MINIMALIST INTERFACES: SELECTED ISSUES IN INDONESIAN AND JAVANESE by Yosuke Sato

MINIMALIST INTERFACES: SELECTED ISSUES IN INDONESIAN AND JAVANESE by  Yosuke Sato
MINIMALIST INTERFACES: SELECTED ISSUES IN INDONESIAN AND JAVANESE
by
Yosuke Sato
_________________________
A Dissertation to the Faculty of the
DEPARTMENT OF LINGUISTICS
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2008
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Yosuke Sato
entitled Minimalist Interfaces: Selected Issues in Indonesian and Javanese
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy
_______________________________________________________________________
Date: April 21st, 2008
Heidi Harley
_______________________________________________________________________
Date: April 21st, 2008
Andrew Barss
_______________________________________________________________________
Date: April 21st, 2008
Andrew Carnie
_______________________________________________________________________
Date: April 21st, 2008
Simin Karimi
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: April 21st, 2008
Dissertation Director: Heidi Harley
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission,
provided that accurate acknowledgement of source is made. Requests for permission for
extended quotation from or reproduction of this manuscript in whole or in part may be
granted by the head of the major department or the Dean of the Graduate College when in
his or her judgment the proposed use of the material is in the interests of scholarship. In
all other instances, however, permission must be obtained from the author.
SIGNED: Yosuke Sato
4
ACKNOWLEDGEMENTS
Having been a student for about 20 years, time has finally come for me to put an end to
this very long journey and thank countless many people who have had influences on my
life before I forget who they are. The real reason I am the way I am is because I have been
blessed with great teachers, scholars, friends, and families that have provided me with
encouragement, trust, care, and love.
My first thanks go to my dissertation committee: Heidi Harley, Andrew Carnie,
Andy Barss, and Simin Karimi. Getting Heidi’s praise on my work and progress has
been the major drive for my life in Tucson. Since I met her back in September 2, 2004,
Heidi always has served as my role model; brilliant, kind, friendly, sensitive, highly
knowledgeable; has interests in whatever topics I have worked on during my graduate
life and always has something very important to say about my work. Among her many
good qualities, Heidi is the most humble person I have ever met. She has taken great
pains not only as my academic supervisor but also as one human being in walking with
me through various stages of my progress, academic or life, sometimes dark and
sometimes challenging, so that I am doing okay. I don’t really remember how many
times I have cried in front of her due to many obstacles that have overwhelmed me, just
because she is so caring and kind, but I do remember that she always has listened to me
very patiently and gave appropriate advice so I could overcome those obstacles and
become a grown-up individual. Her extensive comments, suggestions, criticisms, and
questions on all my work have been unparalleled. They often exceed the amount of texts
I wrote myself! Nobody should be able to expect such high-quality feedback from
anybody in this busy world. In short, she has made me want to be a better scholar.
Andrew has also been a very important person for me. He has provided new directions to
pursue in my work, suggested a list of many papers that independently came up with the
ideas I thought were mine, taught me the “real” state of arts on the journal submission,
paper presentation, job applications and interviews, conference organization, and, most
importantly, how to manage my life as a scholar without getting too stressed out to
produce good work: “Reward yourself; you are worth it!”, that is what Andrew said in
October 2007. It really worked, Andrew! I will keep this sentence as my mantra
whenever I feel I am being overwhelmed with too many stuff to do in the rest of my
academic career. Andy has always given me warm encouragement and provided
extensive comments on my work since I took his seminar on ellipsis and sluicing in Fall
2006, however busy he was with his little Xander. The material in chapter 3 developed
from our discussion on Indonesian sluicing. I will also never forget his wonderful,
5
flowing lectures on formal semantics and syntax (especially, binding theory). Finally,
Simin has just been great in every respect: as my professor, as my colleague, as my
teaching supervisor, and as a human being. She read all of some 40 papers I wrote during
2004-2007 and left extensive feedback on each of them. Noticing as early as Fall 2006
that I would be burned out if I had continued my work habit at that time, she
immediately told me to stop working and relax, drive, watch, enjoy nature in the desert,
and so on. This warning also worked for the better. It is also one of my privileges to have
been able to work with Simin on the paper on modality and as her TA for LING 210. I
will never forget her warm, motherly smile every time we met!
I would like to thank many other professors at the Department of Linguistics for
giving me the best education possible: Amy Fountain, Mike Hammond, Cecile McKee,
Terry Langendoen, Janet Nicol, Massimo Piattelli-Palmarini, Adam Ussishkin, Natasha
Warner, Andy Wedel, and Mary Willie. Special thanks to Amy for being such a great
professor whose teaching philosophy influenced me most, to Andy and Adam for getting
rid of my irrational fear of phonology/phonetics and letting me enjoy it, and to Massimo
for introducing me to many linguistic giants from all over the world so that I can get the
best human networking possible. I also thank the following staff at our department for
making my life much easier: Jennifer Columbus, Marian Wiseley, Jennie Bradley,
Kimberly Young, and Emery Rosemary.
My thanks, of course, also go to my fellow students at the University of Arizona
for sharing joy, happiness, sadness, and many other experiences with me and/or being
willing to give me a hand when I need one badly. I thank the following colleagues:
Michael Anderson, Jeff Berry, Shannon Bischoff, Jordan Brewer, Lindsay Butler, Jerid
Francom, Jason Ginsburg, Erin Good, Jason Haugen, Mosa Hulden, Scott Jackson,
Sunjing Ji, Greg Key, Amy LaCross, Charles Lin, David Medeiros, Kumiko Nakamura,
Peter Norquest, Naomi Ogasawara, Jaime Parchment, Jeff Punske, Sumayya Racy,
Peter Richtsmeier, Dan Siddiqi, Tatyana Slobodchikoff, Azita Taleghani, Alex
Trueman, Mercedes Tubino-Blanco, Stacey Oberly, Sylvia Reed and Dainon Woudstra.
These people have all marked big impacts on my life but among these great colleagues,
I would like to give special thanks to: Mosa for always encouraging me through his
peculiar Finnish-style ironical critique of my work, taking me to Frisbees and drink,
driving me back and forth between the Tucson Int’l airport and my apartment, and,
most importantly, for just being such a brilliant soul mate; to Shannon for helping me
overcome many issues, academic or otherwise, through his diverse experiences in life
and linguistics; to Jaime and Jeff for giving me shoulders to cry on and general help in
6
overcoming many obstacles that got in my way; to Kumiko for deep sympathy and
kindness as another civilian from Niigata Prefecture in Japan.
I have been benefited from many scholars on many occasions whose
comments/criticisms/suggestions/ encouragements via personal discussion, e-mail
correspondences, meetings, lectures and many other forms deeply influenced my work
and academic progress. Here is some loose list of people I would like to thank: David
Adger, Seiki Ayano, Mark Baker, Bob Berwick, Cedric Boeckx, Noam Chomsky,
Sandy Chung, Barbara Citko, Peter Culicover, William Davies, Henry Davis, Marcel
den Dikken, Anna-Maria Di Sciullo, Malcolm Elliott, Catherine Fortin, Shin Fukuda,
Naoki Fukui, Elly van Gleneden, Donna Gerts, Annastasia Ginnakidou, Kleanthes
Grohmann, Youssef Haddad, Nobuko Hasegawa, Hui-Yu Huang, Jiro Inaba, Yuriko
Hatori, Kyle Johnson, Taro Kageyama, Ayasha Kidwai, Masaaki Kamiya, Richard
Kayne, Maki Kishida, Hideki Kishimoto, Paul Kroeger, Richard Larson, Howard
Lasnik, Roger Martin, Bradley McDonnell, Jason Merchant, Shigeru Miyagawa,
Tomohiro Miyake, Takashi Munakata, Pam Munro, Masanori Nakamura, Kunio
Nishiyama, Masayuki Oishi, Masha Polinsky, Jerrold Saddock, Mamoru Saito, Ivan
Ortega-Santos, Lisa Selkirk, Peter Sells, Kayono Shiobara, Tom Stroik, Hisao Tokizaki,
Takashi Toyoshima, Lisa Travis, Juan Uriagereka, James Yoon, Akira Watanabe, and
Paul Willis. I also wish to take this opportunity to express my deep gratitude to the
anonymous reviewers for providing many detailed comments to improve the quality of
my articles: Gengo Kenkyu, English Linguistics, Journal of East Asian Linguistics,
Journal of Linguistics, Journal of South East Asian Linguistic Society, Language,
Lingua, Linguistic Inquiry, Natural Language & Linguistic Theory, Natural Language
Semantics, Snippets, Syntax, Studia Linguistica, and The Linguistic Review.
I have been extremely fortunate to meet many scholars, teachers and colleagues in
Sendai and Niigata whose encouragements and support made it possible for me to continue
my linguistic research. I thank Masaru Nakamura, Yoshiaki Kaneko, Yoshiki Ogawa,
Nobuhiro Miyoshi, Yoshihito Dobashi, Dana Lupsa, Mika Takahashi, Tomohiro Oshima,
Takahiro Tozawa, Kazuyuki Nakano, Shinichiro Hatade, Akito Kurogi, and Yukie Hara. I
don’t know how to thank Nakamura-sensei enough for providing me with support and
encouragement to make sure that I am doing all right in Tucson since I left Sendai in Fall
2004 and for giving me a fundamental sense of direction in linguistics. I regret that I could
not be in Sendai to celebrate his retirement from Tohoku University in March, 2008. I just
hope that this dissertation would serve as a small token of my appreciation for his constant
guidance and support and as a sign that I accomplished what I promised him to do in
America. Special thanks also go to Kaneko-sensei, Ogawa-sensei, Miyoshi-san, and
7
Dobashi-san for providing many critical comments and suggestions on my half-baked
ideas and/or supporting my decision to study abroad. My thanks also extend to Takamichi
Aki and Tsuyoshi Oishi for first introducing me to the exciting field of linguistics called
generative grammar and sending me go to Sendai to study more about this discipline.
Thanks also go to Yuso Yoshi for many conversations regarding English learning/teaching
and for telling me that English is worth living for, way back when I was 18. Your words
will always be in my heart.
My life could not come this far without bread and butter. My first two years of
graduate life have been supported by the Fulbright Fellowship provided by the
International Institute of Education/IIE and the Japan-United States Education
Commission/JUSEC; the other two years have been supported by Teaching Assistantships
from the Department of Linguistics. I am very grateful to these organizations for
supporting my life here in Tucson. Special thanks to Mizuho Iwata and Miyuki Ito (from
JUSEC), Emily Bosio Gutridge (from Denver IIE), as well as Shadi Moqbel and Selim
Ben-Said (my Fulbright fellows), for always making sure that I will complete my program
as promised. I also thank the Graduate & Professional Student Council of the University of
Arizona for the Professional Opportunities Grant to make my planed conference end with a
great success and thereby to start a new tradition that I want to persist after I leave Tucson.
Finally, I want to give deepest thanks to my family back in Niigata, Japan and Kendal,
Central Java, Indonesia. I am enormously grateful to my parents, Toyoki Sato and Yuko
Sato, my grandparents, Sakae Sato and Fusa Sato, my two sisters Asami Sato and Naomi
Sato, and my late great grandmother Torano Sato for their love, trust, and unshakable faith
in me. Whether they know what I have done in Tucson, I can at least tell that I had the
greatest time here thanks to your support. I favorably remember that they have been
always proud to say that I have been doing something worthwhile in Tucson. None of my
family members in Niigata would have ever imagined that the poor kid who got F’s on all
subjects in elementary school and whose only good quality was to play Kendo well would
come this far to finish the PhD program in linguistics. My warm thanks also go to my
family in Indonesia who welcomed me into their house with very big smiles at first sight
and taught me by example that life contains many lessons that are equally or more
important than linguistics. Thank you very much to you all for your care, love, and support.
My life so far has been very dramatic. In retrospect, my stay at SUNY Buffalo, NY
during July 14-August 4, 2008 as the pre-academic orientation for graduate students was
THE turning point in my life, when I finally met the person I wanted to grow old with and
become happy with. (Thanks again to the Fulbright, for that matter, for sending me to
Buffalo!) She was to be my ambassador of happiness from Indonesia. She came from
8
Bowling Green, OH all the way into the southwest desert in Tucson in May 2006 for the sole
purpose of living with me and supporting me in every way imaginable until this very
moment. She has always been with me in times of sadness and happiness in too many forms
to list: as my sister, as my consultant, as my mother, as my wife, as my best cook, as my best
teacher, as my best soulmate, as my best colleague, and as my everything good. She
understands everything I would say before I actually do; she understands every emotion I
have; her love and care for me is unparalleled. She has many superb qualities but her inner
beauty is what I respect the most. I feel in tremendous awe and at a loss for words when I
recall that such a great human being like her chose to complete me. Thinking this way makes
me try to even stand her frequent kentuts, far more potent than my socks... I give my
warmest thanks to Dwi Hesti Yuliani with love and respect. Teri makasi, sayangku tercinta.
Aku cinta kamu.
9
DEDICATION
To my families back in Niigata, Japan and Kendal, Indonesia,
with deepest appreciation, respect, and love
10
TABLE OF CONTENTS
List of Figures…………………………………………………………………………….14
List of Tables……………………………………………………………………………..15
List of Abbreviations……………………………………………………………………...16
Abstract…………………………………………………………………………………. 17
CHAPTER 1: INTRODUCTION………………………………………………………...19
Section 1: Minimalist Interfaces………………………………………………………19
Section 2: Overview of the Dissertation……………………………………………….27
CHAPTER 2: SUCCESSIVE CYCLICITY AND PHASE THEORY AT THE SYNTAXPHONOLOGY INTERFACE: “VOICES” FROM INDONESIAN AND JAVANESE……38
Section 1: Introduction………………………………………………………………..38
Section 2: The Distribution of the Active Voice Morphology in Indonesian and Javanese.........42
Section 2.1: Cole and Hermon’s (1998) Generalization in Malay/Indonesian………42
Section 2.2: The Distribution of the Nasal Prefix ng- in Javanese……………….......53
Section 2.3: A Brief Excursus on Madurese………………………………………. 56
Section 3: Successive Cyclicity and the Role of vP Phases at the Syntax-Phonology Interface…..60
Section 3.1: Kayne’s (1989) Analysis of Participle Agreement in French…………...61
Section 3.2: Phase Theory………………………………………………………… 66
Section 3.3: A Phase-Theoretic Analysis of the Active Voice Deletion……………..73
Section 3.4: meN-/ng-Deletion as Failure of Vocabulary Insertion…………………... 86
Section 3.5: Section Summary…………………………………………………….... 88
Section 4: Other Alternative Accounts…………………………………………………89
Section 4.1: NP Accessibility and the Subject-Only Restriction……………………89
Section 4.1.1: Indonesian………………………………………………………92
Section 4.1.2: Javanese………………………………………………………....98
Section 4.2: Soh’s (1998) Relativized Minimality Account……………………….100
Section 4.3: Voskuil’s (2000) Pro-Based Account………………………………...... 104
Section 4.4: Fortin’s (In press) Anti-Passive Account……………………………....109
Section 4.5: Anti-Agreement-Based Accounts……………………………………. 117
Section 4.6: Case-Based Accounts………………………………………………….119
Section 5: Conclusions……………………………………………………………… 123
Section 5.1: The Derivational Nature of the Syntax-Phonology Interface………...... 124
Section 5.2: Minimalist Interfaces……………………………………………….….125
11
TABLE OF CONTENTS — Continued
CHAPTER 3: PREPOSITION STRANDING UNDER SLUICING AND REPAIR
BY DELETION: WHY INDONESIAN IS (NOT) SO SPECIAL?..................................129
Section 1: Introduction………………………………………………………………129
Section 2: Sluicing Constructions……………………………………………………136
Section 2.1: Merchant’s (2001) Analysis of Sluicing and the P-Stranding Generalization..137
Section 2.2: P-Stranding under Sluicing in languages without P-Stranding……… 141
Section 2.2.1: Brazilian Portuguese……………………………………………142
Section 2.2.2: Polish…………………………………………………………...145
Section 2.2.3: Mandarin Chinese………………………………………………150
Section 2.2.4: Malagasy………………………………………………………..153
Section 2.2.5: Serbo-Croatian………………………………………………… 156
Section 2.2.6: “Noises” in Italian, Greek, and Polish………………………… 163
Section 2.3: Section Summary…………………………………………………… 163
Section 3: The Syntax of Sluicing in Indonesian……………………………………...164
Section 3.1: Indonesian Sluicing = Pseudoslucing?........................................................165
Section 3.2: Cheng (1991): A Reduced Cleft Analysis……………………………....180
Section 3.3: Cole et al. (To appear): A Headless Relative Clause Analysis……............183
Section 3.4: Sluicing in Indonesian ≠ Pseudosluicing: Why Indonesian is So Special…....189
Section 4: Repair by Deletion at the Syntax-Phonology Interface…………………….190
Section 4.1: Feature Percolation, Minimality, D-to-P Incorporation, and Interface Repair.......191
Section 4.1.1: Whether the [+wh] Feature Percolates or Not…………………...192
Section 4.1.2: D-P Coalescence as D-to-P Incorporation……………………….196
Section 4.1.3: Repair of “Imperfections” at the Syntax-Phonology Interface…...202
Section 4.2: Towards an Etiology of P-Stranding Violations across Languages….…207
Section 4.2.1: English………………………………………………………….209
Section 4.2.2: Indonesian……………………………………………………....210
Section 4.2.3: French…………………………………………………………..216
Section 4.3: Pseudogapping in Indonesian and French…………………………..….219
Section 5: Comparison of the Proposed Analysis with Fortin’s (2007) Analysis……..223
Section 5.1: Fortin’s (2007) LF Copy + Long Distance Agree Analysis……………223
Section 5.2: Four Problems with Fortin’s (2007) Analysis……………………….....227
Section 5.3: Derivationalism vs. Representationalism……………………………....229
Section 6: Conclusions………………………………………………………………231
Section 6.1: Chapter Summary…………………………………………………………… 231
Section 6.2: Implications of Indonesian Sluicing for the Theory of the Syntax-Phonology Interface..233
12
TABLE OF CONTENTS — Continued
CHAPTER 4: THE MORPHOSYNTAX OF BARE NOMINALS IN INDONESIAN AND
JAVANESE: A RELATIVIZED PARAMETRIC THEORY OF NOMINAL DENOTATION.....235
Section 1: Introduction……………………………………………………………….235
Section 2: Chierchia’s (1998a, b) Nominal Mapping Parameter……………………...240
Section 3: The Denotation and Morphosyntax of Bare Nominals in Indonesian and Javanese....243
Section 3.1: Bare Nominals in Indonesian………………………………………...244
Section 3.2: Bare Nominals in Javanese…………………………………………..258
Section 4: A Relativized Parametric Theory of Nominal Denotation: From Indonesia to the World….265
Section 4.1: A Relativized Parametric Theory of Nominal Denotation and Morphosyntax….266
Section 4.2: Deriving the Denotation and Morphosyntax of Bare Nominals across Languages..284
Section 5: New Typological Predictions of the Proposed Analysis…………………...311
Section 5.1: Child English Morphosyntax………………………………………....311
Section 5.2: Slavic Morphosyntax………………………………………………...317
Section 5.3: Chinese Morphosyntax………………………………………………319
Section 5.4: Other Languages……………………………………………………..322
Section 6: Chapter Summary: Theoretical Implications for the Syntax-Semantics Interface…323
CHAPTER 5: WHETHER WH-IN-SITU MOVES OR NOT IN INDONESIAN: CHOICE
FUNCTION AND INTERFACE ECONOMY……………………………………… .333
Section 1: Introduction……………………………………………………………….333
Section 2: Wh-in-Situ in Indonesian………………………………………………….336
Section 2.1: “Overt” Syntactic Movement?....................................................................338
Section 2.2: Covert Syntactic Movement?....................................................................343
Section 2.3: Unselective Binding?.................................................................................356
Section 2.4: Section Summary……………………………………………………363
Section 3: Wh-in-Situ in Indonesian is Not an Interrogative Definite Description…….365
Section 3.1: Saddy’s (1991) Analysis of Wh-in-Situ as Interrogative Definite Description…...365
Section 3.2: Problems with Saddy’s (1991) Analysis……………………………...370
Section 4: Choice Function as an Optimal Interface Strategy at the Syntax-Semantics Interface.....375
Section 4.1: Choice functions……………………………………………………..377
Section 4.2: Deriving the Properties of Wh-in-Situ in Indonesian…………………386
Section 4.3: New Predictions: The NP vs. non-NP Asymmetry………………...…393
Section 4.4: Cole and Hermon’s (1998) Unselective Binding vs. Choice Function…….395
Section 5: Conclusions………………………………………………………………...402
13
TABLE OF CONTENTS — Continued
CHAPTER 6: REDUPLICATION ASYMMETRIES IN INDONESIAN AND THEIR
THEORETICAL IMPLICATIONS FOR THE LEXICON-SYNTAX INTERFACE…..411
Section 1: Introduction……………………………………………………………….411
Section 2: Lexicalist vs. Non-Lexicalist Theories…………………………………….417
Section 2.1: Lexicalist Theories…………………………………………………...417
Section 2.2: Non-Lexicalist Theories………………………………………….…..427
Section 3: Asymmetries between Nominal and Verbal Reduplications in Indonesian...432
Section 4: Reduplication Asymmetries in Indonesian and lexicalist Theories………...437
Section 4.1: Chomsky’s (1970) Weak Lexicalist Hypothesis……………………...437
Section 4.2: Anderson’s (1982, 1992) Weak Lexicalist Theory…………………....447
Section 4.3: Kiparsky’s (1982a, b, c, 1985)/Mohanan’s (1986) Lexical Phonology…..449
Section 4.4: Di Sciullo and Williams’ (1987)/Williams’ (2007) Strong Lexicalist Theory…454
Section 4.5: The “Lexicon” as the Source of the Ordering Paradox……………….457
Section 5: A Distributed Morphology Approach to Reduplication Asymmetries in Indonesian...458
Section 5.1: Nominal Reduplication………………………………………………459
Section 5.2: Verbal Reduplication………………………………………………...462
Section 6: Conclusions……………………………………………………………..471
Section 6.1: Implications for the Proper Theory of the Lexicon-Syntax Interface.....474
Section 6.2: Minimalist Interfaces……………………………………………........476
CHAPTER 7:
Section 7.1:
Section 7.2:
Section 7.3:
MINIMALIST INTERFACES………………………………………….478
Summary of the Dissertation……………………………………….........478
Minimalist Interfaces……………………………………………………480
Questions for Future Research and Conjectures about Linguistic Interfaces…..491
REFERENCES………………………………………………………………………….503
14
LIST OF FIGURES
Figure 1: The Architecture of the Interfaces under the Thesis of the Minimalist Interface...21
15
LIST OF TABLES
Table 1:
The Parametric-Theory of P-Stranding at the Syntax-Phonology Interface........131
Table 2:
Pseudosluicing Diagnostics in Indonesian.............................................................166
Table 3:
A Relativized Parametric Theory of Nominal Denotation (preliminary version).........283
Table 4:
A Relativized Parametric Theory of Nominal Denotation (final version)...........325
Table 5:
The Corpus Survey of Four Popular Newspapers in Indonesia..............................433
(approx.160, 000 words)
16
LIST OF ABBREVIATIONS
Abs
Acc
Act
Av
C
Circum
Cl
Class
Conj
Cop
Dat
Det
Emp
Erg
Ez
f
1
Foc
Fut
Gen
Hab
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Absolutive Case
Accusative Case
Actor Topic
Active voice
Complementizer
Circumstantial
Clitic
Classifier
Conjunctive Particle
Copula
Dative
Determiner
Emphatic Particle
Ergative Case
Ezafe particle
Feminine
First Person
Focus
Future Tense
Genitive
Habitual
Hyp
Instr
Intr
Loc
Link
Neg
Nom
Ov
Part
Past
Perf
Pl
Prog
Pv
Q
Red
Rel
Sg
3
Top
Vz
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Hypothetical
Instrumental
Intransitive
Locative
Linker
Negation
Nominative Case
Object voice
Participal
Past Tense
Perfective Aspect
Plural
Progressive
Passive voice
Question morpheme
Reduplication
Relative clause
Singular
Third person
Topic Marker
Verbalizer
17
ABSTRACT
This dissertation is a theoretical investigation of the thesis of Minimalist Interfaces, namely,
that syntax-external linguistic interfaces that interact with the core syntactic computation
and language-independent sound and concept systems play a more critical role in
manipulating syntactic objects to make them legible to those systems than is currently
assumed in the recent minimalist inquiry. The core theme of this thesis lies in the idea that
syntax is not entirely crash-proof but could make a variety of derivational mistakes;
phonological and semantic linguistic interfaces conduct a handful of independent domainspecific operations to attempt to legitimize illicit syntactic objects, if any, for the purposes
of legibility at the language-external sound and concept systems. Evidence is provided that
the syntax-external components use whatever resources they can to repair certain
“imperfections” created by syntax but only within the range of options made available by
the universal principles of syntax in tandem with the language-specific parameter values.
This dissertation explores some of the ramifications and empirical consequences of this
thesis based on the comprehensive description of a sizable portion of the grammar of
Indonesian and Javanese collected by my fieldwork with three native Indonesian and
Javanese consultants. Phenomena discussed here include the distribution of active voice
morphology, P-stranding under sluicing, the denotation and morphosyntax of bare
18
nominals, wh-in-situ questions, and reduplication asymmetries between nominal and
verbal derivational affixes. These diverse ranges of phenomena in the two languages are
analyzed in depth to provide converging evidence that the thesis of minimalist interface as
defined above yields a deep understanding of the way the syntax interacts with the
language-dependent interfaces responsible for phonological and semantic interpretation.
The investigation conducted here, necessitates serious reconsideration of the commonly
held view of linguistic interfaces as passive, merely ornamental components of natural
language grammar ruled by the universal law of syntax.
19
CHAPTER 1 INTRODUCTION
1. Minimalist Interfaces
This dissertation is a theoretical investigation of a thesis I call Minimalist Interfaces. The
goal is to see the extent to which this thesis serves as an adequate hypothesis of the
correspondence between syntactic computation and the linguistic interfaces responsible for
phonological and semantic interpretation.1 The fundamental idea behind this hypothesis is
that syntax-external linguistic components play a critical role in applying a handful of
domain-specific operations to legitimatize otherwise illicit objects created by the universal
combinatorial process of Merge so that they become legible for the purposes of actual use in
the language-independent articulatory-perceptual/AP and conceptual-intentional/CI systems
(Chomsky 1995, 2000, 2001, 2004, 2005). The core intuition behind this thesis is that syntax
is not entirely crash-proof, as argued for in Frampton and Guttmann (1999, 2002), in that it
makes certain derivational mistakes but syntax-external linguistic interfaces make use of
whatever resources they can to attempt to make them legible for the purposes of the external
phonetic and conceptual systems. This thesis also establishes that only certain mistakes
1
I am grateful to Andrew Carnie (personal communication) for originally suggesting the keyword minimalist
interfaces back in October 2007 as one of the potential overarching hypotheses that characterizes this whole
dissertation. Thanks also go to Noam Chomsky (personal communication) for encouraging me to take seriously
the role of linguistic interfaces and their relation with the language-independent concept and sound modules.
20
committed by syntax can be repaired at the interfaces. This idea thus leads to the claim that
syntax-external components that interface both with syntax and the AP/CI systems can make
use of their automous independent operations distinct from syntax to legitimatize certain
syntactically illegitimate objects but only within the range of options made available by the
combination of the universal law of syntax with the language-particular parameter values.
When combined with the recent derivational theory of syntax and its correspondence with
phonology and semantics (see Chomsky 1995, 2000, 2001, 2004, 2005, Epstein et al. 1998,
Uriagereka 1999, and Grohmann 2003), the thesis of minimalist interface defined above yields
the following architecture of the interface between syntax, syntax-external interfaces, and
language-independent AP and CI systems. The small lf and pf should read as the chunk of
syntactic structure that is transferred to syntax-external phonological and semantic interfaces.
The big LF and PF should read as whatever representation that the C-I and A-P systems create
based on the corresponding lf and pf.
21
Figure 1. The Architecture of the Interfaces under the Thesis of the Minimalist Interface
C-I Linguistic Interface Syntactic Computation (CHL) Linguistic Interface A-P
LF1
lf 1
pf1
PF1
LF2
lf2
pf2
PF 2
LF3
lf3
pf3
PF 3
As noted above, this mode of dynamic interpretation of the relation between syntax and
phonology/semantics is itself not a new idea but has been pursued in several different
directions within the framework of the Minimalist Program. Where they differ is primarily
in the exact size of material that undergoes transfer to the syntax-external interpretive
components. The critical feature that distinguishes the derivational model of the minimalist
interface proposed here from other instantiations lies in the idea that phonological and
semantic interfaces are not merely interpretive but also actively participate in the linguistic
computation as broadly conceived, in the sense that they use a handful of domain-specific
operations such as deletion or choice functions to repair certain imperfections created by
the functionally blind derivational process in the syntactic component. This idea, thus,
meets the leading methodological minimalist desideratum of explaining linguistic
22
phenomena in solely in terms of interface conditions. Under the present thesis, then, syntax
cannot be crash-proof in the sense of Frampton and Guttmann (1999, 2002); the objects
sent to the linguistic interfaces from syntax may or may not be legitimate by themselves,
contrary to the minimalist assumption, as first expressed in Chomsky (1995), that syntax
does not take any unnecessary or illegitimate derivational steps.
One primary goal of this dissertation, therefore, is to establish that linguistic interfaces do
whatever operations they can to attempt to create legitimate objects out of illegitimate
syntactic objects, if any, and send them off to the language-independent A-P and C-I systems
in a way that they are usable for these systems, as shown in Figure 1. In other words, syntax
creates objects that may or may not converge from the perspective of the AP and CI systems.
Specifically, when syntax creates objects as its output that do not involve any syntactic
violation and spells them out, the linguistic interfaces add minimal modifications to convert
the syntactic object into a representation that is legible to the AP and CI systems. I show in
chapters 2 and 4 that this scenario is instantiated in the proposed analysis of the distribution
of active voice morphology and the denotation and morphosyntax of bare nouns in
Indonesian and Javanese. However, when syntax creates objects that do involve certain
syntactic violations, the linguistic interfaces conduct domain-specific operations such as
deletion or choice functions to make them legible for the AP and CI systems. This does not
23
mean that this type of interface strategy is omnipotent, in the sense that all kinds of syntactic
violation can be repaired at the interfaces so that the linguistic computation will always be
able to create lfs and pfs that satisfy the demands of the AP and CI systems. For example,
certain imperfections such as the failure of feature percolation and the lack of the diversity of
communicative intent expressed by wh-questions can be repaired/supplemnted by deletion
and choice functions but other violations such as the failure of D-to-P incorporation or the
category-sensitive in-situ option for wh-questions cannot. It is in this type of case that the
minimalist interface thesis allows us to gain a deep understanding of the way syntax interacts
with its neighboring linguistic interfaces and the language-external AP and CI systems. I
show in chapters 3 and 5 that this scenario is instantiated in the proposed analysis of the Pstranding pattern and in-situ wh-questions in Indonesian.
There are several other aspects of the present thesis that bear emphasizing. First, the
present thesis indicates that syntax is functionally blind; it does whatever its abstract
computational processes such as Internal/External Merge, Agree, Spell-Out, Transfer allow
it to do to construct complex objects in a recursive fashion based on a language-particular
subset of the universal set of morphosyntactic features (such as T, v, V, C, etc.), without ever
caring about the fate of the objects thus created, leaving the task of their
interpretability/convergence entirely to the external sound-and meaning-related linguistic
24
components. Therefore, it is natural to expect that syntax creates certain objects that would
be simply unusable from the perspective of the A-P and C-I systems. This view of interfacedriven interpretability is a reasonable one in light of another consideration that what actually
interfaces with the language-independent A-P and C-I systems is not syntax per se but the
intermediate components that connect syntax and the systems, as shown in Figure 1. This
view has also been argued for in recent work as in Boeckx (2007), who proposes to let
linguistic interfaces determine the legitimacy of syntactic objects. This view, therefore,
naturally leads us to the proposal made above, namely, that linguistic interfaces are equipped
with domain-specific operations to legitimatize syntactic objects to make them readable for
the A-P and C-I modules. Second, the proposed architecture of interface summarized in
Figure 1 makes it clear that there is no room for the lexicon as traditionally conceived of as a
storage point for words and their formation processes; the traditional conception of the
lexicon does not find its natural place under the most parsimonious version of the minimalist
view of the linguistic computation, adopted in this dissertation, that what syntax interacts
with is the sound and meaning component. Accordingly, the thesis of minimalist interface
leads us to expect that (part of) the traditional roles of the lexicon in the lexicalist sense be
played by the post-syntactic linguistic interfaces by such means as late insertion of
phonological material, as recently argued for in Distributed Morphology (Halle and Marantz
25
1993, 1994; Harley and Noyer 1999; Embick and Noyer 2007). I show in chapter 6 that this
minimalist view provides a natural account of certain asymmetries between nominal and
verbal reduplication and the ber-prefixation pattern in Indonesian that would pose
empirical/architectural problems for various versions of the lexicalist theory. Third, given
that it is linguistic interfaces that directly “communicate” with the A-P and C-I systems, it is
possible to think that the nature of domain-specific operations conducted at those interfaces
are influenced by properties characteristic of AP and CI systems such as nonparsing/dephoneticiztaion, logic, and set-formation. It is shown in this dissertation that this
line of thinking yields a new analysis of the crosslinguistically peculiar P-stranding pattern
under sluicing via deletion and of wh-in-situ in Indonesian via choice function.
The present dissertation discusses some of the empirical ramifications of the minimalist
interface defined above on the basis of the comprehensive description and theoretical
analysis of a sizable portion of the syntax, semantics, phonology and morphology of two
under-represented
Malayo-Polynesian
languages
from
the
Austronesian
family,
Indonesian/Bahasa Indonesia and Javanese. I examine a wide variety of areas in these
languages where syntax interacts with phonology, morphology, and semantics. They include,
but are not limited to, the interaction of syntactic movement with the distribution of active
voice morphology, the crosslinguistically atypical P-stranding pattern under sluicing, the
26
denotation and morphosyntax of bare nominals, in-situ wh-questions, and nominal vs. verbal
reduplication asymmetries. These apparently disparate phenomena in these two languages
are analyzed in great depth to provide converging evidence for the single idea that syntaxexternal linguistic interfaces make use of whatever domain-specific resources they can to
modify/remedy certain mistakes created by syntax, if any, but only within the logical space
set up by the combination of the architectural design of syntactic computation and the
language-particular values of independently motivated parameters. To mention one case, the
analysis of the typology of P-stranding proposed in chapter 3 draws on two parameters
concerning the percolation of the [+wh] feature of the wh-word onto the PP (Chomsky 1972;
Stepanov 2001) and concerning the D-to-P incorporation in syntax (Law 1998, 2006; van
Riemsdijk 1998) to account for the three-way contrast with respect to P-stranding under whquestions and sluicing among English, French, and Indonesian. It is proposed that the
phonological component can repair the failure of the [+wh] feature percolation by deleting
the offending PP structure, but not the failure of the D-to-P incorporation, thereby providing
an explanation for why possibilities of repairs at the phonological interface are restricted by
language-particular parameter values.
Although the subtitle of this dissertation gives the impression that the database of our
inquiry is limited to Indonesian and Javanese, this dissertation also contains a wealth of
27
examples and descriptions from a far wider range of genetically unrelated languages
encompassing Indo-European, Austronesian, Altaic, and so on, as long as their
investigation bears on the question of the extent to which the thesis of minimalist
interfaces holds.
2. Overview of the Dissertation
Beyond the present introductory chapter, this dissertation is composed of 6 chapters.
Chapters 2 and 3 concern the interface of syntax with phonology, Chapters 4 and 5 the
interface of syntax with semantics, and Chapter 6 the interface of syntax with
morphology/lexicon. Below is the concise summary of the content of each successive
chapter, with particular emphasis on the relevance of the phenomena discussed therein to
the thesis of minimalist interface.
Chapter 2 is a theoretical exploration of the interface between syntax and its neighboring
phonological component within the derivational theory of syntax known as Phase Theory
(Chomsky 2000, 2001, 2004, 2005, 2006) with a case study in the voice-movement interaction
in Indonesian and Javanese. I start by reviewing the descriptive generalization, first made by
Cole and Hermon (1998), that, in Malay/Indonesian, the movement of an NP across the active
voice marker meN- results in the obligatory deletion of the active voice marker, and by
28
showing that this generalization also holds for Javanese in terms of the deletion of the nasal
active voice prefix. Drawing on the core idea behind Kayne’s (1989) analysis of participial
agreement in French, namely, that syntactic movement of an argument affects the form of a
verb within its extraction path, I propose that the obligatory deletion of the active voice
morphology in Indonesian and Javanese is the reflex at the syntax-external phonological
component of the Spec-Head D-feature checking relation that holds between the moved NP
and its local v head at the vP phase. To the extent that this analysis is tenable, the current
investigation provides important evidence for the role of the vP phase at the syntax-phonology
interface. Though numerous types of evidence have been accumulated in recent generative
research for successive cyclic movement through intermediate CPs, evidence supporting the
comparable movement through intermediate vPs has proven difficult to come by. In this regard,
the contribution from under-represented languages such as Indonesian and Javanese is highly
significant. I conclude this chapter by making it clear how the results in this chapter support the
idea that the way the phonological component conducts the deletion of active voice
morphology in Indonesian tells us a lot about the way syntactic derivation proceeds in tandem
with its neighboring phonological interface. This idea received further support from
examination of data from other languages like Irish, Kikuyu, Berber, Irish, Italian, and Turkish,
as these languages employ other interface strategies such as the complementizer alternation,
29
the loss of tonal downstep, and the anti-agreement that serve to diagnose the phase-based local
computation. This observation, then, leads to the conclusion, expected under the minimalist
interface thesis, that the phonological interface is endowed with a handful of domain-specific
operations, but can apply them only within the range of options set by principles of syntax.
Chapter 3 continues the exploration of the interface between syntax and the
phonological component through detailed investigation of the P-stranding pattern and
sluicing in Indonesian. In favor of his movement + TP deletion approach to sluicing
constructions, Merchant (2001) establishes the so-called P-Stranding Generalization that
P-stranding is permissible under sluicing only in those languages that independently allow
this option under regular wh-movement. I start by reviewing the P-stranding pattern in
several languages, such as Brazilian Portuguese, Mandarin Chinese, Polish, SerboCroatian, and Malagasy, that have been recently reported to contradict this generalization. I
show, discussing recent work on sluicing and P-stranding in these languages, that none of
these languages poses genuine counterevidence to the P-Stranding Generalization because
the underlying structure of sluicing before deletion involves some other syntactic structure
(cleft, pseudocleft, resumption) or phonological operation (P-omission at PF) than whmovement of the English kind in these languages. One primary goal of this chapter is to
establish that Indonesian is the first genuine counterexample to the relevant generalization
30
as a non-P-stranding language that nonetheless allows P-stranding under sluicing. I provide
evidence based on the distribution of the question marker -kah, first discovered by Fortin
(2007), and the obligatory lack of the complementizer yang with wh-questions with PPs
that at least PP sluices in this language are based on regular wh-movement, rejecting
alternative cleft (Cheng 1991) and headless relative clause analyses (Cole et al. to appear).
This result, thus, splits languages into three types depending on the availability of Pstranding under wh-questions and sluicing: a) a handful of languages such as English,
Danish, and Scandinavian languages that allow P-stranding under both contexts, many
languages such as French, Spanish, and Italian that disallow P-stranding under either
context, and languages such as Indonesian that allow P-stranding only under sluicing. The
second goal of this chapter is, therefore, to provide a novel analysis for this three-way
contrast across languages. Specifically, I propose that the relevant contrast receives a
straightforward parametric explanation by introducing the following two parameters: a)
whether the [+wh] feature CAN or MUST percolate from the nominal complement of a
preposition onto the dominating PP, and b) whether D-to-P incorporation is attested in the
language or not. The most important idea in the proposed analysis is that of interface
repair, namely, the notion introduced above that syntax is allowed to make certain
mistakes that can be repaired at the phonological interface. Details of the proposed analysis
31
show that certain derivational mistakes such as percolation failure, but not other mistakes
such as failure to incorporate into P heads, can be undone by deletion at the syntaxexternal phonological interface. This result, therefore, provides substantial empirical
support for the core idea of the minimalist interface thesis that syntax is not entirely crashproof and that syntax-external linguistic interfaces apply domain-specific mechanisms
such as deletion to make them legible and usable to the language-independent A-P and C-I
systems. The same result also indicates, however, that syntax still has control over the fate
of potentially illegitimate objects it sent out to its neighboring interfaces in that certain
mistakes deeply rooted in syntax cannot be repaired by interface operations. Finally, I also
provide one analysis regarding the question of why certain syntactic mistakes such as
failure of the [+wh] feature percolation can be repaired by deletion, but not other mistakes
such as the failure of D-to-P incorporation cannot be so repaired, by examining the input
and output structures of the PF component. This analysis sheds a new light of the form and
function of the syntax-external but linguistic-internal phonological component.
Chapter 4 turns to the investigation of the interface between syntax and its other neighboring
linguistic component, semantics, with a case study in the denotation and morphosyntax of bare
nominals in Indonesian and Javanese. Chierchia (1998a, b) recently proposed a restrictive
parameter of the denotation and morphosyntax of bare nominals called the Nominal Mapping
32
Parameter, which states that there is a semantic parameter concerning whether a particular
language allows its bare nominals to denote the name of a kind, the name of a property, or both
in the mapping between syntax and semantics. One attractive feature of this hypothesis lies in
the idea that the setting of this parameter exhaustively determines the morphosyntactic profile of
nominals in a given language and that all languages should be characterized as falling within
one of the three language types. My starting point in this chapter is to show, following Chung
(2000), that Indonesian and Javanese do not fit into any one of these three languages under
Chierchia’s semantic typology by demonstrating that no combination of parameter values in
Chierchia’s system would accommodate the full range of morphosyntactic properties of bare
nominals in these languages. This is an important result since Chierchia’s (1998a, b) discussion
concentrates on detailed comparison of relatively well-studied languages such as English,
Italian, and Chinese. Following the standard assumption within the Principles-&-Parameters
approach to linguistic variation (Borer 1984; Fukui 1986, 1995; Chomsky 1986a, 1995) that the
setting of the values of a parameter is localized in the inventory of functional categories, I
propose a relativized parametric theory of the denotation of bare nominals that draws on two
independently motivated ideas: a) how high a language can allow its nominal functional
superstructure to grow (Grimshaw 1991, 2005; Massam 2001; Guilfoyle and Noonan 1992;
Vainikka 1993/1994) and b) what possible values the Num head can take in a given language
33
(Carson 2000; Chung 2000). I show that this analysis provides a principled explanation for the
crosslinguistically variant semantic and morphosyntactic profile of bare nominals in not only in
Indonesian and Javanese but also in other languages such as English, Italian, Japanese, Chinese,
and Russian. I also argue that the proposed analysis makes correct predictions concerning the
development/maturation of child language grammar with particular attention to the order of
acquisition of functional categories in English-learning children documented by Guilfoyle and
Noonan (1992), Fromkin et al. (2003), and Miller and Ervin-Tripp (1973). When combined
with the Structure Building Hypothesis (Guilfoyle and Noonan 1992) and the Subset Principle
(Wexler and Manzini 1987), the proposed analysis predicts that initial acquisitional stages of all
languages should be like Javanese and Indonesian in the nominal denotation and morphosyntax
of nominals because the latter represent the simplex nominal structure. I show that this
prediction is indeed verified by utterances that are produced in the so-called telegraphic stage.
To the extent that the current analysis holds, we have no need to make recourse to rigid one-toone mapping principles as in Chierchia’s Nominal Mapping Parameter; its effects directly
follow from the complexity of nominal functional structures and the set of possible values for
the Num head parametrically set in each language. This result, therefore, provides substance to
the idea encoded in the minimalist interface, namely, that the syntax-external semantic
34
component can apply a restricted range of domain-specific operations to the output of syntax
but only within the realm of syntactic derivation.
Chapter 5 continues the exploration of the syntax-semantics interface with detailed
investigation of the syntax and semantics of wh-in-situ in Indonesian. In his seminal
work on wh-questions in Indonesian, Saddy (1991) provides extensive arguments that
there is something different about wh-in-situ in this language, whose interpretive and
syntactic properties are quite unpredictable based on the recent study of corresponding
wh-in-situ constructions in other Asian languages such as Japanese and Chinese. I review
his major arguments to show why the two most popular analyses of wh-in-situ in terms
of syntactic movement (Watanabe 1992, 2001; Huang 1982; Richards 2001) and
unselective binding (Pesetsky 1987) fail in the face of wh-in-situ in Indonesian. At the
same time, I make it clear that some of the crucial judgments reported by Saddy to
establish this result could not be reproduced in my field work with native consultants and
contradict the results reported in recent work on Indonesian and Malay in Cheng (1991),
Cole and Hermon (1998, 2000), and others. Based on this result, I seek a new analysis of
wh-in-situ informed by the guiding thesis of minimalist interface, which leads us to the
claim that the very fact that a wh-phrase remains in situ has consequences for the
semantic interface; the mechanism to license wh-in-situ does not lie in syntax per se but
35
rather in the semantic interface that connects it to the language-independent C-I system.
Specifically, I propose that the relevant mechanism is a choice function in the sense of
Reinhart (1992, 1995, 1997, 1998, 2006). I demonstrate that all apparently peculiar
syntactic and semantic characteristics associated with wh-in-situ in Indonesian directly
follow from this single mechanism at work at the linguistic interface. Finally, I compare
the present analysis with the most recent version of the unselective binding approach to
wh-in-situ in Malay/Indonesian proposed by Cole and Hermon (1998, 2000). I show that
the latter has several shortcomings in face of the so-called “Donald Duck Problem”
(Reinhart 1992) and the presence of the intermediate reading in long-distance whquestions (Ruys 1992; Reinhart 1998, 2006) that is successfully resolved under the
former from the very definition of choice function without any ad hoc stipulations on the
syntax-semantics mapping. The proposed analysis, if correct, provides evidence that the
semantic interface also is endowed with a handful of modular-specific operations such as
choice functions to legitimatize incomplete syntactic objects and send them off to the C-I
system in a legible manner. I also briefly address the important question of why the
semantics interface develops this particular semantic operation. I speculate that this state
of affairs is naturally expected because what actually interfaces with the conceptual
system is the linguistic interface, not syntax, and it is reasonable to expect that the
36
linguistic semantic interface “borrows” some mathematical properties characteristic of
the conceptual system such as set formation, predicate logic, and so on.
Chapter 6 turns to the interface of the syntax with the morphology to provide part of the
answer from Indonesian and Javanese facts to the hotly debated issue of the division of labor
between syntactic computation and the so-called lexicon as a storage of words and their
formation processes, as commonly conceived of in the generative enterprise. The empirical
domain on which I base my investigation is a new observation that there is a curious
asymmetry between nominal and verbal reduplication in Indonesian. A corpus study of four
popular newspapers published in Indonesia reveals that verbal derivational affixes have a
strong tendency to feed only stem reduplication whereas nominal derivational affixes allow
either stem reduplication or stem-affix reduplication. I show that this new observation is also
confirmed by the data I elicited with one native Indonesian consultant. I show that the steminternal reduplication pattern as well as the observed asymmetry pose empirical/architectural
paradoxes for several well-known variants of the so-called lexicalist theory as in Chomsky
(1970), Anderson (1982, 1992), Kiparsky (1982a, b, c: 1985), Monahan (1986), and Di Sciullo
and Williams (1987). Since the debate between lexicalist and non-lexicalist approaches to
word formation has quite a long history sometimes coupled with heated rhetoric, I make it
clear what specific aspects of these variants of the lexicalist theory are not tenable with respect
37
to the facts in Indonesian reduplication. Based on this result, I propose a morphosyntactic
analysis of Indonesian reduplication within the uni-modular syntactic approach to word
formation as in Distributed Morphology (Halle and Marantz 1993, 1994; Marantz 1997;
Harley and Noyer 1999; Embick and Noyer 2007). I show that the observed facts receive a
straightforward account under the post-syntactic bottom-up cyclic insertion of phonological
features once we take seriously a hierarchical arrangement of morphosyntactic features and the
underlying syntactic category of input stems for reduplication. The architecture of the syntaxmorphology interface that emerges from this investigation is one where there is in fact no such
interface in the strict sense, because the current analysis indicates that morphological structure
is itself syntactic structure unless otherwise motivated (Harley and Noyer 1999; Embick and
Noyer 2007). This conclusion, I argue, is optimal under the most restrictive view of the
minimalist interface guideline that language must minimally interface with the A-P and C-I
systems, but not with any other language-internal modules such as lexicon, unless empirical
evidence suggests otherwise.
Chapter 7 summarizes the high points of the contents of my investigation in the
previous chapters and draws some conclusions regarding the architecture of the faculty
of human language, the nature of syntactic computation and its networking with
language-independent sensory-motor and articulatory-perceptual systems.
38
CHAPTER 2 SUCCESSIVE CYCLICITY AND PHASE THEORY AT THE SYNTAXPHONOLOGY INTERFACE: “VOICES” FROM INDONESIAN AND JAVANESE 1
1. Introduction
This chapter examines the distribution of the active voice morphology in Indonesian and
Javanese from the perspective of the syntax-phonology interface within the Minimalist
Program. One of the central theses of the theory of generative grammar since its inception is
the notion of successive cyclicity, namely, that syntactic movement occurs in a series of local
steps on its way to the final landing site. Though theoretical implementations of this thesis
have undergone several theoretical changes, e.g., Subjacency (Chomsky 1973), Barriers
(Chomsky 1986b), Shortest Move (Chomsky 1993; 1995:ch. 3), and the Minimal Link
Condition (Chomsky 1993; 1995: ch. 2), it has been generally considered as a centrally
important discovery of generative investigation. This idea has been pushed forward further by
the most recent derivational theory of syntax known as Phase Theory outlined in a series of
work by Chomsky (Chomsky 2000, 2001, 2004, 2005), which takes it that syntactic derivation
proceeds in tandem with semantic and phonological interpretation in a series of sub-chunks.
1
Earlier versions of this chapter were presented at the 2007 Western Conference on Linguistics (WECOL 2007)
held in San Diego (December 2007) and the 82nd Annual Meeting of the Linguistic Society of America (LSA)
held in Chicago (January 2008). An abridged version of this chapter is to appear in Sato (in press a, b).
39
More specifically, Chomsky proposes that, once syntactic derivation constructs the vP and CP
structures headed by phase heads v and C, their respective complements (VP and CP) undergo
Spell-Out/Transfer to the syntax-external interpretive systems responsible for phonological and
semantic interpretation in mid-derivation, contrary to the traditional T-model of syntactic
derivation in which only the final output of the derivation is Spelled-Out once, as assumed in
earlier architectures of the generative grammar such as Government-&-Binding Theory and
so-called Classical Minimalism (Chomsky 1993: 1995: ch.2). In this strictly derivational
model of syntax, elements such as direct object wh-phrases that are ultimately to undergo
movement to the specifier of CP must first be moved into the specifier/edge of vP, for
otherwise they would undergo Transfer when the vP was completed and be inaccessible to
syntactic operations at the higher CP phase level. In this chapter, I discuss the syntactically
governed distribution of the active voice marker in Indonesian and Javanese. I show that it
provides strong support for the phase-based notion of successive cyclicity, in particular, for the
role of the vP phase from the perspective of the syntax-phonology interface.
It has been widely acknowledged in the literature on Indonesian as in Chung (1976), Saddy
(1991), and Cole and Hermon (1998, 2000, 2005, in press), that, in Malay/Indonesian, the
movement of an NP across a verb results in the deletion of the active voice morphology meN-
40
from the verb.2 I show that this generalization, which I dub Cole and Hermon’s Generalization,
holds for Javanese as well. One question that has remained unresolved, however, is why this
generalization holds for these two Javanic languages. I propose, drawing on Kayne’s (1989)
analysis of French participle agreement, that the obligatory deletion of the active voice
morphology in these languages is the reflex at the PF interface of the Spec-Head D-feature
checking relation between the moved NP and its local v head at the vP phase. Specifically,
following the recent late-insertion theory of Distributed Morphology, I propose that the Dfeature-driven movement of an NP across the v head causes a change in the feature content of the
head, resulting in the failure of insertion at the post-syntactic phonological interface. The current
analysis also correctly derives the fact that the movement of non-nominal phrases such as
adjuncts and prepositional arguments does not cause obligatory active voice deletion, because
these phrases lack D-features.
2
I annotate the following allomorphs of the active voice marker in Indonesian collectively as meN- for
simplicity’s sake. (i) lists phonological conditions in which those various forms of meN- are realized.
(i) The Allomorphy of meNa. meN- → meng/menge (if the stem starts with a, e, g, h, i, o, u)
b. meN- → mem (if the stem starts with b, f, p, v)
c. meN- → men (if the stem starts with c, d, j, t, z)
d. meN- → men/meny (if the stem starts with s)
e. meN- → meny (if the stem starts with k, l, m, n, v, w, y)
41
The proposed analysis of the distribution of the active voice morphology allows us to
draw two conclusions for the phase theory of syntax and the syntax-phonology interface. First,
as stated above, the phase-bound system of syntactic derivation predicts that syntactic
movement occurs via the edge of CP and vP until its final landing site. Although evidence
abounds in the generative literature for the movement to the edge of CP, evidence has been
hard to find for comparable movement via the edge of vP; see Legate 2003 and the reply to
Legate by den Dikken 2006). The syntactically governed distribution of the active voice
morphology in Indonesian and Javanese is important in that it provides clear evidence from the
syntax-phonology interface for the role of the vP phase in the syntax. Second, the proposed
analysis provides substance to the thesis of “minimalist interfaces” introduced in the
introductory chapter. It shows that the syntax provides a parametrically defined curve that the
syntax-external interpretive components must follow with whatever domain-specific resources
they can avail themselves of to mirror the way syntax works as closely as possible. In other
words, languages will do whatever they do within their available morphological and
phonological resources to reflect the local step of movement via the edge of vP and CP, as
required by the phase-bound computation.
42
2. The Distribution of the Active Voice Morphology in Indonesian and Javanese
In this section, I provide a comprehensive description of the prefixes, meN- in Indonesian
and ng- in Javanese. I concentrate on the distribution of these prefixes, which has been shown
in recent research (Chung 1976; Saddy 1991; Cole and Hermon 1998, 2000, 2005, in press) to
play a crucial role in diagnosing certain types of syntactic movement. I start in section 2.1 with
the description of meN- in Indonesian and introduce Cole and Hermon’s generalization that the
movement of an NP cannot occur across an active verb unless the active voice meN- prefix is
deleted. I show in section 2.2 that the same generalization also correctly governs the
distribution of the nasal prefix ng- in Javanese. In section 2.3, I review Davies’ (2003)
evidence from Madurese, an Austronesian language spoken in the Madura Island in Indonesia,
which shows that a similar generalization also holds for this language.
2.1. Cole and Hermon’s 1998 Generalization in Malay/Indonesian
A large body of work has been produced in the literature on Indonesian on the exact nature
of the prefix meN-. It is termed the “transitive/agentivity marker” by Chung (1976), the
“agent trigger” by Englebretson (2003), the “[+active] feature” by Postman (2002), the
“actor topic” by Guilfoyle et al. (1992), or the “active voice” by Sneddon (1996) and
Voskuil (2000). For the purposes of this chapter, I assume that meN- is the marker of
43
active voice in Indonesian but is also affected by the syntactic transitivity of a verb that this
affix is attached to. Evidence for this assumption comes from the following two facts. First,
this prefix may appear in an active sentence but not in a passive sentence, as shown by the
contrast between (1a) and (1b).
(1) a. Esti mem-baca/*di-baca
Esti AV-read/PV-read
buku
itu.
book
that
‘Esti read that book.’
b. Buku itu *mem-baca/di-baca oleh Esti.
book
that AV-read/PV-read by
Esti.
‘That book was read by Esti.’
In (1a), an active sentence, the verb baca ‘read’ is prefixed by meN-, but not by di-,
whereas the reverse situation holds in (1b), a passive counterpart to (1a). This contrast
indicates that meN- signals active voice in Indonesian. Second, Fortin (in press) observes
that, though meN- prefixation is optional with many transitive verbs in colloquial
Indonesian, there is one case in which this prefixation is obligatory. The relevance case is
44
when a transitive verb is used in its intransitive use, as in English examples such as John
ate at noon. This is illustrated in (2a, b).
(2) a. Ali
Ali
sedang mem-baca.
Prog
AV-read
‘Ali is reading.’
b. * Ali
Ali
sedang baca.
Prog
read
‘Ali is reading.’
(Fortin in press)
In a normal transitive sentence as in (2a), meN- is optional. However, in its intransitive
counterpart of (2a) as in (2b), this affix is obligatory. These two arguments, therefore,
support my working assumption that meN- is an active voice marker whose realization is
affected by syntactic transitivity (see Gil 2002 for a review of the historical development
of this marker). To be more specific, since the diathesis here is connected to voice, as
shown in (1a, b), and to the presence of overt objects, as shown in (2a, b), it is natural to
associate it with v, which is the locus of external arguments (Hale and Keyser 1993, 2002;
Chomsky 1995) and of accusative Case (Kratzer 1996).
As noted by recent work on the syntax of Indonesian as in Chung (1976), Saddy (1991),
and Cole and Hermon (1998, 2000, 2005, in press), however, this characterization of meNin terms of voice morphology and transitivity alone fails to explain certain interesting
45
restrictions on the obligatory deletion of the prefix caused by a particular type of
movement. Cole and Hermon (1998) observe, drawing on data and observations made
earlier by Chung and Saddy, that the distribution of meN- in Singapore Malay and
Indonesian is governed by the generalization in (3). All the examples from Cole and
Hermon (1998) are based on Malay but the same generalization they developed for Malay
also holds for Indonesian; for example, they note that “the distribution of meng- in
Indonesian…seems to be identical to what we have found in Singaporean Malay” (p. 230).
(3) Cole and Hermon’s Generalization
The obligatory omission of meng- with verbs that would otherwise permit mengindicates the movement of an NP argument over the meng- + verb.
(Cole and Hermon 1998: 233)
This generalization accommodates all existing cases of meN-deletion. Let us start with A′movement. Wh-questions and relativization cause meN- deletion from all verbs that the
moved NP/DP crosses on its extraction path toward its final landing site, as illustrated in
(4a, b) and (5a, b).
46
(4) Wh-Questions
a. Apai yang Bill (*mem)-beritahu
what that Bill
ibu-nya [yang Fatimah (*mem)-beli ti]?
AV-tell
mother-his that Fatimah AV-buy
‘What did Bill tell his mother that Fatimah bought?
b. Siapai yang Bill (*mem)-beritahu
who
that Bill
AV-tell
ibu-nya [ ti
(mem)-beli
buku]?
mother-his
AV-buy
book
‘Who did Bill tell his mother bought a book?’ (Indonesian)
(Indonesian: modeled after the Malay examples from Cole and Hermon 1998: 231, 232)
(5) Relativization
a. [Bukui [OPi yang [John (*mem)-beli ti ]]
book
that
John
AV-buy
itu]
menarik.
the
interesting
‘The book that John bought is interesting.’
b. [Lelakii [OPi yang [ti (mem)-beli buku
man
that
AV-buy
book
itu]]] adik
saya.
that brother
my
‘The man who bought that book is my brother.’
(Indonesian/Malay: slightly modified from Cole and Hermon 1998: 233)
47
In (4a), A′- movement occurs from the embedded object position to the matrix [Spec, CP].
The active voice prefix is obligatorily deleted from both the higher verb beritahu ‘tell’ and the
lower verb beli ‘buy’ in conformity with Cole and Hermon’s generalization. The same
generalization also accounts for the mandatory deletion of meN- from the higher verb, but not
from the lower verb, in (4b); wh-movement of the embedded subject siapa ‘what’ crosses
only the matrix verb beritahu ‘tell’. The same pattern holds for relativization, as shown in (5a,
b). In (5a), operator movement of the object crosses the verb beli ‘buy’, causing the deletion
of meN- from the verb. This deletion does not occur in (5b), because the null operator
movement of the subject does not occur across the same verb.
Further support for Cole and Hermon’s (1998) generalization comes from the movement
options allowed in Malay/Indonesian and their effects on the active voice deletion. As first
discovered by Saddy (1991), Indonesian has three ways of forming wh-questions: a) overt
wh-movement, as illustrated in (4a, b) above, b) partial wh-movement, and c) wh-in-situ; see
chapters 2 and 5 of the present dissertation for various analyses of Indonesian wh-questions.
The three options are illustrated in (6a-c), respectively. (The examples in (6b, c) are from Cole
and Hermon 1998: 233, 237.)
48
(6) Full Wh-Movement, Partial Wh-Movement, and Wh-in-situ in Indonesian/Malay
a. Full Wh-Movement
apa i Ali (*mem)-beritahu
kamu tadi
what Ali
you just now that Fatimah
AV-told
[ti yang Fatimah (*mem)-baca ti]?
AV-read
‘What did Ali tell you just now that Fatimah was reading?’
b. Partial Wh-Movement
Ali
(mem)-beritahu
kamu tadi
Ali
AV-told
you
[apai
yang Fatimah (*mem)-baca ti]?
just now what that
Fatimah
AV-read
‘What did Ali tell you just now that Fatimah was reading?’
c. Wh-in-situ
Ali
(mem)-beri Fatimah apa?
Ali
AV-give
Fatimah what
‘What did Ali give Fatimah?’
(Indonesian/Malay)
In (6b), the movement of the embedded direct object apa ‘what’ targets the embedded,
non-scopal [Spec, CP], though the interpretive outcome is the same as that of fully moved
wh-questions as in (6a). The active voice deletion occurs from both the higher and lower
verbs in (6a) whereas it occurs only from the lower verb in (6b). When the in-situ option is
49
utilized as in (6c), there is no meN-deletion observed from the verb beri ‘give’. All these
patterns, thus, show that the form of a verb is affected by the options of wh-movement
independently available in Indonesian.
Let us now turn to A-movement. As in A′-movement, passivization and object
preposing in Indonesian causes active voice deletion from verbs that it crosses. Consider
(7a, b) (= (1a, b)) and (8a, b).
(7) Passivization
a. Esti (mem)-baca/(*di)-baca buku itu.
Esti AV-read/PV-read
book that
‘Esti read that book.’
b. Buku itu (*mem)-baca/*(di)-baca oleh Esti.
book that AV-read/PV-read
‘That book was read by Esti.’
by
Esti
50
(8) Object Preposing
a. Buku itui
adik
saya (*mem)-beli ti.
book that brother my
AV-buy
‘My brother bought that book./That book was bought by my brother.’
b. Alii saya (*men)-cubit
Ali
I
ti.
AV-pinch
‘I pinched Ali./Ali was pinched by me.’
(Indonesian/Malay: Cole and Hermon 1998: 232)
In (7b), a passive counterpart to the active sentence in (7a), the active voice marker must
be replaced by the passive voice marker di-. 8a, b) might appear to involve topicalization, a
case of A′-movement. However, evidence based on the interaction of object preposition
and Equi NP Deletion first presented in Chung (1976) and further expanded by Hopper
(1983), Musgrave (2001), Aldridge (in press), and Cole and Hermon (2005, in press),
suggests that object preposing in Indonesian is derived by A-movement of a logical object
into [Spec, TP]. Consider (9a-d).3
3
The examples here are given here as they appear in Chung (1978: 46, 47) except slight modifications of the
morphological glosses. The orthography used here reflects the one used prior to the spelling reform of 1972.
51
(9)a.
Dia datang untuk ber-tjakap2 dengan Ali.
he
come
for
Intr-talk
with
Ali
‘He came to talk with Ali.’
b.?* Saja mem-bawa
I
AV-bring
surat
itu untuk teman saja (dapat) (mem)-batja.
letter the for
friend my
can
AV-read
‘I brought the letter for my friends to (be able to) read.’
c.
Saja mem-bawa
I
surat
itu untuk (dapat)
AV-bring letter the for
can
di-batja oleh teman saja.
PV-read by
friend my
‘I brought the letter to (be able to) be read by my friends.’
d.
Saja mem-bawa
surat
I
letter the for
AV-bring
itu untuk (dapat) kau batja.
can
you read
‘I brought the letter to (be able to) be read by you.’
(Chung 1976: 46, 47)
Chung (1976) observes that Equi-NP deletion (subject control) targets only subject NPs, as the
contrast between (9a) and (9b) shows. The example in (9c) shows that the derived subject can be
PRO. With this background in mind, the fact that the preposed object in (9d) can be PRO
indicates that Object Preposing involves movement of the logical object to [Spec, TP], a case of
A-movement, like passivization. Therefore, the examples of object preposing as in (8a, b)
52
provide evidence that A-movement also affects the morphology of a verb that is contained within
the extraction path of the movement.
Finally, as stated in Cole and Hermon’s generalization in (3), what matters for the
deletion of the active voice morphology is the movement of a nominal category. In other
words, movement of non-nominal phrases such as adverbial adjuncts and prepositional
arguments does not cause the active voice deletion from verbs it crosses, as in (10a, b).
(10) Movement of non-Wh-NPs
a.
Kenapai
Mary
(mem)-beli
buku
itu ti?
why
Mary
AV-buy
book
that
‘Why did Mary buy that book?’
b.
Kepada
siapai Mary
(mem)-beri
buku
to
who
AV-buy
book
Mary
‘To whom did Mary give a book?’
ti?
(Cole and Hermon 1998: 231, 232)
In (10a, b), movement of phrases like kenapa ‘why’ and kepada siapa ‘to whom’ does not
result in meN-deletion though the movement itself crosses the verb beli ‘buy’/beri ‘give’ in
both cases.
53
In sum, we have reviewed Cole and Hermon’s (1998) generalization that the movement
of an NP causes the obligatory deletion of the active voice marker meN- in Indonesian.
One obvious question is, then, why is it that this generalization holds for Indonesian?
Before I answer this question, I show in the next two subsections that Cole and Hermon’s
generalization also holds for at least other Javanic languages, Javanese and Madurese.
2.2. The Distribution of the Nasal Prefix ng- in Javanese
The distribution of the active voice morphology ng- in Javanese is also constrained by
Cole and Hermon’s generalization. The examples in (11a-d) illustrate the basic paradigm.
(11) a. Basic Transitive Clause
c. A′-movement (Relativization)
Esti *(nge)-sun Fernando.
Wong lanangi sing Esti (*nge)-sun ti kuwi Fernando.
Esti AV-kiss
person male that Esti AV-kiss
Fernando
‘Esti kissed Fernando.’
Cop Fernando
‘The man that Esti kissed is Fernando.’
b. A′-movement (Wh-Questions) d. A-movement (Passivization)
Sapai Esti
who
Esti
(*nge)-sun ti?
AV-kiss
‘Who did Esti kiss?’
Fernandoi
(*nge)-sun
(karo)
Esti
Fernando
AV-kiss
by
Esti
‘Fernando was kissed by Esti.’
ti.
54
In Javanese, the nasal prefix ng- is obligatorily attached to a transitive verb, as shown in
(11a). This is in contrast with Indonesian -meN, which is optional, as we saw in the
previous subsection. The examples in (11b, c) show that A′-movement of the objects in
wh-questions and relativization leads to the obligatory deletion of the otherwise mandatory
nasal prefix from the verb that it crosses. The example in (11d) further shows that Amovement of the object with passivization has the same morphological consequence for
the form of the verb contained within the extraction path. Consider examples as in (12).
This example is given here to show that long-distance extraction of an NP such as subject
across a series of verbs also triggers ng-deletion from the verbs in the extraction path of the
NP, though it does not trigger ng-deletion on the embedded verb, which the subject does
not cross, exactly as in Malay/Indonesian.
(12) Long-Distance Extraction of Subject across Verbs in Javanese
[CP Sapai sing John
who
that John
(*ng)-ira
[CP ti nge-sun
AV-think
‘Who does John think kissed Fernando?’
AV-kiss
Fernando]]?
Fernando
55
Thus, it is only the higher verb ira ‘think’ that has its nasal prefix obligatorily deleted. This
example also shows that the morphology of a verb is affected by the movement of a phrase
across it.
Finally, as in Indonesian, it is the movement of an NP that leads to the ng-deletion in
Javanese. Thus, movement of non-nominal phrases such as adverbials (e.g., nangapa
‘why’) and prepositional arguments (e.g., ning sapa ‘to whom’) has no effect on the fate of
the active voice marker in this language either, as shown in (13a, b).4
(13) Movement of Non-Wh-NPs
a. Nangapai Esti nge-sun Fernando
why
Esti AV-kiss Fernando
‘Why did Esti kiss Fernando?’
4
ti? b. Ning sapai Esti nge-irim packet ti?
to
whom Esti AV-send package
‘To whom did Esti send a package?’
Heidi Harley (personal communication) asks whether Indonesian allows pseudopassivization like English, as in (i).
(i) This bed was slept in by Washington.
The prediction would be that the meN-deletion should occur in this construction because the movement of an NP
crosses the (reanalyzed) verb. Unfortunately, this prediction is impossible to test for two reasons. First, Indonesian
does not have pseudopassivization. Second, as we saw earlier, passivization introduces the passive suffix di-, which
is in the complementary distribution with meN-. Similarly, P-stranding under wh-movement would be predicted to
cause meN- deletion, but this prediction also cannot be tested since, as we will see in chapter 3, P-stranding is
impossible in Indonesian.
56
The evidence presented here, therefore, shows that the Javanese voice-movement
interaction is also governed by Cole and Hermon’s generalization, originally established
for Indonesian.
2.3. A Brief Excursus on Madurese
Evidence presented in Davies (2003) indicates that Madurese, a language spoken on the
Madura Island of Indonesian, is also subject to the voice-movement alternation governed
by Cole and Hermon’s (1998) generalization. Consider the basic paradigm in Madurese,
given in (14a-c).
(14) The Basic Voice System in Madurese
a. Embi’ juwa
goat
that
ng-ekke’ Ali.
AV-bite Ali
‘The goat bit Ali.’
b. Ali
Ali
embi’ juwa
kekke’.
goat
bite
that
‘The goat bit Ali.’
57
c. Ali
e-kekke’
(bi’) embi’
juwa.
Ali
OV-bite
with goat
that
‘The goat bit Ali.’
(Davies 2003: 246)
In Madurese, the active voice morphology ng- appears on the verb when the agent or actor
of a transitive sentence occurs as the subject, as shown in (14a). In passive examples as in
(14b, c), the active voice marker must be eliminated; either the verb appears as a bare stem
or the overt object voice prefix e- is attached to the verb. The appearance of the object
voice marker in (14c) can be analyzed on a par with the voice-movement interaction
observed in Tagalog and Malagasy (Rackowski and Richards 2005; Pearson 2001, 2005);
see also section 3.4 for related discussion.
With this in mind, the contrast between (15a) and (15b) shows that the syntactic
movement of a phrase across a verb results in the active voice deletion from that verb.
(15) Wh-Movement and the Voice-Movement Interaction in Madurese
a.
Sapai se
e-kera
who
OV-think Ali
Rel
Ali
(ja’/se) ti
melle
motor?
C/Rel
AV.buy
car
‘Who did Ali think bought a car?’
58
b. * Sapai
who
se
ngera
(ja’/se) ti
Rel
AV.think C/Rel
melle
motor?
AV.buy
car
‘Who did Ali think bought a car?’
(Davies 2003: 246)
In these examples, the movement of sapa ‘who’ contains the higher verb within its
extraction path, triggering the deletion of the active voice marker ng-, as in (15b), or its
replacement with the object voice marker.
Further evidence that Cole and Hermon’s generalization is at work in Madurese comes
from the correlation between the deletion of the active voice morphology and whmovement options. As in Indonesian and Javanese, Madurese allows overt wh-movement,
partial wh-movement, and wh-in-situ, as illustrated in (16a-d).
(16) Wh-questions in Madurese and the Voice-Movement Interaction in Madurese
a. Apai se
e-yaken-ne Amir [CP (ja’) e-bala-agi Hasan dha’ Atin [e-baca
what Rel OV-sure-E Amir
Siti ti]]?
C OV-say-BVHasan to Atin OV-read Siti
‘What is Amir sure that Hasan told Atin that Siti read?’
59
b.
Amir yaken [CP apa se e-bala-agi Hasan
Amir sure
dha’ Atin [CP e-baca
what Rel OV-read-BV Hasan to Atin
Siti
ti]]?
OV-read Siti
‘What is Amir sure that Hasan told Atin that Siti read?’
c.
Amir yaken [CP Hasan a-bala
Amir sure
Hasan
dha’ Atin [CP apai se e-baca
AV-say to
Atin
Siti ti]]?
what Rel OV-read Siti
‘What is Amir sure that Hasan told Atin that Siti read?’
d.
Amir yaken [CP Hasan a-bala
Amir sure
dha’ Atin [CP ja’
Hasan AV-say to
Atin
Siti
that Siti
maca
apa]]?
AV.read
what
‘What is Amir sure that Hasan told Atin that Siti read?’
In the full wh-movement case shown in (16a), all the verbs must take object voice morphology.
When the partial wh-movement targets the specifier of the higher intermediate CP as in (16b),
the verbs that are contained within its extraction path all require object voice morphology. When
this type of movement targets the specifier of the lower intermediate CP, as in (16c), only the
most deeply embedded verb must be in object voice. Finally, in the case of the wh-in-situ as in
(16d), all verbs can retain the active voice morphology. These examples, thus, clearly show that
Cole and Hermon’s (1998) generalization serves to adequately characterize the movement-voice
interaction in Madurese. To complete the picture, examples as in (17) show that the movement of
60
non-nominal phrases such as prepositional objects does not result in the active-to-object voice
change, as in Indonesian and Javanese.5
(17) Dha’
to
sapa
Atin
ng-erem
paket?
whom
Atin
AV-send
package
‘To whom did Atin send a package?’
Thus, based on the data available in Madurese, it seems that Cole and Hermon’s
generalization holds for Madurese. Due to the limited availability of the data from
Madurese, however, I concentrate on facts from Indonesian and Javanese in the rest of this
chapter, hoping that the analysis developed below will also hold for Madurese.
3. Successive Cyclicity and the Role of vP Phases at the Syntax-Phonology Interface
We have seen in the previous section that Cole and Hermon’s (1998) generalization
adequately characterizes the voice-movement interaction in Indonesian and Javanese. As
stated above, Cole and Hermon (1998) do not provide a theoretical explanation for why
such a generalization holds for Javanic languages but instead merely speculates (p. 234)
5
Thanks to William Davies (personal communication) for providing me with the example in (17).
61
that “the treatment of meng- [meN- in this chapter—YS] is along the lines proposed by
Chung for wh-agreement in Chamorro.”6 The purpose of this section is to show that this
generalization provides strong morphosyntactic evidence for the role of vP phase at the
syntax-phonology interface. Specifically, I propose that the obligatory deletion of active
voice morphology from a verb that is crossed by the movement of an NP is a reflex in the
syntax-external component of the Spec-Head D-Feature checking relation between the
moved NP and its local v head at the vP phase level. This analysis also correctly captures
the fact that non-nominal phrases such as adverbials and prepositional arguments does not
cause deletion, in tandem with the independently motivated distinction between Internal
Merge and External Merge, made recently by Chomsky (2004, 2005).
3.1. Kayne’s (1989) Analysis of Participle Agreement in French
The fundamental idea I pursue below to account for the voice-movement interaction in
Indonesian and Javanese is part of the theory of “minimalist interface”, a central thesis
developed through in this dissertation, namely, that syntax-external interpretive components do
whatever they can to reflect the path that syntax carves within the parametrically defined set of
6
Cole and Hermon repeat a similar remark in Cole and Hermon (2005: 86): “the omission of meN- should be
viewed as Wh-agreement, similar to that described by Chung 1982, 1994 for Chamorro.” However, I provide
evidence in section 4.6 that the Case-based analysis is not adequate for the active voice deletion in Indonesian or
Javanese.
62
options available to each language. In the present case, this means that (morpho-) phonology is
subservient to the needs of syntax.
The idea that syntactic movement affects the morphology of verbs within its path is not
a new idea. Kayne’s (1989) analysis of participle agreement in French and other Romance
languages, followed by Chomsky (1991: 1995; ch.2), is one of the most well-known
attempts that represent this idea.7 Consider the examples in French given in (18a-c).
(18) French Participle Agreement
a.
Paul a
repeint/*repeintes les chaises.
Paul has
repainted
the chairs
‘Paul has repainted the chairs.’
b.
Je me demande
combien
de
tables
Paul
a
I
how-many of
tables
Paul
has repainted
wonder
repeintes.
‘I wonder how-many of tables Paul has repainted.’
7
I thank Andrew Carnie (personal communication) for directing my attention to Kayne’s (1989) work.
63
c.
Je medemande combien de chaises il
I
wonder
sera
repeint/*repeintes cette année.
how-many of chairs Imp be.Fut painted
this year
‘I wonder how-many of tables will be repainted this year.’
(Kayne 1989: 85, 86, 91)
French does not show overt agreement between a verbal participle and its selected direct
object, as the example in (18a) illustrates. However, when the object undergoes whmovement, as shown in (18b), there occurs overt agreement between the object and the
participle. By contrast, this agreement does not occur in impersonal passive constructions,
as shown in (18c).
Kayne argues that this pattern is naturally accounted for if there is an agreement relation
established between the wh-phrase and Agr in the manner illustrated in (19) for (18b).
(19) The Structure of the Example in (18b)
[CP combien de tablesi [TP Paul … [AgrOP ti [AgrO′ Agr [VP repeintes
Spec-Head Agreement
ti]]]]]
64
Kayne himself assumes that the intermediate position above the VP for the purposes of
agreement in (18b) is an adjoined position to the Agr projection. Chomsky (1991: 1995;
ch.2) revises this original analysis and proposes that there is a Spec-Head Agreement
relation between the moved wh-phrase and the participle that yields overt agreement
morphology on the participle in (18b). Given that agreement is contingent on the SpecHead Agreement relation between the moved NP and the AgrOP (see also Mahajan 1989;
Koopman and Sportiche 1991 for further evidence for this claim), the Spec-Head Agreement
relation in (19) is reflected morphologically on the participle either as the result of V-to-Agr
raising or Agr-to-V lowering, a choice we can leave open here.
As noted above, this participle agreement is impossible in impersonal constructions as in
(18c) above. Kayne argues that this is exactly what we expect under two independently
motivated assumptions of the theory around the early 1990s: a) movement from an A′
position to an A position is prohibited as a case of improper movement (see Fukui 1993 for
an account of this prohibition in terms of chain uniformity) and b) LF expletive replacement
is a form of A-movement (Chomsky 1986b). The presence of overt agreement means under
Kayne’s analysis that the movement of the wh-phrase combien de chaises ‘how-many of
chairs’ has occurred to [Spec, AgrOP]. Kayne takes this movement to be an instance of A′-
65
movement.8 This movement is followed by further movement of the phrase into the position
of the impersonal form il ([Spec, TP]) at LF. Kayne considers this movement a case of Amovement. This sequence of movement, thus, counts as improper movement and hence
renders (18c) ungrammatical with agreement on a par with English examples such as (20).
