. LIBRARY
Morgan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 C'JCIFIC/DataouepGS-p. 15

 

 

ON THE ACQUISITION OF JAPANESE PASSIVES BY ENGLISH SPEAKERS
By

Masahiro Hara

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHYLOSOPHY

Department of Linguistics and Germanic, Slavic, Asian, and African Languages

2002

ABSTRACT
ON THE ACQUISITION OF JAPANESE PASSIVES BY ENGLISH SPEAKERS
By

Masahiro Hara

This study investigates knowledge of Japanese passives that English-speaking
learners develop and it addresses the questions of what aspects of Japanese passives pose
learning difficulties for them, and why. Pedagogically, it seems essential to understand
what learners know and do not know about passives to facilitate their learning.
Theoretically, it is of significance to explore whether or not they are able to attain
knowledge of the m’ passive versus the ni yotte passive contrasts, which underlie
overarching characterizations of Japanese passives.

Following Hoshi (1994a; 1999) and Kuroda's (1979) analysis of Japanese
passives, the present study classifies them syntactically into three types, the ni direct
passive, the ni indirect passive, and the ni yotte passive; semantically, they are grouped
into two classes, the m’ passive (including both direct and indirect passives) and the m“
yotte passive. The subjects of the present study include 81 English-speaking learners
whose proficiency levels range from intermediate to low and highly advanced, as well as
52 native speakers of Japanese who constitute two distinct groups in terms of the
presence or absence of linguistic training received as educational background. The data
consist of five-point scale grammaticality judgments of 84 test sentences, 62 of which

pertain to either syntactic or semantic properties of Japanese passives and the remaining

22 distractor sentences.

 

The study first shows that notwithstanding highly delicate semantic differences
between closely related passive sentences, the native speaker group without linguistic
training made grammaticality judgments of the test sentences in nearly full agreement
with the theory predictions. Yet, somewhat surprisingly, the native speaker group with
such training only agreed with the predictions approximately 70% at best.

The results of the analysis of the non-native speaker subjects’ data Show the
following: 1) Syntactically, the ni direct passive was learned best, followed by the ni
yotte passive; the ni indirect passive was most difficult to learn; 2) At certain stage(s) of
learning passives, learners appear to have heavily relied on the ni-marked O-role
transmitted phrase to judge the grammaticality of passive sentences; 3) The incorrect
analysis of the m‘ indirect passive as involving passivization in its lower clause seems to
have persisted even after learners attained a highly advanced proficiency level in
Japanese; 4) Semantically, English-speaking learners were capable of learning the ni
passive versus ni yotte passive cardinal contrasts, including the ones in which input
indicating the grammaticality contrasts between the ni direct and the ni yotte passives
appears not to be readily available to them; and 5) The only semantic area that even
highly advanced learners failed to fully acquire involves knowledge that the passive verb
rare of the m' direct passive may intrinsically invoke an adversative connotation. On the
basis of these findings, the (highly advanced) non-native Speaker subjects’ knowledge
representation of the Japanese passive morpheme rare is characterized as almost identical
to that held by native speakers except for the optionality analysis of the ni indirect

passive. Finally, pedagogical implications of these findings are considered.

COPyright by
Masahiro Hara
2002

 

For my wife,

Katherine

ACKNOWLEDGMENTS

I would like to express my deep and sincere gratitude to the following people for
their help with this dissertation. First and foremost, I would like to thank my major
advisor, Dennis Preston. His encouragement, support, and insightful academic advice
enabled me to complete this work. He was an exemplary academic advisor in all respects
throughout my graduate studies. I would also like to thank the other members of my
thesis committee, Susan Gass, Alan Munn, and Yen-Hwei Lin. Their comments and
criticisms were invaluable in helping me sharpen data analyses and in improving the
clarity of the work. I am grateful to Patricia Lunn for her comments, too.

I am especially indebted to those people both in the US and Japan who were so
kind and generous as to cooperate me with a yearlong data collection effort. I would like
to express my heartfelt thanks to the following people who solicited volunteers for the
non-native speakers’ data collection: Kumiko Aso, Shoko Emori, Emi Fujiwara, Leopold
Hanami, Tetsuo Harada, Susan Kubota, David Mills, Manabu Mizobe, Hiroyuki Oshita,
Ken'ichi Ujie, Suwako Watanabe; special thanks are due to Mari Noda, who went so far
as to ask her departmental colleagues for cooperation. Sincere gratitude also goes to the
following people who helped me with the native speakers’ data collection: Yasuaki Abe,
Osamu Hashimoto, Kiyoshi Ishikawa, Masaru Nakamura, Yukio Otsu, Izumi Saita,
Takashi Sugimoto, and Kensaku Yoshida. Needless to say, wholehearted thanks are due
to those people in the US and Japan who generously participated in the research by filling
out a rather long questionnaire: the non-native speaker subjects were students from the

George Washington University, the Ohio State University, the Ohio University, Portland

vi

State University, Truman State University, the University of Florida, the University of
Oregon, the University of Michigan at Ann Arbor, the University of Pittsburgh, the
University of Texas at Austin, Washington and Lee University, and Washington
University; and the native speaker subjects were students from Hosei University, Keio
University, Nagoya Univeristy, Nanzan University, Osaka University of Foreign Studies,
Seikei University, Sophia University, Tohoku University, the University of Tokyo, and
the University of Tsukuba. I am also grateful to Hyun-Joo Kim for her assistance in
statistical analyses, and to Mimi Kato, Lily Lok, and Luke Walczewski for their help with
data coding and entry.

I am thankful to my family, particularly my parents, Tatsuo and Teiko Hara, for
their love and support throughout my graduate work. Special thanks go to my sister,
Tomoko Hara, who helped me immensely with the data collection effort in Japan.
Finally, for her personal and academic support as well as for editing my work, I would

like to thank my wife, Katherine, to whom this thesis is dedicated.

vii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. xii
LIST OF FIGURES ............................................................................... xiv
INTRODUCTION ................................................................................... 1
CHAPTER 1
THEORETICAL AND EMPIRICAL BACKGROUND ....................................... 8
1.1. Introduction ................................................................................ 8
1.2. Characterizations of Japanese passives ................................................. 8
1.2.1. Semantic analysis of Japanese passives ....................................... 11
1.2.1.1. Introduction ................................................................. 11
1.2.1.2. Adversity in the ni direct passive ........................................ 11
1.2.1.3. Ni versus ni yotte contrasts ................................................ 13
1.2.1 .3.]. Perfective versus non-perfective readings ........................ 13
1.2.1 .3.2. Verb-induced viewpoint differences .............................. l7
1.2.1.3.3. Adversity in the ni indirect passive ............................... 20
1.2.2. Syntactic analysis of Japanese passives ....................................... 23
1.2.2.1. Introduction ................................................................. 23
1.2.2.2. Syntactic evidence for the difference between the m’ direct and the ni
yotte passives ............................................................... 23
1.2.2.3. Syntactic representations of the ni yotte passive ........................ 25
1.2.2.4. Syntactic representations of the m’ indirect passive ................... 27
1.2.2.5. Syntactic representations of the ni direct passive ...................... 28
1.3. Acquisition studies on Japanese passives in SLA ................................... 32
1.4. Adversative readings of passives by native speakers of Japanese ................. 43
1.5. Issues on native speakers’ grammaticality judgment ............................... 46
1.6. Use of grammaticality judgment tasks in SLA research ............................ 54
1.6.1. Advantages of grammaticality judgment tasks ................................ 54
1.6.2. Validity and reliability issues of grammaticality judgment tasks in SLA
research ............................................................................ 55
1.6.2.1. Validity issues of L2 grammaticality judgment ........................ 55
1.6.2.2. Reliability issues of L2 grammaticality judgment ...................... 61
CHAPTER 2
PREDICTIONS .................................................................................... 64
2.1. Predictions for native speakers of Japanese .......................................... 64
2.2. Predictions for English-speaking non-native speakers of Japanese ............... 65
CHAPTER 3

viii

METHOD ............................................................................................ 70

3.1. Subjects .................................................................................... 70
3.2. Materials .................................................................................. 72
3.2.1. Background information questionnaire ........................................ 72
3.2.2. Japanese language proficiency test ............................................. 73
3.2.3. Vocabulary test .................................................................... 74
3.2.4. Grammaticality judgment test .................................................... 75
3.2.5. Illustrations of test sentences .................................................... 78
3.2.5.1. Testing the knowledge of syntactic properties of Japanese passives 78
3.2.5.1 .1. The ni yotte passive ................................................. 79
3.2.5.1.2. The ni direct passive ................................................ 8O
3.2.5.1.3. The ni indirect passive .............................................. 81
3.2.5.2. Testing the knowledge of semantic properties of Japanese passives 83
32.5.2.1. An adversative reading of the m’ direct and indirect passives ..83

3.2.5.2.2. A grammaticality contrast in perfective versus non-perfective
readings ................................................................. 84

3.2.5.2.3. Unavailability of an adversative reading of the ni yotte
passive ................................................................. 86
3.2.5.2.4. Verb-induced viewpoint differences .............................. 87
3.2.5.3. Distractors ................................................................... 90
3.3. Procedures ................................................................................ 91
3.4. Data coding .............................................................................. 93
CHAPTER 4

RESULTS ........................................................................................... 95
4.1. Introduction .............................................................................. 95
4.2. Judgments by native speakers of Japanese ........................................... 95
4.2.1. Conformity to the theory predictions .......................................... 96
4.2.2. Group judgment consistency .................................................... 100
4.2.3. Judgments of distractor sentences ............................................. 103
4.2.4. Summary of native speakers’ grammaticality judgments .................. 104
4.2.5. Consideration of some problematic test sentences .......................... 105
4.3. Judgments by English-speaking learners of Japanese ............................. 106
4.3.1. Judgments of three syntactic types of passives .............................. 107
4.3.1.1. Grammatical judgments of three syntactic types of passives ....... 107
4.3.1.2. Ungrammatical judgments of three syntactic types of passives 112
4.3.1.3. Judgments of improper syntactic operations .......................... 122
4.3.1.4. Summary of judgments of three syntactic types of passives ........ 129
4.3.2. Judgments of semantic properties of passives .............................. 131

4.3.2.1. Judgments of three types of passives for an adversative reading 131
4.3.2.2. Judgments of adversative versus non-adversative readings 136
4.3.2.2.]. An adversative versus non-adversative reading distinction
within each type of passive ........................................ 137
4.3.2.2.2. Judgments of adversative readings across the m' passives 144
4.3.2.2.3. Summary of judgments of passives for adversative and non-

ix

adversative readings ................................................ 149

4.3.2.3. Judgments of ni passive versus ni yotte passive contrasts ........... 149
4.3.2.3.1. Judgments of perfective versus non-perfective readings 150
4.3.2.3.1.1. Judgments of ta-marked perfective versus past
readings ....................................................... 150
4.3.2.312. Judgments of I'm-marked perfective versus progressive
readings ....................................................... 155
4.3.2.3.1.3. Summary of judgments of perfective versus non-
perfective readings ........................................... 161
4.3.2.3.2. Judgments of the ni yotte indirect and direct passive with an
adversative reading ................................................. 161
4.3.2.3.3. Judgments of verb-induced viewpoint differences ............ 169
4.3.2.331. Judgments of Japanese native versus Sino-Japanese verb
contrasts ....................................................... 170
4.3.2.332. Judgments of affective and unaccusative verb
contrasts ...................................................... 176
4.3.2.333. Summary of judgments of verb-induced viewpoint
differences ..................................................... 181
CHAPTER 5
DISCUSSION .................................................................................... 183
5.1. Introduction ............................................................................. 183
5.2. Linguistically naive and sophisticated native speakers’ grammaticality
judgments ............................................................................... 183
5.3. Non-native speakers’ grammatical and ungrarnmatical judgments of three
syntactic types of passives ............................................................ 185
5.4. Non-native Speakers’ judgments of perfective versus non-perfective readings 194
5.5. Non-native speakers’ judgments of adversative readings ........................ 200
5.6. Native speaker control group’s non-rejection of the m’ yotte indirect passive with
a transitive verb carrying an adversative reading .................................. 205
5.7. Judgments of verb-induced viewpoint differences ................................. 208
CHAPTER 6
CONCLUSION ................................................................................... 215
ENDNOTES ....................................................................................... 231
APPENDECES
Appendix A: Linguistically sophisticated NS subject background information .. .. 243
Appendix B: Linguistically na'r've NS subject background information ............. 245
Appendix C: NNS subject background information .................................... 247
Appendix D: NS subject background information questionnaire ..................... 251

Appendix E: NNS subject background information questionnaire ................... 253

Appendix F: NNS proficiency test ........................................................ 255
Appendix G: NNS vocabulary test ......................................................... 260
Appendix H: Instruction sheet on the grammaticality judgment test ................. 262
Appendix 1: Sample grammaticality judgment sheet ................................... 265
Appendix J: List of test sentences ......................................................... 267
Appendix K: Grammaticality judgment disagreements of linguistically naive N83
with a breakdown of syntactic and semantic sentence types .......... 278

Appendix L: Grammaticality judgment disagreements of linguistically sophisticated
NSs with a breakdown of syntactic and semantic sentence types 280
Appendix M: Grammaticality judgment inconsistencies of the linguistically naive NS
group with a breakdown of syntactic and semantic sentence types .. 282
Appendix N: Grammaticality judgment inconsistencies of the linguistically
sophisticated NS group with a breakdown of syntactic and semantic
sentence types ............................................................... 284

BIBLIOGRAPHY ................................................................................ 286

xi

LIST OF TABLES

Table 1: Syntactic classification of Japanese passives ........................................ 32
Table 2: Subjects’ percentage choice of m’ yotte and ni ....................................... 42
Table 3: Passive sentence types syntactically or semantically manipulated with their
grammaticality status ................................................................... 90
Table 4: Grammaticality judgment disagreements of linguistically naive NSs ............ 98

Table 5: Grammaticality judgment disagreements of linguistically sophisticated NSs 99
Table 6: Grammaticality judgment inconsistency of linguistically naive NS group 102

Table 7: Grammaticality judgment inconsistency of linguistically sophisticated NS

group ..................................................................................... 102
Table 8: Grammatical judgments of three syntactic types of passives ..................... 109
Table 9: Grammatical vs. ungrammatical judgments of the m’ direct passive ............ 113
Table 10: Grammatical vs. ungrammatical judgments of the m” yotte passive ............ 116
Table 11: Grammatical vs. ungrammatical judgments of the m’ indirect passive 1 19
Table 12: The ni direct and indirect passives without 0-role suppression 124
Table 13: The ni yotte passive without 0-role suppression or case absorption ........... 125
Table 14: The ni indirect passive unpassivized and passivized ............................ 127
Table 15: Three types of passives for an adversative reading .............................. 133

Table 16: The ni indirect passive with intransitive verbs with adversative and non-
adversative readings .................................................................. 138

Table 17: The ni indirect passive with transitive verbs with adversative and non-
adversative readings .................................................................. 141

Table 18: The m“ direct passive with adversative and non-adversative readings 143

Table 19: The ni indirect passive with intransitive plus transitive verbs and the m' direct
passive with adversative readings ................................................... 146

xii

Table 20: Ni direct and ni yotte passives with past and perfective readings .............. 152
Table 21: Ni direct and ni yotte passives with progressive and perfective readings 157

Table 22: Ni and ni yotte indirect passives with intransitive verbs with adversative

readings ................................................................................. 162
Table 23: Ni and ni yotte indirect passives with transitive verbs with adversative

readings ................................................................................. 165
Table 24: Ni and ni yotte direct passives with adversative readings ....................... 167

Table 25: The ni direct passive with Japanese native verbs and the ni yotte passive with
Sino-Japanese verbs .................................................................. 171

Table 26: The ni yotte passive with Sino-Japanese verbs and with Japanese native

verbs .................................................................................... 175
Table 27: Ni direct and ni yotte passives with affective verbs .............................. 178
Table 28: Unaccusative verbs .................................................................. 181
Table 29: NNS subjects’ characterizations of the three types of Japanese passive

morphology rare ..................................................................... 226
Table 30: Linguistically sophisticated NS subject background information ............. 244
Table 31: Linguistically naive NS subject background information ...................... 246
Table 32: NNS subject background information ....................................... 248-250

Table 33: Grammaticality judgment disagreements of linguistically naive N88 with
a breakdown of syntactic and semantic sentence types .......................... 279

Table 34: Grammaticality judgment disagreements of linguistically sophisticated N83
with a breakdown of syntactic and semantic sentence types ................... 281

Table 35: Grammaticality judgment inconsistencies of the linguistically naive NS group
with a breakdown of syntactic and semantic sentence types ................... 283

Table 36: Grammaticality judgment inconsistencies of the linguistically sophisticated NS
group with a breakdown of syntactic and semantic sentence types ............ 285

xiii

LIST OF FIGURES

Figure 1: Semantic classifications of Japanese passives ...................................... 22
Figure 2: Three syntactic types of passives ................................................... 109
Figure 3: The ni direct passive (GR vs. UG) ................................................. 114
Figure 4: The ni yotte passive (GR vs. UG) ................................................... 117
Figure 5: The ni indirect passive (GR vs. UG) ................................................ 120
Figure 6: Three types of passives for an adversative reading .......... - .................... 133
Figure 7: The ni indirect passive with intr. verbs with +/- adversative readings . .. 139
Figure 8: The ni indirect passive with tr. verbs with +/- adversative readings ........... 141
Figure 9: The ni direct passive with +/- adversative readings .............................. 143
Figure 10: The ni indirect passive with intr./tr. verbs and ni direct passive with
adversative readings ................................................................. 147
Figure 11: Ni direct and ni yotte passives with past and perfective readings ............. 153

Figure 12: Ni direct and ni yotte passives with progressive and perfective readings 158
Figure 13: Ni and ni yotte indirect passives with intr. verbs with adversative readings .163
Figure 14: Ni and ni yotte indirect passives with tr. verbs with adversative readings 165
Figure 15: Ni and ni yotte direct passives with adversative readings ...................... 167

Figure 16: The m’ direct passive with Japanese native verbs and the ni yotte passive with
Sino-Japanese verbs ................................................................. 172

Figure 17: The ni yotte passive with Sino-Japanese verbs and with Japanese native
verbs ................................................................................... 175

Figure 18: Ni direct and ni yotte passives with affective verbs ............................. 179

xiv

INTRODUCTION

This study investigates knowledge of Japanese passives that English-speaking
learners develop and addresses the question of what aspects of Japanese passives pose
learning difficulties for them, and why. While it has been long observed that learners of
Japanese, particularly English-speaking learners, have great difficulty learning Japanese
passives, second language acquisition (SLA) research has scarcely examined learners’
acquisition of them. In order to facilitate learning, it is surely important to understand
what English-speaking learners know and do not know about passives. Theoretically, it
is of significance to explore whether or not they are able to attain knowledge of the ni
passive versus the ni yotte passive contrasts which underlie overarching characterizations
of Japanese passives. The attainment of this knowledge is critical particularly because,
given that input for certain (especially semantic) properties of Japanese passives appears
not to be readily available to learners, deduction from the cardinal distinction between the
ni and ni yotte passives seems to be the only path through which to learn such properties.
Toward these ends, following Hoshi (1994a; 1999) and Kuroda's (1979) analysis of
Japanese passives, the present study examines grammaticality judgments that English-
speaking learners gave to various types of passive sentences in order to explore their
knowledge of not only syntactic but also semantic properties of Japanese passives.

Chapter 1 first provides the theoretical bases of the present study. Specifically, it
presents Hoshi (1994; 1999) and Kuroda's (1979) analysis of Japanese passives. A
typical classification of Japanese passives holds that there are two types of passives, i.e.,

the ni direct passive and the ni indirect passive, and that the ni yotte passive is

characterized as a simple stylistic variant of the former type of passive. Hoshi and
Kuroda, however, proposed the following analysis of Japanese passives: Syntactically,
they may be classified into three types, the ni direct passive, the ni indirect passive, and
the ni yotte passive, on the basis of types of syntactic operations involved in each of the
three, i.e., the presence or absence of passivization processes and of ‘affectee’ 0-role
assignment. On the other hand, they may be grouped semantically into two classes, the ni
passive (including both direct and indirect passives above) and the ni yotte passive,
depending on whether or not each type of passive carries an ‘affective’ connotation.

This chapter then provides empirical background on the following issues central
to the present study: 1) review of previous studies on the acquisition of Japanese passives
in SLA research, showing that the previous studies found that learners of Japanese had
great difficulty using passive sentences in various writing production tasks. Due to the
nature of the tasks employed in these studies, however, they failed to address in detail the
question of why Japanese passives were difficult for their subject learners; and 2) a report
on how native speakers (N Ss) of Japanese perceive an adversative connotation in various
types of ni passive sentences, demonstrating that they read the adversative connotation
predominantly in the ni indirect passive and somewhat less in the m' direct passive
irrespective of connotation types of verbs used in them and the passive subject animacy
versus inanimacy contrast. Then, the following two issues concerning methodology are
considered: 1) differences in grammaticality judgment performances between NSs with
and without linguistic training; and 2) the validity and reliability for use of
grammaticality judgment in SLA research. This review demonstrates that second

language (L2) learners’ grammaticality judgments can provide valid and reliable data for

L2 research when we exercise care in experimental design and caution in analysis of
data.

In chapter 2, research questions are formulated both in syntactic and semantic
terms. Syntactically, the present study first investigates which type of passive is learned
better than the other(s) and then looks into what is difficult for English-speaking learners
in light of passivization operations involved in the three types of passives: the ni direct,
the ni indirect, and the ni yotte passive. Specifically, the aim is to explore their
knowledge of the presence or absence of external 0-role suppression and accusative case
absorption as effected in each type of passive by means of examining their knowledge of
case marking phenomena resulting from the passivization processes. In semantic areas,
the present study investigates English-speaking learners’ knowledge of the ni passive
versus the ni yotte passive contrasts which arise from their passive subject status
difference, i.e., ‘affectee’ subject versus non-‘affectee’ subject, respectively. Particularly,
the aim is to probe their knowledge of semantic contrasts arising in the following
conditions: 1) an adversative reading is available, indeed required in certain contexts, for
the ni (direct and indirect) passive while it is unavailable for the m’ yotte passive; 2)
grammaticality contrasts result in the ni direct passive with an inanimate passive subject
depending on its perfective versus non-perfective readings , whereas no such contrast
ensues in the ni yotte passive; and 3) the ni direct passive is suitable to express personally
involved situations whereas the ni yotte passive is appropriate for neutral, objective
descriptions.

Chapter 3 presents the research design of the present study. The subjects of the

present study include 81 English-speaking learners whose proficiency levels range from

intermediate to low and highly advanced, as well as 52 N88 of Japanese who constitute
two distinctive groups in terms of the presence or absence of linguistic training received
as educational background. The data consist of five-point scale grammaticality
judgments of 84 test sentences, 62 of which pertain to either syntactic or semantic
properties of Japanese passives as described above (32 grammatical and 30
ungrammatical) together with 22 distractor sentences (12 grammatical and 10
ungrammatical).

Since Kuroda's (1979) analysis of Japanese passives hinges on highly subtle
semantic differences in closely related passive sentences, it is imperative to first
demonstrate that ordinary NSS of Japanese would indeed judge the grammaticality of
passive sentences differing in semantic nuances as his theory predicts. In view of
judgments of such subtlety, it might be advisable to take into consideration the possibility
of grammaticality judgment differences between N85 with linguistic training and those
without it. There are two conflicting views in the literature on the effects of linguistic
training on NS grammaticality judgment performance: one holds that linguistic training
makes a native speaker who has received it capable of recognizing grammatical nuances
which may go unnoticed by a native speaker who has not, whereas the other view states
that a native speaker with linguistic training develops an idiosyncratic sense of
grammaticality only remotely resembling that held by a native speaker without such
training.

Chapter 4 presents analyses of the grammaticality judgment data (both N83’ and
English-speaking learners’). Chapter 5 then offers interpretations and considerations of

implications of the findings obtained in the previous chapter. The present study first

shows that notwithstanding highly delicate semantic differences between closely related
passive sentences under investigation, the NS group without linguistic training made
grammaticality judgments of the test sentences (both syntactic and semantic) in nearly
full agreement with Hoshi (1994; 1999) and Kuroda's (197 9) predictions. Yet, somewhat
surprisingly, the NS group with such training only agreed with the predictions
approximately 70% at best. Thus, in judgment of the passive sentences in question, these
two groups behaved quite differently, e. g., with the former group performing in favor of
the theory predictions.

The results of the analysis of the learner subjects’ data provide the information to
unravel how Japanese passives were represented in English-speaking learners’
knowledge and thus identify what aspects of passives posed learning difficulties for them,
and why. Furthermore, the findings evince that they learners able to attain knowledge of
the critical semantic contrasts between the m’ passive and the ni yotte passive. Specific
findings include the following: 1) Syntactically, the ni direct passive was learned best,
followed by the ni yotte passive; the ni indirect passive was most difficult to learn; 2) At
certain stage(s) of learning passives, learners appear to have heavily relied on the ni-
marked 0-role transmitted phrase to judge the grammaticality of passive sentences, as in
first language acquisition (FLA) (Fox & Grodzinsky, 1998; Sano, Endo, & Yamakoshi,
2001); 3) The incorrect analysis of the ni indirect passive as involving passivization in its
lower clause seems to have persisted even after learners attained a highly advanced
proficiency in the target language; 4) Semantically, English-speaking learners were
capable of learning the ni versus ni yotte passive contrasts as set forth in the above

depicted conditions, including the ones in which input indicating the grammaticality

contrasts between the ni direct and the ni yotte passives appears to be unavailable to
them, i.e., the perfective versus non-perfective reading contrasts. (Not until they reached
a highly advanced proficiency level, however, did they achieve knowledge of all these
contrasts); and 5) The only semantic knowledge that even highly advanced learners failed
to fully acquire involves recognition of the incompatibility between the ni yotte passive
and an adversative reading. In view of the important role the ni phrase apparently plays
in semantic interpretations, this implies that learners had serious difficulty learning that
the passive verb rare of the ni direct passive may intrinsically invoke an adversative
connotation.

Finally, Chapter 6 summarizes the present study and then considers pedagogical
implications of its findings. Japanese passives are indeed very difficult to learn, not only
in terms of their use as observed in the literature, but also in acquiring knowledge of their
various intricate properties; the intermediate learners’ grammaticality judgment data
indicate their slight knowledge of Japanese passives. Yet the highly advanced learners’
judgment data show that they were able to learn the syntactic Operations involved in three
types of passives, as well as the vital ‘affectee’ versus non-‘affectee’ subject contrast
between the m’ and ni yotte passives, as they became more proficient in the target
language. There are, however, two exceptions which seem to have escaped their
learning: 1) that the ni indirect passive does not involve syntactic passivization; and 2)
that the ni direct passive may intrinsically invoke adversity. On the basis of these
findings, the (highly advanced) NNS subjects’ knowledge representation of the Japanese

passive morpheme rare is characterized as almost identical to that held by NSs with the

exception of NNS subjects’ analysis of the ni indirect passive as involving passivization
processes with optional accusative case absorption.

Finally, pedagogical implications are as follows: it is important to explore ways to
help English-speaking learners to learn these syntactic and semantic properties of
passives. It is also imperative to find out how to assist them in learning those two areas

that even highly advanced learners failed to learn.

CHAPTER 1

THEORETICAL AND EMPIRICAL BACKGROUND

1.1. Introduction

This chapter provides both the theoretical and empirical bases of the present
study. First, an analysis of Japanese passives, adopted in the present study, is presented.
Specifically, after typically observed characterizations of them are reviewed, a more
detailed semantic examination of them is given. Then, the syntactic analysis consistent
with their semantic properties is summarized. Second, empirical background on issues
central to the present study is provided. To begin with, previous studies on the
acquisition of Japanese passives in SLA research are reviewed, followed by a report on
how NSS of Japanese perceive them in light of adversativeness. Then, the following two
issues concerning methodology are considered: 1) differences in grammaticality
judgment performances between linguists and non-linguists; and 2) the validity and

reliability of use of grammaticality judgments in SLA research.

1.2. Characterizations of Japanese passives

Japanese passives may be classified into three types: the ni direct passive, the ni
indirect passive, and the ni yotte passive. Typical characterizations of these three types of
Japanese passives are illustrated as follows.

First of all, many Japanese passive sentences seem structurally parallel to their
English counterparts. Specifically, the object of the active sentence (1a) below appears

as the subject of the corresponding passive sentence (lb) as given below:

(1) a. sensee-ga J ohn-o sikat-ta
teacher-NOM -ACC scold-Pst

‘the teacher scolded John’
b. John—ga sensee-ni sikar-are-ta.
-NOM -by scold-Pass-Pst

‘John was scolded by the teacher’1

Following Howard & Niyekawa—Howard (1976), this apparent English passive-like type
of the Japanese passive is referred to as the ‘direct’ passive.

There is, however, another kind of Japanese passive which demonstrates a
striking departure from the English one. For one thing, this second type of Japanese
passive may contain one more noun phrase (NP) than its closest active counterpart.
Stated differently, the subject of this type of passive does not bear any apparent
grammatical relation with its closest active counterpart. Unlike the paired active—passive
sentences of (1) above, there is no active sentence directly corresponding to either of the
passive examples of (2a) and (2c) below. (The closest active counterparts to (2a) and
(2c) are given in (2b) and (2d), respectively.) The additional NP represents the referent
adversely affected by the action or event expressed in the remainder of the sentence

(Howard & Niyekawa-Howard, 1976) as shown below:

(2) a. John-ga sensee-ni kodomo-o sikar-are-ta
-NOM teacher-by child-ACC scold-Pass-Pst
‘John was adversely affected by the teacher’s scolding his child’

b. sensee-ga kodomo-o sikat—ta
teacher-NOM child-ACC scold-Pst
‘the teacher scolded a child’

c. John-ga ame-ni hur-are-ta
-NOM rain-by fall-Pass-Pst
‘John was adversely affected by rain falling on him’

d. ame-ga hut-ta

rain-NOM fall-Pst

‘rain fell’
The passive sentence (2a) above expresses that John (the additional NP) was unfavorably
affected, i.e., upset or embarrassed in this particular case, because his child had been
scolded by the teacher. Similarly, the passive sentence (2c) above states that John (the
extra NP) was adversely affected in some sense, e. g., was in trouble or was upset because
it had rained (which perhaps made him cancel or postpone something he had planned).
For another thing, (20) demonstrates that even intransitive verbs may be passivized in this
type of passive. This second type is referred to as the ‘indirect’ passive.

There are thus two types of ni passives: the first type is the m” ‘direct’ passive as
in (1b), and the second is the ni ‘indirect’ passive such as (2a) and (2c). (See also Kuno,
1973; 1983; McCawley, 1972; Shibatani, 1990 and references cited there for detailed
characterizations of these two types of Japanese passives.)

Lastly, the following is an instance of the m' yotte2 passive along with its related

active counterpart:

(3) a. Fermat no teiri-ga J ohn-ni yotte syoomei-s-are-ta
Fermat’s theorem-NOM -by prove-Pass-Pst
‘F errnat’s Theorem was proved by John’
b. John—ga Fermat no teiri-o syoomei-si-ta
-NOM -ACC prove-Pst
‘John proved Fermat’s Theorem’ (Kuroda, 1979, p. 320)

When the paired sentences of (3a)-(3b) above are compared with those of (la)-(1b), the
ni yotte passive may appear to be identical with the ni ‘direct’ passive simply with the ni

yotte phrase replacing the ni phrase. Yet these two types of passives are in fact both

10

 

 

semantically and syntactically different. These differences are discussed in the

subsequent sections.

1.2.]. Semantic analysis of Japanese passives
1.2.1.1. Introduction

In this section, the three types of Japanese passives introduced in the previous
section are looked at in more detail. As overviewed in the previous section, the m‘ yotte
passive is often treated as a simple stylistic variant of the ni direct passive. Inoue (1976),
however, offered a different analysis, observing that there are cases where the ni yotte
passive is allowed while the ni direct passive is not. Stimulated by her insightfirl
observation on the differences between the two, Kuroda (1979) proposed the following
distinction between them: the ni direct passive carries an “affective” connotation whereas
the m' yotte passive does not. In examining the semantic properties of ni direct and ni
yotte passives in this section, this crucial semantic concept of the ni direct passive, e.g.,
the “affective” connotation, is elaborated on. Finally, the semantic analysis of the ni

‘direct’ passive is extended to the ni ‘indirect’ passive.

1.2.1.2. Adversity in the ni direct passive

One semantic instantiation of “affective connotation” in the ni direct passive, i.e.,
its adversative reading, is examined below. Although, initially, the ni direct passive has
been exemplified as apparently similar in structure to the English passive (see (1)), the
former may carry an adversative connotation whereas the latter does not. According to

Kuroda (1965), “the Japanese passive sentence is, in principle, not neutral as it is in

11

English, but carries an implication of disadvantage for the subject” (p. 160). This
unfavorable connotation in the ni direct passive arises even when there is no such
implication in its active counterpart. Compare the differences in meaning below between
the active and passive sentences in Japanese ((4a) and (4b)), on the one hand, and

between the Japanese and English passive sentences ((4b) and (4c)), on the other:

(4) a. butyoo-ga Satoo-o soomu-ka-ni mawasi-ta
director-NOM Sato-ACC general-affairs section-to transfer-Pst
‘the director transferred Sato to the general-affairs section’

b. Satoo-ga butyoo-ni soomu-ka-ni mawas-are-ta
Sato-NOM director-by general-affairs section-to transfer-Pass-Pst
‘Sato was adversely affected by being transferred to the general-affairs section
by the director’

(Howard & Niyekawa-Howard, 1976, p. 211)

c. Sato was transferred to the general-affairs section by the director.

Howard & Niyekawa-Howard (1976) held that while the active sentence (4a) above may
be used in any kind of transfer, the passive sentence (4b) typically implies that the
transfer to the general-affairs section was a demotion for Mr./Ms. Sato, the person
transferred. Such an unfavorable connotation to Mr./Ms. Sato is totally lacking in (4c),
the English passive counterpart of (4b).

An additional example of the connotation differences follows below:

(5) a. Suzuki-san-ga Ozawa-san-o nizikan mat-ta
Suzuki-NOM Ozawa-ACC two hours wait-Pst
‘Suzuki waited two hours for Ozawa’

b. Ozawa-san-ga Suzuki-san-ni nizikan mat-are-ta
Ozawa-NOM Suzuki-by two hours wait-Pass-Pst
‘Ozawa was adversely affected by being waited for by Suzuki for two hours’
(Howard & Niyekawa-Howard, 1976, p. 210)

c. Ozawa was waited for by Suzuki for two hours.

12

Similarly, in contrast to the neutral statement in the active sentence (5a) above, the
passive sentence (5b) indicates that Ozawa found himself in an inconvenient situation
because Suzuki had been waiting for him for two hours. It implies, for instance, a
context in which the former was trying to avoid meeting the latter. Such an implication is

far-fetched in the English passive sentence (5c).3

1.2.1.3. Ni versus niyotte contrasts
Turning to the contrast between the ni yotte passive and the ni direct passive, in

many cases it may appear that ni yotte is merely a stylistic variant of ni as shown in (6)

below:
(6) a. John-ga Bill-ni hinan-s-are-ta
-NOM -by criticize-Pass-Pst
b. John-ga Bill-ni yotte hinan-s-are-ta
-NOM —by criticize-Pass-Pst
‘John was criticized by Bill’
c. Bill-ga John-o hinan-si-ta
-NOM -ACC criticize-Pst
‘Bill criticized John’

There are, however, significant differences in how the ni direct passive and the m' yotte
passive may be used. Two differences resulting from the fundamental semantic

difference between the two are presented below.

1.2.1.3.1. Perfective versus non-perfective readings
There are cases in which ni yotte, but not ni, may mark an agent NP in passive

sentences (see Inoue, 1976 for more examples):

13

(7)

(8)

kaikai-ga gityoo-ni yotte/*-ni sengen-S-are-ta
opening of the meeting-NOM chairperson-by announce-Pass-Pst
‘the opening of the meeting was announced by the chairperson’

(Inoue, 1976, p.83)

sono hon-wa Ooe Kenzaburoo-ni yotte/*-ni kak-are-ta
that book-TOP Ooe Kenzaburoo-by write-Pass-Pst
‘that book was written by Kenzaburoo Ooe’

The contrast in grammaticality in (7) above shows that if the passive subject is an abstract

noun like kaikai ‘the opening of the meeting’, ni yotte is obligatory. Similarly, as illustrated

in (8), when the subject in a passive sentence is concrete but inanimate, such as sono hon

‘that book’, ni yotte is obligatory. This is so because

m' has the meaning of ‘influence of the agent’ on the passive subject. Its
difference from ni yotte is: ni may be used only in cases where the passive subject
and the agent are closely related to each other in this sense... if the passive
subject is such that it does not feel such influence, or that it does not receive the
direct effect of such influence, one cannot use ni. Therefore, if the passive subject
is inanimate, ni is often excluded (Kuroda, 1979, pp. 309-310, citing Inoue, 1976,

But as Inoue (1976) further noted, not all inanimate passive subjects require ni

yotte. The following examples show exceptions to this generalization:

(9)

a. kono siro-wa Tokugawa-gun-ni koogeki-s-are-ta

this castle-TOP Tokugawa-army-by attack-Pass-Pst
‘this castle was attacked by the Tokugawa army’

b. kono ie-wa itabei-ni kakom-are-te iru

this house-TOP wooden-fence-by surround-Pass-Perf-Pres
‘this house is under the state of being surrounded by a wooden fence’
(Inoue, 1976, p.83)

In both (9a) and (9b) above, despite the inanimate passive subjects, siro ‘castle’ and ie

‘house’ respectively, the ni direct passive is grammatical. In the case of the first type of

14

exception, Inoue proposed that siro ‘castle’ in (9a) means not only a castle as a building,
but also by extension a group of warriors stationed in it, and thus may be understood as
standing for an animate referent. Thus, the ni direct passive sentence (9a) does not count
as a true exception to Inoue’s observation above.

Turning to the second exceptional case illustrated by (9b), Kuroda (197 9)
proposed that an inanimate NP may be the subject of the m‘ direct passive if the passive
sentence is in the perfective aspect, in the sense that it expresses a state of the passive
subject which has resulted from an event or process described by the rest of the sentence.
One way of effecting a perfective reading in the ni direct passive is by means of the te iru
form as shown in sentence (9b). That is to say, it describes the state of the house existing
at the present time as a result of the process of someone having surrounded it with a
wooden fence which took place prior to the reference time.

On the other hand, when the te iru form expresses a progressive aspect, the ni

direct passive with an inanimate subject is not permissible as shown below:

(10) ano mati-wa nihon-gun-ni yotte/*-ni hakai-s-are-te iru
that town-TOP the Japanese army—by destroy-Pass-Prog-Pres
‘that town is being destroyed by the Japanese army’ (Kuroda, 1979, p. 328)

This is so because the passive sentence of the progressive reading (10) above does not
readily express the existing state of the passive subject ano mati ‘that town’ as resulting
from destruction by the Japanese army, but rather describes the on- going process of the -
destruction of the town. Hence, the grammaticality contrast between (9b) and (10),

namely, the grammaticality of (9b) with its perfective te iru reading versus the

15

ungrarnmaticality of (10) with its progressive te iru reading, illustrates that an inanimate
NP may be the subject of the ni direct passive in a perfective reading.

