not which we'll
call negation (Neg). There are other categories that express negation (e.g.,
the determiners no, any, and the noun none). We'll reserve the category Neg
for the word not, however.
You nowhave enough information toanszver General Problem Sets4& 5 and
Challenge Problem Set 4
6 Not all quantifiers can be determiners, for example, the quantifiers lot and least
cannot function in this capacity (and are a noun and adjective respectively).
7The possessive forms mine, yours, hers, theirs, and ours are nouns, as are some uses of
his and its (when there is no other noun in the NP).
48 Preliminaries
3.4 Summary
This concludes our discussion of the major classes of words. We've looked
at the distributional criteria for the open/lexical categories of N, V, Adj,
and Adv, and we've listed the main functional/closed categories of P, D,
Conj, Neg, C, and T.
4. Subcategories and Features
Youmay have noticed that in sections2 and 3, I hinted that each majorpart
of speech category may have sub-types. For example, we listed six different
kinds of D (articles, deictics, quantifiers, numerals, possessive pronouns,
zu/i-pronouns) and three kinds of T (auxiliaries, modals, and the non-finite
marker). The technical term for these subtypes is subcategories. For the most
part, we are going to be interested in the main Parts of SpeechCategories (N,
V, Adj, Adv, P, D, Conj, C, T, and Neg), but sometimes we will want to refer
to the subcategories.
One way to mark subcategories is through the use offeatures. Consider
the case of T. To distinguish among the subcategories we can appeal
to the features [±modal] and ±non-finite]:
17) Auxiliary T|_modalf_nonfinilc|
Modal '|+modal,-nonfinite)
tO *• (-modal, +nonfinite]
There is, of course, one set of possible values of these features
which is missing ([+modal, +nonfinite]). We might similarly distinguish
among tense forms using features like [±past] etc.
Similarly we can distinguish among the various kinds of determiner
using features like [±wh], [±quantifier],[tdeictic], etc. The details of this kind
of analysis aren't crucial to the grammar fragments you are given in this
book, as long as you understand the basic concept behind using features
to mark subcategories. In the rest of this section, we look at some of the
subcategories of N, V and P that will be of use to us in the rest of the book.
I'm not going to discuss subcategories of Adj and Adv, although
they exist. In a grey textbox above, I've suggested that Adj and Adv
are themselves subcategories of a larger category A. We also find
many subcategories within the Adj and Adv categories. These distinctions
are explored in problem sets at the end of the chapter.
Chapter 2:Parts ofSpeech 49
4.1 Subcategories ofNouns
We can slice the pie of English nouns apart along several dimensions
including plural vs. singular, proper vs. common, pronoun vs. lexical noun,
and count vs. mass noun.
First let's distinguish along the line of plurality. English nouns can be
either singular or plural. The distinction between singular and plural
is usually morphologically marked with one of the plural endings (although
it need not be: mice, deer). Singular nouns in English require a D; plural ones
do not require a D, although they allow one:
18) a) *Cat ate the spider.
b) The cat ate the spider.
c) Cats ate the spider.
d) The Cats ate the spider.
We mark this distinction with the feature [±plural].
Closely related to the plural/singular distinction is the count vs. mass
noun distinction. Count nouns represent individual, "countable" elements.
For example, apple is a count noun. "Mass nouns" usually can't be counted
in the same way. For example sincerity and air are mass nouns. There are
two easy distributional tests to distinguish between mass and count nouns.
Mass nouns take the quantifier much, count nouns take the quantifier many.
a) many apples
b) *muchapples/apple1
c) *many sincerity
d) *many air
e) much sincerity
f) much air
Like plurals mass nouns generally don't require a determiner, but count
nouns do:
20) a) *Iate apple.
b) I ate the apple.
c) I ate sugar.
d) I ate the sugar.
e) He is filled with sincerity
8Many native speakers of English will be able to "force" a reading onto much apple.
But what they are doing is using apple as a mass noun (referring to the state of being
an apple or the totality of apples in the universe). It is often possible to force a mass
reading on count nouns, and a count reading on mass nouns (e.g. the water).
50 Preliminaries
f) I doubt his sincerity.
We distinguish between count and mass nouns using the feature [±count].
Next, let us distinguish between proper names and common nouns.
Proper names are nouns like Andrew Carnie. Common nouns are all
other nouns. For the most part proper names resist taking determiners:
21) a) Andrew Carnie
b) *the Andrew Carnie
There are some exceptions to this generalization. For example,
when referring to a family it's common to say the Smiths. In other languages,
proper names can take determiners. For example, in Spanish, it is perfectly
acceptable to say La Rosamaria "the Rosemary." If necessary, we
can distinguish proper names from common nouns using the feature
[icommon], although this feature is less useful than the others.
Finally let's look at the subcategories of pronouns and anaphors.
These classes differ from the others in that they are closed.
They never allow determiners or adjectival modification.
a) he
b) himself
c) *thehe
d) *the himself
e) *big he
f) *big himself
Pronouns belong to the class [+pronoun, -anaphor]. Anaphors are
[+pronoun, +anaphor]. All other nouns are [-pronoun, -anaphor]. For the
purposes of this book we are treating possessive pronouns as determiners,
not as a subcategory of nouns.
You nowhave enough information todoGeneral Problem Set 6
4.2 Subcategories of Verbs
There are really two major ways in which we can divide up verbs
into subcategories. One is along the lines of tense/finiteness (i.e., whether
the verb is left, leaves, (will) leave or (to) leave. We're going to leave
these distinctions aside until chapter 8, although hopefully it is obvious by
now how we'd use features to distinguish among them, even if the precise
features we'd use aren't defined yet. The other way to divvy up verbs is in
terms of the number of Noun Phrases (NPs) and Prepositional Phrases (PPs)
Chapter 2: Parts of Speech 51
or clauses (CPs) they require. This second kind of division is known
as argument structure.
In order to discuss argument structure, we first need to define
some basic terms. If you took grammar in school, you probably learned
that "every sentence has a subject and a predicate." Under your schoolroom
definitions, the subject is usually the first noun phrase (that is, the first noun
and all things that go along with it), and the predicate is everything else. So
for example, in (23) the subject is the dastardly phonologist, and the predicate
would be stole thesyntactician's lunch.
23) [The dastardly phonologist][stole the syntactician's lunch].
subject predicate (traditional definitions)
The definition of subject isn't too bad (we'll refine it later though), but
syntacticians use the term "predicate" entirely differently. The syntactician's
definition of predicate is based on the mathematical notion of a "relation."
The predicate defines the relation between the individuals being talked
about and the real world - as well as with each other. The entities (which
can be abstract) participating in the relation are called arguments. To see
how this works, look at the following example:
24) Gwen hit the baseball.
There are two arguments in this example, Gwen and the baseball. These
are elements in the world that are participants in the action described by the
sentence. The predicate here is hit. Hit expresses a relation between the
two arguments: more precisely, it indicates that the first argument (Gwen)
is applying some force to the second argument (the baseball). This may seem
patently self-evident, but it's important to understand what is going on here
on an abstract level. This usage of the terms predicate and argument
is identical to how they are used in formal logic.
We can speak about any particular predicate's argument structure.
This refers to the number of arguments that a particular predicate requires.
Another name for argument structure is valency. Take, for example,
predicates that take only one argument (i.e., they have a valency of 1). These
are predicates like smile, arrive, sit, run, etc. The property of transitivity refers
to how many arguments follow the verb. In predicates with a valency of 1,
no arguments follow the verb (the single argument precedes the verb),
so these predicates are said to be intransitive. Predicates that take two
obligatory arguments have a valency of 2; some examples are hit, love, see,
kiss, admire, etc. These predicates are said to be transitive, because they have
a single argument after the noun (the other argument precedes the verb).
Finally predicates that take three arguments have a valency of 3. Put and give
52 Preliminaries
are the best examples of this class. These predicates have two arguments
after the verb so are said to be ditransitive.
25)
Transitivity Valency Example
Intransitive 1 argument smile, arrive
Transitive 2 arguments hit, love, kiss
Ditransitive 3 arguments give, put
In determining how many arguments a predicate has, we only consider
the obligatory NPs and PPs. Optional ones are never counted in the list
of arguments. Only obligatory elements are considered arguments.
Did You Run the Race?
The claim that only obligatory arguments are found in argument structure
is not as straightforward as it sounds. Consider the verb run. It has both
an intransitive use (7 ran) and a transitive use (7 ran the race). A similar
problem is raised by languages that can drop the subject argument (e.g.
Spanish and Italian) and by imperative sentences in English (Go home
nowl). The subject is still an argument in these constructions, even though
you can't hear it. In situations like the verb run, we'll simply claim that
there are two verbs torun: one that takes an object and one that doesn't.
Predicates impose other restrictions on their arguments too.
Forexample, they also place restrictions on the categories of the things that
go with them. A verb like ask can take either an NP or a clause (embedded
sentence = CP) as a complement:
26) a) I asked [NI. the question].
b) I asked [CI. if you knew the answer].
But a verb like hit can only take an NP complement:
27) a) I hit [NT the ball].
b)*I hit [CI. that you knew the answer].
With these basics in mind, we can set up a series of features based on how
many and what kind of arguments a verb takes.
Let's start with intransitives. These require a single NP subject. We'll
mark this with the feature [NP ] where the underscore represents where
the verb would go in the sentence. An example of such a verb would be leave.
Most transitive verbs require an NP object, so we can mark these with
the feature [NP NP], an example of this is the verb hit, seen above in (27).
Chapter 2: Parts of Speech 53
Verbs like ask (see 26 above), think, say, etc. allow either an NP object
or a CP (embedded clause) object. We can mark this using curly brackets {}
and a slash. [NP/CP] means "a choice of NP or CP." So the feature structure
for predicates like this is [NP_ {NP/CP}].
Ditransitive verbs come of several major types. Some ditransitives
require two NP objects (the first is an indirect object the other a direct object).
The verb spare is of this category. It does not allow an NP and a PP:
28) a) I spared [NP him] [NP the trouble].
b) *Ispared [NP the trouble] [PP to him].
This category of ditransitive is marked with the feature [NP NP NP].
The opposite kind of ditransitive is found with the verb put. Put requires
an NP and a PP:
29) a) *Iput [NP the box] [NP the book],
b) I put [NP the book] [PPin the box].
This kind of ditransitive takes the feature [NP NP PP]. We also
have ditransitives that appear to be a combination of these two types
and allow either an NP or a PP in the second position:
30) a) I gave [NP the box] [PP to Leah],
b) I gave [NP Leah] [NP the box].
These have the feature [NP NP {NP/PP}]. Finally we have ditranstives
that take either two NPs, or one NP and one CP, or an NP and a PP:
31) a) I told [NP Daniel] [NP the story].
b) I told [NP Daniel] [CP that the exam was cancelled].
c) I told [NP the story] [PP to Daniel].
Verbs like tellhave the feature [NP _ NP {NP/PP/CP}].
The following chart summarizes all the different subcategories of verb
we've discussed here:
54
32)
Preliminaries
Subcategory Example
V(NP ](intransitive) Leave
V|NP NP) (transitive type 1) Hit
V[np_inp/cp)] (transitive type 2) Ask
V[NP NP NP| (ditransitive type 1) Spare
V[NP NPPPj (ditransitive type 2) Put
V[np_npinp/ppi] (ditransitive type 3) Give
V[np_np)np/pp/cpi] (ditransitive type 4) Tell
There are other types of verbs that we haven't listed here. We'll introduce
the features as we need them.
You can now tryGeneral Problem Set7and Challenge Problem Sets5 &6
5. Summary
In this chapter, we've surveyed the parts of speech categories that we will
use in this book. Wehave the Lexical parts ofspeech N, V,Adj, Adv, and the
functional categories D, P, C, Conj, Neg, and T. Determining part of speech
is done not by traditional semantic criteria, but by using morphological
and syntactic distribution tests. We also looked at distributional evidence
for various subcategories of Nouns and Verbs, and represented
these distinctions as feature notations on the major categories.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Parts of Speech (a.k.a. word class, syntactic categories): The labels
we give to constituents (N, V, Adj, Adv, D, P, C, T, Neg, Conj).
These determine the position of the word in the sentence
ii) Distribution: Parts of Speech are determined based on their
distribution. We have both morphological distribution (what affixes
are found on the word) and syntactic distribution (what other
words are nearby).
Chapter 2: Parts ofSpeech 55
iii) Complementary Distribution: When you have two categories
and they never appear in the same environment (context), you have
complementary distribution. Typically Complementary Distribution
means that the two categories are subtypes of a larger class.
iv) Parts of speech that are open class can take new members
or coinages: N, V, Adj, Adv.
v) Parts of speech that are closed class don't allow new coinages: D, P,
Conj, C, T, Neg, and the pronoun subcategory of N.
vi) Lexical Categories express the content of the sentence. N (including
pronouns), V, Adj,Adv.
vii) Functional Categories contain the grammatical information in
a sentence: D, P, Conj, T, Neg, C.
viii) Subcategories: The major parts of speech can often be divided up
into subtypes, these are called subcategories.
ix) Feature notations on major categories are a mechanism for
indicating subcategories.
x) Plurality refers to the number of nouns. It is usually indicated
in English with an -s suffix. Plural nouns in English do not require
a determiner.
xi) Count vs. Mass: Count nouns can appear with determiners
and the quantifier many. Mass nouns appear with much and usually
don't have articles.
xii) The predicate defines the relation between the individuals being
talked about and some fact about them - as well as relations among
the arguments.
xiii) Argument Structure: The number of arguments that a predicate
takes.
xiv) The arguments aretheentities whoareparticipating in thepredicate
relation.
xv) Intransitive: A predicate that takes only one argument.
xvi) Transitive: A predicate that takes two arguments.
xvii) Ditransitive: A predicate that takes three arguments.
56 Preliminaries
Further Reading
Grimshaw, Jane (1990) Argument Structure. Cambridge: MITPress.
Harley, Heidi (2006) English Words: A Linguistic Introduction. Oxford.
Blackwell.
Katamba, Francis (2004) English Words. New York:Routledge.
Levin, Beth (1993) English Verb Classes and Alternations: A Preliminary
Investigation. Chicago:University of Chicago Press.
Williams, Edwin (1983) Semantic vs. syntactic categories. Linguistics and
Philosophy 6,423-46.
General Problem Sets
1. Part of Speech 19
[Application of Skills; Basic]
Identify the main parts of speech (i.e., Nouns, Verbs, Adjectives/Adverbs,
and Prepositions) in the following sentences. Treat hyphenated words
as single words:
a) The old rusty pot-belly stove has been replaced.
b) The red-haired assistant put the vital documents through the new
efficient shredder.
c) The large evil leathery alligator complained to his aging keeper about his
extremely unattractive description.
d) I've just eaten the last piece of chocolate cake.
2. NOOTKA
[Application of Skills;Intermediate]
Consider the following data from Nootka (data from Sapir and Swadesh
1939), a language spoken in British Columbia, Canada and answer
the questions that follow the grey text box.
a) Mamu:k-ma qu:?as-?i.
working-PRES man-DEF
"The man is working."
Problem set contributed by Sheila Dooley-Collberg.
Chapter 2: Parts of Speech 57
b) Qu:Vas-ma mamu:k-?i.
man-PRES working-DEF
"The working one is a man."
(The : mark indicates a long vowel. ? is a glottal stop, pres in the second line
means "present tense," def means "definite determiner" (the).)
Reading Foreign Language Examples
There are three parts to most foreign language examples used in syntax.
Look at the sentences above. The first line is the sentence or phrase in the
language under consideration. The second line, which is the most
important for our purposes, contains a word-by-word translation of the
sentence. Finally, there is a colloquial English translation. The second
line,called the gloss, is the most useful if you don't speak the language.
It shows you the order of elements in the sentence. When reading about
thesyntax of foreign languages, concentrate on the order of elements in
this second line.
Questions about Nootka:
1) In sentence a, is Qu:?as functioning as a verb or a noun?
2) In sentence a, is Mamu:k functioning as a verb or a noun?
3) In sentence b, is Qt/;?as a verb or a noun?
4) In sentence b, is Mamu:k a verb or a noun?
5) What criteria did you use to tell what is a noun in Nootka and what is a
verb?
6) How does this data support the idea that there are no semantic criteria
involved in determining the part of speech?
3. Gender Neutral Pronouns
[Creative and Critical Thinking; Basic]
Most standard varieties of English don't have a gender-neutral singular
pronoun that can refer to humans (other than the veryawkward "one"). There
have been numerous attempts to introduce gender-neutral singular human
pronouns into English. The following list is a subset of the ones found
on John Chao's gender neutral pronoun FAQ :
tie, or, co, e, em, ems, en, es, et, ey, fm, ha, hann, he'er, lieesh, heir, hem, her'n,
lurim, herm, lies, hesh, heshe, hey, hez, hi, himer, hir, hirem, hires, hirm, liis'er,
his'n, lusher, Inzer, ho, horn, hse, hymer, im, ip, ir, iro, jhe, le, km, na, ne, net; nim,
http://www.aetherlumina.com/gnp/index.html
58 Preliminaries
on, per, po, rim, s/he, sap, se, sent, ser, sheehy, shem, shey, shim, sie, sim, ta, tern,
term, tey, thim, thon, uh,ve, vim,vir,vis, xe, z, ze, zie,zim, zir.
None of these have caught on. Instead, the otherwise plural they/them/
their/themselves is usually felt to be more natural by native speakers.
Why have the above forms not caught on, but instead we have co-opted
a plural pronoun for this usage?
4. Functional Categories
[Application of Skills; Basic]
The following is an extract from the preface to Captain Grose's Dictionary
of the Vulgar Tongue (1811) (from the open source Gutenberg project):
The propriety of introducing the university slang will be readily
admitted; it is not less curious than that of the College in the Old
Bailey, and is less generally understood. When the number and
accuracy of our additions are compared with the price of the volume,
we have no doubt that its editors will meet with the encouragement
that is due to learning, modesty, and virtue.
For every word in this paragraph identify its part of speech, and mark
whether part of speech is a lexical or functional part of speech and whether
the part of speech is open or closed.
5. Part of Speech 2
[Application of Skills; Intermediate]
Consider the following selection from Jabberwocky, a poem by Lewis Carroll
(From Through the Looking-Glass and What Alice Found There, 1872):
Twas brillig and the slithy toves
Did gyre and gimble in the wabe:
All mimsy were the borogoves,
And the mome raths outqrabe.
"Beware the Jabberwock, my son!
The jaws that bite, the claws that catch!
Beware the Jubiub bird, and shun
The frumious bandersnatch!"
He took his vorpal sword in hand:
Long time the manxome foe he sought So
rested he by the tumtum tree
And stood a while jn thought.
Chapter 2: Parts of Speech 59
And as in uffish thought he stood
The Jabberwock with eyes of flame,
Came whiffling through the tulgev wood,
and burbled as it came.
For each underlined word, indicate its part of speech (word class), and for
Ns, Vs, Adjs, Advs, explain the distributional criteria by which you came up
with that classification. If the item is a closed class part of speech, indicate
that. Do not try to use a dictionary. Most of these words are nonsense words.
You will need to figure out what part of speech they are based upon what
suffixes and prefixes they take, along with where they appear relative
to other words. Capitalization and punctuation should not be used as a guide
to part of speech.
6. Subcategories of Nouns
[Application of Knowledge; Basic]
For each of the nouns below put a + sign in the box under the features
that they have. Note that some nouns might have a plus value for more than
one feature. The first one is done for you. Do not mark the minus (-) values,
or the values for which the word is not specified; mark only the plus values!
Noun Plural Count Proper Pronoun Anaphor
Cats + +
Milk
New York
They
People
Language
Printer
Himself
Wind
Lightbulb
7. Subcategories of Verbs
[Application of Knowledge; intermediate]
For each of the verbs below, list whether they are intransitive, transitive
or ditransitive and list which features they take (see the list in (32) as
an example). In some cases they may allow more than one feature. E.g., the
verb eat is both [NP NP] and [NP ]. Give an example for each feature:
60 Preliminaries
spray, sleep, escape, throw, wipe, say, think, grudge, thank,
pour, send, promise, kiss, arrive
Challenge Problem Sets
Challenge Problem Set 1: -ian and -ish
[Critical and Creative Thinking; Challenge]
In the text we claimed that the suffixes -ian and -ish mark adjectives.
Consider the following sentences:
a) The Canadian government uses a parliamentary system of democracy.
b) The Canadian bought himself a barbeque.
c) The prudish linguist didn't enjoy looking at the internet.
d) We keep those censored copies of the book available to protect the
sensibilities of the prudish.
What should we make the words ending in -ish and -ian in sentences (b)
and (d)? Are they adjectives? If not, how can we account for the fact
that these words end in -ish and -ian? There are many possible answers
to this question.
Challenge Problem Set 2: Nominal Prenominal Modifiers11
[Critical and Creative Thinking; Challenge]
Part 1: By the syntactic criteria given to you in section, what part of speech
should the underlined words in the following examples be?
a) the leather couch
b) the water spout
Part 2: By contrast what do the following facts tell us about the parts
of speech of leather and water.
a) the leather
b) the water
c) ?the very leather couch (cf. the very red couch)
d) ?the very water spout (cf. the very big spout)
e) The more leather couch / *The leatherer couch (cf. the bigger couch)
f) *The more water spout
Thanks to Jack Martin for suggesting this problem set.
Chapter 2: Parts ofSpeech 61
Challenge Problem Set 3: Intensifiers
[Application of Knowledge; Challenge]
English has a subcategory of Adverbs called intensifiers. This class
includes very, rather, too (when used before an adjective), quite, less,
nearly, partly, fully, mostly, and sometimes.
Question 1: Is this subcategory an open class part of speech or a closed
class part of speech? Explain your answer.
Question 2: Describe the distribution of this subcategory. In particular
describe where it can appear relative to other adverbs (and adjectives).
Can other adverbs appear in this environment?
Challenge Problem Set 4: Complementary Distribution
[Critical Thinking; Challenge]
In a grey textbox in section 2.4, it's argued that Adjectives and Adverbs
are in complementary distribution and thus might be part of the same supercategory
A. Are N and V in complementary distribution? What about Adv
and V? What about N and Adj? Create examples to show whether these
categories are in complementary distribution. If any are in complementary
distribution with the others what does this tell us about the parts of speech?
Next consider whether any functional categories are in complementary
distribution with lexical categories.
Challenge Problem Set 5: Subcategories of Adverbs
[Application of Skillsand Knowledge; Challenge]
Yourgoal inthis problem set is to set up a set of subcategories for Adverbs.
Consider the following adverbs:
luckily, earnestly, intently, hopefully, probably, certainly, frequently,
patiently, always, completely, almost, again, evidently, frankly,
demandingly, yesterday, necessarily
Part 1: For each adverb determine:
1) Can it appear before the subject? (e.g., Unbelievably. I don'tknow any
pixies.)
2) Can itappear between the T (e.g., will, have, is, can, etc.) and the verb?
(e.g., / have often wondered about the existence ofpixies.)
3) Can it appear after the object? Or at the end of the sentence? (e.g.,
Pixies eat mushrooms vigorously.)
4) Can itappear between an object and a PP in a ditransitive (e.g., I put the
book carefully on the table.)
(Note, these adverbs may appear inseveral of these positions.)
62 Preliminaries
Part 2: Group the adverbs together into subcategories based on your
answers to part 1.
Part 3: Within each group you may find more subtle orderings. For example,
within the subcategory of adverbs that can appear between auxiliaries
and verbs there may be an ordering of adverbs. Try putting multiple adverbs
in each position. What are the orderings you find?
Challenge Problem Set 6: Subcategories of Adjectives
[Application of Knowledge; Challenge]
Just as there are positional differences among adverbs (see Challenge set
3), we find an ordering of adjectives with respect to each other. Below is a list
of adjectives. Pair each adjective with every other adjective and see which
must come first in a noun phrase. Try to come up with a general ordering
among these adjectives. (Although in the text I've told you to include
numerals with the class of determiners, I've listed them here as adjectives,
for the rest of the book treat them as determiners.)
deep, big, young, blue, desperate, two, scaly, thick
One word of caution: it is sometimes possible to put some adjectives in any
order. However, many of these orders are only possible if you are using
the adjective contrastively or emphatically. For example, you can say the
old rubber sneaker with a normal non-contrastive meaning, but the rubber
old sneaker is only possible when it has a contrastive emphatic meaning
{the RUBBER old sneaker as opposed to the leather one). Don't let
these contrastive readings interfere with yoursubcategorization.
Constituency, Trees,
nfeawBii 3.nQ iviiies
0. Introduction
Syntaxis about the study of sentencestructure. So let's start by defining what
we mean by "structure." Consider the sentence in (1):
1) The students loved their syntax assignments.
One way to describe this sentence is as a simple linear string of words.
Certainly this is how it is represented on the page. We could describe
the sentence as consisting of the words the, students, loved, their, syntax,
assignments in that order. As you can probably figure out, if that were all
there was tosyntax, you couldput down thisbookhere and notbother with
the next fourteen chapters. Butthat isn't all there is to syntax. Thestatement
that sentence (1) consists of a linear string of words misses several important
generalizations about the internal structure of sentences and how
these structures are represented in our minds. In point of fact we are going
to claim that the words in sentence (1) are grouped into units (called
constituents) and that these constituents are grouped into larger
constituents, and so on until you get a sentence.
Notice that on a purely intuitive level there is some notion that certain
words are more closely related to one another. For example, the word the
seems to be tied more to the meaning of students than it is to loved or syntax.
A related intuition can be seen by looking at the sentences in (2).
64 Preliminaries
2) a) The student loved his phonology readings,
b) The student hated his morphology professor.
Compare these sentences to (1). You'll see right away that the relationship
between the student and his syntax assignments in (1) and the student and
his phonology readings in (2a) is the same. Similarly, the relation between
the student and his morphology professor in (2b), while of a different kind
(hating instead of loving), is of a similar type: There is one entity (the
student) who are either hating or loving another entity (his syntax assignments,
his phonology readings or his morphology professor). In order to capture these
intuitions (the intuition that certain words are more closely connected than
others, and the intuitions about relationships between words in the
sentence), we need a more complex notion. The notions we use to capture
these intuitions are constituency and hierarchical structure. The notion that
the and studetit are closely related to one another is captured by the fact
that we treat them as part of a bigger unit that contains them, but not
other words. We have two different ways to represent this bigger unit.
One of them is to put square brackets around units:
3) [the student]
The other is to represent the units with a group of lines called a tree
structure:
4) / \
the student
These bigger units are called constituents. An informal definition for
a constituent is given in (5):
5) Constituent: A group of words that functions together as a unit.
Constituency is the most important and basic notion in syntactic theory.
Constituents form the backbone of the rest of this book. They capture the
intuitions mentioned above. The "relatedness" is captured by membership
in a constituent. As we will see it also allows us to capture the relationships
between constituents exemplified in (1).
Constituents don't float out in space. Instead they are embedded one
inside another to form larger and larger constituents. This is hierarchical
structure. Foreshadowing the discussion below a bit, here is the structure
we'll develop for (1):
6)
65
AdjP N
j assignments
A
syntax
This is a typical hierarchical tree structure. The sentence constituent
(represented by the symbol TP) consists of two constituents: a subject noun
phrase (NP) Ithestudent] and a predicate or verb phrase (VP) [love his syntax
assignments}. The subject NP in turn contains a noun (N) student and
a determiner (or article) (D) the. Similarly the VP contains a verb (V), and an
object NP [his syntaxassignments]. The object NP is further broken down into
three bits: a determiner his, an adjective syntax, and a noun assignments.
As you can see this tree has constituents (each represented by the point
where lines come together) which are inside other constituents. This is
hierarchical structure. Hierarchical constituent structure can also be
represented with brackets. Each pair of brackets ([ ]) represents a constituent.
We normally put the label of the constituent on the left member of the pair.
The bracketed diagram for (6) is given in (7):
The Psychological Reality of Constituency
In the 1960s, Merrill Garrett and his colleagues showed that constituency
has some reality in the minds of speakers. The researchers developed a
seriesofexperiments that involved placing a click in a neutral place in the
stream of sounds. People tend to perceive these clicks not in the place
where they actually occur, but at the edges of constituents. The italicized
strings of words in the following sentences differ only in how the
constituents are arranged.
i) [Inher hope ofmarrying] An/na was impractical.
ii) [Harry's hope of mam/inf An/na] was impractical.
Syntactic constituency is marked with square brackets [ ]; the placement
of the click is marked with a slash /. People perceive the click in different
places (marked with a Is) in the two sentences, corresponding to the
constituent boundaries - even though the click actually appears in the
same place in each sentence (in the middle of the word Anna).
66 Preliminaries
7)[Tp[Np[DThe][Nstudent]][Vp[vloved][NP[Dhis][Adjp[Adjsyntax]][Nassignments]]]].
As you can see, bracketed diagrams are much harder to read, so for the most
part we will use tree diagrams in this book. However, sometimes bracketed
diagrams have their uses, so you should be able to translate back and forth
between trees and bracketed diagrams.
1. Rules and Trees
Now we have the tools necessary to develop a simple theory of sentence
structure. We have a notion of constituent, which is a group of words
that functions as a unit, and we have labels (parts of speech) that we can use
to describe the parts of those units. Let's put the two of these together
and try to develop a description of a possible English sentence. In generative
grammar, generalizations about structure are represented by rules. These
rules are said to "generate" the tree. So if we draw a tree a particular way,
we need a rule to generate that tree. The rules we are going to consider
in this chapter are calledphrase structure rules (PSRs) because they generate
the phrase structure tree of a sentence.
1.1 Noun Phrases (NPs)
Let's start with the constituents we callnoun phrases (or NPs) and explore
the range of material that can appear in them. The simplest NPs contain
only a noun (usually a proper noun [+proper], pronoun [+pron], mass noun
[-count] or a plural noun [+plural]):
8) a) John b) water c) cats
Our rule must minimally generate NPs then that contain only an N.
The format for PSRs is shown in (9a), we use X, Y, and Z here as variables
to stand for any category. (9b)shows our first pass at an NP rule:
9) a) XP -> XYZ
t t f
the label "consists of" the elements that make up
for theconstituent the constituent
b) NP->N
This rule says that an NP is composed of (written as ->) an N. This rule
would generate a tree like (10):
Chapter 3: Constituency 67
10) NP
I
N
There are many NPs (e.g., those that are [+count]) that are more complex
than this of course:
11) a) the box
b) his binder
c) that pink fluffy cushion
We must revise our rule to account for the presence of determiners:
12) a) NP -» D N
This generates a tree like:
b) NP
D N
the box
Compare the NPs in (8) and (11): You'll see that determiners are optional.
As such we must indicate their optionality in the rule. We do this
with parentheses ( ) around the optional elements:
13) NP -» (D) N
Nouns can also be optionally modified by adjectives, so we will need
to revise our rule as in (14) (don't worry about the "P" in AdjP yet,
we'll explain that below).
14) a) the big box b) his yellow binder
15) NP -> (D) (AdjP) N
Nouns can also take prepositional phrase (PP) modifiers (see below where
we discuss the structure of these constituents), so once again we'll have to
revise our rule:
16) a) the big box of crayons
b) his yellow binder with the red stripe
17) NP -> (D) (AdjP) N (PP)
For concreteness, let's apply the rule in (17):
68 Preliminaries
18)
big of poems
The NP constituent in (18) consists of four subconstihaents: D, AdjP, N
and PP.
For the moment, we need to make one more major revision to our NP
rule. It turns out that you can have more than one adjective and more
than one PP in an English NP:
19) The [AdjP big] [AdjP yellow] box [PP ofcookies] [PP with the pink lid].
In this NP, the noun box is modified by big, yellow, ofcookies, and with the pink
lid. The rule must be changed then to account for this. It must allow more
thanoneadjective and more thanonePPmodifier. We indicate thiswith a +,
which means "repeat this category as many times as needed":
20) NP -> (D) (AdjP+) N (PP+)
Wewillhave causeto slightly revise this rule in later sections of this chapter
and laterchapters, but fornowwecanusethisruleasa working hypothesis.
You now have enough information to try Challenge Problem Set 1
1.2 Adjective Phrases (AdjPs) and Adverb Phrases (AdvPs)
Consider the following two NPs:
21) a) the big yellow book
b) the very yellow book
On the surface, these two NPs look very similar. They both consist of a
determiner, followed by two modifiers and then a noun. But consider what
modifies what in these NPs. In (21a) big modifiesbook, as does yellow. In (21b)
on the other hand onlyyellow modifies book;very does not modifybook (*very
book) - it modifiesyelloiv. On an intuitive level then, the structures of these
two phrases are actually quitedifferent. (21a) has two adjective constituents
that modify the N, whereas (21b) has onlyone [very yellow]. Thisconstituent
1We use a triangle here to obscure the detailsof the PP and AdjP. Studentsshould
avoid using triangles when drawing trees, as you want to be as explicit as possible.
I use it here only to draw attention to other aspectsof the structure.
Chapter 3: Constituency 69
iscalled an adjective phrase (AdjP). Therule forthe adjective phrase isgiven
in (22a):
22) a) AdjP -> (AdvP) Adj
b) AdjP
AdvP
I
Adv
very
Adj
yellow
This will give us the following structures for the two NPs in (21):
23) a) NP
b)
D
the
AdjP AdvP N
| | book
Adj Adj
big yellow
NP
D
the
AdjP
AdvP Adj
| yellow
Adv
very
book
So despite their surface similarity, these two NPs have radically different
structures. In (23a) the N is modified by two AdjPs, in (23b) by only one.
This leads us to an important observation about tree structures:
24) Principle ofModification (informal): Modifiers are always attached within
the phrase they modify.
The adverb very modifies yellow, so it is part of the yellow AdjP in (23b).
In (23a) by contrast, big doesn't modify yellow, it modifies book, so it is
attached directly to the NP containing book.
A very similar rule is used to introduce AdvPs:
25) AdvP -> (AdvP) Adv
26) very quickly
70 Preliminaries
27) AdvP
AdvP Adv
| quickly
Adv
very
Here is a common mistake to avoid: Notice that the AdvP rule specifies that
its modifier is another AdvP: AdvP -»(AdvP) Adv. The rule does NOT say
*AdvP —»(Adv) Adv, so you will never get trees of the form shown in (28):
28) .^SdvP
You might find the tree in (27) a little confusing. There are two Advs
and two AdvPs. In order to understand that tree a little better, let's introduce
a new concept: heads. We'll spend much more time on heads in chapters 6
and 7, but here's a first pass: The head of a phrase is the word that gives
the phrase its category. For example, the head of the NP is the N, the head
of a PP is the P, the head of the AdjP is Adj and the head of an AdvP is Adv.
Let'slook first at an adjective phrase (29a) and compare it toa complex AdvP:
29) a) AdjP<-^ b) AdvP^
^/^"^^^^ head ^^^^*\\ ^ea^
AdvP Adj AdvP Aclv
head C \ yellow head r | quickly
Adv Adv
very very
In (29a), the heads should be clear. The adverb very is the head of the adverb
phrase and the adjective yellow is the head of AdjP. In (29b) we have the
same kind of headedness, except both elements are adverbs. Very is the head
of the lower AdvP, and quickly is the head of the higher one. We have two
adverbs, so we have two AdvPs - each has their own head.
With this in mind, we can explain why the "very" AdvP is embedded
in the AdjP. Above we gave a very informal description of the principle
of modification. Let's try for a more precise version here:
30) Principle ofModification (revised): If an XP (that is, a phrase with some
category X) modifies some head Y, then XP must be a sister to Y (i.e.,
a daughter of YP).
Chapter 3: Constituency 71
31) AdjP = YP ^_ Mother
AdvP =XP Adj =Y< ^Daughters ofAdjP
Adv Sisters to each other
The diagram in (31) shows you the relations mentioned in the definition
in (30). If we take the AdjP to be the mother then its daughters are the AdvP
and the head Adj. Since AdvP and Adj are both daughters of the same
mother then we say they are sisters. In (30) X and Y are variables that stand
for any category. If one phrase, XP (AdvP) modifies some head Y(Adj), then
the XP must be a sister to Y (i.e., the AdvP must be a sister to the head Adj),
meaning they must share a mother. You'll notice that this relationship
is asymmetric: AdvP modifies Adj, but Adj does not modify AdvP.
You nowhave enough information to try General Problem Set 1
1.3 Prepositional Phrases (PPs)
The next major kind of constituent we consider is the prepositional phrase
(PP).Most PPs take the form of a preposition (the head) followed by an NP:
32) a) [PP to [NP the store]]
b) [PP with [NP an axe]]
c) [PP behind [NP the rubber tree]]
The PP rule appears to be:
33) a) PP -> P NP
b) PP
P NP
with / \
D N
an axe
In the rule we've given the NP in the PP is obligatory. There may
actually be some evidence for treating the NP in PPs as optional. There is
a class of prepositions, traditionally called particles, that don't require
a following NP:
34) a) I haven't seen him before.
b) I blew it up.
c) I threw the garbage out.
72 Preliminaries
If these are prepositions, then it appears as if the NP in the PP rule is
optional:
35) PP -> P (NP)
Even though all these particles look similar to prepositions (or are at least
homophonous with them), there is some debate about whether they are
or not. As an exercise you might try to think about the kinds of phenomena
that would distinguish particles from prepositions without NPs.
You nowhaveenough information to tryGeneral Problem Set2
Ifyou read theAppendix, you should beable todoGeneral Problem Set3
1.4 Verb Phrases (VPs)
Next we have the category headed by the verb: the verb phrase (VP).
Minimally a VP consists of a single verb. This is the case of intransitives
(V,np_i):
36) a) VP -> V
b) Ignacious [VP left].
c) VP
I
V
left
Verbs maybe modified by adverbs(AdvPs), whichare, ofcourse, optional:
37) a) Ignacious[VP left quickly].
b) VP->V(AdvP)
c) VP
V AdvP
left |
Adv
quickly
Interestingly, many of these adverbs can appear on either side of the V,
and you can have as many AdvPs as you like:
38) a) Ignacious [VP quickly left].
b) Ignacious [VP [AdvP deliberately] [AdvP always] left [AdvP quietly]
[AdvP early]].
c) VP -> (AdvP+) V (AdvP+)
39)
AdvP
Adv Adv
deliberately always
Chapter 3: Constituency
AdvP
Adv Adv
quietly early
73
You'll recall from chapter 2 that there is a subcategory of verbs that can take
an NP object (the transitive V|NP _ NP)); these NPs appear immediately after
the V and before any AdvPs:
40) a) VP -> (AP+) V (NP) (AP+)
b) Bill [VP frequently kissed his mother-in-law],
c) Bill [VP kissed hismother-in-law quietly], (cf.*Bill [VP kissed quietly his
mother-in-law].)
41) VP
V NP AdvP
kissed
^ \ 1
D N A
his mother-in-law quietly
You now have enough information to try Challenge Problem Set 2
It is also possible to have two NPs in a sentence, for example
with a double objectverb like spare (V[NP_ NPNP]). Boththese NPs must come
between the verb and any AdvPs:
42) I spared[NPthe student] [NP anyembarrassment] [AdvP yesterday].
Note, you are allowed to have a maximum of only two argument NPs.
For this reason, we are not going to use the kleene plus (+) which entails
that you can have as many as you like. Instead we are going to simply list
both NPs in the rule:
43) a) VP -> (AP+) V (NP) (NP) (AP+)
b) VP
V
spare
NP NP AdvP
D N D N Adv
the student any embarrassment yesterday
74 Preliminaries
Verbs can be modified by PPs as well. These PPs can be arguments as in
ditransitive verbs of the type V[NP _ NP PP] (e.g., the PP argument of the verb
put) or they can be simple modifiers PP like for a dollar below. These PPs
can appear either after an adverb or before it.
44) a) Bill[VPfrequently got his buckets [Prfrom the store ] [?Pfor a dollar]].
b) VP -> (AdvP+) V (NP) (NP) (AdvP+) (PP+) (AdvP+)
c) VP
AdvP V NP PP PP
1 got /\ /\ y\
Adv D N P NP P NP
lently his buckets from /\ for /\
D N D N
the store a dollar
The rule in (44b) is nearly our final VP rule for this chapter; we'll need to
make one further adjustment to it once we look at the structure of clauses.
1.5 Clauses
Thus far, we have NPs, VPs, APs, and PPs, and we've seen how they can
be hierarchically organized with respect to one another. One thing that we
have not accounted for is the structure of the sentence (or more accurately
clause).2 A clause consists of a subject NP and a VP. The label we use
for clause is TP.3
45) [tp[np Bill ] [VP frequently got his buckets from the store for a dollar]].
This can be represented by the rule in (46):
46) TP -> NP VP
A tree for (45) is given in (47):
*We'll give a proper definition for clause in a later chapter.
' Inotherbooks youmightfind sentences labeled asSor IP. Sand IPareessentially
the same thing as TP. We'll use TP here since it will make the transition to X-bar
theory (in chapter 6) a little easier.
47)
Chapter 3: Constituency 75
N _ -^^ \ '—
Bill AdvP V NP PP PP
1 got /\ /\ / \
A D N P NP P NP
frequently his buckets from / \ for /\
D N D N
the store a dollar
TPs can also include other items, including unsurprisingly elements of the
category T (such as modal verbs and auxiliary verbs) like those in (48):
48) a) Cedric might crash the longboat,
b) Gustaf has crashed the semi-truck.
It may surprise you that we won't treat these as verbs, the reason for this
will become clear in later chapters. Note that the T in the TP is optional.
49) TP -> NP (T) VP
A tree showing the application of this rule is given in (50):
50) TP
NP T VP
| might /\
N V NP
Cedric crash / \
D N
the longboat
Clausesdon't always have to stand on their own. There are times when one
clause is embedded inside another:
51) [TP Shawn said [TP he decked the janitor]].
In sentence (51) the clause he decked the janitor lies inside the larger main
clause. Often embedded clauses are introduced by a complementizer
like that or //:
52) [tpShawn said [CP [c that ] [TP he decked the janitor]]].
We need a special rule to introduce complementizers (C):
76
53) a) CP -> (C) TP
b) TP
N V
Shawn said
Preliminaries
C
that
N V
he decked
D N
the Janitor
For the moment we will assume that all embedded clauses are CPs, whether
or not they have a complementizer. We'll show evidence for this in chapter 7.
This means that a sentence like Shawn said he decked the janitor will have a CP
in it even though there is no complementizer that.
54) TP
N V
Shawn said
NP
TP
VP
N V NP
he decked
D N
the Janitor
Embedded clauses appear in a variety of positions. In (54),
the embedded clause appears in essentially the same slot as the direct object.
Embedded clauses can also appear in subject position:
55) [TP [CP That he decked the janitor] worried Jeff].
Because of this we are going to have to modify our TP and VP rules to allow
embedded clauses. Syntacticians use curly brackets { } to indicate a choice.
Chapter 3: Constituency 77
So {NP/CP} means that you are allowed either an NP or an CP but not both.
The Modification to the TP rule is relatively straightforward. We simply
allow the choice between an NP and a CP in the initial NP:
56) a)
b)
TP-> {NP/CP} (T)VP
You now have enough information to try Challenge Problem Set 3
TherevisedVPrule requires a little morefinesse. Firstobserve that in verbs
that allow both an NP and a CP (V|NP_ |NP/Cp|] such as ask), the CP follows
the NP but precedes the PP (in the following sentence yesterday and
over the phone should be interpreted as modifying ask, not ate), essentially
in the position of the second NP in the rule:
57) Naomi asked [NP Erin] [CP if [TP Dan ate her Kung-Pao chicken]] yesterday
over the phone.
This gives us the rule :
58) a) VP-> (AdvP+) V (NP) ({NP/CP}) (AdvP+) (PP+)(AdvP+)
78
NP
N V
Naomi asked
N
Erin
Preliminaries
Adj P
yesterday over
D AdjP N
her | chicken
Adj
Kung-Pao
This rule is by no means perfect. There is no way to draw the tree for
sentenceswhere an AdvP canappear before the CP (Naomi asked Erin quietly
if Dan ate her KungPao Chicken). We don't want to add an optional AdjP
before the ({CP/NP}) in the rule because AdvPs cannot appear before
the NP. For the moment, we'll go with the VPrule as it is written, and return
to questions like this later, although we return to the issue in chapter 6.
The last revision we have to make to our PSRs is to add the CP as
a modifier to NPs to account for cases like:
59) a) [NP The fact about Bill[CP that he likes icecream]] bothers Natasha.
b) NP -> (D) (AdjP+) N (PP+) (CP)
c) TP
icecream
Chapter 3: Constituency 79
You nowhave enough information totryGeneral Problem Set4.
Ifyou read the Appendix youwillhave enough information to do
General Problem Sets 5,6 & 7.
Relative Clauses
In addition to the CPs that modify Ns as in the above cases, there is
another kind of CP modifier to an N. These are called relative clauses. We
aren't going to include relative clauses in our rules yet. This is because
they often contain what is called a "gap" or a place where some part of the
clause is missing. For example:
i) The man [whose car I hit last week] sued me.
The underscore in the sentence indicates where the gap is - the object of
the verb hit is in the wrong place, it should be where the underscore is.
Corresponding to the gap we also have the zvh-word whose and the noun
car. These are appearing at the beginning of the clause. Because of these
gaps and fronted ic/i-elements, we aren't going to worry about the internal
structure of these clauses.
Here's a challenge: relative clauses actually appear in a different
position than the CPs that follow nouns like the fact. Can you figure out
what the difference is? (Hint: it has to do with the relative position of the
CP and the PP in the NP rule.)
1.6 Summary
In this section we've been looking at the PSRs needed to generate trees that
account for English sentences. As we'll see in later chapters, this is nothing
but a first pass at a very complex set of data. It is probably worth repeating
the final form of each of the rules here:
60) a) CP -> (C) TP
b) TP-> (NP/CP) (T)VP
c) VP -> (AdvP+) V (NP)({NP/CP)) (AdvP+) (PP+) (AdvP+)
d) NP -> (D) (AdjP+) N (PP+) (CP)
e) PP^P(NP)
0 AdjP -> (AdvP) Adj
g) AdvP -> (AdvP) Adv
so Preliminaries
Recursion
The rules we have written here have a very important property. Notice
the following thing: The TP rule has a VP under it. Similarly, the VP rule
can take a CP under it, and the CP takes a TP. This means that the three
rules can form a loop and repeat endlessly:
i) Fred said that Mary believes that Susan wants that Peter desires
that... etc.
This property, called recursion, accounts partially for the infinite nature of
human language. Because you get these endless loops, it is possible to
generate sentences that have never been heard before. This simple
property of these rules thus at least partly explains the creativity of
human language, which in itself is a remarkable result.
These rules account for a wide variety of English sentences. A sentence using
each of these rules is shown in (61):
61) The big man from NY has often said that he gave peanuts to elephants.
TP
NP T
has
D AdjP N PP
The man / \
A P NP
big from
N
NY
VP
AdvP V
I said
A
often
N V NP
he gave !
N P NP
peanuts to \
N
elephants
This is by no means the only tree that can be drawn by these rules.
PP
Chapter 3: Constituency 81
2. How to Draw a Tree
Younow have the tools you need to start drawing trees. You have the rules,
and you have the parts of speech. I suspect that you'll find drawing trees
much more difficult than you expect. It takes a lot of practice to know which
rules to apply and apply them consistently and accurately to a sentence.
You won't be able to draw trees easily until you literally do dozens of them.
Drawing syntactic trees is a learned skill that needs lots of practice,
just like learning to play the piano.
There are actually two ways to go about drawing a tree. You can start
at the bottom and work your way up to the TP, or you can start with the TP
and work your way down. Which technique you use depends upon your
individual style. For most people who are just starting out, starting at
the bottom of the tree with the words works best. When you become more
practiced and experienced you may find starting at the top quicker. Below,
I give step-by-step instructions for both of these techniques.
2.1 Bottom-up Trees
This method for tree drawing often works best for beginners. Here are some
(hopefully helpful) steps to go through when drawing trees.
1. Write out the sentence and identify the parts of speech:
D Adv Adj N V D N
The very small boy kissed the platypus.
2. Identify what modifies what. Remember the modification relations. If
the word modifies something then it is contained in the same constituent
as that thing.
Very modifiessmall. Very small modifiesboy.
The modifies boy. The modifies platypus.
The platypus modifies kissed.
3. Start linking together items that modify one another. It often helps to
start at the right edge. Always start with adjacent words. If the modifier
is modifying a noun, then the rule you must apply is the NP rule:
NP
/\
D Adv Adj N V D N
The very small boy kissed the platypus.
82 Preliminaries
Similarly if the word that is being modified is an adjective, then you
must apply the AdjP rule:
AdjP
NP
/\
D Adv Adj N V D N
The very small boy kissed the platypus.
Make sure you apply the rule exactly as it is written. For example
the AdjP rule reads AdjP -» (AdvP) Adj. This means that the Adv
must have an AdvP on top of it before it can combine with the Adj.
Keep applying the rules until you have attached all the modifiers to the
modified constituents. Apply one rule at a time. Work from right to left
(from the end of the sentence to the beginning.) Try doing the rules
in the following order:
a) AdjPs & AdvPs b) NPs & PPs
c) VPs d) TP
e) If your sentence has more than one clause in it, start with the most
embedded clause.
NP
/\
D Adv Adj N V D N
The very small boy kissed the platypus.
NP
NP
/\
D Adv Adj N V D N
The very small boy kissed the platypus.
When you've built up the subject NP and the VP, apply the TP (and
if appropriate the CP) rule:
Chapter 3: Constituency
TP
83
D Adv Adj N V D N
The very small boy kissed the platypus.
7. This is the most important step of all: Now go back and make sure
that your tree is really generated by the rules. Check each level
in the tree and make sure your rules will generate it. If they don't,
apply the rule correctly and fix the structure.
8. Some important considerations:
a) Make sure that everything is attached to the tree.
b) Make sure that every category has only one line immediately on top
of it (it can have more than one under it, but only one immediately
on top of it).
c) Don't cross lines.
d) Make sure all branches in the tree have a part of speech label.
e) Avoid triangles.
To Line or Not?
In many works on syntax you will find trees that have the word
connected to the category with a line, rather than writing the word
immediately under its category as we have been doing. This is a
historical artifact of the way trees used to be constructed in the 1950s. The
lines that connect elements in trees mean "created by a phrase structure
rule." There are no phrase structure rules that connect words with their
categories (i.e., there is no rule V-> kissed), so technically speaking any
line between the word's category and the word is incorrect.
CORRECT
NP
N
dogs
INCORRECT
*NP
N
dogs
Don't do this!
84 Preliminaries
Skill at treedrawingcomes only withpractice. At theend ofthis chapter are
a number of sentencesthat you can practiceon. Use the suggestions above if
you find them helpful. Another helpful idea is to model your trees on ones
that you can find in this chapter. Look carefully at them, and use them
as a starting point. Finally, don't forget: Always check your trees against
the rules that generate them.
2.2 The Top-down Method ofDrawing Trees
Most professional syntacticians use a slightly quicker means of drawing
trees. Once you are practiced at identifying the structure of trees, you will
probably want to use this technique. But be warned, sometimes
this technique can lead you astray if you are not careful.
1. This method starts out the same way as the other: write out the sentence
and identify the parts of speech.
D Adv Adj N V D N
The very small boy kissed the platypus.
2. Next draw the TP node at the top of the tree, with the subject NP and VP
underneath:
TP
NP VP
D Adv Adj N V D N
The very small boy kissed the platypus.
3. Using the NP rule, flesh out the subject NP. You will have to look ahead
here. If there is a P, you will probably need a PP. Similarly, if there is an
Adj, you'll need at least one AdjP, maybe more. Remember the principle
of modification: elements that modify one another are part of the same
constituent.
D Adv Adj N V D N
The very small boy kissed the platypus.
Chapter 3: Constituency 85
4. Fill in the AdvPs, AdjPsand PPsas necessary. Youmay need to do other
NPs inside PPs
D Adv Adj N V D N
The very small boy kissed the platypus.
5. Next do constituents inside the VP, including object NPs, and any APs
and PPs inside them.
D Adv Adj
The very small
NP
/\
N V D N
boy kissed the platypus.
Again, the most important step is to go back and make sure that your
tree obeysall the rules, as wellas the golden rule of tree structures.
Some important considerations:
a) Make sure that everything is attached.
Make sure that every category has only one line immediately on top
of it. (It can have more than one under it, but only one immediately
on top of it.)
Don't cross lines.
Make sure all branches in the tree have a part of speech label.
Avoid triangles.
Again, I strongly recommend that you start your tree drawing using the
bottom-up method, but after somepractice, you may find this latter method
quicker.
b)
c)
d)
e)
86 Preliminaries
2.3 Bracketed Diagrams
Sometimes it is preferable to use the bracketed notation instead of the tree
notation. This is especially true when there are large parts of the sentence
that are irrelevant to the discussion at hand. Drawing bracketed diagrams
essentially follows the same principles for tree drawing (see2.1 or 2.2 above).
The exception is that instead of drawing to lines connecting at the top,
you put square brackets on either side of the constituent. A label is usually
put on the left member of the bracket pair as a subscript.
62) NP
D N
the man = [NP[D the] [N man]]
Both words and phrases are bracketed this way. For each point where you
have a bunch of lines connecting, you have a pair of brackets.
To see how this works, let's take our sentence from sections 2.1 and 2.2
above and do it again in brackets:
1. First we mark the parts of speech. This time with labeled brackets:
[DThe] [Advvery] [Adjsmall] [Nboy] [vkissed] [Dthe] [Nplatypus].
2. Next we apply the AP rule, NP and PP rules:
AP:
[DThe] [AdvP[Advvery]] [Adjsmall] [Nboy] [vkissed] [Dthe] [Nplatypus].
[DThe] [AdjplAdvpLdvVery]] [Adj small]] [Nboy] [v kissed] [Dthe] [N platypus].
NP:
[NP[D The][AdjP[AdvP[Advvery]][Adjsmall]][Nboy]] [v kissed] [Dthe] [N platypus].
[NP[DThe][AdjP[AdvP[Advvery]][Adjsmall]][Nboy]][vkissed][NP[Dthe][Nplatypus]].
3. Now the VP and TP rules:
VP:
[NP[DThe][AdjP[AdvP[Advvery]][Adjsmall]][Nboy]][Vp[vkissed][Np[Dthe][Nplatypus]]].
TP:
[Tp[Ni'[DThe][AdjP[AdvI.[Advvery]][Adjsrnall]][Nboy]][Vp[vkissed][Np[Dthe][Nplatypus]]]].
4. Finally, goback and check that thestructure can begenerated by the rules.
Chapter 3: Constituency 87
3. Modification and Ambiguity
Syntactic trees allow us to capture another remarkable fact about language.
Let's start with the following sentence:
63) The man killed the king with a knife.
This sentences turns out to have more than one meaning, but for the moment
consider only the least difficult reading for it (the phrase in quotes in (64)
is called a paraphrase, which is the technical term for "another way of saying
the same thing"):
64) (63) meaning "the man used a knife to kill the king."
Remember the Principle of modification:
65) Principle ofModification (revised): If an XP (that is, a phrase with some
category X) modifies some head Y, then XP must be a sister to Y (i.e.,
a daughter of YP).
In (63) the PP with aknife modifieskilled, so the structure will look like (66):
66) TP
D N / v \ NP
The man Ucilled/ /\
— D N
the king
modifies
[With a knife] describes how the man killed the king. It modifies the verb
killed, so it is attached under the VP. Now consider the other meaning of (63).
67) (63a) meaning "the kingwith theknife waskilled by the man (whoused
a gun)."
The meaning in (67) has the PP with the knife modifying king, and thus
attached to the NP:
88 Preliminaries
68) TP
NP VP
D N V NP
The man killed /T
D/N
the \ kii
modifies
These examples illustrate an important property of syntactic trees.
Trees allow us to capture the differences between ambiguous readings of
the same surface sentence.
You now have enough information to try General Problem Set 8
4. Constituency Tests
In chapter 1, we held linguistics in general (and syntax specifically) up to
the light of the scientific method. That is, if we make a hypothesis about
something, wemust beable totest thathypothesis. In thischapter, wehave
proposed the hypothesis that sentences are composed of higher-level
groupings calledconstituents. Constituentsare represented in tree structures
and are generated by rules. If the hypothesis of constituency is correct,
we should be able to test it in general (aswell as test the specific instances
of the rules).
In order to figure out what kinds of tests we need, it is helpful
to reconsider the specifics of the hypothesis. The definition of constituents
states that they are groups of words that function as a unit. If this is the case,
then we should find instances wheregroups ofwords behave as single units.
These instances can serve as tests for the hypothesis. In other words,
they are tests forconstituency. There are a lot ofconstituency tests listed in
thesyntactic literature. We are going tolook at only four here: replacement,
stand alone, movement, and coordination.
First, the smallest constituent is a single word, so it follows that if you
can replace a group of words with a singleword then we know that group
Chapter 3: Constituency 89
forms a constituent. Consider the italicized NP in (69), it can be replaced
with a single word (in this case a pronoun). This is the replacementtest.
69) a) The man from NY flew only ultra-light planes.
b) He flew only ultra-light planes.
There is one important caveat to the test of replacement: There are many
cases in our rules of optional items (those things marked in parentheses like
the AP in NP —» (D) (AdjP+) N.) When we replace a string of words with
a single word, how do we know that we aren't just leaving off the optional
items? To avoid this problem, we have to keep the meaning as closely related
to the original as possible. This requires some judgment on your part.
None of these tests is absolute or foolproof.
The second test we will use is the stand alone test (sometimes also called
the sentence fragment test). If the words can stand alone in response to a
question, then they probably constitute a constituent. Consider the sentence
in (70a) and repeated in (70b). We are going to test for the constituency
of the italicized phrases.
70) a) Paul ateat a reallyfancy restaurant.
b) Paul ateat a really fancy restaurant.
If we ask the question "What did Paul do yesterday afternoon?" we can
answer with the italicized group of words in (70a), but not in (70b):
71) a) Ate at a really fancy restaurant,
b) *Ateat.
Neither of these responses is proper English in prescriptive terms, but you
caneasilytellthat (71a) is better than (71b).
Movement is our third test of constituency. If you can move a group
of words around in the sentence, then they are a constituent because you
can move them as a unit. Some typical examples are shown in (72). Clefting
(72a) involves putting a string ofwordsbetween Itwas (orItis) and a that at
the beginning of the sentence. Preposing (72b) (also called pseudoclefting)
involves putting the string of words before a is/are what or is/are who at
the front of the sentence. Wediscuss thepassive(72c) at length in chapter 10.
Briefly, it involves putting the object in the subject position, the subject in
a "by phrase" (after the word by) and changing the verb form (forexample
from kiss to was kissed).
90 Preliminaries
72) a) Clefting: It was [a brand new car] that he bought.
(from Hebought a brand newcar)
b) Proposing: [Bigbowls of beans] are what I like.
(from I like bigbowls ofbeans)
c) Passive: [Thebig boy] was kissed by [the slobbering dog].
(from The slobbering dog kissed thebigboy)
Again, the movement test is only reliable when you keep the meaning
PSRs for Conjunction
In order to draw trees with conjunction in them, we need two more rules.
These rules are slightly different than the ones we have looked at up to
now. These rules are not category specific. Instead they use a variable (X).
This X can stand for N or V or A or P etc. Just like in algebra, it is a
variable that can stand for different categories. We need two rules, one to
conjoin phrases ([The Flintstones] and [the Rubbles]) and one to conjoin
words (the [dancer] and [singer]):
i) XP -> XP conj XP
ii) X-^XconjX
These result in trees like:
NP conj NP
iv) V
V conj V
roughly the same as the original sentence.
Finally, we have the test of coordination (also called conjunction).
Coordinate structures are constituents linked by a conjunction like and or or.
Only constituents of the same syntactic category can be conjoined:
73) a) [John] and [the man] went to the store,
b) *Johnand very blue went to the store.
If you can coordinate a group of words with a similar group of words, then
they form a constituent.
You now have enough information totry General Problem Sets9 & 10and
Challenge Problem Set4
Chapter 3: Constituency 91
When Constituency Tests Fail
Unfortunately, sometimes it is the case that constituency tests give false
results (which is one of the reasons we haven't spent much time on them
in this text). Consider the case of the subject of a sentence and its verb.
These do not form a constituent:
i) TP
NP VP
subject
V NP
object
However, under certain circumstances you can conjoin a subject and verb
to the exclusion of the object:
ii) Bruce loved and Kelly hated phonology class.
Sentence (ii) seems to indicate that the verb and subject form a
constituent, which they don't according to the tree in (i). As you will see
in later chapters, it turns out that things can move around in sentences or
be deleted. This means that sometimes the constituency is obscured by
other factors. For this reason, to be sure that a test is working correctly
you have to apply more than one test to a given structure. Always
perform at least two different tests to check constituency; as one alone
may give you a false result.
5. SUMMARY AND CONCLUSION
We've done a lot in this chapter. We looked at the idea that sentences are
hierarchically organized into constituent structures. We represented these
constituent structures in trees and bracketed diagrams. We also developed
a set of rules to generate those structures, and finally we looked at
constituency tests that can be used to test the structures. Parts of speech are
the labeling system for constituent structure. Weshowed that parts of speech
can't be determined by meaning alone. In the appendix to this chapter,
we sketch out some distributional tests for part of speech class.
92 Preliminaries
Appendix: How to Do Foreign Language PSR Problems
There are two kinds of non-English language problems found in syntax:
those that provide a word-by-word gloss and those that don't.
Al. Doing problems with word-by-word glosses
Often, linguistic examples from languages other than English will take
the following form (example from Sinhala - a language spoken in Sri Lanka;
data from Lehmann 1978):
74) Jon ballavo dakka ^ Actual language data
John dog saw < Word-by-word gloss
"John saw the dog." < Idiomatic translation
There are three lines: the actual data, a word-by-word gloss and an idiomatic
translation into English. Of these the most important for doing the problem
set is the second line - the word-by-word gloss. The glosses are lined up
word for word (and sometimes morpheme for morpheme) with the foreign
language on the line above. This line tells you (1) what each word
in the foreign language example means, and more importantly, (2) the order
of the words in the foreign language. When trying to determine the phrase
structure of a foreign language or the behavior of a word or phrase, this
is the line to look at! (However, when drawing trees and citing examples
in your answer it is considered more respectful of the language to use
the actual foreign language words.) Remember: don't do an analysis
ofthe idiomatic translation ofthe sentence, because then you are only doing
an analysis of English!
Here's a more complete paradigm of Sinhala, along with a series of
typical questions:
i) Jon ballavo dakka
John dog saw
"John saw the dog."
ii) Jon janele indola ballavo dakka
John window from dog saw
"John saw the dog from the window."
Chapter 3:Constituency 93
iii) Jon eyage tadi ballava dakka
John his big dog saw
"John saw his big dog."
a) Assume there is an AdjP rule: AdjP -> Adj. What is the NP rule of
Sinhala?
b) What is the PP rule of Sinhala?
c) What is the VP rule of Sinhala? (Assume all non-head material is
optional.)
d) What is the TP rule of Sinhala?
e) Draw the tree for sentences (ii) and (iii).
The first step in analyzing a language like this is to determine the parts
of speech of each of the words. Be very careful here, do not assume
that because English has certain categories that the language you are looking
at has the same categories; however, all other things being equal
you can assume that there will be some parallels (unless we have evidence
to the contrary):
i) Jon ballavo dakka
John dog saw
N N V
ii) Jon janele indala ballavo dakka
John window from dog saw
N N P N V
iii) Jon eyage tadi ballavs dakka
John his big dog saw
N D Adj N V
Next let's answer question (a). We can observe from sentence (i) that an NP
in Sinhala (just like in English) can be an N by itself (e.g., ]0n). This means
that anything other than the noun has to be optional. Consider now the
sentence in (iii); from the literal English translation we can tell that the words
meaning "big" and "his" modify the word "dog," and are thus part of the
NP headed by "dog." We're told in (a) to assume that there is an AdjP rule
(AdjP -> Adj), and we are treating the word for "his" as a determiner. Thus it
follows that the Sinhala NP rule is at least NP -> (D) (AdjP) N. You'll notice
that the order of elements in this rule is the same as the order of elements
in the Sinhala sentence. You should also note that the PP meaning "from the
window" does not modify the N, so is not part of the NP rule at this point.
Since it modifies the V, it will be part of the VP rule.
94 Preliminaries
Question (b) asks us about the PP rule. We have one P in the data the
word meaning "from" in sentence (ii). Pay careful attention here.
This P appears between two nouns; but the noun associated with the P is
the one meaning "window." This means that the P in Sinhalafollows the NP;
so the rule is PP -> NP P. We have no evidence if the NP here is optional.
The VP rule is next in (c). Sentence (ii) is the most informative here.
Looking at what would be in the VP in English, we have the PP meaning
"from the window" and the NP meaning "dog." These both precede the V.
This is true in sentences (i) and (iii) too. The PP is clearly optional, but there
is no evidence in the data about whether the NP is or not. However,
you are told to assume that "all non-head material is optional." So the rule
is VP -» (PP) (NP) V.
Finally we have the TP rule. Like English, the subject NP precedes
the VP. So the rule is TP -> NP VP. We have no evidence for a T node so we
have not posited one.
Here are the trees for (ii) and (iii).
iii)
N PP NP V N NP V
Jon
/\ | dakka Jon /f\ dakka
NP P N D AdjP N
ifidola ballavo eyage ballavo
N Adj
janele tadi
A2. Doingproblems without word-by-word glosses
Sometimes you will be given data without word-by-word glosses, and only
an idiomatic sentence translation. Take the following example from Welsh:
i) Agorodd y dyn y drws.
ii) Collodd y dyn ddwy bunt,
iii) Gyrhaeddodd y dyn.
iv) Gaeth y dyn ddwy bunt,
v) Agorodd Fred ddwy ddrws.
'The man opened the door."
'The man lost two pounds (money).'
'The man arrived."
'The man got two pounds."
'Fred opened two doors."
Since word-by-word glosses are the most important part of a foreign
language problem (see sectionAl), the first thing you have to do is develop
Chapter 3:Constituency 95
a word-by-word gloss (or morphological analysis). To do this you compare
and contrast the sentences in the data set, using the translation as a guide.
Let us do the above sentences as an example. First, look at the first four
of the sentences, what words in the English gloss is common to them all?
"The man". Now, look at the Welsh in the left column. What words
are found in all of these sentences? Vdyn. There is a high probability that the
Welsh words y dyn means "the man." We can deduce that 1/ means "the" by
looking at the other instance of 1/ in sentence (i). Sentence (i) has two "the"s
in it, and the Welsh has two i/s. This means that dyn probably means man.
We might even venture that drws means door. This appears to be consistent
with the fact that sentence (v) has ddrws in it. Although ddrws and drws aren't
identical, neither are their glosses: sentence (i) has the singular "door" in it,
and sentence (v) has the plural "doors."'
Similarly by looking at sentences (ii) and (v) we can see that the only
two words they have in common in both the English and the Welsh is ddwy
"two." Hopefully the meaning of Fred is self-evident, but even if it weren't
we could deduce it by process of elimination. The same is true of each of the
verbs. Sentences (i) and (v) have in common the word agorodd and the gloss
"open." We might suppose that initial position is where verbs like "open"
go. This means that the first word in each of the other sentences is the verb.
Alternately we could have deduced that these were the various verbs,
based on the fact that we had meanings for all the nouns and determiners,
so by process of elimination all that is left in each sentence is the verb.
Once you've done this morphological analysis, you are ready to tackle
the rest of the problem as we did in section Al.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Constituent: A group of words that functions together as a unit.
ii) Hierarchical Structure: Constituents in a sentence are embedded
inside of other constituents.
iii) Syntactic Trees and Bracketed Diagrams: These are means of
representing constituency. They are generated by rules.
AYou might be tempted to think that theAd is a plural marker. While consistent with
the facts above, the dd is actually a result of a special morphophonological process
triggered by dun/ (ddwy) called a consonant mutation. Knowing this isn't necessary
to solving the problem set.
96 Preliminaries
iv) Phrase Structure Rules
a) CP -> (C) TP
b) TP -> (NP/CP) (T)VP
c) VP -> (AdvP+) V (NP) ({NP/CP}) (AdvP+) (PP+) (AdvP+)
d) NP -> (D) (AdjP+) N (PP+) (CP)
e) PP -> P (NP)
f) AdjP -» (AdvP) Adj
g) AdvP -» (AdvP) Adv
h) XP -» XP conj XP
i) X -> Xconj X
v) Head: The word that gives its category to the phrase.
vi) Recursion: The possibility of loops in the phrase structure rules
that allow infinitely long sentences, and explain the creativity
of language.
vii) The Principle of Modification: If an XP (that is, a phrase with some
category X)modifies some head Y, then XP must be a sister to Y(i.e.,
a daughter of YP).
viii) Constituency Tests: Tests that show that a group of words function
as a unit. There are four major constituency tests given here:
movement, coordination, stand alone, and replacement.
Further Reading
Chomsky, Noam (1957) Syntactic Structures. The Hague: Janua Linguarum 4.
Chomsky, Noam (1965) Aspects of the Theory of Syntax: Cambridge: MIT
Press.
General Problem Sets
1. Trees: NPs, AdjPs and AdvPs
[Application of Skills; Basic]
Draw the trees for the following AdjPs, AdvPs, and NPs:
a) very smelly b) too quickly
c) much too quickly d) very much too quickly
e) the old shoelace
f) the soggy limp spaghetti noodle [assume spaghetti = Adj]
g) these very finicky children
Chapter 3: Constituency 97
2. Trees II: English PPs
[Application of Skills; Basic]
Draw the trees for the following English NPs and PPs:
a) The desk with the wobbly drawer
b) In my black rubber boots [assume rubber is an Adj]
c) That notebook with the scribbles in the margin
d) The pen at the back of the drawer in the desk near the bright yellow
painting
3. Swedish NPs
[Application of Skills and Knowledge; Basic]
Consider the following data from Swedish. (If you speak Swedish, please
confine yourself to this data, do not try to include definite forms, e.g.,
the umbrella.) You may wish to review Appendix A before attempting
this problem. (Data courtesy of Sheila Dooley.)
folk "people"
ett paraply "an umbrella"
tre paraplyer "three umbrellas"
ett apple "an apple"
ett rott paraply "a red umbrella"
ett gult apple "a yellow apple"
ett mycket tint paraply "a very fine umbrella"
ett gammalt fint paraply "a fine old umbrella"
ett rott paraply med ett gult handtag "a red umbrella with a yellow handle"
Assume the Adv rule of Swedish is AdvP -> Adv. What is the AdjP rule?
Are determiners obligatory in Swedish NPs?
Are AdjPs obligatory in Swedish NPs?
What is the PP rule for Swedish?
Are PPs obligatory in Swedish NPs?
What is the NP rule for Swedish?
Draw the trees for (g), (h), and (i)
Give the bracketed diagram for (f) and (i)
4. English
[Application of Skillsand Knowledge; Basic to Advanced]
Draw phrase structure trees and bracketed diagrams for each of the
following sentences, indicate all the categories (phrase (e.g., NP) and word
level (e.g., N)) on the tree. Use the rules given above in the "Ideas" summary
of this chapter. Be careful that items that modifyone another are part of the
same constituent. Treat words like can, should, might, was, as instances
of the category T (tense). (Sentences d-h are from Sheila Dooley.)
a) The kangaroo hopped over the truck.
b) Ihaven't seen this sentence before, /before is a P, haven't is a 77
c) Susan will never sing at weddings, /never is an Adv]
d) The officer carefully inspected the license.
98 Preliminaries
e) Every cat always knows the location of her favorite catnip toy.
f) The cat put her catnip toy on the plastic mat.
g) The very young child walked from school to the store,
h) John paid a dollar for a head of lettuce.
i) Teenagers drive rather quickly.
j) A clever magician with the right equipment can fool the audience easily.
k) The police might plant the drugs in the apartment.
I) Those Olympic hopefuls should practice diligently daily.
m) The latest research on dieting always warns people about the dangers of
too much cholesterol,
n) That annoying faucet was dripping constantly for months.
0) Marian wonders if the package from Boston will ever arrive.
p) Isaid that Bonny should do some dances from the Middle East,
q) That Dan smokes in the office really bothers Alina.
r) The belief that syntactic theory reveals the inner structure of sentences
emboldened the already much too cocky professor.
5. Bambara
[Application of Skills; Basic]
Consider the following data from Bambara, a Mande language spoken in
Mali. (The glosses have been slightly simplified.) Pay careful attention to the
second line, where the word order of Bambara is shown. (Data from Koopman
1992.)
a) A kasira.
he cried
"He cried."
b) Den ye ji min.
child past water drink
"The child drank water."
c) N sonna a ma.
I agreed it to
"I agreed to it."
Answer the following questions about Bambara. Do not break apart words in
your analysis.
1) Do you need a T category in Bambara?
2) Do you need a D category in Bambara?
3) What is the NP rule for Bambara? (You do not need any AdjP or PPs in
the rule.)
4) What is the PP rule for Bambara?
5) What is the VP rule for Bambara?
6) What is the TP rule for Bambara? (Keep in mind your answers to the
above questions; be consistent.)
7) Draw trees for (a), (b), and (c) using your rules.
8) Draw bracketed diagrams for (b) and (c).
Chapter 3:Constituency 99
6. Hixkaryana
[Application of Skills; Basic/Intermediate]
Look carefully at the following data from a Carib language from Brazil
(the glosses have been slightly simplified from the original). In your analysis
do not break apart words. (Data from Derbyshire 1985.)
a) Kuraha yonyhoryeno biyekomo.
bow made boy
"The boy made a bow."
b) Newehyatxhe woriskomo komo.
take-bath women all
"All the women take a bath."
c) Toto heno komo yonoye kamara.
person dead all ate jaguar
"The jaguar ate all the dead people."
Now answer the following questions about Hixkaryana:
1) Is there any evidence for a determiner category in Hixkaryana? Be sure
to consider quantifier words as possible determiners (like some and all).
2) Posit an NP rule to account for Hixkaryana. (Be careful to do it for
the second line, the word-by-word gloss, in these examples
not the third line.) Assume there is an AdjP rule: AdjP -> Adj.
3) Posit a VP rule for Hixkaryana.
4) Posit a TP rule for Hixkaryana.
5) What is the part of speech of newehyatxhe? How do you know?
6) Draw the trees for (a) and (c) using the rules you posited above.
(Hint: if your trees don't work, then you have probably made a mistake
in the rules.)
7) Give bracketed diagrams for the same sentences.
7. DUTCH
[Application of Skills: Intermediate]
Consider the following sentences of Dutch. (Data from Ferdinand de Haan.)
a) De man in de regenjas is naar Amsterdam gegaan.
the man in the raincoat is to Amsterdam going
"The man in the raincoat is going to Amsterdam."
b) De man heeft een gele auto met een aanhanger gekocht.
the man has a yellow car with a trailer bought
"The man has bought a yellow car with a trailer."
c) De vrouw heeft een auto gekocht.
the woman has a car bought
"The woman has bought a car."
100 Preliminaries
d) Jan is vertrokken.
John is gone
"John left."
(Ifyou speak Dutch, please confine your answer to the data given above and
do not add any other examples.)
1) Assume an AdjP rule, AdjP -> Adj; What is the NP rule of Dutch?
2) What is the PP rule of Dutch?
3) What is the VP rule of Dutch? (Assume that is and heeft are of the
category T and are not part of the VP.)
4) What is the TP rule for Dutch?
5) Draw the trees for (a) and (b).
8. Ambiguity
[Application of Knowledge and Skills; Basic to Intermediate]
The following English sentences are all ambiguous. Provide a paraphrase (a
sentence with roughly the same meaning) for each of the possible meanings,
and then draw (two) trees of the original sentence that distinguish the two
meanings. Be careful not to draw the tree of the paraphrase. Your two trees
should be different from one another, where the difference reflects which
elements modify what. (For sentence (b) ignore the issue of capitalization.)
You may need to assume that old and seven can function as adverbs.
Sentences (c), (d), (e), and (f) are ambiguous newspaper headlines taken
from http://www.fun-with-words.com/ambiguous_headlines.html.
a) John said Mary went to the store quickly.
b) I discovered an old English poem.
c) Two sisters reunited after 18 years in checkout counter
d) Enraged cow injures farmer with ax
e) Hospitals are sued by seven foot doctors
f) Dealers will hear car talk after noon
9. Structure
[Application of Knowledge; Intermediate]
In the following sentences a sequence of words is marked as a constituent
with square brackets. State whether or not it is a real constituent, and
what criteria (that is constituency tests) you applied to determine that result.
a) Susanne gave [the minivan to Petunia].
b) Clyde got [a passionate love letter from Stacy].
10. English Prepositions
[Critical Thinking; Intermediate]
In the text, we claimed that perhaps the NP in PPs was optional, explaining
why we can say He passed out, where the preposition out has no object.
Consider an alternative: the expression [passed out] is really a "complex"
Chapter 3:Constituency 101
verb. Using constituency tests, provide arguments that the structure
of expressions like (a-d) is really [[V P] NP] rather than: [V[P NP]].
a) He blew out the candle.
b) He turned off the light.
c) He blew up the building.
d) He rode out the storm.