(20) * [TP Maryi seems [CP ti that Mary loved ti]].
However, (18c) is grammatical without agreement because the derivation of such an
example does not involve A′-movement of the wh-phrase to [Spec, AgrOP], thereby
circumventing the danger of improper movement.
In the following section, drawing on Kayne’s analysis of French participle agreement, I
propose that the voice-movement interaction in Indonesian and Javanese receives a
straightforward account within the Phase Theory of Chomsky (2000, 2001, 2004, 2005).
8
Whether this assumption holds across languages is a matter of considerable debate, as Simin Karimi (personal
communication) points out. Research as in Mahajan (1990) and Depréz (1989) argues, contrary to Kayne, that the
specifier of the AgrOP is an A-position. I leave comprehensive discussion of what determines whether the specifier
of AgrOP is an A- or A′-position for another occasion. The proposed analysis of meN-/ng-deletion casts doubt on
the A vs. A′-position.
66
3.2. Phase Theory
Let us review here several central assumptions of the Phase Theory as currently outlined in
a series of Chomsky’s work to lay the groundwork for the analysis presented in the next
subsection. One central thesis of the derivational theory of syntax pursued within the
Minimalist Program (Chomsky 2000, 2001, 2004, 2005, Epstein et al. 1998, Uriagereka 1999;
Grohmann 2003) is that phonological and semantic information is transferred to the PF and LF
interfaces ‘online’ in a piecemeal fashion (see the notion of the syntactic cycle in Bresnan 1971a,
b, Jackendoff 1972, and Lasnik 1972 for an important earlier antecedent). The phase theory
proposed by Chomsky is one particular version of this hypothesis. Chomsky proposes that this
transfer occurs, not after every application of a structure building operation (as in Epstein et al.
1998), but instead at specific derivational cascades called phases. A phase is a mid-derivational
object created by the syntactic computation that is headed by an instance of v or C.9 This theory
significantly reduces computational complexity in that the derivation can forget about material
after it has been transferred to the external systems. According to this theory, the complement of
the phase-defining heads (i.e. VP or TP) is sent to the sound and meaning components for
9
Chomsky (2000, 2001, 2004) assumes that only those verbs that instantiate “full argument structure” (transitive
and experiencer constructions) have strong phase heads, as indicated by v*. In this chapter, however, I assume
that every instance of v is a strong phase. See the following discussion on Legate (2003), who provides a number
of phonological and semantic arguments that all types of verbs, including unaccusative and passives, constitute a
strong phase.
67
phonological and semantic interpretation once a higher phase head is introduced into the
derivation. More concretely, at the point where the vP and CP structures have been built up, the
complement domains of the v and C, i.e. VP and TP, are transferred and interpreted at the
interfaces. This model yields the so-called Phase Impenetrability Condition/PIC, which can
be defined as in (21).10, 11
(21) Phase Impenetrability Condition (adopted from Chomsky 2001: 14)
In [ZP Z … [HP α [H YP]], where HP is a strong phase and ZP is the smallest strong
phase, the domain of H is not accessible to operations at ZP; only H and its edge are
accessible to such operations.
10
Chomsky (2000, 2004) proposes a slightly different formulation of the PIC, as in (i), adopted from Chomsky
(2000:108).
(i)
Phase Impenetrability Condition (adopted from Chomsky 2000: 108)
In phase α with head H, only H and its edge are accessible to operations outside α.
I adopt the definition of the PIC given in (21) for the purposes of exposition in this chapter.
11
As Heidi Harley (personal communication) points out, the PIC may mean that there is a functional processing
benefit associated with overtly marking the phase boundaries within movement chains, as independently proposed
by Givón 1979); see Lasnik 1999 for a critical discussion. Specifically, as long as a gap created by movement
remains unfilled, phase edges must indicate this information, possibly motivating the parser to search further within
the domain for the gap. I discuss this and a related question in chapter 8, when I propose to derive several
architectural properties of the phase theory from interface conditions imposed by the A-P and C-I systems.
68
As stated above, Chomsky proposes that only vPs and CPs form a strong phase. The “edge”
of the phase head H includes any specifiers of and any adjunct to H. Since CPs and vPs form
a strong phase, the PIC states that the complement of the vP phase cannot be a target for any
operations at the CP phase, as shown in a schematic derivation such as (22). This is because,
at the point the phase head C is introduced into the syntactic derivation, the VP domain of the
lower phase head v undergoes Transfer/Spell-Out to the interfaces for semantic and
phonological interpretation.
(22)
CP
C [=Z in (21)]
[=ZP in (21)]
TP
DP1
T′
T
vP
[= HP in (21)]
tDP1 [= α in (21)] v′
v [= H in (21)]
VP [=YP in (21) ]
V
DP2
Transfer/Spell-Out
69
In other words, the PIC requires that all elements base-generated within the complement of
v that are ultimately to move higher than the vP phase must first move to the edge of v
([Spec, vP]) in the manner seen in (23b) for English sentences such as (23a).
(23) a. What did Ashley buy?
b. [CP what [C′ did [TP Ashley [vP twhat [v′ tAshley [v′ v [VP buy twhat ]]]]]]]
In this derivation, the C can access only the edge of the vP due to the PIC. Thus, the direct
object what passes through the escape hatch at the edge of vP so that it may be accessible
for further movement to [Spec, CP] at the CP phase level.
This phase-based computation leads us to a prediction that the local steps of movement,
as illustrated in (23b), should somehow be reflected in the semantics and phonology of a
language by the syntax-external components. Evidence of the first kind involves semantic
facts concerning scope, binding, parasitic gaps, reconstruction, island effects, and so on. To
take a recent case, Legate (2003) provides evidence for the notion of vP phase based on
binding and reconstruction, parasitic gaps, and antecedent-contained deletion (as well as
the nuclear stress pattern of English). I reproduce Legate’s argument concerning binding
70
here. The phase-based theory as explained above predicts that successive cyclic movement
leaves copies in all intermediate vPs (as well as CPs). Drawing on the data and analyses
presented by Lebeaux (1988) and Fox (1998), Legate shows that examples of A′movement like (24a-c) that involve the interaction of binding and reconstruction effects
can only be accounted for by the copy left by movement via the edge of vP phases.
Potential reconstruction sites are indicated by underlining in these examples.
(24) The Interaction of Binding and Reconstruction in Wh-Movement (A′-Movement)
a. [Which of the papers that hei gave Maryj] did every studenti √ ask herj to read * carefully?
(Legate 2003: 507)
b.*[Which of the papers that hei gave Maryj] did shej * ask every studenti to revise * ?
(Fox 1998: 157)
On the one hand, Condition (C) of the Binding Theory requires that, in the example in
(24a), the complex wh-phrase must not reconstruct to the thematic position of the verb
read indicated by
*
. On the other hand, for the pronoun to receive a bound variable
interpretation, it must be the case that the complex wh-phrase must reconstruct below
every student. To satisfy both of these requirements, the wh-phrase must leave a copy
71
below the universal quantifier but above the thematic position of the embedded verb: a
position marked by √ . The copy in this position is naturally available if movement
stops by at the edge of vP phase. In the example in (24b), however, there is no such
position created by successive cyclic wh-movement that satifisfies both Condition C and
the bound variable configuration. Thus, examples as in (24a, b) provide evidence for the
role of vP phase from the syntax-semantics interface. A similar argument can be made
for the phasehood of passive VPs. Consider examples in (25a, b) and (26a, b).
(25) The Interaction of Binding and Reconstruction in Passive Derivation (A-Movement)
a.. [At which of the parties that hei invited Maryj to] was every mani √ introduced to herj * ?
b.* [At which of the parties that hei invited Maryj to] was shej
introduced to every mani * ?
(Legate 2003: 507)
(26) The Interaction of Binding and Reconstruction in Unaccusative Derivation (A-Movement)
a. [At which conference where hei mispronounced the uninvited speakerj’s name]
did every organizeri’s embarrassment √
escape herj * ?
b. [At which conference where hei mispronounced the uninvited speaker’s namek] did
itk *
escape every organizeri entirely * ?
(Legate 2003: 508)
72
In the example in (25a), the copy left at the edge of the lower vP phase by the successive
cyclic movement of the complex wh-phrase meets two binding-theoretic requirements:
he is correctly c-commanded by everyone and Mary is not in a position to be bound
incorrectly by her. The example in (25b) is ungrammatical because there is no single
position that meets both requirements. The same story can be told for the examples in
(26a, b) that involve unaccusative verbs like escape meaning forget (Pesetsky 1995). In
(26a), the copy left in the specifier of the lower vP phase serves to provide the unique
appropriate position for the purposes of Condition C and variable binding. A comparable
position is not available in the example in (26b). Thus, the contrast between (25a)/(26a)
and (25b)/(26b) suggests that passive and unaccusative verbs also constitute strong
phases, contra Chomsky’s (2000, 2001, 2004) claim that only transitive and experiencer
verbs count as strong phases (cf. note 6).
Evidence of the second kind involves reflexes of local movement whose distribution
can only be accounted for by positing successive cyclic movement, as in complementizer
agreement in Irish (McCloskey 1979; 2001) and Chamorro (Chung 1982, 1994, 1998). I
show in the next subsection that the distribution of the active voice morphology in
Indonesian and Javanese adds to this type of evidence for the locality of movement
required by the PIC.
73
3.3. A Phase-Theoretic Analysis of the Active Voice Deletion in Indonesian and Javanese
Let us now come back to the central point of this chapter, the distribution of the active
voice morphology in Indonesian and Javanese, and see how the phase-based analysis
correctly derives Cole and Hermon’s (1998) generalization. Consider (11b) from
Javanese, repeated here as (27a), for illustration. The derivation for the example is given
in (27b).
(27)a. Sapai
who
Esti
(*nge)-sun
Esti
AV-kiss
ti?
(=11b)
‘Who did Esti kiss?’
b. [CP Sapa … [TP Esti [vP tEsti [v′ tsapa [v′ ng-v [VP sun tsapa]]]]]]
Spec-Head D-Feature Checking
In this derivation, the wh-phrase sapa ‘who’ undergoes successive cyclic movement through
the edge of the vP phase into the edge of the CP phase in a manner required by the PIC
reviewed above. The external argument Esti is base-generated in the specifier of vP and
undergoes A-movement into [Spec, TP], as typically assumed for languages such as English.
74
When the wh-phrase stops by in the edge of vP phase, Spec-Head D-feature checking occurs
between the moved phrase and its local little v, as shown in (27b). This syntactic agreement
is realized in the syntax-external phonological component as the deletion of the nasal prefix
ng- in Javanese. More specifically, this analysis claims that only the derivation in which ngis not present on the series of transitive verbs crossed by the movement of an NP converges
at the syntax-external interface. Note that this analysis is strikingly parallel to the analysis for
French participle agreement developed by Kayne (1989). The fact that the movement of
non-nominal phrases does not trigger meN-/ng- deletion also naturally falls out since those
phrases do not have D-features to be checked against their local v heads. The same analysis
holds for the movement-voice interaction in Javanese and Madurese.
It should be noted here that, to the extent that the current analysis is tenable, the facts
examined in this chapter provide direct support from the syntax-phonology interface for
the role of vP phase. This point bears emphasis for the following reason. For reasons
detailed above, the phase-bound derivational system as in Chomsky (2000, 2001, 2004,
2005) predicts that movement occurs at the edge of vP and CP. Evidence abounds in the
literature for successive cyclic movement via intermediate [Spec, CP]s. To mention but a
few, complementizer alternation in Irish and Chamorro (McCloskey 1979, 2001; Chung
1982, 1994, 1998), the complementizer agreement in several dialects of Dutch, German,
75
Frisian, and West Flemish (Haegeman 1992; Zwart 1993, 1997; Watanabe 2000), whcopying in languages such as Frisian, Dutch, Afrikaans, German, and Child English
(Felser 2004; Thornton 1995) as well as the existence of partial wh-movement in
languages such as German, Hungarian, and Indonesian (McDaniel 1989; Horvath 1997;
Saddy 1991; Cole and Hermon 1998, 2000; see also chapter 5) provide morphological
evidence that wh-phrases move through the edge of CP phases. However, evidence has
been somewhat hard to come by for comparable cyclic movement via the edge of vP,
though the WH-agreement in Chamorro (Chung 1982; 1994, 1998), Innu-aimûn
(Branigan and MacKenzie 2002), and Kilega (den Dikken 2001), and the obligatory loss
of tonal downstep in Kikuyu that is triggered by wh-movement (Clements et al. 1983;
Clements 1984; Sabel 2000) have been sometimes interpreted as possible arguments that
wh-movement occurs through the edge of vP phases; see Radford (2004), for example.
The pattern of the active voice deletion in Indonesian and Javanese discussed here,
however, precisely fills this empirical gap and provide clear morpho-syntactic evidence
that the movement of an NP stops at the edge of each vP phase that intervenes between
the launching and landing site of the movement. In this regard, the contribution to
linguistic theory from under-represented languages such as Indonesian and Javanese is
highly significant.
76
Note further that the proposed analysis provides further support for Legate’s (2003)
view that passive vPs are strong phases, contra Chomsky’s (2000, 2001, 2004). In the
derivation of passive and object preposing constructions, the active voice deletion is
triggered by the Spec-Head D-feature checking requirement between the A-moved NP
and its local little v. This means that A-movement also passes through the edge of vP.12
The proposed analysis of the active voice deletion in Indonesian and Javanese is related
to another phase-theoretic account recently proposed by Aldridge (in press). Drawing on
certain parallels between meN- clauses and Tagalog-type anti-passive constructions,
Aldridge proposes that the feature bundle inserted in v and spelled-out as meN- cannot
include an EPP feature, hence blocks objects from moving out of the VP due to the PIC.
To illustrate her analysis, consider the schematic derivation for the example in (4b) in (28).
12
Two notes are in order. First, following Guilfoyle et al. (1992), I assume that the passive voice marker di- in
Indonesian is a clitic-like element that needs to be incorporated into a verb. This assumption receives
morphological support from the fact that di- can be considered as an allomorph of the third person singular
pronoun dia (Guilfoyle et al. 1992; 400). Second, it is impossible to construct a parallel argument for the
phasehood of unaccusative verbs. This argument, in principle, could be made if Indonesian has transitive
unaccusatives like escape in the sense of ‘forget’ (see (26a, b)). However, my informant work so far has not
been successful in identifying the Indonesian analogue of verbs like escape. Unergatives verb such as nyanyi
‘sing’ may occur with or without meN- in both its transitive and intransitive use whereas other verbs such as
dansa ‘dance’ only occurs with the derivational affix ber-.
77
(28)
C [= phase head]
…
vP
v′
v [meN]
VP
Spell-Out/Transfer
no EPP V apa kepada Fatimah
Aldridge assumes that the v head spelled-out as meN- in Indonesian cannot have the EPP
feature. Recall that the PIC dictates that the object that is to be ultimately moved into the
higher CP phase must first move into the edge of the v. Since the v head in (28) does not
carry the EPP feature, the movement of the object into [Spec, CP] violates the PIC. The
example in (4b) is grammatical without meN- on the higher verb because the v head
spelled-out this way can carry the EPP feature that triggers the movement of the direct
object into the specifier of the vP in (28).
Two problems remain with Aldridge’s analysis, however. First, it is not clear what
prevents meN- from having an EPP feature. Under Chomsky’s (2000, 2001, 2004, 2005)
version of the EPP-assignment on phase heads, nothing blocks transitive v heads from
being assigned an EPP feature that would allow movement of an VP-internal element to
78
the its edge.13 Aldridge does develop a diachronic explanation for the lack of EPP features
on v heads in Indonesian by analyzing meN- as anti-passive morpheme that can be
considered a historical residue of the ergative system of the Philippine-type languages as in
Tagalog and Malagasy. However, this explanation could lead to the claim that Indonesian
is an ergative language. This reasoning might be undermined by the argument made by
Ndayiragije (2006) that Kirundi is a nominative-accusative language yet has antipassives
but it is at least a reasonable move given the observation that the antipassive voice is found
mostly in ergative languages. There are two arguments that Indonesian is
morphosyntactically not an ergative language. First, as we will see in section 4.1, Chung
(1976) and Cole and Hermon (2005) present evidence that direct object extraction is
possible in Indonesian. Given that the direct extraction of non-subject argument is
impossible in many of the ergative Austronesian languages such as Tagalog, Malagasy,
and Formosa, Chung’s discovery suggests that Indonesian is not an ergative language of
the Tagalog-type. Second, Chung (in press) claims that the derivation of the Indonesian
SVO word order is derived by simply raising the external argument to [Spec, TP], in much
the same way as the derivation of the English SVO order. Several researchers in
13
Chomsky (2000, 2001, 2004) claims that the EPP-assignment on the phase edge is done only when
movement has an effect on semantic interpretation at the interface. If the proposed analysis is correct, the EPP
assignment on the phase edge is motivated by PF effects because the EPP-checking results in the meNdeletion at the PF interface.
79
Indonesian as in Cartier (1989) and Verhaar (1989) argued that Indonesian is
morphosyntatically an ergative language on the grounds that both the regular di-passive
and the so-called zero-passive/Passive Type Two are available in this language (cf. (8); see
also section 4.1 for detailed discussion). This argument, Chung argues, is incompatible
with the observation confirmed by Anderson (1976: 18-19), Manning (1996), and other
researchers that morphosyntactic ergativity is found in verb-initial or verb-final languages,
but not in verb-medial languages. Thus, to the extent that the Indonesian is neither a verbinitial or verb-final language, as Chung claims, then this language is not a
morphosyntatically ergative language of the Tagalog-type, as proposed by Aldridge.
The second problem with Aldridge’s analysis is that we would predict that non-nominal
phrases such as prepositional and adjunct phrases moving from embedded to matrix
clauses should trigger deletion on the matrix verb, since they must also move through
intermediate phase edges as well in order to extract long distance, and hence must check
EPP features on such phase heads. This prediction is false since, as we saw earlier, the
movement of these phrases does not trigger meN-deletion in this context. This suggests
that it is not an EPP feature but a D-feature that is responsible for this deletion.14
14
This argument was suggested to me by Heidi Harley (personal communication).
80
One important question that arises in this connection is the following.15 Under the
present analysis, it is the D-feature checking relation between the specifier and the head
of the vP that triggers active voice deletion on the verb dominated by the vP. If so, why is
it that the merger of external arguments such as Esti in (11b) does not cause the deletion
of the active voice marker? An answer to this question is available once we introduce the
distinction between External Merge and Internal Merge. Chomsky (2004) proposes that the
concatenative operation of Merge comes free in two forms; when X is merged with Y, X can
be either external to Y (External Merge) or part of Y (Internal Merge). Chomsky (2004) (see
also Richards 2007) argues that pretty much everything is driven by phase heads other than
External Merge, which is triggered solely by selectional features. For example, Chomsky
(2004: 111) maintains that “Argument structure is associated with external Merge (base
structure), everything else with internal Merge (derived structure).” It is clear that whatever
feature that causes meN-/ng-deletion does not belong to argument structural/θ-theoretic
features but instead forms some formal computational feature that is triggered by phase heads
such as v heads. Indeed, we have seen evidence above that the deletion is sensitive to Internal
Merge of an NP across a transitive verb. Also recall from Kayne’s (1989) analysis that French
15
Thanks to Heidi Harley (personal communication) and Simin Karimi (personal communication) for raising this
question and providing possible directions to take with respect to this question such as the External vs. Internal
Merge distinction.
81
participle agreement is sensitive to Internal Merge, as the contrast between (18a) and (18b)
illustrates. Then, the fact that the External Merge of an external argument does not cause active
voice deletion in examples like (11a) follows if only Internal Merge has consequences such as
the active voice deletion for the sound-related syntax-external interpretive component. This
assumption is also in keeping with the “complementarity between θ-theory and formal feature
checking” that has been expressed in one way or another in the minimalist literature (e.g.,
Chomsky 1995), namely, that all arguments must form non-trivial chains by movement to
checking their formal features with an appropriate functional category such as T or v (see
López 2001, though, for an opposing argument from exceptional case-marking subjects).
The assumption that the θ-requirement and formal feature checking is dissociated this
way is also conceptually natural in view of the following observation. The v is considered
both lexical and functional at the same time, lexical in that it selects an external argument
in its specifier and functional in that it assigns accusative Case to the direct object. These
two roles were distributed by AgrO and the V within the Agr-based Case Theory but are
now served by the v in the more recent Agr-less vP-approach. This claim, in turn, has lead
to controversies on whether the A vs. A′-distinction can be established in syntax. Thus, we
need some principled way to encode this dual behavior of the v head. The distinction
between Internal Merge and External Merge seems to be able to make the desired
82
distinction; the v head assigns configurational properties such as Case and Agreement to
only an Internally Merged object but, at the same time, assigns a θ-theoretic interpretation
to an Externally Merged object based on its configuration (Hale and Keyser 1993, 2002).
Given that active voice deletion belongs to the former type of feature-driven computation,
it stands to reason that the External Merge of an external argument as in (11a) does not
have any consequence for the fate of the active voice morphology at the phonological
interface.
It is in this regard that the proposed analysis is crucially different from the most recent
analysis presented by Cole and Hermon (in press). Cole and Hermon propose an account of
the distribution of the active voice morphology in Malay/Indonesian that assimilates the
active voice deletion pattern to the more general Philippine-type voice agreement
discussed in Rackowski and Richards (2005). They propose (p. 15) that “the presence of
meng- indicates that the agent is the highest specifier of vP, and, hence, that object shift has
not occurred.” Conversely, then, the lack of meng- indicates that the non-external
argument is moved to the edge of vP phase to be accessible to movement into [Spec, CP].
Thus, the impossibility of the movement of a non-external argument across the meN-verb
is naturally predicted because the presence of meN- means the lack of the object shift of a
post-verbal argument into the edge of vP, thereby blocking its movement across meN- due
83
to the PIC. This line of approach, in fact, has been already proposed in a slighty different
form by Guilfoyle et al. (1992: 385-387) within the Government-and-Binding Theory.
Assuming that meN- and di- are Actor Topic and Theme Topic, respectively, Guilfoyle et
al. argue that the impossibility of extraction of an object across meN- is a natural
consequence of the general fact observed in many Austronesian languages such as Tagalog
and Malagasy that only the NP which agrees with the verb which has the appropriate
topic/voice morphology is extractable.
Cole and Hermon further note (p.15) that the agreement-based analysis makes a correct
prediction that the extraction of an NP across the higher verb, not the lower verb, in a biclausal environment, as in (29) (=4a), triggers the meN-deletion only from the higher verb.
(29) Siapai Bill (*mem)-beritahu ibu-nya [CP yang ti (men)-cintai Fatimah]? (=4a)
who
Bill
AV-tell
mother-his that
AV-love
Fatimah
‘Who does Bill tell his mother that loves Fatimah?’
According to Cole and Hermon, this deletion pattern is explained as follows: the deletion
of the meN-(or the null prefix ØmeN- in their terminology) signals the agreement between
the higher v head and the CP clause out of which the movement has occurred.
84
Cole and Hermon’s agreement-based analysis appears to make the exact same predictions
concerning the distribution of the active voice morphology as the proposed D-feature
checking analysis. Three considerations suggest, however, that the present analysis is to be
preferred on both theoretical and empirical grounds. One difference between the two
analyses lies in their view that meN- indicates agreement between the v and the external
argument base-generated in the specifier of the v head. Under the complementarity between
θ-theory and formal feature checking noted above, their analysis would amount to the claim
that this complementarity is broken down in Indonesian. Unless evidence is presented for
this view, however, it is at least theoretically desirable to eliminate this extra stipulation, as in
the proposed D-feature checking analysis. Another difference lies in their treatment of the
pattern of meN-deletion in long-distance extraction as illustrated in (29). As we have seen,
Cole and Hermon claim that the null prefix occurs on the higher verb due to the agreement
between the upstairs v and its complement CP. However, they do not make it explicit in what
formal feature they Agree with each other. Let us then consider what feature would be
relevant for their analysis to go through. The possibility that the feature checked is Case is
unlikely because verbs like beritahu ‘tell’ in (29), like their English counterparts, can only
assign one accusative Case to its immediately following Goal DP. Another possibility that
the feature checked is the nominal D-feature also cannot be maintained because CPs as in
85
(29) that occupy non-subject positions have been standardly assumed to lack D-features.
Note that this kind of problem does not arise under the proposed analysis. The proposed
analysis explicitly states that the higher meN- in (29) is elided due to the D-feature checking
relation between the upstairs v and the successively moved NP. The final difference, related
to the second, is the following.16 Cole and Hermon argue that it is the agreement of the
complement CP with the matrix verb in (29) that triggers meN-deletion. Their analysis
predicts that matrix meN-deletion should occur on the matrix verb in the configuration in (29)
even when a non-DP argument is extracted from an embedded CP. This is because the
syntactic category of a moving element should have nothing to affect the agreement between
the verb and its complement CP. This prediction, however, is incorrect, as shown in (30),
which minimally contrasts with (29), in that what moves in the former is a non-nominal
(prepositional) expression, because meN- on the matrix verb does not undergo obligatory
deletion in (30), contrary to what Cole and Hermon’s analysis predicts.
(30) [PP Kepada siapa]i Bill (mem)-beritahu ibu-nya [CP Fatimah (mem)-beri buku ti]
to
whom Bill AV-tell
mother-his Fatimah AV-buy
‘To whom did Bill tell his mother that Fatimah gave a book?’
16
Many thanks to Heidi Harley (personal communication) for suggesting this argument to me.
book
86
On the other hand, the obligatory deletion of meN- in (30) naturally falls out from the
present analysis because non-nominal expressions such as kepada siapa ‘to whom’ do
not have D-features. On the above-mentioned grounds, I conclude that, although Cole
and Hermon’s analysis does seem to make essentially the same set of predictions as the
D-feature checking analysis concerning Indonesian facts, the latter analysis is preferred
on both conceptual and empirical grounds.
3.4. meN-/ng-Deletion as Failure of Vocabulary Insertion 17
So far, I have remained vague on the precise nature of “deletion” involved in the present
phase-theoretic analysis of the active voice deletion in Indonesian/Javanese. The issue of
exactly what “deletion” means in contemporary syntactic theory has been far from settled. In
this section, following Harley (2005), I propose that the active voice deletion is failure of the
post-syntactic insertion of meN-/ng- at the PF interface within the morphosyntactic framework
of Distributed Morphology.
It is claimed within the Distributed Morphology that phonological features are
assigned to abstract morphemes (as well as roots under certain views) post-syntactically.
17
Many thanks to Heidi Harley (personal communication), Mark Baker (personal communication), and
Andrew Carnie (personal communication) for encouraging me to formulate a rule for the deletion and very
useful discussion. See detailed explication of theoretical claims of Distributed Morphology in chapter 6.
87
The mechanism to assign this feature is Vocabulary Insertion in the technical sense; “the
Vocabulary is the list of the phonological exponents of the different abstract morphemes of
the language, paired with conditions on insertion. Each such paring of a phonological
exponent with information about the grammatical (i.e. syntactic and morphological)
context in which the exponent is inserted is called a vocabulary item.” (Embick and Noyer
2007: 297). A list of potential candidates compete for the same abstract morpheme
position in the morphosynatctic derivation; which candidate is selected as the exponent of
that position is determined by the Subset Principle proposed by Halle (1997).
I propose that the Spec-Head D-feature checking between the moved NP and its local v
head changes the D-feature of the v and that this change in the feature content of the head
drives meN/ng- out of the candidate set for insertion under the v head at PF. Suppose for the
sake of argument that the v head has the bundle of c features consisting of [+F(unctional)],
[+Cause/Do/Have], and [+D-(feature)] and that the meN- and its null variant ØmeN- is inserted
under the condition specified in (31).
88
(31) a.
vP
NP
Checking
b.
v′
v
VP
{[F] , [+Cause], [+D]} V
Partial List of the Vocabulary Items
i. meN
← → [v
[+D]]
ii. ØmeN-
← → [v
[…]]
tNP
Under my proposal, the checking relation illustrated in (31a) deletes the [+D] feature of
the v head. Since meN- can only be inserted under the v head with the corresponding
feature, as shown in (31bi), meN-insertion is correctly blocked; then, the null counterpart
ØmeN- is inserted as an elsewhere case, as shown in (31bii). This proposal, thus, sheds
new lights on the precise nature of active voice deletion in Indonesian/Javanese and of
ellipsis processes, more generally.
3.5. Section Summary
To sum up this section, I have proposed, following the insight of Kayne’s (1989)
analysis of participle agreement in French, that Cole and Hermon’s generalization is
naturally derived by analyzing the active voice deletion as the PF reflex of a Spec-Head
Agreement relation between the moved NP and its local v head at the vP phase. The
movement of non-nominal phrases does not cause active voice deletion because they lack
89
D-features to be checked in this configuration. The data examined here, thus, provides
empirical support for the phase-based implementation of successive cyclicity, in particular,
the role of the vP phase at the syntax-phonology interface.
4. Other Alternative Accounts
In this section, I compare the proposed phase-theoretic analysis with several existing and
potential analyses of the meN- deletion in Indonesian and ng-deletion in Javanese, by extension:
a) the NP Accessibility Hierarchy and the Subjects-Only Restriction b) Soh’s (1998)
Relativized Minimality account, c) Fortin’s (in press) antipassive account, d) Anti-Agreementbased accounts, and e) Case-based accounts. I show that each of these analyses encounters a
particular set of empirical and conceptual problems that are successfully resolved in the present
phase-theoretic approach.
4.1. NP Accessibility and Subject-Only Restriction
In this subsection, I review and reject a potential analysis of Cole and Hermon’s
generalization based on the so-called Accessibility Hierarchy. 18 Based on a large-scale
survey of many western Malayo-Polynesian languages, Keenan and Comrie (1977: 70)
18
Thanks to Paul Kroeger (personal communication) for suggesting this possible account.
90
(see also Keenan 1972) observe that only subjects can be relativized. This observation is
also made in the one of the most comprehensive grammars of Indonesian in Snedon (1996),
who notes (p. 286) that “a relative clause can contain any constituent occurring in an
independent clause except the subject, which is identical to the head of the embeddeding
noun phrase.” The accessibility-based analysis is further articulated in the minimalist
framework by Nakamura (1994), which derives this subject-only restriction in terms of the
global economy principle (“minimize chain links” in the sense of Chomsky and Lasnik
1993) that essentially forces non-subject arguments to be first promoted to a subject/topic
position for the purposes of creating the shortest possible link in a set of alternative
competing derivations. Under this analysis, examples that involve extraction of an
apparent non-subject argument all involve passivization of this argument into [Spec, TP]
prior to movement to [Spec, CP]; the active voice morphology, then, is deleted as a natural
result of the passivization.
This analysis seems desirable in two respects. First, it fits the apparent mysterious
restriction on the active voice deletion in Indonesian and Javanese into the independently
motivated restriction or whatever account derives such a restriction. Second, this analysis
appears to be particularly true for Indonesian because this language has what several
Indonesianists have variously called “Passive Type Two,” “Object Preposing”, “Object
91
Voice,” “Subjective Passive”, or “Bare Passive” (Sneddon 1996; see also Chung 1976,
Kana 1986; Guilfoyle et al. 1992; MacDonald and Dardjowidjojo 1967; Sie 1988; Cole
and Hermon 2005, Arka and Manning 1998), in which a bare stem is immediately
followed by a subject, preferably a pronominal subject, as shown in (32b) (recall the
examples in (8a, b), whose derivation also instantiates the Passive Type Two).19
(32) Active and Passive Type Two Sentences
a.
Aku sudah
mem-beli
buku
itu.
I
AV-read
book
that
already
‘I already read that book.’
b.
Buku
itu
sudah
Ø-beli
aku/ku.
book
that
already
OV-buy me
‘I already read that book./That book was already read by me.’
If the subject-only restriction is correct for Indonesian and Javanese, then all the examples
that involve meN-deletion may be analyzed as passivization of the non-subject argument
19
A construction similar to the Passive Type Two in Indonesian is found in Malagasy and Tagalog with the
Theme-Topic marker. See Rackowski and Richards (2005) for Tagalog and Pearson (2001, 2005) for
Malagasy.
92
into [Spec, TP], followed by A′-movement into [Spec, CP]. However, there are many
arguments presented in the literature that this analysis is untenable. I review several of
those arguments below.
4.1.1. Indonesian
The first argument against the accessibility-based analysis comes from its prediction that all
A′-extractions should pass through [Spec, TP], which was shown to be is false in
Malay/Indonesian by Soh (1998) on the basis of the relative position of subjects with respect
to aspectual markers, weak crossover effects, long-distance extraction and the that-trace
asymmetry. To repeat one of her arguments based on weak crossover effects, it has been
widely known that a) a pronoun must be c-commanded by a binder and its variable at the
surface/derived structure to be construed as a bound variable, as shown by the contrast
between (33a) and (33b) and that b) the trace created by A-movement provides a new binder
for a pronoun, as shown in (33c) (Mahajan 1990: 24).
93
(33) Weak Crossover Effects in English
a.
Whoi ti saw hisi mother?
b.* Whoi did hisi mother see ti?
c.
[Whoi ti seems to his mother [ti to have come]].
(Soh 1998: 300)
Then, the prediction of the passivization-based account is that extraction of a non-subject
argument should not cause a weak crossover effect, as in (33c), but, as the ungrammaticality of
(34b) shows, the bare passive counterpart to (34a), shows that this prediction is wrong; note
that, in a ‘real’ passive of the English kind in Indonesian given in (34c), we don’t have weak
crossover effects, a pattern that suggests that the movement of siapa-kah ‘who-Q’ is preceded
by the prior movement to [Spec, TP].
(34) Weakcrover effects in Malay/Indonesian
a.
Emak-nyai
sayang Alii.
mother-his
love
‘His mother loves Ali.’
Ali
94
b.* Siapai-kah yang
who-Q
that
emak-nyai sayang ti?
mother-his love
‘Who does his mother love?’
c.
Siapai-kah yang
who-Q
that
ti di-sayangi
OV-love
‘Who is loved by his mother?’
emak-nyai ti?
mother-his
(Soh 1998: 300)
The presence of the weak crossover effect in the example in (34b) and the lack thereof in
the example in (34c) indicate that the subject-only restriction cannot be the proper account
for the active voice deletion in Indonesian.
The second argument against the accessibility-based account is that made by Cole and
Hermon (2005); see also Musgrave (2001). Following Chung (1976), Cole and Hermon
provide numerous arguments that the distributional restrictions on the Passive Type Two do
not hold for object relativization in Indonesian, contrary to what the Accessibility-based
analysis would lead us to predict. In addition to the two properties noted above, in the Passive
Type Two, the negative marker tidak and auxiliaries must precede the agent. This is illustrated
by the contrast between (35a) and (35b).
95
(35) Negation and Auxiliaries Precede the Agent in Passive Type Two
a.
Buku ini tidak
akan
kami
Ø-baca.
book
will
we
OV-read
this not
‘This book will not be read by us.’
b.*
Buku ini kami
tidak
akan
Ø-baca.
book
not
will
OV-read
this me
‘This book will not be read by us.’
(Cole and Hermon 2005: 62)
These characteristics of the Passive Type Two, thus, restrict the possible word orders to the
[Neg Aux Agent-Verb]. Cole and Hermon show that this word order restriction does not
hold for object relativization. If object relativization were done through Passive Type Two,
followed by null operator movement of the derived subject, as the NP-Accessibility
analysis would argue, then the agent must follow negatives and auxiliaries in cases of
object relativization since the input configuration for this relativization should be Passive
Type Two. This prediction is false because object relativization allows agents to precede
negation and auxiliaries, as shown in (36a, b).
96
(36) Object Relativization in Indonesian
a.
[Buku
[yang Budi
tidak
akan
baca]]
sangat
menarik.
book
that
will
not
read
very
interesting
Budi
‘The book that Budi will not read is very interesting.’
b.
Anak [yang Wati
tidak
pukuki]] itu men-angis.
child
not
hit
that
Wati
that AV-cry
‘The child that Wati didn’t hit is crying.’
(Cole and Hermon 2005: 64)
This order is unacceptable in the Passive Type Two, as in (37a) and (38a). By contrast, the
active counterparts to (37a) and (37a) in (39a, b) are grammatical. Note further that the
relativization based on the Passive Type Two construction in (35a) is, of course,
grammatical, as in (40).
(37) Passive Type Two and Active Sentences in Indonesian
a.* Buku itu dia tidak
book
that he not
akan
baca.
will
read
‘The book will not be read by him.’
b. Buku itu tidak
book
that not
akan
dia baca.
will
he read
‘The book will not be read by him.’
(Cole and Hermon 2005: 64)
97
(38) Passive Type Two and Active Sentences in Indonesian
a. * Anak itu kami
child that you
tidak sedang jemput. b. Anak itu sedang kami
jemput.
not
pick up
Prog
pick up
‘The child is not being picked up by us.’
child that Prog
you
‘The child is not being picked up by us.’
(Cole and Hermon 2005: 64)
(39) Active Counterparts to the Examples in (36a) and (37a) in Indonesian
a. Dia tidak akan mem-baca buku
he not
will AV-read
book
itu.
that
‘He will not read that book.’
b. Kami tidak
you
not
sendag mem-jemput anak itu.
Prog
AV-pick up child that
‘We are not picking up that child.’
(Cole and Hermon 2005: 64)
(40) The Example of Object Relativization Based on the Passive Type Two
Buku [yang tidak
akan
kami Ø-baca] sangat
menarik.
book
will
me
interesting
that Neg
read very
‘The book that I won’t read is very interesting.’
98
The data examined above, therefore, provides evidence that non-subject relativization can
occur without prior promotion/movement to [Spec, TP], hence argues strongly against the
account of meN-deletion in terms of the NP-Accessibility. I refer the reader to Chung (1976),
Cole and Hermon (2005) and Musgrave (2001) for numerous other arguments along these
lines against such an analysis.
Finally, the Accessibility Hierarchy encodes the universal statement that all human languages
allow at least relativization of a subject. Specifically, Keenan and Comrie (1977: 67) make such a
statement in the form of the Hierarchy Constraint “A language must be able to relativize
subjects.” However, the Accessibility-based account by itself predicts that not only local
extraction but also long-distance extraction of a subject should be acceptable with or without
meN on its extraction path because subject relativization is available in all languages. We have
seen above that examples such as (4a, b), (5a, b), and (6a-c) clearly falsify this prediction. This
indicates that the distribution of the active voice morphology in Indonesian is something beyond
Keenan’s (1972)/Keenan and Comrie’s (1977) Accessibility Hierarchy.
4.1.2. Javanese
A similar argument against the NP Accessibility-based account for the active voice
morphology in Javanese was made by Cole et al. (1999). Examples as in (41a, b) show that
99
relativization of an object results in the deletion of the active voice marker ng-. The analysis
based on accessibility would explain this deletion as the result of the passivization that applied
prior to operator movement.
(41) Object Relativization
a. * Buku kuwi
book
[DP [CP sing Budi *m-waca
that
that Budi AV-read
‘That book is the one that Budi read.’
b.
Buku kuwi
book
[DP [CP sing Budi maca]]
that
that Budi read
‘That book is the one that Budi read.’
(Cole et al. 1999: 88)
This analysis, thus, makes a prediction that Javanese should also have a construction akin
to Passive Type Two, as found in Indonesian, a prediction that is incorrect. As Cole et al.
(1996: 92) observe, the bare passive construction does not occur in Javanese. Thus, the
NP-accessibility analysis would wrongly predict that examples as in (42a, b) are
ungrammatical, contrary to facts.
100
(42) Object Relativization in Javanese
a.
Bukune
sing aku waca lucu.
book-ne that I
read
b.
funny
‘The book that I read is funny.’
Montor sing Siti
setir abang.
car
drive read
that Siti
‘The car that Siti drives is red.’
(Cole et al. 1999: 92)
The grammaticality of object relativization in (42a, b), thus, argues against the NP
Accessibility-based account of the ng-deletion.
4.2. Soh’s (1998) Relativized Minimality Account
Soh (1998) provides a new account of the syntactically governed distribution of meN- in
Malay/Indonesian in terms of Relativized Minimality (Rizzi 1990). According to this
account, the active voice marker meN-, base-generated in the A′-position within the VP,
blocks A′-movement of a phrase across the marker, in violation of the minimality
constraint defined as in (43) and (44).
101
(43) Relativized Minimality
Antecedent Government: X antecedent-governs Y iff
(i) X and Y are coindexed
(ii) X c-commands Y
(iii) no barrier intervenes
(iv) Relativized Minimality is respect
(Rizzi 1990: 6-7)
(44) Relativized Minimality: X α-governs Y only if there is no Z such that
(i) Z is a typical potential α-governor for Y,
(ii) Z c-commands Y and does not c-command X.
(Rizzi 1990: 6-7)
To take (4a) for illustration, repeated here as (45), this example is ruled out with the active
voice marker on the higher verb because meN- blocks A′-extraction of the wh-phrase siapa
‘who’ across it, in violation of the Relativized Minimality defined as in (44).
(45) Siapai Bill (*mem)-beritahu ibu-nya
who
Bill
AV-tell
mother-his
[yang ti (men)-cintai
Fatimah]?
that
Fatimah
‘Who does Bill tell his mother that loves Fatimah?
AV-love
102
We have seen in section 2 that meN- does not block movement of non-nominal phrases
such as prepositional objects and locative adjuncts across the verb marked with it, as
shown in (46) (recall also the examples in (10a, b) and (13a, b)).
(46) Movement of non-Wh-NPs
[PP kepada siapa]i Bill (mem-)beritahu ibu-nya
to
who Bill AV-tell
[bahwa Esti (mem-)beri motor
mother-his that
Esti AV-give
ti]?
motor bike
‘To whom did Bill tell his mother that Esti gave a motor bike?’
Soh (p. 304) claims that this NP-PP asymmetry is accounted for if Minimality is
relativized not only in terms of position (A vs. A′) but also category (NP or PP). Then, a
position such as the base-generated position of meN- is restricted to NPs, thus meN- only
blocks extraction of an NP, not PP, over the voice marker.
Soh’s Relativized Minimality account is quite close to the proposed D-feature checking
analysis of the active voice deletion. The two analyses essentially block movement of an NP
across the active voice verb from independently motivated constraints on syntactic
derivation. However, two problems remain unresolved in Soh’s analysis. First, Soh assumes
that meN- is in an A′-position. This analysis, thus, predicts that only A′-movement of an NP
103
across the verb with the active voice morphology is prohibited. We have seen, however, that
A-movement such as object preposing also causes the active voice deletion of a verb that it
crosses, as illustrated in (8a, b), repeated here.
(47) Object Preposing
a. Buku itui
adik
saya (*mem)-beli ti.
book that brother my
AV-buy
‘My brother bought that book./That book was bought by my brother.’
b. Alii saya (*men)-cubit
Ali
I
ti.
AV-pinch
‘I pinched Ali./Ali was pinched by me.’
Thus, Soh’s analysis falsely predicts that examples like (47b) should be grammatical with
meN- marked on the verb that the movement has crossed because the movement involved
is an A-movement that thereby should not blocked by an A′-element like meN-. On the
other hand, as we have shown in section 3, the phase-theoretic D-feature checking account
correctly predicts that whatever movement is involved across an active verb results in the
deletion of meN- across it. Another problem with Soh’s analysis concerns the base-position
104
of meN-. She assumes (p. 297) that “meN- occupies a position above the verb and below
the internal subject position,’ but this assumption is difficult to sustain. Within her
Relativized Minimality account, meN- should be base-generated in an A′-position to block
A′-movement of an NP across the active voice marker; the blocking effect would not arise
if it were base-generated under V because a head position is not normally characterized as
an A vs. A′ position (cf. Li 1990). Updating Soh’s remark above within the more recent
split vP model, the relevant position would arguably correspond to a lower specifier of the
vP. However, the assumption that voice markers such as meN- are base-generated in a
specifier of a functional head (be it v or Voice) is not a commonly held assumption and
requires independent empirical motivation.20 Based on these grounds, I conclude that the
present phase-theoretic analysis is empirically superior to Soh’s Relativized Minimality
account.
4.3. Voskuil’s (2000) Pro-Based Account
Voskuil (2000) attempts to account for the distribution of the active voice morphology in
Indonesian from the interaction of several constraints and principles. Voskuil assumes that
20
True, this assumption is not utterly unreasonable, as long as Fortin’s (in press) argument that meN- in
Indonesian is an object clitic holds. However, I provide evidence in section 4.4 that this treatment of meN- is
problematic.
105
meN- licenses pro in its object position, as stated in (48), an assumption that he claims to be
independently supported by facts concerning left dislocation, topicalization and binding
(Baker 1996).21
(48) pro is licensed if governed by a verb prefixed with meN- or di-. (Voskuil 2000: 206)
He further adopts the assumption in (49), which, combined with the previous one, yields
the result that the object of the verb marked with meN- is always occupied by pro.
(49) Pronoun Principle
An empty category is interpreted as pro if morphological licensing conditions for
pro are met, irrespective of derivational history. (Voskuil 2000: 206)
Working with this set of assumptions, he proposes that the A′-extraction of an NP across
the meN-verb is blocked by the constraint on variables stated in (50). This constraint
essentially states that, when syntactic movement does not violate island constraints, the
operator-variable relation must be established by movement leaving a trace instead of pro.
21
However, this assumption is incorrect because direct objects can occur with meN-, as many of the examples
in this chapter have shown.
106
(50) Constraints on Variables
A variable-operator pair at LF must be derived through movement, unless movement is
prohibited by conditions pertaining to the structural distance between the variable and the
operator. (Voskuil 2000: 206)
According to this proposal, the extraction of an NP across a verb with meN- in a monoclausal context is blocked in the following way. The presence of meN- on the verb
requires that the pro is generated in the object position of the verb. However, since the
syntactic movement involved in such a derivation does not violate any island constraints,
the operator-variable relation must be established by movement rather than a resumptivelike strategy in Indonesian, leaving a trace, not a pro, in the direct object position. The
clash of these two requirements gives rise to ungrammaticality. When meN- is absent,
however, no such clash arises, because the tail of the movement does not have to be a
pro. The NP-only property may well be derived naturally because the only antecedent for
pro is an NP operator.
Voskuil’s analysis does seem to account for the restriction on the movement of an NP
across the meN-verb but it is not clear at all whether the three constraints he devised to derive
this restriction are independently motivated in Indonesian. Granted that these constraints are
107
indeed motivated, there are two serious problems with his account. One problem is that, as is
clear from his generalization in (51), Voskuil’s analysis is designed to block only A′-extraction
of an NP across the meN-verb.
(51) meN- exclusively blocks A-bar movement of the direct object, di- likewise blocks Abar movement of the genitive agent. (Voskuil 2000: 206)
Voskuil’s analysis, therefore, predicts that A-extraction of an NP across such a verb should
be possible acceptable, contrary to facts. The other problem comes from one of his
assumptions, made elsewhere in Voskuil (2000: sec.3), that LF movement also triggers
meN-deletion. Saddy (1991) and Cole and Hermon (1998, 2000) provide evidence that LF
movement in Indonesian is subject to subjacency constraints; see chapter 5 of this
dissertation for detailed discussion on wh-movement in Indonesian. An illustrative
example from Indonesian to make this point is in (52).
(52) * Kamu kira (bahwa)[DP cerita bahwa siapai yang ti men-geritik Jon itu] di-jual.
you
think that
story that
who that
AV-criticize John the Pass-sell
‘Who do you think that the story that criticized Jon was sold.” (Saddy 1991: 195)
108
Both Saddy and Cole and Hermon show that LF movement does not cause meN-deletion
from the verbs, as shown in examples like (6b), repeated here as (53).
(53) Ali
Ali
(mem)-beritahu
kamu tadi
[apai yang Fatimah (*mem)-baca ti]?
AV-told
you just now what that Fatimah
AV-read
‘What did Ali tell you just now that Fatimah was reading?’
Voskuil’s assumption above, thus, incorrectly predicts that the higher verb beritahu ‘tell’
should have its active voice deleted. Therefore, the ban on A-movement of an NP across meNverbs as well as the interaction of meN-deletion with partial wh-movement suggest that his
analysis does not adequately capture the distribution of the active voice morphology in
Indonesian and in Javanese, by extension.
Note that the fact that the “covert” movement does not cause meN-/ng-deletion in
Indonesian/Javanese provides support for the D-feature checking analysis of this deletion.
Under the recent Agree-based framework within the Minimalist Program (see Chomsky
2000, 2001, 2004, 2005), there is no LF movement as conceived of in the traditional Tmodel of grammar. Chomsky proposes instead that overt movement is the byproduct of
Agree + EPP. The data in (53) shows that Agree is dissociated from EPP-feature checking,
109
for otherwise we would make the wrong prediction that the covert movement involved in
this example should be able to cause meN-deletion in the matrix clause. Thus, since
Aldridge’s (in press) EPP-analysis has been shown to be problematic, the active voice
deletion pattern observed in (53) provides indirect support that what is involved in the active
voice deletion in Indonesian/Javanese is D-feature checking.22
4.4. Fortin’s (in press) Anti-Passive Account
Fortin (in press) attempts to derive the syntactically governed distribution of the active
voice marker from the assumption that meN- is an object-clitic antipassivizing morpheme
(Baker 1988); also recall Aldridge’s (in press) treatment of meN-as anti-passive morpheme.
Fortin’s argument for the analysis of meN- as the object clitic, not merely the active
voice/transitivity marker, comes from her observation that meN- is obligatory when verbs
like baca ‘read’ are used intransitively, namely, without any overt syntactic arguments, as
shown by the contrast between (2a) and (2b), repeated here as (54a) and (54b), respectively.
The contrast here naturally follows if the prefix meN- itself is base-generated as the
internal argument of the verb baca ‘read’ and its absence leads to the violation of the
Theta-Criterion.
22
My thanks to Heidi Harley (personal communication) for suggesting this line of argument for the proposed
analysis.
110
(54) a.
Ali
sedang mem-baca.
Ali
Prog
b. * Ali
AV-read
‘Ali is reading.’
Ali
sedang baca.
Prog
read
‘Ali is reading.’
We have seen in the beginning of section 2, however, that this contrast is equally expected under
the analysis of meN- as the active voice whose appearance is affected by syntactic transitivity.
Thus, this contrast itself is mute on whether the morpheme is an example of object-clitic.
According to Fortin’s analysis, for clauses with meN-, the prefix originates in the θ-position
for the internal argument and overly moves to the verb. As is the case of bona fide antipassive
constructions generally, such clauses allow an optionally specified object. In this case, the
specified object is an oblique adjunct in a VP-adjoined position. In clauses without meN-, by
contrast, the object NP is a direct object that is base-generated into the θ-position for the internal
argument. The two derivations for the sentence in (55), then, will be as in (56) and (57).
(55)
Apai
yang
kamu (*mem)-beli
what
that
you
‘What did you buy?’
ti?
AV-buy
(Fortin in press)
111
(56) The Derivation of the Example in (54) with -meN
CP
C′
C
TP
…
vP
D
kamu
v′
v
VP
VP
V
meN-
PP
meN-
apa
V
beli
(violation of the Adjunct Island Condition)
(slightly modified from Fortin in press)
112
(57) The Derivation of the Example in (55) without -meN
CP
D
C′
C
TP
…
vP
D
kamu
v′
v
VP
V
D
apa
(slightly modified from Fortin in press)
Since meN- is present as in the derivation in (56), the prefix is base-generated in the
object position of the verb beli ‘buy’ and later moves to the V position. The specified
object, in turn, is base-generated as an oblique adjunct in the VP-adjoined position. If
meN- is absent in the same clause as in the derivation in (57), apa ‘what’ occupies the
object position of the verb. Fortin argues that the inability of an NP to move across a
verb with meN- in the derivation in (56) is derived by the Adjunct Island Condition
113
because the movement of the specified object involves extraction from within an
adjunct phrase. By contrast, there is no problem with the derivation of the meN-less
clause shown in (57) with respect to the movement of apa ‘what’ from the thematic
position to the specifier of CP.
Fortin’s antipassive account derives the incompatibility of the movement of an NP with
the active voice marker from the independently motivated syntactic constraint, the Adjunct
Condition. Three considerations indicate, however, that this analysis is inadequate. Firstly,
Fortin simply assumes that, in the derivation for meN-clauses in (56), Case on the adjunct
NP is checked against a null preposition, akin to the null preposition that checks the Case
of the agent NP in di-passives. Examples of this type of passive are given in (58).
(58) Piring
dish
itu
sudah
di-cuci
that already PV-wash
oleh/Ø
Pak
Ali.
by
Mister
Ali
‘These dishes were already washed by Pak Ali.’
(slightly modified from Fortin in press)
The crucial assumption for Fortin’s analysis to work is then that the oblique adjunct whphrase in meN-questions would have to have a preposition. However, it is not clear
114
whether we have independent evidence that the wh-phrase is contained within the PP in the
derivation of meN-clauses as in (56). At any rate, evidence from Heavy NP Shift in
Indonesian as in (59a), modeled after the Malay examples given in Soh (1998: 302),
indicates that the assumption that object DPs in meN-marked phrases are PPs seems hard
to sustain.
(59) Heavy NP Shift in Indonesian
a. Saya meny-epak [NP bola yang dia campak kepada saya dengan kuat]
I
AV-kick
ball that s/he throw to
me with
kemarin.
strength yesterday
‘I kicked the ball that s/he threw to me very hard yesterday.’
b. Saya meny-epakti kemarin [NP bola yang dia campak kepada saya dengan kuat]i .
I
AV-kick
yesterday
ball that
s/he throw to
me with
strength
‘I kicked yesterday the ball that s/he threw to me very hard.’
Under the standard analysis of Heavy NP Shift as a case of A′-movement, the example in
(59a) involves rightward movement of the phonologically heavy NP into a vP-adjoined
position. That the rightward movement has occurred can be seen from the relative order of
the heavy NP with respect to the adverb kemarin ‘yesterday’. Fortin’s analysis incorrectly
115
predicts that this sentence should be ungrammatical with meN- because the A′-movement
of the heavy NP here would also be blocked by the Adjunct Condition on a par with meNsentences, as in (56). However, this problem does not arise under the proposed phasetheoretic analysis. The heavy NP moves rightward into the adjoined position, not the
specifier position of the vP. This movement does not cause the mandatory active voice
deletion because it does not bring the moved NP into the Spec-Head D-feature checking
relation in [Spec, vP].
Secondly, a question remains as to why the whole PP, not the specified object NP,
cannot move into the specifier of CP in the derivation in (56). If this movement were
possible, then meN-deletion would not be obligatorily caused in this derivation. However,
(10b) and (46) show that a PP can be fronted in Indonesian. Fortin’s analysis, thus, does
not seem to go unless this available movement option in Indonesian is blocked by
independent principles of syntax in (56).
Finally, Fortin treats meN- in Indonesian as the antipassivizing object clitic. In other
words, meN- clitic-doubles the adjoined PP which gives the full specification of the object.
However, this treatment of meN- seems problematic because meN-+ object NP
constructions do not show properties characteristic of bona-fide double-clitic constructions,
as found in Romance languages. To mention one case, Franco and Landa (2006) show that
116
quantifiers and wh-words cannot be clitic-doubled due to their inherent quantificational
force. This is shown in Spanish examples in (60a, b); see also Suñer (1988), Franco (1993),
Cinque (1990), and Rizzi (1997) for much relevant discussion.
(60)a. Juan (*loi) quiere
Juan
Cl
a todo el mundoi. b. ¿A
love-3s to everybody
to
‘Juan loves everybody.’
quiéni (*loi) viste?
whom
Cl
saw-2
‘Who did you see?’
(Franco and Landa 2006: 37)
Thus, Fortin’s treatment of meN- predicts that quantifiers and wh-expressions should also
not co-occur with the prefix. This prediction, however, is falsified by examples of wh-insitu as in (6c), repeated here as (61a), and examples as in (61b) that have quantifiers in
direct object positions.
(61) a.
Ali
(mem)-beri Fatimah apa?
Ali
AV-give
Fatimah what
‘What did Ali give Fatimah?’
b.
Ali
(mem)-beri
setiap
orang.
Ali
AV-love
every
one
‘Ali loves everyone.’
117
The three considerations above, therefore, suggest that Fortin’s antipassive analysis of the
active voice deletion in Indonesian is difficult to sustain.
4.5. Anti-Agreement-Based Accounts
Another potential analysis of the meN-/ng- deletion in Indonesian and Javanese is to claim
that this deletion instantiates a type of agreement known as anti-agreement since Ouhalla
(1993, 2005), whereby (local) extraction of the subject results in impoverished
morphology.23 Some examples from Berber are given here for illustration.
(62) Anti-agreement in Berber
a. man
tamghart ay yzrin
which woman
Mohand?
C see (Part)
b.*man tamghart ay t-zra
Mohand
‘Which woman saw Mohand?’
Mohand?
whichwoman C 3fs-saw Mohand
‘Which woman saw Mohand?’
(Ouhalla 1993: 479)
In these examples, the subject man tamghart ‘which woman’ is moved into the specifier of
CP. When this extraction happens, the agreement gives rise to ungrammaticality, as
23
I thank Andrew Carnie (personal communication) and Catherine Fortin (personal communication) for
suggesting this potential analysis and relevant work cited here.
118
illustrated in (62b); the verb instead must take its morphologically impoverished form, as
in (62a), which traditional grammarians dub the participle. Ouhalla (1993) derives antiagreement from the interaction of the licensing condition on pro with the A′-disjointness
requirement on the distribution of pronominal elements. It has been often held that pro is
licensed in the subject position in a language with rich agreement, as evidenced by prodrop languages such as Italian and French. He follows this analysis and assumes that pro is
licensed only by rich subject agreement. The A′-disjoint requirement, which he adopts
from Aoun and Li (1990) and defines as in (63), states that pronominal elements, including
pro, must be free within the minimal CP that contains them.
(63) The A′-disjointness Requirement
A pronoun must be free in the smallest Complete Functional Complex (CFC) which contains it.
(Ouhalla 1993: 490)
Ouhalla claims that anti-agreement is the result of the strategy to avoid violating the A′disjointness Requirement. In (62b), regular rich agreement is reflected, which licenses a
pro in the subject position. This pronominal argument, however, violates the disjointness
requirement, giving rise to the ungrammaticality in (62b). The example in (62a) is
119
grammatical, by contrast, because the impoverished morphology of the verb does not
license a pro in the subject position, which is occupied by the trace of the wh-movement.
Although Ouhalla’s (1993) analysis has the potential of deriving the active voice deletion in
Indonesian and Javanese as another manifestation of the anti-agreement effect, there are
several problems in this potential attempt, to the extent that Ouhalla’s analysis is tenable. First,
the anti-agreement effect in a language correlates with the availability of rich agreement in that
language. However, Indonesian and Javanese are both agreement-less languages in the same
sense that languages Chinese and Japanese are. Second, Ouhalla observes that anti-agreement
occurs only in the case of subject extraction. I have provided evidence in section 2 that meN/ng- deletion is triggered (in fact, more typically) by non-subject NPs. Also recall that, as
dicussed in section 4.1, an NP can be extracted from non-subject positions into the specifier of
CP without first passing through the subject position [Spec, TP] in Indonesian and Javanese.
These two differences indicate that the analysis of active voice deletion as an instantiation of
the anti-agreement pattern of the Berber kind is problematic.
4.6. Case-Based Accounts
Let us finally consider a potential Case-based account of the restriction on the movement of
an NP across an active voice verb marked with meN- and ng-. This analysis would say that
120
the moved NP enters into a Spec-Head Case-checking relation with its local v head, which
results in the deletion of the active voice. This analysis correctly predicts that the extraction
of the external argument from a clause headed by meN-/ng- verbs does not trigger the active
voice deletion because that argument checks its nominative Case against T. This analysis
also naturally explains the NP-only property of extraction because non-nominal phrases such
as adverbials and prepositional arguments do not enter into the Spec-Head Case-feature
agreement with their local v heads.
The Case-based approach is also desirable in one important respect. Other Austronesian
languages such as Chamorro, Palauan, Tagalog, and Malagasy exhibit a phenomenon in the
form of WH-agreement, the mood alternation, and case agreement, which is structurally
quite similar to the active voice deletion in Indonesian and Javanese. Indeed, existing
analyses of these phenomena as in Chung (1982, 1994, 1998), Georgopoulos (1985, 1991),
and Rackowski and Richards (2005) propose a Case-based explanation of these phenomena.
Therefore, it seems that the Case-based analysis not only correctly derive Cole and
Hermon’s generalization but also allows a unified explanation of the Austronesian syntaxphonology interaction.
However, there are two reasons to believe that the Case-based analysis is incorrect for the
meN-/ng- deletion in Indonesian and Javanese. First, if the Case-based analysis could be
121
extended to Indonesian and Javanese, it would predict that a single NP should receive multiple
Accusative Cases on its way toward the final landing site in multi-clausal environments since it
triggers meN-deletion on all intervening verbs on its movement path. This is a rather
undesirable outcome. No languages have been found in which a single NP receives more than
two structural Cases. Bejar and Masam (1999) document several instances of multiple Case
checking in languages such as Niuean, Norwegian, and Latin and develop a late-insertion
analysis of Case assignment constrained by markedness, so it is possible that a single NP could
receive multiple cases.24 However, all the cases discussed there are restricted to those in which
a single NP receives either two different values for structural Case [Nominative and
Accusative] or one structural Case [Nominative/Accusative] and one inherent Case [Oblique].
However, if the meN-/ng- deletion were the reflex of the accusative Case-checking in
Indonesian and Javanese, then a single A-bar moved phrase in a multi-clausal environment
would receive a series of the same values for Case, in principle, the number of Cases that
corresponds to the number of transitive verbs present in multi-clausal environments. However,
this type of multiple case checking has always been assumed to be impossible in natural
language, and as long as a better alternative account is possible, should be dispreferred.
24
As suggested by Simin Karimi (personal communication). See also Boeckx and Hornstein (2006) for an
analysis of the multiple case agreement in Icelandic control constructions.
122
Second, this multiple Case-checking account is suspect on theoretical grounds. With the
exception of cases as discussed in Bejar and Masam (1999), it seems correct to think that a
single DP receiving a single Case is the unmarked pattern in natural language syntax. The
modern theory of Case assignment/checking/evaluation in the Government-and-Binding
approach (Chomsky 1981, 1986a, b) or the Minimalist Program (Chomsky 1995, 2000,
2001, 2004, 2005) is indeed so constructed as to exclude a single DP from receiving more
than one Case. Within the Minimalist Program outlined in Chomsky (1995), for example,
the alleged Case-driven movement, as applied to long-distance extraction of wh-phrases in
Indonesian and Javanese, would violate the Last Resort Condition/Greed, which
effectively blocks movement from a Case position to another Case position, as shown in
examples such as (64b).
(64) a.
Johni seems to Mary [TP ti to be ill].
b.* Johni seems to Mary [CP that ti is ill].
This theoretical viewpoint, therefore, casts further doubts on the Case-based analysis of meN/ng- deletion. Therefore, I conclude that the Case-based analysis of the distribution of the
active voice deletion is not tenable for Indonesian or Javanese.
123
5. Conclusions
Let us summarize the present chapter. I have started by reviewing Cole and Hermon’s (1998)
generalization that, in Malay/Indonesian, the movement of an NP across a verb triggers the
active voice deletion from the verb. I have shown that this generalization also adequately
characterizes the distribution of the active voice morphology in Javanese and Madurese. I have
argued, following Kayne’s (1989) analysis of French participle agreement, that this
generalization is naturally explained if the obligatory deletion of the active voice morphology is a
reflex at the syntax-external sound component of the syntactic Spec-Head D-Feature Checking
between the moved NP and its local v head. I have shown that the external argument does not
cause this deletion by introducing the crucial distinction between External Merge and Internal
Merge recently proposed by Chomsky (2004, 2005). The currently analysis naturally predicts
that the movement of a non-nominal phrase such as adverbials and prepositional arguments does
not trigger the deletion due to their lack of nominal D-features.
I have compared the proposed analysis with several existing/potential alternative accounts
of the same phenomena as in Keenan (1972)/Keenan and Comrie’s (1977) NP-Accessibility
account, Soh’s (1998) Relativized Minimality account, Fortin’s (in press) antipassive
account, the anti-agreement-based accounts, and the Case-based account. I have shown that
each of these analyses has a particular set of empirical and cncepual problems that are
124
naturally resolved under the current phase-theoretic analysis. To the extent that this analysis
is correct, the syntactically governed distribution of the active voice morphology in
Indonesian and Javanese provides strong support for the role of the vP phase at the syntaxphonology interface. This result is important because evidence has been deemed hard to find
for the successive-cyclic movement via the edge of vP.
5.1. The Derivational Nature of the Syntax-Phonology Interface
In this subsection, I discuss a theoretical implication of the proposed analysis of the
distribution of the active voice morphology in Indonesian and Javanese for the syntaxphonology interface. The proposed analysis provides clear evidence that the syntaxphonology interaction is fundamentally derivational, meaning that the interpretation of
syntactic objects at the phonological component occurs in a piece-meal fashion rather than
by examining the final output of the syntactic derivation.
The idea that syntax interacts with phonology in a dynamic manner has proven a
fruitful line of inquiry since the seminal work of Bresnan (1971a, b), who shows that
nuclear stress assignment is cyclic. See also Legate (2003), Arregi (2003), Kahnemuyipour
(2004) and Adger (2007) for recent phase-theoretic implementations of Bresnan’s analysis
which was couched within the Standard Theory. I have also shown elsewhere (Sato in
125
press c; see also Carnie in preparation) that the domain for certain phonological rule
applications such as Taiwanese tone sandhi, Giyak lenition, Kinyamboo high tone deletion,
and Irish/Welsh consonant mutation, are isomorphic to domains demarcated by a
derivational model of syntax like Phase Theory. The proposed account of the active voice
deletion in Indonesian and Javanese, therefore, provides further evidence that the syntactic
computation is tightly intertwined with phonology in a derivational manner.
5.2. Minimalist Interfaces
I conclude by noting what the results achieved in this chapter tell us about the thesis of “minimalist
interfaces” outlined in the introductory chapter. Part of the idea behind this thesis as stating that the
syntax-external interpretive components follow a parametrically defined path curved by the
universal syntactic computation and assign whatever interpretation it can to a syntactic object sent
to the components. The results in this chapter give substance for this notion of minimalist interfaces.
Indonesian and Javanese happen to signal successive cyclic movement via the deletion of the active
voice on the v head. This, of course, does not have to be the only way to indicate the universal
locality principle in syntactic computation. Indeed, many other languages have different ways of
accomplishing the same task of successive cyclic movement via the two phase edges at the
interface. Irish (McCloskey 1979, 2001; Duffield 1995) expresses the locality of movement via the
126
edge of CP and vP phases through complementizer agreement and the ag→aL rule in progressives,
respectively.
25
Chamorro (Chung 1982, 1994, 1998), Palauan (Georgopoulos 1985, 1991),
Tagalog (Rackowski and Richards 2005), and Malagasy (Pearson 2001, 2005) exhibit movement
to the vP edge via Case agreement, a special mood alternation, and voice agreement, respectively.
French marks the movement of an NP via [Spec, vP] by participial agreement (Kayne 1989).
Berber (Ouhalla 1993, 2005) marks the intermediate site of the movement in [Spec, CP] by the use
of special impoverished agreement morphology. Kikuyu (Clements 1984; Clements et al. 1983)
signals the movement via the vP edge in the form of the loss of tonal downstep from a verb crossed
by the movement. Innu-Aimûn (Branigan and MacKenzie 2002), and Kilega (den Dikken 2001)
show evidence for the local movement through intermediate vPs in the form of Wh-agreement on
the verbs, as in Chamorro. Hausa and Moore express the occurrence of successive cyclic
movement via the vP edge through the two different forms of modals and affixes. Afrikaans (Du
Plessis 1977) marks successive cyclic movement through [Spec, CP] by prepositional copying in
intermediate specifiers of CPs. Spanish (Torrego 1984) and French (Kayne and Pollock 1979)
provide evidence for intermediate [Spec, CP]s via obligatory stylistic inversion in clauses out of
which movement has occurred. Wh-copying in languages such as Frisian, Dutch, Afrikaans,
German, and Child English (Felser 2004; Thornton 1995) as well as the existence of partial wh-
25
I thank Andrew Carnie (personal communication) for bringing my attention to the ag→aL rule.
127
movement in languages like German, Hungarian, and Indonesian (McDaniel 1989; Horvath 1997;
Saddy 1991; Cole and Hermon 1998, 2000) provide morphological evidence that wh-phrases
move through the edge of CP phases. Even languages with impoverished morphology such as
English, which have no morphological ways to signal the phase-based locality of syntactic
movement, still have alternative methods of reflecting the locality of movement in the other
semantic interface via reconstruction, parasitic gaps, and antecedent-contained deletion (Lebeaux
1988; Fox 1998; Legate 2003).
All these phenomena are but several of the options within the range set by Universal
Grammar the phonological interface can choose to reflect the fundamental principle of
locality in syntactic computation. This is highly reminiscent of the remark below, made by
Chomsky (1995), citing Otto Jespersen.
Jespersen held further that it is only “with respect to syntax” that we expect “that there
must be something in common to all human speech”; there can be a “universal (or
general) grammar,” hence a perhaps a far-reaching account of the initial stage of the
language faculty in this domain, though “no one ever dreamed of a universal
morphology.” (Chomsky 1995: 3)
128
The existence of the apparently diverse, outwardly different surface manifestations that
nonetheless signal the single core property of syntax can be construed as powerful support
for the idea of “minimalist interfaces” that the working of the sound-related interface
component is fundamentally interpretive and subservient to the universal law of syntax in
that it has a handful of domain-specific operations, some of them discussed above, to
reflect the locality of syntactic derivation. I discuss further ramifications of this notion of
minimalist interface in chapter 7.
129
CHAPTER 3 PREPOSITION STRANDING UNDER SLUICING AND REPAIR
BY DELETION: WHY IS INDONESIAN (NOT) SO SPECIAL1
1. Introduction
In this chapter, I explore the issue of the syntax-phonology interface with a case study in
sluicing in Indonesian to see the extent to which the thesis of minimalist interface proposed
in the first chapter holds. I demonstrate that detailed investigation of the sluicing construction
in this language lends further credence to the idea embedded within the proposed thesis that
the syntax-external phonological component avails itself of whatever domain-operation it
can to remedy certain “imperfections” created by the functionally blind computational
process within the curve parametrically defined by the syntax.
My starting observation in this chapter is that sluicing in Indonesian behaves differently
from other languages with respect to P-stranding. In favor of his movement and TP deletion
approach to sluicing constructions, Merchant (2001) proposes, based on his survey of 24
languages, the P(reposition)-Stranding Generalization, namely, that P-stranding under
sluicing is possible only in those languages that independently allow P-stranding under
1
The core observation in this chapter was presented at the Linguistic Colloquium Datablitz held at the
University of Arizona (April 2007) and later published in Sato (2007). An earlier draft of this chapter was
presented at the 34th Annual Meeting of the Berkeley Linguistics Society (BLS 34) held at the University
of California, Berkeley (February 2008) and will appear in Sato (forthcoming a).
130
regular wh-questions. I show that Indonesian presents itself as a clear counterexample to this
cross-linguistic generalization to lay the groundwork for the investigation of the syntaxphonology interface. One of the most important findings in this chapter lies in my
demonstration that Indonesian constitutes the first bona fide counterexample to the
generalization stated above. Although several languages such as Brazilian Portuguese
(Almeida and Yoshida 2007), Polish (Szczegielmiak 2006), Mandarin Chinese (Wang 2006),
Malagasy (Potsdam 2003), and Serbo Croatian (Stepanović 2008) that appear to contract the
P-Stranding Generalization, I point out that none of these languages poses a genuine
counterexample to the generalization, unlike Indonesian. In light of this observation, I
propose a new analysis of the typology of P-stranding under sluicing which takes seriously
the role of prepositions across languages with respect to the syntax-phonology interface.
Specifically, I argue that the apparently atypical P-stranding pattern observed in Indonesian
naturally falls out from independently motivated assumptions concerning the percolation of
the [+wh] feature onto the PP, D-to-P incorporation, and the minimalist constraint on
movement/attraction. The core idea behind the proposed analysis is that certain violations
created by syntactic derivation can be remedied by syntax-external operations such as
deletion at the phonological component, an idea that goes back to Ross’ (1969) global
evaluation of the syntactic derivation and has been developed in recent minimalist inquiries
131
as in Merchant (2001), Boeckx and Lasnik (2006), Lasnik (1999, 2001, 2005, 2007), and
others. The proposed analysis is summarized in Table 1.2
Table 1: The Parametric-Theory of P-Stranding at the Syntax-Phonology Interface
Parameters
Languages
English
[+wh] feature percolation
from the nominal to PP
OPTIONAL
D-to-P incorporation
in the syntax
NO
Indonesian
OBLIGATORY
NO
French
OBLIGATORY
YES
I demonstrate that the proposed analysis correctly captures the three-way contrast between
English, French, and Indonesian with respect to the P-stranding pattern under sluicing and
wh-movement. I also show that the proposed analysis also makes a correction prediction
that not only TP deletion but deletion of smaller constituents saves the otherwise illicit
syntactic configuration through examination of pseudogapping in Indonesian and French.
The core observation in this chapter that Indonesian presents a challenge to the PStranding Generalization has been independently made by Fortin (2007a), around the
2
I discuss the fourth logically possible parameter setting, namely in section 4.2, where I argue that the non-
existence of languages with this setting is fully predicted due to the incompatibility of obligatory D-to-P
incorporation with optional wh-feature percolation.
132
same time I started working on sluicing in this language, a preliminary result of which
was reported in Sato (2007). We have not been aware of each other’s work until quite
recently, when her dissertation (Fortin 2007b) became available to me through our email correspondence. As we will see below, this chapter owes a great deal to her
dissertation, and reaches the same observation that Indonesian is a real counterexample
to the P-Stranding Generalization. However, the analysis developed in this chapter is
quite different from her analysis of the P-stranding pattern, each motivated by two
distinct sets of theoretical assumptions within the minimalist framework. Nonetheless, I
hereby would like to acknowledge that her work represents the first comprehensive
description and analysis of Indonesian sluicing and that the current chapter also draws on
her data and analysis.
The organization of this chapter is as follows. In the next section, I provide an overview
of the analysis of sluicing constructions proposed by Merchant (2001) as the result of the
syntactic movement of the wh-remnant, followed by TP deletion at the PF component, and
introduce one of his core generalizations in support of this analysis, the P-Stranding
Generalization, namely, that a language allows P-stranding/omission under sluicing if and
only if it allows P-stranding under wh-movement. Several languages have been reported in
the literature that appear to be counterexamples to this generalization in that P-stranding is
133
impossible in sluicing but permitted under wh-questions in those languages. These
languages include Brazilian Portuguese (Almeida and Yoshida 2006), Mandarin Chinese
(Wang 2006), Polish (Szczegielniak 2006), Serbo-Croatian (Stjepanovic 2007), and
Malagasy (Potsdam 2003). I review the derivation of sluicing constructions in these
languages with reference to the works cited here and conclude that none of these languages
refutes the P-Stranding Generalization because wh-questions in these languages involve
alternative sources of syntactic derivation such as clefting (Brazilian Portuguese and
Polish), resumption (Mandarin Chinese), pseudoclefting (Malagasy), and P-drop at PF
(Serbo-Croatian), that do not involve wh-movement of the English kind.
In section 3, I turn to a detailed investigation of the syntax of sluicing constructions in
Indonesian. The sole purpose of this section is to establish that sluicing in this language
presents the first real counterexample to the P-Stranding Generalization. I achieve this goal
as follows. I first review Fortin’s (2007b) results of applying some of the operational tests
developed by Merchant (2001) to distinguish genuine sluicing (derived by wh-movement +
TP deletion) and pseudosluicing (derived by clefting + ellipsis), which suggest that
Indonesian sluicing behaves more like English sluicing rather than like English clefting. I
conclude that this result, though indicative of the nature of sluicing in Indonesian, is not
conclusive, as the operational tests were developed on the basis of English and related
134
languages. Thus, I provide language-internal evidence based on the distribution of the
question marker -kah (Fortin 2007b) and the obligatory absence of the complementizer
yang with PP remnants that show that sluicing in this language is derived by true whmovement plus TP deletion, rejecting alternative analyses of wh-questions as reduced
clefts/headless relative clauses, as proposed by Cheng (1991) and Cole et al. (to appear).
Having established that Indonesian sluicing presents itself as the first genuine
counterexample to the P-Stranding Generalization, in section 4, I propose a new analysis
of the typology of P-stranding patterns that draws on independently motivated assumptions
concerning the percolation of the [+wh] feature onto the PP (Chomsky 1972; Stepanov
2001) and the D-to-P incorporation in Romance (Law 1998, 2006; van Riemsdijk 1998).
The core idea pursued here is that certain violations such as the failure of feature
percolation can be ameliorated as long as offending configurations created in syntax can be
nullified via deletion at the syntax-external phonological component. I show that the
proposed analysis straightforwardly derives the three-way contrast between English,
French, and Indonesian with respect to P-stranding under wh-questions and sluicing. The
proposed analysis further predicts that the size of the syntactic object deleted at PF does
not matter for the purposes of amelioration at the interface as long as the offending
135
configuration is eliminated at this level. I show that this prediction is indeed borne out by
Indonesian and French pseudogapping constructions.
In section 5, I compare the proposed interface-based approach to Indonesian sluicing with
the most recent analysis of it by Fortin (2007b) in terms of LF Copying and Long-Distance
Agree. I show that, although her analysis correctly captures the apparently atypical P-stranding
pattern in Indonesian, it encounters serious empirical and conceptual problems with regards to
the case checking of the wh-remnant, the restriction on [+wh] feature percolation, and casematching effects. I also briefly address a number of issues concerning the derivation and
representation of syntax and its interface with the PF component.
In section 6, I provide the summary of this chapter and point out several important
implications of the proposed analysis for the proper theory of syntax-phonology
correspondence. Among them, I point out that the current investigation provides strong
confirmation to the idea in the minimalist interface that the syntax-external phonological
component does whatever it can to save otherwise illegitimate violations created by
universal syntactic mechanisms but only does so within the restricted range of options
parametrically available in a particular language.