Furthermore, a perfective reading may be brought forward by means of the ta
form (which expresses either a perfective or simple past reading). Compare the following
presence and absence of the grammaticality contrast between the ni direct and the ni yotte

passives:

(11) a. ano mati-wa nihon-gun-ni yotte/-ni hakai-s-are-ta
that town-TOP Japanese Army-by destroy-Pass-Perf
‘that town has been destroyed by the Japanese Army’ (perfective reading)

b. ano mati-wa nihon-gun-ni yotte/*-ni hakai-s-are-ta
destroy-Pass-Pst
‘that town was destroyed by the Japanese Army’ (simple past reading)
(Kuroda, 1979, p. 327)

When someone is looking down from a hilltop and sees the town in rubble with his or her
own eyes as a result of destruction by the Japanese Army, as in the perfective reading,
both the ni direct and the m" yotte passives (11a) above are permissible; on the other hand,
to refer to the above destruction as a historical incident that took place in the past, e. g., in
the past reading, the m' yotte, but not the ni direct passive, is used (Kuroda, 1979).
Consequently, Kuroda (1979) claimed that Inoue’s (1976) general observation on
m’ direct passive sentences with an inanimate subject such as (7) and (8) should be
subsumed under the former’s (i.e., Kuroda’s) proposal. It is not possible, for instance, to
conceive that an abstract entity such as ‘the opening of a meeting’ will be in a certain
state as a result of the chairperson’s announcing it, and thus a perfective reading is not
possible for (7). This requires the passive sentence (7) to be in a simple past reading,

resulting in the ungrammaticality of the ni direct passive version of (7). Similarly, the

16

passive sentence (8) clearly states a historical achievement by Kenzaburo Ooe, a Nobel

laureate; thereby it must be in a past reading. This renders the m‘ direct passive version of

(8) ungrammatical.

1.2.1.3.2. Verb-induced viewpoint differences

The other difference between ni direct and ni yotte passives arises depending on
the viewpoint from which an event or process is described in the passive sentence. More
specifically, the ni yotte passive may be used when a passive sentence provides a neutral,
objective description of an event, while the ni direct passive may be used appropriately
when a passive sentence describes a personally-involved event. One means of effecting
such a viewpoint difference is in the choice of a verb used in the passive sentence. Two
types of verb-induced viewpoint differences are given below.

The first case is based on the observation that traditionally, Sino—Japanese verbs
are preferred over Japanese native ones in formal objective writing. Thus the former
may fit with the ni yotte passive while the latter may be appropriate for the ni direct

passive. With this in mind, look at the following set of three passive sentences:

(12) a. Bill—ga John-ni koros-are-ta
-NOM -by kill-Pass-Pst
‘Bill was killed by John’

b. Bill-ga John-ni yotte koros-are-ta
.by
‘Bill was killed by John’

0. Bill-ga John-ni yotte satugai-s-are—ta

murder-Pass-Pst
‘Bill was murdered by John’ (Kuroda, 1979, p. 320)

17

The verb in (12a) and (12b) above, korosu, is a Japanese native verb meaning ‘kill,
murder’, and that in (12c), satugai—suru, is a near-synonymous Sine-Japanese verb
meaning ‘kill, murder, slaughter’. The agentive NP, John, is marked with ni in (12a), but
with ni yotte in (12b) and (12c). Kuroda (1979) maintained that (12b) sounds unnatural
due to the rather inappropriate combination of the Japanese native verb and ni yotte. The
sentence (12c) definitely reads more naturally than (12b) since the main verb is
substituted with the Sino-Japanese one (consonant with ni yotte), but (12c) might still be
felt to be less natural than (12a) wherein the appropriate combination of the Japanese
native verb and ni is used.

Further, Kuroda (1979) claimed that the degree of naturalness of the ni yotte
passive sentence (12c) above is improved if the personal name, John, is replaced by the

name of an impersonal agency, say, the CIA, as given in (133) below:

(13) a. Bill-ga CIA-ni yotte satugai-s-are-ta
-NOM -by murder-Pass-Pst
‘Bill was murdered by the CIA’
b. ‘Bill-ga tuma-ni yotte satugai-s-are-ta
wife-by murder-Pass-Pst

‘Bill was murdered by his wife’

c. Bill-ga tuma-ni koros-are-ta
‘Bill was killed by his wife’
This is so because the impersonal agent, the CIA, firrther enhances the objectivity of the
situation described, as in newspaper reporting. Conversely, (13b) is rendered unnatural
as a ni yotte passive when Bill’s wife, who maintains a close relationship with Bill (the
victim), is substituted for the CIA in (13a). On the other hand, this family tragedy of the

murder of the husband by the wife is appropriately expressed in the ni direct passive

l8

sentence (13c). Based on these and other observations of highly subtle but consistently
apparent semantic differences between ni and m‘ yotte passives, Kuroda suggested that the
m’ yotte passive tends to serve as an objective description whereas the ni direct passive
implies some empathy on the part of the passive subject (from which an ‘affective’
connotation may derive).

The second case of verb-induced viewpoint difference derives from a special verb
class. Verbs of this class, when passivized, carry a clear unfavorable connotation to the
passive subject. These verbs are also unique in that the existence of lexical passive
counterparts, i.e., unaccusative verbs, preempts otherwise possible ni yotte passivization.
One such case involves the transitive verb miru and the unaccusative verb mieru. The
former means “perceive by sight (viz. look at)” and the latter “perceive by sight passively

(or spontaneously) (viz. visible)”.4 Compare the following sentences:

(14)
a. Bill-ga nozokimi site iru tokoro-os J ohn-ni mi-rare-ta
-NOM peeping (into a room) as -by see-Pass-Pst
b. ‘Bill-ga nozokimi site iru tokoro—o John-ni yotte mi-rare-ta

‘Bill was affected by being seen by John as he was peeping (into a room)’

c. John-ni Bill-ga nozokimi site iru tokoro-ga mic-ta
-DAT -NOM peeping (into a room) as visible-Pst
‘Bill was visible to John as he was peeping (into a room)’
(Kuroda 1979, p. 314 & p. 317)

Kuroda (1979) stated that the ni direct passive sentence (14a) above is an appropriate

description of the incident in which Bill was peeping into a room and suddenly perceived
that his peeping had been being witnessed by John. Bill’s highly personal, psychological
reaction to the event is aptly expressed by the ni direct passive. Kuroda also held that the

ni yotte passive sentence (14b) is totally inappropriate due to the incompatibility between

19

the emotional reaction and the objective nature of the ni yotte passive. Indeed, replacing

(14b), sentence (14c) serves as a ‘non-affective’ or impersonal counterpart of (14a).

1.2.1.3.3. Adversity in the m‘ indirect passive

Here the analysis of the semantic contrast between the ni direct passive and the m'
yotte passive is extended to the ni indirect passive. As presented above, the ni indirect
passive has often been characterized as adversative (Kuno, 1973) since this type of
passive predominantly expresses the idea that the referent of the surface subject is
adversely affected by the event or state described by the remaining sentence. In fact, an
adversative connotation of the indirect passive is further underscored in an extended

context such as below:

(15)
a. J ohn-wa warui told-ni tomodati-ni ko-rare-ta
-TOP bad time-at fiiend-by come-Pass-Pst
‘John was unfavorably affected by his friend visiting him at an inconvenient time’
b. I"John-wa yoi told-ni tomodati-ni ko-rare-ta

good time-at
”John was unfavorably affected by his friend visiting him at a convenient time’
(Kuroda 1979, p.355)
In (15a) above, the adversative connotation that John was inconvenienced by his friend’s
visiting is compatible with, and indeed enhanced by, the adverbial phrase warui toki ni ‘at
an inconvenient time’, whereas in (15b), it is incompatible with the adverbial phrase ii
toki ni ‘at a convenient time’, rendering it ungrammatical.

This property of the ni indirect passive is in line with the observations on the

semantic differences between m‘ direct and m“ yotte passives, viz., the ni direct passive

20

expresses an affective connotation whereas the m’ yotte passive does not. Compare the

following pairs of ni and m’ yotte indirect passive sentences:

(16) a. (= Za) J ohn-ga sensee-ni kodomo-o sikar-are-ta
b. *John-ga sensee-ni yotte kodomo-o sikar-are-ta
‘John was adversely affected by the teacher’s scolding his child’
c. (= 2c) John-ga ame-ni hur-are-ta
d. *John-ga ame-ni yotte hur-are-ta

‘John was adversely affected by rain falling on him’

These grammaticality contrasts between the grammatical ni indirect passive sentences
(16a)/(l6c), on the one hand, and the ungrammatical ni yotte indirect passive sentences
(16b)/(16d), on the other, clearly Show that m' but not ni yotte is compatible with the
indirect passive due to its predominantly adversative reading.

Figure 1 summarizes the three types of Japanese passives characterized in this

section according to various semantic features which produce an affective connotation.

21

.28 .3385“. opuoobw-§e 38 o “38.3 «a gamma 88%
E 2.50 085 3 mcaqoampuB 338m 02888 n: montage 2:33 38:. .B 230 99:8 35 an.
.a

 

“50333 028qu voosvevomogw .1.

30.38 0333 3a 83¢qu memos“: mad. I new man 83A. we

we mom 8 93% 98 £8- coz
m

 

I\

manage obflaflopvfi + 02323 39%”: m2

 

 

”509503 #28980” cougarfiocp J
80.3% 253a Bananas?» memos": obnoufiom VI + gamed 6an 5.
mafia“: obfiflopwfi
l\
$350.? mcflaooa 28:32:80
«333m ones 80m 338% 893 “Sign

833.3 omegaahmo meonaomamflo uncaSom A 0:.me

22

1.2.2. Syntactic analysis of Japanese passives
1.2.2.1. Introduction

The previous section has reviewed various kinds of evidence that Japanese
passives may be classified semantically into two types: the ni passive, both direct and
indirect, and the m‘ yotte passive. This section summarizes Hoshi’s (1991; 1993a; 1993b;
1993c; 1994a; 1994b; 1999) proposal of syntactic representations of the ni direct passive,
the ni indirect passive, and the ni yotte passive in line with Kuroda’s (1965; 1979)
semantic analysis.‘5 Specifically, Hoshi’s arguments for the syntactic distinction between
the ni direct passive and the ni yotte passive are examined first, followed by his syntactic

analysis of the three types of Japanese passives.

1.2.2.2. Syntactic evidence for the difference between the ni direct and the m' yotte
passives

This section presents two kinds of syntactic evidence for the difference between
ni yotte and ni direct passives. Hoshi (1991; 1994a; 1994b) observed the following

contrast in grammaticality where the verb phrase idiom tyuui-o harau ‘pay heed’ is

passivized:
(l 7) a. Mary-ga tyuui—o harat-ta
-NOM heed-ACC pay-Pst
‘Mary paid heed’
b. tyuui-ga Mary-ni yotte haraw-are-ta (the ni yotte passive)
heed-NOM -by pay-Pass-Pst
‘heed was paid’
c. *tyuui-ga Mary-ni haraw-are-ta (ni passive)
heed-NOM -by pay-Pass-Pst
“heed was affected by being paid by Mary’ (Hoshi, 1991, p. 70)

23

Following Kuroda’s (1965; 1979) proposal that the passive morpheme rare of the ni
direct passive is a verb requiring the subject of a passive sentence to be an ‘affectee’7
(i.e., an experiencer of the event or action expressed by the lower clause (clarified later)),
whereas the passive morpheme rare of the ni yotte passive is a suffix which imposes no
such restriction on the subject, Hoshi accounted for the above grammaticality contrast
between the grammatical ni yotte passive sentence (17b) above and the ungrammatical ni
direct passive sentence (17c) as follows: It is possible to passivize the object of a verb
phrase idiom in the ni yotte passive while it is not possible in the ni direct passive. This
is so because being assigned the ‘affectee’ 0-role, the subject position of the latter type of
passive cannot take an object NP of the verb phrase idiom (viz. non-affectee). No such
restriction is imposed on the subject position of the ni yotte passive, allowing it to serve
as a passive subject.

Next, Hoshi (1991; 1994a; 1994b) observed the grammaticality contrast put forth
by the subj ect-oriented adverbial modification. In short, subject oriented adverbs such as
orokanimo ‘stupidly’ cannot modify the subject of the ni yotte passive while such adverb

modification is possible in the ni direct passive as given below:

(18) a. daitooryoo-ga orokanimo CIA-ni koros-are-te-simat-ta
president-NOM stupidly -by kill-Pass-should not have-Pst

b. ??daitooryoo-ga orokanimo CIA-ni yotte koros-are-te-simat-ta
‘the president stupidly let the CIA kill him, which he should not have let happen’
(Hoshi, 1994, p. 151)

Hoshi held that this contrast could be accounted for if we assume that the subject in the m‘
direct passive is a 0-position whereas that in the ni yotte passive is a non 0-position and,

furthermore, that subject oriented adverbs are licensed by, or require, a 0-subject.8 Since

24

the subject position in the ni direct passive is a 0-position, a subject oriented adverb such
as orokanimo ‘stupidly’ is licensed, resulting in the grammaticality of (1 8a) above. But
since the ni yotte passive does not provide a 0-subj ect, the licensing condition for the
adverb is not met, rendering (18b) ungrammatical.

Based on these observations, Hoshi (1994a; 1994 b) proposed a syntactic analysis
of the three types of Japanese passives. The analysis of each type of passive is presented

below in the order of the ni yotte passive, the ni indirect passive, and the ni direct passive.

1.2.2.3. Syntactic representations of the niyotte passive9

The ni yotte passive is derived in precisely the same way as the English passive is
derived from its active counterpart. In other words, the passive morpheme of the m' yotte
passive, rare, functions as a passive affix as the English counterpart, -en, does. See the

following ni yotte passive and its corresponding active sentences:

(19) a. Mary-ga sensee-ni yotte home-rare-ta
-NOM teacher-by praise-Pass-Pst
‘Mary was praised by the teacher’
b. sensee-ga Mary-o home-ta
-NOM -ACC praise-Pst
‘the teacher praised Mary’ (Hoshi, 1994a, p. 42)

The structures for the ni yotte passive (19a) above are as follows:

25

(20)

b.
a IP IP
/ \ / \
I' Maryi-ga I'
/ \ /
VP] I VPl
/ \ => / \
V' (t'i) V' I
/ \ / \ / \
(sensee ni yotte) V' (sensee ni yotte) V' VJ

/\ /\ /\

I l I
/ \ tj home rare-ta
V V

home rare-ta
(Hoshi, 1994a, p. 43)

As shown in (20a) above, the passive morpheme rare attaches to the verb at the initial
point of the derivation. It then triggers passivization in which the external 0-role of the
verb homeru ‘praise’ is suppressed and the objective case of the verb is absorbed.10 Due
to the external 0-role suppression, the external argument of the verb, sensee ‘teacher’,
appears as an adverbial adjunct phrase, sensee—ni yotte. Next, as illustrated in (20b), the
internal argument of the verb, Mary, undergoes NP movement to SPEC of IP for case

reasons. Finally, the complex verb home-rare-ta moves up to Ito have its tense features

checked off.

26

1.2.2.4. Syntactic representations of the ni indirect passive

The second type of passive presented is the m“ indirect passive. In this type of
passive, unlike the passive affix rare of the ni yotte passive, the passive morpheme rare
serves as a verb assigning its external ‘affectee’ 6-role and does not trigger passivization.

An example of a ni indirect passive sentence follows in (21) and is derived in (22).

(21) John-ga Mary-ni Bill-o sinyoos-are-ta
-NOM -by -ACC trust-Pass-Pst
‘John was affected by Mary’s trusting Bill’ (Hoshi, 1994a, p. 60)
(22)
a b.
IP IP
I' Johni-ga I'

/K =w/K

\ ‘/\ /\

VP2 V VP2 V Vj I
/\ g /\ g /\
Mary V' Mary-ni V' V V
/ \ / \ sinyoos arrlr-ta
Bill V Bill-o V
/ \ I
V V tj

sinyoos are-ta
(Hoshi, 1994a, p. 64)

27

In (22a), the passive verb of the ni indirect passive, rare, adjoins to the transitive verb

sinyoo-suru ‘tr'ust’ without triggering passivization processes because by assumption it is

not a passivizer. Due to the absence of passivization within VP2 (the lower VP), i.e.,

without 0-role suppression, the external argument of the embedded verb, Mary, appears
in the SPEC of VP; position, retaining an argument status. In (22b), it is ni case-marked

as a last resort (see Saito’s, 1982, theory of case assignment, cited in endnote 9).

Similarly, because of the lack of case absorption within VPZ, the internal argument of the

embedded verb, Bill, remains in its verb complement position, and later in (22b), it is o
case-marked.
Moreover, since the passive verb rare assigns an external O-role

(‘affectee’/experiencer) to the matrix subject, as well as an internal O—role (‘event’ 1 l) to

the embedded clause later in the derivation, the Larsonian VP shell, VP1, is generated
over the lower clause VP2.12 The matrix subject John is generated under the SPEC of

VP1. In (22b), to have its tense features checked off, the complex verb, sinyoos-are-ta, is
forced to move up to I. In the course of the movement up to I, the complex verb initially

moves into the empty V position of VP], marked as e in (22a). In that V position, the

passive verb rare assigns its ‘affectee’ external 0-role to John and the internal 0-role to
the lower clause VP2. The matrix subject John also moves up to the SPEC of I for case

reasons.

1.2.2.5. Syntactic representations of the ni direct passive

28

We now turn to an examination of the structures of the ni direct passive. In this
type of passive, as in the ni indirect passive, the passive morpheme rare functions as a
verb assigning its ‘affectee’ 0-role; however, unlike in the ni indirect passive, this passive
verb triggers passivization. See the ni direct passive sentence (23) and its structures (24)

below:

(23) a. Mary-ga sensee-ni home-rare-ta
-NOM teacher-by praise-Pass-Pst
‘Mary was affected by being praised by the teacher’ (Hoshi, 1994a, p. 35)

29

 

(24)

/K /\
/ I\ Maryi-ga/ I
/K cw/K

/\ ‘/\ /\

M

VPz V VP2 V Vj I
/\ g /\ g /\
(sensee-ni) V' PRO V V

i V'
l l
/ \ / \ sinyoos rare-ta
RO V (sensee-ni) V'
/ \ /
V V ti V
I I '

home rare—ta tj
(PRO within VP2 is discussed shortly.)

P

(Hoshi, 1994a, p. 35)

In (24a), the passive verb rare attaches to the embedded verb homeru ‘praise’ at the
initial point of the derivation. Upon adjoining, the passive verb suppresses the external
0-role and absorbs the objective case of the embedded verb as in the case of m' yotte
passivization in (20a). Because of this external 0-role suppression, the external argument

of homeru appears as an adverbial phrase in a V’ adjoined position, e. g., sensee-ni, under

the embedded clause VPz. As illustrated in the derivation of the ni indirect passive in

(22), the Larsonian VP Shell VP1 is also generated over the embedded clause VP2. In the

30

SPEC of VP1, the matrix subject Mary is generated. Then, in (24b), the complex verb,

home-rare-ta, moves up to I by way of the empty V position within VP1, where the

passive verb assigns its external as well as internal 0-roles. The matrix subject Mary also
moves up to the SPEC of I for case reasons.

In the syntactic analysis of the ni direct passive above, Hoshi (1994a) posited

PRO under V’ of lower VP2 in (24a) for the following reasons. In Kuroda’s (1965;

1979) analysis of the ni direct passive, as shown in (25) below, the matrix subject
receives the external ‘affectee’ 0-role from the passive verb rare. However, it also
involves a deletion operation of ‘EQUI’ type (see below). That is to say, because the
embedded object John is identical with the matrix subject John (indicated by co-

indexation), the embedded object (the second John) is deleted.

(25) ‘affectee’ 0-role assignment

1 l

[S Johni-ga [S Mary-ni John; home I rare-ta]

 

l deletion

Q

Hoshi pointed out, however, that it is not clear, under current grammatical theory, what
the nature of the deletion operation in (25) is. Moreover, it is necessary to specify how
the internal argument of the embedded verb, the second John, is represented in the above

structure (25). While following Kuroda’s analysis of the ni direct passive, Hoshi

resolved these problems by assuming that PRO is generated under V’ of VPz (embedded

31

VP) and receives the internal 0-role of the verb homeru as shown in (24a) above. At a
later point of the derivation, in (24b), PRO moves up to SPEC of VP2 to receive null case

(Chomsky & Lasnik, 1995).13

To recap, the three types of syntactic derivations have been identified for
Japanese passives: the ni yotte passive, the ni indirect passive, and the ni direct passive.
Table 1 summarizes the classification of these three passives according to the presence or
absence of ‘affectee’ (experiencer) external 0-role assignment and passivization as the

functions of the passive morphology rare.

Table l: Syntactic classification of Japanese passives
External 0-role Assignment Passivization

 

ni direct passive + +
ni indirect passive + -
ni yotte passive - +

 

 

1.3. Acquisition studies on Japanese passives in SLA

In this section, previous studies on the acquisition of Japanese passives are
reviewed. To the best of my knowledge, there are only a small number of studies
addressing the acquisition of passives in Japanese in SLA.

First, Tanaka (1992) conducted an error analysis type of examination of uses of
Japanese passives by upper intermediate and advanced learners who were predominantly
speakers of European languages. Forty-six out of the 71 learners were English-speaking

learners. They had all enrolled in either an upper intermediate or an advanced class at a

32

university in Tokyo during the data collection period. Data for the study consisted of
written samples from class writing assignments collected over a period of three years.

A total of 311 instances of uses of passives were extracted from the writing
samples for the error analysis. The initial inspection of overall correct and incorrect uses
indicated that there were 181 correct uses and 130 incorrect instances (i.e., 58.2% correct
use). Next, in order to conduct a more detailed analysis, these passive sentences were
divided into two types: the animate passive subject type (in which a subject of the passive
sentence is animate) and the inanimate passive subject type (in which it is inanimate). 1"
These occurred 117 and 194 times, respectively. The inanimate passive subject type was
used 62.4% of the time and thus appeared more frequently than the animate passive
subject type. This was due perhaps to the nature of essay-type writing assignments which
require stating objective facts and opinions (Tanaka, 1992), and which thus provided
more suitable contexts for the former type of passive over the latter.

In the animate passive subject type, 71 uses were correct and 46 were incorrect
(i.e., 60.7% correct use) whereas in the inanimate passive subject type, 110 instances
were correct and 84 were incorrect (i.e., 56.7% correct use). Hence it appears that both
types of passives were used roughly at the same correct use rate. The indirect passive
was used only four times: twice correctly and twice incorrectly (i.e., 50% correct use),
suggesting the avoidance strategy in use due to its difficulty.”

Furthermore, in order to conduct a qualitative analysis of those incorrect uses of
passives, Tanaka (1992) examined what kinds of errors were made in using the inanimate
subject and the animate subject passives. In the former, errors in the choice of aspect as

well as in particle use were fiequent. Specifically, the stative verb form to iru was not

33

often used in passive sentences where it was needed. Misuses of m' instead of ni yotte to

mark an agent NP of the passive sentence were also identified. Characteristic misuses of

the animate subject passive had to do with confusion with other complex structures such

as causatives, causative passives, etc; that is to say, the animate subject passive was used

where causatives or causative passives should have been or vice versa. Errors in particle

use involving go, ni, 0 were also found.

Actually observed misuses of particles in Tanaka ( 1992) are given below in (26)

together with their corrected counterparts, where the bold-face parts indicate incorrect

and corrected particle uses, and pairs of (3)-(b), (c)-(d), (e)-(f), as well as (g)-(h)

constitute incorrect and corrected pairs:

(26)

*sono ningyoo-wa tanin-ga tukur-are-ta
that doll-TOP a stranger-N OM make-Pass-Pst
sono nigyoo-wa tanin-ni yotte tukur-are-ta
-by
‘that doll was made by a stranger’
‘siken/zyuken benkyoo no kyoosoo-o kyootyoo-s-are-ta
competition in exams/entrance exams-ACC emphasize-Pass-Pst
siken/zyuken benkyoo no kyoosoo-ga kyootyoo-s-are-ta
-NOM
‘competition in entrance examinations was emphasized’
*wakai hito no kaiwa-ga komar-ase-rareru-rasii
conversation by younger people-NOM trouble-Pass-seem
wakai hito no kaiwa-nl komar-ase-rareru-rasii
-by

‘(they) seem to be troubled with conversation by younger people’

*nihon de-wa gaikokuzin-ni Simon-o tor-arete-imasu
Japan in-TOP foreigners-by fingerprint-ACC take-Pass-Pres

34

h. nihon de-wa gaikokuzin-ga Simon-o tor-arete-imasu
-N OM
‘The government in Japan makes resident foreigners get fingerprinted’
(Tanaka, 1992, pp. 155-156)

In (26a), the agent NP, tanin ‘stranger’, in the ni yotte passive is nominative case-marked
where it must be ni yotte marked as in (26b), since the agent NP has undergone 0-role
suppression and has been demoted to the ni yotte phrase. Likewise, in (26c), the complex
patient NP, siken/zyuken benkyoo no kyoosoo ‘competition in exams/entrance exams’, is
wrongly accusative case-marked where the correct particle use is a nominative case
marker, go, as shown in (26d), since the patient NP has undergone accusative case
absorption and has been moved to the subject position. In (26c), the source NP, wakai
hito no kaiwa ‘conversation by younger people’, in the ni direct passive is nominative
case-marked where it should be ni marked as in (260 since the source NP has gone
through 0-role suppression and has been demoted to the ni phrase. Finally, in (26g), the
experiencer NP, gaikokuzin ‘foreigners’, in the ni indirect passive is incorrectly agentive
ni-marked where it should be nominative case ga-marked as the ni indirect passive
subject, indicating a person who has experienced the unpleasant incident of being
fingerprinted for the domicile registration as in (26h).

To further examine the uses of passives and other complex structures, Tanaka
(1993) contrived a discourse completion test based on mis- or non-uses of them, as found
in Tanaka (1992). She administered it to 20 intermediate, 16 upper intermediate, and 44
advanced learners of Japanese. As in her previous study they were all enrolled in
Japanese language classes at a university in Tokyo. The numbers of English-speaking
learners in each group were 18 (90% of the group), 13 (81.25%), and 24 (54.55%),

respectively. Thus English was the predominant L1 of the groups.

35

It was found that learners at the intermediate level failed to use the passive
structure when they were asked to complete the clause where its initial NP was only
given with a verb specified as follows: nihonzin [ ] yoo ni, ‘so that Japanese
people [ ], .....’ (use warau ‘laugh at’). In this example, they had to first supply an
appropriate particle and then conjugate the verb warau ‘laugh’ to a passive negative
form. Half of the intermediate learners completed it as an active sentence as follows:
nihonzin[-ga warawa-nai] yooni ‘So that Japanese people do not laugh at (you)’. The
correct completion required using a passive sentence, namely, nihonzin[-ni waraw-are—
nai] yoo ni ‘so that (you) are not laughed at by Japanese people’. From a purely
grammatical perspective, the former response is correct; however, in order to describe a
potentially embarrassing situation in which a referent of the statement above may be
laughed at (for doing something Japanese people do not normally), the ni direct passive
should be used to express this adversity connotation. This type of misuse of the active
instead of the passive sentence diminished as learners advanced in proficiency levels:
31.3% of the upper intermediate and 11.4% of the advanced group.

From these two studies the following two observations can be made: First,
syntactically, the ni direct and the ni yotte passives were almost equally difficult to learn,
to the extent that even upper intermediate and advanced learners were able to use them
correctly in writing only half of the time. For the ni indirect passive, an apparent
avoidance in use was evidenced. Misuses of case marking were evidenced across all
three types of passives. Second, semantically, an adversative reading of the ni direct
passive was not well learned by the intermediate learners. Yet it appears that the learners

were eventually able to learn it as they advanced in their proficiency levels.

36

The next work under review is Watabe, Brown, & Ueta (1991) in which they
conducted a contrastive analysis study to examine functional transfer effects from
English to Japanese and vice versa in the use of passive sentences. English-speaking
learners of Japanese in the study were 27 advanced learners who had been studying
Japanese for an average of four years at a university in the US and who had all lived in
Japan for 18 to 24 months. The eighteen NSS of Japanese used as controls were students
or their spouses at the university who had very little knowledge of English.

Both the NSS and the NNSs (non-native speakers) of Japanese were asked to
perform the following two writing tasks: 1) to produce a newspaper account of a fire
which was portrayed in a series of five pictures; and 2) to write on the topic of ‘the most
misfortunate event in my life’. For both compositions, subjects were told to write as
much as they could in 20 minutes each. The first topic was chosen to elicit English-type
passive uses such as irnpersonalization (i.e., “the identity of the subject/agent of the
active is suppressed”) and de-transitivization (i.e., “the clause becomes semantically less-
active, less-transitive, more stative”) (Watabe et al., 1991, cited from Givon, 1981). The
second topic was expected to elicit J apanese-type passive uses, particularly the
adversative passive, which expresses that the subject of the passive sentence is adversely
affected by the event or action described by the remainder of the sentence.

Overall there were 27 passive sentences found in the English-speaking learners’
writings, of which thirteen uses were incorrect (48.15% incorrect use). Watabe et al.
found that none of these errors appeared to be attributable solely to grammatical mistakes
such as incorrect morphological change in the verb or a faulty particle choice, except for

the inaccurate lexical choice between transitive and intransitive verbs. On the other hand,

37

a functional analysis of both groups’ compositions identified the source of the incorrect
uses, and revealed a difference in the way that the two groups used Japanese passive
sentences in their writings. Specifically, NSS of Japanese were more likely to use passive
sentences in reporting their personal account (the second task) while English-speaking
learners were more inclined to make use of them in the newspaper account (the first task).
The ratio of the number of passives used in the newspaper account to that in the personal
experience account was .40 to 1.0 for the NS group, and 3.43 to 1.0 for the English-
speaking learners’ group.

Given that the newspaper account provided a functionally appropriate setting for
use of the English passive, and the personal account was well suited to the use of the
Japanese adversative passive, these findings led Watabe et a1. (1991) to the following
conclusion: the English-speaking learners transferred discourse functions of the English
passive when producing Japanese passive sentences in writing although they made very
few grammatical errors in forming them.

An example of such non-use of the Japanese passive due to the functional transfer

from English is given below:

(27)
I"yoi koto-o site-iru noni naze tomodati de aru
good thing-ACC do-Prog-Pres even though why fiiend be
minna-wa (watasi-o) izimete-i-ta no ka to nayan-da
everyone-TOP -ACC tease-Prog-Pst QUE that worry-Pst
‘I wondered why everyone who was my friend was teasing me even though I was doing
good things’ (Watabe et al., 1991, p. 128)

The above sentence is inappropriate because it does not use the passive construction to

express the situation in which the subject of the sentence was adversely affected by the

38

action of the verb izimeru ‘tease’. The appropriate expression for this situation would be
the following use of the passive construction (where bold-face indicates a corrected use

of the ni direct passive):

(28) yoi koto-o site-iru noni naze tomodati de aru minnna-nl izime-rare-te
everyone-by tease-Pass-Prog
ita no ka

Thus, in Watabe et al.'s (1991) study it was demonstrated that it is very difficult to
elicit Japanese passive sentences in production tasks. Even in the task Specifically
designed to elicit Japanese passive sentences, i.e., the personal experience account task,
the NS controls used passives only on average 1.389 times per subject, whereas they
produced them on average as few as .556 times per subject for the newspaper account
task. The English-speaking learners used them .308 times for the former task and 1.056
times for the latter. Next, the widely held view that Japanese passives are difficult for
English-speaking learners to learn received supportive evidence from this study. The
English-speaking subjects of the study who had studied Japanese for as many as four
years on average and who had lived in Japan over a year were successful in using passive
sentences correctly only about half of the time (at a rate of 51.85%). Finally, although
Watabe et a1. ascribed the NNS subjects’ errors in the use of Japanese passives to
functional transfer from English, it is not clear from their study whether the NNS subjects
possessed the semantic knowledge that Japanese passives may express adversity. The
study would have shed light on the question of source(s) of transfer if there had been an

attempt to measure the subjects’ knowledge of adversity in Japanese passives.

39

The final study under review was prompted by frequent reports that Chinese-
speaking learners encounter difficulties learning Japanese passives. Feng (1993)
investigated their acquisition of Japanese passives in light of both promoting and
interfering influences of the L1. Subjects in the study included three groups of Chinese-
speaking learners: 1) 20 Chinese speakers who had either taught Japanese or served as
interpreters in China for an average of five years and two months after obtaining a degree
in Japanese, and who had lived in Japan on average for two years and three months.
Their average length of study amounted to 11 years (= C1); 2) 3O learners who had been
studying Japanese at a university in China for three years, and who had never been in
Japan (= C2); 3) 21 learners who had been studying Japanese at an institution in Japan for
one year and one month on average (= C3). As a control, 130 NSS of Japanese (= J)
participated in the study.

All study participants answered two surveys concerning Japanese passives. The
first survey involved rating the semantic naturalness of a total of 36 passive sentences on
a five-point scale, e. g., ‘very natur ’, ‘somewhat natur ’, ‘not sure’, ‘somewhat
unnatural’, and ‘very unnatural’. Ten sentences in the survey were acceptable both in
Japanese and Chinese, while the remaining 26 sentences were acceptable in one
language, but not in the other. The second survey required choosing appropriate passive
markers from the four choices of ni, ni yotte, kara, and ale“5 in a total of 46 passive
sentences. There are no such equivalents to these passive particle markers in Chinese.

It was expected that where there were similarities in passives between the two
languages, learning had been facilitated; where there were dissirnilarities, learning had

been hindered. The results of the first survey supported this hypothesis. Namely, for the

40

rating of passive sentences exhibiting similarities between the two languages, the most
advanced Group C1 approximated the NS group J in 93.5% of them whereas C2 and C3
also agreed with J in 86.3% and 87.1% of them, respectively. On the other hand, in
rating those displaying dissimilarities, C1 agreed with J in only 43.1% of them, whereas
C2 and C3 agreed in 47.3% and 46.7%, respectively.

The second survey (choice of appropriate passive particle markers) included
passive sentences such as those given below (where correct passive markers (bold-faced)

are provided):

(29) a. karera-wa sinkoo-nl yotte sukuw-are-ta
they-TOP faith-by redeem-Pass-Pst
‘they were redeemed by their faith in GOD’
b. heewa-mo karera-ni yotte yabur-are-ta
peace-also them-by break-Pass-Pst

‘peace was broken by them, too’

Note that passive sentence (29a) above is such that the ni yotte passive is preferred over
the ni direct passive in objective reporting whereas sentence (29b) does not readily allow
a perfective reading and thus requires the ni yotte passive. Table 2 below summarizes the
proportions for the choices of ni and ni yotte in the two passive sentences above by the
subject groups. In Table 2, the figure under each group column indicates the proportions
of subjects who chose ni yotte and ni respectively. (Since they were allowed to choose
more than one passive marker, the total percentage of ni yotte and ni choices may exceed

100%.)

41

Table 2: Subjects’ percentage choice of m' yotte and ni

 

 

 

Sentences J C1 C2 C3
Aa ni yotte 89.2 95 46.7 33.3
ni 7.7 15 40 42.9
B ni yotte 83.1 75 20 33.3
ni 6.2 O 40 57.1

 

 

a. A and B designate sentences (29a) and (29b) respectively.

The table shows that a great majority of NS subjects chose ni yotte for both sentences
(89.2% and 83.1%), whereas only a few of them chose ni (7.7% and 6.2%), consistent
with Kuroda's (1979) theory. The highly advanced Group C1 exhibited a particle-choice
pattern remarkably similar to the NS group J. i.e., 95% and 75% for the ni yotte choice
versus 15% and 0% for the ni one. However, the less proficient groups, C2 and C3,
failed to evince a similar pattern of choices, and opted for ni over ni yotte more
frequently.

F eng's (1993) study found clear evidence for L1 influences, both promoting and
interfering, in Chinese-speaking learners’ acquisition of the semantics of Japanese
passives. Properties of Japanese passives shared with their Chinese counterparts were
easily learned, while those differing from Chinese passives were not learned even by the
highly advanced subjects. By contrast, in learning appropriate passive particle marker
choices, the highly advanced Chinese-speaking subjects approximated the NS controls in
differential uses of ni and ni yotte in passive sentences. The performances of the other
two groups indicated the lack of such knowledge of Japanese passives. (See also Yang &
Akahori, 1998 for similar findings.) (As for FLA on Japanese passives, see Hakuta,

1982; Harada, 1977; Sano, 1977; and, particularly, Clancy, 1985 for an overview of

42

acquisition studies of Japanese passives. For more recent work, see Iitaka, 1989; Ryuzaki
& Ito, 1999; Sano, Endo, & Yamakoshi, 2001; Suzuki, 1998.)

The aforementioned studies on the acquisition of Japanese passives found the
following: 1) the ni indirect passive was most difficult for English-speaking learners to
learn, whereas ni direct and ni yotte passives were similar in difficulty for them; 2) case
marking errors in passive sentences of all three types were all evidenced; 3) the ni direct
passive was often not used where it should have been in order to express an adversative
reading; and 4) highly advanced Chinese-speaking learners were able to learn differential

uses of the ni direct versus the ni yotte passive.

1.4. Adversative readings of passives by native speakers of Japanese

This section explores the extent and the nature of an adversative reading in m'
direct and indirect passive sentences by reviewing an empirical study which examined
how NSS of Japanese rate the adversative connotation of ni direct and indirect passives
when these passive sentences vary in terms of verb connotation and subject animacy.

There is general agreement in the literature that the ni indirect passive primarily
expresses an adversative reading, yet it is still controversial to what extent the ni direct
passive carries an adversative reading as its basic meaning (see Alfonso, 1980; Howard &
Niyekawa-Howard, 1976; Jacobsen, 1992; Kuno, 1973; 1983; Kuroda, 1979; McCawley,
1972; Shibatani, 1990; Wierzbicka, 1979). In order to address this issue (and others)
Specs (1992) investigated the following questions: 1) does the ni direct passive as well as
the ni indirect passive generally carry an adversative reading? 2) is animacy of the

passive subject a necessary condition for an adversative reading? Twenty-one NSS of

43

Japanese participated in her study: eight males and thirteen females whose ages ranged
from 19 to 34. They included mostly students and their spouses or friends at a university
in the US.

They were asked to assess the degree of positiveness or negativeness in the
meaning of a total of 42 Japanese sentences using a seven-point rating scale (i.e., from +3
to -3). The test sentences included pairs or trios of active and ni passive (direct and
indirect) sentences, controlled for the following two factors: 1) animacy of the passive
subject, i.e., animate or inanimate; and 2) verb connotations, positive (ex. ai—suru ‘love’;
seekoo-suru ‘succeed’, etc.), neutral (ax. tukau ‘use; iku ‘go’, etc.), and negative (ex.
hinan-suru ‘criticize’; sinu ‘die’, etc.). Each example of active and passive sentences

with a neutral verb is given below:

(30) i. active: Yasuko-wa sono pen-o tukat—ta
Yasuko-TOP that pen-ACC use-Pst
‘Yasuko used that pen’
ii. ni direct: sono pen-wa Yasuko-ni tukaw-are-ta
passive that pen-TOP Yasuko-by use-Pass-Pst
‘that pen was used by Yasuko’
iii. ni indirect: Kazue-wa pen-o Ziro-ni tukaw-are-ta
passive Kazue-TOP pen-ACC Ziro-by use-Pass-Pst

‘Kazue was negatively affected by Ziro using the pen’
(Specs, 1992)
It was expected that active sentences with neutral verbs such as (3 0i) would be rated
around zero, e.g., neither positive nor negative, while the ni indirect passive versions like
(30iii) would be assessed negatively. A critical case was whether sentences like (30ii), a
ni direct passive containing not only a neutral verb but also an inanimate subject, would

be rated negatively.

It was found, first, that ni indirect passive sentences were rated negatively, i.e.,
read adversely, irrespective of verb connotations: the mean rating scores were -.92 for
positives, -1.52 for neutrals, and -2.60 for negatives. It was found, secondly, that the
degree of negative rating increased significantly progressively from positive verbs, to
neutral, and to negative ones. That is to say, sentences with negative verbs were judged
most negatively and those with positive verbs least negatively. (The ni indirect passive
sentences with an inanimate subject were not tested.)

Next, as for ni direct passive sentences with an animate subject, negative ratings
were given to those with negative verbs (a mean rating score of -2.45) and neutral ones (-
.95), but not with positive ones (2.38). As was found in the ni indirect passive, the degree
of negative ratings decreased significantly from negative to neutral verbs, but unlike the
previous finding, ni direct passive sentences with positive verbs were not rated
negatively. Second, it was evidenced that ni direct passive sentences with an inanimate
subject such as (30ii) above were rated negatively when there were no positive
connotations in the verbs used in them: the mean rating scores were -2.06 for negatives, -
.95 for neutrals, and .40 for positives.