CHALLENGE PROBLEM SETS
Challenge Problem Set 1: Quantifiers
[Critical Thinking; Challenge]
Our NP rule only allows one determiner. How can we deal with NPs like (a)
and (b), but still rule out NPs like (c):
a) the two CDs
b) the many reasons
c) *the those books
Challenge Problem Set 2: Nominal Adverbials
[Critical Thinking; Data Analysis; Challenge]
In the text we observed that NPs must appear adjacent to the Verb in VPs,
they cannot come after a post-verbal AdvP:
a) *Shannon kissed quietly the kitten.
b) Shannon kissed the kitten quietly
However, there appears to be a class of nouns that can appear in this
position. These are nouns expressing quantities of time:
c) Shannon left quietly every day
Other example include last year, every day, each week etc.
Part 1: How do we know that these constituents are NPs and not AdvPs?
(Pay attention to what can modify the N.)
Part 2: Is there a way to incorporate such NPs into our PSR system? Explain
your answer.
Challenge Problem Set 3: Possessive NPs
[Critical Thinking; Challenge]
Part 1: Our NP rule reads NP -» (D) (AdjP+) N (PP+) (CP). Consider the
following NPs. What problem do these NPs cause our rule:
a) Patrick's box
b) the man's box
Part 2: Consider the following data:
102 Preliminaries
c) *Patrick's the box
d) *the man's the box
How might you revise the NP rule to account for NPs like (a) and (b), keeping
in mind that a possessive NP (like Patrick's) cannot appear in the same NP
as a determiner. Given the rule you develop draw the tree for (b).
Challenge Problem Set 4: Constituency Tests5
[Application of Knowledge; Challenge]
Do the words in boldface in the following sentence form a single constituent?
That is, is there a [Barbie and Ken kissing] constituent? How do you know?
Use all the tests available to you.
Barbie and Ken were seen by everyone at the party kissing.
A couple of things may help you in this problem. (1) Remember that
constituents can be inside other constituents. (2) This sentence is a passive,
which means that some movement has happened, so don't let the fact
that there is other stuff in between the two bits throw you off.
Sheila Dooley is the source of this problem set.
Structural Relations
0. Introduction
In chapter 3, we developed the notion of constituency. Constituents
are groups of words that function as single units. In order to systematically
identify these, we proposed a set of rules. These rules generate trees, which
in turn represent constituency. Take a careful look at any tree in the last
chapter and you'll notice that it is a collection of labels and lines; within this
collection of labels there is an organization. In particular, various parts of
the tree are organizedhierarchically withrespect toone another. A collection
of lines and labels with an internal organization like syntactic trees
is a geometric object. It isn't a geometric object like a circle or a square,
but nonetheless it has bits that are spatially organized with respect to one
another. If syntactic trees are geometric objects, they can be studied
and described mathematically - the focus of this chapter. This chapter differs
from all the others in this book. You won't see many sentences or phrases
here, and there is very little data. This chapter is about the purely formal
properties of trees. But don't think you can skip it. The terminology
we develop here is a fundamental part of syntactic theory and will play
an important role in subsequent chapters.
104 Preliminaries
Why Study the Geometry of Trees?
It is worth considering whether it is necessary to concern ourselves with
the mathematics of tree diagrams. There are actually two very good
reasons why we should do this. First, by considering the geometry of trees,
we can assign names to the various parts and describe how the parts relate
to one another. For example, in the last chapter we were only able to give a
vague definition of the term constituent. In this chapter, we'll be able to give
a precise description. Second, it turns out that there are many syntactic
phenomena that make explicit reference to the geometry of trees. One
of the most obvious of these refers to anaphors. Anaphors can only appear
in certain positions in the geometry of the tree. The distribution
of anaphors and other types of nouns is the focus of the next chapter.
1. The Parts of a Tree
Let's start with a very abstract tree drawing:
1) M
X O
D E F H I J
This tree would be generated by the rules in (2):
2) M-^NO
N^DEF
O^HIJ
You can check this byapplying each of the rules to the tree in (1). I'm using
an abstract tree here because I don't want the content of each of the nodes
to interfere with the underlying abstract mathematics. (But if you find this
confusing, you can substitute TP for M, NP for N, VP for O, etc., and you'll
see that this is just a normal tree.) Now we can describe the various parts
of this tree. The lines in the tree are called branches. A formal definition
of branch is given in (3), and the branches are marked in (4):
3) Branch: A line connecting two parts of a tree.
4) M branches
Chapter 4: Structural Relations 105
The end of any branch is called a node. Both ends are called nodes.
For example, N and F are both called nodes of a branch. Any time two
or more branches come together, this is also called a node:
5) Node:The end of a branch.
A node with two or more branches below it is said to be branching; a node
that has a single branch below it is said to be non-branching.
Nodes in a tree are labeled. In the tree above, M, N, O, D, E, F, H, I, J are
the labels for the nodes that make up the tree. This is very abstract of course.
In the last chapter, we looked at the various parts of speech (N, V, A, P,
etc.) and the phrasal categories associated with them (NP, VP, AP, PP, etc.).
These are the labels in a real syntactic tree.
6) Label: The name given to a node.
There are actually different kinds of nodes that we'll want to make reference
to. The first of these is called the root node. The root node doesn't have any
branch on top of it. There is only ever one root node in a sentence. (Theterm
root is a little confusing, but try turning the trees upside down and you'll
see that they actually do look like a tree (or a bush at least). In most trees we
looked at in the last chapter, the root node is almost always the TP (sentence)
node.
7) Root node (preliminary): The node with no line on top of it.
At the opposite end of the tree are the nodes that don't have any lines
underneath them. If the tree analogy were to really hold up, we should call
these "leaves." More commonly, however, these are called terminal nodes.
8) Terminal node (preliminary): Anynode with no branchunderneath it.
Any node that isn't a terminal node is calleda non-terminal node:
9) Non-terminal node (preliminary): Anynode with a branch underneath it.
Notice that the root node is also a non-terminal node by this definition.
After we add some definitions in the next chapter, we'll have reason to
reformulate the definitions of root, terminal and non-terminal nodes, but for
now theseshould give you the basicidea. In (10), we have a tree where the
root node, the terminal nodes, and the non-terminal nodes are all marked.
10) M •< Root node
Non-terminal nodes
D E F H I J *< Terminal nodes
106 Preliminaries
In this tree, M is the root node. M, N, and O are non-terminals, and D, E, F,
H, I, and J are terminal nodes.
We now have all the terms we need to describe the various parts of a
tree. The lines are called branches. The ends of the lines are called nodes, and
each of the nodes has a label. Depending upon where the node is in the tree,
it can be a root node (the top), a terminal (the bottom), or a non-terminal
(any node except the bottom). Next we turn to a set of terms and
descriptions that will allow us to describe the relations that hold between
these parts. Because we are talking about a tree structure here, these relations
are often called structural relations.
2. Domination
2.1 Domination
Some nodes are higher in the tree than others. This reflects the fact that trees
show a hierarchy of constituents. In particular, we want to talk about nodes
that are higher than one another and are connected by a branch. The relation
that describes two nodes that stand in this configuration is called
domination. A node that sits atop another and is connected to it by a branch
is said to dominate that node.
11) Domination1: Node Adominates node Bif and only ifA ishigher up in
the tree than B and if you can trace a line from A to B going
only downwards.
In (12), M dominates all the other nodes (N, O, D, E, F, H, I, J). N dominates
D, E, and F, and O dominates H, I, J.O does not dominate F, as you can see
by virtue of the fact that there is no branch connecting them.
12)
1The definition given here is actually for proper domination (an irreflexive relation). Simple
domination is usually reflexive (nodes dominate themselves). For the most part linguists are
interested in proper domination rather than simple domination, and they use the term
"domination" to mean "proper domination" as we do here. Domination is sometimes also called
dominance.
Chapter 4: Structural Relations 107
Domination is essentially a containment relation. The phrasal category
N contains the terminal nodes D, E, and F. Containment is seen more clearly
when the tree is converted into a bracketed diagram:
13) UnDEFHoHIJ]]
In (13) the brackets associated with N ([N D E F]) contains the nodes D, E,
and F. The same holds true for O which contains H, I, and J. M contains both
N and O and all the nodes that they contain. So domination is a technical
way of expressing which categories belong to larger categories.
You now have enough information totry General Problem Sets 1 &2
2.2 Exhaustive Domination
In the last chapter, we developed an intuitive notion of constituent.
The relation of domination actually allows us to be a little more rigorous and
develop a formal notion of constituency. In order to do this, we need another
definition, exhaustive domination'.
14) Exhaustive domination: Node A exhaustively dominates a set of terminal
nodes {B, C, ..., D}, provided it dominates all the members of the set (so
that there is no member of the set that is not dominated by A) and there
is no terminal node G dominated by A that is not a member of the set.
This is a rather laborious definition. Let's tease it apart by considering an
example.
15) A
B C D
What we are concerned with here is a set of nodes and whether or not
agivennode dominates theentire set.Setsareindicated with curly brackets {}.
Start with the set of terminal {B, C, D}. In (15) all members of the set {B, C,
D} are dominated by A; there is no member of the set that isn't dominated
byA.Thissatisfies the firstpart of thedefinition in(15). Turning to the second
part, A only dominates these terminal nodes and no other terminals. There is
no node G dominated by A that is not a member of the set. This being
the case we can say of the tree in (15) that A exhaustively dominates the set
{B,C, D}. Let's turn to a different tree now.
108 Preliminaries
16) H
B C D
Again let's consider whether A exhaustively dominates the set {B, C, D}. In
(16), one member of the set, D, is not immediately dominated by A. As such
the set (B, C, D} is not exhaustively dominated by A. The reverse situation
is seen in (17):
17)
D
While it is the case that in (17), B, C, and D are all immediately dominated
by A, there is also the node G, which is not a member of the set {B, C, D},
so the set {B, C, D} is not exhaustively dominated by A (although the set
{B, C, D, G} is). On a more intuitive level, exhaustive domination holds
between a set of nodes and their mother. Only when the entire set (and
only that set) are immediately dominated by their mother can we say that
the mother exhaustively dominates the set.
Look carefully at the structures in (15), (16), and (17). In (15) you'll see
that the set {B, C, D} forms a constituent (labeled A). In (16), that set does not
form a constituent, nor does it form a constituent in (17) (although the set is
part of a larger constituent in that tree). In (17), there is no sense in which B,
C, D form a unit that excludes G. It seems then that the notion
of constituency is closely related to the relation of exhaustive domination.
This is reflected in the following formal definition of a constituent.
18) Constituent: A set of terminal nodes exhaustively dominated by a
particular node.
If we look at the tree in (16) again, you can see that each constituent meets
this definition. The set of nodes exhaustively dominated by A is {B, C} which
is the set of terminals that make up the A constituent. Similarly, The
constituent F is made up of the set {D} which is exhaustively dominated by F;
finally, H exhaustively dominates {B, C, D} (remember the definition
is defined over terminals, so A and F don't count) which is the constituent
that H represents.
Before turning to some other structural relations, it is important to look
at one confusing piece of terminology. This is the distinction between
constituent and constituent of. A constituent, as defined in (18), is a set
Chapter 4:Structural Relations 109
of nodes exhaustively dominated by a single node. A constituent of,
by contrast, is a member of the constituent set. Consider the tree in (19):
19) A
B C D
Here we have the constituent A, which exhaustively dominates the set
{B, C, D}. Each member of this set is called a "constituent of A." So B is
a constituent of A. "Constituent of" boils down to domination. Adominates B
therefore B is a constituent of A:
20) Constituent of: B is a constituent of A if and only if A dominates B.
You now haveenough information to try General Problem Set 3
2.3 Immediate Domination
Domination is actually quite a general notion: In (21), M dominates all of
the nodes under it.
21)
H I J
In certain circumstances we might want to talk about relationships that are
smaller and more local. This is the relationship of immediate domination.
A node immediately dominates another if there is only one branch between
them.
22) Immediately dominate: Node A immediately dominates node B if there
is no intervening node G that is dominated by A, but dominates B.
(In other words, A is the first node that dominates B.)
In (21), M dominates all the other nodes in the tree, but it only immediately
dominates N and O. It does not immediately dominate any of the
other nodes because N and O intervene.
There is an informal set of terms that we frequently use to refer
to immediate domination. This set of terms is based on the fact that syntactic
trees look a bit like family trees. If one node immediately dominates another,
it is said to be the mother; the node that is immediately dominated is
called the daughter. In the tree above in (21), N is D's mother and D is
110 Preliminaries
N's daughter. We can even extend the analogy (although this is pushing
things a bit) and call M D's grandmother.
23) Mother: A is the mother of Bif A immediately dominates B.
24) Daughter: B is the daughter of A if Bis immediately dominated by A.
Closely related to these definitions is the definition of sister:
25) Sisters: Two nodes that share the same mother.
With this set of terms in place we can now redefine our definitions of root
nodes, terminal nodes, and non-terminals a little more rigorously:
26) Root node (revised): The node that dominates everything, but is
dominated by nothing. (The node that is no node's daughter.)
27) Terminal node (revised): A node that dominates nothing. (A node that is
not a mother.)
28) Non-terminal node (revised) A node that dominates something. (A node
that is a mother.)
We defined "constituent" in terms of domination, and from that we
derived the "constituent of" relation (essentially the opposite of domination).
We can also define a local variety of the "constituent of" relation that is
the opposite of immediate domination:
29) Immediate constituent of. B is an immediate constituent of A if and only if
A immediately dominates B.
This ends our discussion of the vertical axis of syntactic trees. Next we
consider horizontal relations.
You now have enough information to try General Problem Set 4
3. Precedence
Syntactic trees don't only encode the hierarchical organization of sentences,
they also encode the linear order of the constituents. Linear order refers
to the order in which words are spoken or written (left to right if you
are writing in English). Consider the following rule:
30) M -> A B
Thisrule not only says that Mdominates A and B and iscomposed ofA and B.
It also says that A must precede B in linear order. A must be said before B,
Chapter 4: Structural Relations 111
because it appears to the left of B in the rule. The relation of "what is said
first" is called precedence.2 In order to define this rigorously we have to
first appeal to a notion known as sister precedence:
31) Sister precedence: Node A sister-precedes node B if and only if both are
immediately dominated by the same node, and A appears to the left of B.
The ordering in this definition follows from the order of elements within
a phrase structure rule. If A is to the left of B in a phrase structure rule M —»
A B, then A and B are immediately dominated by M, and are in the relevant
order by virtue of the ordering within that rule. With this basic definition
in mind we can define the more general precedence relation:
32) Precedence: Node A precedes node B if and only if neither A dominates B
nor B dominates A and A or some node dominating A sister-precedes B
or some node dominating B.
This definition is pretty complex, so let's break it apart. The first bit of the
definition says "neither A dominates B nor B dominates A." The reason
for this should be obvious on an intuitive level. Remember, domination is
a containment relation. If A contains B, there is no obvious way in which A
could be to the left of B.Think of it this way. If you have a box, and the box
has a ball in it, you can't say that the box is to the left of the ball. That is
physically impossible. The box surrounds the ball. The same holds true
for domination. You can't both dominate and precede/follow.
The second part of the definition says "A or some node dominating
A sister-precedes B or some node dominating B" This may seems like an
overly complexway to say "to the left," but there is a good reason we phrase
it this. This has to do with the fact that the terminals of a tree don't float out
in space, they are dominated by other nodes that might precede or follow
themselves and other nodes. Consider the following tree drawn
by a sloppy tree-drawer:
2Thanks to Dave Medieros for helpful discussion of these notions.
112
33)
Preliminaries
N
doberman
In this sloppily drawn tree, the verb kissed actually appears to the left of
the noun clown. However, we wouldn't want to say that kissed precedes clown;
this is clearly wrong. The sentence is said "The clown kissed the doberman,"
where kissed follows clown. We guarantee this ordering by making reference
to the material that dominates the nodes we are looking at. Let A = clown
and B = kissed. Let's substitute those into the definition:
34) [N clown] or some node dominating [N clown] (in this case NP) sisterprecedes
[vkissed] or some node dominating [vkissed] (in this case VP).
This means that [N clown] precedes [v kissed], because NP precedes VP.
Note that precedence holds over all nodes not just terminals. So [N clown]
also precedes [NP thedoberman].
The second clause of the definition also allows us to explain an
important restriction on syntactic trees: You cannot allow branches to cross.
Trees like (35) are completely unacceptable (they are also impossible
to generate with phrase structure rules - try to write one and you'll see):
35)
In this tree, Q is written to the left of R, apparently preceding R, but by the
definition of precedence given above, this tree is ruled out. Q is to the left
of R, but O whichdominates Q isnot. In other words, you can't cross branches.
Another way of phrasing this is given below in (36):
36) No crossing branches constraint: If one node X precedes another node Y
then X and all nodes dominated by X must precede Y and all nodes
dominated by Y.
Chapter 4: Structural Relations 113
You now have enough information totry General Problem Set5 and Challenge
Problem Set 1
Just as in the domination relation, where there is the special local
definition called "immediate domination," there is a special local form
of precedence called immediate precedence:
37) Immediate precedence: A immediately precedes Bif there is no node G that
follows A but precedes B.
Consider the string given in (38) (assume that the nodes dominating this
string meet all the criteria set out in (32)):
38) A B G
In this linear string, A immediately precedes B, because A precedes B and
there is nothing in between them. Contrast this with (39):
39) A G B
In this string, A does not immediately precede B. It does precede B, but G
intervenes between them, so the relation is not immediate.
You nowhave enough information totry General Problem Set 6
4. C-COMMAND
Perhaps the most important of the structural relations is the one we call
c-command. Although c-command takes a little getting used to, it is actually
the most useful of all the relations. In the next chapter, we'll look at the
phenomenon of binding, which makes explicit reference to the c-command
relation. C-command is defined intuitively in (40) and more formally in (41):
40) C-command (informal): A node c-commands its sisters and all
the daughters (and granddaughters and great-granddaughters,
etc.) of its sisters.
41) C-command (formal): Node A c-commands node B if every3 node
dominating A also dominates B, and neither A nor Bdominate the other.
Look at the tree in (42). The node A c-commands all the nodes in the circle.
It doesn't c-command any others:
3 The usual requirement on c-command is that every branching node dominating A also
dominate B. This additional branching requirement isn't necessary given the irreflexive
definition of domination (i.e. proper domination) that we've given above. However, students
may run into the branching definition in other works if other definitions of domination are used.
114 Preliminaries
42)
That is, A c-commands its sister (B) and all the nodes dominated by its sister
(C, D, E, F, G, H, I, J). Consider now the same tree without the circle,
and look at the nodes c-commanded by G:
43) M
N O
D
H
I J
G c-commands onlyH (its sister). Notice that it does not c-command C, E, F,
I, or J. C-command is a relation that holds between sisters and aunts and
nieces. It never holds between cousins or between a mother and daughter.
You nowhave enough information to tryGeneral Problem Set 7 andChallenge
Problem Set 2
There are various kinds of c-command. The first of these is when
two nodes c-command one another. This is called symmetric c-command
and is defined in (44):
44) Symmetric c-command: A symmetrically c-commands B, if A c-commands
B and B c-commands A.
Chapter 4: Structural Relations 115
This relation holds only between sisters. The other kind of c-command
is the kind that holds between an aunt and her nieces. This is called
(unsurprisingly) asymmetricc-command:
45) Asymmetric c-command: A asymmetrically c-commands B if A
c-commands B but B does not c-command A.
Consider again the tree in (42); N and O symmetrically c-command
each other (as do all other pairs of sisters). However, N asymmetrically
c-commands A, B, C, D, E, F, G, H, I, and J, since none of these c-command N.
You nowhave enough information to try General Problem Set 8
Just as we had local (immediate) versions of domination and precedence,
there is a local version of c-command. This is typically called government4
(rather than immediate c-command). There are a number of different
definitions for government. If you look back at our definitions for immediate
precedence and immediate domination, you'll see that in both cases
the locality (i.e., the closeness) of the relationship was defined by making
reference to a potential intervening category. So for domination, some node
A immediately dominates B another provided there is no intermediate node
G that A dominates and that dominates B. In (46a) there is no node between
A and B, so A immediately dominates B.In (46b) by contrast G is in between
them, so A does not immediately dominate B.
46) a) A b) A
B
Thesame idea played a role in precedence. In (47a), A immediately precedes
B because there is nothing between them; in (47b) A precedes B,
but it doesn't immediately precede B, because G intervenes.
47) a) M b) M
A B
Government is similarly defined:
4 Technically speaking, government isn't just immediate c-command, it also expresses
a licensingrelationship (that is it has or had the specialstatus of a constraint on the grammar).
In this book, this licensing function isn't going to be used, so we're going to concentrate on
the structural relationship part of the definition only.
116 Preliminaries
48) Government (first version): Node A governs node B if A c-commands B,
and there is no node G, such that G is c-commanded by A
and G c-commands B.
To see this at work, look at the tree in (49):
49) M
In this tree, A governs B. It c-commands B, and there is no node that
c-commands B that A also c-commands. (You should note that A also
governs N under this definition, A c-commands N, and there is no node
that N c-commands that also c-commands A. The reverse is also true N
governs A because the relationship between A and N is symmetric
c-command. B does not govern A, because B does not c-command A.)
Contrast this with the tree in (50):
50)
intervenes between A and B
Here A does not govern B, because the node G intervenes(or more precisely,
A c-commands G and G c-command B,thus violating the definition).
Government is often "relativized" to the particular kind ofelement that's
doing the government. Forexample, if thegovernor (theelementdoing the
governing is a phrase (an NP, a VP, etc.), then what counts as an intervener
are only other phrases, not heads like N, V, etc.
51) MP donotcount as intervenersfor
phrase government
Chapter 4: Structural Relations 117
In (51) the AP phrase-governs5 B. G and M don't count as interveners, even
though they both are c-commanded by AP and they both c-command B.
This is because they are not phrases - they are heads. GP and BPdon't count
as interveners either, because they don't command B; they dominate it.
Similarly, if the governor is a head (head-government), then phrasal
interveners don't count:
52) AP donot countas intervenersfor
head-government
G B
In (52), MP and GP do not count as interveners for A head-governing B
because they are phrases. M and G don't count because they don't
c-command B.6 With this in mind we can revise the definition:
53) Government
Node A governs node B if A c-commands B, and there is no node G,
such that G is c-commanded by A and G c-commands B.
• Phrase-government: If A is a phrase, then the categories that count
for G in the above definition must also be phrases
• Head-government: If A is a head (word), then the categories that
count for G in the above definition must also be heads.
In recentyears, government has to a greater or lesser degree fallen out
of fashion. Instead local relations previouslylinked to government are often
determined by what is called the specifier-head relation. However, it is
important toknow whatgovernment is, because ifyouread many influential
papers in syntax they will refer to this relation.
You now have enough information totry General Problem Set9
5 In the syntactic literature, this is more usually called antecedent government (which has
an additionalconstraintcalledcoindexing on it and is definedover particular categories - so NP
antecedent governs another coindexed NP, provided there is no intervening c-commanding NP
that also is c-commanded by the first). Thisis a refinement that we won't pursue here because
it is rarely used anymore.
6These don't c-command Bonlyifthebranching requirement on c-command doesnot hold.
118 Preliminaries
5. Grammatical Relations
In addition to the structural relations that hold between items in a tree, there
are also some traditional grammatical terms that can be defined structurally.
These are useful terms, and we will frequently make reference to them.
We call these grammatical relations. Technically speaking, grammatical
relations are not structural relations. Some theories of grammar (for example
Lexical Functional Grammar and Relational Grammar) posit primitive
grammatical relations (meaning they are not structurally defined).
In the approach we are developing here, however, grammatical relations
are defined structurally; that is, they are defined in terms of the tree.
In English the subject is always the NP or CP that appears before
the verb or auxiliary:
54) a) The puppy licked the kitten's face.
b) It is raining.
c) Fred feels fine.
d) The kitten was licked.
e) That Bill's breath smells ofonions bothers Erin.
Notice thatthedefinition ofsubject isnota semantic one. Itis notnecessarily
the doer of the action. In (54c) for example, Fred is not deliberately feeling
fine. In sentence (54d), the kitten is the one being licked, not the licker.
Different semantic types7 ofnoun phrases appear tobe allowed tofunction
as thesubject. There is a straightforward structural definition ofthesubject:
55) Subject (preliminary): NPor CPdaughter ofTP
In later chapters, we will have cause to refine this definition somewhat,
but for now, this will do.
Next we have the direct object of the verb and the object of
a preposition. Examples of theseare seen in (56) and (57) respectively:
56) Direct object
a) Susan kissed the clown's nose.
b) Cedric danced ajolly jig.
c) Dale said that the lawn was overgrown.
57) Object ofapreposition
a) Gilgamesh cut the steak with aknife.
b) We drove all the way to Buenos Aires.
7In chapter 8, we will look at different semantic types ofnoun phrases. These types are called
thematic relations.
Chapter 4: Structural Relations 119
Preliminary definitions of these are given in (58) and (59), again we will have
reason to revise these in later chapters.
58) (Direct) object (preliminary): NP or CP daughter of a VP headed by a
transitive verb
59) Object ofpreposition: NP daughter of PP
To see these definitions at work consider the following tree. The NP Les
is the daughter ofTP, and is thus the subject. TheNP Paula is a daughter of
the VP headed by the transitive verb kissed, so Paula is the direct object.
Tuesday is the NP daughter of a PP, thus the object of a preposition.
60) TP
subject >NP VP
N V NP PP
Les kissed
/ N P NP <r- object ofaP
direct object Paula on |
N
Tuesday
In addition to direct objects, when you have a ditransitive verb like give
or put, you also have an indirect object. Indirect objects in English come
ofseveral types in terms ofthe types ofarguments they take. The two most
common types are the direct object which we preliminarily defined above
and indirect object. The indirect object in English shows up in two places.
It can be the PP that follows the direct object:
61)
N V NP PP <£— indirectobject
Les gave
/ N P NP
direct object peanuts to
N
Paula
It canalsobe thefirst NPafterthe verbwhen the verb takestwo NPs:
120 Preliminaries
62)
N V NP NP <— direct object
Les gave 7* |
/ N N
indirect object Paula peanuts
Notice that the directobject is the second ofthe two NPs, roughly the reverse
ofthe treein (61). Thismeanscomplicating our definitions somewhat:
63) Direct Object (second pass):
a) With verbs of type V(NP_NP1, V,NP_CP] and V(NP_ NP PP], the NP or CP
daughter of VP
b) With verbs of type V [NP _ NP ,NP/CP)1, An NP or CP daughter of
VPthat is preceded by another NP daughter of VP. (i.e., the second
NP daughter of VP)
64) Indirect Object (preliminary):
a) With verbs oftype V[NP_ NP PP1, the PP daughter of VP immediately
preceded by an NP daughter of VP.
b) With verbs of type V|NP _ NP {NP/CP)], the NP daughter of VP
immediatelyprecededby V(i.e., thefirst NP daughter ofVP).
In addition to subjects, objects, and indirect objects, you may also
occasionally see reference made to obliques. In English, obliques are
almost always marked with a preposition. The PPs inthe following sentence
are obliques:
65) John tagged Lewis [PP with aregulation baseball][PP on Tuesday].
In many languages, such as Finnish, obliques aren't marked with
prepositions, instead they get special suffixes that mark them as oblique;
so obliqueness isnotnecessarily defined by being a preposition, that isjust
a convenient definition fornow. Notice that obliques canstructurally show
up in the same position as indirect objects (compare (66a) to (66b)).
The difference between the two is inwhether the PP ispartofthe argument
structure oftheverb ornot. Ifverb isoftype V[NP_NPPP, like give, thenthePP
is an indirect object, but if the verb is of type V(NP _ NP, (where the PP
isn'tspecified by the feature), like eat, then the PP is an oblique.
66) a)
b)
Chapter 4: Structural Relations
N V NP PP <r- indirect object
Les gave | l^"""*---^
N P NP
peanuts to |
N
Paula
oblique
N P NP
peanuts with |
N
jam
You now have enough information to try General Problem Sets 10-16and
Challenge Problem Set 3
121
6. Summary and Conclusions
This chapter has been a bit different from the rest of this book. It hasn't
been about Language per se, but rather about the mathematical properties
of the system we use to describe language. We looked at the various parts
of a syntactic tree and then at the three relations that can hold between
these parts: domination, precedence, andc-command. Inall the subsequent
chapters ofthis book, you'll find much utility for theterms andtherelations
described here.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Branch: A line connecting two parts of a tree,
ii) Node: The end of a branch.
122 Preliminaries
iii) Label: The name given to a node (e.g.,N, NP, TP, etc.).
iv) (Proper) Domination: Node A dominates node B if and only if A
is higher up in the tree than Band if you can trace a branch from A
to B going only downwards.
v) Immediately Dominate: Node A immediately dominates node B if
there is no intervening node G that is dominated by A, but
dominates B. (In other words, A is the first node that dominates B.)
vi) A is the mother of BifA immediately dominates B.
vii) Bis thedaughter ofAifBisimmediately dominatedby A.
viii) Sisters: Two nodes that share the same mother.
ix) Root Node (revised): The node that dominates everything, but
is dominated by nothing. (The node that is no node's daughter.)
x) Terminal Node (revised): A node that dominates nothing. (A node
that is not a mother.)
xi) Non-terminal Node (revised): A node that dominates something.
(A node that is a mother.)
xii) Exhaustive Domination: Node A exhaustively dominates a set of
terminal nodes {B, C, ..., D}, provided it dominates all the members
of the set (so that there is no member of the set that is not dominated
by A) and there is no terminal node G dominated by A that is not
a member of the set.
xiii) Constituent: A set of terminal nodes exhaustively dominated by
a particular node.
xiv) Constituent of: Aisa constituent ofBifand only ifBdominates A.
xv) Immediate Constituent of.A is an immediate constituent of Bif and
only if Bimmediately dominates A.
xvi) Sister Precedence: Node Asister-precedes node Bifand only ifboth
are immediately dominated by the same node, and A appears to
the left of B.
xvii) Precedence: Node A precedes node B if and only if neither A
dominates Bnor Bdominates A and A or some node dominating A
sister-precedes Bor some node dominating B.
Chapter 4: Structural Relations 123
xviii) No Crossing Branches Constraint: If node X precedes another node
Y then X and all nodes dominated by X must precede Y and all
nodes dominated by Y.
xix) ImmediatePrecedence: A immediately precedes B if there is no node
G that follows A but precedes B.
xx) C-command (informal): A node c-commands its sisters and all
the daughters (and granddaughters, and great-granddaughters, etc.)
of its sisters.
xxi) C-command (formal): Node A c-commands node B if every node
dominating A also dominates B and neither A nor B dominates
the other.
xxii) Symmetric C-command: A symmetrically c-commands B if A
c-commands B and B c-commands A.
xxiii) Asymmetric C-command: A asymmetrically c-commands B if A
c-commands B but B does not c-command A.
xxiv) Government: Node A governs node Bif A c-commandsB, and there
is no node G, where G is c-commanded by A and G c-commands B.
• Phrase-government: If A is a phrase, then the categories that
count for G in the above definition must also be phrases
• Head-government: IfA is a head (word), then thecategories that
count for G in the above definition must also be heads.
xxv) Subject (preliminary): NP or CPdaughter ofTP.
xxvi) Object ofPreposition (preliminary): NP daughter of PP.
xxvii) Direct Object:
a) With verbs of type V|NP_NP)/ V(NP_ CP) and V[NP_ NP PP1, the NP
or CP daughter of VP.
b) With verbsof type V[NP _ NP [mCe\y an NP or CP daughter of VP
that is preceded by another NPdaughterof VP (i.e., the second
NP daughter of VP).
xxviii) Indirect Object (preliminary):
a) With verbs of type V|NP_ NP m the PP daughter of VP
immediately preceded by an NP daughter of VP.
b) With verbs of type V|NP _ NP {NP/Cv)V tne NP daughter of VP
immediately preceded byV(i.e., thefirstNPdaughter ofVP).
xxix) Oblique: any NP/PP in the sentence that is not a subject, object of
a preposition, or indirect object.
124 Preliminaries
Further Reading
Barker, Chris and Geoffrey Pullum (1990) A theory of command relations.
Linguistics and Philosophy 13, 1-34.
Chomsky, Noam (1975) The Logical Structure of Linguistic Theory. New York:
Plenum.
Higginbotham, James (1985) A note on phrase markers. MIT Working Papers
in Linguistics 6, 87-101.
Reinhart, Tanya (1976) The Syntactic Domain of Anaphora. Ph.D.
dissertation, MIT.
Reinhart, Tanya (1983) Anaphora and Semantic Interpretation. London: Croom
Helm.
General Problem Sets
1. Trees
[Application of Skills; Basic to Intermediate]
Using the rules we developed in chapter 3, draw the trees for the following
sentences. Many of the sentences are ambiguous. For those sentences
draw one possible tree, indicating the meaning by providing a paraphrase.
a) The big man from New Yorkloves bagels with cream cheese.
b) Susan rode a bright blue train from New York.
c) The plucky platypus kicked a can of soup from New York to Tucson.
d) John said Martha sang the aria with gusto.
e) Martha said John sang the aria from La Boheme.
f) The book of poems with the bright red cover stinks.
g) Louis hinted Mary stole the purse deftly.
h) The extremely tired students hated syntactic trees with a passion,
i) Many soldiers have claimed bottled water quenches thirst best,
j) Networking helps you grow your business.
Chapter 4: Structural Relations 125
2. Domination
[Application of Skills; Basic]
Study the following tree carefully and then answer the questions about it
that follow:
IPi
Di
the baker'
Vi
said
VPi
C
that
D2
his
CP
TP2
NP2
N2
bread
VP2
V2
smelled
AdvP
I
Adv
glorious
1) List all the nodes that dominate D, the.
2) List all the nodes that dominate D2his.
3) List all the nodes that dominate Ni baker.
4) List all the nodes that dominate N2 bread.
5) List all the nodes that dominate V, said.
6) List all the nodes that dominate V2 smelled.
7) List all the nodes that dominate Adv glorious.
8) List all the nodes that dominate C that.
9) Listall the nodes that dominate TPi (ifthere are any).
10) List all the nodes that dominate TP2.
11) List all the nodes that dominate NPi.
12) List all the nodes that dominate NP2.
13) List all the nodes that dominate VPV
14) List all the nodes that dominate VP2.
15) List all the nodes that dominate CP.
16) List all the nodes that dominate AdvP.
17) What is the root node?
18) List all the terminal nodes.
19) List all the non-terminal nodes.
20) List all the nodes that VP2 dominates.
21) List all the nodes that CP dominates.
22) List all the nodes that NPi dominates.
3. Exhaustive Domination
[Application of Skills; Intermediate]
Refer back to the tree for problem set 2 to answer this question.
1) In the tree, is the set of terminals {Ni,
a single node? Ifso, which one?
N2} exhaustively dominated by
126 Preliminaries
2) In the tree, is the set {D1( N^ exhaustively dominated by a single node?
If so, which one?
3) In the tree, is the set {V2, Adv}exhaustively dominated by a single node?
If so, which one?
4) In the tree, is the set {D2, N2i V2, Adv} exhaustively dominated by a single
node? If so, which one?
5) In the tree, is the set {D1( N^ V^ exhaustively dominated by a single
node? If so, which one?
6) In the tree, is the set {D^ exhaustively dominated by a single node?
If so, which one?
7) In the tree, is the set {C, D2, N2, V2, Adv} exhaustively dominated by
a single node? Ifso, which one?
8) What is the set of terminal nodes exhaustively dominated by VPi?
9) Is the string that his bread a constituent? Explain your answer using
the terminology of exhaustive domination.
10) Is the string The baker said that his bread smelted glorious a
constituent? Explain your answer using the terminology of exhaustive
domination.
11) Is NPi a constituent of TP!?
12) Is NP2 a constituent of TPi?
13) Is NPi a constituent of TP2?
14) Is NP2 a constituent of TP2?
15) Is V2 a constituent of CP?
16) Is VP2 a constituent of CP?
17) Are both Adv and AdvP constituents of VP2?
4. Immediate domination
[Application of Skills; Basic]
Go back to problem set 2, study the tree again and answer the questions
(1-16) as in problem set 2, except limiting your answer to immediate
domination instead of domination.
5. Precedence
[Application of Skills; Basic]
Go back to problem set 2, study the tree again and answer the questions
(1-16) except changing domination to precedence (i.e., list all the nodes that
precede D-i etc.). For some elements there may be nothing that precedes
them.
6. Immediate Precedence
[Application of Skills; Basic]
Go back to problem set 2, study the tree again and answer the questions
(1-16) except changing domination to immediate precedence (i.e. list all
the nodes that immediately precede D-i, etc.). For some elements there may
be nothing that immediately precedes them.
Chapter 4: Structural Relations 127
7. C-command
[Application of Skills; Basic]
Go back to problem set 2, study the tree again and answer the following
questions:
1) List all the nodes that Di the c-commands (note NOT the nodes that
c-command Di, but the ones that Di c-commands).
2) List all the nodes that D2his c-commands.
3) List all the nodes that Ni baker c-commands.
4) List all the nodes that N2 bread c-commands.
5) List all the nodes that Vi said c-commands.
6) List all the nodes that V2smelled c-commands.
7) List all the nodes that Adv glorious c-commands.
8) List all the nodes that C that c-commands.
9) List all the nodes that TPi c-commands (ifthere are any).
10) List all the nodes that TP2 c-commands.
11) List all the nodes that NPi c-commands.
12) List all the nodes that NP2 c-commands.
13) List ail the nodes that VP! c-commands.
14) List all the nodes that VP2 c-commands.
15) List all the nodes that CP c-commands.
16) List all the nodes that AdvP c-commands.
17) What nodes c-command TP2?
18) What nodes c-command NP!?
19) What nodes c-command C?
8. Symmetric and Asymmetric C-command
[Application of Skills; Basic/Intermediate]
Go back to problem set 2, study the tree again and answer the following
questions. For some questions the answer may be "none":
1) List allthe nodes that Di thesymmetrically c-commands.
2) List allthe nodes that D2 his symmetrically c-commands.
3) List all the nodes that Ni baker symmetrically c-commands.
4) List all the nodes that N2 bread symmetrically c-commands.
5) List allthe nodes that Vi said symmetrically c-commands.
6) List allthe nodes that V2 smelledsymmetrically c-commands.
7) List allthe nodes that Adv glorious symmetrically c-commands.
8) Listall the nodes that C thatsymmetricallyc-commands.
9) List allthe nodes that TP! symmetrically c-commands (ifthere are any).
10) List all the nodes that TP2 symmetrically c-commands.
11) Listall the nodes that NPi symmetricallyc-commands.
12) Listall the nodes that NP2 symmetrically c-commands.
13) Listall the nodes that VPi symmetrically c-commands.
14) List all the nodes that VP2symmetrically c-commands.
15) List all the nodes that CP symmetrically c-commands.
16) Listall the nodes that AdvP symmetrically c-commands.
128 Preliminaries
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
List all the nodes that Di the asymmetrically c-commands.
List all the nodes that D2his asymmetrically c-commands.
List all the nodes that Ni baker asymmetrically c-commands.
List all the nodes that N2 bread asymmetrically c-commands.
List all the nodes that Vi said asymmetrically c-commands.
List all the nodes that V2 smelled asymmetrically c-commands.
List all the nodes that Adv glorious asymmetrically c-commands.
List all the nodes that C that asymmetrically c-commands.
List all the nodes that TPi asymmetrically c-commands (ifthere are any).
List all the nodes that TP2 asymmetrically c-commands.
List all the nodes that NP! asymmetrically c-commands.
List all the nodes that NP2 asymmetrically c-commands.
List all the nodes that VPi asymmetrically c-commands.
List all the nodes that VP2 asymmetrically c-commands.
List all the nodes that CP asymmetrically c-commands.
List all the nodes that AdvP asymmetrically c-commands.
What nodes asymmetrically c-command V2?
What nodes symmetrically c-command NPi?
What nodes asymmetrically c-command C?
What nodes symmetrically c-command C?
9. Government
[Applicationof Skills; Intermediate/Advanced]
Go back to problem set 2, study the tree again and answer the following
questions:
1) Does NPi govern VP2? Why or why not?
2) Does NP! govern C that? Why or why not?
3) What nodes does Ni govern?
4) Does Vi head-govern V2? Why or why not?
5) What node(s) does C that head-govern?
6) Does NPi phrase-govern AdvP? Why or why not?
7) Does VP2 phrase-govern N2?Why or why not?
10. Grammatical Relations I
[Application of Skills; Intermediate]
Go back to problem set 2, study the tree again and answer the following
questions:
1) What is the subject of TP1 ?
2) What is the subject of TP2?
3) What is the object of VP1 ?
4) Does VP2 have an object? Why or why not?
Chapter 4:Structural Relations
11. Grammatical Relations II
[Application of Skills; Intermediate]
Examine the following tree and then answer the questions that follow:
129
N3 V2 NP4 PPi PP2
Alyssa put | /\ / \
N4 Pi NP5 P2 NP6
peanuts in /\ on |
Di N5 N6
his ear Tuesday
1) What is the subject of TPi?
2) What is the subject of TP2?
3) What is the object of Pi?
4) What is the object of P2?
5) What is the direct object of VPi?
6) What is the direct object of VP2?
7) What is the indirect object of VPi?
8) What is the indirect object of VP2?
9) Is PP2 an indirect object or an oblique. How can you tell?
12. Grammatical Relations III8
[Application of Skillsand Data Analysis;Basic]
For each of the following sentences, identify the subject, the object (if there
is one), the indirect object (if there is one), any objects of prepositions,
the verb, and any obliques. Draw the tree for each sentence.
a) It never rains violently in southern California.
b) Soon we should give the family dog another bath.
c) The quiz show contestant bravely made a wild guess about the answer.
Problem set contributed by Sheila Dooley.
130 Preliminaries
13. Structural Relations I9
[Application of Skills; Advanced]
Consider the following tree:
apples from
D3 N3
the grocer
1) What node(s) dominate N3 grocer?
2) What node(s) immediately dominate D3 the?
3) Do T willand V buy form a constituent?
4) What nodes does ^ bullyc-command?
5) What nodes does NPi the big bullyc-command?
6) What is V buy's mother?
7) What nodes does T willprecede?
8) List all the sets of sisters in the tree.
9) What is the PP's mother?
10) Do NPi and VP asymmetrically or symmetrically c-command one
another?
11) Listall the nodes c-commanded by V.
12) What is the subject of the sentence?
13) What is the object of the sentence?
14) What is the object of the preposition?
15) Is NP3 a constituent of VP?
16) What node(s) is NP3 an immediate constituent of?
17) What node(s) does VP exhaustively dominate?
18) What is the root node?
19) List all the terminal nodes.
20) What immediately precedes N3 grocer?
14. Structural Relations II
[Application of Skills; Advanced]
Look at your tree for sentence (a) of problem set 1. Number the nodes
the way Idid for problem set 13.
1) List all the nodes that the subject NP c-commands.
2) List allthe nodes that the subject NP asymmetrically c-commands.
3) List all the nodes that the subject NP dominates.
'Theidea forthis problemset is borrowedfromRadford (1988).
Chapter 4:Structural Relations 131
4) List all the nodes that the subject NP immediately dominates.
5) List all the nodes that the subject NP precedes.
6) List all the nodes that the VP node c-commands.
7) List all the nodes that the VP asymmetrically c-commands.
8) List all the nodes that the VP dominates.
9) List all the nodes that the VP immediately dominates.
10) List all the nodes that the VP precedes.
11) List all the nodes that the VP follows (i.e., is preceded by).
15. Tzotzil
[Data Analysis; Basic]
Tzotzil is a Mayan language spoken in Mexico. Consider the following
sentences, then answer the questions that follow. Glosses have been
simplified and the orthography altered from the original source. (Data from
Aissen 1987.)
a) 'ispet lok'el 'antz ti t'ule.
carry away woman the rabbit
"The rabbit carried away (the) woman."
b) 'ibat xchi'uk smalal Ii Maruche.
go with her-husband the Maruche
"(the) Maruch went with her husband." (Maruche is a proper name.)
c) Pas ti 'eklixa'une.
built the church
"The church was built."
1) What is the NP rule for Tzotzil?
2) What is the PP rule for Tzotzil?
3) What is the VP rule for Tzotzil?
4) What is the TP rule for Tzotzil?
5) What is the subject of sentence (b)?
6) Is [the church] a subject or an object of sentence (c)?
7) Does the verb precede the subject in Tzotzil?
8) Does the object precede the subject in Tzotzil?
9) Does the verb precede the object in Tzotzil?
10) Using the rules you developed in (1-4) above, draw the trees for (b) and
(c).
16. HlAKI
[Data Analysis; Intermediate]
Consider the data from the following sentences of Hiaki (also known as
Yaqui), an Uto-Aztecan language from Arizona and Mexico. Data have been
simplified. (Data from Dedrick and Casad 1999.)
a) Tekil ne-u 'aayu-k.
work me-for is
"There is work for me." (literally: "Work is for me.")
132 Preliminaries
b) Hunaa'a ya'uraa hunaka'a hamutta nokriak.
that chief that woman defend
"That chief defended that woman."
c) Taawe totoi'asb'olam kaamom6lim hiba-tu'ure.
Hawk chickens young like
"(The) hawk likes young chickens."
d) Ta'abwikasu 'ama yepsak.
different-person there arrived
"Adifferent person arrived there." (assume there is an adverb not a N)
Assume the rules AdjP -> Adj and AdvP -> Adj and answer the following
questions.
1) What is the NP rule for Hiaki?
2) Do you need a PP rule for Hiaki? Why or why not?
3) What is the VP rule for Hiaki?
4) What is the TP rule for Hiaki?
5) Using the rules you developed in questions 1-4, draw the tree for
sentences (b, c, d).
6) What is the subject of sentence (b)?
7) Is there an object in (d)? Ifso, what is it?
8) What node(s) does hunaa'a c-command in (b)?
9) What node(s) does hunaa'a ya'uraa c-command in (b)?
10) What does 'dma precede in (d)?
11) What node immediately dominates kiamomdlim in (c)?
12) What nodes dominate kaamomdlim in (c)?
13) What node immediately precedes kaamomdlim in (c)?
14) What nodes precede kaamomdlim in (c)?
15) Does kaamomdlim c-command teawe in (c)?
16) Do hunaka'a and hamutta symmetrically c-command one another in (b)?
Challenge Problem Sets
Challenge Problem Set 1: Discontinuous Constituents
[Critical Thinking; Challenge]
Consider the following data:
a) A woman entered who was eating a chocolate enchiladas.
b) The man that Bill said that Mary disliked loves beef waffles.
With sentence (a) assume that the relative clause [who was wearing a hat] is
a modifier of man. Assume that the man is both the direct object of the verb
disliked and the subject of the verb loves. Is it possible to draw trees
for these sentences without crossing lines? Explain why or why not.
Chapter 4:Structural Relations 133
Challenge Problem Set 2: Negative Polarity Items
[Critical Thinking; Challenge]
There is a class of phrase, such as [a red cent] and [a single thing], that are
called Negative Polarity Items (NPI). These are only allowed in sentences
with a negative word like not. So for example, in sentences (a) and (c)
the NPI is fine, in the (b) and (d) sentences, however, the sentence is at best
strange.
a) I didn't have a red cent.
b) *l had a red cent, (ungrammatical withidiomatic reading)
c) I didn't read a single book the whole time I was in the library.
d) *l read a single book the whole time I was in the library.
It turns out that sentences with NPIs not only must have a word like not, they
also have to be in a particular structural relationship with that not word.
On the basis of the following sentences figure out what that relationship is.
There are two possible answers consistent with this data. Assume that not
and n't are dominated by the VP node.
e) I did not have a red cent.
f) *A red cent was not found in the box.
What kind of data would you need to decide between the two possible
answers to this question?
Challenge Problem Set 3: Irish
[Data Analysis and Critical Thinking: Challenge]
Consider the following data from Modern Irish Gaelic:
a) Phog Liam Sean,
kissed William John
"William kissed John."
b) Phog Sean Liam.
Kissed John William
"John kissed William."
c) Phog an fear an mhuc.
kissed the man the pig
"The man kissed the pig."
d) Chonaic me an mhuc mhor.
Saw I the pig big
"I saw the big pig."
e) Rince an bhean.
Danced the woman
"The woman danced."
On the basis of this data answer the following questions:
134 Preliminaries
1) What is the AdjP rule in Irish (if there is one)? Constrain your answer to
the data here.
2) Write the NP rule for Irish, be sure to mark optional things in
parentheses.
3) Can you write a VP rule for Irish? Assume that if you have a VP then
object NPs (like William in (b) and the big pig in (d)) must be part of the
VP, and that subject NPs (like John in (b) and / in (d)) are never part of
VPs. Is it possible to keep those assumptions and not cross lines? If you
can't, then don't posit a VP.
4) If you don't have a VP rule for Irish, then how do we define direct object
in this language?
5) What is the TP rule for Irish? (Be careful that your TP rule is consistent
with your answer in (3).)
6) Using the rules you developed, draw trees for sentences (c), (d) and (e).
Binding Theory
0. INTRODUCTION
Let's leave syntax for a moment and consider some facts about the meaning
of NPs in English. There are some NPs that get their meaning from
the context and discourse around them. For example, in the sentence in (1),
the meaning of the word Felicia comes from the situation in which the
sentence is uttered:
1) Felicia wrote a fine paperon Zapotec.1
If you heard this sentence said in the real world, the speaker is assuming
that you know who Felicia is and that there is somebody called Felicia who
is contextually relevant. Although you may not have already known that she
wrote a paper on Zapotec, this sentence informs you that there is some paper
in the world that Felicia wrote, and it's about Zapotec. It presupposes
that there is a paper in the real world and that this paper is the meaning
of the phrase afine paper onZapotec. Both afine paper onZapotec and Felicia get
their meaning by referring toobjects intheworld.2 This kind ofNP iscalled
a referring expression (or R-expression):
2) R-expressiojr. An NP that gets its meaning by referring to an entity in
the world.
1Zapotec is a languagespokenin southernMexico.
" This is true whether the world being referred to is the actual world, or some
fictional imaginary world created by the speaker/hearer.
136 Preliminaries
The vast majority of NPs are R-expressions. But it is by no means the case
that all NPs are R-expressions. Consider the case of the NP herself
in the following sentence:
3) Heidi bopped herselfon the head with a zucchini.
In this sentence, Heidi is an R-expression and gets its meaning from the
context, butherselfmust refer back to Heidi. It cannot refer to Arthur, Miriam,
or Andrea. It must get its meaning from a previous word in the sentence
(in this case Heidi). This kind of NP, one that obligatorily gets its meaning
from anotherNP in the sentence, is called an anaphor (as we saw in chapter 1).
4) Anaphor: An NP that obligatorily gets its meaning from another NP in
the sentence.
Typical anaphors are himself, herself, themselves, myself, yourself, and each other.
Types of Anaphors
There are actually (at least) two different kinds of anaphors. One type is
the reflexive pronouns like herself, himself, and themselves. The other kind
are called reciprocals, and include words like each other. For our purposes,
we'll just treat this group like a single class, although there are minor
differences between the distribution of reflexives and reciprocals.
There is yet another kind of NP. These are NPs that can optionally
get their meaning from another NP in the sentence, but may also optionally
get it from somewhere else (including context or previous sentences
in the discourse). These NPs are called pronouns: Look at the sentence in (5):
5) Art said that he played basketball.
In this sentence, the word hecan optionally refer to Art (i.e., the sentence can
mean "Art said that Art played basketball") or it can refer to someone else
(i.e. "Art said that Noam played basketball"). Typical pronouns include: he,
she, it, I, you, me, we, they, us, him, her, them, his, her, your, my, our, their, one.
A definition of pronoun is given in (6):
6) Pronoun: An NP that may (but need not) get its meaning from another
word in the sentence.
There is some discrepancy among linguists in the use of this term. Some linguists
use the term pronominal instead of pronoun and use the term pronoun to cover both
anaphors and pronominals. This distinction, while more precise, is confusing to
the beginner, so for our purposes we'll just contrast pronouns to anaphors, and avoid
the term pronominal.
Chapter 5: Binding Theory 137
Getting back to syntax, it turns out that these different semantic types
of NPs can only appear in certain syntactic positions that are defined using
the structural relations we developed in the last chapter. Anaphors,
R-expressions, and pronouns can only appear in specific parts of the
sentence. For example, an anaphor may not appear in the subject position
of sentence:
7) *Herself bopped Heidi on the head with a zucchini.
The theory of the syntactic restrictions on where these different NP types can
appear in a sentence is calledBinding Theory and is the focus of this chapter
and makes reference to the structural relations we learned about
in the previous chapter. This chapter thus will be your first exposure
to why structural relations are so important to linguists.
You now have enough information totry General Problem Set 1
1. The Notions Coindex and Antecedent
We're going to start with the distribution of anaphors. First, we need some
terminology to set out the facts. An NP that gives its meaning to another
noun in the sentence is called the antecedent:
8) Antecedent4: An NP that gives its meaning to another NP.
Forexample, in sentence (3) (repeated here as 9), the NP Heidi is the source
ofthe meaningfor the anaphor herself, so Heidi is called the antecedent:
9) Heidibopped herselfon the head with a zucchini.
T t
antecedent anaphor
We use a special mechanism to indicate that two NPs refer to the same
entity. After each NP we write a subscript letter. Ifthe NPsrefer to the same
entity, then theyget thesame letter. Iftheyrefer todifferent entities theyget
different letters. Usually we start (as a matter of tradition) with the letter
/ and work our way down the alphabet. These subscript letters
are called indices or indexes (singular: index).
4In Latin the prefix ante means "before." However, in the system we are developing
here, antecedents do not need to precede the noun they give their meaning
to (althoughthey frequently do). In some cases the antecedentmay follow the noun
that it gives its meaning to: e.g., Everyone who knows him loves Dan. Him can get
its meaning from Dan, even though Danfollows him.
138 Preliminaries
10) a) [Colinjj gave [Andreajj [abasketball]k.
b) [Artjj said that [he]j played [basketball]k in [the darkjj.
c) [Artji said that [he]j played [basketball]k in [the dark],.
d) [Heidijj bopped [herself]; on [theheadjjwith [azucchini]k.
In (10a), all the NPs refer to different entities in the world, so they all get
different indexes. The same is true for (10b). Note that with this indexing, the
sentence only has the meaning where he is not Art, but someone else - the
pronoun heand Art have different indexes. Sentence (10c), by contrast, has he
and Art referring to the same person. In this sentence, Art is the antecedent
of the pronoun he, so they have the same index. Finally in (lOd), the anaphor
herself, by definition, refers back to Heidi so they get the same index. Two
NPs that get the same index are said to be coindexed. NPs that are coindexed
with each other are said to corefer (i.e., refer to the same entity in the world).
11) Coindexed: Two NPs are said to be coindexed if they have the same
index.
In (10c) Art and he are coindexed; in (10b) Art and he are not coindexed.
2. Binding
The notions of coindexation, coreference, and antecedence are actually
quite general ones. They hold no matter what structural position an NP is in
the sentence. It turns out, however, that the relations between an antecedent
and a pronoun or anaphor must bear particular structural relations. Contrast
the three sentences in (12).5
12) a) Heidi; bopped herself; on the head with a zucchini.
b) [Heidij's mother]) boppedherselfj on thehead with a zucchini.
c) *[Heidij's mother])bopped herselfj on thehead with a zucchini.
In particular notice the pattern of indexes on (12b) and (12c). These sentences
show, that while the word herself can refer to the whole subject NP Heidi's
mother, it can't refer to an NP embedded inside the subject NP, such as Heidi.
Similar facts are seen in (13).
13) a) [The motherof Heidijjbopped herselfj on the head with a zucchini,
b) *[The mother of Heidijjbopped herself; on the head with a zucchini.
5Inordertoaccount for these sentences we'llhavetoslightly modify our NPrule:
NP -> (ID/NP's)) (AdjP+) N (PP+)
Chapter 5: Binding Theory 139
A Quick Note on Notation
Syntacticians will sometimes abbreviate two sentences that are otherwise
identical, but have different indices. The two possible indices are
separated by a slash (/) and the index that would make the sentence
ungrammatical is marked with an asterisk (*). So the abbreviated form of
the two sentences in (13) would be:
13') [The mother ofHeidijj bopped herselfj/.; onthehead witha zucchini.
This means that that the version of this sentence where herself is indexed j
(i.e., coindexed with [the mother ofHeidi]j) is grammatical; but when it is
indexed i (i.e., coindexed with IHeidi],) it is ungrammatical.
Look at the trees for (12a and b), shown in (14a and b) below, and you
will notice a significant difference in terms of the position where the NP
immediately dominating Heidi is placed.
14) a) (=12a)
b) (=12b)
N
Heidi's
N V NP,
Heidi bopped
N P NP P NP
herself on / \ with / \
D N D N
the head a zucchini
N P NP P NP
herself on /\ with / \
D N D N
the head a zucchini
In (14a) the circled NP c-commands the NP dominating herself, but in (14b) it
does not. It appears that the crucial relationship between an anaphor and its
antecedent involves c-command. So in describing the relationship between
140 Preliminaries
an anaphor and an antecedent we need a more specific notion
than simple coindexation. This is binding:
15) Binds: A binds B if and only if A c-commands B and A and B are
coindexed.
Binding is a kind of coindexation. It is coindexation that happens when one
of the two NPs c-commands the other. Notice that coindexation alone does
not constitute binding. Binding requires both coindexation and c-command.
You nowhave enough information totryGeneral Problem Set2 and Challenge
Problem Set 1
Now we can make the following generalization, which explains the
ungrammaticality of sentences (16a)(=7)and (16b) (=12c):
16) a) (=7) *Herself;bopped Heidi; on the head with a zucchini.
b) (=12c) *[Heidij's mother])bopped herself; on thehead with a zucchini.
In neither of these sentences is the anaphor bound. In other words, it is not
c-commanded by the NP it is coindexed with. This generalization is called
BindingPrinciple A. Principle A determines the distribution of anaphors:
17) Binding Principle A (preliminary): An anaphor must be bound.
Remember, bound means coindexed with an NP that c-commands it. If you
look at the tree in (14b) you'll see that the anaphor herself, and the NP Heidi
are coindexed. However they are not bound, since [NP Heidi]
does not c-command [NP herself]. The same is true in the tree for (16a) (=7)
shown in (18):
18)
N
Herself
V NP;
bopped 1
N
Heidi
PP PP
/\ /\
P NP P NP
on / \ with /\
D N D N
the head a zucchini
Even though the two NPs are coindexed, they do not form a binding
relation, since the antecedent doesn't c-command the anaphor. You might
Chapter 5:Binding Theory 141
think that Heidi binds herself, since the anaphor c-commands the antecedent6
But notice that this is not the way binding is defined. Binding is not
a symmetric relationship. The binder (or antecedent) must do the
c-commanding of the bindee (anaphor or pronoun), not the reverse.
3. LOCALITY CONDITIONS ON THE BINDING OF ANAPHORS
Consider now the following fact about anaphors:
19) *Heidi; said that herself; discoed with Art.
(cf. Heidi; said that she; discoed with Art.)
A tree for sentence (19) is given below:
20) * TP
N V
herself discoed
P NP
with I
N
Art
As you can see from this tree, the anaphor is bound by its antecedent:
[NPHeidi] c-commands [NP herself] and is coindexed with it. This sentence is
predicted to be grammatical by the version of Principle A presented in (17),
since it meets the requirement that anaphors be bound. Surprisingly,
however, the sentence is ungrammatical. Notice that the difference between
a sentence like (19) and a sentence like (12a) is that in the ungrammatical
(19), the anaphor is in an embedded clause. The anaphor seems to need
to find its antecedent in the same clause. This is called a locality constraint.
6In fact, in this tree /ierse//binds Heidi, and therein lies the problem; anaphors must
be bound, they aren't the binders.
142 Preliminaries
The anaphor's antecedent must be near it or "local" in some way.
The syntactic space in which an anaphor must find its antecedent is called
a binding domain. For the moment let's just assume that the binding domain
is the clause (TP).
21) Binding domain: The clause containing the NP (anaphor, pronoun, or
R-expression).
With this in mind, let's revise Principle A:
22) Binding Principle A (revised): An anaphor must be bound in its binding
domain.
This constraint says that anaphors must find an antecedent within the clause
that immediately contains them.
You now have enough information to try General Problem Set 3
Binding Domain
The definition we've given here for "binding domain" is clearly oversimplistic.
For example, when there is an NP that contains an anaphor
and an NP marked with 's, that NP seems to function as a binding
domain:
) Heidi; believes any description of herself;.
i) *Heidi; believes Martha's description ofherself;.
ii) Heidi; believes Martin's description ofherselfj.
The literature on this is extensive and beyond the scope of this chapter.
But you should be aware that the definition given here needs extensive
revision, we will return to this in chapter 15.
4. The Distribution of Pronouns
Anaphors are not the only NP type with restrictions on their syntactic
position. Pronouns can also be restricted in where they may appear:
23) a) Heidi;bopped hei-j on the head with the zucchini,
b) *Heidi; bopped her; on the head with the zucchini.
Pronouns like her in the sentences in (23) may not be bound. (They may
not be coindexed by a c-commanding NP.) The sentence in (23) may only
Chapter 5: Binding Theory 143
have the meaning where the her refers to someone other than Heidi. Contrast
this situation with the one in which the pronoun is in an embedded clause:
24) a) Heidi; said [s- that she; discoed with Art],
b) Heidi; said [s- that shek discoed with Art].
In this situation, a pronoun may be bound by an antecedent, but it doesn't
have to be. It can be bound as in (24a), or not bound as in (24b). Unlike
the case of anaphors, (which must be bound in a particular configuration),
pronouns seem only to have a limitation on where they cannot be bound.
That is, a pronoun cannot be bound by an antecedent that is a clause-mate
(in the same immediate clause). You'll notice that this is exactly the opposite
of where anaphors are allowed. This restriction is called Principle B of the
binding theory. It makes use of the term free. Free is the opposite of bound.
25) Free: Not bound.
26) Principle B: A pronoun must be free in its binding domain.
Given that the binding domain is a clause, the ungrammaticality of (23b)
is explained. Both Heidi and her are in the same clause, so they may not
be bound to each other. The pronoun must be free. In (24) both indexings
are allowed by Principle B. In (24b) the pronoun isn't bound at all (so is free
within its binding domain). In (24a), the situation is a little trickier:
The pronoun is bound, but it isn't bound within its binding domain
(the embedded clause). Its binder lies outside the binding domain,
so the sentence is grammatical.
You now have enough information totry Challenge Problem Sets 2 &3
5. The Distribution of R-expressions
R-expressions have yet another distribution. R-expressions don't seem
to allow any instances of binding at all, not within the binding domain
and not outside it either.
27) a) *Heidi; kissed Miriam;.
b) *Art; kissed Geoff;.
c) *She; kissed Heidi;.7
Note that this sentence is not a violation of Principle B.Heidi does not bind she here,
even though they are coindexed. This is because Heididoes not c-command she.Note
that [Even hert enemies] love Heidi, is well-formed, because neither NP c-commands
the other, so there is no binding, even if they are coindexed. Remember coindexation
is not the same thing as binding.
144 Preliminaries
d) *She; said that Heidi; was a disco queen.
In none of these sentences can the second NP (all R-expressions) be bound
by a c-commanding word. This in and of itself isn't terribly surprising, given
the fact that R-expressions receive their meaning from outside the sentence
(i.e., from the context). That they don't get their meaning from another word
in the sentence (via binding) is entirely expected. We do have to rule
out situations like (27). The constraint that describes the distribution
of R-expressions is called Principle C.
28) Principle C: An R-expression must be free.
Notice that Principle C says nothing about a binding domain. Essentially
R-expressions must be free everywhere. They cannot be bound at all.
You nowhaveenough information to try General Problem Set4 and Challenge
Problem Sets 4-6
A Common Mistake
Consider the sentence *S//c', loves Maryt. Which of the two NPs in this
sentence is the antecedent? Common sense might tell us that Mary is. But
common sense is wrong. The antecedent here is she. This is because she
c-commands Mary,and not vice versa.
One easy way to avoid this mistake is not to think in terms of
antecedent and anaphor/pronoun, but in terms of binder and bindee. The
binder here is she because it is coindexed with Mary and c-commands
Mary. Mary is the thing being bound (the bindee). Note that binding is
typically an asymmetric relationship.
6. CONCLUSION
In this chapter, we looked at a very complex set of data concerning
the distribution of different kinds of NPs. We saw that these different kinds
of NPs can appear in different syntactic positions. A simple set of Binding
Principles (A, B, and C) governs the distribution of NPs. This set of binding
principles is built upon the structural relations developed in the last chapter.
In the next chapter, we are going to look at how we can develop a
similarly simple set of revisions to the phrase structure rules. The constraints
developed in this chapter have the shape of locality constraints (in that they
require local, or nearness, relations between certain syntactic objects). In later
chapters, we'll see a trend towards using locality constraints in other parts
of the grammar.
Chapter 5: Binding Theory 145
The constraints developed in this chapter account for a wide range
of data, but there are many cases that don't work; In particular there is
a problem with our definition of binding domain. You can see some of these
problems by trying some of the challenge problem sets at the end of this
chapter. We return to a more sophisticated version of the binding theory
in chapter 15 in the last part of this book.
Ideas, Rules, and Constraints Introduced in this Chapter
i) R-expression: An NP that gets it meaning by referring to an entity in
the world.
ii) Anaphor: An NP that obligatorily gets its meaning from another NP
in the sentence.
iii) Pronoun: An NP that may (but need not) get its meaning from
another NP in the sentence.
iv) Antecedent: The element that binds a pronoun, anaphor or Rexpression.
When this element c-commands another coindexed NP,
it is a binder of that NP.
v) Index: A subscript mark that indicates what an NP refers to.
vi) Coindexed: Two NPs that have the same index („ vk, etc.) are said to
be coindexed.
vii) Corcfer: Two NPs that are coindexed are said to corefer (refer to
the same entity in the world).
viii) Binding: A binds B if and only if A c-commands B and A and B
are coindexed. A is the binder, B is the bindee.
ix) Locality Constraint: A constraint on the grammar, such that two
syntactic entities must be "local" or near to one another.
x) BindingDomain: The clause (for our purposes).
xi) Free: Not bound.
xii) The Binding Principles
Principle A: An anaphor must be bound in its binding domain.
Principle B: A pronoun must be free in its binding domain.
Principle C: An R-expression must be free.
146 Preliminaries
Further Reading
Aoun, Joseph (1985) A Grammar ofAnaphora. Cambridge: MIT Press.
Chomsky, Noam (1980) On Binding. Linguistic Inquiry 11,1-46.
Chomsky, Noam (1981) Lectures on Government and Binding. Dordrecht: Foris.
Higginbotham, James (1980) Pronouns and bound variables. Linguistic
Inquiry 11, 697-708.
Lasnik, Howard (1989) Essays on Anaphora. Dordrecht: Kluwer Academic
Publishers.
Reinhart, Tanya (1976) The Syntactic Domain of Anaphora. Ph.D.
dissertation, MIT.
General Problem Sets
1. NP TYPES
[Application of Skills; Very Basic]
Identify the type of NP (Anaphor, Pronoun, R-expression) of each of
the following:
their, each cat, folk dancing, oneself, each other, she, her, themselves
2. C-COMMAND AND BINDING
[Application of Skills; Basic]
Draw the trees for each of the following sentences and for the bolded NPs
indicate whether (i) there is a binding relationship between the two nouns,
and (ii) if there is relationship, which noun is the binder and which is the
element that is being bound; if there is no binding relationship explain why
(i.e., state which part of the definition of "binding" is not met). Note, this is
not a question about the binding conditions (A, B, C) but about the definition
of binding itself.
a) [The book about [the president]! ]k didn't bother hirri[.
b) [The book about [the president]! Ikdidn't bother himj.
c) [The book about [the president]) ]k sold itselfk.
d) [Andyi's constant lack of effort]k dismayed [hiSj father]m.
e) [Andy{'s constant lack of effort]k dismayed [hisn father]m.
Chapter 5: Binding Theory 147
3. Binding Domain
[Application of Skills; Basic]
Draw the tree for each of the following sentences. In your tree circle
the binding domain for the boldfaced noun:
a) The students told themselves that the exam wouldn't be too hard.
b) The students told their professor that they weren't worried about binding
theory.
c) Michael said the binding judgments were wrong.
4. Binding Principles
[Application of Skills, Data Analysis; Intermediate]
Explain why the following sentences are ungrammatical. For each sentence,
say what the binding domain of the NP causing the problem is, if it is
c-commanded by its binder (antecedent), and name the binding condition
that is violated.
a) *Michaelj loves himj.
b) *Hei loves Michaelj.
c) *Michaelj's father, loves himselfj.
d) *Michaelj's father loves himj.
e) *Susanj thinks that John should marry herselfj.
f) *John thinks that Susanj should kiss hen.
Challenge Problem Sets
Challenge Problem Set 1: Wh-questions
[Critical Thinking; Challenge]
What problem(s) does the following sentence make for the binding theory
as we have sketched it in this chapter? Can you think of a solution?
(Hint: consider the non-question form of this sentence John despises
these pictures of himself.)
Which pictures of himselfj does Johni despise?
Assume the following tree for this sentence:
CP
NP Comp TP
^^^~~^~~~~-~~~~~^ does
D N PP NPj VP
Which pictures / \ I I
P NPj N V
of | John despise
N
himself
148 Preliminaries
Challenge Problem Set 2: Binding Domain
[Critical Thinking; Challenge]
The following sentence with the assigned indexing is predicted by the theory
we have given so far to be ungrammatical. But it is actually ok. Explain
why our theory says this should be ungrammatical.
Andyj dismayed [hiSj father]m.
Challenge Problem Set 3: Persian8
[Critical Thinking; Challenge]
Does the binding theory account for the following data? Explain. (Ri means
"the" when following object NPs. 3sg means "third person singular.")
a) Janj goft [s- ke [s Meryk ketab-a ro be xodeshj/k bargardune]].
John said that Mary book-PL ra to himself/herself return
"John said that Mary (should) return the books to him/herself."
b) Janj goft [s- ke [s Meryj ketab-a ro be xodeshj/j barmigardune]].
John said that Mary book-PL ra to himself/herself return3SG.FUT
"John said that Mary will return the books to him/herself."
Now consider (c) and (d): in these examples, xod "self instead of xodesh
"himself is used. How do you explain the contrast between (a and b) and
(c and d)? Note that (a and b) are taken from the spoken language, whereas
(c and d) represent the formal written variant.
c) Janj goft [ke [s Meryk ketab r§ baraye xod^ bexanad]].
John said that Mary book ra for self read3SG
"John said that Mary (should) read the book to *himself/herself."
d) Janj goft [ke [s Meryk ketab ra baraye xod*j/k negahdaTad]].
John said that Mary book ra for self keep3sG
"John said that Mary (should) keep the books for *himself/herself."
Challenge Problem Set 4: Japanese
[Data Analysis and Critical Thinking; Challenge]
Japanese has a number of items that can be called pronouns or anaphors.
One of these is zibunzisin. For the purposes of this assignment assume that
any noun that has the suffix -wa c-commands any other NP, and assume
that any noun that has the suffix -ga c-commands any NP with the suffix -o.
Consider the following data (Data from Aikawa 1994):
a) Johnwaj [Cp [tpMarygakzibunzisinot^j hihansita] [c to]] itta.
John Mary zibunzisin criticized that said
"John said that Maryk criticized herselfk."
"*Johnj said that Mary criticized himselfj."
!Thisproblemset was contributed bySimin Karimi.
Chapter 5: Binding Theory 149
Question 1: On the basis of only the data in (a) is zibunzisin an anaphor or
a pronoun? How can you tell?
Now consider this sentence:
b) Johnwaj [cp [tp zibunzisingaj Maryo korosita] [c to]] omotteiru.
John zibunzisin Mary killed that think
"John thinks that himself killed Mary."
(note: grammatical in Japanese.)
Question 2: Given this additional evidence, do you need to revise
your hypothesis from question 1? Is zibunzisin an anaphor, a pronoun
or something else entirely? How can you tell?
One more piece of data:
c) Mohnwaj [Cp[tp zibunzisingak Maryok korosita] [c to]] omotteiru.
John zibunzisin Mary killed that think
"*John thinks that herselfk killed Maryk."
Question 3: Sentence (c) is a violation of which binding principle? (A, B,
or C?) Which NP is binding which other NP in this sentence to cause
the ungrammaticality?
Challenge Problem Set5: Counterexamples?9
[Critical Thinking and Data Analysis; Challenge]
Each of the following examples is problematic for the binding theory we
formulated above. Briefly explain why. For data from languages other than
English, your answer should be based on the facts of the target language,
and not the English translations. Use the word-by-word glosses to determine
whether the Dogrib and Modern Greek NPs should be analyzed
as anaphors, pronouns or R-expressions. Your discussion of Dogrib
should be based on consideration of both sentences taken together.
a) I have no money on me.
b) John knew that there would be a picture of himself hanging in the post
office.
c) Modern Greek
O YaniSj ipe stin Katerina oti i Maria aghapa ton idhiOj.
John said to Catherin that Mary loves himself
"Johnj told Catherine that Mary loves hinrvk."
9Thisproblem setwas contributed byBetsy Ritter. TheDogrib data come from Saxon
(1984).
150 Preliminaries
d) Dogrib
(i) John ye-hk'e ha
John 3sG(=him)-shoot future
"Johnj is going to shoot hinvj."
(ii) *ye-zha sheeti
3sG(=his)-son ate
"His son ate."
Challenge Problem Set 6: C-command or Precedence?
[Critical Thinking and Data Analysis; Challenge]
In the text above, we proposed that binding required both c-command
and coindexation. Consider an alternative: binding requires that the binder
precedes (rather than c-commands) and is coindexed with the element
is bound. Which of these alternatives is right? How can you tell?
You might consider data such as the following:
a) [cp [cp Although hej loves marshmallows] [tp Artj is not a big fan of
Smores10]].
b) [tp [np HiSj yearbook picture] gives Torrij the creeps].
Be very careful about this data. In particular, do not assume
that an R-expression is automatically the binder, pronouns can be binders
for the purposes of binding theory.
10 For those who may not be familiar with the term, Smores are a typical American
camp-fire treat. They involve a marshmallow candy cooked over an open fire,
squished between two layers of graham cracker along with a layer of dark chocolate
candy.
(\J
03
Q.
The Base
X-bar Theory
0. Introduction
As we saw in the last chapter, the theory of sentence structure that we've
developed is quite powerful. It correctly predicts constituency and alongwith
structural relations and the binding theory - it also accounts for
the structural restrictions on the interpretation of pronouns, anaphors and
R-expressions. This said, if we look a little more closely at sentence structure
in manylanguages, we see thatour theory has some empirical inadequacies.
(It can't account for all the data.) Consider, for example, the subject NP
in the sentence in (1):
1) [The bigbookofpoems withthebluecover] ison the table.
Thestructure our NP rule NP -> (D) (AdjP+) N (PP+) assignsto this is:
2) NP
D AdjP N
the A
big
book
of poems with the blue cover
Wecan call this aflatstructure. The PP ofpoems and the PP with the blue cover
are on the same level hierarchically; there is no distinction between them
in terms of dominance or c-command. In other words they are "flat" with
respect to the head word book. Fromthe point ofview ofconstituency, we see
154 The Base
that a number of tests point towards a more complicatedstructure. Consider
first the constituency test of replacement. There is a particular variety of
this process, called one-replacement, that seems to targetprecisely a group
of nodes that don't form a constituent in the tree in (2):
3) I bought the big [book of poems with the blue cover] not the small [one].
Here, one-replacement targets book ofpoems with the blue cover, this group
of words does not form a constituent in the tree in (2). Furthermore,
one-replacement seems to be able to target other subgroups of words
that similarly don't form constituents in (2):
4) I bought the big [book of poems] with the blue cover not the small [one]
with the red cover.
These facts seem to point to a moredeeplyembedded structure for the NP:
5) NP
D
J^<the
AP
J^
N'2
big
N
book
PP with the blue cover
of poems
The one-replacement in(4) targets the node labeled N'3. The one-replacement
in(3) targets the node labeled N'2. We have to change the NP slightly to get
evidence for N\. If we change the determiner the to the determiner that,
we can use one-replacement to target N\.
6) I want [NP this [N. big book of poems with the red cover]] not
[NPthat[N. one]].
Similar evidence comes from conjunction:
7) Calvin is[the [dean ofhumanities] and[director ofsocial sciences]].
8) Give me [the [blue book] and [red binder]].
We need these "intermediate" N'(pronounced "en-bar") categories to
explain the items that are conjoinedin these sentences.
Chapter 6: X-barTheory 155
The flat structure seen in (2) is clearly inadequate and a more articulated
structure is needed. This chapter is about these articulated trees. The theory
that accounts for these is called X-bar theory.
Before getting into the content of this chapter, a few bibliographic notes
are in order. The first presentation of X-bar theory appeared in Chomsky
(1970). Jackendoff's (1977) seminal book X-bar Syntax is the source of many
of the ideas surrounding X-bar theory. Perhaps the most complete
description of X-bar theory comes from an introductory syntax textbook (like
this one). This is Radford's (1988) Transformational Grammar: A First Course.
That textbook presents one of the most comprehensive arguments for X-bar
theory. This chapter draws heavily on all three of these sources. If you are
interested in reading a more comprehensive (although slightly out-of-date)
version of X-bar theory, then you should look at Radford's book.