136
2. Sluicing Constructions
Sluicing is one of the most investigated constructions in recent syntactic literature. It was
identified in seminal work by Ross (1969). “Sluicing” refers to the elided part of sentences
such as Mary bought something but I don’t know what [
]. Research since Ross (1969) has
concentrated on identifying the external and internal syntax of the elided part as well as
syntactic/semantic conditions that license it. The purpose of this section is two-fold. First, I
review the most recent analysis of this construction proposed by Merchant (2001) and
introduce one of his core form-identity generalizations concerning P-stranding under sluicing
known as the P-Stranding Generalization. This generalization states that P-stranding/omission
in a language is possible only if that language independently allows P-stranding under regular
wh-movement. Second, I review several languages that potentially undermine this
generalization and show that none of these languages is a genuine counterexample to this
generalization because alternative syntactic and/or phonological strategies for P-stranding are
available to derive sluicing constructions other than syntactic wh-movement.
137
2.1. Merchant’s (2001) Analysis of Sluicing and the P-Stranding Generalization
Drawing on the data and analysis presented in Ross (1969), Merchant (2001) argues that
sluicing constructions as in (1a) are the product of syntactic wh-movement of the remnant
who, followed by the deletion/dephoneticization of the TP constituent, as shown in (1b).
(1)
a.
Somebody just left. - Guess who.
b.
Somebody just left. - Guess [CP whoi … [TP ti just left]]
Merchant adduces a wide variety of syntactic and morphological effects such as casematching, number agreement, and so on, many of them mentioned in Ross (1969), to support
his movement plus TP deletion analysis. One of the most convincing arguments in its favor
comes from what he calls the Preposition-Stranding Generalization stated as in (2).
(2)
Preposition-Stranding Generalization (Merchant 2001: 92)
A language L will allow preposition stranding under sluicing iff L allows
preposition stranding under regular wh-movement.
138
This generalization states that, if a preposition can be omitted under sluicing in a
particular language, that language should independently allow P-stranding under regular
wh-movement. The logic behind this generalization is clear. Under Merchant’s analysis,
sluicing is derived by regular wh-movement plus TP deletion. Therefore, the availability
of P-stranding under sluicing means that the same option should be independently
available under regular wh-movement.
Merchant (2001) surveys the P-stranding
pattern both under wh-movement and sluicing in 24 languages to show that this
generalization is crosslinguistically correct for these languages. As is well-known,
English allows P-stranding both under wh-movement and sluicing, as shown in examples
as in (3a, c), in conformity with the P-Stranding Generalization. Note that the preposition
can also be pied-piped along into the specifier of the CP, as illustrated in (3b).
(3) a.
Who was he talking with?
b.
With whom was he talking?
c.
Peter was talking with someone, but I don’t know (with) who.
(Merchant 2001: 92)
139
This pattern falls out from Merchant’s analysis because the preposition-less variant of the
sluicing in (3b) is derived when the preposition with is stranded within the TP, as in (3a).
On the other hand, Romance languages such as French are strongly non-P-stranding
languages, as shown by the contrast between (4a) and (4b). Thus, French does not allow
omission of the preposition under sluicing, as shown in (4c), a pattern that is also predicted
by Merchant’s analysis. (The examples in (4a, c) are from Merchant 2001: 98.)3
(4) a. * Qui
who
est-ce
qu’
Q
elle
l’a
offert
she
it-has
offered
à?
to
‘Whom has she offered it to?’
b.
À
qui
l’a-t-elle
offert ?
to
whom
it-has-she offered
‘To whom has she offered it?’
c.
Anne
l’a
offert
à quelqu’un, mais je
Anne
it-has offered to someone
but
I
ne
sais pas *(à) qui.
Neg know not to whom
‘Anne has offered it to someone, but I don’t know (to) whom.’
3
Some speakers readily accept the P-less sluice in (4c). Thanks to Myriam Bouveret (personal
communication) for pointing this out.
140
The type of language that is not predicted by Merchant’s generalization, therefore, have
syntactic wh-movement, disallow P-stranding under this context, but nonetheless allow Pstranding/omission under sluicing. In section 3, I provide arguments based on the
examples as in (5a-c) that Indonesian is precisely of this type.
(5) a. * Siapa
who
yang
kamu
ber-dansa
dengan?
that
you
Vz-dance
with
‘Whom did you dance with?’
b.
Dengan siapa
kamu
ber-dansa?
with
you
Vz-dance
who
‘With whom did you dance?’
c.
Saya ingat
I
Hasan
ber-dansa
dengan seseorang, tapi saya tidak
remember Hasan
Vz-dance
with
someone but I
Neg
tahu (dengan) siapa.
know (with)
who
‘I remember Hasan danced with someone, but I don’t know (with) whom.’
141
The contrast between (5a) and (5b) shows that Indonesian does not allow P-stranding
under wh-questions, as in French. Surprisingly, however, the grammaticality of the Pless sluice in (5c) illustrates that the preposition can be deleted under sluicing. To the
extent that the underlying syntactic source for sluices as in (5c) involves syntactic whmovement, the observed pattern of P-stranding in (5a-c) presents a counterexample to
the P-Stranding Generalization.
Before going into detailed discussion of the precise derivation of sluicing constructions in
Indonesian, however, I review several other languages that have been reported in the
literature to disallow P-stranding under wh-questions but allow this option under sluicing. I
show that this P-stranding pattern in those languages does not undermine the P-Stranding
Generalization because wh-questions in those languages have alternative derivational
sources such as clefting (Brazilian Portuguese, Polish), pseudoclefting (Malagasy),
resumption (Mandarin Chinese), or the post-syntactic P-drop (Serbo-Croatian).
2.2. P-Stranding under Sluicing in Languages without P-Stranding
In this subsection, I discuss the P-stranding pattern in Brazilian Portuguese, Polish, Mandarin
Chinese, Malagasy, and Serbo-Croatian as examples of languages that appear to contradict
the P-Stranding Generalization. Several works published on this issue show that none of
142
these languages undermines the relevant generalization because sluicing in these languages
involve some syntactic or phonological operation other than regular wh-movement, as in
English. I also discuss whether any of the analyses developed in the literature for these
languages can be transported to the Indonesian P-stranding pattern and show that none of
them works for Indonesian. The survey here is far from comprehensive, as many other
languages such as Persian, Russian, and Finnish also seem to behave in a way not predicted
by the generalization.4 It is by now widely acknowledged that the generalization does not
hold across the board either across languages or within a single language.
2.2.1.
Brazilian Portuguese
Brazilian Portuguese is perhaps the most well-known language that seems to directly
contradict the P-Stranding Generalization. Almeida and Yoshida (2007) points out
example as in (6a-c).
4
I thank Simin Karimi (personal communication) and Mans Hulden (personal communication) for bringing
Persian and Finnish to my attention. See Szczegielniak (2006) and Hartman (2005) for some data concerning
German and Finnish. However, whether or not the languages mentioned here are true counterexamples to the
P-Stranding Generalization is unclear and requires detailed work that goes beyond the scope of this chapter.
143
(6) a. Com
with
quemi
que
a
Maria
dançou
who
that
the
Maria
danced
ti ?
‘With whom did Maria dance?’
b.* Quemi
who
que
that
a
Maria
the
Maria
dançou
com ti ?
danced
with
‘Who did Maria dance with?’
c. A Maria dançou com alguém, mas eu não lembro
quemi a
Maria dançou com ti.
the Maria danced with someone but I Neg remember who the Maria danced with
‘Maria danced with someone, but I don’t remember who.’
(Almeida and Yoshida 2007: 350)
Applying those tests applicable to Brazilian Portuguese that were developed by Merchant
(2001) to distinguish between genuine sluicing and elliptical clefts (i.e. prosody, aggressively
non-D-linked wh-phrases, ‘mention-some’ modification, and ‘else’ modification, Almeida and
Yoshida present arguments to show that the example in (6c) is not an instance of elliptical cleft
(as argued for in Erteschik-Shir 1977) but a genuine sluicing. In a recent rebuttal, however,
Rodriguez et al. (2007) argue that the P-stranding pattern observed in Brazilian Portuguese
does not constitute a genuine counterexample to the P-Stranding Generalization by showing
144
that there are two distinct sources of sluicing in Romance, one wh-movement plus TP deletion
as in Merchant’s (2001) theory, and the other clefting plus TP deletion, as in Merchant’s (1998)
analysis of Japanese-type “sluicing” that does not involve P-stranding in the first place.5 Their
cleft analysis is illustrated here with the derivation of the P-less sluice in Brazilian Portuguese
in (7a) that is given in (7b).
(7)
O
João
falou
com
una
the
João
talked
with
a
a. mas eu
but
I
não sei
menina…
girl
qual é a menina com a
Neg know which
is the girl with the
qual João
falou
which João
talked
b. mas eu não sei [CP qual [IP é [DP a menina [RC com a qual João falou]]]
butI Neg know
which is
the girl
with the which João talked
Lit. ‘João talked with a girl, but I don’t know which is the girl with which Joãotalked.’
(Rodrigues et al. 2007: 4)
The derivation in (7b) involves a cleft containing a specificational copular sentence; the
copular verb é is followed by a DP that contains a restrictive relative clause structure. Since the
5
I thank Heidi Harley (personal communication) for directing my attention to Rodriguez et al. (2007).
145
pivot of the cleft derivation here is not introduced by the preposition, deletion of all materials
including the verb and the predicate has the same effect as P-stranding, as commonly
understood. Rodrigues et al. provide evidence internal to Brazilian Portuguese and Spanish that
the tests Almeida and Yoshida used against a psedusluicing analysis for this language are not
applicable to these languages due to the bleached semantics of clefts, the inadequacy of the Dlinking test, and so on, which I will not review here.
Rodrigues et al’s (to appear) cleft analysis for Spanish and Brazilian Portuguese is quite
similar to the headless relative clause analysis proposed by Cole, et al. (to appear) for whquestions in Malay/Indonesian. In section 3, I provide evidence that their cleft analysis is not
correct for Indonesian.
2.2.2.
Polish
Merchant provides data as in (8a, b) that Polish conforms to his P-Stranding Generalization.
(8) a. * Kim
who
rozmawiala
Anna
z?
spoke
Anna
with
‘Who did Anna speak with?’
146
b. * Anna
Anna
rosmawiala
z
kimś
spoke
with someone
ale
nie
wiem
kim.
but
not
I.know
who
‘Anna spoke with someone, but I don’t know who.’
(Merchant 2001: 96)
Szczegielniak (2006) argues, however, that the facts in Polish are more complicated than
what Merchant reports.6 Specifically, he provides examples as in (9a-c) to show that the
omission of the preposition is possible with sluicing only with D-linked wh-phrases.
(9) a. * Którym1
which
Anna
tańczyła
z
Anna
danced
with
t1 męŜczyzną.?
man
‘Which man did Ann dance with?’
b. * [Którym
which
męŜczyzną] 1
Anna
tańczyła
z
man
Anna
danced
with
t1?
‘Which man did Ann dance with?’
6
Thanks to Andrew Carnie (personal communication) for directing my attention to Szczegielniak (2006).
147
c.
Anna
tańczyła z
jednym męŜczyzną, ale
Anna
danced with one
man
but
nie
wiem
Neg know
(z)
którym.
with which
‘Ann danced with one man, but I don’t know which.’
(Szczegielniak 2006: 3)
The ungrammaticality of the examples in (9a, b) under wh-questions shows that Polish
does not allow stranding of prepositions such as z ‘with’. The fact that the same
preposition can be deleted in the sluicing counterpart as in (9c), therefore, poses a problem
for Merchant’s generalization. Szczegielniak proposes, however, that, in cases of sluicing
with D-linked wh-phrases, a non-wh-movement alternative is available for languages such
as Polish and Russian in the form of a cleft construction. The underlying cleft structure for
the P-less sluice in (9c) is given in (10a). Note that the deleted portion of (9c) is itself
grammatical, as shown by the cleft wh-question in (10b).
148
(10)a.
Anna tańczyła z
jednym męŜczyzną ale
Anna danced with one
[z
b.
t2
man
but
nie
wiem
[którym]2
to
not
know
which
it
męŜczyzną] 1
(ona)
tańczyła t1.
with
man
(she)
danced
którym2 to
[z t2
męŜczyzną] 1
(ona)
tańczyła
which
with
man
(she)
danced
it
t 1.
‘With which man was it that she danced?’
(slightly modified from Szczegielniak 2006: 3)
In the derivation in (10a), the PP containing the wh-phrase undergoes movement, which is
followed by clefting of the wh-phrase itself. Crucially, then, this derivation does not
involve the P-stranding configuration in the sense that it does not extract the entire DP from
the complement position of the preposition. This analysis also correctly accounts for why
P-less sluices are ungrammatical with non-D-linked wh-phrases as in (8c); there is no
possible cleft source for such wh-phrases, as shown by the possible but non-convergent
derivation given in (11a). Note that the deleted portion of these derivations is itself
ungrammatical in Polish, as shown in (11b).
149
(11)a. * Anna tańczyła z
jednym męŜczyzną
Ann danced with one
[z
t2]1
with
man
ona
tańczyła
she
danced
ale
nie
wiem
kim2
to
but
Neg
know
who
it
t1.
‘Ann danced with one man but I don’t know who it was that she danced with.’
b. * Kim2
who
to
[z
it
with
t2]1
ona
tańczyła
she
danced
‘Who was it that she danced with?’
t1.
(Szczegielniak 2006: 4)
Based on these data, Szczegielniak concludes that the apparently problematic P-omission
pattern under sluicing naturally makes sense because only D-linked wh-phrases have cleft
sources that do not involve extraction of them from the PP. Szczegielniak extends this
analysis to German sluicing which exhibits the same behavior with Polish in terms of Pstranding under sluicing.
Like Rodriguez et al’s (2006) analysis for sluicing in Brazilian Portuguese, the clefting
analysis could also account for the P-stranding facts in Indonesian, which is superficially
similar to that in Polish in terms of P-stranding under wh-questions and sluicing. This
consideration is important since several researchers in Indonesian as in Cheng (1991) and Cole
150
et al. (to appear) have independently proposed that wh-questions in this language are a form of
cleft or headless relative clause. I show in section 3.2, however, that the cleft analysis is not the
right approach to sluicing in Indonesian.
2.2.3.
Mandarin Chinese
Wang (2006) reports that Mandarin Chinese presents a counterexample to the P-Stranding
Generalization as a non-P-stranding language that nonetheless allows omission of the
preposition under sluicing. Consider examples in (12a, b).
(12)a. * (shi)
Foc/Cop
Lisi
[na-ge
ren]i
which-Class
person Lisi
gen
ti
with
zai
shuohua?
Prog
talk
‘Which one is Lisi talking with?’
b.
Lisi gen mou-ge
ren
qu wan, dan wo bu zhidao shi
Lisi with certain-Class person go play but I
(gen) shei.
Neg know Foc/Cop with who
‘Lisi has a trip with a certain person, but I don’t know who.’
(Wang 2006: 9, 10)
151
(12a) shows that Mandarin Chinese does not allow P-stranding under wh-questions, but (12b)
shows that the P-less sluice is grammatical, contrary to the P-Stranding Generalization. Based
on her observation that all languages that show this P-stranding pattern are reported to make an
extensive use of a resumptive pronoun strategy, she argues that (12b) is grammatical because it
involves the generation of a resumptive pronoun that follows regular wh-movement. This is
evidenced by the acceptability of the resumptive strategy in wh-questions and sluicing
constructions, as shown in (13a) and (13b), respectively.
(13)a.
b.
Lisi
na-ge
reni
which-Class
person Lisi
keshi wo bu
zhidao na-ge
but
hen
zihuan
(ta-ti)?
very
like
him
reni <TP Lisi gen
ta-ti
qu kan dianying>
I Neg know which-Class person Lisi with him go see
movies
Lit. ‘… but I don’t know which person (did) Lisi go to the movies with him.’
(Wang 2006: 10, 11)
To the extent that her resumptive analysis of sluicing in Mandarin Chinese holds, the Pstranding pattern in this language does not contradict the P-Stranding Generalization, as
the syntax of this construction does not involve regular movement of a question operator
152
into [Spec, CP]. The non-movement analysis of sluicing in Mandarin Chinese is also
independently motivated by the fact that this language is a so-called wh-in-situ language. I
do not examine the validity of Wang’s resumptive analysis for Mandarin Chinese here as
the sole purpose here is to see why Mandarin Chinese does not provide a conclusive
counterexample to the P-Stranding Generalization. However, I do wish to note here that
the resumptive analysis cannot be extended to accommodate the Indonesian P-stranding
pattern. If Wang’s analysis were extendable to Indonesian, we predict that this language
also would make use of the resumptive pronoun strategy as in Mandarin Chinese. This
prediction is clearly incorrect, since Indonesian does not use resumptive pronouns even
under contexts where they would ameliorate island violations (Sells 1984), as shown in
examples in (14a-c) provided by Fortin (2007b: 71).
(14)a. * Apa yang Ali
what that Ali
jadi terlalu
gemuk [CP karena dia
be
chubby
too
makan apa/pro]?
because he/she eat
what
‘What did Ali get fat because he ate?’
b. * Apa yang Ali
what that Ali
jadi terlalu
gemuk
be
chubby
too
[CP
karena itu
dimakan+nya]?
because it
Pas-eat+he/she
‘What did Ali get fat because it was eaten by him?’
153
c. * Apa yang Ali
what that Ali
jadi terlalu
gemuk
be
chubby
too
[CP karena dia
because he/she
makan+nya]?
eat+he/she
‘What did Ali get fat because he ate it?’
(14a) shows that wh-movement in Indonesian shows the standard adjunct island effect; see
chapter 5 for details on the syntax and semantics of wh-questions. (14b, c) are attempts to
insert a resumptive pronoun via the clitic pronoun -nya. The fact that there is no way to save
the adjunct island violation by resumptive suggests that the resumptive analysis motivated by
Wang on Mandarin Chinese data cannot be the right approach to the superficially similar Pstranding pattern in Indonesian.
2.2.4.
Malagasy
Malagasy was the first language reported in the literature whose P-stranding pattern under
sluicing poses a potential counterexample to the P-Stranding Generalization. Potsdam
(2003) shows that P-stranding is prohibited under wh-movement but is allowed under
sluicing, as (15a, b) illustrate.
154
(15)a. * inona
ny
trano
no
mitoetra
amina
i
Rasoa?
what
the
house
Foc
live. Act
in
Rasoa
‘Which house does Rasoa live in?’
b.
mitoetra
amin’
ny
trano i Rasoa
fa adinoko
hoe inona
ilay
live.Act
in
the
house
but forget.1sg HOE what
that
trano
no
mitoetra
amin’
house
Foc
live.Act
in
Rasoa
twhich house
‘Rasoa lives in a house but I forgot which house.’
i
Rasoa
Rasoa
(Potsdam 2003: 299)
If syntactic isomorphism is the relevant condition governing the sluicing construction in
Malagasy, as in Merchant’s original generalization, we wrongly predict that the P-less
sluicing in (15b) should be ungrammatical. Potsdam claims instead that the correct
derivation for (15b) before deletion is as in (16), in which the circumstantial voice is used
to promote the object of the preposition to subject position before it undergoes whmovement.
155
(16) mitoetra
amin’
ny
trano i Rasoa
fa adinoko
live.Act
in
the
house
but forget.1sg HOE what that house
(no
itoeran’
Foc
live.Circ
i
Rasoa
Rasoa
hoe inona ilay trano
twhich house)
Rasoa
‘Rasoa lives in a house but I forgot which house (is lived in by Rasoa)’
(Postdam 2003: 300)
In his later work, Potsdam (2007) provides extensive arguments that, in Malagasy, what is
at stake is the semantic identity condition: an elided constituent must be semantically
parallel to its antecedent but does not have to be syntactically parallel. Specifically, after
establishing that wh-questions in Malagasy are pseudoclefts, Potsdam provides
grammatical examples of sluicing in this language where the underlying structure for a
sluice is a pseudocleft but the antecedent full-fledged is not. If this analysis holds, the Pstranding pattern observed above does not constitute a counterexample to the P-Stranding
Generalization, as originally stated in Merchant (2001).
As noted above, a similar analysis has been independently proposed for wh-questions in
Malay by Cole et al. (to appear). I show in section 3, however, that this analysis cannot be
correct for sluicing in Indonesian.
156
2.2.5.
Serbo-Croatian
Merchant (2001) mentions Serbo-Croatian as another non-P-stranding language that disallows Pstranding under sluicing. Stepanović (2008) provides examples to show that this language
actually does not seem to obey the P-stranding generalization but ultimately concludes that the
omission of the preposition under sluicing is due to some conditions at work at PF rather than
stranding in the sense of extraction of the nominal complement from PPs.
The original examples that Merchant reports are given in (17a-c). (The examples in
(17a, c) are from Merchant 2001: 97; the example in (17b) from Stepanović 2008: 179)
(17)a. * Kim
whom.Instr
je
govorila
Ana
sa?
is
spoken
Ana
with
‘Whom did Ana speak with?’
b.
Sa
kim
je
Ana
govorila?
with
whom
is
Ana
spoken
‘With whom did Ana speak?’
c.
Ana
je govoria sa
ekim,
ali ne
znam *(sa) kim.
Ana
is spoken with someone.Instr but Neg I.know with whom.Instr
‘Ana spoke with someone, but I don’t know who with.’
157
Stepanović starts by pointing out that the non-strandability of sa ‘with’ in (17c) is due to
an idiosyncratic property of the wh-phrase kim ‘who’. She illustrates this point with (18a-c).
(18)a.
Marko se
ponosi
Marijom
/nekom djevojkom
Marko Refl takes.pride Marija.Instr some
girl.Instr
/neĉim.
something.Instr
‘Marko is proud of Marija/some girl/something.’
b.
*(sa)
kim
se
Marko
ponosi?
with
whom.Instr
Refl
Marko
takes.pride
‘Who is Marko proud of?’
c.
(?*Sa)
kojom
djevojkom/(*?Sa)
ĉim
with
which
girl.Instr/
what.Instr Refl Marko takes.pride
with
se
Marko ponosi?
‘Which girl/What is Marko proud of?’
(Stepanović 2008: 180)
The example in (18a) shows that the verb ponositi se ‘to take pride’ selects the
instrumental direct object. The fact that the animate, instrumental wh-phrase kim ‘who’
must occur following the preposition sa ‘with’ is due to independent factors idiosyncratic
to this single wh-phrase since it is not required by the selectional properties of the relevant
158
verb, as shown in (18b). Indeed, when we have other wh-phrases such as kojom djevojkom
‘which girl’ and ĉim ‘what’, they cannot occur with the same preposition, as shown in
(18c). This consideration, thus, suggests that it is hasty to draw conclusions on the status of
Serbo-Croatian with respect to the P-Stranding Generalization. Stepanović then provides
examples as in (19a-c) showing that Serbo-Croatian in fact presents a P-stranding pattern
that superficially contradicts the generalization
(19)a. * Čega
what.Gen
je
Petar
glasao
protive?
is
Petar
voted
against
‘What did Petar vote against?’
b.
Protiv
čega
je
Petar
glasao?
against
what.Gen
is
Petar
voted
‘Against what did Petar vote?’
c.
nečega,
ali
Petar je glasao
protiv
Petar is voted
against something but
(protiv)
čega.
Neg I.know against
what
ne
znam
‘Petar voted against something, but I don’t know what.’
(Stepanović 2008: 181)
159
The contrast between (19a) and (19b) shows that the preposition protive ‘against’ cannot be
stranded under wh-questions but the grammaticality of the P-less sluice in (19c), then,
indicates that Serbo-Croatian is a counterexample to the P-Stranding Generalization.
Stepanović argues that this conclusion is not warranted since there is evidence that Pomission under sluicing is not due to the stranding of the preposition followed by TP deletion
but some phonological operations at PF, rejecting potential alternatives such as basegeneration (van Riemsdijk 1978) or cleft strategies on empirical grounds. Assuming that her
demonstration is correct in that the underlying source of sluicing in Serbo-Croatian is sluicing
(movement of a wh-phrase into [Spec, CP] followed by TP deletion), I concentrate here on
evidence presented by her that the P-omission under sluicing cannot be due to P-stranding in
the sense of extraction of the nominal complement of the P out of PPs as in Merchant (2001).
Consider the following example of sluicing that involves coordinated PPs.
160
(20) Petar je sakrio
igračku ispod
petar is hidden toy
ali
ne
znam
but
Neg I.know under
under
(ispod) koje
which
jedne stolice
i
pored
one chair.Gen and beside
stolice
i
jednog zida,
one
wall
(pored) kojeg zida.
chair.Gen and beside
which wall.Gen
‘Petar hide the toy under a chair and beside a wall, but I don’t know which chair and which wall.’
(Stepanović 2008: 183)
In this example, the antecedent PP in the full-fledged clause is a coordinated phrase with
two PP conjuncts, each headed by a different preposition. The remnant in the sluicing
clause also consists in a coordinated phrase that contains two wh-phrases without
prepositions. If P-omission in this example were due to the stranding of the preposition as
in Merchant’s (2001) deletion theory, then no derivation would yield the structure that
feeds P-omission under TP deletion “because under no current theory of movement can the
coordinated remnant phrase move as a constituent, while stranding the two Ps.” (p. 183).
What she intends by this remark can be illustrated in the following schematic derivation
for the example in (20): see Stepanović (2008: 186, 187) for arguments based on the
number of places denoted by (20) that it involves PP coordination, not CP coordination.
161
(21)
CP
…
ConjP
PP
P
ispod
PP
DP
P
jedne stolice
pored
DP
jednog zida,
??
This derivation involves extraction of two different wh-phrases into the specifier of CP but
such a derivation is not possible under the current theory of extraction even under the Acrossthe-Board Movement because it should involve extraction of the same wh-phrase. The P-less
sluice in (20), by contrast, naturally follows if the P-stranding is not due to the wh-extraction of
the complement of the PP out of PPs but rather to some other operations such as P-omission at
work at PF. Stepanović’s (2008) analysis of the P-stranding in Serbo-Croatian, therefore,
indicates that the examples in (18a-c) do not constitute a genuine counterexample to
Merchant’s Generalization. Stepanović (p. 188) speculates that P-stranding in this language “is
a post-syntactic phenomenon, occurring possibly at PF” but leaves the precise implementation
of this conclusion for future research. It is questionable, however, whether this PF P-omission
162
analysis works for Indonesian. It is known that Indonesian has the P-drop option, as shown in
(22a).7 This observation, thus, gives the impression that Stepanovic’s analysis might work for
Indonesian. However, my native language consultants report that this option is not available
for all prepositions. Thus, the free omission of prepositions such as dengan ‘with’ in the
antecedent clause results in ungrammaticality, as shown in (22b).
(22)a. Saya ingat
I
Hasan bicara (tentang) sesuatu,
remember Hasan talk about
tapi saya tidak tahu apa.
something but
I
Neg know what
‘I remember Hasan talked about something, but I don’t know what.’
b. Saya ingat
Hasan ber-dansa (*dengan)seseorang, tapi saya tidak tahu (dengan) siapa.
I remember Hasan Vz-dance with someone but I
Neg know with
who
‘I remember Hasan danced with someone, but I don’t know who.’
This contrast between (22a) and (22b), therefore, indicates that the P-drop analysis is not general
enough to accommodate the consistent P-omission pattern observed in Indonesian sluicing.
7
Thanks to an anonymous reviewer for Snippets for pointing this out and Heidi Harley (personal
communication) for useful discussion on problems with this analytical possibility.
163
2.2.6.
“Noises” in Italian, Greek, and Polish
It has been noticed in several works that appeared as replies to Merchant (2001) that the PStranding Generalization holds neither across the board within a single language nor across
languages. Merchant (2001) himself points out that there is a huge divergence among his
consultants concerning the acceptability of the preposition-less sluice in languages such as Italian
and Hebrew that uniformly prohibit P-stranding under wh-questions.8 In fact, Merchant gives
judgments of the P-less variant of sluicing constructions in only 12 out of 18 non-P-stranding
languages as totally ungrammatical; his reported judgments in the rest of the languages indicate
that TP deletion improves at least a bit an otherwise illegal P-stranding violation under whquestions. See also Tanaka (2007), who observes that the judgments reported by Merchant (2001)
lack general agreement among speakers of languages such as Greek and Polish.
2.3. Section Summary
I have reviewed Merchant’s analysis of sluicing as the product of wh-movement, followed by TP
deletion at PF. I have introduced the P-Stranding Generalization. I have examined several
languages whose P-stranding pattern under wh-questions and sluicing constructions apparently
contradict this generalization and have shown that none of these languages actually runs counter
8
Thanks to Adam Ussishkin (personal communication) for valuable discussion on Hebrew sluicing.
164
to it as the derivation of sluicing in these languages involves alternative syntactic or phonological
means other than syntactic wh-movement. Based on the current survey, it is safe to conclude that
no language seems to pose genuine counterevidence for the generalization. I have also argued
that these alternative analyses cannot be applied to the superficially similar P-stranding pattern in
Indonesian. It is against this background that the P-stranding pattern under wh-questions and
sluicing in Indonesian, as illustrated in (5a-c), becomes very important. In the next section, I
provide a variety of syntactic and morphological arguments to establish that Indonesian is the
first genuine counterexample to the P-Stranding Generalization.
3. The Syntax of Sluicing in Indonesian
This section examines the syntax of sluicing in Indonesian. I start by reviewing Fortin’s
(2007b) results concerning Indonesian sluicing with respect to tests that Merchant developed to
distinguish genuine sluicing and pseudosluicing. Though the overall result does indicate that
Indonesian sluicing is derived by wh-movement, it is not conclusive, since the tests were based
primarily on English, and there is independent evidence that some of the tests do not diagnose
the syntax of sluicing in Indonesian in the same way as they do in English. I turn to Indonesianinternal arguments based on the distribution of the question morpheme -kah discovered by Fortin
(2007b) and the lack of the complementizer yang in questions with non-nominal wh-phrases that
165
sluicing with PP remnants is derived via syntactic wh-movement followed by TP deletion, as in
English. In the course of this discussion, I also reject two analyses of Indonesian wh-questions as
reduced clefts (Cheng 1991) and headless relative clauses (Cole et al. to appear). This result
confirms that Indonesian presents the first genuine challenge to the P-Stranding Generalization.
3.1. Indonesian Sluicing = Pseudosluicing?
Merchant (2001) argues that there are two types of sluicing. Genuine sluicing is derived by
the wh-movement of a remnant in syntax followed by TP deletion. Pseudosluicing is derived
by clefting of the wh-pivot (which involves wh-movement from the cleft predicate position,
not from the complement position of a P), followed by deletion of the copula and the
expletive subject. The two types of constructions are illustrated in (23a, b).
(23)a.
b.
Pat was speaking to someone, but I don’t know [CP who Pat was talking to who].
Pat was speaking to someone, but I don’t know [CP who it was
that just left].
Merchant emphasizes the importance of the distinction between genuine sluicing and pseudosluicing as recent work on in-situ languages such as Japanese as in Nishiyama et al. (1996) and
Kizu (2005) presents evidence that sluicing constructions in Japanese are elliptical clefts.
166
Merchant argues that English has genuine sluices by developing 10 operational tests to
distinguish between pseudosluicing and genuine sluicing. The purpose of this section is to
review Fortin’s (2007b) discussion on Indonesian sluicing in this regard. Fortin applies those
tests that are applicable to Indonesian wh-clefts and wh-questions and concludes that “the
totality of the data appears to indicate that Indonesian sluices are elliptical wh-questions, and
not elliptical clefts.” (pp. 198-199). Her reported results are given in Table 2 (p. 206).
Table 2: Pseudosluicing Diagnostics in Indonesian
Attested in
Attested in sluices?
Diagnostics
wh-clefts?
Attested in
in wh-questions?
English Indonesian English Indonesian English Indonesian
adjuncts
√
√
X
X
√
√
implicit arguments
√
√
X
X
√
√
‘mention-some’
√
√
X
X
√
√
‘mention-all’
X
X
√
X
√
√
‘else’
√
X
X
X
√
√
attributive adjs
√
√
X
X
X
√
167
The first diagnostic concerns the distribution of adjuncts. Merchant (p. 121) observes that
an adjunct cannot be the pivot of a cleft (24a), but can be the remnant of a sluice (24b).
The non-elliptical wh-question patterns with the sluice in this regard, as shown in (24c).9
(24)a. * Pat is crying, but I don’t know why it is.
b.
Pat is crying, but I don’t know why.
c.
Pat is crying, but I don’t know why Pat is crying.
(Fortin 2007b: 199)
Fortin observes that Indonesian clefts are not possible with a wh-adjunct but the
corresponding sluices are grammatical, as shown in (25a, b). (25c) shows that the whquestion patterns with (25b), indicating that Indonesian sluicing is based on wh-questions.
(25)a.* Ali memperbaiki sepeda + nya, tapi saya tidak tahu bagaimana(kah) itu.
Ali AV-fix
bike+3sg
but
I
Neg know how-Q
that
‘Ali fixed his bike, but I don’t know how it was.’
9
One question, raised by Heidi Harley (personal communication), is how what Zagona (2007) analyze as
quasi-PP arguments, such as instrumental, benefactive and locational expressions, behave with respect to this
test. Because Zagona shows that this type of expression passes lots of tests for argumenthood, the prediction
would be that they should be able to occur as the pivot of the cleft construction. I leave this important
question for my future research.
168
b. Ali memperbaiki sepeda + nya, tapi saya tidak tahu bagaimana (kah).
Ali Av-fix
bike + 3sg
but
I
Neg know how-Q
‘Ali fixed his bike, but I don’t know how.’
c. Ali memperbaiki sepeda + nya, tapi saya tidak tahu bagaimana
Ali Av-fix
bike + 3sg but I
Neg know how
dia memperbaiki+nya.
3sg Av-fix-3sg
‘Ali fixed his bike, but I don’t know how he fixed it.’
(Fortin 2007b: 199, 200)
The second diagnostic concerns the distribution of sprouted implicit arguments. Chung et
al. (1995) propose that there are two types of sluicing constructions in English. The first
type, exemplified in (26a), has an overt correlate in the antecedent clause that corresponds
to the wh-remnant in the sluice. The other type, exemplified in (26b), illustrates the
sluicing construction in which the wh-remnant is licensed in the sluice without any overt
correlate in the antecedent clause based on the argument structure of the verb in the
antecedent clause. Chung et al. propose the LF structure building operation of “sprouting”
to accommodate this second pattern of sluicing whereby the extra empty category can be
constructed within the TP recycled from the antecedent TP.
169
(26) a.
b.
She bought something the other day, but I don’t remember what.
They served the guests, but I don’t know what. (Merchant 2001: 121)
Merchant (p. 121) and Fortin observe that sprouted implicit arguments are fine in sluicing and
wh-questions but not in clefts, as the contrast between (27a) and (27b, c) shows.
(27)a. * Robin was reading, but I don’t know what it was.
b.
Robin was reading, but I don’t know what.
c.
Robin was reading, but I don’t know what Robin was reading.
(Fortin 2007b: 200)
Fortin shows that this pattern is replicated in Indonesian, as shown in (28a-c). This is
another indication that sluicing in Indonesian is based on wh-questions, not wh-clefts.
(28)a. * Ali sedang
Ali Prog
memasak, tapi saya tidak tahu masakan
apa(kah)
itu.
Av-cook but I
what(-Q)
that
Neg know dish
‘Ali is cooking, but I don’t know what dish it is.’
170
b. Ali
Ali
sedang memasak, tapi
saya tidak
tahu
masakan
apa.
Prog
I
know
dish
what
Av-cook
but
Neg
‘Ali is cooking, but I don’t know what dish.’
c.
Ali
sedang memasak, tapi saya tidak tahu masakan
apa(kah)
Ali
Prog
what-Q
yang
Ali
that
Ali
Av-cook
but
I
Neg know dish
masak.
cook
‘Ali is cooking, but I don’t know what dish he is cooking.’ (Fortin 2007b: 201)
The next three tests concern different types of modification. The first of these is ‘mentionsome’ modification, such as for example. Merchant (p. 122) and Fortin observe that this
modification is possible with sluices and non-elliptical wh-questions, but impossible with whclefts, as the contrast between (29a) and (29b, c) shows. Fortin observes that Indonesian
patterns with English in this regard, as illustrated in (30a-c).
171
(29) A:
You should talk to someone in the legal department about that.
a.
B:
* Can you tell me who it is, for example?
b.
B:
Can you tell me who, for example?
c.
B:
Can you tell me who I should talk to, for example?
((12a, b) from Merchant 2001: 122, (12c) from Fortin 2007b: 201)
(30) A:
Kamu
harus
makan lebih
you
should eat
more
banyak
sayur-mayur.
many
vegetable-Red
‘You should eat more (different kinds of) vegetables.’
a.
B. # Misalnya,
apa-kah
for example what-Q
b.
B:
Misalnya,
for example
c.
B:
Misalnya,
itu?
‘for example, what is it?’
that
apa?
‘for example, what?’
what
apa
yang harus
for example what that should I
‘For example, what should I eat?’
saya
makan?
eat
(Fortin 2007b: 202)
The second of the three modification tests is a “mention-all” modification. Merchant (p. 122)
observes that this modification is fine with the cleft but not with the sluice. This is because wh-
172
clefts in English have the so-called exhaustivity requirement (Kiss 1998; Groenendijk and
Stokhof 1997). This is illustrated in (31a, b). Note that all is compatible with the non-elliptical
wh-question, as shown in (31c).
(31)a.
A bunch of students were protesting, and the FBI is trying to find out who all it was.
b. * A bunch of students were protesting, and the FBI is trying to find out who all.
c.
A bunch of students were protesting, and the FBI is trying to find out who all was protesting.
((14a, b) from Merchant 2001: 122; (14c) from Fortin 2007b: 203)
This result shows that there is no telling from the all-modification test what the source of the
sluice in Indonesian is. This is because, if all is compatible with the cleft and the full-fledged whquestion but not with the sluice, then there is an additional constraint at work in the sluice.
The Indonesian facts are also not clear in this regard. According to Fortin (p. 203), three of
her four consultants report that saja, the Indonesian equivalent of all, can modify neither
sluices nor clefts, though it can modify wh-elements in a full-fledged wh-question, as in (32a-c).
173
(32) A:
Ada
banyak
tamu
yang
mendatangi
exist
many
guest
that
Av-come-Loc party+1sg
a. B: * Tolong kasih tahu
help
give know who all-Q
b. B: * Tolong kasih
help
c. B:
siapa saja-kah itu.
give
give
‘Please, tell me who all it was.’
that
tahu siapa saja.
‘Please, tell me who all.’
know who all
Tolong kasih tahu siapa saja-kah mereka.
help
pesta+ku.
know who all-Q
‘Please, tell me who all they were.’
they
(Fortin 2007b: 203)
The contrast between (32c) and (32a, b) may be amenable to a semantic analysis independently
of the validity of the ‘mention-all’ modification as a probe into the syntax of sluicing in
Indonesian. As Rodriguez et al. (2007:10) also note for Brazilian Portuguese, Indonesian clefts
do not exhibit the exhaustivity requirement due to some construction-specific bleached
semantics. Evidence that the relevant requirement does not hold in Brazilian Portuguese or
Indonesian comes from the fact that clefts in these languages allow negative quantifiers to serve
as pivots of the cleft, a pattern that is impossible in English and Spanish. The contrast is
illustrated in (33-d). (33b, d) are from Rodriguez et al. (2007: 10).10
10
As pointed out to me by Heidi Harley (personal communication), a possibility remains that ninguém ‘nobody’
in (33a) is a case of negative concord under a negated clause, not a true negative quantifier. If so, (33a) should be
174
(33)a.
Não foi
not
ninguém que bateu
na
was nobody that knocked on.the
porta.
(Brazilian Portuguese)
door.
‘It was nobody that knocked on the door.’
b.
Tak ada
orang
yang
Neg exist person that
mengetuk
pintu.
Av-knock
door
(Indonesian)
‘It was nobody that knocked on the door.’
c. * No
not
fue
nadie
que golpeó en
was nobody that knocked on
la puerta. (Spanish)
the
door
‘It was nobody who knocked on the door.’
d. * It was nobody who knocked on the door.
(English)
The contrast between (33a, b) and (33c, d), therefore, indicates that the exhaustivity
requirement is not a universal property of wh-clefts. If this conclusion is tenable, then the
grammatical only under the specific interpretation (“Somebody knocked on the door but no one of importance”)
because the English equivalent It was not anybody who knocked on the door only allows this interpretation. I
could not test this claim due to the unavailability of native speakers of Brazilian Portuguese. Interestingly,
however, according to my three consultants, the Indonesian example in (33b) only allows a true indefinite
reading (i.e. “Nobody knocked on the door”). Thus, the negative quantifier test could still serve to show that at
least Indonesian clefts do not require exhaustivity. Nonetheless, it is a separate question whether or not the
negative quantifier test proves that Indonesian clefts forbid exhaustivity, unlike English clefts, for otherwise (32a)
would be acceptable. The crucial test case, then, would be one where a cleft in Indonesian is bad in a context
which necessarily requires exhaustivity. My consultant work so far has been unsuccessful in identifying this case,
so I must leave testing this prediction as an important task to be undertaken in my future research.
175
unacceptability of (36a) might follow from the contradictory semantic requirements imposed
by saja ‘all’ and the non-exhaustive bleached semantics of clefts in Indonesian. (32c) is not
surprising because the plural denotation of the subject of the embedded wh-question is
compatible with the exhaustivity requirement.
The last test is else-modification. Expressions such as else cannot co-occur with the whcleft but can co-occur with sluicing and wh-questions in English, as shown in (34a, b).
((34a, b) from Merchant (2001: 122); (34c) from Fortin (2007b: 204).)
(34)a. * Harry was there, but I don’t know who else it was.
b.
Harry was there, but I don’t know who else.
c.
Harry was there, but I don’t know who else was there.
The Indonesian facts do not come out in the same way as in English. lagi ‘again’ receives an
interpretation akin to English else when modifying a wh-phrase. Fortin observes that lagi cannot
modify the wh-pivot of a cleft (35a). However, it also cannot modify the wh-remnant of a sluice
(35b), even though it can modify the wh-phrase in a full-fledged wh-question (35c).
176
(35) Ali
Ali
datang
ke
pesta+ku
come
to
party+1sg
a. … * tapi saya tidak
but
I
ingat
Neg
siapa
remember who
lagi(kah)
itu.
else
that
‘…but I don’t remember who else it was.’
b. … * tapi
but
saya
tidak
ingat
siapa
lagi.
I
Neg
remember
who
else
‘…but I don’t remember who else.’
c. …
tapi
saya
but
I
tidak
Neg
ingat
siapa lagi(kah) yang datang.
remember who else-Q
‘..but I don’t remember who else came.’
that come
(Fortin 2007b: 204)
This result indicates that the lagi-modification test gives contradictory results to the other
tests. Since the preponderance of evidence goes the other way in Indonesian, other factors
seem to be at work in ruling out (35a).
The final test concerns adjectival modification. Merchant (p. 127) observes that extraction
of an attributive adjective in both wh-clefts and non-elliptical wh-questions gives rise to
177
ungrammaticality, in contrast to sluicing, which allows such an extraction. This contrast is
illustrated in (36a-c) from Fortin (2007b: 205); cf. Merchant (2001: 127).
(36)a. * I heard that Pat met a nice guy, but I don’t know how nice it is.
b.
I heard that Pat met a nice guy, but I don’t know how nice.
c. * I heard that pat met a nice guy, but I don’t know how nice Pat met a guy.
The same pattern characterizes the extraction of an attributive adjective in Indonesian, as
shown by the contrast between (37a, c) and (37b). ((20a, b) are from Fortin (2007b: 206).)
(37)a. * Saya mendengar Siti
menikahi
orang
yang
kaya, tapi
saya
Siti
Av-marry
person
that
rich but
I
tidak tahu
[se-kaya
apa
itu].
Neg know
one-rich
what
that
I
Av-hear
‘I heard Siti married a rich man, but I don’t know how rich it is.’
178
b.
Saya
mendengar
Siti
menikah
orang
yang
I
Av-hear
Siti
Av-marry person that
tapi saya
tidak
tahu
se-kaya
apa.
but
Neg
know
one-rich
what
I
kaya,
rich
‘I heard Siti married a rich man, but I don’t know how rich.’
c. * Saya
I
mendengar Siti
menikahi
Av-hear
Av-marry person that rich but
Siti
tidak tahu se-berapa
kaya-kah
Neg know one-how rich-Q
itu
orang
yang kaya, tapi
sehingga dia
that far
she
saya
I
orang.
person
‘I heard that Siti married a rich man, but I don’t know how rich she married a man’
This test requires careful elicitation. The point of this adjective modification test is whether
the example in (37a) is acceptable to Indonesian speakers when the embedded question of
the second clause asks the degree of niceness (i.e. I don’t know how rich it is that he is). This
consideration bears emphasizing because the topic of the sentence in (37a) is clearly a
human, so speakers would find it so easy to access this topic, not the degree of niceness
(which is hardly accessible even in English). With this consideration in mind, all of my three
consultants uniformly reported that (37a, c) are bad while (37b) is fine under the “degree-of-
179
nicess” reading. This result, therefore, show that the sluice can be distinguished from the
cleft in Indonesian, just as in English.
Let us summarize here what the results reported in Table 2 tell about Indonesian
sluicing. Some of the tests (i.e. adjunct/implicit remnants and ‘mention-some’ modification)
are suggestive that Indonesian sluicing is based on the non-elliptical wh-question, as in
English. Some other tests (i.e. ‘mention-all/else’ modification) do not yield clear results in
Indonesian, as they do in English. This is because there is an independent factor related to
the arguably obligatory lack of the exhaustivity requirement of Indonesian wh-clefts on the
pivot, which would be a prerequisite for all these tests to apply. The last test concerning
extraction of the attributive adjective, however, at least serves to distinguish wh-clefts and
sluices in Indonesian, as in English. Therefore, the reasonable conclusion is that although the
results here are suggestive that sluices have different syntax from wh-clefts Indonesian, they
are still conclusive as to the question of what is the underlying syntax of sluicing
constructions in Indonesian. This result is hardly surprising, however, since many of
Merchant’s tests are based on the syntax and semantics of wh-questions, sluicing, and whclefts in English which does not necessarily hold for Indonesian. For this reason, I turn in the
next section to syntactic and morphological arguments internal to Indonesian that
180
conclusively determine whether Indonesian sluices are products of elliptical clefts or regular
wh-movement plus TP deletion.
3.2. Cheng (1991): A Reduced Cleft Analysis
Cheng (1991) proposes the Clausal-Typing Hypothesis, namely, that the interrogative force
of a statement must be marked either as a Q-particle in the scopal C or via the movement of
a wh-operator into the specifier of the same C at S-structure. When combined with the
Economy of Derivation (Chomsky 1995), which prohibits superfluous steps in syntactic
derivation, the Clausal-Typing Hypothesis predicts that if a language has a Q-particle in its
lexical inventory, that language should use it for all types of wh-questions, thereby excluding
the need for a wh-phrase to undergo syntactic movement into the specifier of CP on the
grounds of the Economy of Derivation and yielding the wh-in-situ option across the board. If
a language lacks a Q-particle in its lexical inventory, on the other hand, then that language
should use syntactic wh-movement for the purpose of clause-typing. As shown in chapter 2
of this dissertation (see also chapter 5), however, Indonesian has three ways to form whquestions (overt movement, partial movement, and in-situ) and has a Q-particle, -kah. This
state of affair would appear to contradict the prediction of Cheng’s Clause-Typing
Hypothesis because, as the wh-in-situ option indicates, Indonesian has a Q-particle as a
181
clause-typing morpheme and this option should block overt wh-movement into the scopal
specifier of CP. Accordingly, Cheng (1991) argues that what appears to be an overt whmovement construction in Indonesian is actually a reduced cleft (where the expletive subject
and copula re missing), not a genuine wh-question, as in English wh-questions. According to
this analysis, the sentence in (38a) would have the structure in (38b).
(38)a.
Apai
yang
kamu
beli
what
that
you
buy
ti
?
‘What did you buy?’
b.
[CP1
apai
[CP2
Opi
yang [TP
kamu
beli
ti]]
In the structure in (38b), the wh-phrase apa ‘what’ is base-generated in the specifier of
CP1. The null operator undergoes TP-internal movement from the object of the verb into
the specifier of CP2. If this analysis is tenable, the Indonesian P-stranding pattern does not
contradict the P-Stranding Generalization.
However, there is evidence presented by Fortin (2007b) that the derivation of sluicing
involves true wh-questions, not elliptical clefts. Fortin observes that the question particle -kah
182
can co-occur with the wh-pivot of the cleft construction but not with the wh-remnant of the
sluicing construction. This is illustrated by the contrast between (39a) and (39b).
(39) Ada
exist
a.
seseorang
yang
menelpon
tadi …
someone
that
AV-phone
just now
coba
tebak
siapa-(kah)
itu!
try
guess
who-Q
Dem
‘try to guess who it was!’
b.
coba
tebak
siapa-(*kah)!
try
guess
who-Q
‘try to guess who!’
c.
coba
tebak
siapa(*kah)
yang
menelpon
tadi!
try
guess
who-Q
that
AV-phone
just now
‘try to guess who just called now!’
(Fortin 2007b: 207, 208)
Cheng argues that the ostensible cases of over wh-movement in Indonesian are all reduced
clefts. Then, if the example in (39b) were derived by the cleft construction in (39a), as it
were under Cheng’s cleft analysis, the example in (39b) should be able to allow the
183
question marker -kah to occur with the wh-phrase siapa ‘who’. Thus, the distribution of kah separates “true” clefts from genuine sluices and wh-questions. Note that this pattern
follows straightforwardly if Indonesian has true wh-movement of the English type.
Cheng’s reduced cleft analysis, on the other hand, would need some extra stipulation to
capture the distribution of the question morpheme illustrated in (39a-c).11 Based on this
consideration, I conclude that Indonesian has true wh-movement.
3.3. Cole et al. (To appear): A Headless Relative Clause Analysis
Cole et al. (to appear) argue for a different analysis of wh-questions in Malay (and Indonesian,
by extension). They propose that wh-questions with the complementizer yang and those
without have two different syntactic derivations: the former types of questions involve short
focus movement of the wh-phrase from the post-copula position to the specifier of the matrix
CP while the latter type of questions involve successive cyclic movement of the wh-phrase
from its base position to the specifier of CP as in English wh-questions: see also Davies’ (2003,
2005) proleptic analysis of wh-questions in Madurese that deny the existence of long-distance
movement. Consider (40a) and its derivation in (40b).
11
For example, Cheng’s analysis could be saved by saying that genuine clefts could reduce –kah as well as
the copula, but this would still leave unexplained why kah- must be reduced in the sluice in (39b) if the sluice
were derived through clefting by the further deletion of the expletive subject.
184
(40)a.
Apai
yang
kamu
beli
what
that
you
buy
ti
‘What did you buy?’
b.
?
(Malay/Indonesian)
CP
apai
C′
C
IP
Ø
NP
BE
NP
CP
Ø Opi
C′
ti
C
IP
Ø
yang kamu beli ti
(modified from Cole et al. to appear: 4)
In this derivation, there is TP-internal movement of the null interrogative operator from the
TP internal position to the specifier of the embedded CP. The wh-phrase apa ‘what’
undergoes focus movement from the position following the null copula (BE) to the
specifier of the matrix CP. One of their arguments for this headless relative clause analysis
of wh-questions with yang comes from the categorical restriction on interrogative elements
that can be fronted in this type of question. Consider (41a-e) and (42a-e).
185
(41)a.
Apa yang
di-perbaiki
Ali?
what that
Pv-fix
Ali
‘What was fixed by Ali?’
b.
Siapa
yang
melihat
kau?
who
that
see
you
‘Who saw you?’
c.?? Di
mana
yang
kau
tinggal?
at
where
that
you
live
‘Where do you live?’
d.?? Bagaimana
how
yang Ali
memperbaiki kereta
itu?
that Ali
fix
that
car
‘How did Ali fix that car?’
e. ?? Kenapa
why
yang
Ali
di-pecat?
that
Ali
Pass-fire
‘Why is it that Ali was fired?’
(Indonesian; modeled after the Malay examples from Cole et al. to appear: 6, 7)
186
(42)a.?? Yang
that
aku
tinggal
(ialah)
di K.L.
I
stay
is
at K.L.
‘The place that I live is in K.L.’
b.?? Yang
that
Ali
perbaiki
kereta
itu
(ialah)
Ali
fix
car
that is
dengan
alat-nya.
with
tool-his
‘The way that Ali fixed that car is with his tool.’
c.?? Yang
that
Ali
di-pecat
(ialah)
kerana dia
cuai.
Ali
Pv-fire
is
because he
careless
‘Why Ali was fired is because he was careless.’
d.
Yang
aku
makan
nasi
goreng (-lah).
that
I
eat
rice
fried-focus
‘Fried rice is what I am eating.’
e.
Yang
kau
nampak
Siti(-lah).
that
you
see
Siti-Foc
‘Siti is what you see.’
(Indonesian; modeled after the Malay examples from Cole et al. to appear: 9)
187
It is clear from the examples in (41a-e) that, when wh-questions are formed with yang,
only questions with nominal wh-phrases such as apa ‘what’ and siapa ‘who’ are wellformed. This categorical restriction would remain mysterious under the common analysis
of wh-questions in languages such as English as fronting of an interrogative phrase into the
specifier of the matrix CP, because no such restrictions would be imposed on the kind of
elements to be fronted. This observation, by contrast, directly follows if the underlying
structure of yang-questions is clefting, because the same restriction is independently
observed in cleft constructions, as shown by the examples in (42a-e). Their analysis, thus,
might mean that the underlying syntax of slucing would be a headless relative clause, an
idea that has also been argued for by Potsdam (2003, 2007) for Malagasy sluicing, as we
have seen in the previous section.
However, the headless relative clause analysis would only work for sluicing examples
with nominal wh-phrases. This point is clearly emphasized by Cole et. al’s (to appear: 26)
conclusion that “questions without yang involve potentially long distance movement of the
WH word itself.’ Consider examples in (43a, b).
188
(43)a.
Bilai
Maryam
pikir
[yang Ali akan
datang
ke
sini
when
Maryam
think
that Ali
come
to
here
will
ti]?
‘When does Miriam think that Ali will come here?’
b.
Kenapai Siti
kata
[yang
Fatimah
beli
ikan
itu
why
say
that
Fatimah
buy
fish
that
Siti
ti ]?
‘Why did Siti say that Fatimah bought that fish?’
(Indonesian; modeled after the Malay examples from Cole et al. to appear: 27)
The examples in (43a, b) show that wh-questions with non-nominal wh-elements such as bila
‘when’ and kenapa ‘why’. For Cole et al. (to appear), the lack of yang means that the
questions are derived by regular wh-movement of an interrogative phrase into the specifier
of CP. Then, the obligatory absence of yang in (41c) indicates that the wh-question with the
fronted PP cannot be analyzed as the headless relative clause because there is an independent
restriction that the nominal head of such a clause must be nominal wh-phrases such as apa
‘what’ and siapa ‘who’. Thus, I conclude that at least the PP wh-question is derived by
regular wh-movement, as in English.12
12
Note that the argument for the wh-movement analysis of sluices here is the Indonesian analogue of
Merchant’s (2001) adjunct test reviewed in section 2.1.
189
3.4.
Sluicing in Indonesian ≠ Pseudoslucing: Why Indonesian Is So Special
In this section, I have presented evidence based on the distribution of the question particle
-kah and the obligatory absence of yang in wh-questions in Indonesian that the derivational
source for sluicing in Indonesian (at least with PP remnants) cannot be clefting as in Cheng
(1991) or headless relative clause as in Cole et al. (to appear). The argument made by
Fortin (2007b) from the question particle shows that the slucing construction patterns with
a wh-question but not with a cleft counterpart. The argument from the lack of yang shows
that the derivation of sluicing with the PP remnant involves regular wh-movement of the
PP. This result, therefore, suggests that the Indonesian sluicing with the PP remnant is
derived by regular wh-movement. With this observation in place, consider again the
examples in (5a-c), repeated here as (44a-c).
(44)a. * Siapa
who
yang
kamu
ber-dansa
dengan?
that
you
Vz-dance
with
‘Whom did you dance with?’
b.
Dengan siapa
kamu
ber-dansa?
with
you
Vz-dance
who
‘With whom did you dance?’
190
c.
Saya ingat
I
Hasan
ber-dansa
dengan seseorang, tapi saya tidak
remember Hasan
Vz-dance
with
someone but I
Neg
tahu (dengan) siapa.
know (with)
who
‘I remember Hasan danced with someone, but I don’t know (with) whom.’
We have seen in section 3 that several other languages that behave superficially similar to
this pattern of P-stranding are not genuine counterexamples to the same generalization. I
have shown in this section, however, that at least the Indonesian slucing with PP remnants
is derived by regular wh-movement of the remnant followed by TP deletion. Therefore, I
conclude that the P-stranding pattern in Indonesian presents the first genuine
counterexample to the P-Stranding Generalization.
4. Salvation by Deletion at the Syntax-Phonology Interface
In this section, I propose a novel analysis of the (un-)availability of the P-stranding across
languages that draws on independently motivated assumptions. I show that the proposed analysis
provides a straightforward explanation for the three-way contrast between English, French, and
191
Indonesian with respect to P-stranding under wh-questions and sluicing. In particular, the
apparently atypical pattern of Indonesian is fully expected under the proposed analysis.
What is crucial below is the idea behind the thesis of minimalist interface, introduced in
chapter 1, that certain imperfections created by the syntactic derivation can be ameliorated
by syntax-external operations such as deletion of the offending part of the derivation. This
idea of “salvation by deletion” goes back to Ross’ (1969) seminal analysis of the
ameliorating effect of deletion on subjacency-violating movements in global terms and has
been resurrected in recent minimalist research on the syntax-phonology interface as in
Merchant (2001), Lasnik (1999, 2001, 2005, 2007), and Boeckx and Lasnik (2006). The
proposed analysis, thus, further substantiates the claim that syntax-external components do
whatever they can to save an otherwise illicit syntactic object within the narrow range of
options permitted by the interaction of universal principles and parametrically defined
options available in a particular language.
4.1. Feature Percolation, Minimality, D-to-P Incorporation, and Interface Repair
The proposed analysis adopts three independently motivated assumptions concerning the
percolation of the [+wh] feature borne by a nominal element onto its dominating PP, D-to-
192
P coalescence, and, most importantly, the notion of interface repair. Let us first discuss
each of these assumptions and how they are independently motivated.
4.1.1.
Whether the [+wh] Feature Percolates or Not
The proposed analysis claims that there is a parameter concerning the percolation of the
[+wh] feature of the interrogative D onto its dominating PP. This idea was first proposed
by Chomsky (1972) to answer a criticism raised by Postal (1972). Postal observes that, if
movement is successive-cyclic in the sense defined in chapter 2 of this dissertation, it
would predict that the preposition should be able to be stranded in any one of the
specifiers of intermediate CPs. The ungrammatical examples in (45d, e), however,
indicate that this prediction is incorrect.
(45)a.
I believe Mary thinks Joan talked to someone.
b.
Who do you believe Mary thinks Joan talked to?
c.
To whom do you believe Mary thinks Joan talked?
d. * Who/Whom do you believe to Mary thinks Joan talked?
e. * Who/whom do you believe Mary thinks to Joan talked?
(Postal 1972: 213)
193
The relevant generalization here is that prepositions in English must either be stranded in
situ or be pied-pied into the specifier of the matrix CP. Chomsky argues that this
generalization naturally falls out if we assume that the [+wh] feature of the wh-expression
can percolate onto its dominating PP in English, in the manner seen in (46a, b); see
Stepanov (2001) and Medeiros (2006a) for implementations of this idea within the phasebased framework to raising and pied-piping with inversion, respectively.
(46)a.
PP
P
b.
DP[+wh]
(no percolation)
PP [+wh]
P
DP [+wh]
(percolation)
Rephrasing Chomsky’s original analysis that employs the A-over-A Principle within the
modern minimalist framework, when the [+wh] feature does not percolate as in (46a), the
closest element from the perspective of the interrogative C is the DP. This option, thus, yields
the stranded preposition structure as illustrated in (45b). When the [+wh] feature does percolate
as in (46b), however, it is the PP now marked with that feature that is moved/attracted by the
interrogative C. This option, thus, yields the pied-piping structure as illustrated in (45c). Notice
that, under this feature-based analysis, there is no way in which the preposition can be stranded
194
in intermediate sites because the decision as to whether the relevant feature is percolated onto
PP or not is made when the derivation constructs the PP, as shown in (46a, b): once it
percolates, the shortest attract/movement requirement demands that the PP must move entirely
as it is the closest interrogative element. If it doesn’t, the same requirement demands that the
wh-phrase itself must be carried onto the specifier of the matrix CP. Thus, the feature
percolation operation receives independent empirical motivation.13
The proposed analysis also provides a natural explanation for one well-known fact about
P-stranding. It has been widely acknowledged since the single most important work on Pstranding by van Riemsdjik (1978) that it is a crosslinguistically very marked option that is
observed only in a handful of languages such as English, Frisian, Swedish, Norwegian, and
other Scandinavian languages. It is virtually unattested in Romance languages and in other
branches of Germanic, for example. Riemsdijik argues that the lack of this option in
Romance can be accommodated if we assume that these languages have the syntactically
represented marked option of projecting an escape hatch in the specifier of PPs that wh13
Alternatively, as suggested to me by Andrew Carnie (personal communication), “percolation” could be
recast as whether or not the P head has an unvalued wh-feature which agrees with the interrogative DP or not.
According to this alternative, in non-P-stranding languages like French and Indonesian, the P head has an
unvalued wh-feature which agrees with the wh-DP. Thus, the closet constituent to be attracted is the whole PP.
In contrast, in P-stranding languages, the P-head does not have this feature, hence the DP must move into
[Spec, CP], leaving the P behind. I think that this alternative would need some special mechanism to explain
why English also allows the pied-piping option because the DP would be the closest element to the C.
Therefore, I assume the feature percolation analysis here.
195
phrases can go through to end up in the specifier of CP. I will not review many empirical
problems with this analysis; see Hornstein and Weinberg (1981), Kayne (1981), Stowell
(1981, 1982), Herslund (1984), Takami (1988), Abels (2003), Law (1998, 2006), Salles
(1997), and sources cited therein. I do want to point out, however, that this analysis is not
groundable within the modern theory of phrase structure such as the Bare Phrase Structure
Theory (Fukui 1986, 1995; Speas 1990; Chomsky 1995) as nothing in this theory can block
the projection of specifier as it is automatically created by the internal merge of the NP into
[Spec, PP].14 One could technically block the projection of such a position within the Phase
Theory by not assigning EPP-features to drive the required movement but this analysis is
obviously circular unless independent evidence is provided for it.
By contrast, as Lasnik (2005) observes, Chomsky’s feature percolation analysis provides
a natural account of the unavailability of the P-stranding in Romance languages that is free
from the technical problem noted above. Let us suppose that there is a parameter with
respect to the optionality of the feature percolation; the [+wh] feature a) can percolate in
English but must percolate in Romance and other languages including Indonesian that do
not allow P-stranding under wh-questions. Under this parametric analysis, the latter type of
languages do not allow P-stranding under wh-movement because the closest element to be
14
I thank Heidi Harley (personal communication) and Simin Karimi (personal communication) for pointing this out.
196
attracted by the interrogative C is always the PP, as shown in (46b), hence the wh-phrase
would never be attracted by the same head on the ground of shortest movement; English and
other languages mentioned above allow P-stranding, however, because these languages have
the parametric option of not percolating the feature onto PPs. This parametric view of feature
percolation is also in line with the standard assumption that the values of a parameter must
be learnable from the visible cues and localized to the properties of lexical items; see also
chapter 4 for relevant discussion. This consideration, therefore, provides further independent
motivation for the feature percolation analysis. 15
4.1.2.
D-P Coalescence as D-to-P Incorporation
It is well-known that, in Romance languages, a preposition sometimes coalesces with the
following determiner element into a suppletive form. Consider (47) from French and (48) from
Italian. A list of examples of other examples of D-P coalescence in these languages is also
given below each example.
15
See also our discussion of Law’s (1998, 2006) analysis of the impossibility of P-stranding in Romance
shortly below. I argue in section 4.2 that both Law’s analysis and the minimality constraint on movement are
independently necessary to derive the three-way contrast between English, French, and Indonesian with
respect to P-stranding.
197
(47) Jean a
parlé
Jean have talked
du
sujet
le
about-the
subject the
plus
difficile.
most
difficult
(French)
‘Jean talked about the most difficult subject.’
suppletive forms:du = de le, des = de les, duquel = de lequel, à les = aux, à le = au,
desquels = de lesquels ‘of the’, à lequel = auquel, à lesquels = auxquels ‘to the’
(Law 1998: 226)
(48) Gianni ha
parlato del
Gianni have talked
about-the
soggetto
più
difficile.
subject
most
difficult
(Italian)
‘Gianni talked about the most difficult subject.’
suppletive forms: al = a il, alla = a la ‘to the’, sul = su il, sulla = su la ‘on the’, nel =
in il, nei = in i ‘in the’, del = di il, dello = di lo ‘of the’, col = con il ‘with the’
(Law 1998: 226)
In the French example given in (47), the preposition de coalesces with its following
determiner le to yield a suppletive form du. Similar observations characterize the D-P
coalescence in the Italian example in (48) ((di + il = del); see also German examples below
in (52a, b). Law (1998, 2005) (see also Beermann 1990 and van Riemsdijk 1998: 639)
198
proposes that there is a syntactic constraint on suppletion, as defined in (49), to account for
the impossibility of P-stranding under wh-movement in Romance languages.16
(49) Syntactic Constraint on Suppletion (Law 1998: 22)
Elements undergoing suppletive rules must form a syntactic unit X0.
This constraint essentially states that determiners must incorporate onto their governing
prepositions to be reanalyzed in the post-syntactic component as a suppletive element. Of
course, there are cases (e.g., de la in French) where the D-P coalescence does not occur,
but it is not likely that general syntactic operations such as D-to-P incorporation should be
constrained by this type of unpredictable morpholexical gap. Rather, a more plausible
analysis would be one in which the D-to-P incorporation occurs across the board in French
and Italian whether or not its effect is morphophonologically realized in the form of
coalescence.
16
To account for the fact that V bears agreement and tense inflection despite the lack of V-to-T movement in
English, Bobaljik (1995) proposes that V undergo Morphological Merger with T post-syntactically under
adjacency. The principle in (49), thus, might not be consistent with Bobaljik’s proposal. However, as Heidi Harley
(personal communication) notes, this depends on where the representation is checked. If the constraint in (49) is
checked before Morphological Merger, then it would be violated, but if it is checked at PF, after morphological
operations had applied, the Morphological Merger operation would have indeed created a single V + T head and
the constraint would be fulfilled. I will not pursue this matter further in this chapter.
199
The constraint in (49) provides a straightforward answer for why Romance prohibits Pstranding; once the D incorporates into the P, the N and D no longer form a constituent. As
a result, the movement of the D + N becomes impossible, as illustrated in a schematic
derivation in (50).
(50)
CP
…
PP
P+ D
D and NP no longer form a constituent
DP
tD
NP
Law’s analysis predicts that the D-P coalescence should be impossible when independent
syntactic conditions block incorporation of D onto P in the syntax. This prediction is confirmed
by the observation that the coalescence is impossible in examples shown in (51a, b) from French.
(51)a.
Je
lui
ai
demandé
I
him have asked
[CP
de le/*du
lire] ‘I asked him to read it.’
to it
read
200
b.
Nous
sommes
prêts
we
are
ready
[CP
à le/*au
faire] ‘We are ready to do it.’
to it
do
(Law 1998: 227, 228)
(51a) illustrates that coalescence does not occur between the head of the embedded
complementizer and its following clitic subject which is attached to the verb. Similarly, (51b)
illustrates that it also does not happen between the complementizer and the embedded clitic
object which is dependent on the verb. The impossibility of D-P coalescence in (51b) is what
is expected under the constraint in (49) above because the incorporation of the complement
of the verb into the C/T head à would be an instance of ex-corporation, which is generally
assumed to be impossible. 17
It is important to clarify the status of the constraint in (49). Law (1998, 2006) maintains
that this constraint is a necessary condition that must be satisfied in the syntactic
component for the D-P sequence to be reanalyzed as a suppletive form at the post-syntactic
component. van Riemsdijk (1998) shows, based on examples structurally similar but more
17
Alternatively, the impossibility of D-P coalescence in (51a, b) could be blocked by the intervention of the CP
projection between the D and P if CPs serve to demarcate possible domains for prosodic phonology, as
independently proposed in Bošković and Lasnik (2003), An (2006), Richards (2006), and Sato (in press c). See
van Riemsdijk (1998: 655) for a similar observation. For the purposes of the present discussion, however, I
simply follow Law’s exposition here.
201
complicated than examples such as (47-48), that there are cases in where D-P coalescence
are blocked even though D and P are phonologically adjacent when certain syntactic
configurations are not met. I repeat one construction in German to illustrate this point.
Consider the following examples.
(52)a.
von [DP
[D e]
[AP
of
dem
König
treu
ergeben]
theDat
king
faithfully devoted
[N Dienern]]
servant
‘of the servant that is faithfully devoted to the king’
b. * vom
of-theDat
König
true
ergebenen
Dienern.
king
faithfully
devoted
servant
‘of the servant that is faithfully devoted to the king’
(van Riemsdijk 1998: 655)
In (52a), the preposition von ‘of’ selects the DP complement. Within this complement, the
adjective ergeben ‘devoted’ governs the dative DP complement dem König ‘the king’ to its left.
If phonological adjacency were the only relevant condition that governs the D-P coalescence in
German, we would predict that the contraction of von and dem would yield the suppletive form
vom (see van Riemsdijk 1998: 653 for the basic morphophonemics of coalescence in German),
since the determiner is linearly adjacent to the preposition without any prosodic boundary such
202
as CP in between. This prediction is not borne out by the ungrammaticality of the example
shown in (52b). On the other hand, the failure of D-P coalescence in this example naturally
falls into place if we assume that independent syntactic constraints, either the Empty Category
Principle, the Head Movement Constraint or the phasehood of DPs, that blocks the head
movement from within the embedded AP to the P over the intervening null D. For example,
the Head Movement Constraint correctly blocks the required head movement on a par with
examples such as *Havei you could ti left? This observation, therefore, shows that D-to-P
coalescence has its source in the syntax, even though its morphophonolgogical effect is
realized post-syntactically in the form of suppletion. This point becomes important in section
4.2 when I propose a new analysis of P-stranding.
4.1.3.
Repair of “Imperfections” at the Syntax-Phonology Interface
The final and most important idea I pursue in this chapter is the idea that syntax makes
“mistakes”, some of which can be remedied later by syntax-external operations such as
deletion. This idea has been circulated since the late 1960s, when Ross (1969) observed
that the sluicing transformation ameliorates island-violations that would otherwise yield
ungrammatical sentences. Some original examples noted by Ross are given in (53-54) with
his own judgments indicated.
203
(53) The Complex NP Constraint
a. * She kissed a man who bit one of my friends, but Tom doesn’t realize which one of
my friends she kissed a man who bit.
b. ? She kissed a man who bit one of his friends, but Tom doesn’t realize which one of my
friends.
(Ross 1969: 276)
(54) The Sentential Subject Constraint
a. * That he’ll hire someone is possible, but I won’t divulge who that he’ll fire is possible.
b. ?? That he’ll hire someone is possible, but I won’t divulge who.
(Ross 1969: 277)
The contrast between (53a) and (53b) shows that the sentence that would be
ungrammatical due to the Complex NP Constraint is improved when sluicing deletes the
part (namely the embedded TP) that involves the violation of this constraint. Similarly, the
grammaticality of the example in (54b) shows that the Sentential Subject Constraint is also
ameliorated under sluicing. Ross argues based on this observation for the necessity of
global trans-derivational comparsion, as stated in (55).
204
(55) If a node is moved out of its island, an ungrammatical sentence will result. If the
island-forming node does not appear in surface structure, violations of lesser severity
will (in general) result. (Ross 1969: 277)
In other words, “ungrammaticality is a property not of merely deep or surface structure, or of pairs
of trees which are related by rules, but rather of derivations.” (p. 277). In contrast to Ross’s original
judgments cited above, recent researchers (e.g., Lasnik 1999, 2001, 2007, Fox and Lasnik 2003;
Boeckx and Lasnik 2006) are in agreement that sluiced versions as in (53b) and (54b) are perfect
rather than marginal. Following Chomsky (1972), Merchant (2001) proposes a revision of Ross’s
statement by arguing that sluicing ameliorates certain island violations as illustrated in (53a) and
(54a) because they constitute islands only at PF; thus, (53b) and (54b) become grammatical
because the island-violation is nullified at PF by deleting the structure that encodes such a violation.
Elaborating this point further, Boeckx and Lasnik (2006) claim (cf. Aoun et al. 1987 and
Aoun and Li 2003) that both derivational and representational constraints must be admitted into
the theory of grammar. They provide data concerning wh-island effects and superiority under
sluicing in Serbo-Croatian and resumption in Egyptian Arabic. I just review their argument based
on the examples from Serbo-Croatian, given in (56-59). (The angled bracket in (58) means that
the words are deleted/unpronounced.)
205
(56) Ivan
Ivan
a.
i
Marko
ne
znaju….
and
Marko
Neg
know.
ko
je
šta
kupio.
who
is
what
bought
‘Who bought what’
b. * šta
what
je
ko
kupio.
is
who
bought
‘What who bought’
(57)a.
b.
(Boeckx and Lasnik 2006: 152)
Somebody bought something, but…
i.
Ivan
i
Marko
ne
znaju
ko
šta.
Ivan
and
Marko
Neg
know
who
what
ko.
‘Ivan and Marko know who what.’
ii.*
Ivan
i
Marko
Ivan
and Marko
ne
znaju
šta
Neg
know
what who
‘Ivan and Marko don’t know what who.’
(Boeckx and Lasnik 2006: 152)
206
(58) * idio sam knjigu
seen am
book
koju
se
pitam
which
SE
wonder.1sg who sells
‘I saw a book which I wonder who sells.’
(59)a.
b.
ko prodaje <knjigu
koju>.
(Boeckx and Lasnik 2006: 151)
Every journalist went out today to find out who was selling a certain book….
ali
ne
znam
koju
(knjigu).
but
Neg
know
which
book
‘but I don’t know which (book).’
(Boeckx and Lasnik 2006: 152)
The examples in (56a, b) illustrate that Serbo-Croatian exhibits the superiority effect in
multiple wh-questions. The contrast in grammaticality between (57a) and (57b) shows that this
effect persists after sluicing. Interestingly, however, the wh-island effect behaves different from
superiority in this respect. Serbo-Croatian exhibits the wh-island effect, as shown in (58).
However, when sluicing is applied to eliminate the TP that contains the island violation, the
result becomes grammatical, as illustrated in (59b). Boeckx and Lasnik suggest one possible
approach to the observed difference in “reparability” between superiority and wh-island effects.
Specifically, they propose that superiority is a derivational constraint within syntax that thereby
is immune to interface operations such as deletion whereas the wh-island effect is a
207
representational constraint imposed on chains that thereby is subject to repair at the syntaxexternal system. This bipartite approach to syntactic violations is also hinted at in Almeida and
Yoshida (2007) and Lasnik (2007) in light of the P-stranding pattern in Brazilian Portuguese,
who speculate that there is more than one source of P-stranding ban, one derivational and
hence irreparable, and the other representational and hence reparable.
The survey of the literature concerning syntactic violations conducted here provides
support for the assumption that syntax could make certain mistakes, some of which external
interpretive components can remedy by domain-specific operations such as deletion. This
assumption plays an important role in the following section. We return to the issue of
derivational vs. representational theories of grammar in section 5 when we discuss Fortin’s
(2007b) analysis of Indonesian sluicing.
4.2. Towards an Etiology of P-Stranding Violations across Languages
In this section, I propose a novel, parametric analysis of the typology of P-stranding under
sluicing that draw on these independently motivated assumptions. The proposed analysis
was in Table 1, repeated here.
208
Table 1: The Parametric-Theory of P-Stranding at the Syntax-Phonology Interface
Parameters
Languages
English
[+wh] feature percolation
from the D to PP
OPTIONAL
D-to-P incorporation
in the syntax
NO
Indonesian
OBLIGATORY
NO
French
OBLIGATORY
YES
The proposed analysis claims that the three-way contrast in P-stranding possibilities under
sluicing and wh-movement among English, Indonesian, and French is naturally derived by
two simple parametric choices: a) whether or not the [+wh] feature MUST percolate onto
PPs and b) whether there is D-to-P coalescence/syntactic incorporation. English can
optionally percolate the wh-feature onto the PP and lacks syntactic incorporation of D to P.
Indonesian must percolate the feature onto the PP and lacks syntactic incorporation of D to P.
French must percolate the feature onto the PP and has syntactic incorporation of D to P.
This cross-classification of the two parameters might predict languages that optionally
percolate the feature and have syntactic incorporation of D to P, the fourth logically possible
combination of the above features. In such a language, the [+wh] feature would optionally
percolate to P, while D would mandatorily incorporate into D. Note the present parametric
analysis correctly predicts that the existence of this type of languages could not be detected,
209
because even though the non-percolation option would potentially allow the wh-phrase to be
directly accessible to the interrogative C as in English, P-stranding would be independently
blocked by the D-to-P incorporation, as in French and Italian. The only convergent derivations
would have [+wh] percolation with the D-to-P incorporation, and so in practice this grammar
would produce a P-stranding pattern indistinguishable from the French pattern.
4.2.1. English
Consider first why English allows P-stranding both under wh-movement and sluicing in
conformity with Merchant’s (2001) P-Stranding Generalization. The answer is quite
straightforward under the proposed parametric analysis of P-stranding summarized in
Table 2. English allows P-stranding under wh-movement because this language has the
option of not percolating the [+wh] feature of the nominal complement of P onto the PP.
When this option is chosen, the interrogative C attracts the closest element, namely, the
wh-phrase, onto its specifier, deriving the P-stranding configuration. This yields the
example in (3a), repeated here as (60a). When the relevant feature is percolated, then the
pied-pied counterpart of (3a) results in (3b), repeated here as (60b). English also allows Pstranding under sluicing because the preposition left behind within the PP is elided by the
deletion of the TP that contains this constituent. This yields (3c), repeated here as (60c).
210
(60)a. Who was he talking with?
b. With whom was he talking?
c. Peter was talking with someone, but I don’t know (with) who. (Merchant 2001: 92)
4.2.2. Indonesian
Consider now the Indonesian P-stranding paradigm. I have shown in the present chapter
that Indonesian allows P-stranding under sluicing but nonetheless disallows this option
under regular wh-movement, in violation of Merchant’s P-Stranding Generalization.
Consider the examples in (5a-c), repeated here as (61a-c).
(61)a. * Siapa
who
yang
kamu
ber-dansa
dengan?
that
you
Vz-dance
with
‘Whom did you dance with?’
b.