These findings led Specs to conclude 1) that the ni indirect passive primarily
carries an adversative reading as asserted in the literature, and furthermore that this is so
even for passive sentences with verbs carrying positive connotation; 2) that the ni direct
passive may be read adversely when verbs used in it carry either negative or neutral
connotations; and 3) that animacy of the subject is not necessary for an adversative

reading in the ni direct passive; that is to say, the ni direct passive with an inanimate

45

subject may carry an adversative reading in the same way that with an animate subject

does.

1.5. Issues on native speakers’ grammaticality judgment

This section reviews empirical studies which investigated various aspects of the
difference in grammaticality judgment performances between linguists and non-linguists.
There are two positions put forth concerning the question of what effect linguistic
training exerts on NS’S grammaticality judgment ability: Does a linguist come to have a
sharper sense of grammaticality than a non-linguist, or alternatively, does he or she come
to develop a sense of grammaticality quite different from that of a non-linguist? Bradac,
Martin, Elliott, & Tardy (1980), arguing in favor of the latter view, noted: “as a result of
their special training, linguists may tend to judge strings differently from nonlinguists.
Training in linguistics may produce beliefs or attitudes which are not shared by those
who have not received such training” (p. 968). Levelt (1974), by contrast, maintained
that linguists’ training indeed enables them to factor out various irrelevant noise sources
influencing grammaticality judgments to which naive NSS may be susceptible. (See
Schfitze, 1996 for an extensive review of various reliability and validity issues
surrounding NS grammaticality judgment.) Results of empirical studies addressing this
issue have been found inconclusive.

In her widely cited seminal work on grammaticality judgment differences
between linguists and non-linguists, Spencer (1973) adopted a skeptical stance on
linguists’ use of their own intuitions for linguistic research. The subjects of her study

included 43 (non-linguist) NSS of English taking an introductory psychology course and

46

22 NNSS of English who had taken at least one course in generative grammar at a
university in the US.17 They were instructed to rate 150 sentences drawn from six
linguistic articles as either good or bad, with no other option available. The test
sentences were those judged as either clearly grammatical or ungrammatical by the
original linguist authors.

It was found first that an average of 81 .4% of the 150 test sentences were clear
cases, in that at least 65% of the subjects in both (NS and NNS) groups gave the same
judgment to a given sentence. The two groups differed by only 6% in the proportion of
those sentences accepted; in other words, they accepted roughly the same number of
sentences.18 Second, since it was established that the subjects’ ratings were overall
consistent, the linguist authors’ judgments, i.e., judgments in the source articles, were
compared to those of the subjects (i.e., 65 non-linguists). The result was that 48.67% of
the test sentences (73 out of a total of 150) exhibited judgment disagreements between
the two groups. The disagreement was counted when more than 35% of the subjects did
not agree with the author in rating any given sentence. (See Greenbaum, 1988 for similar
results.)

Thus, Spencer (1973) showed that approximately half of the time, sentence
judgments given her subjects were different from those given by the articles’ authors
(from whom the test sentences were taken). But as Schutze (1996) has pointed out, this is
quite different from showing that linguists’ grammaticality judgments as a group differ
systematically from non-linguists’.

Next, Snow & Meijer (1977) conducted three experiments to substantiate their

claim that syntactic intuitions are methodologically secondary to spontaneous speech by

47

Showing that metalinguistic performances may vary depending on how data are collected.
(Only the first two experiments, relevant to the present study, are reported on below.)
The subjects of their first experiment involved 25 na'r've NSS of Dutch who were in their
first or second year of a linguistics program, but had not taken any courses in syntactic
theory. The subjects of the second experiment included eight practicing syntacticians
whose primary data consisted of their own intuitions.

Their test materials involved 12 structures concerning three word order issues in
Dutch. There were two sentence items for each of the 12 structures in question, with a
total of 24 test sentences in the experimental material set. They were presented in two
different conditions: absolute judgments and rank-orderings. In the former, the subjects
were instructed to indicate whether they thought each of the 24 sentences was “good”
Dutch, where “good” meant acceptable in the spoken language. There were three options
from which to choose: good, not good, and in-between or not sure. In the second
condition, the 24 sentences were divided into four sets of six sentences each containing
the three word order problems in question. The subjects were told to reorder them, by
rewriting what they considered the best Dutch at the top, the worst Dutch at the bottom,
and multiple sentences on a single line if they were equally good or bad.

Two sessions, separated by a one-week interval, were convened in which half of
the subjects took the absolute judgment test and the other half took the rank-ordering test
in the first session, and vice versa in the second session. Both the first and second
experiments employed precisely the same test materials and experimental procedures in

order to produce comparable findings.

48

In the first experiment with the 25 non-linguists, there was some variability
observed in their absolute ratings in both between-subjects and within-subj ects
agreements. In terms of the between-subjects, at the highest agreement, one test structure
was agreed upon by 23 of them: at the lowest agreement, two sentence types received
almost an equal split with respect to the subjects’ judgments, i.e., approximately half
rated them as good and the other half as bad on both occasions. There was disagreement
over half of the ten structures tested19 where at least two subjects both times gave ratings
different from those made by the majority of the subjects. The within-subjects
consistency was 70.8%, with a given subject rating about seven out of the ten structurally
identical sentence pairs consistently, i.e., good, bad, or unsure on both occasions.

The results of the rank-orderings condition also showed some variability. The
between-subj ects agreement was found significant for all of the sets of six sentences as
measured by Kendall’s coefficient of concordance ranging from .466 to .670 (which is
not extremely high). The most agreed-upon sentence showed disagreements by only
three of 25 subjects, while all the other sentences showed at least seven disagreements as
compared to their mean rank; i.e., at least seven subjects placed them differently from
their mean rank. Moreover, no single test sentence was ranked by all the subjects either
at the best or the worst in a total of six sentences. In short, variability was found in the
non-linguists’ metalinguistic performances under two data collection conditions.

In the second experiment with eight linguists, greater consistencies were found in
both conditions, although some variability remained. The linguist subjects showed
greater between-subj ects consensus in the rank-orderings condition with Kendall’s

coefficients ranging between .581 and .844 (as opposed to those from .466 to .670 for the

49

non-linguist subjects), suggesting that the linguist subjects were more likely to rank a
group of six test sentences in the same order as the non-linguist subjects did. The
linguists also demonstrated Significantly greater within-subj ects consistency in the
absolute judgments condition, e.g., 94.3% compared to 70.8% for the non-linguist group.
(No report on sentence-by-sentence comparisons was given as in the first non-linguists’
experiment.)

Finally, when a direct comparison was made between the linguists and the non-
linguists, the mean rank-orderings of both groups demonstrated a high correlation
(Spearman p = .89) and so did the absolute judgments (Spearman p = .84), indicating that
these two groups performed the two tasks similarly.

By demonstrating variability in sentence judgments under two different data
presentation conditions, Snow & Meijer (197 7) claimed that it was necessary to establish
better methodological principles in syntactic research data collection. In contrast to
Spencer's (1973) finding that there was a judgment difference between a group of non-
linguists and an individual linguist, Snow & Meij er found that the judgments given by the
linguist subjects as a group were in accord with those of the non-linguists. In fact, the
linguists manifested more consistent judgments than the non-linguists, indicating the
beneficial effects of linguistic training (see also Levelt, 1974).

In a departure from Snow & Meijer's (197 7) approach to the issue of linguist and
non—linguist grammaticality judgment performances, Ross (1979) investigated perceptual
differences in grammaticality judgments. He asked 30 subjects (15 linguists and 15 non-
linguists) to assess the grammaticality of 12 sentences (see below) on a four-point scale,

i.e., 1 (perfectly grammatical) to 4 (perfectly ungrammatical). Moreover, he elicited their

50

perceptions about the judgments they gave. Specifically, they were asked to indicate 1)
how confident they were of their judgment by choosing one of the three options of ‘pretty
sure’, ‘middling’, and ‘pretty unsure’; and 2) how normative they thought that judgment
was in comparison to most speakers by rating themselves as ‘liberal’, ‘middle-of-the-
road’, or ‘conservative’. By way of illustration, test sentences rated as grammatical,
ungrammatical, and somewhere in between (in mean judgment score terms) are given

below in (31a) to (31c), respectively:

(31) a. the doctor is sure that there will be no problems
b. what will the grandfather clock stand between the bed and?
c. the idea he wasn’t in the store is preposterous.

Four differences between linguists and non-linguists were observed in the
responses each group gave to the 12 sentences exemplified by the three examples in (31)
above. First, it was found that, on average, the linguists were slightly less confident of
their judgments than the non-linguists, as reflected in their indices of confidence
assessing the sentences. The linguists marked the ‘pretty unsure’ option three times more
frequently than their counterparts (6% versus 2%) whereas both groups chose the ‘pretty
sure’ choice with almost identical frequency (73% for the former versus 75% for the
latter). Second, the linguists rated themselves as less conservative than the non-linguists
in judging the sentences. Although both groups were similar with respect to the
percentage of liberal judgments (22% for the linguists versus 24% for the non-linguists),
the former gave ‘conservative’ judgments only about half as often as the latter (13% vs.
24%). Third, the linguists rated the sentences as overall more grammatical than did the

non-linguists. While both groups used roughly the same number of the ‘somewhat

51

grammatical’ and the ‘perfectly ungrammatical’ choices, the linguists selected the
‘perfectly grarnmatical’ option approximately 35% more times than the non-linguists
whereas the latter opted for the ‘somewhat ungrammatical’ option about 45% more times
than the former. Fourth, the linguists made fuller use of the four levels of grammaticality
than the non-linguists: 33% of the linguists used each category of grammaticality (1 to 4)
at least twice whereas only 23% of the non-linguists did so.

Although no statistical significance was considered in Ross's (1979) analyses, it
appears that linguists, while less confident and less stringent sentence raters than non—
linguists, nevertheless attempted to make finer grammaticality distinctions than their
counterparts.

More recently, Coppieters (1987) addressed the issue of ultimate attainment in
SLA by investigating whether near-native speakers of French attained essentially
identical grammars of the language as developed by NSS. In the course of data analysis,
he examined NSs’ judgments closely and found variation among the NSS.

Data were collected from 21 near-native speakers of French whose linguistic
abilities were indistinguishable from NSS in language use, and from 20 NS respondents
with varied backgrounds ranging from five French linguists and five professors of
language or literature to one caretaker and one retired public employee. Each subject in
the two groups was interviewed individually by the researcher using a questionnaire of
107 sentences. The questionnaire consisted of two components: 1) a preference test in
which subjects were asked to indicate and provide reasons for a preference of one string
over the other in 20 pairs of closely related sentences and 2) an acceptability judgment

test in which they were requested to judge the acceptability of 66 sentences.

52

In his analysis, Coppieters (1987) first developed a prototypical profile of
grammaticality for the test sentences on the basis of the majority of the NSS’ judgments.
More than 80% of the NS subject group were found to agree on their judgments of 90 out
of 107 test sentences (an agreement rate of 84.1%). Individual NSS’ judgments of the test
sentences ranged in disagreement with the established profile from 5% to 16%. Next,
when their responses were examined in light of sentence types, the subjects diverged
considerably from the prototypical norm (e. g., 10% or above) in four out of a total of nine
areas, including article usage, contrasts between preposed and postposed adjectives, and
the like. By contrast, the NS subjects conformed to the established norm more than 90%
in the remaining five areas such as contrasts between imparfait and passé compose, A-
over-A Constraint, etc.

Although Coppieters (1987) did not attempt to provide an overarching
characterization of the sentence types on which the NS subjects tended to agree or
disagree in their responses, his study clearly demonstrated that NSS’ (both linguists and
non-linguists) grammaticality judgments vary in consensus by differing degrees
depending on the linguistic features being tested.

This section has summarized a number of representative studies addressing the
issue of differences in grammaticality judgment between linguists and non-linguists.
Despite a considerable body of evidence that these two groups of NSS differ in
performing grammaticality judgments, the question of how they differ remains

unresolved.

53

1.6. Use of grammaticality judgment tasks in SLA research
This section considers various issues concerning the use of grammaticality
judgment tasks in SLA since the present study employs a grammaticality judgment task

for data collection.

1.6.1. Advantages of grammaticality judgment tasks

There are three practical reasons often given to justify the use of grammaticalin
judgment data over spontaneous speech data in an investigation of interlanguage (IL)
grammar (see Juffs, 1996; White, 1989). First, certain linguistic behaviors of L2 (second
language) learners are not readily accessible to researchers in production data (such as
naturally occurring conversations) because they occur either rarely or not at all.
Furthermore, learners have been observed to avoid using structures that they find difficult
(Schachter, 1974). In the case of Japanese passives, it was found that learners rarely use
ni indirect passive sentences in compositions at least (Tanaka, 1992; Watabe et al., 1991).
Thus it would be highly problematic to gather large samples of various kinds of passive
sentences from L2 production data. Second, related to this, the use of grammaticality
judgment tasks allows a researcher to obtain information on strings that do not exist in
language, i.e., ungrammatical strings constructed to violate certain principle(s) of the
language. Third, in observing and collecting naturally occurring speech samples, it is
difficult to reliably distinguish slips of the tongue, unfinished utterances, etc. from

ungrammatical production reflecting IL grammar.

54

 

1.6.2. Validity and reliability issues of grammaticality judgment tasks in SLA
research

Following the adoption of this particular methodology from the research of
theoretical linguistics, a considerable body of L2 research has subjected to close scrutiny
the assumption that grammaticality judgment tasks provide indirect measures of learners’
linguistic competence from which inferences can be made about learners’ linguistic
development. (For views advocating the use of grammaticality judgment tasks in L2
research, see Bley-Vroman, Felix, & Ioup, 1988; Chaudron, 1983; Cowan & Hatasa,
1994; Gass, 1983; 1994; Hedgcock, 1993; Juffs, 1996; Munnich, Flynn, &
Martohardjono, 1994; Schachter & Yip, 1990; White, 1989; and, for reserved views,
Birdsong, 1989; Davies & Kaplan, 1998; Ellis, 1991; Goss, Zhang, & Lantolf, 1994.)
The various findings have led to the conclusion that L2 learners’ grammaticality
judgments reflect complex processes in which linguistic competence interacts with
performance variables in intricate ways. The emerging consensus is that they can provide
valid and reliable data for L2 research when careful and appropriate care has been taken
to reduce confounding variables, both linguistic and extralinguistic, and when L2
judgment data are examined as phenomena involving linguistic, metalinguistic, and
cognitive behaviors. In the empirical studies under review below, attempts were made to

factor out various factors involved in L2 leamer’s grammaticality judgment processes.

1.6.2.1. Validity issues of L2 grammaticality judgment

In addressing the validity issue of grammaticality judgment in SLA research,

Davies & Kaplan (1998) compared strategies involved in arriving at judgments in L1 and

55

L2 to determine if there were any similarities or differences in the two contexts. The
subjects of their study comprised 37 English-speaking learners of French in a fourth
semester college-level course at a university in the US. These subjects performed both
L2 French and L1 English grammaticality judgment tasks: in both tasks they were
instructed to rate 12 sentences on a three-point scale: OK, Not OK, and Not Sure. The
two tasks were separated by a five-minute intermission. Thirteen dyads of them
performed the tasks in a speak-aloud format in which they were instructed to verbalize
whatever they were thinking in arriving at grammatical judgments. The remaining eleven
subjects carried out the task in an individual format. This format was included to insure
that the judgments derived through dyadic interaction matched those made by the
individuals.

Upon inspection of the transcription of the audio tape-recorded dyads, Davies &
Kaplan (1998) identified strategies employed for the L2 grammaticality judgment task
highly similar to those reported by Ellis (1991) (where the study adopted individual
speak-loud protocols). These included linguistic intuitions-based judgment, use of
explicit or learned knowledge, a meaning-based response, repair of a perceived problem,
translation, analogy from some perceived similar structure, and guessing. Similarly, Ll
judgments involved linguistic intuitions-based judgment, a meaning-based response,
repair, and use of learned knowledge.

On the other hand, Davies & Kaplan (1998) identified two differences in use of
strategies between L1 English and L2 French grammaticality judgment contexts: l) the
number of strategies employed per sentence and 2) particular types of strategies

frequently used in each context. The first difference refers to the finding that in judging

56

Ll sentences, the dyads used, on average, 1.29 different strategy types per sentence while
in rating L2 sentences, they employed 1.9 strategy types on average. The second
difference is that the dyads gave linguistic intuitions-based judgments to 86.5% of the L1
sentences judged, by far the most frequently used strategy. By contrast, in the L2
context, their most frequently used strategy was use of explicit or learned knowledge
(57.8%) followed by meaning-based responses (43.7%). Linguistic intuitions-based
judgments were third (39.3%). These findings led Davies & Kaplan (1998) to conclude
that the subjects of their study made L1 and L2 grammaticality judgments in different
ways. They thus claimed that “L2 GJS [grammaticality judgments] are not necessarily as
representative of a speaker’s [linguistic] competence as it is assumed that L1 GJS are” (p.
198).

There are, however, several problems in their study. First, a concern has been
raised over the validity and reliability of this line of research methodology. Specifically,
an identification of strategies from transcribed data may risk susceptibility to subjective
interpretation when there are no standard data coding procedures established (Cowan &
Hatasa, 1994). Second, Davies & Kaplan's (1998) study simply showed that there were
quantitative differences in use of strategies between L1 and L2 grammaticality
judgments; it failed to demonstrate a clear qualitative difference between the two
contexts. Moreover, Davies & Kaplan themselves acknowledged a possible positive
correlation between the use of Ll-like strategies and learners’ proficiency levels, and
speculated that “strategies used for L2 GJS are reminiscent of strategies used for
judgments of very complex and difficult-to-process L1 structures” (p. 202) (see also Goss

et al., 1994 who made a similar observation). If so, contrary to the authors’ claim, the

57

two considerations above suggest there is no fundamental difference in the ways L1 and
L2 grammatical judgments are given.

Next, in consideration of grammaticality judgment as a complex product of
interactions between grammatical and processing factors, Schachter & Yip (1990) further
investigated Schachter's (1989) finding that both NSs and NNSs of English found subject
extraction sentences more difficult to judge than object extraction sentences. An example
of the former is ‘What did the nurse say she reported t had happened to the patient?’;
‘What did the nurse say she reported the patient had taken t?’ exemplifies the latter type
(Schachter & Yip, 1990, p. 387): the t in the exemplar sentences indicates the extraction
site of a wh-word. The subjects of their study included 20 English-speaking freshman
university students, 20 Chinese-speaking and 20 Korean-speaking NNSs who had passed
the English proficiency exams required for the university enrollment. They rated 54 test
sentences on a four-point grammaticality scale: clearly grammatical, probably
grammatical, probably ungrammatical, and clearly ungrammatical. A questionnaire
included, inter alia, wh-subj ect and wh-object extraction type questions as shown above.
There were three tokens of each type, all grammatical, at four-clause, three-clause (such
as the above examples), and two-clause levels, respectively.

Individual judgment scores were obtained by assigning scores of three to zero to
responses of ‘clearly grammatical’ to ‘clearly ungrammatical’, respectively. Since there
were three tokens of each type, the maximum score was nine. It was found that NSs’
mean scores were consistently higher for wh-object extraction questions than for wh-
subject ones throughout the clause levels. They also showed an inverse relationship

between the level of embeddedness and the degree of grammaticality for both wh—subj ect

58

and wh-object extraction questions. Specifically, these ranged from a mean of 7.8 for
one-clause movement to 5.6 for three-clause movement for objects, and from 6.5 to 4.0
for subjects. Identical phenomena were observed in NNSS’ data. Chinese-speaking
NNSS’ mean scores ranged from 6.9 to 5.1 for objects, and from 5.1 to 2.8 for subjects.
Korean—speaking NNSs demonstrated the same judgment pattern.

Clearly, the inverse relationships between embeddedness and grammaticality
were due to the processing difficulty of identifying a wh—word with its embedded
extraction site. Schachter & Yip (1990) offered a further account for the judgment
differences between subject and object extractions in processing terms. In brief, they
argued that the subject extraction sentence is more difficult for on-line processing since it
requires a more demanding revision of a phrase marker being constructed at the initial
parsing, in order to arrive at an eventually correct parsing (see Schachter & Yip, 1990,
pp. 387-3 89 for more details).

On the basis of these findings, they claimed that both NSS and NNSs, when
confronted with a highly complex or lengthy sentence to judge, may reject or accept it
regardless of its grammaticality status. This is due to processing difficulty, not the ability
or inability to assess grammaticality based on linguistic knowledge. This study clearly
demonstrated the importance of taking into consideration the complex interplay of both
linguistic and processing factors in eliciting and interpreting metalinguistic data.

Finally, there have been various attempts to demonstrate that grammaticality
judgment data are a reliable measure of learner’s linguistic knowledge by advancing
other performance data that converge with those from metalinguistic judgments. Leow

(1996) conducted one such study that investigated whether a relationship exists between

59

learners’ grammaticality judgments and their performances in production tasks at two
substantially different stages of their L2 development. In that study 30 English-speaking
learners of Spanish who were enrolled in the first semester of a college-level Spanish
course not only carried out a grammaticality judgment task, but also performed both oral
and written production tasks in two occasions separated by 11 weeks.

The linguistic item selected for the study was the agreement feature between
nouns/noun phrases and adjectives or past participles (which function like adjectives in
Spanish). In the first metalinguistic task, there were 33 test sentences to discriminate in
grammaticality, of which 20 contained ungrammatical target forms. The subjects were
told to correct the errors and provide a reason for the correction. In the second
production tasks, they carried out the oral and written activities in which they provided
responses to questions about a series of drawings depicting everyday episodes that lent
themselves to the use of the target agreement forms. Forty-one questions (including 21
distractors) were also designed to encourage the use of target items in the responses. The
three tasks described above were administered as a set on two occasions: Session 1 took
place during the third week of the language course and Session 2 in the fourteenth week
of the semester.

The mean scores for oral and written performance on the three tasks in Session 1
and 2, respectively, were as follows (a full score was 20 for each task): for the
grammaticality judgment task, 4.68 and 9.13; for the written production task, 8.68 and
11.17; and for the oral production, 9.50 and 12.60. Significant positive correlations were
found for all the four comparisons between the grammaticality judgment tasks and the

oral/written production tasks in Session 1 and 2 (i.e., r = .737 and .733, p < .01 between

60

the grammaticality judgment and the written production task at Session 1 and 2,
respectively; r = .515 and .602, p < .01 between the grammaticality judgment and the oral
production, respectively).

Thus, this study showed not only that the judgment data mirrored other types of
L2 data, namely, the oral and written production data, but also that the performances on
grammaticality judgments appeared to reflect developing IL grammars, measured at
Session 1 and 2, eleven weeks apart. This led Leow (1996) to conclude, “grammaticality
judgments appear to reflect behavioral patterns of L2 development and, consequently,
may be used as a reliable instrument for SLA research” (p. 135). (See also Mandel],
1999; Murphy, 1997 for studies validating L2 grammaticality judgment with other L2

data; and Christine & Lantolf, 1992 for a study invalidating the use of grammaticality

judgment.)

1.6.2.2. Reliability issues of L2 grammaticality judgment

As reviewed above, empirical evidence indicated that some within- and between-
subjects’ variability exists in NSs’ grammaticality judgments. Some degree of L2
judgment variability thus comes as no surprise (Birdsong, 1989; Ellis, 1990; 1991).
There is, furthermore, one potential source for judgment inconsistency peculiar to SLA,
i.e., IL indeterminacy, which refers to “the learner’s incomplete knowledge or absence of
knowledge of parts of the second-language grammar” (Gass, 1994, p. 305). It is thus
important to take into consideration this evolving, indeterminate nature of IL grammar

(Sorace, 1988) in dealing with L2 grammaticality judgment data.

61

Gass (1994) demonstrated one way of doing so. Her study included 23 subject
learners of English (Chinese, Japanese, and Korean speakers) who were enrolled in ESL
classes to satisfy the language requirement for university admission. They were told to
assess a set of 24 relative clause sentences (half of which were grammatical and the other
half ungrammatical) comprised of two tokens of each type together with 6 distractors.
This instrument was administered twice, with one week separating the two (Time 1 and
Time 2, respectively). In both sessions subjects made two kinds of assessments: first to
make categorical grammaticality judgments, either Correct or Incorrect, and, second, to
rate the confidence in their first judgment using a seven-point scale ranging from —3
(definitely incorrect), to 0 (unsure), to +3 (definitely correct).

The test sentences were constructed to include all six types of relative clauses
based on the Accessibility Hierarchy (Keenan & Comrie, 1977)”. There is empirical
evidence indicating that the Accessibility Hierarchy reflects degree of acquisition
difficulties (Doughty, 1991; Gass, 1979); that is to say, a relative clause type lower on the
Hierarchy is more difficult to learn than any higher.

Three major findings emerged. First, Gass (1994) reported that the overall
reliability coefficients between Time 1 and Time 2 were significant (r = .5979, p < .01 for
the categorical judgments, and r = .6443, p < .01 for the scalar judgments). Second,
when the subjects’ responses were examined according to the relative clause types, a
greater consistency was found in judgments of relative clause types high on the Hierarchy
and the least consistency in those low on the Hierarchy (i.e., in the categorical judgments
r = .7620, p < .001 for the subject type and r = .4789, p < .05 for the object of

comparative type; and in the scalar judgments r = .8302, p < .001 for the subject and r =

62

.5442, p < .01 for the object of comparative type). Third, when those judgments which
showed erratic changes between Time 1 and Time 221 were eliminated, even greater
reliability coefficients were obtained (e. g., in the categorical judgments r = .8206, p <
.001 for the subject type and r = .5933, p < .01 for the object of comparative type; and in
the scalar judgments r = .9025, p < .001 for the subject and r = .7055, p < .001 for the
object of comparative type).

Conclusions drawn from these findings were three-fold. First, unlike Ellis (1990),
this study did not find that L2 grammaticality judgments given on two separate occasions
were unreliable. Second, support was found for the view that the degree of reliability of
L2 judgment data is a fimction of syntactic factors. That is to say, lower reliability of the
judgment data of the study was found where greater learning difficulty is predicted by the
Accessibility Hierarchy. Third, it was experimentally demonstrated that L2 judgments
deriving from indeterminate IL grammar may be isolated and eliminated, resulting in
increased reliability of the data. (See also Johnson, Shenkrnan, Newport, & Medin, 1996
for discussion of indeterminacy in adult asymptotic (i.e., highly advanced) L2 learners’
grammars in view of qualitative (as well as quantitative) differences fiom NSs’
grammar.)

On the basis of the findings and suggestions offered in the extant studies reviewed
above, great effort and care have been given to reducing linguistic and extralinguistic

variables in designing the data collection method employed in the present study.

63

CHAPTER 2

PREDICTIONS

Predictions are made separately for NSS and NNSs of Japanese since the research

questions addressed are different for the two groups.

2.1. Predictions for native speakers of Japanese

Because Kuroda's (1979) analysis of the semantic properties of Japanese passives
hinges on the recognition of highly subtle differences in connotation in passive
sentences', it is of critical importance to demonstrate that ordinary NSs’ grammaticality
judgments on semantic properties of passives in particular mirror his predictions. At the
same time, an interesting question arises as to whether or not there would be any
grammaticality judgment differences among NSS of Japanese in assessing semantic
properties of passive sentences in view of such delicate nuances under consideration.
This question may be addressed by collecting data from multiple NS groups who may
differ in judgment performance. One such group difference may be elicited by gathering
data from NSS both with and without linguistic training.

There are experimental studies which investigated the effect of linguistic training
on NS’s grammaticality judgment ability as one of the judge’s internal factors influencing
grammaticality judgment processes. The literature has been inconclusive, however, on
such effects. Some studies found that NSS with linguistic training (henceforth referred to

as linguists) could give more informative grammaticality judgments than NSS without

64

such background (abbreviated as non-linguists), while other studies showed that
linguists’ judgments were highly distinct from non-linguists’.

Thus, two predictions are given below: Prediction 1a concerns whether or not
ordinary NSS of Japanese (defined as non-linguists) observe Hoshi (1994a; 1999) and
Kuroda's (1979) characterizations of Japanese passives; Prediction 1b has to do with the
issue of the presence or absence of grammaticality judgment differences between

linguists and non-linguists.

Prediction 1a: Ordinary NSs of Japanese will give theory-consistent judgments in

assessing the properties of Japanese passives;

Prediction lb: The judgments of linguists and non-linguists will differ in their

assessments of the properties of Japanese passives.

2.2. Predictions for English-speaking non-native speakers of Japanese

The present study investigated English-speaking NNSs’ knowledge of syntactic
and semantic pr0perties of Japanese passives at different proficiency levels. Predictions
on their syntactic and then semantic knowledge of them follow.

First, in examining English-speaking NNSS’ syntactic knowledge of Japanese
passives, their knowledge of case-marking phenomena in passivization processes is

investigated (see Introduction and Materials for illustrations of such phenomena):

65

Prediction 2: The ni direct passive will be rated better than the ni yotte and the m’ indirect
passives in judging both grammatical and ungrammatical sentences in terms of case

markings resulting from passivization processes;

Prediction 3: The m' yotte passive will be rated better than the ni indirect passive in
judging both grammatical and ungrammatical sentences in terms of case markings

resulting from passivization processes.

Of the three types, the ni direct passive generally is learned first by English-
speaking learners owing to its surface similarity to the English passive and its frequency
of use compared to the other two types. The ni yotte passive and the ni indirect passive,
on the other hand, present different kinds of learning problems. Because the ni yotte
passive is preferred and used mostly in formal writing, non-advanced learners may suffer
input paucity and thus seem to encounter some learning difficulty despite the fact that the
structure is syntactically an exact match to the English passive. In contrast, the ni
indirect passive is syntactically entirely different from the English passive, which poses a
primary learning difficulty. The difficulty with this type of passive was reported in the
literature as learners’ avoidance of its use (Tanaka, 1992).

Next, in gauging English-speaking NNSS’ knowledge of the semantic properties
of passives, the following features are pertinent to the present study: the presence and
absence of an adversative reading, the presence and absence of the grammaticality
contrast in perfective versus non-perfective readings, and verb-induced viewpoint

differences (see Introduction and Materials for illustrations of these properties). These all

66

have to do with the difference between a 0 (affectee)- and a non-O-passive subject status,
reflected in the ni versus ni yotte marking contrast. In the following, a prediction is given

concerning each of the properties above:

Prediction 4: An adversative connotation will be read clearly in the ni indirect passive in
the sense that a grammaticalin distinction will be made between adversative and non-
adversative readings: To a lesser extent, an adversative connotation will be read in the ni

direct passive.

The observation has been made in the literature and has received empirical
support in Spees (1992) that the ni indirect passive carries a predominantly adversative
reading. In contrast, the degree to which the ni direct passive is read with an adversative
connotation is weaker (see Wierzbicka, 1979). Thus, a grammaticality contrast will be
made when the adversative and non-adversative readings are matched with the ni indirect
passive. Similarly, but to a lesser extent, the ni direct passive will be judged contrastively

in grammaticality between the adversative and non-adversative readings.

Prediction 5: The ta- and iru-marked ni direct passive will be judged as grammatical

when it is read as perfective; however, it will not be judged as ungrammatical when read

as non-perfective;

Prediction 6: The ta- and iru-marked ni yotte passive will be judged similarly whether it

carries a perfective or a non-perfective reading.

67

The first part of Prediction 5 is straightforward: if NNS subjects have learned the
syntactic basics of the ni direct passive, they will accept it. On the other hand, the second
part holds that subjects will not be able to show a contrastive judgment of the ni direct
passive with a non-perfective reading vis-a-vis perfective one. Rather, they will accept
both perfective and non-perfective versions of the ni direct passive. This distinction is so
subtle that it is unclear if they may be able to recognize it. Indeed, until Kuroda (1979)
unraveled it, this phenomenon had been entirely unknown to Japanese linguists and still
requires more thorough explanation. It is highly unlikely that subjects have received
input, in a systematic fashion conducive to learning, which would indicate to them that
non-perfective versions of the ta- and iru-marked ni direct passive is ungrammatical. In
contrast, without such restriction on possible readings, the ni yotte passive will be judged

without any difference irrespective of perfective and non-perfective readings.

Prediction 7: The indirect passive with an adversative reading will be rejected clearly
when marked by ni yotte; and, to a lesser extent, the direct passive with an adversative

reading will be rejected when marked by ni yotte.

As noted above, a predominant adversative reading in the indirect passive
strongly renders the indirect passive incompatible with a ni yotte marking which indicates
a non-O-passive subject status. With a reduced tendency for an adversative reading, the
degree of incompatibility between the direct passive and a ni yotte marking will be less

aggravated.

68

 

Prediction 8: The ni yotte passive will be judged as grammatical when used in a context
inducing an objective viewpoint, while it will be judged as ungrammatical when used in a

personal context where the ni direct passive is appropriate.

In the literature (see Alfonso, 1980; Jacobsen, 1992; Kuno, 1983; Kuroda, 1979;
Martin, 1975; Miyagi, 1999; Shibatani, 1990; Wierzbicka, 1979) the ni yotte passive is
characterized as preferred in formal writing. Perhaps it is this area in which learners of
Japanese are most likely to be exposed to the ni yotte passive. It is also a most common
pedagogical characterization of this type of passive. Hence, it is in this area that learners

initially learn about the ni yotte passive.

69

CHAPTER 3

METHOD

3.1. Subjects

The subjects of the present study included native speakers of Japanese and
English-speaking learners of Japanese. There were two groups of NS subjects: the first
group included 21 native speakers of Japanese who at the time of the data collection had
been studying or had studied Japanese theoretical linguistics, particularly syntax and
semantics, in graduate programs at eight different institutions.‘ These 21 NSS included
nine doctoral students, six MA students, two university professors holding a PhD degree,
two visiting researchers with an MA degree, and two university seniors. This group was
classified as a linguistically sophisticated group owing to their solid linguistic training
(see Appendix A for more detailed relevant background information). The second NS
group included 31 native speakers studying sociology at the undergraduate level in a
university in Tokyo. It was comprised of 10 sophomores, 15 juniors, and 6 seniors. A
check of the background information sheet they filled out at the outset of the data
collection procedure assured that no one in this group had ever taken a linguistics
course.2 Note also that their subject major, sociology, is not much concerned with
language, which justified terming this group as linguistically na'r've.3 (See Appendix B
for more detailed relevant background information.)

There were 81 subjects in the English-speaking NNS group“, including
undergraduate students, graduate students, and professors from twelve major universities

in the US along with three subjects residing in Japan who were all US university

70

graduates.5 The majors or specializations of forty-nine of them (60.5%) were J apan-
related fields such as Japanese linguistics, Japanese literature, Japanese language, Japan
studies, and East Asian Studies.

These 81 NNS subjects were assigned to three groups, (tentatively) termed as
Group A (25 subjects), B (36 subjects), and C (20 subjects), based on their scores on part
of the Japanese Language Proficiency Test (J LPT) taken during the data collection
procedure. Group A JPLT scores ranged from 100% to 88% with a mean score of 23.76
out of a full score of 25; Group B from 84% to 64% with a mean of 18.57; and Group C
from 60% to 32% with a mean of 11.40 (see the data coding section for scoring
procedures). The cutoff line between groups A and B was set at approximately 90%, that
between groups B and C at 60% which corresponds to the official pass-fail cutoff line set
for the Level Three J LPT adopted in the present study. There was a highly significant
difference in group mean scores (F(2, 78) = 320.403, p < .0005). The Tukey HSD result
showed that all three groups were significantly different from one another.

The average length of study of each group was 9.74 years for Group A, 4.16 years
for Group B, and 4.32 years for Group C, a calculation based on the self-reported
information in the background information sheets. Twenty-three members of Group A
(92%), 27 of Group B (75%), and 13 of Group C (65%) each had visiting experiences in
Japan in the capacity of either research, work, exchange student programs, study,
homestays, or vacations. The average duration in Japan of each group was 4.2 years for
Group A, 20 months for Group B, and 9.9 weeks for Group C. Hence, groups A, B, and

C are henceforth called the highly advanced group, the low advanced group, and the

71

intermediate group, respectively. (See Appendix C for complete background information

on the NNSs’ subjects.)

3.2. Materials

The data collection materials for the NSS of Japanese consisted of a background
information questionnaire and a grammaticality judgment test. The NNSS’ materials
included a background information questionnaire, a Japanese language proficiency test

including a vocabulary test, and a grammaticality judgment test.

3.2.1. Background information questionnaire

Two different formats for the background information questionnaire were
developed for the NSS and the NNSs of Japanese respectively. The questionnaire given
to the NS subjects of Japanese was designed to obtain information regarding their
training in linguistics and other language-related experiences such as foreign language
learning and study abroad experiences, as well as general demographic information. It
included the following questions: linguistics-related courses previously taken, foreign
languages learned, study abroad experiences, major, the highest attained educational
level, age, and gender.

The questionnaire given to the NNS subjects of Japanese was devised to collect
information on learning experiences of Japanese and a general demographic background.
It contained questions concerning the length of learning Japanese, textbooks used,
previous visiting experiences in Japan, out-of—classroom contact with Japanese, the most

recent Japanese language-related course taken and its grade, foreign/second language

72

learning experiences other than Japanese, primary language(s), age, gender, the highest
attained educational level, and major. (See Appendices D and E for the actual

questionnaires in both formats.)

3.2.2. Japanese language proficiency test

It was critical to obtain an independent measure of the English—speaking subjects’
general proficiency levels in Japanese to serve as the basis upon which they would be
assigned to different proficiency groups. Part of the 1998 Japan Foundation Japanese
Language Proficiency Test—which is essentially a Japanese language version of
TOEFL— served this purpose (see Nabei & Busch, 1999 for discussion of the validity
and reliability of the JLPT as a standardized proficiency measurement test). After pilot
testing to determine which level of the JLPT6 would be appropriate for the proficiency
levels of learners likely to participate in the present study, the Level 3 JLPT was chosen.

The general criteria for Level 3 state that an examinee “knows about 300 kanji
(Chinese characters) and 1,500 vocabulary words, and has the ability to take part in
everyday conversation and to read and write simple sentences” (Japan Foundation &
Association of International Education, 1994; Japan Foundation & Association of
lntemational Education, 1999). This level is broadly characterized as reachable after
studying Japanese for approximately 300 hours. Since it was originally expected that the
majority of the NNSs’ subjects in the present study were likely to be third and fourth year
learners of Japanese at universities, the Level 3 JLPT appeared to match their proficiency

levels. Another important consideration was that the Level 2 JLPT was heavily inclined

73

to test exarninee’s lexical knowledge of fairly specialized vocabulary rather than syntax
or morphology.

Since subjects’ syntactic and semantic knowledge of Japanese passives was
under investigation, the grammar component of the JLPT was selected: the other JLPT
components of vocabulary, reading comprehension, and listening comprehension were
not included in the administered proficiency test. This consisted of fill-in-the-blank,
multiple-choice type of questions: namely, a subject had to fill in a blank found either in
a single sentence or in a conversation with two turns, by choosing and circling the
appropriate word(s) from four choices given. It was designed to elicit his or her
knowledge of syntax and morphology as well as lexicon. Twenty-five test sentences out
of a total of 38 were randomly selected. (The actual proficiency test appears in Appendix

F.)

3.2.3. Vocabulary test

A vocabulary test was necessary to ensure that the English-speaking subjects were
familiar with words used in certain grammaticality judgment test material. This was
particularly important in order to gauge their knowledge of certain semantic properties of
Japanese passives since grammaticality status of such passive sentences was conditioned
in some cases largely by the lexical properties of words used, particularly verbs. In other
words, it was essential to exclude from analyses those subjects who were not familiar
with key verbs since it would be impossible to infer from their grammaticality judgments
if they had knowledge of the relevant properties of passives. The test was devised

specifically for this purpose.

74

It consisted of a translation task from Japanese to English which required
choosing an appropriate English equivalent for a Japanese word from four choices

presented in the test and circling it as shown below:

(32) satugai—suru

harm poison assault

This type of test was adopted in view of time constraints. It had six vocabulary items all
of which were used in test sentences to characterize the nature of those passive sentences.
As shown in the example above, a special effort was made to construct choices of verbs
with highly similar meanings, intended to only elicit the subjects’ solid lexical
knowledge. (The actual vocabulary test appears in Appendix G.) Both the proficiency
test and the vocabulary test constituted pre-tests in the present study. A suggested test
time of 20 minutes for this portion of the NNS test materials was given on the first page

of the pre-test.