1. Bar-level Projections
In order to account for the data seen above in the introduction, let us revise
our NP rules to add the intermediate structure:
9) NP->(D)N"
10) N" -» (AP) N" or N' (PP)
11) N' -> N (PP)
Equivalent Notations
The name "X-bar theory" comes from the original mechanism for
indicating intermediate categories. N' was written N with a bar over the
letter. This overbar is the origin of the "bar" in the name of the theory.
"X" is a variable that stands for any category (N, Adj, V, P, etc.). The
following notations are all equivalent,^-- ournotations
Phrase level NP = N"= N"=
Intermediate level N' = N'=N
Word/Head level N = N°
= Nr
The same is true of all other categories as well (e.g., PP =*P = P"= P"=
P"13*, etc.). Since overbars are hard to type, even with Unicode fonts, most
people use a prime (') or apostrophe (') for the intermediate level and
write the phrasal level as NP (or more rarely, N").
These rules introduce a new character to our cast of nodes, seen briefly
above. This is the N' node. It plays the role of the intermediate constituent
156 The Base
replaced by one above. The tree in (5) is repeated here showing how
these rules (9-11) apply.
(12) NP
D N\
the
AP
J^
N'2
big
N
book
PP
of poems
with the blue cover
Rule (9) generates the NP node of this tree, with its daughters D and N'.
The first version of rule (10) generates N\. The second version of rule (10)
generates N'2. Finally the last rule (11)spells out N'3 as N and its PP sister.
We can now straightforwardly account for the one-replacement
sentences. One-replacement is a process that targets the N' node:
13) One-replacement: Replacean N' node with one.
Without the intermediate N' node, we would have no way of accounting
for one-replacement or conjunction facts. With N', explaining these sentences
is easy, since there is more structure in each phrase.
Therulesystem in (9-11) hasa number ofstriking properties (including
the facts that it is binary branching and the first N' rule is iterative or
self-recursive). We will return to these properties in a later section and show
how they account for a number of surprising facts about the internal
structure of phrases. First, however, let's see if any other categories
also have intermediate structure.
1.1 V-bar
There is a similar process to one-replacement found in the syntax of VPs.
This is the process of do-so-1 (or did-so-)replacement. Consider first the VP
in the following sentence, which has both an NP and a PP in it.
14) I [eat beans with a fork].
Depending upon which dialect of English you speak, you may prefer "did too" over
'did so too" or "did so." If the VPs below sound odd, try substituting "did too" for
'did so."
Chapter 6: X-barTheory 157
The rule we developed for VPsin chapter 3 generates the following flat tree:
15) VP
V NP PP
eat beans with a fork
In this tree, there is no constituent that groups together the V and NP and
excludes the PP. However, do-so-replacement targets exactly this unit:
16) I [eat beans] with a fork but Janet [does (so)] with a spoon.
Let's formalize this rule as:
17) Do-so-replacement: Replace a V with doso (or door doso too or dotoo).
For this to work we need the following rules2:
18) VP -> V
19) V -> V (PP) or V (AdvP)
20) V -> V (NP)
The tree structure for the VP in (14) will look like (21).
21) VP
V*
V PP
V NP
Rule (18) generates the VP and the V under it; the next rule (19) expands
the top V into another V and a PP. Finally, the lower V is expanded into V
and NP by rule (20).
The rule of rfo-so-replacement seen in (17) targets the lower V
and replaces it with doso. Evidence for the higher V comes from sentences
like (22):
2 The rule in (18) may appear a little mysterious right now (since it appears
to introduce a vacuous structure) but we will have need of it in a later chapter. For
the moment, just assume that it is necessary, and we will provide additional
justification for it later. You can note for now that in order to account for sentences
like (22)below, we will need to assume that the entire replaced structure is a V, if we
assume that rfo-so-replacement only targets V nodes (and not VPnodes).
158 The Base
22) Kevin [ate spaghetti with a spoon] and Geordi [did so] too.
In this sentence, did soreplaces the higher V (which includes the V, the lower
V, the NP, and the PP).
Similarly, conjunction seems to show an intermediate V projection:
23) The chef [eats beans] and [tosses salads] with forks.
The tree for a structure like this requires a V node (a description
of the conjunction rule can be found below in the additional rules
in the Ideas section at the end of the chapter):
24)
V Conj V with forks
/\ and f\
V NP V NP
eats /\ tosses /\
beans salads
You now have enough information to try Challenge Problem Set 1
1.2 Adj-bar andAdv-bar
The arguments for intermediate structure in AdjPs are a little more tricky,
as Englishseems to limit the amount of material that can appear in an AdjP.
However, we do see such structure in phrases like (25):
25) the [very [[brightblue] and [dull green]]] gown
In this NP, bright clearly modifies blue, and dull clearly modifies green. One
possible interpretation of this phrase (although not the only one) allows very
to modify both bright blue and dull green. If this is the case then the structure
must minimally look like (26) (note: we will have reason to revise
this tree later).
Chapter 6: X-barTheory 159
26) AdjP
This must be the structure so that the AdvP can modify both bright blue and
dullgreen.
Under certain circumstances, some adjectives appear to allow
prepositional modifiers to follow them:
27) I am afraid/frightened of tigers.
28) I am fond of circus performers.
Thesepost-adjectival PPs parallel the direct object of related verbs:
29) I fear tigers.
30) I like circus performers.
Consider now:
31) I am [[afraid/frightened of tigers] and [fond of clowns] without
exception].
Under one reading of this sentence, without exception modifies both afraid
of tigers and fond of circus performers. Again this would seem to suggest
that the sentence has the constituency represented by the above bracketing,
which points towards an intermediate category of Adj'.
There is also a replacement phenomenon that seems to target Adj's.
This is so-replacement:
32) Bob is [very[serious about Mary]], but [less [so]] than Paul.
The adjective phrase here is very serious about Mary, but so-replacement only
targets serious about Mary.
The rules that generate these structures are:
33) AdjP -> Adj'
34) Adj' -> (AdvP) Adj'
35) Adj' -> Adj (PP)
160 The Base
For reasons of parsimony, we might presume that a similar set of rules
governs Adverbs as well, although the evidence is very scarce.
2.3 P-bar
Consider the following sentences:
36) Gwen placed it [right [in the middle of the spaghetti sauce]].
37) Maurice was [[in love] with his boss].
38) Susanna was [utterly [in love]].
In these examples, we have what appear to be prepositional phrases (in the
middle ofthe spaghetti sauce, in love) that are modified by some other element:
right, with his boss, and utterly respectively. Note, however, that you can
target smaller units within these large PPs with constituency tests:
39) Gwen knocked it [right [off the table] and [into the trash]].
40) Maurice was [[in love] and [at odds] with his boss].
41) Susanna was [utterly [in love]], but Louiswas only [partly [so]].
Examples (39) and (40) show conjunction of the two smaller constituents.
Example (41) is an example of so-replacement. Let us call the smaller
constituent here P' on a parallel with N', Adj', and V. Therules that generate
PPs are given below:
42) PP -> P'
43) P' -> P' (PP)
44) P' -> P (NP)
With this, we complete our tour of intermediate structure. In developing
our phrase structure system, we've managed to complicate it significantly.
In the nextsection we look at ways to simplify the rule system yet capture
all the constituency facts we've considered here.
2. Generalizing the Rules: The X-bar Schema
For each of the major phrase types (NPs, VPs, AdjPs, AdvPs and PPs)
we have come up with three rules, where the first and second rules serve to
introduce intermediate structure. Lefs repeat all the rules here. (Rules (48),
(51), (54), and (57) are admittedly here simply by stipulation, we've seen
no evidence for them. We're positing them now for reasons of parsimony
with rule (45), but we'll see in the next chapter and the chapter that follows
that the structures these rules introduce will be useful to us. Please allow me
Chapter 6: X-barTheory 161
this one mysterious stipulation for the moment, I promise we'll return
to the issue later in the book.)
45) NP -> (D) N'
46) N" -> (AdjP) N' or N' (PP)
47) N' -> N (PP)
48) VP->V
49) V' -> V (PP) or V (AdvP)
50) V -> V (NP)
51) AdvP -> Adv'
52) Adv' -> (AdvP) Adv'
53) Adv' -> Adv' (PP)
54) AdjP -»Adj'
55) Adj' -> (AdvP) Adj'
56) Adj' -> Adj (PP)
57) PP -> P'
58) P' -> P" (PP)
59) P' -> P (NP)
Thisis quite a complicated set, but seems to be more empirically motivated
than the set of rules we set out in chapter 3. We can now ask, are we missing
any generalizations here?
Indeed, we seem to be missing several. First, note that in all the rules
above, the category of the rule is the same as the onlyelement that is not
optional. For example, in the NP rule, the element that isn't optional is N'.
This is the same part ofspeech. Similarly, the onlyobligatory element in N'
is either another N' or N. This is a very general notion in phrase structure;
we call this headedness. All phrases appear to have heads. Heads are the
most prominent element ina phrasal category and give their part of speech
category tothe whole phrase. Note that we don'thave any rules oftheform:
60) NP -> V AdjP
This rule not only seems meaningless, it is unattested in the system we've
developed here. This property iscalled endocentricity, meaning thatevery
phrase has ahead. The only obligatory element ina phrase isthe head.
Second, note that with the exception of the determiner in the NP rule,
all non-head material in the rules is both phrasal and optional.
We never find rules of the form:
61) V -> Adv V
162 The Base
With the exception of the determiner (an exception that we'll resolve
in chapter 7), anything in an X-bar rule that isn't a head must be a phrase
and optional.
Finally, notice that for each major category, there are three rules, one
that introduces the NP, VP, AdvP, AdjP and PP, one that takes a bar level
and repeats it (e.g., NJ. -» N (PP)), and one that takes a bar level and spells
out the head (e.g., N' -» N (PP)). We seem to be missing the generalization
that for each kind of phrase, the same kinds of rules appear. X-bar theory
is an attempt to capture these similarities between rules.
We can condense the rules we've proposed into a simple set. To do this
we are going to make use of variables (like variables in algebra) to stand for
particular parts of speech. Let X be a variable that can stand for any category
N, V, Adj, Adv, P. An XP is a catch-all term to cover NP, VP, AP, PP,
similarly X' stands for N', V, Adj', Adv', P', and X represents N, V, Adj,
Adv, and P.
Using this variable notation we can capture the generalizations that we
have missed. Let's start with the rules that introduce heads:
62) a) N' -> N (PP)
b) V -> V (NP)
c) Adj' -> Adj (PP)
d) Adv' -> Adv (PP)
e) P' -> P (NP)
By using the variable notation we can generalize across these rules with
the single general statement:
63) X' -» X (WP) (to be revised)
BothXand W here are variables for categories. This rule says that some bar
level (on the left of the arrow) consists ofsomehead followed by an optional,
phrasal3 element.
Now turn to the recursive N', A', V and P' rules:
64) a) N' -> (AdjP) N' or N' (PP)
b) V -> V (PP) or V (AdvP)
c) Adj' -> (AdvP) Adj'
d) Adv' -> (AdvP) Adv'
e) P' -> P' (PP)
3TheD in the NP rule is,of course, not phrasal. This is a problem we will returnto
in the next chapter.
Chapter 6: X-bar Theory 163
For each of these rules, a category with a single bar level is iterated
(repeated), with some optional material either on the right or the left.
Again using X as a variable, we can condense these into a single statement:
65) X' -» (ZP) X' or X" (ZP) (to be revised)
Again the Xshere must be consistent in part of speech category. The material
that is not the head (i.e., not X) must be phrasal and optional. Note
that the categories of these non-head items are also indicated with variables
(in this case: ZP).
Finally, let's consider the rules that introduce the last layer of structure:
66) a) NP -> (D) N'
b) VP->V
c) AdjP -> Adj'
d) AdvP -> Adv'
e) PP -> P'
These rules can also be collapsed into a single rule:
67) XP -> (YP) X" (to be revised)
We haven't motivated the existence of the YP here, except in the form
of determiners. I know you're naturally suspicious of me saying "trust me on
this" (and rightly so); but I promise we will resolve this in the next chapter.
The system we've come up with here is simple. We've reduced our
phrase structure rules down to three general rules using variables: Because
they use variables, these rules can generate the correct constituent
of the sentences of English. This analysis isn't without problems, however.
Before we turn to resolving these problems and drafting a final version
of the X-bar rules, we need to introduce some new terminology.
3. COMPLEMENTS, ADJUNCTS, AND SPECIFIERS
Consider now the two prepositional phrases that are subconstituents of
the following NP:
68) the book [PP of poems] [PP with the glossy cover]
Using the X-barrules,4 we can generate the following tree for this NP:
4Specific instructions on drawing trees using the X-bar rules are found at the end
of this chapter.
164
69)
The Base
poems
sister to a bar level
sister to a head
the glossy cover
(I've used triangles in this tree to obscure some of the irrelevant details,
but you should not do this when you are drawing trees, until you have
a confidentgrasp ofhow treenotationworks.) You'll note that thetwo PPs in
this tree are at different levels in the tree. The lower PPi is a sister to the head
N (book), whereas the higher PP2 is a sister to the N' dominating the head N
and PPi. You'll also notice that these two PPs were introduced by different
rules. PPX is introduced by the rule:
70) X' -> X (WP)
and PP2 is introduced by the higher level rule:
71) X' ^ X' (ZP)
An XP that is a sister to a head (N, V, A, or P) is called a complement. PPi
is a complement. Complements roughly correspond to the notion "object" in
traditional grammar. XPs that are sisters to single bar levels (N', V, A', or P')
and are daughters of an N' are called adjuncts. PP2 is an adjunct. Adjuncts
often have the feel of adverbial or obliques.
72) Adjunct: An XP that is a sister to a single bar level (N', V, A', or P") and
a daughter of a single bar level (N', V, A', or P').
73) Complement: An XP that is a sister to a head (N, V, A, P), and a daughter
of a single bar level (N', V, A', or P').
The rules that introduce these two kinds of XPs get special names:
74) Adjunct rule: X' -> X'(ZP)
75) Complement rule: X' -> X(WP)
Chapter 6: X-bar Theory 165
A tree showing the structural differencebetween these is given below:
76) XP
adjunct
complement
If there really are two different kinds of PP within an NP, then we expect
that they will exhibit different kinds ofbehavior. It turns out that this is true:
There are significant differences in behavior between adjuncts and
complements.
3.1 Complements andAdjunctsin NPs
Take NPs as a prototypical example. Consider the difference in meaning
between the two NPs below:
77) the book of poems
78) the book with a red cover
Although both these examples seem to have, on the surface, parallel
structures (a determiner, followed by a noun, followed by a prepositional
phrase), in reality, they have quite different structures. The PP in (77) is
a complement and has the following tree:
79)
of poems
You'll note that the circled PP is a sister to N, so it is a complement.
Bycontrast, the structure of (78)is:
166 The Base
80)
D N'
the
N'< (PP)
N with a red cover
book
Here the PP with a red cover is a sister to N', so it is an adjunct.
The differences between these two NPs is not one that you can hear. The
difference between the two is in terms of the amount of structure in the tree.
In (80), there is an extra N'. While this difference may at first seem abstract, it
has important implications for the behavior of the two PPs. Consider first the
meaning of our two NPs. In (77), the PP seems to complete (or complement)
the meaning of the noun. It tells us what kind of book is being referred to.
In (78), by contrast, the PP seems more optional and more loosely related to
the NP. This is a highly subjective piece of evidence, but it corresponds to
more syntactic and structural evidence too.
An easy heuristic (guiding principle) for distinguishing complements
from adjunct PPs inside NPs, is by lookingat what preposition they take.
In English, almostalways (although thereare someexceptions) complement
PPs take the preposition of. Adjuncts, by contrast, take other prepositions
(such asfrom, at, to, with, under, on, etc.). This test isn't 100 percentreliable,
but will allow you to eyeball PPs and tell whether they are complements or
adjuncts for the vast majority of cases. With this in mind, let's look at some
of the otherbehavioral distinctions between complements and adjuncts.
Think carefully about the two rules that introduce complements and
adjuncts. There are several significantdifferences between them. These rules
are repeated here for your convenience:
81) Adjunct rule: X' -> X'(ZP)
82) Complement rule: X -> X(WP)
First observe that because the complement rule introduces the head (X),
the complement PP willalwaysbeadjacent tothehead. Or moreparticularly,
it will always be closer to the head than an adjunct PP will be. This is seen
in the following data:
83) the book [of poems] [with a red cover]
head complement adjunct
84) *thebook [with a red cover][of poems]
head adjunct complement
Chapter 6: X-barTheory
You can see how this is true if you look at the tree for sentence (83):
85) NP
167
created by theadjunct rule
book
PP jwi
of poems created bythe complement rule
Since the adjunct rule takes an X' level category and generates
another X' category, it will always be higher in the tree than the output
of the complement rule (which takes an X' and generates an X). Since lines
can't cross, this means that complements will always be lower in the tree
than adjuncts, and will always be closer to the head than adjuncts.
There is another property of the rules that manifests itself
in the difference between adjuncts and complements. The adjunct rule, as
passingly observed above, is an iterative rule. That is, within the rule itself,
it shows the property of recursion(discussed in chapter 3): On the left-hand
side of the rule there is an X' category, and on the right hand side there is
another X'. This means that the rule can generate infinite strings of X' nodes,
since you can apply the rule over and over again to its own output:
86)
The complement rule doesnot havethisproperty. On theleftside oftherule
there is an X', but on the right there is only X. So the rule cannot apply
iteratively. That is, it can only apply once within an XP. What this means
for complements and adjuncts is that you can have any number of adjuncts
(87), but you can only ever have one complement (88):
168 The Base
87) the book [of poems] [with a red cover] [from Blackwell] [by Robert Burns]
head complement adjunct adjunct adjunct
88) *the book [of poems] [of fiction] [with a red cover]
head complement complement adjunct
The tree for (87) is given below; you'll note that since there is only one N,
there can only be one complement, but since there are multiple N's,
there can be as many adjuncts as desired.
89)
PP by Robert Burns
of poems
Related to the facts that the number of adjuncts is unlimited, but only
one complement is allowed, and complements are always adjacent
to the head, observe that you can usually reorder adjuncts with respect
to oneanother, but you cannever reorder a complement with theadjuncts:
90) a)
b)
c)
d)
e)
f)
g)
h)
i)
thebookofpoemswitha redcover from Blackwell by Robert Burns
thebookofpoemsfrom Blackwell with a red coverby Robert Burns
the book of poems from Blackwell by RobertBurnswith a red cover
the book of poems by Robert Burns from Blackwellwith a red cover
the book of poems by RobertBurns with a red cover from Blackwell
the book of poems with a red cover by Robert Burns from Blackwell
*thebook with a red cover of poems from Blackwellby Robert Burns
*thebook with a red coverfromBlackwell of poemsby RobertBurns
*the bookwitha red cover from Blackwell by Robert Burns ofpoems
(etc.)
Note that adjuncts and complements are constituents of different types.
The definition of adjuncthood holds that adjuncts are sisters to X'. Since
conjunction (see under additional rules at the end of this chapter) requires
that you conjoin elements of the samebar level, you could not, for example,
Chapter 6: X-bar Theory 169
conjoin an adjunct with a complement. This would result in a contradiction:
Something can't be both a sister to X' and X at the same time. Adjuncts
can conjoin with other adjuncts (other sisters to X'), and complements
can conjoin with other complements (other sisters to X), but complements
cannot conjoin with adjuncts:
91) a) thebookofpoemswitha red cover and witha bluespine5
b) the book of poems and of fiction from Blackwell
c) *the book of poems and from Blackwell
There is one final difference between adjuncts and complements that we
will examine here. Recall the test of one-replacement:
92) One-replacement: Replacean N' node with one.
Thisoperationreplaces anN' node with theword one. Look at the tree in (93):
93)
canbereplaced by one
cannot bereplaced byone
If you look closely at this tree you'll see that two possibilities for onereplacement
exist. We caneithertargetthehighest N',and get:
94) the one
or we can target the lower N' and get:
95) the one with a red cover
Butwe cannot target the N head; it is not an N'. This means that one followed
by a complement is ill-formed:
96) *the oneof poemswith a red cover6
5Ifthis NP sounds odd to you, try putting emphasis on the and.
6 Not everyone finds this NP ill-formed. There is at least one major US dialect
that allows sentence (96). One possible explanation for this is that different dialects
170 The Base
Since complements are sisters to X and not X', they cannot stand next to the
word one.Adjuncts, by definition, can.
So far in this chapter, we've covered a huge range of facts, so a quick
summary is probably in order. In section 1, we saw that constituency tests
pointed towards a more articulated structure for our trees than the one we
developed in chapter 3. In section 2, we introduced the X'notation to account
for this more complicated structure. In X-bar structure, there are three levels
of categories. There are XPs, X's, and Xs. In this section - focusing exclusively
on NPs - we introduced special terms for elements that are sisters to X'
and X: adjuncts and complements. These two different kinds of modifier
have different properties. Adjuncts but not complements can be iterated
and reordered and can stand next to one. Complements, by contrast,
must be located next to the head and can't be reordered. We also saw that we
could conjoin complements with complements and adjuncts with adjuncts,
but that we couldn't mix the two. All of these data provide support
for the extra structure proposed in X-bar theory. In the next subsection,
we'll briefly consider evidence that the complement/adjunct distinction
holds for categories other than NP as well.
You can nowtryGeneral Problem Sets 1,2 &3 and Challenge Problem Set2.
You may wanttoread section 6 (on tree drazving) before attempting
General Problem Set 3.
3.2 Complements andAdjuncts in VPs, AdjPs, AdvPs, andPPs
The distinction between complements and adjuncts is not limited to NPs;
we find itholds in all the major syntactic categories. Thebestexample is seen
in VPs. Thedirectobject ofa verb isa complement of the verb. Prepositional
and adverbial modifiers of verbs are adjuncts:
97) I loved [the policeman] [intensely] [with all my heart].
V direct object adverbial PP phrase
complement adjunct adjunct
have different one-replacement rules. The dialect that finds this NP well-formed
allows either N or N' to be replaced. The dialect that finds this ill-formed (or at least
odd), only allows N' to be replaced.
171
98)
AdvP with all my heart
the policeman
Direct objects must be adjacent to the verb, and there can only be one of
them.
99) a) *I loved intensely the policeman with all my heart.
b) *I loved the policeman the baker intensely with all my heart.
Did-so (did-too) replacement targets V. Like one-replacement, this means that
it can only apply before an adjunct and not before a complement:
100) Mika loved the policemen intensely and
a) Susan did so half-heartedly.
b) *Susan did so the baker.
This is classic adjunct/complement distinction. In general, complements of
all categories (N, V, A, P, etc.) are the semanticobjects of the head. Consider
for example all the complements below:
101) a) John fears dogs. (verb)
b) John is afraid of dogs, (adjective)
c) John has a fear of dogs, (noun)
In all these sentences, (of) dogs is a complement.
You now have enough information totry General Problem Set 4
The evidence for the adjunct/complement distinction in adjective
phrases and prepositional phrases isconsiderably weaker than that ofnouns
and verbs. Adverbs that modify adjectives have an adjunct flair - they can
be stacked and reordered. Other than this, however, the evidence for the
distinction in PPs and AdjPs, comes mainly as a parallel to the NPs and VPs.
This may be less than satisfying, but is balanced by the formal simplicity
of having the same system apply to all categories.
172 The Base
3.3 The Notion Specifier
In the section 3.1 above, we introduced two structural notions: adjuncts and
complements. These correspond to two of the three X-bar rules:
102) a) Adjunct rule: X'-> X'(ZP) or X'-> (ZP) X'
b) Complement rule: X'->X(WP)
The third rule also introduces a structural position: the specifier.
103) Specifier rule: XP -»(YP) X'
We have only seen one specifier so far - the determiner in NPs:
104) [the] [book] [of poems] [with a red cover]
specifier head complement adjunct
105) NP
specifier
adjunct
with a red cover
head of poems complement
The specifieris defined as the daughter of XPand sister to X':
106) Specifier: AnXP7 thatisa sister toan X' level, and a daughterofan XP.
We can show that specifiers are different from adjuncts and
complements. Since the specifier rule is not recursive, you can only have
one specifier:8
107) *the these red books
The specifier rule has to apply at the top of the structure, this means that
the specifier will always be the left-mostelement (in English anyway):
7 If you are being observant you'll notice that the single example we have of
a specifier is not a phrase, but a word (the), so it may seem odd to say XP here.
We return to this issue in later chapters.
8One possible exception to this is the quantifier all, as in all the books. In the next
chapter, we discuss the idea that determiners head their own phrase (called a DP),
which might provide a partial explanation for this exception.
Chapter 6: X-barTheory 173
108) *boring the book
The above example also shows that specifiers can't be reordered with respect
to other adjuncts or complements. As the final difference between specifiers
and other types of modifier, specifiers can only be conjoined with other
specifiers:
109) a) two or three books
b) *two or boring books
On the surface, the usefulness of this position may seem obscure, since
only determiners appear in it. But in later chapters we will have important
use for specifiers. (In particular, we will claim that they are the position
where subjects are generated in a variety of categories.)
4. SOME DEFINITIONAL HOUSEKEEPING
With the refinements to the grammar we've made by adding X-bar theory
to our system, we need to make some minor modifications to the rules
and definitions that we introduced in previous chapters.
First, we need some terminology to describe the new parts of the phrase
that we have added. We can refer to all the elements in an NP, for example,
that are introduced (i.e. appear on the left side of the rule) by the
threephrase structure rulesas projections ofthe head. In the following tree,
all the N's and the NP said to be projections of the head N.
110)
Projections ofN
of poems
The NP is called the maximal projection; the N's are called intermediate
projections.
Recall our definition of the principle of modification from chapter 3;
a modifier must be a sister to the head it modifies:
174 The Base
111) Principle ofModification (old)
If a YP (that is, a phrase with some category Y) modifies some head X,
then YP must be a sister to X.
This definition no longer works for us. If you look at the tree in (110) you'll
see that only the complement is actually a sister to the head. Modifiers
that are adjuncts and specifiers aren't. To fix this we need to again revise
the principle of modification:
111') Principle ofModification (revised)
If a YP modifies some head X, then YP must be a sister to X or
a projection of X (i.e., X', XP).
By adding extra layers of structure we also need to revise our definitions
of object and indirect object. In chapter 4, these relations are defined in terms
of being immediately dominated by VP. But with the new X-bar structures
these NPs aren't immediately dominated by VP anymore, so we need
to change them so they are defined in terms of intermediate structure.
This is easy to do for the direct object of transitive verbs:
112) Direct Object (partly revised):
With verbs of type V(NP_NPJ, and V|NP_CPF the NP or CP sister to the V.
The definitions of indirect and direct objects in other types (e.g.,
ditransitives) are muchharder. Thisisbecause our rules as stated onlyallow
binary branching; only one complement is allowed. Yet for ditransitives
we would ideally like to have two complements. This is a problem for X-bar
theory. We willreturn to ditransitives and indirectobjects in chapter 13.
5. PARAMETERS OF WORD ORDER
In thischapter, and thusfar inthisbook, we'vebeenconcentrating primarily
on English. The reason for this is that, sinceyou are reading this book, it is
the language most accessible toyou. However, syntacticiansaren't interested
only in English. One of the most interesting parts of syntax is comparing the
sentence structure of different languages. The X-bar rules we've developed
so far for English do an acceptable job of accounting for the order of
constituents and hierarchical structure of English:
113) a) Specifier rule: XP -> (YP)X'
b) Adjunct rule: X' -> X'(ZP) orX' -> (ZP) X'
c) Complement rule: X' -» X (WP)
Chapter 6:X-barTheory 175
They don't, however, account well for other languages. Consider
the position of direct objects (complements) in Turkish. In Turkish,
the complement precedes the head:
114) Hasan kitab-i oku-du.
Hasan-SUBJ book-OBj read-PAST
"Hasan read the book."
If you look carefully at sentence (114) you notice that the word kitabi "book"
precedes the word okudu "read."
Not all languages put the complement on the right-hand side
like English. Not all languages put the specifier before the head either. Our
rules, while adequate for English, don't really get at the syntactic structure
of languages in general. Remember, syntax is the study of the mental
representation of sentence structure, and since we all have the same basic
gray matter in our brains, it would be nice if our theory accounted for both
the similarities and the differences among languages.
X-bar theory provides us with an avenue for exploring the differences
and similarities among languages. Let's start by generalizing our rules a little
bit. Let's allow specifiers and adjuncts to appear on either side of the head:
115) a) Specifier rule: XP->(YP) X'orXP-> X' (YP)
b) Adjunct rule: X' -»X' (ZP) orX' -> (ZP) X'
c) Complement rule: X' -> X (WP) orX -> (WP) X
Each of these rules has two options, the specifier/complement/adjunct
can all appear on either side of their head. Obviously, these rules are now
too general to account for English. If these rules, as stated, were adopted
straight out, they would predict the grammaticality of sentences like:
116) *[NP Policeman the] [VP Mary kissed],
(meaning The policeman kissed Mary.)
It would be a bad thing to do this. At the same time, constituent orders
like those of Turkish, in fact, exist, so this clearly is an option. Our theory
must capture both facts: The fact that the object-verb(OV) order is an option
that languages use, and that it isn't the option used by English.
The way that generative syntacticians accomplish this is by claiming that
the rules in (115) are the possibilities universally available to human beings.
When you acquire a particular language you select one of the options in the
rule, based upon the input you hear from your parents. Take, for example,
the complement rule. In English, complements of verbs follow the verbal
head. In Turkish, they precede the head. There are two options in the rule:
176 The Base
117) a) X'->X(WP)
b) X1 ->(WP)X
The child learning English will adopt option (a), the child learning Turkish
will adopt option (b). These options are called parameters. The proposal
that word order is parameterized finds its origins in Travis (1984).
Here is an analogy that might help you understand this concept. Imagine
that in your head you have a box of switches, just like the box of master
breaker switches which controls the electricity in your house. These switches
can be set on or off. The options in the X-bar rules are like these switches,
they can be set in one direction or the other (and in some situations such
as adjuncts in English - allow both settings).
118) X-bar parameters switch box
Specifie
XP->(\
i o
jj 1
l e
;|
r
T)X
Adjunct
X -> (ZP
i
)X'
Complement
X -> (WP) X
IT
1! •
XP->X" (YP) X' -» X' (ZP) X' -> X (WP)
When you are a child acquiring your language, you subconsciously set
these switches, to tell you which version of the rules to use.
Notice that this gives us a very simple system for acquiring the word
order of our languages. There are a finite set of possibilities, represented
by the different settings of the parameters. English sets its complement
parameter so that the complement follows the head. Turkish sets it the other
way. The child only has to hear a small amount of data (perhaps even
as little as one sentence) to know what side of the head complements go in
their language. Once children have set theparameter, they can apply the right
version of the rule and generate an unlimited number of sentences. In the
problem sets at the end of this chapter, you have the opportunity of looking
at some data from a variety of languages and determining how their X-bar
parameters are set. For your reference, the English settings are given below:
119) a) Specifier specifier on left, head on right (XP -»(YP) X')
e.g., thebook
Chapter 6: X-barTheory 177
b) Adjunct both options allowed (X'-> (ZP) X'and X' -> X' (ZP))
e.g., yellow roses
booksfrom Poland
c) Complement head on left, complement on right (X' -» X (WP))
e.g., books ofpoems
John kissed his mother.
You nowhaveenough information to try General Problem Sets5 & 6 (although you
may want to read section 6 (on tree drawing) before attempting Problem Set4).
You alsohave enough information toanswer Challenge Problem sets 3 &4
6. Drawing Trees in X-bar Notation
6.1 Important Considerations in Tree Drawing
In this section, we'll run through the steps for drawing trees in X-bar
notation. The basics of tree drawing that you learned in chapter 3 hold
here too. However, some special principles apply to X-bar tree drawing:
i) When identifying what modifies what, it is also important to know
whether it is a complement, adjunct, or specifier. This is important
because you have to know whether to make it a sister to the head,
to an X', etc.
ii) When linking material up, start with the modifiersclosest to the head.
Because X-bar structure is formulated the way it is, material closest
to the head will be the most deeply embedded material - so it will
have to attach to the head first.
a) Identify the head.
b) Attach the complement (must be a phrase!).
c) Attach any adjuncts (which must be phrases themselves) working
progressively away from the head, each adjunct gets its own
X' mother. (See points (iv) and (v) below for dealing with cases
when you have either no adjuncts or have an adjunct on either side
of the head.)
d) When there are no more adjuncts, attach the specifier, if there is one.
This gets an XP on top.
e) Even if there is no specifier, put an XP on top of the projections.
This indicates that there are no more modifiers of the head X.
178 The Base
iii) Keep inmindthatnone oftheX-bar rulesareoptional. Thatis, theymust
all apply. This results in a fair amount of vacuous or non-branching
structure. Even if the phrase has only a single word in it you willhave
at least the following structures:
120) a) NP b) VP
N'
N
Peter
V
V
left
c) AdjP
I
Adj'
Adj
red
d) AdvP e) PP
Adv'
Adv
badly
P'
P
before
iv) Perhaps one of the most common errors of new syntacticians is in
drawing trees for phrases with an adjunct but no complement. Consider
the NP [notebook with a red cover]. With a red cover is an adjunct - that
means that it has to be a sister to N' and a daughter to N' (by definition).
This is seen in the following tree:
121) NP
N
notebook
The circled N' here must be present in order to make the PP an adjunct.
Bevery careful to draw in these vacuous nodes that distinguish adjuncts
from complements.
v) Another common issue that arises for new syntacticians is how to tree
a sentence when there is an adjunct on either side of the head. Consider
the sentence in (123):
122) Andy [VP frequently eats sushi with his boss].
adjunct head complement adjunct
We start by attaching the complement to the head:
Chapter 6: X-barTheory 179
123) V
V NP
eats I
N'
I
N
sushi
Our next step should be to attach an adjunct. But there are two adjuncts.
Which one comes first? Interestingly, the answer is either. Two possible
trees can come out of this VP:
124) a) VP
V
AdvP
frequently V
eats
NP
sushi
b)
V V*
PP AdvP
VP
V
PP
V with his boss
NP
sushi
with his boss frequently V
eats
In (124a), the AdvPfrequently is attached higher than the PP with his boss.
In (124b), the PP is attached higher than the AdvP. Both of these trees
are acceptable, because the adjunct rule iterates. This means that either
version of it can appear in either order. Since we have two structures
for this sentence you might wonder if there is any semantic ambiguity
in this phrase. The distinction is subtle, but it is there. For (124a),
we can identify a set of events of sushi-eating with the boss, and then
we identify those events as occurring frequently. The meaning of (124b)
is very subtly different, there is a set of frequent events of sushi eating,
and we are identifying those as occurring with his boss. This distinction
is a little easier to see in NPs:
125) The red dress with the pink stripes
This can be treed two different ways:
180 The Base
126) a) NP
N' with pink stripes
dress
The first tree corresponds to a meaning where we are picking a red
member out of the set of dresses with pink stripes. (126b) corresponds
to the situation where we are picking a dress with pink stripes out of
the set of red dresses. These two trees may pick out the same individuals
in the world, but they do so in different contexts (one where we are
trying to distinguish among red dresses; and the other where we are
distinguishing among pink-striped dresses).
With these additional X-bar theoretic considerations in mind we can now
draw a sample tree:
6.2 A Sample Tree
The sentence we'll draw is:
127) Thej ugly man from Brazil found books of poems in the2puddle.
Our first step, as always, is to identify the parts of speech:
128) DAdjNP NV NPNPDN
The! ugly man from Brazil found books of poems in the2puddle.
Next, and most importantly, we have to identify what modifies or relates
to what, and whether that modification is as an adjunct, complement,
or specifier. This is perhaps the most difficult and tedious step, but it is also
the most important. You will get better at this with practice. You can use the
tests we developed above (stacking, coordination, etc.) to determine whether
the modifier is a complement, adjunct, or specifier.
129) [Thej modifies [man] as a specifier,
[ugly] modifies [man] as an adjunct.
[Brazil] modifies [from] as a complement,
[from Brazil] modifies [man] as an adjunct.
[Poems] modifies [of] as a complement.
Chapter 6: X-bar Theory 181
[of Poems] modifies [books] as a complement,
[books of poems] modifies [found] as a complement.
[the2] modifies [puddle] as a specifier,
[the puddle] modifies [in] as a complement,
[in the puddle] modifies [found] as an adjunct.
Keeping in mind the (revised) principle of modification, and the strict X-bar
structure, we next start to build the trees. I suggest you generally start
with AdjPs and AdvPs. We have one Adj here. Note that nothing modifies
this Adj. As such, it has the minimal structure given above in (120c).
130) AdjP
I
Adj'
I
D Adj N PN V NPNPD N
The! ugly man from Brazilfound books of poems in the2puddle.
Next we do NPs and PPs. Again, we'll also start on the right hand side of the
sentence. The first NP is the puddle, be sure to apply all three of the NP rules
here. Don't forget the N' node in the middle. The determiner is the specifier
of the NP, so it must be the sister to N' and daughter of NP.
131) AdjP
I
Adj'
I
D Adj N PN VNPNPDN
The,ugly man fromBrazil foundbooksofpoemsin the2 puddle.
There are two nouns in this sentence that aren't modified by anything (Brazil
and poems). Let's do these next. Even though they aren't modified by
anything theyget thefull X-bar structure, withNP, N' and N:This isbecause
the rules are not optional.
132) AdjP NP NP NP
I I I
Adj* N" N'
D Adj N PN VNPNPDN
The! ugly man from Brazilfound books of poems in the2 puddle.
There are two more nouns in this sentence (man and books), but if you look
carefully at our list of modifications (129), you'll see that they are both
modified by PPs. So in order to do them, we have to first build our PPs.
182 The Base
There are three Ps in this sentence (and hence three PPs), each of them takes
one of the NPs we've built as a complement. The objects of prepositions are
always complements. That means that they are sisters to P, and daughters
ofP':
133)
AdjP
I
Adj'
DAdjN PN VNPNPDN
Thej ugly man from Brazil found books of poems in the2 puddle.
Now that we've generated our PPs, we'll go back to the two remaining NPs.
Let's first observe that the PP in the puddle does not modify an N (it modifies
the Vfound), so it is ngi attached at this stage. Now, turn to the N books.
We start with the complement. Of poems is the complement meaning
that the PP will be the sister to the N head, and the daughter of N'.
134)
AdjP
I
Adj"
NP
DAdjN PN VNPNPDN
Theiugly man from Brazil found books ofpoems in the2 puddle.
Nothing else modifies books. When there are no more modifiers we close off
the projection with the phrase level:
135)
AdjP
I
Adj'
Chapter 6: X-barTheory
NP
183
DAdjN PN VNPNPDN
Thei ugly man fromBrazil foundbooksofpoemsin the2 puddle.
Finally, we have the NP the ugly man from Brazil. There is no complement
here, so we project to N' without any branching. Were there a complement
in the NP we would attach it first.
136)
AdjP
I
Adj' N'
DAdjN PN VNPNPDN
Thet ugly man from Brazil found books ofpoems in the2 puddle.
There aretwoadjuncts inthis NP: from Brazil andugly. As perpoint (v) (and
see the trees in (126), this can be treed two different ways. We can attach
eitheradjunct first. I'll arbitrarily pick to attach the PP first here. Because
it is an adjunct, it has to be a sister to N' and a daughter of N'.
(Note the difference between this NP and books ofpoems.)
184
137)
AdjP
I
Adj' N'
The Base
DAdjN PN VNPNPDN
Thej uglymanfrom Brazil found books ofpoemsin the2 puddle.
Nextwe attach theAdjP. Note thatbecause it is an adjunct, it has tobe sister
to an N' and daughter ofan N'. The N' it is a sister to is already in the tree
(having been added in the previous step).
138) N* NP
DAdjN PN VNPNPDN
Thej ugly man from Brazil found books ofpoems inthe2 puddle.
We're nearly finished with this NP. The determiner is a specifier, which is
a daughter of NP and a sister to N'.
Chapter 6: X-barTheory 185
139) NP
DAdjN PN VNPNPDN
The! uglymanfrom Brazil found books ofpoems in the2 puddle.
Now we turn to the VP. The verbfound has two modifiers. Books ofpoems is
a complement, and in the puddle isanadjunct. You should always start with
the complement, and then follow with the adjuncts, because complements
are closer to the head. Remember, complements are sisters to V, and adjuncts
aresisters toV. Notice that thecomplement NP, which iscloser to thehead,
isattached lower than theadjunct PP, itis thesister toVand daughter ofV.
140)
D Adj N
Thei uglyman from Brazil found books ofpoemsin the2 puddle.
186 The Base
Now we attach the adjunct PP, it has to be a sister to V (the one just created
by the previous step) and daughter of a V (which we will add here).
Since there are no other modifiers of theV, wewill also complete this phrase
with the VP:
141)
DAdjN PN VNPNPDN
The, ugly man from Brazil found books ofpoems inthe2 puddle.
Last, but not least, we apply the TP rule, and then check the tree against the
X-bar rules. Making sure that everything isattached; there are nocrossing
lines; adjuncts are sisters to a bar level, complements are sisters to a head;
and finally every head has at least an X, X', and XP on top ofit.
142)
Chapter 6: X-bar Theory
TP
187
DAdjN PN VNPNPDN
Theiugly man fromBrazil foundbooksofpoemsin the2 puddle.
Each tree will be different, of course, but with practice and patience you will
develop the skill quite easily.
You now have enough information totry General Problem Set7
7. X-bar Theory: A Summary
Let's summarize the rather lengthy discussion we've had so far in this
chapter. We started off with the observation that there seemed to be more
structure to our trees than that given by the basic phrase structure rules we
developed in chapter 3. In particular, we introduced the intermediate levels
of structure called N', V, Adj', and P'. The evidence for these comes from
standard constituency tests like conjunction, and from processes like onereplacement,
and do-so-replacement. We also saw that material on different
levels of structure behaved differently. Complements exhibit one set
of behaviors and adjuncts a different set. Next we observed that our rules
were failing to capture several generalizations about the data. Firstwas the
endocentricity generalization: allNPs, haveanNhead, allAdjPs anAdj head,
188 The Base
etc. There is no rule like NP -> V Adj. Next, there was the observation that
all trees have three levels of structure. They all have specifiers (weak
evidence here), adjuncts and complements. In response to this, we
proposed the following general X-bar theoretic rules:
143) a) Specifier rule: XP ->(YP) X1 or XP -> X' (YP)
b) Adjunct rule: X ->X' (ZP) or X ->(ZP) X'
c) Complement rule: X ->X(WP) or X -» (WP) X
These rules use variables tocapture cross-categorial generalizations. In order
tolimit thepower ofthese rules, and in order tocapturedifferences between
languages, we proposed that the options within these rules were
parameterized. Speakers of languages select the appropriate option for
their language.
This is, you'll note, a very simple system. There are, of course,
some looseends, and in the next couple ofchapters we'll try to tidy these up.
Firstofall we have the problem ofspecifiers, we only haveone specifier (the
determiner). In the next chapter we'll suggest that in fact determiners aren't
specifiers at all, instead they aretheir ownheads. Then we'll reserve specifier
positions for something else: subjects. We'll also try to integrate our TP
and CP rules into the system.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Specifier: Sister to X',daughter of XP.
ii) Adjunct: Sister to X', daughter of X'.
iii) Complement: Sister to X, daughter of X'.
iv) Head:The word that gives its category to the phrase.
v) Projection: The string of elements associated with a head that bear
the same category as the head (N, N', N', N', NP etc).
vi) Maximal Projection: The topmost projection in a phrase (XP).
vii) Intermediate Projection: Any projection that is neither the head nor
the phrase (i.e.all the X' levels).
viii) One-replacement: Replacean N' node with one.
ix) Do-so-replaccment: Replacea V with do so.
x) Specifier Rule: XP -> (YP) X' or XP ->X'(YP)
Chapter 6: X-bar Theory 189
X' -> X' (ZP) or X' -> (ZP) X'
X' -> X (WP) or X' -> (WP) X
xi) Adjunct Rule:
xii) Complement Rule:
xiii) Additional Rules:
CP -> (C) TP
TP -> NP VP
XP -> XP Conj XP
X' -> X' Conj X"
X -> XConj X
xiv) Parameterization: The idea that there is a fixed set of possibilities in
terms of structure (such as the options in the X-bar framework), and
people acquiring a language choose from among those possibilities.
xv) Principleof Modification (revised): If a YP modifies some head X,
then YP must be a sister to Xor a projection of X(i.e., X', XP).
Further Reading
Baltin, Mark and Anthony Kroch (1989) Alternative Conceptions of Phrase
Structure. Chicago: University of Chicago Press.
Borsley, Robert (1996) Modern Phrase Structure Grammar. Oxford: Blackwell.
Carnie, Andrew (1995) Head Movement and Non-Verbal Predication. Ph.D.
dissertation, MIT.
Chametzky, Robert (1996) A Theory of Phrase Markers and the Extended Base.
Albany: SUNY Press.
Chomsky, Noam (1970) Remarks on nominalization. In Roderick Jacobs and
Peter Rosenbaum (eds.), Readings in English Transformational Grammar.
Waltham: Ginn. pp. 184-221.
Jackendoff, Ray (1977) X-bar Syntax: A Theory ofPhrase Structure. Cambridge:
MIT Press.
Kayne, Richard (1994) The Antisymmetry ofSyntax. Cambridge: MIT Press.
Lightfoot, David (1991) How to Set Parameters: Evidencefrom Language Change.
Cambridge: MIT Press.
Speas, Margaret (1990) Phrase Structure in Natural Language. Dordrecht:
Kluwer Academic Publishers.
Stowell, Tim (1981) Origins of Phrase Structure. Ph.D. dissertation, MIT.
Travis, Lisa de Mena (1984) Parameters and Effects of Word Order
Derivation. Ph.D. dissertation, MIT.
190 The Base
General Problem Sets
1. Complements vs. Adjuncts in NPs
[Application of Skills; Basic]
Using the tests you have been given (reordering, adjacency, conjunction of
likes, one-replacement) determine whether the PPs in the following NPs are
complements or adjuncts; give the examples that you used in constructing
your tests. Some of the NPs have multiple PPs, be sure to answer
the question for every PP in the NP.
a) A container [of flour]
b) A container [with a glass lid]
c) The collection [of figurines] [in the window]
d) The statue [of Napoleon] [on the corner]
e) Every window [in the building] [with a broken pane]
2. Adjectives
[Critical Thinking; Intermediate]
Are adjectives complements or adjuncts to the N? Use the tests you have
been given to determine ifadjectives are complements or adjuncts. Do NOT
use the reordering test - it will not work because adjectives in English
are strictly ordered by other principles. Also confine yourself to the adjectives
listed below. (Other adjectives, such as leatherin leathershoes or Chemistry
in Chemistry Professor, behave differently. However, you can use these
adjectives as interveners ifyou need to check adjacency to the head.)
hot, big, red, tiny, ugly
3. German Noun Phrases
[Data Analysis; Intermediate/Advanced]
Consider sentence (a) from German:.9
a) Die schlanke Frau aus Frankreich isst Kuchen mit Sahne.
the thin woman from France eats cake with cream
"The thin woman from France eats cake with cream."
The following sentences are grammatical ifthey refer to the same woman
described in (a):
Thanks to Simin Karimi for providing the data for this question, and to Susi
Wurmbrand for clarifying the facts.
Chapter 6: X-barTheory 191
b) Die Schlanke aus Frankreich isst Kuchen mit Sahne.
"The thin one from France eats cake with cream."
c) Die aus Frankreich isst Kuchen mit Sahne.
"The one from France eats cake with cream."
d) Die Schlanke isst Kuchen mit Sahne.
"The thin one eats cake with cream."
e) Die isst Kuchen mit Sahne.
"She eats cake with cream."
Now consider sentences (f-i):
f) Die junge Koenigin von England liebte die Prinzessin.
The young queen of England loved the princess
"The young queen of England loved the princess."
g) Die Junge liebte die Prinzessin.
"The young one loved the princess."
h) Die liebte die Prinzessin.
"She loved the princess."
i) *Die von England liebte die Prinzessin.
"*The one of England loved the princess."
(Native speakers of German should assume the judgments given even if
they don't agree with them.)
Assume the following things:
i) Der/Die are always determiners, they are never nouns or pronouns.
ii) Schlanke and junge are always adjectives, even in sentences (d) and
(g) - assume they never become nouns. (Ignore the rules of German
capitalization.)
The questions:
1) Describe and explain the process seen in (a-e) and (f-i), be sure
to make explicit reference to X-bar theory. What English phenomenon
(discussed in this chapter) is this similar to? Make sure you analyze
the German sentences not the English translations.
2) Draw the trees for sentences (a) and (f). Sentence (a) requires two
different trees (important hint: the relevant ambiguity in (a) is inside
the subject NP, not in the position of the PP mit sahne).
192 The Base
3) Explain the ungrammaticality of (i) in terms of X-bar theory. In particular
explain the difference between it and sentence (c). Draw trees
to explicate your answer.
4. Complements and Adjuncts in VPs
[Application of Skills; Basic]
Using the tests you have been given (reordering, adjacency, conjunction
of likes, c/o-so-replacement) determine whether the marked NPs, PPs and
AdvPs in the following VPs are complements or adjuncts; give the examples
that you used in constructing your tests. Some of the VPs have multiple PPs
and AdvPs, be sure to answer the question for every PP, NP, and AdvP
in the VP.
a) Erin [VP Udvp never] keeps [NP her pencils] [PP in the correct drawer]].
b) Dan [VP walked [PP to New Mexico] [PP in the rain] [AdvP last year]].
5. Japanese
[Data Analysis; Basic]
Consider the following data from Japanese:
a) Masa-ga kita. "Masa came."
b) Toru-ga shinda. "Toru died."
c) Kumiko-ga yonda. "Kumiko read."
d) Kumiko-ga hon-o yonda. "Kumiko read the book."
e) Toru-ga Kumiko-o mita. "Toru saw Kumiko."
f) Kumiko-ga Toru-o mita. "Kumiko saw Toru."
g) Hon-ga akai desu. "The book is red."
h) Toru-ga sensei desu. "Toru is a teacher."
i) Masa-ga ookii desu. "Masa is big."
j) Sono hon-ga ookii desu. "That book is big."
k) Toru-ga sono akai hon-o mita. "Toru saw that red book."
1) What is the function of the suffixes -o and -ga?
2) What is the word order of Japanese?
3) Does the complement precede or followthe head in Japanese?
4) Do adjuncts precede or follow the head in Japanese?
5) Do specifiers precede or followthe X' node in Japanese?
6) Draw the tree for sentence (k) using X-bar theory. Keep in mind your
answers to questions (1-5).
Chapter 6: X-barTheory 193
6. Parameters
[Data Analysis; Basic to Intermediate]
Go back to the foreign language problems from chapters 3 and 4, (Hiaki,
Irish, Bambara, Hixkaryana, Swedish, Dutch, Tzotzil) and see if you can
determine the parameter settings for these languages. You may not be able
to determine all the settings for each language. (Suggestion: put your answer
in a table like the one below. English is done for you as an example.)
Assume the following: Determiners are typical examples of specifiers,
Adjectives and many PPs (although not all) are adjuncts, "of PPs and direct
objects are complements. Be sure to check the complement/adjunct relation
in all categories (N, Adj, Adv, V, P etc.) if you can.
Specifier Adjunct Complement
English (YP) X' Both X(ZP)
7. Trees
[Application of Skills; Basic to Advanced]
Draw the X-bar theoretic trees for the following sentences. Treat possessive
NPs like H4loi'se's as specifiers. Several of the sentences are ambiguous;
draw only one tree, but indicate using a paraphrase (or paraphrases)
which meaning you intend by your tree.
a) Abelard wrote a volume of poems in Latin for HeloTse.
b) Armadillos from New Yorkoften destroy old pillowcases with their
snouts. (NB: assume "their" is a determiner)
c) People with boxes of old clothes lined up behind the door of the building
with the leaky roof.
d) That automobile factories abound in Michiganworries me greatly.
e) No-one understands that phrase structure rules explain the little
understood phenomenon of the infinite length of sentences.
f) My favorite language is a language with simple morphology and
complicated syntax.
g) Ivan got a noogie on Wednesday from the disgruntled students of
phonology from Michigan.
h) The collection of syntax articles with the red cover bores students of
syntax in Tucson
i) The red volume of obscene verse from Italyshocked the puritan soul of
the minister with the beard quite thoroughly yesterday,
j) The biggest man in the room said that John danced an Irishjig from
County Kerry to County Tipperary on Thursday.
194 The Base
k) A burlap sack of potatoes with mealy skins fell on the professor
of linguistics with the terrible taste in T-shirts from the twelfth story
of the Douglass Building last Friday.
I) The bright green filing cabinet was filled to the brim with the most boring
articles from a prestigious journal of linguistics with a moderately large
readership
Challenge Problem Sets
Challenge Problem Set 1: Intermediate Structure
[Application of Knowledge, Critical Thinking; Advanced-Challenge]
The following verb phrase is ambiguous in its structure
Adam [VP frequently buys paintings from Natasha].
The ambiguity has to do with where UdvP frequently] and [PP from Natasha]
are attached in the string of V categories. Note that the V rule can be either
V -» V (PP) or V ->V (AdvP) and these rules can apply in either order.
Using the do-so/did-so/did-too replacement test, provide some sentences
that show that there is an ambiguity in structure here.
Challenge Problem Set 2: ComplementAdjPs?
[Data Analysis and Critical Thinking; Challenge]
You should do general problem set 2 above before attempting this problem.
Part 1: Consider the following adjectives:
leather (as in leather shoe), Chemistry (as in Chemistrystudent)
Using your tests for complements and adjuncthood in NPs (adjacency,
one-replacement, coordination of likes, only one complement - the test
of reordering doesn't work since adjectives in English are ordered by
other principles), decide whether these adjectives are functioning more like
complements or adjuncts. Contrast them explicitly to adverbs such as red
and big. Provide the relevant examples to support your claim.
Part 2: Two analyses of these adjectives have been proposed. One is that
they are complements (a); the other, more common, analysis is that
these aren't adjectives at all, but are noun-noun compounds (notice that both
leather and Chemistry can function as nouns in their own right) as in (b).
Chapter 6: X-barTheory 195
a) NP b) NP
I
N'
N
leather shoes
Which of these proposals is right? Try to come up with some arguments
to distinguish between them. The following data may be helpful to you.
But you should look for arguments beyond this data (i.e., come up with data
of your own).
c) plastic and leather shoes
d) ?very leather shoes
e) "Very Chemistry professor
Challenge Problem Set 3: Ambiguous AdjPs ?
[Data Analysis and Critical Thinking; Challenge]
Before trying this problem set you should try Challenge Problem Set 2
above.
Part 1: No matter what your answer to challenge problem set 2 was above,
assume for the moment that some AdjPs can function as adjuncts and others
can function as complements. This requires that we modify our parameter
settings for English. Propose a revised set of parameters for English to allow
for the possibility that Chemistry in Chemistry professor is a complement.
Note: your proposal must explain why in English object NP complements
cannot appear before verbs but AdjP complements can appear before nouns
(i.e., your proposal must account for why complement-head order is allowed
in NPs but not in VPs).
Part 2: Consider the following ambiguous NP:
a) the German teacher
It can mean either a teacher (say of math) who is German, or it can mean
someone (of any nationality) who teaches the German language. Using the
complement/adjunct distinction and the following data, explain this ambiguity
in meaning. Pay attention to the meaning of German (whether it refers
196 The Base
to the nationality or the subject) in each of the following sentences.
Draw trees to explain your answer.
b) the French German teacher
c) the math and German teacher
d) ... not the American teacher but the German one
Challenge Problem Set4: Complements to Adj heads™
[Critical Thinking and Application of Skills; Challenge]
Part 1: Consider the word sick. This word seems to have two or more
meanings. One meaning corresponds to the meaning "ill," as in / feel sick.
The other meaning is something like "I've had enough," as in the expression
/ am sick of it. This second meaning seems to take a complement PP. The
evidence for this is twofold: (1) To get this meaning of sick the complement
must be present in the sentence. Otherwise we understand the physical
meaning of sick. (2) The preposition that we find, of, is the most common
preposition used with complement PP of adjectives and nouns. Judging from
their meanings and other properties, do any of the adjectives below regularly
occur with complements?
delightful, familiar, sensitive, adjacent, full
Part 2: Draw the trees for the following two sentences according
to the principles of X-bar Theory. Think about the reasons you would use
for considering a PP either a complement or an adjunct of an adjective. This
is what you need to think about to get each PP attached in the right place.
a) The director is as aware of the problems as the committee members.
b) Everyone was curious about it to the nth degree.
10 Thanks to LeslieSaxon for contributing this problem set.
Extending X-bar
Theory to Functional
Categories
0. INTRODUCTION
In the last chapter, we looked at a refinement of our phrase structure rules
that not only accounted for intermediate structure, but also generalized
patterns across categories. This refinement is X-bar theory:
1) a) Specifier rule: XP -> (YP)X'orXP -> X' (YP)
b) Adjunct rule: X -> X' (ZP) orX -> (ZP) X"
c) Complement rule: X -»X (WP) orX -> (WP) X
These rules not only generate most of the trees we need for the sentences
of the world's languages, they also capture the additional properties
of hierarchical structure found within the major constituents. This said, you
may have noticed that this system is far from perfect. First, there is the status
of specifiers. In particular, the specifier rule we proposed above requires that
the specifier be a phrase (XP) level category. However, the only instances
of specifiers we've looked at are determiners, which appear notto be phrasal.
In this chapter, we will look at determiners, and specifiers, and propose
a new category that fits X-bar theory: a determinerphrase (DP). We will see
that determiners are not specifiers. Instead, we'll claim that the specifier
position is used to mark a particular grammatical function: that of subjects.
You'll see that specifiers (of all categories) are where subjects go.
Another troubling aspect of the X-bar theory is the exceptional CP and
TP rules that we have yet to incorporate into the system:
198 The Base
2) CP->(C)TP
TP -> NP (T) VP
These rules do not fit X-bar theory. In the X-bar rules in (1), you'll note
that the only obligatory element is the head. In the sentence rules in (2),
the opposite istrue: theonlyoptionalelement isthehead itself. In thischapter,
we will look at how we can modify these so that they fit into
the more general pattern.
1. Determiner Phrases (DPs)
In the last chapter, for lack of a better place to put them, we put determiners,
like the, a, that, this, those, these in the specifiers of NPs. This however, violates
one of the basic principles underlying X-bar theory: All non-head material
must be phrasal. Notice that this principle is a theoretical rather than
an empirical requirement (i.e., it is motivated by the elegance of the theory
and not by any data), but it is a nice idea from a mathematical point of view,
and it would be good if we could show that it has some empirical basis.
One thing to note about determiners is that they are heads. There can
only be one of them in an NP (this isn't true cross-linguistically, but
for now let us limit ourselves to English):
3) *the that book
In other words, they don't seem to be phrasal.1 If our requirement says
that the only thing that isn't a phrase in an NP is the N itself, then we
have a problem. One solution, perhaps not obvious, to this is to claim that
the determiner is not actually inside the NP. Instead, it heads its own phrasal
projection. This was first proposed by Abney (1987):
1 In the last chapter we used this exact same piece of evidence to distinguish
specifiers from adjuncts. As an exercise, you could try to construct an argument
that distinguishes these two accounts of the same data.
4)
D
Old view
NP
N'
N
Chapter 7: Extending X-barTheory
DP hypothesis
DP
199
Determiners, in this view, are not part of the NP. Instead the NP is
the complement to the determiner head. This solution solves the theoretical
problem of the non-phrasality of the D, but we still need empirical evidence
in its favor.
One piece of evidence comes from the behavior of genitive (possessive)
NPs. There are two kinds of possessive NPs. The first is of less interest to us.
This one is often called the free genitive or of-genitive:
5) a) the coat of the panther
b) the roof of the building
c) the hat of the man standing over there
The free genitive uses the preposition of to mark the possessive relation
between the two NPs. More important in terms of evidence for DP is
the behavior of the other kind of possessive: the construct or 's-genitive.
6) a) the panther's coat
b) the building's roof
c) the man standing over there's hat
There are a couple of important things to note about this construction. Notice
first that the 's marker appears after the entire possessor NP. For example,
it attaches to the whole phrase the man standing over there not just to
the head man:
7) a) [the man standing over therej's hat
b) *the man's standing over there hat
This means that '$ is not a suffix. Instead it seems to be a small word
indicating possession. Next, note that it is in complementary distribution
with (i.e., cannot co-occur with) determiners:
8) a) *thebuilding's the roof (cf.the roof of the building)
b) *the panther's the coat (cf.the coat of the panther)
200 The Base
c) *the man standing over there's the hat (cf. the hat of the man
standing over there)
Unlike the o/-genitive, the 's-genitive does not allow both the nouns to have
a determiner. In other words, 's and determiners are in complementary
distribution. As in other domains of linguistics, when two items are in
complementary distribution, they are instances of the same thing. (Take
for example, phonology, where when two phones are found in different
environments - in complementary distribution - then they are allophones
of the same phoneme.) Determiners like the and 's are different tokens of
the same type. Assuming that 's is a determiner, and assuming the DP
hypothesis holds true, we can now account for the positioning of the 's
relative to the possessor. The 's occupies the head D position, and
the possessor appears in its specifier:
9)
possessor D
's
A tree for (8c) shows this
10)
possessed
man standing over there
Chapter 7: Extending X-bar Theory 201
The possessor [DP2 the man standing over there] sits in the specifier of DPX,
which is headed by 's. So 's follows the whole thing.2 Notice that with our
old theory, where determiners were specifiers of NP, there is no way at all
to generate 'sas a determiner and to also have the possessor NP preceding it.
11)
the man standing over there
The X-bar rules don't provide any place to attach this pre-determiner NP,
if determiners are specifiers.
Notice that in the tree in (10) there is a specifier of DPX (filled by DP2).
Note further that this specifier is phrasal (projects to an XP). Which means
that it meets with our requirement that all non-head material be phrasal.
You might ask if by moving determiners out of the specifier we have
completely destroyed the empirical justification for the specifier rule.
Actually, we haven't. Again if you look closely at the tree in (10) we still
have a specifier, it just isn't D„ instead it is the DPpossessor (DP2). Further,
as we will see below, there are other related uses for the specifier positions.
In particular, we will come to associate specifiers with subjects of various
kinds of constituents.
You now have enough information to try Challenge Problem Sets 1 &2
2. A Descriptive Tangent into Clause Types
A clause is essentially a subject (usually a DP that has the property indicated
by the predicate; this is what the clause is about) and a predicate phrase
(a group of words that assign a property to the subject). The most obvious
kind of clauseis the simplesentence. In the following examples, the subject
is indicated in italics and the predicate phrase is in bold:
12) a) Theboyran.
b) Howard is a linguistics student.
2 We might extend this analysis to possessive pronouns. Take the pronoun his
for example, we could analyze this as he occupying the specifier of 's, then there is
a morphological operation that turns he+'s into his. (This of course is less plausible
with pronouns likeher or your; unlessthe morphological rule is extremelyabstract.)
202 The Base
As we'll see below, there are many other kinds of clauses. But we can use
this as a working definition.
A clause that stands on its own is called a root, matrix, or main clause.
Sometimes, however, we can find examples of clauses within clauses.
Examples of this are seen below:
13) a) [Petersaid [that Danny danced]].
b) [Bill wants [Susan to leave]].
In each of these sentences there are two clauses. In sentence (13a), there is the
clause (that) Danny danced which isinside theroot clause Peter said that Danny
danced. In (13b), we have the clause Susan to leave which has the subject
Susan, and the predicate phrase (to) leave. This is contained within the main
clause Bill wants Susan to leave.
Both of these clauses within clauses are called embedded clauses.
Another name for embedded clause is subordinate clause. The clause
containing the embedded clause is still called the main or root clause.
Embedded Clauses are Part of Main Clauses
A very common error among new syntacticians is to forget that
embedded clauses are contained within main clauses. That is, when
faced with identifying what is the main clause in a sentence like
i) Peterthinks that Cathyloves him.
most students will properly identify theembedded clause, as (that) Cathy
loves him, but will claim that the main clause is only Peter thinks. This is
completely incorrect. Peter thinks is not a constituent. The main clause is
everything under the root TP node. So the main clause is Peter thinks that
Cathy loveshim. Be very careful about this.
Using the TP and CP rules we developed in chapter 3, the structure of
a root clause containing an embedded clause is given below (I've obscured
the irrelevant details with triangles):
Chapter 7:ExtendingX-bar Theory 203
14) TP < Main Clause
/
Subject
DP
Peter
VP <— Predicate phrase
Embedded Clause
VP <—Predicate phrase
Danny danced
In addition to the distinction between main and embedded clauses,
we can also distinguish among specifier, complement, and adjunct clauses.
Here are some examples of complement clauses:
15) a) Heidi said [that Art loves peanut butter],
b) Colin asked [if they could get a mortgage].
These complement clauses (CPs) are sisters to the verb, and thus
complements. Clauses can also appear in adjunct positions. Relative clauses
are one example of adjunct clauses:
16) [The man [I saw get into the cab]] robbed the bank.
The relative clause in (16) [I saw get into the cab] modifies the head man.
Specifier clauses are ones that serve as the subject of a sentence (why these
are specifiers will be made clear below):
17) a) [[Peopleselling their stocks] caused the crash of 1929].
b) [[ForMary to love that boor] is a travesty].
To summarize, we have two basic kinds of clauses, main and embedded.
Embedded clauses are contained within main clauses. Further, there are
three types of embedded clauses: specifier clauses, complement clauses
and adjunct clauses. This is summarized in the following table:
18)
Main clauses Embedded clauses
specifier
clauses
complement
clauses
adjunct clauses
204 The Base
There is another way ofdividing up the clause-typepie. We classclauses
into two groups depending upon whether they are tensed or not.3 Clauses
with predicates that are tensed are sometimes called (obviously)
tensedclauses, but you may more frequently find them called finite clauses.
Clauses without a tensed verb are called tenseless or non-finite clauses
(sometimes also infinitival clauses).4
19) a) I said [that Mary signed my yearbook]. tensed orfinite
b) I want [Mary to sign my yearbook]. tenseless ornon-finite
There are a number of tests for distinguishing finite from non-finite clauses.
These tests are taken from Radford (1988). The embedded clause in sentence
(20a) is tensed, the one in (20b) is untensed. I have deliberately selected
a verb that is ambiguous between tensed and untensed in terms of
its morphology (suffixes) here as an illustration:
20) a) I know [you eat asparagus]. finite
b) I've never seen [you eat asparagus]. non-finite
One way to tell if a clause is finite or not is to look for agreement and tense
morphology on the verb. These include the -s ending associated with third
person nouns (he eats) and the past tense suffixes like -ed. The above
examples don't show any such suffixes. However, if we change the tense
to the past a difference emerges:
21) a) I know you ate asparagus. finite
b) *I've never seen you ate asparagus. non-finite
Finite clauses allow past tense morphology (the ate form of the verb eat),
non-finite clauses don't. The same effect is seen if you change the person
of the subject in the embedded clause. Third person subjects trigger
the -s ending. This is allowed only in finite clauses.
3 There is a third kind of clause that we won't discuss here, called "small clauses."
Small clauses don't have verbal predicates (that is, a DP, PP, or AP serves as
the predicate. These generally don't get tense marking. An example is the embedded
string in:
i) [Maurice considers [Jason a fine upstanding gentleman]].
Small clauses are an important part of syntactic theory, but they are notoriously
difficult to spot until you have some practice. For the purposes of this text we'll just
ignore small clauses, but if you pursue syntax at a higher level you'll have to learn
how to identify them.
4 In many languages, the form of a verb found in a non-finite clause is called the
infinitive. In English, infinitives are often marked with the auxiliary to, as in tosign.
Chapter 7: Extending X-barTheory
22) a) I know he eats asparagus.
b) *I've never seen him eats asparagus.
finite
non-finite
205
The case on the subject of the noun is often a giveaway for determining
whether or not a clause is finite. Case refers to the notions nominative
and accusative introduced in chapter 1, repeated here:
23)
Nominative Accusative Anaphoric
Singular Plural Singular Plural Singular Plural
1 I we me us myself ourselves
2 you you you you yourself yourselves
3masc he
they
him
them
himself
themselves3 fern she her herself
3neut it it itself
If the clause is finite, then a subject pronoun will take the nominative case
form:
24) I know heeats asparagus. finite
If the clause is non-finite then the subject will take the accusative form:
25) I've never seen himeat asparagus. non-finite
One test that works most of the time, but is not as reliable as the others,
is to see if the subject is obligatory. If the subject is obligatory, then the clause
is finite. If the subject is optional, or is not allowed at all, then it is non-finite.
(Note: this test only works for English; in many languages, such as Spanish,
subjects of finite clauses are optional.)
26) a) I think that he eats asparagus. finite
(cf.*Ithink that eats asparagus.)
b) I want (him) to eat asparagus. non-finite
(cf.I want to eat asparagus.)
Another way to tell if a clause is finite or not is by looking at
the complementizer. The complementizer for is only found with non-finite
clauses. By contrast that and if are only found with tensed clauses:
27) a) I wonder if he eats asparagus. finite
b) I think that he eats asparagus. finite
c) [For him to eat asparagus] is a travesty. non-finite
d) I asked for him to eat the asparagus. non-finite
As a final test, we can note that finite and non-finite clauses take
different kinds of T elements. The T in tensed clauses can contain auxiliaries
206 The Base
and modals like will, can, must, may, should, shall, is, have. By contrast
the only auxiliary allowed in non-finite clauses is to.5
28) a) I think [he willeat asparagus].
b) I want him to eat asparagus, (cf. *I want him will eat asparagus.)
This last property gets at the heart of the distinction between finite and nonfinite
clauses. In structural terms the difference between a finite and a nonfinite
clause lies in terms of what kind of T the clause has. If a clause is finite
it bears some tense feature (like [±past] or [ifuture]). If it is non-finite,
it doesn't have any of these features. The question of how this works for
clauses where there is no auxiliary, we'll leave as a bit of a mystery for now,
but will return to later in this chapter.
Let's summarize the discussion we've had thus far. We've been looking
at a number of terms for describing various kinds of clauses. We defined
clauses as a subject and a predicate phrase. We distinguished root or main
clauses from embedded clauses. Embedded clauses come in three types:
specifier clauses, complement clauses and adjunct clauses. The other
dimension along which we can describe clauses is the finite/non-finite
distinction.
With this terminology under our belts, we'll now turn to the structure
of clauses, and see if we can make them fit better into X-bar theory.
You nowhaveenough information to try General Problem Sets 1,2 &3
3. COMPLEMENTIZER PHRASES (CPS)
We've observed that the TP rule and the CP rule stand out, since they
don't fit X-bar theory. In X-bar theory, the head is always obligatory.
This is not true of these two rules:
29) a) CP -> (C) TP
b) TP->DP(T)VP
In fact, it is a fairly trivial matter to change these rules into X-bar theoretic
format. Let us deal with the CP rule first. If we take X-bar theory to extend
to CPs, we can assimilate the rule in (29a) to get a tree like that in (30):
English has two words to. One is a preposition,the other is non-finiteT.
30) CP
C
that
Chapter 7:Extending X-barTheory 207
C
TP
This CP structure has C as the head, a TP complement and an empty
specifier position (this empty specifier position will become very important
later for us when we do zy/i-movement in chapter 11).
We might ask how pervasive this rule is in our mental grammars. That
is, do all clauses have CPs, or do only embedded clauses have CPs? On the
surface, the answer to this question seems obvious: Only embedded clauses
have CPs, since only embedded clauses appear to allow complementizers:
31) a) John thinks that asparagus is yummy.
b) That asparagus is yummy, (cf. Asparagus is yummy.)
However, there is evidence that all clauses, even root clauses like (31),
require some kind of complementizer.
32) Asparagus grows in California.
In particular, we'll claim that some sentences have null complementizers.
Don't assume that I'm crazy. No matter how strange this proposal sounds,
there is actually some good evidence that this is correct. The tree in (33)
shows one of these null complementizers.
33)
Asparagus grows in California
The evidence for this claim comes from cross-linguistic comparison
of questions among languages. In particular, we'll focus on yes/no questions
(see chapter 9 for more discussion on these). These are questions that can
be answered with either yes, no or maybe. Examples of yes/no questions
in English are given below:
34) a) Did John leave?
b) Have you seen Louis?
208 The Base
In English, to form a yes/no question you either insert some form of the verb
do (do, does, did) before the subject, or you invert the subject and the auxiliary
(You have seen Louis. —» Have you seen Louis?). This operation is called subjectaux
inversion (more on this in chapter 9). In many other languages, however,
yes/no questions are formed with a complementizer particle that precedes
the verb. Take for example, Irish, which indicates yes/no questions
with a special particle Ar (or its allomorph An):
35) ArthitSean?
Q fall John
"Did John fall?"
Languages like English that use subject-aux inversion don't have special
complementizer question particles. The opposite also holds true. If a
language has complementizer question particles, then it won't have subjectaux
inversion. The phenomena are in complementary distribution. It seems
reasonable to claim then, that question complementizers and subject-aux
inversion are part of the same basic phenomenon. In order to make
this concrete, let's make the following proposal: There is a question
complementizer particle in English, just like there is in Irish. The difference
is that in English this complementizer particle is null (has no phonological
content). We will represent this null complementizer with the symbol 0[+Q].
It has no phonological content, but it must be realized or pronounced
someway. The way English satisfies this requirement is by moving T
into the C head:
36)
This results in the correct order, where the auxiliary (in T) now appears
before the subject. By contrast, languages like Irish don't utilize this
mechanism. Instead they have a particle that fills their [+Q] complementizer
(like Ar/An in Irish).
English does, in fact, have an overt [+Q] complementizer, but it is only
found in embedded questions. This complementizer is if. Unsurprisingly,
subject-aux inversion is completely disallowed when if is present:
37) a) Fabio asked if Claus had run a marathon.
Chapter 7: Extending X-barTheory 209
b) *Fabio asked if had Claus run a marathon.
c) *Fabio asked had if Claus run a marathon.
d) ?Fabio asked had Claus run a marathon.
//occupies the [+Q] complementizer, so no subject-aux inversion is required
(or allowed).
Given the existence of overt root complementizers in other languages
and the evidence that subject-aux inversion patterns like these overt root
complementizers, we can conclude that, for questions at least, there are
complementizers (and CPs) present, even in main clauses.
Of course, we haven't yet shown that non-question sentences have a
root complementizer. For this, we need to add an extra step in the argument.
You can only conjoin identical categories. If sentences showing subject-aux
inversion use a null complementizer and if you can conjoin that question
with a non-question (such as a statement), then that statement must
also include a (null) complementizer and CP. It is indeed possible to conjoin
a statement with a question:
38) [You can lead a horse to water] but [will it drink]?
Since the second clause here shows subject-aux inversion, we know there is
a 0|+Q| question complementizer present. By extension, we know that
the clause it is conjoined with must also have a complementizer - this time,
a non-question 0|_q|. A CP can only be conjoined with another CP.
0 [-Q]
DP
A
TP
T
can
C
"\. 0l+Ql
VP A
You leada \^
horse to water
This is an argument for null complementizers attached to root clauses, even
in simple statements. From this point forward, we will assume that there is
a CP on top of every clause. For brevity's sake, I may occasionally leave
this CP off my trees, but the underlying assumption is that it is always there.
You should always draw it in when you are drawing your trees.
210 The Base
4. Tense Phrases (TPs)
The other rule that doesn't fit the X-barpattern is our S rule:
40) TP -> DP (T) VP
Assimilating this rule to X-bar theory results in a structure like the following:
41) TP
DP
A
(subject) T VP
A
In this tree, S is replaced by TP; the subject DP sits in the specifier of TP,
and the VP is the complement. (This is our first clear instance where
the notion of specifier corresponds to the notion of subject. We will consider
some other cases below.) Again the problem here is that the element
that we have designated as the head of the phrase (T) is apparently optional.
In X-bar theory, heads are the only obligatory element.
In chapter 2, we equated T with auxiliaryverbs. But we might ask what
happens in clauses where there is no auxiliary: Is there a TP? Is there a T?
Can we make the same claim we did for CPs that the C is obligatory?
In order to answer this question, let's make the following observation:
Tense inflection on a verb is in complementary distribution with auxiliaries
(you never get both of them at the same time):
42) a) The roadrunner walks funny.
b) The roadrunner is walking funny.
c) *Theroadrunner is walks/walkings funny.
Recall that when two elements are in complementary distribution then they
are instances of the same category. This means that T is both auxiliaries and
inflectional endings on verbs. Similar evidence comes from coordination.
Recall that you can only coordinate two items that are of the same category
and bar level. In the following sentence, we are conjoining a T' that
has an auxiliary with a T that has a tensed verb. The tense inflection and
auxiliary are italicized.
43) [tpI [T[T kissed the toad] and [Tmust go wash my mouth now]]].
This evidence suggests that the two T's are identical in some deep sense: that
is they both involve a T node: one an auxiliary, the other a tense inflectional
ending.
Chapter 7: Extending X-barTheory 211
If you think about the basic order of the elements we seem to have
argued ourselves into a corner. Auxiliaries appear on the left of verbs,
and inflectional suffixes (like -ed, and -s) appear on the right:
44) a) He will go.
b) He goes.
There are other differences between auxiliaries and inflectional suffixes.