Dengan siapa
kamu
ber-dansa?
with
you
Vz-dance
who
‘With whom did you dance?’
211
c.
Saya ingat
I
Hasan
ber-dansa
dengan seseorang, tapi saya tidak
remember Hasan
Vz-dance
with
someone but I
Neg
tahu (dengan) siapa.
know (with)
who
‘I remember Hasan danced with someone, but I don’t know (with) whom.’
Indonesian does not allow P-stranding since the [+wh] feature of the nominal complement
of P must percolate onto the PP. As a result, the PP, which is closest to the interrogative C,
is attracted to its specifier. The P-stranding example in (61a) is thus deemed
ungrammatical. The question is, then, why P-stranding does not yield ungrammaticality
under sluicing, as illustrated in (61c).
It is at this point that the role of the syntax-external phonological system plays an
important role in remedying imperfections created by syntactic computation. Consider the
schematic derivation in (62) for the P-less sluice in (61c).
212
(62)a. Syntax-Phonology Interface (no repair)
b. Syntax-Phonology Interface (repair)
CP
CP
NP
siapa
C′
C [+wh]
NP
…
siapa
C′
C [+wh]
…
PP
P
dengan
PP [+wh]
tNP [+wh]
P
tNP [+wh]
dengan
Let us propose that a failure of the [+wh] feature to percolate is repaired at the syntaxphonology interface and that a representational constraint to verify percolation rules out
the offending PP at the interface. 18 If the offending PP remains at PF, the representational
constraint is violated, as shown in (62a). If the offending PP is deleted at the interface as
shown in (62b), however, the representational constraint has nothing to apply to. Thus, the
failure of percolation is repaired. One could, of course, think of an alternative formulation
of the Indonesian data where what is repaired is minimality violations; when the NP moves
to [Spec, CP], this movement violates the minimality constraint because the PP is a closer
18
I thank Heidi Harley (personal communication) for suggesting this idea and useful discussion.
213
element to C than the NP. Under this alternative, the derivation of the Indonesian facts
would be as in (63a, b).
(63)a. Syntax-Phonology Interface (no repair)
b. Syntax-Phonology Interface (repair)
CP
CP
NP
siapa
C′
C [+wh]
NP
…
siapa
C′
C [+wh]
…
PP [+wh] *
P
dengan
tNP [+wh]
PP [+wh] *
P
tNP [+wh]
dengan
Within the first analysis, the representational constraint to verify percolation is violated but is
repaired at the interface. Within the second analysis, minimality violations occur but are repaired
at the interface. There is a number of ways to choose between the two analyses. First, if
minimality is a derivational constraint and if derivational constraints are syntax-internal and
inviolable, then the failure-of-percolation approach to interface repair is empirically superior to
the minimality-based approach because the minimality-violating movement should not be able to
occur in the first place. It is true that a growing body of work in recent minimalist research, as
214
mentioned in section 4.1, accumulates evidence that “true” minimality violations such as
subjacency violations are in principle tolerable within syntax and only representationally
illegitimate at the interface. But, crucially, the violation caused in (63a, b) seems to be of a
different sort than subjacency; rather, it is a superiority-type/A-over-A-type violation (Chomsky
1964). As we saw earlier, Boeckx and Lasnik (2006) claim that superiority violations are
inviolable derivational constraints within syntax. Thus, the alternative analysis of P-stranding in
Indonesian would lead to the conclusion that superiority violations behave as representational
constraints in Indonesian but behave as derivational constraints in Serbo-Croatian, which is a
highly undesirable conclusion in the view of learnability. Second, for the minimality-based
analysis of Indonesian to go through, it must be the case that feature percolation is feature
copying; even after the percolation takes place, the [+wh] feature should still remain on the whphrase, for otherwise minimality violations would not occur in the first place. This original
formulation of percolation (percolation-as-copying approach) was illustrated earlier in (46b),
repeated here as (64a). True, Lieber (1980) and Williams (1981) provide cases based on the
alternation of strong verbs such as stand-stood and withstand-withstood in which feature
percolation could be construed as feature copying in morphology, it is not clear whether
comparable phenomena exists in syntax to support this copying mechanism. This problem,
however, does not arise if it is a failure to percolate that is repaired,; feature copying is not
215
necessary component of percolation. This alternative idea (percolation-as-percolation approach),
which I adopt, is illustrated in (64b).
(64)a. Percolation as Copying Approach b. Percolation as Percolation Approach
PP [+wh]
P
PP [+wh]
DP [+wh]
P
DP
These two considerations, therefore, suggest that the failure-of-percolation analysis is superior to
the minimality-based analysis on both empirical and conceptual grounds.
The proposed repair-based analysis of the Indonesian P-stranding pattern is quite close in
spirit to Lasnik’s (1995, 1999, 2001) analysis of pseudogapping that also draws on the idea
of “repair-by-deletion”; see section 4.3 for further discussion of his analysis. Lasnik argues
that the overt V raising to v that is obligatory under non-elliptical contexts in English does
not happen in pseudogapping constructions in English precisely because the constituent
(namely, VP) that contains the violation is eliminated by PF deletion. 19 Therefore, the
19
As Heidi Harley (personal communication) pointed out, the analogy here is not completely parallel because what is
repaired at the PF interface is different between the two analyses; it is failure of feature percolation that is repaired
within the present analysis whereas it is the defectiveness of verbs (i.e. verbs with scattered features) that is repaired
within Lasnik’s analysis. See Chomsky (1995) and Lasnik (1995, 1999, 2001) for details on why verbs with scattered
feature cause crash at the interface.
216
present analysis provides a natural explanation for the apparently atypical P-stranding pattern
under sluicing in Indonesian from the interaction of independently motivated assumptions
concerning feature percolation and interface repair strategies.
4.2.3. French
Let us finally consider why French does not allow P-stranding under wh-movement or
sluicing. The examples that illustrate this pattern are repeated here as (65a-c) from (4a-c).
(65)a. * Qui
who
est-ce
qu’
Q
elle
l’a
offert
à?
she
it-has
offered
to
‘Whom has she offered it to?’
b.
À
qui
l’a-te-ell
offert ?
to
whom
it-has-she offered
‘To whom has she offered it?’
c.
Anne
l’a
offert
à quelqu’un, mais
Anne
it-has offered to someone
but
je ne
sais pas *(à) qui.
I Neg know not to whom
‘Anne has offered it to someone, but I don’t know (to) whom.’
217
French does not permit P-stranding under regular wh-movement as in Indonesian because
the [+wh] feature obligatorily percolates onto the PP that dominates the wh-phrase. As
shown in the example in (65c), French also does not allow P-less sluices. What is crucial
here is that languages such as French have D-to-P syntactic incorporation, as we saw earlier
in section 4.1. Consider the derivation in (66a, b) for the unacceptable P-less sluice in (65c).
(66)a.
Syntax
b.
CP
PF
CP
C′
C′
C [+wh]
…
C [+wh]
…
PP [+wh] *
P [+wh]
P [+wh]
à
D i [+wh] ti
PP [+wh]*
DP
P [+wh]
N
P [+wh]
Di[+wh] ti
DP
N
à
In the derivation in (66a), the D head undergoes syntactic incorporation into the P in
accordance with Law’s (1998, 2006) constraint given in (49). The [+wh] feature of the
moved D head percolates into the dominating PP, which is thereby now marked with the
218
same feature. This derivation crashes, because, when the interrogative C with the [+wh]
feature attracts the element with the matching feature, the DP is no longer a constituent that
includes the D, hence cannot be attracted by C. Notice that the P-less sluice could be
potentially derived if the D head underwent excorporation to be attracted by the C head.
However, this possibility is blocked because the excorporation would cause both the
minimality violation on attraction (because the PP is closer to C than the D head) and the
ECP-like violation (because the trace of the excorporating element cannot be properly
licensed). The point here is that whatever derivation would possibly yield the P-stranding
sluice in French crashes because of the interaction of independently motivated syntactic
constraints on D-to-P incorporation. Thus, when the derivation in (66a) reaches the
phonological component, as shown in (66b), it is too late to repair violations associated with
D-to-P incorporation within this derivation by deletion because the violations are within
syntax. This pattern is different from that in Indonesian because this language does not have
D-to-P incorporation, as the lack of D-to-P coalescence shows.
It is clear, then, that the notion of interface repair by way of deletion plays a crucial role
in the proposed account of the three-way contrast between English, Indonesian, and French
with respect to P-stranding. The most important point of the proposed analysis, therefore,
is that the syntax-external phonological component can repair certain illicit configurations
219
created in syntax by deleting them but not all configurations: it cannot undo mistakes
concerning the D-to-P coalescence that are syntactically derivationally conditioned. The
present analysis, therefore, provides powerful support for the idea behind the thesis of
minimalist interface, namely, that interface components can conduct domain-specific
operations they avail of to repair certain syntactic imperfections but only within the
parametrically defined curve set by syntax.
4.3. Pseudogapping in Indonesian and French
The proposed analysis makes certain interesting predictions regarding pseudogapping in
Indonesian and French. The proposed analysis argues that the P-stranding is tolerated in
Indonesian only under sluicing because the offending part of syntactic structure (namely,
the PP that records the violation of the minimality constraint on movement) is removed by
deleting it at the syntax-external phonological component. The proposed analysis, thus,
makes a prediction that not only deletion of TP but also deletion of smaller constituents
than TP would also have the ameliorating effect.20 One construction in point in Indonesian
is pseudogapping. Following the analysis of pseudogapping in English proposed by Lasnik
(1995, 1999, 2001) (see also Jayaseelan 1990 and Merchant 2008), let us assume that the
20
Thanks to Heidi Harley (personal communication) for pointing this out.
220
focused element undergoes movement into the specifier of a higher projection FP (such as
AgroP in Lasnik’s analysis or the vP-internal focus position) that dominates the VP: a
pseduogapping construction arises when the VP remnant is deleted at PF. The derivation
of Indonesian pseudogapping examples with the focused object DP as in (67a) is shown in
(68), ignoring the structure above FP that would contain the TP whose specifier the
external argument Hasan moves and the split vP configuration. (67b) shows that the PP
remnant in pseudogapping is also grammatical. 21
(67)a. Esti ber-danca dengan Fernando dan Hasan [DP Rifi ] i [VP ber-danca dengan ti].
Esti Vz-dance with
Fernando and Hasan
Rifi
‘Esti danced with Fernando, and Hasan (danced with) Rifi.’
b. Esti ber-danca dengan Fernando dan Hasan [PP dengan Rifi]i [VP ber-danca ti].
Esti Vz-dance with
Fernando and Hasan
with
Rifi
‘Esti danced with Fernando, and Hasan (danced) with Rifi.’
21
The nature of the functional projection into whose specifier the remnant moves does not matter for the
purposes of the present discussion. It might be AgrOP, AspP, or vP-internal FocP. Thus, I notate it here as FP.
221
(68)
FP
DP
Rifi
F′
F
VP
V
ber-dansa
PP
P
tDP [+F]
dengan
In this derivation, the object of the preposition undergoes movement/object shift into the
specifier of the FP. The VP portion of this structure undergoes deletion at PF. This
sequence of operations yields the P-less pseudogapping construction in (67a). Note that
this derivation yields the P-stranding violation that is active in Indonesian. To achieve
this result, let us suppose that there is a feature such as [+focus] that must obligatorily
percolate from the focused NP onto the dominating PP and violation of this condition is
recoded in the PP that dominates the launching site of the movement, as argued
independently by Lasnik (1999, 2001). Then, the movement illustrated in (68) violates
the representational interface constraint that detects the failure of percolation if no
deletion of the offending PP applies. The grammaticality of the example in (67a), then,
222
clearly indicates that the relevant violation can be repaired if constituent that contain the
PP (namely, VP) is eliminated by deletion at the phonological component. The
acceptability of the pseudogapping without the preposition, thus, provides further
confirmation for the proposed repair-based approach to Indonesian P-stranding.22
Note that the current analysis predicts that P-less pseudogapping should be
ungrammatical in French because it has the D-to-P syntactic incorporation that cannot be
ameliorated by interface repair. This prediction is indeed borne out by the contrast
between (69a) and (69b).23
(69) a. * Jean
Jean
a
danca
avec Marie et Robert [DP Suzanne]i [VP danse avec ti].
has
danced with Marie and Robert
Suzanne
‘Jean has danced with Marie, and Robert (danced with) Suzanne.’
b.
Jean a danca
avec Marie et
Robert [PP avec Suzanne]i [VP danse ti].
Jean has danced with Marie and Robert
with Suzanne
‘Jean has danced with Marie, and Robert (danced) with Suzanne.’
22
This analysis predicts that languages which allow P-stranding under pseudogapping should also behave like
Indonesian with respect to sluicing. Javanese is certainly one language of this type. Whether this prediction holds also
true for other Austronesian languages closely related to Indonesian (Madurese, Balinese, etc.) is an important task to be
undertaken in my future research.
23
Thanks to Summaya Racy (personal communication) for her help in constructing the data in (69a, b).
223
I take this difference in P-stranding under pseudogapping between Indonesian and French
to provide further confirmation for the proposed parametric theory of P-stranding.
5. Comparison of the Proposed Analysis with Fortin’s (2007b) Analysis
In this section, I compare the proposed analysis of the P-stranding pattern in Indonesian
with the most recent alternative presented by Fortin (2007b). As mentioned in the first
section of this chapter, Fortin shows with me that the P-stranding pattern in Indonesian
presents a genuine counterexample to Merchant’s P-Stranding Generalization. I show,
however, that there are several serious empirical shortcomings within her analysis that are
successfully resolved under the present interface-based approach to P-stranding.
5.1.
Fortin’s (2007b) LF Copy + Long Distance Agree Analysis
Fortin’s (2007b) analysis follows the tradition of the LF Copy theory of sluicing as
originally proposed by Chung et al. (1995) but presents a refinement of their original
theory within the Minimalist Program. She argues that the sluice is a deficient syntactic
structure with no TP. The wh-remnant is base-generated in the specifier of CP, with the
sluiced clause being supplied semantic content by copying the antecedent TP into the TP
224
part of the sluice via sideward movement in the sense of Nunes (2004).24 To illustrate the
specifics of her LF-Copy analysis, consider the syntactic derivation of the example in (70)
that involves a DP-remnant, as shown in (71).
(70) Pak Guru
ber-bicara dengan seseorang, tapi saya tidak
Mr. teacher Vz-speak
with
someone
but
‘Pak Guru spoke to someone, but I don’t know who.’
24
I
Neg
tahu siapa.
know who
(Fortin 2007b: 326)
There are several assumptions made by Fortin such as her treatment of sprouted arguments. Since those
assumptions are not germane to the discussion, I just mention only those assumptions relevant to P-stranding.
225
(71)
siapa
[+Q]
[+wh]
CP
C′
C
[uQ]
TP
Pak Guru
T′
vP
T[Past]
Pak Guru
v′
Agree
ber-bicara
VP
ber-bicara
dengan
PP
DP[+wh] (no feature
percolation)
seseorang [-Q]
(Fortin 2007b: 327)
In this derivation, the wh-phrase siapa ‘who’ is base-generated in the specifier of the CP.
The TP is reconstructed by copying the antecedent TP into the derivation via sideward
movement. The probe C, then, agrees with the NP seseorang ‘someone’. Fortin assumes
that there is a three-way matching relation between the wh-phrase in [Spec, CP], the
interrogative Q, and the indefinite NP, such that the syntactic category of the wh-phrase
delimits the syntactic category of the goal NP that C agrees with. In the derivation above,
226
the syntactic category of the indefinite matches with the wh-phrase in [Spec, CP] because
the [+wh] feature does not percolate onto PP. The P-less sluice in examples as in (70), thus,
can be naturally accommodated under her analysis.
Consider now the derivation of the sluice with the PP remnant in (72), as shown in (73).
(72) Pak Guru
ber-bicara [PP dengan seseorang], tapi saya tidak tahu dengan siapa.
Mr. teacher Vz-talk
with
someone but I
Neg know with
who
‘Pak Guru spoke to someone, but I don’t know with who.’
(slightly modified from Fortin 2007b: 327)
(73)
CP
PP
[+wh] [+Q]
C′
C
dengan siapa [uQ]
TP
Pak Guru
T′
T[Past]
vP
Pak Guru
v′
ber-bicara
ber-bicara
Agree
VP
PP [-Q] (with feature
percolation)
227
In this derivation, the prepositional wh-phrase dengan siapa ‘with whom’ is base-generated
in the specifier of the CP, with the empty TP being supplied with semantic content by
copying the antecedent TP into the empty TP of the sluice. Again, the three-way matching
requirement between the interrogative PP, the goal C, and the indefinite PP in the recycled
TP requires that the correlate match the interrogative PP in syntactic category. This result is
achieved here by the percolation of the [+wh] feature onto the PP, as also assumed in the
interface-based approach proposed in the previous section, so that the whole PP becomes the
closest goal from the perspective of the probe C.
5.2. Four Problems with Fortin’s (2007b) Analysis
There are three serious empirical difficulties, some of them noted by Fortin herself, which
show that her LF Copy + Agree-based analysis is not the right way to understand P-stranding
under sluicing. The first problem concerns the Case checking of the wh-phrase base-generated
in [Spec, CP], which has been shown to be a perennial difficulty for the general LF Copy
Theory of sluicing as in Chung et al.(1995) by Merchant (2001: 151). Under the standard
minimalist assumption that the Case feature of a wh-phrase is checked via Agree with an
appropriate functional head such as v and T, Fortin’s analysis would not be able to provide a
natural explanation for how the Case of the base-generated wh-phrase can be checked. One
228
could get around this problem by introducing new mechanisms such as Case Transmission but
that is clearly an ad hoc assumption unless it is given independent motivation.
Second, Ross (1969) and Merchant (2001) observe that, in many languages with rich
case morphology, the wh-remnant in a sluice is marked for the case that it would bear in a
corresponding, non-elliptical wh-question. As Ross originally notes, this case-matching
effect receives a straightforward account if the sluicing is derived from the regular whquestion, as in the movement + deletion approach adopted in this chapter. It is not clear,
however, whether Fortin’s analysis derives this result as naturally as the deletion analysis.
The third problem concerns the percolation of the [+wh] feature of the complement of a
preposition onto the PP. As in the proposed analysis, Fortin assumes, following Chomsky
(1972) and Stepanov (2001), that the [+wh] feature may or may not percolate depending
on languages. As is clear from the derivations in (71) and (73) above, she assumes that the
feature percolation is optional in Indonesian, a view that the proposed analysis does not
support, as shown in Table 3. However, Fortin’s assumption overgenerates. For example,
in the derivation in (71), the wh-feature does not percolate whereas in the derivation in (73),
it does. The question here is what prevents the percolation in (71) but not in (73). To put
the same question in a different way, if the feature percolation were truly optional in
Indonesian, we would predict that the P-stranding should be acceptable under non-
229
elliptical wh-questions in Indonesian because nothing in Fortin’s analysis seems to
necessitate the percolation. The present analysis, by contrast, is sufficiently constrained in
this regard: the [+wh] feature always percolates onto PP in Indonesian whether the
construction in question is a non-elliptical wh-question, sluicing, or pseudogapping.
Based on these four considerations, I conclude that the present interface-based approach
is superior to Fortin’s LF Copy + Agree-based approach on empirical grounds.
5.3. Derivationalism vs. Representationalism
It is interesting in this connection to think about Fortin’s primary motivation for pursuing
the LF Copy analysis explicated above. It comes from the idea, as expressed by Chomsky
(1995, 2000, 2001, 2004, 2005), that the shortest movement /minimal search requirement
is included in the definition of Move/Attract: movement of an element to some position
over some other closer position, where closeness is defined in terms of asymmetric ccommand, is simply impossible in syntactic computation. Note that this idea is in striking
contrast with the idea behind Merchant’s (2001) repair theory of PF island violations (see
also Lasnik 1999, 2001, 2005, 2007). This is because, for a PF island to be ameliorated at
all, the syntax should in principle allow violations of the minimality constraint. Indeed, this
kind of representational view of subjacency violation has been adopted since Ross (1969)
230
and Chomsky (1972) and still around in recent work as in Lasnik and Saito (1992),
Chomsky and Lasnik (1993), Merchant (2001), and Lasnik (1999, 2001, 2005, 2007). The
amelioration-based account of the island effects, therefore, seems to show that subjacency
violations are tolerable in syntax, in contrast to Chomsky’s own idea that syntactic
computation embedded shortest move/minimal search into the definition of Move.
Nonetheless, the proposed account could still be formulated in derivational terms. For
example, Richards (2001) proposes the Principle of Minimal Compliance; once syntactic
constraints such as subjacency are satisfied by movement in a certain stage of the
derivation, violations of the same constraints in later stages of the same derivation can be
tolerated.25 Thus, whether subjacency should be best formulated in (fully) representational
or (strictly/partially) derivational terms is far from settled. See chapter 7 for further
discussion on this debate.
The debate between derivational and representational theories of syntax has been active
since the advent of generative enterprise and teasing apart two theories on empirical grounds
has proven quite subtle, depending on complex considerations of extremely intricate data. To
the extent that the proposed analysis of the typology of P-stranding is tenable, however, our
25
Thanks to Andrew Carnie (personal communication) for directing my attention to Richards’ (2001)
Principle of Minimal Compliance in this context.
231
current investigation indicates that certain violations such as subjacency, minimality, etc. are
tolerated and repaired at the interface whereas other violations such as the failure of the
syntactic D-to-P incorporation, superiority violations, etc. are not. Thus, it seems that both
derivational and representational constraints are needed to describe and explain the full range
of data in natural language. See also Aoun and Li (2003) for much relevant discussion.26
6. Conclusions
Let us summarize the content of this chapter and consider the implications of the idea
proposed here for the syntax-phonology interface and the thesis of minimalist interfaces.
6.1. Chapter Summary
In the current chapter, I have observed that the P-stranding pattern in Indonesian presents
the first genuine challenge to the P-Stranding Generalization established by Merchant
(2001) in support of his movement + TP deletion analysis of sluicing. I have presented
arguments based on the distribution of the question particle -kah and the obligatory lack of
the complementizer yang in non-nominal wh-questions that the source of PP-sluicing in
this language is via regular wh-movement, as in English, rejecting alternative accounts of
26
Thanks to Simin Karimi (personal communication) for useful discussion on the debate and bringing my
attention to the work by Aoun and Li (2003).
232
wh-questions as reduced clefts as in Cheng (1991) or headless relative clauses as in Cole et
al. (to appear). This observation is important because, to the extent that it is correct, the Pstranding pattern in Indonesian provides the first genuine counterexample to the PStranding Generalization.
Based on this result, I develop a new theory of the typology of P-Stranding that
crucially relies on the recent idea of interface repair, coupled with two independently
motivated assumptions concerning the percolation of the [+wh] feature of the nominal
complement of the preposition onto the PP and the D-to-P coalescence as the result of
syntactic D-to-P incorporation. I have demonstrated that the three-way contrast in Pstranding between English, Indonesian, and French receives a straightforward
explanation in a way that the apparently atypical pattern of Indonesian sluicing is
naturally accommodated. I have also shown that the current analysis makes a correct
prediction that deletion of smaller categories than TP should also ameliorate P-stranding
violations in Indonesian but not in French. I have shown that this prediction is verified
by the grammaticality of the pseudogapping construction with the DP remnant in
Indonesian and the ungrammaticality of the corresponding construction in French.
I have compared the proposed analysis with the most recent alternative LF copy
analysis of P-stranding pattern in Indonesian by Fortin (2007b) and shown that there are
233
several empirical shortcomings concerning the case-matching effects documented in Ross
(1969) and Merchant (2001), the Case checking of the wh-remnant base-generated in the
specifier of CP, the optionality of the [+wh] feature percolation, and the effects of the
morphophonology of certain prepositions on the reparability of the P-less sluice, that are
successfully resolved under the current analysis.
6.2.
Implications of Indonesian Sluicing for the Theory of the Syntax-Phonology Interface
The proposed analysis of sluicing and P-stranding in this chapter provides strong
evidence that syntax could make certain derivational mistakes and some of them can be
repaired by operations such as deletion at the syntax-external phonological component. It
is important that only certain mistakes such as failure of percolation and subjacency-type
violations can be repaired; some others such as the failure of the syntactic D-to-P
incorporation cannot.
One important question is why the PF component is not omnipotent in that it can repair
both the failure of D-to-P incorporation and the minimality violation. I believe that one answer
to this question is available once we take seriously the interaction between syntactic
computation, the PF computation and the A-P system. The representation that the A-P system
receives from the PF interface would not contain information about syntactic violations of
234
minimality, since the A-P cannot read such language-specific information; it would be simply
a string of words as in #I#wonder#what#who#bought, with accompanied information on
intonation and stress. Then, violations such as failure of feature percolation/subjacency must
be repairable by deleting relevant information at the PF interface before they reach the A-P
system. On the other hand, the failure of the D-to-P incorporation often has repercussions in
the form of coalescence (as in Romance). Thus, the effects of such a syntactic operation are
easily detectable in pronunciation and orthography of the D-P combination, and hence relevant
to the working of the A-P system; failure to apply this operation in syntax necessarily turns out
to cause failure in the proper communication between the phonological component and the AP system. Then, the failure of D-to-P incorporation should not be repairable at the PF interface
since coalescence and cliticization, which must be reflected at the A-P system, are syntactically
conditioned.
The analysis proposed in this chapter is one clear demonstration of how the thesis of
minimalist interface works; interface components do whatever domain-specific operations
they avail of to make them legible/usable for the language-independent sound system (A-P
system in Chomsky’s terms) but only within the range of options that is parametrically set
by a particular language. I discuss further ramifications of the repair-based theory of Pstranding in chapter 7.
235
CHAPTER 4 THE MORPHOSYNTAX OF BARE NOMINALS IN INDONESIAN AND
JAVANESE: A RELATIVIZED PARAMETRIC THEORY OF NOMINAL DENOTATION 1
1. Introduction
This chapter discusses the issue of syntax-semantics interface with a case study in the
relation between the denotation and morphosyntax of bare nominals in Indonesian and
Javanese. Though the first part of this chapter is devoted to the descriptive analysis of bare
nouns in these two languages, it ultimately becomes clear in the rest of the chapter that the
primary goal of this chapter is to construct a universal theory of the denotation of bare nouns
within the Principles & Parameters approach to language variation (Chomsky 1986a, 1995).
In a series of his recent work (Chierchia 1998a, 1998b), Gennaro Chierchia proposes a
semantic parameter, known as the Nominal Mapping Parameter, which states that
languages differ in terms of what they allow their bare nouns to denote in the syntaxsemantics mapping, namely kinds ([+arg, -pred] languages), properties ([-arg, +pred]
languages), or both ([+arg, +pred] languages) under certain conditions. Chierchia argues
that setting of this parameter uniquely determines the morphosyntactic profile of bare
1
This chapter is a substantially modified version of the paper read at the Inaugural Meeting of the Arizona
Linguistics Circle (ALC1) held at the University of Arizona, Tucson and the Mid-America Linguistic
Conference (MALC) held at the University of Kansas, Lawrence. The papers read at those two meetings are
to appear in Sato (in press d, forthcoming b).
236
nominals in a particular language with respect to the availability of bare arguments, the
generalized classifier system, and plural morphology. This claim, therefore, includes a
strong statement that all natural languages can be classified as one of three and only three
languages with respect to the denotation of bare nominals.
In this chapter, I provide arguments, modeled after Chung (2000), that Indonesian and
Javanese do not fit into any one of the three language types identified under Chierchia’s semantic
typology. I further argue that the very notion of “semantic parameter” as varying the nature of the
mapping between the syntactic and semantic representation of a bare noun across languages is
hardly groundable within the standard generative conception of the locus of parameter as
restricted to properties of the lexicon (Chomsky 1995) or, more precisely, properties of
functional categories (Borer 1984; Fukui 1986, 1995).
Based on these results, I propose an alternative, relativized parametric theory of the
effects of the Nominal Mapping Parameter, whereby languages differ in terms of the
complexity
of
nominal
functional
structures
each
language
allows
(i.e.
DP>QP>ClP>NumP>NP) and in terms of the possible set of values the Num Head in each
language can take (i.e. {singular, plural} or {neutral, plural}). Neither of these ideas is
anything new. The idea that languages differ in terms of the complexity of nominal
structure has been already proposed on various grounds in recent work as in Grimshaw
237
(1991, 2005), Massam (2001), Guilfoyle and Noonan (1992), and Vainikka (1993/1994).
The observation that languages differ in their possible set of Num values has been recently
made in extensive studies of the number system in languages such as Malay and
Indonesian as in Carson (2000) and Chung (2000). The results achieved in this chapter,
therefore, provide further evidence for these proposals.
Given the Subset Principle of Wexler and Manzini (1987), the proposed parametric
theory of bare nominals makes an interesting prediction that early acquisitional stages of all
languages should look like Indonesian and Javanese since these languages instantiate the
simplest nominal structure (NumP) under the proposed theory. I show that this prediction is
borne out by the so-called telegraphic speech produced by English-learning children, at
which English nouns share several fundamentally morphosyntactic properties with bare
nouns in Indonesian and Javanese. The proposed analysis also sheds light on the recent
debate on the nominal structure of Slavic languages such as Russian. Our discussion of the
morphosyntax of bare nouns in Russian shows that bare nouns in this language only project
up to NumP as do Indonesian and Javanese bare nouns and leads us to conclude that the
universal DP hypothesis for article-less languages is incorrect.2
2
The proposed parametric theory of nominal denotation contrasts sharply with the proposal made by Gil
(2005). Gil argues that Riau Indonesian, the variety of Indonesian spoken by local people in east-central
Sumatra, has no distinction between lexical and functional categories and capture its many correlates of a
verb-initial language by the interaction of the single principle (heads precede modifiers) with iconicity and
238
I also discuss the broader theoretical consequences of the proposed analysis of bare
nominals across languages by situating it within the context of the syntax-semantics
interface, a central subject matter of this dissertation. In particular, empircal findings in this
chapter provide strong support for the notion of “minimalist interfaces”, whereby the
syntactic computation provides a parametrically defined curve that the universal semantics
interface blindly follows, without any extrinsically determined mapping between the
syntax and semantics of a particular expression as in Chierchia’s (1998a, b) Nominal
Mapping Parameter.
The present chapter is organized in the following manner. In the next section, I review
in detail the idea behind the Nominal Mapping Parameter proposed by Chierchia (1998a,
b). In section 3, I turn to the denotation and morphosyntax of bare nominals in Indonesian
and Javanese. The purpose of this section is empirical. I provide evidence that neither of
these two languages fits into any one of the three language types under Chierchia’s
semantic classification. I discuss Chung’s (2000) analysis of Indonesian bare nouns in
information flow. It is not clear whether Gil’s radical analysis applies to the dialect of Indonesian discussed in
this dissertation. Although this chapter argues that this dialect represents the simple nominal functional
structure (NumP) among languages of the world, it is an SVO language, unlike many dialects in Sumatra
such as Riau Indonesian and Toba Batak, as argued for convincingly in Chung (to appear). Thus, whether
Gil’s analysis could even apply to the SVO dialect under investigation is far from clear and needs extensive
discussion that goes beyond the scope of this chapter. Thanks to Andrew Carnie (personal communication)
for directing my attention to Gil (2005).
239
section 3.1. Though Indonesian allows bare arguments and has plural morphology, I argue,
against Chung, that Indonesian is not a classifier language. I turn to parallel facts in
Javanese in section 3.2, where I show that Javanese behaves as Indonesian in all relevant
respects; it allows bare arguments, has plural morphology, and lacks a generalized
classifier system. These empirical findings strongly indicate that the Nominal Mapping
Parameter imposes too tight a mapping between the syntax and semantics of bare nouns in
natural language and falsely exclude the morphosyntactic patterns of bare nominals in
Indonesian and Javanese that are actually attested. These results lead us to pursue a
different account of what Chierchia’s Nominal Mapping Parameter is supposed to capture.
In section 4, starting with the observation that the notion of “semantic parameter” as
constraining the mapping between the syntax and semantics of nominals is at odds with the
standard view of the locus of parameter, I propose a novel, relativized parametric theory of
the denotation of bare nominals across languages. According to this theory, languages
differ in two dimensions: how high a language allows its bare noun to grow and which set
of feature values each language allows its Num head to choose. I show that the
morphosyntactic profile of bare nouns in languages such as Javanese, Indonesian, Japanese,
English, and Italian is naturally derived from the combination of these two parameters. In
section 5, I consider a number of predictions made by the proposed theory of bare
240
nominals. The proposed analysis combines with the Subset Principle proposed by Wexler
and Manzini (1987) to predict that initial stages of all languages should go through stages
where bare nominals behave similarly to those in Indonesian and Javanese. Drawing on the
data and analysis presented by Guilfoyle and Noonan (1992) and Vainika (1993/1994), I
show that this prediction is indeed borne out by the so-called telegraphic speech produced
by English-learning children. I further show that the proposed analysis sheds a new light
on the debate concerning the nominal syntax of Slavic languages such as Russian.
Specifically, the proposed analysis, if correct, indicates that Russian nominals project only
up to NumP as Indonesian/Javanese nominals and that the universal DP hypothesis for
article-less languages needs to be seriously reconsidered. In section 6, I summarize major
theoretical implications of the parametric theory presented here as they relate to the theory
of syntax-semantic interfaces, a central subject matter of the dissertation.
2. Chierchia’s (1998a, b) Nominal Mapping Parameter
The Nominal Mapping Parameter, recently proposed by Chierchia (1998a, b), claims that
languages differ in terms of what they let their bare nouns denote in the syntax-semantics
mapping: kinds, properties, or both under certain conditions. Chierchia claims that the setting
of this semantic parameter serves to uniquely identify syntactic and morphological properties
241
of a bare noun in a given language. Chierchia’s Nominal Mapping Parameter, which
identifies three and only three language types, includes a strong statement that all languages
should belong to one of them. In the first type of languages such as Chinese and Japanese,
which he calls [+arg, -pred] languages, bare nouns are mapped onto kinds (type <e>). A kind
is defined by Chierchia (1998a: 349) as “function[s] from worlds (or situations) into
pluralities, the sum of all instances of the kind.” This has three morphosyntactic
consequences. First, since a kind is saturated in the Fregean sense, this type of language
allows bare nominal arguments. Second, languages of this type also lack plural morphology
for the following reason. As Chierchia (1998a: 351) puts it, “Fido is as good an instance of
the dog-kind as Fido and Barky are. This means that the property corresponding to a kind
comes out as being mass.” The notion of mass, in turn, is defined in Chierchia (1998a: 347)
to “come out of the lexicon already pluralized… a mass noun, such as, say, furniture, will be
true in an undifferentiated manner of singular pieces of furniture as well as of pluralities
thereof… quite literally the neutralization of the singular/plural distinction.” Since a kind is
essentially mass in that it cannot differentiate between singular and plural instances of a kind,
[+arg, -pred] languages should not have plural morphology; mass terms, being “prepluralized”, cannot be further pluralized. In other words, the extension of all bare nouns ends
up being mass in this type of language. Third, [+arg, -pred] languages also develop a
242
generalized classifier system. This is because kinds cannot be individuated and hence need
an appropriate counting level for each bare noun. Note that all the three morphosyntactic
characteristics of bare nouns observed here are the automatic consequence of the denotation
of a bare noun as a name of a kind.
The second type of language is what Chierchia (1998a, b) calls [-arg, +pred] language
such as Italian and French, where bare nouns are mapped onto properties (type <e, t>). This
type of language does not allow bare nominal arguments; they need to be combined with
determiners (either covert or overt depending on the language) to be able to serve as a
saturated argument of type <e>.
The third type of language is termed [+arg, +pred] language, and Chierchia mentions
English and Russian as examples that belong to this language type. As the setting [+arg,
+pred] indicates, this type of language expectedly shows a mixed morphosyntactic profile. It
behaves as Chinese and Japanese in that mass and bare plurals are mapped onto kinds (hence
[+arg]) whereas it behaves as French and Italian in that count nouns are mapped onto
properties (hence [+pred]).
Bare nouns must be mapped onto some semantic type, either kinds, properties, or both,
under Chierchia’s theory of a semantic parameter. The fourth possibility that bare nouns
243
are not mapped onto any type (i.e. [-arg, -pred]) is excluded because such a type does not
denote anything.
One important aspect of Chierchia’s Nominal Mapping Parameter is that, as noted above,
a particular setting of the denotation of a bare noun in a language uniquely predicts its
morphosyntactic profile. In other words, Chierchia’s theory imposes a rigid one-to-one
mapping between the morphosyntax and the denotation of a bare noun in a given language.
Thus, for example, if a language L develops a generalized classifier system, it must be the
case that that language also allows bare nominal arguments and lacks plural morphology.
This observation is emphasized by Chierchia (1998a: 354), who explicitly states that “for
example, a language with the plural-singular contrast and a generalized classifier system is
certainly logically conceivable; it could, in principle, exist. The point of view we are
adopting offers a seemingly principled way for ruling it out.”
3. The Denotation and Morphosyntax of Bare Nominals in Indonesian and Javanese
In this section, I show, based on evidence from Indonesian and Javanese, that these two
languages do not fit into any one of the three language types that should exhaust all natural
languages under Chierchia’s semantic theory. The arguments presented below owe a great
deal to and are modeled on those developed by Chung (2000), who argues against the
244
Nominal Mapping Parameter from facts in Indonesian. Chung shows that Indonesian is a
language with bare nominal arguments and a generalized classifier system, but nonetheless
does have plural morphology marked via full reduplication of the root. I review her analysis
of Indonesian bare nouns below in section 3.1. Though Indonesian does have generalized
bare arguments and plural morphology, I argue against Chung (2000) that Indonesian is not a
generalized classifier language. In section 3.2, I turn to bare nominals and their
morphosyntax in Javanese. I show that Javanese is identical to Indonesian in all criteria
pertinent to Chierchia’s Nominal Mapping Parameter; It allows bare nominals in all
argument positions and has plural morphology marked by full reduplication but does not
develop a generalized classifier system of the kind found in Japanese or Chinese.
3.1. Bare Nominals in Indonesian
Let us review evidence from Chung (2000) that Indonesian runs counter to the predictions of
the Nominal Mapping Parameter. Her basic argument is that, in Indonesian, bare nominal
arguments occur freely and numeral classifiers are required under certain conditions like a
[+arg, +pred] language, but bare nouns have a singular-plural contrast like a [-arg, +pred]
language; this cluster of morphosyntactic characteristics, thus, would remain mysterious
under Chierchia’s semantic theory.
245
First of all, Chung notes that in Indonesian, bare nouns can occur freely as direct
objects, object of prepositions and subjects, as shown in (1a-c), respectively. Notice that
there are no elements in Indonesian that correspond to definite or indefinite articles, with the
relevant distinction being only made by reference to contexts.
(1) a. Dia membeli buku.
he buy
(direct object)
book
‘He bought a book.’
(Wolff et al. 1992: 715; Chung 2000: 159)
b. Tutup-lah pintu dengan kunci.
lock-Emp door with
key
‘Lock the door with a key.’
c. Rem depan, tanpa
brake.front
(oblique object)
(McDonald 1976: 128; Chung 2000: 159)
aku rem,
without I
mengerem
sendiri.
(subject)
brake put.on.brake itself
‘The front brake, without being braked by me, braked itself.’
(McDonald 1976: 102; Chung 2000: 160)
The fact that bare nominal arguments occur freely as shown in (1a-c), therefore, indicates
that Indonesian is either a [+arg, -pred] such as Japanese or [+arg, +pred] language such as
246
English. Suppose now that Indonesian is a [+arg, -pred] language for the sake of argument.
Chung argues that Indonesian is a classifier language, in which numerals precede bare
nouns and are immediately followed by a classifier, consistent with the [+arg, -pred]
language. Interestingly, though Dardjowidjojo (1978) notes that Indonesian has as many as
sixty classifiers, only three of them are in frequent use in contemporary Indonesian,
according to Chung (p. 162); orang ‘person’ for counting persons, ekor ‘tail’ for counting
animals, birds and fish, and buah ‘fruit’ for counting other objects. Her evidence that
Indonesian is a classifier language is two-fold. First, classifiers are obligatory with the
numeral se- ‘one’ in Indonesian. She observes (p. 163) that se- ‘one’ must either be
followed by a classifier or else occur in the fixed expression s(u)atu, in which it is
combined with the obsolete classifier watu ‘stone’ (Hopper 1986: 311)”. In the absence of
a classifier, se- cannot occur. This is illustrated in examples as in (2a, b).
(2) a. Kemudian di-ambil-nya
later
se-helai
serbet kertas yang
baru. (se + classifier)
Pv-take-by.her one-Class napkin paper which new
‘Then she got a new napkin.’
(Purwo 1989: 318; Chung 2000: 163)
247
b. Kemudian di-ambil-nya
later
Pv-take-by.her
kertas
baru.
paper
new
‘Then she got a new napkin.’
(Purwo 1989: 312; Chung 2000: 163)
Second, Chung notes (p. 164) that, at an earlier stage of the development of Indonesian,
overt classifiers were more frequent than they are today after dua ‘two’ and higher numerals.
To support this, she points out that “statistics reported in Hopper’s 1986 careful study of
classifier use in the 19th century Malay of the Hikayat Abdullah, an autobiography published
in 1849, suggest that roughly 80% of the numerals that combine with NPs are accompanied
by an overt classifier.” The following examples illustrate this point.
(3) a. Maka
then
di-tembak-lah
dua-bělas
puchok
měriam
di-bukit.
Pv-fire-Emp
twelve
Class
gun
from-hill
‘[A salute of] twelve guns was fired from the hill.’
(Abdullah 1963 [1849]: 222; Chung 2000: 164)
248
b.
Ada pun takala měmbuat rumah itu tiga orang orang China kuli
as
for when make
jatoh dari atas.
house the three Class person Chinese laborer fell from top
‘In the course of its construction three of the Chinese workmen fell from the top.’
(Abdullah 1963 [1849]: 222; Chung 2000: 164)
Granted that Indonesian is a classifier language with generalized bare arguments, the only
setting that would account for these two properties under Chierchia’s Nominal Mapping
Parameter is a [+arg, -pred] type language. This setting, thus, predicts that Indonesian should
not have plural morphology. Chung (p. 164) shows that this prediction is false because
Indonesian does have a way of expressing plurality via full reduplication of the root, as
illustrated in (4) (see chapter 6 for further discussion of the semantics of reduplication).
(4)
Buah-lah
kalimat-kalimat berikut
make-Emp
sentence-Pl
menjadi kalimat-kalimat
following become
sentence-Pl
negatif.
negative
‘Please make the following sentences negative.’
(Dardjowidjojo 1978: 27; Chung 2000: 16)
249
It is widely known in the literature on Indonesian (Dardjowidjojo 1978: Dyen 1964) that,
though an unreduplicated bare noun can be constructed either as singular or plural, its
reduplicated counterpart is necessarily interpreted as plural. This traditional observation
becomes important in section 4. The crucial point here is that, under Chierchia’s system, a kind,
by definition, cannot differentiate between singular and plural instance of that kind. The fact
that reduplication has the function of denoting plurality, as in (4), thus, constitutes strong
evidence that Indonesian cannot be a [+arg, -pred] language such as Japanese or Chinese.3
Chung further points out two problems that would arise with the alternative potential
analysis of Indonesian as a [+arg, +pred] language such as English. The first problem concerns
the scopelessness of bare nominal arguments in Indonesian. Following Carlson (1977),
Chierchia (1998a: 368) observes that in English, bare plurals in object position behave as kinds
in that they obligatorily take narrow scope with respect to negation and intensional operators
3
In Indonesian, reduplication does not co-occur with numerals more than 2, as shown by the
ungrammaticality of (i):
(i) * Esti
Esti
mem-beli
tiga
buku-buku
kemarin.
Av-buy
three
book-Red
yesterday
‘Esti bought three books yesterday.’
One analysis is that the reduplicative process is blocked by numerals on the grounds of expressive economy.
Since numerals more than 2 denotes a more specific function (e.g., 2, 3, 4, etc.) than nominal reduplication (n,
n ≥2, defined in contexts), the use of a numeral independently blocks reduplication. This particular analysis,
however, does not affect the content of this chapter. Thanks to Heidi Harley (personal communication) for
useful discussion.
250
whereas indefinite singulars can take wide scope over these scope-bearing elements. This
contrast is shown in (5a, b) with respect to negation.
(5) a. I didn’t see spots on the floor
Neg>Indefinite (narrow scope): I did not see any spot on the floor.
* Indefinite>Neg (wide scope): There is a spot/are certain spots that I failed to see on the floor.
b. I didn’t see a spot on the floor.
* Neg>Indefinite (narrow scope): I did not see any spot on the floor.
Indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.
(Chierchia 1998a: 368)
If Indonesian is a [+arg, +pred] language as in English, we predict that bare nominals should
also be able to take scope over negation as do English indefinites, under their indefinite
reading. This prediction is clearly false, as evidenced by (6a, b), where the bare nouns buku
‘book’ and perempuan ‘woman’ must take narrow scope with respect to the negative
element tidak ‘not’. To express the wide scope reading of the bare nominal, a relative clause
existential sentence must be used as in (6c).
251
(6) a. Ali
tidak
jadi
membeli
buku.
Ali
Neg
finished
buy
book
‘Ali didn’t finish any book(s).’/*’There was a book that All didn’t finish.’ (Chung 2000: 161)
b.
Ia
tidak
he
Neg
melihat
see
woman
‘He saw no women.’
c.
perempuan.
(Purwo 1989: 303; Chung 2000: 161)
Ada
sebuah
buku
yang
Ali
tidak
jadi
beli.
exist
one
book
that
Ali
Neg
finish
buy
‘There is a book that Ali didn’t finish.’
The second problem concerns the lack of reduplicated forms for generic statements.
Chierchia (1998a: 362-368) observes that bare nominals in English can be inflected for plural in
generic statements as in (7).
(7) Dogs bark.
plural interpretation: There is more than one dog that barks/are barking.
generic interpretation: It is a general property of dogs that they bark.
(adopted from Chierchia 1998: 367)
252
We have seen above that Indonesian has a way of expressing plurality by full reduplication of
the root, as shown in examples such as (4). Then, if Indonesian is a [+arg, +pred] language as
in English, we predict that reduplicated nominals in this language should also be able to feed
generic interpretation. Examples as in (8a), however, show that this prediction is false, because
the reduplicated counterpart of the root anjing ‘dog’ can yield only a plural interpretation
(Sneddon 1996: 17). To express the generic interpretation, the unreduplicated bare noun must
be used instead, as shown in (8b), which gives rise to both plural and generic interpretations.
(8) a. Anjing-anjing
dog-Red
menggonggong.
bark-Red
‘Dogs bark/are barking.’
plural interpretation: There is more than one dog that barks/is barking.
* generic interpretation: It is a general property of dogs that they bark.
b.
Anjing
menggonggong.
dog
bark
‘Dogs bark/are barking.’
plural interpretation: There is more than one dog that barks/is barking.
generic interpretation: It is a general property of dogs that they bark.
253
Chung (p. 168) briefly considers the final analytic possibility compatible with the
Nominal Mapping Parameter whereby Indonesian is currently in the transition stage from a
[+arg, -pred] language to a [+arg, +pred] language. Chung rejects this analysis on the ground
that it would lead us to “expect the singular-plural contrast to be less in evidence at earlier
stages of the language, when overt classifiers were more frequent.” Chung argues that
prediction is not borne out by the 19th century Malay from Hikayat Abdullah, because
examples as in (9) that contain nouns with overt plural inflection occur frequently.
(9)
Maka tukang-tukang kayu pun měnarah-lah akan sěgala pěrkakas rumah itu.
then
worker-Pl
wood also smooth-Emp for all
part
house that
‘Carpenters started shaping planks of wood for various parts of the building.’
(Abdullah 1963 [1849]: 221; Chung 2000: 169)
To test Chung’s prediction, one would need a translation of the same autobiography in
contemporary Indonesian so that we can compare the number of reduplicated nouns in
the 17th century Malay with that of the corresponding nouns in contemporary Indonesian.
This prediction, however, is impossible to test for the moment due to the absence of the
254
Indonesian translation of the text to date. Thus, I leave the fourth analysis of Indonesian
aside in this chapter.
To sum up, Indonesian does not fit into any one of the three language types under
Chierchia’s (1998a, b) semantic typology. The free occurrence of determinerless, bare
arguments shows that Indonesian is not a [-arg, +pred] language such as Italian or French.
The presence of plural morphology marked by full reduplication shows that this language
is also not a [+arg, -pred] language such as Japanese or Chinese. The obligatory narrow
scope reading of bare nominals with respect to negation and the lack of the reduplicated
form of a bare noun for generic statements means that Indonesian is also not a [+arg, +pred]
language such as English or Russian.
Although Chung’s argument against the Nominal Mapping Parameter from Indonesian is
clear and I develop similar arguments against it based on parallel facts from Javanese in the
next subsection, I point out here that it is problematic to analyze Indonesian as a (generalized)
classifier language, as Chung does. In fact, Chung (p. 162-164) provides two arguments that
suggest that Indonesian is not a classifier language. First, she observes (p. 162) that classifiers
in Indonesian are more often than not omitted in colloquial Indonesian after dua ‘two’ and
some number greater than two (Dardjowidjojo 1978: 64, 65; MacDonald 1976: 82, 83;
Sneddon 1996: 134, 135). Second, she points out that “even in formal registers of the
255
contemporary language, an overt classifier need not occur after dua ‘two’ or higher numerals.”
Some examples to illustrate these points are given in (10), which is “excerpted from
Indonesian translations of English-language articles on Indonesian syntax” (p. 163).
(10) Muda-mudahan makalah ini telah
hopefully
then
memenuhi
this already fulfill
dua
tujuan
pokok-nya.
two
goal
principal-its
‘Hopefully, this paper has fulfilled its two major goals.’(Purwo 1989: 333, Chung 2000: 163)
How about the two arguments made by Chung based on the obligatory presence of a
classifier with the numeral se- ‘one’ and the classifier use in the 19th century Malay of the
Hikayat Abdullah? The obligatory presence of a classifier with the numeral can be accounted
for independently without necessarily assuming that Indonesian is a classifier language
because the numeral se- ‘one’ is a clitic that needs a classifier as a host. Indeed, the non-clitic
free morpheme meaning ‘one’, satu, can occur without any classifier, as the comparison
between (2a) and (11) shows.
256
(11) Kemudian di-ambil-nya
later
satu serbet kertas yang
baru.
(satu + no classifier)
Pv-take-by.her one napkin paper which new
‘Then she got a new napkin.’
Chung’s second argument for her claim that Indonesian is a classifier language came from
her observation that overt classifiers were more frequent than they are today after dua ‘two’
and higher numerals. We have seen above that this is supported by Hopper’s 1986 careful
study of classifier use in the 19th century Malay of the Hikayat Abdullah, an autobiography
published in 1849, according to which roughly 80% of the numerals that combine with NPs
are accompanied by an overt classifier. This argument seems hard to evaluate at present for
two reasons. First, as already noted above, there is no translation of the relevant
autobiography in modern standard Indonesian to see whether this argument is still
substantiated. Second, granted that we will have such a translation at hand, it is still not clear
whether the comparison of 17th century Malay and the contemporary Indonesian versions of
the same text can yield a meaningful result. Given a variety of differences between
contemporary Malay and Indonesian that pose a formidable obstacle to field linguists
working on dialects of Malay/Indonesian, it is possible that there would be even more
dramatic differences between 17th century Malay and contemporary Indonesian.
257
At any rate, Chung (p. 164) concludes, based on the two arguments made above, that
“when NP combines with a numeral, a classifier must be syntactically present even though it
need not be phonetically overt.” This final conclusion, however, seems not to be supported by
facts from other [+arg, -pred] languages such as Japanese or Chinese, which have a bona-fide
classifier system. In Japanese, for example, when a noun combines with a numeral, a classifier
must be overtly expressed. This is shown by the contrast between (12a) and (12b).
(12) a.
Taro-ga
san-nin-no
gakusei-o
mita.
Taro-Nom
3-Class-Link
student-Acc
saw
‘Taro saw three students.’
b. *
Taro-ga
san-(no)
gakusei-o
Taro-Nom
3-Class-Link
student-Acc
mita.
saw
‘Taro saw three students.’
As we will see in the next section, there is clear evidence that Javanese is not a classifier
system. Given the large-scale linguistic and social interaction between Indonesian and
Javanese in the Java island of Indonesia (Poedjosoedarmo 1982: 84; see also Chung 2000:
163), one likely scenario, compatible with the all the facts noted above, is that contemporary
258
Indonesian has lost a classifier system due to the linguistic influence from Javanese. Based
on the considerations above, I assume that Indonesian is a non-classifier language like
Javanese, contra Chung (2000).
3.2. Bare Nominals in Javanese
In this subsection, I develop a similar argument as made by Chung from Indonesian against
the Nominal Mapping Parameter from Javanese, a closely related Austronesian language
spoken in Indonesia. I show that Javanese behaves like Indonesian in all relevant respects.
Specifically, the free occurrence of bare nominal arguments shows that Javanese cannot be a
[-arg, +pred] language such as Italian or French under Chierchia’s system; it must be either a
[+arg, -pred] language such as Japanese or Chinese or a [+arg, +pred] language such as
English and Russian. I reject the first possibility on the ground that Javanese has plural
morphology marked via reduplication and lacks a generalized classifier system. I reject the
second possibility based on the obligatory narrow scope of bare nominals under their
indefinite interpretation as well as the lack of reduplicated forms of bare nouns for generic
statements. These results, therefore, show that Javanese presents itself as another
counterexample to the predictions of Chierchia’s Nominal Mapping Parameter.
259
Like Indonesian, Javanese permits bare, determinerless arguments to occur rather freely
in any argument position, as illustrated in (13a-d).
(13) a.
Buku larang.
book
expensive
‘A book/the book/books {is/are} expensive.’
b.
Esti
tuku
buku.
Esti
buy
book
‘Esti bought a book/the book/books.’
c.
Esti nukokke uwong buku.
Esti buy
man
book
‘Esti bought a man/the man/men a book/the book/books.’
d.
Esti entuk
informasi
seko
buku.
Esti get
information
from
book
‘Esti got information from a book/the book/books.’
In these examples, the bare noun buku ‘book’ occurs as the subject (13a), the direct object
(13b), the indirect object (13c), and the object of preposition (13d). The free occurrence of bare
260
arguments in Javanese, thus, shows that this language cannot be a [-arg, +pred] language such
as Italian or French. In other words, it should be either a [+arg, -pred] language or a [+arg,
+pred] language. However, I demonstrate, following Chung’s (2000) argument based on
Indonesian, that Javanese does not fit into either one of the language types.
Let us now consider first the analysis whereby Javanese is a [+arg, -pred] language.
Recall that, under Chierchia’s semantic theory, [+arg, -pred] languages should have three
morphosyntactic properties due to the specification of bare nouns as denoting a kind; a bare
nominal argument, a generalized classifier system, and no plural morphology. Thus, if
Javanese is a [+arg, -pred] language, the Nominal Mapping Parameter predicts that a) this
language should have a generalized classifier system and that b) it should have no plural
morphology, as in Japanese or Chinese.4 Both of these predictions are falsified by examples
as in (14a, b) and (15a, b).
4
In fact, we will see in the next section that Japanese also does have plural morphology marked by tachi-
suffixation to or reduplication of a nominal root, as shown in (ia, b).
(i) a.
Shonen-tachi-ga kooen-de
asonda.
boy-Pl-Nom
played
park-Loc
‘Boys played in the park.’
b.
Hito-bito-ga
kooen-ni
kita.
people-Red-Nom
park-Loc
came
‘People came to the park.’
This fact poses a problem for Chierchia’s Nominal Mapping Parameter because Japanese, being a [+arg, —
pred] language, should not be able to have plural morphology due to the kind-denoting requirement of bare
nouns. See Nakanishi and Tomioka (2004) for a detailed analysis of the semantics of –tachi. See also note 30.
261
(14) a. Esti
tuku
buku
telu.
Esti
buy
book
three
b.
‘Esti bought three books.’
(15) a. Esti
tuku
buku-buku.
Esti
buy
book-Red
‘Esti bought books.’
Esti mangan
pelem
loro.
Esti eat
mango two
‘Esti ate two mangos.’
b.
Esti
nata
meja-meja.
Esti
arranged table-Red
‘Esti arranged tables.’
(14a, b) show that Javanese is not a classifier language (Poedjosoedarmo 1982; Robson
2002). This utter lack of a classifier system is in contrast with the optionality of classifiers in
Indonesian as we have seen in the previous section. The examples in (15a, b) show that
Javanese marks plurality via full reduplication of the root, as in Indonesian. These two facts
clearly indicate that Javanese is not a [+arg, -pred] language such as Chinese and Japanese.
Let us next consider another analysis whereby Javanese is a [+arg, +pred] language such as
English or Russian. Again applying the two arguments developed by Chung to bare nominals in
Javanese, we can see that this analysis is incorrect. The first argument concerns the scopelessness
of bare nouns in Javanese under their indefinite interpretation. Recall Carlson’s (1977)/Chierchia’s
(1998a: 368) observation from the last subsection that bare plurals in English behave as kinds and
262
obligatorily take narrow scope with respect to negation. This is not the case with indefinite singular
nouns which can take wide scope over negation. The relevant contrast was illustrated in (5a, b).
Now, if Javanese is a [+arg, +pred] language, a bare nominal argument in this language should also
allow a wide scope reading of the argument with respect to negation on its indefinite singular
interpretation, as do English indefinites. This prediction is incorrect, as shown by (16a), in which
the bare noun kotoran ‘spot’ must take narrow scope with respect to negation. Just as in Indonesian,
a relative clause sentence is used as in (16b) to express the wide scope reading.
(16)a. Aku ora
I
Neg
weruh
kotoran
ning jubin.
see
spot
on
floor
‘I did not see spots on the floor.’
Neg>Indefinite (narrow scope): I did not see any spot on the floor.
*Indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.
b.
Ana kotoran
sing aku
ora
weruh
ning
jubin.
exist spot
that I
Neg
see
in
floor
‘There is a spot on the floor that I failed to see.’
* Neg>Indefinite (narrow scope): I did not see any spot on the floor.
Indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.
263
The second argument against the classification of Javanese as a [+arg, +pred] language
concerns the absence of reduplicated forms in Javanese for the generic use of bare nouns. We
have seen above that bare nominals in English can be inflected for plural in generic statements
as in (7a, b). If Javanese is a [+arg, +pred] language, then the prediction is that bare nominal
arguments in Javanese should also be able to reduplicated when interpreted as generic. This
prediction is incorrect, as shown in (17a).
(17) a.
Asu-asu
njegug.
dog-Red
bark
“Dogs are barking.’
plural interpretation: There is more than one dog that barks/are barking.
* generic interpretation: It is a general property of dogs that they bark.
b.
Asu
njegug.
dog
bark
“A dog/the dog/dogs bark.’
plural interpretation: There is more than one dog that barks/are barking.
generic interpretation: It is a general property of dogs that they bark.
In the example in (17a), the reduplicated noun asu-asu ‘dogs’ only allows plural
interpretation. Instead, the non-reduplicated bare nominal is used for generic statements
264
in Javanese, as in (17b). True, this fact is naturally accounted for if Javanese is a [+arg, pred] language as in Japanese because kinds are known to yield a universal reading
(Chierchia 1998: 363a) but we have seen already above that this analytic possibility is
incorrect in light of the lack of a generalized classifier system and the presence of plural
morphology marked by reduplication. Thus, the two arguments concerning the
mandatory narrow scope reading of bare nominals under their indefinite readings and the
lack of reduplicated bare nouns for generic statements in Javanese provide evidence that
this language cannot be a [+arg, +pred] language such as English. The
last
analytic
possibility that is compatible with the Nominal Mapping Parameter is to resort to a finergrained distinction within Chierchia’s semantic typology. Under this possibility, Javanese
may well be similar to Russian.5 In the next section, I show that this comparison is on the
right track by demonstrating that Javanese is minimally different from Russian in the
possible values for the Number head that each language is allowed to choose.
To summarize this section, I have shown, using the arguments developed by Chung
(2000), that neither Indonesian nor Javanese can classified as any one of the three language
types that should serve to categorize all human languages under Chierchia’s Nominal
Mapping Parameter. This result is important because it raises important questions about the
5
Thanks to Heidi Harley (personal communication) and Simin Karimi (personal communication) for this observation.
265
extent to which the morphosyntactic properties of a bare noun in a language are predictable
from its denotation and vice versa. In particular, the descriptive results achieved here clearly
suggest that the Nominal Mapping Parameter imposes too tight a mapping between the
denotation and morphosyntax of NPs.
4. A Relativized Parametric Theory of Nominal Denotation: From Indonesia to the World
In this section, I develop a purely syntactic account for the effects of the Nominal
Mapping Parameter of Chierchia (1998a, b) within the framework of the Minimalist
Program (Chomsky 1995). I start by pointing out a conceptual problem with Chierchia’s
notion of semantic parameter. I show that this view of parameters is hardly groundable
within the standard conception of the locus of parameters within the Principles &
Parameters approach to language variation. Resolution of this problem leads us to seek an
alternative explanation in the realm of the morphosyntactic variation of the lexicon in each
language, which is an unreducible source of linguistic variation (Borer 1984; Fukui
1986,1995; Chomsky 1995). I propose a relativized parametric theory of nominal
denotation, whereby languages differ in terms of a) the height/complexity of the functional
super-structure above bare nominals and b) the possible set of Number values. Point (a)
has been independently argued for by recent work such as Grimshaw (1991, 2005),
266
Massam (2001), Guilfoyle and Noonan (1992), and Vainikka (1993/1994). Point (b) is
independently supported by extensive study of the Number system in languages such as
Malay and Bahasa Indonesia as in Carson (2000) and Chung (2000). The proposed
analysis, thus, provides further evidence for these claims.6
4.1. A Relativized Parametric Theory of Nominal Denotation and Morphosyntax
Let us start by pointing out a conceptual problem with Chierchia’s notion of “Semantic
Parameter”. The Nominal Mapping Parameter essentially proposes that parameters can be
postulated within the mapping between syntactic and semantic/LF representations. This
proposal, however, is clearly at odds with the standard conception of the locus of the parameter
as in the Principles & Parameters approach to language variation as outlined in Chomsky
(1995). Instead, this approach takes it as a fundamental heuristic that parametric variation be
restricted to the properties of the lexicon (Chomsky 1995, 2000) or, more specifically,
functional categories alone, a hypothesis known as the Functional Parameterization
Hypothesis (Borer 1984; Fukui 1986, 1995). A natural approach then should be one that
6
Thanks to Heidi Harley (personal communication) for suggesting the idea of “growth of bare nominals” and
to Andrew Carnie (personal communication) for bringing my attention to several important works such as
those cited here that independently proposed such an idea.
267
derives crosslinguistic variation in the denotation and morphosyntax of bare nominals from
independently motivated variation in the makeup of the lexicon in each language.
With this agenda in mind, I propose that languages differ in two dimensions: a) how
high a bare nominal can “grow” across languages and b) what set of the binary values the
Num head in each language selects. For the first part of this claim, I propose the universal
nominal morphosyntactic hierarchy DP>QP>ClP>NumP>NP, from which languages set
the appropriate height of nominal projections for their bare nouns. Specifically, I argue that
languages such as Javanese and Indonesian project up to NumP; languages such as
Japanese project up to QP; languages such as Italian project up to DP. Finally, languages
such as English project either up to QP like Japanese or DP like Italian, depending on the
nature of a noun inserted into the N head in a manner to be explained below. A question
arises as to whether there are languages which instantiate the ClP option. 7 I address this
question in section 5, where I claim, drawing on evidence from Cheng and Sybesma
(1999), that definite bare nouns in Cantonese and Mandarin instantiate this option. This
proposal is summarized in (18).
7
Thanks to Heidi Harley (personal communication) for this question.
268
(18) The Universal Nominal Morphosyntactic Hierarchy
DP ITALIAN/ENGLISH (for count nouns)
QP JAPANESE/ENGLISH (for bare plurals/mass nouns)
D
ClP MANDARIN/CANTONESE (for definite bare nouns)
Q
NumP JAVANESE/INDONESIAN
Cl
Num
N
As stated at the outset of this section, this idea of “growth of NPs” is not a new idea but
rather has been argued for by recent work as in Grimshaw (1991, 2005), Massam (2001),
Guilfoyle and Noonan (1992), and Vainikka (1993/1994) on various grounds. Grimshaw
(1991, 2005) proposes an influential theory of extended projection whereby lexical heads
such as V and N form a single projection with functional heads such as T/C and D/P on top
of them conditioned by two constraints on projection. First, two categories in a projection
chain must be categorically identical in terms of verbal and nominal features. Second, the
F-value of X must not be higher than the F-value of YP” when X is the head of YP and YP
is a projection of X (Grimshaw 2001: 4, 5).8 “F-value” is a functional status assigned as
8
Grimshaw (p. 4) defines the notions of a head and a projection as in (i):
269
follows: F0 is assigned to the lexical categories, F1 is assigned to the lowest level
functional category; and Fn to the next successively higher functional categories.9 Thus,
the lexical category V forms an extended projection with T and C because the F-value of V,
being associated with the verbal feature, is not higher than that of a T or C. For the same
reason, N (F1) forms an extended projection with D (F2) and P (F3). One important
respect in which Grimshaw’s approach is similar to the proposed analysis based on the
universal nominal functional structure in (18) is her claim that, under a universal tree
structure/inventory (as in Cinque 1999), languages can choose whether or not to include
each functional specification as a head in their possible inventory, as long as possible
combinations of the functional head and their complement do not violate the two
constraints noted above. Thus, if we have 20 functional categories in the universal
inventory, a language might choose {F20, F9, F5, F0} while another language might
(i) X is a head of YP, and YP is a projection of X iff:
a. YP dominates X
b. The categorial features of YP and X are consistent.
c. There is no inconsistency in the categorial features of all nodes intervening between X and YP (where
a node N intervenes between X and YP if YP dominates X and N, N dominates X.)
For example, TP forms an extended projection with V because the F-value of V (F0) is not higher than the Fvalue of TP (F1) and no projection intervenes between TP and V. The same holds for the CP. See Grimshaw
(2005) for details.
9
Grimshaw (2005: 64-71) also provides a useful summary of a variety of papers that appeared after the
publication of Grimshaw (1991) that address important issues that arise in the theory of extended projections.
270
choose {F8, F4, F2, F0}. However, the two principles of extended projections above
impose strict structure constraints on the relative hierarchy of these categories. Grimshaw’s
particular application of the theory of extended projection to nominal domains is different
from the proposed analysis in that the highest (extended) projection of nominals is PP
under her analysis whereas it is DP under the current analysis. Nonetheless, in a
fundamental respect, her proposal is quite similar to the analysis I develop here, according
to which languages choose a subset of the universal nominal structure given in (18) (or the
universal morphosyntactic pool), and this choice determines the fate of the
morphosyntactic behavior of elements embedded within the structural configuration
parametrically chosen. Therefore, the proposed analysis can be understood as one
particular way of implementing Grimshaw’s theory of extended projection to nominal
domains within a more elaborated array of functional projections such as Num, Cl, and Q.
Massam (2001) reaches similar conclusions based on her careful study of what she calls
Pseudo Noun Incorporation/PNI in Niuean, an ergative-absolutive language of the Tongic
subgroup from the Oceanic family. This language is a VSOX language, as illustrated in (19a).
As Massam observes, this language also allows VOSX order as shown in (19b), which she
analyzes as a case of PNI.
271
(19) a.
Ne
kai
e
Sione
e
tau
Past
eat
Erg
Sione
Abs Pl
‘Sione ate the taros with a fork.’
b.
aki
e
huki.
taro
with
Abs
fork
(Massam 2001: 155)
Ne inu
kofee
kono
a
Mele.
Past
drink
coffee
bitter
Abs
‘Mary drank bitter coffee.’
talo
(Massam 2001: 158)
Massam assumes that both examples are derived by the VP fronting into [Spec, TP]. For the
VSOX order as in (19a), the DP argument undergoes movement into [Spec, AbsP] to check
Absolutive Case. This movement is followed by the fronting of the VP that contains the verb
and the trace of the direct object into [Spec, TP]. This derivation is illustrated in (20).
272
(20) The Derivation of the VSOX Order
TP
VPx
V
T′
ti
T
vP
DP(erg)
v′
v
AbsP
DPi(abs)
Abs′
K(abs)
tVPx
(slightly modified from Massam 2001: 163)
For the VOSX/PNI order as in (19b), Massam crucially assumes that the direct object is an
NP, not a DP. Under the assumption that Case features appear on an extended functional
head such as KP, not on an NP, the direct object in (19b) cannot (and need not) check
Absolutive Case, which is checked by the external argument of the sentence. The VOSX
order is derived by fronting of the VP that contains both the verb and the NP object. This
derivation is illustrated in (21).
273
(21) The Derivation of the VOSX Order
TP
VPx
V
T′
NP
T
AbsP
DP(abs)
Abs′
K(Abs)
tVPx
(slightly modified from Massam 2001: 165)
Massam shows that the analysis above that draws on the NP vs. DP distinction provides a
natural account of several differences between the two types of Pseudo Noun Incorporation
known as Generic PNI and Existential PNI first noted by Seiter (1980). The properties of the
two types of PNI that are relevant for present purposes are summarized in (22) and (23),
together with illustrative examples in each type of the PNI in (24a) and (24b).
274
(22) The characteristics of the Generic PNI Construction (Massam 2001: 172)
a. NP is non-specific and non-referential
b. no extended nominal categories or dependants of extended nominal categories
(i.e., tensed relative clauses) appear over N.
c. durative/frequentative meaning
(23) The characteristics of the Existential PNI Construction (Massam 2001: 177)
a. occurs with closed class of verbs fai ‘have/be’, muhu ‘have plenty/be plentiful’
b. relative clause can appear at the right edge of the sentence
c. NP is referential, non-specific and indefinite
(24) a. Example of the Generic PNI (= (19b)) b. Example of the Existential PNI
Ne
inu
kofee kono
a
Mele.
kua fai
nakai e
umu haau?
Past drink coffee bitter Abs
Perf make Q
Abs oven
‘Mary drank bitter coffee.’
‘Have you made your oven yet?’
your
(Massam 2001: 173)
275
The three characteristics of the Generic PNI in (22a-c) directly follow from the NP status of
the pseudo-incorporated nominal. First, property (22b) falls out from the fact that it is only a
bare NP that undergoes movement into [Spec, TP] with the verb; if it were a DP, it would
need to check Absolutive Case, contrary to facts. This property, Massam argues, also
accounts for the fact that relative clauses cannot attach to pseudo-incorporated nominals
under the assumption (Finer 1998; Ghomeshi 1996; Larson 1994; Kayne 1994) that relative
clauses appear high within DPs. The property in (22a) is also naturally predicted under the
standard assumption that a bare nominal has a non-referential, non-specific denotation. This
property, in turn, yields the durative/frequentative meaning for the Generic PNI because the
lack of referentiality ensures unbounded/non-delimited interpretation of the event denoted by
the sentence, as argued for in Jackendoff (1990) and Tenny (1994), as cited in Massam
(2001:171) Massam further shows that the three properties of Existential PNI in (23a-c) can
also be derived if the nominal is topped with a quasi-determiner projection contributed by
verbs of existence such as fai ‘have/be’ and humu ‘have plenty/be plentiful’ (23a). She
proposes (p. 186) that “fai is essentially at one and the same time a verb and a determiner, in
that it serves simultaneously as the main lexical predicate head for the sentence, and as a
functional category which confers referentiality on its complement NP” in the sense of
Higginbotham (1985) (see also Nichols 1997 and Johns 1999 for similar ideas as applied to
276
Noun Incorporation in Zuni and Inuktitut, respectively). Specifically, verbs like fai bind the
referentiality index of the NP as a typical determiner does to its complement. This proposal
thus not only derives the fact (23b) that relative clauses can appear in the Existential NPI
(because verbs of existence serve the same function as D heads) but also yields the
referential, non-specific, indefinite reading to the NP (23c). It is clear by now that Massam’s
analysis of the differences between the Generic PNI and the Existential PNI crucially
depends on the fine-grained distinction between the NP and DP. Therefore, her analysis can
be considered as another variation of the analysis proposed here whereby a parametrically
different height of nominal functional projections across languages accounts for different
morphosyntactic properties of their bare nominals.
Guilfoyle and Noonan (1992) argue for what they call the Structure Building
Hypothesis, namely, that, in the early stages of language acquisition, children have a UGconstrained grammar solely consisting of lexical categories (N/V/A/P), Lexical Grammar,
with functional categories such as I, C, D, and KASE gradually emerging after lexical
categories based on the positive evidence from input data and the emergence of what they
call Functional Grammar.10 They show that this hypothesis receives strong support from
10
Guilfoyle and Noonan (p. 241) acknowledge that Lebeaux (1988) and Radford (1990) developed similar ideas.
For example, they state, citing Lebeaux, that “the early stages of acquisition are a pure representation of thetatheory.” See also Clahsen (1991), Clahsen and Penke (1992), and Platzack (1991) for related discussion on
the Structure Building Hypothesis.
277
the optionality of subjects in child language as well as the absence of NP-movement and
subject-auxiliary inversion and several stages concerning the position of verbs in German
because these properties are exactly what we expect if child English syntax lacks
functional categories such as T and C. The proposed idea of “growth of bare nominals”,
thus, can be naturally regarded as applying Guilfoyle and Noonan’s Structure Building
Hypothesis to the domain of nominal syntax.
Vainikka (1993/1994) provides further evidence for the non-maturational gradual
development of phrase structure as in Guilfoyle and Noonan’s Structure Building
Hypothesis, based on the parallel acquisition of nominative case and tense-related
materials. Her study carefully observes four of the developmental stages the child “Nina”
went through. During the first stage of her acquisition (age 1; 11-2; 1), Nina used
predominantly genitive subjects. During the second stage (age 2; 1-2; 2), she instead began
to use nominative subjects predominantly. This shift in case marking for subjects makes
sense if Nina’s grammar developed a phrase structure that includes functional material in
the V projection up to TP; the subject is assigned genitive case in the specifier of VP but
when the TP structure is learned, the subject is assigned nominative case in the specifier of
TP as in Adult English. Vainikka shows that this analysis is supported by the fact that the
TP-related material such as modals, auxiliaries, third person singular -s, and past tense -ed,
278
also gradually emerge in the second stage. This view of phrase structure thus can be
considered another variation of the proposed analysis of nominal syntax.
Getting back now to the second part of the claim concerning the Num head, I claim
that there are two possible sets of values for the Num P; {singular, plural} or {neutral,
plural}. Languages such as English and Italian select {singular, plural} or {neutral, plural}
values whereas languages like Japanese, Javanese, and Indonesian select {neutral, plural}
values. There is independent evidence that the possible values for the number slot in Italian
and English are significantly different from those for the same slot in Indonesian, Javanese,
and Japanese. Thus, in her extensive study of the number system in Malay, Carson (2000)
shows that bare nouns in Malay are neutral with respect to number unless reduplication
tells us otherwise, and concludes that Malay chooses {neutral, plural} values for the
Number head. For example, a (unreduplicated) bare nominal can denote either a singular
or plural instance of the entity denoted by that nominal whereas its reduplicated form
specifically denotes more than one instance of the same entity. As noted in section 3.1, the
same argument was independently made by Chung (2000: 165, 167) for Indonesian.
Consider examples as in (25a) (= (4)) and (25b).11
11
The translation for (25b) is slightly modified to reflect the fact that kalimat can be interpreted either as
singular or plural, a judgment confirmed by my native language consultant.
279
(25)a. Buat-lah
make-Emp
kalimat-kalimat
berikut
menjadi
sentence-Pl
following become
kalimat-kalimat
negatif.
sentence-Pl
negative
‘Please make the following sentences negative.’ (Dardjowidjojo 1978: 27; Chung 2000: 165)
b. Kalimat
Dasar.
sentence basic
‘Basic sentence(s) “ (Wolff et al, 1992; Chung 2000: 165)
In the example in (25a), the reduplicated noun kalimat-kalimat ‘sentences’ must be
construed as ‘more than one sentences’ whereas, in the example in (25b), the
corresponding bare noun kalimat ‘sentence(s)’ can be construed either as a singular or
plural. These examples, thus, illustrate that a bare nominal in Indonesian takes {neutral,
plural} values for the Num head as in Malay. See also Dyen (1964: 7a.-10), cited by
Chung (2000: 166-167), who makes a similar observation. .
There is evidence that Japanese and Javanese also select the {neutral, plural} values for
the Num head as in Indonesian though dominant morphological processes to denote plurality
seem to be different between Japanese (tachi-suffixation) and Javanese/Indonesian
(reduplication) (see note 3 for relevant discussion). Consider examples in (26a, b) from
Javanese and (27a, b) from Japanese.
280
(26) a. Jaran
lagi
horse Prog
mangan.
b.
eat
‘A horse is eating./Horses are eating.’
(27) a.
Uma-ga
hasitteiru.
horse-Nom
play
Jaran-jaran
lagi
mangan.
horse-Red
Prog
eat
‘*A horse is eating.’/Horses are eating.’
b. Uma-tachi-ga
‘A horse is running./Horses are running.’
horse-PL-Nom
hasitteiru.
run
‘*A horse is running./Horses are running.’
In the example in (26a), the unreduplicated bare nominal jaran ‘horse(s)’ can denote either
singular or plural instances of the horse. Its reduplicated correspondent jaran-jaran
‘horses’ specifically denotes plural, as shown by the English translation given to the
example in (26b). Exactly the same observation holds for Japanese examples as in (27a)
and (26b), where the bare nominal uma ‘horse(s)’ can be construed as singular or plural
depending on non-linguistic contexts but must denote plural once it is suffixed with -tachi.
This semantic contrast between (26a)/(27a) and (26b)/(27b), therefore, provides evidence
that bare nominals in Javanese and Japanese are specified as {neutral, plural} for the Num
head as in Indonesian. Examples as in (28a, b) from Javanese and (29a, b) from Japanese
make the same point.
281
(28) a.
Callie lan
Tisa kuwi
Callie and Tisa Cop
kucing.
b.
cat
‘Callie and Tisa are cats.’
(29) a.
Callie to Tisa-wa
neko-da.
Callie
kucing
Callie
cat
‘Callie is a cat.’
b.
Callie-wa
neko-da.
Callieand Tisa-Top cat-Cop
Callie-Top
cat-Cop
‘Callie and Tisa are cats.’
‘Callie is a cat.’
Notice that there is no morpheme in Indonesian/Javanese/Japanese that specifically
denotes singularity. As stated above, these languages make the distinction between
singularity and plurality of a noun only based on contexts in which it is found. I take this to
suggest that there is no grammaticalized notion of singularity in these languages.