3.2.4. Grammaticality judgment test

A grammaticality judgment test was employed to investigate both J apanese-
speaking and English-speaking subjects’ knowledge of syntactic and semantic properties
of Japanese passives. In this test the subj ects had to make judgments on sentences using
a five point-scale of grammaticality and complete it by circling the response they chose

as illustrated below:7

75

(33) Mary-wa J ohn-ga itumo soodan-s-are-ru

Acceptable Unacceptable Not Sure

Somewhat Acceptable Somewhat Unacceptable

In order to ensure that the subjects understand what ‘acceptable’ and
‘unacceptable’ sentences were intended to be in the present study, an instruction Sheet for
the grammaticality judgment test was provided (see Schiitze, 1996 for the importance of
clarifying for sentence-rating subjects what is meant by acceptability in the theoretical
construct.) It was designed to serve the following three purposes (following Bley-
Vroman et al., 1988): l) to inform the subjects that the objective of the test was to seek
their intuitions about what constitutes acceptable and unacceptable sentences in Japanese;
2) to illuminate the concepts of ‘acceptable’ and ‘unacceptable’, ‘somewhat acceptable’
and ‘somewhat unacceptable’ with contrastive pairs of exemplar sentences, stressing that
intuitions about acceptable or unacceptable sentences often result from a ‘feel’ for them
rather than knowledge of a rule, and pointing out that different speakers may have
different intuitions about a particular sentence; and 3) to provide instruction on test-
taking procedures. (The actual instruction sheet is included in Appendix H.)

A total of 84 test sentences (44 grammatical and 40 ungrammatical) were selected
for the grammaticality judgment test on the basis of extensive pilot testing of NSS of
Japanese. It contained 62 test sentences pertaining to Japanese passives, including 16
sentences devised to investigate knowledge of syntactic properties (6 grammatical and 10
ungrammatical) as well as 46 sentences testing knowledge of semantic properties (26

grammatical and 20 ungrammatical). Additionally, it contained 22 distractors (12

76

grammatical and 10 ungrammatical). It was expected to take the NNS subjects 45
minutes to over an hour to complete it.

There were two test tokens for each of the syntactic and semantic properties under
investigation. More tokens for each sentence type would have been ideal to increase the
reliability of the study (Cowan & Hatasa, 1994), but this would have amounted to an
unrealistically long test. Considering the fact that the present study is the first of its kind
to look closely into English-speaking NNSS’ knowledge of Japanese passives, it was
decided that the test should encompass a wider range of properties of Japanese passives at
the expense of a fewer number of tokens. The distractors were devised to include among
other things, (other-than-passive) structures used in test sentences that were important in
arriving at correct judgments on passive sentences. Subjects’ judgments of such
distractor items were taken into account, as with the vocabulary test, at the time of
analyses of the relevant passive test sentences.

Test sentences for a given syntactic or semantic property of passives were
constructed to be of comparable length and structural complexity; vocabulary items were
also carefully chosen so that they were likely to be familiar to the subjects at intermediate
and advanced proficiency levels. There were, however, sentences in which some
complex structures or highly advanced vocabulary items were necessary to embed certain
semantic properties of passives. In these cases, as well as when there was any concern
that intermediate-level subjects might not be familiar with certain words used, English
glosses were provided for those lexical items. To minimize ordering effects and the
potential effects of subjects’ fatigue on their judgments, the test sentences were arranged

in two random orders with half of the subjects presented each order. (See Birdsong,

77

1989; Bley-Vroman et al., 1988; Cowan & Hatasa, 1994; Cowart, 1997; Ellis, 1991;
Schiitze, 1996, for discussions of appropriate measures for controlling subj ect-internal
and -external factors which may increase judgment variability.)

The test sentences were identical for the NS and NNS versions with the following
two exceptions: 1) when a proper noun such as a personal name was used in a test
sentence, a Japanese name and an English one were used for each version respectively.
For example, T anaka ‘Mr./Ms. Tanaka’ appeared in the Japanese version and John in the
English version: and 2) in the English version, each test sentence was given both in
Japanese and its Romanized equivalent (see Appendix I for the sample grammaticality

judgment test sheet in the English version).

3.2.5. Illustrations of test sentences

The following illustrates each of the syntactic and semantic properties of Japanese
passives under investigation as well as some distractors with their actual test sentence
exemplar(s). (Appendix J provides a complete list of the test sentences used in the
grammaticality judgment questionnaire together with their English translations.8 See also

the Introduction for the theoretical framework adopted in the thesis.)

3.2.5.1. Testing the knowledge of syntactic properties of Japanese passives

Test sentences were constructed to investigate subjects’ syntactic knowledge of
the three types of Japanese passives: the ni yotte passive, the ni direct passive, and the ni
indirect passive.9 The aim was attempted to examine their knowledge of case marking

resulting from passivization operations. Empirical evidence in L1 performance data

78

(including grammaticality judgment data) has revealed that the knowledge (or lack
thereof) of syntactic case marking in Japanese reflects the presence (or absence) of
knowledge of syntactic operations (including passivization) which underlie such case
assignment phenomena (Hagiwara, 1993; Hagiwara & Caplan, 1990; Otsu, 1999;
Ryuzaki & Ito, 1999). It would therefore be safe to draw inferences about their
knowledge of syntactic operations of passives from that of case assignment phenomena
resulting fi'om passivization processes. In the following, test sentences for each type of

passive are illustrated in the order given above.

3.2.5.1.]. The niyotte passive

The ni yotte passive involves two passivization operations: accusative case
absorption and external 0-role suppression, a process identical with English passivization.
In brief, in the derivation of the ni yotte passive, as a result of the first operation, the
internal argument of a verb gets moved to assume the nominative case ga rather than the
accusative case 0. On the other hand, due to the second operation (i.e., external 0-role
suppression), the external argument of the verb appears as an adverbial ni yotte phrase.

These operations derive the ni yotte passive sentence (34b) below from its active

counterpart (34a):
(34) a. biitoruzu-ga utukusii uta-o takusan tukut-ta
the Beatles—NOM beautiful songs-ACC a lot make-Pst
‘the Beatles made many beautifirl songs’
b. utukusii uta-ga biitoruzu-ni yotte takusan tukur-are-ta
beautiful songs-NOM the Beatles-by a lot make-Pass-Pst

‘many beautiful songs were made by the Beatles’

79

c. "biitoruzu-ni yotte utukusii uta-o takusan tukur-are-ta
-by -ACC

d. *utukusii uta-wa biitoruzu-ga takusan tukur-are-ta
-TOP -NOM

If either of these two operations is absent in the derivation, an ungrammatical sentence
results. Indeed, it was attested in Tanaka (1992) that learners of Japanese produce both
types of ungrammatical ni yotte passive sentences (see (26a) for the lack of external 0-
role suppression and (26c) for that of accusative case absorption).

The first type of ungrammatical ni yotte passive sentences occurs with the lack of
accusative case absorption. The sentence (34c) above shows that the internal argument of
the verb uta ‘song’, maintains its original grammatical function as an objective NP as
shown by the accusative marker 0 being attached to it. This in turn suggests non-
application of the case absorption operation. On the other hand, the external argument
biitoruzu ‘the Beatles’, is demoted to the ni yotte adverbial phrase, suggesting the proper
operation of the external 0-role suppression.

The second type of non-operation is, conversely, concerned with the lack of
external G-role suppression. In the sentence (34d) above, the external argument of the
verb biitoruzu ‘the Beatles’, assumes a nominative case go. This retention of ga-marking
suggests the lack of external argument demotion to the adverbial phrase, i.e., non-
application of external B-role suppression. In contrast, the internal argument, uta ‘song’,
has been moved to get wav-marked10 suggesting the proper operation of accusative case

absorption.

3.2.5.1.2. The ni direct passive

80

 

The ni direct passive involves two passivization operations: external O-role
suppression and PRO movement. Unlike the ni yotte passive, these operations take place
under the lower VP domain, i.e., the complement clause of rare. Specifically, as a result
of external O-role suppression, the external argument of a lower clause verb demotes to an
adverbial ni phrase. PRO in turn moves to SPEC of the lower VP to get licensed. These

passivization processes are involved in grammatical ni passive sentences such as (35a)

below:
(35) a. Mary-wa John-ni itumo soodan-S-are-ru
-TOP -by always consult-Pass-Pres
‘Mary is affected by being always consulted by John’
b. *Mary-wa John-ga itumo soodan-s-are-ru
-TOP -NOM

When external 0-role suppression is not applied, however, such derivation results
in an ungrammatical sentence. It was observed in Tanaka (1992) that learners of
Japanese produce this kind of ungrammatical ni direct passive sentences (see (26e) for
her attested sentence). For instance, sentence (35b) above indicates that the external
argument of the lower clause verb, John, retains its nominative case marking as opposed
to the adverbial ni marking in (35a). This suggests that John has not undergone the
application of the external 0-role suppression operation, rendering sentence (35b)

ungrammatical. ' ‘
3.2.5.1.3. The ni indirect passive

The ni indirect passive does not involve passivization operations as do the ni

yotte passive and the ni direct passive. Thus, the internal argument of an embedded

81

 

clause verb retains its accusative case and the external argument of the verb is ni-marked
because the external ‘affectee’ argument of the passive verb must assume the nominative
case marker go. These processes lead to the grammatical ni indirect passive sentence

(36a) below: ’2

(3 6) a. J ohn-wa kodomo-ni kuruma no mado-o war-are-ta
-TOP child-by car window-ACC break-Pass-Pst
‘John was affected by a child breaking the car window’
b. *John-wa kuruma no mado-ga kodomo-m war-are-ta
-TOP car window-NOM child-by
c. *John-wa kodomo-ga kuruma no mado-o war-are-ta
-TOP child-NOM car window -ACC

Yet it was found in Tanaka (1992) that learners of Japanese produce incorrectly case-
marked ungrammatical ni indirect passive sentences (see (26g)).

For example, sentence (3 6b) above suggests that passivization operations have
applied in the lower clause, for the internal argument of the lower clause verb kuruma no
mado ‘car window’, has moved and been nominative ga-marked, resulting from
accusative case absorption, and the external argument kodomo ‘child’, has been demoted
and adverbial ni—marked due to external O-role suppression. On the other hand, sentence
(3 6c) above indicates that passivization operations have not applied, for the internal
argument kuruma no mado ‘car window’, retains its accusative case marker 0, and the
external argument kodomo ‘child’, assumes a nominative case marker ga. However, this
nominative case marking is in direct conflict with that of the matrix ‘affectee’ subject

John, resulting in the ungrammaticality of (36c).

82

3.2.5.2. Testing the knowledge of semantic properties of Japanese passives

Test sentences were constructed to investigate subjects’ semantic knowledge of
the three types of passives under investigation.” Specifically, the aim was to probe their
knowledge of the following semantic properties of Japanese passives which result from a
passive subject status difference, i.e., a 0 (affectee)—subject versus a non-O-subject as
reflected in the ni versus ni yotte marking contrast: 1) an adversative reading of the ni
direct and indirect passives; 2) a grammaticality contrast in perfective versus non—
perfective readings; 3) unavailability of an adversative reading of the ni yotte passive; 4)
verb-induced viewpoint differences. In the following, test sentences for each semantic

feature of passives are presented in the order given above.

3.2.5.2.]. Arr adversative reading of the ni direct and indirect passives
It is a widely accepted view that the ni indirect passive carries a predominantly

adversative reading as illustrated below:

(37) a. John-ga warui toki-ni tomodati-ni ko-rare-ta
—NOM at an inconvenient time friend-by come-Pass-Pst
‘John was adversely affected by his friend visiting him at an inconvenient time’
b. *John-ga ii toki-ni tomodati-ni ko-rare-ta
at a convenient time

""John was adversely affected by his fiiend visiting him at a convenient time’
(Kuroda 1979, p.314 & p.317)

The temporal adverbial phrase of negative connotation, warui toki ni ‘at an inconvenient
time’, is consistent with the adversative reading of the ni indirect passive sentence (37a);

on the other hand, the positive connotation of ii toki ni ‘at a convenient time’ in (37b) is

83

inconsistent with an adversative reading. Hence the former is grammatical while the
latter is rendered ungrammatical.
Similarly, an adversative reading may manifest itself clearly in the ni direct

passive such as in (38a) below:

(3 8)
a. Jane-ga mukasi no kare-ni ni-zikan mo mat-arete komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst
‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

b. * Jane-ga mukasi no kare-ni ni-zikan mo mat-arete uresikat-ta
happy-Pst
“Jane was happy by being adversely affected by having been waited for by her ex-boyfriend
for as many as two hours’

For the same reasons outlined above regarding the grammaticality contrast in the ni
indirect passive sentences, the ni direct passive sentence (38a) is grammatical while (38b)
is ungrammatical. This is so because the matrix verb of (38a) komatta ‘was annoyed’,
carries a negative connotation consonant with the adversative reading of the ni direct
passive subordinate clause of (38a); by contrast, the verb in (3 8b) uresikatta ‘was happy’,
connoting positivity, is not consistent with the adversative reading of the ni direct passive

clause of (38b).

3.2.5.2.2. A grammaticality contrast in perfective versus non-perfective readings14

The ni direct passive expresses the state of the passive subject somehow affected
by the event or state of affairs described by the remaining part of the passive sentence.
When it takes an inanimate subject it needs to satisfy a certain condition, e. g., a

perfective reading. This is so because a perfective reading makes the above subject-

84

affectedness connotation readily available since by definition the perfective describes the
state resulting from an event or state of affairs entering into its earlier situation (Comrie,
1976; 1981). In a non-perfective reading, by contrast, such an affected state is not
expressed straightforwardly. Hence, the ni direct passive with an inanimate subject is
inconsistent with a non-perfective reading. On the other hand, the ni yotte passive with
an inanimate subject does not impose such a restriction on possible readings of the
passive sentence since it does not express the subject-affected state in the same way as
the ni direct passive does. Thus, the ni yotte passive may be compatible both with
perfective and non-perfective readings.

There are two cases where such a perfective versus non-perfective reading
contrast is brought about in the ni direct and the ni yotte passives, namely, te iru and to
markings.” Both cases are illustrated here with test sentences.

The to iru form may express either a progressive or a perfective reading. In the
case of a progressive reading which does not express the affected state, the m' yotte

passive sentence with an inanimate subject (39a) is grammatical whereas the ni direct

passive (39b) is not:
(39)
a. atarasii konpyuutaa no puroguramu-ga John-ni yotte tukur-are-te iru
new computer program-NOM -by make-Pass-Prog-Pres
b. *atarasii konpyuutaa no puroguramu-ga J ohn-ni tukur-are-te iru

-by
‘a new computer program is being made by John’

On the other hand, with the perfective reading, both ni yotte and ni direct passive

sentences, (40a) and (40b), are grammatical as given below:

85

(40)
a. intaanetto-wa sekai-zyuu no hito-bito-ni yotte tukaw-are-te iru
Internet-TOP all over the world people-by use-Pass-Perf-Pres
‘Internet has been used by people all over the world’
b. intaanetto-wa sekai-zyuu no hito-bito-ni tukaw-are-te iru
.by
‘Internet is under the state affected by having been used by people all over the world’

Similarly, the to form may carry either a simple past or a perfective reading. In
the case of a simple past reading which expresses an event in the past in totality, the ni

yotte passive sentence with an inanimate subject (41a) is grammatical while the ni direct

passive (41b) is not:

(4 1)
a. ‘Hamlet’-wa Shakespeare-mi yotte kak-are-ta
‘Harnlet’-TOP Shakespeare-by write-Pass-Pst
b. *‘Harnlet’-wa Shakespeare-ni kak-are-ta
.by

‘Hamlet was written by Shakespeare’

By contrast, with the perfective reading both ni yotte and ni direct passive sentences (42a)

and (42b) are grammatical as shown below:

(42)

a. kimitu-syorui-ga tekikoku no supai-ni yotte nusum-are-ta
secrest documents-NOM spy from an enemy country-by steal-Pass-Perf
‘secret documents have been stolen by a spy from an enemy country’

b. kimitu-syorui-ga tekikoku no supai-ni nusum-are-ta

.by .
‘secret documents are under the state affected by having been stolen by a spy from an
enemy country’

3.2.5.2.3. Unavailability of an adversative reading of the ni yotte passive

86

In contrast to the ni indirect and direct passives, the ni yotte passive cannot carry
an adversative reading due to its non-G-subj ect status. Therefore, ni yotte simply cannot
replace ni either in the ni indirect or the ni direct passives connoting an adversative '

reading as shown below:

(43)
a. (= 373)
J ohn-ga warui toki-ni tomodati-m ko-rare-ta
-NOM at an inconvenient time fiiend-by come-Pass-Pst
b. *John-ga warui told-ni tomodai-ni yott ko-rare-ta
-by
‘John was adversely affected by his friend visiting him at an inconvenient time’
c. (= 38a)
Jane-ga mukasi no kare-ni ni-zikan mo mat-arete komat-ta
-NOM ex-boyfriend—by as many as 2 hours wait-Pass annoy-Pst
d. *Jane-ga mukasi no kare-ni yotte ni-zikan mo mat-arete komat-ta
.by

‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

Ni passive sentences, both indirect and direct, with an adversative reading such as (43 a)
and (430) respectively, are grammatical while their ni yotte passive counterparts (43b)

and (43d) are ungrammatical.

3.2.5.2.4. Verb-induced viewpoint differences

The last ni versus m' yotte contrast involves viewpoint differences effected
primarily by the choice of the verb used in a passive sentence. There are two cases where
different verb choices bring about a difference in viewpoint from which to describe a
situation: 1) Japanese native versus Sino-Japanese verbs; and 2) miru ‘see’-type verbs

(referred to as affective verbs).

87

First, Japanese native verbs together with the contextualization adding up a
personal touch are consonant with the ni direct passive to express the passive subject’s
personal and psychological involvement in the event described by the passive sentence.
On the other hand, Sino-Japanese verbs along with the contextualization signaling an
impersonal touch are consistent with the ni yotte passive to indicate a neutral and

objective perspective. Thus:

(44) a. Bill-wa kawaii musuko-ni koros-are—ta
-TOP beloved son-by kill-Pass-Pst
‘Bill was affected by being killed by his own beloved 3011’
b. *Bill-wa kawaii musuko-ni yotte koros-are-ta
.by
‘Bill was killed by his own beloved son’
c. Bill-wa CIA-ni yotte satugai-s-are-ta
-TOP the CIA-by murder-Pass-Pst

‘Bill was murdered by the CIA’

The farme tragedy of ‘Bill being killed by his own beloved son’ would naturally be
described by a ni direct passive such as (44a) above whereas it sounds much less natural
in the ni yotte passive sentence (44b). But by replacing kawaii musuko ‘his own beloved
son’ with CIA (an impersonal agent) and korosu ‘kill’ (a Japanese native verb) with
satugai-suru ‘murder’ (a Sino-Japanese verb) to increase the objectivity of the
expression, the grammaticality of the ni yotte passive sentence (44c) results.

Second, besides the general verb-induced difference illustrated above, there are
some verbs (viz. affective verbs) which are compatible with the ni direct passive but not

with the ni yotte passive:

88

(45)

a. Mary to hanasi-te iru tokoro-o, Bill-ga gaaruhurendo-ni kik-are-ta
talking to Mary as -NOM his girlfriend-by hear-Pass-Pst
‘Bill was affected by being heard by his girlfriend as he was talking to Mary’

b. *Mary to hanasi—te iru tokoro-o, Bill-ga gaaruhurendo-ni yotte kik—are-ta
.by

c. gaaruhurendo-ni, Bill-ga Mary to hanasi-te iru tokoro—ga kikoe-ta

his girlfiiend-to -NOM talking to Mary as audible-Pst

‘Bill was audible to his girlfriend as he was talking to Mary’

The ni direct passive sentence (45a) above is grammatical while the ni yotte passive
counterpart (45b) with the passivized verb kikareta ‘was heard’ is ungrammatical. This is
so not only because the passivized verb kikareta ‘was heard’ in itself expresses a
psychologically affected state of the passive subject rendering the ni yotte passive neutral
stance in conflict with such a reading; but also because there exists an unaccusative verb
which carries an objective connotation and thus serves as an impersonal version of
kikareru ‘be heard’, i.e., kikoeru ‘be audible’ as shown in (45c) above.

Table 3 below summarizes the passive test sentence types under investigation.
The left-hand column presents a list of sentence types organized in terms of syntactic and
semantic properties of passives manipulated, whereas the right-hand column provides a

summary of predicted grammaticality status of each sentence type according to passive

types.

89

Table 3: Passive sentence types syntactically or semantically manipulated
with their grammaticality status

Sentence T ypesa

SYNTAX

Ni yotte

Passive T ypesb

Ni direct

Ni indirect

 

-Proper syntactic operations

 

-- Case absorption w/+ O-role
suppression

0- O-role suppression w/+ case
absorption

 

-- 0-role suppression

 

-+ Passivization
0- Passivization

SEMANTICS
oAdversative reading
Non-adversative reading

 

-Progressive reading w/iru-marking
~Perfective reading w/iru-marking

 

-Past reading w/ta—marking
~Perfective reading w/ta-marking

 

-Adversative reading w/ni yotte
marking

 

-Personal viewpoint w/Japanese
native verb

°Objective viewpoint w/Sino-
Japanese verb

 

-Personal viewpoint w/affective verb

 

‘1

*

Ni yotte

444.4

\I

*

V

Ni direct

*4i4-fl- *4

4.

‘j

‘1

Ni indirect
\I

*

a. The markings of + and — in the syntax colurrm represent the presence or absence of

a relevant syntactic operation.

b' The \l- and ’-markings in the passive types column stand for predicted
grammatical and ungrammatical status, respectively. (Empty cells indicate that there

are no test sentences pertinent to a given passive type.)

3.2.5.3. Distractors

Distractors included the following: sentence structures with various kinds of case

marking, adverbial o, te iru perfective, ta perfective, and causatives. The adverbial o,

phonetically identical with the accusative case marker 0, indicates a temporal or spatial

90

place where an event or activity denoted by the verb of a sentence takes place. For

instance:

(46) Mary-wa hitori de sizuka na kooen-o/*de sanpo-sita
-TOP by herself quiet park-in take a walk-Pst

‘Mary took a walk by herself in a quiet park’
In sentence (46) above, the place where Mary took a walk, kooen ‘parrk’, must be marked
by the adverbial o in place of the postposition de which typically marks the place of an
activity. The distractor sentence with adverbial 0 was used to sort out NNS subjects
according to their correct and incorrect responses since a proper understanding of this
adverbial phrase presupposed a grammaticality judgment of test sentences involving one

type of verb-induced viewpoint differences.16

3.3. Procedures

All participants in the present study, both NSS and NNSs, were studying at
various universities in Japan as well as the US and were recruited by instructors whose
courses they were taking at the time of data collection. Some NNS participants were
contacted through their academic department colleague in a university. All participation
was on a voluntary basis and there was no compensation provided for their
participation.17 For the NNS subjects, however, it was hoped that by answering the
grammaticality judgment questionnaire, extensive exposure to Japanese passives in
various forms would enable them to learn something about these structures.

The NS participants received an envelope containing a general instruction sheet, a

consent form, a background information questionnaire, and a grammaticality judgment

91

test in conjunction with its instruction sheet. The general instruction sheet requested that
they carry out each task in the order given above. The grammaticality judgment
questionnaire was produced on a set of index card-sized paper. Each of the eighty-four
sheets contained one test sentence with the five choices underneath it: ‘Acceptable’,
‘Somewhat Acceptable’, ‘Somewhat Unacceptable’, ‘Unacceptable’, and ‘Not Sure’.
They were instructed to decide what they thought about the sentence, circle the relevant
choice on the sheet of paper, and place it back in the envelope. An example sheet with a
correct response to the item (irrelevant for the present study) was included as the first
sheet of the questionnaire set. Additionally, they were told not to refer back to previously
answered sheets. They were then informed that there was no time limit and that they
could spend as much time on the test as needed, performing it at their leisure. It was
expected to take the NS participants 20 to 30 minutes to go through the entire procedure.
Similarly, the NNS participants received an envelope containing a general
instruction sheet, a consent form, a background information questionnaire, a set of
proficiency and vocabulary tests, and a grammaticality judgment test in conjunction with
its instruction sheet. The general instruction sheet stated that they carry out each task in
the order given above. Both the pre-test (proficiency and vocabulary) and the
grammaticality judgment instruction sheet emphasized the importance of taking the tests
by themselves and that it was totally unnecessary to consult any other material such as a
dictionary, a grammar reference book, a Japanese NS friend, etc. in completing the
procedures. The general procedures for the grammaticality judgment test were the same
as the NS version above. It was expected to require an hour to an hour and a half for the

NNS participants to complete the entire procedure.

92

It may have been ideal that a proctor should administer the entire session of data
collection. In part this was a trade-off for obtaining a much larger amount of data. As
mentioned elsewhere, however, I took careful steps to reduce the possible introduction of
unnecessary variables in the absence of a proctor. Moreover, it is not fully clear that
supervised questionnaire procedures are best to obtain subjects’ metalinguistic judgment
data. For instance, Carroll, Bever, & Pollack (1981) Showed that the metalinguistic
performances of two groups of NSS differed depending on the test-taking conditions
arranged for each group. The subjects of one group were tested with a mirror placed in
fiont of them, while there was no mirror facing the subjects of the other group. They
attributed the judgment differences evidenced between the two groups to different types
of self-awareness induced by the mirror’s presence or absence. In the similar vein, it is
highly conceivable that the proctor’s presence in and of itself introduces certain
unnecessary variables for NNS subjects’ metalinguistic performances (see also Birdsong,

1989).

3.4. Data coding

The pre-test consisted of a version of the Japanese Language Proficiency Test and
a vocabulary test. The JLPT was comprised of 25 question items each of which carried
one point. Every NNS subject was assigned his or her J LPT score, equivalent to the total
number of questions he or she answered correctly. Since the test was in a multiple-choice
format, the possible scores for each item were either one or zero point. The possible

JLPT score range was from zero (0%) to 25 points (100%). The vocabulary test was also

93

in a multiple-choice format containing six questions. Those NNS subjects who did not
respond correctly were excluded from analyses of test sentences using such lexical items.

The grammaticality judgment test included 62 test sentences pertaining to
Japanese passives together with 22 distractor sentences. There were five choices for each
test sentence. When a subject chose the ‘acceptable’ option, he or she was assigned +2
points for that particular test sentence irrespective of its congruence to the theory
prediction. Likewise, -2 points were given to the choice of ‘unacceptable’, whereas the
‘Somewhat acceptable’, ‘not sure’, and ‘somewhat unacceptable’ options counted +1, 0,
and -1 point, respectively. When a subject did not respond to a test sentence, he or she
was not included in the analysis of that particular item. Subsequently, all subjects’
responses to each test sentence were averaged according to groups in order to make them
subject to statistical analyses. The range of each mean score was between -2 and +2
points. As in the vocabulary test, some distractor items were used to exclude subjects
with incorrect responses from analyses of those test sentences involving linguistic

devices used in them.

94

CHAPTER 4

RESULTS

4.1. Introduction
This chapter presents the results of analyses of the grarmnaticality judgment data
collected from NS and NNS subjects. First, the NSs’ data are examined; second, the

NNSS’ data are analyzed.1

4.2. Judgments by native speakers of Japanese

In this section we first examine the grammaticality judgments of passive test
sentences, and then turn to distractor sentences. Since Kuroda's (1979) analysis of the
semantic properties of passives hinges on the recognition of highly subtle differences in
connotation of passive sentences, it is imperative first to establish, prior to proceeding
with the investigation of NNSs’ judgments, that such semantic as well as syntactic
properties of passives are indeed reflected in ordinary NSS’ grammaticality judgments.
In investigating grammaticality judgments of such delicate nuances in passive sentences,
it might also be fruitful to take into consideration the possibility of the presence or
absence of NS grammaticality judgment differences between linguists and non-linguists.
The findings in the literature are inconclusive as to whether linguist’s judgments better
reflect a knowledge of language than non-linguist’s. That is to say, does linguistic
training make a linguist capable of recognizing grammatical nuances which may go
unnoticed by a non-linguist, or does a linguist develop an idiosyncratic sense of

grammaticality only remotely resembling that held by an ordinary NS?

95

We examine the grammaticality judgments given by ordinary, viz. linguistically
na'r've, NSS, followed by those given by linguistically sophisticated NSS in the following
two respects: 1) to what extent these two groups’ judgments conformed to the predictions
of the analyses adopted in the present study; and 2) how consistent their judgments were
as a group in assessing the test sentences of each type in question. At the end of the

section, some problematic sentences are considered.

4.2.1. Conformity to the theory predictions

We first look at the question of judgment conformity to the theory predictions
taking into consideration judgment disagreements between the two NS groups. In so
doing we need to identify judgment differences (defined below) between them as well as
from the predictions. To qualify as a judgment difference the following two conditions
must be met: 1) the judgment of a given test sentence was in disagreement with the
prediction2 and 2) there was a significant mean judgment score difference in rating it
between the two groups.

The first condition serves to glean judgments incongruent, both considerably and
marginally, to the predictions. The second aims to isolate those judgments in marginal
disagreement (defined below) with the predictions from significant and thus actual
judgment differences. The marginal disagreement includes cases where one group was in
disagreement with the theory prediction about the grammaticality status of a given test
sentence while the other group agreed with it, but where there was no statistically
significant difference between the two groups’ mean judgment scores. Consequently,

this particular judgment was disqualified as a significant judgment difference.

96

For instance, in a hypothetical example, the ungrammatical sentence A was rated
on average as -1.5 and +1.5 with a significant difference in the mean score. This counts
as one instance of a judgment difference. Namely, one group agreed with the theory
prediction while the other disagreed. On the other hand, the grammatical sentence B was
rated on average as +1.0 and -.01 with no significant difference. This is regarded as a
marginal judgment difference and thus does not count. That is to say, one group agreed
with the prediction, but the other did not disagree.

When the judgment difference defined in this way occasioned for a given test
sentence, we believe that it is reasonably safe to conclude that the particular judgment
one group gave was inconsistent with the theory prediction as well as the other group’s
judgment.3

Table 4 below presents the judgment difference occurrences identified for the
linguistically na'r've NS group in rating the passive test sentences. The left-hand column
of the table summarizes the sentence types grouped together in syntactic and semantic
terms as well as their predicted grammaticality status (grammatical or ungrammatical).
The right-hand column displays judgment disagreements this group made with theory
predictions, summarized according to the passive types (ni yotte, ni direct, and ni indirect
passives). Each x in the right-hand column indicates a single occurrence of judgment
disagreement in the given area into which it falls. The figure expressed as a denominator
in the same column indicates the number of test sentences used for a given area. For
instance, the cell in the upper left-hand corner below ‘ni yotte’ in the passive types
column stands for the area of judgments of ‘syntactically’ ‘grammatical’ ‘ni yotte’

passive sentences. (There are two passive sentences for this area, indicated by /2.) No x

97

in this cell shows full agreement with the predictions (i.e., all sentences in this area were
judged consistently with the predictions). Similarly, the cell in the lower left-hand comer
under ‘ni yotte’ represents the area of judgments of ‘semantically’ ‘ungrammatical’ ‘ni
yotte’ passive sentences. (There are nine passive sentences in this area.) Three x’s in this
cell show that the linguistically naive NS group disagreed with the predictions about

three sentences out of nine in this area.

Table 4: Grammaticality judgment disagreements of linguistically naive

 

 

 

NSS
Sentence Types Passive Types

SYNTAX Ni yotte Ni direct Ni indirect
Grammatical /2 /2 /2
Ungrammatical /4 /2 /4

SEMANTICS
Grammatical /10 x /9 /4
Ungrammatical x x x /9 /5 /4

 

 

When we look at linguistically naive NSS’ judgments of passive sentences, we find that
their judgments were rarely in disagreement with the theory predictions. This group only
rated four sentences (viz. a total of four x’s in the table) differently from the predictions
out of 57 test sentences (a 7.02% disagreement rate): one sentence out of 18 for the ni
direct passive (a 5.56% disagreement rate) and for the ni yotte passive three out of 25
sentences (12.0%).4 These four sentences were all in the semantics area which contains a
total of 41 sentences (a 9.76% disagreement rate), and one of them was predicted to be
grammatical out of a total of 29 grammatical sentences (a 3.45% disagreement rate),

whereas the other three were ungrammatical out of a total of 28 ungrammatical sentences

98

(10.7%). (See Appendix K for a table with a breakdown of syntactic and semantic
sentence types for linguistically naive NSS’ judgment disagreements.)
Similarly, Table 5 below displays a summary of the judgment difference

occurrences identified for the linguistically sophisticated NS group.

Table 5: Grammaticality judgment disagreements of linguistically

 

 

 

sophisticated NSS
Sentence Types Passive Types

SYNTAX Ni yotte Ni direct Ni indirect

Grammatical /2 x /2 x /2

Ungrammatical x x /4 x x /2 x /4
SEMANTICS

Grammatical x x x / 10 x x x /9 /4

Ungrammatical x x x x x /9 x x x /5 x x x /4

 

 

In contrast to the results in Table 4, Table 5 shows that their disagreements with
theory predictions are found over most types of passive sentences. In terms of overall
sentence tokens, this group judged 24 sentences (viz. a total of 24 x’s in the table)
differently fi'om the predictions in a total of 5 7 passive sentences (3 42.11% disagreement
rate). In syntactic and semantic terms, there were seven syntactic sentences out of 16 (a
43.75% disagreement rate) and 17 semantic sentences out of 41 (41.46%) on which this
group disagreed with the predictions. Furthermore, in light of grammaticality status, this
group disagreed with the theory predictions on eight out of 29 grammatical sentences (a
27.59% disagreement rate) and on 16 out of 28 ungrammatical sentences (57.14%).
Finally, in examining their judgments according to passive types, there were 10

disagreements out of 25 sentences (a 40% disagreement rate) for the ni yotte passive; nine

99

judgment difference occurrences were found out of 18 (50%) for the ni direct passive;
and there were five disagreements out of 14 (35.71%) for the ni indirect passive. Hence,
linguistically sophisticated NSs’ judgment disagreements with the theory predictions
occasioned approximately 30% to 55% of the time depending on the measures of
differences. (See Appendix L for a table with a breakdown of syntactic and semantic
sentence types for linguistically sophisticated NSs’ judgment disagreements.)

Therefore, it has been found that the linguistically naive NSS gave grammaticality
judgments of the passive test sentences conforming approximately 90% or more to the
theory predictions; that is to say, it can safely be concluded that their passive sentence
assessments indeed reflected Hoshi (1994a; 1999) and Kuroda's (1979) characterizations
of Japanese passives. On the other hand, the linguistically sophisticated NSS agreed with

the theory predictions approximately 45% to 70% of the time.

4.2.2. Group judgment consistency

Second, we investigate the question of group judgment consistency; namely, how
consistently did linguistically naive and s0phisticated NSS rate two tokens of each passive
test sentence type? In doing so, we need to identify their inconsistent responses to two
test sentences of each type. Specifically, two responses are classified to be inconsistent
when the following two conditions are met: 1) subjects (as a group) judged one of the two
as grammatical and the other as ungrammatical; and 2) the two judgment mean scores
were significantly different fi'om each other. The first condition excludes cases of
consistent judgments. The second condition is necessary to ensure that the judgment

inconsistency observed in the first condition was large enough in mean score terms to

100

indicate their non-marginal contradictory judgments. Since there are 27 (syntactic or
semantic) passive sentence pairs in question, there are a total of 27 cases where such
inconsistent judgments may have occurred.

Tables 6 and 7 below summarize the two NS groups’ inconsistent judgments in
rating pairs of passive test sentences. As in Tables 4 and 5 above, the left-hand column
of the tables summarizes the sentence types grouped together in syntactic and semantic
terms as well as their respective predicted grammaticality status (grammatical and
ungrammatical). The right-hand column displays judgment inconsistency occurrences
summarized according to the passive types (ni yotte, ni direct, and ni indirect passives).
Each # under the passive types columns indicates a single occurrence of inconsistent
judgments of a sentence pair. The figure expressed as a denominator in the same
columns shows the number of sentence pairs formed in a given area. (There are two
tokens for each sentence type.) For example, a single # in the third row under ‘ni direct’
in Table 6 shows that there was a single judgment inconsistency occurrence in the area of
‘Semantically’ ‘grammatica ’ ‘ni direct’ passive sentence pairs; namely, the linguistically
naive NS group judged one of the four sentence pairs in this area inconsistently. Empty

cells in these tables indicate that sentence pairs in a given area were judged consistently.

101

Table 6: Grammaticality judgment inconsistency of linguistically naive

 

 

 

NS group
Sentence Types Passive Types

SYNTAX Ni yotte Ni direct Ni indirect
Grammatical / 1 /1 /1
Ungrammatical /2 / 1 / 2

SEMANTICS
Grammatical /5 # /4 /2
Ungrammatical # /4 /2 /2

 

 

Table 7: Grammaticality judgment inconsistency of linguistically

 

 

 

sophisticated NS group
Sentence Types Passive Types

SYNTAX Ni yotte Ni direct Ni indirect

Grammatical / l # /1 /1

Ungrammatical /2 /1 # /2
SEMANTICS

Grammatical # # /5 # /4 /2

Ungrammatical # /4 # /2 # /2

 

 

It is apparent from Tables 6 and 7 that the two groups differed in terms of how
inconsistent their responses were in rating two test sentences of each type. There'were
only two cases (a total of two #’s in Table 6) out of 27 (a 7.41% inconsistent rate) where
the linguistically naive group responded inconsistently. These cases were both in
semantic areas, and were equally distributed over grammaticality status (one grammatical
and one ungrammatical). One was for the at yotte passive and the other for the ni direct
passive. In contrast, there were eight cases (a total of eight #’s in Table 7) out of 27 (a

29.63% inconsistent rate) where the linguistically sophisticated group made inconsistent

102

judgments. They occasioned widely across syntactic and semantic areas (two syntactic
and six semantic), grammaticality status (half grammatical and half ungrammatical), and
types of passives (three for the ni yotte passive, three for the ni direct passive, and two for
the ni indirect passive). (See Appendices M and N for tables with a breakdown of
syntactic and semantic sentence types for linguistically native and sophisticated NS
groups’ judgment inconsistencies.)

Thus, in light of group judgment consistency, the linguistically naive group has
been found more consistent in judging the test sentences of passives than the

linguistically sophisticated one.

4.2.3. Judgments of distractor sentences

Next, we look at the two groups’ judgments of the distractor sentences and
unaccusativess. Since the distractor items generally deal with syntactically (and
semantically) basic structures in comparison to the passive ones, their judgments of these
items are expected to be similar to or better than those found for the passive test
sentences. As examined above for the passive test sentences, we inspect how differently
one group rated them {Tom the theory predictions, taking into consideration judgment
disagreements with the other group. The linguistically naive group only rated two test
sentences out of 24 (8.33%) contradictory to the predictions with a significant difference
from the linguistically sophisticated group. On the other hand, the latter group judged 10
out of 24 test sentences (3 41.67% disagreement rate) in disagreement with the
predictions with a significant difference from the former group. Hence, in judgment of

the distractor sentences, both the linguistically naive and the sophisticated groups

103

demonstrated prediction-agreement rates comparable to those found for the passive
sentences. This in turn shows that they behaved properly on the task of judging the
passive test sentences.

Although the linguistically naive and sophisticated NSS of the present study
evinced rather distinct judgments about the test sentences, there was one aspect of
judgment patterns both groups had in common; namely, their judgments tended to be
expressed in a self-assured manner. The absolute judgments, i.e., the options of
‘grammatical’ and ‘ungrammatical’, accounted for a large majority of their judgments,
e.g., 76.78% (62.19 cases out of 81 on average) for the former group and 75.54% (61.19
cases out of 81 on average) for the latter group. Moreover, both groups rarely opted for
the ‘unsure’ option: .80% (.65 case in 81 on average) for the naive group and 1.11% (.90
case out of 81 on average) for the sophisticated group (see Bley-Vroman et al., 1988 for
similar findings on NSs’ judgment characteristics). These findings imply that the two
groups were not heterogeneous in judgment characteristics and provide some evidence
against the possible concern that the linguistically sophisticated group of the present
study exhibited eccentric judgment patterns and should not be considered representative

of a group of NSS with linguistic training.