For example, auxiliaries, but not suffixes undergo subject-aux inversion.
If we are to claim that inflectional suffixes and auxiliaries are both instances
of T we have to account for these differences.
One possibility is to claim that both inflectional suffixes and auxiliaries
are indeed generated under T. They differ, however, in terms of whether
they can stand alone or not. Auxiliaries are independent words and can
stand alone. By contrast, suffixes like -s and -ed have to be attached to a verb.
Much like the case of moving T to C in order to pronounce 0[+q], we might
hypothesize that endings like -s and -ed can't be pronounced in isolation,
so they move to attach to the verb. In particular they seem to lower onto the
verb: The following tree shows how this would work for the simple sentence
He walked. This sentence starts out as [he-edwalk] then the -edending lowers
to attach to the end of the verb:
45)
Notice that both the movements we have proposed (T-affix lowering, and
T —» C) have morphophonological motivations. Auxiliaries move to 0[+qj
to pronounce it, inflectional endings lower to V since they are verbal suffixes.
There is much more to the study of T and C and movement of these
elements. (For example, the issue of what happens when you have both
a 0j+q] and an inflectional suffix that need to be pronounced leaps to mind.)
We will return to these issues in chapter 9.
You can now try General Problem Sets4 &5 andChallenge Problem Set 3
212 The Base
5. CP,TP, DP Tree
Here is the tree drawn in section6.2of chapter 6, but with CP, TP, and DP:
46) CP
N' PP
^\ I
AdjP N' P'
I I
Adj' N P DP
J man from I
Adj D'
ugly
D NP
0
N'
I
N
Brazil
This tree has the subject DP in the specifier of TP. The past tense ending is
in T, and lowers to the verb (we of course have to assume that there is
some morphological readjustment process that turnsfinded into found). You
will also notice that we have a null 0[_q] complementizer. In addition you'll
note that all NPs are complements to DPs. In a move parallel to having null
Cs, I have drawn in null 0 D heads as well, although this is a matter of some
controversy.
Chapter 7: Extending X-barTheory 213
Ideas, Rules, and Constraints Introduced in this Chapter
i) DeterminerPhrase (DP): D is not in the specifier of NP. D heads its
own phrase: [DP [D D NP]].
ii) Complementizer Phrase (CP): C is the head of CP and is obligatory
in all clauses, although sometimes phonologically null:
[CP[CCTP]].
iii) Tense Phrase (TP): T is the head of TP and is obligatory in all
clauses, sometimes it involves lowering of the affix to the V.
Thesubject DP occupies the specifier position: [yp DPsubject [T T VP ]].
iv) Free Genitive/oi-Genitive: Possessed of the possessor.
v) Construct Genitive/'s-Genitive: Possessor 's possessed.
vi) Subject: A DP which has the property indicated by the predicate
phrase. What the sentence is about. In most sentences, this surfaces
in the specifier of TP.
vii) Predicate Phrase: A group of words that attributes a property to the
subject. (In most sentences this is the VP, although not necessarily
so.)
viii) Clause: A subject and a predicate phrase. Always a CP in our system
ix) Root, Matrix, or Main Clause: A clause (CP) that isn't dominated by
anything.
x) Embedded Clause/SubordinateClause: A clause inside of another.
xi) Specifier Clause: An embedded clause in a specifier position.
xii) Adjunct Clause:An embedded clause in an adjunct position.
xiii) Complement Clause: An embedded clause in a complement
position.
xiv) Tenseless or Non-finite Clause: A clause that isn't tensed (e.g., I
want [Mary to leave]).
xv) Tensed or Finite Clause: A clause that is tensed.
xvi) Yes/No Question: A question that can be answered with a yes, a noor
maybe.
214 The Base
xvii) Subject-Aux Inversion: A means of indicating a yes/no question.
Involves movement of T to 0(+Q| complementizer for morpho
phonological reasons.
xviii) Affix Lowering: The lowering of inflectional suffixes to attach to
their verb.
mmmmmammmmmm
Further Reading
Abney, Steven (1987) The English Noun Phrase in its Sentential Aspect.
Ph.D. dissertation, MIT.
Chomsky, Noam (1991) Some notes on economy of derivation and
representation. In Robert Freidin (ed.), Principles and Parameters in
Comparative Grammar. Cambridge: MIT Press, pp. 417-54.
Emonds, Joseph (1980) Word order in Generative Grammar. Journal of
Linguistic Research 1, 33-54.
Pollock, Jean-Yves (1989) Verb-movement, Universal Grammar, and the
structure of IP. Linguistic Inquiry 20, 365-424.
General Problem Sets
1. Subjects and Predicate Phrases
[Data Analysis; Basic]
In each of the following clauses identify the subject and the predicate phrase.
Some sentences contain multiple clauses, be sure to identify the subjects
and predicate phrases of a//clauses.
a) The peanutbutter has got6 moldy.
b) The duffer's swing blasted the golf ball across the green.
c) That Harry loves dancing is evidenced by his shiny tap shoes.
d) The Brazilians pumped the oil across the river.
2. Clause Types
[Data Analysis; Basic]
The following sentences are "complex" in that they contain more than one
clause. For each sentence, identify each clause. Remember main clauses
include embedded clauses. Identify the complementizer, the T, and the
You may prefer gotten to gothere. The choice is dialect-dependent.
Chapter 7:ExtendingX-bar Theory 215
subject of the clause; be sure to identify even null (0) complementizers
and Ts with suffixes in them. State whether each clause is a finite clause
or a non-finite clause.
a) Stalin may think that Roosevelt is a fool.
b) Lenin believes the Tsar to be a power-hungry dictator.
c) Brezhnev had said for Andropov to leave.
d) Yeltsin saw Chemyenko holding the bag.
3. English That7
[Critical Thinking; Basic]
Discuss the status of the word that in each of the following two sentences.
Explain the differences between the two sentences. If you assign a different
category status to that in each sentence, explain why. Draw the tree
(use X-bar theory) for each of the sentences.
a) Robert thinks that students should eat asparagus.
b) Robert thinks that student should eat asparagus.
4. Trees
[Application of Skills; Basic to Intermediate]
Draw the trees for the following sentences. Use X-bar theory, show all CPs,
DPs, and TPs.
a) The very young child walked from school to the store.
b) Linguistics students like phonetics tutorials.
c) John paid a dollarfor a head of lettuce.
d) Teenagers drive rather quickly.
e) Martha said that Bill loved his Cheerios in the morning.
f) Eloise wants you to study a new language, [assume to =T]
g) ForMaurice to quarrelwith Joel frightened Maggie,
h) John's drum will always bother me.
5. Trees II
[Application of Skills; Basic toIntermediate]
1) Go back to chapter 3, general problem set 1, and draw the trees using
X-bar theory, including DPs.
2) Go back to chapter 3, general problem set 4, and draw the trees using
X-bar theory, including DPs, TPs, and CPs.
'Thanks to Eithne Guilfoyle for contributing this problem set.
216 The Base
3) Go back to chapter4, general problem set 1, and draw the trees using
X-bar theory, including DPs, TPs, and CPs.
Challenge Problem Sets
Challenge Problem Set 1: Hungarian DPs
[Data Analysis; Challenge]
In the text above, we argued that the structure of genitive constructions in
English looks like:
DP
DP D'
possessor D NP
I
N'
N
possessed
Consider the follow data from Hungarian. Does the possessor DP appear in
the same place as the English ones? Assume the determiners az and a
modify the possessed noun, not the possessor. Theending on the word hat
varies depending upon the possessor, this does not affect the answer to
this question. (Data from Szabolcsi 1994.)
a) az en kalapom
the I hat-1SG
"my hat"
b) a te kalapod
the you hat-2sG
"your hat"
Hungarian has another possessive construction, seen in (c).
c) Marinak a kalapja
Mary the hat-3SG
"Mary's hat"
Where is the possessor DP in (c)? Explain your answer.
Chapter7: Extending X-bar Theory 217
Challenge Problem Set 2: NPI Licensers
[Data Analysis and Critical Thinking; Challenge]
The adverb ever is a negative polarity item. Negative polarity items must
stand ina c-command relationship with a negative licenser. Assume that the
properties ofthe head uniquely determine theproperties ofa phrase. Explain
how the following sentences are an argument for the subject being a DP
rather than an NP:
a) No man has ever beaten the centaur.
b) *Some man has ever beaten the centaur.
c) *Everyman has ever beaten the centaur.
Challenge Problem Set 3: English Modals and Auxiliaries
[Data Analysis and Critical Thinking; Challenge]
In traditional grammar, two different kinds of T are found: modals and
auxiliaries. Modals include words like can, must, should, would, could,
may, will and in some dialects shall. Auxiliary verbs, by contrast, include
such words as have and be. In this book, we've treated both modals and
auxiliaries as T. An alternative is that only modals are really of category T,
and that auxiliaries are real verbs. Auxiliary and verb combinations are
actually a stacked set of VPs:
VP
I
V
V VP
be I
V
I
V
running
Construct an argument in favor ofthe idea that modals are ofcategoryT, but
auxiliaries are really verbs. Assume the following: You may have as many
Vcategories as you like, butthere is only one T in anytensed clause's tree.
Constraining X-bar
Theory: The Lexicon
0. INTRODUCTION
In chapters 6 and 7, we developed a very simple and general theory of
phrase structure: X-bar theory. Using only three rules, this theory accounts
for the distinction between adjuncts, complements, and specifiers.
It incorporates the more articulated view of sentence hierarchy required
by constituency tests, and it captures cross-categorial generalizations
(i.e., thefactthat allkinds ofphrases-NPs, VPs, APs,PPs, CPs,DPs, and TPs
- have the same basic properties). Most importantly, it allows us to draw
trees for most of the sentences of any language.
This said, there is a significant problem with X-bar theory: it also
generates sentences thatarenotacceptable orgrammatical. Take for example
the following pairs of grammatical and ungrammatical sentences:
1) a) Rosemary hates New York,
b) *Rosemary hates.
2) a) Jennie smiled.
b) *Jennie smiled the breadbox.
3) a) Traci gave the whale a jawbreaker.
b) Traci gave the whale.
c) Traci gave a jawbreaker.
Sentence (lb) should be perfectly acceptable (compare it to Rosemary ran).
X-bar theory says that complements are optional. Therefore, direct objects,
220 The Base
which arecomplements, should always beoptional. The opposite type offact
is seen in the pair in (2). X-bar theory optionally allows a complement.
So having a direct object here should be fine too. The same kind of effect
isseen in(3), where both the direct object andindirect object areobligatory contra
X-bar theory.
What seems tobe at work here is thatcertain verbs require objects and
others don't. It appears to be a property of the particular verb. Information
about the peculiar or particular properties of verbs is contained in our
mental dictionary or lexicon. In this chapter, we'll look at how we can use
thelexicon toconstrain X-bar theory, so thatit doesn't predict theexistence
of ungrammatical sentences.
1. Some Basic Terminology
Inchapter 2, wediscussed howdifferent verb types take a different number
ofarguments. For example, an intransitive verb like leave takes a single DP,
which is the subject. A transitive verb such as hit takes a DP subject
and a DP object. Below are the subcategories we came up with inchapter 2
(substituting DP for NP):
4)
Subcategory Example
V|DP_, (intransitive) Leave
vidp_ dpi (transitive type 1) Hit
V(dp_{dp/cph (transitive type 2) Ask
vidp_ dpdpi (ditransitive type1) Spare
vidp_dp ppi (ditransitivetype 2) Put
vidp_dp{dp/pp>i (ditransitivetype 3) Give
vidp_ dp{dp/pp/cp}] (ditransitive type4) Tell
In addition to these restrictions, we also find semantic restrictions on what
can appear in particular positions:
5) a) #My comb hates raisonettes.
b) #A bolt of lightning killed the rock.
There is something decidedly strange about these sentences. Combs can't
hate anything and rocks can't be killed. These semantic criteria are called
selectional restrictions.
Chapter 8:Constraining X-bar Theory: The Lexicon 221
In the next section, we'll look at the theory of thematic relations,
which is a particular way of representing selectional and subcategorizational
restrictions.
2. Thematic Relations and Theta Roles
One way of encoding selectional restrictions is through the use of what are
called thematic relations. These are particular semantic terms that are used
to describe the role that the argument plays with respect to the predicate.
This section describes some common thematic relations (this list is by
no means exhaustive, and the particular definitions are not universally
accepted).
The initiator or doer of an action is called the agent. In the following
sentences, Ryan and Michael are agents.
6) a) Ryan hit Andrew.
b) Michael accidentally broke the glass.
Agents are most frequently subjects, but they can also appear in other
positions.
Arguments that feel or perceive events are called experiences.
Experiencers can appear in a number of argument positions including
subject and object:
7) a) Leah likes cookies.
b) Lorenzo saw the eclipse.
c) Syntax frightens Kenna.
Entities that undergo actions, are moved, experienced or perceived are
called themes.
8) a) Alyssa kept hersyntax book.
b) The arrow hit Ben.
c) The syntactician hates phonology.
Theentity towards which motion takes placeis calleda goal. Goals may
involve abstract motion:
9) a) Doug went to Chicago.
b) Dave was given the pina colada mix.
There is a special kind of goal called recipient. Recipients only occur with
verbs that denote a change of possession:
10) a) Mikaela gave Jessica the book.
222 The Base
b) Daniel received a scolding from Hanna.
The opposite of a goal is the source. This is the entity from which a motion
takes place:
11) a) Bob gave Steve the syntax assignment.
b) Stacy came directlyfrom sociolinguistics class.
The place where the action occurs is called the location:
12) a) Andrew is in Tucson'sfinestapartment.
b) We're all at school.
The objectwith which an action is performed is called the instrument:
13) a) Chris hacked the computer apart with an axe.
b) This key will open the door to the linguisticsbuilding.
Finally, the one for whose benefit an event took place is called the
beneficiary:
14) a) He bought these flowers for Aaron.
b) She cooked Matt dinner.
You now have enough information to try General Problem Sets 1&2
Notice that any given DP can have more than one thematic relation.
In the following sentence, the DPJason bears the thematic relations ofagent
and source (at the very least).
15) Jason gave the books to Anna.
There is not a one-to-one relationship between thematic relations and
arguments. However, linguists have a special construct called a theta role (or
0 role) that does map one-to-one with arguments. Theta rolesare bundles of
thematic relations that clusteron oneargument. In (15) above, Jason gets two
thematic relations (agent and source), but only one theta role (the one that
contains the agent and source thematic relations). Somewhat confusingly,
syntacticians often refer to particular theta roles by the most prominent
thematic relation that they contain. So you might hear a syntactician refer to
the "agent theta role" of [DPJason]. Strictly speaking, this is incorrect: Agent
refers to a thematic relation, whereas the theta role is a bundle of thematic
relations. But the practice is common, so we'll do it here. Remember,
thematic relations are things like agent, theme, goal, etc., but theta roles
are bundles of thematic relations assigned to a particular argument.
Let's now see how we can use these theta roles to represent
the argument structure of a verb. Take a ditransitive verb like place.
Chapter 8: Constraining X-barTheory: The Lexicon 223
Place requires three arguments, a subject that must be an agent (the placer),
a direct object, which represents the theme (the thing being placed),
and an indirect object, which represents a location or goal (the the thing
on which the theme is being placed). Any variation from this results
in ungrammaticality:
16) a) John placed the flute on the table.
b) *placed the flute on the table.
c) *Johnplaced on the table.
d) *John placed the flute.
e) *John placed theflute theviolin onthetable.1
f) The rock placed the sky with the fork.
g) *John placed the flute the table.
Examples (16a-e) show that either having too many or too few arguments
results in ungrammaticality. Example (16f) shows that using DPs with the
wrong theta roles does the same (the rock can't be an agent; the sky can't
be a theme - it can't be given to anyone; and with thefork is an instrument,
not a goal). (16g) shows us that the category of the argument is important
(this we already knew from chapter 2), the goal argument of the verb place
must be a PP. It appears as if the verb place requires three arguments,
which bear precisely the theta roles ofagent (DP), theme (DP), and goal (PP).
We represent this formally interms ofwhat iscalled a theta grid.2
17) © ®
\ V
place ~
This grid consists of several parts. First of all, we have the name of the
predicate (A). Next, for each argument that the predicate requires, there is
a column (with two rows). Each of these columns represents a theta role.
Notice that a column can have more than one thematic relation in it (but only
one theta role). The number of columns corresponds exactly to the number
of arguments the predicate requires. The first row (B) tells you the thematic
Source/Agent
DP
Theme
DP
Goal
PP
i i k
1This sentence would be OK if there were a conjunction between the flute and the
violin. What does this tell us about what conjunction does to theta roles?
2There are many ways to formalize theta grids, but I adopt here the indexing box
method that Haegeman (1994) uses, since it seems to be the most transparent.
224 The Base
relations and the categories associated with each of these theta roles.
The second row (C), gives you what are called indices (singular: index)
for each theta role. These are not the same as the indices in binding theory.
When a predicate appears in an actual sentence, we mark the DP bearing
the particular theta role with that index. Applying our grid to sentence (18),
we get the following indexed sentence:
18) Johns placed [theflutejj [onthe table]k.
The | index maps the agent theta role to John. The }index maps the theme
theta role to the book, etc.
Theta roles actually come in two types. The first is the external theta role
(D). This is the one assigned to the subject. External theta roles are usually
indicated by underlining the name of the theta role in the theta grid.
Theother kind are internal theta roles (E). Theseare the theta roles assigned
to the object and indirect object. There is a semantic reason for the distinction
between internal and external theta roles (see Marantz 1984 for extensive
discussion), but we will leave that issue aside here. We will have use for
the external/internal distinction in chapter 10, when we do DP movement.
Fornow, however, youshouldsimply indicate whichargument isthesubject
by underlining its name.
If you look carefully at the theta grid in (17) you'll notice that it only
contains a specifier (subject) and complements (direct object and indirect
object). There are no adjuncts listed in the theta grid. Adjuncts seem
to be entirely optional:
19) a) John put the book on the table(with a pair of tongs). instrument
b) (In the classroom)John put the book on the table. location
Thiscorrespondsto our observation in chapter6, that you canhave as many
or as few adjuncts as you like,but the number ofcomplements and specifiers
are more restricted. Adjuncts are never arguments, and they never appear in
theta grids.
You can nowtryGeneral Problem Sets 3 &4 (you may wish toreview section 4
below before trying these)
Up until now, we have been representing our grammar solely through
the mechanism ofrules (phrasestructure, then X-bar rules). In order to stop
X-bar rules from overgenerating, we need a constraint. Constraints are like
filters. They take the output of rules, and throw away any that don't meet
the constraint's requirements. In essence, we are going to allow the X-bar
rules to wildly overgenerate, and produce ungrammatical sentences. Those
sentences, however, will be thrown out by our constraint. The constraint
Chapter 8:Constraining X-barTheory:The Lexicon 225
we are going to use is called the Theta Criterion. The theta criterion ensures
that there is a strict match between the number and types of arguments
in a sentence and the theta grid.
20) The Theta Criterion
a) Eachargument is assigned one and only one theta role.
b) Each theta role is assigned to one and only one argument.
This constraint requires that there is a strict one-to-one match between
argument DPs and theta roles. You can't have more arguments than you
have theta roles, and you can't have more theta roles than you have DPs.
Furthermore, since theta roles express particular thematic relations,
thearguments will have to be of appropriate semantic types for the sentence
to pass the constraint.
Let's look at some examples to see how this works. Consider the verb
love. It has the theta grid given in (21). I haven't written in the indices here,
because we'll add them when we compare the grid to a particular sentence.
21) love
Experiencer
DP
Theme
DP
When a sentence containing the predicate love is produced, we apply indices
to each of the arguments, and match those arguments to theta roles
in the grid. The sentence in (22) is grammatical with the correct number
of arguments. It is matched to the theta grid in (23). There is a one-to-one
matching between arguments and theta roles. So the theta criterion is
satisfied, and the sentence is allowed to pass through the filter and surface.
22) Megani loves Kevinj.
23) love
Experiencer
DP
Theme
DP
Contrast this with the ungrammatical sentence in (24):
24) *Megan; loves.
Thissentencelacksa theme argument, as seen in the following theta grid:
226
25) love
Experiencer
DP
The Base
Theme
DP
The theme theta role is not assigned to an argument (there is no index in its
lower box). This violates the second condition of the theta criterion: Every
theta role is assigned to an argument. There is not a one-to-one matching
of the theta roles to the arguments in this sentence. Since the theta criterion
is violated, the sentence is filtered out (marked as ungrammatical). Notice,
our X-bar rules can generate this sentence; it is ruled as ungrammatical
by our constraint.
The next sentence shows the opposite problem: A sentence with too
many arguments.
26) *Meganj lovesJasonj Kevin
27) love
k-
Experiencer
DP
Theme
DP
i i k
Here, the argument Kevin doesn't get a theta role. There are only two theta
roles to be assigned, but there are three arguments. This violates the first part
of the theta criterion: the requirement that every argument have a theta role.
Again, the theta criterion filters out this sentence as ungrammatical.
To summarize, we can constrain the output of the X-bar rules using a
semantic tool: theta roles. The theta criterion is a constraint or filter that rules
out otherwise well-formed sentences. The theta criterion requires that there
is a strict one-to-one matching between the number and kind of theta roles
and the number and kind of arguments.
You now have enough information totry General Problem Set5 and Challenge
Problem Set 1
3. The Lexicon
Let's take a step back from these details and look at the big picture. We have
developed a model of grammar where we have three simple rules (the X-bar
rules) that can generate a hierarchical constituent structure. These rules
are constrained by the theta criterion, which uses the semantic notion of
theta roles. Recall that our theory ofsyntax is meant to be a cognitive theory,
Chapter 8:Constraining X-barTheory: The Lexicon 227
so let's consider the question of where these rules and these theta roles
are stored in the mind. Chomsky proposes that the part of the mind devoted
to language is essentially divided into two parts. One part, which he calls
the computational component, contains all the rules and constraints.
This part of the mind does the work of building sentences and filtering out
any ill-formed ones. The computational component can't work in a vacuum,
however. It needs access to information about theta roles and the like.
Chomsky claims that this information is stored in the lexicon, the other part
of the human language faculty. The lexicon is your mental dictionary or list
of words (and their properties). If you think about it, this is the obvious place
for theta grids to be stored. Which theta role is assigned to which argument
is a property of each predicate. It is information that must be associated
with that predicate and that predicate only. The obvious place to store
information about particular words (or more properly lexical items) is in
the lexicon.
The lexicon contains all the irregular and memorized parts of language.
Each lexical entry (dictionary entry) must contain at least the following
information):
• the meaning of the word
• the syntactic category of the word (N, V, A, P, T, C, etc.)
• the pronunciation of the word
• exceptional information of all kinds (such as morphological
irregularities)
• the theta grid (argument structure).
When you learn a new word, you memorize all this information.
On an abstract level we can diagram the grammatical system as looking
something like:
28)
The Lexicon
(theta grids)
The Computational Component
X-bar rules
I
Theta Criterion
~J
Output
The lexicon feeds into the computational component, which then combines
words and generates sentences. The fact that lexical information affects the
form of the sentence is formalized in what we call the Projection Principle:
228 The Base
29) The Projection Principle
Lexical information (such as theta roles) is syntactically represented at
all levels.
4. EXPLETIVES AND THE EXTENDED PROJECTION PRINCIPLE
Beforeleaving the topic of the lexicon, I'd like to point out two special classes
of predicates. Consider first the following "weather" verbs. These predicates
don't seem to assign any theta roles:
30) a) It rained.
b) It snowed.
c) It hailed.
What theta role does the pronoun itget in these sentences? If you are having
a problem figuring this out, ask yourself what it refers to in the above
sentences. It appears as if it doesn't refer to anything. In syntax, we refer to
pronouns like this as expletive or pleonasticpronouns. These pronouns don't
get a theta role (which of course is a violation of the theta criterion - a point
we will return to below). The theta grid for weather verbs is empty.
They don't assign any theta roles.
There is another class of predicates that take expletive pronouns. These
are predicates that optionally take a CP subject:
31) [CP That Billloves chocolate] is likely.
The predicate is likely assigns one theta role. It takes one argument (the
clause). (We will notate clausal arguments with the theta role proposition.)
32) is likely
Proposition
CP
You'll note that in (32) the theta role is not underlined. This is because the
clause bearing the theta role of proposition is a complement. This can be seen
in the following example:
33) It is likely that Billlikes chocolate.
In this sentence, we again have an expletive it, which gets no theta role.
In order to maintain the theta criterion, we need to account for these
expletive DPs without theta roles. Expletive pronouns usually appear
Chapter 8: Constraining X-bar Theory: The Lexicon 229
in subject position. When it appears in other positions, it usually bears
a theta role:
34) a) I love it. (it is a theme)
b) I put a book on it. (itis a goal or location)
Expletives seem to appear where there is no theta marked DP (or CP)
that fills the subject position. This is encoded in a revised version
of the Projection Principle: The Extended Projection Principle (EPP):
35) Extended Projection Principle (EPP)
All clauses must have subjects, (i.e. the specifier of TP must be filled by
a DP or CP)
The EPP works like the theta criterion, it is a constraint on the output of
the X-bar rules. It requires that every sentence have a subject. Next, we
must account for the fact that expletives violate the theta criterion.
One way of doing this is by claiming that expletives are not generated by
the X-bar rules. Instead they are inserted by a special expletive insertion rule:
36) Expletive insertion rule
Insert an expletive pronoun into the specifier of TP.
This rule applies when there is no other subject. If there is no theta marked
subject and no expletive subject, then the EPP will filter the sentence out.
The way in which we get around the theta criteria is by ordering the expletive
insertion rule after the theta criterion has applied.
The Lexicon
(theta grids)
The Computational Component
X-bar rules
Theta criterion (constraint)
I
Expletive insertion rule
EPP (constraint)
Grammaticality Judgments
Since expletives are inserted after the theta criterion has applied, they can't
be filtered out by it.
The model we've drawn here is very preliminary. In the next chapter, we
will introduce a new kind of rule (the transformation - of which expletive
insertion is a very special case) that will cause us to significantly revise
this diagram.
230 The Base
You nowhaveenough information to try Challenge Problem Sets2, 3, 4, & 5
Two Kinds of If
There are two it pronouns in English. One is the expletive found with
weather verbs. The other is the neuter pronoun itfound in sentences like:
i) It bit me on the leg.
If you contrast the it in (i) with the ones in the weather verbs, you'll see
that the it in (i) does take a theta role (agent) and does refer to something
(probably an insect or some other animal). Not every sentence with an it
involves an expletive.
5. Summary
We started this chapter off with the observation that while X-bar rules
capture important facts about constituency and cross-categorial
generalizations, they overgenerate (that is they generate ungrammatical
sentences). One way of constraining X-bar theory is by invoking lexical
restrictions on sentences, such that particular predicates have specific
argument structures, in the form of theta grids. The theta criterion rules out
any sentence where the number and type of arguments don't match up one
to one with the number and type of theta roles in the theta grid.
We also looked at one apparent exception to the theta criterion: theta
role-less expletive pronouns. These pronouns only show up when there is
no other subject, and are forced by the EPP. They escape the theta criterion
by being inserted after the theta criterion has filtered out the X-bar rules.
By using lexical information (like theta roles) we're able to stop the X-bar
rules from generating sentences that are ungrammatical. Unfortunately,
as we'll see in the next chapter, there are also many sentences that the X-bar
rules cannot generate. In order to account for these, we'll introduce a further
theoretical tool: the movement rule.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Selectional Restrictions: Semantic restrictions on arguments.
ii) Thematic Relations: Semantic relations between a predicate and
an argument- used as a means of encoding selectional restrictions.
Chapter 8: Constraining X-barTheory: The Lexicon 231
iii) Agent: The doer of an action (under some definitions must be
capable of volition).
iv) Experiencer: The argument that perceives or experiences an event or
state.
v) Theme: The element that is perceived, experienced or undergoing
the action or change of state
vi) Goal:The end point of a movement.
vii) Recipient: A special kind of goal, found with verbs of possession
(e.g.,give).
viii) Source: The starting point of a movement.
ix) Location:The place an action or state occurs.
x) Instrument: A tool with which an action is performed.
xi) Beneficiary: The entity for whosebenefit the action is performed.
xii) Proposition: The thematic relation assigned to clauses.
xiii) Theta Role: A bundle of thematic relations associated with a
particular argument (DPsor CPs).
xiv) Theta Grid: The schematic representation of the argument structure
of a predicate, where the theta roles are listed.
xv) External Theta Role: The theta role associated with subject DPs or
CPs.
xvi) Internal Theta Role: The theta role associated with objects or indirect
objects.
xvii) The Theta Criterion:
a) Each argument is assigned oneand onlyone theta role.
b) Each theta role is assigned to one and only one argument.
xviii) Lexical Item: Another way of saying "word." A lexical item is
an entry in the mental dictionary.
xix) The Projection Principle: Lexical information (like theta roles) is
syntactically represented at all levels.
xx) Expletive (or Pleonastic) Pronouns: A pronoun (usually it or there)
without a theta role. Usually found in subject position.
xxi) Extended Projection Principle (EPP): Allclauses must have subjects.
Lexical information is syntactically represented.
232
xxii)
xxiii)
xxiv)
xxv)
The Base
Expletive Insertion: Insert an expletive pronoun into the specifier
ofTP.
The Lexicon: The mental dictionary or list of words. Contains
all irregular and memorized information about language, including
the argument structure (theta grid) of predicates.
The Computational Component: The combinatorial, rule-based, part
of the mind. Where the rules and filters are found.
The Model:
The Lexicon
(theta grids)
The Computational Component
X-bar rules
I
Theta criterion (constraint)
Expletive insertion
EPP (constraint)
Grammaticality Judgments
Further Reading
Gruber, Jeffrey (1965) Studies in Lexical Relations. Ph.D. dissertation, MIT.
Haegeman, Liliane (1994) introduction to Government and Binding Theory.
Oxford: Blackwell.
Marantz,Alec(1984) On the Nature ofGrammatical Relations. Cambridge:MIT
Press.
Williams, Edwin (1980) Predication. Linguistic Inquiry 11,203-38.
Williams, Edwin (1994) Thematic Structure in Syntax. Cambridge: MITPress.
msse^ss aeAa: B
General Problem Sets
1. Identifying Thematic Relations
[Data Analysis and Application of Skills; Basic]
Part 1: Identify the thematic relations associated with each DP or embedded
CP in the following sentences. Each DP or CP may have more than one
thematic relation associated with it.
Chapter 8:Constraining X-bar Theory: The Lexicon 233
a) Shannon sent Dan an email.
b) Jerid thinks that Sumayya cooked some beef waffles for him.
c) Stacy hit a baseball to Yosuke.
d) Jaime danced a jig.
e) Yuko rubbed the pizza with a garlic clove.
f) It's foggy in San Francisco.
Part 2: Draw the trees for (b-f), use CPs, DPs and TPs.
2. Warlpiri3
[Data Analysis; Basic]
Consider the following data from Warlpiri:
a) Lungkarda ka ngulya-ngka nguna-mi.
bluetongue aux burrow-A lie-NON.PAST
"The bluetongue skink is lying in the burrow."
b) Nantuwu ka karru-kurra pamka-mi.
horse aux creek-B run-NON.PAST
"The horse is running to the creek."
c) Karli ka pirli-ngirli wanti-mi.
boomerang aux stone-C fall-NON.PAST
"The boomerang is falling from the stone."
d) Kurdu-ngku ka-jana pirli yurutu-wana yirra-rni.
child-D aux stone road-E put.NON.PAST
"The child is putting stones along the road."
What is the meaning of each of the affixes (suffixes) glossed with -A, -B, -C,
-D,and -E. Can you relate these suffixes to thematicrelations?Which ones?
3. Theta Grids
[Data Analysis; Basic]
For each of the sentences below identify each of the predicates (including
non-verbal predicates like is likely). Provide the theta grid for each.
Don'tforget: include only arguments in the theta grid; DPs and PPs that are
adjuncts are not included. Index each DP, PP, CP argument with the theta
role it takes. Assume that there are two different verbs give (each with their
own theta grids) to account for (c) and (d); two different verbs eat (each with
theirown theta grids for(e) and (f); and two different verbs ask for(i) and Q).
The data for this problem set comes from KenHale via BarbBrunson.
234 The Base
a) The stodgy professor left with his teaching assistant.
b) I played a tune on my iPod.
c) Molly gave Calvin a kiss.
d) Mercedes gave a test to the students in the lecture hall.
e) Pangur ate a cat treat.
f) Susan ate yesterday at the restaurant.
g) Gwen saw a fire truck.
h) Gwen looked at a fire truck.
i) Michael asked a question.
j) Adam asked if Hyacinth likes pineapples.
k) It is sunny in the dining room.
I) Ifeel it is unfortunate that television is so vulgar these days.
m) That Angus hates sushi is mysterious.
4. Sinhala4
[Data Analysis; Basic/Intermediate]
Two forms of the Sinhala verb appear in the data below and are identified
in the glosses as A or B. (Data fromGair 1970.)
1) Provide a complete theta grid for each of the verbs in the following data.
Be sure to primarily look at the second line of each piece of data,
not the English translation.
2) Using indexes identifywhat theta role is assigned to each DP.
3) Discuss briefly (no more than 2 sentences) what kind of DP the suffix -£
attaches to.
4) What is the difference between mama and ma^l (Hint: the answer to
this question is related to the answer to question (3).)
5) In terms of theta roles, what is the difference between the A and the B
verb forms?
a) Mams kawi kianawa.
I poetry tell-A
"I recite poetry."
b) Mat? kawi kiswenawa.
I poetry tell-B
"I started reciting poetry (despite myself)."
This problem is loosely based on one given to me by Barb Brunson. However,
the data and questions havebeen altered. Thedata in this version of the problem set
is taken directly from Gair, with some minor modifications to the glosses.
Chapter 8: Constraining X-barTheory: The Lexicon 235
c) Lamea kataawa ahanawa.
child story hear-A
"The child listens to the story."
d) Lameata kataawa aehenawa.
child story hear-B
"The child hears the story."
e) Mams natanawa.
I dance-A
"I dance."
f) Mats naetaenawa.
I dance-B
"I dance (I can't help but do so)."
g) Haema irida ma mama kolamba yanawa.
every Sunday EMPH I Columbo go-A
"Every Sunday I deliberately go to Colombo."
h) Haema irida ma mats kolamba yaewenawa.
every Sunday EMPH I Columbo go-B
"Every Sunday I experience going to Colombo."
i) Malli nitarama andanawa.
brother always cries-A
"Brother always cries."
J) Mallita nitarama aenrfanawa.
brother always cries-B
"Brother always bursts out crying without control."
k) Mama unta baninawa.
I them scold-A
"I deliberately scold them."
1) Mat? unt9 baenenawa.
I them scold-B
"I experienced scolding them."
236 The Base
m) Api]o pansala peenawa.
we temple see-B
"We saw the temple."
5. Theta Criterion
[Data Analysis; Intermediate]
Show how each of the following sentences are violations of the theta
criterion. Use theta grids to explain your answers.
a) *Rosemary hates.
b) *Jennie smiled the breadbox.
c) *Traci gave the whale.
d) *Traci gave a jawbreaker.
e) *placed the flute on the table.
f) *John placed on the table.
g) *John placed the flute.
h) *John placed the flute the violin on the table,
i) *The rock placed the sky with the fork,
j) \John placed the flute the table.
Challenge Problem Sets
Challenge Problem Set 1: Irish and the Theta-Criterion
[Data Analysis and Application of Skills; Challenge]
What problems do each of the following examples give for the theta
criterion? (As a starting point, it may help to draw the theta grid for each verb
and show what DP gets what role.) Please, not more than 3-4 sentences
of discussion per example.
a) an fear a bhfaca me e
the man who saw I him
"the man who I saw"
b) Rinceamar.
Dance.1pl
"We danced."
c) Ba-mhaith Horn an teach a thogail.
coND-good with-me the house its building
"I would like to build the house."
Chapter 8:Constraining X-bar Theory: The Lexicon 237
Challenge Problem Set 2: Object Expletives
[Critical Thinking; Challenge]
In the text above, it was observed that theta-role-less expletives primarily
appear in subject position. Consider the following sentence. Is it here
an expletive?
I hate it that you're always late.
How could you tell?
Challenge Problem Set 3: Passives
[Data Analysis; Challenge]
Part 1: Write up the theta grids for the verbs in the following sentences.
Assume there are two verbs give (give! is seen in (d), give2 in (e)).
a) John bit the apple.
b) Susan forgave Louis.
c) The jockey rides the horse.
d) Phillip gave the medal to the soldier.
e) Phillip gave the soldier the medal.
Part 2: English has a suffix -en, which when attached to verbs changes the
structure of the sentence associated with them. This is called the passive
morpheme. The following sentences are the passive equivalents
of the sentences in part 1. The bracketed PPs starting with by are optional.
f) The apple was bitten (by John).
g) Louis was forgiven (by Susan).
h) The horse was ridden (by the jockey).
i) The medal was given to the soldier (by Phillip).
j) The soldier was given the medal (by Phillip).
Describe in your own words what the -en passive suffix does to
the theta grids of verbs. Pay careful attention to the last two examples,
and to the optionality of the by-phrases.
Challenge Problem Set 4: Hiaki -wa5
[DataAnalysis and Critical Thinking; Challenge]
Part 1: Consider the function of the suffix -wa in Hiaki (also known as Yaqui),
a language spoken in Southern Arizona and Mexico. Look carefully at the
data below, and figure out what effect this suffix has on the theta grids of
5Thanks to Heidi Harley for contributing this problem set.
238 The Base
Hiaki verbs. What English phenomenon is this similar to? (Data from Escalante
1990 and Jelinek and Escalante 2003.)
(Notes: Sometimes when -wa attaches to a verb, the form of the root changes
(usually lei becomes l\l). This is a morphophonological phenomenon that you don't
need to worry about, acc refers to accusative case, inst means instrument, and perf
means perfective aspect (aspect plays no role in the answer to this problem). There is
no nominative suffix in Hiaki.)
a) Peo Huan-ta chochon-ak.
Pete John-Ace punch-PERF
"Pete punched John."
a') Huan chochon-wa-k.
John punch-WA- perf
"John was punched."
b) 'Ume uusi-m uka kuchu-ta kuchi'i-m-mea bwa'a-ka.
the children-PL the-ACC fish-ACC knife-PL-iNST eat- perf
"The children ate the fish with knives."
b') 'U kuchu kuchi'i-m-mea bwa'a-wa-k.
the fish knife-PL-iNST eat-WA-PERF
"The fish was eaten with knives."
c) Peo bwiika.
Pete sing
"Pete is singing."
c') Bwiik-wa.
sing-WA
"Singing is happening." or "There is singing going on." or "Someone is
singing."
Part 2: Not all verbs allow -wa. Consider the following pairs of sentences
that show verbs that don't allow -wa. In terms of theta grids, what do
these sentences have in common with each other that differentiates them
from the ones that allow -wa (above in part 1).
a) 'U wikia chukte.
the rope come.loose
"The rope is coming loose."
a') *Chukti-wa.
come.loose-WA
"Coming loose is happening." or "There is coming loose going on." or
"Something is coming loose."
Chapter 8:Constraining X-bar Theory: The Lexicon 239
b) 'U kaaro nasonte.
the car damage
"The car is damaged."
b') *Nasonti-wa.
damage-WA
"Damage is happening." or "There is damage going on." or "Something is
getting damaged."
c) 'U kari veete-k.
The house burn-PERF
"The house burned."
c*) *Veeti-wa-k.
Burn-WA-PERF
"Burning happened." or "There was burning going on." or "Something is
getting burned."
d) 'U vachi bwase'e.
The corn cook
"The corn is cooking."
d') *Bwase'i-wa.
cook-WA
"Cooking happened." or "There was cooking going on." or "Something is
being cooked."
Challenge Problem Set 5: Antipassives
[Data Analysis and Critical Thinking; Challenge]
In many languages there is an operation that changes the theta grid of
certain verbs, this operation is called the antipassive.
Part 1: Here is some data from Inupiaq, an Inuit language of Canada and
Alaska. Explain what adding the antipassive morpheme does to the theta
grid of the verb. Verbs in Inupiaq agree with both their subjects and their
objects. 3subj-3obj means that the verb agrees with both a 3rd person
subject and a 3rd person object. 3 means that the verb only agrees
with a 3rd person subject. (Data from Seiler 1978.)
a) Arjuti-m umiaq qinig-aa tirrag-mi. Active
man-ERG boat-ABS see-3suBJ.30BJ beach-at
"The man sees the boat at the beach."
240 The Base
b) Arjun (umiag-mik) qiniq-tuq tirrag-mi. Antipassive
man-ABS boat-iNST see-3 beach-at
"The man sees (with a boat) at the beach."
Part 2; The following is some data from English. This might also be called
an antipassive construction. How is it similar or different from the Inupiaq
antipassive?
c) Iate a basket of apples.
d) late.
00
03
d
Movement
Head-to-Head
Movement
0. Introduction
Consider the relation between a verb and its object: According to X-bar
theory, an object is the complement to V (sister to V, daughter of V). This
means that no specifier or adjunct can intervene between the complement
and the head (if they did, the object would no longer be a complement).
The following sentence is from Modern Irish Gaelic, this is a VerbSubject-Object
(VSO)word order language:
1) Phog Maire an lucharach^n.
Kissed Mary the leprechaun
"Mary kissed the leprechaun."
In this sentence, the subject (a specifier) intervenes between the verb and
the object; this sentence cannot be generated by X-bar theory. (Try to draw
a tree where the specifier intervenes between the head and the complement
- it's impossible.)
Now consider the following sentence from French:
2) Je mange souvent des pommes.
I eat often of.the apples
"I often eat apples."
Souvent "often" intervenes between the verb and the object. If souvent
is an adjunct it is appearing between a head and its complement.
X-bar theory can't draw the tree for this one either.
244 Movement
In sum, X-bar theory undergenerates, it does not produce all the possible
grammatical sentences in a language.
Although based on very different problems than the ones in (1) and (2),
Chomsky (1957) observed that a phrase structure grammar (such as X-bar
theory) cannot generate all the sentences of a language. He proposed that
what was needed was a set of rules that change the structure (in very limited
ways) generated by phrase structure rules. These rules are called
transformational rules. Transformations take the output of X-bar rules
(and other transformations) and change them into different trees.
The model of grammar that we are suggesting here takes the following
form. You should read this like a flow chart. The derivation of a sentence
starts at the top, and what comes out at the bottom is what you say.
3)
The Lexicon
The Computational Component
> X-bar rules 'the base'
D-structure
(constrained by the Theta Criterion)
Transformational rules
> <
S-structure
(constrained by EPP)
T
Grammaticality Judgments
X-bar theory and the lexicon conspire together to generate trees.
This conspiracy is called the base. The result of this tree generation is a level
we call D-structure (this used to be called Deep Structure, but for reasons
that need not concern us here, the name has changed to D-structure).
You will never pronounce a D-structure. D-structure is also sometimes
called the underlying form or underlying representation (and is similar in
Chapter 9: Head-to-Head Movement 245
many ways to the underlying form found in phonology). The theta criterion
filters out ungrammatical sentences at D-structure.
D-structure is then subject to the transformational rules. These
transformational rules can move words around in the sentence. We've
actually already seen two of these transformational rules. In Chapter 7,
we looked briefly atT to C movement in subject-aux inversion constructions,
and affix lowering, which gets inflectional suffixes to lower to their verb.
(In this chapter, we're going to look in more detail at these two rules.)
The output of a transformational rule is called the S-structure of a sentence.
The S-structure is filtered by the EPP, which ensures that the sentence
has a subject. What are left are grammatical sentences.
In the version of Chomskyan grammar we are considering, we will
look at two different kinds of transformations: movement rules and insertion
rules. Movement rules move things around in the sentence. Insertion
rules put something new into the sentence. This chapter is about one kind
of movement rule: the rules that move one head into another, called
head-to-head movement. These transformational rules will allow us to
generate sentences like (1) and (2) above. X-bar theory by itself cannot
produce these structures.
Generative Power
Before we go any further and look at an example of a transformation,
consider the power of this type of rule. A transformation is a rule that can
change the trees built by X-bartheory. Ifyou think about it, you'll see that
such a device is extremely powerful; in principle it could do anything.
Forexample you could write a changingrule that turns all sentences that
have the word "red" in them to sentences with SOV order.
i) [TP...red...]=>[TI>S[OV]]
This rule would take a sentence like (ii) and change it into a sentence like
(iii):
ii) The red book bores me.
iii) The red book me bores.
Similarly we could allow X-bar theory to generate sentences where the
work "snookums" appears after every word, then have a transformation
that deletes all instances of "snookums" (iv). (v) shows the D-structure
of such a sentence, (vi) would be the S-structure (output) of the rule.
iv) "snookums" => 0
v) I snookums built snookums the snookums house snookums.
246 Movement
vi) I built the house.
These are crazy rules. No language has a rule of these types. However, in
principle, there is no reason that rules of this kind couldn't exist. We thus
need to restrict the power of transformational rules. We do this two ways:
vii) Rules must have a motivation. Frequently these motivations are output
constraints, like the EPP that we saw in the last chapter, or
morphophonological, like the ones we will propose in this chapter.
viii) Not only are rules motivated by output constraints, they are
restricted by them. You cannot write a rule that will create a violation of
an output constraint.
As we go along we will consider specific ways to constrain
transformational rules so that they don't overgenerate.
1. Verb Movement (V -> T)
1.1 French
Let's return now to the problems we raised in the introduction to this
chapter. Let's start with the sentence from French:
4) Je mange souvent des pommes.
I eat often of.the apples
"I often eat apples."
In this sentence, an adjunct surprisingly appears between the head of VP
and its complement. Compare this sentence to the English sentence in (5):
5) I often eat apples.
In the English sentence, the adjunct does not intervene between the verb
and the complement. The tree for (5) would look like (6).
6)
Chapter 9: Head-to-Head Movement
CP
often V
eat
247
Notice thefollowing thingaboutthisstructure. There is a head position that
intervenesbetween the subjectDP and the adverb often: this is the T position.
T, you will recall, gives its inflection to the verb or surfaces as an auxiliary.
Notice that in French (4), the thing that appears between the subject
and the adverb is not T, but the tensed main verb.
Keeping thisidea in theback ofyourmind now consider the following
chart, which shows the relative placement of the major constituents
of a French sentence with a tensed main verb (b), and English sentence with
a tensed verb (a), and bothlanguages withauxiliary constructions (cand d):
7)
a) I T often eat apples
b) Je mange souvent des pommes
c) I Have often eaten apples
d) r Ai souvent mange des pommes
There are several things to observe about this chart. Recall from chapter 2,
that auxiliaries are instances of the category T; as such, V adjuncts
arepredicted to invariably follow them. This seems to be thecase (c and d).
What is striking about the above chart is that tensed main verbs in French
also seem to occupy this slot, whereas in English, they follow the adverb.
How can we account for this alternation? Let's assume that the form which
meets X-bar theory (and happens to be identical to the English tree in (6))
is what is generated in both French and English. The difference between
248 Movement
the two is that French has a special extra rule which moves its verbs out of
the VP. More precisely, it moves them into the slot associated with T.
This is the transformational rule we will call V -» T; it is also known as
verb movementor verb raising. This rule is informally stated in (8):
8) V —>T movement: Move the head V to the head T.
Before looking at an example, consider for a moment why this rule might
apply. Much like the rule of affix lowering we introduced for English in
chapter 7, this rule exists to get an inflectional affix on the verb. In fact, let's
go one step further, let's claim that affix lowering and verb raising are really
the same operation. Notice that they are in complementary distribution a
language either has one or the other.Thedifference between a verb raising
language (French) and an affix lowering language (like English) might
simply be one of a parameter. All languages have some version of this rule,
some set the parameter to raise the verb to T, others set it to lower the T
to the V.
9) Verb movement parameter: Verbs raise to T or T lowers to V.
This provides a simple account ofthedifference betweenEnglish and French
adverbial placement.
Now, let's do a derivation for the French sentence Je mange souvent
des pommes. The first step in the derivation is to build an X-bar structure,
and insert all the words. This gives us the D-structure of the sentence:
souvent V
mange
des pommes
Chapter 9: Head-to-Head Movement 249
Notice that this D-structure is not a grammatical sentence of French (yet).
In factit has exactly the same word order as the Englishsentence in (5).
The next step in the derivation is to apply the transformation of Verb
Movement. One typical way of representing a movement transformation
is to draw a tree with an arrow.
des pommes
This results in the correct S-structure string:
12) Je mangej souvent f, des pommes.
Yet at the same time, we can maintain X-bar theory. The f, here stands for
"trace" and sits at the D-structure position of the verb.
Consider now the related derivation for the English sentence Heoften eats
apples. The D-structure is thesame, except withEnglish words:
250 Movement
AdvP V
A ^
often V DP
eat A
apples
Since English is parameterized for affix lowering rather than verb raising,
the inverse movement to Frenchapplies:
This results in the grammatical S-Structure:
15) He often eats apples.
Chapter 9: Head-to-Head Movement 251
What we have seen so far is a rather technical solution to a relatively
small problem. Now, I'm going to show you that this solution
can be extended. Recall our chart with adverb above in (7). Consider now
the identical chart with negatives:
16)
a)
b)
c)
d)
I
Je
I
Je
do
ne-mange
have
n'ai
not
pas
not
pas
eat
eaten
mang£
Apples
des pommes
Apples
des pommes
Ignore for the moment the French morpheme ne-, which is optional
in spoken French in any case. Concentrate instead on the relative positioning
of the negatives pas and not and the verbs. The situation is the same as with
the adverb often. All auxiliaries in both languages precede negation, as does
themain verb in French. But inEnglish, themain verb follows thenegation.1
We can apply the same solution to this word order alternation that we
did for adverbs: we will move the verb around the negation. The tree here
will be slightly different, however. Let us assume that not heads a projection
called NegP, and this projection is the complement ofTP, and dominates VP.
1For the moment, ignore the doverb. We will return to this below.
252 Movement
The transformation of verb movement then raises the verb around pas as
represented by thearrow in (17).2 Again thisderives thecorrect word order.3
You nowhaveenough information totry General Problem Sets 1 &2
Observe that the alternation in position between an auxiliary and
a tensed verb is not limited to French. Many (if not most) languages show
this same alternation. Take for example the language Vata, a Kru language
of West Africa. The underlying word order of Vata is SOV (data from
Koopman 1984).
18) a) A la saka li.
we have rice eaten
"We have eaten rice."
b) A li saka.
we eat rice
"We eat rice."
In the sentence with the overt auxiliary, the verb appears to the far right.
When there is no auxiliary, the verb appears in the structural slot otherwise
occupied by the auxiliary. This alternation can be attributed to verb raising.
When there is an auxiliary (la), T does not require "support" from the verb,
so the verb remains in its base generated position (19).
2 An alternative to this is often found in the literature. In this alternative ne- heads
the NegP and pas is in the specifier of NegP. The verb raises and stops off at the Neg
head, (picking up ne- on the way) and then moves up to T. This alternative was
presented in Pollock (1989).
3You might note that in English the comparable operation (affix lowering) does not
apply around negation. We explore this issue in more detail below.
Chapter 9: Head-to-Head Movement 253
19) CP
When there is no auxiliary, T requires support, and the verb raises around
the object to T:
This, of course, is the correct word order (A li saka).
The transformational rule of V —> T movement thus provides a simple,
elegant and motivated account of cases where the verb shows up in
the "wrong" position. The motivation for the verb to move (or the affix
to lower) is intuitive: the need for the verb to get its inflection. This seems
to correlate with the fact that in many languages there are positional
alternations where auxiliaries (T) and tensed verbs alternate and are in
254 Movement
complementary distribution. This also gives a straightforward account of
certain cross-linguistic differences. We can account for the fact that English
and French consistently differ in the relative placement of adverbs
and negation with respect to tensed verbs. We derived this difference
by appealing to a parameter which either has the verb raise toT, orT-affixes
lower to the verb.
1.2 Irish
Now we'll turn to the other (more difficult) problem raised in the
introduction to this chapter. This is the Verb-Subject-Object (VSO) order
of Irish.
21) Phog Maire an lucharachan.
Kissed Mary the leprechaun
"Mary kissed the leprechaun."
As we observed above, there is no way that X-bar theory can generate
a sentence of this type. This is true of every basic sentence in Irish. VSO
order is found in every tensed sentence in Irish. It is also the basic order of
about 9 percent of the world's languages, including languages from many
different language families such as Tagalog, Welsh, Arabic, Mixtec, Mayan,
Salish, Turkana, Maasai (to name only a few).
Digression on Flat Structure
Up until the early 1980s, most linguists considered VSO languages to
simply be exceptions to X-bar theory. They proposed that these languages
had a flat structure:
i) TP
V DP DP
This structure is called "flat" because there are no hierarchical differences
between the subject, the object, and the verb. In other words, there are no
structural distinctions between complements, adjuncts and specifiers.
These sentences don't have a VP constituent. In (i) there is no single node
dominating both the Vand the second DP, but excluding the subject DP.
There is a delicate balance between a theory that is empirically
adequate (one that accounts for all the data), like a theory that has both
flat structure languages and X-bar languages, and one, which is
explanatorily adequate and elegant (like pure X-bar theory). By claiming
that these languages were exceptions, linguists were left with
a considerably less elegant theory. Thus the race was on to see if there
Chapter 9: Head-to-Head Movement 255
was some way to incorporate these languages into X-bar theory. Notice,
however, that pure elegance alone is not sufficient cause to abandon
an empirically adequate but inelegant theory like flat structure - we must
also have empirical evidence (data) in favor of the elegant theory.
Flat structure makes the following predications:
a) There is no VP constituent.
b) There is no evidence for a hierarchical distinction between subjects
and objects - they both have the same mother and mutually ccommand
one another.
It turns out that both these predications are wrong. First, if VSO
languages have no VP in simple tensed clauses they should have no VPs
in other clause types either. McCloskey (1983) observed for Irish, and
Sproat (1985) for Welsh, that this is false.
ii) Ta Maire [ag-pogail an lucharachan].
Is Mary ing-kiss the leprechaun
"Mary is kissing the leprechaun."
In auxiliary sentences in Irish, there is a plausible candidate for a VP:
the words bracketed in (ii). If this V + O sequence is a constituent,
it should obey constituency tests. Two typical constituency tests from
chapter 3, coordination and movement (clefting), show this:
iii) Ta Maire [ag-pogail an lucharachan] agus [ag-goidu a 6r].
Is Maiy [ing-kiss the leprechaun] and [ing-steal his gold]
"Mary is kissing the leprechaun and stealing his gold."
iv) Is [ag-pogail an lucharachan] ata Maire.
It-is [ing-kiss the leprechaun] that.be Mary
"It's kissing the leprechaun that Maiy is."
These sentences show that the bracketed [V + O] sequence in (ii) is
indeed a constituent, and a plausible VP.
Now, turn to the second prediction made by flat structure, where all
the DPs are on a par hierarchically. This too we can show is false. Recall
from chapter 5, that there is at least one phenomenon sensitive to
hierarchical position: the distribution of anaphors. Recall that the
antecedent of an anaphor must c-command it. If flat structure is correct,
then you should be able to have either DP be the antecedent and
either DP be the anaphor, since they mutually c-command one another
(they are sisters):
256 Movement
v) TP
V DP DP
The data in (vi) and (vii)show that this is false. Only the objectDP can be
an anaphor. This means that the object must be c-commanded by
the subject. Further it shows that the subject cannot be c-commanded by
the object. Flat structure simply can't account for this.
vi) Chonaic Silej 1-fekv
Saw Sheila her-self
"Sheila saw herself."
vii) *Chonaic i-feinj Sflej.
Saw her-self Sheila
"Sheila saw herself."
The flat structure approach, if you'll pardon the pun, comes up flat. It
makes the wrong predictions. The verb raising approach proposed in the
main text doesn't suffer from these problems. It maintains X-bar theory so
both has a VPand a hierarchicaldistinctionbetween subjectsand objects.
You nowhave enough information totryGeneral Problem Set3
The failure of X-bar theory to account for 9 percent of the world's
languages is a significant one! However, the theory of transformations gives
us an easy out to this problem. If we assume that VSO languages are
underlyingly SVO (at D-structure), then a transformational rule applies
which derives the initial order.
22) SVO => VSO
How might we actually structurally implement this rule? Given
the discussion in section 1.1 above, the answer should be obvious: we can
use verb movement.
There is some straightforward evidence in favor of a verb movement
approach to Irish word order: First, we see the same type of positional
auxiliary/tensed verb word order alternations.
23) Td Maire ag-p6gdil an lucharachan.
Is Mary ing-kiss the leprechaun
"Mary is kissing the leprechaun."
Chapter 9: Head-to-Head Movement 257
24) Phog Maire an lucharachan.
kissed Mary the leprechaun
"Mary kissed the leprechaun."
As in the French and Vata cases, with respect to a certain position (in Irish
the initial position), auxiliaries and main verbs are in complementary
distribution - evidence for V —> T movement.
Unfortunately the situation here is not as straightforward as the French
and Vata cases. If we try to draw the tree for (24) we immediately run into
a problem.
an lucharachan
While moving the verb to T certainly accounts for the alternation between
verbs and auxiliaries, it does not derive the correct VSO word order. Instead
we get incorrect SVO order.
In all the sentences of Irish we've looked at, T (in the form either of
anauxiliary ora raisedtensed verb)seems toprecede itsspecifier(the subject).
One possibility to resolve this might be in exercising the parameters we
looked at in chapter 6. Sowe might try putting the specifierofTP to the right
in Irish:
258 Movement
an lucharachan
But this doesn't work, if you look carefully at the order of elements
in 26you'll see this results in VOS order, which is completely ungrammatical
in Irish:
27) *Ph6g an lucharachan Mdire.
kissed the leprechaun Mary
(ungrammatical with the reading "Mary kissed the leprechaun.")
So X-bar parameters clearly aren't the solution. The only alternative
is to claim that we've been generating subjects in the wrong position. That is,
subjects are not generated in the specifier ofTP, like we have been assuming.
Instead, they are underlyingly generated in the specifier of VP.
The idea that subjects are generated in the specifier of VP is called the
VP-internal subject hypothesis, and was first proposed by Hilda Koopman
and DominiqueSportiche (1991). The idea has some thematic motivation. By
assuming that subjects are generated inside the VP we can make the strong
claim that theta roles are assigned entirely within the VP. We can encode
this in the following constraint:
28) The Locality Constraint on Theta Role Assignment
Theta roles are assigned within the projection of the head that assigns
them (i.e., the VP or other predicate).
If we assume the VP-internal subject hypothesis, the derivation of VSO
order is trivial: It involves a straightforward instance of V -» T movement:
Chapter 9: Head-to-Head Movement 259
29) CP
an lucharachcin
This derives the correct VSO order of Irish.
Now at this point your head is probably spinning and you are saying
to yourself "Hold on, what about English, French, and Vata! In all
those languages the subject precedes T." Alas, this is true. The solution to the
conundrum lies easily within our grasp, however. Perhaps it is the case that
in English, French, and Vata (but not the VSO languages) subject DPs move
from the specifier of VP to the specifier of TP. A simple French sentence
then would have two movements: one for the verb, one for the subject.
des pommes
260 Movement
This second kind of movement is called DP movement and is the topic of the
next chapter, where we'll discuss further evidence for VP-internal subjects.
The correct formulation and motivations for DP movement are set out there.
For now, we'll just observe that we have not argued ourselves into a corner;
there is a way out.
You nowhave enough information totry General Problem Sets4 &5 and Challenge
Problem Set 1
Let us summarize the (quite complicated) discussion up to now. In
section 0, we saw that there are instances where X-bar rules fail to generate
the correct orders of sentences. To solve this problem, we looked at a new
rule type: the transformation. Transformations take a structure generated
by X-bar theory and change it in restricted ways. We've looked at one such
transformation: V -> T. This rule has the function of movement a verb to
the T head. It does so in order that the verb can support inflection. We also
looked at the mirror image of verb movement: affix lowering, which lowers
an inflectional suffix to the verb. These are in complementary distribution, so
serve as tokens of the same rule. A language is parameterized as to whether
it takes the raising or the lowering variant. The difference in word order
between French and English negatives and sentences with adverbials can
be boiled down to this parameter. The rule of verb movement itself can
explain the fact that an adjunct (the adverb) appears between a head
and its complement. Taken together with the VP-internal subject hypothesis,
verb movement can also explain the very problematic basic VSO word order.
This simple straightforward tool thus allows us to account for a very wide
range of complicated facts.
2. T MOVEMENT (T -> C)
Before leaving the topic of the movement of heads, we briefly return
to a phenomenon somewhat obliquely discussed in chapter 7. This is
the phenomenon known as T —> C movement or subject-aux inversion.
In yes/no questions in English (questions that can be answered with either
a yes or no), auxiliary verbs invert with their subject:
31) a) Youhave squeezed the toilet paper,
b) Have you squeezed the toilet paper?
In chapter 7, we claimed that this alternation is due to the presence
of a special null question complementizer 0[+Q]. We observed that in many
languages (such as Polish and Irish) yes/no questions aren't indicated with
Chapter 9: Head-to-Head Movement 261
subject-aux inversion, but with a special form of the initial complementizer
(recall Irish is VSO to start with, so subject-aux inversion would do nothing):
32) An bhfaca tu an madra?
Q See you the dog
"Did you see the dog?"
We claimed that subject-aux inversion is a special case of these question
complementizers. English doesn't have an overt (pronounced) question
complementizer like the Irish an. Instead, English has a null 0[+qj
complementizer. Being phonologically null, however, is a bit of a problem,
since the difference in meaning between a statement and a question
is encoded in something you can't hear. English employs a mechanism
(which we now know is a transformation), that gives phonological content
to that 0[+q] by moving T to it, around the subject:
This kind of analysis is supported by the fact subject-aux inversion
(T -» C) is in strict complementary distribution with overt question
complementizers as seen in the following embedded clauses:
34) a) Iasked have yousqueezed thetoilet paper.4
b) I asked whether you have squeezed the toilet paper.
c) *Iasked whether have you squeezed the toilet paper.
4 For many people this sentence is not grammatical unless the embedded clause is
a direct quote. (That is, it would properly be written with " " around it.) This fact
muddies the waters somewhat in this argument, as it may not be the case that T —> C
movement is allowed at all in embedded clauses in English. However, the same facts
do hold true in other languages where subject-aux inversion in embedded clauses
is more clearly instantiated.
262 Movement
So the process of subject-aux inversion must be a property triggered by
complementizers. This rule is very similar to the rule of V —> T movement.
It is triggered by morphophonological requirements (such as the fact that
something contentful must be pronounced, or that an affix needs a host).
Both movements are instances of moving one head into another, so are
considered instances of the same basic operation: head-to-head movement.
This is a cover term for both V -» T and T —> C.
VSO as Raising to C?
In the previous section we claimed that Irish VSO order involves raising
the verb to T. We were also forced to claim that subjects were generated
VP internally. Notice that in English, we also have a VS order, found in
yes/no questions. These VS orders we analyze as T -> C movement, with
the subject remaining in its more typical place in the specifier of TP. Why
don't we analyze Irish VSO order the same way? Instead of having VPinternal
subjects, why don't we simply have verbs raise to T, then do T —>
C in all Irish clauses. This too would derive VSO order. There is a very
good reason for this. Recall that in English T —> C movement is blocked
when there is an overt complementizer. (You don't move T into the C,
because it already has phonological content.) If Irish VSO really involves
raising to C, then it should be the case that you do not get VSO order
when there is an overt complementizer. This is false. You get VSO order
even when there is a complementizer.
i) Duirt me gur phog Maire an lucharachan.
Said I that kissed Mary the leprechaun
"I said that Mary kissed the leprechaun."
This means that VSO must result from movement of the verb to some
position lower than the complementizer. This is the analysis we argued
for above, where V raises to T, and the subject is in the specifier of VP.
It appears as if V —> T and T -» C interact. In English, only auxiliaries
ever occupy the T head as free-standing entities. Main verbs do not raise to T
in English. So only auxiliaries undergo T —> C movement. Main verbs
never do:
35) a) Have you squeezed the toilet paper?
b) *Squeezed you the toilet paper?
Contrast this to French. In French, main verbs undergo V —> T movement.
This means that when French does T -> C movement, main verbs
Chapter 9: Head-to-Head Movement 263
are predicted to also invert (because they are in T). This can be seen
in the following tree:
des pommes
Movement [l] isV-»Tmovement. Movement [2] issubsequent movement of
the verb (in T) to C as part of T —> C movement.
This prediction is borne out. Main verbs in French do invert in questions,
but English main verbs do not.
37) a) Mangez-vous des pommes?
b) *Eatyou the apples?
To summarize, we have looked (again) at the transformation of T —> C
movement in more detail. We saw that it has a phonological motivation,
and is similar in some ways to V -» T movement. We also noticed that in
a language (such as French) where V -» T movement applies main verbs
as well as auxiliary verbs undergo T —> C.
You nowhave enough information to tryGeneral Problem Sets6,7 &8 and
Challenge Problem Sets2, 3,4 &5
3. DO-SUPPORT
In English, an interesting effect emerges when we try to question a sentence
with no auxiliary:
38) a) You eat apples,
b) Do you eat apples?
264 Movement
In sentences with no auxiliary, we insert a dummy (=meaningless) auxiliary
in yes/no questions. There must be a reason for this. We have argued
that in English, T lowers to attach to V, at the same time in questions,
the transformation of T —> C movement forces the same T to raise. This is
a contradiction: we want T to raise and lower at the same time. The
phenomenon of do-support appears to be an escape hatch for T. If we insert
a dummy (contentless) auxiliary to support the inflectional affixes, then this
dummy can undergo T -> C movement. This is an insertion transformation.
This transformation is called do-insertion or do-support:
39) Do-insertion: When there is no other option for supporting inflectional
affixes, insert the dummy verb dointo T.
What triggers this transformation is different than what triggers the
movement transformations. The movement transformations are motivated
(triggered) by morphophonological concerns. Insertion transformations
apply only in the case that there is nothing else you can do. They are, in
essence, operations of last resort, you only apply them when you absolutely
have to and when no movement transformation can apply.
There are Two Verbs Do in English
Quite confusingly, English has two verbs to do. One is a main verb,
meaning roughly "accomplish something," "perform an action." The
other is a dummy (meaningless) auxiliary, which is inserted under "dosupport."
These are quite distinct entities. As can be seen by the fact that
you can have both of them in one sentence:
i) Did you do your homework?
Main verb dois not an auxiliary and is not in T, this can be seen by the fact
that it cannot undergo T -* C movement, and it follows often and not.
ii) *Do you your homework?
iii) You often do your homework.
iv) You have not done your homework.
When invoking the do-insertion transformation, be careful that you only
do it when dummy do is involved - not main verb do.
Do-support doesn't apply only in questions; it also shows up in negative
sentences.
40) a) I ate the apple.
b) I didn't eat the apple.
Chapter 9: Head-to-Head Movement 265
The negative morpheme not blocks the operation of affix lowering.
The reasons for this are obscure. We will simply state it here as a stipulation.
4. Multiple Auxiliaries and Affix-hopping in English
4.1 MultipleAuxiliaries
One issue still remaining in our discussion comes from the behavior
of sentences in English with more than one auxiliary. Take the following
sentence as an example:
41) Shannon should have been being fed at the table.
The placement of should is easy, it's in T, but where should the other
auxiliaries in this sentence go? In chapter 7, challenge problem set 3, we
suggested that one solution to this problem might be to treat auxiliaries as Vs
that take other VPs as complements.
Shannon T
j^ should
tsubj fed at the table
266 Movement
One piece of evidence for this approach is that these auxiliaries can follow
negation and adverbs:
43) a) Shannon should not have been being fed at the table.
b) Shannon should [Adv never] have been being fed at the table.
Notice that negation cannot precede modals (which we will continue to treat
as belonging to category T):
44) a) *Shannonnot should have been being fed at the table.
b) *Shannondid not should have been being fed at the table.
This shows that these auxiliaries are not in T.
With this then, we have argued ourselves into a corner, since there are
many instances where the auxiliary verbs be and have do appear to be in T
because they precede negation:
45) a) Shannon has not eaten,
b) Shannon isnot eating.
There is a simple solution to this, auxiliary verbs (but not main verbs)
raise in English. This means that the verb movement parameter needs to be
modified:
46) Verb Movement Parameter: Option 1:All tensed verbs raise to T; Option 2:
tensed Auxiliaries raise to T and T lowers to tensed main verbs.
English chooses the second option.
Let's tree a number of sentences and see how this might work. Lefs start
with a simple sentence with a modal verb. We're continuing to treat modals
as instances of the category T:
47) Shannon should eat.
Chapter 9: Head-to-Head Movement
DP
Shannon T
4 should
267
There is no verb movement in this sentence, the only movement is the shift
of the subject DP from the specifier of the VP into the specifier of TP. Next
let's contrast this with a sentence with a tense suffix and no auxiliary at all:
49) Shannon ate.
50) CP
Here the -ed suffix lowers and attaches to the verb (this is followed
by a morphological rule that turns eat-ed into ate). Finally here's the tree
for a sentence with an auxiliary:
268
51) Shannon has eaten.
52) CP
Movement
Here, because there is no modal in T and no suffix in T, the auxiliary have
raises to T.
4.2 Affix-hopping
This analysis of multiple auxiliary constructions brings to light another
property of English. Consider the shape of the verb that follows each of most
common usages of the auxiliaries. When a verb follows a modal (category T
with no lowering), it takes the same form as it would in a non-finite context
(i.e., after to). We call this form the base form.
53) Shannon should eat. (cf. I want to eat.) modal
When the verb follows the auxiliary have in its perfective usage (that is, when
it marks that an event has a clear end-point), then the verb must appear
in it's past participle form. This is usually marked with the -en or -ed suffix,
although there are many verbs that have irregular participial forms.
Be careful not to mix up the -ed suffix here with the past tense suffix;
they are homophonous but theymean differentthings. Toavoid this confusion
we will indicate the past participle with -en.
54) Shannon has eaten. perfective
Chapter 9: Head-to-Head Movement 269
When the verb is indication that the action has a portion that is on-going,
we use the progressive form (this is sometimes also called a gerund, and
the aspect is sometimes called imperfect).
55) Shannon is eating. progressive
Finally, we have the form of the verb that is used for the passive voice. We'll
look at the passive in more detail in later in chapter 10. Like the perfective,
the passive uses the past participle, but follows the verb be:
56) The bread was eaten. passive
For each English type of inflection, we have a distinct set of verbal forms.
The suffix on the verb and the auxiliary together determine the meaning
of the expression. When we have a tensed main verb (that is, a simple past or
simple present), then the verb bears the appropriate tense morphology (-s, -d,
or -0). When a modal (category T) such as can or willappears, then there is
no morphology on the verb. When you have a perfective structure, we pair
the auxiliary have with an -en participle. The progressive is a pairing of
the verb be with an -ing progressive form. Finally the passive is a pairing
of the verb be with an -en participle. This is summarized in the following
chart.
57) a) Simple Past and Present V+ed/ V+s/ V+0
b) Modal modal V
c) Perfective have V+en
d) Progressive beV+ing
e) Passive be V+en
Notice what happens when we combine these patterns, the information
about aspect and voice gets distributed in a systematic way:
perfective passive
58) Shannon [has
progressive
Wehave a pattern of interleavinghere, where the suffixassociated with each
verbal inflection is actually one verb or auxiliary down from the auxiliary
that is associated with it. Chomsky, in what is perhaps his most famous
book, (Syntactic Structures 1957) provided an account of this phenomenon
known as Affix-hopping. The analysis we provide here, in terms of the affix
lowering transformation we already have, is a modified version
of Chomsky's.
270 Movement
Observe that the meaning "passive" cannot be described by the verb
be alone, since be is found in both passives and progressives, and it cannot
be described by just the suffix -en, since that is associated with both passives
and perfectives. The notion of "passive" is only encoded when both the be
and the -en suffix are present. Conversely, the meaning of be in isolation
isn't clear as it can be found with multiple inflection types. One way of
accounting for this is by having special lexical items that are syntactically
decomposable (that is, they can be split into two different syntactic units).
We can encode these as follows:
59) a) VPerf
have en
b) Vprog
A
be -ing
c) V pass
be
A
-en
Of course using these lexical items would result in the wrong word order.
Chomsky's insight was that this contradiction could be resolved using
a transformation: affix lowering (in combination with V—»T movement).
Let's start with a simple perfective sentence.
In this tree, we have both verb (auxiliary) raising and affix lowering.
The derivation of simple progressives and passive verb morphology is the
same substituting be+ing and be+en in for the higher V. A more complicated
example in (61):
Chapter 9: Head-to-Head Movement 271
61) CP
There are some things about these derivations which are still mysterious
of course, For example, why lexical items that can be split this way exist
and the precise motivation for the lowering the affixes (when they could,
in principle, attach to the auxiliary they come with). Nevertheless this
seems like a a good first approximation at capturing the complex behavior
of English auxiliaries.
You now have enough information to try General Problem Set 9
5. Summary
In this chapter we've looked at a range of phenomena (subject-aux inversion,
word order differences among languages, and do-support) that support
the basic notion that we need more than X-bar rules. We have introduced
272 Movement
the transformational movement rules of V -> T and T -> C and the insertion
rule of do-support to account for these phenomena.
Appendix: Tests for Determining if a Language has V -» T
or Affix Lowering
The following are tests that you can use to determine if a particular language
shows verb raising or not. These tests work well on SVO languages,
but don't work with SOV languages (such as Japanese).
A) If the language shows Subj V often O order then it has V —> T.
Ifthe language shows Subj often VO order then it has affix lowering.
B) If the language shows SubjV not O order then it has V -»T.
If the language shows Subjnot VO order then it has affixlowering.
C) If main verbs undergo T -» C movement, then the language has
V->T.
B^B^BDBBBBBBnl
Ideas, Rules, and Constraints Introduced in this Chapter
i) Transformation: A rule that takes an X-bar generated structure and
changes it in restricted ways.
ii) D-structure: The level of the derivation created by the base, and has
had no transformations applied to it.
iii) S-structure: The output of transformations. What you say.
iv) V —> T movement: Move the head V to the head T (motivated by
morphology).
v) Verb Movement Parameter: Option 1: all tensed verbs raise to T;
Option 2: tensed auxiliaries raise to T and T lowers to tensed main
verbs.
vi) The VP-internal Subject Hypothesis: Subjects are generated in the
specifier of VP.
vii) The Locality Constraint on Theta Role Assignment: Theta roles
are assigned within the projection of the head that assigns them
(i.e.,the VP or other predicate).
Chapter 9: Head-to-Head Movement 273
viii) T ^> C Movement: Move T to C, when there is a phonologically
empty 0[+qj complementizer.
ix) Do-insertion (Do-support): When there is no other option for
supporting inflectional affixes, insert the dummy verb do into T.
x) Stipulation: Affix lowering is blocked by the presence of not in
English.
xi) The Model:
The Computational Component
"the base"The Lexicon >• X-bar rules
D-structure
(constrained by Theta Criterion)
Transformational rules
> f
S-structure (constrained by EPP)
I
Grammaticality Judgments
Further Reading
Carnie, Andrew and Eithne Guilfoyle (2000) The Syntax of Verb Initial
Languages. Oxford: Oxford University Press.
Emonds, Joseph (1980) Word order in Generative Grammar. Journal of
Linguistic Research 1, 33-54.
Koopman, Hilda (1984) The Syntax of Verbs: From Verb Movement Rules in the
Km Languages to Universal Grammar. Dordrecht: Foris.
Koopman, Hilda and Dominique Sportiche (1991) The position of subjects.
Lingua85, 211-58.
274 Movement
Lightfoot, David and Norbert Hornstein (eds.) (1994) Verb Movement.
Cambridge: Cambridge University Press.
McCloskey, James (1983) A VP in a VSO language. In Gerald Gazdar,
Geoffrey Pullum, and Ivan Sag (eds.), Order Concord and Constituency.
Foris, Dordrecht, pp. 9-55.
McCloskey, James (1991) Clause structure, ellipsis and proper government in
Irish. Lingua 85, 259-302.
Ritter, Elizabeth (1988) A head movement approach to construct state noun
phrases. Linguistics 26, 909-29.
General Problem Sets
1. Italian
[Data Analysis; Basic]
Consider the following data from Italian. Assume non is like French neand
is irrelevant to the discussion. Concentrate instead on the positioning
of the word piu, 'anymore.' (Data from Belletti 1994.)
a) Gianni non ha piu parlato.
Gianni non has anymore spoken
"Gianni does not speak anymore."
b) Gianni non parla piu.
Gianni non speaks anymore
"Gianni speaks no more."
On the basis of this very limited data, is Italian a verb raising language or
an affix lowering language?
2. Haitian Creole Verb Placement
[Data Analysis; Basic]
Consider the following sentences from Haitian Creole. Is Creole a verb
raising language or an affix lowering language? Explain your answer.
(Data from DeGraff 2005.)
a) Bouki deja konnen Boukinet
Bouki already knows Boukinet
"Bouki already knows Boukinet."
b) Bouki pa konnen Boukinet
Bouki neg knows Boukinet
"Bouki doesn't know Boukinet."