The number system in languages such as English and Italian exhibits a different picture
from that in Japanese/Indonesian/Javanese. In English, there is a purely grammatical
distinction between mass and count nouns that seems not predictable from their conceptual
structures, though I could not make a similar argument for the number system in Italian due
to the unavailability of the data. Consider the pair of words, wheat and oat, in English. (30a,
282
b) and (31a, b) show that wheat is a mass noun whereas oat is a count noun, even though
they both denote “a grain of cereal”.12
(30)a.
David did not eat many oats.
(31)a. * David did not eat many wheat(s).
b. *David did not eat much oat.
b.
David did not eat much wheat.
(30a, b) show that the word oat can go only with the determiner (many) that requires a count
noun. (31a, b) show that the word wheat can go only with the determiner (much) that
requires a mass noun. These examples, thus, suggest that the count-mass distinction in
English is grammaticalized in certain pairs of words such as oat and wheat. Based on this
consideration, I assume that English and Italian can take one of the following three values for
the Num head: singular, plural, and neuter.13
12
Many thanks to Heidi Harley for (30a, b) and (31a, b) and very useful discussion. (31b) is acceptable if
many wheats is interpreted as “many types of wheat”.
13
As Heidi Harley (personal communication) notes, an intriguing issue remains with the English number system.
Borer (2005) claims that the alleged English plural morpheme –s is a kind of classifier. One argument for this claim
comes from the fact that the proposition in (i) is true even if only one dog is removed; if -s denoted a plurality of the
noun it attaches to, (i) should come out as a false statement.
(i) Any dogs will be removed.
I would like to address this issue in my future work.
283
With all the results discussed thus far in place, the proposed relativized parametric
theory of nominal denotation can be summarized as in Table 3.
Table 3: A Relativized Parametric Theory of Nominal Denotation (preliminary version)
Languages
Indonesian/Javanese
Height of Nominal Projections
Possible Sets of Num Values
NumP
{neutral, plural}
Japanese
QP
{neutral, plural}
Italian
DP
{singular, plural, neuter}
English
DP or QP
{singular, plural, neuter}
Three assumptions are in order here before we move onto the actual structural analysis of
bare nominals across languages. First, I assume that the N in the universal nominal
morphosyntactic hierarchy in (18) is underspecified with respect to its denotation (cf.
Distributed Morphology; see, for example, Marantz 1997 and Harley and Noyer 1999).
Second, I assume the economy of projection as proposed on various conceptual and empirical
grounds in recent work (Chomsky 1995, Fukui 1986, 1995, Fukui and Speas 1986, Law 1991;
Speas 1994; Bošković 1997; Grimshaw 1993; Radford 1990; Safir 1993). Specifically, if a
language like Italian and French selects the {singular, plural} set for the Num head, the
284
Cl(assifier) P does not project on the ground of expressive economy. The individuation
function encoded by the singular value of the {singular, plural} set has the same function as
that encoded by the classifier. Under the theory that semantic composition is computed in the
bottom up fashion in a strictly local manner, the projection of the NumP with the relevant
value makes the projection of the dominating ClP redundant. Finally, I assume that there is a
feature checking/valuation relation of some sort between the Num head and its complement N.
To be precise, the Num head with the {singular, plural} set values its complement as a count
noun while the Num head with the {neutral, plural} set values it as a mass noun. I will return to
this last assumption in the end of this section.
4.2. Deriving the Denotation and Morphosyntax of Bare Nominals across Languages
Consider first the nominal syntax of bare nouns in Javanese and Indonesian. We have seen
in sections 2 and 3 that these two languages exhibit the following four morphosyntactic
characteristics: i) they allow bare nominal arguments, ii) they lack a generalized classifier
system, iii) all nouns are a type of mass nouns, and iv) they allow only narrow scope of
bare nominals. All these properties straightforwardly follow if bare nominals in Javanese
and Indonesian project up to NumP with the set {neutral, plural}, as shown in (32).14
14
See chapter 6 for a detailed explication of how reduplication is realized in Indonesian.
285
(32) The Nominal Structure of Bare Nouns in Indonesian and Javanese
NumP
Num
{neutral, plural}
N
buku ‘book’
First, the two languages in question allow bare nominal arguments because there is no DP,
as shown in (32). Second, they don’t have a generalized classifier system because they
project only up to NumP for bare nominals. Third, the extension of all nouns is mass for
the following reason. If the Num value is specified as neutral, the denotation of the NumP
is a kind because it does not differentiate between singular and plural instances of the
NumP. If the Num value is plural, the denotation of the Num P still comes out as a bare
plural, which is also a kind under Carlson’s (1977)/Chierchia’s (1998a, b) theory. Thus,
whichever value the Num head selects yields a kind, hence mass interpretation to the
NumP. Finally, the obligatory narrow scope reading of bare nouns with respect to negation
follows from the kind-denoting requirement (or whatever principle blocks the wide scope
reading of such nouns). In this way, the clustering of the morphosyntactic properties
observed in Javanese and Indonesian, which was shown to be unpredictable under
286
Chierchia’s (1998a, b) Nominal Mapping Parameter, naturally follows from the interaction
of the Num P structure and the {neutral, plural} set for the Num head.
One potential problem that remains with the proposed analysis is how to accommodate
simple Javanese and Indonesian expressions as in (33a, b) and (34a, b), respectively.15
(33) Indonesian
a. tiga
three
buku.
b. buku
book
book
‘three books’
ini
this
‘this book’
(34) Javanese
a. telung buku.
three
book
‘three books’
b.
buku
iki.
book
this
‘this book’
The issue here is that expressions as in (33a, b) and (34a, b), which include numerals and
demonstratives, might be incorrectly ruled out under present assumptions because of the lack of
15
This question was raised by Andrew Carnie (personal communication) and Heidi Harley (personal
communication).
287
the QP or DP that would host the numeral and demonstrative words as their head in languages
such as English and Italian. The claim that demonstratives are semantically functioning as
determiners at least in the contemporary Indonesian has been made by MacDonald (1976: 85),
who observes that the demonstrative itu ‘that’ is “coming to fulfill a function very much like that
of the definite article.” A related question, of course, comes from our earlier observation that a
bare noun in Indonesian/Javanese can have a singular indefinite reading, as we have seen in the
examples in (6a, b) and (16a). Notice that these two questions, in fact, are two special cases of the
more general problem of how it is possible that a bare noun, which otherwise would denote a
mass extension, is also able to denote a singularity in Indonesian or Javanese.
I maintain, however, that this question receives a principled answer once we recall our
earlier assumption that the singularity of a noun is not encoded in the grammar; rather, it is
determined by contexts in which it appears. Consider first the issue raised by demonstratives.
Fukui (1986, 1995) and Fukui and Speas (1986) argue that Japanese entirely lacks functional
categories such as C, T, and D of the English-kind, or at best has a quite impoverished
system of these function items, with elements such as subjects and modifiers all being
attached to lexical projections as an X′ adjunct (see also Guilfoyle and Noonan 1992 and
Vainikka 1993/1994; see discussion in section 5). If this analysis can be extended to
Indonesian and Javanese, then one can maintain that numerals and demonstratives as in (33a,
288
b) and (34a, b) are also modifiers of lexical projections such as NPs in the structure in (32) in
these languages. There is independent evidence that demonstratives in Indonesian and
Javanese are not D heads but instead modifiers of NPs. The evidence concerns the iterativity of
such expressions. It is well known that lexical heads such as nouns do not impose any
structural limit on iterating modifiers as long as they can be semantically interpreted and
licensed. This is illustrated in examples in (35a-c) in English.
(35) a. a big balloon.
b. a red big balloon
c. a red big expensive balloon
Demonstrative words in English such as this and that, by contrast, are instances of the D
head, not modifiers of an NP, by the same criteria. This is shown by the contrast between
(36a, b) and (36c), where this/that can never co-occur with other functional D elements
such as John’s. In other words, D elements are uniterable.
289
(36) a.
b.
this/that book
John’s book
c. * this/that John’s book
Importantly, demonstratives in Indonesian and Javanese pattern with modifiers of the NP
rather than D heads since they can co-occur with other possessor elements, as in (37c)
and (38c).16
(37) Indonesian
a. buku
book
ini
this
‘this book’
b. buku
John
c. buku
John
ini
book
John
book
John
this
‘John’s book’
‘this John’s book’
(38) Javanese
a. buku
iki
book
‘this book’
16
this
b. buku-ne
John
c. buku
John
iki
buku-Poss John
book
John
this
‘John’s book’
‘this John’s book’
It is important that the example in (37c) is not acceptable under the following reading: There are several
people named John and the speaker talks about the book owned by one of them in contrast to the other. The
only reading available to this sentence is roughly “this (deictic) book which belongs to John.”
290
This contrast between (37c)/(38c) and (36c), therefore, provides independent support that
there is no D in Javanese and Indonesian and that demonstratives in these languages are
modifiers of the lexical NP projection. Further cross-linguistic evidence for this conclusion
is provided by Bernstein (1997). Bernstein points out that, in languages such as Arabic and
Greek, demonstratives and determiners can co-occur, suggesting that the former is not a D
head. Bernstein further observes that, in other languages such as Spanish, Swedish,
Norwegian and Scottish Gaelic, the demonstratives are the same words as ‘here’ and
‘there’. 17 Crucially, this observation holds at least for Indonesian as well, as shown in
examples in (39a, b), assuming that sini is an allomorph of ini.
17
Interestingly, Bernstein (p. 93) points out Javanese examples as in (i) as a case where a demonstrative and a
determiner may co-occur.
(i)
ika
n
anak
this
the child
‘this child’
(Bernstein 1997: 93)
This example, however, is unacceptable according to my consultants, who speak modern Javanese. First,
modern Javanese does not have any determiners like n ‘the’ in (i); recall our observation on Javanese made in
section 3.2. Second, the demonstrative iki ‘this’ is used instead of ika in (i). Thus, Bernstein’s second point does
not apply to Javanese.
291
(39) Indonesian
a. buku
di
sini.
book
in
here
‘A book/the book/books/the books are here.’
b.
buku ini
book this
‘This book/these books’
The expression in (39a) is a TP due to the lack of copula in Indonesian where the word sini is
used as a preposition. The expression in (39b) is a nominal phrase modified by the related
word ini. This homophony between words meaning ‘here’ in Indonesian, thus, provide
further support for the view that demonstratives are not necessarily D-elements but locative
modifiers arguably attached to NPs.
More generally, it may well be that natural language uses whatever syntactic resources
that are independently available to them to express the same meanings that other languages
would express with a different (more articulated) syntactic structure. 18 The fact that a bare
nominal can, in principle, occur with numerals despite the fact that it denotes a mass
extension also follows from the lack of the grammatical encoding of the singularity in
Indonesian/Javanese. Syntax simply makes the denotation of a bare nominal
underdetermined. It is the semantic interface that actually determines whether that particular
18
The idea in this paragraph owes a great deal to my discussion with Heidi Harley (personal communication).
292
instance of bare nominal denotes a count or mass extension. If a bare nominal occurs with
numerals such as 2, 3, 4, etc, then the interface coerces its otherwise mass denotation into
count denotation so it may be compatible with a specific numeral; if contexts in which a bare
nominal X is found make it clear that it is intended to denote a singularity, then it will come
out as denoting one instance of X at the semantic interface. This observation is supported
from the fact that once coupled with a numeral, a bare noun triggers scope interaction with
negation, as in English. This is illustrated by examples in (40) and (41), which minimally
contrast with (6a) and (16a) in that the bare nominals are accompanied with a numeral.
(40) Ali
tidak
jadi
membeli
tiga
buku.
Ali
Neg
finished
buy
three
book
Neg>Indefinite (narrow scope): Ali did not buy three books.
Indefinite>Neg (wide scope): There are three books that Ali failed to buy.
(41) Aku ora
I
Neg
weruh telung
see
three
kotoran
spot
ning jubin.
on
floor
Neg>Indefinite (narrow scope): I did not find three spots on the floor.
Indefinite>Neg (wide scope): There are three spots that I failed to find.
293
This observation, thus, provides important support for the idea behind the thesis of
minimalist interface that the semantic component is endowed with domain-specific
operations to yield an appropriate semantic interpretation to an otherwise underdetermined
denotation of a particular noun. Notice that this underdertmined interface theory of nominal
denotation predicts that there should not be any noun in Indonesian or Javanese whose
count/mass denotation cannot be computed solely from its (prototypical) conceptual
manifestation in the real world. This prediction seems to be borne out from the fact that
Indonesian does not have pairs of words such as oat and wheat in English, though I must
leave comprehensive examination of this prediction for my future research.19
19
Another important issue that the proposed analysis does not fully resolve is what restrictions are imposed
on this recursion. For example, it is impossible to freely concatenate more than one numeral in Indonesian, as
in (i), the intended meaning being that dua tiga buku means 2 x 3 = 6 books.
(i) * Esti
Esti
mem-beli
dua
tiga
buku.
Av-buy
two
three book
‘Esti bought
I believe that the impossibility of this type of recursion violates certain general semantic or pragmatic constraints
such as expressive economy and Gricean Maxims. In terms of expressive economy, enam ‘six’ is a more
economical way of denoting 6 than multiplication of 2 x 3. The same word also satisfies the Maxim of Quantity
better than 2 x 3 because it is more concise than 2 x 3. This question might lead us to a new comparative analysis of
the syntax-semantics interface with respect to complex cardinals. See Ionin and Matushansky (2006) for a
promising compositional approach to the semantics of numerals which do behave in this way, for example, three
hundred. The comprehensive discussion of the cross-linguistic typology of number systems goes beyond the scope
of this chapter. I thank Heidi Harley (personal communication) for useful discussion on the question raised here.
294
Let us now turn to the nominal syntax of bare nouns in Japanese. Japanese has the
following morphosyntactic characteristics; i) bare arguments, ii) the generalized classifier
system, iii) the extension of all nouns is mass, and iv) the obligatory narrow scope of bare
nouns with respect to negation. The first three properties were noted by Chierchia (1998a,
b); the last property is illustrated in the contrast between examples in (42a, b).
(42)a.
John-ga
yuka-de
John-Nom floor-Loc
yogore-o
mituke-naka-tta
(koto)
dirt-Acc
find-Neg-Past
(fact)
‘(The fact that) John did not find a spot/any spot/spots on the floor.’
Neg>indefinite (narrow scope): I did not see any spot on the floor.’
* indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.’
b.
John-ga yuka-de
mituke-naka-tta
yogore-ga
aru
John
find-Neg-Past
dirt-Nom
exist
floor-Loc
(koto).
‘(The fact that) There is a spot that John did not find on the floor.’
* Neg>indefinite (narrow scope): I did not see any spot on the floor.’
indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.’
295
The bare noun yogore ‘dirt’ cannot take scope over the negative morpheme nai ‘not’ as
shown in (42a). The wide scope reading is expressed by the relative clause structure, as
illustrated in (42b).
I propose that the nominal structure in Japanese is as shown in (43) where bare
nominals project up to QP with the Num specification being {neutral, plural}, as in
Javanese and Indonesian. I assume, following Kitagawa and Ross (1982) and Watanabe
(2006) (see also Saito and Murasugi 1990 for relevant discussion), that what otherwise
looks like a genitive case marker no in the expression san satu-no hon ‘three-Cl books’
that intervenes between the classifier and the head noun is a linker that is inserted postsyntactically after any non-clausal pre-nominal element.20
20
Evidence for this assumption is that this morphological treatment is able to capture the distributional
properties of no better than a purely syntactic treatment. Specifically, Kitagawa and Ross (1982) make the
generalization that the marker no is attached to a non-clausal prenominal element of any kind. Watanabe (p.
256) also notes that this marker “can even iterate after every such non-clausal element”, as shown in (ia, b):
(i) a.
san-satsu-no
Chomsky-nitsuite-no
hon
3-Class-Gen
Chomsky-about-Gen
book
‘three books about Chomsky’
b.
tsugi-no
suugaku-no
mondai
next-Gen
math-Gen
problem
‘(the) next math problem’
(Watanabe 2006: 256)
These examples indicate that it is difficult, if not impossible, to specify a unique structural position for noinsertion in Japanese grammar. Rather, the distribution of this maker is better captured by stating that it is
inserted after any non-clausal element in the post-syntactic morphological component.
296
(43) The Nominal Structure of Bare Nouns in Japanese
QP
Q
ClP
san
Cl
‘three’
satu
NumP
Num
no {neutral, plural}
N
hon ‘book’
The above-noted morphosyntactic profile of nominals in Japanese directly follows. First,
Japanese allows bare nominals in any argument position because there is no DP projection in
the structure in (43). Second, Japanese has a generalized classifier system due to the
projection up to QP which dominates the ClP. Third, the extension of all nouns is mass for
the same reason that the extension of all nouns is mass in Javanese and Indonesia: whichever
value the Num head takes, the denotation of the NumP is a kind, which is mass. Finally, bare
nominals in Japanese can only take narrow scope with respect to negation due to their kinddenoting requirement that is independently known to block wide scope readings. Thus, the
morphosyntactic profile of [+arg, -pred] languages as in Japanese follows as an automatic
consequence of the fact that Japanese nominals project only up to QP without the DP
projection with the number values {neutral, plural}. Notice that the proposed analysis of
297
nominal structure in Japanese predicts that there are no grammaticalized mass-count
distinctions in nominals, as exhibited by the oat vs. wheat pair, that is not solely computable
from its conceptual realization. This prediction receives support from the observation that the
translation equivalents of English mass nouns such as furniture are count nouns in the
lexicon of Japanese speakers; it represents a conceptually identifiable discrete object such as
couch, chair, and so on. The proposed analysis also predicts that bare nominals in Japanese
should be able to exhibit scope interaction with negation if we introduce numerals. The
comparison of (42a) and (44) shows that this prediction is confirmed.
(44) John-ga
John-Nom
yuka-de
mittu-no
yogore-o
floor-Loc
three-Link dirt-Acc
mituke-naka-tta
(koto)
find-Neg-Past
(fact)
‘(The fact that) John did not find a spot/any spot/spots on the floor.’
Neg>indefinite (narrow scope): I did not see three spots on the floor.’
indefinite>Neg (wide scope): There are three spots that I failed to see on the floor.’
298
This fact thus provides strong support for the assumption that singularity is not a grammarinternal property of Japanese but rather is the byproduct of semantic disambiguation at the
semantic interface.21
The idea that Japanese does not have anything like D heads in English has been a
traditional one within generative research on this language since the seminal work by Fukui
(1986, 1995) and Fukui and Speas (1986). Noguchi (1997) provides further arguments in
favor of this position based on the contrast between English and Japanese with respect to the
availability of variable binding for pronouns. One potential candidate for the D head in
Japanese is a class of demonstratives such as kono ‘this’ and ano ‘that’. However, the
iterativity test we introduced above in our discussion of Indonesian and Javanese shows that
these elements are not D heads but instead modifiers of some other projection such as NP or
the whole QP in the structure in (43) because, as Fukui (1986, 1995) notes, these elements
can co-occur with other possessor or pronominal adjectives, as illustrated in (45c).
21
As Heidi Harley (personal communication) points out, the present analysis of scope variability in
Indonesian/Javanese and Japanese allows us to raise a new question of why the phrase headed by the numeral
phrase can undergo QR but bare nominals cannot. I speculate that numerals serve (part of) the same discourse
function as Ds; D has the function of mapping the property denoted by a common noun (NP) onto whatever
real-world entity the property holds true for. This discourse-oriented function is also shared by numerals.
299
(45) a. ano kuruma
that car
‘that car’
b. John-no ano kuruma
John-Gen that car
‘*John’s that car’
c. ookina John-no
big
ano kuruma
John-Gen that car
‘*big John’s that car’
(Fukui 1995: 106, 107)
Let us now consider the nominal syntax of bare nouns in Italian. Italian, being one of the
examples of the [+arg, -pred] language under Chierchia’s (1998a, b) Nominal Mapping
Parameter, does not allow bare arguments. The unavailability of bare nominal arguments in
this language directly follows if we assume that Italian nouns must always project up to DPs,
hence instantiate the maximally complex nominal structure among languages of the world. We
have also seen that languages such as Italian can take either {singular, plural} or {neutral,
plural} for the Num value. The structure for Italian nominals, thus, will be as in (46) or (47),
depending on whether the Num value is specified either as {singular, plural} or {neutral,
plural}. Note that nominals in Italian project up to DP in both cases. I assume that de ‘of’ in the
structure in (47) is inserted post-syntactically as a linker.
300
(46) Minimal Nominal Projection in Italian (for count nouns)
DP
D
il
QP
Q
NumP
‘the’
Num
N
{singular, plural}
libro ‘book’
(47) Minima Nominal Projection in Italian (for mass nouns)
DP
D
e
QP
Q
tre
ClP
Cl
‘three’ bicchiere
NumP
Num
‘glasses’ d’ {neutral, plural}
N
acqua ‘water’
We saw in section 2 that Italian disallows bare nominal arguments. This property falls out
because Italian nominals always project up to the DP.
301
Evidence that Italian nominals always project up to DPs comes from the subject-object
asymmetry in Italian noted by Chierchia (1998a: 356), who observes that bare nominal
arguments are allowed in direct object positions in certain cases but never permitted in subject
positions, as illustrated by the contrast between (48a) and (48b). The same observation is also
made by Longobardi (1994: 616), who points out the contrast between (49a) and (49b, c).
(48) a. * Bambini sono venuti da noi. b. Ho
kids
be
come by us
water
‘I ate cookies with my milk.’
viene
giù
dalle
colline.
comes
down from
the-hills
‘Water comes down from the hills.’
b.
il
mio latte.
I-have taken cookie with the my
‘Kids came by us.’
(49) a. * Acqua
preso bicotti con
Viene
giù
acqua dalle
colline.
comes
down water from
the-hills
‘Down from the hills comes water.’
milk
302
c.
Ho presco
acqua
dalla
soregente.
I
water
from
the-spring
took
‘I took water from the spring.’
(Longobardi 1994: 616)
The asymmetry observed here naturally follows if Italian nouns always project up to a DP with
an empty head; see Contreras (1986) for a similar analysis in Spanish. One standard
assumption in the generative framework has been that empty heads must be properly licensed
by appropriate heads (Chomsky 1981, 1986a, b; Rizzi 1990). Under this assumption, the null
D head in (48b) that dominates the bare noun bicotti ‘cookie’ is correctly licensed by the verbal
head preso ‘take’. This licensing option is unavailable for the empty D head that dominates
bambini ‘kids’ in (48a). A similar story holds for the contrast between (49a) and (49b, c). Thus,
the subject object asymmetry provides support for the DP structure for Italian nominals.
Alternatively, one could maintain that nominals in Italian themselves do not project to DP but
the specifier position of TP requires a D head for EPP reasons. Then, nominals in subject
position must be headed by a D head whereas nominals in object position need not (hence
cannot) be headed by a D head.22
22
I thank Andrew Carnie (personal communication) for suggesting this alternative analysis and Simin Karimi
(personal communication) for directing my attention to Contreras (1986).
303
Consider finally the syntax of bare nominals in English. We have seen in the previous
section that, under Chierchia’s Nominal Mapping Parameter, English belongs to the [+arg,
+pred] language type. This means that this language behaves as Japanese and Chinese in
that the extension of its bare plural and mass nouns is a kind (hence [+arg]) whereas
behaving as Italian and French in that it prohibits count nouns from occurring without
determiners (hence [+pred]). I propose that this dual behavior of English nouns is exactly
what we predict under the proposed analysis if English can choose the Japanese-type QPstructure or the Italian-type DP structure. Consider first the Japanese-type structure assigned
to English when bare plurals and mass nouns are involved. In this case, English allows bare
arguments, requires a classifier system, and does not have plural morphology, as in (50a-c).
These properties mirror exactly those observed in Japanese.
(50) a.
b.
I drank water.
I drank three glasses of water.
c. * I drank waters.
Thus, I propose the nominal structure for bare plurals and mass nouns as shown in (51),
which is the Japanese-type nominal structure; it projects up to QP with the Num value
304
being chosen from the {neutral, plural} set. I assume that the preposition of is inserted in
the post-syntactic morphological component between classifiers and their nominal heads
as in three glasses of water.
(51) The Nominal Structure in English (for bare plurals and mass nouns)
QP
Q
three
ClP
Cl
NumP
glasses Num
of {neutral}
N
water
The three morphosyntactic properties of bare plurals and mass nouns in English noted above are
derived automatically by virtue of the fact that English has the Japanese-type QP structure in this
context. The bare nominal option is possible because there is no DP on top of the QP. The Num
specification in (51) requires that the denotation of the NumP be a kind. Thus, a certain set of
classifier-like expressions such as glass, cup, and piece is required for nouns in (51) to set up an
appropriate counting level for each noun, just as in languages like Indonesian, Javanese, and
Japanese wherein all bare nouns denote a kind. There is no plural morphology observed in bare
305
plurals or mass nouns because they are true in an undifferentiated manner of a singular or plural
instance of the entity denoted by this type of noun. 23
If English is like Japanese, the proposed analysis also leads us to the prediction that
bare plurals and mass nouns cannot take wide scope over negation due to their kinddenoting requirement. This prediction is indeed confirmed by (5a), repeated here as (52).
(52) I didn’t see spots on the floor.
Neg>Indefinite (narrow scope): I did not see any spot on the floor.
*Indefinite>Neg (wide scope): There is a spot/are certain spots that I failed to see on the floor.
Consider now the structure for count nouns in English. When count nouns are involved,
English does not allow bare nominal arguments (53a, b), lacks any classifier (53c), has plural
morphology (53d), a cluster of properties that we have seen to characterize nominals in Italian.
23
I have no good answer at this point of research for what wound be wrong with using the null varieties of Q, Cl,
and the Num heads in the derivation in (51) if D could be phonologically unrealized in Italian. This question is one
special case of the broader question of why there are certain tendencies in natural language for a particular value
(such as third person, not first or second person) to be null rather than overt. One might find clues to this question
from learnability, but this is a speculation.
306
(53) a.
I bought a pen.
b.* I bought pen.
c. * I bought a piece of pen.
d.
I bought pens.
Thus, English takes the Italian-type nominal functional structure shown in (54). I assume
that the indefinite article a is base-generated under the D head. Alternatively, a realizes the
Q head (that is raised to the D head), since it denotes a singular instance of a discrete,
countable entity; recall Chierchia’s (1998b: 91, 92) observation noted above; see also
Longobardi (1994).
(54) The Nominal Structure in English (for count nouns)
DP
D
a
QP
Q
NumP
Num
{singular}
N
chair
307
The determiner-less bare option for count nouns is impossible for count nouns because
the nominal structure projects up to the DP. When the singular value is selected for the
Num head, the denotation of the NumP is a singular instance of the chair, which is
compatible with the function of the indefinite article a. When the plural value is selected
instead, the denotation of the NumP is bare plurals, which is a name of kind, as argued
for in Carlson and Chierchia.
There are two potential problems with the proposed analysis for nominal syntax in English.
The first problem concerns the selectional relation between the Num head and its N
complement.24 The question is, why it is that only bare plurals and mass nouns are inserted in
the structure in (51) whereas only count nouns are inserted in the structure in (54)? What
blocks count nouns and bare plurals/mass nouns from being inserted in the structures in (51)
and (54), in that order? I maintain that there is actually nothing wrong with this choice as far as
syntax is concerned; the syntax-external component interprets whatever syntactic object the
narrow syntax creates and sends out. This position is related to what Hinzen (2006) calls
Semantic Blindness; “as if syntax carved the path interpretation must blindly follow”
(Uriagereka 2002: 275, as quoted in Hinzen 2006: 250). In other words, "the human language
faculty provides forms that a possible human structured meaning may have, leaving a residue
24
I thank Heidi Harley (personal communication) and Vicki Carstens (personal communication) for this
question and useful discussion on how to address this question.
308
of non-structured meanings (concepts)." (Hinzen 2002: 235). It has been widely known that,
when the meaning of an open class nominal element clashes with that of determiners, it is
always determiners whose interpretation molds that of nouns (Harley and Noyer 2000: Harley
2006: 213-214). For example, mass nouns could occur with determiners that specifically select
count nouns; thus, not only (55a) but also (55b) is possible in English.
(55) a. I don’t drink much coffee.
b. I bought two coffees this morning.
(56) a. I had a cookie for breakfast.
b. That baby has cookie all over his face.
With an appropriate context, two coffees in (55b) can be interpreted as packaged coffees
in cups or bags. Similarly, we can use count nouns together with determiners that
specifically select mass nouns. In (56b), the bare nominal cookie is not interpreted as a
discrete entity but instead as amorphous substance that cookies are generally made of.
((55a, b) and (56a) are from Harley 2006: 213.) These examples illustrate that the
meaning of the noun is always bent to be compatible with the semantic contribution of
the determiner which it co-occurs, not the other way around. In other words, there is no
ungrammatical combination of nouns and functional elements within syntax per se, with
the semantics trying its best to get a felicitous interpretation that is compatible with
309
world knowledge. Thus, when we have sentences like (57), to which no stretch of our
encyclopedic knowledge can assign any reasonable interpretation, they are anomalous
solely by our syntax-external criteria.
(57) # I had three oxygens in the kitchen.
(Harley and Noyer 2000: 21)
The second potential problem with the proposed analysis concerns the modifiability of
elements like glasses, cups and pieces, which I have thus far analyzed as a classifier on a
par with Japanese classifiers.25 This unified treatment appears to be incorrect, given that
the former may be modified by adjectives while the latter may not, as the contrast between
(58a) and (58b) shows.26
(58) a.
three big cups of coffee
b. * ni
two
ookina hai-no
kohii
big
coffee
Class-Gen
‘two big cups of coffee’
25
As pointed out by Jaeshil Kim (personal communication) when I presented material in this chapter at the
Mid-America Linguistic Conference held at the University of Kansas, Lawrence in October 2007.
26
(58d) is acceptable in a context where the speaker talks about brands of coffee beans, for example.
310
c.
ookina
ni-hai-no
kohii
big
two-Class-Gen
coffee
‘two big cups of coffee’
d. # ookina kohii
big
coffee
‘big coffee’
The contrast between (58a) and (58b) receives a natural treatment in prosodic terms. We saw
in section 3.1 that the numeral se- ‘one’ in Indonesian needs an overt classifier on its right due
to its clitic/boundedness requirement. Following this line of thinking, (58b) is ungrammatical
because the adjacency between the numeral ni- ‘two’ to the classifier hon is blocked by the
intervening adjective ookina ‘big’ (cf. Bobaljik’s (1995) Morphological Merger under
Adjacency). Thus, if we change the relative order of the adjective and numeral so that the two
elements are phonologically adjacent, the result is grammatical, as in (58c). Notice that the
adjective ookina ‘big’ modifies the classifier, not the head noun kohii ‘coffee’, in (58c), since
the same adjective cannot co-occur with the same noun without the classifier hai, as in (58d).
To sum up this section, I have proposed a novel, relativized parametric theory of the
morphosyntax and denotation of bare nominals across languages whereby different
311
morphosyntactic profiles exhibited by nominals in different languages are derivable from the
interaction of two independently morphosyntactic/morphosemantic parameters: how high each
language allows its bare nominal to grow and what set of Numb values each language can
choose. The proposed analysis does not require any rigid mapping between the syntax and
semantics of bare nominals, as in Chierchia’s (1998a, b) Nominal Mapping Parameter but
captures its effects through independently motivated observations about the available set of
functional items and of the possible number values in each language, in a way that is
compatible with the standard generative assumption about the locus of parameters.
5. New Typological Predictions of the Proposed Analysis
The relativized parametric theory of the denotation and morphosyntax of bare nominals
makes certain predictions concerning several other languages and their morphosyntactic
profile. In this section, I mention three languages, Child English, Russian and Chinese, whose
morphosyntactic profile is naturally predicted by the proposed theory.
5.1. Child English Morphosyntax
The Subset Principle, proposed by Wexler and Manzini (1987), states that child language
acquisition starts with the setting of the parameter values that rules out the most, so that
312
children can revise and modify their working hypothesizes solely on the basis of positive
evidence. Following the standard assumption that only positive evidence is available to
children learning a language, Wexler and Manzini define the Subset Principle as in (59).
(59) The Subset Principle
Suppose one value of a parameter yields a language L(i) and another value of the
parameter yields a language L(j). Suppose further that L(i) is a smaller language that L(j),
that is, that L(i) is contained in L(j). L(i) is a strict subset of L(j). Then the learning strategy
specified by the Subset Principle is that the learner select the value which yields L(i) first. If
this is the correct choice, there will never be evidence that it isn’t, and the learner will stay
with the value. If this is the wrong choice, then there will be positive evidence (sentences
from L(j)) which are not in L(i)) which the learner will eventually hear; this evidence must
exist, because L(i) is a strict subset of L(j).The Subset Principle specifies that when positive
evidence which shows that L(i) is the wrong language is encountered, the learner will
switch to the parameter value that yields language. In short, the Subset Principle is a
method for specifying a markedness hierarchy when alternative values yield languages that
are in a subset relation. (Wexler and Manzini 1987: 44)
313
As mentioned in the previous section, the idea that children learn nominal syntax in a bottomup way has been independently proposed by Guilfoyle and Noonan (1992) and supported by
Vainikka (1993/1994). Guilfoyle and Noonan argue that Functional Grammar is not present in
the early stages of language acquisition based on well-known acquisitional properties such as
the optionality of subjects, the absence of NP-movement, subject-auxiliary inversion, and verb
placement facts. To repeat one argument based on the absence of NP-movement, Borer and
Wexler (1987) observe that children treat all passives as adjectival passives and claim that this
observation naturally follows under the syntactic NP-movement account of verbal passives (se
Baker et al. 1989 for a detailed analysis) if children cannot form A-chains. This maturation
analysis is untenable according to Guilfoyle and Noonan because principles of UG such as the
θ-Criterion, the EPP, and the Projection Principle (Chomsky 1981, 1986a) built into their
grammar would not be able to block their association of two structural positions with a single
θ-role. Instead, Guilfoyle and Noonan argue that the above observation is straightforwardly
derived if there is simply no IP in the child language syntax, so that the movement of an NP
can target: as a result, children treat all cases of passive as adjectival because formation of this
type of passive is a process that does not require IPs (Wasow 1977; Levin and Rappaport
1986). This ‘bottom-up’ incremental view of phrase structure is also supported by Vainikka’s
314
(1993/1994) developmental evidence that the emergence of nominative subjects occurs around
the same time as that of TP-related material like modals, auxiliaries, and past tense.
The relativized parametric theory, in combination with the Subset Principle, makes an
interesting prediction; early stages of acquisition of nominal syntax in all languages should
mirror the morphosyntactic profile of bare nominals in Javanese and Indonesian because
these two languages instantiate the simplest nominal functional structure (NumP).
Specifically, we predict that initial stages of all child languages should allow bare nominal
arguments, lack a classifier system for nouns, and iii) have plural morphology. This
prediction is indeed borne out by what the literature on acquisition calls the Telegraphic
Speech, as illustrated in English examples as in (60); see also Chierchia (1998a: 400) for a
similar remark made on the basis of his Nominal Mapping Parameter.
(60) Telegraphic Speech 27
25
27
[dan/ I/ tsI/]
“don’t eat (the) chip”
[bwd/ tat]
“block (is on) top”
The number on the left-hand side indicates the age in months at which the utterances on the right were uttered.
315
26
27
28
[mamis tu hQs]
“Mommy’s two hands”
[mo b√s go]
“Where bus go?”
[dQdi go]
“Where Daddy go?”
[/aj gat tu djus]
“I got two (glasses of) juice”
[do baj/ mi]
“don’t bite (kiss) me”
[k√d´r s√ni ber]
“Sonny color(ed a) bear”
[/aj gat
pwedIs] “I (‘m) play(ing with) this.’
[mamIs tak mEns]
“Mommy talk(ed to the) men”
(Fromkin et al. 2003: 365)
The examples illustrate use of generalized bare arguments (i.e. chip, block, bus, juice, bear),
lack of classifiers for mass nouns (i.e. juice), and presence of plural morphology (hands, men),
a clustering of properties that we have seen to hold for Javanese and Indonesian. This behavior
of nominals in Child English therefore provides evidence for the proposed cross-linguistic
analysis of bare nominals and for the Structure Building Hypothesis of Guilfoyle and Noonan
(1992) from the domain of nominal syntax.
The current analysis also makes two other predictions concerning child English syntax.
First, nominals in child English syntax should show the interpretive property of taking narrow
316
scope with respect to negation and lack plural forms for generic statements, as in Indonesian
and Javanese. I have not found any literature that focus on this specific property in child
language acquisition.28 However, the experiment conducted by Pérez-Leroux and Roeper
(1999) provides strong experimental evidence that English-learning children are sensitive to
the presence or absence of D heads in calculation of differences in scope between the
following pair of examples: Everybody went home vs. Everybody went to his home. Second,
recall that I have argued above that expressions such as demonstratives should be analyzed as
modifiers of the NP projection in D-less languages such as Indonesian, Javanese, and Japanese,
on the grounds that this type of element can co-occurr with other modifying elements such as
possessives and adjectives. If child English starts as Indonesian and Javanese, we predict that
there should be instances where what are analzyed as D elements in Adult English should be
able to co-occur with similar types of elements. This prediction is borne out by examples as in
(61a-c), docmuneted by Miller and Ervin-Tripp (1973), where “determiners” co-occur with
possessives and adjectives.
28
As Cecile McKee (personal communication) points out, there is a linkage problem as well because children in
the scope experiments conducted in current acquisitional study often tend to be well past the telegraphic stage.
317
(61) a. Is that the blue mine?
b. this a Bonnie pants
c. I know a that.
d . These a Lidz pants
e.
mine, all a mine.
(Miller and Ervin-Tripp 1973: 363)
The iterability of “functional” elements in child syntax, therefore, provides further
evidence that nominal syntax in Child English does not project up to the DP for count
nouns. This property also supports my argument, made in section 4, that languages do
whatever they can do with their available syntactic and lexical resources available to their
grammars to arrive at the same interpretation that other languages would yield with more
elaborated functional structures.
5.2. Slavic MorphoSyntax
The relativized parametric theory proposed in this chapter also sheds new light on the nominal
syntax in Slavic languages such as Russian, which Chierchia (1998a, b) briefly mentions as a
language of the [+arg, +pred] type on a par with English. The morphosyntactic profile of
Russian is as follows: i) bare nominals are permitted in argument positions, ii) a generalized
318
classifier system is missing, iii) bare nominals may take either wide or narrow scope with
respect to negation. This last scope-related property is illustrated in (62a, b).29 Note that this
contrasts sharply with the behavior of bare nominals under negation in Indonesian and
Javanese illustrated above.
(62) a.
Ya ne
I
vizhu ni
Neg see
odnogo pyanta na polu.
no one
spot
on floor
Neg>indefinite (narrow scope): I did not see a single spot on the floor.
*Indefinite>Neg (wide scope): There is a spot on the floor that I failed to see.’
b. Ya ne
I
zametila ni
Neg notice
odnogo pyatna na polu.
no single
spot
on floor
‘There might have been one spot on the floor that I failed to notice.’
Neg>indefinite (narrow scope): I did not see a single spot on the floor.’
indefinite>Neg (wide scope): There is a spot on the floor that I failed to see.’
Under the proposed analysis, the morphosyntactic profile of Russian falls into place if bare
nouns in this language project up to NumP, as in Javanese and Indonesian, but with its
29
Thanks to Tatyana Slobodchikoff (personal communication) for useful discussion on Russian syntax.
319
possible set of the Num values being either {singular, plural} or {neutral, plural}, as in
English and Italian. The availability of bare nominal arguments and the lack of a generalized
classifier system results from the NumP nominal structure. The scope variability of bare
nominals as shown in (62a, b) results when a N is selected by the Num head with the
singular specification, as in English and Italian count nouns. In this way, the proposed
analysis can serve to classify languages like Russian in terms of the complexity of nominal
projections and the possible number values and derive their morphosyntactic profile from the
interaction of the two parameters.
5.3. Chinese Morphosyntax
The proposed theory also has something to say about the nominal syntax in Chinese. Recall the
universal nominal morphosyntactic hierachy shown in (17). I have provided evidence that the
DP, QP, and NumP options are instantiated by Italian/English (for count nouns),
Japanese/English (for bare plurals and mass nouns) and Javanese/Indonesian, respectively. The
proposed analysis predicts that there are languages that instantiate the ClP option. Of course,
there may be conceptual reasons why languages of this type are not easy to find. As Iljic (1994:
104) notes (see also Croft 1994), clasisifers have the function of individuation, which makes it
possible to extract “discrete occurences.’ This individuation function may well naturally tie with
320
that of numerals to produce expressions as in three cups of coffee, thereby blocking the ClP
option from being utilized across the board in natural language syntax. However, as we have
seen above, English can choose among structures depending on the nature of the bare nouns
involved. Therefore, the proposed analysis still leads us to expect that some languages might
instantiate the ClP option in a restricted range of circumstances. Importantly, Cheng and
Sybesma (1999) show that definite bare nominals in Mandarin and Cantonese project up to
ClPs whereas indefinite bare nominals in these languages project up to Numeral P (QP under
the proposed analysis). According to their view, the definite interpretation is derived from the
ClP structure under their assumption that N-to-Cl movement feeds the generation of the ιoperator, which, according to Chierchia (1998a: 359), “defines selects the greatest element from
the extension of a predicate and constitutes typically the meaning of a definite article.” (see
Cheng and Sybesma 1999; 524 for detailed discussion). The indefinite interpretation, on the
other hand, is derived from the NumeralP/QP option since “the numeral apparently has the
effect of undoing the definiteness.” (Cheng and Sybesma 1999: 524). The argument above
therefore shows that Cantonese and Mandarin are optional ClP languages.
This analysis further allows us to correcly predict several other morphosyntactic
properties of bare nominals in Madarin and Cantonese that are indeed borne out by
Cheng and Sybesma’s findings. First, these two languages allow bare nominal arguments
321
because these languages do not project up to DPs. Second, these languages have a
generalized classifier system due to the projection up to ClPs or QPs that dominate them.
Third, Cheng and Sybesma (1999: 519) observes that “Chinese bare nouns can be
interpreted as both singular and plural.” This property follows from the choice of the
feature set [neutral, plural] in these languages.30 Finally, bare nominals should not be
able to take scope over negation, a prediction that is borne out by the obligatory narrow
scope of bare nominals as in (63a) in contrast to (63b).31
(63)a. di
ban
floor
shang yil
board on
ge
ban-dian wo mei kan-dao.
one Class spot
I
Neg see
‘I did not see a spot on the floor.’
Neg>Indefinite (narrow scope): I did not see any spot on the floor.
* Indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.’
30
As stressed by Heidi Harley (personal communication), though, this feature set itself does not exclude the
possibility of plural morphology in Mandarin and Cantonese because that is exactly the feature set in
Indonesian, Javanese, and Japanese, which all have such morphology. One might regard (yī) xiē and –men as
two possible markers of plurality in Mandarin but this view is quite controversial because Iljic (1994) argues
that these morphemes are not plural but rather collectives that refer to wholes. For this reason, I leave the
issue of whether Mandarin/Cantonese has genuine plural morphology open here.
31
I thank Sunjing Ji (personal communication) for her help in constructing the examples in (63a, b).
322
b. di
floor
ban
shang
you
ge
ban-dian
wo
mei
kan-dao.
board
on
exist
Class
spot
I
not
see
‘There is a spot that I failed to see on the floor.’
* Neg>Indefinite (narrow scope): I did not see any spot on the floor.
Indefinite>Neg (wide scope): There is a spot that I failed to see on the floor.
In this way, the proposed analysis correcly characterizes the morphosyntactic profile of
bare nominals in Mandarin/Cantonese by assuming that they project up to ClPs/QPs
depending on the definiteness with the Num value being set as {neutral, plural}.
5.4. Other Languages
The relativized parametric theory of nominal morphosyntax may well predict some other
language types once the research goes beyond those languages discussed in this chapter. Thus,
in a series of recent work, Schmitt and Munn (1991, 2002) present evidence that Brazilian
Portuguese provides a counterexample to Chierchia’s Nominal Mapping Parameter and argue
for a purely syntactic account of the morphosyntactic profile of bare nominals in this language
by extending Bobaljik’s (1995) Free Agr Parameter to nominal domains. Importantly,
Schemitt and Munn observe that bare singulars in this language must take narrow scope with
323
respect to negation and are underspecified for number. Under present assumptions, this
observation means that the Num value must be selected from {neutral, plural} set as in
Javanese, Indonesian, Japanese, and Mandarin/Cantonese. Schmitt and Munn conclude that
bare singulars in Brazilian Portuguese are DPs. Within the present framework, combining this
conclusion with the above-noted observation on Number, we can conclude that bare nominals
always project up to DP, as in Italian and English, with the Num head being specified for
{neutral, plural}, as in Javanese, Indonesian, Japanese and Mandarin/Cantonese. Schmitt and
Munn note many detailed interpretive subtlities involved in the use of bare singulars caused by
the nature of predicates, espisodic contexts, and so on, so the situation is more complex than
stated here. Howerver, the fact that the core morphosyntactic property of bare nominals in
Brazilian Portuguese can be characterized as the interaction of the morphosyntactic complexity
with the possible Num values indicates that the proposed analysis is on the right track.
6. Chapter Summary: Theoretical Implications for the Syntax-Semantics Interface
Let us summarize what we have done in this chapter. I have discussed the issue of syntaxsemantics interface with reference to the denotation and morphosyntax of bare nominals in
several languages, in particular, Indonesian and Javanese. I have shown that these two
languages do not fit into any one of the three language types predicted by Chierchia’s (1998a,
324
b) Nominal Mapping Parameter. First, the free occurrence of bare arguments in Indonesian and
Javanese shows that these languages are not [-arg, +pred] languages such as Italian. Second,
the absence of a generalized classifier system and the presence of plural morphology marked
by reduplication of a root argue against catgeorizing the two languages as [+arg, -pred]
languages such as Japanese and Chinese. Finally, the obligatory narrow scope of bare
nominals with respect to negation and the lack of pluralized/reduplicated forms for generic
statements suggests that Indonesian and Javanese are also not [+arg, +pred] languages such as
English. This result casts serious doubts on the the rigid mapping between the syntax and
semantics of bare nominals of the kind assumed in Chierchia’s Nominal Mapping Parameter.
Following the standard conception of the locus of parameters in the Principles-&Parameters approach to language variation (Borer 1984; Fukui 1986, 1995; Chomsky 1981,
1986b, 1995), I have proposed a relatived parametric theory of nominal denotation that
derives different morphosyntactic profiles of bare nominals in different languages from the
relative complexity of nominal functional structures and the possible set of Num values
that are available in each language. The complete summary of this theory is summarized in
Table 4 with a full range of languages dicussed in this paper.
325
Table 4: A Relativized Parametric Theory of Nominal Denotation (final version)
Languages
Indonesian, Javanese
Russian
Chinese
Chinese, Japanese
Num Values
{neutral, plural}
{singular, neutral, plural}
Nominal Syntax
NumP
NumP
{neutral, plural}
ClP
{neutral, plural}
QP
Italian
{singular, plural, neutral}
English
{singular, plural, neutral}
DP
DP or QP
I have also shown that the proposed analysis makes new predictions concerning a few
other languages that have not been dicussed in detail in Chierchia (1998a, b). Coupled with
the Subset Principle of Wexler and Manzini (1987), the proposed analysis predicts that
nominals in the early stages of language acqusiition should show the same set of
morphosyntactic properties as those in Javanese and Indonesian because these languages
instantiate the simplex nominal functional structure. I have shown that this prediction is
borne out by telegraphic speeches prodcued by English-learning children. I have also
argued that Russian is minimally difference from Javanese and Indonesian in that the
326
former projects only up to NumP as in Indonesian and Javanese but with the Num values
being either {singular, plural} or {neutral, plural}, as in English and Italian.
As is true with the other chapters in this dissertation, the major goal of this chapter has
been to propose a new theory of language variation within the Principles & Parameters
approach with special refenrece to the morphosyntax of bare nominals. The case study
reported here highlights the importance of under-studied languages as in Indonesian and
Javanese because generalizations concerning these languages often deviate from those that
have been found among relatively well-studied languages such as English, Romance, and
East Asian languages, hence reveal crucial gaps that otherwise would go unnoticed in
modern syntactic theorizing. I have shown here that detailed examination of these
languages as conducted here is instructive in informing a more “balanced” parametric
theory of human language syntax, a desidaratum in the current generative enterprise.
The proposed analysis has several imporant implications for the proper theory of syntaxsemantics interface. First of all, the proposed analysis of bare nominals across languages
provides support for a certain conception on the economy of derivation and projection at the
syntax-semantics interface: the syntax-external interpretive component employs whatever
syntatic recourses are avalailable to the language to express the same denotation that other
languages would express with a more complex nominal functional structures. This was seen,
327
for example, in the treatment of demonstratives and numerals as adjuncts/modifiers of lexical
projections such as NPs in Indonesian, Javanese and Japanese, which would consitute D and Q
heads in languages such as Adult English and Italian. This point also applies to the
morphosyntactic development of bare nominals in Child English, which suggests that child
language acquisiiton also respects structural economy in the sense that it posits minimal
structures necessary to analyze the data available to children unitil it ends with the QP/DP
structure (Radford 1990). We also have noted that various denotations assigned to a particular
nominal element (kind, predicate, indefinite, etc.) are the interpetive outcome of the semantics
component that interpretes a different height of the nominal functional projection and of the
Num values. Similarly, when the functional requirement of a determiner clashes with the
conceptual structure of its complement nouns (mass vs. count), it is always the former that
prevails, forcing the semantic interpretive component to interpret the output of syntax in every
way compatible with our knowledge of whether particular nouns can be conceptualized as
discreet individualizable objects or amorphous discreet mass and how. These results provide
strong support for the notion of “minimalist interfaces” introduced in chapter 1, whereby the
syntactic computation provides a parametrically defined curve that the conceptual/semantic
interfaces must blindly follow, without any extrinsically determined mapping between the
syntax and semantics of a particular expression as in Chierchia’s (1998a, b) Nominal Mapping
328
Parameter. The manner in which the syntactic representation is mapped onto the semantic
representation is language-invariant. The narrow syntactic computation will do whatever it can
within a parametrically chosen set of morphosyntactic features and their projections in each
language, and the universal semantics will just come up with an interpretation that is
compatible with our conceptual knowledge of how things are represented in the external world.
Second, the proposed parametric theory of bare nominals argues against the common
assumption in the generative liteature (e.g., Higginbotham 1985; Stowell 1989; Szabolcis
1987, 1994; Longobardi 1994; Heim and Kratzer 1998) that it is only DPs that can serve as
arguments; see also Baker (2003: 113) for the claim that D is not necessary for
argumenthood. 32 This assumption is understandable, given that NPs denote <e, t> whereas
DPs denote a type <e>; as a result, a NP must combine with a D to be saturated and
computed as type <e>. This assumption also makes sense if D can be considered to have the
function of mapping the set of enitities denoted by the N-set to a specific/definite entity
thereof. To the extent that my analysis is correct, however, the proposed syntax of nouns in
languages such as Javanese, Indonesian, Japanese, Russian, and English (for bare plurals and
mass nouns) provides strong evidence against this commonly held view.
32
Recall that Chierchia’s Nominal Mapping Parameter as well, DP is not necessary for argumenthood.
329
It is important to note in this context that there have been considerable debates concerning
whether nouns in Slavic/article-less languages such as Serbo-Croatian, Bosnian and Polish
are associated with DPs or NPs. See Progovac (1998), Bašić (2004), Leko (1999), Rappaport
(2001), Rutkowski (2002), Rutkowski and Maliszewska (to appear), Franks and Pereltsvaig
(2004), Trugman (2005), and Pereltsvaig (2007) for the DP analysis; see Corver (1992),
Willim (1998, 2002), Trenkic (2004), Bošković (2003, 2004, 2005), Zlatić (1997), and
Stepanović (1999) for the bare NP analysis. If the proposed analysis of Russian
morphosyntax in section 5 is correct, it argues for the parametrized NP hypothesis for Slavic,
hence necessiates a large-scale re-examination of the data in this language family that have
been adduced in favor of the DP hypothesis.
I conclude this chapter with the following important question, which the presnet
analysis would bring to light if it is on the right track. As stated above, the proposed theory
claims that the denotation of a functional catgeory is language-invariant. For languages
like Italian and English (in certain cases) whose nominals always project to DP, the
syntax-semantics mapping is strictly compositional, for the reason stated above. Then, a
question arises in those languages which I have argued to lack a D as to how a determinerless nominal can denote an inidvidual in the same way that a determiner-headed nominal
does. To put in a more informal way, how can humans express information such as
330
definiteness and specificity in languages without D heads that other languages with D
heads will express in fully syntactic ways given that those pieces of information are
syntatically represented? 33 I believe that the present thesis of minimalist interfaces
provides an interesting answer to this question. It is possible that some sort of repair takes
place at LF in just as much as at PF (chapter 2). Under the present case, the linguistic
semantic component may well develop general type-shifting operations to solve a type
mismatch problem that necessarily arises in D-less languages. In fact, this is the tack that
Chierchia (1998a) develops to account for the morphosyntactic behavior of Russian. More
generally, given that all languages have some or other ways to express a particular
thought/message, D-less languages will activate certain domain-specific operations that
would remain unsed in D-languages to express the same though, whether via
demonstratives that seem to have a similar function as D, general type-shifting operations,
or even extra-linguistic contexts that impose the definite/specificity on a D-less noun. Of
course, one could imagine that there should be certain restrictions on the kind of operations
that the semantic component perform to remedy the crude syntactic representation; see
Chierchia (1998a), for example, who claims that type shifting is a Last Resort operation at
33
Thanks to Andrew Carnie (personal communication) and Simin Karimi (personal communication) for
raising this question and useful discussion/suggestions.
331
the LF component. I leave detailed examination of the nature of such restrictions as an
important task for future research.34
The investgation conducted in this chapter, therefore, is construed as a case of the
thesis of minimalist interface; the semantic component is subservient to the needs of the
syntax, assigning whatever reasonable interpretation it can to the output of the syntax
with several domain-specific operations in hand (type-shifting, contexts, pragmatics) but
only within the parametrically defined carve set up by syntactic computation. I discuss
further ramifications of this conjecture in chapter 7.
34
It would be interesting to extend the parametric theory of nominal projections proposed here to clausal
projections as well in order to see whether any parallelism holds between the nature of functional nominal
structures and the nature of functional clause structures. Fukui (1986, 1995) argue that Japanese lacks
functional categories such as T and C. The observation that Indonesian, Javanese, and Chinese all lack
morphemes such as tense and agreement overtly manifested in Indo-European languages provides evidence
that these languages may also lack verbal functional categories such as T and C. I wish to conduct a fuller
examination of this parallelism in my future research. I am very grateful to Simin Karimi (personal
communication) for suggesting this extension.
332
CHAPTER 5 WHETHER WH-IN-SITU MOVES OR NOT IN INDONESIAN:
CHOICE FUNCTIONS AND INTERFACE ECONOMY 1
1. Introduction
In the last three chapters, I have explored the consequences of the thesis of the
minimalist interface for the way syntax interfaces with the external phonological and
semantic components. The results in chapters 2 and 3 based on the interaction of the active
voice deletion with syntactic movement in Indonesian and Javanese as well as the
amelioration of P-stranding violations by deletion show that the phonological component
is fundamentally interpretive, subservient to the universal law of syntax, doing whatever it
can within the parametrically defined range of options available to the linguistic interface.
The results in chapter 4 based on the denotation and morphosyntax of bare nominals in
Indonesian and Javanese as well as other languages show that the same characterization
holds for the other syntax-external semantic component, which, I have argued, comes up
with whatever semantic interpretation it can within the restricted set of morphosyntactic
1
This is a substantially expanded version of the paper presented at the Mid-America Linguistic Conference
held in the University of Kansas, Lawrence (October 2007). The paper presented there is to appear in Sato
and Yuliani (forthcoming).
333
structures that each language parametrically selects from a universal set of
morphosyntactic features and of possible number values.
The purpose of this chapter is to further substantiate the validity of the thesis of
minimalist interfaces from the perspective of the syntax-semantics interface. For this
purpose, I concentrate on the question of what the proper licensing mechanism of wh-insitu is in Indonesian. The current minimalist interface thesis expects that this
phenomenon is naturally explained as an optimal realization of the needs of the syntactic
computation on the part of the semantic component that is responsible for instructions
for the conceptual-intentional system. I argue that this is indeed what the facts on wh-insitu in Indonesian actually attest by showing that it has syntactic and semantic properties
that would remain mysterious under purely syntactic approaches to wh-questions and
that their interpretation is straightforwardly derived by the asymmetric interaction of
syntax with semantics via the notion of Interface Economy in the sense of Reinhart (1992,
1995, 1997, 1998, 2006), Fox (1998, 2000), and others.
The present chapter is organized as follows. In the next section, I provide a detailed
overview of the syntactic and semantic properties of wh-in-situ in Indonesian. The
discussion in this section owes a great deal to the seminal work on this construction by
Saddy (1991), who observes that wh-in-situ in this language shows a spectacular range of
334
interesting properties such as the lack of island/ECP effects, the absence of pair-list
readings and weak/strong crossover effects, and the quantificational uninformativiness.2
Drawing on Saddy’s description and analysis, I show that commonly accepted analyses of
wh-in-situ in other Asian languages such Chinese and Japanese as in overt/covert whmovement (Huang 1982; Nishigauchi 1986, 1990; Watanabe 1992, 2001; Richards 2001)
and unselective binding (Baker 1970; Pesetsky 1987) cannot adequately account for these
intriguing syntactic and interpretive characteristics uniquely associated with wh-in-situ in
Indonesian. This discussion leaves us with the question we started with: what is the proper
analysis of wh-in-situ in Indonesian. In section 3, I provide Saddy’s own analysis of this
construction as an interrogative definite description. Saddy argues, based on the
interpretive and syntactic parallelisms between wh-in-situ in Indonesian and the
formulation of “questions” on the game show “Jeopardy” in the United States, that the
former receives the interpretation it has due to the fact that it is interpreted in the same way
as a definite description is. I provide several empirical arguments against this treatment of
Indonesian wh-in-situ. The relevant arguments concern the source of interrogative force in
the two cases, the indefinite nature of wh-in-situ in Indonesian in terms of reduplication,
and, most importantly, the observation that this type of wh-phrase behaves more like
2
Although I present evidence against some of Saddy’s reported results and analysis.
335
existential indefinites rather than definite descriptions, which do not contain a variable
position. Based on this consideration, I show in section 4 that all the observed properties of
wh-in-situ in Indonesian are naturally derived once we assume that the appropriate
licensing mechanism of the wh-construal is via a choice function in the sense of Reinhart
(1992, 1995, 1997, 1998, 2006). I further demonstrate that this analysis makes several
correct empirical predictions, in particular, in the area of NP vs. non-NP asymmetries with
respect to putative ECP effects. I conclude this section by comparing the choice function
analysis with the fine-grained unselective binding approach to wh-in-situ in Indonesian
recently developed by Cole and Hermon (1998, 2000). Though the two analyses initially
appear to amount to the same predictions in the core cases, I argue that the choice function
makes correct predictions concerning the so-called “Donald Duck” problem and the
intermediate scope reading of wh-in-situ in multi-clausal contexts that would only be
accounted for by ad hoc stipulations under Cole and Hermon’s unselective binding
analysis. I conclude the present chapter in section 5 by discussing several theoretical
implications of wh-in-situ in Indonesian and the proposed analysis for the proper view of
the syntax-semantics interface and for the thesis of minimalist interfaces. I argue that the
choice function analysis developed in this chapter is one optimal satisfaction on the part of
the semantic component of the needs of the syntactic computation, further vindicating the
336
validity of the minimalist interface thesis, because a choice function interpretation can be
considered as a syntax-external manifestation in the interface component of a failure that
syntax itself cannot make up for.
2. Wh-in-Situ in Indonesian
This section provides an overview of the structural and interpretive properties of wh-insitu in Indonesian. The discussion in this section draws heavily on the description and
analysis of this construction presented by Saddy (1991). Accordingly, all the examples in
this section are drawn from his work, unless otherwise indicated, though I have checked the
same data with three native-speaker consultants for grammaticality judgments. I show that
the two most widely held analyses of this construction in other Asian languages such as
Japanese and Chinese − syntactic movement and unselective binding − are not the right
approach to wh-construal for in-situ wh-phrases in Indonesian.
As we saw in chapters 2 and 3, Indonesian has three ways to form wh-questions: i) overt
syntactic movement to the matrix, scopal [Spec, CP], ii) partial syntactic movement to the
embedded, non-scopal [Spec, CP]s, and iii) wh-in-situ. The three strategies are illustrated in
(1a-c), respectively.
337
(1) Wh-Questions in Indonesian
a. [CP1 Apai yang kamu pikir [CP2 Esti kira [CP3 Pak Yanto beli ti kemarin]]]?
what that you
think
Esti expect
Mr. Yanto
buy
yesterday?
‘What do you think Esti expects Mr. Yanto bought yesterday?’
b. [CP1 Kamu pikir [CP2 apai yang Esti kira [CP3 Pak Yanto beli ti kemarin]]]?
you
think
what that Esti expect
Mr. Yanto buy
yesterday
‘What do you think Esti expects Mr. Yanto bought yesterday?’
c. [CP1 Kamu
you
pikir
think
[CP2 Esti kira [CP3 Pak Yanto beli apa kemarin]]]?
Esti think
Mr. Yanto buy what yesterday
‘What do you think Esti expects Mr. Yanto bought yesterday?’
In the example in (1a), the wh-phrase apa ‘what’ undergoes overt syntactic movement to
the scopal, matrix [Spec, CP]. This option is always available for nominal wh-phrases such
as siapa ‘who’ and apa ‘what’ and prepositional wh-phrases such as cara apa ‘in what
way’, dimana ‘in what place, where’, and untuk apa ‘for what reason’, but obligatory for
non-nominal wh-phrases such as kenapa ‘why’, bagaimana ‘how’, and kapan ‘when’; see
section 4.3 for related discussion. The example in (1b) illustrates the partial syntactic
movement option in Indonesian, where the same wh-phrase undergoes movement into the
338
intermediate, non-scopal [Spec, CP], though the example itself has a matrix whinterpretation as the fully moved example in (1a). This option is available for nominal whphrases but not for non-nominal wh-phrases. Finally, the example in (ic) illustrates the insitu option in Indonesian. This option is always possible for nominal wh-phrases but never
possible for non-nominal wh-phrases, according to my consultants.
Saddy (1991) observes that the wh-in-situ construction in Indonesian exhibits a spectacular
range of syntactic and semantic characteristics that would not be accounted for under standard
analyses of the corresponding constructions in other languages such as English, Chinese, and
Japanese. I review his main arguments in the rest of this section to show that the two most
widely assumed analyses of wh-construal, syntactic movement and unselective binding, are not
applicable for wh-in-situ in Indonesian.
2.1. “Overt” Syntactic Movement?
The first analysis of wh-in-situ in Indonesian, which is most easily dismissed, is the null
operator/Q-feature movement in the syntactic component, as proposed by Watanabe
(1992, 2001) and Richards (2001) for wh-in-situ constructions in Japanese (cf. Kayne
1998). Watanabe proposes that Japanese has an overt syntactic movement of a
phonologically invisible wh-operator into the specifier of the scopal C, as in English.
339
Richards proposes, following the Single Cycle Model (Bobaljik 1995; Groat and O’Neil
1996; Pesetsky 1998), that the difference between “overt” and “covert” movement is
purely phonological, determined by whether the head or the tail of a movement chain is
pronounced; if the head is pronounced, we have “overt” movement whereas, if the head
is pronounced, we have “covert” movement. This model, thus, eliminates the traditional
distinction between overt and covert movement drawn in the classical minimalist period
(Chomsky 1993, 1995). Working within this model, Richards argues that Japanese whin-situ occurs when the tail of the movement of the in-situ wh-phrase into the scopal
[Spec, CP] is pronounced at the PF component.
The following consideration shows that the analysis of wh-in-situ presented in
Watanabe (1992, 2001) and Richards (2001) for Japanese is not transportable to the
corresponding construction in Indonesian. Syntactic island effects (Ross 1967) have been
widely acknowledged as a textbook case for the occurrence of syntactic movement in the
generative framework. Expectedly, overt syntactic movement in Indonesian shows various
island effects, as in (2a-d).
340
(2) a.* Apai yang
what that
kamu katakan
[dimana kita beli ti]?
you
where
mention
we
(wh-Island)
buy
‘What do you mention where we bought?’
b.* Siapai yang kamu suka [cerita yang mengkritik ti itu]? (Complex NP Island)
who
that you like
stories that criticize
the
‘Who do you like the stories that criticized?’
c.* Siapai
who
yang
kamu
kira gambar ti dijual?
Foc
you
think pictures
(Subject Island)
be-sold
‘Who do you think pictures of were sold?
d.* Dengan siapai yang kamu cemburui Bill [karena saya berbicara ti]?(Adjunct Island)
with
who Foc you get jealous of Bill because I
spoke
‘With who did you get jealous of Bill because I spoke?’
(Saddy 1991: 190, 191)
If Watanabe’s/Richards’ analysis is correct for wh-in-situ in Indonesian, in the wh-in-situ
counterparts to these examples, the relevant wh-phrase would undergo overt movement of
the null operator into the matrix, scopal [Spec, CP] (for Watanabe) or the tail of the
movement is pronounced at PF (for Richards). Thus, this analysis would predict that the
341
in-situ counterparts to (2a-d) should be as ungrammatical as (2a-d). This prediction is false,
however. Saddy shows that wh-in-situ in Indonesian is a well-formed question in this
environment, freely taking the matrix interrogative interpretation outside of syntactic
islands. This point is illustrated by the grammaticality of the examples in (3a-d). 3
(3) a.
Kamu
katakana
[kita mem-beli
you
mention
we AV-buy
apa dimana]?
(wh-Island)
what where
‘What did you mention where we bought?’
b.
Kamu suka
[cerita
you
stories that
like
yang
mengkritik
siapa itu]? (Complex NP Island)
criticize
who
the
‘Who do you like the stories that criticized?’
c.
Kamu
men-gira
gambar siapa dijual?
You
AV-think
pictures who
(Subject Island)
be sold
‘Who do you think pictures of were sold?’
3
(3a) is modified from Saddy (p. 190) by changing the verb from ingat ‘remember’ to katakan ‘mention.’
342
d. Kamu men-cemburui Bill [karena saya berbicara dengan siapa]? (Adjunct Island)
you
AV-jealous of Bill because I
spoke
with
who
‘With who did you get jealous of Bill because I spoke?’
(Saddy 1991: 190, 191)
The fact that these examples are all well-formed wh-questions, thus, provides strong
evidence that Watanabe’s/Richards’analysis is not the correct licensing mechanism for whin-situ in Indonesian.4
In the next section, I consider whether the alternative LF movement analysis works for
Indonesian and provide several arguments from Saddy that this analysis is also not tenable.
I point out that those arguments also provide further evidence against extending the
movement-based analysis proposed by Wanatabe and Richards to wh-in-situ in Indonesian.
4
Of course, it is a separate issue whether Watanabe’s (1991, 2001) analysis still holds for Japanese, if not for
Indonesian. The main argument for Watanabe’s analysis comes from the wh-island effect, illustrated in (i):
(i) ?? John-wa
[Mary-ga
nani-o
katta
kadooka] Tom-ni
John-Top Mary-Nom what-Acc bought whether
‘What did John ask Tom whether Mary bought?’
tazuneta
Tom-Dat asked
no?
Q
(Watanabe 2001: 208)
Provided that LF movement is immune to subjacencty (Baker 1970), the fact that (i) cannot be construed as the matrix
wh-question indicates that movement of the embedded wh-phrase has taken place in overt syntax. Importantly, however,
Toyoshima (2004) claims that the apparent wh-island effect is purely semantic in nature, hence does not necessarily
support Watanabe’s analysis. Specifically, since the embedded question denotes a set of existentially closed propositions,
the wh-interpretation of the cardinal predicate cannot be associated with the matrix set of propositions. See also note 11.
343
2.2. Covert Syntactic Movement?
Saddy provides four arguments that the covert syntactic movement analysis of wh-in-situ in
languages as in Chinese and Japanese as proposed in Huang (1982) and Lasnik and Saito
(1984, 1992) is also incorrect for wh-in-situ in Indonesian. The first argument comes from his
observation, well-known among Indonesianists (see Cole and Hermon 1998, 2000 and Fortin
2007b, to mention a few), that the covert/LF movement in this language obeys the same set of
island constraints as does the overt/syntactic movement, a position that has been also supported
from work on other languages such as Japanese, Korean, and English as in Nishigauchi (1986,
1990), Choe (1987), Pesetsky (1987), and Reinhart (1991). This observation is based on the
ungrammaticality of examples as in (4a-c) that involve partial wh-movement, that wh-phrases
that remain within syntactic islands in overt syntax still give rise to ungrammaticality.
(4) Partial Wh-Movement in Indonesian
a. * Kamu kira (bahwa) [cerita bahwa siapai yang ti mengkritik Jon itu] di-jual?
you
think that
story that
who that
criticized Jon the PV-sold
‘Who do you think that the story that t criticized Jon was sold?’
344
b. * Kamu kira (bahwa) [cerita bahwa siapai yang Jon mengkritik ti itu]di-jual.
you
think that
story that
who
that Jon criticized
the PV-sold
‘Who do you think that the story that John criticized t was sold?’
c. * Kamu men-cemburui
you
Bill [karena [PP dengan siapa]i yang saya berbicara ti]?
AV-get jealous of Bill because
with
who
Foc I
spoke
‘With whom did you get jealous of Bill because I spoke (to) t?’
(Saddy 1991: 195, 196)
In (4a), the wh-phrase siapa ‘who’ undergoes partial wh-movement into the intermediate,
non-scopal specifier of CP. Since this short extraction itself does not cross any syntactic
island, it cannot be the source of the ungrammaticality. Saddy argues that the
ungrammaticality here directly follows if we assume that the LF/covert movement of the
partially moved wh-phrase into the matrix specifier of CP obeys island constraints in
Indonesian. According to this analysis, the wh-phrase siapa ‘who’ undergoes covert
movement into the scopal specifier of CP for the purposes of scope taking. This
movement, thus, renders the example in (4a) ungrammatical due to the Complex NP
island. A similar story holds for (4b). In the example in (4c), the extraction of the
prepositional wh-phrase dangen siapa ‘with who’ remains within the adjunct island but
345
the example is still ungrammatical. This is naturally predicted if the phrase undergoes
further wh-movement into the matrix [Spec, CP], crossing the island at LF. Thus,
examples as in (4a-c) provide us with independent reason to believe that covert
movement (as well as overt syntactic movement, as we have seen in the previous
subsection) obeys island constraints in Indonesian.
Now, if the covert movement analysis of wh-in-situ in languages such as Chinese,
Japanese, and English (in multiple interrogative questions) as proposed by Huang and
Lasnik and Saito is correct for Indonesian, we predict that the in-situ counterparts of the
partially-moved examples in (4a-c) should also be ungrammatical because the LF
representation of the in-situ variants would be identical to that of the examples in (4a-c) in
that it involves violation of one of the island constraints. Crucially, however, Saddy shows
that this prediction is falsified by the grammaticality of (5a-d), in-situ counterparts to (4a-d).
(5) a. Kamu kira (bahwa) [cerita
you
think that
story
bahwa
siapa mengkritik
Jon
itu]
di-jual?
that
who
Jon
the
PV-sold
criticized
‘Who do you think that the story that t criticized Jon was sold?’
346
b.
Kamu kira (bahwa) [cerita bahwa Jon
mengkritik
siapa itu]
di-jual?
you
criticized
who
PV-sold
think that
story that
Jon
the
‘Who do you think that the story that John criticized t was sold?’
c.
Kamu
men-cemburui
Bill [karena saya berbicara
you
AV-get jealous of Bill
because I
spoke
dengan siapa]?
with
who
‘Who did you get jealous of Bill because I spoke with t?’
(Saddy 1991: 195, 196)
All these examples are impeccable as matrix wh-questions, contrary to what the LF
movement analysis would predict. Thus, the grammaticality of (5a-d) provides strong
evidence against the application of the LF movement hypothesis to wh-in-situ constructions in
Indonesian. Note that the same examples are also problematic for the Watanabe/Richardsstyle overt movement analysis reviewed in the previous subsection because they would also
incorrectly render (5a-d) ungrammatical. Thus, the contrast between (4a-c) and (5a-c) argues
against a covert movement analysis of wh-in-situ in Indonesian, whether formulated as LF
movement or as by Watanabe and Richards.
The second argument against the LF movement approach to wh-in-situ in Indonesian is
based on the fact that this language does not allow complements that contain a wh-in-situ
347
for verbs such as ingin tahu ‘want to know, wonder’ that are obligatorily subcategorized
for a +WH feature, as in English (e.g., I wonder what you bought. vs. *I wonder you
bought what.). This is illustrated by the contrast between (6a) and (6b).
(6) a. * Saya
I
ingin tahu
want
Jon
know Jon
men-cintai
siapa.
AV-love
who
‘I want to know who Jon loves.’
b.
Saya
ingin
tahu
siapa yang
I
want
know who
Foc
‘I want to know who Jon loves.’
Jon
Ø-cintai.
Jon
love
(Saddy 1991: 207)
It is a fact of English grammar that the +WH complement property of interrogative verbs
such as wonder in English can only be satisfied by the movement of a wh-word into the
specifier of the CP complement of such verbs to check its inherent Q-feature against that
of C. If the [+WH] subcategorization of the verb ingin tahu ‘wonder’ [+WH] must be
satisfied by the [+WH] feature within its complement CP, then the fact that in-situ whelements do not satisfy the [+WH] requirement of this verb in (6a) suggests that they do
not substitute into the specifier of CP, in contrast to overtly moved wh-phrases in (6b).
348
This contrast would remain mysterious under the LF covert movement analysis because
the interrogative CP requirement would be satisfied by the covert movement of the in-situ
phrase siapa ‘who’ into the specifier of the embedded CP. Note that this contrast is also
problematic for Watanabe’s (1992, 2001)/Richards’ (2001) approach because the LF
configurations for the moved and in-situ structures in (6a) and (6b) would be identical at
LF. By contrast, the difference in grammaticality here naturally follows if we assume that
the in-situ wh-phrase in (6a) literally remains in situ, meaning that no features of the phrase
undergo any movement either in overt syntax or at LF; the subcategorization requirement
is simply violated in (6a) because nothing moves to satisfy it, in contrast to (6b).5
5
Notice that the unselective binding approach predicts that (6a) should be grammatical, contrary to facts
because this in-situ mechanism would allow the wh-phrase siapa ‘who’ to be interpreted in situ, rendering the
embedded clause a well-formed question. Therefore, the contrast between (6a) and (6b) shows that
unselective binding is inadequate to license an embedded question interpretation in Indonesian.
Carrying this conclusion over to Chinese, the embedded clause in (i) must involve covert/LF movement
as proposed in Huang (1982), rather than unselective binding, because the latter operation does not work for
creating embedded questions in Indonesian.
(i) Zhangsan
Zhangsan
xiang-zhidao
Lisi
mai-le
shenme.
wonder
Lisi
bought what
‘Zhangsan wonders what Lisi bought.’
I thank Heidi Harley (personal communication) for pointing this out.
(Watanabe 2001: 204)
349
Thirdly, Saddy claims that wh-in-situ constructions in Indonesian do not show strong or
weak crossover effects, as illustrated in (7a, b).6
(7) a. (*) Diai
he
meng-harap
Jon
AV-expect
Jon
men-cintai
AV-love
siapai?
who
‘Whoi does hei expect Jon to love?’
diai
meng-ira
saya
professor his
AV-think
I
b. (*) Prof
men-cintai
siapai?
AV-love
who
‘Whoi does hisi professor thinks I love?’
(Saddy 1991: 207, 208)
As we have seen briefly in chapter 2, the standard assumption on the crossover effect (Postal
1971; Wasow 1979) is that it arises when a variable-bound pronoun fails to be c-commanded
both by a binder and by its variable at the surface/derived structure. Under this assumption,
this effect can be formalized as the filter of the form * whi …. pronouni… ti (see discussion
below, though, for a clarificatory remark). The strong and weak crossover effects arise in
examples in (8a) and (8b), respectively, because the pronoun coindexed with the binder is
6
The examples are (7a, b) are reported to be ungrammatical in Cheng (1991), Cole and Hermon (1998),
Rogayah (1995). Thus, I have put the star in parenthesis here. See discussion below for more detailed
discussion on these examples.
350
not c-commanded by the variable, namely, t. The ungrammatical examples in (9a, b) show
that the relevant effect is also caused by quantifier raising (May 1985), a case of LF
movement.
(8) a. * Whoi does hei love ti?
(strong crossover effect in overt syntax)
b. *? Whoi does hisi mother love ti ? (weak crossover effect in overt syntax)
(9) a. * Hei loves everyonei.
(strong crossover effect at LF)
b. * Hisi mother loves everyonei. (weak crossover effect at LF)
Under the standard assumption, the (putative) lack of the weak/strong crossover effects in (7a, b)
can be construed as evidence that the wh-phrase siapa ‘who’ remains in its thematic position
both in overt syntax and at LF. If the syntactic movement in the sense of Watanabe (1992, 2001)
and Richards (2001) occurred into the specifier of CP that c-commands the pronoun coindexed
with the wh-operator, then the resulting configuration would cause the strong/weak crossover
effect, as shown in (8a, b), contrary to facts. If the covert movement in the sense of Huang (1982)
and Lasnik and Saito (1984, 1992) were correct, then the LF movement would cause the same
violation as quantifier raising would, as shown in (9a, b). Thus, the (putative) absence of the
351
crossover effects in examples as in (7a, b) reported by Saddy cast further doubts on the validity
of the syntactic movement as the mechanism of in-situ wh-construal in Indonesian.
Importantly, the preceding argument against the LF movement analysis based on the crossover
effect crucially depends on the grammaticality of the examples in (7a, b) as reported by Saddy. It
is quite debatable, however, whether this observation holds for Indonesian and Malay. For
example, Cole and Hermon (1998) provide examples as in (10) to show that a crossover effect is
observed in wh-in-situ in the dialect of Malay they document, contrary to what Saddy reports for
Indonesian; see Cheng (1991) and Rogayah (1995) for the same observation.
(10) * Prof
diai
fikir
saya
Prof
his
think I
meny-intai
siapai?
AV-love
who
‘Whoi does his professor think I love ti?’