4.2.4. Summary of native speakers’ grammaticality judgments

On the whole, it has been convincingly demonstrated that ordinary (defined as
linguistically naive) NSS judged the passive test sentences which were either semantically
or syntactically manipulated according to Hoshi (1994a; 1999) and Kuroda's (1979)

theory predictions. Furthermore, in all three terms examined above, i.e., judgment

104

congruence to the theory predictions on passive as well as distractor test sentences, and
group judgment consistency, the linguistically naive group demonstrated better

performances than the linguistically sophisticated group.

4.2.5. Consideration of some problematic test sentences
A set of three test sentences requires special consideration before proceeding. It

concerns ni direct and ni yotte direct passive sentences with an adversative reading, given

below:
(47)

a. zannen na koto ni, Tokyo daigaku-wa Tanaka-kyoozyu-ni yammer-are-ta
unfortunately Tokyo University -NOM Prof. Tanaka-by quit-Pass-Pst
‘unfortunately, Tokyo University was adversely affected by being quit by Prof. Tanaka’

b. *zannen na koto ni,Tokyo daigaku-wa Tanaka-kyoozyu-ni yotte yammer-are-ta

.by

c. *Saiwai na koto ni, Tokyo daigaku-wa Tanaka-kyoozyu-ni yammer-are-ta

fortunately Tokyo University -NOM Prof. Tanaka-by quit-Pass-Pst

‘fortunately, Tokyo University was adversely affected by being quit by Prof. Tanaka’

Despite Howard & Niyekawa-Howard's (1976) assertion of sentence (47a) above
(without the phrase zannen no koto ni ‘unfortunately’) as an example of the ni direct
passive with an adversative reading and Spees's (1992) empirical finding that NSS of
Japanese indeed tend to read an adversative connotation in the ni direct passive, it was
judged as ungrammatical both by the linguistically naive group (a mean judgment score
of -.84) and the linguistically sophisticated group (that of -1 .38). In subjects-proportion
terms, 74.19% of the former group (23 out of 31 subjects) and 85.71% of the latter group

(18 out of 21 subjects) rated it as ungrammatical.

105

It appears that the problem with this test item lay in the interpretation of the
sentence subject. An inspection of the notes some subjects left on questionnaire sheets
indicated that they conceived the sentence subject, Tokyo University, as an inanimate
entity, which renders the sentence as ungrammatical due to the incompatibility between
the insentient subject and an adversative reading. It was perhaps unclear to them how the
insentient institution, Tokyo University, was able to be adversely affected. In Howard &
Niyekawa-Howard's (1976) interpretation, on the other hand, it was considered to be an
animate entity, say, the university community, which was adversely affected by Prof.
Tanaka’s quitting; and thus the sentence was rendered as grammatical.

Since the other two sentences (47b) and (47c) had been contrived on the basis of
the grammaticality of sentence (47a), it would be extremely difficult to infer on what
basis subjects judged these two sentences when they apparently did not interpret sentence
(47a) in the intended sense. Both sentences (47b) and (47c) (expected to be rej ected)"
were rejected as ungrammatical by the linguistically naive group (mean judgment scores
of -1 .58 and -l .13) whereas the linguistically sophisticated group judged (47b) as
grammatical (score of 1.76) and (47c) as ungrammatical (score of -1 .43). But because it
is unclear if they rejected them on the intended basis, it was decided to exclude this set of

test sentences from the analyses."8

4.3. Judgments by English-speaking learners of Japanese
In this section we analyze the grammaticality judgments given by the NS control
group (referred to as Group 1) and three NNS groups, namely, the highly advanced group

(Group 2), the low advanced group (Group 3), and the intermediate group (Group 4).

106

First, judgments concerning syntactic properties of Japanese passives are examined,

followed by analysis of those on semantic properties.

4.3.1. Judgments of three syntactic types of passives

In the following we attempt to determine which type of passive was judged more
accurately than others by examining all groups’ judgments of grammatical as well as
ungrammatical sentences. Then, we investigate more specifically which violated

syntactic operation of passivization was difficult to judge as improper.

4.3.1.1. Grammatical judgments of three syntactic types of passives

We examine subjects’ (both NS3’ and NNS learners’) judgments of grammatical
sentences of three syntactic types of passives, i.e., the ni yotte, the ni direct, and the ni
indirect passives, in order to identify which type of passive was judged higher (thus more
accurately). We turn first to how each group judged each type of passive relative to the
others and second, compare judgments by learners’ groups to those of N85’ to find out
how close the former groups approximated the latter in judgment of each type of passive.

For the sake of exposition, example sentences of the three syntactic types of

passives are repeated below in the order of the ni yotte, the ni direct, and the ni indirect

passive:

(48)

a. utukusii uta-ga biitoruzu-ni yotte takusan tukur-are-ta
(= (34b)) beautifirl songs-NOM the Beatles-by a lot make-Pass-Pst

‘many beautiful songs were made by the Beatles’

107

b. Mary-wa J ohn-ni itumo soodan-s-are-ru
(= (35a)) -TOP -by always consult-Pass-Pres
‘Mary is affected by being always consulted by John’

c. J ohn-wa kodomo-ni kuruma no mado-o war-are-ta
(= (36a) -TOP child-by car window-ACC break-Pass-Pst
‘John was affected by a child breaking the car window’

Table 8 presents all groups’ mean judgment scores for each type of passive and
their respective standard deviations together with the number of subjects in each group.
Figure 2 displays the statistically estimated means of judgment scores of the four groups
for three types of passives. Each line represents the mean ratings of each group. The
figures on the y-axis signify the grammaticality rating scale (within the range of +2 to -2).
Numbers 1, 2, and 3 on the x-axis indicate the judgments of the ni yotte passive, the ni
direct passive, and the ni indirect passive, respectively. As shown in Figure 2, across
each of the learners’ groups, the ni direct passive was judged higher than the other two,

whereas ni yotte and ni indirect passives were judged similarly.

108

Table 8: Grammatical judgments of three syntactic

types of passives

 

 

 

 

 

 

Sentence Groups Mean SD N
T mes
ni yotte 1 1.21 .79 31
passivea 2 .80 1.30 25
3 .17 1.45 36
4 .15 1.26 20
ni direct l 1.85 .35 31
passiveb 2 1.66 .62 25
3 1.43 .79 36
4 .83 1.24 20
ni indirect l 1.85 .35 31
passivec 2 .98 1.31 25
3 .39 1.32 36
4 .25 1.19 20
a., b., c.

The test sentences for ni yotte, ni direct and ni
indirect passives include #47 and #50; #53 and #55; and
#57 and #60, respectively.9

 

 

 

 

 

 

2.0
fill-III-IIIIIIIIICI(J
.t' a
.‘ S
2 .° ,’ s
(U 1.5“ o.
0 ”9 ’I R \
E . ‘ I 0 ~ ‘ Groups 14
5 / \ I I I I
g) D NS control
'0
3 (G 1)
£5 - -
'5 a High adv.
f“ (G 2)
g _ I
Lowadv.
5 D
(G 3)
D lntenn.
o.o . (G 4)
1 2
l. ni yotte 2. ni direct 3. ni indirect

Figure 2: Three syntactic types of passives

109

A Repeated-Measures ANOVA was performed on the three syntactic types of passives as
a within-subjects variable and on the four groups as a between-subjects variable.
Significant main effects were found both in passive-types (F(2, 216) = 21.491, p < .0005)
and groups (F (3, 108) = 17.154, p < .0005); yet there was no significant interaction effect
between passive-types and groups (F(6, 216) = 1.819, p = .101).

For the purpose of identifying the precise nature of these main effects, first, a
Repeated-Measures ANOVA for each group on the three syntactic types of passives was
run. The results for Group 1 showed that there was a passive-type main effect (F(2, 60) =
15.579, p < .0005). Tests of within-subjects contrasts showed that both the ni direct and
the ni indirect passives were judged significantly higher than the ni yotte passive, but that
there was no significant difference in judgment between the ni direct and the ni indirect
passives.lo The ANOVA results for Group 2 indicated that there was a passive-type main
effect (F (2, 48) = 4.966, p = .011). Tests of within-subjects contrasts indicated that the hi
direct passive was judged significantly higher than both ni yotte and ni indirect passives
whereas there was no significant difference in judgment between ni yotte and ni indirect
passives. Group 3’s judgments exhibited identical patterns to Group 2’s. The results for
Group 3 showed that there was a passive-type main effect (F (2, 70) = 12.765, p < .0005).
Tests of within-subj ects contrasts demonstrated that the ni direct passive was rated
significantly higher than both ni yotte and ni indirect passives, but that there was no
significant judgment difference between at yotte and ni indirect passives. On the other
hand, the Group 4 AN OVA results showed that no significant main effect was found for
three syntactic types of passives (F(2, 38) = 2.023, p = .146), suggesting that Group 4

subjects did not judge any one type of passive higher than the other two.

110

In all, the ni direct passive was judged higher than both ni yotte and ni indirect
passives by all learners’ groups except for Group 4 (which did not make any differential
judgments of the three types of passives). On the other hand, ni yotte and ni indirect
passives were rated indiscriminately in grammaticality by learners’ groups.

Second, a Oneway AN OVA for each syntactic type was conducted on all four
subject groups to explore how closely Groups 2, 3, and 4 (the learners’ groups)
approximated Group l’s (the NS control group) judgments. There was a significant
difference between groups in judging the ni yotte passive (F (3, 108) = 5.154, p = .002).
The Tukey HSD results showed that Group 2 was not significantly different from Group
1, but that Groups 3 and 4 were significantly different from Group 1. A significant
difference was also found between groups in judging the ni direct passive (F (3, 108) =
7.814, p < .0005). The Tukey HSD results indicated that Groups 2 and 3 were not
significantly different from Group 1 whereas Group 4 judged significantly differently
from Group 1. Finally, there was a significant difference between groups in judging the
ni indirect passive (F (3, 108) = 12.560, p < .0005). The Tukey HSD results showed that
all learners’ groups were significantly different from Group 1 in judging this type of
passive.

Overall, the ni direct passive was judged by Groups 2 and 3 without significant
difference from Group 1 whereas only Group 2 rated the ni yotte passive without
significant difference from Group 1. Lastly, the ni indirect passive was judged by all
three learners’ groups significantly differently from Group 1.

In sum, in examining how each learners’ group judged the three syntactic types of

passives, the following two results were obtained: 1) the ni direct passive was judged

111

higher than the other two types of passives by the highly advanced and the low advanced
groups; 2) ni yotte and ni indirect passives were not judged differently by any group.
Next, the following three results were found by investigating whether learners’ groups’
judgments of each syntactic type of passive approximated those of the NS group: 1) in
judging the ni direct passive the highly advanced and the low advanced groups
approximated the NS group; 2) in rating the ni yotte passive only the highly advanced

group did so; 3) in judging the ni indirect passive no group successfully did so.

4.3.1.2. Ungrammatical judgments of three syntactic types of passives

We now address the question of which type of passive was judged more
accurately than the other(s) by examining subjects’ judgments of ungrammatical
sentences of three syntactic types of passives. Specifically, we investigate for which type
of passive each group made a grammaticality distinction between ungrammatical and
grammatical sentences.

First, we look at the ni direct passive. For exposition purposes, examples of its
grammatical and ungrammatical sentences are provided in (49) below (the ungrammatical

example of (49b) results from the lack of 0-role suppression):

(49) a. Mary-wa J ohn-ni itumo soodan-s-are-ru
(= (35)) -TOP -by always consult-Pass-Pres
‘Mary is affected by being always consulted by John’
b. *Mary-wa J ohn-ga itumo soodan-s-are-ru
-TOP -NOM

Table 9 gives all groups’ mean judgment scores for the grammatical and

ungrammatical sentences of the ni direct passive as well as their respective standard

112

deviations together with the number of the subjects of each group. Figure 3 displays
statistically estimated means of judgment scores of the four groups for the grammatical
and ungrammatical ni direct passive sentences. Each line represents the mean ratings of
each group. The figures on the y-axis signify the grammaticality rating scale (within the
range of +2 to -2). Numbers 1 and 2 on the x-axis indicate the judgments of the
grammatical and ungrammatical ni direct passive sentences, respectively. As
demonstrated in Figure 3, Groups 1, 2, and 3 correctly judged the grammatical and
ungrammatical sentences of the ni direct passive accordingly. It is also indicated that the
more advanced subject groups made a more pronounced distinction in judging them than
their less advanced counterparts. Group 4 rated grammatical and ungrammatical

sentences similarly (i.e., both as grammatical).

Table 9: Grammatical vs. ungrammatical judgments of the
ni direct passive

 

 

 

Sentence Types Groups M can SD N
Grammaticala 1 1.85 .35 31
2 1 .66 .62 25
3 1.43 .79 36
4 .83 1.24 20
Ungrammaticalb 1 '1-90 .37 31
2 -1.44 1.08 25
3 —.60 1.45 36
4 .33 1.55 20

 

 

., b. . .
a The test sentences for the grammatical and ungrammatical
ni direct passives include #53 and #55, and #54 and #56,
respectively.

113

3.0

 

 

 

 

1

2

 

2.0-
.. I.
C .0
(U E ‘.
a) K .
2 ‘ \9
.. 10. Groups 1-4
8 E \. I I I a
g) ’3’. . \ NS control
'0 0" ‘ \ :1 Cl
3 0.04 ‘c‘, ‘x. (G1)
a. q.‘ \ - —
"T .0 \ .
3 ".‘~ ‘13 a High adv.
m _ . (32
E 1.0 , _( )
‘3. I! D Lowadv.
0 .
-2.o ‘3 (G 3)
D lntenn.
-3.0 (G 4)

l. grammatical (GR) 2. ungrammatical (U G)

Figure 3: The ni direct passive (GR vs. UG)

A Repeated-Measures ANOVA was performed on the two types of sentences
(grammatical and ungrammatical) as a within-subjects variable and on groups as a
between-subj ects variable. A significant main effect was found in sentence-types (F(l ,
108) = 327.870, p < .0005). As expected from Figure 3, there was a significant
interaction effect between sentence types and groups (F(3, 108) = 27.178, p < .0005).
Thus, a Repeated-Measures AN OVA for each group on the two sentence types was run.
The results for Group 1 showed that there was a significant sentence-type main effect
(F (l, 30) = 1766.453, p < .0005), reflecting that Group 1 subjects made an extremely
clear distinction in judging the grammatical and ungrammatical sentences. The ANOVA

results of Group 2 indicated a significant sentence-type main effect (F (1 , 24) = 153.760,

114

p < .0005), suggesting that Group 2 subjects made highly distinctive judgments in
grammaticality between two types of sentences. The results for Group 3 showed a
significant sentence-type main effect (F(1, 35) = 55.726, p < .0005), indicating that
Group 3 subjects also made an unambiguous distinction in grammaticality between them.
Finally, in Group 4’s judgments, there was no significant sentence-type main effect (F (1 ,
19) = 1.712, p = .206), reflecting that Group 4 subjects made no grammaticality
distinction between the grammatical and ungrammatical sentences.

In all, Groups 2 and 3 made a clear—cut distinction in judging the grammatical and
ungrammatical sentences of the ni direct passive; on the other hand, Group 4 did not
make differential judgments between the two.

Second, we turn to the ni yotte passive. Examples of its grammatical and
ungrammatical sentences are given in (50) below (the first ungrammatical type (50b)
results from the lack of accusative case absorption, while the second (50c) from the lack

of 0-role suppression):

(50) a. utukusii uta-ga biitoruzu-ni yotte takusan tukur-are-ta
(= (34)) beautiful songs-NOM the Beatles-by a lot make-Pass-Pst
‘many beautiful songs were made by the Beatles’
b. I"biitoruzu—ni yotte utukusii uta-o takusan tukur-are-ta
-by ~ACC
c. *utukusii uta-wa biitoruzu-ga takusan tukur-are-ta
-TOP -NOM

Table 10, as before, provides the statistical data on all groups’ judgments of
grammatical and two types of ungrammatical sentences of the ni yotte passive.

‘Ungrammatical l and 2’ in the sentence types column represent the ungrammatical

115

sentence types given in (50b) and (50c) above, respectively. Figure 4 exhibits them
graphically. Numbers 1, 2, and 3 on the x-axis indicate the judgments of grammatical
and two types of ungrammatical ni yotte passive sentences, respectively. As shown in
Figure 4, Groups 1 and 2 appear to demonstrate an identical judgment pattern, but by
different degrees where they judged the grammatical and ungrammatical sentences
accordingly. In contrast, Groups 3 and 4 seem to pattern similarly in that they did not

make a grammaticality distinction between the grammatical and ungrammatical ones.

Table 10: Grammatical vs. ungrammatical judgments of the

 

 

 

 

ni yotte passive
Sentence T ”68 Groups Mean SD N
Grammaticala 1 1-21 .79 31
2 .80 1.30 25
3 .17 1.45 36
4 .15 1.26 20
b
Ungrammatical 1 l -1.34 .95 31
2 -.56 1.42 25
3 .24 1.53 36
4 .50 1.09 20
Ungrammatical 2c 1 '1-90 -33 31
2 -.96 1.14 25
3 .17 1.52 36
4 .20 .95 20

 

 

a., b., c. .
The test sentences for the grammatrcal and

tmgrammatical ni yotte passives include #47 and #50; and #48
and #51 (for Ungrammatical l) as well as #49 and #52 (for
Ungrammatical 2), respectively.

116

Grammaticality Judgment Means

 