Chapter 9: Head-to-Head Movement 275
3. Flat vs. Hierarchical Structure: Berber
[Data Analysis; Advanced]
Consider the following data from Berber. Using your knowledge of binding
theory, construct an argument that there is a VP in Berber, even though it
is a VSO language. (Hint: When there is a VP, the subject c-commands
the object, but when there is no VP, the two NPs asymmetrically c-command
each other.) (Data From Choe 1987.)
a)
b)
Yutut wrbak ixfnnsk
hit boy-NOMk himselfk
"The boyk hit himk"
*yutut ixfnnsk arbak
hit himselfk boyk
"himselfk hit the boyk"
4. Welsh
[Data Analysis; Basic]
Using the very limited data from Welsh below, construct an argument
that Welsh has V to T movement. Do not worry about the alternation in the
form of the word for "dragon," it is irrelevant to the answer to the question.
(Data from Kroeger 1993.)
a) Gwelodd Sion ddraig.
saw.PAST John dragon
"John saw a dragon."
b) Gwnaeth Sion weld draig.
do.PAST John seen dragon.GEN
"John saw a dragon."
5. VP Internal Subjects: Practice
[Application of Skills; Basic]
Using VP internal subjects, with movement to the specifier of TP
where appropriate, and verb movement or affix lowering where appropriate,
draw the trees for the following sentences:
a) Tiffany is not taking her syntax class until next year.
b) Christine likes wood furniture with a dark finish.
c) Les enfants n'ont pas travaille. (French)
the children have not worked
"The children haven't worked."
276 Movement
d) Les enfants (ne)-travaillent pas. (French)
the children work not
"The children don't work."
6. American vs. British English Verb Have
[Critical Thinking; Basic/Intermediate]
English has two verbs to have. One is an auxiliary seen in sentences like (a):
a) I have never seen this movie.
The other indicates possession:
b) I never have a pen when I need it.
You will note from the position of the adverb never that the possessive verb
have is a main verb, whereas the auxiliary have is raises to T.
Part 1: Consider the following data from American English. How does
it support the idea that auxiliary have ends up in T, but possessive have is
a main verb, and stays downstairs (i.e., has affix lowering applied)?
c) I have had a horrible day.
d) I have never had a pencil case like that!
e) Have you seen my backpack?
f) *Have you a pencil?
Part 2: Consider now the following sentence, which is grammatical in some
varieties of British English:
g) Have you a pencil?
Does the possessive verb have in these dialects undergo V —» T movement?
How can you tell?
7. Hebrew Construct State (N -> D)
[Data Analysis; Intermediate]
Background; In the text above we considered two variations
on head movement: V -> T, and T -> C. In an influential article in 1988, Ritter
proposed that head movement might also apply inside DPs. More particularly
she proposed that in many Semitic languages there is a rule of N -> D
movement. This applies in a possessive construction called the construct
state. (Based on the analysis of Ritter1988, data from Borer 1999.)
a) beit ha-more
house the-teacher
"the teacher's house"
Chapter 9: Head-to-Head Movement 277
In the construct state, the noun takes on a special form (the construct):
b) Free form bayit "house"
Construct beit "house"
Ritter proposes that the construct arises when the noun moves into the
determiner. The construct morphology indicates that this noun is attached
to the determiner. A tree for sentence (a) is given below. The possessor
noun sits in the specifier of the NP, the possessed N head undergoes
head movement to D, where it takes on the construct morphology:
This results in the surface DP [beit ha-more].
Part 1: Consider now the following evidence, how does this support Ritter's
N -» D analysis?
c) *ha-beit ha-more
the house the teacher
"the house of the teacher"
Part 2; Now look at the positioning of adjectives. How does this support
Ritter's analysis? Note in particular what noun the adjective modifies. (Ifyou
are having trouble with this question, trying drawing the tree of what
the whole DP would look like before N -> D movement applied.) M stands
for "masculine", and f stands for feminine:
d) more kita xadaS
teacher-M class-F new-M
"a class's new teacher" or "the new teacher of a class"
but: "*a new class's teacher" or "*the teacher of a new class"
278 Movement
8. English5
[Data Analysis; Intermediate]
Consider the italicized noun phrases inthe following sentences:
a) I ate something spicy.
b) Someone tallwas looking for you.
c) Idon't like anyone smart.
d) Iwill read anything interesting.
One analysis that has been proposed for noun phrases like the ones
above involves generating elements like some and any as determiners,
and generating elements one and thing as nouns (under N), and then
doing head-to-head movement of the Ns up to D. The tree below illustrates
this analysis:
r
Give an argument in favor of this analysis, based on the order of elements
within the noun phrase in general, and the order of elements in the noun
phrases above.
9. English Trees
[Application of Skills; Basic toAdvanced]
Draw trees for the following English sentences; be sure to indicate all
transformations with arrows.
a) I have always loved peanut butter.
b) I do not love peanut butter.
c) Martha often thinks Kim hates phonology.
d) Do you like peanut butter?
e) Have you always hated peanut butter?
'ThankstoJilaGhomeshi forcontributing this problemset.
Chapter 9: Head-to-Head Movement 279
f) Are you always so obtuse? (Assume that AdjP can be a complement
to Tifit is a predicate, as in this case)
g) Will you bring your spouse?
h) Has the food been eaten?
i) Mike is always eating peanuts.
Challenge Problem Sets
Challenge Problem Set 1: Floating Quantifiers
[Critical Thinking; Challenge]
In English, quantifiers normally appear before a DP. Up to this point in the
book, we've been treating them as determiners. However, certain quantifiers
can appear before determiners. One example is the quantifier all:
alt the men. In section 4 above, we argued that we can have stacked VPs.
Let's extend that analysis and claim that we can have stacked DPs in certain
circumstances (limited by the particular determiners involved). The structure
of all the men is given below:
DP2
I
D'2
D2 DPt
all I
D'i
Di NP
the '
N'
I
N
men
There are two DPs here (DP! and DP2), in principle either of them could be
moved to the specifier of TP. With this in mind provide an argument using the
following data to argue that subjects in English start in the specifier of VP:
a) All the men have gone.
b) The men have all gone.
280 Movement
Challenge Problem Set 2: Verb Movement6
[Data Analysis; Challenge]
Based on the following data, do German and Persian exhibit V —> T
movement? Explain how you came to your answer.
German
a) Sprechen Sie Deutsch?
speak you German
"Do you speak German?"
b) 1st er nach Hause gegangen?
is he to home gone
"Has he gone home?"
c) Er sitzt nicht auf diesem Tisch.
he sits not on this table
"He does not sit on this table."
d) Sie soil nicht auf diesem Tisch sitzen.
she must not on this table sit
"She must not sit on this table."
Persian
a) Rafti to madrese?
went you school
"Did you go to school?"
b) Bayad un biyad?
must he come
"Must he come?"
c) Man keyk na-poxtam.
I cake not-cooked
"I did not bake cakes."
d) Un na-xahad amad.
he not-will come
"He will not come."
6ThankstoSiminKarimi forcontributing thisdata.
Chapter 9: Head-to-Head Movement 281
Challenge Problem Set 3: Germanic Verb Second
[Data Analysis and Critical Thinking; Challenge]
Background: Many of the languages of the Germaniclanguage family exhibit
what is known as verb second order (also known as V2). With V2, the main
restriction on word order is that, in main clauses, the constituents may
appear in essentially any order, as long as the verb is in the second position
in the sentence. This is seen in the following data from German:
German (Vikner 1995)
a) Die Kinder haben diesen Film gesehen.
the children have this film seen
"The children have seen this film."
b) Diesen Film haben die Kinder gesehen.
One analysis of this phenomenon uses the specifier of CP as a "topic"
position. The most topical constituent (the bit under discussion) is put in the
specifier of CP (i.e., is moved there - we'll discuss this kind of movement
in chapter 12). Whatever is in T then moves to the C head by T -» C
movement:
c)
This puts T in second position. For the tree above and this problem set,
assume that the VP and the TP have their heads on the right, but CP
is left headed.
Part 1: Now consider the following data from embedded clauses in German.
d) Er sagt [dad die Kinder diesen Filmgesehen haben].
He said that the children this film saw have
"He said that the children saw this film."
e) *Ersagt [dali die Kinder haben diesen Film gesehen].
How does this data support the T -> C analysis of V2? (Having trouble?
Think about embedded yes/no questions in English.)
282 Movement
Part 2: Consider now the following sentence of German and compare it
to the embedded clauses in part 1 above.
f) Gestern sahen die Kinder den Film.
Yesterday saw the children the film
"The children saw the film yesterday."
Given what you now know about V2 and T -> C movement
in these languages, isGerman aV->T raising language or an affix lowering
language?
Bonus: Is the data in part 1 above consistent with your answer? If not how
might you make it consistent?
Challenge Problem Set4: Proper Names and Pronouns
[Data Analysis; Challenge]
Consider the following data from English:
a) Lucy
b) *The Lucy
c) 'Smiths
d) The Smiths
e) Him
f) *The him
g) We linguistslove a good debate over grammar.
Part 1: One possible analysis ofproper names in English isthat they involve
head movement from an N position into a D position. How does the data
in (a-d) above support this idea?
Part 2: Consider now the pronouns in (e-g). What category are they? Nor
D? Is there any evidence for movement?
Challenge Problem Set 5: Italian N->D7
[Data Analysis and Critical Thinking; Challenge]
(You may want to do Challenge Question 4before attempting this problem.)
In English, proper names cannotco-occur with determiners (e.g. *the John).
However, in Italian proper names of human beings can occur
with determiners as the following example shows. (The presence or absence
ofthe determiner seems to be free or perhapsstylistically governed.)
7Jila Ghomeshi contributed this problem setbased ondata from Longobardi (1994).
a) i)
ii)
Chapter 9: Head-to-Head Movement
Gianni mi ha telefonato.
Gianni me has telephoned
"Gianni called me up."
II Gianni mi ha telefonato.
the Gianni me has telephoned
"Gianni called me up."
283
Now, it has been argued that in the cases where the determiner does
not occur, the proper name has moved from N to D. Provide an argument to
support this view, based on the following examples. (Note: for the purposes
of this question treat possessive pronouns such as my as adjectives.)
b) i) II mio
my
Gianni
Gianni
mio
my
ii)
iii)
c) i)
ii)
the
*Mio
my
Gianni
Gianni
E'venuto
came
*E'venuto
came
iii) E'venuto
came
Gianni
Gianni
ha
has
ha
has
the
vecchio
older
Cameresi
Cameresi
d) i) L' antica Roma
the ancient Rome
"Ancient Rome"
ii) *Antica
ancient
iii) Roma
Rome
Roma
Rome
antica
ancien
ha
has
finalmente
finally
telefonato.
telephoned
finalmente
finally
finalmente
finally
vecchio
older
telefonato.
telephoned
telefonato.
telephoned
Cameresi.
Cameresi
Cameresi.
Cameresi
vecchio.
older
DP Movement
0. INTRODUCTION
In the last chapter, we looked at how certain basic word order facts could not
be generated by X-bar theory alone. Instead, we saw that we need another
rule type: the transformation. Transformations take X-bar trees and move
elements around in them. The kind of transformation we looked at there
moved heads into other heads. In this chapter, we are going to look
at transformations that move NPs and DPs. (For the sake of convenience,
I'm going to use NP to mean either NP or DP. Nothing turns on this usage.
We could equally call the phenomenon DP-movement.)
Unlike head-to-head movement, where movement is motivated by word
orders that cannot be generated using X-bartheory, the movement described
here frequently takes X-bar generated trees and turns them into
other acceptableX-bar generated trees. What motivates the movement is not
a failure of X-bar theory, but instead the fact that certain DPs can appear
in positions we don't expect from a thematic(theta role) perspective.
1. A Puzzle for the Theory of Theta Roles
Try to sketch out the theta grid for the verb to leave. Leave requires one
obligatory argument: an agent:
286 Movement
1) leave
Agent
DP
This can be seen from the following paradigm.
2) a) Bradleyj left.
b) Stacyj left Tucson.
c) SlavkOj left his wife.
d) *Itleft, (where ifis a dummy pronoun, not a thing)
The only obligatory argument for the verb leave is the agent, which is
an external (subject) argument. Other arguments are possible (as in 2b and c)
but not required. Now, note the following thing about the obligatory agent
theta role. The agent role must be assigned to an argument within the clause
that contains leave:
3) a) *[Iwant Bradleyj[that left]],
b) *Johnj thinks [that left].
When you try to assign the theta role to a DP that is outside the clause
(such as theobject Bradley orJohn in (3)) you get a stunninglyungrammatical
sentence. We already have an explanation for this fact: in the last chapter
we posited the following constraint:
4) The Locality Constraint on Theta Role Assignment
Theta roles are assigned within the projection of the head that assigns
them (i.e., the VPor other predicate).
This constraint requires that the DP getting the theta role be local
to the predicate that assigns it. In the sentences in (3) the DP is actually
in a different clause than the predicate that assigns it, so (4) predicts
them to be ungrammatical.
Now, look at the following sentence:
5) [Johnjis likely [to leave]].
John here is the agent of leaving, but the DPJohn appears in the main clause,
far away from its predicate. Even more surprising is the fact that there seems
to be no subject of the embedded clause. This is in direct violation of (4).
The solution to this problem is simple: there is a transformation that takes
the DP John and moves it from the lower clause to the higher clause.
Let's spell this out in more detail. The theta grid for is likely includes
only one argument: the embedded clause. This is seen in the fact that it can
appear as the sole theta marked argument:
Chapter 10: DP Movement
6) a) [[ThatJohn will leavejj is likely].
b) It is likely [that John will leave],.
c) is likely
Proposition
CP
J
287
Predicates Like Is Likely
In this chapter we're going to look at a number of predicates that consist
of the auxiliary beand an adjective such as likely or obvious, as in It is likely
that Daphne like cremefraiche. A few words are in order on how to tree this
structure. In the last chapter, we argued that auxiliaries like is are
generated in a V and then raise to the T node. The adjective likely (this is
an adjective even though it ends in -ly, as only other adjectives, like
obvious, eager, easy, etc., can appear in this position) is the complement of
this verb. These adjectival predicates typically take a CP as a complement.
We can tree these forms as below. We'll revise this slightly in chapter 14.
If this is the case, then in sentence (5), John is not receiving its theta role from
is likely. This should be obvious from the meaning of the sentence as well.
288 Movement
There is nothing about John that is likely, instead it is what John is doing
(his leaving) that is likely. The sentence is a clear violation of the locality
condition on theta role assignment in its surface form. In chapter 8, we
argued that the theta criterion applies before the transformation of expletive
insertion occurs. Translated into our new terminology, this means that the
theta criterion holds of D-structure. This means that theta role assignment
must also happen before all transformations. We can arrange for John's
theta role to be assigned clause internally, at D-structure. The D-structure
of the sentence would then look like (7) (Theta marking is indicated with
a dotted large arrow):
7)
Adj CP
likely* *\
'proposition
c
0
c
TP
9ageNnV .-'leave
Chapter 10:DP Movement 289
The subject DP is generated in the specifier of the embedded VP where it
is assigned the agent theta role. How then do we derive the surface order?
We need a transformation that moves this DP to the specifier of the main
clause TP. This transformation is called DP movement:
8) DP movement
Move a DP to a specifier position.
Notice that in the D-structure tree in (7) the specifier of the higher clause's
TP is unoccupied. We can thus move the DPJohn into that position resulting
in the tree in (9). (Note: the movement stops off in the specifier
of the embedded TP and then moves on to the higher TP; we'll discuss why
this happens in two hops shortly.)
9) CP
290 Movement
This particular instance of DP movement is frequently called raising,
because you are raising the DP from the lower clause to the higher.
The surface structure of this tree looks like (10) where there is a trace
(marked t) left in each position that the DP has occupied.
You nowhave enough information to tryGeneral Problem Set 1
As we stated in the last chapter, transformations are very powerful tools,
and we want to limit their use. In particular we want to ensure that they only
apply when required. Transformations thus need motivations or triggers.
Look at the sentences in (11).
Chapter 10: DP Movement 291
11) a) [That John will leave] is likely,
b) It is likely that John will leave.
Recall back to the chapter on the lexicon, the presence of the theta-role-less it
in (b) is forced by the Extended Projection Principle (EPP) - the requirement
that the specifier ofTP be filled by something (i.e., the requirement that there
is a subject in every sentence). We might speculate then that the absence of
a subject is the trigger for DP movement. The DP moves to the TP to satisfy
the EPP. Since we have twoTPs thisapplies twice. [A] TheDPmoves from its
theta position in the specifier of the embedded VP to the lower TP to satisfy
this TP's EPP requirement. Then it moves on to the higher TP to satisfy
its requirements |B].
12) EPP EPP
IcpItp T+is [VP tv[AP likely [CP [TP to [VP John leave]]]]]]]
This explanation seems to work at least partially well. And we'll adopt it for
theory-internal reasons to motivate the movement to the embedded specifier
at the least. In section 3, we revisit this question and see that the EPP
is only a partly satisfactory motivation for DP movement, and will posit
an approach using "Case." First,however, let's look at the other main situation
that involves DP movement: Passives.
2. Passives
The sentence given in (13) is what is called an active sentence in traditional
grammar:
13) The policeman kissed the puppy. Active
The sentence given in (14)by contrast is what is called a passive:
14) The puppy was kissed by the policeman. Passive
These two sentences don't mean exactly the same thing. The first one
is a sentence about a policeman (the policeman is the topic of the sentence);
by contrast (14) isa sentence about a puppy (thepuppy is the topic). However,
they do describe the same basic event in the world with the same basic
participants: there is some kissing going on, and the kisser (agent) is the
policeman and the kissee (theme) is the puppy. At least on the surface then,
these two sentences seem to involve the same thematic information.
292 Movement
On closer examination however, things change. Notice that in the passive
sentence, the agent is represented by an optional prepositional phrase
headed by by. This is an adjunct; as discussed in the chapter on the lexicon,
adjuncts are not included in the basic theta grid and are not subject to
the theta criterion. If the agent here is an adjunct and not subject to the theta
criterion it should be optional. This is indeed the case:
15) The puppy was kissed.
It thus seems that passives and actives have different thematic properties.
Actives have an agent and a theme, whereas passives lack the agentive theta
role in their theta grids.
The explanation for this is not syntactic, instead it is a morphological
issue. The passive form of a verb takes special morphology. In English, there
are two main passive suffixes. One is (unfortunately) homophonous with the
past tense suffix -ed. The other is the -en suffix. These two are allomorphs of
each other. We will use -en as the basic form, so as not to confuse the passive
morpheme with the past tense. There is a simple morphological operation
that derives a passive verb from an active one:
16) kiss+en —» kissed, beat+en —»beaten, etc.
This morphological operation doesn't only affect the outward pronunciation
of the word, it also affects the meaning. More particularly it affects the theta
grid of the verb. Whenever the -en suffix is present, there is no DP. One way
of thinking of this is that the -en absorbs (or is itselfassigned) the agent role.1
17) a) kiss
b)
Agent
DP
Theme
DP
kiss+en (—> kissed)
Agent
DP
Theme
DP
-en
Now, let's look at the word order in the passive and active. In the active,
the theme argument appears in object position; in the passive it appears in
the subject position. One possible analysis of this is to claim that the theme is
1This technically is a violation of our locality constraint as the -en is introduced by
a different V than its theta assigner. For the moment we'll simply assume that any VP
inside a clause counts as the domain for theta role assignment for that main verb
in that clause.
Chapter 10:DP Movement 293
generated in object position in both actives and passives, but then is moved
to subject position in passives.
Here is a sample derivation. The D-structure of the passive sentence
looks like (18). The dotted arrows in this tree represent theta (0)
assignment, not movement. Because -en absorbs the agent role, there is
only one DP in this sentence (the puppy), the one that gets the theme role.
Even if there is a byphrase (e.g., by the policeman) it does not get its theta role
from the verb, it is an adjunct, and adjuncts are never included in theta grids.
The theme is the internal argument (i.e., it is not underlined in
the theta grid), so it does not appear in the specifier of the VP, it must appear
as the complement, like other internal theta roles .
was +en
agent
~\ v e^e „,^DP
kiss***'"^^^^
The puppy
Now, like the raising sentences we looked at in section 1, the EPP is
not satisfied here. There is nothing in the specifier of TP. The surface order
of the passive can then be derived by DP movement (and head movement
of the auxiliary and lowering of the affix).
294 Movement
19) CP
The DP the puppy moves to satisfy the EPP.
As mentioned above, passives often also occur with what appears to be
the originalexternal argument in a prepositional phrase marked with by.
20) The puppy was kissed by the policeman
We treat these by-phrases as optional adjuncts. We draw these by-phrases in
by adjoining them to V:
21)
by the policeman
You nowhave enough information totryGeneral Problem Set2. You can alsotry
Challenge Problem Sets 1 &2
Chapter 10: DP Movement 295
Movement or Underlying External Theme?
One might ask why it isn't simpler to say that the passive morpheme just
deletes the agent and makes the theme an external argument in the theta
grid.
i)
Agent
DP
Theme
DP
Theme
DP
Then the D-structure of the sentence will put the theme into the subject
position right from the start with no movement. This is impossible,
however, if you look at passives of sentences that take clausal
complements. Take the active sentence in (ii):
ii) Wilma considers [Fredrick to be foolish].
In this sentence, Wilma is the experiencer of consider, and Fredrick is the
external theta role of the predicate isfoolish. When consider is made into a
passive, the subject of the lower clause raises to become the subject of the
main clause:
iii) Fredrickj is considered f, to be foolish.
Notice that Fredrick is never theta-marked by the verb consider. As such
there is no way to make it the external argument like in (i). Because of
cases like (iii), the movement account is preferred.
3. Case
Up until now, we've motivated the movement of DPs using the EPP. In this
section, we look at some data that shows that we might need an additional
mechanism to account for movement.
Let's start with raising: as we saw in the last chapter, one way to satisfy
the EPP is by inserting an expletive. For some reason this option isn't
available in raising environments:
22) *Itis likely John to leave. (cf. // is likely that John left)
Nor does it explain why only the subject DP of an embedded clause can
satisfy the EPP; objectDP may not move to satisfy this requirement:
23) *Billi is likely John to hit f,.
296 Movement
The same kind of mystery appears in passives. It isn't clear why it isn't
simply permissible to satisfy the EPP by inserting an expletive:
24) *It waskissed the puppy.2
Our theory predicts that such sentences should be acceptable. In order to
explain why they are not, we are going to have to add a new theoretical tool:
Case.
In many languages, nouns bearing various grammatical relations take
special forms. For example, in Japanese, subjects are marked with the suffix
-ga, objects are marked with -o and indirect objects and certain adjuncts
with -ni:
25) Asako-ga ronbun-o kai-ta.
Asako-NOM article-ACC wrote-PAST
"Asako wrote the article."
26) Etsuko-ga heya-ni haitte-kita.
Etsuko-NOM room-DAT in-came
"Etsuko came into the room."
These suffixes represent grammatical relations (see chapter 4). The three
most important grammatical relations are subject, object, and indirect object.
Notice that these are not the same as thematic relations. Thematic relations
represent meaning. Grammatical relationsrepresent how a DP is functioning
in the sentence syntactically. The morphology associated with grammatical
relations is called case. The two cases we will be primarily concerned
with here are the nominative case, which is found with subjects,
and the accusative case, found with objects.
English is a morphologically poor language. In sentences with full DPs,
there is no obvious case marking. Grammatical relations are represented
by the position of the noun in the sentence:
27) a) Jennifer swatted Steve,
b) Steve swatted Jennifer.
There is no difference in form between Jennifer in (27a), where the DP
is functioning as a subject, and (27b), where it is functioning as an object.
With pronouns, by contrast, there is a clear morphological difference,
as we observed in chapter 1.
2This sentence becomes grammatical if you put a big pause after kissed, but notice
that in this circumstance, the it is not a dummy, but refers to the puppy.
Chapter 10: DP Movement 297
28) a) She swatted him.
b) He swatted her.
Most pronouns in English have different forms depending upon what case
they are in:
29) Nominative I you
Accusative me you
Can this be extended to full DPs? Well, consider the general poverty
of English morphology. The first and second persons in the present tense
form of verbs don't take any overt suffix:
30) a) I walk.
b) You walk. (cf. He/She/It walks. You walked.)
But one wouldn't want to claim that (30a and b) aren't inflected for tense.
Semantically they are. These forms can only refer to the present, they can't
refer to the past or the future. We are thus forced to claim that there is
an unpronounced or null present tense morpheme in English. It seems
reasonable to claim that if there are null tense suffixes, there are also null
case suffixes in English. Indeed, in the system we are proposing here all
nouns get case - we just don't see it overtly in the pronounced morphology.
This is called abstract Case. (Abstract Case normally has a capital C
to distinguish it from morphological case.)
Case, then, is a general property of Language. Furthermore it seems
to be associated with a syntactic phenomenon - the grammatical function
(relations) of DPs. If it is indeed a syntactic property, then it should have a
structural trigger. In the case theory of Chomsky (1981), DPs are given Case
if and only if they appear in specific positions in the sentence. In particular,
nominative case is assigned in the specifier of finite T, and accusative case
is assigned as a sister to the verb (prepositions also assign what is often
called "Prepositional case" to theircomplement DP):3
31) NOMinative case Specifierof finite T
ACCusative case Sister to transitive V
PREPositional case Assigned by a preposition
Case serves as our motivation for DP movement. You can think of Case
as being like a driver's license. You can't drive without a license, and you
can only get a license at the Department of Motor Vehicles. So you have to
e she it we you they
im her it us you them
3This is an almost ridiculous oversimplification. There are many prepositional cases
(datives, locatives, ablatives, jussives, etc.). We abstract away from this here. We are
also ignoring the genitive case normally associatedwith possessive constructions.
298 Movement
go there to get the license. A DP needs a license to surface in the sentence,
and it can only get a license (Case) in specific positions. If it isn't in one
of those positions, it must move to get Case. A DP without Case can't drive.
This is called the Casefilter.
32) The Casefilter
All DPs must be marked with a Case.
If a DP doesn't get Case the derivation will crash.
One standard way of implementing the Case filter is by using a mechanism
known as feature checking. This is based on a notion taken from phonology.
The idea is that words are composed of atomic features. A word like he
is composed of features representing its person, its number, its gender etc.
We can represent these features in a matrix:
33) he [""masculine
3rd person
singular
jiominative
Similarly, we will claimthat CaseassignerslikeT have a feature matrix:
34) T(is) ["present -
3rd person
singular
nominative
You'll notice that both of these feature matrices have a feature [nominative].
The Case filter becomes a requirement that a noun like he be close enough
to a Case assigner like is, to check that the noun has the right features.
The noun must be close to its Case assigner:
35)
Nominative Case
checking
36)
checking
Chapter 10: DP Movement
Accusative Case
299
Ergative/Absolutive Languages
In this book, we are looking exclusively at languages that take
nominative and accusative cases. These are a fairly common kind of
language in the western hemisphere. In nominative/accusative
languages, the same case is assigned to the subjects of transitives and the
subjects of intransitives (nominative case); a different case (accusative) is
assigned to the objects of intransitives.
i) Norn/Ace languages
Nom Ace
Trans Subject Object
In trans Subject
However, there is a huge class of languages that does not use this case
pattern, including many Polynesian, Australian, and Central American
languages. These languages, called "Ergative/Absolutive" languages,
mark the object of transitives and the subject of intransitives using the
same case (absolutive); subjects of transitives are marked with a different
case: ergative.
ii) Erg/Abs languages
Er* Abs
Trans Subject Object
Intrans Subject
Fror
mys
pres
Non
This
n the perspec
tery and the
ented here. E\
n/Acc and Er
is a topic of a
tive of structu
subject of gre
en more myste
g/Abs case sy
lot of current re
*al case the
at debate,
rious are th
stems (undt
search in sy
Dry, these languages are a
They don't fit the theory
ose languages that use both
r different circumstances),
ntax now.
300 Movement
37) PP Prepositional Case
I
F
p DP
to
A
[prep] John
[prep]
checking
If the noun and the Case assigner are not local (that is, the noun is not in
the specifier or complement of the Case assigner), then the feature won't
be checked and the Case filter violated. We'll use this notion of locality
in feature checking again in chapter 11, when we look at w/i-movement.
You now have enough information to try General Problem Set 3
4. Raising: Reprise
Let's now return to the raising sentences we were looking at in section 1, and
we'll expand the paradigm to include the following:
38) a) It is likely that Patrick left.
b) That Patrick left is likely.
c) *Patrick is likely that f, left.
d) *Itis likely Patrick to leave.
e) *Patrick to leave is likely.
f) Patrick is likely t{ to leave.
Sentences (38a-c) involve a tensed (finite) embedded clause. Sentence
(38a) shows that one can satisfy the EPP with an expletive, provided
the embedded clause is finite. Sentence (38d) shows that an expletive won't
suffice with a non-finite embedded clause. Sentence (38b) shows that
a tensed clause can satisfy the EPP, but a non-finite one cannot (38e). Finally,
we see that raising is possible with a non-finite clause (38f) but not a finite
one (38c). This is quite a complicated set of facts, but it turns out that
the distribution turns on a single issue. Above we saw that DPs are assigned
nominative Case only in the specifierof finiteT. (In other words, non-finiteT
does not have a [nom] feature, whereas finite T does.) Sentences (38d-f)
are non-finite. This means that the DPPatrick cannotget nominative Case in
Chapter 10:DP Movement 301
the specifierof the embedded clause.The ungrammaticality of (38d and e) are
now explained: Patrick is not getting Case, so the sentence violates the Case
filter. In sentence (38f) by contrast, the DP has moved to the specifier of the
finite main clause T; it can receive Case here, so the sentence is grammatical:
39) CP
C
C TP Nominative Case
0 ^^^
C TP ., No case here! (But movement to
0 /^x this position happensfor EPP)
T
John starts out where it gets its theta role (the specifier of VP), then it moves
to the specifier of the embedded TP, where it satisfies the EPP for that TP.
But this is not a case position, the T to shows that the clause is non-finite.
So the DP moves from this position to the specifier of the higher TP, where it
can check its nominative Case. This is a pattern that is repeated over and
302 Movement
over again. DPs always move from positions where they can't check Case
(but where they get a theta role) to positions where they get Case.
The distribution of raising in sentences (39a-c) is also now explained.
These clauses have an embedded finite T. As such the DP Patrick can get
nominative Case in the specifier of embedded T. It does not have to move.
If it did move, it would move without reason, as it already has Case.
You nowhave enough information totryGeneral Problem Set4 andChallenge
Problem Set 3
5. Passives: Reprise
Case theory also allows an explanation of passive constructions. However,
this requires an additional piece of machinery to be added to the passive
morphology. Only active transitive verbs can assign accusative Case:
40) He kissed her.
Passive verbs cannot:
41) a) She was kissed.
b) *Shewas kissed him.4
c) *It was kissed her. (where it is an expletive)
It thus appears that not only does the passive suffix absorb the verb's
external theta role, it also absorbs the verb's ability to assign accusative Case.
This is a rough version of what is called Burzio's Generalization
(after Burzio 1986): A predicate that assigns noexternal theta role cannot assign
accusative Case. The passive morpheme thus has the following two functions:
42) The passive morpheme -en
a) absorbs a verb's external theta role.
b) checks a verb's [ACC] Case feature.
Recall that the -en suffix lowers and attaches to the V. This movement also
creates a local configuration. The verb and the -ensuffix are adjacent. In fact,
the -en is closer to the verb than the DP.
4This sentence is also a violation of the theta criterion.
Chapter 10:DP Movement
43)
kiss + en
[acc] [acc]
localconfiguration
Since the -en absorbs the accusative case, there is now no case for the DP, so
it must move to get Case. With this in mind, reconsider the passive sentence
we looked at in section 2:
44) CP
the puppy T
[nom]
. no longerlocalto thecase
^ assigner
303
The passive morphology has conspired to absorb both the accusative case
and the external theta role. This means that there is no DP in the specifier
of the finite T. There is a Case position open so the theme DP can move to
the specifier of TP. Now we have the trigger for DP movement in passives:
A DP moves to get Case from its Caseless theta position to the nominative
304 Movement
Case assigning specifier of TP. Notice that this DP now moves for two
reasons. First it moves to satisfy the EPP, but it also must move to get Case.
Inherently Passive Verbs: Unaccusatives
One of the interesting discoveries of the 1980swas the fact that there is a
set of verbs in many languages that are inherently passive. That is they
have only an internal argument, and they don't assign accusative case.
These are called unaccusative verbs (or less commonly ergative verbs).
Compare the two sentences in (i) and (ii)
i) Stacy danced at the palace,
ii) Stacy arrived at the palace.
The first sentence is a regular intransitive (often called unergative) where
Stacy bears an external agent theta role. The sentence in (ii) by contrast
has no external theta role. Stacy is a theme that originates in the object
position of the sentence. Stacy is then raised to subject position to satisfy
the Case filter, just like a passive. These predicates are passive without
having any passive morphology. The arguments for this are well beyond
the scope of this textbook. But note the following two differences between
the predicates in (i) and (ii).The unergative predicate in (i) can optionally
take a direct object. Unaccusative predicates cannot (something that is
predicted, if their subject is underlyingly an object):
iii) Stacy danced a jig.
iv) *Stacy arrived a letter.
Unaccusatives also allow an alternative word order (called there
inversion) where the underlying object remains in object position. Since
unergative subjects aren't generated in object position, they aren't
allowed to appear there with there inversion.
v) *Theredanced three men at the palace,
vi) ?There arrived three men at the palace.
You now have enough information to try General Problem Set 5
6. Closing Up a Loose End
In the last chapter, we were forced to argue (on the basis of evidence from
the VSO language Irish) that subject DPs were generated in the specifier
of VP not TP.
45) TP
Chapter 10:DP Movement
Irish
305
The problem, then, was why subject DPs appear before T in languages
like English. The solution should now be clear: All subject DPs move
to the specifier of finite T to get Case. In actives and intransitives, this is from
the specifier of VP. In passives, the movement is from the underlying object
position.
English
The difference between SVOlanguages like English and VSOlanguages is in
where nominative Case is assigned. In SVO languages, nominative Case
is assigned in the specifier of finite T. In VSO languages, nominative Case
is assigned when the DP is immediately c-commanded by finite T (which
allows it to remain inside VP).
_ Nominative Case position for English
English DPsubjJT
T VP ^^ Nominative Case position for Irish
Irish DP,subj V
V DPobj
306 Movement
You nowhaveenough information to try General Problem Sets 6-9
7. Conclusion
In this chapter, we've looked at situations where DPs don't appear in the
positions we expect them to (given our knowledge of theta theory). We have
argued that these sentences involve movement of DPs to various specifier
positions. The motivation for this comes from Case. The Case filter requires
all DPs to check a Case in a specific structural position. We looked at two
situations where DPs don't get Case in their D-structure position. In raising
structures, a DP is in the specifier of an embedded clause with non-finite T.
In this position, it can't receive Case so it raises to the specifier of the finite T
in the higher clause. We also looked at passive structures. The passive
morpheme does two things: it takes the role of external argument and
absorbs the verb's ability to assign accusative Case. This results in a structure
where there is no subject DP, and the object cannot receive Case in its base
position. The DP must move to the specifier of T to get Case.
Ideas, Rules, and Constraints Introduced in this Chapter
i) DP Movement: Move a DP to a specifier position.
ii) Raising: A specific instance of DP movement. The DP moves from
the specifier of an embedded non-finite T to the specifier of a finite T
in the main clause where it can get Case.
iii) case (lowercase c): The special form DPs get depending upon their
place in the sentence.
iv) Case (capital C): The licensing that a DP requires: Nominative
is found on subjects (specifier of finite T). Accusative is found
on objects (complement to V).
v) The Case Filter: All DPs must be marked with Case.
vi) Passives: A particular verb form where the external argument (often
the agent or experiencer) is suppressed and the theme appears
in subject position. The movement of the theme is also an instance
of DP movement.
vii) The Morphology ofPassives: The suffix -en:
a) absorbs a verb's external theta role
b) absorbs a verb's ability to assign accusative Case to its sister.
Chapter 10:DP Movement 307
viii) Burzio's Generalization: The idea that if a verb does not assign
an external argument (i.e., is passive or unaccusative), then it can't
assign accusative case.
ix) Unaccusatives: Inherently passive verbs like arrive.
Further Reading
Baker,Mark, KyleJohnson, and Ian Roberts(1989) Passive arguments raised.
Linguistic Inquiry 20, 219-51.
Burzio, Luigi (1986) Italian Syntax. Dordrecht: Reidel.
Chomsky, Noam (1995) The Minimalist Program. Cambridge: MITPress.
Jaeggli, Osvaldo (1986)Passive. Linguistic Inquiry 17,587-622.
Perlmutter, David and Paul Postal (1984) The 1-Advancement Exclusiveness
Law. In David Perlmutter and Carol Rosen (eds.), Studies in Relational
Grammar. Chicago: University of Chicago Press, pp. 81-125.
Sportiche, Dominique (1988) A theory of floating quantifiers and its
corollaries for constituent structure. Linguistic Inquiry 19,425-49.
General Problem Sets
1. Haitian Creole
[Data Analysis and CriticalThinking; Intermediate]
In the text, we suggested that DP movement leaves what is called a trace (t)
at the D-structure position of the DP. In English, you can't hear this trace.
Now consider the following data from Haitian Creole. (Data from Deprez 1992.)
a) Sanble Jan pati.
seems John left
"It seems that John left."
b) Jan sanble li pati.
John seems he leave
"John seems he to have left."
c) Man sanble pati.
Questions:
1) How does this data support the idea that raising constructions involve
movement from the lower clause to the higher clause, and the movement
leaves a trace?
308 Movement
2) Is sentence (b) a violation of the theta criterion? How might we make
sure that it isn't?
2. Arizona Tewa
[Data Analysis; Basic]
The following data is from ArizonaTewa (Data from Kroskrity 1985):
a) h§'i sen ne'i 'enu mankhw|di.
that man this boy 3.3.hit
"That man hit this boy."
b) ne'i 'enu h§'i sen-di '6khwfdi.
This boy that man-DAT 3.PASS.hit
"This boy was hit by that man."
c) na:bf kwiy6 h§'i p'o mansunt'6.
my woman that water 3.3.drink
"My wife will drink that water."
d) h§'i p'o nasunti.
that water 3.PASS.drunk
"That water was drunk."
1) Determine the X-bar parameter settings for Tewa.
2) Draw trees for (a) and (c). Assume Tewa is an affix lowering language.
3) Describe in your own words the differences between (a) and (b) and
between (c) and (d)
4) Draw the trees of (b) and (d) showing all the movements.
3. Persian Accusative Case5
[Data Analysis and Critical Thinking; Intermediate]
In the text above, we claimed that some verbs have an accusative feature
[acc] that must get checked by a complement DP. In English, we only
see the realization of this feature on pronouns. This question focuses
on the [acc] feature in Persian.
Background: Persian is an SOV language. There is no Case distinction
among Persian pronouns. For example, the pronoun man "I, me" doesn't
change whether it is a subject, object of a preposition or possessor
(see (a) below), (iii) shows that possessors are linked to head nouns
with a vowel glossed as ez (for Ezafe).
' Thanks to Jila Ghomeshi for contributing this problem set.
Chapter 10:DP Movement 309
a) i) Man ruzname xarid-am.
I newspaper bought-1SG
"I bought a newspaper."
ii) Sima az man ruzname xast.
Sima from me newspaper wanted.3sg
"Sima wanted a newspaper from me."
iii) Ruzname-ye man inja-st.
newspaper-EZ me here-is
"My newspaper is here."
Hypothesis: It looks like the clitic -ra (which is pronounced as -o or -ro,
depending on whether the preceding word ends in a vowel or not)
is the realization of the [acc] feature based on examples like the following:
b) i) Man
1
"1 saw
jian-o
Jian-RA
Jian."
didam.
saw.lSG
ii) *Man
1
jian
Jian
did-am.
saw-1SG
c) i) Jian man-o did.
Jian I-rA saw.3SG
"Jian saw me."
ii) *Jian
Jian
man
1
did.
saw.3SG
d) i) Jian in ketab-io xarid.
Jian this book-RA bought.3sG
"Jian bought this book."
ii) *Jian in ketab xarid.
Jian this book bought.3sG
One possible analysis is that Persian verbs have an [acc] feature that gets
checked by -ra. That is, -ri contributes the [acc] feature to the DP that can
be used to check the feature of the verb.
The problem: Not all direct objects show up with -ra. Yet we don't want to say
that the ones without -ra don't check the [acc] feature of the verb.
e) i) Jian ye ketab xund.
Jian a book read.3sG
"Jian read a book."
310 Movement
ii) Jian ketab-o xund.
Jian book-RA read.3SG
"Jian read the book."
f) i) Man se-ta qalam xarid-am.
I three pen bought-1SG
"I bought three pens."
ii) Man se-ta qalam-o xarid-am.
I three pen-RA bought-1SG
"I bought the three pens."
g) i) Jian pirhan xarid.
Jian shirt bought.3sG
"Jian bought a shirt."
ii) Jian pirhan-o xarid.
Jian shirt-RA bought.3sG
"Jian bought the shirt."
Suggest a solution to this problem.
4. Turkish
[Data Analysis, Critical Thinking; Advanced]
In this chapter, we argued that the reason DPs raise from embedded clauses
to main clauses is that they cannot get Case in the embedded clause.
Consider the following data from Turkish. What problems does this cause
for our theory? Is there a simple way to explain why Turkish nouns raise?
(Data from Moore 1998.)
a) Biz sut iciyoruz.
we milk drink
"We are drinking milk."
b) Biz;sana [cp t, sut ictik ] gibi gorunduk.
We you-DAT milk drank like appear
"We appear to you [Cp drunk milk]."
5. IMPERSONALS IN UKRAINIAN, KANNADA, AND IRISH
[Data Analysis; Intermediate]
(The Ukrainian and Kannada data are taken from Goodall 1993. The Ukrainian data
originally comes from Sobin 1985. The Kannada data is originally from Cole
and Sridhar 1976. The Irish data is slightly modified from Stenson 1989.)
Chapter10: DPMovement 311
Many languages contain a construction similar to the passive called the
impersonal passive. Consider the following data from Ukrainian, Kannada,
and Irish. Pay careful attention to the Case marking on the various nouns.
a) Cerkvu bulo zbudovano v 1640roc'i. Ukrainian
Church-ACC was built in 1640 year
"The Church was built in the year 1640."
b) Rama-nannu kollalayitu. Kannada
Ramma-Acc kill.PASS
"Rama was killed."
c) Buaileadh iad sagcluife deireanach. Irish
beat.PAST.PASS them.ACC in the game last
"They were beaten in the last game."
What is the difference between these impersonal passive constructions and
more traditional passives of English? Suggest a parameter that will account
for the difference between languages like Ukrainian, Kannada, and Irish
and languages like English. (Hint: the parameter will have to do with the way
the passive morphology works.)
6. English
[Application of Skills; Basic to Advanced]
Draw the D-structure trees for the following sentences. Be explicit about what
transformations derived the S-structure tree (if any). Recall that we have
the following transformations: Expletiveinsertion, DP movement (both raising
and passive), affix lowering, verb movement, T -> C movement,
and ofo-support/insertion. Annotate the D-structure tree with arrows to show
the derivation of the S-structure.
a) Marie is likely to leave the store.
b) The money was hidden in the drawer.
c) Donny is likely to have been kissed by the puppy.
d) It seems that Sonny loves Cher.
e) Has the rice been eaten?
312 Movement
7. English Ungrammatical Sentences
[Application of Skills; Basic to Intermediate]
Explain why the following sentences are ungrammatical. Some sentences
may have more than one problem with them.
a) *ltseems Sonny to love Cher.
b) *Bill was bitten the dog.
c) *Donny is likely that left.
8. Unaccusatives and Passives
[Critical Thinking; Advanced]
In a textbox above, we mentioned the existence of a class of verbs that are
essentially inherently passive. These are called unaccusatives. A surprising
property of unaccusative verbs is that they don't allow passivization.6
(Data from Perlmutter and Postal 1984.)
a) The Shah slept in a bed.
b) The bed was slept in by the Shah.
c) Dust fell on the bed. unaccusative
d) *The bed was fallen on by the dust, unaccusative
Similar effects are seen inthe following Dutch sentences. Sentence (e) is not
unaccusative (we call these "unergatives"), while sentence (f) is. Both these
sentences are impersonal passives. English doesn't have this construction,
so they are difficult to translate into English.
e) In de zomer wordt er hier vaak gezwommen.
"In the summer, there is swimming here."
f) In de zomer wordt er hier vaak verdronken.
"In the summer, there is drowning here."
Your task is to figure out why passives of unaccusatives (like c, d, and f) are
not allowed. The following data might help you:
g) Bill was hit by the baseball.
h) *Was been hit by Bill by the baseball, (passive of a passive)
i) Billgave Sue the book.
j) Sue was given the book by Bill.
k) *The book was been given by Bill by Sue. (passive of a passive)
6Strictly speaking, thedatain (a-d) donotinvolve passivization, since theNP thatis
moved comes from inside a PP. The technical term for these constructions is pseudopassivization.
The differences between pseudo-passivization and passivization are
not relevant to this problem set.
Chapter 10: DPMovement 313
9. Icelandic Quirky Case
[Data Analysis and Critical Thinking; Advanced]
In Icelandic, some verbs assign irregular case marking to particular
arguments. For example, the verb hjalpad "help" assigns dative case
to its theme argument. (Data from Zaenen, Maling, and Thrainsson 1985.)
a) Eg hjalpadi honum.
I helped him-DAT
"I helped him."
This kind of irregular case marking is called quirky Case and it seems
to be linked to the theta grid of the particular predicate. The dative case
is obligatorily linked with whatever noun takes the theme role:
hfilpad "heir/
Aaent
DP
Theme
DP
i K
Dative Case
Now consider the following data from Icelandic DP movement constructions.
b) Honumk var hjalpad fchim-DAT
was helped
"He was helped."
c) £g tel honumk [tk hafa verid hjalpad tk i profinu].
I believe him-DAT have been helped in the-exam
"I believe him [to have been helped in the exam]."
What problem does this cause for the theory of DP movement we have
proposed above? Can you think of a solution? (A number of possibilities
exist, be creative.)
Challenge Problem Sets
Challenge Problem Set 1: Middles and Passives
[Critical Thinking; Challenge]
Middles are English constructions that are little bit like passives. An example
of an active/middle pair is seen below:
314 Movement
a) I cut the soft bread.
b) The soft bread cuts easily.
In (b), the theme appears in the subject position. One analysis of this order
has the theme undergoing DP movementto subject position.
Consider now the following triplet of sentences. The first sentence is
called a middle, the second an active, and the third a causative.
c) The boat sank. middle
d) The torpedo sank the boat. active
e) The captain sank the boat (with a torpedo), causative
Part 1: Describe the relationship between the active, middle, and causative
in terms of their theta grids.
Part 2: Now consider the passives of sentences (c-e). Why should sentence
(f) be ungrammatical, but (g)and (h) grammatical?
f) *Was sunk (by the boat).
(also *Itwas sunk bythe boat,where itis an expletive)
g) The boat was sunk by the torpedo.
h) The boat was sunk bythe captain (with a torpedo).
Challenge Problem Set 2: Passives and Double Objects
[Critical Thinking; Challenge]
(For more information on the phenomenon discussed in this problemset, see Larson
1988.) English has two constructions that surface with ditranstive verbs.
One is called the prepositional construction, the other the double object
construction:7
a) Isent a book to Louis. prepositional
b) Isent Louis a book. double object
It is possible to make passives out of these constructions. But some
additional restrictions on how passives work are needed. Consider the
following data and posit a restriction on DP movement in passives to account
for the ill-formedness of the ungrammatical sentences. Pay careful attention
to sentence (g).
Thereis a great deal of literature that triesto derive the double object construction
from the prepositional construction using NP movement (see for example Larson
1988). The relationship between the two constructions is not relevantto the question
in this problem set, but is an interesting puzzle in and of itself.
Chapter 10:DP Movement 315
c) A book was sent to Louis.
d) *Louis was sent a book to.
e) *To Louis was senta book.8
f) Louis was sent a book.
g) *A book was sent Louis.
Challenge Problem Set 3: Two Kinds of Raising
[Critical Thinking; Challenge]
In the text, we proposed that subjects of non-finite clauses can raise to
the subject position of finite clauses in sentences like (a):
a) Johni seems [ti to have left].
This kind of raising is sometimes called subject-to-subject raising. Now
consider the following sentence:
b) Bill wants John to leave.
This sentence should be ungrammatical, because to is a non-finite T,
so can't assign Case to John. One hypothesis that has been proposed
to account for this says there is also a process of subject-to-object raising:
leave
How does the following data support this analysis?
8Thismaybe marginally acceptable in poetic or flowery speech.Assumefor
the purposes of this problem set that this is ungrammatical.
316 Movement
c) John wants Bill to leave.
d) John wants him to leave.
e) John believes him to have been at the game.
f) ?Johnibelieves himselfj to have been at the game.
g) *Johnibelieves himjto have been at the game.
h) He is believed (by John) to have been at the game.
^_
VWz-movement
0. Introduction
Inchapter10, we looked at DPsthat wereappearing in positions where they
didn't get theta roles. Instead, the DPsurfaced in a derived position. That is,
they were moved from the position where they got a theta role to a position
where they could get Case. The trigger for this movement was the
requirement that DPs check their Casefeature, as Case can only be assigned
in specific structural positions. In this chapter, we turn to another kind
of phrasal movement, one where DPs already have Case. DPs (and other
phrases) can move for a different reason to form what are called
-wh-questions.
There are several different kinds of questions, only two of which we are
concerned with in this book. The first kind is the familiar yes/no question
that we looked at in the chapter on head movement:
1) a) Are you going to eat that bagel?
b) Do you drink whisky?
c) Have you seen the spectrograph for that phoneme?
The answers to these questions cannot be other than yes, no, maybe or J don't
know. Any other response sounds strange:
318 Movement
1') a') #Pizza/ /yes
b') #Scotch/ /no
c') #Syntactic tree/ /no
The other kind of question is called a ic/z-question. These questions take
their name from the fact that the words that introduce them (mostly) begin
with the letters <wh> in English: who/whom, what, when, where, why, which,
and how. The responses to these kind of questions cannot be yes or no.
Instead they must be informative phrases.
2) a) When did you do your syntax homework? #yes / /yesterday
b) What are you eating? #no/ /a bagel
c) How is Louise feeling? #yes/ /much better
How these questions are formed is the focus of this chapter.
Wlw and Whom
In traditional prescriptive grammar, there are two ^//-phrases that refer to
people: who and whom. Who is used when the ze//-phrase originates in
subject position and gets nominative Case. Whom is the accusative version.
In most spoken dialects of Canadian and American English this
distinction no longer exists, and who is used in all environments. For the
sake of clarity, I use wlio(m) to indicate that the wh-phrase originated in
object position, but you should note that from a descriptive point of view
who is perfectlyacceptable in objectposition for most speakers today.
1. Movement in Wh-questions
If you look closely at a statement and a related (('//-question, you'll see
that the wh-phiase appears in a position far away from the position where
its theta role is assigned. Take for example:
3) a) Becky bought the syntax book,
b) What did Becky buy?
The verb buy in English takes two theta roles, an external agent
and an internal theme. In sentence (3a), Becky is the agent, and the syntaxbook
is the theme. In sentence (3b) Becky is the agent and what is the theme. In the
first sentence, the theme is the object of the verb, in the second the theme is
at the beginning of the clause. The situation becomes even more mysterious
when we look at sentences like (4):
4) What did Stacy say Beckybought?
Chapter 11: Wh-movement 319
In this sentence what is still the theme of bought, yet it appears way up at
the beginning of the main clause. This would appear to be a violation
of the locality constraint on theta role assignment introduced in chapter 9.
The situation becomes murkier still when we look at Case. Recall that
accusative Case is assigned when a DP is the sister to a V:
5) Matt [Vp kissed herACC],
Butin zy/Vquestions the accusativeform (likewhom) is not a sister to V:
6) WhomACC did Matt kiss?
Soit appears as if not only are thesew/Vphrases not in their theta positions,
but they aren't in their Case positions either.
Given what we've seen in the previous two chapters, this looks like
another case of movement - this one with different triggers again. Let's start
with the issue of where w/Vphrases move to. One position that we've had
for a while, but have not yet used, is the specifier of CP. This is the place
ly/i-phrases move to:
Notice that what moves here is anentire phrase. You canseethis ifyou look at
slightly more complex w/i-questions:
8) a) [To whom] did Michael give the book?
b) [Which book] did Michael give to Millie?
When you move an entire phrase, it cannot be an instance of head-to-head
movement (by definition), so this must be movement to a position other
320 Movement
than a head, in this case we have the empty specifier of CP. The element
that is moved can be a DP, a PP, or an AdvP.
The movement to the specifier of CP accounts for another fact about
the word order of w/i-questions: they also involve T —» C movement
(in main clauses):
9) a) Who(m) are you meeting?
b) *Who(m) you are meeting?
The wh-phrase appears to the left of the auxiliary in C. This means that the
wh-phrase must raise to a position higher than C. The only position available
to us is the specifier of CP:
The fact that w/Vmovement is to the CP specifier position can also be
seen in languages that allow both a wh-phrase and an overt complementizer,
such as Irish:
11) Cad aL td sa seomra?
What Q-wh is in-the room
"What is in the room?"
In Irish, the wh-phrase cad "what" appears to the left of the complementizer
aL, supporting the idea that the wh-phrase is in the specifier of CP, the only
position available to it. A similar fact is seen in Bavarian German
(Bayer 1984):
12) I woass ned wann dass da Xavea kummt.
I know not when that the Xavea comes
"I don't know when Xavea is coming."
Chapter 11: W/Vmovement 321
In English the only thing allowed to appear in C is an inverted auxiliary,
other complementizers are not:
13) a) *I asked what that she kissed?
b) *I asked what whether she kissed?
This follows simply from the assumption that the only complementizer
that is compatible with w/i-movement in English is null. In other languages
this complementizer has phonological content (e.g., Irish a1 or Bavarian
German dass).
Let's now consider the possible motivations for w/Vmovement.
In chapter 9, we triggered T -> C movement with a [+Q] feature that was
part of the complementizer. DP movement, in chapter 10, was triggered by a
Case feature. We can do the same thing, here, for wJi-questions, by proposing
a feature that triggers w/i-movement. Let's call this feature [+WH]. It resides
in the C of a w/Vsentence. In some languages (such as Irish), there are special
forms of complementizers that represent these features:
14) [-Q,-WH] go
[+Q,-WH] an
[+Q, +WH] a1
You get the go complementizer when the sentence is not a yes/no or
wh-qaestion. You get the an complementizer in yes/no questions and aL in
w/Vquestions. Theform ofthe complementizer isdependentupon the features
it contains (McCloskey 1979).
A wh-phrase moves to the specifier of CP to be near the [+WH] feature.
Another way to phrase this is to say that wh-phrases move into the specifier
of CP to check the wh-feature, just like we moved DPs to the specifier of TP
to check a [nom] Case feature in chapter 10. We can formalize w/i-movement
the following way:
15) Wh-movement
Move a wh-phrase to the specifier of CP to check a [+WH] feature in C.
Let's do a derivation for the following sentence:
16) Who(m) did Matt kiss?
The D-structure of this sentence will look like (17):
1 [nom]
-ed
Movement
VP
DP,
A
(nomJ
Matt V,[ACC]
kiss
V
DP,|+WH, ACC|
A
whom
Matt and whom both get their theta roles in these D-structure positions.
Whom also gets its Case in this base position. Three other operations apply:
There is DP movement of you to the specifier of TP to check the [nom]
feature. There is insertion of do to support the -ed and we get T -» C
movement to fill the null [+Q] complementizer:
(+WH, ACC)
W/7-movement applies to check the [+WH] feature:
Chapter 11: W/i-movement
This results in the surface string in (20a) and the tree in (20b).
20) a) Who(m) did Matt kiss?
b) CP
DP[/WH, /ACC] C'
A
whom did+C[/Ci /WH] TP
323
324 Movement
Traces and Wnnwa-contraction
You may have noticed that I have been marking the position that
movement occurs from with a t (coindexed with the word it replaces).
The t here stand for "trace." Later in this chapter we'll see that traces are
required to block certain kinds of illicit movement. But an important
question is whether there is any reality behind the notion "trace." This is
especially important in a theory like Generative Grammar which claims
psychological underpinnings. Finding evidence for something that is
not pronounced is remarkably difficult. However, there is some
straightforward evidence for traces. First a little background: In spoken
varieties of English (both standard and non-standard), function words
often contract with nearby words. One such contraction takes non-finite T
(to) and contracts it with a preceding verb like want:
i) I want to leave -» I wanna leave.
This phenomenon is called wanna-contraction. Now consider what
happens when you have (('//-movement and 7(w//m-contraction going on at
the same time. Wanna-contraction is permitted when the zy/z-movement
applies to an object:
ii) Who(m)i do you wanna kiss £,.
But look what happens when you try to do zwzzz/zrt-contraction, when whmovement
targets the subject:
iii) Whoi do you want tt to kiss the puppy?
iv) *Whodo you wanna kiss the puppy?
English speakers have very strong judgments that zw?/z/zz7-contraction is
impossible when the subject is zy/z-questioned. Why should this be the
case? If we have traces, the explanation is simple: the trace intervenes
between the to and the verb. It blocks the strict adjacency between
the verb and the to, thus blocking contraction:
v) WhOj do you want tjto kiss the puppy?
The theory of traces, provides a nice explanation for this fact. For
an alternate view see Pullum (1997).
Now let's do a more complicated example. This one involves DP
movement, zy/z-movement and T —> C movement:
21) Who was kissed?
The D-structure of this sentence is given in (22). This sentence is a passive.
Chapter 11: W/i-movement 325
22) CP
l+WH, NOM]
Who is the only argument in the sentence (a theme) and it starts out as a
complement to the verb. The suffix -en lowers to the verb and absorbs both
the agent theta role and the verb's accusative case. Also the auxiliary raises
to T to support the tense:
23)
326 Movement
Since -en has taken the verb's [ACC] feature, who cannot get Case in its base
position. It must move to the specifier of TP to check nominative Case:
DP movement
Once this DP has checked its Case features, it can move on to the specifier
of CP for zy/z-feature checking (|A|). The auxiliary also undergoes T —» C
movement ([B]) for the [+Q] feature:
tbe ten kissed f,
These two movements are "vacuous" in that who and was are in the order
who was ... both before movements (A] and |B] and after them. However,
the feature checking requirements force us to claim that both movements
occur anyway.
Chapter 11: W/i-movement 327
You now have enough information to try Challenge Problem Set 1
W/z-movement can also apply across clauses. Next, we'll do a derivation
of a sentence where the wh-phrase moves from an embedded clause
to the specifier of a main clause CP.
26) Who(m) do you think Jim kissed?
The D-structure of this sentence will look like (27). In this tree, who(m)
is theta marked by the verb kiss, and gets its internal theme theta role
in the object position of that verb:
27) CP
C
C[+q, +wh| TP
PRES
T[NOMl VP
-ed y^\
DP|NOM] V1
A s~
Jim V[+ACCl *-'t (+VVH, +ACC]
kiss
A
who
The T suffix -ed needs support, so it undergoes affix lowering and
the present tense feature on the higher T that requires do-insertion. The [+Q]
feature on the C also triggers T —> C movement. The DP Jim moves from
328 Movement
the specifier ofthe embedded VP to the specifier ofthe embedded TP for EPP
and Case reasons. The DP you does the same in the higher clause.
28)
C
Q+O, +WH1 TP
T—>C\ . _^C>^-"'DP movement
[pres]
do V_ DP V
insertion A /~^\
you V CP
think y^
C
C[_q_wh| TP
DP movement
affix lowering
ccj Ul [+Wn# +acc]
Finally we have zy/z-movement. For reasons that will become clear towards
the end of this chapter, we do this movement in two hops, moving first to
the specifier of the embedded CP, then on to the higher CP to check
that Cs [+WH] feature.
Chapter 11:W/i-movement 329
29) CP
This derives the correct S-structure, where the wh-phrase is in initial
position:
Movement
M-Q.-WH1 TP
DP[/NOM|T'
A / \
Jim f„rf VP
V
V
kissed
lwh
Let's do one more derivation, this time a sentence like the one above, but
where thewh-phrase stopsin thespecifier position oftheembedded CP:
31) I wonder who Jim kissed.
The main difference between this sentence and (26) lies in the nature of the
main verb. In (26) the verb was think, that subcategorizes for a CP headed
by C(_Q/ _WH1 (32a). The verbwonder1 differs in that it subcategorizes for a CP
headed by C,_q +wh), that is the embedded clause has w/i-movement
in it (32b):
1Wehave to assume that there is another verb wonder, found in sentencessuch as
/ wonder ifBill left thatselects for a CPheaded byC(+a.WH).
Chapter 11: W/z-movement
32) a) think
Agent
DP
Proposition
CP|_o_WHl
b) wonder
Agent
DP
Proposition
CP|_o.+WHI
331
The D-structure for (31) is given in (33); it differs minimally from (27)
only in the main verb and the feature structures of the two complementizers.
The DPs all get their theta roles in these D-structure positions.
33)
PRES
I V CP
wonder y^^
C
^-|-Q,+WH] *• *
Jim V|+ACC]
kiss
l+WH, +ACC]
332 Movement
Just as in the previousexample, who gets its case in its baseposition; the -ed
and null PRES affixes lower to the Vs and the two agent DPs (you and Jim)
move to their respectivespecifiers of TP to get case.
34)
ACC] L/l [+WH, +ACC]
Finally we have movement of the wh-phrase. Notice that it only goes to
the specifier of the embedded CP. This is because of the featural content
of the Cs. The embedded CP is [+WH], the main clause CP is [-WH].
Chapter 11: W//-movement 333
No T —> C Movement in Embedded Clauses
In the main text, we've noticed that zy/z-movement and T —> C movement
often go hand in hand. One surprising fact about English is that this is not
true of embedded zc/z-questions. When a zy/z-question is embedded the
subject does not invert with the auxiliary (i.e.,no T —> C movement):
i) I wonder what he has done?
ii) *I wonder what has he done?
In other words, in embedded clauses there is no C[+Q/ +WH]- One simple
explanation for this is that theta grids simply can't contain C1+Qj.
35) CP
C
C|-Q,-WH| '
tDP v
V CP
wonder+0PRES
M/acc] twh
kissed.
You now have enough information to try General Problem Sets 1-3
334 Movement
2. Islands
W/z-movement isn't entirely free. There are constraints on what categories
you can move out of (the categories that contain the wh-phrase). Compare
thefollowing twosentences, oneofwhich haszy/z-movement out ofa simple
complement CP (36a). The other (36b) moves a wh-phrase out of a clause
that is contained inside a DP:
36) a) What; did Bill claim [CP thathe read f, in the syntaxbook]?
b) *Whatj did Bill make [DP the claim [CP that he read f, in the syntax
book]]?
In sentence (36a), we see that zy/z-movement out of a complement clause
is grammatical, but movement out of a CP that is dominated by a DP
is horrible (36b). Thisphenomenon, first observed by Ross (1967), has come
to be known as the Complex DP Island phenomenon. The word island
here is meant to be iconic. Islands are places you can't get off of (without
specialmeans likea plane), they are surrounded by water, so you are limited
in where you can move: You can only move about within the confines of
the island. Islands in syntax are the same. Youcannot move outof an island,
but you can move around within it. DPs are islands.
37) *Whatj did Bill make [DP theclaim [CP thathe read f, in the syntaxbook]] ?
Complex DP Island
The example in (37) involves a CP that is a complement to the N head.
The same effect is found when the CP is a relative clause (i.e., an adjunct
to the N):
38) *[Whichcake]jdid you see [DP the man [CP who baked f,]]
We can characterize this phenomenon with the following descriptive
statement:
39) The Complex DP constraint
* wh, [...[„, - *,...]...]
You now have enough information to try General Problem Set4
There are many other kinds of islands. One of the most important
is called a wh-island. First, observe that it is possible to move a wh-phrase
to the specifier of an embedded CP, when the C is [+WH]:
40) I wonder [CP whatsC,.^ +WH, [•„> John bought f, with the $20bill]].
Chapter 11: W/z-movement 335
It is also possible to move (another) wh-phrase to the specifier of the main
CP:
41) [CP Howk do [tp you think [Johnbought the sweater ffc]]]?
However, look at what happens when you try to do both (move one
wh-phrase totheembedded specifier, and theother tothemainCPspecifier):
42) *[CP Howk do [TP you wonder [CP whatj[TP Johnbought r, tk]]]]7
This sentence is wildly ungrammatical - even though we have only
done two otherwise legitimate transformations. Now this isn't a constraint
on having two zy/z-phrases in a sentence. Two zy/i-phrases are perfectly
acceptable in other contexts:"
43) a) How do you think John bought what?
b) I wonder what John bought how.
It seems then to be a constraint on moving both of them. The same kind
of example is seen in (44a)and (44b):
44) a) I wonder [cpwhatj [TP John kissed f,]].
b) [CP Whok did [TP you think [TP tk kissed the gorilla]]]]?
Movement of either the subject (44b) or the object (44a) to the specifiers
of the CPs is acceptable. However, movement of both results in terrible
ungrammaticality:
45) *[CP1 Whok did [TP you wonder [CP2 whatj [w tk kissed r,]]]]?
The central intuition underlying an account of these facts is that once you
move a wh-phrase into the specifier of a CP, then that CP becomes an island
for further extraction:
46) I asked [CP whatj John kissed t,] zy/z-island
Movement out of this zy/z-island results in ungrammaticality. We can express
this with the following descriptive statement:
47) Wli-island Constraint
*whi[...[CPwhk[...f, ...]...]...]
This constraint simply says that you cannot do zy/z-movement
(in the schematic in (47) this is represented by the whj and the coindexed ti)
and skip around a CP that has another wh-phrase (whk) in its specifier.
2If you have trouble with this judgment, try stressing the word what in (41) and how
in (42).
336 Movement
We're going to discuss this particular island in much greater detail in
the next section.
Subjects are another kind of island. Consider the following sentence;
it has a CP in its subject position (48a). When you try to zy/z-move
the zy/z-equivalent to several rioters (who in 48b), the sentence becomes
ungrammatical.
48) a) [TP [CPThat thepolice wouldarrestseveral rioters] wasa certainty],
b) *WhOj was [TP [CP thatthepolice wouldarrestf,] twas a certainty]?
This is called the subject condition:
49) The Subject Condition.
*whj ...[TP[CP...tj...]T...]
We have one final island to consider. Consider a conjunction like
that in (50a). Here we have twoDPs conjoined with eachother. W/z-moving
either of these DPs results in ungrammaticality (50b and c).
50) a) I liked Mary and John.
b) *WhOj did you like Mary and f,?
c) *WhOj did you like f, and John?
The same is true if you try to do w/t-movementfrom within another structure
that is conjoined, such as a conjoined VPin (51):
51) a) I [VP ate some popcorn] and [VP drank some soda].
b) *Whatj did you eat f, and drink some soda?
c) *Whatj did you eat some popcorn and drink ti?
The island condition that governs these situations is called the Coordinate
Structure Constraint:
52) Coordinate Structure Constraint:
*whj ... [XP [XP ... r,... ] conj [XP ... ]] ...
or *whj... [XP [XP ... ] conj [XP ... f,... ]] ...
or *whj...[XP [XP...]conj f,] ...
or *whi...[XPf,conj[XP...]] ...
We thus have four environments out of which zy/z-movement cannot occur:
Complex DPs, Subjects, CPs with a wh-word in their specifier and out
of a conjunct in a coordination. These environments are the subject of
much research in syntactic theory right now. In the next section, we will look
at one possible explanation for some of these island effects (although
the account does not account for all of them by any means). This account
refers to a constraint known as the Minimal Link Condition.
Chapter 11: W/z-movement 337
3. The Minimal Link Condition
3.1 Wh-islands and the Minimal Link Condition
Island phenomena beg for explanation. Let's consider w/z-islands in some
detail. As we noticed above, in questions with multiple zy/i-phrases,
the movement of eachwh-phrase is allowed independently ofeach other:
53) a) I wonder [CP whatj [TP John kissed f,]].
b) [Cp Whok did [TP you think [^ tk kissed the gorilla]]]]?
However, when you combine the movements the sentence becomes nearly
incomprehensible:
54) *[Cpi Whok did [TP you wonder [CP2 whatj[tp tk kissed f,]]]]?
{ * I I
Recall from earlier discussion that syntactic operations like to either be local
(forexample, anaphors must be bound within their clause - a local relation;
similarly theta roles are assigned within their VP - another local relation)
or create localities (for example, DPs move to get close or local to their case
assigner; affixes move to get close or adjacent to theirhost; and zy/i-phrases
move to get near a [+WH]). In the nextchapter, we will consider a unified
approach tomovement that triestocapture at leastthelastset ofcases. What
is important here is that our grammarsseem to like relations that are close.
With this intuition in mind, think about zy/z-islands. Wft-phrases move to get
in the specifier of a CI+WH1 so let's hypothesize that there is a further
restriction, movement must always target the nearest potential position.
This is another localitycondition: the Minimal Link Condition (MLC):
55) Minimal Link Condition (MLC) (intuitive version)
Move to the closest potential landing site.
In (54) there are two CPs, but both the wh-phrases start in the embedded
clause. This means that for both zy/z-phrases the embedded CP (CP2)
is the closestpotential landing site. Here's an abbreviated D-structure of (54),
the potential landing sites for the wh-phrase are underlined:
56) [CPi _ C|+WH] [Tp you 0|r>RESl wonder [CP2 _ Cl+wh, [TP who -ed kiss what]]]]?
If we start by moving what to this position, we can check off what's
zy/z-feature, and this move meets the minimal link condition because
the movement has targeted the closest potential landing site:
57) [CP1 _ CI+WH] [Tp you 0,PRES) wonder [CP2 whatk C|/wh, [^ who -ed kiss rJ]]]?
f. I
338 Movement
Now the other wh-phrase in this sentence has to check3 its zy/i-features,
but the closest potential position is filled by what. While movement to
the specifier of CP1 would allow it to check its [+WH] feature, this would
beaviolation of theMLC, as themovement skips thefirstpotential position:
58) [CPi _ C|+WH] [Tp you0|PRES, wonder [CP2 whatk CKwh, [TP who-ed kiss tJ]]]?
t
*firstpotential position
Even though the specifier of CP2 is filled with what, it still counts
as the closest position. Butsince it isoccupied, who can't movethere, so there
is no way for the [+WH] feature to be checked. Notice that it doesn't matter
whatorder weapply theoperations in. Ifwemove who first, stopping offin
the specifier of CP2 (thus meeting theMLC), then thatspecifier is occupied
by the trace, so there is no place for the what to move to:
59) [Cpi whOj C,+WH, [TP you 0[PRES, wonder[CP2 f, C|/wh) [^ f, -ed kisswhat]]]]?
i I* I
? <The
MLC thusexplains theungrammaticality ofzy/z-islands: When youhave
multiple zy/z-phrases that require movement, movement of at least one
ofthemwill beblocked bytheMLC because theclosest potential landing site
will be occupied by the other.
Before moving on to look at the utility of the MLC in other domains,
it's worth noting how the grammaticality of a sentence like (60) is derived
when we have a constraint like the MLC. This sentence looks like we
have non-local movement. Theword who gets its theta role in the embedded
clause yet ends up in the specifier of the higher CP:
60) [CP WhOj do you think [CP2 [TP r, kissed the gorilla]]]?
This should be a violation of the MLC, since the wh-phrase ends up in
the specifier of a CP that is higher up in the tree. You will recall from the tree
in (29) that we said that zy/z-movement that crosses clause boundaries does so
in two hops: First to the specifier of the lower CP, then on to the higher CP.
61) [CP Who; do you think [CP2 f, [wf, kissed the gorilla]]]?
l lf_J
3 We return to sentences like / zoonder ivho loves what where there appears to be
no movement in section 4.
Chapter 11:W/z-movement 339
We now have an explanation for why this is the case: The MLC requires
that all movement be local. In order to maintain this locality the movement
happens in two hops. This phenomenon is called successive cyclic movement.
In the problem sets section of this chapter there is a question
on Irish (General Problem Set 6) that shows a morphological correspondence
to successive cyclic movement.
You now have enough information to try General Problem Sets 5 &6and Challenge
Problem Sets 2 &3
It should be noted before we go on that the MLC does not explain
all island effects, only the zy/z-islands. Accounting for other island types
is a hot topic of research in syntax today.
3.2 The MLC in DP Movement and Head Movement
The MLC has usage above and beyond that of zy/z-islands. We can use it
to account for a variety of other locality effects with DP and head movement
too.
The verbs is likely and seem both have empty subjectpositions and allow
the subject-to-subject raising variant ofDP movement.
62) a) Markjis likely [tjto have left],
b) Markj seems [tj to have left].
Consider now whathappens whenyouembed one ofthesein the other. It is
only possible for DP movement to occur to the lower of the two case
positions. (63a) shows the D-structure. (63b) shows the grammatical
S-structure where the DP shifts to the lower position, and expletive insertion
applies at the higher position. (63c) and (63d) show ungrammatical forms,
where the DP has shifted to the higher of the two positions. This kind of
movementis ungrammatical, whether or not (63d vs.63c) expletive insertion
applies in the lower specifier of TP.
63) a) seems [that is likely [Mark to have left]].
b) It seems [that Markj is likely [tjto have left]].
c) *Markjseems that is likely [tj to have left].
d) *Markj seems that it is likely [tjto have left].
When two Case positions are available movement has to target the closer
(lower) one. The MLC condition explains these facts as well.
340 Movement
64) *[tp Markj seems that [•„, it is likely [tj to have left]]].
f £ I
firstpotential nominative position
Sentences (63c and d) are ungrammatical because the movement goes
beyond the closest potential position, which is occupied by an expletive
into a higher position. It is as if the expletive creates a "case-island"
for the purposes of DP movement.
A similar effect is seen in head-movement. Recall from chapter 9
that in French we have both T -> C movement and V -> T movement.
These two operations had tohappen intandem ifwehave a yes/no question
with a main verb and no auxiliary:
65) a) [CP C^Qifyp vous T(prcs, [VP tvous mangez despommes]]]
you eat of.the apples
b) [Cp Cj+qjItp vous T[pres, [VP rraus mangez despommes]]]
4 It I
T->C V->T
c) Mangez vous des pommes?
"Do you eat apples?"
(65a) is roughly theD-structure of thesentence (with the subject DP moved
from the specifier of TP for case). We have two instances of head movement
(65b). FirstV -> T applies, then T -> C to give the C content, this results in
thesurface string in (65c). Consider what would happen if the intermediate
T position were occupied by an auxiliary (66a) and we tried to do headmovement
oftheverb aroundit(66b), thiswouldgive us theungrammatical
string in (66c):4
66) a) [CP Cj+QjfTp vous avez[VP tV0IIS mange* despommes]]]
you have eaten of.theapples
b) [cp Q+Q] [tp vous avez[VP tvous mange" despommes]]]
-t | I
first potential position
4 For ease of reading these diagrams I'm leaving out the stacked verb analysis
of auxiliary constructions and just generating auxiliaries in T; this does not change
the MLC effect.
Chapter11: W/z-movement 341
c) *Mang£ vous avez des pommes
eaten you have
The ungrammaticality of (66c) follows easily, the V -> C movement
has skipped the intermediate T(occupied by avez). This Tposition isthe first
potential landing site for the verb. This is thus a violation of the Minimal
Link Condition.5
You now have enough information totry Challenge Problem Set 4
The MLC then doesn't only explain zy/z-islands, it also extends
to other locality restrictions on movement, such as the requirement that
DP movement always target the closest case position and the requirement
thathead movement notskipintervening heads. Notice that in each ofthese
cases what counts as a "potential landing site" is different. The same basic
constraint holds, but the conditions for each type of movement are different.
This discovery was made by Luigi Rizzi in his famous book Relativized
Minimality (1990). Thereare two things that are vague about our preliminary
definition of the MLC above: the precise definition of "closest" and the
precise definition of "potential landing site." Nevertheless for most people
ourpreliminary definition should be intuitive and sufficient. For people who
prefer more precision here is a definition that is more formal and defines
closest in terms of c-command and relativizes the landing site to the type
of movement:
67) The MinimalLink Condition (formal)
Movement of some item (5 can target some position a of type 8 if and
only if
i) a c-commands (3
ii) there is no y, also of type 8, such that a c-commands y, and y
c-commands (3.
iii) 8 is defined as:
(a) A head if (3 = a head.
(b) Thespecifier ofTPif p= a DPwithan unchecked [NOM].
(c) Thecomplement of theVif|3 = a DPwithan unchecked [acc].
(d) The specifier of CP if [5 = zy/i-phrase with an unchecked [+WH]
feature.
Conditions (i) and (ii) of this version of the MLC firm up what is meant
by"closest." Here closeness isdefined in terms ofc-command, where there
5 This instance of the MLC is sometimes known by an older name: The Head
Movement Constraint (HMC), which was invented by Travis (1984); the HMC was
the inspiration behind the MLC.
342 Movement
can be no intermediate c-commanding landing site of the relevant type
intervening between the item that's moving and its landing site. The way
that this is defined is actually nearly identical to the definition ofgovernment
given way back in chapter 4. In fact, the MLC is often assumed to be the
successor tothe government relation.6 Condition (iii) defines the kind ortype
of landing site relative to each type of movement.