(Malay: Cole and Hermon 1998: 234)
My language consultants also concur with Cole and Hermon, reporting that the examples in
(7a, b) are unacceptable when the pronominal dia is construed as a variable whose value covaries with that of the wh-operator. It is not clear at this moment what causes this huge
variation in the acceptability of the examples in (7a, b). However, provided the judgment cited
by Saddy represents the minority one in the literature in Malay/Indonesian linguistics and in
352
the face of my own consultant work, I take it that the examples in (7a, b) are ungrammatical in
(certain varieties of) Indonesian. This, in turn, might mean that we lose an argument against
the covert (and null operator) movement approach to wh-in-situ in Indonesian.
Quite to the contrary, I maintain, following the suggestion made in Cole and Hermon (1998:
234), that the presence of the crossover effect does not automatically mean that the in-situ whphrase in (7a, b) undergoes syntactic movement because the crossover effect can be
formulated in non-movement terms as a constraint on representations. Specifically, Cole and
Hermon argue that the crossover effect can be analyzed as the byproduct of the Bijection
Principle of Koopman and Sportiche (1983) that prohibits a single operator from binding
more than one variable. This principle allows us to correctly block the examples in (7a, b)
without also assuming syntactic movement because the base-generated wh-operator in [Spec,
CP] binds both the pronoun and the variable (see detailed discussion of their unselective
binding approach to wh-in-situ in Malay in section 4.4).7 For this reason, I conclude, contrary
7
As Heidi Harley (personal communication) points out, this representational analysis predicts that crossover
effects should exist in wh-in-situ constructions in other in-situ languages such as Japanese. This prediction is
correct, as shown in (i) in Japanese.
(i) a. * Soitui-wa
that person-Top
John-ga
darei-o
aisiteiru-to
John-Nom who-Acc love-C
omotteiru
no?
think
Q
‘Whoi does hei think that John loves?’
b. * Soitui-no-sensei-wa
John-ga
darei-o
aisiteiru-to omotteiru
that person-Gen-teacher-Top John-Nom who-Acc love-C
‘Whoi does hisi professor think John loves?’
think
no?
Q
353
to Saddy, that the presence of the crossover effect is mute on the question of whether wh-insitu in Indonesian undergoes syntactic movement or not.
Let us now come back to the last argument made by Saddy against the LF movement
analysis of wh-in-situ in Indonesian. This argument is based on his observation that wh-insitu constructions do not support a pair-list reading. Consider examples as in (11).
(11)
Siapa
mem-beli
apa ?
Who
AV-bought
what
‘Who bought what?’
(Saddy 1991: 208)
Saddy reports that this multiple wh-question can only be interpreted as a request for a single
pair as in John bought a book; thus, answers such as John bought a book, Mary bought a
magazine, Bob bought a shirt are not a possible reply to this question. Let us suppose for the
sake of argument that this observation holds for Indonesian multiple wh-questions. Since
Higginbotham and May’s (1981) work on English multiple interrogatives, the availability of
the pair-list reading for English sentences such as who bought what has been taken to be
driven by the association of the two wh-phrases in the same Comp at LF (or the multiple
specifiers of the same C, in the more modern terminology) via the semantic process known
354
as absorption. Higginbotham and May’s central assumptions are summarized as in (12a-c),
adopted from Barss’ (2000) exposition (his (7-9), respectively) with a slight modification.
(12) Higginbotham and May’s 1981 Absorption
a. Absorpotion has a syntactic precondition of structural adjacency, which is defined as
mutual m-command of the two operators.
b. Absorption creates a complex n-ary quantifier, within which the original restrictions on
the unary quantifiers are conjoined as a complex restriction and to which all the
variables bound by the input quantifier are bound.
c. Absorption is a sufficient, and necessary, precondition to the multiple-pair
interpretation of multiple questions with singular wh-phrases.
(adopted from Barss 2000: 33, 34)
To the extent that this absorption analysis is correct, the lack of the pair-list reading in (11)
indicates that apa ‘what’ does not undergo movement either in overt syntax or LF.
Accordingly, this example provides evidence against the overt or covert movement approach
to the wh-construal in Indonesian.
355
Again, however, my language consultants have all reported that the pair-list reading is
readily available in sentences such as (11). This is also the judgment elicited from speakers
of Malay by Cole and Hermon (1998: 225), who report that their Malay informants had no
problem in giving a list interpretation for sentences as in (13).
(13)
Siapa
kamu
fikir
beli
apa?
who
you
think
buy
what
‘Who did you think bought what?’
(Cole and Hermon 1998: 225)
This judgment indicates that the argument against the LF movement based on the pair-list
reading is not strong as Saddy wanted it to be. We come back to this point again in section 4.
To sum up, I have reviewed four arguments presented in Saddy (1991) that the covert
movement analysis as proposed in Huang (1982) and Lasnik and Saito (1984, 1992) for
Chinese and Japanese is not an adequate mechanism of licensing wh-in-situ in Indonesian.
Though the arguments based on the crossover effect and the availability of the pair-list reading
in multiple questions do not necessarily argue for or against the LF movement analysis due
differences in judgment between Saddy’s and my work, the two other arguments based on the
356
lack of island effects and the [+WH] subcategorization requirement provide relatively solid
evidence that this analysis is not transportable to wh-in-situ in Indonesian.
2.3. Unselective Binding?
The third potential analysis of wh-in-situ, which is perhaps the most widely held analysis
for wh-in-situ in languages such as Japanese and Chinese is that of unselective binding
(Baker 1970; Pesetsky 1987); I defer discussion of the variant of this approach presented
recently by Cole and Hermon 1998 until section 4.4). Drawing on the analysis of multiple
wh-questions and their interpretations in English by Baker (1970), Pesetsky (1987)
proposes that wh-interpretation may be achieved not only by syntactic movement but also
a non-movement mechanism called unselective binding. Pesetsky claims that the choice
between these two options is determined by the notion of D(iscourse)-Linking, which
roughly corresponds to the morphological distinction in English of wh-words between
“which-X” (which man, which book, etc) and everything else (who, what, etc). (recall our
discussion of Cinque’s (1990) notion of referentiality in chapter 2). As Pesetsky (p. 107108) remarks, which-phrases are discoursed-linked (D-linked), because “when a speaker
asks a question like which book did you read?, the range of felicitous answers is limited by
a set of books both speaker and hearer have in mind” whereas “no such requirement is
357
imposed on wh-phrases like who, what, or how many books.” Pesetsky argues that if a whphrase is D-linked, it contains a variable that is unselectively bound by a Q-morpheme
located in the scopal C head position, and thereby is licensed without syntactic movement.
On the other hand, if a wh-phrase is not D-linked, it must undergo syntactic movement, be
it overt or covert, to be properly licensed by the scopal C. Pesetsky provides various types
of evidence based on the presence/absence of superiority effects in English questions as
well as the behavior of what he calls aggressively non-D-linked wh-phrases such as what
the hell in English and Japanese to support this hybrid approach to wh-construal.
Saddy, however, points out a couple of potential empirical problems with Pesetsky’s
version of unselective binding analysis when applied to wh-in-situ in Indonesian. The first
problem concerns the morphological composition of wh-phrases in Indonesian. As we have
seen above, Pesetsky’s analysis rests upon the correlation between the morphological
composition of a wh-phrase and its interpretive mechanism: if a wh-phrase is D-linked, it is
interpreted by unselective binding without movement whereas, if it is non-D-linked, it must
be interpreted by syntactic movement. This correlation, however, does not hold in Indonesian
because almost all wh-phrases in this language have “D-linked” expressions corresponding to
English “which-X” form. For example, Saddy observes that orang siapa ‘which person”,
which would be analyzed as a D-linked phrase in Pesetsky’s terms, is used interchangeably
358
with the non-D-linked form siapa ‘who’ but this difference in morphological composition
does not change the interpretive and structural constraints observed so far in this section.
Though this observation may not be a serious problem for Pesetsky’s theory, it at least
indicates that Pesetsky-style D-linking is not directly applicable to Indonesian wh-questions.
The second potential problem with the extension of Pesetsky’s analysis to Indonesian insitu questions is based on what Saddy reports as the quantificational uninformativeness of
wh-in-situ in Indonesian. Pesetsky employs D-linking to account for the triplet interpretation
available for sentences as in (14), so that the D-linked phrase which prize may get matrix
scope without movement by being bound by the matrix Q in the manner seen in (15).
(14) Who did every athlete expect to win which prize?
triplet answer: Gretsky expected Milli Vanilli to win an Oscar, Gefrion expected
George Burns to win Grammy, etc.
(Saddy 1991: 204)
(15) [S′ [Comp Qi, j whoi [S ei every athlete expect…. win which prizej]]
Importantly, this analysis crucially assumes that D-linked in-situ phrases such as which
prize must be able to interact in scope with other scope-bearing elements such as every
359
athlete; for otherwise, the triplet interpretation would be unavailable in examples such as
(14). As Saddy (p. 205) put it, “it is a necessary property of Pesetsky’s Q-bound D-linked
WH expressions that they interact quantificationally with other elements in the martin
clause” When applied to Indonesian wh-in-situ constructions akin to (14) above,
Pesetsky’s analysis leads us to predicts that this type of construction also should allow the
triplet interpretation. Saddy observes that this prediction is false in Indonesian because this
reading is precisely the kind of interpretation that BI wh-in-situ resists, as in (16a).
(16) a.
Setiap
orang
men-cintai
siapa?
every
person
AV-love
who
‘Who did every person love?’ who>every, *every>who
b.
Siapai
yang
setiap
orang
Ø-cintai ti?
who
Foc
every
person
love
‘Who did every person love?’ who>every, every>who
(Saddy 1991: 199)
According to Saddy, the example in (16a) with the wh-phrase in situ only allows the wide
scope reading of the in-situ phrase with respect to the universal quantifier setipa orang
‘every person’ in subject position; the reading where the value of the person loved co-varies
360
with that of the lover is impossible. This latter reading becomes available only when the whphrase must undergo overt syntactic movement, as illustrated in (16b).
Saddy notes that the same contrast can be seen in multi-clausal environments, as the
possible scope interpretation in the triplet of examples in (17a-c) illustrates.
(17) a.
Setiap
orang tahu
Tom
every
person knows Tom
mem-beli
apa?
AV-buy
what
‘What does every person know Tom bought?’
b.
Setiap
orang tahu apa yang
Tom
beli ti?
every
person know what Foc
Tom
buy
‘What does every person know Tom bought?’
c.
what>every, *every>what
what>every, every>what
Apa yang
setiap orang tahu Tom
beli ti ?
what Foc
every person know Tom
buy
‘What does every person know Tom bought?’
what>every, every>what
(judgments as reported by Saddy 1991: 200)
In (17a), the in-situ wh-phrase apa ‘what’ necessarily takes wide scope over the universal
quantifier setiap orang ‘every person’, even though the relative structural height of the
361
latter with the latter leads us to expect the opposite reading. Again, the wide scope reading
of the universal quantifier over the wh-phrase is only possible when the latter undergoes
syntactic movement, either partial movement, as shown in (17b), or full movement, as
shown in (17c). The pattern of (un-)available scope readings seen in these examples,
therefore, shows that wh-in-situ in Indonesian is quantificationally uninformative with
respect to other scope-bearing expressions, unlike moved wh-phrases. Accordingly, this
result would remain mysterious under Pesetsky’s Q-binding analysis of triplet questions.
Note that this contrast in scope interaction between moved wh-phrases and wh-in-situ also
shows that the scope of the latter cannot be determined by overt or covert syntactic movement;
if this were the case, there should be no scopal difference between these two expressions.
Based on this conclusion, Saddy concludes that Pesetsky’s analysis is incorrect.
An important remark I need to add here, however, is that I could not reproduce the same
judgments as elicited by Saddy from his Indonesian consultants. According to my language
consultants, both (16a) and (17a) allow the narrow scope reading of the in-situ wh-phrase with
respect to the universal quantifier, the reading where the value of the thing bought and the
person loved can vary with the value of the universal quantifier. To check this judgment, I
asked my language consultants to provide scope judgments for the sentence in (15a) in the
following manner: “Suppose that there are only five persons in the world in 2100 due to
362
critical food shortage: John, Bob, Susan, Mary, and Amy. Then, observing this world, I said
the sentence in (15a). In this case, are the following scenarios compatible with the meaning of
the sentence in (15a)?” Then, I gave two scenarios given below that diagnose narrow scope
and wide scope reading of the in-situ wh-phrase apa ‘what’.
(18) Scenario A (narrow scope reading of in-situ wh-phrase)
: John loves Susan, Susan loves Bob, Bob loves Amy, Amy loves Bob, and Mary loves John.
Scenario B (wide scope reading of in-situ wh-phrase)
: John, Bob, Susan, Mary all love Amy (and, Amy loves herself).
All the three consultants reported to me that the sentence in (15a) can be used to describe
both situations in (18a, b). A similar judgment task was conducted for the in-situ question
in (16a), for which the same consultants reported the same ambiguity. This result, therefore,
shows that the wh-in-situ in Indonesian is scopally informative, quite contrary to what
Saddy reports. At this moment, I have no idea how scope judgments can vary in such a clear
manner, as I do not know the linguistic backgrounds of Saddy’s language consultants. For the
purposes of this chapter, I assume that my consultants represent the majority judgment,
363
keeping in mind, though, that Saddy’s reported judgments might also hold for certain dialects
of Indonesian. We return to this in section 4.
Now if this is so, we have lost one argument against Pesetsky’s version of the unselective
binding approach to wh-in-situ in Indonesian. However, in section 4.1, I review Reinhart’s
(1992, 1995, 1997, 1998, 2006) arguments based on the scope behavior of wh-in-situ in
multiple questions in English and the “Donald Duck” Problem that cast serious doubts on the
general applicability of unselective binding as a possible non-movement strategy to license insitu expressions. Anticipating this forthcoming discussion, I assert here that Pesetsky’s Dlinking-based unselective binding analysis is not suitable for wh-in-situ in Indonesian.
2.4. Section Summary
Let me summarize what we have done in this section. I have reviewed here Saddy’s (1991)
arguments that none of the existing movement and non-movement-based accounts of whquestions motivated in other in-situ languages can adequately account for a series of
somewhat peculiar syntactic and semantic properties associated with wh-in-situ in
Indonesian. The lack of island effects in Indonesian wh-in-situ in contrast to overtly moved
wh-phrases and facts concerning verbal subcategorization show that neither “overt”
movement analyses as in Watanabe (1992, 2001) and Richards (2001) nor covert movement
364
analyses as in Huang (1982) and Lasnik and Saito (1984, 1992) are viable ways of
interpreting this type of interrogative phrase in Indonesian. The putative lack of strong/weak
crossover effects, and the potential absence of the pair-list reading suggest that wh-in-situ
and its associated features in Indonesian literally remain in their base-generated thematic
position. The language-particular morphological composition of wh-phrases and the (alleged)
scopally uninformative behavior of this class of expression in Indonesian indicates that
Pesetsky’s (1987) version of the non-movement approach to in-situ questions in terms of
unselective binding is also problematic. I have also made it clear that some of the judgments
reported by Saddy himself are contradictory to the judgments elicited in the present study
and those reported in the literature on Indonesian/Malay wh-questions as in Cheng (1991),
Rogayah (1995), and Cole and Hermon (1998).
Now that three widely held analyses of wh-construal have been eliminated as analytical
options for wh-in-situ in Indonesian, we are left once again with the original question with
which I started the present chapter: what is the correct licensing mechanism for wh-in-situ
in Indonesian that captures all the apparently mysterious syntactic and semantic properties
such as those discovered by Saddy (1991). The rest of this chapter is devoted to showing
that the answer to this question can only be found in the syntax-external interpretive
component and to arguing that the particular semantic analysis proposed below is an
365
optimal realization of the needs of syntax on the part of the component in the way that our
minimalist interface thesis leads us to expect. Before presenting such an analysis,
however, let us consider first in the next section what Saddy (1991) himself said about
the now mysterious wh-in-situ construction in Indonesian.
3. Wh-in-Situ in Indonesian is Not an Interrogative Definite Description
In this section, I review Saddy’s novel analysis of wh-in-situ in Indonesian as an
interrogative definite description that draws on the Jeopardy game show question in English.
Although I provide several arguments that such a treatment of the relevant class of whexpressions is inadequate, I also stress that his particular analysis contains a number of
important insights that the analysis to be developed in section 4 attempts to capture.
3.1. Saddy’s (1991) Analysis of Wh-in-Situ as Interrogative Definite Description
Saddy proposes that wh-in-situ in Indonesian behaves as interrogative definite descriptions,
drawing on an impressive range of syntactic and semantic parallelisms that hold between this
class of expressions and words of the form “this-X/these-Xs” in game Jeopardy show
questions in English. An example of English game show quizzes is given in (19).
366
(19) Question: For $100, every armchair general watched this television station.
Answer: What is NBC?
(slightly modified from Saddy 1991: 208)
The “question” in (19) does not have interrogative force in the standard sense as a wh-question
because it is syntactically a declarative statement; rather, it gains such force from the very
context that this sentence is uttered in a game show; a host utters this sentence to competitors,
expecting them to make a question such that it constitutes an appropriate answer to the definite
DP this television station. In other words, the interrogative requirement here is that competitors
come up with the member (s) of the definite description of the form ‘this-X/these-Xs.’ The
reason Saddy brought this type of game show scenario into his work is because of his
observation that statements as in (19) under the game show context exhibit exactly the same
range of structural and interpretive properties that we have seen to characterize wh-in-situ in
Indonesian. Those properties are illustrated in (20-23), with some examples.
367
(20) “This-X” Form Cause No Island Effects.
a. For $100, the viewing public wondered whether Reagan believed the claim that this
man was a hero.
b. For $100, the viewing public accepted the circumstances tat this man cited as extenuating.
c. For $100, many people refused to take showers after seeing this movie.
(Saddy 1991: 210)
(21) “This-X” Form Cannot Satisfy Subcategorization Requirements.
* Bill wonders (for $100), every armchair general watched this television station.
(Saddy 1991: 210)
(22) “This-X” Form Cause No Weak or Strong Crossover Violations.
a. For $100, hisi mother likes this mani.
b. Oliver North has been hailed as a great American patriot. ? For $100, hei most
admires this mani.
(Saddy 1991: 211)
368
(23) “This-X” Form Blocks Pair-List Reading in Multiple Wh-Questions
Question: For $100, this man married this woman
Response:* John Smith married Sally Jones, Tod Wilks married Martha Sachs…
(Saddy 1991: 211)
Based on these facts, Saddy concludes that the same range of properties of Indonesian in-situ whphrases naturally fall out if they are interrogative definite descriptions. This analysis of wh-in-situ
in Indonesian also makes the correct prediction that it should be scopally uninformative because
it is widely known that bona fide definite descriptions are not, as shown in (24), where the
definite DP the woman must take wide scope over the quantified subject every person.
(24) The Quantificational Uninformativeness of Definite Descriptions in English
Every person likes the woman with red hair.
*every> the woman with red hair, the woman with red hair>every
(Saddy 1991: 212)
Saddy argues that the definite description analysis of wh-in-situ in Indonesian, informed by
the behavior of “this-X” form in game shows, provides a unified account of all the properties
we have seen to hold for this class of wh-questions. As we have just seen, the scopally
369
uninteractive behavior of in-situ wh-phrases is a natural consequence of the fact that it is a
definite description, namely, that they “pick out a specific individual or a set of individuals.”
(p. 212). Wh-in-situ does not satisfy the WH-complement requirement of verbs such as ingin
tahu ‘wonder’ because it does not move into the specifier of the complement CP but instead
is licensed in situ by a non-quantificational mechanism by virtue of its definite nature.
Similarly, the (putative) lack of pair-list readings and weak/strong crossover effects and the
insensitivity to syntactic islands for the purposes of scope taking are derived because wh-insitu is interpreted in situ as an interrogative definite description.
Saddy’s analysis is extremely ingenious in a number of important ways. First, it provides
one unified, non-stipulatory account of all the otherwise mysterious syntactic and semantic
characteristics associated with wh-in-situ in Indonesian from the single fact that this class of
wh-words is an interrogative definite description. Second, as hinted at the end of his paper (p.
213), his analysis leads to the conclusion that, once a wh-element undergoes syntactic
movement, that element becomes quantificationally, as evidenced by the sudden emergence
of new scope interaction between it and other scope-bearing elements such as universal
quantifiers. Though he does not have an answer to this question, he speculates (p. 212) that
“part of the mechanism of quantificational WH construal is derivative from the simple fact
of movement having taken place.” Finally, his analysis suggests that natural languages may
370
well develop a non-syntactic mechanism of licensing wh-in-situ without relying on syntactic
movement. The last two points become crucial in the rest of this chapter, in which I develop
a different implementation of Saddy’s insight in a more sophisticated manner that is driven
by the considerations of the thesis of minimalist interface.
Though Saddy’s analysis has many things that inform us about the proper analysis of whin-situ in Indonesian as well as other in-situ languages, there are, nonetheless, several
arguments that suggest that his most fundamental claim that wh-in-situ in Indonesian is a
definite description is incorrect. I review those arguments in the next subsection.
3.2. Problems with Saddy’s (1991) Analysis
There are three problems that cast doubts on the validity of Saddy’s treatment of wh-in-situ in
Indonesian as an interrogative definite description. The first potential problem analysis is that
the source of interrogative force is quite different between English game shows and wh-in-situ
questions in Indonesian.8 In game show questions, the moderator already knows the value(s) of
the definite description provided by “this-X/these-Xs”. Speakers of Indonesian, by contrast,
utter wh-in-situ questions precisely because they would like to know what is the correct value
for the wh-phrase. In normal discourse situations, they always utter these sentences to seek
new information. It has been generally assumed (see Chomsky 1977 and Zubizarreta 1998 for
8
Thanks to Heidi Harley (personal communication) for discussion on this point.
371
extensive discussion) that wh-questions request new information. Thus, in English sentences
such as Who bought the book yesterday, the wh-phrase who requests new information while
the rest of the clause [x bought the book yesterday] is old information. This means that,
semantically, the wh-phrase is not a definite description. This qualitative difference between
“this-X/these-Xs” form in game shows and the wh-phrase in Indonesian, thus, poses a potential
problem for Saddy’s attempt to unify the two phenomena.
The second argument against Saddy’s analysis is that there is evidence internal to Indonesian
that wh-in-situ in this language contains a variable. Cole and Hermon (1998) observe that
nominal wh-words in Malay can be used as a variable bound by non-wh-operators, as in (25a, b)
and (26a, b). This observation also holds for Indonesian, according to my consultants.
(25) a.
Dia
tidak
mem-beli
apa-apa
untuk saya.
he
not
AV-buy
what-Red for
me
‘He did not buy anything for me.’
b.
Dia
tidak
mem-beli
apa-pun
he
not
AV-buy
what-also for
‘He did not buy anything for me.’
untuk saya.
me
(Malay: Cole and Hermon 1998: 239)
372
(26) a.
Saya
tidak
kenal
siapa-siapa di
I
not
recognize who-Red
at
universitas
itu.
university
that
‘I didn’t recognize anyone at that university.’
b.
Saya
tidak
kenal
siapa-pun di
universitas
itu.
I
not
recognize
who-who
university
that
at
‘I didn’t recognize anyone at that university.’ (Malay: Cole and Hermon 1998: 239)
In (25a) and (26a), the wh-word is bound by the existential quantifier that is overly
represented by the reduplication of the question word itself (see Travis 1999, 2001, 2003 for
the observation that reduplication also has the parallel function of creating existential
quantification in Malagasy). Similarly, in the examples in (25b) and (26b), the wh-word is
bound by the existential quantifier realized in the form of -pun ‘also’. This use of the in-situ
wh-words in Indonesian as an element bound by quantifiers, therefore, shows that this class
of words contains a variable. This result is problematic for Saddy’s treatment of the same
words as interrogative definite descriptions because definite descriptions as rigid designators
do not contain a variable under the most commonly held assumption. At the same time, this
result indicates that wh-in-situ in Indonesian, like its equivalent in Chinese, behave more like
normal wh-phrases whose movement leaves a variable in the launching site.
373
The final argument against Saddy’s analysis is that it misses the important generalization
that wh-in-situ in Indonesian behaves as existential indefinites rather than definite descriptions
in terms of scope taking. It is widely acknowledged in the literature that certain
weak/existential indefinites such as singular NPs (e.g., someone, something) and cardinal
plurals (e.g., two men, many women) are insensitive to syntactic islands for scope-taking, as
shown by the contrast between (27a-c) and (28a-c).
(27) a.
Someone reported that Max and all the ladies disappeared.
some>all, * all>some
b.
Someone will be offended if we don’t invite most philosophers.
some>most, *most>some
c.
Many students believe anything that every teacher says.
many>every, * every>many
(Reinhart 1997: 338)
374
(28)a.
Everyone reported that Max and some lady disappeared.
every>some, some>every
b.
Most guests will be offended if we don’t invite some philosopher.
most>some, some>most
c.
All students believe anything that many teachers says.
all>many, many>all
(Reinhart 1997: 339)
(27a-c) show that the strong quantifiers such as all, most, and every cannot violate one or the
other island constraints to take wide scope over another scope-bearing element in the matrix
clause. This is not surprising if Quantifier Raising, an instance of LF movement, is constrained
by the island constraints, as is overt syntactic movement. What is surprising, then, is the fact
illustrated in (28a-c), that weak existential indefinites such as some and many take wide scope
over the quantifier in the matrix subject position, in apparent violation of the island constraints
that we have just seen to constraint the Quantifier Raising operation. This wide scope reading
of certain existential indefinites has been a source of endless controversies in the semantic
literature; see Reinhart (2006: ch2.) for a concise summary of these controversies. Whatever
the ultimate analysis might turn out to be, this island-insensitive scope-taking behavior is
similar to that of wh-in-situ in Indonesian. We have seen in section 2 that this class of phrases
375
can freely take wide(st) scope in a massive violation of the set of standard island constraints on
movement. Given this parallelism, the null hypothesis seems to be that wh-in-situ in
Indonesian should be treated as existential indefinites. This indeed has been a standard
assumption on wh-phrases since the earliest days of generative research as argued for in
Chomsky (1964); Kats and Postal (1964); Klima (1964), and Kuroda (1965); see also
Karttunen (1977), and Engdahl (1986).
Based on the afore-mentioned three considerations, I reject Saddy’s (1991) approach to
wh-in-situ in Indonesian and seek an alternative account that nonetheless captures the three
fundamental insights behind his work mentioned in the end of section 3.1. 9
4. Choice Function as an Optimal Interface Strategy at the Syntax-Semantics Interface
We have seen thus far that wh-in-situ in Indonesian remains in situ throughout the
syntactic derivation. Drawing on the evidence presented in Saddy (1991), I have shown
9
Of course, what is the proper analysis of the game show question with the properties as observed by Saddy is a
separate question that I leave aside in this chapter. I note here, however, that the choice function analysis of the kind
developed Kratzer (1998) and Matthewson (1999) (see section 4.2 for discussion of their analysis) has suitable
theoretical properties to accommodate this type of question. Kratzer argues that indefinites in English are divided into
specific and quantificational and that they must take widest scope when interpreted as specific in the form of choice
function. Matthewson provides evidence for Kratzer’s analysis from evidence in St’át’imcets. Since Saddy’s core
claim is that all the peculiar properties of the “this-X/these-Xs’ in the game show question are derivable from their
denotation as a definite description, it may well be that Kratzer/Matthewson-style analysis provides a unified account
of the observed properties. See Sato (in prep) for such an analysis.
376
that two analyses that have been most often adopted for wh-in-situ in other languages
such as English (for multiple wh-questions), Japanese, and Chinese in terms of syntactic
movement and unselective binding have several serious shortcomings in face of several
structural and semantic properties associated with this construction. I have also provided
evidence based on the reduplication and the island-insensitive behavior of wh-words for
the purposes of scope-taking that Saddy’s analysis of this class of expressions as an
interrogative definite description cannot be upheld. Thus, the fundamental question that
we still need to address is what mechanism is available to license wh-in-situ in
Indonesian without moving it while, at the same time, capturing the observed properties
associated with this construction.
Now, let us think what we can do about this state of affair under the thesis of
minimalist interface. Given that the syntax-external semantic component is fundamentally
interpretive, utilizing whatever domain-specific operations available within this component
to assign a semantic interpretation to the output of syntactic derivation, the current thesis
expects that the fact that wh-in-situ remains in situ in Indonesian is simply the result of the
syntactic computation: under the Last Resort view of movement, for example, this class of
expressions remains in situ since there is no morphological feature that legitimizes such
movement. The minimalist interface thesis, therefore, entails that the syntactic and
377
semantic properties associated with wh-in-situ in Indonesian should be solely accounted
for in terms of the syntax-external semantic interface.
Crucially, an important series of work by Reinhart (1992, 1997, 1998, 2006) appeared
soon after the publication of Saddy (1991), which claims that there is an independent
interpretive mechanism available for licensing indefinite expressions, including whphrases, in terms of allowing existential quantification over choice functions in the sense
of Heim (1982).10 Reinhart argues that introducing this way of licensing allows for a
unified explanation for the set of syntactic and semantic problems that would arise with
analyses in terms of LF movement, unselective binding, and absorption. Let us, therefore,
first review her work on choice function.
4.1. Choice Functions
Reinhart (1992, 1995, 1997, 1998, 2006) starts by showing that neither the LF movement
(Huang 1982) nor unselective binding (Pesetsky 1987)/absorption (Higginbotham and May 1981)
analyses of wh-in-situ in multiple questions in English are tenable on the grounds that they
cannot derive several syntactic and interpretive properties associated with this type of
expressions. Consider first examples as in (29a-d).
10
I am grateful to Heidi Harley (personal communication) for suggesting this idea to me.
378
(29) a.
Who fainted when you attacked whom?
b. *
Who fainted when you behaved how?
c. *
How did Max faint when you behaved?
d.
Who fainted when you behaved what way?
(Reinhart 1998: 31, 44)
It has been standardly assumed since the seminal work by Huang (1982) that overt syntactic
movement obeys both subjacency and the ECP whereas covert LF movement is only
constrained by the ECP. This line of analysis correctly predicts the asymmetry between (29a)
and (29b) once we assume that in-situ wh-phrases undergo covert movement into the matrix
C. The example in (29a) is grammatical because the trace of whom is head-governed by the
verb attacked. The example in (29b) is ungrammatical because the trace of how does not
satisfy the antecedent-government requirement of the ECP, on a par with the example in (29c).
Reinhart notes, however, that this line of analysis inspired would, then, have no way of
accounting for why where is a contrast between (29b) and (29d); it would incorrectly predict
the latter to be ungrammatical because what way is just as much an adjunct as how. This
contrast, therefore, shows that what matters is not LF movement but instead the difference in
syntactic category (NP vs. non-NP) for the purposes of licensing wh-in-situ expressions in
English (see also section 4.3 for relevant discussion on Indonesian examples akin to (29d)).
379
Another problem with this LF movement analysis is that this way of assigning matrix
scope by LF movement is inconsistent within the Minimalist Program (Chomsky 1995).
One of the central ideas of this framework is that movement is subject to the “shortest
steps” requirement (recall also chapter 2). To illustrate, consider examples as in (30a),
which is assigned the LF representation in (30b) under the matrix scope reading of the insitu wh-phrase what (in which case, the question can be answered by sentences like Max
knows where to find bicycles).
(30) a. Who knows where to find what.
b. for which <x, y>, x knows where to find y
(Reinhart 1998: 33)
The LF movement analysis would claim that the relevant in-situ phrase would take the
matrix scope by undergoing LF movement into the specifier of the matrix, scopal C.
Reinhart notes that this very movement is impossible within the minimalist framework
because it is less economical in terms of the shortest movement requirement than its
potential movement into the specifier of the embedded CP. Thus, the example here shows
that the scope assignment of wh-in-situ in LF movement is untenable and that a nonmovement licensing is needed.
380
Reinhart further shows that the non-movement approach to wh-in-situ in terms of
unselective binding/absorption also fails in light of the interpretation of examples as in (31).
(31) Who will be offended if we invite which philosopher?
(Reinhart 1998: 36)
Reinhart assumes the semantics of questions proposed by Karttunen (1977) and Engdahl
(1986), namely, that the denotation of a question is the set of propositions which constitute
true answers to it. The unselective binding/absoportion mechanism would assign the
interpretation in (32a), which would be more formally represented as in (32b) under
Karttunen’s model, for the sentence in (31).
Wrong:
(32)a. for which <x, y>, if we invite y and y is a philosopher, then x will be offended.
b. {Pé(∃<x, y>)} & P = ^ ((we invite y and y is a philosopher) → (x will be offended) & true (P))
c. Lucie will be offended if we invite Donald Duck.
(Reinhart 1998: 36)
381
It is important to note that, in the representation in (32a), the restriction is contained in the
implication of an if-clause. Given the truth-theoretic conditions on such a clause, the
sentence in (31) would come out true in cases where the value of y is a member of the nonphilosopher set; for example, the sentence would be true if Donald Duck is inserted since
he is not a philosopher, as in (32c). This is because the only case where a conditional is a
false statement is when a true antecedent if-clause leads to a false consequence clause; if
Donald Duck were the value of y, then the antecedent clause would be false, hence the
truth value of the sentence in (31) as a whole would be true. Clearly, this is not what the
English sentence means. What we need to ensure, then, is thus to pull out the restriction
from the implication as in (33a) or its Karttunen-style equivalent in (33b).
Right:
(33) a. for which <x, y>, y is a philosopher, and if we invite y, x will be offended.
b. {Pé(∃<x, y>)(y is a philosopher) & P = ^ ((we write y) → (x will be offended)) & true (P))}
(Reinhart 1998: 36)
Thus, examples as in (31) show that the absorption/unselective binding is not adequate for
assigning scope for wh-in-situ phrases in English wh-questions. At the same time, we can see
382
from these examples that the ultimate semantic mechanism should be able to ensure that the
value of the wh-in-situ will be necessarily chosen from the set of members that satisfy its
accompanying restriction.
Reinhart argues that all the problems noted above with LF movement or unselective
binding/absorption are straightforwardly resolved once we allow existential quantification
over choice functions in the form of existential closure in the sense of Heim (1982); see
also Winter (1997) for a quite similar approach. Choice function is defined as in (34).
(34) A function f is a choice function (CH (f)) if it applies to any non-empty set and yields
a member of that set.
(Reinhart 1997: 372)
Let me illustrate how this choice function mechanism works to solve the problems noted
above. For example, the LF representation of examples as in (35a) under the wide scope
reading of the indefinite is shown in (35b) under the choice function approach.
(35) a. Every lady read some book.
b. ∃f (CH(f) & (∀z) (lady (z) → z read f (book))
383
In (35b), the indefinite book is replaced by a function variable to be bound by an existential
operator that is base-generated in the highest level. The choice function here, thus, applies to
the non-empty set of books and picks up one member out of this set. This representation says
that there is a function f such that for every z, if z is a lady, z reads the book selected by this
function. More informally, then, this representation states that there is a book that is read by
every lady, i.e. the wide scope reading of the indefinite expression.
This analysis provides a straightforward solution to the problems we have noted above
with LF movement or unselective binding/absorption approaches. First, the contrast between
(29a, d) and (29b) naturally follows under the standard assumption since Szabolcsi and
Zwarts (1993) that adverbial wh-phrases do not have an N-set and that they denote functions
ranging over higher-order entities. The examples in (29a, d) are both grammatical because
whom and what way contain an N-set (see also Higginbotham (1985) for his notion of
argument index and discharge), which is a necessary condition for choice function to work.
On the other hand, the example in (29b) is ungrammatical because how cannot be evaluated
by choice function due to its lack of N-set; the movement that is required would be blocked
by the shortest movement requirement.11 The same conclusion is reached in Tsai (1999). He
11
This is not what Reinhart actually argues for. Specifically, Reinhart (1997, 2006) argues, drawing on her
earlier work in Reinhart (1981), that adverbial wh-phrase, base-generated in [Spec, QP], can only be licensed
in [Spec, CP] and that a single sentence can host only one wh-phrase in such a position. Since the sole
purpose here is to show that the contrast is a direct consequence of choice function, I have stated here that the
384
argues that the adjunct vs. argument asymmetry here is actually the nominal wh vs. adverbial
wh asymmetry on the ground that adverbials do not contain a variable, hence must undergo
covert syntactic movement that would result in island/ECP violations, acknowledging (p.
183) that this possibility has been actually entertained in Huang (1982).
The “Donald Duck” problem, which we have seen to arise with unselective
binding/absorption above, is also directly solved as the consequence of the choice
function because, as defined in (34), a choice function applies to a non-empty set of
individuals and yields a member out of this set. Accordingly, the value for y in
examples such as (31) is correctly chosen from the set of philosophers.12
We should add two other remarks on Reinhart’s analysis at this point. First, the choice
function analysis closely mirrors the intuition expressed by Saddy (1991) that natural
languages may well develop a non-syntactic mechanism of licensing wh-in-situ in its base
position without relying on syntactic movement. Second, this licensing strategy seems to be
a natural analysis developed by the syntax-external semantic component to reflect the needs
of the conceptual-intentional system, given that the wh-in-situ literally remains in situ in
ungrammaticality of (29b) is due to the shortest movement as one possibility. This move is conceivable given
her conception of the shortest movement requirement made earlier in the text. At any rate, this decision does
not affect our discussion in this chapter.
12
As mentioned in note 4, Toyoshima (2004) applies the choice function analysis to Japanese wh-questions.
See also Lin (2004), who extends the choice function analysis to wh-questions in Mandarin Chinese.
385
Indonesian as far as syntax goes. Therefore, it stands to reason that the characteristic
properties of wh-in-situ are derived as the result of the semantics developing its own
interpretive strategy. This line of thinking is exactly what we expect under the thesis of
minimalist interfaces, according to which the semantic interface does whatever it can to
assign an interpretation to the output of syntax. Recall that the results of my earlier chapters
have suggested that semantics assign interpretation to the output of syntax in a way that is
required by the syntactic information contained within the output. Similarly, in the present
case, semantics not only accepts the needs of syntax (in the sense that it must interpret wh-insitu in its base position) but also must be rich enough to yield a number of new
interpretations to satisfy the needs of the conceptual-intentional system that would be
unavailable by the purely syntax-driven mechanisms of wh-scope assignment as in LF
movement, unselective binding, and absorption. I believe that this state of affairs is quite
natural in light of the fact that the semantics is the interface between the language faculty and
the conceptual-intentional system in the sense of Chomsky (1995, 2000, 2001, 2004, 2005).
It is undeniable fact that it serves as the information access point between language and the
general concept system (belief, context, intention, etc.). I will return to this point in section 5.
Of course, it is a different question whether this conceptual expectation is satisfied or not
in the particular case of wh-in-situ in Indonesian. In the rest of this chapter, I demonstrate
386
that this expectation is fulfilled by showing that a) Reinhart’s version of the choice function
approach straightforwardly derives all the properties associated with wh-in-situ and that b) it
allows us to make further empirical predictions that are indeed correct with regards to the NP
vs. non-NP asymmetry with regards to the putative ECP effects.
4.2. Deriving the Properties of Wh-in-Situ in Indonesian
Let us now see whether all the syntactic and semantic properties we discussed so far with
respect to wh-in-situ in Indonesian follow under the choice function approach by Reinhart
(1992, 1995, 1997, 1998, 2006). The relevant properties are summarized in (36a-e). Recall
that the status of the properties given in (36c-e) is unclear, as shown in section 2.
(36) Syntactic and Semantic Properties of Wh-in-Situ in Indonesian
a.
Wh-in-situ is insensitive to syntactic island/ECP effects.
b.
Wh-in-situ is not able to satisfy verbs’ +wh subcategorization requirement.
c.
Wh-in-situ is immune to weak/strong crossover effects.
d.
Wh-in-situ is scopally uninteractive, always taking non-overt wide scope.
e. Wh-in-situ is not able to support a pair-list reading for multiple wh-questions.
387
The property in (36a) is directly derived from the simple fact that wh-in-situ in
Indonesian remains in situ throughout the syntactic derivation (including on the mapping
to LF). The property in (36b) is derived for the same reason; this class of expression
cannot satisfy the [+WH] subcategorization requirement of verbs such as ingin tahu
‘wonder’ because it does not undergo movement nor form a formally clause-typed
question. Similarly, to the extent that the property in (36c) holds in Indonesian (recall our
earlier conclusion that this property itself is silent about whether the movement has
occurred or not), the lack of crossover effects is a natural consequence of the fact that
wh-in-situ does not undergo movement.
The properties in (36d, e) need more elaborate discussion. We have seen in section 3.3
that, according to Saddy (1991), unlike overtly moved wh-phrases, in-situ wh-phrases do
not show scope interaction with quantifiers that c-command them but instead take only
widest scope. To the extent that this judgment is real, his observation naturally follows
from Reinhart’s assumption that existential closure of a function variable introduced by
an NP can be introduced by the highest possible position. Thus, sentences as in (37a),
repeated here, would receive an LF representation in (37b) under choice function.
388
(37) a.
Setiap
orang
men-cintai
every
person TR-love
‘Who did every person love?’
siapa?
who
who>every, *every>who
b. {Pé∃<f> (CH(f)) & (∀x) & P = ^ (person (x) → x loves f (person)) & true (P))}
In the LF representation in (37b), the choice function applies to a set of persons in a
given world and picks out one of them from this set. The representation states informally
that the denotation of the sentence in (37a) is the set of true propositions, each stating
that there is a single function f such that, for every x, if x is a person, x loves the person
selected by f. This representation, therefore, corresponds to the wide scope reading of the
in-situ wh-phrase over the universal quantifier in subject position, thereby yielding
Saddy’s observation about the widest scope.
We have seen in section 3, however, that Saddy’s characterization of the scope of
wh-in-situ in Indonesian not only could not be reproduced in the present study but also
goes against the judgments reported by Cole and Hermon (1998) and others. This
raises the question of whether the existential operator that binds the function variable
can be introduced in the scope of another quantifier. If the answer is yes, the scope
interaction between wh-in-situ and c-commanding universal quantifiers, as reported by
389
my language consultants, is naturally predictable. Indeed, Reinhart shows, citing
examples as in (38a) from Ruys (1992), that the operator can be inserted within the
scope of another operator, as in (38b). See also Winter (1997) for the claim that
function variables can be bound by an operator at any level.
(38)a. Most linguists have looked at every analysis that solves some problem.
b. For most linguists x, (∃(f)) (CH (f) & (∀y) (analysis (y) and y solves f(problem)) → (x looked at y)).
(Reinhart 1998: 40)
According to Reinhart (1997: 40), in the example in (38a), “the choice of a problem
may vary with the choice of a linguist, in which case some problem is not “specific.”
Nevertheless it can take scope over every analysis.” Thus, this intermediate reading of
the indefinite some problem is naturally accounted for if the existential operator is
below another quantifier most, as shown in (38b). This analysis, of course, predicts that
the corresponding wh-in-situ in Indonesian should also be able to take this intermediate
scope in Indonesian sentences structurally akin to the English sentence in (39a). This
390
prediction is indeed confirmed by the same consultant. The LF representation for (39a)
then looks like (39b). 13
(39)a. Tiga
three
siswa
mempertimbangkan
student consider
masalah
yang
mana.
problem
that
which
setiap
analisis
yang
memecahkan
every
analysis
that
solve
‘Three students considered every analysis that solved which problem?’
three> which>every (the nature of problem is different depending on each student; the
nature of analysis is different depending on each problem)
b. For three students x, (∃(f)) (CH (f) & (∀y) (analysis (y) and y solves f(problem)) → (x consider y)).
13
For this case, I have elicited the judgment reported in (39) as follows. I asked the three consultants whether the
sentence can be used for the following scenario and wrote the following diagram: there are three students (A-C),
each one of the students works on three different problems (1-3), and each one of the problems has two different
analyses that have been proposed in the history of academia (U-Z).
(i) Student A
Problem 1
Student B
Problem 2
Student C
Problem 3
Analysis U
Analysis W
Analysis Y
Analysis V
Analysis X
Analysis Z
The consultants all reported that the sentence can be used to describe this scenario, which instantiates the
intermediate reading we are interested in.
391
This fact would remain mysterious under Saddy’s (1991) account because it crucially depends
on the observation that wh-in-situ in Indonesian always takes the widest possible scope.
The same observation also allows us to tease apart several competing theories of choice
function. Thus, Kratzer (1998) and Matthewson (1999) (see relevant discussion in note 5) both
argue that choice function variables can be bound only by an operator that is in the highest
possible position. Drawing on Fodor and Sag’s (1982) observation that indefinites in English
are ambiguous between a specific and a quantificational expression, Kratzer argues that an
indefinite must have widest scope by being evaluated by choice function when it is specific.
The consequence of this analysis is that the intermediate scope reading of the kind we
observed in English and Indonesian should be impossible. Matthewson provides further
arguments for Kratzer’s ambiguity-based approach to the scope behavior of indefinites from
her analysis of indefinite determiners in St’át’imcets (Lillooet Salish). Observing that indefinite
determiners in this language come in two types, the polarity determiner ku and non-polarity
determiners, Mattewson shows that a non-polarity determiner-headed element such as ta
twíw’t-a ‘a child’ must take the widest possible interpretation, as shown in (40).
392
(40) [tákem i
wa7 tsunám’-cal]cuz’
wa7 qwenúxw-ahhts’a7 ih-káw-lec-as
[all
Det.Pl Prog teach-Intr going.to Prog sick-inside
[ta
twíw’t-a]
Det
child-Det
Hyp-far-Intr-3Conj
‘Every teacher will be sad if a child quits.’
i. Accepted in context: There is one child, who every teacher doesn’t want to leave.
ii. Rejected in context: For each teacher, there’s one child who s/he doesn’t want to leave.
iii. Rejected in context: Every teacher will be sad if any child leaves.
√ widest
* intermediate
* narrowest
(Matthewson 1999: 119)
We have seen above, however, that this kind of intermediate scope reading is possible in
English and Indonesian. Thus, examples as in (38a) and (39a) serve to tease part the
predictions of Reinhart’s/Winter’s analyses vs. Kratzer’s/Matthewson’s analyses regarding
the possible position of the existential quantifier and support the idea in the former type of
analysis that nothing in principle blocks the introduction of existential closure in the scope of
another quantifier/operator.
393
Let us now turn to the property in (36d), namely, that wh-in-situ in Indonesian is not
able to support a pair-list reading for multiple wh-questions, as reported by Saddy (1991).
In section 3, we have seen that this property is also controversial, as my consultant work
has indicated. Nonetheless, whichever judgment is correct in certain dialects of
Indonesian follows under the choice function approach because nothing blocks basegeneration of the existential operator in the highest position or in the scope of another
quantifier; the first option yields Saddy’s observation while the second option yields the
scope interaction reported by my consultants.
Thus, the choice function approach proposed by Reinhart allows for a unified
explanation of a set of structural and interpretive properties associated with wh-in-situ in
Indonesian,
some
of
which
would
remain
mysterious
under
syntactic
movement/unselective binding/absorption/definite description-based accounts.
4.3. New Predictions: The NP vs. non-NP Asymmetry
Given the definition of choice function given in (34), we can make the prediction that the
availability of licensing via choice function crucially depends on whether a given in-situ
wh-phrase in Indonesian can denote an N-set. In other words, we predict that the NP vs.
non-NP constraint should be observed essentially in the same way as in the English
394
examples in (29a, b, d). Cole and Hermon (1998: 226) show that this prediction is indeed
borne out. Examples are in (41a-g) below are constructed in Indonesian based on the
related but partial paradigm from Malay reported in Cole and Hermon (1998: 226).
(41) a.
Siapa mem-beli buku?
who
e.
AV-buy book
Esti AV-buy book in
‘Who bought a book?’
b.
Esti mem-beli
apa?
Esti AV-buy
what
way what
‘In what way did Esti buy a book?’
f. * Esti mem-beli
Esti AV-buy
‘What did Esti buy?’
c.
Esti mem-beli buku dengan cara apa?
buku
mengapa?
book
why
‘Why did Esti buy?’
Esti mem-beli
buku
dimana? g.
Esti mem-beli
buku
untuk apa?
Esti AV-buy
book
where
Esti AV-buy
book
for
‘Where did Esti buy a book?’
d.* Esti mem-beli
Esti AV-buy
buku
bagaimana?
book
how
‘How did Esti buy a book?’
‘For what did Esti buy a book?’
what
395
The examples in (41a, b) are naturally predicted to be grammatical because the in-situ whphrases siapa ‘who’ and apa ‘what’ denote an N-set (the set of persons and the set of things,
respectively). The example in (41c) also falls into place because demana ‘where’ is
actually bimorphemic, consisting of di ‘in’ and mana ‘place’. What is crucial for us is
the contrast between (41d, f), on the one hand, and (41e, g), on the other. The examples
in (41d, f) are both ungrammatical because bagaimana ‘how’ and mengapa ‘why’ are
wh-adverbials, hence need to undergo movement to be licensed but there is no
morphological reason to legitimize this movement. The examples in (41e, g) are both
grammatical because cara apa ‘what manner’ and untuk apa ‘for what’ clearly contain
an NP that contributes an N-set required for choice function to apply. Note that this
contrast provides another argument against the Huang-style ECP-based LF movement
analysis of wh-in-situ in Indonesian, as the LF movement of the in-situ wh-phrases in
(41e, g) would uniformly violate the ECP, rending these examples ungrammatical.
4.4. Cole and Hermon’s 1998 Unselective Binding vs. Choice Function
A different implementation of the non-syntactic, in-situ approach to wh-in-situ has been
independently proposed by Cole and Hermon (1998, 2000) on the basis of data from
Malay. In this subsection, I consider whether the choice function approach and their fine-
396
grained unselective binding approach would make any different empirical predictions
concerning wh-in-situ in Indonesian and show that the “Donald Duck” Problem and the
intermediate scope reading of wh-in-situ in multi-clausal wh-questions pose non-trivial
problems for Cole and Hermon’s analysis.
Cole and Hermon (1998: 240) propose that “in wh-in-situ in Malay the (wh-OP) question
operator is merged at the root Spec CP, and, therefore, unselectively binds a wh-variable in its
scope.” This analysis derives essentially the same set of facts concerning Indonesian wh-in-situ
as the choice function analysis. There are no island effects in syntax or at LF because this class of
expression does not undergo syntactic movement but instead gets interpreted in situ by being
bound by an operator base-generated in the matrix specifier of CP. The fact that wh-adverbials
cannot be interpreted in situ also follows from the inability of such elements to be bound by
unselective binding along lines suggested by Reinhart and Tsai. It appears, then, that their
analysis might amount to the same thing as our choice function approach. Indeed, they note
(Cole and Hermon1 998: 240) that “since it does not affect the issues under consideration in this
paper, we will maintain the pretense that the question operator binds the wh-variable directly
rather than through the mediation of a choice function, and shall continue to employ the term
‘unselective binding’.” The same position is maintained in Cole and Hermon (2000: 106), who
remark that “Reinhart 1995 argues that the correct mechanism for in situ interpretations of wh is
397
a choice function rather than unselective binding. We leave this issue open since the precise
mechanism for in situ interpretation is irrelevant for our analysis.” Cole and Hermon (2000),
however, is more explicit in their analysis of wh-in-situ, as shown in their schematic
representation of their unselective binding analysis (p. 109).
(42) Unselective Binding of Wh-in situ:
[CP OPi [... [CP ...whi]]]
where wh is a variable in a base-generated position and OP is base-generated in scopal
position and binds wh.
(Cole and Hermon 2000: 109)
The explication of Cole and Hermon’s (1998, 2000) analysis above, therefore, might
give the impression that the choice function approach is a notational variant of their
proposed version of unselective binding. Nonetheless, I show below that the predictions
do diverge in a number of domains related to the “Donald Duck” Problem and the
intermediate scope reading and that the choice function analysis is superior to Cole and
Hermon’s analysis on these grounds.
The first divergence in prediction between the choice function approach and Cole and
Hermon’s (1998, 2000) version of the unselective binding concerns the interpretation of
398
in-situ wh-phrases in Indonesian contained within an if-clause. Recall that Reinhart showed
that the LF representation of examples such as (31), repeated here as (43), which would be
derived under selective binding, fails to express the fact that the value of the denotation of
the in-situ wh-phrase must be selected from the set of philosophers because this approach
would leave the restriction (philosopher) in the implicational clause at LF, as shown in
(32a, b) (= (44a, b)), and render the sentence in (32c) (=(44c)), true even though the value
of y is Donald Duck, contrary to facts.
(43) Who will be offended if we invite which philosopher?
(Reinhart 1998: 36)
Wrong:
(44)a. for which <x, y>, if we invite y and y is a philosopher, then x will be offended.
b. {Pé(∃<x, y>)} & P = ^ ((we invite y and y is a philosopher) → (x will be offended) & true (P))
c. Lucie will be offended if we invite Donald Duck.
(Reinhart 1998: 36)
As we saw in section 4.1, this Donald Duck problem won’t arise under the choice function
approach because the value of y must be selected from the non-empty set of philosophers in a
399
given model/world by virtue of the very definition of choice function given in (34). Therefore,
Cole and Hermon’s analysis cannot capture the correct interpretation unless it is accompanied
with special mechanisms of pulling out the restriction out of the antecedent of an implicational
clause to circumvent the “Donald Duck problem. Several technical additions, of course, would
not be inconceivable that would potentially solve this problem; for example, the
representations in (44a, b) could be converted to something like (45a) below, in which the
value of y can only be selected from the set of philosophers. However, as Reinhart (1992)
shows, whatever addition that yields this presentation would rule in not only the desired
reading in (45b) but also the undesired reading in (46c).
(45) a. Lucie will be offended if Donald Duck is a philosopher and we invite him.
b. Lucie will be offended if we invite Kripke.
c. Lucie will be offended if Kripke is a philosopher and we invite him. (Reinhart 1992: 3)
The point here is that no such special addition is required under the proposed approach; instead,
the effects of such stipulations are automatically derived from the definition of choice function
as applying to a (non-empty) set of individuals and picking up one member from that set.
400
The second empirical domain in which the predictions of the two competing
approaches would diverge concerns the intermediate scope reading illustrated by data as in
(39a), repeated as (46a). The LF representation in (46b) is derived under the choice
function, which correctly captures the reading where masalah yang mana ‘which problem’
takes wide scope over setiap analisis ‘some analysis’ but takes narrow scope with respect
to tiga siswa ‘three students.’
(46)a. Tiga
three
siswa
mempertimbangkan setiap
student consider
masalah
yang
mana.
problem
that
which
every
analisis
yang
memecahkan
analysis
that
solve
‘Three students considered every analysis that solved which problem?’
three> which>every ( the nature of problem is different depending on each student;
the nature of analysis is different depending on each problem)
b. For three students x, (∃(f)) (CH (f) & (∀y) (analysis (y) and y solves f(problem)) → (x consider y)).
It is not clear, however, whether this intermediate scope reading would be derived under
Cole and Hermon’s version of the unselective binding approach. In their 1998 paper, they
401
assume that the wh-in-situ in Malay is bound unselectively by the operator base-generated in
the root [Spec, CP]. Accordingly, it would falsely predict that the intermediate scope reading
observed in (46a) would be impossible. The same problem remains with their analysis
updated in their 2000 paper explicated in (42) because it base-generates the operator in the
scopal [Spec, CP]. Crucially, however, the intermediate reading in (46a) requires that the
operator must be base-generated in a position in the matrix clause that is lower than the
specifier of the matrix TP but higher than the complement of the matrix VP. Thus, the
relevant reading would be impossible, contrary to facts.
Based on these two divergences between Cole and Hermon’s (1998, 2000) analysis and the
choice function analysis, I conclude that the two analyses are not entirely the same; the latter
analysis makes better empirical predictions concerning the “Donald Duck” Problem and the
intermediate scope reading in multi-clausal wh-questions. The two problems could be technically
solvable by several special amendments on the mapping from syntax to LF under Cole and
Hermon’s analysis but the fact that these amendments are not necessary but instead derived from
the way choice function independently works provide support in favor of the choice function
analysis adopted in this chapter.
402
5. Conclusions
Let me summarize the whole chapter. I have started by reviewing several peculiar
structural and interpretive characteristics associated with wh-in-situ in Indonesian
originally discovered by Saddy (1991). The lack of island effects and the failure of the
[+WH] subcategorization requirements show that neither overt movement analyses as in
Watanabe (1992, 2001) and Richards (2001) nor covert movement analyses as in Huang
(1982) and Lasnik and Saito (1992) are not suitable mechanisms of licensing wh-in-situ in
Indonesian. The difference in morphological composition between wh-phrases in English
and Indonesian also casts doubts on the general applicability of Pesetsky’s (1987) variant
of the unselective binding approach to Indonesian. During the course of this demonstration,
I have reported judgments from my language consultants that are contradictory with those
reported by Saddy with respect to the lack of pair-list readings in multiple questions and
the obligatory widest scope reading of in-situ wh-phrases.
I have proceeded to show that there are several problems with Saddy’s own analysis of
wh-in-situ in Indonesian as an interrogative definite description. In particular, this analysis
is in direct contrast with the assumption since the earliest days of generative grammar that
wh-phrases are indefinite expressions that contain an individual variable. Of course, the
island-insensitive behavior of wh-in-situ in Indonesian could follow from Saddy’s
403
treatment of this class of expressions as definite descriptions, but this analysis directly
contradicts the motivated assumption about wh-phrases stated above; rather, a more natural
possibility is to show that the scope-taking property is a special case of existential
indefinites, which are independently known to freely take wide scope, in massive violation
of known island constraints on syntactic movement. The evidence based on reduplication
from Cole and Hermon (1998) constitutes further evidence against the characterization of
the in-situ wh-phrases in terms of definite description.
In light of these considerations, I have proposed that all the apparent peculiar
properties of wh-in-situ in Indonesian receives a unified explanation once we allow
existential quantification over choice functions along the lines suggested by Reinhart
(1992, 1997, 1998, 2006). I have further compared this analysis with a fine-grained
analysis of unselective binding developed by Cole and Hermon (1998, 2000). Based on
facts regarding on the “Donald Duck” Problem and the intermediate scope reading of
wh-in-situ in multi-clausal contexts, I have argued that the former approach is superior
to the latter approach in that it requires no special technical stipulations that would be
required under Cole and Hermon’s analysis. The overall result of this chapter, therefore,
provides support that choice function should be admitted as a natural mechanism of
wh-licensing on the part of the semantic interface that does not depend on syntax.
404
One question that the proposed analysis leaves unresolved is, why choice function
seems not to apply for embedded questions headed by interrogative verbs?14 We have
seen from the contrast between (6a) and (6b) that verbs like ingin tahu ‘wonder’
necessarily requires overt movement of a wh-phrase into the specifier of its
complement CP (see also notes 5 and 12 for relevant discussion). This question could
be resolved if “part of the mechanism of quantificational WH construal is derivative
from the simple fact of movement having taken place.” (Saddy 1991: 212). Specifically,
the fact that overt movement occurs means that the (regular) quantificational whconstrual is available for the LF interface. Therefore, starting up a new domain-specific
operation (choice function) is more costly than the already activated syntactic
operation. Notice that the former option is also prohibited independently because of the
minimalist interface thesis that the semantic interface attempts to blindly follow the
carve parametrically defined by syntax. In Chinese and Japanese, however, syntactic
movement as a way of wh-construal is impossible for independent syntactic reasons
(such as strength of Q-features). Thus, the LF component activates the choice function
option as a case of Interface Economy (Reinhart 2006) to derive a variety of
14
Thanks to Heidi Harley (personal communication) for raising this important question.
405
interpretations that other movement languages would derive by regular wh-movement
(but cf. the discussion in note 5 above for an alternative viewpoint).
The results achieved in this chapter have certain important implications for the way the
syntactic computation interacts with the semantic component. As I have shown in this chapter,
the choice function analysis along the lines proposed by Reinhart (1992, 1995, 1997, 1998,
2006) provides a straightforward solution to various syntactic and semantic problems in
Indonesian with movement-based and unselective binding analyses proposed for other Asian
languages such as Chinese and Japanese. The plausibility of the choice function analysis
appears to suggest that the semantic component is relatively independent from the needs of
syntactic computation per se but serves more the needs of the system of concepts (inferences,
contexts, concepts) that interacts with the faculty of language. Of course, the fact that the NP
vs. non-NP distinction is crucial for the application of choice function indicates that certain
purely syntactic information is used for the purposes of semantic interpretation on the part of
the linguistic interface between syntax and the conceptual system, but it may well be the other
way around; the NP vs. non-NP distinction is a syntactic manifestation of the application of
choice function and the lack thereof; syntax happens to provide sufficient tools to work out the
semantic interpretation right on the part of the syntax-external interpretive component. One
406
speculation that would emerge from the present study of wh-in-situ in Indonesian is, then, that
the relevant interface does whatever it can to meet the needs of communication.
A natural question to ask here, then, is why doesn’t this option manifest itself with English
wh-questions. More specifically, why does Indonesian allow three types of questions (overt
wh-movement, partial wh-movement, and wh-in-situ) whereas English allows only overt whmovement and wh-in-situ? This difference between Indonesian and English as well as its
semantic corollaries may well be captured as another case in which the minimalist interface
thesis is at work; it is because the LF interface blindly follows the carve parametrically defined
by syntax. Cole and Hermon (1998, 2000) propose a parametric theory of wh-questions
whereby languages differ in whether a question word consists in an operator and variable
combined in a single word or of a variable generated in a thematic position separately from a
base-generated Q-operator. Malay/Indonesian can take both feature bundling options. Suppose
that the variable is generated separately from the base-generated wh-operator. Then, since there
is no morphological feature to legitimize the movement of the variable, the Economy of
Derivation dictates that there is no movement. To satisfy the communicative needs of the C-I
system that would have been met by this movement, the LF interface applies choice functions
to this in-situ object. If the variable and the operator are bundled together in a single word, then
full syntactic movement must occur to create an operator-variable structure and obtain
407
appropriate interrogative scope. The partial movement option could arise in Malay/Indonesian
either as a type of prolepsis (cf. Davies 2003, 2005) or due to the existence of the null whexpletive (Cole and Hermon 1998) in these languages. In English, on the other hand, the whoperator and variable must be bundled together but there is an additional requirement that only
one wh-phrase suffices to type a clause as a question; other wh-expressions must remain in situ
and hence evaluated at the interface via choice functions. In this way, the difference in
diversity of wh-questions between Malay/Indonesian and English follows from the way that
certain morphosyntactic features are bundled in each language and the way LF reacts to it.
Whether there is independent morphological/syntactic evidence for this type of lexical
parameterization in domains other than wh-questions is an important task to be undertaken.
We have seen thus far in this dissertation that the syntax-external linguistic interfaces are
subservient, doing the best they can to meet the needs of the universal computational system,
syntax, within the parametrically defined range of options available to a particular language.
We have seen in the preceding chapters that this characterization is on the right track with
respect to the interaction of syntax with phonology and certain aspects of semantics such as
the denotation of bare nominals. The results reported in this chapter seem slightly deviant
from those reported in the earlier chapters in that they suggest that the semantic component
is relatively more independent from the requirements of syntax and serves the
408
communicative needs imposed from our language-independent conceptual system, as hinted
at the outset of section 4. However, it is also possible to draw a different conclusion from the
current investigation of wh-in-situ in Indonesian, namely, that the choice function approach
is developed in the syntax-external semantic component due to the very fact that syntax does
not move wh-in-situ in tandem with the language-external requirements from the conceptual
system. Within the deterministic theory of movement, as adopted in the Minimalist Program,
movement occurs only to satisfy the morphological features of the element that moves/is
moved to. Thus, what syntax can do is to leave a wh-in-situ in its base position and send the
derivation that contains this structure to the semantic component. It seems, then, that there
might be no way that semantics would interpret the relevant structure. However, under the
minimalist interface thesis argued for in this dissertation that semantics is purely interpretive,
the very fact that syntax sends this configuration means that semantics should be rich enough
to be able to accommodate and interpret this configuration. There are several ways in which
this can be done in the semantic interface. The approach I have pursued in this chapter is that
the semantic component independently applies the interpretive mechanism of choice
function to satisfy the communicative needs of the language-external conceptual system.
This interaction between the linguistic interface and the system of concepts is not unreasonable
on two counts. First, one primary way language can be used is communication from the
409
perspective of the conceptual system, hence the exact form of the interpretive method might be
heavily influenced by language-external requirements imposed from the conceptual system.
Second, the system of concepts takes care of information such as concepts, logical reasoning,
contexts, beliefs and complex thoughts. Thus, the choice function and its accompanying
notions such as existential quantification, set formation, non-empty set, and so can be analyzed
as the linguistic manifestation of the interface strategy that the linguistic semantic component
happens to have selected on the basis of its interaction with the concepts system.
To the extent that the above noted scenario is tenable, it indicates that syntax is actually an
“imperfect” system for the purposes of communication; the faculty of language is poorly
designed for this purpose. This is because it does not provide sufficient information that the
semantics could access in order to find a unique way to accommodate a multitude of
communicative needs as required by the conceptual system. This is hardly surprising, of
course, under the minimalist interface thesis, according to which syntax is totally blind to the
external linguistic interface such as phonology and semantics, as I have argued for so far in
this dissertation. Evidence abounds that this is the correct characterization of syntactic
system. Embedding in the syntax as in Why did you think that you kissed me? cannot help
creating ambiguities that should be ideally avoided for the purposes of communication (as in
the Gricean theory of cooperative maxims). See also Hinzen (2006) for many other examples
410
to illustrate this blind nature of syntactic computation, some of them already discussed in the
previous chapter of this thesis. I provide one possible explanation for why syntax has this
property of “imperfection” in terms of the evolution of the language faculty.
Based on this consideration, I conclude that syntax is indeed an autonomous computational
system that is blind to the consequences it has for the syntax-external semantic component. Syntax
sends its final outputs (e.g., structures that contain in-situ wh-phrases) to semantics, which assigns
an interpretation to them via whatever domain-specific mechanism it happens to have at its
disposal (e.g., choice function) from the way it interacts with the language-external conceptual
system. However “poor” the result of syntactic computation may be from the perspectives of the
conceptual system, the semantic component devises ways to legitimatize, repair and improve them
to satisfy the requirements of human communication. This scenario, therefore, allows us to
maintain the minimalist interface, argued for thus far in this thesis, namely, that the syntax-external
components that connect meaning and sound to structure are purely interpretive, doing whatever
domain-specific operations it can to make certain syntactic objects legible for language-external
systems to use but in a way that it is totally subservient to the principles and parameters active in a
given language. I return to further ramifications of this conclusion in chapter 7.
411
CHAPTER 6 REDUPLICATION ASYMMETRIES IN INDONESIAN: THEIR
THEORETICAL IMPLICATIONS FOR THE SYNTAX-MORPHOLOGY INTERFACE 1
1. Introduction
In this chapter, I explore issues revolving around the morphology-syntax interface
with a case study in reduplication in Indonesian. Through detailed discussion of a certain
asymmetry between nominal and verbal reduplication in this language, I propose further
evidence for the thesis of minimalist interface that the syntax-external phonological
component is fundamentally interpretive, assigning phonological interpretation solely
based on the input created by the universal syntactic computation.
I couch the examination of the validity of the thesis of minimalist interface within the longstanding debate in the contemporary linguistic theory between lexicalist and non-lexicalist
theories of the lexicon-syntax interface. I argue that the minimalist interface thesis leads us to
the non-lexicalist view as hold in the recent morphosyntactic framework of Distributed
Morphology (Halle and Marantz 1993, 1994) that words are not a triplet of sound, meaning,
and their correspondence, as commonly assumed, but rather nothing but the morpho1
This chapter is a substantially expanded version of the paper presented at the 33rd Annual Meeting of the
Berkeley Linguistic Society held at the University of California, Berkeley with Bradley McDonnell and reported in
Sato and McDonnell (in press) and Sato (2008). I am very grateful to Bradley for our corroboration, in particular,
his timely help in conducting a corpus study mentioned in Table 5 that forms the basis of this chapter.
412
phonological output of the complex objects created by syntactic derivation. According to this
view, the syntactic derivation constructs whatever objects it can, based on a particular
arrangement of morpho-syntactic features available in a given language; what the syntaxexternal phonological component can do is to assign a language-particular surface realization
to this output object post-syntactically. In this chapter, I argue that this view is correct by
showing that reduplication in Indonesian is sensitive to syntactic structure, not just morphemes,
thereby providing empirical support for the framework of Distributed Morphology.
With the phonological feature assignment being purely interpretive and post-syntactic
and governed by the way syntactic computation unfolds, the notion of minimalist
interface leads us to another claim that there is no such thing as the lexicon in the
traditional sense in which a word is constructed by processes different from syntactic
combinatorial processes and assigned a meaning and sound pre-syntactically. I show that
this non-lexicalist approach to the lexicon-syntax interface provides a natural account of
reduplication asymmetries in Indoensian. At the same time, I show that several variants
of the lexicalist theory have difficulties in accounting for the attested pattern of
reduplication in Indonesian precisely because they postulate the traditional lexicon as a
pre-syntactic/autonomous generative component. This result, therefore, provides a
413
morphological piece of evidence for the non-lexicalist view of the lexicon-syntax
interface entailed by the thesis of minimalist interface.
The empirical domain I investigate is reduplication in Indonesian. A corpus survey of
four popular newspapers published in Indonesia reveals a curious asymmetry between
nominal and verbal reduplication that has not been reported in the literature on Indonesian
morphology: nominal stems allow both stem and stem-affix reduplication whereas verbal
stems allow only stem reduplication. I show how this asymmetry as well as the word-internal
stem reduplication pattern pose non-trivial empirical and architectural difficulties for several
versions of the lexicalist theory as presented in Chomsky (1970), Anderson (1982, 1992),
Kiparsky (1982a, b, c, 1985), Mohanan (1986), and Di Sciullo and Williams
(1987)/Williams (2007). 2 I show that these difficulties arise as the artifact of traditional
2
A similar problem has been independently noted in Yaqui reduplication by Haugen and Harley (2006),
though with certain properties that distinguish it from the Indonesian reduplication pattern to be discussed in
this chapter. Haugen and Harley observe that, in Yaqui, the inflectional process of reduplication targets the
head of the verbal head of an N + V compound rather than the compound itself, as shown in (ia-c).
(i) Word-internal head-reduplication in Yaqui
a. kuta-siute
‘wood-split’
stick-tear
b. chit-wat-te
stick-Red-tear
‘spitting’
saliva-throw-Intr
c. hiavih-muuke
breathe-die
kuta-siu-siute
chit-wat-wat-te
saliva-Red-throw-Intr
‘gasping’
hiavih-mu-muuke
breathe-Red-die
(Haugen and Harley 2006:6, 7)
414
theories of the syntax-lexicon interface as in the lexicalist theory that postulates the lexicon
as an autonomous pre-syntactic generative component or system whose relation with respect
to syntax is atomic and asymmetric. The thesis of minimalist interface, by contrast, leads us
to the alternative view that words as well as sentences are objects created by the sole
generative syntactic component. Given the role of the morpho-phonological component as
subservient to the needs of syntax, the same thesis also entails that morpho-phonology is
located after syntax and assigns interpretation to the objects created by syntax. These two
views amount to the core claims that have been independently made in the morphosyntactic
theory of Distributed Morphology (Halle and Marantz 1993, 1994; Marantz 1997; Harley
and Noyer 1999; Embick and Noyer 2007). I show that this conception of the lexicon-syntax
interface provides a principled explanation of the reduplication asymmetry and the wordinternal reduplication pattern in Indonesian; these facts are straightforwardly derived as a
natural consequence of a particular hierarchical arrangement of morphosyntactic features,
including Asp and Num, which is independently motivated by the semantic/selectional
properties of certain derivational affixes such as ber- and an-. This result, therefore, provides
a strong piece of evidence against the traditional lexicalist architecture, and, at the same time,
They argue that this word-internal reduplication pattern presents evidence against lexicalist views of the
lexicon-syntax interface as in Di Sciullo and Williams (1987) and Anderson (1982, 1992) and present a new
analysis of this pattern within the framework of Distributed Morphology.
415
argues in favor of more recent non-lexicalist theories of the syntax-lexicon interface as in
Distributed Morphology that attempt to locate all types of word formation within the sole
realm of the syntactic derivation. Accordingly, I construe this result as further empirical
evidence for the thesis of minimalist interface that the syntax-external interface component is
a fundamentally interpretive system whose role is to assign whatever interpretation it can to
the outputs delivered by the universal syntactic computation.
The present chapter is organized in the following manner. In the next section, I provide an
overview of the so-called lexicalist theory in the generative tradition and compare its
grammatical architecture with more recent, non-lexicalist theories of the lexicon-syntax
interaction as in Distributed Morphology. I highlight one primary difference between the two
theories concerning the way the lexicon interacts with the syntax. Lexicalist theories predict
that no “lexical” processes, defined along a variety of dimensions, can follow syntactic
combinatorial processes such as Merge and Move (Chomsky 1995) due to their postulation
of the lexicon as a pre-syntactic generative component. In contrast, non-lexicalist theories
predict that there is no inherent ordering between the two types of operations because they
do not posit such a component, relegating the roles of the lexicon to the sole generative
syntactic component. This difference becomes important in later sections. In section 3, I
report the results of a corpus survey of four popular newspapers in Indonesian. The most
416
important among them is my new finding initially made by the pilot study in Sato and
McDonnell (in press) that there is an asymmetry between nominal and verbal reduplication
in Indoensian, namely, that derivational nominal affixes are productive in reduplicating both
the stem-affix combination and the stem alone whereas derivational verbal affixes allow
stem reduplication, but never stem-affix reduplication. In the next two sections, I consider
implications of this asymmetry for the proper theory of the syntax-lexicon interface. In
section 4, I show that this asymmetry as well as the word-internal reduplication pattern poses
non-trivial architectural/empirical difficulties for several, well-known versions of the
lexicalist theory as in Chomsky (1970), Anderson (1982, 1992), Kiparsky (1982a, b, c
1985)/Mohanan (1986), and Di Sciullo and Williams (1987)/Williams (2007). Based on this
consideration, I show, in section 5, that the two observations concerning reduplication in
Indoensian can be given a straightforward account under non-lexicalist, morpho-syntactic
theories of the syntax-lexicon interface as in Distributed Morphology if we take into account
a particular hierarchical arrangement of certain morphosyntactic features such as Asp and
Num as well as the underlying syntactic category of input stems for reduplication. In section
6, I provide a summary of the contents of this chapter and discuss their implications for the
proper theory of the syntax-lexicon interface and minimalist interfaces.
417
2. Lexicalist vs. Non-Lexicalist Theories
In this section, I compare two competing theories of the syntax-lexicon interface,
restricting attention to the debate within the framework of generative grammar: lexicalist
theories and non-lexicalist theories.
2.1. Lexicalist Theories
In this subsection, I introduce the main claims of the lexicalist theory. I show that the most
important theoretical tenet of this hypothesis for the purposes of this chapter is that the lexicon is
postulated as an autonomous generative component prior to the syntax that is responsible for
certain morphologically derived complex objects that are identified along several dimensions.
The theory of the syntax-lexicon interface most often termed the Lexicalist Hypothesis
(Chomsky 1970, 1993, 1995; Lieber 1980; Williams 1981; 2007; Anderson 1982, 1992;
Farmer 1982; Lapointe 1980, 1981; Jensen and Stong-Jensen 1984; Di Sciullo and
Williams 1987; Pesetsky 1979; Kiparsky 1982a, b, c, 1985; Mohanan 1986) claims that a)
there are two independent generative components for the formation of words and phrases
and that b) there is a strict division of labor between the two components. Under this
traditional architecture of the lexicon-syntax interaction, the products of the operations in
the lexical component serve as atomic indivisible units that syntactic processes operate on
418
as terminal nodes. As a result, the lexicalist theory adopts one or the other version of the
so-called Lexical Integrity Hypothesis, according to which principles of syntax are not
operative in generating the structure of words, the products of the lexical component. This
hybrid approach to word formation stems primarily from the time-honored observation that
“words” are somehow different from “phrases” along several (somewhat unclear)
dimensions, including semantic and phonological idiosyncrasies/compositionality,
gaps/productivity, and derivation/inflection. As a natural consequence of the strict division
of labor between the lexicon and syntax, under this lexicalist archicture of grammar, there
is no reason to expect that the interface of the syntax and the lexicon is direct; rather, the
interface between the two components may well be opaque.
Under the standard interpretation of the history of the theories of the lexicon-syntax
interface from the early 1970s to the present within the framework of generative grammar,
the Lexicalist Hypothesis comes in two varieties, strong and weak versions. The strong
version of the Lexicalist Hypothesis, represented by work as in Lieber (1980), Lapointe
(1980, 1981), Williams (1981), Farmer (1982), Pesetsky (1979), Kiparsky (1982a, b, c,
1985), Mohanan (1986), Di Sciullo and Williams (1987) and Chomsky (1993, 1995),
holds that all word formation processes occur in the pre-syntactic lexical component. The
weak version of the Lexicalist Hypothesis, which has been most often associated with
419
Chomsky (1970) (“Remark on Nominalizations”) and Anderson (1982, 1992) in the
literature, maintains that certain (regular, productive) word formation processes occur in
the syntax whereas other (irregular, non-productive) processes occur in the pre-syntactic
lexicon in a way that is conditioned by a variety of criteria, including productivity,
derivation vs. inflection, and semantic and/or morphological idiosyncrasies. This is one
interpretation of Chomsky’s position, for example, that by Spencer (1991):
(1) Spencer’s (1991: 69) interpretation of Chomsky’s (1970) work
Chomsky argued that transformations should capture regular correspondences between
linguistic form, and that idiosyncratic information belonged in the lexicon … derived
nominalizations are morphologically, syntactically and semantically idiosyncratic…”
Against this interpretation of Chomsky’s work as the birthplace of the (weak) lexicalist
theory, Marantz (1997) claims that Chomsky actually argued against a lexicalist treatment
of derived nominalizations by showing that such a treatment needs to stipulate a uniform
pattern concerning the unacceptability of the transitive use of internally caused change of
state predicates (Levin and Rappaport-Hovav 1995) such as growth (as in *John’s growth
of tomatoes). In contrast, this pattern would be naturally explained by the lack of (a certain
420
type of) v in a syntactic approach to word formation. In this chapter, I follow Spencer’s
interpretation of Chomsky’s work for the purposes of discussion.
Scholars working within the Minimalist Program (that do not adopt the theory of
Distributed Morphology; see section 2.2) seem to assume the strong version of the hypothesis
as defined above, essentially following the idea of Chomsky (1993, 1995) that syntax selects
fully inflected lexical items from the Numeration and combines them by the recursive process
of Merge.3 Within the morphological research, however, the lexicalist hypothesis is split into
the weak and strong versions as defined above. Non-Chomskyan declarative frameworks,
including Lexical Functional Grammar (Kaplan and Bresnan 1982; Bresnan 1982), HeadDriven Phrase Structure Grammar (Pollard and Sag 1994) and Construction Grammar
(Goldberg 1995, 2006), all adopt what can be termed ‘Hyper-Lexicalism”, according to which
all operations for word and phrasal formation occur in the lexical component.4
Given below are some definitions of the Lexicalist Hypothesis from the lexicalist literature.