~
~~~
~~

1.0-I ‘.

 

 

aGroups 1-4

0 NS control

(GI)

~ ~ it D High adv.

 

 

 

‘. (62)
h.. . — I
-15. NH..." D Lowadv.
..O.‘J (G 3)
-2.0'
CI lnterm.
-2.5 . (G4)
1 2 3

1. grammatical (GR) 2. ungrammatical (UG 1) 3. ungrammatical (UG 2)

Figure 4: The ni yotte passive (GR vs. UG)

A Repeated-Measures AN OVA for all groups was performed on the three types of

117

sentences: one grammatical and two ungrammatical. There was a significant sentence-
type main effect (F(2, 216) = 31.199, p < .0005) as well as a significant interaction effect
between sentence types and groups (F (6, 216) = 15.219, p < .0005). Thus, a Repeated-
Measures AN OVA on sentence types was performed on each group. The results for
Group 1 showed that there was a significant sentence-type main effect (F (2, 60) =
194.516, p < .0005). Tests of within-subjects contrasts demonstrated that the
grammatical sentences were judged significantly differently from both types of
ungrammatical sentences, reflecting that Group 1 subjects made a highly distinctive

differentiation between the grammatical and two types of ungrammatical sentences. The

ANOVA results of Group 2 indicated a significant sentence-type main effect (F (2, 48) =
15.297, p < .0005). Tests of within-subj ects contrasts revealed that the grammatical
sentences were judged differently fiom both types of ungrammatical sentences,
suggesting that Group 2 subjects also made a good distinction in grammaticality between
them. The results for Groups 3 and 4 showed that there was no sentence-type main effect
for either (F(2, 70) = .026, p = .958 for Group 3; F (2, 38) = .745, p = .482 for Group 4),
showing that Groups 3 and 4 subjects did not make any distinction in judging between
the grammatical and ungrammatical sentences.

Overall, Group 2 made a clear grammaticality distinction in judging the
grammatical and the two types of ungrammatical sentences (where either accusative case
absorption or 0-role suppression did not take place) of the ni yotte passive; on the other
hand, Groups 3 and 4 did not make differential judgments of them.

Third, we look at the ni indirect passive. Examples of its grammatical and
ungrammatical sentences are found in (51) below (the first ungrammatical type (51b)

results fi'om its lower clause being unpassivized, while the second one (52c) from

passivized):
(51) a. John-wa kodomo-ni kuruma no mado-o war-are-ta
(= (36)) -TOP child-by car window-ACC break-Pass-Pst
‘John was affected by a child breaking the car window’
b. *John-wa kodomo-ga kuruma no mado-o war-are-ta
-TOP child-NOM car window -ACC
c. *John-wa kuruma no mado-ga kodomo-ni war-are-ta

-TOP car window-NOM child-by

118

Table 11 gives the statistical data on all groups’ judgments of the grammatical
and two types of ungrammatical sentences of the ni indirect passive. ‘Ungrammatical l
and 2’ in the sentence types column represent the ungrammatical sentence types given in
(51b) and (51c) above, respectively. Figure 5 illustrates them graphically where numbers
1, 2, and 3 on the x-axis indicate the judgments of the grammatical sentences as well as
ungrammatical sentences 1 and 2, respectively. Figure 5 shows that in judgment of the
grammatical sentences and ungrammatical sentences 1, as with the ni yotte passive,
Groups 1 and 2 made a clear distinction in grammaticality between them. Group 3
showed a weak distinction whereas Group 4 did not. On the other hand, there seems to
be a considerable difference in judging ungrammatical sentences 2 between Group 1 and

the other groups.

Table 11: Grammatical vs. ungrammatical judgments of the
ni indirect passive

 

 

 

 

 

 

Sentence Types Groups Mean SD N
Grammaticala 1 1 .85 .35 3 l
2 .98 1.31 25
3 .39 1.32 36
4 .25 1.19 20
b
Ungrammatical 1 1 -l.92 .37 31
2 -l.60 .68 25
3 -.64 1.31 36
4 .05 1.19 20
C
Ungrammatical 2 1 -1 .69 .68 3 1
2 4.29 x 10'17 1-55 25
3 2.90 x 10'17 1-44 36
4 .25 .99 20
a., b., c.

The test sentences for the grammatical and ungrammatical
ni indirect passives include #57 and #60; and #58 and #61 (for
Ungrammatical 1) as well as #59 and #62 (for Ungrammatical 2),
respectively.

119

 

3.0

 

 

 

 

Groups 1-4

0 NS control

(G1)

High adv.
(G 2)

D

(G 3)

c1 lnterm.
(G 4)

2.0
In ET‘
5 .0
Q
2 a
*-' 1.0
5 \
E O
8’ .
:3 00‘ .
a . x ’.
3‘ \‘P‘ I"-”
= \o \~ .i. ’
8 B" ”
a: I
(U -10 ’
E '1 ’
E I
E ’.}g’ lufi D Lowadv.
o .DIII‘I IIIII |...g.
-2.0-i
-3.0 ‘
1 2 3

l. grammatical (GR) 2. ungrammatical (UG l) 3. ungrammatical (UG 2)

Figure 5: The ni indirect passive (GR vs. UG)

A Repeated-Measures ANOVA for all groups was performed on the three types of

sentences: one grammatical and two ungrammatical. There was a significant sentence-

type main effect (F(2, 216) = 97.503, p < .0005) as well as a significant interaction effect

between sentence types and groups (F(6, 216) = 25.262, p < .0005). Thus, a Repeated-

Measures AN OVA for each group was performed on the three sentence types. The

results for Group 1 showed a significant sentence-type main effect (F(2, 60) = 568.076, p

< .0005). Tests of within-subjects contrasts demonstrated that the grammatical sentences

were judged significantly differently from both types of ungrammatical sentences,

indicating that Group 1 subjects made a clear-cut grammaticality distinction in judging

the grammatical and ungrammatical sentences. The results for Group 2 indicated a

120

significant sentence-type main effect (F(2, 48) = 30.983, p < .0005). Tests of within-
subjects contrasts showed that the grammatical sentences were judged differently from
both types of ungrammatical sentences; however, ungrammatical sentences 2 (a mean
judgment score of 4.29 x 10'”) were rated significantly higher than ungrammatical
sentences 1 (a mean judgment score of -1 .60). Thus, these findings suggest that Group 2
subjects made differential judgments in grammaticality between the grammatical
sentences and ungrammatical sentences 1, but that they failed to reject ungrammatical
sentences 2 as ungrammatical. The results for Group 3 displayed a significant sentence-
type main effect (F (2, 70) = 5.973, p = .004). Tests of within-subjects contrasts indicated
that ungrammatical sentences 1 were judged significantly differently fi'om the
grammatical sentences whereas ungrammatical sentences 2 were not rated differently
from the grammatical ones. This shows that Group 3 subjects made a distinction in
grammaticality between the grammatical sentences and ungrammatical sentences 1, but
not between the grammatical ones and ungrammatical sentences 2. Finally, in Group 4’s
judgments, there was no significant sentence-type main effect (F(2, 38) = .414, p = .664),
suggesting that Group 4 subjects did not make any distinctions in judging the three types
of sentences.

In short, Groups 2 and 3 made a grammaticality distinction in judging the
grammatical sentences and ungrammatical sentences 1 (where the lower clause was not
passivized) of the ni indirect passive. On the other hand, in judging ungrammatical
sentences 2 (where the lower clause was passivized), Groups 2 and 3 failed to do so.

Group 4 did not make any differential judgments between the three.

121

In summary, when the judgments of the ungrammatical sentences of the three
syntactic types of passives were considered with respect to those of grammatical
counterparts, the following three results were obtained: 1) in judging the ni direct passive
the highly advanced and the low advanced groups made grammatical versus
ungrammatical distinctions; 2) in rating the ni yotte passive only the highly advanced
group discriminated between the grammatical and two types of ungrammatical sentences;
3) in judging the ni indirect passive both the highly advanced and the low advanced
groups made a grammaticality differentiation between the grammatical and one of the
two ungrammatical sentence types, but for the other ungrammatical sentence type, they
did not make a grammaticality distinction. (See the following section for further

investigation of the nature of judgments of these ungrammatical sentences.)

4.3.1.3. Judgments of improper syntactic operations

The previous section compared the subjects’ judgments of the ungrammatical
sentences to the grammatical ones. In this section we investigate their judgments of the
ungrammatical sentences of the three syntactic types of passives from a different
perspective. Specifically, we investigate which syntactic operation of passivization was
difficult to judge as improper when it was violated. In doing so, the ungrammatical
sentences were classified into the following three cases depending on the kind(s) of
passivization processes involved: 1) the ni direct and indirect passive sentences where
external 0-role was not suppressed; 2) ni yotte passive sentences in which either external
0-role suppression or accusative case absorption did not take place; and 3) ni indirect

passive sentences in which passivization either took place or did not. There are two types

122

of ungrammatical sentences in each of the three cases. These two ungrammatical types in
each case are analyzed in terms of how one type of ungrammatical sentences was judged
with respect to the other, and how closely the NNS groups approximated the NS control
group.

First, with respect to the ungrammatical sentences of the ni direct and indirect
passives where their external 0-roles were not suppressed, examples of both types are
provided below, i.e., (52a) for the ni direct passive and (52b) for the ni indirect passive

(the English glosses indicate their intended meanings):

(52)
a. (= (3 5b)) II‘Mary-wa John-ga itumo soodan-s-are-ru
-TOP -NOM always consult-Pass-Pres
‘Mary is affected by being always consulted by John’
b. (= (36c)) *John-wa kodomo-ga kuruma no mado-o war—are-ta

-TOP child-NOM car window-ACC break-Pass-Pst
‘John was affected by a child breaking the car window’

Table 12 shows the statistical data of subjects’ judgments of these ungrammatical
sentences for all groups. (The asterisk attached to the passive type in the table stands for
its predicted ungrammaticality status.) As is clearly shown, the subjects’ mean judgment
scores and their proficiency levels are in an inverse relationship and parallel for both
types of passives. That is to say, these ungrammatical sentences were rated more

strongly ungrammatical as the subjects’ proficiency levels increased.

123

Table 12: The ni direct and indirect passives
without 0-role suppression

 

 

 

 

 

Sentence Groups M can SD N
Types
*ni direct 1 -l .90 .37 31
passivea 2 -1.44 1.08 25
3 -.60 l .45 36
4 .33 l .55 20
*ni indirect 1 -l .92 .37 31
passive!) 2 -1.60 .68 25
3 -.64 1.31 36
4 .05 1 .19 20
a b

" ' The test sentences for ni direct and indirect
passives include #54 and #56; and #58 and #61.

A Repeated-Measures ANOVA for all groups was performed on both types of
passives. The results indicated that there was no passive-type main effect (F (1 , 108) =
1.154, p = .285), nor was there an interaction effect between passive types and groups
(F(3, 108) = .248, p = .863). A significant group main effect was found (F (3, 108) =
26.561, p < .0005), however. The Tukey HSD results showed that Groups 1 and 2 were
indistinguishable from each other and that Groups 3 and 4 were significantly different
both from each other and from Groups 1 and 2.

In brief, these findings suggest that learners’ groups learned the ungrammaticality
of both the ni direct and the ni indirect passive sentences without 0-role suppression as
they became more proficient in TL (target language) and furthermore that the highly
advanced group successfully approximated the NS control group.

Second, with regard to the two ungrammatical sentence types of the ni yotte
passive, examples of both types are given in (53) below, i.e., (53a) in which 0-role

suppression took place but case absorption did not, and (53b) where 0-role suppression

124

did not occur, but case absorption did (the English gloss indicates their intended

meaning):

(53) a. ‘biitoruzu-ni yotte utukusii uta-o takusan tukur-are-ta

(= (34)) the Beatles-by beautiful songs-ACC a lot make-Pass-Pst
b. *utukusii uta-wa biitoruzu-ga takusan tukur-are-ta

-TOP -NOM
‘many beautiful songs were made by the Beatles’

Table 13 displays the statistical data on all groups’ judgments of these
ungrammatical sentences”. It appears that if Groups 3 and 4 are considered together,
group grarmnatical judgment scores and proficiency levels, i.e., Groups 1, 2, and 3/4, are
inversely related. Note also that there is a considerable judgment difference between the
sentences without 0-role suppression and those without case absorption for Group 1 and,

to a slightly lesser extent, for Group 2.

Table 13: The ni yotte passive without 0-role
suppression or case absorption

 

 

 

Sentence Groups Mean SD N
T )pes

*+ 0-role l -1.34 .95 31

._ Casea 2 -.56 1.42 25

3 .24 1.53 36

4 .50 1.09 20

*- 0-role l -1.90 .33 31

+ Case 2 -.96 1.14 25

3 .17 1.52 36

4 .20 .95 20

 

 

a., b. . .

The test sentences for the ungrammatical nr yotte
passive include #48 and #51 (without case absorption) as
well as #49 and #52 (without 0-role suppression).

125

A Repeated-Measures AN OVA for all groups was performed on both types of
sentences. The results indicated that there were significant main effects both in sentence-
types(F(1, 108) = 5.506, p = .021), and in groups (F(3, 108) = 26.721, p < .0005). There
was no interaction effect between sentence types and groups (F (3, 108) = .662, p = .577).
The Tukey HSD results showed the following: 1) Group 1 was significantly different
from all learners’ groups; 2) Group 2 was different from Groups 3 and 4; and 3) Groups 3
and 4 were indistinguishable fiom each other. Thus, Group 2 failed to approximate in
judgment to the NS control group, although it still rated better than Groups 3 and 4.

Next, Repeated-Measures AN OVA results for each group showed that there was a
significant sentence-type main effect for Group 1 (F (1, 30) = 10.634, p = .003); however,
no effect was found for all learners’ groups (F( 1, 24) = 2.286, p = .144 for Group 2; (F (1,
35) = .048, p = .828 for Group 3; and (F(1, 19) = .932, p = .347 for Group 4). These
findings all suggest that Group 2 indiscriminately judged both types of sentences as
ungrammatical whereas Groups 3 and 4 did not rate either of them as ungrammatical; on
the other hand, Group 1 subjects judged the sentences without 0-role suppression
significantly lower (thus as more ungrammatical) than those without case absorption.

In sum, in contrast to Groups 3 and 4’s failure, Group 2 subjects squarely rejected
both types of ungrammatical sentences of the ni yotte passive although they failed to
approximate Group 1. The learners’ groups rated them indiscriminately while Group 1
judged the sentences without 0-role suppression worse than those without case absorption.

Third, with respect to the two ungrammatical sentence types of the ni indirect

passive, examples of both types are repeated in (54) below, viz., (54a), unpassivized in its

126

embedded clause and (54b), passivized (the English gloss below shows their intended

meaning):

(54) a. *John-wa kodomo-ga kuruma no mado-o war-are-ta

(= (34)) -TOP child-NOM car window-ACC break-Pass-Pst
b. *John—wa kuruma no mado-ga kodomo-ni war-are-ta

-TOP car window-NOM child-by
‘John was affected by a child breaking the car window’

Table 14 presents the statistical data on all groups’ judgments of these
ungrammatical sentences". There is a marked difference in judgment between these two
types of sentences for Groups 2 and 3. The unpassivized sentences were rejected clearly
as ungrammatical while the passivized ones were not. Note also that Group 1 did not

make such a differential judgment and rated both of them as squarely ungrammatical.

Table 14: The ni indirect passive unpassivized and

 

 

 

passivized
Sentence Types Groups Mean SD N
*Unpassivizeda l -l .92 .37 3 l
2 -1.60 .68 25
3 -.64 1.31 36
4 .05 1.19 20
tpassivizedb 1 -1.69 .68 31
2 4.29 x 10’17 1-55 25
3 2.90 x 10'17 1-44 36
4 .25 .99 20

 

 

a., B' The test sentences for the ni indirect passive include #58
and #61 (for unpassivized) and #59 and #62 (for passivized).

A Repeated-Measures ANOVA for all groups was performed on both types of
sentences. The results showed that there was a significant interaction effect between

sentence types and groups (F (3, 108) = 26.721, p < .0005). Thus, for the purpose of

127

exploring the nature of this interaction, to begin with, a Repeated-Measures AN OVA on
these ungrammatical sentence types was run for each group. It was found that there was
no significant sentence-type main effect for Groups 1 and 4 (F ( 1, 30) = 2.506, p = .124
for Group 1; and F (1, 19) = .655, p = .428 for Group 4), suggesting that both groups did
not judge the two types of sentences differently. On the other hand, significant sentence-
type main effects were found for Groups 2 and 3 (F (1, 24) = 23.814, p < .0005 for Group
2; and F(1, 35) = 4.188, p = .048 for Group 3), indicating that both groups judged them
significantly differently from each other. Thus, the unpassivized and passivized
sentences were judged differently by Groups 2 and 3, but not by Groups 1 and 4.

Next, Oneway AN OVAs for both ungrammatical sentence types on all four
groups were conducted to examine how closely learners’ groups approximated the NS
control group. There was a significant difference between groups in judging the
unpassivized sentences (F (3, 108) = 21 .515, p < .0005). The Tukey HSD results showed
that Group 2 was not significantly different from Group 1, but that Groups 3 and 4 were
significantly different from Group 1. A significant difference was also found between
groups in judging the passivized sentences (F(3, 108) = 15.503, p < .0005). The Tukey
HSD results indicated that all learners’ groups were significantly different from Group 1
and that learners’ groups were not different from one another. Thus, Group 2 subjects
approximated Group 1 only in judging the unpassivized sentences, but they failed to do
so for the passivized ones. Groups 3 and 4 judged both unpassivized and passivised
ungrammatical sentences differently from Group 1.

In all, the unpassivized sentences were judged clearly as ungrammatical by Group

2 subjects, approximating Group 1; on the other hand, the passivized ones were not

128

rejected as ungrammatical. Group 3 subjects rated the unpassivized sentences as
ungrammatical to a lesser extent than Groups 1 and 2, whereas they failed to do so for the
passivized ones. Group 4 subjects did not reject either types of sentences as
ungrammatical.

On the whole, when the subjects’ judgments of ungrammatical sentences were
examined in light of the improper syntactic operations responsible for their
ungrammaticality, the following three results were obtained: 1) the lack of 0-role
suppression served as a progressively clearer indication of ungrammaticality for the ni
direct and the ni indirect passives as learner subjects became more proficient in the TL;
2) in judging the two types of ungrammatical sentences of the ni yotte passive, only the
highly advanced group subjects rated them as ungrammatical. However, they did so with
no differentiation between them unlike the NS group who judged the one without 0-role
suppression worse than the other without case absorption; 3) in rating the unpassivized
and passivized ungrammatical sentences of the ni indirect passive, the former were
judged as squarely ungrammatical by the highly advanced group (approximating the NS
group) and, to a lesser extent, by the low advanced group, whereas the latter sentences

were not rejected as ungrammatical by any NNS group.

4.3.1.4. Summary of judgments of three syntactic types of passives

A summary of the analyses of the subjects’ judgments of three syntactic types of
passives follows. First, of all three types of grammatical sentences, the ni direct passive
was judged highest and the other two types of passives, ni yotte and ni indirect passives,

were rated indistinguishably by the highly advanced and the low advanced groups.

129

Moreover, the highly advanced and the low advanced groups approximated the NS
control group in rating the ni direct passive. In contrast, only the highly advanced group
approximated the NS group in rating the hi yotte passive while no NNS group did so in
assessing the ni indirect passive.

Second, in comparison of ungrammatical sentences to grammatical ones, 1) the
ungrammatical sentences of the ni direct passive (where 0-role suppression did not take
place) were judged significantly differently from the grammatical ones by the highly
advanced and the low advanced groups; 2) two types of ungrammatical sentences of the
ni yotte passive (where either 0-role suppression or accusative case absorption did not
occur) were only differentiated from the grammatical ones by the highly advanced group;
3) one type of ungrammatical sentences of the ni indirect passive (the unpassivized
version) was discriminated fiom the grammatical ones by the highly advanced and the
low advanced groups whereas the other type (the passivized version) was not
distinguished in grammaticality from the grammatical ones by any learners’ group.

Third, in the analysis of the ungrammatical sentences in terms of improper
syntactic operations, 1) the ungrammatical sentences in which 0-role suppression did not
take place were judged as ungrammatical both in the ni direct and the ni indirect passives
all the more clearly as the subjects’ proficiency levels increased; 2) only the highly
advanced group rejected two ungrammatical types of the ni yotte passive: one where 0-
role suppression did not take place and the other in which accusative case absorption did
not occur; 3) the unpassivized sentences of the ni indirect passive were correctly rejected

by the highly advanced group and, to a lesser extent, by the low advanced group as well,

130

while the passivized counterparts were not rejected by any NNSs’ group. In all analyses

above, the intermediate group did not give any differential judgments.

4.3.2. Judgments of semantic properties of passives

In the following we attempt to investigate NNS subjects’ knowledge of semantic
properties of Japanese passives by examining 1) their grammaticality judgments of the
three types of passives for an adversative reading, viz., the ni yotte, the ni direct, and the
ni indirect passives; 2) those of adversative versus non-adversative readings of the ni
direct and the ni indirect passives; and 3) those of ni passive versus ni yotte passive

contrasts effected in various conditions.

4.3.2.1. Judgments of three types of passives for an adversative reading

As an initial investigation of NNS subjects’ knowledge of semantic properties of
Japanese passives, we obtain an overview of their judgment patterns of the three types of
passives in semantic terms just as we did for the syntactic investigation. Namely, we
examine how the NNS groups judged each type of passive with respect to the other two
as well as how closely they approximated the NS control group in rating them. In doing
so, their judgments of the three types of passives (i.e., the ni yotte, the ni direct, and the ni
indirect passives) for an adversative reading are investigated since this semantic property
is the only type which manifests its semantic effect across all these passives. For
expository purposes, an example sentence of each passive type is provided in (55) below.
Due to its non-0-subject status, the ni yotte passive (55a) is incompatible with an

adversative reading (which expresses a situation unfavorable for its passive subject) and

131

thus is ungrammatical, whereas the ni passives, both direct and indirect ((55b) and (55c)
respectively), are compatible with an adversative reading owing to their ‘affectee’ 0-

subject status:

(5 5)
a. (= (43d))
*Jane-ga mukasi no kare-ni yotte ni-zikan mo mat-arete komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst
b. (= (43 c))
J ane-ga mukasi no kare-ni ni-zikan mo mat-arete komat-ta
-by

‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

. c. (= (3 7))
J ohn-ga warui told-ni tomodati-ni ko-rare—ta
-NOM at an inconvenient time friend-by come-Pass-Pst

‘John was adversely affected by his friend visiting him at an inconvenient time’

Table 15 provides the statistical data on all groups’ judgments of these three types of
grammatical or ungrammatical sentences.13 In Figure 6 numbers 1, 2, and 3 on the x-axis
indicate the judgments of passives in the order given above for the example sentences.
(The asterisk attached to the passive type both in the table and the figure indicates its
predicted ungrammaticality status.) As displayed in Figure 6, Groups 1 and 2 seem to
have made differential judgments between the ni yotte passive and the ni passives: Group
1 appears to have made a clearer differentiation between the ni yotte passive and the ni

_ direct passive than Group 2, while both groups gave almost identical judgments to the ni
indirect passive. On the other hand, Group 3 seems to have judged the ni direct passive
differently from both the ni yotte and the ni indirect passives. Finally, Group 4 appears
not to have made any differential judgments between any two of the three types of

passives.

132

Table 15: Three types of passives for an adversative

 

 

 

 

 

 

reading
Sentence Groups Mean SD N
Types
*ni yotte l -.42 1.36 31
passivea 2 .24 1 .67 25
3 .67 1 .62 36
4 .65 1 .42 20
ni direct 1 1.84 .58 31
passiveb 2 1.52 1.08 25
3 1 .39 l .05 36
4 .65 1 .42 20
ni indirect 1 1.3629 .6796 31
passivec 2 1 .4000 .6250 25
3 .1597 1.1880 36 .
4 .4125 .9293 20
a.,T., c.

The test sentences include #26 for the ni yotte passive;

#25 for the ni direct passive; and for the ni indirect passive #13
and #16 with intransitive verbs as well as #19 and #22 with

 

 

 

 

 

 

transitive verbs.
2.0
0.. o ,

1.5- .‘ i-—-_:'__
2 ’0’ I}. ‘lfia
m . ’ \
§ /’ '
.3 1.0< / . f \, GROUP1_4
é ' / . I’ .0 \ I I I I
.3: [3’ ’ a .0. n \ . [:1 NS control
3: .5 . , ,0 \ (G 1)
g ’ .0 I \ - -
g If .‘ .m 0 High adv.
r __._..,
E .
9 0 Low adv.
9 In

”5 -I (G 3)

D lnterrn.
-1.0 . (G 4)
1 2 3
1. *ni yotte 2. ni direct 3. ni indirect

Figure 6: Three types of passives for an adversative reading

133

A Repeated-Measures AN OVA for all groups was performed on the three types of
passives. A significant main effect was found in passive-types (F (2, 216) = 21.362, p <
.0005), but there was a significant interaction effect between passive-types and groups
(F (6, 216) = 7.988, p < .0005). Thus a Repeated-Measures ANOVA on the three types of
passives was run for each group. The results for Group 1 showed that there was a
passive-type main effect (F (2, 60) = 49.011, p < .0005). Tests of within-subjects
contrasts showed that each of the three passives was judged differently from one another.
Namely, the ni yotte passive was rated significantly lower than both types of ni passives,
but there was a significant judgment difference between ni direct and ni indirect passives.
The AN OVA results for Group 2 indicated that there was a passive-type main effect (F (2,
48) = 8.982, p < .0005). Tests of within-subjects contrasts indicated that the ni yotte
passive was judged significantly lower than both types of ni passives whereas there was
no significant difference in judgment between ni direct and ni indirect passives. Next, the
results for Group 3 showed that there was a passive-type main effect (F (2, 70) = 7.953, p
= .001). Tests of within-subjects contrasts demonstrated that the ni direct passive was
rated significantly higher than both ni yotte and ni indirect passives, but that there was no
significant judgment difference between ni yotte and ni indirect passives. Finally, Group
4 ANOVA results showed that no Significant main effect was found for the three types of
passives (F(2, 38) = .222, p = .802), suggesting that Group 4 subjects did not judge any
one type of passive higher than the other two.

In all, the ni direct passive was judged indiscriminately from the ni indirect
passive by Group 2. On the other hand, the ni direct passive was rated differently from

both ni yotte and ni indirect passives by Group 3; furthermore, the ni yotte passive was

134

judged differently from the ni indirect passive by Group 2, but similarly by Group 3.
(Group 2’s judgments of this property of the ni yotte passive are examined more closely
in a later analysis.) Group 4 did not make any differential judgments of the three types of
passives.

Second, a Oneway AN OVA for each passive type was conducted on all four
subject groups to explore how closely the learners’ groups (viz. Groups 2, 3, and 4)
approximated the NS control group (Group 1) in judgments. There was a significant
difference between groups in judging the ni yotte passive (F (3, 108) = 3.321 , p = .023).
The Tukey HSD results showed that Groups 2 and 4 were not significantly different from
Group 1, but that Group 3 was significantly different from Group 1. Considering the
ungrammaticality status of the ni yotte passive and Group 4’s scant knowledge
(evidenced thus far) of Japanese passives, Group 4’s approximation to Group 1 seems
accidental in that the former group rated it low due to the absence of relevant knowledge,
whereas the latter group rated it negatively owing to the presence of such knowledge. A
significant difference was also found between groups in judging the ni direct passive
(F (3, 108) = 5.474, p = .002). The Tukey HSD results indicated that Groups 2 and 3
were not significantly different from Group 1 whereas Group 4 judged significantly
differently from Group 1. Finally, there was a significant difference between groups in
judging the ni indirect passive (F(3, 108) = 14.783, p < .0005). The Tukey HSD results
showed that Group 2 was not different from Group 1, but Groups 3 and 4 were

significantly different from Group 1 in judging this type of passive.

135

Overall, the ni direct passive was judged by Groups 2 and 3 without significant
difference fiom Group 1, but only Group 2 rated the ni yotte passive as well as the hi
indirect passive similarly to Group 1.

In sum, in examining how learners’ groups judged each type of passive with
respect to the other two, the following three results were obtained: 1) the highly advanced
group judged the two types of ni passives, i.e., direct and indirect, without significant
difference while this group rated the ni yotte passive significantly lower than both types
of ni passives; 2) the low advanced group judged the ni direct passive significantly higher
than ni yotte and ni indirect passives, whereas this group rated the latter two types of
passives indiscriminately; 3) the intermediate group did not judge any one type of passive
higher than the other two. Next, the following two results were found by investigating
whether learner groups’ judgments of each type of passive approximated those of the NS
group: 1) in judging the ni direct passive the highly advanced and the low advanced
groups approximated the NS group; 2) in rating ni yotte and ni indirect passives only the
highly advanced group did so.

Therefore, on the whole, the highly advanced group seems to have demonstrated
judgment patterns indistinguishable from the NS control group in rating the three types of
passives for an adversative reading, whereas the low advanced group showed good
knowledge of only the ni direct passive, and the intermediate group failed to demonstrate
any evidence of knowledge. (See the subsequent analyses for closer examinations of

learner groups’ judgments of adversative readings for the three types of passives.)

4.3.2.2. Judgments of adversative versus non-adversative readings

136

The ni indirect passive predominantly carries an adversative reading whereas the
ni direct passive does also, but to a lesser extent. Thus, the ni indirect and the ni direct
passives are both compatible with a context denoting an unfavorable situation for their
passive subject, and not in harmony with a context denoting a favorable one. We
examine subjects’ judgments of the at indirect and the ni direct passives with an
adversative reading vis-a-vis those of a non-adversative reading to find out if judgments
of them reflected the above noted context-compatibility contrast. We look first into
whether or not they made differential judgments between adversative and non-
adversative readings within each type of passive; second, we compare their judgments of
adversative readings across the types. We look at three types: i.e., the ni indirect passive
with intransitive verbs, the ni indirect passive with transitive verbs, and the ni direct

passive.

4.3.2.2.]. An adversative versus non-adversative reading distinction within each
type of passive

First, we examine subjects’ differential judgments of the ni indirect passive with
intransitive verbs in both unfavorable and favorable contexts for its passive subject.
Example sentences of each type are provided in (56) below, i.e., (56a) with an

unfavorable context (thus grammatical), and (56b) with a favorable context (thus

ungrammatical):
(5 6) a. J ohn-ga warui told-ni tomodati-ni ko-rare-ta
(= (37)) -NOM at an inconvenient time friend-by come-Pass-Pst

‘John was adversely affected by his friend visiting him at an inconvenient time’

137

b. *John-ga ii toki-ni tomodati-ni ko-rare-ta
at a convenient time

“John was adversely affected by his friend visiting him at a convenient time’
(Kuroda 1979, p. 314 & p.317)

Table 16 provides the statistical data on all groups’ judgments of these
grammatical and ungrammatical sentences. In Figure 7 numbers 1 and 2 on the x-axis
indicate the judgments of respective readings. As shown in Figure 7, Groups 1 and 2
seem to have made a distinction in judgment between adversative and non-adversative

readings while Groups 3 and 4 did not.

Table 16: The ni indirect passive with intransitive verbs
with adversative and non-adversative readings

 

 

 

Sentence Types Groups Mean SD N
intransitive 1 1 .23 .84 3 1
indirect with 2 1.46 .85 25
adversative 3 .0139 1.37 36
readinga 4 .58 1.38 20
‘intransitive 1 -.758 1.032 31
indirect with 2 -.0800 1.087 25
non-adversative 3 -.01389 1.296 36
readingb 4 .575 .977 20

 

 

a., b° The test sentences include #13 and #16 with adversative
reading; and #15 and #18 with non-adversative reading.

138

Grammaticality Judgment Means

1. iii indirect with adversity

 

2.0 ’

(“'1

J

 

0.0Eh-—._.—I—I_I.*

 

‘ri

 

 

2

Groups 14

NS control

(61)

High adv.
(G 2)

D

U

B Low adv.
(G 3)

 

D lnterm.
(G 4)

2. *ni indirect without adversity

Figure 7: The ni indirect passive with intr. verbs with +/- adversative readings

A Repeated-Measures ANOVA for all groups was run on the passive types. The

results showed that there was a significant interaction effect between passive types and

groups (F(3, 108) = 15.569, p < .0005). Thus a Repeated-Measures ANOVA was

performed for each group. The results for Groups 1 and 2 indicated that there was a

significant passive-type main effect (F (l, 30) = 105.355, p < .0005 for Group 1; F (1, 24)

= 27.125, p < .0005 for Group 2), suggesting that both groups judged the ni indirect

passive with an adversative reading differently from those with a non-adversative

reading. On the other hand, Groups 3 and 4 AN OVA results showed no significant

judgment differences (F(1, 35) = .013,p = .910 for Group 3; F(1, 19) < .0005,p = 1.000

139

for Group 4), suggesting that these groups made no differentiation in rating them
irrespective of their readings.

Next, we look at the judgments of the ni indirect passive sentences with transitive
verbs in both unfavorable and favorable contexts for the passive subject. Example
sentences of each type are given in (57) below, i.e., (57a) with an unfavorable context

(thus grammatical), and (57b) with a favorable context (thus ungrammatical):

(57)

a. warui koto ni, Bill-wa J ane-ni girlfriend kara no tegarni-o yom-are-ta
unfortunately -TOP -by letter from his girlfriend-ACC read-Pass-Pst
‘unfortunately, Bill was adversely affected by Jane reading a letter from his girlfriend’

b. *saiwaina koto ni, Bill-wa Jane-ni girlfriend kara no tegami-o yom-are-ta
fortunately

”fortunately, Bill was adversely affected by Jane reading a letter from his girlfriend’

Table 17 presents the statistical data on all groups’ judgments of these
grammatical and ungrammatical sentences. In Figure 8 numbers 1 and 2 on the x-axis
represent the judgments of respective readings. Figure 8 indicates that the subjects’
judgment patterns in the ni indirect passive with transitive verbs replicated those found
for the ni indirect passive with intransitive verbs. That is, it appears that Groups 1 and 2
made differential judgments between adversative and non-adversative readings whereas

Groups 3 and 4 did not.

140

Table 17: The ni indirect passive with transitive verbs

with adversative and non-adversative readings

 

 

 

Sentence Types Greys Mean SD N
transitive 1 1 .50 .79 3 1
indirect with 2 1.34 .73 25
adversative 3 .3 1 1 .56 36
readinga 4 .25 1.28 20
*transitive l -.23 1.25 31
indirect with 2 -.36 1.30 25
non-adversative 3 .40 1 .34 36
readingb 4 .38 1.20 20

 

 

a., b' The test sentences include #19 and #22 with adversative
reading; and #21 and #24 for non-adversative reading.

 

2.0 ’

1.0+

Grammaticality Judgment Means

 

i

 

 

 

1. indirect with adversity

2

Groups 1-4

NS control
(G 1)

High adv.
(G 2)

D

U

B Low adv.
(G 3)

 

CI lnterrn.
(G 4)

2. *indirect without adversity

Figure 8: The ni indirect passive with tr. verbs with +/- adversative readings

A Repeated-Measures AN OVA for all groups was run on the passive types. The

results showed that there was a significant interaction effect between passive types and

141

groups (F(3, 108) = 12.754, p < .0005). Thus a Repeated-Measures AN OVA was
performed for each group. The results for Groups 1 and 2 indicated that there were
significant passive-type main effects (F (1, 30) = 36.01 1, p < .0005 for Group 1; F (1, 24)
= 28.195, p < .0005 for Group 2), indicating that both groups judged the ni indirect
passive with transitive verbs differently depending on adversative and non-adversative
readings. Likewise, Groups 3 and 4 ANOVA results showed no significant judgment
differences (F(1, 35) = .140,p = .710 for Group 3; F(1, 19) = .152,p = .701 for Group 4),
suggesting that these groups made no distinction in rating them regardless of their
readings.

Finally, we examine the ni direct passive sentences in both unfavorable and
favorable contexts for the passive subject. Example sentences of each type are repeated
in (58) below, viz., (58a) with an unfavorable context (thus grammatical), and (58b) with

a favorable context (thus ungrammatical):

(5 8) (= (3 8))
a. Jane-ga mukasi no kare-ni ni-zikan mo mat-arete komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst
‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

b. *Jane-ga mukasi no kare-ni ni-zikan mo mat-arete uresikat-ta
happy-Pst

”Jane was happy by being adversely affected by having been waited for by her ex-boyfriend
for as many as two hours’
Table 18 presents the statistical data on all groups’ judgments of these
grammatical and ungrammatical sentences. In Figure 9 numbers 1 and 2 on the x-axis

signify the judgments of respective readings. Figure 9 indicates that groups’ patterns in

judgment of the ni direct passive were different from those of the ni indirect passive. It

142

seems that Group 3 as well as Groups 1 and 2 made distinctions in judging the ni direct

passive between adversative and non-adversative readings whereas Group 4 did not.

Table 18: The ni direct passive with adversative and non-
adversative readings

 

 

 

Sentence Types Groups Mean SD N

direct with 1 1.84 .58 31
adversative 2 1 .52 l .08 25
readinga 3 1.39 1.05 36
4 .65 1 .42 20

*direct with l -. 16 1.68 31
non-adversative 2 -.76 1 .5 1 25
readingb 3 -.36 1.57 36
4 .10 1.33 20

 

 

a., b. . . .
The test sentences include #25 With an adversative

reading and #27 with a non-adversative reading.

 

 

.0

15 ...‘
m . &\ 0..
g 5‘ °.
0 \

A
E 1 o 1 \ .. Groups 14
E NS
35 C] control
3 (G 1)
B. - —
8 C) High adv.
1’5 (G 2)
E — I
g 3 B Low adv
(G 3)
D lnterm.
-1.0 (G 4)

 

 

 

2

1. direct with adversity 2. *direct without adversity

Figure 9: The ni direct passive with +/- adversative readings

143

A Repeated-Measures ANOVA for all groups was run on the passive types. The
results showed that there was a significant interaction effect between passive types and
groups (F(3, 108) = 3.507, p = .018). Thus a Repeated-Measures ANOVA was
performed for each group. The results for Groups 1, 2, and 3 showed that there was a
significant passive-type main effect (F (l, 30) = 48.947, p < .0005 for Group 1; F (1, 24) =
37.561, p < .0005 for Group 2; F(1, 35) = 24.617, p < .0005 for Group 3), indicating that
these three groups made differential judgments of the ni direct passive between
adversative and non-adversative readings. However, Group 4 ANOVA results showed
no significant judgment difference (F (1, 19) = 1.620, p = .218), suggesting that Group 4
made no distinction in rating the ni direct passive irrespective of their readings.

In summary, as for the ni indirect passive both with intransitive and transitive
verbs, the highly advanced group judged differently between adversative and non-
adversative readings whereas the other two groups failed to distinguish between them.
For the ni direct passive, the highly advanced and the low advanced groups made
distinctions in judgment according to whether they carried adversative or non-adversative

readings, but the intermediate group did not.

4.3.2.2.2. Judgments of adversative readings across the ni passives

Second, we compare subjects’ judgments of the following three types of ni
passives with adversative readings, i.e., the ni indirect passive with intransitive and
transitive verbs, and the ni direct passive, in order to determine if any particular type of
passive was judged higher than the others. For the sake of exposition, example sentences

of each type are provided in (59) below in the order given above:

144

(59)

a. (= (3 73))
J ohn-ga warui told-ni tomodati-ni ko-rare-ta
-NOM at an inconvenient time friend-by come-Pass-Pst

‘John was adversely affected by his friend visiting him at an inconvenient time’

b. (= (56a))
warui koto ni, Bill-wa Jane-ni girlfiiend kara no tegami-o yom-are-ta
unfortunately -TOP -by letter from his girlfriend-ACC read-Pass-Pst
‘unfortunately, Bill was adversely affected by Jane reading a letter from his girlfriend’
c. (= (3 8))
Jane-ga mukasi no kare-ni ni-zikan mo mat-arete komat-ta
-NOM ex-boyfiiend-by as many as 2 hours wait-Pass annoy-Pst

‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

Table 19 provides the statistical data on all groups’ judgments of these three types
of passive sentences. In Figure 10 numbers 1, 2, and 3 on the x-axis represent the
judgments of the respective passive above. As illustrated in Figure 10, Groups 1 and 2
appear to have judged these passives without much difference from each other (with the
exception of Group 1’s slightly larger judgment difference between the ni indirect passive
with intransitive verbs and the ni direct passive). On the other hand, Group 3 seems to
have rated the ni direct passive much higher than either type of the ni indirect passive.

Group 4 appears to have judged all three types of passives indiscriminatingly.

145

Table 19: The ni indirect passive with intransitive plus
transitive verbs and the ni direct passive with adversative

 

 

 

 

 

 

readings
Sentence Groups Mean SD N
T was
ni indirect 1 1.23 .84 31
passive w/ 2 1.46 .85 25
intr. verbsa 3 .0139 1.37 36
4 .58 1 .38 20
ni indirect 1 1.50 .79 31
passive w/ 2 1.34 .73 25
tr. verbs 3 .31 1.56 36
4 .25 1 .28 20
ni direct l 1.84 .58 31
passivec 2 1.52 1.08 25
3 1 .39 l .05 36
4 .65 1 .42 20
a., b., c.

The test sentences include #13 and #16 for the ni
indirect passive with intransitive verbs; and #19 and #22 for
the ni indirect passive with transitive verbs; and #25 for the ni
direct passive.

146

 

2.0

 

 

 

 

(0
Fa a
a -———“ i
’15 Groups 1-4
a) I I I I
CE» 0 NS control
'0
i (G 1)
E 0 High adv.
g 3 (G 2)
s ”I. ..
r. U 0W3 V.
(D
(G 3)
U lnterrn.
, (G 4)
1 2 3
1. indirect w/ intr. 2. indirect w/ tr. 3. direct

Figure 10: The ni indirect passive with intr./tr. verbs and ni direct passive
with adversative readings
A Repeated-Measures ANOVA for all groups was run on these passive types.

The results showed that there was a significant interaction effect between passive types
and groups (F (6, 216) = 2.624, p = .018). Thus a Repeated-Measures ANOVA was
performed for each group. The results for Group 1 indicated that there was a significant
passive-type main effect (F (2, 60) = 5.481, p = .007). Tests of within-subjects contrasts
showed 1) that there was no significant difference in judging the ni indirect passive with
intransitive and transitive verbs; and 2) that the ni direct passive was not rated differently
from the ni indirect passive with transitive verbs, but was rated higher than the ni indirect
passive with intransitive verbs.” These findings suggest that when sentences carried an

adversative reading, Group 1 judged the ni indirect passive with intransitive and

147

transitive verbs as well as the ni direct passive similarly (except for the judgment
difference between the ni indirect passive with intransitive verbs and the ni direct
passive). The AN OVA results for Group 2 showed that there was no significant passive-
type main effect (F(2, 48) = .319, p = .728), suggesting that Group 2 subjects judged
these three types of passives not differently. Note that in examining their ratings of three
syntactic types of passives (i.e., the ni yotte, the ni direct, and the ni indirect passives), it
was found that the ni indirect passive was judged lower than the ni direct passive. Thus,
it appears that a pronounced adversative reading in the ni indirect passive sentences
contributed to promoting Group 2 judgments of them.”

The ANOVA results for Group 3 showed that there was a significant passive-type
main effect (F (2, 70) = 11.448, p < .0005). Tests of within-subjects contrasts indicated
that the ni direct passive was judged significantly higher than both types of the ni indirect
passive, but that there was no significant difference in rating between the two types of the
ni indirect passive, suggesting that Group 3 subjects judged the m' direct passive higher
than the hi indirect passive. Note that this replicates the Group 3 judgment pattern found
for the analysis of three syntactic types of passives. Therefore, it seems that unlike
Group 2, a pronounced adversative reading in the ni indirect passive sentences did not
promote Group 3 judgments of them. Finally, no significant passive-type main effect
was found in Group 4 judgments of them (F(2, 38) = .426, p = .656), showing that Group
4 subjects rated these types of passives indiscriminately.

In all, the highly advanced group rated the ni indirect passive with intransitive and
transitive verbs as well as the ni direct passive all carrying adversative readings with no

judgment differences from each other. The low advanced group judged the ni direct

148

passive higher than the ni indirect passive while the intermediate group did not judge

them differently.

4.3.2.2.3. Summary of judgments of passives for adversative and non-adversative
readings

In sum, the highly advanced group demonstrated judgment patterns closely
conforming to those of the NS control group in rating the three types of passives (the ni
yotte, the ni direct, and the ni indirect passives) for an adversative reading while the low
advanced group did so only in judging the ni direct passive. The intermediate group did
not show any knowledge of them. (Their judgments of the ni yotte passive with an
adversative reading are examined in greater detail in subsequent analyses.)

Next, the highly advanced group made differential judgments between adversative
and non-adversative readings in all three types of passives: i.e., the ni indirect passive
with intransitive as well as transitive verbs, and the ni direct passive. Furthermore, this
group rated both types of the ni indirect passive with an adversative reading as high as its
ni direct counterpart, suggesting that a pronounced adversative reading in ni indirect
passive sentences promoted their judgments of them. On the other hand, the low
advanced group made a differential judgment between adversative and non-adversative
readings only in the ni direct passive. A pronounced adversative reading in the ni indirect
passive did not improve this group’s judgments. The intermediate group did not evince

evidence of their knowledge of an adversative reading of the three types.

4.3.2.3. Judgments of 711’ passive versus niyotte passive contrasts

149

This section presents the analyses of the NNS subjects’ judgments of ni versus ni
yotte passive contrasts effected in the following areas: 1) a perfective versus non-
perfective reading; 2) an adversative reading; 3) verb-induced viewpoint differences. We

examine each area in the order given above.

4.3.2.3.1. Judgments of perfective versus non-perfective readings

Due to its 0-subj ect status, the ni direct passive with an inanimate subject allows a

perfective reading, but not a non-perfective one, while owing to its non-O-subj ect status,
the ni yotte passive does not impose such restriction on possible readings. We investigate
if the NNS subjects made differential judgments of the ni direct and the ni yotte passives
depending on whether they carried perfective or non-perfective readings. Specifically,
we first look at their judgments of the ni direct and the ni yotte passives with the to-
marked perfective and past readings and second, examine those with the iru-marked
perfective and progressive readings to find out if their judgments reflected the perfective

versus non-perfective reading distinction when it was effected.

4.3.2.3.1.1. Judgments of to-marked perfective versus past readings

We first examine if the subjects observed the perfective versus non-perfective
contrast in judgment of to-marked ni direct and ni yotte passives as effected. Recall that
according to Kuroda (1979) the ni direct passive with a past reading is ungrammatical
while the others are grammatical. Example sentences of each type of passive are given in

(60) below, namely, (60a) for the ni direct passive with a past reading, (60b) for the ni

150

direct passive with a perfective reading, (60c) for the ni yotte passive with a past reading,

and (60d) for the ni yotte passive with a perfective reading:

(60)

a. (= (41b))
*‘Hamlet’-wa Shakespeare-ni kak-are-ta
‘Hamlet’-TOP Shakespeare-by write-Pass-Pst
‘Hamlet was written by Shakespeare’

b. (= (42b))
kimitu-syorui-ga tekikoku no supai-ni nusum-are-ta
secrest documents-NOM spy from an enemy country-by steal-Pass-Perf
‘secret documents are under the state affected by having been stolen by a spy from an
enemy country’

c. (= (41a))
‘Hamlet’-wa Shakespeare-n1 yotte kak-are-ta
‘Hamlet’-TOP Shakespeare-by write-Pass-Pst
‘Hamlet was written by Shakespeare’

d. (= (423))

kimitu-syorui-ga tekikoku no supai-ni yotte nusum-are-ta
secrest documents-NOM spy from an enemy country-by steal—Pass-Perf
‘secret documents have been stolen by a spy from an enemy country’

As before, Table 20 provides the statistical data on all groups’ judgments of these
four types of passives: the numbers 1 to 4 on the x-axis of Figure 11 represent the
judgments of the four types in the order given above in (60). As displayed in Figure 11,
Groups 1 and 2 appear to have judged the ni direct passive with a past reading differently
from the other three grammatical passive sentences, i.e., the ni direct passive with a
perfective reading and the ni yotte passive with both perfective and past readings. Group
3 seems to have judged the ni direct passive with a past reading slightly lower than the ni
direct passive with a perfective reading, but still rated the former higher than the ni yotte
passive. Finally, Group 4 appears to have rated all four types of passives without much

difference.

151

Table 20: Ni direct and ni yotte passives with past and
perfective readings

 

 

 

 

 

 

 

Sentence Types Groups Mean SD N
*ni direct 1 -1.290 .920 31
with past 2 .260 1.487 25
readinga 3 .625 1.556 36
4 .850 1.171 20
ni direct 1 1.98 .0898 31
with perfective 2 1.74 .77 25
reading 3 1.29 1.07 36
4 .95 1.04 20
ni yotte 1 1.74 .56 31