4. Echo Questions (Wh-in-situ) in English
You may have noticed in the previous section that the MLC, when applied
to zy/z-movement, in essence prevents any clause from having two moved
zy/z-phrases. Does this mean that a clause can't have two w/z-phrases at all?
Obviously not:
68) Who loves who(m)?
This sentence is grammatical, even though the second w/z-phrase
does not move. This is a phenomenon called ztz/z-in-situ (from the Latin in situ
"inplace"). We also see wh-in-sihi insentences with only one wh-phrase:
69) Shelly loves who? (If this isnotgrammatical for you, stress who.)
We might ask why the wh-phrase in (69) and the second who in (68)
don't move to check their [+WH] features. The answer is simple, these are
not zy/i-questions and these apparent zy/z-phrases are [-WH]. These are
echo-questions. Echo questions are notrequestsfornew information; instead
they are requests for confirmation of something someone has heard.
Consider sentence (68) in a conversational context:
70) Daniel: Hey, Ijustheard thatShelly loves Ferdinand.
Andrew: Shelly loves who?
Daniel: Youheard me,Shellyloves Ferdinand.
It's clear from this snippet of discussion that Andrew is incredulous about
Shelly loving Ferdinand and wasasking for confirmation of what he heard.
This is very different from a request for information. There are two relevant
properties of echo-questions (i) they don't involve movement, but (ii)
they do involve a special intonation, where the in-situ wh-phrase is stressed.
Since echo questions don'tinvolve movement, they aren't going tobesubject
6The term "govern" had two usages in the1980s and early 1990s: one as a structural
relation (asdefined in chapter 4); the otherwasas a constraint on the grammar. It is
this latter usagethat the MLC replaces.
Chapter 11: WZz-movement 343
to the MLC (explaining the grammaticality of(68) and otherexamples like it).
While yes/no questions and zy/z-questions have some kind of syntactic
licensing echo questions seem to be licensed by intonation and stress. In this
regard they are similar to intonational questions that don't have subject/aux
inversion such as (71) (where the rising curve is meant to indicate raised
intonation and the italics represent stress on the words):
Fred saw a spaceship in the linguistics lounge?
Note that this question again has a subtly different meaning from the one
with subject/aux inversion and do-support (Did Fred see a space ship in the
linguistics lounge?). The sentence with subject/aux inversion isa request for
information. (71) is anexpression of doubt and a request for confirmation.7
How such phonological licensing is encoded into the syntactic tree is very
controversial. One solution is that, like zy/z-questions and yes/no questions,
echo questions andintonational questions involve a special complementizer.
We can indicate this as C,+INI). The [+1NT] feature doesn't trigger any
movements, but it instructs the phonology to put a rising intonation curve
on the clause that follows the C. The stress has to do with contrastive focus.
In English contrastively focused material is stressed.
W/z-in-situ in English (and in closely related languages) seems to be
largely limited toecho-question contexts.8 However, zy/z-in-situ is the norm
for real zy/z-questions in languages such as Chinese and Japanese.
These languages appear to have no zy/z-movement at all. This will be
a major topic of the next chapter.
5. CONCLUSION
In thischapter, we looked at a third kind ofmovement transformation: Whmovement.
This process targets zy/z-phrases and moves them to thespecifier
7 In fact, many people distinguish these using special punctuation in emails
and informal writing. A yes/no question is indicated with a simple question mark
(?), but an echo or intonational question is usually indicated with a combination of
two (or more)question marks with an exclamation mark in between (?!?).
8 There are contexts however, which involve a multiple zy/z-question interpretation.
For example, a police officer might ask a suspect "Wlien did you convince your
accomplice tohide the money where?" or a parent might ask a teenager "Wlio were you
with where?". These kinds of in-situ questions will receive the same treatment as
in-situ zy/z-questions in languages such as Japanese and Chinese to be discussed in
the next chapter.
344 Movement
ofCPs. This movement istriggered by the presence ofa [+WH] feature inC.
Movement is always from a Case position to the specifier of CP.
W/z-movement is not totally unrestricted; there is a locality constraint
on themovement: theMLC. Movement mustbelocal, wherelocal isdefined
in terms of closest potential landing site. We saw further that the MLC
mightbe extended to other types ofmovement.
In the next chapter, we're going to continue this trend and look
at movement processes in general and the similarities between them, as well
as briefly delve into the interaction between the syntax and the formal
interpretation (semantics) of the sentence.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Wh-movement: Move a wh-phrase to the specifier of CP to check
a zy/t-feature in C
ii) Island: Aphrase that contains (dominates) the zy/z-phrase, and that
you may not move out of.
iii) The Complex DPConstraint: Vh, [ ... [DP ... tt... ] ...]
iv) Wh-island Constraint: *whj [ ... [cf whk [ ... f, ...]...]...]
v) The Subject Condition: *whs ... [Tl, [CI,... tj... ]T ...]
vi) Coordinate Structure Constraint:
*wh,...Ixp [xp- */.»] conj [xp...]]...
or *whi...[XP [xp ...] conj[XP ...*,-...]] ...
or *whj... [xp [XP...]conj /,] ...
or *whi... [xpf/conjlxp...]] ...
vii) Minimal Link Condition (MLC) (intuitive version): Move to the closest
potential landing site.
viii) The Minimal LinkCondition (formal):
Movement ofsome item|3 cantargetsome position a oftype8 ifand
only if
i) a c-commands |3
ii) there is no y, also of type 5, such that a c-commands y, and y
c-commands (3.
iii) 5 is defined as
(a) A head if (3 = a head
Chapter 11: W/z-movement 345
(b) The specifier ofTP if(3 =a DP withan unchecked [nom].
(c) Thecomplement of theVif(3 = a DPwith an unchecked
| ACC].
(d) The specifier of CP if [3 = wh-phrase with an unchecked
[+WH] feature.
ix) Wh-in-situ: when a wh-phrase does not move.
x) Echo-Questions and Intonational Questions: Question forms
that are licensed by the phonology (intonation and stress) and not
by the syntax, although they may involve a special C.
Further Reading
Baltin, Mark (1981) Strictbounding. In C. L. Baker and JohnMcCarthy (eds.)
The Logical Problem of Language Acquisition. Cambridge: MIT Press.
pp. 257-95.
Cheng, Lisa (1997) On the Typology of Wh-Questions. New York: Garland
Press.
Chomsky, Noam (1977) On zy/j-movement. In Peter Culicover, Thomas
Wasow, and Adrian Akmajian (eds.), Formal Syntax. New York:
Academic Press, pp. 71-132.
Chomsky, Noam(1986a) Barriers. Cambridge: MIT Press.
Chomsky, Noam and Howard Lasnik (1978) A remark on contraction.
Linguistic Inquiry 9, 268-74.
Cinque, Guglielmo (1981) Types ofA' Dependencies. Cambridge: MIT Press.
Koopman, Hilda (1984) The Syntax ofVerbs. Dordrecht: Foris.
Lasnik, Howard and Mamoru Saito (1984) On the nature of proper
government. Linguistic Inquiry 15,235-89.
Lightfoot, David (1976) Trace theory and twice moved NPs. Linguistic Inquiry
7, 559-82.
Manzini, Maria Rita (1992) Locality: A Theory and Some of its Empirical
Consequences. Cambridge: MIT Press.
Richards, Norvin (1997) What Moves Where When in Which Language?
Ph.D. dissertation, MIT.
Rizzi, Luigi (1990) Relativized Minimality. Cambridge: MITPress.
Ross, J. R. (Haj) (1967) Constraintson Variables in Syntax. Ph.D.dissertation,
MIT.
Travis, Lisa deMena (1984) Parameters and Effects of Word Order Variation.
Ph.D. dissertation, MIT.
346 Movement
General Problem Sets
1. English Movement Sentences
[Application of Skills;Basic to Advanced]
For each of the following sentences, give the D-structure tree and annotate
it with arrows indicating what transformations have applied. The sentences
may have head-to-head movement, c/o-insertion, expletive insertion,
DP movement and wh-movement.
a) What is bothering you?
b) Who has seen my snorkel?
c) How was the plot discovered by the authorities?
d) Which animals appear to have lost their collars?
e) What did Jean think was likely to have been stolen?
f) Car sales have surprised the stockbrokers.
g) Have you seen my model airplane collection?
h) Can you find the lightbulb store?
i) John was bitten by an advertising executive.
j) It is likely that Tami will leave New York.
k) Tami is likely to leave New York.
I) Itseems that Susy was mugged.
m) Susy seems to have been mugged.
n) What did you buy at the supermarket?
o) Iasked what Beth bought at the supermarket.
p) What is itlikely for Beth to have bought at the supermarket?
(Treat the PP for Beth as appearing the specifier ofthe embedded TP.)
q) What is likely to have been bought at the supermarket?
2. Binding Theory
[Critical Thinking; Basic]
In chapter 5, you were asked why the sentence below causes a problem
for the binding theory. Remind yourself of your answer, and then explain
how the model of grammar we have proposed in this chapter accounts
for this fact.
Which pictures of himself does John despise?
Chapter 11: W/z-movement 347
3. Bindingand Scrambling9
[Critical Thinking; Intermediate/Advanced]
You should complete Problem Set 2 before attempting this problem set.
Modern Persian has a kind of movement often called scrambling. Your task
in this problem set is to figure out whether scrambling is DP movement,
head-to-head movement or wn-movement. The Persian word hamdiga
means "each other" and is an anaphor. Assume that anaphors are subject to
the binding theory ofchapter 4, and that they must be in argument positions
to be bound. Sentence (a) shows the basic order. Sentences (b) and (c)
show the surface word order after scrambling has applied. The scrambled
sentences mean almost exactly the same thing as (a), hab stands
for "habitual." Recall that j/.k means that the sentence is okay with the DP
having the index -t but not with the index k.
a) Mo'allem-ak fekr mi-kon-an [cp ke [T- [vp bachche-haj
teacher-PL thought HAB-do-3PL that child-PL
[VP aks - a -ye hamdigaj/*k - ro be modir neshun dad-an]]]].
picture-PL-EZ each other - ra to principal sign gave-3PL
"The teachersk think that the children showed [each other's]^ pictures to
the principal."
b) Mo'allem-ak [aks-a-ye hamdigaj/.k-ro]m fekr mi-kon-an [Cp ke [r [vp
[bachche-haj ] [Vptm be modir neshun dad-an]]]].
c) [Aks-a-ye hamdigaj/-k-ro]m mo'allem-aj fekr mi-kon-an [Cp ke [T [vp
bachche-hsii [Vptm be modir neshun dad-an]]]].
4. Picture DPs
[Critical Thinking; Intermediate/Advanced]
Why is the grammaticality of the following sentence surprising? Does
the theory we have presented in this chapter predict this to be acceptable?
What constraint should this sentence violate.
Who(m) did you see a picture of?
5. LOCALITY
[Data Analysis; Basic]
Why is the following sentence ungrammatical?
*WhOj did [tpGeorgetry to find out [CP what, [tptj wanted tj]]]?
1This problem set was contributed by Simin Karimi.
348 Movement
Draw a tree showing the exact problem with thissentence. Be precise about
what constraint rules it out.
6. Irish10
[Data Analysis; Advanced]
Irish has a number of different complementizer forms. In declarative clauses
(statements), it uses the complementizer go/gur. As discussed in the text
above, when there is a question, this complementizer switches to the whform
a . (The idea behind this problem set is taken from McCloskey 1979.)
a) Ceapann tu go bhuailfidh an piobaire an t-amhran
think you that play.FUT the piper the song
"You think that the piper will play the song."
b) Caide aLcheapann til aL bhuailfidh an piobaire?
What wh think you wh play.FUT the piper
"What do you think the piper will play?"
Note carefully the number of aL complementizers in sentence (b). (b)
provides evidence that wh-phrases stop off in intermediate specifiers of CP
(for MLC reasons). Explain why. You need to make the assumption that the
complementizer aL only shows up when a wh-phrase has atone point shown
up in its specifier.
Challenge Problem Sets
Challenge Problem Set 1: Who A te the Pizza ?
[Critical Thinking; Challenge]
In the text we suggested that subject questions involving an auxiliary
have vacuous movement of both the wh-phrase (to the specifier of CP) and
the auxiliary (to C[+Q. +WH]), even though that leaves the subject wh-phrase
and auxiliary in the same order they'd be in if they'd stayed in the TP:
Thisproblem set wassuggested byan anonymous Blackwell reviewer.
Chapter 11: Wh-movement 349
a)
We suggested this was the case so that we could maintain the feature
checking requirements hold here too.
Consider however the following sentence. Notice that the tense
morphology is not realized as did, but on the main verb ate.
b) Who ate the pizza?
Question 1: Explain why this is an argument against the vacuous movement
shown in (a).
Question 2; If there is no vacuous movement, what then are we to make of
the complementizer insubject questions? Is it[+Q, +WH]? Is there some way
to explain why the complementizers with subject w/?-questions are different
than complementizers with object and adjunct wh-questions (which do show
both wn-movement and T -> C movement)?
Challenge Problem Set 2: Irish
[Data Analysis; Challenge]
Some dialects of English allow a kind of wn-construction, where the base
position of the wn-phrase is filled by a resumptive pronoun (the idea behind
this problem set is taken from McCloskey 1991):
This is the bookj that the police are arresting everyone who reads it.
In Modern Irish, this kind of construction is very common. Modern Irish
has two different wft-complementizers (notice that these are not wh-phrases,
which go in the specifier of CP, these are complementizers): aL, aN.
The complementizer aL is found in sentences like (a). Sentence (i) shows
a simple sentence without wh-movement using the non-wn-complementizer
go. Sentences (ii) and (iii) show two possible forms of the question,
(ii) has the question moved only to an intermediate CP specifier,
(iii) has the wh-phrase moved to the topmost specifier.
350 Movement
a) i) Bionn fios agat igconai [CP go bhuailfidh an piobaire an t-amhran].
be.HAB know at.2.s always that play.FUT the piper the song
"You always knowthat the bagpiper will playthe song."
ii) Bionn fios agat igconai [CP caidei aL bhuailfidh anpiobaire ti].
be.HAB know at.2.s always what comp play.FUT the piper ti
"Youalways know what the bagpiper will play."
iii) [CpCaidejaL[Tpbhionnfios agat igconai [CP ft aL bhuailfidh an piobaire ft]]]?
What compbe.HAB know at.2.s always tj compplay.FUT the piperft
"Whatdo you always know the piper will play?"
Now the distribution of the complementizer aN seems to be linked to the
presence ofa resumptive pronoun. Considerthe (ii) sentences in(b) and (c).
Both show resumptive pronouns and the complementizer aN:
b) i) Bionn fios agat igconai [CP caidej aL bhuailfidh anpiobaire ft].
be.HAB knowat.2.s always what* compplay.FUT the piper ft
"Youalways know what the bagpiper will play."
ii) [CPCen Piobairej aN [tp mbionn fios agat igconai [CPcaidej aL bhuailfidh sty tj]]]?
Which pipercompbe.HAB know at.2.s always whatj compplay.FUT he
"Which bagpiper do you always knowwhat he will play?"
c) i) Ta mathair an fhir san otharlann.
Be.PRES mother the man.GEN in.the hospital
"The man's mother is in the hospital."
ii) C6j aN bhfuil a/ mhathairsan otharlann?
who comp be.PRES his mother in.the hospital
"Who is (his) mother in the hospital?"
The a complementizer and the resumptive pronouns are boldfaced in the
above examples. Where precisely does the aw-resumptive strategy appear?
In what syntactic environment do you get this construction?
Challenge Problem Set 3: Serbian/Croatian/Bosnian11
[Critical Thinking; Challenge]
In this chapter, we have proposed that a wh-phrase appears in the specifier
of CP, to check a [WH] feature. Our analysis of locality conditions requires
that only one wh-phrase can appear in the specifier of CP. Consider
the following data from Serbian/Croatian/Bosnian. Assume that this language
is SVO at D-structure. (Data from Boskovic 1997 as cited in Lasnik 1999a.)
11 Thisproblemset was contributed bySimin Karimi.
a)
b)
c)
d)
Chapter 11: W/i-movement
Ko sta gdje kupuje?
who what where buys
"Who buys what where?"
*Ko kupuje sta gdje?
*Ko sta kupuje gdje?
*Ko gdje kupuje sta?
351
What problems does this data raise for our analysis? Can you see a way
around these problems?
Challenge Problem Set 4: French Negation
[Critical Thinking; Challenge]
We've argued that in French the verb raises to T, and that T raises to C in
yes/no questions. Further in this chapter, we've argued that head movement
is subject to the Minimal Link Condition. In previous chapters we've treated
the French word pas as the head of NegP as in (a). Consider an alternative
where pas isn't the specifier of NegP but is an adjunct to the VP as in (b):
a) CP
TP
b) CP
TP
NegP
Neg
pas
VP
NegP
A
pas
Neg'
VP
V
V
Keep in mind the restriction in the MLC that the moving element can't skip
potential landing sites that c-command it. Recall that V -> T movement jumps
over the word pas:
352 Movement
c) Je n'aime pas tv des pommes.
I like not of.the apples.
"I don't like apples"
Does this argue for the analysis in (a) or the analysis in (b)? Explain your
answer.
A Unified Theory of
Movement
0. Introduction
At the beginning of this book, we looked at rules that generate the basic
phrase structure ofhuman syntax. These rulesgenerated trees that represent
hierarchical structure and constituency. These trees have particular
mathematical properties that we investigated in chapters 4 and 5. In chapter
6, we saw that stipulated phrase structure rules missed some very basic
generalizations, and refined them into X-bar phrase structure rules. The
X-bar rules, being very general, allow us (informed by parameter settings) to
generate a wide variety of trees, and capture structural differences between
heads, complements, adjuncts, and specifiers. In chapter 7, we extended
the rules to various clause types, complementizers, and DPs. In chapter 8,
we saw that, in fact, the X-bar rules actually generated too many structures,
and that we had to constrain their power. The device we use to limit them is
a semantic one: the thematic properties of predicates (stored in the lexicon)
and the theta criterion. What results from the output of the X-bar schema
and the lexicon is called D-structure. The theta criterion holds of
D-structures, as do the binding conditions. In chapters 9, 10, and 11, we saw
a variety of cases where lexical items either could not be generated where
they surfaced by X-bar theory (e.g., head-adjunct-complement ordering
in French) or appeared in positions other than the ones predicted
by theta theory. We developed a new kind of rule: the movement rule
or transformation, which moves items around from their base position in
354 Movement
the D-structure to the actual position they appear in on the surface.
There are three movement transformations: Head-to-head movement (T -> C
and V —» T), DP movement, and lo/i-movement. In each of these cases
movement occurs because it has to. Each movement has a trigger or
motivation. Heads move to fill empty [Q] features or to take an inflectional
suffix. DPs move to check case features. W/i-phrases move to be near the
[WH] feature. We also posited two insertion transformations: Do-support
and expletive insertion. The output of the transformations is called
S-structure, which is itself subject to several constraints: the Case filter, the
EPP and the Subjacency Constraint.The model (flowchart) of the grammar
looks like (1).
1)
The Lexicon
The Computational Component
"^ X-bar rules
i
D-structure
(constrained by theTheta Criterion
and binding conditions)
Transformational rules
DP movement
head movement
wh-movement
expletive insertion
do-insertion
I
S-structure
(constrained by theEPP,Casefilter
and MLC)
'the base'
Grammaticality Judgments
This straightforward model can generate a large number of sentence types.
It is also a system with explanatory adequacy, which makes specific
predictions about how a child goes about acquiring a language
(via parameters).
This is a fairly complete and simple system, but we might ask if the
system can be made even simpler. Are we missing any generalizations here?
Chapter 12: A Unified Theory of Movement 355
Recent work in syntactic theory that answers this question suggests
we might unify the different types of movement into a single rule type
with a slight reorganization of the architecture of the grammar.
The Minimalist Program
The system of grammar described in this chapter is a very cursory look at
some of the principles underlying the most recent version of generative
grammar: The Minimalist Program (Chomsky 1993,1995).The Minimalist
Program is motivated not only by the search for explanatory adequacy but
also for a certain level of formal simplicity and elegance. What is outlined
here is by no means a complete picture, but is meant to give you a taste of
what current work is striving towards.
1. Move
In this book we've proposed the following motivations for transformations:
2) a) Head movement toget a suffixorfill null f+Q]
b) DP movement to check casefeatures [nom] or [ACC]
c) W//-movement to check a[+WH]feature
Notice that while there are significant differences between the motivations
for the various types of movement, there is one overwhelming similarity.
The movements all occur so that one item can appear near another. In the
case of head movement the V orT head needs to appear as part of the same
word as the place it moves to. With zy/z-movement, the <<>/;-phrase needs to be
near the [WH| feature, just as DP movement occurs so the DP can check its
Case feature with T or V. All the motivations for movement then seem to be
locality constraints. That is, two items must be near or local to one another.
This is a trend that we've seen before in previous chapters.
If all the movement types are motivated by locality, then there isn't
really a significant difference between the rule types. Perhaps we can unify
them into a single rule: Move. Move says simply "move something"
(bul only if you have to):
3) Move(very informal version)
Move something somewhere.
Now of course, this is a bit vague and we'll have to sharpen it up in some
way. In particular, we will want to constrain the rule so there isn't just
random movement all over the sentence. So the next step is to formulate
356 Movement
a constraint that motivates and forces this transformation to apply (in all
the different circumstances).
Let's take ly/j-movement as our paradigm case. In lu/r-movement the
wh-phrase moves to the specifier of CP so as to be local with a [WH] feature.
Another way to think of this, as we suggested in chapter 11, is to say that
both the wh-phrase and the complementizerhave a [+WH] feature, and they
need to compare them, or check them. Checking of features can only occur
in a local configuration. In this case we have what is called a specifier-head
configuration for reasons that should be obvious from the following
partial tree.
4)
Specifier-head checking configuration
The constraint that forces this movement to occur is called the Principle of
Full Interpretation (Chomsky 1993,1995).
5) FullInterpretation (Fl)
Features must be checked in a local configuration.
6) Local Configuration (preliminary)
[WH] features: Specifier-head configuration.
We can extend this to the other cases of movement too. As we suggested in
chapter 10, imagine that Case is not simply an ending, but is also a feature.
A subject DP bears a [nom] Case feature. Imagine also that the heads
of the phrases that assign case (T and V) also bear this feature (although
they don't show it morphologically). We can thus reduce the Case filter
to full interpretation: Nominative Case is feature checking like that in (7)
and accusative Case is like that in (8):
7)
8)
Chapter 12:A Unified Theory of Movement
TP
Specifier-head checking configuration
V
V ,Acc DP,ACC]
357
Head-complement checking configuration
Case and Agreement
The notion that T bears some kind of Case feature often troubles people,
since Case is an inherently nominal kind of inflection and T seems to be
associated with verbal material. One clever solution to this problem is to
claim that verbal items like T do in fact bear Case; we just call case on
verbs "agreement." In fact, cross-linguistically, there does seem to be
some kind of correlation between the kinds of agreement markers that are
found on verbs and the case marking on the subjects. Languages with
ergative/absolutive case marking systems often also have
ergative/absolutive agreement. So we could claim that [NOM] when on a
noun is case, but when on T or a V is agreement, thus at least partly
motivating the structure in (7) and (8).
Finally, we can extend this to the head movement cases. Instead of claiming
that verbs move to pick up inflectional suffixes in V —> T movement, let's
claim that both the V and the T head bear some kind of abstract inflectional
features (e.g., [±past]). This allows us to capture the behavior of verbs
with null T morphology as well as that of ones with affixes. When the verb
and T check these features against one another then the suffix representing
that tense feature (or agreement feature) is allowed to surface on the verb.
The local configuration in this setting is within the head itself (a relationship
which is called a head-head configuration):
358 Movement
Head/Head checking configuration
Similarly, both T and C bear a [+Q] feature, and they must be in a head-head
checking relationship:
Head/Head checking configuration
Local Configuration is thus defined in terms of features. The particular
configuration required is determined by which feature is being used. This is
very similar to the way in which we formally defined the conditions on
the Minimal Link Condition in the previous chapter, where the intervening
categories were relativized to the kind of feature being checked by the
element that is moving. (11) is a summary of these local configurations.
11) Local Configuration:
[WH], [NOM] features: Specifier-head configuration.
[ACC] features: Head-complement configuration.1
[PAST], etc., [Q] features: Head-head configuration.
With this in place we actually have a very elegant transformational system.
We have combined our three movement rules into one: Move and two
constraints: full interpretation and the Minimal Link Condition. In previous
chapters, we've already argued that constituent structure iscreated by a few
verylimited phrase structurerules,which areconstrained by thetheta criterion
1In the next chapter we will claim that [ACC] is actually checked in a specifier-head
configuration like [nom], this will allow us to create a phrase structure system
that accounts for double object verbs and dative constructions.
Chapter 12:A Unified Theory of Movement 359
and the lexical entries of categories. Computationally speaking, this is a
surprisingly simple system of grammar.
2. Explaining Cross-linguistic Differences
The system outlined above in section 1, is simple and elegant. It
does however, make the unfortunate prediction that all languages will have
exactly the same set of transformational rules (although they can still differ
in phrase structure, due to the parameters). This is clearly not the case.
English does not have V —> T movement. Many other languages lack passive
and raising. Still others appear to lack w/i-movement. Take the case of
Mandarin Chinese (data from Huang 1982; tone is omitted).
12) a) Ni xiang chi sheme?
you want eat what
"What do you want to eat?"
b) *Sheme ni xiang chi?
what you want eat
"What do you want to eat?"
c) Ni kanjian-le shei?
you see-ASP who
"Who did you see?"
d) *Shei ni kanjian-le?
who you see-ASP
"Who did you see?"
Chinese appears to have no zWj-movement. As we discussed in the
last chapter this is called wh-in-situ. The Chinese case differs from English,
however, in that these are not echo-questions. These are real w/j-questions.
As such they should have [+WH] features on their Cs and on the zy/i-phrases.
Why, then, is it the case that the unchecked [+WH] features on the
w/i-phrasesdon't violate Full Interpretation? They have not moved so they
are not in a local configuration with their C. Full Interpretation predicts
that (12a) should be ungrammatical and (12b) should be grammatical the
exact opposite of the facts.
Our solution to this problem is going to surprise you. We're going
to claim that in Chinese the 7t>/*-phrase does move, you just don't hear it!
This requires a refinement of our grammar model.
Ferdinand de Saussure (a linguist from the late nineteenth century)
observed that every linguisticexpression consists of two parts: the signifier
360 Movement
and the signified. For our purposes, this roughly corresponds
to the phonological or phoneticform of the sentence (abbreviated as PF) and
its semantic or logicalform (LF). We call these "forms" the interface levels,
because they represent the interface with the phonological system and with
theinterpretive system respectively. This means thatwhen we'recomputing
thegrammaticality ofa sentence we're really computing two distinct things:
its sound (for the purposes of a syntactician this means the sequence
of the words) and its meaning. To a certain degree these interfacelevels are
computed together, but they also diverge from one another. When we look
at the question cross-linguistically, we see that any particular PF order
of elements does not directly correspond to some specific meaning. For
example, the English sentence / sawthe man and the Irish sentence Chonaic me
an fear (literally Saw I the man) mean the same thing, but they have different
word orders. One way to represent this conundrum is by having two
separate levels in our model of grammar that correspond to these interface
levels. These levels represent the final products of our computation, so
they should appear at theend of the derivation. Thisgives us a more refined
model of the grammar than the one we saw in (1):
13) The Lexicon
overt movement
SPELLOUT
Judgments about Form
•> X-bar rules The Base
I
D-structure
•—>•
"A
Move &
V Insertion
Logical Form (LF)
Full Interpretation
Judgments about Meaning
Let me draw your attention to some of the important differences between
this model2 and the model we had in (1). First of all you'll note that there is
2This model isn't entirely an accurate representation of Chomsky's minimalist model.
I have retained the notion of D-structure here. In Chomsky's version,
Chapter 12:A Unified Theory of Movement 361
no S-structure. In the old model, S-structure was the level from which we
drew grammaticality judgments. In this new model it is the interface levels
PF and LF that give us well-formedness judgments (judgments about form)
and semantic judgments respectively. Some of the derivation applies
in tandem generating both the PF and the LF; these operations apply on the
"stem" of the upside down Y. Then there is a point at which the derivation
branches into operations that are purely about form and sound
and operations that are purely about meaning. This branching point is
called SPELLOUT (usually in all capital letters). After SPELLOUT, the derivation
proceeds along two distinct paths, one generating the PF, the other the LF.
Chomsky (1993) makes two important claims. First he claims that Full
Interpretation is a constraint that holds of sentences at LF. Second, he claims
that exactly the same operations that happen between D-structure and
SPELLOUT can also happen between SPELLOUT and LF (in (13) this is indicated
by means of the double line that extends from D-structure to SPELLOUT
and from SPELLOUT to LF). Move can apply anywhere along this double line.
At first this may seem counterintuitive, but the kinds of movement can
happen after SPELLOUT; that is, after the pronounced order has been created,
and the "form" portion of the sentence is sent off to PF. This is a mindbending
notion, but actually allows us to make the following remarkable
claim about cross-linguistic variation: Every instance of feature-checking
motivated movement happens in every single language. Why would we want
to do such a thing? Let us assume, not uncontroversially, that the kinds
of meaning determined by the syntax are universally held by all humans.
This is not to say that all humans have identical world-views or identical
perceptions of events etc.; we are not making such a strong claim here.
This is not a claim about cultural or personal interpretations. This is simply
a limited statement that all humans have a notion of what it means
to express a declaration, a yes/no question, a wh-question, a passive,
a sentence with raising, etc. These kinds of constructions, and relationships
such as constituency and the binding conditions do seem to be universal in
interpretation even if they do have different forms cross-linguistically.
the X-bar rules are replaced by an operation called Merge and there is no D-structure.
In Chomsky's model the constraints we've claimed to be D-structure constraints
(The Theta Criterion and the binding conditions) are handled differently. Getting rid
of D-structure involves some tricky argumentation that lies beyond the scope of
this textbook. Once you are finished reading this book have a look at some of
the suggested readings at the end that will take you onto this more advanced
material. This aside the diagram in (13)is a fair representation of how many linguists
working within the minimalist program is structured.
362 Movement
Let us call such basic interpretations universal semantics. If, as we
have hypothesized throughout this book, this basic semantic content
is determined by our X-bar phrase structure system (which creates
constituents) and the movement operations (which check to make sure
that there is a featural correspondence among the words in the constituent
structure), then universal semantics should be generated the same way
in every language. Yet it goes without saying, that every language has
different (yet narrowly limited) ways of expressing that universal semantics.
The Y model gives us a straightforward way of accounting for this. The
differences between languages are in when that movement occurs: before
SPELLOUT or after. Essentially there are two kinds of movement, movement
that happens between D-structure and SPELLOUT (called overt movement)
and movement that happens between SPELLOUT and LF (called covert
movement). Since covert movement happens after the branching off to the
PF(phonology) component, you simply can't hear it happen! The differences
between languages then, are in when they time specific instances
of the general operation Move. English times those instances involving [WH]
features overtly before SPELLOUT, Chinese times those same movements
covertly, after SPELLOUT. This can be simply encoded in a parameter:
14) Wh-parameter: Overt/Covert
(English sets at "Overt," Chinese sets at "Covert")
This parameter determines whether movement applies before S-structure
or after:
15) D-structure
^ English wh-movement happens here
1SPELLOUT^ «< theform of thesentencethat is sent to PF
r" ~*i *- Chinese wh-movement happens here
PF LF^
^ "what the sentence means
what the sentence sounds like
To make this clearer let us abstractly show each step of the derivation for
Mandarin Chinese and English (I'm abstracting away from do-insertion, etc.):
Chapter 12: A Unified Theory of Movement 363
16) English
D-structure you did see who?
LF
overt movement
SPELLOUT Who did you see f,?
Who did you see tj?
Mandarin Chinese
Ni kanjian-le shei?
you see what
Ni kanjianle shei?
covert movement
Shei ni kanjianle tj?
You'll notice that the order of elements in the LFs of this sentence is identical
in both languages, but the SPELLOUT form is different. You never hear
the LFs. You only hear the SPELLOUT forms. The LFs are identical because
the two sentences (in the two languages) mean the same thing.
This kind of analysis has a nice side effect. It also allows us to get rid
of the odd-man-out of transformations: Affix (T) lowering. This was the only
movement that we looked at that ever went downwards. It also appeared to
be in complementary distribution with V-movement. If a language has one,
then it doesn't have the other. This is suspicious; when we find items in
complementary distribution, it suggests they are really instances of the same
thing. With the system described above we can get rid of the affix lowering
account ofEnglish. English doesn't have affixlowering. Instead, it has V —¥ T
movement like any other language, only in English it is timed covertly,
so you never hear it.
17) French
D-structure Je souvent mange des pommes.
I often eat of.the apples
overt movement
SPELLOUT Je mangej souvent f,des pommes.
English
I often eat apples.
I often eat apples.
covert movement
LF Je mange; souvent tjdes pommes. I eat, often f,apples.
Again, these sentences mean the same thing, so they have identical LFs. The
word order in the two sentences is different, so the SPELLOUTs are different.
Again this is encoded in a parameter:
364 Movement
18) Verb movement parameter: Overt/Covert
(French sets at Overt; English sets at Covert)
In this view of things, differences in the word order of the world's languages
reduce to a matter of timing in the derivation.
You nowhave enough information totryGeneral Problem Sets1 &2 andChallenge
Problem Sets 1 & 2
3. Scope, Covert Movement, and the MLC
At first blush the whole notion of a movement you cannot hear seems pretty
suspicious (just like empty words that you can't hear seems suspicious).
There is some evidence it does exist, however. This evidence comes
from the behavior of w/i-questions in Japanese and from English sentences
with quantifiers.
3.2 MLC Effects in Wh-in-situ Languages
Let's compare two hypotheses. One is the covert movement hypothesis
proposed above in section 1. That is, in languages like Mandarin Chinese
and Japanese, w/i-phrases move just as in English, but they move covertly, so
that you don't hear the movement. The other hypothesis (on the surface less
suspicious) is simply that wh-phrases don't move in Mandarin and Japanese.
Consider the predictions made by these two hypotheses with respect
to island conditions and the MLC. Island effects are seen in English precisely
because there is movement. Too long a movement (violating the MLC)
causes the sentence to be ill-formed. When there is no movement, obviously,
no violations of the MLC will occur. Now compare our two hypotheses
about Japanese. The first hypothesis, according to which there is (covert)
movement, predicts that Japanese will show MLC violations. The other
hypothesis predicts that no violations will appear since there is no w/imovement.
The following sentence is only ungrammatical with the meaning
3Of courseifyou assume this, you can't claim that tensesuffixes are generated in T,
otherwise there is no way to get them onto the verb overtly. Chomsky (1993) gets
around this problem by hypothesizing that the features that trigger movement aren't
actual morphological items, like -ed, but instead are merely abstract features [+past].
The morphology on the verb is simply base generated on the verb in the lexicon and
the features associated with it are checked at LF.
Chapter 12:A Unified Theory of Movement 365
indicated (it is grammatical with other meanings, such as an echo-question
interpretation).
19) *[Nani-o doko-de katta ka] oboete-iru no?
what-ACC where-at bought Q remember Q
"What do you remember where we bought?"
If this data can be explained by the MLC, then we know that movement
occurs - even if we can't hear it - because this constraint is sensitive
to movement.
You nowhave enough information totr\f General Problem Set3 and
Challenge Problem Set3
3.2 English Quantifiers and Scope
If LF is truly a semantic construct, we expect to find some semantic
correlations to covert movement. One typical assumption about semantics
is that there are some similarities between it and the semantics expressed by
formal logic. With this in mind consider the following discussion of English
quantifier scope.
We call words like every and all universal quantifiers. In formal logic
these are represented by the symbol V. Words like some are existential
quantifiers and are represented by the symbol 3. In logic, quantifiers are said
to hold scope over constituents containing variables. Variables are items
that stand for arguments in the meaning of the sentence. The logical
representation of an expression like Everyone dances is:
20) Vx [x dances]
This means that for every (V) person you choose (represented by x),
then that person (x) dances. The quantifier V has scope over the variable x.
This is indicated by the brackets that surround [x dances]. One popular
interpretation of the logical relation of scope is that it corresponds directly
to the syntactic relation of c-command. So at LF the structure of (20) is (21):
21)
x dances
366 Movement
The quantifier phrase (QP) c-commands the TP, thus is holds scope over it.
The quantifier is said to bind the variable it holds scope over. (In logic, this
is represented as having the first x next to the V, and then the other x inside
the brackets.) In the notation we have used up to now in this book,
we could translate this using indexes:
22) Vx, [x, dances]
However, since the logical notation is more common for drawing LFs
we will use it here.)
An interesting phenomenon arises when we look at sentences with more
than one quantifier. The following sentence is ambiguous in English:
23) Everyone loves someone.
This can have two meanings. The first meaning is that for every person in
the world there is some other person who they love: Mary loves John, Bill
loves Susy, Rose loves Kim, ..., etc. The other meaning is that there is one
person in the world that everyone else in the world loves: Mary loves John,
Bill loves John, Rose loves John, ..., etc. Using a pseudo-logical paraphrase
we can represent these two meanings as (24). The actual logical
representations are given in (25):
24) a) For every person x, there is some person y, where x loves y.
b) For some person y, every person x loves y.
25) a) Vx(3y[x loves y])
b) 3y(Vx[x loves y])
In logical notation, you'll notice that the difference between the two
meanings lies in the order of the quantifiers, which reflects the embedding
of the structure. The universal quantifier in (25a) is said to have wide scope,
whereas in (25b) it has narrow scope.
In chapter 3, we claimed that if a sentence is ambiguous, then there are
two tree structures for the sentence. It would be nice if we could draw two
different trees that represent the two meanings of these sentences. As
mentioned above, one hypothesis about scope phenomena is that they reflect
c-command. That is, a quantifier has scope over everything it c-commands.4
Consider the meaning in (24a). We can easily see this scope when we draw
the simplified tree (QP stands for Quantifier Phrase):
4 A full discussion of the semantics and structure of scope lies well beyond the
purview of this book. See Heim and Kratzer (1998) for more on this complicated
topic.
27)
QP
someone
Chapter 12:A Unified Theory of Movement
TP
367
CP
C
QPi
Everyone T
TP
someone
Everyone here c-commands someone, so the wide scope reading for the
universal quantifier is derived. The narrow scope reading is more difficult, if
this hypothesis is correct, then in the tree for (24b)someone must c-command
everyone. The only way to get this fact is to move the quantifier. This kind
of movement is called quantifier raising or QR.
VP
V f,
loves
In this QR sentence, someone c-commands everyone, and so has scope over it.
This derives the narrow- scope reading for the universal quantifier.
Obviously, this quantifier raising analysis cannot be overt. The surface string
for this ambiguous sentence is everyone loves someone (not *someone everyone
loves). In order to get the second meaning for this sentence we need
to propose movement you can't hear, in other words, covert movement.
Covert movement thus has independent justification.
You now haveenough information to try Challenge Problem Set 4
368 Movement
4. CONCLUSION
In this chapter we made the big jump from three movement rules with
different but similar motivations to a single rule with a single motivation
(Full Interpretation). We also claimed that cross-linguistic variation in
movement, when we assume a universal semantics, requires that movement
can both be overt (before SPELLOUT) and covert (after SPELLOUT). The Y model
with Saussurian interface levels (LF and PF) allows this to occur. We also
looked very briefly at an example from quantifier scope that provides
independent support for the notion of covert movement.
Ideas, Rules, and Constraints Introduced in this Chapter
i) Move (very informal version): Move something somewhere.
ii) Full Interpretation: Features must be checked in a local
configuration.
iii) Local Configuration:
[WH], [NOM] features:Specifier-head configuration.
[ACC] features: Head-complement configuration.
[PaST] etc, [Q] features: Head-head configuration.
iv) Logical Form (LF): The semantic/interpretive system.
v) Phonetic Form (PF): The componentof grammar where word order
is expressed.
vi) SPELLOUT: The point at which the derivation divides into form (PF)
and meaning deriving structures (LF).
vii) Overt Movement: Movement between D-structure and SPELLOUT
(heard/pronounced movement).
viii) Covert Movement: Movement between SPELLOUT and LF (silent
movement).
ix) Wh-Paratneter: Overt/Covert.
x) Verb Movement Parameter: Overt/Covert.
xi) Universal Quantifier (V): Words such as every, each, all, any.
Identifies all the members of a set.
Chapter 12: A Unified Theory of Movement 369
xii) Existential Quantifier (3): Words like some, or a. Identifies at least one
member ofa set.
xiii) Scope: A quantifier's scope is the range of material it c-commands.
xiv) Wide vs. Narrow Scope.Wide scope is when one particular
quantifier c-commands another quantifier. Narrow scope
is the opposite.
xv) Quantifier Raising (QR): A covert instance of Move that moves
quantifiers.
Further Reading
Cheng, Lisa (1997) On the Typology ofWIi-questions. New York: Garland Press.
Chomsky, Noam (1991) Some notes on economy of derivation
and representation. In Robert Friedin (ed.), Principles and Parameters
in Comparative Grammar. Cambridge: MIT Press, pp. 417-54.
Chomsky, Noam (1993) A Minimalist program for linguistic theory. In
Kenneth L. Hale and Samuel J. Keyser (eds.), The View from Building 20.
Cambridge: MIT Press, pp. 1-52.
Chomsky, Noam (1995) The Minimalist Program. Cambridge: MIT Press.
Heim, Irene and Kratzer, Angelika (1998) Semantics in Generative Grammar.
Oxford: Blackwell.
Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of
Grammar. Ph.D. dissertation, MIT.
May, Robert (1985) Logical Form. Cambridge: MIT Press.
Saito, Mamoru and Howard Lasnik (1994) Move Alpha: Conditions on Its
Application and Output. Cambridge: MIT Press.
General Problem Sets
1. French
[Data Analysis; Basic]
Go back and look at all the French data in chapter 9 and determine if French
has overt or covert DP movement. Explain your answer.
370 Movement
2. Irish
[Data Analysis; Basic]
Go back and look at all the Irish data in chapters 9, 11 and determine
whether Irish has overt or covert wn-movement, overt or covert DP
movement and overt or covert head movement.
3. PF Movement
[Creative and Critical Thinking; Advanced]
In the text above, we proposed that some movement was covert. That is,
it happened between spellout and LF. This movement affects meaning, but
it doesn't affect how the sentence is pronounced. Can you think of any kind
of movement that might occur just on the PF branch of the model? That is,
are there any phenomena that affect only how the sentence is pronounced,
but not its meaning?
Challenge Problem Sets
Challenge Problem Set 1: Serbian/Croatian/Bosnian
[Data Analysis; Challenge]
Compare the data in challenge problem set 3 of chapter 11 with their English
Equivalents:
a) *Who what where buys?
b) Who buys what where?
c) *Who what buys where?
d) *Who where buys what?
Using the terms "covert movement" and "overt movement," explain the
difference in parameter setting between Serbo-Croatian and English.
Challenge Problem Set 2: Nepaliand Mongolian
[Data Analysis; Challenge]
Consider the following data from Nepali and Mongolian (data from Erin Good
and Amy LaCross respectively). Do these languages have overt
or covert wn-movement? How can you tell?
Nepali:
a) Timilai uu kahile aunche jasto-lagcha?
you she when coming think
"When do you think she is coming?"
b) Timi kahile aaunchau?
you when coming
"When are you going to come?"
Chapter 12:A Unified Theory of Movement 371
c) Ramie Sitale kun manche ayecha bhaneko sochecha?
Ram Sita which man came said think?
"Which man did Ram think that Sita said came?"
Mongolian:
d) Ek5 jamar hoi hix ve?
older-sister which-one food make C|+Q)
"Which food will the older sister make? "
Challenge Problem Set 3: Echo Questions in English
[Critical Thinking; Challenge]
Give an argument that echo questions in English involve no movement
at all (neither overt nor covert), and thus are very different from the covert
movement found in languages like Chinese and Japanese. Hint: The
evidence will come from the MLC: The following sentence might help you:
a) Who does John think loves what?
Challenge Problem Set 4: Scope of Negation
[Data Analysis, Creative and CriticalThinking; Challenge]
The following sentence is ambiguous:
a) The editor did not find many mistakes in the paper.
This can either mean
i) The editor isn't very good, and although there were many mistakes he
didn't find them,
ii) The editor searched thoroughly for mistakes, but the paper didn't have
many mistakes in it.
We can express these variations in meaning using scope. With meaning (i),
we have a situation where many has scope over negation (i.e. many
c-commands not (-•) or in logic: MANYx [-• find (editor, x)]). (That is, many
has wide scope). By contrast the narrow scope reading (ii), not c-commands
many (-> find (editor, MANYx).
Part 1: Draw the LF tree for each of the meanings. Keep in mind
that the word order for (i) will not be the same as the spellout order,
you are drawing the tree for the LF which includes movement that is covert.
Also assume that only quantifiers move; negation does not move.
Part 2: Consider now the passive form of this sentence:
b) Many mistakes were not found in the paper by the editor.
This sentence is not ambiguous. It only has one meaning: wide scope
for many (that is, the meaning in (i) above). This sentence can never have
the meaning in (ii) above. Why should this be the case? (Hint: ask yourself
372 Movement
if it is possible to create an LF with negation c-commanding many for (b).
Remember negation does not move.)
Part 3: English passives involve overt DP movement from the complement
of V to the specifier of TP. Explain why it is crucial that this movement be
overt, in explaining why many cannot have narrow scope (reading ii) at LF.
Optional really advanced question: Are the traces of DP movement
variables? How can you tell?
4-»
05
0Advanced
Topics
Expanded VPs
0. Introduction
In chapters 1-12, we sketched out the major theoretical tools assumed
by the majority of syntacticians operating in the principles and parameters
framework. The next few chapters take us away from these agreed-upon
areas, and focus on important material that is both more controversial
and more advanced. The discussion in these chapters is going to be more
open-ended. Do not expect a perfect answer or even an answer that can
beconsidered "right." Instead our discussion will consider somemajor lines
of thought about these more difficult topics.
1. The Problem of Ditransitive Verbs1
Inchapters 2and8, we discussed anumber ofditransitive verbs, such asput
of subcategory V(DP _ DP PP1, give of subcategory V(DP _ DP {DP/pP]y and tell
of subcategory V(DP _ DP (Cp/dp)]- m many cases the third argument of these
verbs seems to function like a complement, aside from the fact that it is
not immediately adjacent to the verb. Forexample, no adjunct may intervene
between the two post-verbal DP arguments of the verb give:
1Many thanks to Heidi Harley for allowing me access to her teaching materials
for the preparation of this chapter.
376 Advanced Topics
1) a) *Josh gave Clay carefullya book,
b) Joshgave Claya bookcarefully.
However, we know from our study ofX-bar theory in chapters 6 and 7, that
(i) we are only everallowed one complement and (ii) complements ofverbs
must be adjacent to that verb. This follows from the fact that X-bar theory
requires trees to be strictly binary branching. So the place to attach
these "second" complements is a mystery (2):
2)
DPI DP2
Even if we were to allow ternary branching as in (3), a different problem
emerges. In(3) thetwoDPs c-command one another, thuswe might expect
a symmetrybetween them in terms ofbindingrelationships.
3)
DPI DP2
Barss and Lasnik (1986) showed that there is actually a clear asymmetry
between these two DPs, as if the first one c-commanded the second, but not
vice versa. This can beseen inexamples (4a andb)where we have a typical
anaphor-antecedent relationship (principle A). Asyou can see the indirect
object Justin can bind a direct object anaphor; but thereverse isnotpossible.
If the structure of the sentence were (3), then the anaphor should be able
to appear in either position because the twoDPs symmetrically c-command
one another.
4) a) Briana showed Justinjhimselfj in the mirror,
b) *Briana showed himself;Justinj in the mirror.
These facts show that in terms of c-command relationships the two DPs
must be in a configuration like that in (5):
377
DP2
It's only in a configuration like (5) where DPI c-commands DP2, but DP2
does not c-command DPI. Of course the obvious question that arises then
lies in the nature of the category labeled Xin (5). We address this question
in the next section.
You now have enough information totry General Problem Set1
2. Light Verbs
Thereis reason to think that what appear tobe morphologically simpleverbs
in English may in fact be morphologically complex. In particular, we're
going to claim that verbs that assign an agentive theta role consist of two
parts; a verbal root and what we call a light verb. Light verbs areessentially
auxiliaries (inthat theyhead theirownVP) and theyare part ofthecomplex
thatsurfaces as a simple verb inEnglish. Tosee thisat work,weneed to turn
to other languages where the morphological complexity of verbs is more
obvious on the surface.
In Japanese (6a), Hiaki (6b), Malagasy (6c), we see that certain simple
verbs in English correspond to morphologically complex structures
inthese languages. These each consist of(at least) a verb root andsome other
morpheme that speakers report as either marking agentivity, or making
the root into a full verb. I have abbreviated this light verb element, following
Chomsky (1995) as "v". This is usually called little v to contrast it to
lexical verb roots which are of category big V.
6) a) Keiko-wa pizza-o ag-e-ta.
Keiko-TOP pizza-ACC rise-v-PAST
"Keiko raised the pizza."
b) Huan u'usit-ta ee-tua-k.
Juan child-ACC feel-v-PAST
"Juan teased the child."
378 Advanced Topics
c) M-an-sasa nylamba amin ny savony Rasoa.
PAST-v-wash the clothes with the soap Rasoa
"Rasoawashesthe clotheswith the soap."
A number ofscholars have suggested that even in English, agentive verbs
are bimorphemic. There is a verb root that indicates the lexical meaning
of the word and a lightverb thatroughly means"cause." Soa verb likeclean
really means something like "to cause tobe clean." Kratzer (1996) suggests
that agentive theta roles are notassigned bythe verb, butby the light verb2
contained within it. So if we take a verb like clean, this is really composed
of the little verb v meaning "cause" (CAUSE) which assigns the agent role,
and takes a VP as a complement (we will refer to the theta role assigned
to this VP as "predicate") and the lexicalroot /CLEAN, which takes the theme
as a complement:
7) CAUSE
Agent
DP
Predicate
VP
i i
/CLEAN
Theme
DP
K
So the tree for the sentence Ryan cleaned the window contains avP dominating
a VP. The subject DP moves for the usual case and EPP reasons. In order to
create the verb clean out of CAUSE and /CLEAN there is head movement of
the V into the v category. The affix -ed either lowers on to the v+V head
(as we did in chapter 9) or is attached to the V to start with, and there is
covert movement of the v+V to T (aswe did in chapter 12) - we'll indicate
these options with a dotted curved line.
You now have enough information totry General Problem Set2
2Kratzer actually calls the category voice, and suggests that it isofthe same category
astheauxiliary verbthatintroduces thepassive discussed in chapter 10.
Chapter 13: Expanded VPs 379
the window
The evidence for this separation of agents from their roots comes from the
behavior of phrasal idioms. Marantz noticed that while there are sentential
idioms (the verb plus all the arguments as in The pot called the kettle black*)
and verb+object idioms (such as kick the bucket4), there are no subject+verb
idioms. Similarly, we find that while the meaning of the object can change
theinterpretation oftheverb, as in(9), thesubject never doesso(10).
9) a) killa bug =end the lifeof the bug
= cause the conversation to end
= while away the time span of the evening
= empty the bottle
= entertain the audience
10) a) John laughed
b) The audience laughed
c) The manager laughed
d) The bug laughed
a) kill a bug
b) kill a conversation
c) kill an evening
d) kill a bottle
e) kill an audience
1 For the information of non-native English speakers: this means "to speak
hypocritically."
4For the information of non-native English speakers: this means "to die."
380 Advanced Topics
This suggests that there is a tight link between the verb and its object that
it doesn't share with itssubject. If weadopt thelittlev approach, thesefacts
follow directly: the verb root has the object in its theta grid but the agent
is never in the theta grid of the verb root, so an idiomatic meaning cannot
form around it.
You now have enough information totry Challenge Problem Sets 1 &2
With this light verb approach for simple transitives, let's considerhow
this might extend to datives. It is a relatively simple matter to substitute
littlev and big V in for the verb and Xin the tree in (5) giving(11):
11)
a book
Head movement ofthe V to v gives thecorrect spellout order (Josh gave Clay
a book). It also accounts for the asymmetric c-command effects between
thetwopost-verbal DPs. There arehowever, a number ofproblems withthis
analysis still. One thing that is a bit ofa mystery is how case gets assigned
to the two NPs, especially the one that is labeled DPI in (11). To find
an explanation for this, we turn now to a different, yet equally puzzling
phenomenon involving object positioning.
3. Object Shift
The following pair of sentences with embedded infinitive clauses from two
dialects of Irish show an interesting alternation in the position of the object
Chapter 13:Expanded VPs 381
and in the case marking of the object. Sentence (12a) represents the order of
elements in the northern dialects of Irish (mainly Ulster). You will note that
the object appears before the verb (and the particle a1) and bears accusative
case- the case we normally associate with being the complement to the verb.
In sentence (12b), which is found in literary Irish and in southern (Munster)
Irish, by contrast, the objectappears after the verb, but it takes a genitive case
marking.5
12) a) Bamhaith Horn [CP Sean an abairt aL scriobh].
C goodwith.l.S John the sentence.ACC TRAN write
"I want John to write the sentence."
b) Bamhaith liom [CP Sean aL scriobh na habairte].
C goodwith.l.S John TRAN write the sentence.GEN
"I want John to write the sentence."
Both these sentences are surprising. As we've seen in earlier chapters, Irish
seems to typically put its heads beforeitscomplements(determinersprecede
Ns, prepositions precede DPs, etc.). With this in mind, (12b) displays
the expected word order; but we get the unexpected genitive case marking
on the object. (12a) has the opposite problem, the order of the verb
and itscomplement DP are reversed from what we'd expectin a head-initial
language, but the object atleast bears thecorrect case.6
We find a similar variation in literary Irish when we look at main clauses
in differentaspects.7 In the progressive aspect(13b), objects follow the main
verb and take thegenitive case. In therecentperfective (13a), objects precede
the main verb (and the particleaL), and take the accusativecase:
13) a) T3 Sedntareis an abairt aL scriobh.
be.PRES John PERF the sentence TRAN write
"John has just written the sentence."
b) T3 Sedn ag scriobh na habairte.
be.PRES John PROG write the sentence
"John is writing the sentence."
5 In traditional grammars, this is typically taken to be a result of the fact that
the infinitival verb in Irish is "nominal" in nature. This account doesn't explain
the accusative case in the Northern dialects, so we will leave it aside here.
6This order and case marking are also available in the Southern dialects if there is
no overt subject for the embedded clause. (See McCloskey 1980 for a survey of
the phenomenon).
7Aspect refers to the duration of an event and whether it is completed or not.
382 Advanced Topics
Thiskind ofalternation, known as object shift, is not an esoteric property of
Irish; it is found in a wide variety of languages. Take the embedded clauses
below taken from German (data from Diesing 1992); in particular focus
on the order of negation and the DP referring to "the cat":
14) a) ... weil ich nicht [DP eine einzige Katze] gestreichelt habe
since I not a single cat petted have
"... since I have not petted a single cat"
b) ... weil ich [DP die Katze] nichtstreichle
since I the cat not pet
"... since I did not pet the cat"
The conditions for object shift here are different from the Irish example
(the alternation is around negation instead of around the verb; and the
alternation seems to be linked to definiteness/specificity rather than case),
but this also appears to be a case of object alternation.
We can even find a related alternation in English. Consider complex
verbs like blow up in English. With full NPs, the direct object can either
precede or follow theparticle up (15a and b),but with pronouns theparticle
must appear in the middle of the complexverb (15c and d):
15) a) I blew up the building.
b) I blew the building up.
c) *Iblew up it.
d) I blew it up.
These alternations inobject position alldiffer intheir specifics, butclearly we
have some kindofmovement operation thataffects theposition ofobjects.
The Irish case is particularly illustrative, as it shows an alternation
in case marking. Accusative case is only available in the shifted position.
Building upon a proposal of Pollock (1989) and Chomsky (1991) we can
propose that there is a special functional category whose sole purpose
is accusative case assignment. The name of this category is AgrO (standing
for Object Agreement); (the basic idea behind this name being that the
caseassigner is usually the constituent that agrees with the object). The head
of AgrO in Irish is the particle aL,8 which follows the shifted object.
For Irish object shift then, we have a structure where the object moves
from the position where it gets its theta role to this shifted position
where it gets accusative case:
8This is a different aL from the oneseenin the exercises in chapter 11. It is simply
homophonousin the same way that the English infinitive marker to is homophonous
to the preposition to.
Chapter 13:Expanded VPs 383
16) AgrOP
anabairt
AgrOP seems to be located between VP and vP.9 This can be seen in
thefollowing sentence from Scots Gaelic (alanguage closely related toIrish),
where the shifted object appears before the a1 AgrO morpheme but after
the light verb bhith:
17) Bu toigh learn [CP sibhaL bhithair an dorus aL dhunadh].
be like with me you AGR v PERF the door AGR close
"I'd likeyou to have shut the door."10
You now have enough information to try General Problem Sets 3 &4 and Challenge
Problem Set 3
Leaving asidetheupper portionofthe tree, thestructureoftheexpandedVP
then is at least as follows
18)
As first observed by Koizumi (1993).
0Data from Adger (1996).
Agent theta role assigned here
Accusative caseassigned here
384 Advanced Topics
Thisis quite a radical shift in structurefromthesimpleVP we've had up
to now. So it is worth considering if this proposal makes any predictions.
One elegant argument for this expanded structure comes from the work of
Lasnik (1999a). In many languages there are phenomena where a string that
has already been uttered is omitted in subsequent structures where it would
otherwise haveto be repeatedword forword. Forexample, we getsentences
like (19):
19) Darin will eat a squid sandwich but Raiza won't.
The second CP here [Raiza won't] is obviously missing the VP [eat a squid
sandwich] as that is what Raiza won't do. One typical explanation for this
phenomenon is that there is a specialrule of ellipsis that allows the deletion
of VPs under identity with an antecedent VP:
20) Darin will [eat a squid sandwich]; but Raiza won't [eat a squid
sandwich];.
There is one variant of the ellipsis phenomenon that puzzlingly doesn't
delete the entire VP. Thisis called, for reasons that need not worry us here,
pseudogapping. In pseudogapping constructions the accusative DP is left
behind:
21) Darinwilleat a squid sandwich, and Raiza willapeanut butter one.
This isn't simply deleting a verb:everythingbut the accusative-markedDP is
deleted from the second VP (examplefrom Lasnik 1999a).
22) TheDAprovedJones guilty and theassistant DAwill pt&ve-Smith
guilty.
The splitvP-AgrOP-VP architecture provides a straightforward analysis
of this. Pseudogapping is indeedVPellipsis - but the object has moved out
of the VP into AgrOP, but the verb and all other material remains inside
the VP, and gets deleted. The object survives ellipsis because it has shifted
outside of the VP.
23) AgrOP
elided
Chapter 13:Expanded VPs 385
Regular VP ellipsis is really vP ellipsis, which is why the object disappears
in those contexts.
Object shift to a VP-external AgrOP also explains a puzzling fact
in a phenomenon called Antecedent Contained Deletion (ACD). ACD is
a special case ofellipsis, where an elidedVPiscontained within a DP that is
itself contained within the VP that serves as the antecedent for the ellipsis,
as schematized in (24). An example of ACD is given in (25):
24) [vp ... [dp ...[cp-WttH, ...]]],
25) Brandon [VP read every book that Megan did [VP...]].
ACD has the property of infinite regress. The antecedent of the ellipsis
contains the ellipsis, so how can the content of the elided VP ever be
recovered? The antecedent of the gap also contains that gap. Object shift
provides an elegantsolution to thisproblem.11 TheDPcontaining thegap is
always an object, so it shifts out of the antecedent VP into the specifier
of AgrOP. After this happens the elided VP is no longer contained within
its own antecedent so the problem of infinite regress vanishes, as the actual
gap is no longer contained within the VPantecedent.
26) AgrOP
"SeeHornstein(1994).
antecedentfor theelided VP
V"
V t,
4. DITRANSITIVES: REPRISE
Let's now return to ditransitive verbs. For verbs where the final argument is
a PP and CP we have a straightforward account of where all the arguments
get theta roles and case. First let's look at a verb like tell which can take
both a DP and a CP complement. First we have the theta grids for the little v
and the root /TELL.
386 Advanced Topics
27) CAUSE
Agent
DP
Predicate
AgrOP
i J
/TELL
Theme
DP
Proposition
CP
k 1
Given a sentence like Nate told Brandon that Kimberly drove an SUV, we have
a D-structure tree as in (28) (leaving aside the details of tense inflection).
28) CP
Brandon W
/TELL
CP
that Kimberlydrives an SUV
The root moves through AgrO and into CAUSE; it must stop in AgrO
on its way up to the little v in order to meet the MLC. The DP Brandon shifts
to the specifier of AgrOP to get case.
A similar analysis can be applied to verbs like putthat take a PP second
complement,as in the sentenceBriana put the mug on the counter:
29) CP
Chapter 13:Expanded VPs
vP
DP v'
A
Briana v AgrOP
CAUSE
the mug V PP
/PUT
on the counter
The last kind of ditransitive is more difficult. This is the case oigive. Give
allows two possible argument structures one with a DP theme and PP goal
(as in Jason gave the tape recorder toMaria). This version of give presumably
has a structure like that in (28). The other version of give takes two DP
complements (Jason gave Maria the tape recorder). There are two puzzles with
this kind of construction. First we have the obvious question of the source
of accusative case for the goal DP. Second, more curious, is the fact that
the indirect object goal precedes the direct object theme. If the theme moves
to the specifier of AgrOP for case, then the goal must be moving to a higher
position than that. One possibility that has been proposed is that goals can
be introduced by two distinct mechanisms. One is via a preposition like to.
The other is using another light verb, this one meaning LOCATE or POSSESS
instead of CAUSE. This second mechanism could be paired with another case
assigning functional category, this time for indirect objects (AgrlOP).12
12 Collins and Thr^insson (1996).
387
388 Advanced Topics
Under such a story the architecture of the complex VP for a verb like give
looks like (30).
30) CP
Maria v AgrOP
LOCATE / \
AgrO'
the tape recorder
EachDP moves to its case position, and the root /GrvE moves through AgrO,
Locate,AgrlO, into CAUSE. This gives the correct surface order.
There are, of course, a number ofopen issues here, not the least of which
is the relationship between the two structures for the verb give. However,
these complex split VP structures seem to provide a mechanism for
explaining the hierarchical properties of ditransitive verbs, the case marking
of their arguments and provide an explanation for such diverse facts as
Chapter 13: Expanded VPs 389
pseudogapping, Antecedent Contained Deletion, and objectshift. In the next
chapter, we see that the AgrO category also helps us in providing a landing
site for a new kind of DP movement (subject to object raising).
You now have enough information to try General Problem Sets 5 &6
Ideas, Rules, and Constraints Introduced in this Chapter
i) Light verbs (little v): the higher part of a complex verb, usually
meaning CAUSE (or LOCATE in the case of ditransitive double
object verbs).
ii) Object Shift: the phenomenon where accusatively marked objects
shift leftwards.
iii) AgrO: the head that checks accusative case in the split VP system.
iv) (VP) Ellipsis: A process that deletes a VP (or vP) under identity with
a previously uttered identical VP.
v) Pseudogapping: A variety of ellipsis where the accusative object is
not deleted.
vi) Antecedent Contained Deletion (ACD): A kind of ellipsis where
the antecedent of the ellipsis contains the ellipsis site.
Further Reading
Adger, David (1996) Aspect, agreement and Measure Phrases in Scottish
Gaelic. In Robert Borsley and Ian Roberts (eds.), The Syntax of the Celtic
Languages. Cambridge: Cambridge University Press, pp. 200-2.
Barss, Andy and Howard Lasnik (1986) A note on anaphora and double
objects. LinguisticInquiry 17, 347-54.
Beck, Sigrid and Kyle Johnson (2004) Double objects again. Linguistic Inquiry
35, 97-124.
Chomsky, Noam (1991) Some notes on economy of derivation
and representation. In Robert Freidin (ed.), Principles and Parameters
in Comparative Grammar. Cambridge: MIT Press, pp. 417-54.
Chung, Sandra (1976) An object creating rule in Bahasa Indonesia. Linguistic
Inquiry 7,1-37.
390 Advanced Topics
Collins, Chris and Hoskuldur Thrainsson (1996). VP internal structure and
object shift in Icelandic. Linguistic Inquiry, 27,391-444.
Den Dikken, Marcel (1995) Particles: On the Syntax of Verb-Particle, Triadic, and
Causative Constructions. Oxford: Oxford University Press
Diesing, Molly (1992) Indefinites. Cambridge: MIT Press.
Harley, Heidi (2002) Possession and the double object construction. Yearbook
of LinguisticVariation 2, 29-68.
Hornstein, Norbert (1994) An argument for Minimalism: the case of
antecedent contained deletion. Linguistic Inquiry 25, 455-80.
Iatridou, Sabine (1990) About Agr(P). Linguistic Inquiry 21, 551-7.
Johnson, Kyle (1991) Object positions. Natural Language and Linguistic Theory
9, 577-636.
Koizumi, Masatoshi (1993) Object Agreement Phrases and the split VP
hypothesis. MIT Working Papers in Linguistics 18,99-148.
Kratzer, Angelika (1996) Severing the external argument from its verb. In
Johan Rooryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon.
Dordrecht: Kluwer. pp. 109-37.
Larson, Richard (1988) On the double object construction. Linguistic Inquiry
19,335-91.
Lasnik, Howard (1999a)Minimalist Analyses. Oxford: Blackwell.
Marantz, Alec (1984). On the Nature ofGrammatical Relations. Cambridge: MIT
Press.
McCloskey,James (1980) Is there raising in Modern Irish? Erin 31,59-99.
Pesetsky, David (1994) Zero Syntax: Experiencers and Cascades. Cambridge:
MIT Press.
Pollock, Jean-Yves (1989) Verb movement, Universal Grammar and the
structure of IP. Linguistic Inquiry 20, 365-424.
Stechow, Arnim von (1996) The different readings of wieder 'again': A
structural account. Journal ofSemantics 13, 87-138.
General Problem Sets
1. NPIs and Double Object Verbs
[Data Analysis and Critical Thinking; Intermediate]
The words anything and anyone are negative polarity items, and must be
licensed by a negative word like no one or nothing. Explain how the following
data supports the structure given above in the main text in (5).
(Hint: think structural relations). (Data from Barss and Lasnik 1986.)
a) Amanda gave no one anything.
b) *Amanda gave anyone nothing.
Chapter 13:Expanded VPs 391
2. Complex Verbs
[Data Analysis and Critical Thinking; Intermediate]
Sentence (a) is from Persian and sentence (b) is from Chichefta. Explain
how these data support the idea that verbs are really composed of a v
and a V.
a) Kimea az ra'ise edare da'vat kard
Kimea of boss office invitation v
"Kimea invited the office boss."
b) Mtsikana anagw-ets-a kuti mtsuko
Girl fall-v that waterpot
"The girl knocked over that waterpot."
3. Particles
[Data Analysis and Critical Thinking; Advanced]
Using the split vP-AgrOP-VP system, explain the verb-particle facts of
Englishgiven in example (15). Assume that a verb like blow up is structured
as in (i) and that the blow portion of this complex verb can move
independently of the preposition/particle up. You do not have to explain
why the shifted order is obligatory with pronouns and not with DPs.
i) V
V P
blow up
4. Theta Grids
[Application of Skills; Intermediate]
The theory involving AgrOP requires that we modify the lexicalentries given
in the text above in (7). Provide new theta grids for little v meaning cause
and the root Vclean that take into account AgrOP.
5. Trees
[Application of Skills; Intermediateto Advanced]
Using split VP structures and AgrOP draw the trees for the following
sentences:
a) Susan sent the package to Heidi.
b) Carolyn sent Heidi a package.
c) Peter placed the letter in the envelope.
d) I asked Mike if he had seen the Yeti.
e) I bought some flowers for Manuel.
f) I bought Manuel some flowers.
392 Advanced Topics
6. Applicatives in Bahasa Indonesia
[Data Analysis; Advanced]
Consider the following data from Bahasa Indonesia (Chung 1976).
This language has two orders that are similar to the prepositional order and
the double object orders of English give type verbs. What is interesting is the
presence in the construction with two DPs of a morpheme in the verb that is
typically called the applicative (appl), explain how this data is evidence
for the split VP approach proposed in (30).
a) Saja mem-bawa surat itu kepada AN.
I CAUSE-bring letter the to AH
"I brought the letter to AN."
b) Saja mem bawa-kan AN surat itu.
I CAUSE-bring-APPL AN letter the
"I brought AN the letter."
tmmm
Challenge Problem Sets
Challenge Problem Set 1: That Dog Doesn't Hunt
[Critical Thinking; Challenge]
Consider the idiom: That dog doesn't hunt (meaning "that solution doesn't
work"). Is this a counter-example to the claim that there are no subject-verb
idioms in English? (As a matter of contrast: notice that in verb + object
idioms the subject can be any possible DP: John kicked the bucket,
The man kicked the bucket.
Challenge Problem Set 2: Again
[Critical Thinking; Challenge]
As discussed in von Stechow (1996) and Beck and Johnson (2004) the
following sentence is ambiguous:
John opened the door again.
Itcan have either of the following meanings:
i) The door was open before (perhaps opened by Susan) and now it's open
again due to John's action,
ii) John opened the door before, and he did itagain.
Keeping in mind the principle of modification, explain how this data is
evidence for a little v meaning cause and the split VP hypothesis.
Challenge Problem Set 3: AgrS
[Creative Thinking; Challenge]
In the chapter above we proposed AgrO and AgrlO, there may well
be evidence that case is not assigned by TP, but by an AgrSP. In particular
Chapter 13: Expanded VPs 393
it has been proposed that the EPP is a property of TP, but case is assigned
lower in the structure, in an AgrSP:
EPP checked here
T AgrSP^.Nominative case checked here
AgrS'
AgrS vP
A
Part 1: Using the following data from English argue for an AgrS in the
position suggested above. Assume there is an expletive (without a theta
role.)
a) There was a man arriving at the station when Ipulled up.
b) There were four men arriving at the station when I pulled up.
Part2: The following data from Scots Gaelic was given above as evidence
for the position of AgrO between v and V. This data also contains evidence
for AgrS. Explain what it is. (Scots Gaelic data from Adger1996.)
c) Bu toigh learn [Cp sibh aL bhith air an doras aL dhunadh.]
be likewith me you AGR v perf the door AGR close
"I'd like you to have shut the door."
^
Raising, Control,
and Empty Categories
0. Introduction
The following two sentences look remarkably alike:
1) a) Jean is likely to leave,
b) Jean is reluctant to leave.
But these sentences are structurally very different. Sentence (la) is a raising
sentence like those we saw in chapter 10. Sentence (lb), however, is
a different matter. This is what we call a control sentence; it does not involve
any DP movement. We will claim there is a special kind of null DP
in thesubject position oftheembedded clause. Syntacticians call thisspecial
DP PRO, which stands for "null pronoun." The differences between
these two constructions are schematized below.
2) Jeanjis likely [f,to leave].
* I
3) Jean is reluctant [PRO to leave].
The bracketed diagram in (3) shows the DP raising construction we looked
at in chapter 10. Thestructure in (4), which hasno movement, is the control
construction. The evidence for this kind of proposal will come
from the thematic properties of the various predicates involved. In addition
to contrasting the sentences in (la and b), we'll also look at the differences
between sentences like (4a and b):
subject-to-subject raising
(subject) control
396 Advanced Topics
4) a) Jean wants Brian to leave.
b) Jean persuaded Brian to leave.
Again, on the surface these two sentences look very similar. But, again,
once we look at these in more detail we'll see that they have quite different
structures. We will claim that Brian in (4a) raises to the object position of
the verb wants. This is called subject-to-object raising, and was discussed
in an exercise in the last chapter. The structure of the sentence in (4b)
parallels the structure of the control sentence in (lb). Both Jean and Brian
are arguments of the verb persuade, there is no raising, but there is a PRO
in the subject position of the embedded clause.
5) JeanwantsBrianj [tt toleave]. subject-to-object raising
± I
6) Jean persuaded Brian[PROto leave]. object control
The construction in (6) is called object control (because the object "controls"
what the PRO refers to).
This chapter ends with a short discussion of the various kinds
of empty elements we've looked at so far (null heads, PRO, traces, etc.), and
introduces a new one which is found in languages like Spanish and Italian.
1. Raising vs. Control
1.1 Two Kinds ofTheta Gridsfor MainPredicates
If you look at the following two sentences, you will see that the predicate
is likely only takes one argument: a proposition.
7) a) [ThatJeanleft] islikely. clausal subject
b) It is likely [that Jean left]. extraposition
Sentence (7a) shows the proposition that Jean left functioning as the
predicate'ssubject. Sentence (7b) has thisembedded clause as a complement,
and has an expletive in subject position. For reasons having to do with
the history of generative grammar, but that need not concern us here,
the first construction (7a) is often called a clausal subject construction, and
thesecond(7b) an extraposition construction. The thetagrid for thepredicate
is given in (8). As is standard (see chapter 8), expletives are not marked
in the theta grid, as they don't get a theta role.
Chapter 14:Raising,Control, and Empty Categories
8) is likely
Proposition
CP
397
We assume that the D-structures of the sentences given in (7) are identical.
These sentences have the embedded clause as a complement to the predicate,
and nothing in the subject position:
9) TP
r
vP
T
-ed
VP2
^1
DP V
A 1Jean V
leave
1This little v isn't CAUSE because it doesn't introduce an agent. It probably means
something like "be in the state of..."
2Tokeep this tree down to a reasonable size, I'm abbreviating the vP-VP tree here
simply as VP.
398 Advanced Topics
In the clausal subject construction, the embedded CP moves to the specifier
of TP, presumably to satisfy the EPP requirement that every clause have
a subject:3
(10) shows the SPELLOUT for sentence (7a). Sentence (7b) has a slightly
different derivation, instead of moving the clause to satisfy the EPP, an
expletive it is inserted into the specifier of TP as seen in the SPELLOUT in (11):
3We haven't discussed the possibility ofmoving CPs before. Since this ismovement
for the EPP, it may well be a variant of DP movement. This analysis of clausal
subjects (involving movement) is not uncontroversial. Some researchers generate
these CPs directly in the specifier of TP at D-structure. We move it from the
complement positionto ensure parsimony with the analysisof expletive and raising
constructions discussed below. We should also note that not all raising verbs allow
theclausal subject construction. Forexample, seem and appear do not *[[that Jean left]
seems}. I leave it as an exerciseforyou to figureout why this might be the case.
Chapter 14:Raising, Control, and Empty Categories 399
11) TP
that Jean left
Observe that the embedded clause is finite in both these sentences. This
means that its subject gets nominative Case. As we saw in the chapter 10,
if the embedded clause is non-finite (as in 12), then the subject must move to
get Case. Fortunately, islikely does not have an external (subject) theta role,
but does have a nominative Case feature to check. This means that
the specifier of the higher TP is available for Case feature checking. This is
a typical raising construction.
12) is likely [Jean to leave].
400
13) TP
Advanced Topics
Ends herechecking caseand EPP
/Stops herefor EPP
T VP
to ^
_DP V
A 1Jean V
leave
As we noted in chapter 10, the Jean in this sentence gets its theta role
from leave. Jean is going to leave, she isn't likely. What is likely is the whole
proposition of Jean leaving. With is likely then, there is only one theta role
assigned (to the embedded clause). Three possible sentences emerge
with this structure: clausal subject, extraposition and raising.
Let's contrast this with the predicate is reluctant. If you think carefully
about it, you'll notice that this predicate takes two arguments. The person
who is reluctant (the experiencer) and what they are reluctant about
(the proposition):
14) is reluctant
Experiencer
DP
Proposition
CP
Chapter 14: Raising, Control, and Empty Categories 401
This means that, unlike is likely, is reluctant assigns a theta role to its subject.
Because of this both clausal subject and extraposition (expletive)
constructions are impossible. The specifier of TP of the main clause is
already occupied by the experiencer (it moves there to get Case), so there is
no need to insert an expletive or move the CP for EPPreasons. Thisexplains
why the following two sentences (an extraposition and a clausal subject
example) are ill-formed with the predicate is reluctant:
15) a) *It is reluctant [that Jean left],
b) *[that Jean left] is reluctant.
(where it is an expletive)
Bothof these sentencesseem tobe "missing" something. More precisely they
are both missing the external experiencer role: the person who is reluctant.
Consider now the control sentence we mentioned above in the introduction:
16) Jean is reluctant to leave.
Jean here is the experiencer, and the embedded clauseis the proposition:
17) a) is reluctant
Experiencer
DP
Proposition
CP
i k
b) Jeanj is reluctant [to leave]k.
So Jean is theta marked by is reluctant. Note, however, that this isn't
the only predicate in this sentence. We also have the predicate leave, with
the following theta grid:
18) leave
M
Who is this theta role assigned to? It also appears to be assigned to the DP
Jean:
19) Jeani/m is reluctant [to leave]k.
As we saw in chapter 8, the theta criterion only allows one theta role per DP.