See section 4 for detailed discussion of the weak and strong lexicalist theory presented in
3
This remark seems to be true until quite recently, though, in his most recent work (Chomsky 2006),
Chomsky also hints at the following non-lexicalist view of “words”, which was in fact what Chomsky
(1970) meant to adopt, if Marantz (1997) is correct.
“Possibly the functional category v determines the verbal character of the root R that is its complement, along
lines suggested by Alec Marantz, in which case verbal phrases are of the form v-RP.” (Chomsky 2006: 12)
4
As pointed out to me by Andrew Carnie (personal communication).
421
Chomsky (1970), Kiparsky (1982a, b, c, 1985), Mohanan (1986), and Di Sciullo and Williams
(1987)/Williams (2007).
(2) The Generalized Lexical Hypothesis (Lapointe 1981: 125)
No syntactic rule can refer to an element of morphological structure where element of
morphological structure here refers to any morphological feature, any morphological
category or any element dominated by such a category.
(3) The Thesis of the Atomicity of Words (Di Sciullo and Williams 1987: 48, 49)
Although syntactic rules can access the categorial status and argument structure of a
lexical item, they will never depend on how that categorial status or argument structure
was arrived at through morphological derivation or on the internal constituency of
words. The rules of syntax can see that a word has such and such properties, but they
cannot see how it came to have those properties. …Words are “atomic” at the level of
phrasal syntax and phrasal semantics. The words have “features,” or properties, but
these features have no structure, and the relation of these features to the internal
composition of word cannot be relevant in syntax.
422
(4) The Lexical Integrity Principle (Bresnan and Mchombo 1995: 181, 182)
...words are built out of different structural elements and by different principles of composition
than syntactic phrases. Specifically, the morphological constituents of words are lexical and
sublexical categories-stems and affixes-while the syntactic constituents of phrases have words
as the minimal, unanalyzable units; and syntactic ordering principles do not apply to
morphemic structures. As a result, morphemic order is fixed, even when syntactic word order is
free; the directionality of ‘headedness’ of sublexical structures may differ from supralexical
structures; and the internal structure of words is opaque to certain syntactic processes.
(5) The Thesis of the Atomicity of Words (Di Sciullo and Williams 1987: 48, 49)
Although syntactic rules can access the categorial status and argument structure of a
lexical item, they will never depend on how that categorial status or argument structure
was arrived at through morphological derivation or on the internal constituency of
words. The rules of syntax can see that a word has such and such properties, but they
cannot see how it came to have those properties. …Words are “atomic” at the level of
phrasal syntax and phrasal semantics. The words have “features,” or properties, but
these features have no structure, and the relation of these features to the internal
composition of word cannot be relevant in syntax.
423
Although the precise claims about the lexicon-syntax interaction vary from author to author
within the lexicalist researchers, the essence of the Lexicalist Hypothesis remains the same,
namely, that a) there are two independent components, the lexicon and syntax, and that b)
the type of primitives and processes involved in each component are distinct, with the
processes of one module being not operative to the structures or operations of the other.
Most commonly, the Lexicalist Hypothesis is implemented in a generative model of
grammar (Chomsky 1981, 1986a, 1993, 1995), as shown in (6).
(6) The Lexicalist Hypothesis (embedded within the Government-and-Binding Model)
LEXICON
SYNTAX
-Lexical Rules-
-Phrasal Rules-
idiosyncrasy
compositionality
derivational
inflectional
accidental gaps
productivity
SOUND (PF)
MEANING (LF)
This architecture entails that the lexical word-building component is sequentially ordered
prior to D-structure, the interface of the lexicon and syntax. In other words, the hypothesis
endorses a particular sequence of word formation, namely, that all word-building rules in
424
the Lexicon should precede all phrase-building rules in syntax. Rules in the lexicon are
associated with properties such as semantic/phonological idiosyncrasies, derivation and
accidental gaps. Rules in the syntax have the complement of these properties such as
semantic/phonological compositionality, inflection and productivity. This point is also
clarified by Borer’s (1998) remark below:
(7)
Borer’s (1998) Statement of Chomsky’s (1970) Weak Lexicalist Hypothesis
The way in which L[exical] I[ntegirty] H[ypothesis] is enforced in many of these models is
by assuming that the W[ord] F[ormation] component, as a block of rules, is ordered with
respect to the syntax. The WF component and the syntax thus interact only in one fixed
point. Such ordering entails that the output of one system is the input to the other. This
notion of the autonomy of the syntax and the WF component, and the restricted interaction
between them, thus mimics the notion of autonomy developed for the interaction between
the syntax and the phonology, where it is the output of the former which interacts with the
latter. (Borer 1998: 152, 153)
Lapointe (1981) is one example of work in the lexicalist literature that explicitly mentions the
relative sequential ordering of lexical processes with respect to syntactic processes. After the
425
quote given in (2) above, Lapointe (p. 125) continues as: ‘This framework has the general
organization outlined in Figure 1.’ The relevant figure is reproduced in (8), which is one
particular instantiation of the general lexicalist architecture in the sense that it posits a presyntactic lexical component (lexical System) prior to the generative syntactic component
(Syntactic System).
(8) Organization and interactions of the lexical and syntactic systems with the lexical
insertion and semantic translation mappings in the grammatical theory of Lapointe 1980.
Lexical system
Syntactic System
MS rules
PS rules
Lexical transformation rules
(Transformations)
Lexical entries
Syntactic Structures
Logical Forms
(SSs)
(LFs)
Lexical Insertion
Semantic translation
of MSs for words
mapping φ
(Lapointe 1981: 126)
426
In this architecture, the Lexical System is located prior to the Syntactic System. Based on this
kind of remark together with the generative architecture of the generative grammar (e.g. the
Government-and-Binding/Minimalist Program of Chomsky 1981, 1986a, 1995) within which
the lexicalist hypothesis has been commonly couched, it seems reasonable to make a prediction
that lexical processes cannot follow syntactic processes under any type of the lexicalist theory of
the syntax-lexicon interface that is tied with the generative model of grammar in which a Dstructure/Numeration serves as the interface between the lexicon and syntax.
This characterization is also supported in light of the following consideration. It has
been commonly assumed within the lexicalist literature (see Di Sciullo and Williams 1987
and Bresnan and Mchombo 1995; Chomsky 1981, 1986a, 1993, 1995) that principles in
the Lexicon operate only on zero-level categories that serve as atomic unanalyzable units
which the syntactic derivation uses to create phrasal-level complex objects; in other words,
the output of the lexicon seems to be the input for the syntax. It cannot be the case that the
output of syntax becomes the input for the lexical processes in the lexicalist sense because
this would mean that the lexicon deals with non-zero-level categories, contrary to what
lexicalists generally have agreed upon. Therefore, in any lexicalist model of grammar that
allows the interaction of lexicon and syntax (in whatever constrained manner it may be),
syntax cannot do its job unless the lexicon first provides atomic units. This point is in fact
427
agreed upon by the lexicalists; for example, Williams (2007: 351) notes that “the channel
of communication [between the word system and the phrasal system-YS] is asymmetrical,
by virtue of the fact that phrases are made out of words, but not vice versa. Based on this
consideration, I assume for the rest of this chapter that, under the lexicalist hypothesis, the
lexical component logically precedes the syntactic component, hence, that processes in the
lexicon can never follow processes in the syntax.
To sum up, I have introduced the general theoretical assumptions of the lexicalist theory as
presented in the literature. I have pointed out that one crucial prediction of this hypothesis is that
no lexical processes should be able to apply after syntactic processes due to the way that this
hypothesis is commonly couched within the generative architecture of grammar where Dstructure/Numeration serves as the interface between the lexicon and syntax and the lexicon
provides unanalyzable terminal units that syntax operates on to create complex phrasal-level
objects. As we will see in the next subsection, this lexicalist view is radically different from more
recent non-lexicalist views of the syntax-lexicon interface such as Distributed Morphology.
2.2. Non-Lexicalist Theories
The morphosyntactic framework of Distributed Morphology (Halle and Marantz 1993, 1994;
Marantz 1997; Harley and Noyer 1999; Embick and Noyer 2007) claims that there is a
428
single generative component– syntax – that assembles ‘words’ as well as sentences. This
theory of word formation is in direct contrast to the Lexicalist Hypothesis introduced above,
which posits two generative components – lexicon and syntax – for word formation. The
grammatical architecture of this non-lexicalist framework is given in (9).
(9) The Architecture of Distributed Morphology (adopted from Embick and Noyer 2007: 301)
LISTS ACCESSED
STAGES OF THE DERIVATION
Access to
Syntactic Derivation
Syntactic Terminals
Access to
(Spell Out)
The Vocabulary
PF
Access to
LF
(interpretation)
The Encyclopedia
The model in (9) makes it clear that, within the theory of Distributed Morphology, information
that in other theories is assumed to be solely included in the pre-syntactic lexical component is
429
“distributed” across several components of the grammar: syntax, post-syntactic vocabulary
insertion, and the Encyclopedia. The primitive elements in Distributed Morphology that syntax
manipulates come in two types. One type is roots, which are atomic unanalyzable elements;
the other type is functional heads such as n/v/a as well as other ordinally postulated heads such
as Asp, Tense, C, Num, etc (l-morphemes and f-morphemes in the sense of Harley and Noyer
1999, 2000 and roots and abstract morphemes in Embick and Noyer 2007). Roots are
considered acatgeorial; their syntactic category is contextually specified by combining with
category-defining functional heads such as v, n, and a. For example, the root √destr is realized
as the noun destruction under the nominalizating environment ([nP n [√destr]]); it is realized as
the verb destroy the verbalizing environment ([vP v [√destr]]); it is realized as the adjective
destructive in the adjectivalizing environment ([aP a [√destr]]) (Marantz 1997, in press; cf.
Chomsky 1970, 2006). It is claimed within the Distributed Morphology that phonological
features are assigned to abstract morphemes (as well as roots under certain views) postsyntactically. The mechanism to assign this feature is Vocabulary Insertion in the technical
sense; “the Vocabulary is the list of the phonological exponents of the different abstract
morphemes of the language, paired with conditions on insertion. Each such paring of a
phonological exponent with information about the grammatical (i.e. syntactic and
morphological) context in which the exponent is inserted is called a vocabulary item.”
430
(Embick and Noyer 2007: 297). A list of potential candidates compete for the same abstract
morpheme position in the morphosynatctic derivation; which candidate is selected as the
exponent of that position is determined by the Subset Principle proposed by Halle (1997); see
Embick and Noyer (2007: 298, 299) for concise illustrations of how this principle serves to
constrain vocabulary insertion; also recall our analysis of meN-/ng- deletion in
Indonesian/Javanese as failure of phonological insertion presented in chapter 2. The
Enclyopedia, whose role is not directly relevant to the present chapter (though see section 5), is
defined as a component that “lists the special meanings of particular roots, relative to the
syntactic context of the roots, within local domains” (Marantz 1997: 204). Thus, the
encyclopedic entry for the verb kick specifies that it means ‘kill’ in the environment of “
the
bucket” (Harley and Noyer 1999: 4).
Three ingredients of this theory of the syntax-lexicon interface that will become important
in the following sections are as follows. First, the framework of Distributed Morphology
claims that all types of word formation, including those that would be treated in the presyntactic lexical component in the lexicalist theory, are conducted within the sole realm of
the syntactic computation in much the same way as sentences and phrases are. For this
reason, there is no sense in which lexical processes must precede syntactic processes because
there is no pre-syntactic generative component in the lexicalist sense in the first place; rather,
431
cases in which “lexical” processes in the lexicalist sense follow or are interleaved with
syntactic processes are predicted to be possible.
Second, this framework claims that phonological features for a morphosyntactic head are
inserted post-syntactically with reference to syntactic environments that surround the head, as
illustrated above in the destroy/destruction/destructive alternation. This second claim is a
special case of the more general claim, entailed by the thesis of the minimalist interface, that
the syntax-external phonological component is interpretive, assigning a language-particular
interpretation to the output of the syntax in the way that syntax demands. Accordingly, the
thesis does not go well with the lexicalist theory, under which a word constructed in the
lexicon is inserted into a terminal node in syntax as a triplet of sound, meaning and their
correspondence. In this sense, Distributed Morphology provides a theoretical model that
meshes nicely with the expectation of the minimalist interface defended in this thesis.
Finally, in contrast to the lexicalist theory of the lexicon-syntax interface, the theory of
Distributed Morphology claims that the syntax-lexicon interface is direct. To borrow the
phrase from Embick and Noyer (2007: 302) , “there is no syntax/morphology ‘interface’
because “words and phrases are assembled by the same generative system, and there is thus no
sense in which words must ‘interface’ with the syntax; rather they are derived by the rules of
syntax.” This view of the syntax-lexicon interface, thus, allows us to derive the effects of the
432
so-called Mirror Principle of Baker (1985), which stipulates a condition on how syntactic and
morphological structures relate to one another, as the automatic consequence of the
architecture of the theory.
In the rest of this chapter, I show that the non-lexicalist theory of Distributed Morphology
provides a natural explanation of certain facts concerning reduplication in Indonesian that pose
difficulties for several variants of the lexicalist hypothesis. This result, thus, constitutes
evidence for the interpretive nature of the syntax-external linguistic interface, as entailed by the
thesis of minimalist interfaces.
3. Asymmetries between Nominal and Verbal Reduplication in Indonesian
To find out existing patterns in nominal and verbal reduplication in Indonesian, Sato and
McDonnell (in press) conducted a corpus survey of four popular newspapers published in
Indonesian. The present corpus survey contains approximately 160, 000 words, taken from the
archives of the following four newspapers: Tempointeraktif (www.tempointeraktif.com),
Suarapembaruan (www.suarapembaruan.com), Mediaindo (www.mediaindo.co.id), and
Kompas (www.kompas.com). The result of this survey is shown in Table 5. I have included
here the results concerning derivational affixes; see Sato and McDonnell (in press) for results
that cover inflectional affixes.
433
Table 5: The Corpus Survey of Four Popular Newspapers in Indonesia (approx.160, 000 words)
Stem Reduplication
Total Tokens
No Affix
1014
Stem-Affix Reduplication
Unique Forms Total Tokens Unique Forms
312
N/A
N/A
ber-
89
37
0
0
Verbal
meN-
30
23
0
0
Affixes
di-
23
20
0
0
ter-
13
9
0
0
-an
32
22
19
15
Nominal
peN-
0
0
8
5
Affixes
peN-an
0
0
2
2
per-an
6
2
9
6
ke-an
1
1
10
8
The results given in Table 5 reveal that there is a curious asymmetry between nominal and
verbal reduplication in Indonesian, which has escaped attention in the literature on the
morphology of Indonesian. As is clear from this survey, derivational verbal affixes such as
ber-, meN-, di- and -an allow only stem reduplication. By contrast, derivational nominal
434
affixes behave differently from verbal affixes in that they potentially allow both types of
reduplication. Specifically, peN, peN-an, and ke-an have strong tendency to feed stem-only
reduplication, whereas -an and per-an allow stem-only and stem-affix reduplication. As is
true for the corpus studies in general, it is difficult to know what forms cannot be produced
in Indonesian, though the study does provides an indication that the reduplication
asymmetry is real.5 To address this concern, I have conducted grammaticality judgment
tasks with one informant to confirm whether the forms not found in the corpus study are
actually unacceptable to the native language speaker of Indonesian.
The following examples show that the corpus study in Table 5 reflects the grammatical
intuition of the actual speaker. Consider the reduplication pattern found in the verbal prefix
ber-. Table 5 above indicates that this prefix only allows stem-reduplication. This result is
confirmed by the contrast in acceptability between (10a-c) and (11a-c).
(10) Stem Reduplication with the Derivational Verbal Prefix ber
[ber [belit-belit]]
b. cakap ‘talk’
[ber [cakap-cakap]] ‘chat’
c. jalan
[ber [jalan-jalan]]
a. belit
5
‘twist’
‘walk
As pointed out by an anonymous Morphology reviewer.
‘meander’
‘stroll’
435
(11) Stem-Affix Reduplication with the Derivational Verbal Prefix ber
* [[ber-belit]-[ber-belit]]
b. cakap ‘talk’
* [[ber-cakap]-[ber-cakap]] ‘talk’
c. jalan
* [[ber-jalan]- [ber-jalan]]
a. belit
‘twist’
‘walk
‘meander’
‘stroll’
(10a-c) show that the prefix ber- allows stem reduplication. (11a-c) show that stem-affix
reduplication is unacceptable for this prefix. This contrast, therefore, shows that the results
given in Table 5 are real. A similar argument can be made for the observation made in Table 5
that derivational nominal affixes allow both stem and stem-affix reduplication. To take the
suffix -an, we have seen above that this suffix allows the two types of reduplication. That this
is correct is evidenced by the grammaticality of both examples in (12a-c) and (13a-c) below.
(12) Stem Reduplication with the Derivational Nominal Suffix -an
a. sayur ‘vegetable’
[[sayur-sayur]-an]]
‘many types of vegetables’
* [[sayur-an]-[sayur-an]]
b. buah
‘fruit’
[[buah-buah]-an]]
* [[buah-an]-[buah-an]]
‘many types of fruit’
436
c. biji
‘seed’
[[biji-biji]-an]]
‘many types of seeds’
*[[biji-an]-[biji-an]]
(13) Stem-Affix Reduplication with the Derivational Nominal Suffix -an
a. pikir
‘think’
[[pikir-an]-[pikir-an]]
‘thoughts’
*[[pikir-pikir]-an]]
b. tulis
‘write’
[[tulis-an]-[tulis-an]]
‘writings’
*[[tulis-tulis]-an]]
c. masuk ‘enter’
[[masuk-an]-[masuk-an]] ‘inputs’
*[[masuk-masuk]-an]]
(12a-c) show that the nominal suffix -an allows stem reduplication while (13a-c) show that the
same suffix can also feed stem-affix reduplication. It is important to observe that the choice
between the two forms of reduplication is not entirely free with this suffix; rather, the choice is
affected by the type of stem that it is identified with. Thus, when this suffix is combined with
nominal stems as in (12a-c), it only allows stem-reduplication. On the contrary, when this
suffix is combined with verbal stems as in (13a-c), it only allows stem-affix reduplication.
Thus, it is not the case that a single nominal affix allows both types of reduplication; it allows
437
both types when an appropriate class of elements combines with a particular nominal suffix.
We can observe this effect in the behavior of circumfixes such as peN-, peN-an, and ke-an,
whose dominant reduplication pattern is stem reduplication, as shown in the results reported in
Table 5. This point will become very important in section 5.
4. Reduplication Asymmetries in Indonesian and Lexicalist Theories
The purpose of this section is to see whether the lexicalist theory might be able to
accommodate the existing reduplication patterns in Indonesian. I show that the nominal vs.
verbal reduplication asymmetry and the existence of a word-internal reduplication pattern
that targets the non-edge of a complex stem cannot be accounted for by several versions of
the lexicalist theory as in Chomsky (1970), Anderson (1982, 1992), Kiparsky (1982a, b, c,
1985), Mohanan (1986), and Di Sciullo and Williams (1987)/Williams (2007).
4.1. Chomsky’s (1970) Weak Lexicalist Hypothesis
Chomsky (1970) proposes, based on his examination of several syntactic and semantic
contrasts between derived nominalization as in destroy destruction and genrundive
transformations (destroy destroying), that non-productive, irregular processes take place in
the pre-syntactic lexical component while productive, regular processes take place in the
438
syntactic/transformational component. This separation of two types of complex word
formation in terms of their regularity/productivity has been widely taken in the generative
literature to define the classical version of the weak lexicalist theory (though, recall our earlier
discussion of Marantz 1997). For example, Fabb (1984) considers productivity as the defining
criteria with which to distinguish lexical and syntactic word formation.
If we adopt Chomsky’s version of the lexicalist hypothesis, ber-/an- affixation as observed in
examples like (10-13) counts more as a lexical/pre-syntactic process for the following reasons.
First, the literature on the morphology of Indonesian as in McDonald (1967), Sneddon (1996),
and others points out that the verbal prefix ber- may attach to nominal, numeral, and verbal bases
that yield unpredictable/irregular semantic outcomes. First, predicates consisting of this prefix
and a nominal base refer to a customary possession of, or to characterization by the referent of
the noun, as shown in (14a, b). This type of prefixed predicate can also be used to refer to the act
of producing the reference of the noun or making use of it, as shown in (14c, d). If the nominal
base refers to a profession or way of life of an animate being, the derived predicate refers to the
property of making a living with that possession or by that way of life, as shown in (14e, f).
439
(14) ber-prefixation: Input = Noun/Output= Verb
a. anak
‘child’
→
[ber [anak]]
‘have children’
b . kaki
‘foot’
→
[ber [kaki]]
‘have feet’
c. kokok
‘cackle’
→
[ber [kokok]]
‘produce a cackle’
d. sepeda ‘bicycle’
→
[ber [sepeda]]
‘use a bicycle’
e. kuli
‘coolie’
→
[ber [kuli]]
‘work as a coolie’
f. tukang ‘artisan’
→
[ber [tukang]]
‘work as an artisan’
(MacDonald 1967: 44, 45)
Second, the prefix can combine with a numeral, unreduplicated or reduplicated, to derive
the complex noun meaning ‘forming a group of’ and ‘in groups of’, as shown in (15a-c).
(15) ber-prefixation: Input = Numeral/Output= Numeral
→
[ber [dua]]
‘two together’
a. dua
‘two’
b. ratus
‘hundred’ →
[ber [ratus]] ‘in hundreds’
→
[ber [karung]] ‘in sackfuls’
c. karung ‘sack’
(MacDonald 1967: 47)
440
Finally, the prefix may attach to verbal bases to create intransitive verbs; it works as a verbalizer
of bases that otherwise do not occur alone, as shown in (16a, b). If the root is reduplicated, an
additional meaning of variety, repetition or lack of purpose is added, as in (16c, d).
(16) ber-prefixation: Input = Verb/Output= Verb
a. -henti- ‘stop’
→
[ber [henti]]
‘come to a stop’
b. -pikir-
‘think’
→
[ber [pikir]]
‘be cogitating’
c. belit
‘twist’
→
[ber [belit-belit]]
‘meander’
d. cakap
‘talk’
→
[ber [cakap-cakap]] ‘have a chat’
(MacDonald 1967: 47,48)
The function of the derivational nominal suffix -an is no more or less complex. It has the
role of nominalization when it attaches to verbal bases. It serves as a kind of classifier,
meaning ‘types of”, as reflected in the English translations in examples in (12a-c) when it
attaches to nominal bases. These semantic considerations, therefore, suggest that the two
affixes constitute an irregular process and that the affixation involved is a lexical/pre-syntactic
process in Chomsky’s sense. In section 5, however, I show that the suffix -an is polysemous,
441
but with its dual functions being determined by different attachment sites at which it merges
with two types of complements (NumP vs. vP).
By contrast, reduplication in Indonesian is a fully productive, hence syntactic process under
Chomsky’s proposed productivity-based division of two types of word formation. Reduplication
of any countable noun produces a grammatical form that is specifically plural; recall our
observation made in chapter 4 based on Carson (2000) and Chung (2000). Thus, reduplication in
Indonesian is a productive realization of the Num head in the nominal domain. On the other hand,
it is not apparently as clear whether the corresponding argument can be made for the verbal
domain to show that verbal reduplication is really a productive process. However, the following
two considerations show that it is more like a syntactic process rather than a lexical process in the
lexicalist sense. First, the literature on the verbal reduplication in Indonesian as in MacDonald
(1976) and Sneddon (1996) notes that reduplication of a verb adds emphasis to an action denoted
by the base stem and yields outcomes related to variety, multiplicity, and atelicity. Sneddon
(1996) gives a variety of meanings as follows:
442
(17) The Semantic Effects of Verbal Reduplication
a. With some verbs reduplication gives a connotation of action done in a causal or
leisurely way .
Examples: duduk
‘sit’
berjalan ‘walk’
duduk-duduk
‘sit about’
berjalan-jalan
‘walk about, go for a stroll’
b. With many verbs reduplication indicates continued action, either an action done over a
period of time or an action performed repeatedly
Example:
Bu Yem
mengurut-urut
rambut anaknya.
Mrs Yem
stroked-Red
hair
child-her
‘Mrs.Yem stroked her child’s hair.’
c. With some verbs reduplication gives a meaning somewhat different from that of the
single form, usually conveying a sense of intensity.
Examples: menjadi
‘become’ menjadi-jadi
‘get worse’
meminta ‘request’ meminta-minta ‘beg’
443
d. Accompanied by tidak ‘not’ reduplication of the verb can indicate that the action has
not occurred, usually implying that this is contrary to expectation.
Example:
Sudah
dua hari Pak Tanto tidak
muncul-muncul.
yet
two day Mr
turn up-Red
Tanto Neg
‘Mr Tanto has not turned up for two days now.’
(Sneddon 1996: 20)
(16a-c) all belong to the type (17a) in Sneddon’s classification. This semantic effect as well as
the other three in (16b-d) indeed are in keeping with the general notion of plurality/emphasized
quantity, a crosslinguistically attested effect of reduplication, as evidenced by the extensive
investigation of the function of reduplication conducted by Moravcsik (1978). Though
Moravcsik herself concludes (p. 325) that “no explanatory or predictive generalization about the
meanings of reduplicative constructions can be proposed,” as Travis (1999, 2003) argues, her
extensive cross-linguistic investigation of the functions of reduplication across languages
suggests that reduplication has some abstract quantificational function which is diversely
instantiated as plural, distributivity, multiple iterative event readings, reciprocals, emphasis, and
so on. This argument, thus, suggests that reduplication in Indonesian can be regarded as a
syntactic process that relates to quantification. Travis (1999, 2003) also argues, while drawing a
444
clear separation between phonological and syntactic reduplication, that all types of reduplication
are underlying caused by their surrounding syntactic configurations.
The second argument to support the syntactic nature of the reduplication in Indonesian comes
from work on Madurese reduplication made by Davies (1999, 2000). Davies shows that
reduplication in this language forces the multiple event reading of a verb based on his
examination of reduplicative constructions in Madurese. There seems to be a general agreement
in the lexicalist literature, at least tacitly, that the lexicon creates complex words based solely on
lexical categories (N, V, A) but never on functional categories (T, C). This assumption is natural
because time or event reference must make crucial reference to the rules of sentence formation.
The following examples from Indonesian, modeled after the corresponding examples in
Madurese provided by Davies (2000: 127-129), show that reduplication of a verb in Indonesian
creates a variety of new interpretations unavailable to its unreduplicated counterpart, such as
multiple event readings, interleaved activity readings, and temporally displaced readings.
(18) Reduplication in Indonesian
a. Esti
Esti
meng-elus(-elus)
rambut anak-nya.
AV-stroke-Red
hair
child-her
‘Esti stroked her child’s hair many times.’
445
b. Aini dan Lina me-motong(-motong) kayu selama dua jam dan
Aini and Lina AV-cut-Red
wood for
menanam bibit
two hours and plant
seed
‘Aini and Lina cut down trees for two hours and planted seeds.’
c.
Aini dan Lina men-cubit (*-cubit) adik-nya yang lucu. Aini mem-cubit-nya hari Senin
Aini and Lina AV-pinch-Red child-their that cute Aini AV-pinch-her day Monday
Lina hari selasa.
Lina day Tuesday
‘Aini and Lina pinched their cute baby. Aini did so on Monday and Lina did so on Tuesday.’
The example in (18a) illustrates the multiple event reading whereby the telic event of stroking
a child’s hair occurred several times. If reduplication does not occur, by contrast, the sentence
is ambiguous between the single event reading and the multiple event reading, a pattern that
we have also seen to characterize nominal reduplication in chapter 4 of this dissertation. This
event-related property caused by reduplication can also be seen in the example in (18b).
Although judgments are subtle, according to my two language informants, the example in (18b)
with reduplication allows the interpretation where the event of tree-cutting is interspersed with
the event of seed-planting; for example, this sentence is true in the situation where Aini and
Lina continued the activity of tree cutting for one hour, then did seed-planting for some time,
446
and then resumed the tree-cutting activity for another hour. This interspersed activity reading is
impossible without reduplication of the verb in (18b). Similarly, the acceptability of the
example in (18c) shows that the activity of the reduplicated verb can be spaced over time. For
example, (18c) is acceptable with reduplication under the reading where Aini pinched her baby
on Monday but Lina did so on Tuesday. The acceptability of this example with reduplication is
what we predict precisely because the reduplication of a verb feeds multiple event readings.
This reading, however, is unacceptable without verbal reduplication, as shown in (18c).
What is important about these examples is that the availability of these three readings,
derived by verbal reduplication, makes crucial reference to the notion of time or event.
Again, this reference should not be possible in the lexical component to the extent that the
implicit but natural assumption holds, namely, that the lexicalist sense of lexicon does not
contain functional elements such as T and C. The readings forced by reduplication in
Indonesian as in (18a-c), thus, provide an independent argument for treating Indonesian
reduplication as a syntactic/non-lexical process.
With the two facts noted above in mind, consider now whether the examples of stemreduplication and the nominal vs. verbal reduplication asymmetry might be accounted for under
Chomsky’s theory. Examples of stem-reduplication as illustrated in (10a-c) and (12a-c) instantiate
the word-internal reduplication, namely, that an affix (either ber- or –an) is attached to the complex
447
stem created by reduplication. In other words, the affixation applies word-internally. This pattern of
reduplication poses an inverse ordering problem for Chomsky’s version. The formation of stemreduplicated forms such as belit-belit and sayur-sayur requires the syntactic process of
reduplication because reduplication is a productive process. The ber-/-an affixation applies to the
stem-reduplicated form to yield the grammatical forms such as [[ber-[belit-belit]] and [[sayursayur]-an]]. This ordering should be impossible, however, under the lexicalist architecture of the
lexicon-syntax interface because the generation of these forms requires that the syntactic process of
reduplication be followed by the lexical/pre-syntactic process of affixation. Furthermore, it seems
that Chomsky’s variant of the weak lexicalist hypothesis does not have anything to say about why
there is an asymmetry between nominal and verbal reduplication in Indonesian, as illustrated in the
examples in (10-13) and Table 5, where nouns allow both stem and stem-affixation whereas verbs
only allow stem reduplication. Based on these considerations, I conclude that Chomsky’s (1970)
version of the weak lexicalist theory have non-trivial architectural and empirical problems in face
of the existing reduplication patterns in Indonesian.
4.2. Anderson’s (1982, 1992) Weak Lexicalist Theory
Anderson (1982, 1992) develops a different version of the weak lexicalist theory from
Chomsky’s that does not depend on the somewhat vague notion of productivity. He argues that
448
inflectional morphology is treated in the syntax whereas derivational morphology is treated in
the lexicon. He defines the inflectional/ syntactic nature of any element as follows:
(19) The Definition of Inflectional Morphology in Anderson (1982, 1992)
Inflectional morphology is what is relevant to syntax.
(Anderson 1982: 587)
This definition, thus, allows any affixation that has relevance to syntax such as
agreement, tense, event structure to be treated in the syntactic component. 6 This
conception of the weak lexicalist theory is particularly problematic in face of the
Indonesian reduplication facts. The affixation of ber- counts as a lexical process because
it does not seem to have syntactic effects such as agreement, tense, and event structure.
However, we have seen in section 3 that reduplication in Indonesian has clearly syntactic
outcomes in the form of multiple event readings and discontinuous time-interval reading.
6
As Di Sciullo and Williams (1987) point out, however, this definition does not work in the way Anderson
wants it. If the inflectional nature of a particular affix is determined by its relevance to syntax, then almost all
affixes could be considered as syntactic, a state of affairs that Anderson does not want. For example, they
note (p. 69) that “this definition would seen to consign all nominalizing affixes, such as -ion, to inflection
because nouns and verbs have different syntactic properties, and the affix makes the difference.” Similarly,
ber-suffixation should be treated as syntactic, contrary to what Anderson’s weak lexicalist theory would
actually intend, because it creates a related verb from the verbal stem, and the notion of verb is relevant to
syntax. The problem is more general. For the purposes of this paper, I maintain the pretense that Anderson’s
theory would treat ber-affixation as lexical in the intended sense.
449
This means that reduplication is an inflectional process to be treated in the syntax. Then,
the word-internal reduplication pattern illustrated in (10a-c) and (12a-c) should be
ungrammatical because the generation of such a pattern requires the application of the
syntactic rule to be followed by the application of the lexical rule. Anderson’s version of
the weak lexicalist theory also has little to say about why the reduplication is asymmetric
between nominal and verbal suffixes in Indonesian.
4.2. Kiparsky’s (1982a, b, c, 1985)/Mohanan’s (1986) Lexical Phonology
The same reduplication asymmetry and the word-internal reduplication pattern also refute
one well-known version of the strong lexicalist theory known as Lexical Phonology
(Kiparsky 1982a, b, c, 1985; Mohanan 1986; see also Pesetsky 1979).
This theory
maintains that morphology and phonology interact in tandem, with each stratum/cycle
governing operations with certain characteristics. Specifically, affixational/inflectional
processes with irregular phonological and morphological consequences occur in Stratum 1
while regular inflectional processes with transparent consequences occur in a later Stratum
(Stratum 3 in Kiparsky/Stratum 4 in Mohanan). Kiparsky’s (1982a) model of the Lexical
Phonology is given in (20). See also Mohanan (1986) for a further development of
Kiparsky’s original model, which I am not going to discuss here.
450
(20) Kiparsky’s (1982a) Model of Lexical Phonology in English (Kiparsky 1982a: 133)
underived lexical entries
“+boundary” inflection and derivation
stress, shortening
Level 1
“#-boundary” derivation and compounding
compounding stress
Level 2
“#-boundary” inflection
syntax
laxing
Level 3
postlexical phonology
This model assumes that the word formation rules and the lexical phonological rules are
partitioned into an ordered series of levels/strata/cycles. “+boundary” inflectional affixes in Level
1 include the umlaut of tooth-teeth, the ablaut of sing-sang and other stem-changing morphology
whereas “+boundary” derivational affixes includes what have been called Level 1 affixes in the
Level-Ordering Hypothesis of Siegel (1973) and Allen (1978) such as -al, -ous, and -im, as in
refusal, pious, and impotent. “#-boundary” derivation in Level 2 involves what have been called
451
Level 2 affixes in the Level-Ordering Hypothesis such as un-, -ness, and –er; also compounding, a
process of combining two independent root elements such as black board, nurse shoes, and red
coat. Finally, “#-boundary” inflection in Level 3 deals with the affixation involving the rest of the
regular inflectional affixes such as plural -s, and past tense -ed in English. To illustrate, consider
the derivation of codifiers. The base stem code is submitted to the phonological rules of Level 1,
where the word formation rule attaches the Level 1 affix -ify to the stem. This derived stem is then
assigned stress as códify in the same Level. The resulting object is submitted now to the
phonological component of Level 2, in which the word formation rule attaches the agentive suffix
-er to derive the complex form codifier. Finally, when the resulting object enters Level 3, the
regular plural formation process applies to this object to yield the final output codifiers. In this way,
a set of phonological and morphological processes that apply to complex word formation is
ordered in a series of strata/cycles. Kiparsky assumes that the derivation of all words should go
through all these three Levels, even though relevant phonological and morphological processes
might apply vacuously to a given form at any of these three levels.
This ordered block of rule application correctly predicts why forms such as *un[ept]
is ill-formed in contrast to in[ept]. As we have seen above, the prefix un- is a Level 2
affix. Thus, un- prefixation occurs in Level 2. To create the form *unept, however, the
bound morpheme -ept- must traverse the word formation process in Level 1 that would
452
assign the appropriate Level 1 affix in- to the stem to create the grammatical form inept.
The ill-formedness of *unept, thus follows, as a natural consequence of the ordered
series of morphophonological rule application.
One theoretical tenet of Lexical Phonology which is important for the purposes of this
chapter lies in the Bracketing Erasure Convention (Pesetsky 1979). This convention deletes all
brackets at the end of each stratum/level of word formation, thus has the effect of rendering
access to the previously available internal structure of complex words opaque in later
strata/cycles. This convention, thus, derives a lexicon-internal version of the Lexical Integrity
Hypothesis defined in section 2, namely, that word formation processes in Level 2 and 3
cannot look into the morphological makeup of complex morphological objects created by
word formation processes in Level 1 and Level 2, respectively. Lexical Phonology, therefore,
makes an explicit prediction that no processes in a particular level should be able to apply to a
complex object that is derived by word formation processes characteristic of earlier levels. This
prediction is clearly falsified by the reduplication pattern attested in Indonesian. We have seen
in section 2.1 that reduplication is a fully productive process. Under Kiparsky’s model, this
process should be located in Level 3 on a par with the regular inflectional affixes such as plural
-s, and past tense -ed: recall that any countable noun and verb can be input for reduplication
just as any countable noun and verbs can be affixed by -s and -ed in English, respectively
453
(except irregular nouns and verbs, of course). We have also seen in the same section that
affixes such as ber- and -an yield a set of semantic irregularities when attached to a stem. This
unpredictable behavior leaves affixation of these pieces in Level 1 on a par with the irregular
umlaut and ablaut rules as in brother-brethren. Now, to derive the word-internal reduplication
pattern as illustrated in (10a-c) and (12a-c) under Kiparsky’s model, the Level 1 affixation
(ber-affixation and –an suffixation) must be preceded by the Level 2 inflectional process
(reduplication), an ordering that should be impossible in Lexical Phonology due to its central
hypothesis that each level/stratum is strictly ordered, hence cannot be traversed. To illustrate
the issue at hand with ber-belit-belit, the base belit is submitted to Level 1, at which the berprefixation would apply to yield [ber-belit]. This complex object is submitted to Level 3, at
which reduplication applies to the whole object to create the output [[ber-belit]-[ber-belit]].
Importantly, this output is ill-formed in Indonesian even though this is the only output that is
predicted to be possible under the strict layering of levels in Lexical Phonology.
This type of word-internal reduplication pattern is also problematic for Lexical Phonology
in three other respects. First, due to the Bracketing Erasure Convention, Kiparsky’s model
makes a prediction that reduplication should must target the right or left edge of the whole
complex object because at the time this process applies in Level 3, the input transferred from
Level 2 enters the Level 3 as an atomic unanalyzable element as the result of the erasure of all
454
word-internal constituent boundaries. Thus, the existence of forms such as [ber-[belit-belit]]
shows that reduplication targets part of the complex stem rather than the left or right edge of it.
Second, Kiparsky assumes that the output of each level is itself a full-fledged lexical item.
However, the ill-formedness of forms such as *belit-belit shows that this is not always the case.
Finally, as in Chomsky’s (1970) weak lexicalist theory, Kiparsky’s theory also does not seem
to provide us with any way of explaining why the asymmetry between nominal and verbal
reduplication observed in section 2 obtains in Indonesian.
4.3. Di Sciullo and Williams’ (1987)/Williams’ (2007) Strong Lexicalist Theory
Di Sciullo and Williams (1987) develop the most comprehensive defense of the strong lexicalist
theory, which is conceptually very different from other instantiations of the theory as in Kiparsky
(1982a, b, c, 1985) and Mohanan (1985). They maintain that morphology and syntax are two
different domains of inquiry with two different primes (e.g., stems, affixes, roots vs. NPs, VPs,
CPs) and operations (compounding, θ-identification vs. movement, quantification). Thus, for Di
Sciullo and Williams, the so-called lexicalist hypothesis/the lexical integrity hypothesis/the
lexical atomicity “is not a principle of grammar but rather a consequence of the conception that
grammar contains two subparts, with different atoms and different rules of formation” (p.2). The
word “lexicon” takes an entirely different sense in their framework from the most common
455
usage as the generative system that stores words and their formation rules: the lexicon is the
storage house for listemes, “objects of no single specifiable type (words, VPs, morphemes,
perhaps intonation patterns, and so on) that “fail to conform to interesting generalizations.” (p. 3)
Assuming this strict division of labor between the word system and the phrase system,
Di Sciullo and Williams maintain that the morphology and syntax can still communicate
with one another through a restricted range of shared vocabulary, specifically, the
“topmost properties of words, the features and argument structure of the topmost words.”
(p. 45), as stated in their proposed Thesis of Atomicity of Words defined as in (3). Williams
(2007), the most recent update of Di Sciullo and Williams’s lexicalist hypothesis,
maintains essentially the same position.
Di Sciullo and Williams illustrate this limited cross-modular communication with
compounding in English. Compounding involves the creation of what they call morphological
objects that derive their agreement features from the percolation of the features of the right-hand
head (Williams 1981). Crucially, it is this output agreement recorded on the top-most level of the
compound (namely, the topmost N in (21a, b)) that is used for the purposes of syntactic subjectverb agreement, as the contrast between (22a, b) shows.
456
(21) English N + V compounds
a.
N [sg]
N [sg]
parts
b.
N [pl]
N [sg]
supplier
(Parts-supplier is/*are…)
N[sg ]
part
N[pl]
suppliers
(Part-suppliers *is/are…)
(22) a.
Parts-supplier is/*are mean to me.
b.
Part-suppliers *is/are mean to me.
This agreement pattern correctly falls out from Di Sciullo and William’s system
because the feature specification for the non-head member of the compound is invisible
from the perspective of syntax. Thus, this pattern is one way in which the syntax and
morphology can communicate in a restricted range of share vocabulary though the
Thesis of Atomicity of Words still blocks the syntax from accessing the internal
morphological composition of compounds.
At this point of the present research, it is not clear whether any aspect of Indonesian
reduplications facts could prove Di Sciullo and William’s (1987) version of the lexicalist
theory untenable. For Di Sciullo and William’s lexicalist theory and other variants of the
457
hyper-lexicalist approach as in HPSG, it is extremely difficult, if not impossible, to find
certain morphological patterns that could tease apart the predictions of their theory and
some other theories; see Haugen and Harley (2006), though, who provide a detailed
discussion of how certain compound-internal reduplication patterns in Yaqui pose a
problem for their lexicalist theory; see also Harley to appear for relevant discussion). I
suspect, however, that Di Sciullo and William’s hypothesis is not incorrect; when carried
to its extreme, it boils down to the syntactic approach to reduplication to be proposed in the
next section in a different module of grammar (that they call the Word System).
4.4. The Lexicon as the Source of the Ordering Paradox
To summarize I have shown that reduplication within lexically/pre-syntactically derived
complex stems in Indonesian poses non-trivial empirical and architectural problems for a
number of well-known versions of the weak and strong lexicalist theory as presented in
Chomsky (1970), Anderson (1982, 1992), Kiparsky (1982a, b, c, 1985), Mohanan (1986),
and Di Sciullo and Williams (1987)/Williams (2007). I have also shown that those lexical
approaches would have little to say about how the asymmetry between nominal and verbal
reduplication arises in this language. Thus, those facts on Indonesian reduplication provide
strong arguments against certain versions of the weak/strong lexicalist theory.
458
It is important to point out that this type of inverse ordering is a problem only when we
postulate the lexicon/morphology as the pre-syntactic generative component that is responsible
for certain types of word formation characterized by productivity, semantic/phonological
compositionality, the relevance of morphological primes to the syntax, and so on, as assumed in
Chomsky (1970), Anderson (1982, 1992), and Kiparsky (1982a, b, c), and Mohanan (1986). In
other words, this problem does not (or cannot) arise in non-lexicalist theories of the lexiconsyntax interface that do not posit such an independent component prior in addition to the
generative system of syntax. In light of this consideration, in the next section, I pursue an
alternative, non-lexicalist analysis of the reduplication in Indonesian within the more recent
morphosyntactic framework of Distributed Morphology as reviewed in section 2.2.
5. A Distributed Morphology Approach to Reduplication Asymmetries in Indonesian
In this section, I show that the asymmetry between nominal and verbal reduplication and
the word-internal reduplication pattern receive a straightforward account within the theory of
Distributed Morphology (Halle and Marantz 1993, 1994; Harley and Noyer 1999; Embick
and Noyer 2007). Specifically, I propose that these facts are explained as a natural
consequence of a particular hierarchical arrangement of morphosyntactic features such as
Aspect and Number in Indonesian. I assume, in line with much recent work on reduplication
459
in a number of different theoretical frameworks, that this process consists in affixation of the
reduplicative null morpheme RED (UPLICATION) that triggers copying on a stem on its
local environment; see Marantz (1982), McCarthy and Prince (1986, 1993, 1995); see also
Travis (1999) and Haugen (2004) for a syntactic approach to reduplication. I argue that this
particular analysis has an important bearing on the validity of the minimalist interface that
has been defended and substantiated in the preceding chapters.
5.1. Verbal Reduplication
Consider first verbal reduplication. As we have seen in section 2, derivational verbal affixes
can only allow stem reduplication. This pattern is naturally explained if verbal reduplication
is mediated by the Inner Aspect head (Travis 1999) that dominates the reduplicative null
morpheme and if ber—prefixation represents the addition of the higher v projection to the
Inner Aspect structure. This assumption is supported by the fact that, as noted in section 2,
verbal reduplication has various effects related to the Aktionsart of the predicate. Under this
set of assumptions, then the morphosyntactic derivation for the example in (10a), [ber-[belitbelit]], will be as in (23).
460
(23) The Morphosyntactic Derivation of the Stem-Reduplication in (10a)
Morphosyntax
Phonology
vP
v
ber-
AspP
Asp
√
RED
belit
[ber-[[belit]-[belit]]]
[[belit]-[belit]]
In this derivation, the Asp head merges with the acatgeorial root belit ‘twist’. The object that
results from this merger is phonologically realized as the reduplicative form, [[belit]-[belit]],
because the only stem that the RED morpheme in the Asp head triggers copying of is the
root belit on its local c-commanding environment. The Asp head undergoes further merger
with the verbalizing prefix ber-. The complex morphosyntactic object, then, is interpreted at
the syntax-external phonological component as [ber-[[belit]-[belit]]], as desired.
It is important to note that the reduplicative morpheme intervenes between the v head
and the root in this derivation. Accordingly, the RED morpheme cannot reach up to the
position of the v head to include the verbalizing prefix in its domain for reduplication to
yield the ungrammatical form as in [[ber-belit]-[ber-belit]]. This derivation, thus, correctly
predicts the unavailability of the stem-affixation reduplication pattern for derivational
461
verbal affixes such as ber-. In this way, the fact that verb stems only allow stem
reduplication naturally falls into place by assuming a particular hierarchical arrangement of
morphosyntactic features/heads.
It is also to be stressed here that the state of affairs observed above in which the functional
heads are linearized in the direction predicted by the hierarchical alignment of
morphosyntactic features is exactly what is expected under the theory of Distributed
Morphology. As we have seen, the word formation of all kinds is conducted by the single
generative procedure as the sentence formation of any kind is in this non-lexicalist theory.
Accordingly, the verbal reduplication pattern in Indonesian is simply the direct consequence
of the grammatical architecture of the Distributed Morphology. On the contrary, under nonlexicalist views of the syntax-lexicon interface, there is no reason to expect that the syntactic
structure and the morphological structure match in this manner, as the interface between the
lexicon and syntax is indirect for reasons mentioned in section 2.1. Thus, the reduplication
for verb stems in Indonesian can be construed as one good testing ground to tease apart the
predictions of the two competing theories.
The proposed analysis of verbal reduplication in Indonesian also supports the locality of
post-syntactic phonological feature assignment at the syntax-external interface. The proposed
analysis crucially rests on the idea that the post-syntactic late insertion of phonological material
462
at the interface closely mirrors the way the syntactic derivation proceeds; ber- cannot be
included as part of input for verbal reduplication because it is merged in a structurally higher
position than the object (AspP) that becomes the target for reduplication. The root must be
included for reduplication because it is in the c-commanding domain of the RED morpheme.
By contrast, the stem-affix reduplication pattern as in [[ber-belit]-[ber-belit]] is simply
underivable under the syntax-driven interpretive nature of the phonological component. The
proposed analysis, thus, can be considered another particular instantiation of the theoretical
position entailed by the thesis of minimalist interface that the syntax-external sound
component is fundamentally interpretive, doing the best it can to satisfy the structural
requirements imposed on it by the syntactic derivation. This point also holds true for the
analysis of nominal reduplication presented in the next subsection.
5.2. Nominal Reduplication
Let us now turn to nominal reduplication. We have seen in section 2 that derivational
nominal suffixes allow both stem and stem-affix reduplication. At the same time, we have
also noted that the choice between the two types of reduplication is not entirely free but
rather is governed by the syntactic category of the input stem. This latter point is crucial for
the account presented below.
463
Consider again (12a-c) and (13a-c), repeated as (24a-c) and (25a-c), respectively.
(24) Stem Reduplication with the Derivational Nominal Suffix –an
a. sayur ‘vegetable’
[[sayur-sayur]-an]]
‘many types of vegetables’
* [[sayur-an]-[sayur-an]]
b. buah
‘fruit’
[[buah-buah]-an]]
‘many types of fruit’
* [[buah-an]-[buah-an]]
c. biji
‘seed’
[[biji-biji]-an]]
‘many types of seeds’
*[[biji-an]-[biji-an]]
(25) Stem-Affix Reduplication with the Derivational Nominal Suffix –an
a. pikir
‘think’
[[pikir-an]-[pikir-an]]
‘thoughts’
*[[pikir-pikir]-an]]
b. tulis
‘write’
[[tulis-an]-[tulis-an]]
‘writings’
*[[tulis-tulis]-an]]
c. masuk ‘enter’
[[masuk-an]-[masuk-an]] ‘inputs’
*[[masuk-masuk]-an]]
464
The input nominals in (24a-c) that allow only stem reduplication are all simplex nominals
(i.e. sayur ‘vegetable’, buah ‘fruit’, and biji ‘seed’) whereas the input nominals in (25a-c)
that allow only stem-affix reduplication are all complex deverbal nominals (i.e. pikir
‘think’ pikir-an ‘thought’, tulis ‘write’ tulis-an ‘writing’, and masuk ‘enter’ masuk-an ‘input’). This difference, I claim, holds a key to a full understanding of why
nominal derivational affixes in principle allow two types of reduplication unlike their
verbal counterparts.
Let us assume that nominal reduplication consists in the copying of a nominal stem by
the reduplicative null morpheme located in the Num head. The Num head selects a
nominal stem as its complement, a rather natural assumption provided that reduplication
of a nominal element yields the form that is specifically plural in Indonesian, as we have
seen in section 3 and chapter 4 of this dissertation. This assumption also meshes nicely
with the analysis of the denotation of bare nominals presented in chapter 4, under which
bare uninflected nominals in Indonesian project up to the Num P that selects an NP as its
complement. Alternatively, we might assume that a nominal stem is selected by the
Q(uantity) Phrase in light of the fact that one function of the nominal reduplication is to
yield an emphasized quantity reading for the reference of the noun stem. For the
465
purposes of this chapter, however, I assume, following the analysis in chapter 4, that the
RED morpheme is hosted by the Num head.
Under this analysis, then, simplex nominal stems as in (24a-c) can directly merge with the
Num head. Verbal stems as in (25a-c), by contrast, cannot merge with the Num head this way
because this head only selects a nominal stem as its complement, as stated in the previous
paragraph. Thus, they are nominalized by the suffix -an before they can merge with the Num
head. The morphosyntactic derivations for the examples in (24a) and (25a), then, will be as in
(26) and (27), respectively. I designate the head that host the suffix -an in (26) as the F,
returning to its precise nature shortly.
(26) The Morphosyntactic Derivation of the Stem-Reduplication in (24a)
Morphosyntax
Phonology
FP
F
-an
NumP
Num
RED
nP
n
Ø
[[sayur]-[sayur]-an]]
[[sayur]-[sayur]]
[sayur]
√
sayur
466
(27) The Morphosyntactic Derivation of the Stem-Reduplication in (25a)
Morphosyntax
Phonology
NumP
Num
RED
nP
n
-an
vP
[[[pikir]-an]]-[[pikir]]-an]]
[[pikir]-an]
[pikir]
v
√
Ø
pikir
In the derivation in (26), the root sayur ‘vegetable’ is instantiated as a noun by adjoining to
the null nominalizing head. This stem, being a nominal, can directly merge with the Num
head as input for reduplication, which creates the form [[pikir]-an] at the syntax-external
sound component. The derivation further continues by merging the nP with the Num heat
that hosts the RED morpheme. Since the RED morpheme can have access to the nP in its
local c-commanding domain, the output of the nP must be included for reduplication at the
Num head level. This information is realized at the post-syntactic phonological component
as [[[pikir]-an]]-[[pikir]]-an]]. Note that the form [[pikir]-[pikir]-an]] cannot be created at the
interface because the syntactic derivation here dictates that late insertion of phonological
material at the interface must mirror the history of syntactic derivation in a cyclic manner, as
467
expected under the thesis of minimalist interface. The derivation in (27) is crucially different
from that in (26), in that the base stem is verbal. Accordingly, the stem must undergo zeroderivation into nominal stems by the suffixation of the nominalizing suffix -an to serve as
the complement that can satisfy the categorial restriction imposed by the Num head. Since
the RED morpheme contained in this head includes the nominalizing suffix as well as the
base stem in its local c-commanding domain, the syntactic derivation dictates that the
phonological component include both elements as input for reduplication, thereby closely
following the path curved by syntactic derivation in a local manner. Under this derivation,
then, the stem reduplication pattern as in the hypothetical *[[[pikir]-[pikir]]-an] is simply
underivable due to the way syntactic derivation proceeds and the way a particular set of
morphosyntactic features is organized as shown in the derivation in (27). In this way, the
proposed non-lexicalist, morphosyntactic analysis provides a straightforward explanation for
the fact that the choice between the stem and stem-affix reduplication correlates with the
underlying category of the input stem (verbal vs. nominal).
There is one important question that remains under the proposed account of nominal
reduplication, which concerns the status of the suffix -an.7 As is clear from the position of this
suffix in the two morphosyntactic derivations in (26) and (27), the proposed analysis entails
7
Thanks to Heidi Harley (personal communication) and an audience member at the 33rd Meeting of the
Berkeley Linguistic Society for asking this question.
468
that this single suffix has two separate functions in the two cases. It is clear that the function of
-an is that of nominalization in the stem-affix reduplication. The question is what role the
suffix plays in the stem reduplication. It is important to recall in this connection that, in
examples such as (10a-c), -an yields the reading that can be roughly “many types of”. Thus,
buah-buah-an, derived from buah ‘fruit’, means ‘many types of fruits’. This observation is
related to the remark made in MacDonald Dardjowidjojo (1967: 66), who observes that “when
added to a noun root (usually reduplicated) the suffix -an forms a noun which refers to a
collection of the referents of the simple noun, or of related referents,” as shown in (28a, b).
(28) The Function of -an in Stem Reduplication
a. kaleng
‘tin can’
[[kaleng]-[kaleng]-an]
‘canned goods’
b. pohon
‘tree’
[[pohon]-[pohon]-an]
‘trees, the vegetable kingdom’
(MacDonald Dardjowidjojo 1967: 66)
Interestingly, however, the same suffix cannot be attached to stems like jeruk ‘lemon’ to
derive jeruk-jeruk-an; according to my native language consultants, the form itself is not
ungrammatical but means something entirely different, something like “fake orange.” This
is presumably related to the kind-denoting nature of bare nominals in Indonesian, as
469
argued extensively in chapter 4.8 Since the name of a kind refers to an amorphous, masslike property of that kind under Carlson’s (1977)/Chierchia’s (1998a, b) sense of the term,
it is natural that -an attached to a bare nominal yields readings such as “many types of X”.
However, this line of thinking still leaves unexplained why bare nominals such as
jeruk ‘lemon’ do not yield “many types of lemons”-readings when the stem is
reduplicated and suffixed by -an. I maintain that the suffixability of -an is determined by
the world-knowledge of what constitutes the most natural kind in classificatory terms in
the mind of Indonesian speakers. The same language informants point out that the noun
buah admits -an when reduplicated because it refers to what they would take to be a
natural classificatory term; on the other hand, the noun kejur ‘lemon’ does not in their
opinion because it is a specific instance of the classificatory term buah and does not
presumably serve a natural classificatory function in the light of their encyclopedia
knowledge. This observation is highly reminiscent of Chierchia’s (1998a) discussion of
the nature of a kind. For example, he notes (p. 348) makes the following remark, citing
Carlson (1977: 26ff) and Krifika et al. (1995).
8
Thanks to an anonymous Morphology reviewer for suggesting this possibility.
470
(29) Chierchia’s (1998a) discussion on the nature of a kind (Chierchia 1998a: 348)
By ‘natural’ kinds, we do not necessarily mean, in the present context, just biological
ones or even ‘well-established’ ones. Artifacts (like chairs or cars) or complex things
(like intelligent students or spots of ink) can qualify as kinds, to the extent that we can
impute to them a sufficiently regular behavior…What counts as kind is not set by
grammar, but by the shared knowledge of a community of speakers.
Note that this line of thought provides somewhat interesting support for the Distributed
Morphology view of roots as assigned idiomatic meanings in a language-particular manner
in the post-syntactic Enclopedic component (see Harley 2008 for much relevant discussion
on the nature of roots in Distributed Morphology from a case study in English compounds).
The decision on whether a particular root such as sayur ‘vegetable’ constitutes a natural kind
or not depends on the speaker’s perception of whether the root makes a natural classification
in his/her mind. The Distributed Morphology model predicts precisely this state of affair
because the relevant decision is a matter of the encyclopedic component that deals with
linguistically unpredictable meanings for expressions such as cats, kick the bucket, and so on.
Therefore, I conclude that a) -an is an exponent of what native speakers of Indonesian
take it to a natural kind in a classicatory sense and that b) it is generated in the head of
471
whatever projection that selects the phrase that denotes a kind. This conclusion is also
independently supported by the morphosyntax of bare nominals in Indonesian discussed
in chapter 4, where I have provided independent evidence that whatever value the Num
head takes (specifically, be it plural or neutral), the denotation of the NumP is computed
at the interface as a kind in languages such as Indonesian, Javanese, and Japanese. Thus,
these considerations thus further support my conclusion that -an occupies the head that
selects the NumP as its complement as in the derivation in (26).
6. Conclusions
Let us now take stock of what I have done so far in this chapter. I have started this
chapter by introducing the results of my corpus study of four popular newspapers
published in Indonesian. This study has revealed that a) nominal derivational affixes
such as -an in principle allow both stem and stem-affixation reduplication whereas
verbal derivational affixes such as ber- allow only stem reduplication and that b) both
nominal
and
verbal
stems
may allow
reduplication
to
target
part
of
a
morphologically/lexically derived complex word rather than its left or right edge. I have
also shown that these results of the corpus study are indeed verified by native speakers’
intuition by conducting grammaticality judgment tasks.
472
Then, I have demonstrated that these two facts concerning Indonesian reduplication
pose non-trivial architectural and empirical challenges for a number of well-known
versions of the weak and strong lexicalist theory as in Chomsky (1970), Anderson (1982,
1992), Kiparsky (1982a, b, c, 1985)/Mohanan (1986), and Di Sciullo and Williams
(1987)/Williams (2007). I have also emphasized that the inverse ordering paradox caused
by the word-internal reduplication pattern only arises in a theory of the lexicon-syntax
interface that postulates the generative lexicon as an autonomous pre-syntactic component.
Accordingly, the inverse ordering problem ceases to be a problem under non-lexicalist
theories of the interface because we do not have any pre-syntactic word-formation
component prior to the syntactic component in the first place.
Based on this theoretical consideration, I have argued that the two facts about
Indonesian reduplication noted above receive a straightforward explanation within the
more recent, non-lexicalist, morphosyntactic theory of Distributed Morphology outlined in
Halle and Marantz (1993, 1994), Harley and Noyer (1999), and Embick and Noyer (2007)
once we take seriously a particular hierarchical arrangement of certain morphosyntactic
features/heads such as Asp and Num as well as the underlying syntactic category of input
stems for reduplication. The key assumption of the proposed analysis is that the postsyntactic phonological feature assignment closely mirrors the bottom-up derivation of
473
morphosyntactic structures; the phonological component requires the reduplicative
morpheme to target only the constituent within its c-commanding domain and the
assignment of phonological feature applies in a bottom-up fashion, much in the way that
syntax works. According to this analysis, the stem-affix reduplication in cases such as
*[[[sayur]-an]-[[sayur]-an]]] and the stem reduplication in cases such as *[[pikir]-[[pikir]an]]] are simply underivable.
The overall result in this chapter, therefore, provides a strong piece of evidence against the
traditional lexicalist architecture of the syntax-lexicon interface, and, at the same time, argues
in favor of non-lexicalist theories as in the recent Distributed Morphology framework that
attempt to locate all types of word formation within the sole realm of the syntactic derivation.
In the next section, I show that a similar argument can be made against the lexicalist model of
the syntax-lexicon interface from the phenomenon of what I term “phrasal inclusion within
lexical words” in Indonesian and many other languages.
In the rest of this chapter, I briefly discuss further implications of the results achieved in
this chapter for the proper theory of the syntax-lexicon interface and, most importantly, for
the thesis of minimalist interface.
474
6.1. Implications for the Proper Theory of the Lexicon-Syntax Interface
Let me first note that minimalist considerations of parsimony, elegance, and simplicity,
known as the methodological minimalism, “the drive for simple and nonredundant
theories of the world” or “seeking the best way to theorize about a particular domain of
inquiry” (Martin and Uriagereka 2001: 1), favor the non-lexicalist hypothesis in which
the single generative syntactic component is responsible for all types of word formation
rather than the lexicalist hypothesis in which a separate pre-syntactic component is
postulated for certain types of word formation in addition to the syntactic component.
That is, the non-lexicalist position is the null hypothesis in light of this formal
consideration. In terms of the empirical question of which theory is right, as Embick and
Noyer (1999: 291, 321, 322) note, the lexicalist claim that the generative lexical
component be posited in addition to the syntax can only be motivated to the extent that a
certain important generalization cannot be stated in a syntactic approach to word
formation but instead must be treated in a lexicalist approach to word formation; the
simple observation that a lexicalist approach works is simply irrelevant to the debate
between lexicalism and non-lexicalism. Without this type of demonstration, the
methodological parsimony noted above always tells us that the single-engine hypothesis
of the non-lexicalist theory is the null hypothesis. A similar argument holds for the other
475
part of the lexicalist claim that ‘words” are different from “phrases”. Again, the burden is
on lexicalist theorists to show that certain patterns of “words” cannot be accommodated
within the syntactic approach to word formation and that rules of word formation must
be treated differently from rules of phrase formation.
Importantly, the results achieved in this chapter provide a new type of empirical
evidence against the lexicalist theory from another perspective; the postulation of a presyntactic level such as lexicon actually would lead us to miss certain important
generalizations that could be most naturally statable within the morphosyntactic structure.
In this sense, the patterns of reduplication in Indonesian provide a unique empirical
demonstration that the general architecture of the lexicalist theory is untenable inasmuch as
it postulates a pre-syntactic lexical component prior to the syntactic component. In other
words, the proper theory of the lexicon-syntax interface must be one of the variants of the
grammatical models such as Distributed Morphology that posits a single level, namely, the
syntactic component, for all types of word formation that has been independently shown to
be a generative system in the framework of generative grammar. This point is to be kept in
mind in the next chapter as well, where I provide another empirical argument of the same
type against the lexicalist view of the lexicon-syntax interaction based on the
phenomenon of “phrasal inclusion within lexically derived words” in Indonesian.
476
6.2. Minimalist Interfaces
The conclusion reached at the end of the previous subsection is quite in line with the idea of
minimalist interface that the syntax-external components are purely interpretive, doing the
best they can to assign interpretation to the output of the generative syntactic system in the
manner required by the system. This chapter has shown that the thesis of minimalist
interface is substantiated in three domains. First, we have seen that the late insertion model
as in Distributed Morphology is what the minimalist would lead us to adopt, given that the
phonological component is purely interpretive. Second, we have seen that the way that late
phonological feature assignment works in the syntax-external interpretive component
crucially depends on the structural notions such as locality and c-command that have been
shown to be important theoretical ingredients in the generative syntactic research. Finally,
the proposed non-lexicalist analysis within the Distributed Morphology framework leads to
the theoretical consequence that there is no such thing as the syntax-lexicon interface, as also
pointed out by Embick and Noyer (2007), because words are constructed in the same way as
sentences and phrases are by the same set of rules and principles. This is an optimal
consequence of the thesis of the minimalist interface, which views the role of syntax-external
linguistic interface as autonomous systems that interpret whatever objects are sent out by the
sole generative engine of syntax and add certain modifications to make them usable for the
477
language-external conceptual and articulatory systems within the range of options set up by
syntax. In this way, close examination of the existing reduplication patterns in Indonesian
conducted in this chapter provides further substantiation for the thesis of minimalist interface
defined as above. I discuss further ramifications of this thesis in the next chapter.
478
CHAPTER 7 MINIMALIST INTERFACES
In this final chapter, I summarize the high points of the contents of my investigation in
the previous chapters and identify potentially significant conclusions to be drawn from
this dissertation on the interface between syntax and its neighboring phonological and
semantic interfaces.
7.1. Summary of the Dissertation
The leading idea in this whole dissertation has been that of minimalist interfaces, namely, that
the syntax-external linguistic components conduct a handful of modular-specific operations to
arrange complex objects created by the universal law of syntax to be usable for languageindependent articulatory and conceptual systems. All the apparently disparate phenomena
drawn from the sizable portion of the grammars of Indonesian and Javanese discussed in detail
in the preceding chapters have been shown to provide support for this notion of interface
strategies. This idea has been instantiated in a variety of ways. The active voice deletion caused
by the movement of an NP (chapter 2) is one clear case where the syntax-external
phonological component closely mirrors the local computation of syntax as required by the
Phase Theory. The denotation and morphosyntax of bare nominals in Indonesian and Javanese
479
as well as several other languages (chapter 4) led to a relativized parametric theory of nominals
that proposes that there is only one semantic interpretation that can be conducted at the
semantic interface; what differs in the height of functional structures above the lexical
projection and the possible set of values that the Num feature can take in each language. The
cross-classification of these two parameters yields different outcomes at the semantic interface.
This result, therefore, shows that what the interface can do is to accept whatever objects syntax
sends off and give an interpretation to the objects in a way that is compatible with our
encyclopedic/world knowledge of how nouns are conceptualized. Investigation of nominal vs.
verbal reduplication (chapter 6) also suggests that the phonological component is
fundamentally interpretive, doing whatever operations it can to mirror the way syntactic
derivation proceeds, but within the range of options set up by syntax, such as c-command,
locality, cyclic assignment of phonological features and so on. P-stranding under sluicing in
Indonesian (chapter 3) constitutes one clear case where the role of the syntax-external
phonological component is clearly at work. I have argued that certain violations created by
syntax such as percolation failure can be repaired at the interface by deletion. At the same time,
I have also shown that other violations such as the failure of D-to-P incorporation in syntax
cannot be repaired at the interface. This analysis, if correct, is a natural outcome of the
proposed thesis of minimalist interfaces: interfaces can conduct operations to legitimatize
480
otherwise illegitimate syntactic objects but only within the range of options admitted by
language-particular parameters. Similarly, the semantic component behaves as an autonomous
component that applies choice functions to wh-in-situ questions in Indonesian (chapter 5).
Particular details of the choice function analysis, I showed, provides further confirmation for
the minimalist interface thesis defined above in that the semantic component does conduct the
domain-specific operation of choice function but crucially because syntax is not fine-tuned to
create a multitude of interpretations that would be required by the communicative demands
from the language-independent conceptual system.
7.2. Minimalist Interfaces
The thesis of minimalist interface leads to a particular understanding of the way syntax
interfaces with its neighboring linguistic components, semantics and phonology. Couched
within the recent derivational theory of syntax and its correspondence with phonology and
semantics (see Chomsky 1995, 2000, 2001, 2004, 2005, Epstein et al. 1998, Uriagereka 1999,
and Grohmann 2003), the thesis of minimalist interface defined above yields the architecture
of the interface between syntax, syntax-external interfaces, and language-independent AP and
CI systems, as shown in Figure 1 from chapter 1.
481
Figure 1. The Architecture of the Interfaces under the Thesis of the Minimalist Interface
C-I Linguistic Interface Syntactic Computation (CHL) Linguistic Interface A-P
LF1
lf 1
pf1
PF1
LF2
lf2
pf2
PF 2
LF3
lf3
pf3
PF 3
The small lf and pf should be read as the chunk of syntactic structures that are transferred to
syntax-external phonological and semantic interfaces. The big LF and PF should read as
whatever representation that the C-I and A-P systems use based on their corresponding lf and pf.
One crucial character of the dynamic view of the linguistic interface embedded within the
derivational model in Figure 1, which has been a recurrent theme throughout this dissertation,
was that syntax is not entirely crash-proof; it could make certain derivational mistakes; some
of them can be repaired by domain-specific operations at the interfaces such as deletion or
choice functions so that they may still become usable at the language-independent
articulatory and conceptual systems. This was clearly demonstrated in our investigation of
the repair-by-deletion analysis for the P-stranding and the choice function analysis for wh-in-
482
situ questions. What is crash-proof is the linguistic system as a whole including the syntactic
computation plus the linguistic interfaces.
The present model of linguistic interface driven by considerations of minimalist
interfaces suggests a partial return to the earlier model of grammar as in the Governmentand-Binding Theory, which reduces core syntactic computations to the single operation of
Move-α (Chomsky 1981; Chomsky and Lasnik 1977; Lasnik and Saito 1992; see also
Prince and Smolensky 1993). Under this generation + filter approach, the syntactic
computation itself creates a multitude of syntactic objects and all ungrammatical sentences
are filtered out by domain-/level-specific conditions such as government, binding, case
theory, control theory, and bounding theory. The present model, however, is conceptually
quite different from the GB conception of the “overgenerate +filter approach” in that the
sole role of interfaces is not to constrain the forms of objects created by syntactic
derivation but rather to improve them so that they become legible to the languageindependent thought and production modules.
The thesis of minimalist interfaces serves not only as a hypothesis about the way
syntactic computation networks with its neighboring linguistic interfaces but also opens a
new line of inquiry on the proper understanding of the nature of syntactic derivation. For
example, recall that this dissertation adopts the view that syntax is a functionally blind
483
combinatorial system (Uriagereka 1998, 2002; Hinzen 2006; Chomsky 2004, 2006) that
cannot know about the fate of its own generated objects; they may or may not be
convergent derivations. This means that it is the task of linguistic interfaces to determine
the status of such objects in terms of usability on the part of language-external sound and
concept systems and do whatever they can to modify/ornament/repair/remedy them so that
they may become legible and usable on the part of the A-P/C-I systems. This view of
convergence yields a new insight on the current debate on what syntactic objects constitute
a phase domain. As stated in chapter 2, it has been currently proposed in Chomsky (2000,
2001, 2004, 2005, 2006) that certain syntactic objects such as vPs and CPs form a phase
head; once these objects have been arranged in syntax, the complements of the phase heads
v and C undergo cyclic Spell-Out to the linguistic interfaces for semantic and phonological
interpretation. Chomsky draws on considerations of computational efficiency in support of
the phasehood of these functional heads; syntactic derivation can forget about what it
constructed in the past if derivation works in this way. Under this conceptual view of phase
theory, then, specific syntactic objects have privileged status in that they serve as a unit of
objects not only for syntax but also for purposes of phonological and semantic
interpretation. The present thesis of minimalist interface, however, leads us to the view
quite different from the view just mentioned: it is a property of the linguistics interfaces
484
that requires these syntactic objects to be chunked into phases. This is because syntax is
just about combining a subset of morphosyntactic features culled into numeration via the
recursive process of Merge and create a hierarchical object out of them. There is nothing
within syntax per se to tell us why vPs and CPs are phases, not others (see Epstein and
Seely 2002 for critical discussion of Chomsky’s view of phases). Then, we naturally
expect that interfaces are actively “invasive”: they actively participate in linguistic
computation in such a way that vPs and CPs turn out to behave as phases; in this way, vPs
and CPs do not need to be arbitrarily identified as phases, but rather the fact that they are
phases falls out as an epiphenomenon of the interface computations. This interface-driven
view has also been recently argued for in a different context by Boeckx (2007), who
proposes to let interfaces determine the convergence of objects created by syntax. This
view is conceptually natural, as mentioned several times in this dissertation, in the light of
the fact that what actually directly interacts with the language-independent sound and
conceptual modules are semantic and phonological interfaces in Figure 1; to put it
differently, by virtue of its place in Figure 1, syntax cannot worry about what will happen
in the “negotiation” between these components and the A-P/C-I systems. It is in this
respect that Chomsky’s characterization of these objects in terms of interface properties is
correct. vPs and CPs are not inherent phases in the sense that their construction
485
automatically entails Spell-Out; rather, they are determined as phases contextually by the
syntax-external semantic interface because they instantiate what have been variously
characterized as having “full argument structure” (Chomsky 2000, 2001, 2004, 2005,
2006), “complete functional complex” (Chomsky 1986b), or “propositional content”; they
happen to create good reconstruction sites that are necessary for proper interpretation of
binding relations at the interface (Legate 2003). The same story holds for the other
phonological interface: vPs and CPs are phases not by virtue of their inherent privilege
within the syntax but by virtue of the fact that they happen to demarcate possible domains
for phonological rules such as pauses and parentheticals (Uriagereka 1999), phonological
phrasing (Dobashi 2003), nuclear sentence stress (Legate 2003; Kahnemuyipour 2004;
Wagner 2005), and sandhi/mutation phenomena (Sato in press c; Carnie in preparation);
they happen to provide good escape hatches for various sorts of agreement such as WHagreement in Chamorro (Chung 1982, 1994, 1998), the realis/irrealis alternation in Palauan
(Georgopoulos 1985, 1991), the voice-movement interaction in Tagalog and Malagasy
(Rackowski and Richards 2005; Pearson 2001, 2005) and the active voice deletion in
Indonesian/Javanese/Madurese and many other phenomena briefly mentioned in chapter 2.
The present interface-based approach to determination of phasal constituents, of course,
also opens a new possibility that other syntactic objects than vPs and CPs can be phases
486
depending on external requirements from linguistic interfaces. For example, the current
minimalist interface is expected to derive the recent claim that DPs are one such candidate,
as argued for independently in recent work such as Svenonius (2004) and Hiraiwa (2005),
because certain types of DPs (action nominals, for example) could encode the same
amount of information as their verbal counterparts in terms of argument structure. Other
categories such as TP may well be strong phases given that the information contained
within vPs forms a proper subset of that contained by TPs. Category-defining derivational
morphemes such as -al, -ous, -ful and many others also may form phase heads, as proposed
independently by Marantz (in press) within the framework of Distributed Morphology (see
also Arad 2003 for potential arguments from the denominal verb formation in Hebrew),
because the determination of syntactic category has direct relevance on the linguistic
interfaces, as evidenced by categorial selection (# John broke black) and category-sensitive
stress contours (PROduceN vs. proDUCEV). This line of thought, of course, does not take
into consideration the recent conceptual argument made by Richards (2007) for the
phasehood of vPs and CPs based on the computational efficiency and feature inheritance.
The issue, however, is still on the jury.
487
One could maintain a slightly different view of the connection between syntax and its
linguistic interfaces. 1 For example, Chomsky (2006:8-9) argues that “the relation of the
generative procedure to the interfaces is asymmetrical, CI taking precedence: optimization is
primarily to the CI interface” (see also Boeckx 2008). Chomsky illustrates this primary of the
CI interface from the way natural language deals with Internal Merge. The initial copy of an
item is created by External Merge whereas all other copies of the same item are created by
Internal Merge. Studies on reconstruction (recall our discussion of Legate’s (2003) study in
chapter 2) show that all these copies play a role in semantic interpretation at the C-I interface.
This state of affairs does not obtain at the A-P interface because in normal cases, all the copies
except the highest one are deleted when the derivation is externalized at the interface. However,
as is well known, this deletion pattern at the phonological interface causes serious problems in
language processing (e.g., garden paths), a difficulty that would be easily overcome if all
copies were pronounced at the interface. Chomsky takes this conflict between computational
efficiency and communicative needs to support the primacy of the semantic interface in
language design over the phonological interface. This primary of the C-I interface in language
design may also be reflected in the phasehood of particular syntactic objects such as vPs and
1
The primacy of the C-I interface in language design, expressed in the present paragraph, is also shared by
Heidi Harley (personal communication, May 1, 2008), though in a different context related to the phasehood
of particular syntactic objects. The comments that follow after my exposition of Chomsky’s (2006) position
are based on my interpretation of her written comments on an earlier draft of this chapter.
488
CPs. Recall that Chomsky (2000, 2001, 2004, 2005, 2006) proposes that these particular
objects constitute a natural characterization in terms of “full argument structure”, “proposition”,
and other conceptual-semantic notions. In other words, this particular chunking receives
independent justification in terms of what little is known about the way meaning is calculated
in natural language. This characterization is not easy to come by at the phonological interface
because there is nothing inherently special about these particular syntactic objects in terms of
phonology that warrant their special treatment; other chunks such as AspP, TP, and VP may
well count as phases as are CPs and TPs. True, one could make a case for the independent role
of the phonological interface in the determination of phasehood of vPs and CPs by their
relative phonological isolability such as VP-fronting, pseudoclefting, intonational boundary,
fragments, etc. (Chomsky 2004; see Bošković (2001) for a critical discussion on this point) but
this behavior of these particular chunks may well be an epiphenomenon that arises from the
fact that CPs and vPs have conceptually identifiable propositional/argument-structural
properties that are intrinsic to the way semantic computation works in parallel with narrow
syntactic computation.
A similar interface-driven approach also sheds a new light on the status of the PIC
discussed in chapter 2 of this dissertation.2 This condition states that the complement of a
2
I am grateful to Heidi Harley (personal communication), Cecile McKee (personal communication), and Mosa
Hulden (personal communication) for sharing their perspectives on the content of this paragraph with me.
489
phase head undergoes Transfer to the interface components once the higher phase head is
introduced into the syntactic workspace. Chomsky’s only argument for this condition is
computational complexity, namely, that this condition allows syntax to “forget about”
material it deals with at earlier cycles of the syntactic derivation. A natural question to ask
is, why do we need to worry about whether syntax forgets about material if the phase
theory is intended to be an abstraction of syntactic computation? Of course, one could
devise an argument for this particular condition on theory-internal grounds. The
reductionist approach to syntax entailed by the minimalist interface thesis actually forces
us to take a different approach to this condition: it is deeply rooted in the way parsers work.