with past 2 1.10 1.13 25
readingc 3 .35 1.32 36
4 .33 1.12 20
ni yotte 1 1.89 .38 31
with perfective 2 1.32 .83 25
readingd 3 .29 1.22 36
4 .82 .89 20
a., b., c., d.

The test sentences for the hi direct passive include
#40 and #42 (for a past reading); and #44 and #46 (for a
perfective reading) whereas the test sentences for the ni yotte
passive involve #39 and #41 (for a past reading); and #43 and
#45 (for a perfective reading).

152

 

3.0

 

 

 

 

 

(0 Us I
c IIIII'II. 'lll"'..'£J
a o-
2
3, ~ ’_ ’ _ .mGroups 1-4
5 ~? — — I I I I
g) 3 0 NS control
'0
g (G 1)
a J - -
= High adv.
3% D (G 2)
E _ I
§ Low adv
(1.5 D
(G 3)
D lntenn.
-2.0 . . (G 4)
1 2 3 4
l. .ni direct w/past reading 3. ni yotte w/past reading
2. ni direct w/perfective reading 4. ni direct w/perfective reading

Figure 11: Ni direct and ni yotte passives with past and perfective readings

A Repeated-Measures ANOVA for all groups was run on the passive types. The
results showed that there was a significant interaction effect between passive types and
groups (F (9, 324) = 19.016, p < .0005). Thus a Repeated-Measures AN OVA was
performed for each group. The results for Group 1 indicated that there was a significant
passive-type main effect (F(3, 90) = 260.634, p < .0005). Tests of within-subjects
contrasts showed that the ni direct passive with a past reading was judged significantly
lower than the other three types of passives; and that there was no difference in judging
the remaining three types of passives (except for the judgment difference between the ni
direct passive with a perfective reading and the ni yotte passive with a past reading“).

These suggest that Group 1 clearly judged the ni direct and ni yotte passives with past and

153

perfective readings on the basis of the perfective versus non-perfective distinction as
effected. For Group 2, there was a significant passive-type main effect (F (3, 72) = 9.421,
p < .0005). Tests of within-subj ects contrasts indicated that the ni direct passive with a
past reading was judged significantly lower than the other three types of passives;l7 and
that no significant difference was found in judging the remaining three types of passives
(except for the judgment difference between the ni direct passive with a perfective
reading and the ni yotte passive with a past reading). These findings suggest that there
was a tendency for Group 2 subjects to judge these four types of passives according to the
perfective versus non-perfective distinction as it arose.18
For Group 3, there was also a significant passive-type main effect (F (3, 105) =
5.260, p = .002). Tests of within-subjects contrasts indicated 1) that the ni direct passive
with a past reading was judged significantly lower than the one with a perfective
reading;19 2) the ni direct passive with both readings was rated significantly higher than
the ni yotte passive with both readings; and 3) that there was no significant judgment
difference between the ni yotte passive with past and perfective readings. However, the
fact that their respective mean judgment scores of the ni yotte passive were .35 and .29,
suggests that Group 3 subjects did not have knowledge of this type of passive (see also
the syntactic analysis of the ni yotte passive). All these findings indicate that there was a
tendency (though not as clear as evinced for Group 2) for Group 3 subjects to judge the ni
direct passive with a past reading differently from the one with a perfective reading.
Finally, there was not a Significant passive-type main effect for Group 4 (F (3, 57)
= 1.563, p = .208). Since the other possible classification of these four types of passives

is in a ni direct passive and ni yotte passive contrast, another Repeated-Measures

154

AN OVA was run on this contrast with two readings of each passive type combined. This
did not reveal a significant judgment difference (F(1, 19) = 1.969, p = .177), either.
These results indicate that Group 4 subjects did not make either the perfective versus
non-perfective or the ni direct passive versus ni yotte passive distinctions in rating these
four types of passives.

In summary, in judging the ni direct and the ni yotte passives with past and
perfective readings, there was a tendency for Group 2 subjects to base their judgments on
the perfective versus non-perfective distinction when it arose just as the NS control group
did. For Group 3, there was a weak tendency for them to rate the ni direct passive
differently depending on their past and perfective readings. Group 4 subjects did not
exhibit either the perfective versus non-perfective or the ni direct passive versus ni yotte

passive distinctions in judging them.

4.3.2.3.1.2. Judgments of iru-marked perfective versus progressive readings

We now examine if the NNS subjects observed the perfective versus non-
perfective contrast in judgment of iru-marked ni direct and ni yotte passives as effected.
Recall that the ni direct passive with a progressive reading is ungrammatical while the
others are grammatical according to Kuroda (1979). Example sentences of each type of
passive are repeated in (61) below, i.e., (61a) for the ni direct passive with a progressive
reading, (61b) for the ni direct passive with a perfective reading, (61c) the ni yotte
passive with a progressive reading, and (61d) the ni yotte passive with a perfective

reading:

155

(61)

a. (= (39b))
*atarasii konpyuutaa no puroguramu-ga J ohn-ni tukur-are-te iru
new computer program-NOM -by make-Pass-Prog-Pres
‘a new computer program is being made by John’
b. (= (4%))
intaanetto-wa sekai-zyuu no hito-bito-ni tukaw-are-te iru
Internet-NOM all over the world people-by use-Pass-Perf-Pres
‘Internet is under the state affected by having been used by people all over the world’
c. (= (3%))
atarasii konpyuutaa no puroguramu—ga J ohn-ni yotte tukur-are-te iru
new computer program-NOM -by make-Pass-Prog-Pres
‘a new computer program is being made by John’
d. (= (40a))
intaanetto-wa sekai-zyuu no hito-bito-ni yotte tukaw-are-te iru
Internet-NOM all over the world people-by use-Pass-Perf-Pres

‘Internet has been used by people all over the world’

Table 21 gives the statistical data on all groups’ judgments of these four types of
passives: the numbers 1 to 4 on the x-axis of Figure 12 indicate the judgments of the four
types in the order given above in (61). As is illustrated in Figure 12, Group 1 seems to
have judged the ungrammatical ni direct passive with a progressive reading differently
from the other three grammatical passives, i.e., the ni direct passive with a perfective
reading and the ni yotte passive with both perfective and progressive readings. Group 2
appears to have rated the ungrammatical ni direct passive with a progressive reading
differently from the grammatical ni direct passive with a perfective reading; however,
their judgments of the ni yotte passive grammatical sentences with both readings seem to
be somewhat lower than that of the ni direct passive grammatical sentences with a
perfective reading, but higher than that of the ni direct passive ungrammatical sentences
with a progressive reading. Group 3 seems to have judged the ni direct passive with a

progressive reading marginally (at best) lower than the ni direct passive with a perfective

156

reading, but rated the former noticeably higher than the ni yotte passive with both
readings. Lastly, Group 4 appears to have rated all four types of passives without much

difference.

Table 21: Ni direct and ni yotte passives with progressive
and perfective readings

Sentence Types Groups Mean SD N

 

 

 

 

 

 

*ni direct with 1 -1.15 1.04 31
progressive 2 .48 1 .29 25
readinga 3 .79 1.19 36
4 .98 1.08 20
ni direct 1 1.74 .58 31
with perfective 2 1.66 .77 25
reading)? 3 1.13 1.15 36
4 1.12 .86 20
ni yotte with l 1.71 .57 31
progressive 2 .88 1.17 25
readingc 3 .33 1.41 36
4 .70 1.25 20
ni yotte 1 1.61 .60 31
with perfective 2 1.20 1.02 25
reading 3 .53 1.32 36
4 .72 1.06 20
a.,IS., c., d.

The test sentences for the ni direct passive include
#32 and #34 (for a progressive reading); and #36 and #38 (for
a perfective reading) whereas the test sentences for the ni yotte
passive involve #31 and #33 (for a progressive reading); and
#35 and #37 (for a perfective reading).

157

 

 

 

 

 

 

 

2.0
EIIIIIDIII-Qfl......-.
\ ' ‘ ' -CJ
2
m I]
2
._. Groups 1-4
8 ‘13 I I I I
g, . “’53 0 NS control
3 (G 1)
ET - -
E .. B High adv.
E .' (G 1)
g 351 ‘. _ L dv
b ' 0W3 .
e .‘ D
. (G 3)
-101:
E!) D lnterrn.
-1.5 . ' (G 4)
1 2 3 4
1. .ni direct w/progressive reading 3. ni yotte w/progressive reading
2. ni direct w/perfective reading 4. ni yotte w/perfective reading

Figure 12: Ni direct and ni yotte passives with progressive and perfective readings

A Repeated-Measures ANOVA for all groups was run on the passive types. The
results showed that there was a significant interaction effect between passive types and
groups (F(9, 324) = 15.673, p < .0005). Thus a Repeated-Measures AN OVA was
performed for each group. The results for Group 1 indicated that there was a significant
passive-type main effect (F(3, 90) = 140.591, p < .0005). Tests of within-subjects
contrasts showed that the ni direct passive with a progressive reading was judged
significantly lower than the other three types of passives, and that there was no
significant judgment difference for the latter three types of passives. These suggest, as

evidenced in the judgments of ta-marked passives, that Group 1 subjects judged the ni

158

direct and ni yotte passives with progressive and perfective readings on the basis of the
perfective versus non-perfective distinction as effected.

The results of Repeated-Measures AN OVA for Group 2 showed that there was a
significant passive-type main effect (F (3, 72) = 6.691, p < .0005). Tests of within-
subjects contrasts indicated 1) that the ni direct passive with a progressive reading was
judged significantly lower than the ni direct passive with a perfective reading as well as
the ni yotte passive with a perfective reading, but was not judged differently from the ni
yotte passive with a progressive reading; 20 and 2) the ni yotte passive with both readings
was not rated differently from each other, but judged lower than the ni direct passive with
a perfective reading. In all, the ni direct passive with a progressive reading was judged
lower than the ni direct passive with a perfective reading whereas the judgments of the ni
yotte passive were located between those of two readings of the ni direct passive. These
findings all indicate that there was a tendency for Group 2 subjects to judge the ni direct
and ni yotte passives with progressive and perfective readings on the basis of the
perfective versus non-perfective distinction as it arose although their relatively lower
ratings of the iru-marked ni yotte passive with a progressive reading made this
differentiation pattern less transparent than that of to-marked passives.

For Group 3, there was a significant passive-type main effect (F (3, 105) = 3.002,
p = .034). Tests of within-subjects contrasts showed 1) that the ni direct passive with a
progressive reading was not rated differently from the ni direct passive with a perfective
reading and from both readings of the ni yotte passive; 2) that the ni direct passive with a
perfective reading was judged significantly higher than the ni yotte passive with a

progressive reading and the rating of the former approached a significant level with

159

respect to that of the ni yotte passive with a perfective reading (p = .054); and 3) two
readings of the ni yotte passive were not judged differently from each other. These
indicate that Group 3 did not judge these types of passives depending on their perfective
versus non-perfective reading contrast as effected. On the other hand, 3 Repeated-
Measures ANOVA on the syntactic classification, i.e., the ni direct passive and the ni
yotte passive with two readings of each passive type combined, showed a significant
passive-type main effect (F(1, 35) = 4.919, p = .033). This suggests that Group 3’s
judgments were based on the ni direct passive versus the ni yotte passive distinction.

Finally, a Repeated-Measures AN OVA for Group 4 showed that there was not a
significant passive-type main effect (F (3, 57) = .778, p = .511). Nor did a Repeated-
Measures AN OVA on the ni direct passive and the ni yotte passive (combined for two
readings) reveal a significant judgment difference (F (1, 19) = 1.639, p = .216). These
results suggest that Group 4 subjects did not make either a perfective versus non-
perfective or a ni direct passive versus ni yotte passive distinction in rating iru-marked ni
direct and ni yotte passives.

In short, in judging the ni direct and the ni yotte passives with progressive and
perfective readings, the tendency emerged that Group 2 subjects based their judgments on
the perfective versus non-perfective distinction when effected as did the NS control
group, although it was less clear than that found for to-marked passives. Group 3
subjects, responding differently from their judgments of to-marked passives, based their
judgments on the ni direct passive versus the ni yotte passive distinction. Group 4 did not

exhibit any differential judgments of these four types of passives.

160

4.3.2.3.1.3. Summary of judgments of perfective versus non-perfective readings

In summary, in judging to- and iru-marked ni direct and ni yotte passives with an
inanimate subject carrying perfective and non-perfective readings (i.e., to-marked past
and iru-marked progressive readings), the tendency was found that the highly advanced
group based their judgments on the perfective versus non-perfective distinction as
effected although the tendency for iru-marked passives was less pronounced than that for
to-marked passives. The low advanced group demonstrated a perfective versus non-
perfective distinction in rating the to—marked ni direct passive, while they judged iru-
marked passives on the basis of the ni direct passive versus the ni yotte passive
distinction. The intermediate group failed to Show any distinction in judging all these

types of passives.

4.3.2.3.2. Judgments of the ni yotte indirect and direct passive with an adversative
reading

The ni yotte marking is incompatible with an adversative reading since the former
indicates a non-0—subj ect status while the latter requires a 0-subject. We examine
whether the NNS subjects made differential judgments in grammaticality between the ni
passive and the ni yotte passive with an adversative reading on the basis that the former is
consonant with an adversative reading whereas the latter is not. Specifically, we compare
their judgments of the ni indirect passive (with intransitive and transitive verbs) to those
of the ni yotte indirect passive as well as their rating of the ni direct passive to that of the

ni yotte direct passive.

161

First, we look at the subjects’ judgments of ni and ni yotte indirect passives with
intransitive verbs carrying an adversative reading. Example sentences of both types are
provided in (62) below, viz., (62a) for the ni-marked, hence grammatical, indirect passive

and (62b) for the ni yotte-marked, hence ungrammatical, indirect passive:

(62)
a. (= (43a))
J ohn-ga warui toki-ni tomodati-n1 ko-rare-ta
-NOM at an inconvenient time fi'iend-by come-Pass-Pst
b. (= (43b))
‘John-ga warui toki-ni tomodai-ni yotte ko-rare-ta
.by

‘John was adversely affected by his friend visiting him at an inconvenient time’

Table 22 provides the statistical data on all groups’ judgments of these two types
of passives. In Figure 13 numbers 1 and 2 on the x-axis represent the judgments of them.
As shown in Figure 13, Groups 1 and 2 appear to have made a distinction in judging them
depending on whether the indirect passive is marked by ni or ni yotte; in contrast, Groups

3 and 4 did not exhibit such distinction in judgment.

Table 22: Ni and ni yotte indirect passives with
intransitive verbs with adversative readings

 

 

 

Sentence Types Groups M can SD N
ni indirect 1 1.23 .84 31
w/intr. verbsa 2 1:46 '85 25
3 .0139 1.37 36
4 .58 1 .38 20
*ni yotte l -1.63 .61 31
indirect 2 -.60 1.27 25
w/intr. verbsb 3 “51 1-32 36
4 .22 1 .09 20

 

 

., b. . . .
a The test sentences 1nclude #13 and #16 for the ni mdrrect
passive; and #14 and #17 for the ni yotte indirect passive.

162

Grammaticality Judgment Means

 

2.0 ’

1.5

1.0"

J

 

 

1. ni indirect w/intr.

 

2

Groups 1-4

0 NS control

(G1)

High adv.
(G 2)

El

1:) Low adv.
(G 3)

D lnterm.
(G 4)

2. ‘ni yotte indirect w/intr.

Figure 13: Ni and ni yotte indirect passives with intr. verbs with adversative readings

A Repeated-Measures ANOVA for all groups was run on the passive types. The

results showed that there was a Si grrificant interaction effect between passive types and

groups (F(3, 108) = 18.262, p < .0005). Thus a Repeated-Measures AN OVA was

performed for each group. The results for Groups 1 and 2 indicated that there was a

significant passive-type main effect (F (l, 30) = 335.428, p < .0005 for Group 1; F(1, 24)

= 42.323, p < .0005 for Group 2), suggesting that Groups 1 and 2 made a grammatical

distinction in rating ni and ni yotte indirect passives with intransitive verbs carrying

adversative readings depending on whether they are marked by ni or ni yotte. On the

other hand, the AN OVA results of Groups 3 and 4 showed no significant judgment

differences (F(1, 35) = 3.251,p = .080 for Group 3; F(1, 19) = .831,p = .374 for Group

163

4), indicating that these two groups made no distinction in rating them irrespective of the
marking difference.

Second, we examine the subjects’ judgments of ni and ni yotte indirect passives
with transitive verbs carrying an adversative reading. Example sentences of both types
are given in (63) below, i.e., (63a) for the ni-marked, hence grammatical, indirect passive

and (63b) for the ni yotte-marked, hence ungrammatical, indirect passive:

(63)
a. (= (56a))
warui koto ni, Bill—wa Jane-n1 girlfriend kara no tegarni-o yom-are-ta
unfortunately ~TOP -by letter from his girlfriend-ACC read-Pass-Pst
b. *warui koto ni, Bill-wa Jane-ni yotte girlfriend kara no tegami-o yom-arc-ta
.by

‘unfortunately, Bill was adversely affected by Jane reading a letter from his girlfriend’

Table 23 presents the statistical data on all groups’ judgments of these two
passives. In Figure 14 numbers 1 and 2 on the x-axis represent the judgments of each
type of passive. As illustrated in Figure 14, Groups 1 and 2 appear to have made a
distinction in rating them, as found in their judgments of the indirect passive with

intransitive verbs,21 whereas Groups 3 and 4 rated them similarly.

164

Table 23: Ni and ni yotte indirect passives with transitive
verbs with adversative readings

 

 

 

 

Sentence Groups Mean SD N
Types
ni indirect l 1.50 .79 31
w/tr. verbsa 2 1.34 .73 25
3 .31 1.56 36
4 .25 1 .28 20
*ni yotte 1 .26 1.22 31
indirect 2 .0600 1 .45 25
Wm. verbsb 3 -0.833 1.48 36
4 .0250 1.1 1 20
Tb

 

" ' The test sentences include #19 and #22 for the ni indirect
passive; and #20 and #23 for the ni yotte indirect passive.

 

2.0 ’

1.0“

Grammaticaility Judgment Means

 

 

 

 

1. ni indirect w/tr.

Groups 1-4

NS control
(G 1)

High adv.
(G 2)

D

2. .ni yotte indirect w/tr.

Figure 14: Ni and ni yotte indirect passives with tr. verbs with adversative readings

A Repeated-Measures AN OVA for all groups was run on the passive types. The

results showed that there was a significant interaction effect between passive types and

165

groups (F(3, 108) = 3.575, p = .016). Thus a Repeated-Measures ANOVA was
performed for each group. The results for Groups 1 and 2 indicated that there was a
significant passive-type main effect (F ( l, 30) = 21.430, p < .0005 for Group 1; F ( l, 24) =
28.461 , p < .0005 for Group 2), suggesting that these two groups judged the ni yotte
indirect passive differently from the ni indirect passive. In contrast, the AN OVA results
of Groups 3 and 4 showed no significant judgment differences (F (l, 35) = 1.549, p = .222
for Group 3; F (1 , 19) = .865, p = .364 for Group 4), indicating that they did not make
differential j udgrnents of them.

Third, we investigate subjects’ judgments of the ni direct and the ni yotte direct
passives carrying an adversative reading. Example sentences of both types are repeated
in (64) below, viz., (64a) for the ni—marked, hence grammatical, direct passive and (64b)

for the ni yotte-marked, hence ungrammatical, direct passive:

(64)
a. (= (43c))
Jane-ga mukasi no karc-ni ni-zikan mo mat-aretc komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst
b. (= (43d))
*Janc-ga mukasi no kare-ni yotte ni-zikan mo mat-arete komat-ta
.by

‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

Table 24 provides the statistical data on all groups’ judgments of these two types
of passives. In Figure 15 numbers 1 and 2 on the x-axis represent the judgments of each
type of passive. As displayed in Figure 15, Groups 1, 2, and 3 appear to have made a
distinction in rating ni and ni yotte direct passives more clearly as their proficiency levels

increased. Group 4 seems not to have rated them differently.

166

Table 24: Ni and m‘ yotte direct passives with adversative

 

 

 

readings
Sentence Groups Mean SD N
Types
ni direct l 1.84 .58 31
passivea 2 1.52 1.08 25
3 1.39 1.05 36
4 .65 1 .42 20
*ni yotte l -.42 1.36 31
direct 2 .24 1.67 25
passive 3 .67 1.62 36
4 .65 1.42 20

 

 

a., b. . . . .
The test sentences 1nclude #25 for the m direct passrve

and #26 for the ni yotte passive.

 

 

 

 

 

2.0;
1.5 ’°.
‘0 EK‘~ ..
5 'x .
0) N1,
2 \'\.
E 1.0- °>a \.\
I) ’~ '
E 0.. ~\ I\.
O) n .4 ~‘ “ Fl
'0 [4 I ‘\
3 ’. §
_’ 5" Q. ~
g 0.. ‘~i
g; 00 .
E ' ..
E 0,
(U o
h .0
(D E]
-.5‘
-1.0
1 2
1.nidirect

Groups 14

l
D NS control

(61)

High adv.
(G 2)

U

B Low adv.
(G 3)

 

U lnterm.
(G 4)

2. ‘ni yotte direct

Figure 15: Ni and ni yotte direct passives with adversative readings

A Repeated-Measures AN OVA for all groups was run on the passive types. The

results showed that there was a significant interaction effect between passive types and

167

groups (F (3, 108) = 7.704, p < .0005). Thus a Repeated-Measures ANOVA was
performed for each group. The results for Groups 1, 2, and 3 indicated that there was a
significant passive-type main effect (F (1, 30) = 74.168, p < .0005 for Group 1; F (1, 24) =
10.798, p = .003 for Group 2; F(1, 35) = 6.017, p = .019 for Group 3), suggesting that
these three groups judged ni and ni yotte direct passives differently. In contrast, the
AN OVA results of Group 4 showed no significant judgment difference (F (1, 19) = .000,
p = 1.000), indicating that Group 4 subjects did not make differential judgments of them.
However, in order to fully determine that Groups 2 and 3 ’s differential judgments
between ni and ni yotte direct passives with adversative readings reflected their
knowledge that the direct passive is compatible with a ni marking but not with a m' yotte
marking, it is necessary to show that the judgment difference between ni and ni yotte
direct passives with pronounced adversative readings was greater than that between m'
direct and ni yotte passives without them as exemplified by the test sentences used for the
syntactic analysis. This is so because the results of the syntactic analysis showed that the
ni yotte passive was judged lower than the ni direct passive by Groups 2 and 3. In other
words, it is essential to demonstrate that the (ungrammatical) ni yotte direct passive
sentences with an adversative reading received a significantly lower grammaticality
judgment than the (grammatical) ni yotte sentences lacking in such connotation?”
Thus, a Repeated-Measures AN OVA for all groups was run on two types of ni
yotte passive sentences: one type with an adversative reading and the other without it
(used for the syntactic analysis)“. The results showed that there was a significant
interaction effect between passive types and groups (F (3, 108) = 8.653, p < .0005). The

Repeated-Measures AN OVAs for each group indicated that Group 1 judged these two

168

types of the ni yotte passive differently (F(1, 30) = 33.815, p < .0005), but that Groups 2,
3, and 4 judged them indiscriminately (F(1, 24) = 1.988, p = .171 for Group 2; F(1, 35) =
1.938, p = .173 for Group 3; F( 1, 19) = 1.776, p = .198 for Group 4). These suggest that
NNS groups did not judge the two types of ni yotte passives differently depending on
whether or not they carried an adversative reading. Hence, it does not appear that Groups
2 and 3’s differential judgments between ni and ni yotte direct passives with adversative
readings clearly reflected their knowledge of the ni versus ni yotte marking difference.

In all, only the highly advanced group made contrastive judgments in
grammaticality between ni and ni yotte indirect passives carrying adversative readings.
On the other hand, no NNS group demonstrated clear differential judgments in
grammaticality between the ni direct passive and the ni yotte direct passive with an
adversative reading in that their judgment differences were not significantly larger than

those found between the grammatical ni direct and ni yotte passives.

4.3.2.3.3. Judgments of verb-induced viewpoint differences

The ni direct passive is harmonious with an expression of a personally involved
situation whereas the ni yotte passive is appropriate to depict a situation in an impersonal,
newspaper-report like manner. This viewpoint difference between the ni direct and the m'
yotte passives may be effected by lexical choice, particularly verbs. This section presents
the analysis of NNS subjects’ judgments of two such cases: 1) Japanese native versus
Sino-Japanese verb contrasts; and 2) affective (defined later) versus unaccusative verb

COHtl'flStS.

169

4.3.2.3.3.1. Judgments of Japanese native versus Sino—Japanese verb contrasts

Japanese native verbs are in harmony with the ni direct passive where situations
are expressed in an involved, personal manner, while Sino—Japanese verbs are suitable for
the ni yotte passive in which situations are depicted in an objective report-like fashion.
We investigate if the subjects made appropriate grammatical judgments of the ni direct
passive and, in particular, the ni yotte passive in accordance with the above stated
characterization. Specifically, we first examine the judgments of the ni direct passive
with Japanese native verbs and those of the ni yotte passive with Sino-Japanese verbs to
see if the subjects observed a basic distinction between them. Second, we ask if they
made a grammaticality distinction in judgment of ni yotte passive sentences used with
Japanese native verbs vis-a-vis those containing Sino-Japanese verbs in order to further
investigate whether or not they had the knowledge that the ni yotte passive is restricted to
an objective viewpoint expression.

Since lexical knowledge of verbs used in these passive sentences was critical in
judging them, NNS subjects who failed to choose correct meanings of the verbs in the
vocabulary test were excluded from the subject population for those sentences where
incorrectly selected verbs were used.”

First, we examine the subjects’ judgments of the ni direct passive with Japanese
native verbs and the ni yotte passive with Sine-Japanese verbs (both grammatical).
Example sentences of both types are provided in (65) below, i.e., (65a) for the former and

(65b) for the latter:

170

(65) a- (= (443))

Bill-wa kawaii musuko-ni koros-are-ta
-TOP beloved son-by lcill-Pass-Pst
‘Bill was affected by being killed by his own beloved son’
b. (= (44c))
Bill-wa CIA-ni yotte satugai-s-are-ta
-TOP the CIA-by murder-Pass-Pst

‘Bill was murdered by the CIA’

Table 25 provides the statistical data on all groups’ judgments of them. In Figure
16 numbers 1 and 2 on the x-axis represent the judgments of each type of passive. As
illustrated in Figure 16, for each group, there does not seem any great difference in

judging them.

Table 25: The ni direct passive with Japanese native
verbs and the ni yotte passive with Sino-Japanese verbs

 

 

 

Sentence Groups Mean SD N
Types
ni direct 1 1.161 1.150 31
with Japanese 2 1.087 1.125 23
native verbsa 3 .717 1.636 23
4 .500 1.179 10
ni yotte 1 1.61 .60 31
with Sino-Jap- 2 .80 1.38 23
anese verbsb 3 .46 1.55 23
4 .100 1.84 10

 

 

a.,? The test sentences include #1 and #4 for the ni direct
passive; and #3 and #6 for the ni yotte passive.

171

Grammaticality Judgment Means

1. ni direct with Japanese verbs

 

2.0

1.5‘

 

 

 

Groups 14

0 NS control

(61)

#3 C) High adv.
(G 2)

J B Low adv.
(G 3)

 

D lnterm.
(G 4)

 

 

1

2

2. ni yotte with Chinese verbs

Figure 16: The ni direct passive with Japanese native verbs and the m' yotte passive

with Sino-Japanese verbs

172

A Repeated-Measures ANOVA for all groups was run on the passive types. The
results showed that there was neither a significant passive—type main effect (F (1, 83) =
.389, p = .534) nor a significant interaction effect between passive types and groups (F (3,
83) = 1.322, p = .273), suggesting that all groups judged the ni direct passive with
Japanese native verbs and the nz' yotte passive with Sino-Japanese verbs without
difference. This finding contrasts with that obtained in the analysis of three syntactic
types of passives where the ni direct passive was judged higher than the ni yotte passive
by Groups 2 and 3. There are two possible sources for the different results arising

between the current analysis and the syntactic analysis: 1) the ni direct passive was

judged lower in the former analysis than in the latter; or 2) the m' yotte passive in the
former analysis was judged higher than in the latter.

In order to test the first possibility, a Repeated-Measures AN OVA for all groups
was run on two types of m‘ direct passive sentences: one type used for the current analysis
and the other for the syntactic analysis. The result showed that there was a significant
passive-type main effect (F (1, 105) = 12.940, p < .0005), but that there was no interaction
effect (F (3, 105) = .715, p = .545), suggesting that ni direct passive sentences used for the
current analysis were rated lower than those used for the syntactic analysis. Next, to test
the second possibility, a Repeated-Measures AN OVA for all groups was similarly
performed on two types of ni yotte passive sentences: one type used for the current
analysis and the other for the syntactic analysis. The result showed that there was no
significant passive-type main effect (F (1, 86) =.409, p = .524) nor was there an
interaction between passive types and groups (F (3, 86) = .632, p = .597), suggesting that
all groups rated the two types of m’ yotte passives without difference from each other.
These all suggest that the similar judgment between the ni direct passive with Japanese
native verbs and the ni yotte passive with Sine-Japanese verbs was due to the lower
ratings of the former type of passive, not because of the higher ratings of the latter. Note,
however, that the ni direct passive with Japanese native verbs was judged clearly as
grammatical as shown in Table 25.

In summary, Groups 2 and 3 judged both the ni direct passive with Japanese
native verbs and the ni yotte passive with Sino-Japanese verbs as grammatical while

Group 4 judged both of them without differentiation, but very low in grammaticality.

173

Second, we compare the subjects’ judgments of ni yotte passive sentences with
Japanese native verbs (describing a personal perspective) vis-a-vis those with Sino-
Japanese verbs (depicting an impersonal one) in order to explore if their judgments of
them were based on the observation that the m“ yotte passive was consonant with Sino-
Japanese verbs in an impersonal context, but not with Japanese native verbs in a
personalized one. Example sentences of both types are given in (66) below, viz., (66a) is
the m’ yotte passive with Sino-Japanese verbs (grammatical), and (66b) is that with

Japanese native verbs (ungrammatical):

(66)

3- (== (440))

Bill-wa CIA-ni yotte satugai-s-are-ta
-TOP the CIA-by murder-Pass-Pst

‘Bill was murdered by the CIA’

b. (= (44b)
*Bill-wa kawaii musuko-ni yotte koros-are-ta

-TOP beloved son-by kill-Pass-Pst

‘Bill was killed by his own beloved son’

Table 26 provides the statistical data on all groups’ judgments of them. In Figure
17 numbers 1 and 2 on the x-axis represent the judgments of each type of passive. As
shown in Figure 17, Groups 1, 2, and 3 (but not Group 4) seem to have made a distinction
in grammaticality between m“ yotte passive sentences with Sine-Japanese verbs and with

Japanese native ones.26

174

Table 26: The ni yotte passive with Sino-Japanese verbs
and with Japanese native verbs

 

 

 

Sentence Groups Mean SD N
Types
ni yotte l 1.61 .60 31
with Sino-Jap- 2 .80 1.38 23
anese verbsa 3 .46 1.55 23
4 . 100 1 .84 10
*ni yotte 1 .23 1.03 31
with Japanese 2 -. 15 1.18 23
native verbs 3 "43 1'60 23
4 -.0500 1 .48 10

 

 

., b. . .
a The test sentences 1nclude #3 and #6 for Sino-Japanese
verbs; and #2 and #5 for Japanese native verbs.

 

 

 

 

 

 

 

 

2.0
[3 .
1.5 - ' ~ . ,
U) ‘ o .
c - .
m ’ a
o ~ . .
2 ‘ ° ~ G 14
E 1.0 - 0 . . roups
C . ' O I U I
CE» $‘~~ "0.. Cl NScontrol
3 ~ ‘ .q a
'1 5 fix ‘ ~ . . . . (G 1)
b o \ ~ ~ 0. - -
g C] High adv.
1‘5 “L (G 2)
E W _...
5 B Low adv.
-.5 . (G 3)
D lnterm.
-1.0 (G 4)
1 2
l. ni yotte w/Chinese verbs 2. *m’ yotte w/Japanese verbs

Figure 17: The ni yotte passive with Sino-Japanese verbs and with Japanese
native verbs

175

A Repeated-Measures AN OVA for all groups was run on the passive types. The
results showed that there was a significant passive-type effect (F (1 , 83) = 25.103, p <
.0005) without an interaction effect (F (3, 83) = 1.951, p = .128). The Repeated-Measures
ANOVAs’ results for Groups 1, 2, and 3 indicated that there was a significant passive-
type main effect (F(1, 30) = 55.304, p < .0005 for Group 1; F(1, 22) = 8.660, p = .008 for
Group 2; F (1, 22) = 15.474, p = .001 for Group 3), suggesting that each of these three
groups judged the ni yotte passive with Japanese native verbs differently from those with
Sino-Japanese ones. In contrast, the AN OVA results of Group 4 showed no significant
judgment difference (F (1, 9) = .033, p = .859), indicating that Group 4 made no
distinction in rating between the two.

In summary, Groups 2 and 3 made differential judgments of the ni yotte passive
with Sino-Japanese and Japanese native verbs: the former was rated as grammatical
whereas the latter was rejected as ungrammatical. On the other hand, Group 4 failed to
demonstrate such a distinction.

On the whole, Groups 2 and 3 judged the ni direct passive and the ni yotte passive
based on the observation that Japanese native verbs (expressing a personal standpoint) are
fit with the former type passive whereas Sino-Japanese verbs (expressing an objective
one) are suitable for the latter type. Furthermore, Groups 2 and 3 rejected the ni yotte
passive when used with Japanese native verbs as ungrammatical, reflecting that the ni
yotte passive is not highly compatible with a personalized expression. Group 4 failed to

show any of these distinctions.

4.3.2.3.3.2. Judgments of affective and unaccusative verb contrasts

176

Verbs such as miru ‘see’and kiku ‘hear’ (referred to as affective verbs henceforth)
may be used in the ni direct passive but not in the ni yotte passive, partly because a
passive sentence with this type of verbs expresses the state of the passive subject
personally affected by the event described in it, and in part because there are
unaccusative versions corresponding to affective verbs without affected connotation,
which replace the ni yotte passive. We investigate if the subjects based their judgments
of these sentences with affective and unaccusative verbs on the observation given above.
Specifically, we first examine whether they made a grammaticality distinction between
the ni direct and the ni yotte passives with affective verbs; and, second, ask if they
accepted the unaccusative versions of the corresponding sentences.

Again, it was necessary to exclude those NNS subjects who failed to respond
correctly to the distractor sentence (#75) testing knowledge of the temporal adverbial
phrase marker 0 (phonetically identical with, yet syntactically different from, an
accusative marker 0)”, since knowledge of it was important to judge the relevant passive
sentences.

We first examine the subjects’ judgments of the ni direct and the ni yotte passives
with affective verbs. Example sentences of both types are repeated in (67) below, i.e.,

(67a) for the former (grammatical) and (67b) for the latter (ungrammatical):

(67)
a. (= (45a))
Mary to hanasi-te iru tokoro-o, Bill-ga gaaruhurendo—ni kik-are-ta
talking to Mary as -NOM his girlfi’iend-by hear-Pass-Pst
b. ( =(45b))
*Mary to hanasi-te iru tokoro-o, Bill-ga gaaruhurendo-ni yotte kik-are-ta
his girlfriend-by

‘Bill was affected by being heard by his girlfriend as he was talking to Mary’

177

Table 27 provides the statistical data on all groups’ judgments of these two types
of passive sentences. In Figure 18 numbers 1 and 2 on the x-axis represent the judgments
of each type of passive. As shown in Figure 18 Groups 1 and 2 seem to have made a
distinction in grammaticality between the two, while Group 3 did so to a much smaller

degree. Group 4 appears not to have made any differentiation between them.28

Table 27: Ni direct and ni yotte passives with affective

 

 

 

 

 

verbs
Sentence Groups Mean SD N
Types
ni direct 1 .08065 1.1 1 1 31
passivea 2 .864 1.037 22
3 .450 1.276 20
4 -.346 1.231 13
*ni yotte l -.77 1.12 31
passive 2 -.0682 1.29 22
3 -. 15 1.38 20
4 -.35 1.03 13
a b

" ' The test sentences include #7 and #10 for the ni direct
passive; and #8 and #11 for the ni yotte passive.

178

 

 

 

 

 

 

2.0
Iii.
1.51
ID ‘0.
c c.
(U a
a
.5: 101 Groups 1-4
.' III.
E [F‘s‘ 'u. 0 NScontrol
a N
-. 5k ~. (G1)
g ~\. ‘ .',. _—
g E \,\.\ \“~ I? D Highadv.
g 0.0. , M3 (G2)
\. ~ ‘-
E \.
6‘; \~\. [:1 Lowadv.
-5. ‘3 (63)
U Intenn.
-1.0 (G4)
1 2

1.ni direct with affective verbs

2. .ni yotte with affective verbs

Figure 18: Ni direct and ni yotte passives with affective verbs

A Repeated-Measures AN OVA for all groups was run on the passive types. The

results showed that there was a significant passive-type main effect (F(1, 82) = 11.269, p

.001) without an interaction effect between passive types and groups (F (3, 82) = 1.144, p

= .336). Repeated-Measures ANOVAs for Groups 1 and 2 indicated that there was a

significant passive-type main effect (F(1, 30) = 18.827, p < .0005 for Group 1; F(1, 21) =

6.507, p = .019 for Group 2), suggesting that these two groups judged the ni direct

passive differently from the ni yotte passive when used with affective verbs. In contrast,

the ANOVA results of Group 3 and 4 showed no significant judgment differences (F (1 ,

19) = 1.736, p = .203 for Group 3; (F(1, 12) < .0005, p = 1.000 for Group 4), indicating

that Groups 3 and 4 did not make a distinction in rating the two.

179

Furthermore, it was necessary to ensure that Group 2’s lower judgments of the ni
yotte passive sentences relative to its ni direct passive counterparts were based on the
observation that the ni yotte passive is not compatible with affective verbs. Hence, Group
2 judgments of the ni yotte passive with affective verbs were compared to those of the ni
yotte passive in the syntactic analysis. There was a significant passive-type main effect
(F (1, 21) = 9.240, p = .006), suggesting that Group 2 subjects did judge the
ungrammatical former type significantly lower than the grammatical latter type.

In summary, Group 2 made a differential judgment between the ni direct and the
ni yotte passives with affective verbs: the former type was rated as grammatical whereas
the latter was not accepted. On the other hand, Groups 3 and 4 failed to demonstrate such
a distinction.

Second, we look at the subjects’ judgments of unaccusative versions
corresponding to the ni direct passive sentences examined above. As before, it was
necessary to exclude from the subject population those NNS subjects who failed to
indicate their knowledge of unaccusative verbs, i.e., mieru ‘visible, be seen’ and kikoeru
‘audible, be heard’ in the vocabulary test.29 An example sentence of the unaccusative

version is given in (68) below:

(68) gaaruhurendo-ni, Bill-ga Mary to hanasi-te iru tokoro-ga kikoe-ta
his girlfriend-to -NOM talking to Mary as audible-Pst
‘Bill was audible to his girlfriend as he was talking to Mary’

Table 28 provides the statistical data on all groups’ judgments of the unaccusative

sentences. As shown, Group 1 unexpectedly rejected these sentences with unaccusative

180

verbs whereas Groups 2, 3, and 4 weakly accepted them (see Discussion for the NS

control group rejection).

Table 28: Unaccusative verbs

 

 

Groups Mean SD N
1 -.806 1.167 31
2 .318 1.341 22
3 .355 1.462 31
4 .125 1.234 20

 

 

Note: test sentences included #9 and #12.

A Oneway AN OVA was conducted on all four subject groups to explore how
different each group judgments were from each other. The results showed a significant
difference between groups (F (3, 100) = 5.129, p = .002). The Tukey HSD results showed
that Group 1 was different from Groups 2 and 3, but not from Group 4; and that Groups
2, 3, and 4 were not different from one another. Thus, Groups 2, 3, and 4 weakly
accepted the unaccusative sentences while Group 1 did not.

On the whole, in judging the ni direct and the ni yotte passives with affective
verbs, Group 2 made a differential judgment between the two by accepting the former
and not accepting the latter whereas Groups 3 and 4 did not demonstrate such a contrast.
However, in rating the unaccusative sentences, Group 1 failed to judge them as

grammatical while Groups 2, 3, and 4 did so, but only weakly.

4.3.2.3.3.3. Summary of judgments of verb-induced viewpoint differences
In summary, by examining the subjects’ judgments of the ni direct and, in
particular, the ni yotte passive sentences with respect to verb-induced viewpoint

differences, it was found that the highly advanced and low advanced groups accepted

181

both the ni direct passive sentences with Japanese native verbs (expressing an involved
viewpoint) and the ni yotte passive sentences with Sino-Japanese verbs (expressing an
objective one). Furthermore, both groups rejected the ni yotte passive when used with
Japanese native verbs, suggesting these subjects’ knowledge that they are not compatible
in viewpoint terms. Next, the highly advanced group, but not the low advanced group,
made a differential judgment between the ni direct passive grammatical sentences and the
ni yotte passive ungrammatical sentences with affective verbs. The intermediate group
failed to show any distinction examined above. Finally, in rating unaccusative sentences,

all NNS groups accepted them weakly.

182

CHAPTER 5

DISCUSSION

5.1. Introduction

This chapter first discusses the results of the analyses of linguistically naive and
sophisticated NSs’ grammaticality judgments and then considers the implications of the
findings of the NNSs’ judgments of syntactic and semantic properties of Japanese

passives.

5.2. Linguistically naive and sophisticated native speakers’ grammaticality
judgments

In examining the grammaticality judgments by linguistically naive and
sophisticated NSs, it was found, somewhat surprisingly, that the non-linguists made
judgments in nearly full agreement with the theory predictions whereas the linguists only
agreed with the predictions in approximately 45% to 7 0% (depending on measurement)
of 57 passive test sentences under investigation. In terms of the group judgment
consistency, the non-linguists gave inconsistent judgments only about 7% of the time
(two cases out of 27) whereas the linguists did so about 30% of the time (eight out of 27).
Thus, the present study clearly demonstrated that ordinary NSS of Japanese observe
Hoshi (1994a; 1999) and Kuroda's (1979) characterizations of passives including delicate
semantic nuances in grammaticality judgments.

Why then did the linguistically sophisticated group render judgments differently

from the theory predictions and exhibit unstable judgments? There are at least two

183

possible factors which may have influenced linguistically sophisticated NSs’ judgment
processes but not linguistically naive NSs’. First, the subjects of the former group may
have undergone the phenomenon described tersely in Fraser (1971) as follows: “I think
this issue is fairly clear. It will be resolved by speakers whose intuitions about the
sentences in question are sharper than mine, which have been blunted by frequent
worrying about these cases” (p. 178). (See Chomsky (1962) for a similar comment on a
loss of linguistic intuitions in the short term.) That is to say, linguistically sophisticated
NSs may have fallen into a state where their judgments become blurred, at least in the
short term, due to exposure to and reflection upon closely related sentences with subtle
differences. Indeed, the questionnaire developed for the present study contained as many
as 62 sentences all of which concerned properties of Japanese passives.l Perhaps, an
increased number of distractor sentences in the grammaticality judgment questionnaire
could have led at least to a reduction of this phenomenon.

Second, perhaps not surprisingly, linguistically sophisticated NSs may have
attempted to look for reasons guiding their judgment of a given sentence owing to their
training and practice in performing such analysis (Greenbaum, 1977a; 1977b, cited in
Schiitze, 1996). In so doing, they may have fallen susceptible to the situation where they
inadvertently focused on aspect(s) of the test sentence irrelevant to the present study,
thereby misleading them to an incorrect assessment of it. (See Schfitze, 1996 for a
general discussion of the possible pitfalls linguists may fall into during judgment
processes; and Valian, 1982 for an insightfirl consideration of the parallels between

linguists’ use of their own judgments and expert judgment in other fields.)

184

5.3. Non-native speakers’ grammatical and ungrammatical judgments of three
syntactic types of passives

For the sake of illustration, both grammatical and ungrammatical example
sentences of three syntactic types of passives are provided in (69)-(71) below. Sentences
in (69) illustrate the ni direct passive; the ungrammatical example (69b) results from the

lack of O-role suppression:

(69) a. Mary-wa John-hi itumo soodan-s—are-ru
(= (35)) -TOP -by always consult-Pass-Pres
‘Mary is affected by being always consulted by John’
b. *Mary-wa John-ga itumo soodan-s-are-ru
-TOP -NOM

Sentences in (70) show the ni yotte passive; the first ungrammatical type (70b) results
from the lack of accusative case absorption, while the second (70c) from the lack of 0-

role suppression:

(70) a. utukusii uta-ga biitoruzu-ni yotte takusan tukur-are-ta
(= (34)) beautiful songs-NOM the Beatles-by a lot make-Pass-Pst
‘many beautiful songs were made by the Beatles’
b. *biitoruzu-ni yotte utukusii uta-o takusan tukur-are-ta
-by -ACC
c. *utukusii uta-wa biitoruzu-ga takusan tukur-are-ta
-TOP -NOM

Sentences in (71) exemplify the ni indirect passive; the first ungrammatical type (71b) is

unpassivized in its embedded clause, and the second one (7 lo) is passivized:

185

 

(71) a. J ohn-wa kodomo-ni kuruma no mado—o war-are-ta
(= (36)) -TOP child-by car window-ACC break-Pass-Pst
‘John was affected by a child breaking the car window’

b. *John-wa kodomo-ga kuruma no mado-o war-are-ta
-TOP child-NOM car window -ACC
c. *John-wa kuruma no mado-ga kodomo-ni war-are-ta

-TOP car window-NOM child-by

First, through analysis of the NNS groups’ judgments of the grammatical
sentences of these three syntactic types of passives, it was found that the ni direct passive
such as (69a) above was judged highest whereas the other two types of passives were
rated indistinguishably. It was also evidenced that both the highly advanced and the low
advanced groups approximated the NS group in judging the ni direct passive. On the
other hand, only the highly advanced group approximated the NS control group in rating
the ni yotte passive such as (70a) while no NNS group did so in judging the ni indirect
passive such as (71a).

Second, the analyses of the ungrammatical sentences vis-a-vis their grammatical
counterparts revealed that the ni direct passive ungrammatical sentences (e.g., (69b))
were most clearly differentiated from the grammatical ones, and that the ni yotte passive
ungrammatical sentences (e. g., (70b) and (70c)) were distinguished only by the highly
advanced group, while one type of the ni indirect passive ungrammatical sentences (e. g.,
the unpassivized one (71b)) was distinguished from the grammatical ones by both the
highly advanced and the low advanced groups. Yet the other type of the ni indirect
passive ungrammatical sentences (e. g. the passivized one (71c)) was not differentiated

from its grammatical counterparts by any NNS group. Thus, in all, the ni direct passive

186

was judged best both for grammatical and ungrammatical sentences, while the ni yotte
passive and the ni indirect passive presented different kinds of judgment problems.

The ni direct passive was expected to be the first type to be learned in syntactic
terms due to its high input frequency and apparent similarity to the English passive. On
the other hand, the ni yotte passive, identical with the English passive, posed some
learning problems. These perhaps arose from the fact that the ni yotte passive is likely to
be used in formal reading and writing containing objective description (Howard &
Niyekawa-Howard, 1976; J acobsen, 1992; Kuno, 1986), thus leading to insufficient
exposure to it, particularly for the non-advanced learners. The difficulty with the ni
indirect passive presumably derived from its unique syntactic structure drastically
different from that of the English passive.

Third, a close investigation of the subjects’ judgments of the ungrammatical
sentences revealed the following: 1) for the ni direct passive, as the subjects’ proficiency
levels increased, the ungrammatical sentences (e.g., (69b)) in which 0-role suppression
did not take place were rejected all the more strongly; 2) for the ni yotte passive only the
highly advanced group rejected both types of ungrammatical sentences where either case
absorption or 0-role suppression had not occurred ((70b) for the former and (70c) for the
latter) ; and 3) for the ni indirect passive the unpassivized sentences (e. g., (71b)) were
correctly rejected by the highly advanced group and, to a lesser extent, by the low
advanced group. However, the passivized counterparts (e.g., (71c)) were not rejected by
any NNS group.

Since the highly advanced and the low advanced groups exhibited different

judgment patterns, we discuss their results separately. (The intermediate group never

187

made successful ungrammaticality judgments.) We turn first to the low advanced
group’s judgments. If we look at the types of ungrammatical sentences this group
successfully rejected, we find that they were able to recognize problems with
ungrammatical sentences where 0-role suppression did not take place, namely, those of
the ni direct passive (e. g. (69b) above) and the ni indirect passive without passivization
(e.g., (71b)). On the other hand, when 0-role suppression did take place, they did not
reject such ungrammatical sentences, i.e., those of the ni indirect passive with
passivization (e. g., (71 c)). Therefore, this group seems to have focused on the presence
or absence of 0-role suppression to determine whether a given passive sentence was
ungrammatical or not.

Several findings in FLA research in Japanese lend support to this claim. First,
Hakuta (1982) conducted an experimental study investigating the roles of case marking
and word order in children’s comprehension and production of active as well as passive
sentences. As part of the experiment, children (from 2;3 to 6;2 of agez) performed an act
out activity using toys, based on sentences they heard. He found that all the children
except for those most advanced performed better by approximately at least 15% for the
OSV passive over the SOV passive3—where the OSV passive indicates [agent-ni +
patient-ga + verb] while the SOV passive represents [patient-go + agent-ni + verb]. This
was perhaps due to the fact that the OSV passive places the ni-marked NP in the
sentence-initial position and can thus serve as an early indicator of the passive. (The