This sentence seems to be a violation of the theta criterion, as its subject DP
gets two theta roles. How do we resolve this problem? The theta criterion
says that there must be a one-to-one mapping between the number oftheta
roles and the number of arguments in a sentence. This sentence has three
theta roles (agent, experiencer, and proposition), but only two arguments.
The logical conclusion, if the theta criterion is right - and we have every
402 Advanced Topics
reason to believe it is, since it makes good predictions otherwise - is that
there is actually a third DP here (getting the surplus agent theta role); you
just can't hear it. This DPargument is calledPRO (written in capital letters).
PRO only appears in the subject positions of non-finite clauses. The structure
of a control construction like (19) is given below. Indices mark the theta roles
from the theta grids in (17) and (18):
T VP
to ^1
DPra V
~A 1
PRO V
leave
You'll notice that PRO is appearing in a position where no Case can be
assigned. We return to this below, as well as to the question of why PRO
must obligatorily refer to Jean.
Before looking at any more data it might be helpful to summarize
the differences between control constructions and raising constructions.
The main predicate in a raising construction does not assign an external
4Again thisisnotCAUSE, this little v probably means something like "perceive."
Chapter 14:Raising, Control, and Empty Categories 403
theta role (it has an empty specifier of TP at D-structure). The subject of
the embedded clause is Caseless, and raises to this empty position for Case
checking (and to satisfy the EPP). In control constructions, the main clause
predicate does assign an external argument. There is no raising; the external
theta role of the embedded predicate is assigned to a null Caseless PRO.
This is summarized in the following bracketed diagrams:
21) a) no theta role Agent
i *
[ is likely [Jean to leave ]]. raising
t I
b) Experiencer Agent
[Jean is reluctant [PRO to leave]]. control
1.2 DistinguishingRaisingfrom Control
One of the trials of being a syntactician is learning to distinguish among
constructions that are superficially similar, but actually quite different
once we dig a little deeper. Control and raising constructions are a perfect
example. There are, however, some clear tests we can use to distinguish
them. First, note that whether you have a raising or control construction
is entirely dependent upon the main clause predicate. Some main clause
predicates require raising, others require control (and a few rare ones
can require both). The tests for raising and control then, mostly have to do
with the thematic properties of the main clause's predicate.
To see this we'll contrast our two predicates is likely, which is a raising
predicate, and isreluctant, which takes a control construction.
The most reliable way to distinguish raising constructions from control
constructions is to work out the theta grids associated with the matrix
predicates. If the matrix predicate assigns an external theta role (the one
that is underlined, the one that appears in subject position), then it is not
a raising construction. Take for example:
22) a) Jean is likely to dance,
b) Jean is reluctant to dance.
Contrast the role of Jean in these two sentences (as we did above in section
1.1). In the second sentence is reluctant is a property we are attributing
to Jean. In (22a), however, there is nothing about Jean that is likely. Instead,
what is likely is Jean's dancing.
404 Advanced Topics
One nice test that works well to show this is the behavior of idioms. Let's
take the idiom the catis outofthe bag. This construction only gets its idiomatic
meaning ("the secret is widely known") when the expression is a whole.
When ifs broken up, it can only get a literal interpretation ("the feline is out
of the sack"). You can see this by contrasting the meanings of the sentences
in (23):
23) a) The cat is out of the bag.
b) The cat thinks that he is out of the bag.
Sentence (23b) does not have the meaning "the secret is widely known."
Instead our first reading of this sentence produces a meaning where there
is actual cat-releasing going on. The subject of an idiom must at some point
be local to the rest of the idiom for the sentence to retain its idiosyncratic
meaning. We can use this as a diagnostic for distinguishing raising from
control. Recall that in the D-structure of a raising construction the surface
subject of the main clause starts out in the specifier of the embedded TP.
Therefore in raising constructions, at D-structure, the subject of an
embedded sentence is local to its predicate:
24) [ is likely [Jean to dance]].
If D-structure is the level at which we interpret idiomatic meaning, then we
should get idiomatic meanings with raising constructions.5 With control
constructions, on the other hand, the subject of the main clause is never
in the embedded clause, so we don't expect to get idiomatic readings.
This is borne out by the data.
25) a) The cat is likelyto be out ofthe bag. (idiomatic meaning)
b) The cat is eager to be out of the bag. (non-idiomatic meaning)
We can thus use idiom chunks like the cat in (25) to test for raising versus
control. If you get an idiomatic reading with a predicate, then you know
raising is involved.
Another test you can use to distinguish between raising and control
constructions is to see if they allow the extraposition construction.
Extraposition involves an expletive it. Expletives are only allowed
in non-thematic positions, which are the hallmark of raising:
26) a) It is likely that Jean will dance.
5This is notan implausible hypothesis. Idioms have the feel oflexical items (thatis,
their meaning must be idiosyncratically memorized, just like the meanings of words).
Remember that the lexicon is the source of the material at D-structure, so it makes
sense that D-structure is when idiomatic meanings are inserted.
Chapter 14:Raising, Control, and Empty Categories 405
b) *It is reluctant that Jean will dance.
At the end of this chapter, there is a problem set (general problem set 4)
where you are asked to determine for a list of predicates whether or not
they involve raising or control. You'll need to apply the tests discussed
in this section to do that exercise.
1.3 What is PRO?
You may have noticed a fairly major contradiction in the story we've been
presenting. In chapter 10, we claimed that DPs always need Case. However,
in this section we've proposed that PRO can appear in the specifier of nonfinite
TP. This is not a Case position, so why are we allowed to have PRO
here? Shouldn't PRO get Case too? It is, after all, a DP. Chomsky (1981)
claims that the reason PRO is null and silent is precisely because it appears
in a Caselessposition. In other words PRO is a very special kind of DP, it is
a Caseless DP, which explains why it can show up in Caseless positions,
like the specifier of non-finite TP.
Why do we need PRO? If we didn't have PRO, then we would
have violations of the theta criterion. Notice that what we are doing here
is proposing a null element to account for an apparent hole in our theory
(a violation of either the theta criterion or the Case filter). There is good
reason tobe suspicious ofthis: Itseems like a technical solution toa technical
problem thatis raised only by our particular formulation oftheconstraints.
Nonetheless, it does have a good deal of descriptive power. It can account
for most of the data having to do with embedded infinitival clauses. Until
abettertheory comes along, thePRO hypothesis wins because it can explain
so much data.
You now have enough information to try General Problem Set 1
2. Two Kinds of Raising, Two Kinds of Control
2.1 TwoKinds of Raising
Up to this point we havebeen primarily looking at raising from the subject
of an infinitive complement clause to the specifier of a main clause TP.
This raising happens so theDPcangetCase. However, raising doesn't have
to target the specifier of TP; there are other instances of DP raising where
the DP ends up in other positions. Consider the verb want. Want can take
an accusatively marked DP:
406 Advanced Topics
27) a) I want cookies.
b) Jean wants Robert.
c) Jean wants him.
Want can also take an infinitive CP complement (sentence (28) is an instance
of a control construction.)
28) Ijwant [PRO, to leave].
This flexible verb can also show up with both an accusatively marked DP
and an infinitive complement:
29) Ij want [Jearij to dance]k.
Think carefully about the theta grids of the verbs here. Jean is the agent
ofdance, I is theexperienceroiwant, and theproposition Jean to dance takesup
the second theta role of want.
30) a) dance
b) want
Agent
DP
J
Experiencer
DP
Proposition
CP
i k
Notice that Jean does not get a theta role from want; it only gets one from
dance. This means that this is not a control construction. You can see this if
we apply our idiom test to the sentence: 6
31) I want the cat to be let out of the bag.
Although the judgment isn't as clear here, it is possible to get the idiomatic
reading of the cat tobe letoutofthe bag.
Since this isn't a control construction, then how does the DP Jean get
Case? The embedded TP is non-finite, so its specifier is not a Case position.
The answer to this puzzle is the DP raises to the object position of want,
where it can get accusative Case.
6The extraposition test will not work here. Remember expletives are usually only
found in subject position (because of the EPP).Jean here is found in object position,
so extraposition can't apply.
32) TP
T
Chapter 14: Raising, Control,and Empty Categories
T vP
,PRES ^/\,
DPi v
A / \
I v7 AgrOP „- stops hereforCase
Perceive^>\
AgrO'
AgrO|ACC| VP
V CPk
VWANT / \
407
C TP ^ stops here for EPP
r
The verb root raises through AgrO into v. The DP Jean moves first to the
specifier of the embedded TP for EPP reasons, then moves on tothe specifier
of AgrOP where it gets accusative case.
We can see that this is the right analysis of these facts by looking at
the Case-marking a pronoun would get in these constructions. Since theDP
shows up as thespecifier ofAgrOP with an [acc] Case feature, we predict
it will take accusative Case. This is correct:
7This v isnotcause,asthere isnoagent role here. This v probably means something
like perceive.
Againthis VPis an abbreviation forvPand VP.
408 Advanced Topics
33) a) I want her to dance,
b) *I want she to dance.
Binding theory also provides us with a test for seeing where the DP is.
Recall the fundamental difference between a pronoun and an anaphor.
In the binding theory wedeveloped inchapter 5, ananaphor must bebound
within its clause, whereas a pronoun must be free. What clause a DP is in
determines whether it is an anaphor or a pronoun. Wecan use this as a test
for seeing where a DP appears in the tree structure. We are considering two
hypotheses: (34a) has the DP in the object position ofwant (just as in (32)),
whereas (34b) has the DP in thesubject position ofthe non-finite TP.
34) a) I want Jeanj [t(to dance],
b) I want [Jean to dance].
If we can have a bound anaphor, instead of Jean, then we know that the
pronoun must be in a different clause from its antecedent, since pronouns
cannot be bound within their own clause. Similarly we predict that if
an anaphor is OK, then the DP is within the same clause as its antecedent.
The data supports (34a).
35) a) *Jeanj wantshe^ tobeappointed president,
b) Jeani wants herj tobe appointed president.
b) ?Jeani wants herself tobeappointed president.
These forms exhibit asecond kind of raising, which we might call subject-toobject
raising.
Subject-to-object Raising = Exceptional Case Marking (ECM)
In the early work on generative grammar, in the 1960s and 1970s, the
construction we have been looking at here was treated in a very similar
manner to the analysis presented here. It was also called subject-to-object
raising. Inthe 1980s and early 1990s (in what was called GB theory), there
was a period of time when these constructions got a different analysis.
Instead of raising the infinitival subject to object position, the subject was
left inside the embedded clause (in the specifier of TP), and the verb was
allowed to "exceptionally" Case mark into the embedded clause. Thus for
that period of time, these constructions were called Exceptional Case
Marking (or ECM) constructions. Today, we have gone back to the original
subject-to-object raising analysis. Can you think of some way that we can
distinguish the ECM from subject-to-object raising analyses?
You now have enough information to try General Problem Sets 2&3
Chapter 14: Raising, Control,and Empty Categories 409
2.2 Two Kinds ofControl
In section 1, we contrasted sentences like (36a) and (36b). These sentences
differed in terms of their argument structure and in what movement if any
applies. (36a) is a raising construction, where Jean gets its theta role only
from to leave, and raises for Case reasons to the specifier of the main clause
TP. In(36b), Jean getsa theta role from isreluctant, and there is no movement.
Instead there is a null Caseless PRO in the specifier of the tenseless clause.
36) a) Jeanjis likely [tj to leave].
b) Jeanjis reluctant [PROj to leave].
In this subsection, we'll makea similarclaimabout the structures in (37).
37) a) Jean wants Roberts [f,to leave].
b) Jeanpersuaded Robert; [PROi to leave].
Sentence (37a) is an instance of subject-to-object raising. Sentence (37b),
while on the surface very similar to (37a), is actually also a control
construction. There are twomajor kinds ofcontrol constructions. To seethis
I'll putthe two (b) sentences side by side in(38). (38a) iswhat we call subject
control, because thesubject DP ofthemain clause isco-referential withPRO.
(38b) is object control, where the main clause object is co-referential
with PRO.
38) a)(=36b) Jea^ isreluctant [PROj toleave]. subject control
b)(=37b) Jean persuaded Robert. [PROj toleave]. object control
Considerfirst the thematicpropertiesof the raisingconstruction:
39) Jean; wants Robert, [tj toleave]k.
We are now well familiar with the theta grid for to leave, which takes
asingle agent argument. The theta grid for the subject-to-object raising verb
want is repeated below:
40) a) leave
b) want
Experiencer
DP
Proposition
CP
i k
410 Advanced Topics
Robert is the agent of leave, but is not an argument of want. In section 2.1
above, we used the idiom test to show that this is the case. Now, contrast
thissituation with theobject control verbpersuade:
41) Jean; persuaded Robert,,, [PRO( toleave]k.9
The DP Robert in this sentence is theta marked by persuade. So in order
not to violate the theta criterion we have to propose a null PRO to take
the agent theta role of leave.
42) a) leave
Agent
DP
b) persuade
Agent
DP
Theme
DP
Proposition
CP
i m k
We can see this again by comparing the idiomatic readings of subject-toobject
raising vs. object control.
43) a) Jean wants the cat to gethis/Bill's tongue.
b) #Jean persuaded thecat togethis/Bill's tongue.
Sentence (43a) is slightly odd, but it does allow the idiomatic reading, but
(43b) only takes the literal (non-idiomatic) meaning.
Control = Equi
In early versions ofgenerative grammar - in particular, the ones before the
invention oftheta roles - the phenomenon we are calling control was called
Equi-NP Deletion or Equi for short. This is just another name for the same
phenomenon.
2.3 Summary ofPredicate Types
In this section we've argued for four distinct types ofembedded nonfinite
constructions: subject-to-subject raising, subject-to-object raising, subject
control and object control. Which construction you get seems to be
dependent upon what the main clause predicate is. For example, is likely
9The indices on this sentence mark theta roles (as marked in the grid in (42)). They
donot mark coindexing. In this sentence, the index m=,(m andjare the same index).
Chapter 14: Raising, Control,and Empty Categories 411
requires a subject-to-subject raisingconstruction whereasisreluctant requires
a subject control construction. It should be noted that some verbs allow
more than one type of construction. For example, the verb zoant allows
either subject control, or subject-to-object raising:
44) a) Jeanj wants [PROjto leave]. subject control
b) Jean wants Bill; [tj to leave]. subject-to-object raising
Anexample ofthese typesisgiven in (45) and a summary oftheirproperties
in (46):
45) a) Jean is likelyto leave. subject-to-subject raising
b) Jean wants Robert to leave. subject-to-object raising
c) Jean is reluctant to leave. subject control
d) Jean persuaded Robertto leave. object control
46) a) subject-to-subject raising
• Main clause predicate has one thetarole (tothe proposition), and no
external (subject) theta role
• DP movement of embedded subject to the specifier of TP for EPP
and Case
• Allows idiomatic readings
• Allows extraposition
b) subject-to-object raising
• Main clause predicate assigns two theta roles (an external agent or
experiencer and a proposition)
• Main clause predicate has an [ACC] Casefeature
• DP movement of the embedded clause subject to the specifier of
AgrOP for Case reasons
• Allows idiomatic readings
c) subject control
• Main clause predicate assigns two theta roles (external agent or
experiencer and proposition)
• Caseless PRO in embedded clause
• No DP movement for Case
• Does not allow idiomatic readings or extraposition
d) object control
• Main clause predicate assigns three theta roles (external agent or
experiencer, an internalthemeand a proposition)
• Caseless PRO in embedded clause
• No DP movement for Case
• Does not allow idiomatic readings or extraposition
412 Advanced Topics
You nowhave enough information totry General Problem Sets4 &5, and
Challenge Problem Set1
3. Control Theory
In chapter 5, we developed a set of noun types (anaphors, pronouns, Rexpressions)
thathave different properties withrespect to howtheygettheir
meanings. R-expressionsget their meaning from the discourse or context and
can never bebound; anaphors arebound by antecedents within their clauses;
and pronouns can either be bound by antecedents outside their clause or
be free. In this section, we consider the troubling question of what kind
of DP PRO is. Unfortunately, we are going to get a bit of a mixed answer.
Let us start by defining some terminology. This terminology is subtly
similar to that of the binding theory, but it is different. If PRO gets its
meaning from another DP, then PRO is said to be controlled. This is identical
to the notion coreferent and verysimilar to the notion bound (we will make
this distinction clearer below). The DP that serves as PRO's antecedent
is called its controller.
We are going to contrast two different kinds of PRO. The first kind is
called arbitrary PRO (orPROarb). The meaning ofthis pronoun isessentially
"someone":
47) [PROarb to find a new mate], go to a datingservice.
Arbitrary PRO is not controlled by anything. Arbitrary PRO is a bit like
an R-expression or a pronoun, in that it can get its meaning from outside
the sentence.
Non-arbitrary PRO(henceforthsimply PRO) also comes in two different
varieties. On onehand we have what is called obligatory control. Consider
the sentence in (48). Here, PRO must refer toJean. It can'trefer to anyone
else:
48) Jeanj tried PRO^ to behave.
There are other circumstances where PRO does not have to be (but can be)
controlled. This is called optional control, and is seen in (49):
49) Robert; knows thatit isessential [PRO^ tobewell-behaved].
PRO here can mean two different things. It can either refer to Robert or it
can have anarbitrary PROarb reading (indicated in(49) with the subscript j).
You can see this by looking at the binding of the following two extensions
of this sentence:
Chapter 14:Raising, Control,and Empty Categories 413
50) a) Robert;knows that it is essential [PRO; to be well-behaved on his;
birthday],
b) Robert; knows that it is essential [PROj to be well-behaved on one'Sj
birthday].
(50a) has the controlled meaning (as seen by the binding of his), (50b) has
the arbitrary reading (as seen by the presenceof one's).
With this in mind let's return to the central question of this section.
Is PRO an anaphor, a pronoun, or an R-expression? We can dismiss the
R-expression option right out of hand. R-expressions must always be free.
PRO is only sometimes free (= not controlled). This makes it seem more like
a pronoun; pronouns can be both free or bound. The data in (49) seems
to support this, PRO is behaving very much like a pronoun. Compare (49)
to the pronoun in (51).
51) Robert; knows it isessential [that he^ is well-behaved].
You'll notice that the indexing on (51) which has a pronoun, is identical
to the indexing on PRO in (49). We might hypothesize then that PRO is
a pronoun. This can't beright, however. Recall that wealso havesituations
where PRO must be bound (= controlled) as in the obligatory control
sentence Jean, tried PROirj to behave. This makes PRO look like an anaphor,
since anaphors are obligatorily bound. Williams (1980) suggests that
in obligatory control constructions PRO must be c-commanded by its
controller, just as an anaphor must be c-commanded by its antecedent.
However, as should be obvious, this can't be right either. First, as noted
above, we have situations where PRO is free (as in 47); anaphors can never
be free. Second, if we take the binding theory we developed in chapter 5
literally, PRO and its controller Jean, are in different binding domains,
violating Principle A.10 We thus have a conundrum: PRO doesn't seem to
be an R-expression, a pronoun, or an anaphor. It seems to be a beast of
an altogether different color.
Since the distribution of PRO does not lend itself to the binding theory,
an entirely different module of the grammar has been proposed to account
for PRO. This is called control theory. Control theory is the bane
ofprofessional theoreticians andstudents alike. Itis, quite simply, the least
10 Recall from chapter5 that our definition of bindingdomain as a clauseis probably
wrong. One might even hypothesize on the basis of data like Jean is likely to behave
herself that the definition of binding domain requires some kind of tensed clause,
rather than just any kind of clause. I leave as an exercise the implications of such
a move.
414 Advanced Topics
elegant part of syntactic theory. We'll have a brief look at it here, but
will come to no satisfying conclusions.
First let's observe that some parts of control are sensitive to syntactic
structure. Consider what can control PRO in (52):
52) [Jean/sfather^ is reluctantPROj/.; toleave.
If you draw the tree for (52), you'll see that while the whole DPJean's father
c-commands PRO, Jean by itself does not. The fact that Jean cannot control
PRO stronglysuggests that thereis a c-command requirement on obligatory
control, as argued by Williams (1980). This said, the structure of the sentence
doesn't seem tobe theonlythingthatcomes intoplay withcontrol. Compare
now a subject control sentence to an object control one:
53) a) Robert; is reluctant[PRO; tobehave]. subject control
b) Susarij ordered Robert; [PROj/.j tobehave]. object control
In both these sentences PRO must be controlled by Robert. PRO in (53b)
cannot refer to Susan. This would seem to suggest that the closest DP
that c-commands PRO must control it. In (53a), Robert is the only possible
controller, so it controls PRO. In (53b), there are two possible controllers:
Susan and Robert. But only Robert, which is structurally closer to PRO,
cancontrol it. This hypothesis works well in most cases, but the following
example shows it must be wrong:
54) Jean; promised Susanj [PROj/.j tobehave]. subject control
In this sentence it is Jean doing the behaving, not Susan. PRO must be
controlled by Jean, even though Susan is structurally closer. So structure
doesn't seem to be the only thing determining which DP does
the controlling.
One hypothesis is that the particular main clause predicate determines
which DP does the controlling. That is, the theta grid specifies what kind
of control is involved. There are various ways we could encode this.
One is to mark a particular theta role as the controller:
55) a) is reluctant
Experiencer
DP
controller
Proposition
CP
Chapter 14: Raising, Control, and Empty Categories
b) persuade
c) promise
Agent
DP
Theme
DP
controller
Proposition
CP
Agent
DP
controller
Theme
DP
Proposition
CP
415
In this view of things, control is a thematicproperty. Buta very carefullook
at the data shows that this can't be the whole story either. The sentences
in (56) alluse the verb beg, whichis traditionally viewed as an object control
verb, as seen by the pair of sentences in (56a and b), where the (b) sentence
shows an embedded tense clause paraphrase:
56) a) Louisbegged Kate; [PRO; to leaveher job].
b) Louisbegged Katethat she leaveher job.
c) Louis; begged Kate [PRO; tobeallowed [PRO; to shave himself]].
d) Louis; begged Katethat he be allowed to shave himself.
Sentences (56c and d), however, show subject control. The PROs in (c) must
becontrolled bythesubject Louis. The difference between the(a) and the(b)
sentence seems to be in the nature of the embedded clause. This is mysterious
at best. Examples like these might be used to argue that control
is not entirely syntactic or thematic, but may also rely on our knowledge
of the way the world works. This kind of knowledge, often referred to as
pragmatic knowledge," lies outside the syntactic system we're developing.
The study of the interaction between pragmatics, semantics and syntax
isone that is being vigorously pursued rightnow, but lies beyond the scope
of this book. See the further reading section below for some places you can
go to examinequestions like this in more detail.
You now have enough information to try Challenge Problem Set 2
11 Seeforexample Landau's (1999) dissertation.
416 Advanced Topics
4. Another Kind of Null Subject: "Little" pro
In chapter 8, we made the claim that all sentences require subjects, and
encoded this into the EPP. However, many languages appear to violate this
constraint. Take, for example, these perfectlyacceptable sentences of Italian:
57) a) Parlo.
speak.lSG
"I speak."
b) Parli.
speak.2SG
"You speak."
The subject DP in these sentences seems to be missing. But there is no
ambiguity here. We knowexactly whois doing the talking. This is because
the verbs are inflected with endings that tell us who the subject is. This
phenomenon is called eitherpro-drop or null subjects. Ideally, we would
like to claim that a strong constraint like the EPP is universal, but Italian
(and many other languages) seem tobe exceptions. One technical solution to
this issue istoposit that sentences in(57) actually dohave DPs which satisfy
the EPP. Notice again that this is merely a technical solution to a formal
problem.
You mightthink that theobvious candidate forthisemptyDP wouldbe
PRO. But in fact, PRO could not appear in this position. Remember PRO
only appears in Caseless positions. We know that Italian subject position
is a Case position, because you can have an overt DP like io in (58).
58) Io parlo.
I speak.lSG
"I speak."
So linguists have proposed the categorypro (written in lower-case letters).
pro (called little pro or baby pro) appears in Case positions; PRO (called
big PRO) is Caseless.
English doesn't havepro. This presumably is due to the fact that English
doesn't have a richagreement system in its verbal morphology:
59) a) I speak.
b) You speak.
c) He/ she/it speaks.
d) We speak.
e) They speak.
Chapter 14: Raising, Control, and Empty Categories 417
In English, only third person forms of verbs take any special endings. One
of the conditions on pro seems to be that it often appears in languages with
rich agreement morphology.12 The means weuse toencode variation among
languagesshould now befamiliar: parameters. Weuse thisdevice here again
in the null subject parameter, which governs whether or not a language
allows pro. Italian has this switch turned on. English has it set in the off
position.
You now have enough information to try General Problem Set 6
5. Summary
We started this chapter with the observation that certain sentences,
even though they lookalike on the surface, can actually have very different
syntactic trees. We compared subject-to-subject raising constructions
to subject control constructions, and subject-to-object raising constructions to
object control constructions. You can test for these various construction types
by working out their argument structure, and using the idiom test. Next
under consideration was the issue of what kind of DP PRO is. We claimed
thatit only showed up in Caseless positions. We alsosaw that it didn't meet
any of the binding conditions, and suggested it is subject, instead, to control
theory. Control theory is a bit of a mystery, but may involve syntactic,
thematic, and pragmatic features. Wc closed the chapter by comparing
two different kinds of null subject categories: PRO and pro. PRO is Caseless
and is subject to the theory of control. On the other hand, pro takes Case
and is often "licensed" by rich agreement morphology on the verb.
Ideas, Rules, and Constraints Introduced in this Chapter
i) PRO (big PRO): A null (silent) DP found in Caseless positions
(the specifier of non-finite TP).
ii) pro (Little pro or Baby pro): A null (silent) DP often found in
languages with "rich" agreement, pro does get Case.
iii) Clausal Subject Construction: A sentence where a clause appears in
the specifier of TP. E.g.,[That Jean danced the rumba} is likely.
12 This not a universally true statement. Many Asian languages allow pro-drop even
thoughthey don't have richagreementsystems. Fordiscussion, see Huang (1989).
418 Advanced Topics
iv) Extraposition: A sentence (often an alternate of a clausal subject
construction) where there is an expletive in the subject position
and a clausalcomplement. E.g., It is likely that Jean danced the rumba.
v) Subject-to-subject Raising: A kind of DP movement where
thesubject ofan embedded non-finite clause moves to thespecifier
ofTP ofthe main clause togetnominative Case. E.g., Jeant is likely t,
to dance.
vi) Subject-to-object Raising (also called Exceptional Case Marking
or ECM): A kind ofDPmovement where the subjectof an embedded
non-finite clause moves to the complement of the verb in the main
clause to get accusative Case. E.g., Jean wants Bi7//t, todance].
vii) Control Theory: The theory thatgoverns howPRO gets itsmeaning.
There appear to be syntactic factors (the controller must c-command
PRO), thematic factors (what DP does the controlling is dependent
upon what main clause predicate is present), and pragmatic factors
involved.
viii) Pragmatics: The science thatlooks at how language and knowledge
of the world interact.
ix) Subject Control (also called Equi): A sentence where there is a PRO
in the embedded non-finite clause that is controlled by the subject
argument of the main clause. E.g., john{ isreluctant PRO{ to leave.
x) Object Control: A sentence where there is a PRO in the embedded
non-finite clause that is controlled by the object argument
of the main clause. E.g., John persuaded B///, PROi toleave.
xi) Obligatory vs. Optional Control: Obligatory control is when
the PRO must be controlled: Jean{ is reluctant PROi to leave. Optional
control is when the DP can be controlled or not: Robert, knows
that it isessential [PROy to be well behaved].
xii) PROarb: Uncontrolled PRO takes an "arbitrary" reference. That is,
it means something like someone.
xiii) Null Subject Parameter. The parameter switch that distinguishes
languages like English, which require an overt subject, from
languages like Italian that don't, and allow pro.
Chapter 14: Raising, Control, and Empty Categories 419
Further Reading
Brame, Michael (1976) Conjectures and Refutations in Syntax and Semantics.
Amsterdam: Elsevier.
Bresnan, Joan (1972) Theory of Complementation in English. Ph.D.
dissertation, MIT.
Chomsky, Noam (1965) Aspects of the Theory of Syntax. Cambridge: MIT
Press.
Chomsky, Noam (1981) Lectures on Government and Binding. Dordrecht: Foris.
Hornstein, Norbert (1999) Movement and control. Linguistic Inquiry 30, 69-
96.
Hyams, Nina (1986) Language Acquisition and the Theory of Parameters.
Dordrecht: D. Reidel.
Jaeggli, Osvaldo and Kenneth Safir (eds.) (1989) The Null Subject Parameter.
Dordrecht: Kluwer Academic Publishers.
Landau, Idan (1999) Elements of Control. Ph.D. dissertation, MIT.
Manzini, Maria Rita (1983) On control and control theory. Linguistic Inquiry
14,421-46.
Petter, Marga (1998) Getting PRO under Control. The Hague: Holland
Academic Graphics.
Postal, Paul (1974) On Raising. Cambridge: MIT Press.
Rizzi, Luigi (1982) Issues in Italian Syntax. Dordrecht: Foris.
Rosenbaum, Peter S. (1967) The Grammar of English Predicate Complement
Constructions. Cambridge: MIT Press.
Williams, Edwin (1980) Predication. Linguistic Inquiry 11, 203-38.
General Problem Sets
1. The Existence of PRO
[Critical Thinking; Intermediate]
How does the following sentence provide support for the existence of PRO
in the subject position of the non-finite clause?
a) [To behave oneself in public] is expected.
Consider now the following sentence. Does it provide support for the
existence of PRO? How?
b) Roberti knew [cp that it was necessary [CP PROj to behave himselfj].
420 Advanced Topics
2. Raising to Object
[Critical Thinking; Intermediate]
We claimed that subject-to-object raising targets the specifier of AgrOP as
the landing site of the movement for Case. Consider the following sentences,
keeping in mind that out and incorrectly modify the main verb. How do these
sentences support the idea that subject to object raising lands in AgrOP?
Draw the tree for sentence (b).
a) She made Jerry out to be famous.
b) Mike expected Greg incorrectly to take out the trash.
3. Icelandic PRO and Quirky Case
[Data Analysis and Critical Thinking; Intermediate/Advanced]
Background. In order to do this question it will be helpful to have reviewed
the discussion of floating quantifiers in chapter 10, and to have done
the question on Icelandic quirky Case in chapter 10.
As discussed in chapter 10, in English, it is possible to "float" quantifiers
(words like all) that modify subject arguments:
a) The boys don't all want to leave.
Icelandic also allows floating quantifiers, but with a twist. The quantifier takes
endings indicating that it has the same Case as the DP it modifies.
Recall from the last chapter that certain verbs in Icelandic assign irregular or
"quirky" Cases to their subjects. The verb lelddist "bored" is one of these. In
sentence (b), the subject is marked with its quirky dative Case. The floating
quantifier dllum "all" is also marked with dative. (Data from Sigurflsson 1991.)
b) Strakunum leiddist dllum i skola.
boys.DAT bored all.DAT in school
"The boys were all bored in school."
We might hypothesize then, that floated quantifiers must agree with the noun
they modify in terms of Case.
The question. Now consider the following control sentence. What
problems does the following sentence hold for our claim that PRO does not
get Case? Can you relate your solution to the problem of Icelandic passives
discussed in the problem sets of chapter 10? Note that the noun
in the main clause here is marked with nominative rather than dative Case.
c) Strakamir vonast til ad PRO leidast ekki 6llum i skdla.
boys.NOM hope for to bore not all.DAT in school
"The boys hope not to be bored in school."
Chapter 14: Raising, Control, and Empty Categories 421
4. English Predicates
[Application of Skills; Intermediate]
Using your knowledge of theta theory and the tests of extraposition and
idioms determine if the predicates listed below are subject-to-subject raising
(SSR), subject-to-object raising, (SOR), subject control (SC), or object
control (OC). Some predicates tit into more than one category. (The idea
for this problem set comes from a similar question in Soames and Perlmutter 1979.)
is eager is believed seems is ready
persuaded urged requested hoped
expect force tell advise
ask assure imagine promise
want is likely consent imagine
encouraged intended
5. Trees and Derivations
[Application of Skills; Intermediate to Advanced]
Draw trees for the following sentences, annotate your trees with arrows
so that they show all the movements, and write in all PROs with appropriate
coindexing indicating control. You may wish to do this problem set after
you have completed problem set 4.
a) Jean wants Bill to do the Macarena.
b) Robert is eager to do his homework.
c) Jean seems to be in a good mood.
d) Rosemary tried to get a new car.
e) Susan begged Bill to let her sing in the concert.
f) Susan begged to be allowed to sing in the concert.
g) Christina is ready to leave.
h) Fred was believed to have wanted to try to dance,
i) Susan consented to try to seem to have been kissed,
j) Alan told me who wanted to seem to be invincible,
k) What did John want to eat?
6. Irish pro
[Data Analysis; Advanced]
Irish is a null subject language.
a) Rinceamar.
Dance.3PL.PAST
"We danced."
422 Advanced Topics
Consider the following Irish sentences and discuss how Irish pro-drop differs
from that found in Italian:
b) Ta me.
Am I
"I am."
c) Taim.
Am.lSG
"I am."
d) *Taim me.
Am.lSG I
"I am."
Challenge Problem Sets
Challenge Problem Set 1: Is Easy
[Critical Thinking; Challenge]
Consider the following sentences:
a) This book is easy to read.
b) John is easy to please.
Is is easy a raising or a control predicate or both? Ifit is a raising predicate,
which argument is raised? If it is a control predicate, where is the PRO?
What kind of PRO is it?
Challenge Problem Set 2: Controllers
[Critical Thinking; Challenge]
Williams (1980) claimed that obligatorily controlled PRO requires
a c-commanding controller. What problem do the following sentences hold
for that hypothesis?
a) To improve myself is a goal for next year.
b) To improve yourself would be a good idea.
c) To improve himself, Bruce should consider therapy.
d) To improve herself, Jane went to a health spa.
. = Advanced Topics
in Binding Theory
0. A Quick Review of Chapter 5 Binding Theory
In chapter 5, we sketched out a brief version of the binding theory that
allowed us to see the utility of structural relations and to give us a tool to
probe the c-command structures in a tree. Thatchapter contained a number
of simplifications. It isn't very hard to find counter examples to the theory
proposed there. In this chapter, we give binding relations a slightly more
nuanced look.
First let us review some of the basics from chapter 5. Binding theory
concerns the distribution of types of DPs with respect to each other. It is
partly a semantic study (looking at coreference) and partly a syntactic study
(looking at the structural considerations that govern the coreferential
possibilities). Recall that coreference is defined in terms ofbinding which
is both coindexing and c-command. When a DP is both c-commanded by
and coindexed with another DP, we say the first DP is bound by the second.
We have three major types of DPs. R-expressions are full DPs that get
their reference from external sources and may not be bound (Principle C);
Anaphors are DPs that must get their meaning fromanother local DP in the
sentence (i.e., must be bound) and are governed by Principle A. Pronouns
lie between the two. Pronouns may be coreferent with another DP in
the sentence, but need not be. To the extent that they may be coreferent they
must find their antecedent outside the clause they are contained in (Principle
B). In chapter 5, we defined locality in finding an antecedent by making
424 Advanced Topics
reference to the clauseas thebindingdomain. In thischapter, this is perhaps
the most major part of the theory that we'll have cause to revise. Just as we
defined the MLC relativistically by the type of element we were moving,
we will define binding domain relativistically according to the type of DP
that is involved. We will also consider the question of what level of
representation the binding principles hold at in the model of the grammar
we've been exploring.
The version of the binding theory I give you here is loosely based
on the one found in Chomsky's (1986b) Knowledge of Language, but with
an eye towards more recent developments in the theory.
1. Levels of Representation
In chapter 5, we claimed that the binding domain was a clause (CP). This
nicelyaccounts for the ungrammaticality of sentences like (1)below:
1) *ChriSj said [CP that himselfj was appealing].
However, on the face of it, this runs into trouble with sentences such as:
2) Chris; wants himselfj to be appealing.
Assuming that himself is the subject of the predicate to be appealing, here
the binding relation seems tocross clause boundaries. However, the analysis
we developed of subject to object raising in the previous chapter solves
this problem. If the DP himself moves to thespecifier oftheAgrOP for case
reasons it moves out of the CP where it gets its theta role. Onceit is part of
the higher clause structure its new binding domain contains the antecedent
Chris. In (3) the old binding domain is shown with the rightmost arc.
Theraising of the DPextends this to the higher CP
3) f[CPChriSj wants [AgtOPhimselfj [..f[CP •.. thimsel£ to be appealing]]]].
^- binding domain V. binding domain
after movement before movement
This fact then suggests that the binding principleshold after movement has
applied. Given the model that we sketched in chapter 12, this would be
the level of LF. This makes a fair amount of sense, sincebinding is at least
partly a semantic relation and LF is the level that interacts with
the conceptual/semantic component of the grammar.
Chapter 15: AdvancedTopics in BindingTheory 425
Were things so simple however! If you did general problem set 2
in chapter 11, you will have learned that there is at least some evidence
that bindingprinciples hold before movement. Take the sentence in (4):
4) [Whichpictures of himselfj] did Chris;like?
The wh-moved DP here contains an anaphor (himself). This anaphor is only
c-commanded by its antecedent before the movement:
5) Ci+Q/+whi ChriSj didlike [which pictures ofhimself;] (D-structure)
In chapter 11, this wastaken tobeevidence that thebinding principles held
at D-structure, before movement.
So we have an apparent contradiction here. Raising sentences such
as(2) provide an argument for theclaim thatthebinding principles happen
aftermovement, but w/i-questions such as (4) suggest that binding principles
happen before movement. The theory of movement we have, however,
provides a straightforward solution, if we make a minor adjustment to our
assumptions. We have up to now been marking the source of movement
with a t for trace. Let us consider the possibility that these traces have more
to them than this. Chomsky (1993) suggested that movement was really
an operation of copying, where you don't pronounce the original copy.
So for example, when we do zy/i-movement in a sentence like (4), the LF
of the sentence is really as in (6) where the DP in outline font is the trace
and isn't pronounced.
6) [Which pictures ofhimselfj] didChris; like [Which pictures ofhimselfd?
In this view, movement consists of two parts, a copying operation which
duplicates part of the tree and then puts the copy somewhere in the tree
andthen anoperation that(usually) silences the original. The technical name
for the two DPs in (6) is the chain. Chains are the combination of the moved
copy and any silentoriginals (traces) they leave behind.
With this technology in hand we have a simple account of the timing
dilemma we sketched above. Bindingprinciples all hold at LF. We can claim
that at LF at least one link in the chain (one copy or original) is subject
to thebinding principles. In a sentence like (6), the version of the anaphor
in the trace is c-commanded by Chris, this version is present at LF, it just isn't
pronounced. Binding principle A is met because one copy of the anaphor
is c-commanded by a local antecedent.
In the case of sentences like (2), a different copy of the anaphor is locally
c-commanded by its antecedent. This time it is the moved copy that meets
the binding principle A:
7) ChriSj wants himself;[... [CP ... himselfj to be appealing]]].
426 Advanced Topics
Defining the binding principles over the chains of DPs rather than over DPs
themselves solves this timing problem. We can claim that the binding
principles hold of LF representations and that, in the case of anaphors,
at least one copy must appear in the right binding configuration:
8) Binding Principle A (revised): One copy of an anaphor in a chainmust
be bound within its binding domain.
An exercise at the end of the chapter asks you to consider if the same
property is true of pronouns and Principle B.
You now have enough information to try Challenge Problem Set 1
2. The Definition of Binding Domain
2.1 A Miscellany of Domain Violations
There issignificant reason for thinking thatourdefinition ofbinding domain
in terms of clauses is far too simplistic. Consider the following sentences.
In most major dialects of English, sentence (9b) is ungrammatical with
co-indexation between Heidi and herself, despite the fact that sentence (9a),
which is very similar in structure to (9b) is entirely grammatical. (There is
adialect ofEnglish, spoken mainly inthewestern US, where (9b) isacceptable.
If you are a speaker of that dialect, bear with me and assume that the
judgments given here are correct. Thereis a problem set about this alternate
dialect at the end of this chapter.)
9) a) Heidi; believes any description ofherself;.
b) *Heidij believesMarthak's description ofherself;.
In both (9a) and (9b), the anaphor is c-commanded by a coindexed
antecedent contained in the same clause as the anaphor. By all of
the definitions and constraints we have looked at so far (9b) should
be acceptable, but it isn't.
The next set ofsentences we need to consider has the reverse problem.
Recall from chapter 5 that pronouns must be free within their binding
domain. (10) is ungrammatical with the coindexation given because
the pronoun and its c-commanding antecedent are in the same clause:
10) *Heidi; likes her;.
But a pronoun inside an embedded CP is okay with coindexation or without
it:
11) Heidi; thinks that shei/k has won.
Chapter 15:Advanced Topics in Binding Theory 427
We explained this phenomenon in terms of the clause serving as the binding
domain. Pronouns must be free within their immediate clause. Consider now
the following problem sentence:
12) Heidi likes her violin.
(12)is ambiguous in precisely the same way that (11)is. Hercan be bound by
Heidi or not:
13) a) Heidi; likes her; violin,
b) Heidij likes herk violin.
The interpretation in (13a) is particularly surprising, since her and Heidi
are both dominated by the same CP, so are both in the same binding domain.
The indexation in (13a) should be a violation of principle B, yet the sentence
is entirely acceptable.
To round off our survey of binding puzzles, consider the sentence
in (14). This sentence is acceptable, contrary to all principles of the binding
theory:
14) Heidij thinks that pictures of herselfjshould be hung in the Louvre.
The anaphor is not in the same clauseas its antecedent at all. This should be
a clearprinciple A violation,yet the sentenceis reasonablyacceptable.
2.2 Anaphors
The problem in every case that we have looked at has to do with the
definition of binding domain. Let us start to probe this question by looking
more closely at the difference between (9a and b). The main difference
between the two sentences seems to be the presence of the extra DP Martha
that intervenes between the anaphor and its antecedent in the unacceptable
form. However, not just any intervening DP will do:
15) Heidijgave a present to herselfj.
In (15) a present intervenes between the antecedent and the anaphor
(and furthermore it c-commands the antecedent, but it doesn't intervene
in the binding possibilities the way the middleDP does in (9b). The DP that
causes the problems seems to be the DP in the specifier of another DP
(i.e., the possessor DP). PossessorDPs in thespecifierposition ofanother DP
are a little like the "subject" of those DPs.
16) a) The army's destruction of the palace
b) The army destroyed the palace.
428 Advanced Topics
There seems to be a real parallel between (16a) and (16b). The DP that
interferes with binding seems to be the "subject" of the DP. A similar pattern
with the subject of a TP is seen in control vs. subject-to-object raising
constructions in (17):
17) a) Heidi wants to kiss herself.1
b) *Heidij wants Fred to kiss herselfj.
The ungrammaticality of (17b) seems to be due to the presence of Fred. One
hypothesis that might work is to claim that the binding domain is either a CP
(to account for cases like (9a, 10 or 11) or a DP (as in 9b and 17). But then we
are left with the question of determining which one is appropriate in which
context. The answer to this seems to make reference to this intermediate
"subject." Chomsky (1986b) proposes a revision to the binding theory where
the binding domain of an anaphor can change depending upon whether
there is a potential antecedent (= "subject"). What we find when we look at
sentences like those in (9) is that the binding domain seems to be a DP,
when that DP has a potential antecedent (as in 9b):
9') b) *Heidij believes ([DPMarthak's description ofherselfj].
potential antecedent
This sentence is ungrammatical because the DP here is the binding domain
-it contains the potential antecedent Martha (note, not the actual antecedent,
justa potential2 one). This means thattheanaphor isnotbound initsbinding
domain, and is a violation of principle A.
But now consider (9a) again. Here the anaphor can be bound by Heidi.
The main difference between this sentence and (9b) is that the object DP
contains no potential antecedent for the anaphor. The first potential
antecedent is the actual antecedent. So the binding domain for this sentence
is the whole clause.
9') a) ([CP Heidij believes [DP any description of herselfj].
potential antecedent
1 (17) of course has a PRO in it that does the binding, but it is worth noting
the parallel in any case.
2The notion of "potential" antecedent is very loose. The potential antecedent need
not agree with the anaphor, nor must it even be a semanticallyplausible antecedent.
a) *Heidi;believes Art's description of herselfj.
b) *Heidij dislikes the TV's depiction of herselfj.
The only principle seems to be that it is in the specifier of the DP or TP.
Chapter 15: Advanced Topicsin BindingTheory 429
The surprising result that binding domains for anaphors seem to be able
to shift around depending upon whether there is an antecedent or not
is captured in our revised principle A below:
18) Binding Principle A (final): One copy of an anaphor in a chain must
be bound within the smallest CP or DP containing it and a potential
antecedent.
This version of principle A makes an interesting prediction about the
distribution of anaphors that appear in the subjectposition of an embedded
clause. Let us make the reasonable assumption that an anaphor can't serve as
its own antecedent, or that a DP dominating the anaphor can't serve as that
anaphor'santecedent.3 If we have an embedded clause where the anaphor
isin thesubject, the smallest CPcontaining a subject is the mainclause. This
means that a DP canbind an anaphor in an embedded clauseif that anaphor
is inside the subject position. Quitesurprisingly, this is true: such sentences
are grammatical (19 below and 14 above):
19)([cpHeidij said [CP that [DP pictures ofherselfj] were embarrassing]].
• I
This DP does notcountas a potential antecedent
because it dominates theanaphor.
This CP does not count as the binding domain for the anaphor
because it does notcontain apotential antecedent.
This is the first potential antecedent for the anaphor (it is also the actual
antecedent).
This is the binding domain for the anaphor as it is the smallest CP or DP
containing a potential antecedent.
When we add a possessor within the embeddedsubject, the binding domain
shifts:
20) *[CP Heidi; said UCP that [DP Marthak's pictures of herselfj] were
embarrassing]]. w I
\ potential antecedent
smallest DP orCPcontaining the anaphor anda potential antecedent
This is a truly surprising result, but one that follows directly from the
binding principle in (18).
1This is known as the i-within-i condition, the details of which need not concern us
here.
430 Advanced Topics
Before leaving this topic, it's worth noting that this binding principle
does leave one sentence unexplained, and this is a fairly important sentence
at that. The ungrammaticality of sentence (1) (repeated here as 21) is now
a mystery:
21) *Chris; said [CP that himselfjwas appealing].
According to the principle in (18) this should be acceptable. If himself can't
count as its own potential antecedent, then the smallest CP or DP containing
a potential antecedent for the anaphor is the main clause (with the actual
antecedent Chris serving as the potential antecedent). This means that
himself would be bound within its binding domain so the sentence should
be grammatical contrary to fact. In order to account for (21) we are going
to have to appeal to something other than the binding principle in (18).
Fortunately, there is a relatively simple solution to this problem. The
anaphor in (21) is in the specifierofTP, this is the position where nominative
case isassigned. Notice that English does nothaveany nominative anaphors
(*heself, *sheself, *Iselfetc.) Perhaps the ungrammaticality of (21) is not due to
any binding principle violationsbut is a simplecase conflict instead. Himself
is accusative in case, but it is in a position that is associated with nominative
case.
You can now try General Problem Set 1and Challenge Sets 2&3
2.3 Pronouns
Our definition of binding domain as the smallest CP or DP containing
a potential antecedent seems to work well for anaphors, but unfortunately
it doesn't fare so well for our examples with pronouns. Take the examples
in (13) (repeated here as 22):
22) a) Heidij likes herj violin,
b) Heidij likes herk violin.
Hereagain we have a casewherea DPisactinglikea binding domain. Recall
that pronouns must be free within theirbinding domain. In order toexplain
the grammaticality of (22a), the pronoun must be in a different binding
domain than its antecedent. The obvious candidate for this is the DP
[her violin]. But in the previous section we argued the binding domain
was the smallest DP or CP containing a potential antecedent. Assuming
that pronouns can't be their own antecedents, the DP [her violin] contains no
such potential antecedent, so it can't be a binding domain. By the potential
antecedent definition the binding domain is the whole CP, which would
Chapter 15:Advanced Topicsin BindingTheory 431
mean that in (22a) the pronoun would be bound by Heidi within its domain
in violation of principle B. Yet the sentence is grammatical. Chomsky (1986b)
came up with an ingenious solution to this problem. He suggested that that
binding domains for pronouns and anaphors are defined differently. The
difference lies in the inherent nature of the DP types. Anaphors are DPs that
want to bound so they are going to look for the closest potential antecedent.
Pronouns by contrast want to be free! So they are going to look for structures
without an antecedent if they can find them. So the DP [her violin]
is the smallest DP or CP that does not contain a potential antecedent:4
23) [CP Heidij likes( [DP herj violin]].
V tsmallest DP orCPnotcontaining a potential antecedent
This constraint is encoded in (24):
24) Binding Principle B (nearly final): A pronoun must be free within
the smallest CP or DP containing it but not containing a potential
antecedent.
The fact that binding domain is defined differently for pronouns
and anaphors not only reflects that they are different animals with different
requirements, but more importantly explains the contrasts outlined
in section 2.1.
Butthe definition in (24) still has a problem.Under this definition there
isno way to explain the ungrammaticalityof (25):
25) *Heidij likes her;.
This is because under (24) there is no way to define a binding domain for her.
In (25) there is no DP or CP that contains her, is not her itself, and does
not contain a potential antecedent. Thebinding domain for (25) is undefined
under (24). In order to correct for this we need to add a rider clause to (24),
such that if no CP or DP without a potential antecedent can be found,
then the root (main clause) CP counts:
26) Binding Principle B (final): A pronoun must be free within the smallest
CP or DP containing it but not containing a potential antecedent. If no
such category is found, the pronoun must be free within the root CP.
With this definition, the ungrammaticality of (25) is explained. The main
clause CP acts as the binding domain, and the pronoun is bound within
this domain in violation of principle B.
4Violinisn't a potential antecedent as it doesn't asymmetrically c-command her, and
neither her nor the DP containing her can be the potential antecedent.
432 Advanced Topics
This approach makes a very interesting prediction. Recall sentence (19)
from above, repeated here as (27):
27) [CP Heidi; said [CP that [DP pictures of herselfj] were embarrassing]].
One surprising fact is that for most speakers, anaphors like those in (27)can
freely alternate with pronouns:
28) [CP Heidij said [CP that [DP pictures of he^] were embarrassing]].
Under our old theory - where the binding domains for pronouns and
anaphors were identical - the fact that both (27) and (28) are grammatical
would be a real puzzle. Under the older approach, pronouns and anaphors
were by definition in complementary distribution (pronouns had to be free
in their clause, anaphors had to be bound in their clause). The fact that (27)
and (28) can both exist shows that the domains for the binding priaciples
are more nuanced. The definitions have to allow for a situation, where
the anaphor in (27) is bound by Heidi in its binding domain, but where the
pronoun is free in its binding domain in the structurally identical (28). But
if binding domains are defined relativistically (relative to the type of the DP
involved), then (27) and (28) do not form a contradiction. In (27) the smallest
DP or CP containing a potential antecedent is the main clause CP. In (28)
the smallest DP or CP not containingan antecedent is the DP [pictures ofher].
So the anaphor in (27) can be bound in its domain, while the pronoun
in the exact same position in (28) can be free in its domain.
You now have enough information totry General Problem Set2
Ideas, Rules, and Constraints Introduced in this Chapter
i) The Copy Theory of Movement: Movement is a two-part operation.
First the moved element is copied and put into the surface position;
second the original is made silent (but is still structurally present).
ii) Chain:The moved copy and all its traces.
iii) Potential Antecedent: A DP in the specifier of TP or another DP.
The potential antecedent cannot be the anaphor or pronoun itself,
nor can it be a DP that contains the anaphor or pronoun.
iv) BindingPrincipleA (final): One copy of an anaphor in a chain must
be bound within the smallest CP or DP containing it and a potential
antecedent
Chapter 15:Advanced Topicsin Binding Theory 433
v) Binding Principle B (final): A pronoun must be free within
the smallest CP or DP containing it but not containing a potential
antecedent. If no such CP or DP can be found, the pronoun
must be free within the root CP.
Further Reading
Buring, Daniel (2005) Binding Theory. Cambridge: Cambridge University
Press.
Chomsky, Noam (1986b) Knowledge of Language: Its Nature, Origins and Use.
New York: Praeger.
General Problem Sets
1. Binding Domain for Anaphors
[Application of Skills; Intermediate]
Draw the trees for each of the following sentences, then identify the binding
domain of the anaphors. For the ungrammatical forms, explain why
the sentence is ungrammatical. In all cases assume that John and himself
are coindexed. Assume the judgments given:
a) John loves himself.
b) John loves pictures of himself.
c) *John loves Mary's pictures of himself.
d) *John thinks that Mary loves himself.
e) *John thinks that Mary's depiction of himself is wrong.
f) John thinks that most depictions of himself are wrong.
g) Which pictures of himself does John like?
h) John seems to like pictures of himself.
i) John believes himself to be the best at baseball,
j) John wants to congratulate himself.
2. Binding Domain for Pronouns
[Application of Skills; Intermediate]
Draw the trees for each of the following sentences, then identify the binding
domain of the pronouns. For the ungrammatical forms, explain why the
sentence is ungrammatical. In all cases assume that John and the pronoun
are coindexed. Assume the judgments given:
a) *John loves him
b) John loves his puppy.
434 Advanced Topics
c) John asked if the unflattering description of his work would be published
in the paper.
d) John asked if his essay would be published in the paper.
e) *John wants to kiss him.
f) *John believes him to be fantastic.
Challenge Problem Sets
Challenge Problem Set 1: Pronouns
[Critical Thinking; Challenge]
We argued above that at least one link of a movement chain containing
an anaphor must meet principle A of the binding theory and be bound
within its binding domain. Is this true for pronouns as well? Provide examples
to support your answer.
Challenge Problem Set 2: Possessive Pronouns
[Critical Thinking; Challenge]
Up to now we've treated possessive pronouns as being of category D.
One alternative is that pronouns like his are really bimorphemic and take
the form below:
DP
DP D'
A
he D NP
A
And then there are morphological rules that turn he's into his (and she's
into her, etc.). Using our definition of "potential antecedent," how does the
following sentence argue for the tree above instead of treating his as a D?
Assume the judgment given.
*Mary doesn't like his pictures of himself.
Challenge Problem Set 3: Western American Dialects of English
[Critical Thinking;Challenge]
For most speakers of English the sentence below is ungrammatical with the
indexation given. However, there is a significant dialect area in the western
United States (Andy Barss has found speakers from Arizona, California
and New Mexico that all have this judgment) where this sentence is typically
judged as fully acceptable. What minor adjustment must we make
to principle A to explain the grammaticality of this sentence in this dialect?
Heidij doesn't like Nate's pictures of herself;.
if)
03
0-
Alternatives
C&faMmAiCT Lexical-Functional
Grammar
0. Alternative Theories
The first fifteen chapters of this textbook are an introduction to syntactic
theory from one particular perspective. That is the perspective of the
Chomskyan Principles and Parameters (P&P) approach (and its descendant:
Minimalism). While a large number of syntacticians (perhaps even
a majority) operate using theassumptions and formalisms we've developed
so far in this book, not everyone does. In this chapter and the next, we look
at two other popular formalisms: Lexical-Functional Grammar (LFG)
and Head-Driven Phrase Structure Grammar (HPSG).
In many ways, these theories have the samebasicgoals and assumptions
as P&P syntax. LFG and HPSG are considered to be generative grammars,
just like P&P. Where all these theories differ is in the precise formulation
of the rules and constraints. We will have something to say about choosing
among formalisms at the end of chapter 17, but to a great degree it comes
down to a matter of the range of phenomena one wants to account
for and one's preferred means of formal expression. My inclusion of these
approaches is not meant to imply that theyarebetter than P&P, nor to imply
that P&P is better than them. They should simply be viewed as alternatives.
Asa beginning syntactician, you mightwonder why you should bother
looking at alternative approaches. After all, a significant part oftheliterature
and research being done in syntax uses the assumptions and formalisms
developed in the first fifteen chapters. Shouldn't we just pick one formalism
438 Alternatives
and stick with it? To be honest, most researchers do just this; they do their
work within only oneformalism. Forexample, I do almost all of my research
within the P&P approach. But this doesn't mean that I shouldn't be familiar
with other formalisms too. An important body of work is conducted in these
formalisms, and their results are oftendirectly relevant to work being done
in Chomskyan P&P syntax. Being able to interpret work done in these
alternative approaches is a very useful skill (and unfortunately, one rarely
taught to beginning syntacticians). The results found in other approaches
to syntax have oftenaffected the development of P&P theory. Forexample,
the lexical approach to passives (whereby the passive morphology affects
the thematic and case assigning properties of the underlying verbal
morphology) replaced an earlier purely transformational approach
to passives. This idea was borrowed from LFG. Similarly, the idea offeature
checking is directly related to the notion of unification found in both LFG
and HPSG. So I encourage you to keep an open mind and consider
the advantages of being familiar with more than one formalism.
1. C-STRUCTURE
One major part of LFG is almost identical to the approach taken in the rest
of the book. This is the idea that the words of a sentence are organized
intoconstituents, which are represented by a tree, and generated by rules.
In LFG, these trees are called thec-structure, and are roughly equivalent to
the S-structure inP&P.1 Many LFG theorists adopt X-bar theory, butitisnot
as strictly applied to all languages. For example, LFG posits a flat (VP-less)
structure for many VSO languages (Kroeger 1993) and "free word order"
or "non-configurational" languages like Warlpiri (Simpson1991). This said,
most LFG c-structures look just like the S-structure trees we have built
elsewhere in this book. There is one major exception to this: since there is
no movement, there are (for the most part) no traces (nor are there any
other null elements).
Because LFG doesn't use transformations, there is no D-structure. Phrase structure
rules directly build the c-structure (= S-structure). Displaced items are dealt with
in other ways. See below.
Chapter 16: Lexical-Functional Grammar 439
2. FUNCTIONS
As you might guess from its name, there are two driving forces in LexicalFunctional
Grammar: the lexicon (which we explore in section 3 below)
and functions. The notion of function is borrowed from math and computer
science. Afunction isa rule that maps from one item to another. There are
really two kinds of functions inLFG, which can be abitconfusing. The first
kind arecalled grammaticalfunctions andarethings like subject, object, etc.
We called these grammatical relations in chapter 4. When a practitioner
of LFG talks about functions, they are primarily talking about grammatical
functions. The other kind of function refers to the principles that map
between the different parts of the grammar, such as the mapping between
the c-structure and the structure that represents the grammatical functions.
This is called anf-structure. We are going to look at grammatical functions
here, then turn to the mapping relations in section 4.
InP&P syntax, grammatical functions orrelations areread off ofthe tree.
That is, you know what the subject of a sentence is by looking for the NP
thatis in thespecifier ofTP. In LFG, grammatical functions are not defined
bya tree; instead, they are primitive notions (meaning they can't bederived
somehow). Every sentence has an f-structure that represents grammatical
functions. In the f-structure, a particular NP will be identified as being
thesubject ofthesentence, quite independent ofthetreestructure associated
with the sentence. In the sentence Diana loves phonology, Diana is equated
with the SUBJ grammatical function. This equation is usually represented
in what is called an Attribute Value Matrix (AVM); the item on the left
is the attribute or function, the item on the right is the value attributed
to that function:
1) [SUBJ [pred3 'Diana']]
Attributes can have various kinds of values, including other AVMs.
For example, the value for the SUBJ function in sentence (2) is itself
a matrix containing other functions:
2) The professor loves phonology.
2 If you are unfamiliar with this notion you might want to consult a good
"mathematics forlinguists" textbook, suchasAllwood, Andersson and Dahl(1977).
3The term pred here is a bit confusing, since 'Diana' is an argument of the clause.
Pred herecanbe,very loosely, translatedas thesemantichead ofthe AVM.
440 Alternatives
3)
SUBJ pred 'professor'
DEF +
_NUM sng
Inthe embedded AVM, the function PRED tells you the lexical content ofthe
subject NP, DEF tells you ifitisdefinite ornot, NUM tells you thenumber, etc.
These are all properties of the subject. You'll notice that more than just
grammatical functions are represented in these structures; various kinds of
other features (likedefiniteness, etc.) are as well.
Once you combine all the AVMs for all the parts of a sentence you get
the f-structure, containing all the sentence's featural and functional
information. Where does all this featural information come from? Most of
the information that is combined into the f-structure comes from the lexical
entries ofall the words inthe sentence. The lexicon thus plays animportant
role in this theory.
3. The Lexicon
The lexicon is where a lot of the work in LFG is done. All the information
that ends up in an f-structure starts out in the lexical entries of the words
that compose the sentence.
The rough equivalent to a theta grid in LFG is the a-structure
or argument structure.4 As in P&P syntax, the a-structure is part of
the lexical entry for the predicate. Simplifying somewhat, the lexical entry
for the inflected verb loves isseen in(4). For the moment, ignore the arrows;
I'll explain these below in section 4.
4) loves: V (Tpred) = 'love<(Tsubj),( Tobj)>'
(Ttense) =present
(Tsubj NUM) =sng
(Tsubj pers) =3rd
This lexical entry says that love is a verb that means 'love', and takes
two arguments, an obligatorysubject and an obligatoryobject (as contained
within the < > brackets). The parentheses here do not mean "optional" as
both arguments are, indeed, obligatory. It also tells us that loves is
This is a gross over-simplification, I'm ignoring a large number of mapping
operations here. I'm also ignoring the complexities of inflectional morphology and
how they interact with the featural structure. See Bresnan (2001) or Falk (2001)
for extensive discussion.
Chapter 16:Lexical-Functional Grammar 441
the present tense form. (The lexical entry for loved would have a different
value for TENSE.) Finally it tells us that the subject is third person singular
((Tsubj num) means "my subject'snumber is ...").
All lexical items bring some information to the sentence. For example,
we know that the determiner the is definite and such information
is contained in its lexical entry.
5) the: D (Tdef) = +
Functional and featural information comes with the lexical item when it is
inserted into the c-structure (as we will see below).
4. F-STRUCTURE
F-structures, as noted above, are the set of all the attribute value pairs for
a sentence. Perhaps the easiest way to see this is to look at an example.
We use again the sentence the professor loves phonology. An f-structure for
this sentence is given in (6):
6) PRED 'love <SUBJ,OBJ>'
TENSE present
SUBJ DEF +
NUM sng
PRED 'professor^
OBJ [PRED 'phonology'
The topmost PRED function tells you what the predicate of the sentence is.
It also contains information about the a-structure of the sentence. The TENSE
feature tells you about the tense of the sentence. The SUBJ and OBJ functions
have submatrices (containing information on their internal structure)
as values.
C-structures must be related to f-structures somehow. This is
accomplished with the use of variables. Consider the simple c-structure
in (7). Each lexical item is followed by the information it contributes by
virtue of its lexical entry. Each node in the tree is marked with a variable
(fvfvh> •••' etc-)- These will be used in mapping to the f-structure. Again,
ignore the arrows for the moment.
442
7)
the
(Tdef) =+
Alternatives
TP/,
NP/2 yp/3
fs
r
N/7
professor
(Tpred) ='professor'
(Tnum) =sng
loves
(Tpred) ='love <...>'
(Ttense) =present
(Tsubj num) =sng N/„
(Tsubj pers) =3rd phonology
(Tpred) ='phonology'
Each of these variables corresponds to a pair of matrix brackets in the
f-structure. There is no one-to-one correspondence here. Multiple nodes
in the tree can correspond to the same (sub)AVM:
8)
fvUfff* FRED 'love <SUBJ, OBJ>'
tense present
SUBJ f2,f4,fStf7, DEF +
NUM sng
_pred 'professor^
OBJ fsifwfiu [pRED 'phonology']
This means that the information contained in nodes f2, f4, f5, f7 contribute
the SUBJ features to the sentence./9,/,0,/,„ provide the OBJ info,etc.
The mapping formally happens through what is called anf-description.
The f-description is set out with functional equations. These equations
tell us, for example, that the subject of the sentence f corresponds
to the constituentf2. This is written as:
9) (f,SUBj)=/2
Thefact that thesubject NP(f2) isdefinite isencoded in Dnode(f4):
10)/2=/4
When a node is a head, or simply passes information up the tree (e.g., V
or V), then a simple equivalence relation is stated:
NP/9
V'fio
Chapter 16:Lexical-Functional Grammar 443
ll)/3=/6
These functional equations control how information is mapped through
the tree and between the tree and the f-structure. Each piece of information
in the f-structure comes from (various places in) the tree as controlled
by the functional equations in the f-description.
In order to make this clearer, as a notational device, c-structures are
often annotated with their functional equations. There is a useful device that
is used to clarify theseannotations. These are metavariables. Metavariables
are variables that stand for other variables. There are two metavariables:
12) a) I means "this node"
b) Tmeans "my mother" (immediately dominating node)
So the equation T=i means "all of the features I have also belong to my
mother" - in other words, a head. Thenotation (TsUBj)=i means "I represent
the subjectfunction of my mother".
We also saw these arrows in lexical entries. They mean the same thing
here. (Tpred) = 'love' means "the terminal node that I fill has the predicate
value of 'love'."
Here is the c-structure in (7) repeated below as (13) with the functional
equations annotated using metavariables:
13) TP/,
(Tsubj) =1 Ui
fsJP^ yp/3
T=i T=i T=4
D/4 NY5
the 1
JC£_
(Tdef) =+ 1=1 Ta (Tobj)=>1
N/7 V/s NjP/9
professor loves 1
(Tpred) = 'professor' (Tpred) ='love <...>' T=i
(Tnum) = sng (Ttense) = present N,Y,o
(Tsubj num) == sng 1
(Tsubj pers) ==3rd T=4
phonology
(Tpred)=='phonology'
Unification is the idea that features and functions coming from
disparate places in the tree must be compatible with one another. Take, for
444 Alternatives
example, the fact that the verb specifies that its subject must be third person
singular (as marked by the suffix -s). If the subject has number or person
features, then those features must match the verb's subject features in the
f-structure. This is forced by the fact that the f-structure (in mathematical
terms) solves for, oris the solution to, the setoffunctional equations known
as the functional description. If the features didn't match, then the f-structure
wouldn't unify. Notice that this is a very similar notion to the idea
offeature checking, discussed in chapter 12. Feature checking also ensures
compatibility of features coming from different parts of the sentence.
Minimalism simply uses movement, rather than functional equations,
toguarantee this. Both systems have their advantages; LFG's system has a
certainmathematical precision and elegance that P&P movement and feature
checking do not. By contrast, P&P/Minimalism is able to derive word
order differences between languages from feature checking and movement.
Minimalism thusprovides a slightly more explanatory theory ofwordorder
than LFG, which uses language-specific phrase structure rules. It is not
at all clear which approach is preferable.
There are a number ofconstraints onf-structures. Three ofthese conspire
together to result in the LFG equivalentof the theta criterion:
14) a) Uniqueness: Ina given f-structure, a particular attribute may have at
most one value.
b) Completeness: An f-structure must contain all the governable
grammatical functions that its predicate governs.
c) Coherence: All the governable grammatical functions in an
f-structure must be governed by a localpredicate.
(14a) is also the constraint that forces unification. All f-strucrures must meet
these constraints.
4.1 WhyF-structures?
We now have a fairly detailed sketch of the basics of LFG. Before turning
to some implementations of the model dealing with some of the empirical
issues we've looked at elsewhere in this book, it is worth considering
why the LFG model uses f-structures. The answer is fairly straightforward:
Information about a particular grammatical function may come from more
than one place in the tree and, more importantly, the sources ofinformation
do not have to be constituents. Falk (2001) gives the example of the pair
of following sentences:
15) a) The deer are dancing,
b) The deer is dancing.
Chapter16: Lexical-Functional Grammar 445
The form of the subject noun is identical in both of these sentences.
However, in terms of meaning, it is plural in (15a) and singular in (15b).
Let's assume that the lexical entry for the form 'deer' lacks any specification
for number:
16) deer N (Tpred) = 'deer7
The number associated with the subject function in the sentences in (15)
comes from the auxiliary:
17) a) are T (Ttense) = present
(Tsubj num)= pi
b) is T (Ttense) = present
(Tsubj num) = sg
While the number comes from the auxiliary, it is only really a property of
the subject. The num feature still gets mapped to the SUBJ function, because
of the functional annotation (Tsubj NUM) = pi. Similar facts are seen in "free
word order" or "non-configurational"languages, like the aboriginal Warlpiri
language spoken in Australia. Contra the principle of modification we
discussed in chapter 3, in Warlpiri, words that modify one another do not
have to appear as constituents on the surface. Since information can come
from various parts of the tree, this points towards a system where functional
information is not read directly off the tree; instead, an f-structure-like level
is motivated.5
You can now try General Problem Set 1
5. Assorted Phenomena
Having now quickly laid out the fundamentals of LFG (c-structure,
f-structure, the lexicon, etc.), we turn to how various phenomena discussed
in other parts of this book are treated in LFG.
5It is worth briefly mentioning how P&P deals with thesesame facts. At D-structure,
these units do form constituents. A transformation, known as scrambling (see
the problem sets in chapter 11), then reorders the elements so they don't surface
as constituents. Both approaches achieve essentially the same results with different
underlying assumptions.
446 Alternatives
5.2 Head Mobility
Inchapter 9, weanalyzed alternations between theposition ofa tensed verb
and the position of auxiliaries in languages like French, Vata, or Irish
as involving head-to-head movement. In French, the main verb alternates in
its position relative to adverbs, depending upon the presence or absence of
an auxiliary. When anauxiliary ispresent, theverb stays initsbaseposition.
When there is no auxiliary, the verb moves to T. Although LFG has no
movement, it has a related account ofthese phenomena. LFG simply posits
that tensed verbs and untensed participial forms belong to different
categories. This is called head mobility. Tensed verbs are of category T,
whereas untensed forms are of category V:
18) a) mange T
b)
c)
19) a)
mange
ai
V
T
(Tpred) = 'eat <(Tsubj), (Tobj)>'
(Ttense) =present
(Tpred) = 'eat <(Tsubj), (Tobj)>'
(Ttense) = present6
AdvP
A
souventV
mang£
VP
V
V
NP
des pommes des pommes
You'll notice that in (19a) the VP has no head V. This is allowed in LFG, but
not in P&Ptheory (as it violatesthe endocentricproperties of X-bar theory).
In English, both participles and tensed verbs are of the category V; only
auxiliaries are of category T:
20) a) eat V (Tpred) = 'eat <(Tsubj), (Tobj)>'
(Ttense) =present
b) eaten V (Tpred) = 'eat <(Tsubj), (Tobj)>'
c) have T (Ttense) =present
6 The sentence J'ai souvent mange des pommes also bears what are called aspect
features. We leave these aside here.
Chapter 16: Lexical-Functional Grammar 447
This means that both the participle and the tensed form will appear inthe
VPand no outward appearance ofhead movementwill arise.
5.2 Passives
LFG's basic grammatical functions allow us to do passives in just one step,
which all happens in the lexicon. In LFG there is no syntacticcomponent to
passives. Instead there isasimple lexical change associated with the passive
morphology:
21) a) kiss V (Tpred) ='kiss<(Tsubj),(Tobj)>'
+en y ^
b) kissedpassV (Tpred) ='kiss < 0 (Tsubj) >'
When the lexical entry in (21b) is inserted into a c-structure, the original
object is directly placed intothesubject position. There isno movement.
5.3 Raising and Control
Before getting into LFG's analysis of raising and control constructions, we
need to introduce the theory's treatment of non-finite complements. In most
versions of LFG, these are not treated as CPs. Instead, they are most often
treated asVP constituents. The special category VP' hasbeencreated to host
to. Nearly identical c-structures areused for bothraising and control:
22)
A VP'
likely/reluctant
to VP
I
V*
I
V
dance
448 Alternatives
The differences between control and raising constructions are expressed
in lexical entries and f-structures, not in the phrase structure (c-structure)
as in P&P.
Let's start with control constructions.7 LFG uses a special grammatical
function to represent non-finite complements. Since embedded predicates
in these constructions don't have their subject expressed overtly, they
are considered open functions (thatis, they are missing an argument). The
grammaticalfunction used to represent open functions isXCOMP (orXADju in
thecase ofadjuncts). Predicates like islikely and isreluctant select forXCOMPS
intheir lexical entries. The lexical entry for reluctant isgiven in (17):
23) reluctant A (Tpred) ='reluctant<(Tsubj), (Txcomp)>'
Reluctant takes a thematic subject and an open function as its arguments.
When wecombine thiswith a predicate such as dance, weget an f-structure
like the following for the sentence Jean is reluctant todance:
24) *[SUBJ [PRED 'Jean' ]
PRED 'reluctant<(TsUBj), (TxCOMP)>'
TENSE present
XCOMP [SUBJ ?????? ~~|
_ LPRED 'dance <(Tsubj)>J
This f-structure is ill-formed, because it violates the principle of
completeness: The subject function of the XCOMP is not filled. This is resolved
using the LFG equivalent of control:functional control. Functional control
is indicated with a curved line linking the two functions.
25) SUBJ [ pred 'Jean' ]PRED
'reluctant <(TsUBj), (TxCOMP)>'
TENSE present
xcomp Tsubj [ j—•— -i
[pred 'dance <(Tsubj)>'J
This indicates that the subject of the main clause is also the subject of
the non-finite complement. Functional control is licensed by the lexical
I am simplifying the situation here. The range of phenomena P&Ptheory groups as
"control" constructions divides into two groups in the LFG analysis: functional
controland anaphoric control,which have differentproperties. We won't distinguish
these here.SeeFalk(2001) or Bresnan(2001) forexplanationsof the difference.
Chapter 16: Lexical-Functional Grammar 449
entry of the main clause verb. Here is arevised lexical entry for reluctant that
contains a control equation (astatement that licenses the curved line in(25):
26) reluctant A (Tpred) ='reluctant <(Tsubj), (Txcomp)>'
(Tsubj) = (Txcompsubj)
In English, the second line of this entry stipulates that the subject
of the main predicate is identical to the subject of the XCOMP. Notice that
this isessentially the equivalent ofthe thematic analysis ofcontrol discussed
in chapter 14, and suffers from the same empirical problems as a thematic
analysis (see the discussion inchapter 14 toremind yourself ofthese facts).
In control constructions, the controller must f-command the controllee.
One node f-commands another if it is less deeply embedded in the
f-structure (inside fewer square brackets).
Interestingly, raising constructions have a similar analysis. They also
involve functional control. The difference between them and more
traditional control constructions is simply that the subject argument of
the raising predicate is non-thematic (doesn't geta theta role - indicated by
writing it outside the< >brackets), and it is linked to a particular argument
in its complement's argument structure (as indicated by the second line
of the lexical entry). To see this, consider the lexical entry forlikely:
27) likely A (Tpred) ='likely <(Txcomp)> (Tsubj)'
(Tsubj) = (Txcompsubj)
Thefirstline ofthis lexical entryputs the subject argument outside theangle
brackets (< >), indicating that it doesn't get a theta role. The second line
specifies that the argument which fills this function is the thematic subject
oftheopenfunction. Again, in thef-structure thisisindicated witha curved
line. The following is the f-structure for Jean is likely todance:
28) subj [ pred 'Jean' ]-
PRED
TENSE
XCOMP
'likely <(TxCOMP)>, (TsUBj)'
present
"subj [ ] ~|
_PRED 'dance <(TsUBj)>d
The difference between the raising and the control sentence simply reduces
to a matter of argument structure.
450 Alternatives
5.4 Wh-movement: Long Distance Dependencies
The sharing of feature structures (as expressed by the curved line) is also
used to define long distance dependencies such the relationship between
a wh-phrase and the gap (or trace) it isassociated with. There is a special
grammatical function: FOCUS, which is found in w/i-constructions. I'll abstract
away from the details, but in English, this function is associated with the
specifier of CP. The element taking the FOCUS function must share features
with some argument (this is forced on the sentence by the constraint
of coherence). The following is the f-structure for the sentence Which novel
do you think Anne read? (COMP is the function assigned tensed embedded
clauses.)
29) FOCUS PRON wh
PRED 'book'
_NUM Sg _
TENSE present
PRED 'think<(Tsubj), (Tobj)>'
SUBJ _["you"]
COMP SUBJ ["Anne"]
TENSE past
PRED 'read <(Tsubj), (Tobj)>'
-OBJ [ ] - ""J
The FOCUS shares thefeatures oftheOBJ function ofthe complement clause,
indicating that they are identical.
There is much more tow/i-dependencies inLFG thanthissimple picture.
For example, LFG has an account of island conditions based on f-structures.
There is also an explicit theory of licensing iy/i-dependencies. For more on
this see the readings inthe further reading section atthe end ofthe chapter.
You can now try Challenge Problem Sets 1-3
6. Conclusion
This concludes ourwhirlwind tour ofLexical-Functional Grammar. Asingle
chapter does notgive menearly enough space to cover therange and depth
ofthistheory, and as suchdoesn'tgive LFG the opportunity toreally shine.
I've glossed over many details and arguments here, but I hope this chapter
gives you enough information to pursue further reading in this alternative
framework. Remember that the results derived in frameworks like LFG
Chapter 16:Lexical-Functional Grammar 451
can give us insights about the way human language works. If we're clever
enough, we should be able to incorporate these insights into the P&P or
Minimalist framework, too. Similarly, a student of LFGshould be able to use
the insights ofMinimalism orP&P to inform their theorizing. Remember that
syntactic theories are only hypotheses about the way syntax is organized.