Let us suppose that human language parsers are top-down local parsers in that they cannot
wait too long to calculate filler-gap dependencies. Let us also suppose with a leap of faith
that “too long” means a filler-gap dependency that crosses either a vP and CP. Then, the
local movement of a phrase required by the phase theory could receive a principled syntaxexternal motivation in terms of local processing; the parser must locate as fast as possible
where the real gap is. If the filler-gap dependency is created by one-fell-swoop movement
in a tri-clausal environment, parsers are unable to process this dependency. This is
essentially the idea by Givón (1979). In a similar vein, developing a comprehensive
principle-based approach, Pritchett (1991: 302) argues that island violations not
490
ungrammatical but rather “exceed the capacity of the human sentence processor to perform
certain structural analyses during parsing.” This potential worry on the part of human
language processor is easily eliminated if the edge serves to provide a local signal (either
morphological or semantic) in a step-wise fashion that the real gap is still to be expected
down below in a more deeply embedded clause. To take the distribution of the active voice
morphology in Indonesian and Javanese, the PIC dictates that the movement is phasedependent, with its reflexes manifested in these languages in the form of the meN-/ngdeletion. This analysis is couched in the syntax-phonology interaction. However, a deeper
motivation for why this phase-based derivation is enforced in the computational
component of human language might lie elsewhere; it is a good solution provided by the
formal architecture of the interface to the local parser. If this speculation is real, the
computational constraint such as the PIC might turn out to provide quite an important
intersection for theoretical syntacticians and psycholinguists alike.
The present reductionast approach to various known properties of syntactic computation
goes even deeper. Consider the notions of phrase and word. These notions have been
considered as primitives of syntax since the advent of generative grammar. The current
framework of Minimalist Program also seems to implicitly adopt this premise. Though the
Bare Phrase Structure Theory of Chomsky (1995) and Speas (1990) have contributed to a
491
substantial simplification of the phrase structure in syntax, the XMax = phrase/XMin = word
equations are still assumed in much of the current research, at least tacitly. Carnie (1995, 2000)
argues that the notions of XP/phrase and X/word are not primitives of syntax; rather, they
behave as they are because of the properties they exhibit in various components of grammar
and their interactions. Under this view, the verb kick, for example, is specified as what we
normally call an X element not because it is inserted into a terminal node in syntactic
derivation but because the other components of grammar require such a particular status for
various (language-particular) reasons related to θ-marking, tense and agreement features,
complement selection, reference, among many others. Carnie’s theory of phrasality, thus, is
quite interface-oriented in that it proposes to reduce the alleged phrasality in syntax to
independently necessary conditions to be satisfied at the PF and LF components. This theory,
therefore, allows for linguistic interfaces to choose and specify the status of an object that is
left phrasally ambiguous within syntax.
7.3. Questions for Future Research and Conjectures about Linguistic Interfaces
I conclude this chapter by mentioning several important theoretical questions the model of
linguistic interfaces in this dissertation brings to light. First, given the view adopted in this
model that syntax is, as it were, a functionally blind computational system solely consisting of
492
abstract processes such as Internal/External Merge, Agree, Spell-Out that apply to a languageparticular set of morphosyntactic features, then at what point do we know whether a particular
syntactic object created by syntax converges or not? The proposed model suggests that it is the
task of linguistic interfaces in Figure 1 to make sure that they will be legible to their
neighboring C-I and A-P systems. This position seems natural under the particular conception
of syntax as a generative system which does not care about the fate of its own syntactic objects.
This conception is relatively easy to support at the syntax-semantics interface from numerous
cases, some of them discussed in chapter 4, where semantics tries to give a reasonable
interpretation (sometimes a coarse interpretation) to whatever objects syntax sent off to the
interface. Evidence for the comparable position at the syntax-phonology interface is not easy to
come by due to the ill-understood nature of the phonological interface at this point, but my
guess is that this interface takes the form of syntactic representation just as the syntactic
structure is because we have seen that domain-specific operations such as deletion can target
only syntactic constituents such as TPs and vPs.
The second question is what type of operations the syntax-external linguistic interfaces
can conduct to repair syntactic failures. I have discussed only two such operations, deletion
and choice functions, as two good candidates that these interfaces employ to save syntax.
The thesis of minimalist interface should at least lead us to find many other operations.
493
Several operations on the phonological side of interface immediately come to mind:
resumption (Sells 1984; Aoun and Li 2003; Boeckx and Lasnik 2006), copy spell-out
(Bošković 2001; Landau 2007), focus intonation (Neelman and Reinhart 1998; Reinhart
2006), and phonological restructuring (Dobashi 2003). Some candidates for the other side of
interface that might serve the role of repair for the purposes of communicative needs from
the C-I system include topic-focus articulation (Kiss 1998; Reinhart 2006), metaphors (cf.
Chomsky 2004), and so on. Detailed investigation of what other linguistic phenomena can
be brought to bear on the ameliorating role of linguistic interfaces is an important task to
undertake in future research.
The third question is what determines that a particular object, that is created by syntax
and passed through domain-specific operations of its neighboring interface to the A-P/CI systems, is grammatical or not? Under one standard view of the minimalist program,
syntax never makes superfluous derivational steps (Economy of Derivation) or creates
unnecessary
representations
(Economy
of
Representation);
derivations
and
representations that involve mistakes are deemed simply underivable, hence
ungrammatical. Under another view of the syntactic derivation, as in the Crash-Proof
Syntax of Frampton and Guttmann (1999, 2002), syntax is so constructed that every
object generated by this perfect component is geared to be grammatical at the A-P and
494
C-I systems. We have seen that this view seems untenable to the extent that my analysis
of the P-stranding under sluicing presented in chapter 3 and the choice function analysis
of wh-in-situ questions given in chapter 5 is on the right track.
The minimalist interface thesis, of course, forces the conclusion that the convergence of a
particular syntactic object is entirely for the A-P and C-I systems to decide because these and
only these components use the object for purposes of communication but we have quite a
limited understanding of what is in these systems. We do have some intuitive ideas about
both of these domains such as the two dimensional nature of sound strings, the world
knowledge, how things are naturally categorized and our understanding seems more clear in
the C-I system than the A-P system. For example, we have seen a bit about how our C-I
systems are supposed to work in chapter 4 where the denotation of a bare noun is sometimes
coerced in service of the denotation of a functional element that goes with it or in contexts
that force definite interpretation to the bare noun in languages such as Indonesian, Javanese,
and Japanese that arguably lack DP projections. Things are less clear about the other system.
For example, why is the failure of D-to-P incorporation, but not the failure of feature
percolation, so critical a mistake so that the resulting string that contains the former violation
is judged ungrammatical? The minimalist interface thesis could allow us to answer this
495
question as follows.3 Let us suppose that there are two types of violations in the syntaxphonology interface. One is a strictly syntactic/derivational “violation” that cannot be simply
created in the syntactic computation. To take D-to-P incorporation, this operation is
conducted in the syntax immediately once the preposition is introduced into the workspace
and serves as a probe to attract the D head within its minimal search domain (its complement
domain); whatever uninterpretable/unvalued feature of the P (e.g. strong D-feature in
languages with D-P coalescence) needs to be checked must be checked, since that is the sole
driving force for mechanical computation. The failure of the D-to-P incorporation, therefore,
is simply an impossible scenario in the minimalist vision of syntactic computation. Thus, it is
fully expected that there is no sense in which the failure of D-to-P incorporation could ever
be repaired at the PF interface. The situation could be different with the other type of
violation, interface violation. To take [+wh] feature percolation, failure of this process is a
representational violation whose severity for linguistic computation could vary from
language to language. Therefore, it is possible, in principle, that the failure of this percolation
in languages with the obligatory value of this percolation mechanism (as in Indonesian)
could be tolerated within syntax per se but rather is checked later at the PF interface. Under
this view, syntactic representations that contain failures of percolation could still have
3
I owe the exposition of the following answer to this question to the written suggestions provided by Heidi
Harley (personal communication, May 1, 2008) on an earlier draft of this chapter, which I paraphrase here.
496
chances to converge at the interface depending on what happens at this interface. If PF does
not do anything about it, then this type of representation would persist at the interface: the
representational constraint then applies to this representation and rules it out as
ungrammatical. That was seen to be the case with P-stranding under wh-movement in
Indonesian. If PF does conduct its domain-specific operation to the otherwise ill-formed
object by deleting the offending part of the representation, then the representational
constraint has nothing to apply to. As a result, the derivation can still continue to converge to
yield a grammatical output at the A-P system. That was seen to be the case with P-stranding
under sluicing/pseudogapping (wh-movement + TP/vP deletion) in Indonesian. Therefore,
under this bipartite conception of violations, the contrast in “reparability” between the failure
of D-to-P incorporation and the failure of feature percolation falls out naturally from the very
architecture of the syntax-phonology interface proposed in this dissertation. One could think
of a similar bifurcation in the interaction of syntax with the semantic interface, but
examination of this possibility is left for another occasion.
In a similar vein, our present thesis also allows us to reach a better understanding of
the distinction between grammaticality and acceptability in a way that the parsimonious
(Chomsky 1995, 2000, 2001, 2004, 2005, 2006)/ crash-proof view (Frampton and
Guttmann 1999, 2002) of minimalist syntax cannot achieve. To take one example, it has
497
been widely known (see Chomsky 1986b, Lasnik and Saito 1984, 1992, Chomsky and
Lasnik 1993, and many references cited therein) that there is a contrast between
extraction of an argument and that of an adjunct from a syntactic island. Many papers
have been written that attempt to come up with technical mechanisms to capture this
such as gamma-marking, star-marking convention, and so on. The thesis of minimalist
interfaces leads us to expect that this line of inquiry is misguided; rather, this contrast has
nothing to do with syntactic mechanisms (in fact, a possibility that cannot be entertained
anymore due to the Inclusiveness Condition) but tells us something about the way the CI system independently works. One might entertain the idea that an inherently referential
nominal object (such as who, what) is easier to retrieve than adjunct expressions (such as
why and how), which do not have independent reference. This is not entirely an
unreasonable possibility because I have provided evidence in chapter 5 that only those
wh-phrases in languages such as English and Indonesian that contain an N-set can (or
must) interpreted in situ by the choice function whose theoretical properties are not
unlike those of the conceptual module.4 This possibility, in turn, invites another idea that
often-varying degrees of unacceptability among speakers for island-violating examples
are not a matter of syntax but of language-independent considerations about reference,
4
Thanks to Juan Uriagereka (personal communication) for useful discussion in this regard.
498
basic templates for conversation, frequency of forms, contexts of utterance, and many
related semantic factors. Thus, the present thesis may even provide quite a unique
interdisciplinary corroboration between pure theoretical linguists, psycholinguists, and
philosophers/external semanticists.
However, a fundamental question still remains. What determines that a particular
violation is reparable or not? The failure of feature percolation was argued to be
repairable whereas the ECP-type/constituency-violation was not. Ross (1969) and
Merchant (2001) argue that certain violations such as the that-Trace Effect, the Left
Branch Condition, the Complex NP Constraint are repairable whereas Boeckx and
Lasnik (2006) argue that superiority violations are not repairable. Where is this type of
discrepancy rooted? As I have argued in chapter 2, the derivational vs. representational
violation provides a first approximation in seeking an answer to this question. One might
entertain the hypothesis that the former cannot be generated in syntax hence are never
violable whereas the latter can be tolerated in syntax but cannot be processed at linguistic
interfaces. This hypothesis thus relegates “repairable” violations to general cognitive
limitations on language parsing. Finding the right cutting-off point between repairable
and irreparable violations is a quite important research agenda for minimalist researchers.
499
The final and the most important question that the present dissertation brought to light is,
why would language be organized in the way suggested in this dissertation? Why would a
language not be a truly “perfect” system? 5 This dissertation argued in several places,
especially in chapters 3, 4, and 5, that the objects created by the core syntactic computation
are quite imperfect and need to be remedied/modified/repaired to be legible and usable on
the part of the language-external A-P and C-I systems. One possible answer may be sought
in the recent observation, made within the context of the Minimalist Program, that
recursivity, the fundamental property of natural language ensured by the combinatorial
process of Merge, is not a language-specific computational procedure; its workings could be
seen in other human activities such as mathematics (number quantification), music, tools,
spatial navigation, foraging, tracking, social interaction, among other relevant abilities
(Hauser et al. 2002). This observation suggests that the general recursive system has been
recruited into language, in much the same way it has been into other aspects of human
activity such as those just mentioned above. Thus, it is reasonable to expect that this general
system is not well adapted/fine-tuned to the particular way natural language should work in
its actual use as communicative tool, even though this is the primary purpose of language,
5
I thank Andrew Carnie (personal communication) for asking this question and Dave Medeiros (personal
communication) for useful comments. See also Uriagereka (1998, 2002), Hinzen (2006), and many works cited
therein for much relevant discussion.
500
viewed from the perspective of the A-P and C-I systems. Once Merge has been wired into
the faculty of language, however, the functionally blind purely mechanical combinatorial
machine got connected to the A-P and C-I systems. The objects created by this component
are so crude, hence are not tailored for the A-P and C-I systems to use. At the same time,
those language-external systems are not general enough to be able to create representations
solely based on the objects created by syntactic derivation. I conjecture that the linguistic
interfaces PF and LF are developed in the language faculty, as broadly construed (including
the sensory-motor and conceptual-intentional system; see Hauser et al. 2002 for the
distinction between the FLB (the Faculty of Language in the Broad sense) and the FLN (the
Faculty of Language in the Narrow sense), to solve this two-way discrepancy and to connect
the syntax and the A-P/C-I modules in the best possible and most economical way.
Specifically, if syntax creates an object that requires a minimum amount of work, then the
PF/LF adds a correspondingly minimum amount of modification to send it off to the A-P/C-I
systems. This was the case in the proposed analysis of the distribution of active voice
morphology in Indonesian and Javanese (chapter 2), of the denotation and morphosyntax of
bare nominals across languages (chapter 4), and of reduplication asymmetries between
nominal and verbal affixes in Indonesian (chapter 6). If syntax creates an object that requires
substantive modification for legibility, on the other hand, then the PF and LF conduct a
501
handful of domain-specific operations such as deletion and choice functions to the object to
make it usable for the external systems. This was the case in the proposed analysis of the Pstranding pattern under sluicing in Indonesian (chapter 3) and of in-situ wh-questions in
Indonesian (chapter 5). This conjecture leads us to adopt a particular minimalist vision of
linguistic interfaces: PF and LF do only a minimum amount of work to enable the
connection the purely mechanical computation and the language-external A-P/C-I systems
for convergence. Whence came Minimalist Interfaces, the title of this dissertation.
The view of the language design expressed in the previous paragraph depends on a
certain conception of “perfect language”, as often used in Chomsky’s minimalist work
(Chomsky 1995, 2000, 2001, 2004, 2005, 2006).6 One of the important agendas of the
minimalist inquiry since its inception (Chomsky 1995) has been to clarify and substantiate
the question of “how well-designed/perfect language could be”. If we take “perfect
language” to mean “how well-designed language could be for interfaces”, then the answer
would be in the negative because the syntax is poorly designed for the purposes of actual
language use; it creates failure of percolation (chapter 2), creates type mismatches (chapter
4), and lacks communicative diversities required for actual communication (chapter 6).
6
I am very grateful to Heidi Harley (personal communication, May 1, 2008) for extensive written comments
on the content of this paragraph and clarifying two interrelated notions of “perfect language” here. The
content in this paragraph is based on my understanding of her comments on an earlier draft of this chapter.
502
This is the view expressed in the last paragraph. In addition to this interface-based
conception of language design, however, there is another sense of “perfect language” in
which minimalist researchers ask “how well-designed language is in terms of
computational elegance, mathematical beauty, and simplicity.” This seems to be the view
that many minimalist/biolinguistic researchers (Medeiros 2006b), including Chomsky
himself, currently adopt. Under this conception of “perfect language”, then, one could say
that all imperfections actually lie in the linguistic interfaces, not within syntax. The AP and
CI interfaces are so biologically specialized for the purposes of actual communication,
parsing/processing, etc., that the objects created by such a mathematically elegant syntax
would be too perfect to be usable for externalization at all. Resolution of conflicts between
the two competing conjectures of language design, of course, requires a large-scale
corroboration in comparative research, neuroscience, psycholinguistic, and related fields.
What has been said above in this section is all interesting but quite premature.
However, there seems to be some converging evidence, as is clear from the results of
earlier chapters, that this view of linguistic interfaces is not terribly a wrong idea.
Whether this idea turns out to be a good heuristic of the further exploration of linguistic
interfaces and their networking with the language-external A-P and C-I systems, of
course, only time will tell.
503
REFERENCES
Abdullah bin Abdul Kadir Munshi (1963 [1849]). Hikayat Abdulla, Vol. 2. Singapore:
Pustaka Antara.
Abels, Klaus (2003). Successive cyclicity, anti-locality, and adposition stranding. Doctoral
dissertation, University of Connecticut.
Adger, David (2007). Stress and phasal syntax. Linguistic Analysis 33 (2003): 238-266.
Aldridge, Edith (In press). Phase-based account of extraction in Indonesian. Lingua.
Allen, Margaret (1978). Morphological investigations. Doctoral dissertation, University of
Connecticut.
Almeida, Diogo and Masaya Yoshida (2007). A Problem for the preposition stranding
generalization. Linguistic Inquiry 38: 349-362.
An, Duk-Ho (2007). Clauses in non-canonical positions at the syntax-phonology interface.
Syntax 10: 38-79.
Anderson, Stephen (1976). On the notion of subject in ergative languages. Subject and
topic, ed. Charles Li, 1-24. New York: Academic Press.
Anderson, Stephen (1982). Where’s morphology. Linguistic Inquiry 13: 571-612.
504
Anderson, Stephen (1992). A-morphous morphology. Cambridge: Cambridge University Press.
Aoun, Joseph and Audrey Li (1990). Minimal disjointness. Linguistics 28: 189-203.
Aoun, Joseph and Audrey Li (2003). Essays on the representational and derivational
nature of grammar: The diversity of wh-constructions. Cambridge, MA: MIT Press.
Aoun, Joseph, Norbert Hornstein, David Lightfoot, and Amy Weinberg (1987). Two types
of locality. Linguistic Inquiry 18: 537-577.
Arad, Maya (2003). Locality constraints on the interpretation of roots: The case of Hebrew
denominal verbs. Natural Language and Linguistic Theory 21: 737-778.
Arka, Wayan and Christopher Manning (1998). Voice and grammatical relations in
Indonesian: A new perspective. The Proceedings of the LFG’98 Conference, ed
Miriam Butt and Tracy King. Stanford: CSLI Publications.
Arregi, Karlos (2002). Focus on Basque movement. Doctoral dissertation, MIT.
Baker, Carl. L. (1970). Notes on the description of English questions: The role of an
abstract question morpheme. Foundations of Language 6: 197-219.
Baker, Mark (1985). The mirror principle and morphosyntactic explanation. Linguistic
Inquiry 16: 373-415.
Baker, Mark (1988). Incorporation: A theory of grammatical function changing. Chicago:
University of Chicago Press.
505
Baker, Mark (1996). The polysynthesis parameter. New York: Oxford University Press.
Baker, Mark (2003). Lexical categories: Verbs, nouns, and adjectives. Cambridge:
Cambridge University Press.
Baker, Mark, Kyle Johnson, and Ian Roberts (1989). Passive arguments raised. Linguistic
Inquiry 20: 219-251.
Barss, Andrew (2000). Minimalism and asymmetric wh-interpretation. Step by step:
Essays on minimalism in honor of Howard Lasnik, ed. by RogerMartin, David
Michaels, and Juan Uriagereka, 31-52. Cambridge, MA: MIT Press
Bašić, Monika (2004). Nominal subextractions and the structure of NPs in Serbian and
English. Master’s thesis University of Tromsø.
Beerman, Dorothee (1990). Semantic and syntactic constraints on German determiner
movement within the prepositional phrase. Master’s thesis, University of Texas, El Paso.
Bejar, Susana and Diane Massam (1999). Multiple case checking. Syntax 2: 65-79.
Bernstein, Judy (1997). Demonstratives and reinforcers in Romance and Germaic
languages. Lingua 102: 87-113.
Bobaljik, Jonathan (1995). Morphosyntax: The syntax of verbal inflection. Doctoral
dissertation, MIT.
Boeckx, Cedric (2007). Eliminating spell-out. Linguistic Analysis 34 (2003): 414-425.
506
Boeckx, Cedric (2008). Approaching parameters from below. Ms., Harvard University.
Boeckx, Cedric and Norbert Hornstein (2006). Control in Icelandic and theories of control.
Linguistic Inquiry 37: 591-606.
Boeckx, Cedric and Howard Lasnik (2006). Intervention and repair. Linguistic Inquiry
37:150-155.
Borer, Hagit (1984). Parametric syntax. Dordrecht: Foris.
Borer Hagit (1998). Morphology and syntax. The handbook of morphology, ed. Andrew
Spencer and Arnold Zwicky, 151-190. Oxford: Blackwell.
Borer, Hagit (2005). Structuring sense volume 1: In name only. New York: Oxford
University Press.
Borer, Hagit and Kenneth Wexler (1987). The maturation of syntax. Parameter setting, ed.
Thomas Roeper and Edwin Williams, 23-172. Dordrecht: Leidel.
Bošković, Željko (1997). The syntax of nonfinite complementation: An economy approach.
Cambridge, MA: MIT Press.
Bošković, Željko (2001). On the nature of the syntax-phonology interface: Cliticization
and related phenomena. Amsterdam: North-Holland/Elsevier.
507
Bošković, Željko (2003). On left branch extractions. Investigations into formal Slavic
linguistics: Contributioons of the Fourth European Confernece on Formal
Descriptions of Slavic Languages-FDSL IV, ed. Peter Kosta, Joanna Błaszczak, Jens
Frasek, Ljudimila Geits, and Marzena śygis, 543-577. Frankfurt an Main: Peter Lang.
Bošković, Željko (2004). Topicalization, focalization, lexical insertion, and scrambling.
Linguistic Inquiry 35: 613-638.
Bošković, Željko (2005). On the locality of left branch extraction and the structure of NP.
Studia Linguistica 59: 1-45.
Bošković, Željko and Howard Lasnik (2003). On the distribution of null complementizers.
Linguistic Inquiry 34: 528-546.
Branigan, Phil and MacKenzie Maguerite (2002). Altruism, A-movement and object
agreement in Innu-Aimûn. Linguistic Inquiry 33: 385–407.
Bresnan, Joan (1971a). Sentence stress and syntactic transformations. Language 47: 257-281.
Bresnan, Joan (1971b). Contraction and the transformational cycle. Ms., MIT.
Bresnan, Joan (1982). The passive in lexical theory. The mental representation of
grammatical relations, ed. Joan Bresnan, 3-86. Cambridge, MA: MIT Press.
Bresnan, Joan, and Sam Mchombo (1995). The lexical integrity principle: Evidence from
Bantu. Natural Language and Linguistic Theory 13: 181-254.
508
Carlson, Gregory (1977). Reference to kinds in English. Doctoral dissertation, University
of Massachusetts, Amherst.
Carnie, Andrew (1995). Non-verbal predication and head movement. Doctoral dissertation, MIT.
Carnie, Andrew (2000). On the definitions of X0 and XP. Syntax 5: 59-106.
Carnie, Andrew (in preparation). Consonant mutation as evidence for a vP phase. Ms.,
University of Arizona.
Carson, Jana (2000). The semantics of number in Malay noun phrases. Master’s thesis,
University of Calgary.
Cartier, Alice (1989). De-voiced transitive verb sentences in formal Indonesian. Serpihserpih telaah pasif Bahasa Indonesia [Studies on Passives in Indonesian], ed. Purwo
Bambang Kaswanti , 84-145. Yogyakarta: Penerbit Kanisius.
Cheng, Lisa (1991). On the typology of wh-questions. Doctoral dissertation, MIT.
Cheng, Lisa and Rint Sybesma (1999). Bare and not-so-bare nouns and the structure of NP.
Linguistic Inquiry 30: 509-542.
Chierchia, Gennaro (1998a). Reference to kinds acorss languages. Natural Language
Semantics 6: 339-405.
Chierchia, Gennaro (1998b). Plurality of nouns and the notion of “semantic parameter.
Events and grammar, ed. Susan Rothstein, 53-103. Dordrecht: Kluwer.
509
Choe, Jae-Woong (1987). LF movement and pied-piping. Linguistic Inquiry 18: 348-353.
Chomsky, Noam (1964). Current issues in linguistic theory. The Hague: Mouton.
Chomsky, Noam (1970). Remarks on nominalizations. Readings in English
transformational grammar, ed. Roderick Jacobs and Peter Rosenbaum, 184-221.
Waltham, MA: Blasdell.
Chomsky, Noam (1972). Some empirical issues in the theory of transformational grammar.
Goals of linguistic theory, ed. Stanley Peters, 63-130. Englewood Cliffs: Winston.
Chomsky, Noam (1973). Conditions on transformation. A festschrift for Morris Halle, ed.
Stephen Anderson and Paul Kiparsky, 232-286. New York: Holt, Rinehart and Winston.
Chomsky, Noam (1977). On wh-movement. In Formal syntax, ed. Peter Culicover,
Thomas Wasow, and Adrianne Akmajian, 71-131. New York: Academic Press.
Chomsky, Noam (1981). Lectures on government and binding. Dodrecht: Foris.
Chomsky, Noam (1986a). Knwoledge of language: Its nature, origin, and use. New York: Praeger.
Chomsky, Noam (1986b). Barriers. Cambridge, MA: MIT Press.
Chomsky, Noam (1991). Some notes on economy of derivation and representation.
Principles and parameters in comparative grammar, ed. Robert Freidin, 417–454.
Cambridge, MA: MIT Press. [Reprinted as chapter 2 in Chomsky (1995)]
510
Chomsky, Noam (1993). A minimalist program for linguistic theory. A view from Building
20: Essays in linguistics in honor of Sylvain Bromberger, ed. Kenneth Hale and
Samuel Jay Keyser, 1-52. Cambridge, MA: MIT Press.
Chomsky, Noam (1995). The minimalist program. Cambridge, MA: MIT Press.
Chomsky, Noam (2000). Minimalist inquiries: The framework. Step by step: Essays on
minimalist syntax in honor of Howard Lasnik, ed. R. Martin, D. Michaels & J.
Uriagereka, 89–155. Cambridge, MA: MIT Press.
Chomsky, Noam (2001). Derivation by phase. Ken Hale: A life in language, ed. Michael
Kenstowicz, 1-52. Cambridge, MA: MIT Press.
Chomsky, Noam (2004). Beyond explanatory adequacy. Structure and beyond: The
cartography of syntactic structures, volume 3, ed. A. Belletti, 104–131. Oxford:
Oxford University Press.
Chomsky, Noam (2005). On phases. Ms., MIT. [To appear in Foundational issues in
linguistic theory, ed. Robert Freidin, Carlos Otero, and Maria-Luisa Zubizarreta.
Cambridge, MA: MIT Press.
Chomsky, Noam (2006). Approaching UG from below. Ms., MIT.
Chomsky, Noam and Howard Lasnik (1977). Filters and Control. Linguistic Inquiry 8: 425-504
511
Chomsky, Noam and Howard Lasnik (1993). Principles and parameters theory. Syntax: An
international handbook of contemporary research, ed. Joachim Jacobs, Arnim von
Stechow, Wolfgang Sternefeld, and Theo Vennemann, 506-569. [Reprinted as
chapter 1 in Chomsky (1995)]
Chung, Sandra (1976). An object-creating rule in Bahasa Indonesia. Linguistic Inquiry 7: 41–87.
Chung, Sandra (1982). Unbounded dependencies in Chamorro grammar. Linguistic
Inquiry 13: 39–77.
Chung, Sandra (1994). Wh-agreement and “referentiality” in Chamorro. Linguistic Inquiry
25:1–44.
Chung, Sandra (1998). The design of agreement: Evidence from Chamorro. Chicago:
University of Chicago Press.
Chung, Sandra (2000). On reference to kinds in Indonesian. Natural Language Sematics 8:
157-171.
Chung, Sandra (In press). Indonesian clause structure from an Austronesian perspective. Lingua.
Chung, Sandra, William Ladusaw, and James McCloskey (1995). Sluicing and Logical
Form. Natural Language Semantics 3: 239-282.
Cinque, Guglielmo (1990). Types of Aì-dependencies. Cambridge, MA: MIT Press.
512
Cinque, Guglielmo (1999). Adverbs and functional heads-A crosslinguistic perspective.
Oxford: Oxford University Press.
Clahsen, Harold (1991). Constrains on parameter setting: A grammatocal analysis of some
acqusition stages in German child language. Language Acquisition 1: 361-391.
Clahsen, Harald and Penke Matina (1992). The acquisition of agreement morphology and
its syntactic consequences: New evidence on German child language from the
Simone-Corpus. The acquisition of verb placement, ed. Jürgen Meisel, 5-22.
Dordrecht: Kluwer.
Clements, George (1984). Binding domains in Kikuyu. Studies in the Linguistic Sciences
14: 37–56.
Clements, George, James McCloskey, Joan Maling and Annie Zaenen (1983). Stringvacuous rule application. Linguistic Inquiry 14: 1-17.
Cole, Peter and Gabriella Hermon (1998). The typology of WH movement: WH questions
in Malay. Syntax 1, 221–258.
Cole, Peter and Gabriella Hermon (2000). Partial wh-movement: Evidence from Malay.
Wh-scope marking, ed. Uli Lutz, Gereon Müller, and Arnim. von Stechow, 101-130.
Amsterdam: John Benjamins.
Cole, Peter and Gabriella Hermon (2005). Subject and non-subject relativization in
Indonesian. Journal of East Asian Linguistics 14: 59-81.
Cole, Peter and Gabriella Hermon (In press). Voice in Malay/Indonesian. Lingua.
513
Cole, Peter, Jonczyk Elizabeth., & Lilley Jason (1999). A note on extraction from object
position in Javanese and other Javanic languages. Proceedings of AFLA VI, ed.
Carolyn Smallwood & Catherine Kitto, 87–93. Toronto: University of Toronto.
Cole, Peter, Gabriella Hermon, and Norhaida Aman (To appear). Clefted questions in
Malay. Malay/Indonesian Linguistics, ed. David Gil. London: Curzon Press.
Contreras, Heles (1986). Spanish bare NPs and the ECP. Generative studies in Spanish syntax,
ed. Ivonne Bordelois, Heles Contreras, and Karen Zagona, 25-49. Dordrecht: Foris.
Corver, Norbert (1992). On deriving left brach extraction asymmetries: A case study in
parametric syntax. Proceedings of NELS 22, ed. by Kimberly Broderick, 67-84.
University of Massachusets, Amherst: GLSA.
Croft, William (1994). Semantic universals in classifier systems. Word 45: 145-171.
Dardjowidjojo, Soenjono (1978). Sentence patterns of Indonesian. Honolulu: University of
Hawaii Press.
Davies, William (1999). Reduplication, multiple events, and Madurese reciprocals. In:
Carolyn Smallwood and Catherine Kitto (eds.) Toronto Working Papers in
Linguistics: Proceedings of the Sixth Meeting of the Austronesian Formal Linguistics
Association, 95-106. Toronto: University of Toronto.
Davies, William (2000). Events in Madurese reciprocals. Oceanic Linguistics 39: 123-143.
Davies, William (2003). Extreme locality in Madurese WH-questions. Syntax 6: 237–259.
514
Davies, William (2005). Madurese prolepsis and its implications for a typology of raising.
Language 81: 645-665.
Den Dikken, Marcel. (2001). ‘Pluringulars’, pronouns and quirky agreement. The
Linguistic Review 18, 19-41.
Den Dikken, Marcel. (2006). A reappraisal of vP being phasal: A reply to Legate. Ms.,
CUNY.
Déprez, Viviane (1989). On the typology of syntactic positions and the nature of chains:
Move a(alpha) to the specifier of functional projections. Doctoral dissertation, MIT.
Di Sciullo, Anna-Maria and Edwin Williams (1987). On the definition of words.
Cambridge, MA: MIT Press.
Dobashi, Yoshihito (2003). Phonological phrasing and syntactic derivations. Doctoral
dissertation, Cornell University.
Duffield, Nigel (1995). Particles and projections in Irish syntax. Dordrecht: Kluwer.
Du plessis, Hans (1977). Wh-movement in Afrikaans. Linguistic Inquiry 8: 723-726.
Dyen, Isidore (1964). Beginning Indonesian: Lessons 1-24, Vol. 1-3. U.S. Department of
Education: Language Research Section.
515
Embick, David and Rolf Noyer (2007). Distributed morphology and the
syntax/morphology interface. The Oxford handbook of linguistic interfaces, ed.
Gillian Ramchand and Charles Reiss, 289-324. Oxford: Oxford University Press.
Engdahl, Elisabet (1986). Constituent questions. Dordrecht: Reidel.
Englebretson, Robert (2003). Searching for structure: The problem of complementation in
colloquial Indonesian conversation. Amsterdam/Philadelphia: John Benjamins.
Epstein, Samuel D. and Daniel Seely (2002). Rule applications as cycles in a level-free
syntax. Derivation and explanation in the minimalist program, ed. Samuel D.
Epstein and Daniel Seely, 65-89. Oxford: Blackwell.
Epstein, Samuel D., Erich. M. Groat, Kawashima Ruriko & Kitahara Hisatsugu (1998). A
derivational approach to syntactic relations. Oxford: Oxford University Press.
Erteschik-Shir, Nomi (1977). On the nature of island constraints. Bloomington: Indiana
University Linguistics Club.
Fabb, Nigel (1984). Syntactic affixation. Doctoral dissertation, MIT.
Farmer, Ann (1982). On the interaction of morphology and syntax. Doctoral dissertation, MIT.
Felser, Claudia (2004). Wh-copying, phases and successive cyclicity. Lingua 114: 543–574
516
Finer, Daniel (1998). V to D raising in Sulawesi relatives. Arguments and structure in
Austronesian languages: Thematic issue of the Candian Journal of Linguistics 43, ed.
Diane Massam and Barry Miller, 283-306.
Fodor, Janet, and Ivan Sag (1982). Referential and quantificational indefinites. Linguistics
and Philosophy 5: 355-398.
Fortin, Catherine (2007a). Indonesian sluicing. Paper presented at the 81st Annual Meeting
of the Linguistic Society of America, Anaheim, CA. January 2007.
Fortin, Catherine (2007b). Indonesian sluicing and verb phrase ellipsis: Description and
explanation in a minimalist framework. Doctoral dissertation, University of Michigan.
Fortin, Catherine (In press). Reconciling meng- and NP movement in Indonesian.
Proceedings of the 32nd Annual Meeting of the Berkeley Linguistic Society.
University of California, Berkeley: Berkeley Linguistics Society.
Fox, Danny (1998). Economy and semantic interpretation. Doctoral dissertation, MIT.
Fox, Danny (2000). Economy and semantic interpretation. Cambridge, MA: MIT Press.
Fox, Danny and Howard Lasnik (2003). Successive-cyclic movement and island repair:
The difference between sluicing and VP-ellipsis. Linguistic Inquiry 34: 143-154.
Frampton, John and Samuel Guttmann (1999). Cyclic computation, a computational
efficient minimalist syntax. Syntax 2: 1-27.
517
Frampton, John and Samuel Guttmann (2002). Crash-proof syntax. Derivation and
explanation in the minimalist program, ed. Samuel Epstein and Daniel Seely, 90105. Oxford: Blackwell.
Franko, Jon (1993). On object agreement in Spanish. Doctoral dissertation, University of
Southern California.
Franko, Jon and Alazne Landa (2006). Preverbal N-words and anti-agreement effects.
Selected Proceedings of the 9th Hispanic Linguistics Symposium, ed. Nuria Sagarra
and Almeida Jacqueline Toribio, 34-42. Somerville, MA: Cascadilla.
Franks, Steve and Asya Pereltsvaig (2004). Functional catgeories in the nominal domain.
Proceedings of FASL 12: The Ottawa Meeting 2003, ed. Olga Arnaudova, Wayles
Browne, Maria Luisa Rivero and Danijela Stojanović, 109-128. Ann Arbor, MI:
Uiversity of Michigan Press.
Fromkin, Vicki, Robert Rodman, and Nina Hyams (2003). An introduction to language,
seventh edition. Boston: Heinle.
Fukui, Naoki (1986). A theory of category projections and its application. Doctoral
dissertation, MIT.
Fukui, Naoki (1993) A note on improper movement. The Linguistic Review 10: 111-126.
Fukui, Naoki (1995). Theory of projection in syntax. Stanford, CA: CSLI Publications.
518
Fukui, Naoki and Margaret Speas (1986). Specifiers and projections. Papers in Linguistics
8: Papers on Theoretical Linguistics, ed. Naoki Fukui, Tova Rapoport and Elisabeth
Sagey, 128-172. Department of Linguistics and Philosophy: MITWPL.
Georgopoulos, Carol (1985). Variables in Palauan syntax. Natural Language & Linguistic
Theory 3: 59–94.
Georgopoulos, Carol (1991). Resumptive pronouns and A′-binding in Palauan. Dordrecht:
Kluwer.
Ghomeshi, Jila (1996). Projection and inflection: A study of Persian phrase structure.
Doctoral dissertation, University of Toronto.
Gil, David (2002). The prefixes di- and N- in Malay/Indonesian dialects. The history and
typology of Western Austronesian voice systems, ed. Fay Wouk and Malcolm Ross,
241-283. Canberra: Pacific Linguistics.
Gil, David (2005). Word order without syntactic categories: How Riau Indonesian does it.
Verb first: On the syntax of verb initial languages, ed. Andrew Carnie, Heidi Harley,
and Sheila Dooley, 243-263. Amsterdam: John Benjamins.
Givón, Talmy (1979). On understanding grammar. New York: Academic Press.
Goldberg, Adele (1995). Constructions: A construction grammar approach to argument
structure. Chicago: University of Chicago Press.
Goldberg, Adele (2006). Constructions at work: The nature of generalization in language.
Oxford: Oxford University Press.
519
Grimshaw, Jane (1991). Extended projections. Ms., Brandeis University.
Grimshaw, Jane (1993). Minimal projection, heads and optimality. Ms. Rutgers University
Center for Cognitive Science.
Grimshaw, Jane (2005). Words and structure. Stanford, CA: CSLI Publications.
Groat, Erich and John O’Neil (1996). Spell-out at the LF interface. Minimal ideas: Syntactic
studies in the minimalist framework, ed. Abraham Werner, Samuel Epstein, Höskuldur
Thráinsson, and Jan-Wouter Zwart, 113-139. Amsterdam/Philadelphia: Benjamins.
Groenendijk, Jeroen and Martin Stokhof (1997). Questions. Handbook of logic and
language, ed. Johan van Benthem and Alice ter Meulen, 1055-1124.
Amsterdam/Cambridge, MA, Elsevier/MIT Press.
Grohmann, Kleanthes (2003). Prolific domains: On the anti-locality of movement
dependencies. Amsterdam: John Benjamins.
Guilfoyle, Eithne, Henrietta Hung, and Lisa Travis (1992). Spec of IP and Spec of VP: Two
subjects in Austronesian languages. Natural Language & Linguistic Theory 10, 375–414.
Guilfoyle, Eithne and Maire Noonan (1992). Functional categories and language acquision.
Canadian Journal of Linguistics 37: 241-272.
Haegeman, Liliane (1992). Theory and description in genertative syntax: A case study in
West Flemish. Cambridge: Cambridge University Press.
520
Hale, Kenneth and Samuel J. Keyser (1993). On argument structure and the lexical
expression of syntactic relations. The view from building 20: Essays on linguistics in
honor of Sylvain Bromberger, ed. Kenneth Hale and Samuel J. Keyser, 53-109.
Cambridge, MA: MIT Press.
Hale, Kenneth and Samuel J. Keyser (2002). Prolegomenon to a theory of argument
structure. Cambridge, MA: MIT Press.
Halle, Morris (1997). Distributed morphology: Impoverishment and fission. MIT
Working Papers in Linguistics 30: Papers at the Interface, ed. Benjamin Bruening,
Yoonjung Kang, and Martha McGinnis, 425-449. Department of Linguistics and
Philosophy: MITWPL.
Halle, Morris and Alec Marantz (1993). Distributed morphology and the pieces of
inflection. A view from building 20: Essays in linguistics in honor of Sylvain
Bromberger, 111-176. Cambridge, MA: MIT Press.
Halle, Morris and Alec Marantz (1994). Some key features of Distributed Morphology.
MIT Working Papers in Linguistics 21: Papers on phonology and morphology, ed.
Andrew Carnie and Heidi Harley, 275-288. Department of Linguistics and
Philosophy: MITWPL.
Harley, Heidi (2005). One-replacement, unaccusativity, acatgeorial roots and bare phrase
structure. Harvard Working Papers in Linguistics 11, ed. Slava Gorbachov and
Andrew Nevins. Harvard University: Department of Linguistics.
Harley, Heidi (2006). English words: A linguistic introduction. Oxford: Blackwell.
521
Harley, Heidi (2008). Compounding in distributed morphology. Ms., University of
Arizona, Tucson. [To appear in The Oxford handbook of Compounding, ed.
Rochelle Lieber and Pavol Stekauer, Oxford: Oxford University Press]
Harley, Heidi and Rolf Noyer (1999).State-of-the-Article: Distributed morphology. Glot
International 4: 3-9.
Harley, Heidi and Rolf Noyer (2000). Formal versus encyclopedia properties of
vocabulary: Evidence from nominalizations. The lexicon-encyclopedia interface, ed.
Bert Peters, 349-374. Amsterdam: Elsevier.
Hartman, Jeremy (2005). Sluicing in Finnish. Ms., Harvard University.
Haugen, Jason (2004). Issues in comparative Uto-Aztecan morphosyntax. Doctoral
dissertation. University of Arizona.
Haugen, Jason and Heidi Harley (2006). Reduplication in Yaqui (Hiaki) compound verbs
and the lexicalist hypothesis. Ms., University of Arizona.
Hauser, Mark, Noam Chomsky, and Tecumseh Fitch (2002). The language faculty: What
is it, who has it, and how did it evolve? Science 298: 1569-1579.
Heim, Irene (1982). The semantics of definite and indefinite nouns phrases. Doctoral
dissertation, University of Massachusetts, Amherst.
Heim, Irene and Angelika Kratzer (1998). Semantics in generative grammar. Oxford:
Blackwell.
522
Hersulund, Michael (1984). Particles, prefixes and preposition stranding. Topics in Danish
syntax, ed. Finn Sørensen and Lars Heltoft, 34-71. Copenhagen: Akademisk Forlag.
Higginbotham, James (1985). On semantics. Linguistic Inquiry 16: 547-594.
Higginbotham, James and Robert May (1981). Questions, quantifiers, and crossing. The
Linguistic Review 1: 41-79.
Hinzen, Wolfram (2006). Mind design and minimal syntax. Oxford: Oxford University Press.
Hiraiwa, Ken (2005). Dimensions of symmetry in syntax: Agreement and clausal
architecture. Doctoral dissertation, MIT.
Hopper, Paul (1983). Ergative, passive and active in Malay narrative. Discourse
perspective on syntax, ed. Flora Kelin-Andreu, 67-88. New York: Academic Press.
Hopper, Paul (1986). Some discourse functions of classifiers in Malay. Noun classes and
categorization, ed. Chaudron Craig, 309-325. Amsterdam: John Benjamins.
Hornstein, Norbert and Amy Weinberg (1981). Case theory and preposition stranding.
Linguistic Inquiry 12: 55-91.
Horvath, Julia (1997). The status of ‘wh-expletives’ and the partial movement construction
of Hungarian. Natural Language & Linguistic Theory 15: 509–572.
Huang, J. C.-T. (1982). Logical relations in Chinese and the theory of grammar. Doctoral
dissertation, MIT.
523
Iljic, Robert (1994). Quantification in Mandarin Chinese: Two markers of pluraity.
Linguistics 32: 91-116.
Ionin, Tania and Ora Matushansky (2006). The composition of complex cardinals. Journal
of Semantics 23: 315-360.
Jackendoff, Ray (1972). Semantic interpretation in generative grammar. Cambridge, MA:
MIT Press.
Jackendoff, Ray (1990). Semantic structures. Cambridge, MA: MIT Press.
Jayaseelan, Karattuparambil (1990). Incomplete VP deletion and gapping. Linguistic
Analysis 20: 64-81.
Jensen, John, and Stong-Jensen Margaret (1984). Morphology is in the lexicon! Linguistic
Inquiry 15: 474-498.
Johns, Alana (1999). The lexical basis of noun incorporation in Inuktitut. Paper presented
at the Workshop on the Structure and Constituency in the Languages of the Americas,
University of British Columbia, March 1999.
Kahnemuyipour, Arsalan (2004). The syntax of sentential stress. Doctoral dissertation,
University of Tronto.
Kana, Vamarasi M. (1986). Grammatical relations in Bahasa Indonesia. Doctoral
dissertation, Cornell University.
524
Kaplan, Ronald M., and Joan Bresnan (1982). Lexical-functional grammar: A formal
system for grammatical representation. The mental representation of grammatical
relations, ed. Joan Bresnan, 173-281.Cambridge, MA: MIT Press.
Karttunen, Lauri (1977). Syntax and semantics of question. Linguistics and Philosophy 1: 3-44.
Kats, Jerrold and Paul Postal (1964). An integrated theory of linguistic descriptions.
Cambridge, MA: MIT Press.
Kayne, Richard (1981). On certain differences between English and French. Linguistic
Inquiry 12: 349-371.
Kayne, Richard (1989). Facets of Romance past participle agreement. Dialect variation in
the theory of grammar, ed. Paola Benincà, 85–103. Dordrecht: Foris.
Kayne, Richard (1994). The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kayne, Richard (1998). Overt vs. covert movement. Syntax 1: 128-191.
Kayne, Richard and Jean-Yves Pollock (1979). Stylistic inversion, successive cyclicity and
Move NP in French. Linguistic Inquiry 9: 595-621.
Keenan, Edward (1972). Relative clause formation in Malagasy (and some related and
some not so related languages). The Chicago which hunt: Papers from the Relative
Clause Festival, ed. Paul Peranteau, Judith Levi, and Gloria Phares, 169-189.
Chicago: Chicago Linguistic Society.
525
Keenan, Edward and Bernard Comrie (1977). Noun phrase accessibility and universal
grammar. Linguistic Inquiry 1: 63–99.
Kiparsky, Paul (1982a). From cyclic phonology to lexical phonology. The structure of
phonological representations, ed. Hulst H. van der and Norval Smith, 131-175.
Dordrecht: Foris.
Kiparsky, Paul (1982b). Lexical morphology and phonology. Linguistics in the morning
calm, ed. The Linguistic Society of Korea, 3-91. Seoul: Hanshin.
Kiparsky, Paul (1982c). Word formation and the lexicon. Proceeding of the 1982 MidAmerica Linguistics Conference, ed. Frances Ingemann, 3-22. University of Kansas:
Department of Linguistics.
Kiparsky, Paul (1985). Some consequences of lexical phonology. Phonology Yearbook 2,
ed. Kaisse, Ellen M. and Patricia A. Shaw, 83-138.
Kiss, Katalin (1998). Identificational Focus versus Information Focus. Language 74: 245-268.
Kitagawa, Chisato and Claudia Ross (1982). Prenominal modification in Chinese and
Japanese. Linguistic Analysis 9: 19-53.
Kizu, Mika (2005). Cleft Constructions in Japanese Syntax. New York: Palgrave
Macmillan.
Klima, Edward (1964). Negation in English. The structure of language: Readings in the
philosophy of language, ed. Jerry Fodor and Jerrold Katz, 246-323. Englewood Cliffs,
NJ: Prentice-Hall.
526
Koopman, Hilda and Dominique Sportiche (1983). Variables and the bijection principle.
The Linguistic Review 2: 139-160.
Koopman, Hilde and Dominique Sportiche (1991). The position of subjects. Lingua 85:
211-258.
Kratzer, Angelika (1996). Severing the external argument from its verb. Phrase structure
and the lexicon, ed. Johan Rooryck and Laurie Zaring, 109-137. Dordrecht: Kluwer.
Kratzer, Angelika (1998). Scope or pseudoscope? Are there wide-scope indefinites?
Events in grammar, ed. S. Rothstein, 163-196. Dordrecht: Kluwer.
Krifka, Manfred, Francis Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro Chierchia,
and Godehard Link (1995). Genericity: An introduction, ed. Gregory Carlson and
Francis Pelletier. 1-124. London/Chicago: University of Chicago Press.
Kuroda, Shigeyuki (1965). Generative grammatical studies in the Japanese language.
Doctoral dissertation. MIT.
Landau, Idan (2007). Constraints on partial VP-fronting. Syntax 10: 127-164.
Lapointe, Steven (1980). A theory of grammatical agreement. Doctoral dissertation,
University of Massachusetts, Amherst.
Lapointe, Steven (1981). General and restricted agreement phenomena. The scope of
lexical rules, ed. Michael Moortgat, Harry van der Hulst and Teun Hoekstra, 125-159.
Dordrecht: Foris.
527
Larson, Richard (1994). The projection of DegP and DP structure. Ms., SUNY, Stony Brook.
Lasnik, Howard (1972). Analysis of negation in English. Doctoral dissertation, MIT.
Lasnik, Howard (1995). A note on pseudogapping. MIT Working Papers in Linguistics 27:
Papers on Minimalist Syntax, ed. Robert Pensalfini and Hiroyuki Ura, 143-163.
Department of Linguistics and Philosophy: MITWPL.
Lasnik, Howard (1999). On feature strength: Three minimalist approaches to overt
movement. Linguistic Inquiry 30: 197-217.
Lasnik, Howard (2001). When can you save a structure by destroying it? Proceedings of
NELS 31, ed. Min-Joo Kim and Uri Strauss, 301-320. University of Massachusetts,
Amherst: GLSA
Lasnik, Howard (2005) “Speculations on P-Stranding,” Class lecture handout, Linguistic Society
of America Summer Institute, MIT/Harvard University, July 18-August 3 [available at
http://www.ling.umd.edu/~lasnik/LSA%20course/Lasnik%20LSA%20HO%20Sect%202%2
0Appendix%20P-stranding.pdf]
Lasnik, Howard (2007). On ellipsis: The PF approach to missing constituents. University of
Maryland Working Papers in Linguistics 15, ed. Anastasia Conroy, Chunyuang Jing,
Chizuru Nakao, and Eri Takahashi, 143-153. Department of Linguistics: UMWPIL.
Lasnik, Howard and Mamoru Saito (1984). On the nature of proper government. Linguistic
Inquiry 15: 235-289.
528
Lasnik, Howard and Mamoru Saito (1992). Move α: Conditions on its application and
output. Cambridge, MA: MIT Press.
Law, Paul (1991). Effects of head movement on theories of subjacency and proper
government. Doctoral dissertation, MIT.
Law, Paul (1998). A unified analysis of P-Stranding in Romance and Germanic.
Proceedings of NELS 28, ed. Pius Tamanji and Kiyomi Kusumoto, 219-234.
University of Massachusetts, Amherst: GLSA.
Law, Paul (2006). Preposition stranding. The Blackwell companion to syntax, volume III,
ed. Martin Everaert, Henk van Riemsdijk, Rob Goedemans, and Bart Hollebrandse,
631-684. Oxford: Blackwell.
Lebeaux, David (1988). Language acquisition and the form of grammar. Doctoral
dissertation, University of Massachusetts, Amherst.
Legate, Julie (2003). Some interface properties of the phase. Linguistic Inquiry 34: 506–516.
Leko, N (1999). Functional categories and the structure of the DP in Bosnian. Topics in
south Slavic syntax and semantics, ed. Mila Dimitrova-Vulchanova and Lars Hellan,
229-252. Amsterdam: John Benjamins.
Levin, Beth and Malka Rappaport H (1986). The formation of adjectival passives.
Linguistic Inquiry 17: 623-661.
Levin, Beth and Malka Rappaport Hovav (1995). Unaccusativity: At the Syntax-Lexical
Semantics Interface. Cambridge, MA: MIT Press.
529
Li, Yafei (1990). Conditions on X-movement. Doctoral dissertation, MIT.
Lin, Jo-Wang (2004). Choice functions and scope of existential polarity wh-phrases in
Mandarin Chinese. Linguistics and Philosophy 27: 451-491.
Longobardi, Giuseppe (1994). Reference and proper names: A theory of N-movement in
syntax and logical form. Linguistic Inquiry 25: 609-655.
López, Luis (2001). On the (non)complementarity of θ-theory and checking theory.
Linguistic Inquiry 32: 694-716.
MacDonald, Ross (1976). Indonesian reference grammar. Washington, DC: Georgetown
University Press.
MacDonald, Ross and Soenjono Dardjowidjojo (1967). Indonesian reference grammar.
Washington, DC: Georgetown University Press.
Mahajan, Anoop (1989). Agreement and agreement phrases. MIT Working Papers in
Linguistics 10: Functional heads and clause structure, ed. Itziar Laka and Anoop
Mahajan, 217-252. Department of Linguistics and Philosophy: MITWPL.
Mahajan, Anoop (1990). The A/A-bar distinction and movement theory. Doctoral
dissertation, MIT.
Manning, Christopher (1996). Ergativity: Argument structure and grammatical relations.
Stanford, CA: CSLI Publications.
530
Marantz, Alec (1982). Re reduplication. Linguistic Inquiry 13: 435-482.
Marantz, Alec (1997). No escape from syntax: Don’t try morphological analysis in the
privacy of your own lexicon. Upenn Working Papers in Linguistics 4: Proceedings of
the 21st Annual Penn Linguistics Colloquium, ed. Alexis Dimidiates, Laura Siegel,
Clarissa Sure-Clark, and Alexander Williams, 201-225. Philadelphia: University of
Pennsylvania, Department of Linguistics.
Marantz, Alec (In press). Phases and words. Ms., New York University.
Massam, Diane (2001). Pseudo noun incorporation in Niuean. Natural Language and
Linguistic Theory 19: 153-197.
Matthewson, Lisa (1999). On the interpretation of wide-scope indefinites. Natural
Language Semantics 7: 79-134.
May, Robert (1985). Logical form. Cambridge, MA: MIT Press.
McCarthy, John, and Alan Prince (1986). Prosodic morphology. Ms. University of
Massachusetts, Amherst & Rutgers University.
McCarthy, John, and Alan Prince (1993). Prosodic morphology I: Constraint interaction
and satisfaction. Ms. University of Massachusetts, Amherst & Rutgers University.
McCarthy, John, and Alan Prince (1995). Faithfulness and reduplicative identity.
University of Massachusetts Occasional Papers in Linguistics 18: Papers in
optimality theory, ed. Jill Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk,
249-384. University of Massachusetts, Amherst: GLSA.
531
McCloskey, James (1979). Transformational syntax and model theoretic semantics: A case
study in Modern Irish. Dordrecht: Reidel.
McCloskey, James (2001). The morphosyntax of WH-extraction in Irish. Journal of
Linguistics 37, 67–100.
McDaniel, Dana (1989). Partial and multiple wh-movement. Natural Language &
Linguistic Theory 7: 565–604.
Medeiros, David (2006a). Pied-piping with inversion and labeling. Paper presented at the 80th
Annual Meeting of the Linguistic Society of America, Albuquerque, New Mexico.
Medeiros, David (2006b). A case of ontological minimalism: X-bar as an optimal
packaging solution. Ms., University of Arizona, Tucson.
Merchant, Jason (1998). Pseudosluicing: Elliptical clefts in Japanese and English. ZAS
Working Papers in Linguistics 10, ed. Artemis Alexiadou, Nanna Fuhrhop, Paul Law,
and Ursula Kleinhenz, 88-112. Zentrum für Allgemeine
Merchant, Jason. (2001) The syntax of silence: Sluicing, islands, and the theory of ellipsis.
Oxford: Oxford University Press.
Merchant, Jason (2008). An asymmetry in voice mismatches in VP-ellipsis and
pseudogapping. Linguistic Inquiry 39: 169-179.
Miller, Wick and Susan Ervin-Tripp (1973). The development of grammar in child
language. Studies in child language development, ed. Dan Slobin and Charles
Ferguson, 355-380. New York: Holt, Rinehart and Winston.
532
Mohanan, K.P. (1986). The theory of lexical phonology. Dorecht: Reidel.
Moravcsik, Edith (1978). Reduplicative constructions. Universals of human language volume
3, ed. Joseph H. Greenberg, 297-334. Stanford, CA: Stanford University Press.
Musgrave, Simon (2001). Non-subject arguments in Indonesian. Doctoral dissertation,
University of Melbourne.
Nakamura, Masanori (1994). An economy account of wh-extraction in Tagalog. The
Proceedings of the Twelfth West Coast Conference on Formal Linguistics, ed. Erin
Duncan, Donka Farkas, and Phillip Spaelti, 405-420. Stanford: CSLI Publications.
Nakanishi, Kimiko and Satoshi Tomioka (2004). Japanese plurals are exceptional. Journal
of East Asian Linguistics 13: 113-140.
Ndayiragije, Juvénal (2006). The ergativity parameter: The view from antipassive.
Ergativity: Emerging issues, ed. Alana Johns, Diane Massam, and Juvénal
Ndayiragije, 271-292. Dordrecht: Kluwer.
Neeleman, Ad and Tanya Reinhart (1998). Scrambling and the PF interface. The
projection of arguments: Lexical and compositional factors, ed. Miriam Butt and
Wilhelm Geuder, 309-353. Stanford: CSLI Publications.
Nichols, Lynn (1997). DP and polysynthesis. Proceedings of NELS 27, ed. Kiyomi
Kusumoto, 305-320. University of Massachusetts, Amherst: GLSA.
Nishigauchi, Taisuke (1986). Quantification in syntax. Doctoral dissertation, University of
Massachusetts, Amherst.
533
Nishigauchi, Taisuke (1990). Quantification in the theory of grammar. Dordrecht: Kluwer.
Nishiyama, Kunio, John Whitman, and Eun-Young Yi (1996). Syntactic movement of
Overt wh-phrases in Japanese and Korean. Japanese/Korean Linguistics 5, ed. Noriko
Iwasaki, Susan Strauss, and Shoichi Iwasaki, 337-351. Stanford: CSLI Publications.
Noguchi, Tohru (1997). Two types of pronouns and variable binding. Language 73: 770-797.
Nunes, Jairo (2004). Linearization of chains and sideward movement. Cambridge, MA:
MIT Press.
Ouhalla, Jamal (1993). Subject extraction, negation and the anti-agreement effect. Natural
Language & Linguistic Theory 11: 477-518.
Ouhalla, Jamal (2005). Agreement features, agreement and antiagreement. Natural
Language and Linguistic Theory 23: 655-686.
Pearson, Matthew (2001). The clause structure of Malagasy: A minimalist approach.
Doctoral dissertation, University of California, Los Angeles.
Pearson, Matthew (2005). The Malagasy subject/topic as an A′-element. Natural
Language & Linguistic Theory 23: 381–457.
Pereltsvaig, Asya (2007). The universality of DP: A view from Russian. Studia Linguistica
61: 59-94.
534
Pérez-Leroux Ana and Thomas Roeper (1999). Scope and the structure of bare nominals:
Evidence from child language. Linguistics 37: 927-960.
Pesetsky, David (1979). Russian morphology and lexical theory. Ms. MIT.
Pesetsky, David (1987). Wh-in-situ: Movement and unselective binding. The
representation of (in)definites, ed. Eric. Reuland and Alice. ter Meulen, 98-129.
Cambridge, MA: MIT Press.
Pesetsky, David (1995) Zero syntax: Experiencers and cascades. Cambridge, MA: MIT Press.
Pesetsky, David (1998). Phrasal movement and its kin. Cambridge, MA: MIT Press.
Platzack, Krister (1991). A grammar without functional catgeories: A syntactic study of
early child language. Nordic Journal of Linguistics 13: 107-126.
Poedjosoedarmo, Soepomo (1982). Javanese influence on Indonesian. Materials in
Languages of Indonesia, No. 7. Pacific Linguistics, Series D., No. 38.
Pollard, Carl, and Ivan Sag (1994). Head-driven phrase structure grammar. Chicago:
University of Chicago Press.
Postal, Paul (1971). Cross-over phenomena. New York: Holt, Rinehart, and Winston.
Postal, Paul (1972). On some rules that are not successive cyclic. Linguistic Inquiry 3: 211-222.
535
Postman, Whitney (2002). Thematic role assignment in Indonesian: A case study of
agrammatic aphasia. Doctoral dissertation, Cornell University.
Potsdam, Eric (2003). Evidence for Semantic Identity under Ellipsis from Malagasy,”
Proceedings of NELS 33, ed. Makoto Kadowaki and Shigeto Kawahara, 285-302.
University of Massachusetts, Amherst: GLSA.
Potsdam, Eric (2007) “Malagasy Sluicing and its Consequences for the Identity
Requirement on Ellipsis,” Natural Language and Linguistic Theory 25, 577-613.
Prince, Alan and Paul Smolensky (1993). Optimality theory: Constraint interaction in
generative grammar. Rutgers University Center for Cognitive Science.
Pritchett, Bradley (1991). Subjacency in a principle-based parser. Principle-based parsing:
Computation and psycholinguistics, ed. Robert Berwick, Steven Abney, and Carol
Tenny, 301-345. Dordrecht: Kluwer.
Provovac, Ljiljana (1998). Determiner phrase in a language without determiners. Journal
of Linguistics 34: 165-179.
Purwo, Bambang Kaswanti (ed.) (1989). Serpih-serpih telaah pasif bahasa Indonesia.
[Studies on Passives in Indonesian]. Yogyakarta: Penerbit Kanisius.
Rackowski, Andrea & Norvin Richards (2005). Phase edge and extraction: A Tagalog case
study. Linguistic Inquiry 36: 565–599.
Radford, Andrew (1990). Syntatcic theory and the acquision of English syntax. Oxford:
Backwell.
536
Radford, Andrew (2004). Minimalist syntax: Exploring the structure of English.
Cambridge: Cambridge University Press.
Rappaport, Gilbert (2001). Extraction from nominal phrases in Polish and the theory of
determiners. Journal of Slavic Linguistics 8: 159-198.
Reinhart, Tanya (1981). A second COMP position. Theory of markedness in generative
grammar, ed. Adrianna Belletti, Luciana Brandi, and Luigi Rizzi, 517-557. Pisa:
Scuola Normale Superiore.
Reinhart, Tanya (1991). Elliptic conjunctions-non-quantificational LF. The Chomskyan
turn, ed. Asa Kasher, 360-384. Oxford: Blackwell.
Reinhart, Tanya (1992). Wh-in-situ: An apparent paradox. In Proceedings of the Eighth
Amsterdam Colloquium, ed. Paul Dekker and Martin Stokhof, 483-491. University of
Amsterdam: LLC.
Reinhart, Tanya (1995). Interface strategies. OTS Working Papers in Linguistics. Research
Institute for Language and Speech, Utrecht University.
Reinhart, Tanya (1997). Quantifier scope: How labor is divided between QR and choice
function. Linguistic and Philosophy 20: 335-397.
Reinhart, Tanya (1998). WH-in-situ in the framework of the minimalist program. Natural
Language Semantics 6: 29-56.
Reinhart, Tanya (2006). Interface strategies: Reference-set computation. Cambridge, MA:
MIT Press.
537
Richards, Marc (2006). Weak pronouns, object shift, and multiple spell-out: Evidence for
phases at the PF-interface. Minimalist essays, ed. Cedric Boeckx, 160-181.
Amsterdam: John Benjamins.
Richards, Marc (2007). On feature inheritance: An argument from the phase
impenetrability condition. Linguistic Inquiry 38: 563-572.
Richards, Norvin (2001). Movement in language: Interactions and architectures. Oxford:
Oxford University Press.
Rizzi, Luigi (1990). Relativized minimality. Cambridge, MA: MIT Press.
Rizzi, Luigi (1997) The fine structure of the left periphery. Elements of grammar:
Handbook of generative syntax, ed. Liliane Haegeman, 281-337. Dorddrecht: Kluwer.
Robson, Stuart (2002). Javanese grammar for students. Victoria: Monash University Press.
Rodrigues, Cilene, Andrew Nevins, and Luis Vicente (2007). Cleaving the Interactions between
Sluicing and Preposition Stranding,” ms., Unicamp, Harvard, and University of California,
Santa Cruz. [To appear in Proceedings of Going Romance 20, Vrije Universiteit.]
Rogayah, Razak (1995). The syntax and semantics of quantification in Malay: A
government and binding approach. Doctoral dissertation. Universiti Sains Malaysia,
Penang, Malaysia.
Ross, John. R. (1967). Constraints on variables in syntax. Doctoral dissertation, MIT.
538
Ross, John (1969). Guess Who. Papers from the 5th Regional Meeting of the Chicago
Linguistic Society, ed. Robert Binnick, Alice Davison, Georgia Green, and Jerry
Morgan, 252-286. Chicago: Chicago Linguistic Society.
Rutkowski, Paweł (2002). Noun/pronoun asymmetries: Evidence in support of the DP
hypothesis in Polish. Jezikoslovlje 3: 159-170.
Rutkowski, Paweł and Hanna Maliszewska (To appear). On prepositional phrases inside
numeral expressions in Polish. Lingua.
Ruys, Eddy (1992). The scope of indefinites. Doctoral dissertation, Utrecht University.
Sabel, Joachim (2000). Partial wh-movement and the typology of wh-questions. Wh-scope
marking, ed. Uli Lutz, Gereon Müller & Arnim von Stechow, 409–446. Amsterdam:
John Benjamins.
Saddy, Douglas (1991). WH-scope mechanisms in Bahasa Indonesia. MIT Working
Papers in Linguistics 15: More Papers on Wh-Movement, ed. Lisa Cheng and
Hamida Demirdache, 183-218. Department of Linguistics and Philosophy: MITWPL.
Safir, Kenneth (1993). Perception, selection, and structural economy. Natural Language
Semantics 2: 47-70.
Saito, Mamoru and Keiko Murasugi (1990). N′-deletion in Japanese: A preliminary study.
Japanese/Korean Linguistics 1, ed. Hajime Hoji, 285-301. Stanford: CSLI
Publications.
539
Salles, Heloisa Maria Moreira-Lima (1997). Prepositions and the syntax of
complementation. Doctoral dissertation, University of Wales, Bangor
Sato, Yosuke (2007). P-stranding generalization and Bahasa Indonesia: A myth? Snippets
16: 17-18.
Sato, Yosuke (2008). Reduplication assymetries in Bahasa Indonesia and the organization
of the lexicon-syntax interface. Ms., University of Arizona, Tucson.
Sato, Yosuke (In press a). The distribution of the active voice morphology in Javanese and
vP phases. Proceedings of the 2007 Western Conference on Linguistics. University of
California, San Diego: Department of Linguistics.
Sato, Yosuke (In press b). The distribution of the active voice morphology in Bahasa Indonesia
and vP phases. In: Yoshiaki Kaneko, Akira Kikuchi, Daiko Takahashi, Yoshiki Ogawa,
and Etsuro Shima (eds.) The State of the Art in Linguistic Research, Interface of Form
and Meaning: Festschrift for Professor Masaru Nakamura. Tokyo: Kaitakusha.
Sato, Yosuke (In press c). Spelling-out prosodic domains: A multiple spell-out account.
InterPhases: Phase-Theoretic Investigations of Linguistic Interfaces, ed. Kleanthes
Grohmann. Oxford: Oxford University Press.
Sato, Yosuke (In press d). The denotation and morphosyntax of bare nouns in Javanese:
Against the nominal mapping parameter. Proceedings of the 2007 Mid-America
Linguistics Conference: Kansas Working Papers in Linguistics 30. University of
Kansas: Linguistic Graduate Student Association.
540
Sato, Yosuke (Forthcoming a). Sluicing in Bahasa Indonesia, P-stranding, and interface
repair. Proceedings of the Thirty-Fourth Annual Meeting of the Berkeley Linguistics
Society. Department of Linguistics: Berkeley Linguistics Society
Sato, Yosuke (Forthcoming b). The denotation and morphosyntax of NPs in Javanese: A
problem for the nominal mapping parameter. Proceeding of the First Meeting of the
Arizona Linguistics Circle. University of Arizona: Arizona Linguistics Circle.
Sato, Yosuke (In preparation). A choice function approach to English Jeopardy game
questions. Ms., University of Arizona, Tucson.
Sato, Yosuke and Bradley McDonnell (In press). Reduplication in Indonesian and the
lexicalist hypothesis. Proceedings of the Thirty-Third Annual Meeting of the Berkeley
Linguistics Society. University of California, Berkeley: Department of Linguistics
[with Bradley McDonnell]
Sato, Yosuke and Dwi Hesti Yuliani (In press). Wh-in-situ in Bahasa Indonesia and choice
function. Proceedings of the 2007 Mid-America Linguistics Conference: Kansas Working
Papers in Linguistics 30. University of Kansas: Linguistic Graduate Student Association.
Schmitt, Cristina and Alan Munn (1999). Against the nominal mapping parameter: Bare
nouns in Brazilian Portuguese. Proceedings of NES 29, ed. Pius Tamanji, Masako
Hirotani, and Nancy Hall, 339-353. University of Massachusets, Amherst: GLSA.
Schmitt, Cristina and Alan Munn (2002). The syntax and semantics of bare arguments in
Brazilian Portuguese. Linguistic Variation Yearbook 2: 185-216.
Seiter, William (1980). Studies in Niuean syntax. New York: Garland.
541
Sells, Peter (1984). Syntax and semantics of resumptive pronouns. Doctoral dissertation,
University of Massachusetts, Amherst.
Sie, Ing Djiang (1988). The syntactic passive in Bahasa Indonesia. Doctoral dissertation,
University of Amsterdam.
Siegel, Dorothy (1973). Topics in English morphology. Doctoral dissertation, MIT.
Sneddon, James (1996). Indonesian: A comprehensive grammar. Routledge: London.
Soh, Hooi-Ling (1998). Certain restrictions on A-bar movement in Malay. UCLA
Occasional Papers in Linguistics 21: Proceedings of the third and fourth meetings of
the Austronesian Formal Linguistics Association (AFLA), ed. Matthew Pearson, 295–
308. Department of Linguistics: UCLA.
Speas, Magaret (1990). Phrase structure in natural language. Dordrecht: Kluwer.
Speas, Margaret (1994). Null arguments in a theory of economy of projection. University of
Massachusetts Occasional Papers in Linguistics 17: Functional Projections, ed. Ellena
Benedicto and Jeffrey Runner, 179-208. University of Massachusets, Amherst: GLSA.
Spencer, Andrew (1991). Morphological Theory. Oxford: Blackwell.
Stepanov, Arthur (2001). Late adjunction and minimalist phrase structure. Syntax 4: 94-125.
Stepanović, Sandra (1999). Scrambling: Overt movement or base generation and LF
movement. Journal of Slavic Linguistics 7: 305-324.
542
Stepanović, Sandra (2008). P-stranding under sluicing in a non-P-stranding language?
Linguistic Inquiry 39:179-190.
Stowell, Timothy (1981). Origins of phrase structure. Doctoral dissertation, MIT.
Stowell, Timothy (1982). Conditions on reanalysis. MIT Working Papers in Linguistic 4:
Papers on Syntax, ed. Alec Marantz and Timothy Stowell, 245-269. Department of
Linguistics and Philosophy: MITWPL.
Stowell, Timothy (1989). Subjects, specifiers, and X-bar theory. Alternative conceptions of
phrase structure, ed. Mark Baltin and Anthony Kroch. 232-262. Chicago: University
of Chicago Press.
Suñer, Margarita (1988). The role of agreement in clitic-doubled constructions. Natural
Language and Linguistic Theory 6: 391-434.
Svenonius, Peter (2004). On the edge. Peripheries, ed. David Adger, Cecile de Cat and
George Tsoulas, 259-287. Dordrecht: Kluwer.
Szabolcsi, Anna (1987). Functional categories in the noun phrase. Approaches to
Hungarian, ed. Istvan Kenesei, 167-190. JATE: Szeged.
Szabolcsi, Anna (1994). The noun phrase. Syntax and Semantics 27: The syntatcic
structure of Hungarian, ed. Ferenc Kiefer and Katalin Kiss,170-274. San Diego, CA:
Academic Press.
Szabolcsi, A. and Frans Zwarts (1993). Weak islands and an algebraic semantics for scope
taking. Natural Language Semantics 1: 235-285.
543
Szczegielniak, Adam (2006). All sluiced up, but no alleviation in sight,” Ms., Harvard
University.
Takami, Ken-ichi (1988). Preposition stranding: Arguments against a syntactic analysis
and an alternative functional explanation. Lingua 76: 299-335.
Tanaka, Hidekazu (2007). Ellipsis under syntactic isomorphism. Ms., University of York.
Tenny, Carol (1994). Aspectual roles and the syntax-semantics interface. Dorecht: Kluwer.
Thornton, Rosalind (1995). Referentiality and wh-movement in Child English: Juvenile Dlinkuency. Language Acquisition 4, 139–175.
Torrego, Esther (1984). More effects of successive cyclic movement. Linguistic Inquiry 14:
361-365.
Toyoshima, Takashi (2004). Wh-Tideland. Paper presented at the Second International
Theoretical East Asian Linguistic Workshop. National Tsing Hua University, Taiwan.
June 12-13, 2004.
Travis, Lisa (1999). A syntactician’s view of reduplication. In: Carolyn Smallwood and
Catherine Kitto (eds.) Toronto Working Papers in Linguistics: Proceedings of the
Sixth Meeting of the Austronesian Formal Linguistics Association, 312-331. Toronto:
University of Toronto
Travis, Lisa (2001). The syntax of reduplication. In: Min-Joo. Kim and Uri Strauss (eds.)
Papers from the Thirty-First Annual Meeting of the Northeastern Linguistic Society.
455-469. Amherst, MA: GLSA.
544
Travis, Lisa (2003). Reduplication feeding syntactic movement. In: Sophie Burelle and
Stanca Somesfalean (eds.) Proceedings of the 2003 Annual Conference of the
Canadian Linguistic Association 236-247. Montreal: UQAM.
Trenkic, Danijela (2004). Definiteness in Serbian/Croatian/Bosnian and some implications
for the general structure of the nominal phrase. Lingua 114: 1401-1427.
Trugman, Helen (2005). Syntax of Russian DPs and DP-internal agreement phenomena.
Doctoral dissertation, Tel-Aviv Uiversity.
Tsai, Dylan. W.-T (1999). On economizing the theory of A-bar dependencies. New York:
Garland.
Uriagereka, Juan (1998) Rhyme and reason: An introduction to minimalist syntax.
Cambridge, MA: MIT Press.
Uriagereka, Juan (1999). Multiple spell-out. Working minimalism, ed. Samuel Epstein &
Norbert Hornstein, 251-282. Cambridge, MA: MIT press.
Uriagereka, Juan (2002). Derivations: Exploring the dynamics of syntax. London/New
York: Routledge.
Vainikka, Anne (1993/1994). Case in the devlopment of English syntax. Language
Acquisition 3: 257-324.
van Riemsdijk, Henk (1978). A case study in syntactic markedness: The binding nature of
prepositional phrases. Dordrecht: Foris.
545
van Riemsdijk, Henk (1998). Head movement and adjacency. Natural Language and
Linguistic Theory 16: 633-678.
Verhaar, Jolen (1989). Syntactic ergativity in contemporary Indonesian. Serpih-Serpih
Telaah Pasif Bahasa Indonesia [Studies on Passives in Indonesian], ed. Purwo,
Bambang Kaswanti, 200-281. Yogyakarta: Penerbit Kanisius.
Voskuil, Jan (2000). Indonesian voice and A-bar movement. Formal issues in
Austronesian linguistics, ed. Ileana. Paul, Vivianne Phillips, and Lisa Travis, 195-213.
Dordrecht: Kluwer.
Wagner, Michael (2005). Prosody and recursion. Doctoral dissertation, MIT.
Wang, Chyan-An (2006). Sluicing and Resumption. Paper presented at the 18th
Conference of the Northeast Linguistic Society. [To appear in Proceedings of NELS
37, University of Massachusetts, Amherst: GLSA]
Wasow, Thomas (1977). Transformations and the lexicon. Formal syntax, ed. Peter Culicover,
Thomas Wasow, and Adrian Akmajian, 327-360. New York: Academic Press.
Wasow, Thomas (1979). Anaphora in generative grammar. Ghent: E. Story-Scientia.
Watanabe, Akira (1992). Subjacency and S-structure movement of wh-in-situ. Journal of
East Asian Linguistics 1: 255-291.
Watanabe, Akira (2000). Feature copying and binding: Evidence from complementizer
agreement and switch reference. Syntax 3: 159-181.
546
Watanabe, Akira (2001). Wh-in-situ languages. The handbook of contemporary syntactic
theory, ed. Mark Baltin and Chris Collins, 203-225. Oxford: Blackwell.
Watanabe, Akira (2006). Functional projections of nominals in Japanese: Syntax of
classifiers. Natural Language and Linguistic Theory 24: 241-306.
Wexler, Kenneth and Rita Manzini (1987). Parameters and learnability in binding theory,
Parameter setting, ed. Thomas Roeper and Edwin Williams, 41-76. Dordrecht: Reidel.
Williams, Edwin (1981). On the notions “lexically related” and “head of a word.”
Linguistic Inquiry 12: 234-274.
Williams, Edwin (2007). Dumping lexicalism. In The Oxford Handbook of Linguistic
Interfaces, Gillian Ramchand and Charles Reiss (eds.), 353-381. Oxford: Oxford
University Press.
Willim, Ewa (1998). On the DP-hypothesis in Polish, an article-less language. Projections
and mappings: Studies in syntax, ed. Piotr Stalmaszczyk, 137-158. Lublin: Folium.
Willim, Ewa (2000). On the grammar of Polish nominals. Step by step: Essays on
minimalist syntax in honor of Howard Lasnik, ed. R. Martin, D. Michaels & J.
Uriagereka, 319-346. Cambridge, MA: MIT Press.
Winter, Yoad (1997). Choice functions and the scopal semantics of indefinites. Linguistics
and Philosophy 20: 399-467.
Wolff, John, Dede Oetomo, and Daniel Fietkiewicz (1992). Beginning Indonesian through
self-instruction, vol. 2-3. Ithaca, NY.: Cornell University Southeast Asia Program.
547
Zagona, Karen (2007). “Adjunct theta-roles” and the configurational determination of roles.
Phrasal and clausal architecture: Syntactic derivation and interpretation, ed. Simin
Karimi Vida Samiian and Wendy K. Wilkins, 396-411. Amsterdam: John Benjamins.
Zlantić, Larisa (1997). The structure of the Serbian noun phrase. Doctoral dissertation,
University of Texas, Austin.
Zubizarreta, Maria-Luisa (1998). Prosody, focus, and word order. Cambridge, MA: MIT Press.
Zwart, Jan-Wouter (1993). Verb movement and complementizer agreement. MIT Working
Papers in Linguistics 18: Papers on Case and Agreement I, ed. Jonathan Bobaljik
and Colin Phillips, 297-340. Department of Linguistics and Philosophy: MITWPL.
Zwart, Jan-Wouter (1997). Morphosyntax of verb movement. Dorecht: Kluwer.
Was this manual useful for you? yes no
Thank you for your participation!

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

Download PDF

advertisement