We really don't know whichapproach, ifany ofthese, isright. Soby looking
at other theoretical approaches we question our basic assumptions,
and consider new alternatives.
a_
Ideas, Rules, and Constraints Introduced in this Chapter
i) C-structure: Constituent structure. The tree in LFG. Roughly
equivalent to S-structure in P&P.
ii) Grammatical Function: Same thing as a grammatical relation.
Common grammatical functions: SUBJ = subject; OBJ = object; PRED
= predicate; XCOMP = open complement (non-finite clause); COMP
= closed complement(finite embedded clause); FOCUS = the function
associated with wh-phrases
iii) F-structure: The level of representation where grammatical functions
are unified.
iv) Attribute Value Matrix (AVM): A matrix that has an attribute
(or function) on the left and its value on the right. The set of all
AVMs for a sentence form the sentence's f-structure.
v) A-structure: Argument structure. The LFG equivalent of the theta
grid.
vi) Variables: LFG uses variables (fv f2, f3, ..., etc.) for each node on
the c-structure which are used in the mapping between c-structure
and f-structure.
vii) Functional Equation: An equation that maps one variable to another
(e.g.,(/,subj) =f2says that/2 maps to/,'s SUBJ function).
viii) F-description: The set of all functional equations. Defines the
mapping between c-structure and f-structure.
ix) Annotated C-structure: A c-structure annotated with the functional
equations which map it to the f-structure.
x) Metavariable: A variable over variables. T = my mother's variable,
X= my variable.
452 Alternatives
xi) T=i: "All the functional information I contain, my mother also
contains."
xii) (TsUBj)=i, (ToBj)=i: "Iam the subject of the node that dominates
me" or "I am the objectof the node that dominates me."
xiii) Unification: All the features and functions associated with the
f-structure must be compatible. (Similar to feature checking in P&P.)
xiv) Uniqueness: In a given f-structure, a particular attribute may have
at most one value.
xv) Completeness: An f-structure must contain all the governable
grammatical functions that its predicate governs.
xvi) Coherence: All the governable grammatical functions in an
f-structure must be governed by a localpredicate.
xvii) Head Mobility: The idea that lexical items can take different
categories depending upon their features. E.g., a tensed verb in
French is of category T, whereas an untensed one is a V. This derives
head-to-head movement effects.
xviii) Lexical Rule of Passives: Passives in LFG are entirely lexical. There
is no syntactic movement:
xix) Open function (XCOMP): A function with a missing argument (e.g.,
a non-finite clause).
xx) Functional Control: The LFG equivalent of control, indicated with
a curved line linking two AVMs in an f-structure.
xxi) Raising vs. Control: In LFG raising vs. control reduces to a lexical
difference. The SUBJ function in raising constructions isn't thematic,
but is in control constructions.
Further Reading
Bresnan, Joan (2001) Lexical-Functional Syntax. Oxford: Blackwell.
Dalrymple, Mary, Ronald Kaplan, John Maxwell and Annie Zaenen (eds.)
(1995) Formal Issues in Lexical-Functional Grammar. Stanford:
CSLI Publications.
Falk, Yehuda N. (2001) Lexical-FunctionalGrammar: An Introduction to Parallel
Constraint-Based Syntax. Stanford: CSLI Publications.
Chapter16: Lexical-Functional Grammar 453
Kaplan, Ronald (1995) The formal architecture of Lexical-Functional
Grammar. In Mary Dalrymple et al. (eds.), Formal Issues in LexicalFunctional
Grammar. Stanford: CSLI Publications, pp. 7-27.
Kaplan, Ronald and Joan Bresnan (1982) Lexical-Functional Grammar:
A formal system for grammatical representation. In Joan Bresnan (ed.),
The Mental Representation ofGrammatical Relations. Cambridge: MITPress,
pp. 173-281.
A lot of material on LFG can be found on the web:
http:/ / www-lfg.stanford.edu / lfg
http:/ / www.essex.ac.uk/ linguistics/LFG/
Further reading on other theories:
Edmondson, Jerold and Donald A. Burquest (1998) A Survey of Linguistic
Theories (3rd ed.). Dallas: Summer Institute of Linguistics.
Sells, Peter (1985) Lectures on Contemporary Syntactic Theories. Stanford: CSLI
Publications.
General Problem Set
1. English
[Application of Skills; Intermediate]
Draw the annotated c-structures and f-structures for the following sentences.
(It may also be helpful to write out lexical entries for each of the words
detailing the information each word contributes):
a) Susie loves the rain.
b) Joan thinks that Norvin is likely to write a paper.
c) What have you read?
Challenge Problem Sets
Challenge Problem Set 1: Icelandic (Again)
[Critical Thinking; Challenge]
Go back to the questions on quirky case in Icelandic in previous chapters
and review the data. These data caused problems for us with our case driven
theory of movement and our theory of PRO. Do these same problems arise
in LFG? Why or why not?
454 Alternatives
Challenge Problem Set 2: Transformations or Not?
[Critical Thinking; Challenge]
Construct the design for an experiment that would distinguish between
a transformational approach, and a non-transformational approach like LFG.
Challenge Problem Set 3: Wanna-co/vt/?/»ct/ow
[Critical and Creative Thinking; Challenge]
How might LFG account for wa/wa-contraction (see chapter 11) if it doesn't
have movement or traces?
Head-Driven Phrase
Structure Grammar
0. Introduction
Another major formalism for syntactic theory is Head-Driven Phrase
Structure Grammar or HPSG. HPSG is also a generative theory of grammar.
It shares with P&P and LFG the goal of modeling how human Language
is structured in the mind. HPSG and LFG in particular have many things
in common. Forexample they both make use of a highly enriched lexicon,
and the Attribute Value Matrix (AVM) notation we saw with LFG.1
As with our discussion of LFG, a short chapter like this cannot hope to
properly cover therich variety ofwork done in HPSG. Inorder togeta fuller
picture you'll need tolook at some oftheprimary sources ofmaterial listed
in the further reading section at the end of this chapter; in particular, Sag,
Wasow and Bender (2003) is a very accessible work. Another small caveat
is in order before we launch into the details of the theory. For pedagogical
reasons, I have couched the presentation here so that someonewho has read
the first 15 chapters of this book can relate the material here to what
they already understand. Sometimes in order to do this, I've had to use
metaphors and analogies that many practitioners of HPSG would disagree
with. For example, I often state that some theoretical device in HPSG is
the "equivalent" ofsomethingelse inP&P or LFG. By this, I generallymean
"does roughly the same kind of work;" I do not mean that they are
necessarily notationally or empiricallyequivalent - as they are not.
With some significant differences in notation and assumptions.
456 Alternatives
1. Features
The basic tool of linguistic description in HPSG is features. There are
acouple ofnotarional systems for features; Iadopt here the one used inSag,
Wasow and Bender (2003).2 Much of the argumentation in this chapter
is also taken from that book.
Features enable linguistics to talk about such information asthe category
of a word, what other words it must appear with (i.e., its theta grid), and
what level inthe tree the node is(in HPSG, barlevels aretreated asfeatures).
As in LFG, features arepaired with a value inanAVM, andagain like LFG,
features can take feature structures as values (1):
1) [AGR [num pi]]
The AVM in (1) says that the agreement feature for the word involves
a number feature that is plural in value.
Feature structures come ofa variety oftypes. First we have types that
indicate the word vs. phrase status of every constituent inatree (thus roughly
equivalent to the notion of bar level). The features for a node are next
divided into three major classes: the values of the feature SYN are structures
relevant to the syntax, ARG-ST (argument structure) feature structures
represent the theta grid, and the value ofa SEM feature structure represents
the semantic properties of the node.
Let us first talk about SYN feature structures. SYN feature structures tell
us about the formal grammatical properties of the node. They tell us the
syntactic category of the node, any inflectional properties of the node, what
other elements the node must combine with, etc. The feature that determines
the category of the node and its inflectional properties is called the HEAD
feature. The feature that restricts what kind ofnodes appear inthe specifier
position is called the SPR feature, and the feature that restricts what kind
of nodes appear in the complementposition is the COMPS feature. Tosee how
this works, letustake apartial lexical entry for the word letter. This example
is taken from Sag and Wasow (1999: 132):
2It should be noted that many of the ideas in Sag, Wasow and Bender (2003) diverge
from the conception of HPSG presented in Pollard and Sag (1994); many HPSG
researchers would disagree with the particulars presented here. In particular
see Richter (2000) fordiscussionof the differences among various kinds of HPSG.
2)
<letter,
Chapter17: Head-Driven PhraseStructureGrammar
word
HEAD
SYN SPR
COMPS
noun
AGR
< D
< (PP) >
a
GEND
AGR J|
COUNT +
3sng
neut
457
This looks much more intimidating than it actually is. The lexical entry is
an ordered list of the form of the word and the large AVM. Ordered lists are
represented with angled brackets (< >). At the top we have word, which tells
us, obviously, that this is a word and not a phrase. Below this we have the
SYN feature whose value is described by the AVM containing the HEAD, SPR,
and COMPS features. The HEAD feature tells us what kind of lexical item this
is. It is a noun, it triggers neuter and thirdperson agreement. TheSPR feature
gives usanordered list ofthe items that may or must appear inthe specifier
position ofany projection ofthis noun. The ordered list in this lexical entry
contains only oneitem: a count noun determiner. The boxed number found
both in the HEAD feature and in this SPR feature ® is called a (coreference)
tag. Tags are used to indicate that certain substructures of an overall
structure are identical. The tagUin the description of the HEAD feature refers
to the object described by the AVM that follows it. The |l|in the SPR feature
indicates that whatever the agreement features of the head are, they must
be identical for the specifier. This means that since the noun is third person
singular, the determiner must also be singular:
3) a)
b)
this (sg) letter (sg)
*these (pi) letter (sg)
Theidea of structural identity (asexpressed by coreference tags) does much
of the work that Functional Control does in LFG and feature checking
does in P&P - and more. For instance, it allows HPSG to have a nontransformational
analysis of raising and to eliminate PRO in control
constructions.
Notice that the specifier is not optional in this lexical entry. In English,
with singular count nouns, specifiers are not optional:
4) a) I gave her the letter.
b) I gave her a letter.
c) *Igave her letter.
458 Alternatives
Finally, we have the COMPS feature, which says that we may have anoptional
PP complement:
5) a) a letter about computer software
b) a letter from the president
The next major feature is the ARG-ST feature. Its value is an ordered list
of all the arguments associated with the word and represents the theta grid
ofthe word. You might observe that there is a redundancy between ARG-ST
features, and the SPR/COMPS features. As we will seebelow in section2, this
isacceptable because theyare treateddifferently by the rules whichcombine
words into sentences. As we will see, we need the ARG-ST feature for binding
reasons, independent of the SPR/COMPS features.3 Anexample oftheARG-ST
feature for the verb love is given in (6):
6) <love, [ ARG-ST < NP, NP> ]>
We can impose various kinds of selectional restrictions on this
representation, ofcourse. For example, theverb loves requires thatitssubject
be third person singular. We can encode this by inserting an AVM with
this specification into the first NP slot in the ARG-STR list.
7) <loves, [ ARG-ST < [NP [AGR 3s]],NP> ]>
Finally we have the SEM (semantic) features. These give us information
about how the word andsentence are tobeinterpreted. For example, MODE
tells us thesemantic type ofthe node (proposition, interrogative, directive,
referential item). The INDEX features are like theindices weused inchapter 5:
They mark the various participants or situations described in the sentence.
Last, we have the RESTR (restriction) feature, which tell us the properties
that must hold true for the sentence to be true. Again, this looks a bit like
our theta grids from chapter 8. Unfortunately we don't have the space
to cover this interestingaspect of HPSG in any detail.
The complete lexical entry for thenoun letter isgiven in (8), showing all
these features. The largest AVM in this structure is known as the noun's
SYN-SEM structure.
3Other arguments for distinguishing ARG-STR from COMPS and SPR can be found
in Manning and Sag (1998).
8)
<letter,
Chapter17: Head-Driven PhraseStructureGrammar
word
HEAD
SYN SPR
COMPS
ARG-ST @©@
MODE
INDEX
SEM
RESTR
noun
AGR E
@< D AGR D >
COUNT +
@<(PP)>
ref
k
"reln letter
SIT s
INST k
3sng
GEND neut
459
Each word in the lexicon has a rich lexical entry like this that specifies
the features and feature structures it brings to the sentence. This lexicalentry
tells us that this item is a word; its SYN feature specifies that it is a noun with
a 3sng, neuter HEAD feature. Since it is a count noun, the SPR feature requires
that it take a count determiner, which must agree with the HEAD I AGR
feature (indicated by the tag [!]). It may also take an optional PP COMPS
(complement). The specifier and complement features arerelated in theARGST
feature by the tags @andg (I'll discuss the©symbol below). TheSEMantic
features, which don't really concern us much here, indicate that the item
is referential and is assigned the index k. The RESTR feature tells us what
the word means and what context it can appear in. (Again see Sag, Wasow
and Bender (2003) for a more completedescription of all these features.)
2. The Lexicon
Needless to say, there is a lot of information stored in lexicalentries in HPSG
grammars. In many cases this information is redundant or predictable.
Forexample, for any count noun (such as letter or ball or peanut), there are
two forms: a singular form and a plural form. Ideally we don't really want
two complete lexical entries, as the information in one is predictable from
the other. While memory is cheap, it would be nice to reduce the amount
460 Alternatives
of redundancy or predictable information contained in each lexical entry.
HPSG does this in a number ofways.4
First, like LFG, HPSG has lexical rules that can change the shape
of lexical entries. Take for example the plural rule that derives the lexical
entries for plurals out of uninflected nouns. (The FNPL notation refers to
a function that does theactual morphology. Forexample, FNPL applied tocat,
will result incats, but FNPL applied tochild will give youchildren.)
9) Plural Rule
<^Hioun ~|>=><F
Urg-ST <[count+]>
This rule says that given any basic count noun root U you can create
a plural lexical item thathasanidentical feature structure, except with plural
number.
Similar rulescanbeapplied todo derivation (such as deriving thenoun
dancer from the verb dance), or to do grammatical function-changing
operations like thepassive. Inthefollowing rule (taken from Sag andWasow
1999: 235) thesymbol ©means "an append oflists." Formore on thisnotion,
see the references below, but for ourpurposes it roughly corresponds toan
ordering ofthe AVMs. You'll see that the rule does three things: (i) it puts
the verb @, into a passive form; (ii) itputs the first argument (NP;) into an
optional by-phrase (don't worry about the details of the feature structure for
theby-phrase); (iii) itmoves all the other arguments @ up intheARG-ST list,
the ©symbol showing that the arguments are strictly ordered with respect
to one another. The big gray arrows in the following rule are not part
ofthe rule itself; I've just written them in to point out the important parts
of the rule.
10) Passive Rule
<(lj, transitive verb > =>
[ARG-ST <NPj> ©0
' Zl—Ciii>
^ [HEAD [FORM pass]]
<Fpsp(E),
npl®, word
5YN [HEAD [AGR [NUM pi]]]
FORM by 1)>
•NP,
One of which we don't have the space to discuss here: inheritance hierarchies.
Theseare discussed at length in Sagand Wasow(1999) and other sourceson HPSG.
Chapter17: Head-Driven PhraseStructureGrammar 461
There is onefurther area of redundancyin thelexicon, which we already
noted in section 1 above. The information about the number of arguments
a word takes appears twice in the SYN-SEM structure. It appears once in
the ARG-ST feature and once in the SPR/COMPS features. This redundancy
is solved by theArgument Realization Principle (11), which builds theSPR
and COMPS features out of the ARG-ST feature:
11) Argument Realization Principle (ARP): A word structure satisfies the
following feature structure description:
'"SYN [SPR 0-1
COMPS g
ARG-STU © g
The equivalence of the SPR/COMPS features and the ARG-ST feature
is indicated by the tagsg] and @. The first argument is mapped to the SPR
feature, the second to the COMPS feature. This principle says that you
map the first argument of the argument structure into the SPR position,
and the second one into the COMPS position.
HPSG, then, shares with LFG a rich lexicon with many lexical rules.
This is where the similarities end, however. In LFG, we used functions and
functional equations to map constituency (c-structure) onto the functional
representation (f-structure). In HPSG, like P&P theory and as we'll see
below, functional (= featural) information is read directly off the constituent
tree instead of using mapping principles.
3. Rules, Features, and Trees
HPSG differs from LFG in at least one significant regard: it is compositional.
By that we mean that the way in which the meaning of the sentence
is determined is by looking at the constituent structure associated with the
sentence. HPSG shares this assumption with P&P.The meaning of a sentence
in HPSG and P&P can be calculated by taking each node, and using
constituency to determine what modifies what. Relationships between
words are directly encoded into constituency. In LFG, we instead used
functional equations to calculate the relationships between words.
In all three theories (HPSG, LFG, and P&P), the end of the sentence
should involve a saturation, satisfaction, or unification of the features
462 Alternatives
introduced by the words.5 In Minimalism, this was accomplished by feature
checking via theconstruction ofthephrase structure tree andviamovement.
In LFG, this is accomplished by functional equations that map to an
f-structure. In HPSG, by contrast, all feature satisfaction occurs by combining
words into constituents. Constituency is introduced by phrase structure
rules.6 These rules look alittle different than the ones we used in chapters 3,
6, and 7, but they do similar work. There are three basic rules (which are
roughly equivalent to the complement rule, the adjunct rule and the specifier
rule ofX-bar theory). The first ofthese is the Head Complement Rule. This
takes a head (marked with H) word, and ifit is insequence with a number
of arguments that match its COMPS requirements, licenses a phrase with
an AVM like its own, except that those COMPS features have been checked
off and deleted.7
12) Head Complement Rule
[phrase ~| [word ~1
comps<>J -» h[comps<|,...,0 >J J] ... 0
Notice that the tags |J ..., @on the head must be identical to the tags of the
phrases that follow. This means that the element(s) onthe right hand side
of the head must be selected for in the COMPS portion of the head word.
The resulting category (the phrase) has had its COMPS features erased.
(Thus making the satisfaction of features like COMPS very much like
feature checking.) The output ofthis rule - when applied toa transitive verb
The exact characterization of how this works is a matter of some debate in HPSG,
with various formal proposals indifferent versions ofthe theory. In early HPSG this
was done with unification (see chapter 16), asisthe version found inSag andWasow
(1999). In Pollard and Sag (1994) well-formed feature structures are licensed by
conjunction (or more accurately the conjunctive satisfaction of all the principles of
the grammar). Thedistinctions between these approaches need not concern us here,
which is why I've adopted the neutral term satisfaction (or saturation) to get at
the underlying idea. See Richter (2000) for a discussion of the formal apparatus
underlying these issues.
Phrase structure rules are actually a shorthand notation for more complicated
principles. See Pollard and Sag (1994) for more discussion.
Forexpository ease, I'm occasionally lapsing into the metaphors and terminology
ofMinimalism, which arenotaccepted by practitioners ofHPSG. Idothis sothatyou
can relate the ideas of HPSG to what you have previously seen in this book;
this doesn't mean that the ideas are entirely equivalent. For example, the notions
of checking and deletion suggest a derivational approach to syntax which is not
necessarily a part of HPSG. For more on the philosophical and methodological
assumptions underlying HPSG see Pollardand Sag(1994).
Chapter 17: Head-DrivenPhrase Structure Grammar 463
and an NP - is seen in (13). For ease of exposition, I've omitted most of
the featural information here, leaving just the relevant features present.
13)
Hyword
SYN HEAD
COMPS <[]]>
|_SPR <NP>
ARGST <NP, NP>
SEM [...]
The resulting phrase is what we would label a V in P&P syntax. The rule
combines a head with an item that satisfies one of its COMPS requirements.
It then licenses a phrase lacking that COMPS requirement.
One thing to notice about (13) is that the head features of the head word
become the head features of the entire phrase, just like the syntactic category
of a head is passed up the tree to the phrase level in the X-bar theory
of chapter 6. In HPSG this due to the Head Feature Principle:
14) Head Feature Principle (HFP): The HEAD value of any headed phrase is
identical to the HEAD value of the head daughter.
You'll also notice that the SPR feature of the head daughter is also transferred
up to the mother node. This is triggered by the Valence Principle (Sag,
Wasow and Bending 2003):
15) Valence Principle: Unless the rule says otherwise, the mother's SPR and
COMPS values are identical to those of the head daughter.
HEAD [...]
464 Alternatives
Semantic Feature Flow
The distribution of syntactic feature values is governed by our three
phrase structure rules. Semantic feature valuesalso flowup the tree.This
is governed by two principles. We won't go into these in detail, but here
they are for your reference:
i) Semantic Compositionality Principle
In any well-formed phrase structure, the mother's RESTR value is the
sum of the RESTR values of the daughters.
ii) Semantic Inheritance Principle
In any headed phrase, the mother's mode and index values are
identical to those of the head daughter.
Adjuncts are introduced by the Head Modifier Rule. This rule makes
reference to the special feature MOD. The MOD feature is contained in the SYNSEM
structure of the modifier and is linked to the thing it modifies with a tag,
allowing modifiers to impose selectional restrictions on the phrases
they modify. The rule takes a phrase (equivalent to our X' level) and licenses
another phrase (X').
16) Head Modifier Rule
[phrase] -> H| [phrase] piirase
HEAD[MOD
Specifiers are introduced using the third rule, which also takes
a phrase (X') and licenses a phrase (however, this time equivalent to our XP).
The S node in HPSG is a projection of the verb (i.e., is licensed as a mother
of the VP by this rule). Just as in X-bar theory, the HFP allows this rule
to be non-category specific.
17) Head Specifier Rule
phrase
SPR < > —> H
phrase
SI'R < HI >
This rule takes a phrase with a non-empty SI'R value and combines it with
an item that satisfies that value, and generates a phrase without an SPR value.
On an intuitive level, these rules take as inputs the lexical entries
for words and output sentences where the information has been combined
into a meaningful whole. In the next two sections, we turn to a variety of
phenomena discussed in this book and look at how HPSG accounts for them.
Chapter 17: Head-Driven PhraseStructure Grammar 465
4. Binding
HPSG does not use the notion c-command to determine binding relations.
Instead, binding makes reference to the ARG-ST list in the SYN-SEM structures.
Because of the rules discussed above in section 3, arguments on the left side
of an ARG-ST ordered list will always be higher in the tree than ones further
to the right on the list. As such the binding conditions in HPSG are based on
precedence in the ARG-ST list. Thisis accomplished with the notion outrank.
18) Outrank: A phrase A outranks a phrase Bjust in the casewhere A's SYNSEM
structure precedes B'sSYN-SEM structure on some ARG-ST list.
Anaphors are marked with a special feature: ANA. [ANA +] nodes are subject
to the HPSG equivalent of principle A:
19) Principle A: An [ANA +] SYN-SEM structure must be outranked by
a co-indexed SYN-SEM structure.
Pronouns, which are [ana -] are subject to principle B.
20) Principle B: An [ana -] SYN-SEM structure must not be outranked by a coindexed
SYN-SEM structure.
Becauseof the direct mapping between the ARG-ST and the tree, this will give
usessentially the sameresults as the c-command analysis given in chapter4.
For a discussion of the important differences between a c-command analysis
and an ARG-ST analysis, see Pollard and Sag (1992).
5. Long Distance Dependencies
In this section, we look briefly at how HPSG deals with long distance
dependencies between the surface position of w/i-phrases and the position
with which they are thematically associated. P&P uses movement to capture
this relation. LFG uses functional control. HPSG uses a feature: GAP.8 This is
a feature like COMPS or SPR that indicates that an argument is required by the
SYN-SEM structure and is missing. The presence of a non-empty GAP feature
indicates that an argument is not filling the expected complement position.
This is encoded in a revised version of the argument realization principle,
wherethesequence § © || isa list where theelements on the list|] havebeen
removed from§. This principle allows you to optionally map an argument
to he GAP feature rather than the COMPS feature.
Also called slash.
466 Alternatives
21) Revised ARP9
SPR s
SYN COMPS
.GAP
@e
ARGST ffl©|
Thisguarantees that any argument that couldappear on the COMPS list can
appear on theGAP listinstead. Justas weneededprinciples forpassing head
features, valence features and semantic features up the tree, we also need
a principle tomakesure GAP features make it up to the top of thetree: The GAP
Principle. (Formulationagain taken fromSag, Wasow and Bender (2003).)
22) The Gap Principle [gap|©...©0]
[GAPp [GAP0]
This principle encodes the idea that a mother's GAP feature represents
the union ofall the GAP valuesofits daughters.
Let's do an example. Because we don't have the space to introduce
HPSG analyses of head movement or do-support, we'll use topicalization,
rather than a w/i-question as an example of a long distance dependency.
Topicalization has the samebasicproperties as other w/i-movement (subject
to island constraints, etc.). The sentence we'll do is seen in (23). This sentence
is grammatical if we put contrastive stress on the first NP.
23) That boy, we saw ...
In a normal sentence, an NP (filled by that boy) occupies the COMPS position
of the verb saw. In this sentence, by contrast, the COMPS position is empty.
Instead there is an NP in the GAP feature. The NP we satisfies the verb's SPR
feature. The GAP value is percolated up the tree by the GAP principle
which results in a tree like (24):10
9 This particular formulation only allows w/i-extraction from object position.
As something to think about, you might consider how HPSG would go about dealing
with subject extraction (as seen in the Irish data in chapter 11).
10 You might consider whether gap feature percolation is really different from
movement or is simply a notational variant. What kind of evidence might you
propose to distinguish the two approaches?
24)
Chapter 17: Head-DrivenPhrase Structure Grammar 467
TP
[GAP^<NP>]
NP
Awe
VP
[GAP <NP>]
I
V
[GAP <NP>]
saw
The GAP feature associated with the S node must be satisfied someway. This
is accomplished with the Head FillerRule.
25) The Head Filler Rule ^ _
phrase[phrase | —» (l][p/7rase ~~I H
[gap<>J [gap<>J FORM fin
SPRO
GAP <0 >_
This rule satisfies the GAP feature, by adding the missing NP at the top of the
tree:
26)
NP TP
/\ [GA£<NP>]
That boy
NP VP
/\ [GAP<NP>]
we |
V
[GAP <NP>]
saw
You can now tryGeneral Problem Set1 andChallenge Problem Sets 1 &2
Ideas, Rules, and Constraints Introduced in this Chapter
Some of the definitions here are taken from Sag, Wasow and Bender (2003)
or the first edition of that book (Sag and Wasow 1999).
468 Alternatives
ii)
iii)
iv)
v)
Features: These do the work of determining what can combine with
what.
a) Bar-level-like features tell us what hierarchical level the node is at.
b) The SYN feature structures givesus the syntacticinfoabout the node,
i) The HEAD feature givesthe categoryand inflectional info.
ii) The COMPS feature tells us what complements appear in the
structure,
iii) The SPR feature tellsus what appears in the specifier,
iv) TheGAP feature tellsus if thereis a long distancedependency.
c) TheARG-ST featureis the HPSG equivalentof the theta grid. Binding
relations are defined against this.
d) The SEM feature structures tell us the semantic information about the
constituent, and come in a variety of types.
(Coreference) Tags: Numbers written inboxes(e.g., [J) that show that
two items are identical in a SYN-SEM structure or between SYN-SEM
structures.
SYN-SEM Structure: The set of AVMs for a node, containing all the
SYN, SEMand ARG-STfeatures.
Plural Rule:
<B noun
ARG-ST < [count +]>
word |>
YN [HEAD [AGR [NUM pi]]]
>=>< FNPL(|),(w
PassiveJRule
:B transitive verb
ARG-ST <NPj> ©
<FPSp(S),
word
SYN [HEAD [FORM
ARG-ST 0 © < ( PP
?ass]]
form by
P-OBJ NP ]
1
vi) ArgumentRealization Principle(ARPX,(revised):
SPR 0
SYN COMPS @©H
-CAP | _
ARG ST 0 © H
Chapter 17: Head-Driven Phrase Structure Grammar 469
vii) Compositional: The idea that the semantics of the sentence
can be read off of the constituency tree. This idea is shared by P&P
and HPSG, but is rejected by LFG.
viii) Feature Satisfaction (sometimes loosely called Unification): The idea
that all the features in a SYN-SEM structure must match. The rough
equivalent offeature checking in P&P/Minimalism.
ix) Head Complement Rule:
[phrase ~~| [word ~j
compsoJ -> h[comps<|I|,...,0 >J S ... 0
x) Head Feature Principle: The HEAD value of any headed phrase is
identical to the HEAD value of the head daughter.
xi) Valence Principle: Unless the rule says otherwise, the mother's SPR
and COMPS values are identical to those of the head daughter.
xii) Semantic Compositionality Principle: In any well-formed phrase
structure, the mother's RESTR value is the sum of the RESTR values
of the daughters.
xiii) Semantic Inheritance Principle: In any headed phrase, the mother's
mode and index values are identical to those of the head daughter.
xiv) Head Modifier Rule:
[phrase) -> HU[phrase] fphrase H
[head[mod [Jj J
xv) Head Specifier Rule:
[phrase ~~| [phrase ~~|
jsproJ -> | h[spr<|i]>J
xvi) Outrank: A phrase A outranks a phrase Bjust in the case where A's
SYN-SEM structure precedes B's SYN-SEM structure on some ARG-ST
list.
xvii) Principle A: An [ANA +] SYN-SEM structure must be outranked by
a coindexed SYN-SEM structure.
xviii) Principle B:An [ANA -] SYN-SEM structure must not be ranked by
a coindexed SYN-SEM structure.
xix) The GAP Principle: [GAP g] ©...©0 ]
[GAP H] ... [GAP0]
470 Alternatives
xx) The Head Filler Rule:
[phrase ~~| -> |l]|/?frrase ~~|
[gap<>J [gap<>J
H phrase
FORM fin
SPR<>
GAP <J] >
Further Reading
Borsley, Robert (1996) Modern Phrase Structure Grammar. Oxford: Blackwell.
Gazdar, Gerald, Ewan Klein, Geoffrey Pullum and Ivan Sag (1985)
Generalized Phrase Structure Grammar. Cambridge: Harvard University
Press.
Manning, Christopher and IvanSag(1998) Argumentstructure,valence and
binding. The Nordic Journal ofLinguistics 21,107-44.
Pollard, Carl and Ivan Sag (1992) Anaphors in English and the scope of
BindingTheory. Linguistic Inquiry 23, 261-303.
Pollard, Carl and Ivan Sag (1994) Head-Driven Phrase Structure Grammar.
Stanford: CSLI Publications and Chicago: The University of Chicago
Press.
Richter, Frank (2000) A Mathematical Formalism for Linguistic Theories
with an Application in Head-Driven Phrase Structure Grammar.
Ph.D. dissertation, University ofTubingen.
Sag, Ivan and Thomas Wasow (1999) Syntactic Theory: A Formal Introduction
(1st ed.). Stanford: CSLI Publications.
Sag, Ivan, Thomas Wasow and Emily Bender (2003) Syntactic Theory: A
Formal Introduction (2nd ed.). Stanford: CSLI Publications.
General Problem Set
1. English
[Application of Skills; Intermediate]
Create an HPSG-style lexicon for the words in (a) and then draw a tree using
the rules and lexical entries for the sentence (b).
a) the, kitten, tore, toilet, paper
b) The kitten tore the toilet paper.
You may abbreviate your syn-sem structures using tags.
Chapter 17: Head-Driven Phrase Structure Grammar 471
Challenge Problem Sets
Challenge Problem Set 1: Subject-Aux Inversion
[Critical Thinking; Challenge]
How might HPSG go about doing subject-aux inversion? (Hint: consider
a lexical rule.) Assume that auxiliary verbs select for other verbs
in their arg-st features.
Challenge Problem Set 2: Island Constraints
[Critical Thinking; Challenge]
In this chapter, wedidn't talk at all about how HPSG might account for island
constraints. Propose a constraint on the gap principle that might account
for them.
Conclusions
and Directions
for Further Study
We started this textbook with the question of what a person needs to
know about their language in order to understand a simple sentence.
We hypothesized that some of language is innate and other parts are
parameterized. In the first twelve chapters of this book, we sketched out
some ofthe major research threads inone approach tosyntax: the principles
and parameters (P&P) view. In part 1, we looked at how rules generate
hierarchical tree structures. These structures are geometric objects with
mathematical properties. We looked atone set of phenomena (binding) that
issensitive to those properties. Inpart2, welooked at a more sophisticated
view of tree structures, developing X-bar theory, and the thematic (lexical)
constraints on it such as theta theory. In part 3, we looked extensively at
how problematic word orders, such aspassives, raising, VSO languages, and
w/i-questions could all be accounted for using movement. Inchapter 12, we
brought these threads together and started looking at the unified approach
to movement. Part 4 addressed three more advanced topics in syntax. We
looked atsplitVP/vPsand theway they account for ditransitives and object
shift; welooked at raising and control; and werevisited binding theory and
came up with a more sophisticated version of the binding theory. Part 5
of this book changed direction slightly. We looked, ever so briefly, at two
popular alternatives toP&P/Minimalism. This was so that you could read
papers and books written in those alternatives, as well asgiving you a taste
for other, related, ways we can approach topics in syntax.
Congratulations for getting through all this material. I hope
this book has whetted your appetite for the study of syntax and sentence
structure and that you will pursue further studies in syntactic theory.
To thisend, I've appended a list ofbooks thatcantake youto the nextlevel.
Cook, V. J. and Mark Newson (1996) Chomsky's Universal Grammar: An
Introduction (2nd ed.). Oxford: Blackwell.
Cowper, Elizabeth (1992) A Concise Introduction to Syntactic Theory: The
Government and Binding Approach. Chicago: Universityof ChicagoPress.
474 Conclusions
Haegeman, Liliane (1994) Introduction to Government and Binding Theory.
Oxford: Blackwell.
Ouhalla, Jamal (1990) Introducing Transformational Grammar (2nd ed.).
London: Edward Arnold.
Radford, Andrew (1997a) Syntactic Theory and The Structure ofEnglish: A
Minimalist Approach. Cambridge: Cambridge University Press.
Radford, Andrew (2004) Minimalist Syntax: Exploring the Structure ofEnglish.
Cambridge: Cambridge University Press.
Roberts, Ian (1997) Comparative Syntax. London: Edward Arnold.
References
Aarts, Bas (1997) English Syntax and Argumentation. NewYork: St.Martin'sPress.
Abney, Steven (1987) The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, MIT.
Adger, David (1996) Aspect, agreement and Measure Phrases in Scottish Gaelic. In Robert
Borsley and Ian Roberts (eds.). The Syntax of the Celtic Languages. Cambridge: Cambridge
University Press, pp. 200-22.
Aikawa, Takako (1994) Logophoric use of the Japanese reflexive zibun-zisin 'self-self. MIT
Working Papers in Linguistics 24,1-22.
Aissen, Judith (1987) Tzotzil Clause Structure. Dordrecht:Reidel.
AUwood, Jens, Lars-Gunnar Andersson, and Osten Dahl (1977) Logic in Linguistics. Oxford:
Oxford University Press.
Anderson, Stephen and Paul Kiparsky (eds.) (1973) AFestschriftfor Morris Halle. New York: Holt,
Rinehart and Winston.
Aoun,Joseph(1985) AGrammar ofAnaphora. Cambridge: MITPress.
Baker, C. L. and John McCarthy (eds.) (1981) The Logical Problem of Language Acquisition.
Cambridge: MIT Press.
Baker, Mark (2001) The Atoms ofLanguage: The Mind's Hidden Rules ofGrammar. New York: Basic
Books.
Baker, Mark, Kyle Johnson, and Ian Roberts (1989) Passive arguments raised. Linguistic Inquiry
20,219-51.
Baltin, Mark(1981) Strict bounding. InC.L. Baker andJohn McCarthy (eds.), The Logical Problem
ofLanguage Acquisition. Cambridge: MIT Press,pp. 257-95.
Baltin, Mark and Chris Collins (2000) The Handbook ofContemporary Syntactic Theory. Oxford:
Blackwell.
Baltin, Mark and Anthony Kroch (1989) Alternative Conceptions of Phrase Structure. Chicago:
University of Chicago Press.
Bard, EllenG., Dan Robertson, and AntonellaSorace(1996) Magnitude estimation of linguistic
acceptability. Language 72,32-68.
Barker, Chris and Geoffrey Pullum (1990) A theory of command relations. Linguistics and
Philosophy 13,1-34.
Barsky, Robert (1997) Noam Chomsky: ALife ofDissent. Cambridge: MIT Press.
Barss, Andy and Howard Lasnik (1986) A note on anaphora and double objects. Linguistic
Inquiry17, 347-54.
Bayer, J.(1984) COMP in Bavarian syntax. The Linguistics Review 3,209-4.
Beck, Sigrid and Kyle Johnson (2004) Double objects again. Linguistic Inquiry 35,97-124.
Belletti, Adriana (1994) Verb positions: Evidence from Italian. In David Lightfoot and Norbert
Hornstein (eds.), Verb Movement. Cambridge:Cambridge University Press, pp. 19-40.
Borer, Hagit(1999) Deconstructing theconstruct. In Kyle Johnson and Ian Roberts (eds.), Beyond
Principles and Parameters. Dordrecht: Kluwer Academic Publishers, pp. 43-89.
Borsley, Robert (1996) Modern Phrase Structure Grammar. Oxford: Blackwell.
BoSkovid, Zeljko (1997) Superiority and economy of derivation: Multiple zw/i-fronting. Paper
presented at the 16th West Coast Conferenceon Formal Linguistics.
Brame, Michael (1976) Conjectures and Refutations inSyntax and Semantics. Amsterdam: Elsevier.
Bresnan, Joan (1972) Theoryof Complementationin English. Ph.D.dissertation, MIT.
476 References
Bresnan, Joan, ed. (1982) The Mental Representation of Grammatical Relations. Cambridge: MIT
Press.
Bresnan, Joan(2001) Lexical-Functional Syntax. Oxford: Blackwell.
Buring, Daniel (2005) Binding Theory. Cambridge: Cambridge University Press.
Burzio, Luigi(1986) Italian Syntax. Dordrecht: Reidel.
Carnie, Andrew (1995) Head Movement andNon-Verbal Predication. Ph.D. dissertation, MIT.
Carnie, Andrew and Eithne Guilfoyle (2000) The Syntax ofVerb Initial Languages. Oxford: Oxford
University Press.
Chametzky, Robert (1996) ATheory ofPhrase Markers and the Extended Base. Albany: SUNY Press.
Cheng, Lisa (1997) On the Typology ofWIi-Questions. New York: Garland Press.
Choe, Hyon Sook (1987) An SVO analysis of VSO languages and parameterization: Astudy of
Berber. In Mohammed Guerssel and Ken Hale (eds.), Studies in Berber Syntax. Lexicon Project
Working Paper 14,121-58.
Chomsky, Noam (1957) Syntactic Structures. The Hague: Janua Linguarum 4.
Chomsky, Noam (1965) Aspects ofthe Theory ofSyntax. Cambridge: MIT Press.
Chomsky, Noam (1970) Remarks onnominalization. In Roderick Jacobs and Peter Rosenbaum
(eds.), Readings in English Transformational Grammar. Waltham: Ginn. pp.184-221.
Chomsky, Noam (1973) Conditions on transformations. In Stephen Anderson and Paul Kiparsky
(eds.), AFestschriftfor Morris Halle. New York: Holt, Rinehart and Winston, pp. 232-86.
Chomsky, Noam (1975) The Logical StructureofLinguistic Theory. New York: Plenum.
Chomsky, Noam (1977) On w/i-movement. In Peter Culicover, Thomas Wasow, and Adrian
Akmajian (eds.), Formal Syntax. New York: Academic Press, pp.71-132.
Chomsky, Noam (1980) OnBinding. Linguistic Inquiry 11,1-46.
Chomsky, Noam (1981) Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam (1986a) Barriers. Cambridge: MIT Press.
Chomsky, Noam (1986b) Knowledge ofLanguage: Its Nature, Origins and Use. New York: Praeger.
Chomsky, Noam (1991) Some notes oneconomy ofderivation and representation. In Robert
Friedin (ed.), Principles and Parameters in Comparative Grammar. Cambridge: MIT Press, pp.
417-54.
Chomsky, Noam (1993) A Minimalist program for linguistic theory. In Kenneth L. Hale and
Samuel J. Keyser (eds.), The View from Building 20: Essays in Honor ofSylvain Bromberger.
Cambridge: MITPress, pp. 1-52.
Chomsky, Noam (1995) The Minimalist Program. Cambridge: MIT Press.
Chomsky, Noam and Howard Lasnik (1978) Aremark on contraction. Linguistic Inquiry 9, 268-
74.
Chung, Sandra (1976) An object creating rule inBahasa Indonesia. Linguistic Inquiry 7,1-37.
Cinque, Guglielmo (1981) Types ofA'Dependencies. Cambridge: MIT Press.
Cole, Peter and S. N. Sridhar (1976) Clause union and relational grammar: Evidence from
Hebrew and Kannada. Studies inthe Linguistic Sciences 6, 216-27.
Collins, Chris and Hoskuldur Thrdinsson (1996). VP internal structure and object shift in
Icelandic. Linguistic Inquiry 27,391-444.
Cook, V. J. and Mark Newson (1996) Chomsky's Universal Grammar: An Introduction (2nd ed.).
Oxford: Blackwell.
Cowper, Elizabeth (1992) AConcise Introduction to Syntactic Theory: The Government and Binding
Approach. Chicago: UniversityofChicagoPress.
Culicover, Peter (1997) Principles and Parameters: An Introduction to Syntactic Theory. Oxford:
Oxford University Press.
References 477
Culicover, Peter,ThomasWasow, and AdrianAkmajian (eds.) (1977) Formal Syntax. New York:
Academic Press.
Dalrymple, Mary, Ronald Kaplan, John Maxwell, and Annie Zaenen (eds.) (1995) Formal Issues
in Lexical-Functional Grammar. Stanford: CSLI Publications.
Dedrick, John andEugene Casad (1999) Sonora Yaqui Language Structures. Tucson: University of
Arizona Press.
DeGraff, Michel (2005) Morphology andwordorderin'creolization' andbeyond. InGuglielmo
Cinque and Richard Kayne (eds.), The Oxford Handbook ofComparative Syntax. Oxford:
Oxford University Press.
Den Dikken, Marcel (1995) Particles: On the Syntax of Verb-Particle, Triadic, and Causative
Constructions.Oxford: Oxford University Press.
Deprez, Vivienne (1992) Raising constructions inHaitian Creole. Natural Language and Linguistic
Theory 10,191-231.
Derbyshire, Desmond (1985) Hixkaryana and Linguistic Typology. Dallas: Summer Institute of
Linguistics.
Diesing,Molly(1992) Indefinites. Cambridge:MITPress.
Edmondson, Jerold and Donald A. Burquest (1998) A Survey of Linguistic Theories (3rd ed.).
Dallas: Summer Institute of Linguistics.
Emonds, Joseph (1980) Word orderin Generative Grammar. Journal ofLinguistic Research 1,33-
54.
Escalante, Fernando (1990) Voice and Argument Structure in Yaqui. Ph.D. dissertation,
University of Arizona.
Fabb, Nigel(1994) Sentence Structure. London: Routledge.
Falk, Yehuda N. (2001) Lexical-Functional Grammar: An Introduction to Parallel Constraint-Based
Syntax. Stanford: CSLIPublications.
Friedin, Robert (ed.) (1991) Principles and Parameters in Comparative Grammar. Cambridge: MIT
Press.
Gair,James(1970) Colloquial Sinhalese Clause Structure. The Hague:Mouton.
Garrett, Merrill (1967) Syntactic Structures and Judgments of Auditory Events. Ph.D.
dissertation, University of Illinois.
Gazdar, Gerald, Geoffrey Pullum, and Ivan Sag (eds.) (1983) Order Concord and Constituency.
Dordrecht: Foris.
Gazdar, Gerald, EwanKlein, Geoffrey Pullum, and IvanSag(1985) Generalized Phrase Structure
Grammar. Cambridge: Harvard University Press.
Goodall, Grant(1993) On caseand the passive morpheme. Natural Language and Linguistic Theory
11,31-44.
Grimshaw, Jane (1990) Argument Structure. Cambridge: MITPress.
Gruber, Jeffrey (1965) Studies in Lexical Relations. Ph.D. dissertation, MIT.
Haegeman, Liliane (1994) Introduction to Government and Binding Theory. Oxford: Blackwell.
Haegeman, Liliane and Jacqueline Gueron (1999) English Grammar: A Generative Perspective.
Oxford: Blackwell.
Hale, KennethL.and SamuelJay Keyser(eds.)(1993) The Viewfrom Building 20: Essays inHonor
ofSylvain Bromberger. Cambridge: MITPress.
Halle, Morris and Alec Marantz (1993) Distributed morphology and the pieces of inflection. In
Kenneth L. Hale and Samuel JayKeyser (eds.) The Viewfrom Building 20: Essays in Honor of
Sylvain Bromberger. Cambridge: MITPress, pp. 111-76.
Harley, Heidi (2002) Possession andthe double object construction. Linguistic Variation Yearbook
2,29-68.
478 References
Harley, Heidi (2006) English Words: ALinguistic Introduction. Oxford. Blackwell.
Heim, Irene andKratzer, Angelika (1998) Semantics in Generative Grammar. Oxford: Blackwell.
Higginbotham, James (1980) Pronouns and bound variables. LinguisticInquiry 11,697-708.
Higginbotham, James (1985) Anote on phrase markers. MIT Working Papers in Linguistics 6 87-
101.
Holzman, Mathilda (1997) The Language ofChildren (2nd ed.). Oxford: Blackwell.
Hornstein, Norbert (1994) An argument for Minimalism: the case of antecedent contained
deletion.Linguistic Inquiry 25,455-80.
Hornstein, Norbert (1999) Movement and control. Linguistic Inquiry 30,69-96.
Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of Grammar. Ph.D.
dissertation, MIT.
Huang, C.-T. James (1989) PRO-drop in Chinese. In Osvaldo Jaeggli and Kenneth Safir (eds.).
The Null Subject Parameter. Dordrecht: Kluwer Academic Publishers, pp. 185-214
Hyams, Nina (1986) Language Acquisition and the Theory ofParameters. Dordrecht: D. Reidel.
Iatridou, Sabine (1990) About Agr(P). Linguistic Inquiry 21,551-7.
Jackendoff, Ray (1977) X-bar Syntax: ATheory ofPhrase Structure. Cambridge: MIT Press.
Jackendoff, Ray (1993) Patterns inthe Mind. London: Harvester-Wheatsheaf.
Jacobs, Roderick and Peter Rosenbaum (eds.) (1970) Readings in English Transformational
Grammar. Waltham: Ginn.
Jaeggli, Osvaldo (1986) Passive. Linguistic Inquiry 17,587-622.
Jaeggli, Osvaldo and Kenneth Safir (eds.) (1989) The Null Subject Parameter. Dordrecht: Kluwer
Academic Publishers.
Jelinek, Eloise and Fernando Escalante (2003) Unergative and unaccusative verbs in Yaqui. In
Gene Casad and Thomas L. Willett (eds.), Uto-Aztecan: Structural, Temporal and Geographic
Perspectives: Papers inHonorofW. Miller. Hermosillo: Universidad Autonoma deSonora.
Johnson, Kyle (1991) Object positions. Natural Languageand Linguistic Theory 9,577-636.
Johnson, Kyle and Ian Roberts (eds.) (1991) Beyond Principles and Parameters. Dordrecht: Kluwer
Academic Publishers.
Kaplan, Ronald (1995) The formal architecture of Lexical-Functional Grammar. In Mary
Dalrymple, et al. (eds.), Formal Issues in Lexical-Functional Grammar. Stanford: CSLI
Publications,pp. 7-27.
Kaplan, Ronald and Joan Bresnan (1982) Lexical-Functional Grammar: Aformal system for
grammatical representation. In Joan Bresnan (ed.). The Mental Representation ofGrammatical
Relations. Cambridge: MIT Press, pp.173-281.
Katamba, Francis (2004) English Words. New York: Routledge.
Kayne, Richard (1994) The Antisymmetry ofSyntax. Cambridge: MIT Press.
Koizumi, Masatoshi (1993) Object Agreement Phrases and the split VP hypothesis. MIT Working
Papers in Linguistics 18,99-148.
Koopman, Hilda (1984) The Syntax of Verbs: From Verb Movement Rules in the Kru Languages to
Universal Grammar. Dordrecht: Foris.
Koopman, Hilda (1992) On the absence of case chains in Bambara. Natural Language and
Linguistic Theory 10,555-94.
Koopman, Hilda and Dominique Sportiche (1991) The position ofsubjects. Lingua 85,211-58.
Kratzer, Angelika (1996) Severing the external argument from its verb. In Johan Rooryck and
Laurie Zaring (eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer. pp. 109-37.
Kroeger, Paul (1993) Phrase Structure and Grammatical Relations in Tagalog. Stanford: CSLI
Publications.
Kroskrity, Paul (1985) Aholistic understanding ofArizona Tewa passives. Language 61,306-28.
References 479
Landau, Idan (1999) Elements of Control. Ph.D. dissertation, MIT.
Larson, Richard (1988)On the double object construction. Linguistic Inquiry 19,335-91.
Lasnik, Howard (1989) Essays onAnaphora. Dordrecht: KluwerAcademicPublishers.
Lasnik, Howard (1999a) Minimalist Analyses. Oxford: Blackwell.
Lasnik, Howard (1999b)On feature strength. Linguistic Inquiry 30,197-219.
Lasnik, Howard and Mamoru Saito (1984) On the nature of proper government. Linguistic
Inquiry 15, 235-89.
Lehmann, Winfred (1978) The great underlying ground-plans. In Winfred Lehmann (ed.),
Syntactic Typology. Austin: University of Texas Press, pp. 3-56. ,
Levin, Beth (1993) English Verb Classes and Alternations: A Preliminary Investigation. Chicago:
University of Chicago Press.
Lightfoot, David (1976) Tracetheoryand twicemoved NPs.Linguistic Inquiry 7,559-82.
Lightfoot, David (1991) How to Set Parameters: Evidencefrom Language Change. Cambridge: MIT
Press.
Lightfoot, David and Norbert Hornstein (eds.) (1994) Verb Movement. Cambridge: Cambridge
University Press.
Longobardi, Giuseppi (1994) Reference and proper names: A theory of N-movement in syntax
and Logical Form. Linguistic Inquiry 25,609-65.
Manning, Christopher and IvanSag(1998) Argument structure, valence and binding.The Nordic
Journal of Linguistics 21,107-44.
Manzini, Maria Rita (1983)On control and control theory. Linguistic Inquiry 14,421-46.
Manzini,Maria Rita (1992) Locality: A Theory and Some ofIts Empirical Consequences. Cambridge:
MIT Press.
Marantz, Alec (1984). Onthe Nature ofGrammatical Relations. Cambridge: MIT Press.
Marcus, Gary, Steven Pinker, Michael Ullman, Michelle Hollander, T.J.Rosen, and Fei Xu (1992)
Overregularization in language acquisition. Monographs of the Society for Research in Child
Development 57.
May, Robert (1985) Logical Form. Cambridge: MITPress.
McCloskey, James (1979) Transformational Syntax and Model Theoretic Semantics: A Case Study in
Modern Irish. Dordrecht: Reidel.
McCloskey, James (1980) Isthere raising inModern Irish? triu 31,59-99.
McCloskey, James (1983) A VP ina VSO language. InGerald Gazdar, Geoffrey Pullum, and Ivan
Sag(eds.), Order Concord and Constituency. Dordrecht: Foris. pp. 9-55.
McCloskey, James (1991) Clause structure, ellipsis and propergovernment in Irish. Lingua 85,
259-302.
Moore, John (1998) Turkish copy raising and A-chain locality. Natural Language and Linguistic
Theory 16,149-89.
Ouhalla,Jamal(1990) Introducing Transformational Grammar (2nded.). London:Edward Arnold.
Perlmutter, David and Paul Postal (1984) The 1-Advancement Exclusiveness Law. In David
Perlmutter and Carol Rosen (eds.). Studies in Relational Grammar. Chicago: University of
Chicago Press, pp. 81-125.
Perlmutter, David and Carol Rosen (eds.) (1984) Studies in Relational Grammar. Chicago:
University of Chicago Press.
Pesetsky, David (1994) Zero Syntax: Experiences and Cascades. Cambridge:MITPress.
Petter, Marga (1998) Getting PRO under Control. TheHague: Holland Academic Graphics.
Pinker, Steven (1995)The Language Instinct. New York: Harper Perennial.
Pollard, Carl and Ivan Sag (1992) Anaphors in English and the scope of Binding Theory.
Linguistic Inquiry 23, 261-303.
480 References
Pollard, Carl and Ivan Sag (1994) Head-Driven Phrase Structure Grammar. Stanford: CSLI
Publications and Chicago:University of Chicago Press.
Pollock, Jean-Yves (1989) Verb-movement, Universal Grammar, and the structure of IP.
LinguisticInquiry 20,365-424.
Postal, Paul (1974)OnRaising. Cambridge: MITPress.
Pullum, Geoffrey (1997) The morpholexical nature of English to-contraction. Language 73, 79-
102.
Pullum, Geoffrey and Barbara Scholz (2005) Contrasting applications of logic in natural
language syntactic description. In Petr Hdjek, Luis Vlades-Villanueva and Dag Westerstal
(eds.), Logic, Methodology and Philosophy of Science: Proceedings of the Twelfth International
Congress. London: King's CollegePublications, pp. 481-503.
Radford, Andrew (1988) Transformational Grammar: A First Course. Cambridge: Cambridge
University Press.
Radford, Andrew (1997a) Syntactic Theory and The Structure ofEnglish: A Minimalist Approach.
Cambridge: Cambridge University Press.
Radford, Andrew (1997b) Syntax: AMinimalist Introduction. Cambridge: Cambridge University
Press.
Radford, Andrew (2004) Minimalist Syntax: Exploring the Structure of English. Cambridge:
Cambridge University Press.
Reinhart, Tanya(1976) TheSyntactic Domain ofAnaphora.Ph.D. dissertation,MIT.
Reinhart, Tanya(1983) Anaphora and Semantic Interpretation. London: CroomHelm.
Richards, Norvin (1997) What Moves Where When inWhich Language? Ph.D. dissertation, MIT.
Richter, Frank (2000) AMathematical Formalism for Linguistic Theories with anApplication in
Head-Driven Phrase Structure Grammar. Ph.D. dissertation. University ofTubingen.
Ritter, Elizabeth (1988) Ahead movement approach toconstruct state noun phrases. Linguistics
26,909-29.
Rivero, Maria-Louisa (1991) Longhead movementand negation:Serbo-Croatian vs. Slovakand
Czech. The Linguistic Revieio 8,319-51.
Rizzi, Luigi (1982) Issues inItalian Syntax. Dordrecht: Foris.
Rizzi,Luigi(1990) Relativized Minimality. Cambridge:MITPress.
Roberts, Ian (1997) Comparative Syntax. London: Edward Arnold.
Rosenbaum, Peter (1967) VieGrammar ofEnglish Predicate Complement Constructions. Cambridge:
MIT Press.
Ross, J. R.(Haj)(1967) Constraintson Variables in Syntax. Ph.D. dissertation,MIT.
Sag, Ivan and ThomasWasow (1999) Syntactic Theory: A Formal Introduction (1sted.). Stanford:
CSLI Publications.
Sag, Ivan,Thomas Wasow and Emily Bender (2003) Syntactic Theory: AFormal Introduction (2nd
ed.). Stanford: CSLI Publications.
Saito, Mamoru and Howard Lasnik (1994) Move Alpha: Conditions onIts Application and Output.
Cambridge: MIT Press.
Sampson, Geoffrey (1997) Educating Eve: The Language Instinct Debate. London: Cassell.
Sapir, Edward and Morris Swadesh (1939) Nootka Texts, Tales, and Ethnological Narratives, ivith
Grammatical Notes and Lexical Materials. Philadelphia: Linguistic Society of America.
Saxon, Leslie (1984) Disjointanaphora and the binding theory.Proceedings ofWCCFL 3,242-51.
Seiler, Wolf (1978) The modaliscase in Inupiat. Work Papers ofthe Summer Institute ofLinguistics
22, 71-85.
Sells,Peter (1985) Lectures onContemporary Syntactic Theories. Stanford: CSLIPublications.
References 481
Sigur6sson, Halld6r Armann (1991) Icelandic case-marked PRO and the licensing of lexical
arguments. Natural Language and Linguistic Theory 9, 327-65.
Simpson, Jane (1991) Warlpiri Morpho-Syntax: ALexicalist Approach. Dordrecht: Kluwer Academic
Publishers.
Soames, Scott and David M. Perlmutter (1979) Syntactic Argumentation and the Structure of
English. Berkeley:University of California Press.
Sobin, Nicholas (1985) Case and agreement in the Ukrainian morphological passive
construction. Linguistic Inquiry 16,649-62.
Speas, Margaret (1990) Phrase Structure in Natural Language. Dordrecht: Kluwer Academic
Publishers.
Sportiche, Dominique (1988) A theory of floating quantifiers and its corollaries for constituent
structure. Linguistic Inquiry 19,425-49.
Sproat, Richard (1985) Welsh syntax and VSO structure. Natural Language and Linguistic Theory 3,
173-216.
Stechow,Arnim von (1996) The different readings of wieder 'again': A structural account. Journal
ofSemantics 13,87-138.
Stenson, Nancy (1989) Irishautonomous impersonals. Natural Language and Linguistic Theory 7,
379-406.
Stowell,Tim (1981) Origins of Phrase Structure. Ph.D. dissertation, MIT.
Szabolcsi, Anna (1994) The noun phrase. In Sijntax and Semantics 27: The Syntax of Hungarian.
New York: Academic Press, pp. 179-279.
Travis, Lisa de Mena (1984) Parameters and Effects of Word Order Derivation. Ph.D.
dissertation, MIT.
Uriagereka, Juan (1998) Rhyme and Reason: An Introduction to Minimalist Syntax. Cambridge: MIT
Press.
Vikner, Sten (1995) Verb Movement and Expletive Subjects in the Germanic Languages. Oxford:
Oxford University Press.
Webelhuth, Gert (1995) Government and Binding Theory and the Minimalist Program. Oxford:
Blackwell.
Weerman, Fred (1989) The V2 Conspiracy. Dordrecht: Foris.
Williams, Edwin (1980) Predication. Linguistic Inquiry 11, 203-38.
Williams, Edwin (1983) Semantic vs. syntactic categories. Linguistics andPhilosophy 6,423-46.
Williams, Edwin (1994)Tliematic Structure in Syntax. Cambridge: MIT Press.
Zaenen,Annie,JoanMaling, and HoskuldurThrdinsson (1985) Caseand grammaticalfunctions:
The Icelandic passive. Natural Language andLinguistic Theory 3,441-83.
Index
3, seeexistential quantifier
T, seemetavariable
i, seemetavariable
V,seeuniversal quantifier
*, see asterisk
[+NOM] 297,355
[+ACC] 297,309,355
[±anaphor], seeanaphor
[±pronoun], seepronoun
[±Q] 208,321,355
[±WH] 321,355
{.../...} 53,76
©460
9 roles, see theta roles
A (syntactic category) 44
A, principle, seeprinciple B
absolutive (Abs) 299
abstract Case, see Case
accusative, 11,27,205,296 see also
[+ACC]
acquisition 14-5,28,30
actives 291
Adj, seeadjective
Adj' 158-9
adjective phrase (AdjP) 68-71,170-2,
194-6
adjectives (Adj) 43,60,62,190
AdjP, seeadjective phrase
adjunct 153-73,188,190,192-6
adjunct clause 201-6,213
Adv, see adverb
Adv' 158-9
adverb phrase (AdvP) 68-71
adverbs (Adv) 44-5,61-2,101,243-60
AdvP, seeadverb phrase
affix hopping 265-71
affix lowering 210-1,214,265-71
agent 221,231
agreement 357
AgrO 382,389,407,420
AgrS 392
ambiguity 87-88,100
anaphor 9,27,31-2,136,137-42,145,
419,427-30,433
anaphora 9,31-2
antecedent 9,27,137-8,145,432
antecedent contained deletion (ACD)
385,389
antipassive 239-40
applicative 392
arg-st, see features
Argument Realization Principle 461,
466, 468
argument structure 51,55,219-40
arguments 51,55,201-6
Arizona Tewa 308
aspect 269
asterisk (*) 9,27
a-structure 440
asymmetric c-command 115,123,127
Attribute Value Matrix 439
Aux, seeauxiliary
auxiliary 217,265-71
AVM, see Attribute Value Matrix
B,principle, seeprinciple B
Bahasa Indonesia 392
Bambara 98
bar levels 155
bar-level features, see features
base 151-240,244
Bavarian German 321
beneficiary 222, 231
Berber 275
bind 138-41,145
bindee 141,144
binder 141,144
binding domain 141-2,145,147,148,
426-32
binding principles, seeprinciples A, B,
and C
binding theory 135-50,346,423-34
Bosnian 349,370
bounding theory, seeLocality
constraints and Minimal Link
Condition
bracketed diagrams 65-6,86,95
branch 104,121
branching 105
Burzio's generalization 302,307
C, see complementizer
C, principle, seeprinciple C
case 11,27,205,306
Case 295-300,306,405
Case filter 298,306
categories, seeparts of speech
causatives 313-4
c-command 113-7,123,127,146,150,
376,422
chain 425,432
Chichewa 391
Chinese, see Mandarin Chinese
Chomsky, Noam 6
clausal subject construction 396,417
clause 74-9,201-6,213,214
click experiment 65
closed class parts of speech 45,55
cognitive science 4
Coherence 444,452
coindex 137-8,145
common nouns 50
complement 163-73,188,190,192-6
complement clause 201-6
complementary distribution, 44,55,61
complementizer (C)47, 75, 205,214
complementizer phrases (CP) 75-9,
206-9,213
Completeness 444,452
Complex DP constraint 332,344
compositional 461,469
comps, see features
computational component 227,232
Conditions A, B, and C see principles
A,B,C
Index 483
conjunction 90,154,158
conjunctions (Conj) 47
constituency 64
constituency tests 88-91,96,100,102
constituent 63-4,95,108,122
constituent of, 109,122
construct state 199 276-7
control 395-422,449,452
controller 412,422
Coordinate Structure Constraint 336,
344
coordination, see conjunction
copy theory of movement 425,432
corefer 137-8,145
coreference tags, see tags
corpus (corpora) 12,27
covert movement 350-64,368
CP, seecomplementizer phrase
Creole, see Haitian Creole
Croatian 349,370
c-structure 438-9,451
count nouns 49,55
D, see determiner
data sources 12-4
daughter 109
Deep structure, see D-structure
descriptive grammar 26
descriptive rules 8
descriptively adequate grammar 24,
28,30
determiner (D) 46
determiner phrases (DP) 198-210,213
derivational morphemes 40
direct object, seeobject
discontinuous constituents 132
distributional criteria for parts of
speech, 40-5,54-5, 58-9
ditransitive 52, 55, 220,375-7, 385-9
Dogrib 150
do-insertion, see do-support
domination 106-10,122,125
do-soreplacement 156-157m 177
do-support 263-5,273
484
double object constructions 314,375-
94
DP, seedeterminer phrase
DP movement 259,285-316,339
drawing trees, seetree drawing
D-structure 244,272
dummy noun, seeexpletive
Dutch 99-100,312
echo questions 342-3,345,371
ECM,seeDP movement (subject to
object raising)
E-language, see language
ellipsis 384
embedded clause 75-9,201-6,213,333
endocentricity 161
EPP, seeExtended Projection
Principle
Equi, see control
ergative (Erg) 299
examples, foreign language 92-5
exhaustive domination 107-9,122,
125-6
existential quantifier 365-7
expanded VPs 375-94
experiencers 221,231
explanatorily adequate grammar 24,
28,30
expletive 228-30,231,237,404
expletive insertion 229,232,404
Extended Projection Principle (EPP)
229,231,291
Extended Standard Theory (EST) 5
external theta role 224
extraposition 396,404,418
falsifiable prediction 8,26
Farsi, see Persian
f-command 451
f-description 442, 451
features 49-54,55,456-9,468
feature checking 298-9
feature satisfaction 461,469
finite clause 204-6,213
Index
flat structure 153,254-6,275
floating quantifiers 279,420
foreign language examples 92-5
free 145
free genitive 199,213
French 33,243-60,275-6,351,363,369
f-structure 440-5
Full Interpretation 356,368
functional categories 45-7,55,58,197-
218
functional control 448,452
functional equation 242-3,451
Gaelic, Irish, see Irish
Gaelic, Scottish 383,393
gap, see features
Gap Principle 466,469
GB,seePrinciples and Parameters
gender 9,27,57
Generative Grammar 5
generative power 245-6
German 190-3,280,281-2,382
goal 231
government 115-7,123,128
Government and Binding Theory, see
Principles and Parameters
grammar 8,26
grammatical functions 439,451
grammatical relations 118,128-9,296
grammaticality judgment task 12-4,
27,28
Greek 149
Haitian Creole 274,307
head 70,96,161,188
head, see features
head complement configuration 357
Head Complement Rule 462,469
Head-Driven Phrase Structure
Grammar (HPSG) 5,455-72
Head Feature Principle 463,469
Head Filler Rule 470
head government 117
head-head configuration 358
head-to-head movement 243-84,339,
see also V •* T; T -> C; and N -» D
head mobility 446,452
Head Modifier Rule 464,469
Head Specifier Rule 464,469
Hebrew 276-7
Hiaki 131-2,237-9
hierarchical structure 64,95,
Hixkaryana 23,99
HPSG, see Head-Driven Phrase
Structure Grammar
Human Language Capacity, see
Language (capital L)
Hungarian 216
hypothesis 7-8
i/.j notation 139
Icelandic 313,420,453
idioms 375,392,404
I-language, seeLanguage
immediate constituent of 110,122
immediate precedence 113,123,126
immediately dominate 109-10,122,
126
impersonal constructions 310-1
index 137-8,145
indirect object 119-21,123
infinitival clause, see non-finite clause
Infl see Tense (T)
inflectional morphemes 40
innateness 15-9,28,30
instrument 222,231
intensifiers 61
interface levels 360
intermediate projections 173,188
internal theta roles 224
intonation 342-3
intransitive 51,55, 220
intuitions, seegrammaticality
judgment task
Inupiaq 239-40
Irish 133-4,236,243-60,310-1,321,
348,349-50,370,381-2,421-2
island 334-39,344,471
Index 485
Italian 33-4,274,282-3
Japanese 148-9,192
Kannada 310-1
label 105,122
Language 4,26
language 4,26
language variation 22-4
last resort 264
learning 14-5,28,30
levels of adequacy 24, 28,30
levels of representation 424, seealso
interface levels
Lexical-Functional Grammar (LFG) 5,
437-54
lexical item 227,231
lexical part of speech 45-6,55
lexicon 219-40,440,459-61
LF, seeLogical Form
LFG, see Lexical-Functional Grammar
light verb (v), 377-80,389
Local Configuration 356-9,368
locality Constraint 145,272,347,355
location 222
Logical Form (LF)360,368
Logical Problem of Language
Acquisition 15-9,28,33^
long distance dependencies, 450,465-
7, see also wh-movement
main clause 201-6,213
Mandarin Chinese 359
mass nouns 49,55
matrix clause, see main clause
maximal projection 173,188
metavariable 443,51-2
middles 313-4
Minimal Link Condition 337-68
Minimalism 5,26,355
MLC, see Minimal Link Condition
modal 217,269
Modern Irish, see Irish
modification 69-71,87,96,174
486
Mongolian 370
morphological distribution tests 40-5,
55,58-9
morphology 4
mother 109,122,seealso immediately
dominate
Move 355-72
movement rules 241-72
multiple auxiliaries 265-71
multiple w/i-movement 350-1
N, see noun
N' 155-6
N -> D movement 277-8,282
narrow scope 367,369
negation 47,243-60,351-2,371
negative polarity items (NPI) 133,217,
390
Nepali 370
no crossing branches constraint 112,
123
node 105,121
nominative 11,27,205,296 see also
[+NOM]
non-branching 105
non-finite clause 204-6,213,395-422
non-terminal node 105,110,122
Nootka 56-7
noun phrase (NP) 66-8,135-50,165-
70,seealso determiner phrase (DP)
nouns (N) 42,59
NP, seenoun phrase anddeterminer
phrase
NP movement, see DP movement
NP raising, see DP movement
NPI, seenegative polarity item
null subject parameter 34,417-8
number 10, 27
object (direct) 118-21,123
object control, see control
object expletives 237
object of a preposition 118-9,123
object shift 380-4,389
object-subject-verb (OSV)order 23
Index
object-verb-subject (OVS) order 23,99
obligatory control 412,418
oblique 121,123
observationally adequate grammar 24,
28,30
o/-genitive 199,213
one-replacement 154,169,188
open class parts of speech 45,55
open function 448,452
optional control 412,418
OSV,seeobject-subject-verb
outrank 465,469
overgeneration 219
overt movement 350-64
OVS,seeobject-verb-subject
P, seepreposition
P'160
P&P, seePrinciples and Parameters
parameters 22-4,34-5,174-77
particles 391
parts of speech 37-62
passive 237,269,285-316,447,452
Passive Rule 468
Persian 148,280,308-10,391
person 11,27
PF, see Phonetic Form
Phonetic Form (PF) 360,368,370
phonetics 4
phonology 4
phrase 64-5
phrase government 117
phrase structure rule (PSR)66,96
picture DPs, 347
pleonastic, seeexpletive
plurality 49,55
Plural Rule 468
possessive 101-2,199-200,213,214,
434
potential antecedent 428,432
PP, seeprepositional phrase
pragmatics 415,418
precedence 110-3,122,126
predicate 51,55
predicate phrase 201-6,213,214
prediction 8,26
preposition (P) 46,100,382
prepositional case 298,300
prepositional phrase (PP)71-2,170-2
prescriptive grammar 26
prescriptive rules 8, 29
principle A 140,145,147,429,432,465,
469
principle B142-3,145,147,431,433,
465,469
principle C 143-4,145,147
Principle of Full Interpretation, see
Full Interpretation
Principle of Modification 69-71,87,
96,174,189
Principles and Parameters approach 5,
26
PRO 395-422
pro416,417
PROarb412,417
pro-drop, seenull subject parameter
progressive 269
projection 173,188
Projection Principle 227,231
pronominal, seepronoun
pronoun 136,142-3,145,282,430-1,
434
pronouns: part of speech 46
proper domination, seedomination
proper names 50, 282
proposition, 231
pseudogapping 384,389
pseudo-passive 312
QR, seequantifier raising
quantifier 101,279,365-7
quantifier raising (QR)365-8
quirky case 313,420
raising 306, seealsoDP movement
recipient 231
reciprocal 136
recursion 16-7,28,32-3,80,96
Index 487
recursivity, see recursion
referring expression, seeR-expression
reflexive, seeanaphor
relative clause 79
resumptive pronouns 349-50
R-expression 143-4,145
root (lexical) 378
root clause, see main clause
root node 105,110,122
rules 5, 8
S,seetense phrase
S',seecomplementizer phrase
Sapir-Whorfhypothesis 5
scientific method, the 6-14,24-5
scope 363-9,371
scrambling 347
selectional restrictions 220
sem, see features
Semantic Compositionality Principle
464,469
semantic ill-formedness 14
Semantic Inheritance Principle 464,
469
semantic judgment 28, 29
semantic relations, see thematic
relations
semantics 5
Serbian 349-5,370
's-genitive, see free genitive
Sinhala 92-4,234-6
sister 109,122
sister precedence 111, 122
slash, see features
source 222,231
SOV,seesubject-object-verb
specifier 172-3,188
specifier clause 201-6, 213
specifier-head configuration 356
SPELLOUT350-64,368
S-structure 245, 272
statistical probability 22
structural relations 103-34
structure see also constituency; trees
488
subcategories 49-50,55,59-60
subconscious knowledge 13
subject 123,213,214
subject condition 336,344
subject control, see control
subject to object raising, seeDP
movement
subject to subject raising, seeDP
movement
subject-aux inversion, seeT —» C
movement
subject-object-verb order (SOV) 23,
131
subject-verb-object order (SVO) 23,
256
subordinate clause, see embedded
clause
Surface structure, see S-structure
SVO,seesubject-verb-object
Swedish 97
symmetric c-command 114,123,127
SYN,see features
SYN-SEM structure 468
syntactic categories, seeparts of
speech
syntactic distribution tests, 40-5,55,
58-9
syntactic ill-formedness 14
syntactic judgment 28,29
syntax (definition) 4,26
T -> C movement 207-10,214,260-3,
272,471
T movement/raising, see T -> C
movement
tags (coreference tags) 468
Tense (T) 47,204
tense phrase (TP) 74-9,210-1
tenseless clause, see non-finite clause
terminal node 105,110,122
Tewa, see Arizona Tewa
f/w/-trace effect 20
thematic relations 221,230,232-3
theme 221,231
Index
theta criterion 225,231,236,285-90
theta grid 223,233,391,396
theta roles 222,231,285-90
topicalization 446
traces 249,290,307,321,454
transformations 244,272,454, see also
various movement rules
transitive 51,55,220
tree drawing 81-6,177-87
trees 64,95,104
triangles 68
Turkish 175,310
Tzotzil 131
UG, see Universal Grammar
Ukrainian 310-1
unaccusative hypothesis 304,307,312
unary features, see features
underdetermination of the data 19-20,
28
undergeneration 219
underlying form, seeD-structure
unergative 304
ungrammaticality 9,27
unification 444,452,461
Uniqueness 444,452
universal 21,28,30
Universal Grammar (UG) 15-22,28
universal quantifier 365-7
universal semantics 362
untensed clause, see non-finite clause
v, seelight v
vP shells 375-94
V, see verb
V156-8
V -» T movement 243-60,272
V2, see verb second
Valence Principle 463,469
valency, 51
variables 365,372,451
variation 22-4
Vata 252-3
verb: light, seelight verb (v)
verb movement, see V -> T
verb movement parameter 248-9,269,
272
verb phrase (VP),72-4,170-2,265-71,
375-394
verb second (V2) 281-2
verbs (V) 43,59-60
verb-object-subjectorder (VOS) 23,
131
verb-subject-object (order VSO) 23,
133-4,243-60
VOS, seeverb-object-subject order
VP, seeverb phrase
VP-internal subject hypothesis 258
VSO,seeverb-subject-object order
zvanna-contraction 324,454
Warlpiri 39,233
Welsh 94-5,275
wh-m-situ 342-3,344,359 364-8
w/j-island 334-40,344
wJi-movement 317-52,450
w/i-parameter 362,368
wh-questions 147,317-52
wide scope 366,369
word order 175-77
X-bar theory 153-218
Yaqui, seeHiaki
yes/no questions 207-10,213,261
Yoeme, see Hiaki
Zapotec 135
Index 489
"Andrew Carnie's Syntax has quickly become the standard textbook in generative syntax
because it is neither overly technical nor artificially simple. This second edition is substantially
better and more complete. The original discussion is expanded and there are a number of new
chapters on advanced topics like raising and control, and the book continues to include
chapters that introduce alternative theories like LFG and HPSG. To my mind, this is by far the
best choice on the market today." Peter Cole, University of Delaware
"From first-hand experience, Carnie's book provides a highly readable and engaging initiation
into the mindset and preoccupations of current syntactic theory. It is useful in tying the cognitive
implications and background of current Chomskian work together with the increasing crosslinguistic
emphasis in syntax. The problem sets alone were extremely appreciated by my
undergraduates." Mark Baltin, New York University
Building on the success of the bestselling first edition, the new edition ofthe Syntax textbook
provides a comprehensive and accessible introduction to the major issues in Principles and
Parameters syntactic theory, including phrase structure, the lexicon, case theory, binding,
movement, and locality conditions. Unique among syntax texts, this book also contains short
chapters on both HPSG and LFG. Syntax succeeds in strengthening the foundational knowledge
of its readers, preparing them for moreadvanced study.
The fully revised second edition includes:
• Extended discussion in the foundational chapters, including many more sample trees
• New and extended problem sets in every chapter, all ofwhich have been annotated for
level and skill type
• A new foundational chapter on parts ofspeech
• Three new chapters on advanced topics including vP shells, object shift, gapping and
ellipsis, control, and an additional chapter on advanced topics in binding.
Syntax, Second Edition is supported by an instructor's manual and online resources for students
and instructors, available atwww.blackwellpublishing.com/carnie.
Andrew Carnie is Associate Professor of Linguistics atthe University of Arizona. He is the
editor of several books including Verb First (2005), Formal Approaches to Fonction (2003), and
The Syntax ofVerb Initial Languages (2000).
Cover image: Vasily Kandinsky, Composition 8, July 1923. Oil on canvas, 140 x201 cm. Solomon R. Guggenheim
Museum, New York, Gift, Solomon R. Guggenheim, 1937. 37.262. ©2006 ADAGP, Paris and DACS, London.
Cover design by Baseline Arts, Oxford
Printed in Singapore
For information, news, and content about
Blackwell books and journals in linguistics please visit
www.blackwellpublishing.com/linguistics
ib
. Blackwell
9J Publishing
ISBN 1-4051-3384-8
90000>
9 781405"133845