word order strategy in which they interpret the NNV sequence as agent-patient-action
irrespective of case markers attached to the two NPs was consistent with the above

finding, but was not with other findings of the study. It was therefore rejected.)

188

Second, in part of an experiment where children at the age of 3;3 to 6;8 were told
to imitate passive sentences given orally, i.e., perform an elicitation imitation task, Sano
(1977, cited in Clancy, 1985) found that they often supplied ni if it was omitted from the
model sentence or substituted it for some other particle, suggesting that they expected ni
to be present in passive sentences.4 Based on these findings, Clancy (1985) suggested
that a critical factor in learning the Japanese passive may be the recognition of the ni-
marked NP as indicating the agent in passive sentences.

In a more recent study Fox & Grodzinsky (1998) shed some insight on the
strategy by which the lower advanced group may have operated in making
grammaticality judgments. They addressed the issue of children’s (about 4 years old)
asymmetrical performance on comprehension of the English passive: namely, that
children fail to correctly comprehend passive sentences involving ‘nonactional’ verbs
while they are capable of understanding those containing ‘actional’ verbs. In their
experiment, children (3;6 to 5;5 of age) heard stories and then were asked to judge the
truthfulness of statements about them such as ‘the rock star is being chased by the koala
bear’. The statements included non-truncated passive sentences with ‘actional’ and
‘nonactional’ verbs as well as truncated passive sentences with ‘nonactional’ verbs. (The
non-truncated and truncated passive refers to a passive with and without the by-phrase,
respectively.) They found, as echoed in the literature, that children were able to tell
whether the statement was true or false when it was given in non-truncated passive
sentences with ‘actional’ verbs, but failed to do so when the statement was given in non-

truncated passive sentences with ‘non-actional’ verbs such as ‘the boy is seen by the

189

horse’. However, they were capable of telling the truthfulness of the statement given in
truncated passive sentences with ‘nonactional’ verbs such as ‘the bear is seen’.

Based on these findings, Fox & Grodzinsky (1998) proposed that these children
performed poorly on non-truncated (i.e., with the by-phrase) passive sentences with
‘nonactional’ verbs because they had not acquired the mechanism to transmit the
suppressed external 0-role to the by-phrase.5 In performing the task, these children did
not ignore the by-phrase, incomprehensible to them, and interpret the entire sentence by
relying on the passive subject and verb. They instead ofien explicitly commented that
they did not know what the test statement meant. Thus it indicates that the processing of
the by-phrase plays an important role in comprehending the English passive. (The
operations of external B-role suppression and transmission to the ni phrase also apply to
Japanese passives. See Marantz, 1984; Sano, Endo, & Yamakoshi, 2001; Washio,
1989/1990.)

All these above findings in FLA research clearly emphasize that the ni/by phrase
in passives serves a critical role in processing passive sentences. Along the lines of Fox
& Grodzinsky (1998), the low advanced group’s grammaticality judgment strategy could
be interpreted as follows: the lack of the 0-role transmitted ni phrase served to determine
the ungrammaticality of the ni direct and indirect passive sentences. There is also
evidence in the present study suggesting that the presence or absence of the G-role
transmitted ni phrase plays an important role in NS grammaticality judgment processes.
Namely, the NS control group judged the ungrammatical ni yotte passive sentences

lacking the 0-role transmitted ni yotte phrase significantly worse than the other ones

190

without case absorption (but with the ni yotte phrase): a mean judgment score of.-1.90
versus -1.34; F(1, 30) = 10.634, p = .003).

As for their judgments of the ni yotte passive, the finding that the low advanced
group did not demonstrate any significant judgment difference between grammatical and
ungrammatical sentences with the mean judgment scores all clustering around zero
(i.e., .17 for the grammatical and .24/.17 for the ungrammatical) strongly suggests that
they had not learned the ni yotte passive at the time of data collection. Perhaps the
morphological difference between ni and ni yotte made it very difficult for them to
recognize the latter as an external O-role assigner/transmitter.

Next, we examine the highly advanced group’s judgments. The subjects of this
group successfully rejected all but the passivized ni indirect passive sentences such as
(710) above. (Indeed, this was one of the two types out of all the syntactic and semantic
sentences under investigation where they showed a clear divergence from the NS control
group.) Judging from their highly successful judgment performances in the syntactic
area, it is reasonable to conclude that they had known passivization operations in
Japanese. But why did they fail to reject the ungrammatical passivized ni indirect passive
sentences (while accepting the grammatical unpassivized ones)?

In order to consider this question, let us first review the structures of the ni direct

and the ni indirect passives. Note the following examples:

(72) a. the m’ direct passive (= (lb)):
J ohn-ga sensee-ni sikar—are-ta.
-NOM teacher-by scold-Pass-Pst
‘John was affected by being scolded by the teacher’

191

b. the ni indirect passive (= (2a)):
J ohn-ga sensee-ni kodomo-o sikar-are-ta

-NOM teacher-by child-ACC scold-Pass-Pst
‘John was affected by the teacher scolding his child’

Hoshi (1994b) characterized the passive verb rare of the ni direct passive as [+
Experiencer, + Passivization] whereas that of the ni indirect passive as [+ Experiencer, -
Passivization]. The former type of rare not only triggers Passivization in the lower
clause (where the internal argument of an embedded verb is posited as PRO), but also
assigns its external Experiencer 0-role to the passive subject; the latter type does not
passivize, but assigns its external 0-role to the passive subject.

Now notice that, as far as the above characterization of rare of the ni direct
passive is concerned, nothing theoretically intrinsic prevents an NP of non-empty
category from appearing as the object of the embedded verb. This situation in effect
yields a passivized ni indirect passive.6 Such structural representation of the ni indirect

passive is given below together with that of the ni direct passive:

(73)
a. the ni indirect passive: [[p John [W kodomo; sensee-ni ti sikar] are ta]]
b. the ni direct passive: [[p John; [W PRO; sensee-ni ti sikar] are ta]] (= (71a))

(The case marking of (73a) above is considered shortly.) Notice that the only structural
difference between (73a) and (73b) is that the lower-clause subject position of (73b) is

filled with PRO, while that of (73a) contains kodomo ‘child’. In other words, there is no

essential structural difference between these two representations. If the internal argument

192

of the lower verb, sikaru ‘scold’, is phonetically unrealized, the ni direct passive sentence
(73b) results, whereas if it is phonetically realized, the ni indirect passive sentence (73a)
obtains. Given that the ni direct passive is learned earlier than the ni indirect passive, it
would be highly plausible that learners of Japanese attempt, at least initially, to extend
their analysis of the ni direct passive to that of the ni indirect passive when learning the
latter. (And this learning strategy appears to work effectively as described above.) That
is to say, they came to hold the view that the ni indirect passive involves passivization
just like the ni direct passive. There are, in fact, analyses of the ni indirect passive in
which this passive takes on passsivization in its lower clause (see Marantz, 1984;
Washio, 1989/ 1990). (See also Hoshi, 1994a for discussion of the relative merits of
Hoshi's, 1994a and Washio's, 1989/ 1990 analyses of Japanese passives.)

If this claim turns out to be on target, the following consequence follows in the
analysis of the ni indirect passive: the subjects of the highly advanced group came to
analyze that accusative case absorption is optional in the ni indirect passive. This is so
because when accusative case absorption applies to the ni indirect passive, we obtain
(74a) below, whereas when it does not, we alternatively get ((72b) repeated here as (74b)

with word order modification for ease of comparison):
(74) a. .John-wa(/-ga) kodomo-ga sensee-ni sikar-are-ta
b. John-ga kodomo-o sensee-ni sikar-are-ta

As noted above, highly advanced subjects accepted the ni indirect passive sentences such
as (74b) and, critically, did not reject ones like (74a). That is, their IL grammar seems to

accommodate both types of the ni indirect passive. Such grammar is possible if it dose

193

not take accusative case absorption as obligatory (under the claim that they analyzed the
ni indirect passive as involving passivization). If they analyzed the ni indirect passive
along these lines, this may be one potential area for fossilization, or it will be very
difficult to eradicate this faulty analysis since it seems that no input readily available to

them allows them to disconfirm the idea of accusative case absorption being optional.

5.4. Non-native speakers’ judgments of perfective versus non-perfective readings
Example sentences of both ta- and iru-marked ni direct and ni yotte passives are
provided in (75) and (76) below. Sentences in (7 5) illustrate the ta-marked versions; i.e.,
(75a) is the ni direct passive with a past reading, (75b) is the ni direct passive with a
perfective reading, (75c) is the ni yotte passive with a past reading, and (75d) is the ni
yotte passive with a perfective reading (recall that only the first type, viz., the ni direct

passive with a non-perfective reading (75a), is ungrammatical, while the others are

grammatical):
(75)
a. (= (41b))
*‘Hamlet’-wa Shakespeare-ni kakeare-ta
‘Hamlet’-TOP Shakespeare-by write-Pass-Pst
‘Hamlet was written by Shakespeare’
b. (= (42b))
kimitu-syorui-ga tekikoku no supai-ni nusum-are-ta

secrest documents-NOM spy from an enemy country-by steal-Pass-Perf
‘secret documents are under the state affected by having been stolen by a spy from an
enemy country’

c. (= (41a))
‘Hamlet’-wa Shakespeare-ni yotte kak-are-ta
‘Hamlet’-TOP Shakespeare-by write-Pass-Pst

‘Hamlet was written by Shakespeare’

194

d. (= (42a))
kimitu-syorui-ga tekikoku no supai-ni yotte nusum-are-ta
secrest documents-NOM spy from an enemy country-by steal-Pass-Perf
‘secret documents have been stolen by a spy from an enemy country’

Similarly, sentences in (76) exemplify the iru-marked versions; viz., (76a) is the ni direct
passive with a progressive reading, (76b) is the ni direct passive with a perfective reading,
(76c) is the ni yotte passive with a progressive reading, and (76d) is the ni yotte passive
with a perfective reading (recall that only the first type, i.e., the ni direct passive with a

non-perfective reading (76a), is ungrammatical, while the others are grammatical):

(76)

a. (= (3 9b))
*atarasii konpyuutaa no puroguramu-ga J ohn-ni tukur-are-te iru
new computer program-NOM -by make-Pass-Prog-Pres
‘a new computer program is being made by John’

b. (= (40b))
intaanetto-wa sekai-zyuu no hito-bito-ni tukaw-are-te iru
Intemet-NOM all over the world people-by use-Pass-Perf-Pres
‘Internet is under the state affected by having been used by people all over the world’

c. (= (3 9a))
atarasii konpyuutaa no puroguramu-ga J ohn-ni yotte tukur-are-te iru
new computer program-NOM -by make-Pass-Prog-Pres
‘a new computer program is being made by John’

(I. (= (40a))
intaanetto-wa sekai-zyuu no hito-bito-ni yotte tukaw-are-te iru
Internet-NOM all over the world people-by use-Pass-Perf-Pres

‘Internet has been used by people all over the world’

Although the highly advanced group on the whole did not reject outright the ni
direct passive with an inanimate subject carrying a non-perfective reading (exemplified
by (75a) and (76a) above) as ungrammatical, the present study found evidence that
subjects of this group had the knowledge that the ni direct passive is not compatible with

a non-perfective reading by virtue of their differential judgments of it from the ni direct

195

passive with a perfective reading ((75b) and (76b)). The low advanced group also
evidenced incipient knowledge of this property of the ta-marked, but not the iru-marked,
ni direct passive, namely, this group made a weak differential judgment between (75a)
and (75b), but not between (76a) and (76b). The intermediate group did not exhibit any
differential judgments of them.

This finding thus has not confirmed the second part of Prediction 5 which stated
that the ni direct passive will not be judged as ungrammatical when read as a non-
perfective. It was hypothesized that since negative input indicating the ungrammaticality
of this type of ni direct passive perhaps has not been readily available to them in a
systematic manner such as in language instruction and daily language use, the subjects of
the present study have not come to know this aspect of the ni direct passive.

If negative input had been indeed unavailable to them, this finding offers strong
evidence that those subjects who based their judgments on the perfective versus non-
perfective distinction had acquired the knowledge that the ni direct passive requires a 0-
subject. This is because with learning-promoting input unavailable, they perhaps
deduced the ungrammaticality of the ni direct passive with a non-perfective reading from
the 0-subj ect status requirement on the ni direct passive along the following lines: the
description of a given situation in a non-perfective (i.e., past or progressive) aspect would
not allow a reading in which an inanimate subject of the ni direct passive is ‘affected’ as
a result of the process or event described in that situation (while such a reading is readily
available in a perfective description). With an ‘affective’ effect required on a ni direct
passive subject, the former description is not permissible. It would be fruitful to further

investigate more precisely the nature of the knowledge of the perfective versus non-

196

perfective distinction in the ni direct passive which English-speaking learners have
demonstrated to be capable of learning (see below for a more directed future research
suggestion).

Another noteworthy finding (which is presupposed in the first one) is that these
two groups of advanced subjects knew that the ni direct passive may take an inanimate
subject notwithstanding the predominant use of an animate subject in this type of passive.
They in fact accepted as grammatical the inanimate-subject ni direct passive with a
perfective reading to the same extent as they did the one with an animate subject used for
the syntactic analysis (F(2, 48) = .129, p = .782 for the highly advanced group; (F (2, 70)
= .957, p = .389 for the low advanced group). Perhaps the fact that an inanimate subject
can freely appear in the English passive facilitated their learning of this aspect of the ni
direct passive.

In contrast to the differential acceptability of the ni direct passive between
perfective and non-perfective readings, the lack of such a distinction in the ni yotte
passive was also observed in the highly advanced group (see (75c)-(75d) and (76c)-(76d)
for the absence of grammaticality contrast). The subjects of this group rated the ni yotte
passive with perfective and non-perfective readings without much difference. This
suggests their knowledge that the ni yotte passive does not impose such restriction on the
subject status as the ni direct passive does. The low advanced and the intermediate
groups generally rated ni yotte passive sentences very low elsewhere, suggesting the lack
of syntactic knowledge of this type of passive.

Therefore, in sum, the findings regarding the contrastive grammaticality

judgments between the two types of passives made by the highly advanced group

197

strongly indicate that these subjects possessed the knowledge that the subject status
differs between the ni direct and the ni yotte passive, i.e., the former takes a O-passive
subject whereas the latter a non-9 one. Neither the low advanced nor the intermediate
group demonstrated evidence for such knowledge (although the former group showed the
presence of incipient knowledge of the perfective versus non-perfective reading contrast
in the ta-marked ni direct passive).

There is an additional recurrent pattern observed in the present study deserving
consideration. There were cases where ta-marked passive sentences, ni direct or ni yotte,
were judged in accord with the theory predictions by one group which in turn failed to do
so with their iru-marked counterparts. In other words, there were instances implying that
certain ta-marked passive sentences were learned earlier than their iru-marked
counterparts. (The to form may mark either a perfective or simple past reading whereas
the iru form either a perfective or progressive reading.) Such cases include: 1) the low
advanced group judged the ta-marked ni direct passive with a past reading (e.g., (75a))
differently from the one with a perfective reading (e. g., (75b)), but failed to differentiate
its iru-marked counterparts (e. g., (76a) and (76b)); and 2) the highly advanced group’s
rating of the ta-marked ni yotte passive with a perfective reading (e. g. (7 5d)) was as high
as that of the ni direct passive counterpart (e. g. (75b)) whereas their judgment of the iru-
marked ni yotte passive version (e.g., (76d)) was not as high as that of its ni direct
counterpart (e. g., (76b)). In other words, assuming that the ni direct passive was learned
earlier than the ni yotte passive, the highly advanced group approximated the judgment of

the ta-marked ni yotte passive with a perfective reading (e. g., (75d)) with that of its ni

198

 

 

 

direct passive counterpart (e. g., (7 5b)), but failed to do so with the iru-marked (ni yotte
and ni direct passive) version (e.g., (76d) and (76b)).

Why was this so? To date no semantic analysis of Japanese passives provides an
answer to the question of how aspectual differences between to and iru forms may
interact with the ni direct and the ni yotte passives (see Alfonso, 1980; Inoue, 1976;
Jacobsen, 1992; Klaiman, 1987; Kuroda, 1979; Masuoka, 1982; Wierzbicka, 1979 for
analyses of semantics of the ni direct passive versus the ni yotte passive). We may,
however, locate one possible factor influencing the observed learning differences
between ta- and iru-marked passives in input frequency differences in passive sentences.
Specifically, perfective ta-marked passive sentences, either the ni direct passive or the ni
yotte passive, may appear more frequently than their perfective iru-marked counterparts.

One factor which may be contributing to the input frequency differences is the
verb-type requirement on the perfective reading of the iru form. Although the precise
conditions necessary for the perfective iru form to be met is still debated (see Inoue,
1976; Jacobsen, 1992; Okuda, 1978a; 1978b; Shirai, 1998; 2000; Yoshimoto, 2000),
there is general agreement that the iru form will be assigned a perfective reading when
used with achievement and accomplishment verbs (J acobsen, 1992). Stated differently, it
will not be read as a perfective with activity verbs, in contrast to the to form whose
perfective reading is not restricted in any sense.

Hence, the occurrence of perfective iru-marked passives, both ni direct and ni
yotte, is disfavored over the ta-marked counterparts since a limited range of verbs is
available for an iru perfective reading, while any verb (except for stative verbs) is

available for a to perfective reading. That is to say, passive sentences encountered by

199

advanced subjects were more likely marked with the perfective to form than with the
perfective iru form. This skewed input frequency difference is perhaps partially
responsible for the findings that the low advanced group exhibited a weak tendency for
judgment differentiation in the ta-marked ni direct passive between perfective and past
readings, but did not demonstrate it in the iru-marked ni direct passive version; and that
the highly advanced group accepted the ta-marked ni yotte passive with a perfective
reading as strongly as its ni passive counterpart, but did not do so with the perfective iru-
marked version. This apparent learning difference between ta- and iru-marked (ni direct

and ni yotte) passives merits further research.

5.5. Non-native speakers’ judgments of adversative readings

Examples of ni— and ni yotte-marked direct and indirect passive sentences with
adversative as well as non-adversative readings are given in (77) and (78) below.
Sentences in (77) illustrate the indirect passive versions, i.e., (77a) is the ni indirect
passive carrying an adversative reading, (77b) carries a non-adversative reading (thus,
ungrammatical), and (77c) is the ni yotte indirect passive with an adversative reading

(thus, ungrammatical):

(77) a. J ohn-ga warui toki-ni tomodati-ni ko-rare-ta
(= (3 7)) -NOM at an inconvenient time friend-by come-Pass-Pst
‘John was adversely affected by his friend visiting him at an inconvenient time’

b. *John-ga ii told-ni tomodati-ni ko-rare-ta
at a convenient time

""John was adversely affected by his friend visiting him at a convenient time’
(Kuroda 1979, p.314 & p.317)

200

c. (= (43b))
*John-ga warui toki-ni tomodai-ni yotte ko-rare-ta
-by
‘John was adversely affected by his friend visiting him at an inconvenient time’
Sentences in (78) exemplify the direct passive versions, viz., (78a) is the ni direct passive

carrying an adversative reading, (78b) carries a non-adversative reading (thus,

ungrammatical), and (78c) is the ni yotte direct passive with an adversative reading (thus,

ungrammatical):
(73)
a. (= (3 83))
J ane-ga mukasi no kare-ni ni-zikan mo mat-aretc komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst

‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

b. (= (3 8b))
* J ane-ga mukasi no kare-ni ni-zikan mo mat-arete uresikat-ta
happy-Pst

”Jane was happy by being adversely affected by having been waited for by her ex-boyfriend
for as many as two hours’

c. (= (43d))
*Jane-ga mukasi no kare-ni yotte ni-zikan mo mat-arete komat-ta
.by
‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

In examining the subjects’ grammaticality judgments of the ni direct and the ni
indirect passives carrying adversative and non-adversative readings, as well as their
acceptability of the ni yotte direct and indirect passives with an adversative reading, it has
been found that the highly advanced group observed both that the ni (direct and indirect)

passive carries an adversative reading (by making differential judgments between

(7Sa)/(77a) and (78b)/(77b) above), and that the ni yotte passive is not compatible with

201

such a reading (see (77c)). There was one exception to this general finding, i.e., that the
ni yotte direct passive with an adversative reading (e. g., (78c)) was not rejected outright
(see below for a possible reason). The subjects of the low advanced group showed
evidence of knowledge that the ni direct passive tends to carry an adversative reading (by
making differential judgments between (78a) and (78b)), but they failed to demonstrate
the other relevant distinctions in judgment. This perhaps does not indicate that they had
come to have the knowledge that the ni direct passive carries an adversative reading
while the ni indirect passive does not, but simply suggests that they did not have
sufficient syntactic and semantic knowledge of the latter. In the similar vein, their failure
to reject the ni yotte-marked (direct and indirect) passive with an adversative reading
merely indicates that they had learned very little (at best) about the properties of the ni
yotte marking. The intermediate group did not indicate differential judgments of any
kind.

It is noteworthy that the highly advanced group rated the ni indirect passive with
an adversative reading as high as it did the ni direct passive with such reading. This is so
because in the syntactic analyses (where the ni passives did not carry enhanced
adversative readings), the subjects of this group judged the ni indirect passive
significantly lower than the ni direct one. Thus, a pronounced adversative reading in the
ni indirect passive served to promote their rating of it, suggesting that they had
established a clear association between the ni indirect passive and an adversative reading
just as NSs do (see Specs, 1992 for N83’ rating of adversity in the nz’ indirect and the ni

direct passives).

202

This evinced connection between the ni indirect passive and an adversative
reading sheds some light on the question of why the subjects of the highly advanced
group did not reject more clearly the ni yotte direct passive with an adversative reading.
Namely, why is it that they did not rate ni yotte direct passive sentences with an
adversative reading as ungrammatical strongly enough so that there would have been a
significant judgment difference fiom grammatical ni yotte passive sentences without an
adversative reading?

One possible reason for the lack of the highly advanced group’s clear rejection
seems to have to do with the nature of the verb used in the test sentences (79) below
((79a) exemplifies the ni yotte direct passive with an adversative reading, and (79b) the ni

direct passive with a non-adversative reading):

(79)
a. (= (43d))
*Jane-ga mukasi no kare-ni yotte ni-zikan mo mat-arete komat-ta
-NOM ex-boyfriend-by as many as 2 hours wait-Pass annoy-Pst
‘Jane was annoyed by being adversely affected by having been waited for by her ex-
boyfriend for as many as two hours’

b. (= (3 8b))
*Jane-ga mukasi no kare-ni ni-zikan mo mat-arete uresikat-ta
happy-Pst
”Jane was glad by being adversely affected by having been waited for by her ex-boyfiiend
for as many as two hours’

Clearly, the verb matu ‘wait’ whose passivized form is used in sentence (79a) does not
intrinsically carry a negative connotation. Thus, those subjects of the highly advanced
group who failed to rate it as ungrammatical may have felt that the passivized verb mat-
are-te ‘being waited’ does not carry a negative connotation. The presence of the ni yotte

phrase in it is consistent with this interpretation, possibly even enforcing it. Thus, they

203

might have felt that the passive subordinate clause was an impersonal, report-like
description of the situation and accepted it. In this view the overall negative connotation
in (79a), derives fiom the matrix verb komat-ta ‘was annoyed’, but not from the
passivized verb mat—are-te ‘having been waited’. If so, the problem with their judgments
resides in their failure to associate the passivized verb whose original form carries a
neutral connotation with an adversative reading just as NSs may do (Spees, 1992).

By contrast, the highly advanced group successfully rejected the ni direct passive
with a non-adversative reading (79b) above (with a mean judgment score of -.76). That
is to say, they derived a negative reading fi'om its passive subordinate clause, which is
incompatible with the overall positive connotation expressed by the matrix verb
uresikatta ‘was glad’. Why did they read the negative connotation in the passive
subordinate clause of (79b) but not in that of (79a)? This is perhaps because the former
is marked by ni while the latter by ni yotte. Their remarkably good metalinguistic
performances on perfective versus non-perfective reading contrasts and verb-induced
viewpoint differences strongly suggest that they knew that there is an essential semantic
difference indicated by ni and ni yotte markings in passive sentences.

It has been claimed above that the ni phrase served a critical role for subjects of
the low advanced group to determine the ungrammatical status of passive sentences in
syntactic terms. It is not unwarranted to go a step further and hypothesize that the ni
phrase also plays an important role in interpretation of the semantics of passive sentences
in that the ni marking, not the passive verb rare, is associated with an adversative
reading. In consideration of the possible role either a verb connotation or the ni marking,

or both, may play in interpretation of the ni direct passive, then, an interesting question

204

arises as to whether learners of Japanese at such a high proficiency level as the highly
advanced group would be able to read an adversative connotation in the ni direct passive
where the ni phrase is missing and the passivized verb is derived from a neutral-
connotation verb. If indeed the subjects of the highly advanced group relied on the ni
marking to read an adversative connotation in the ni direct passive, they may not be able
to read it in a ni direct passive sentence where the ni marking not only is unavailable, but
its passivized verb in itself does not connote negativity, i.e., the passive verb rare is the
only one for an adversative reading. This path of inquiry may well illuminate further the
remarkably good knowledge of Japanese passives demonstrated by the highly advanced
group. (Recall also that Tanaka (1993) and Watabe et al. (1991) found that English-
speaking learners of Japanese often failed to employ the ni direct passive in production
tasks to express an unfavorable situation for a referent of the passive subject where a NS

of Japanese would make use of the ni direct passive.)

5.6. Native speaker control group’s non-rejection of the ni yotte indirect passive
with a transitive verb carrying an adversative reading

It was found that the subjects of the NS control group made a judgment
distinction between ni and ni yotte indirect passive sentences with transitive verbs
carrying adversative readings by giving a significant judgment difference between them.
Examples of both types are given below; (80a) is the ni indirect passive (grammatical),

and (80b) is the ni yotte indirect passive (ungrammatical):

205

(80)

a.

huyu no samui hi no asa, Mary-wa J ane-ni sinsitu no mado-o ake-rare-ta

in the cold winter morning -NOM —by bedroom window-ACC open-Pass-Pst
b.

*huyu no samui hi no asa, Mary-wa J ane-ni yotte sinsitu no mado-o ake-rare-ta

‘Mary was adversely affected by Jane opening iii: bedroom window in the cold winter
mornrng’
However, contrary to the theory prediction, they barely succeeded in rating the latter type
(e. g., (80b)) as ungrammatical (with a mean judgment score of .26)7, whereas they
accepted the former type (e.g., (80a)) clearly (with a mean judgment score of 1.50) as
predicted. Thus the question arises as to why the NS control group subjects did not reject

the ni yotte indirect passive sentences with transitive verbs?

In addressing this question, first, consider the following ni yotte passive sentence:

(81) J ohn-ga kolcumu-syoo ni yotte ryoken-o toriage-rare-ta
-NOM State Department by passport-ACC take away-Pass-Pst
‘John had his passport revoked by the State Department’ (Kuroda, 1979, p. 339)

Kuroda (1979) gave ni yotte passive sentences like (81) above as an example of a
potential problem with his claim that the ni passive versus the ni yotte passive
classification is an overarching dichotomous characterization of Japanese passives.
Sentence (81) is syntactically an indirect passive because it contains an accusative case-
marked NP and a passive morpheme rare along with a so called extra NP as the passive
subject. However, its agentive NP kokumu-syoo ‘the State Department’ is marked with ni
yotte not ni. Given that the ni yotte indirect passive cannot be a mere variant of the ni
indirect passive, how is a ni yotte indirect passive such as (81) possibly derived since

there does not seem to be any active counterpart of (81) available?

206

Following Shibatani’s (1990) proposal that Japanese underlyingly has a very wide
range of double/multiple accusative structures as in Korean (Yoon, 1990, cited in
Shibatani, 1990); and further hypothesizing that “predicates in Japanese can take an
“additional” object/“affected” argument, typically a theme or patient ....” (Hoshi 1999, p.

219), Hoshi (1999) proposed the underlying structure of sentence (81) as below:

(82) *kokumu-syoo-ga John-(o) ryoken-(o) toriage-ta
Sate Dept.-NOM -(ACC) passport-(ACC) take away-Pst
‘the State Department revoked John’s passport’

(Note that sentences such as (82) never surface due to the Double-o Constraints, indicated
by *-marking.) Now, sentence (81) is derived by John being moved to the subject
position in the ni yotte passivization process, i.e., NP movementg, and by kokumu-syoo
‘the State Department’ being ni yotte-marked due to O-role suppression. Hence, given
Hoshi’s (1999) movement analysis of the ni yotte (indirect) passive like (81), the
syntactic difference between the ni passive and the ni yotte passive is maintained.

Yet there seems to be some semantic consequence in this analysis. Namely, the
semantic difference between the ni passive ‘affectee’ subject and the ni yotte passive non-
‘affectee’ one seems to be weakened. This is so because the subject of the ni indirect
passive is required to be ‘affectee’ by the passive morpheme rare whereas in the
proposed analysis the subject of the ni yotte indirect passive is to be interpreted as an
“affected” argument.10

The NS control group subjects’ failure to clearly reject the ni yotte indirect
passive like (80b) as ungrammatical may be partly attributable to this reduced semantic

difference between the ni yotte passive and the ni indirect passive with a transitive verb.

207

Specifically, if they did not read the overtone of (80b) as strongly adversative (invoked
by the passive verb rare), they may not have rejected it as an instance of the ni yotte
indirect passive, but instead perceived its overall weak adversative connotation and the

‘6

hint of affectedness given by a ni yotte passive subject’s affected” role as compatible.
On the other hand, if they felt (80b) to be exceedingly adversative, they may have
rejected it on the basis of the general incompatibility between the ni yotte indirect passive
and an adversative reading.

Interestingly enough, in their judgment of the ni yotte indirect passive and the ni
indirect passive with an intransitive verb carrying an adversative reading, NS control
group subjects clearly rejected the former (with a mean judgment score of -1.63) and
accepted the latter (at a mean judgment score of 1.23). Along the lines considered above,
this was so because intransitive verbs would not allow an “additional” object/“affected”
argument even underlyingly and thus the clear semantic division between the ni passive

and the ni yotte passive was intact. In other words, the alternative reading of the (ni)

indirect passive as the ni yotte passive was not available.

5.7. Judgments of verb-induced viewpoint differences

There were two types of verb-induced viewpoint differences resulting in the
grammaticality distinction between the ni direct and the ni yotte passive: i.e., 1) the
Japanese native versus Sino-Japanese verb contrast and 2) the affective and unaccusative
verb contrast. Example sentences for the first type are provided in (83) below, i.e., (83a)

is the ni direct passive with a Japanese native verb, (83b) is the ni yotte passive with a

208

 

Sino-Japanese verb, and (83c) is the ni yotte passive with a Japanese native verb (thus

ungrammatical):

(83) a- (= (443))

Bill-wa kawaii musuko-ni koros-are-ta
-TOP beloved son-by kill-Pass-Pst
‘Bill was affected by being killed by his own beloved son’
b- (= (440))
Bill-wa CIA-ni yotte satugai-s-are-ta
-TOP the CIA-by murder-Pass-Pst
‘Bill was murdered by the CIA’
c. (= (44b)
*Bill-wa kawaii musuko-ni yotte koros-are-ta
-TOP beloved son-by kill-Pass-Pst

‘Bill was killed by his own beloved son’ *

Not only the highly advanced group but also the low advanced group judged as
grammatical both the ni direct passive with a Japanese native verb (e. g., (83a))
characterizing a personal viewpoint and the ni yotte passive with a Sine-Japanese one
(e.g., (83b)) signalizing an objective standpoint. Furthermore, both groups rejected the ni
yotte passive with a Japanese native verb (e. g., (83c)), indicating that they knew that the
ni yotte passive is suitable for an objective description but not for a personal one. The
intermediate group failed to make any distinction set forth by the different verb types.

This is the only distinction between the ni direct and the ni yotte passive that the
low advanced group successfully made. This group, unlike the highly advanced group,
failed to give differential judgments to these two types of passives in the other semantic
areas investigated. As noted earlier, their failure to demonstrate such contrasts was, at
least, partly due to the slight syntactic knowledge the subjects of this group had attained

of the ni yotte passive. Why then were they successful in making a grammaticality

209

 

distinction between ni yotte passive sentences with a Sino-Japanese verb and those with a
native Japanese verb? Namely, why were they able to accept the former and reject the
latter despite their apparent lack of syntactic knowledge of the ni yotte passive?

As held in the Predictions, this verb type-related distinction between the ni direct
and the ni yotte passives is often singled out and characterized as being a stylistic
difference between the two; namely, the ni yotte passive is preferred to the ni direct
passive in formal writing such as academic reporting (Howard & Niyekawa-Howard,
197 6; J acobsen, 1992). Perhaps this is the type of knowledge developed by the low
advanced group, which does not touch on the crucial semantic property of the ni yotte
passive as having a non-G-subj ect.

Essentially, the subjects of this group may have developed a lexicalized
knowledge of the ni yotte marking, such that a ni yotte marking in Japanese passives is
associated with formal writing effected by Sino-Japanese verbs, but is incompatible with
a non-formal situation set forth by Japanese native verbs. This hypothesis then suggests
that they analyzed a ni yotte marking as a mere lexical variant of a ni marking of the ni
direct passive, considering that they evinced (syntactic) knowledge of the ni direct
passive, but not that of the ni yotte passive: More specifically, they may have perceived
this type of ni yotte passive as a ni yotte-marked direct passive without conceiving of the
full semantic consequences for matching the ni yotte marking with the ni direct passive.
Consequently, as a variant of the ni direct passive, they were able to make differential

judgments of these ni yotte passive sentences.

210

 

Next, example sentences for the second type of verb-induced viewpoint
differences are given in (84) below: viz., (84a) is the ni direct passive with an affective

verb, and (84b) is the ni yotte passive with an affective verb (thus ungrammatical):

(84)
a. (= (45a))
Mary to hanasi-te iru tokoro-o, Bill-ga gaaruhurendo-ni kik-are-ta
talking to Mary as -NOM his girlfriend-by hear-Pass-Pst
‘Bill was affected by being heard by his girlfriend as he was talking to Mary’
b. ( =(45b))
*Mary to hanasi-te iru tokoro-o, Bill-ga gaaruhurendo-ni yotte kik-are-ta

.by

In judging these sentences, the highly advanced group made a differential judgment
between the two by accepting the former and rejecting the latter, whereas the low
advanced and intermediate groups did not demonstrate such contrast. Surprisingly,
however, in rating the ni direct passive sentences with affective verbs such as (84a)
above, NS control subjects gave them either a very low grammaticality or
ungrammaticality status, contrary to Kuroda's (1979) theory according to which these
sentences are rendered grammatical.

In consideration of this unexpected assessment given by the NS control group, see

the following sentences (85) (adopted from Kuroda, 1979, p. 314):

(85)
a. Bill-ga nozokimi site iru tokoro-o John-ni mi-rare-ta
-NOM peeping (into a room) as -by see-Pass-Pst
‘Bill was affected by being seen by John as he was peeping (into a room)’
b. A! John-ni mi-rare-ta

My God, -by see-Pass-Pst
‘My God, I’ve been seen by John’

211

 

Kuroda (197 9) delved in great detail into how the subject of a ni direct passive with an
affective verb would feel psychologically affected by the event described by the passive
sentence. In doing so he started out with the ni direct passive sentence (85b) stated from
the first person’s viewpoint, since the psychological experience depicted in this type of
passive sentence tends to be highly personal. In (85b) it conveys a kind of self-shame
caused by “being-seen-by-the-Other” (Kuroda, 1979, p. 314) because one has been
secretly peeping at something forbidden. Then he moved on to a more general situation
like (85a) referring to a third person’s experience. Because of this elaborated transition
in discoursal viewpoint from first to third person, the reader of Kuroda’s paper perhaps is
able to adjust his or her viewpoint to the third person’s in sentence (85a) and perceive it
as a quasi-personal experience invoking highly personal psychological affectivity.

In the present study, on the other hand, (85a) above was merely one of the 84 test
sentences to judge; the subjects were presented with it plainly without any accompanying
discourse information with which they could establish a close identity with the passive
subject. (See also endnote 4 in this chapter for related consideration.) This possibly led
to a conflict in the reading: namely, on the one hand, it expressed a very personal
psychological state (conveyed by the passivized affective verb mir-are-ta ‘was seen’),
and, on the other, it appeared to describe a personally detached situation where the judge
of the sentence had no identity or relation to the passive subject.

This view receives some support from the finding that linguistically sophisticated
NSs judged these sentences better than linguistically naive NSs: the former group judged
them with mean judgment scores of .71 and 1.19, whereas the latter rated them with .58

and -.42, respectively. This was probably because the linguistically trained group was

212

 

 

more readily able to adjust their viewpoint to a different perspective due to their practice
and ability to analyze sentences in various ways.ll

Furthermore, it may be explained in a similar line why the highly advanced group
performed better on average in judging these ni direct passive sentences with affective
verbs than the linguistically naive NS group (the mean judgment scores of .864 versus
.08065, respectively). If the former group subjects lacked the knowledge that such a
subtle psychological connotation may be invoked by passivized affective verbs,12 then,
they perhaps were blind or insensitive to the potential conflict in viewpoint the latter
group may have experienced. This may have led the highly advanced group to construe
the ni direct passive sentences with affective verbs as typical ni direct passive sentences
(in which a deep, highly personal psychological effect is not generally read),
consequently resulting in their superficially better performance without suffering the
viewpoint conflict described above. (See the following for discussions of the role of
discoursal information on the use of passives: Watabe, Brown, & Ueta, 1991 for SLA;
Otsu, 1999; Suzuki, 1998 for FLA; and Ferreira, 1994 for adult NSs.)

In the end, it was found that the NS control group rejected the unaccusative
versions corresponding to the ni direct passive sentences with affective verbs such as
(84a) despite Kuroda's (1979) prediction that they are grammatical. An example sentence

of its unaccusative version is exemplified below:

(86) gaaruhurendo-ni, Bill-ga Mary to hanasi-te iru tokoro-ga kikoe-ta
his girlfriend-to -NOM talking to Mary as audible-Pst
‘Bill was audible to his girlfriend as he was talking to Mary’

213

 

 

They gave mean judgment scores of -.23 and -1.39 to these unaccusative sentences. Why
was this so? Indeed, a similar unexpected result was obtained in Hirakawa (1999)13 in
which she investigated whether English- and Chinese-speaking learners of Japanese have
difficulty acquiring surface unaccusativity in J apanese14 using a grammaticality judgment
task (see also Hirakawa, 2001). It was surprisingly found that Japanese controls did not
show the unaccusative and unergative distinction with respect to a case drop
phenomenon15 . Namely, they treated unaccusative verbs on a par with unergative ones,
rejecting unaccusative as well as unergative sentences with a nominative case marker ga
dropped. (It was expected that unergative sentences with a nominative case marker
dropped would be rejected, while surface unaccusative counterparts would be accepted
since its sole NP remains in the object position and thus satisfies the case marker drop
conditions.) Thus both studies looking at unaccusativity in Japanese by employing a
grammaticality judgment task obtained results incongruent to theory predictions. More

research, both empirical and theoretical, is necessary to find out why this was the case.

214

 

 

 

CHAPTER 6

CONCLUSION

The goal of the present study has been to examine knowledge of passives that
English-speaking learners of Japanese develop in order to address the question of what
aspects of Japanese passives are difficult for them to learn both in syntactic and semantic
areas, and why. Toward this end, based on Hoshi (1994a; 1999) and Kuroda's (1979)
analysis of Japanese passives, a grammaticality judgment questionnaire was devised in
which there were 62 passive-related test sentences as well as 22 distractor sentences to
rate on a five-point grammaticality scale. Data were gathered both from linguistically
sophisticated and naive NSs of Japanese as well as from intermediate to low and highly
advanced English-speaking learners of Japanese.

To begin with, we examined judgment data collected from ordinary (characterized
as linguistically naive) NSs of Japanese (viz. 31 subjects without linguistic training).
This analysis was particularly critical to demonstrate that they indeed judged the passive
test sentences according to Hoshi (1994a; 1999) and Kuroda's (1979) predictions with
keen attention to their judgments of sentences differing in delicate semantic nuances.
This was so because Kuroda’s semantic analysis of Japanese passives hinges on highly
subtle differences in closely related sentences. In this connection, judgment data were
also gathered fi'om linguistically sophisticated NSs (viz. 21 subjects with linguistic
training). This additional data collection was motivated by the suspicion that there might
be differences in assessing the grammaticality of passive sentences in which semantic

features were manipulated, depending on whether or not NS judges had received

215

linguistic training. (Note that the literature has been inconclusive on the effects of
linguistic training for N83’ grammaticality judgment performances.)

It was found that the linguistically naive NS group gave grammaticality
judgments of the passive sentences conforming more than 90% to the theory predictions,
while the linguistically sophisticated NS group did so approximately 45% to 70% of the
time depending on measurement. Therefore, the ordinary NSs gave grammaticality
judgments in close conformity to Hoshi (1994a; 1999) and Kuroda's (197 9) analysis of
Japanese passives including passive sentences in which subtle semantic features were
manipulated: Additionally, in j udgrnent of the questionnaire sentences of the present
study, these two groups behaved quite differently (see Predictions la and lb). Having
established the linguistically naive NS group’s very high correspondences to the theory
predictions, we proceeded to investigate the knowledge English-speaking learners of
Japanese had attained of the properties of Japanese passives.

In the second phase of the study, we analyzed grammaticality judgment data
gathered from 20 intermediate, 36 low advanced, and 25 highly advanced English-
speaking learners of Japanese to examine their syntactic and semantic knowledge of
Japanese passives. First of all, the widely held view that Japanese passives are
exceedingly difficult for English-speaking learners to learn received strong supportive
evidence in the present study. Despite their studying Japanese for an average of 4.32
years and 70% of them majoring in J apanese—related specializations (which normally
suggests their commitment to learning the language), the subjects of the intermediate
group failed to demonstrate their knowledge of Japanese passives in any syntactic or

semantic terms. Namely, in rating grammatical sentences of the ni yotte, the ni direct,

216

 

and the ni indirect passive, they did not judge any one type significantly higher than the
other two; nor did they indicate any sensitivity to grammaticality contrasts reflected in
the judgment of contrastive pairs of grammatical and ungrammatical passive sentences
for either a syntactic or a semantic property. Their mean judgment scores of all of them
stayed very close to zero.

On the other hand, the low and highly advanced groups’ grammaticality judgment
data provided the information to unravel how Japanese passives were represented in their
knowledge and thus what aspects of passives posed learning difficulties for them, and
why. First, an investigation of their syntactic knowledge of passives, particularly that of
passivization, showed that the grammatical ni direct passive was rated significantly
higher than both its ni indirect and ni yotte passive counterparts, while there was no
significant difference in rating the latter two types of passives. When compared with the
NS control group’s judgments, the low advanced as well as the highly advanced groups
approximated to it in judging the ni direct passive; and in rating the ni yotte passive, the
latter group approximated to it, but the former group did not. In judging the ni indirect
passive, in contrast, neither group approximated the NS control group. Therefore,
clearly, the grammatical ni direct passive was rated best of all. On the other hand, there
was some evidence suggesting that the grammatical ni yotte passive appears to have been
more accurately judged than the grammatical ni indirect passive (see Predictions 2 and 3).

Second, in rating the syntactically ungrammatical ni direct passive, both the low
advanced and the highly advanced groups made a clear-cut grammaticality distinction
between the grammatical and ungrammatical ni direct passive sentences. In the case of

the ni yotte passive, the latter group demonstrated a clear grammaticality distinction, but

217

 

 

the former group did not. Finally, both groups failed to reject the ungrammatical ni
indirect passive sentences where passivization operations applied, while they successfully
differentiated from the grammatical ones the other type of ungrammatical ni indirect
sentences where passivization operations did not apply. Hence, as with the judgments of
the grammatical sentences of the three types of passives, clearly, the ungrammatical ni
direct passive was rated best of all. Although less clear, it appears, overall, that the
ungrammatical ni yotte passive was rated more correctly as ungrammatical than the
ungrammatical ni indirect passive (see Predictions 2 and 3).

F urtherrnore, a close examination of the low advanced group’s judgment patterns
in rating all ungrammatical sentences of the three passive types indicated that the subjects
of this group had determined the ungrammaticality of them on the basis of the absence of
the external 0-role transmitted ni phrase (see Fox & Grodzinsky, 1998; Sano, Endo, &
Yamakoshi, 2001 for discussion of the role of the external 0-role transmitted phrase for
processing passive sentences). This was perhaps why this group was spuriously
successful in rejecting the ungrammatical ni indirect passive where passivization
operations, particularly external 0-role suppression, did not take place, because the lack
of external O-role suppression resulted in the absence of the 0-role transmitted ni phrase
in the ni indirect passive. It was also claimed in consideration of the learning difficulty
with the ni yotte passive that the morphological difference between ni and ni yotte had
made it very difficult for the subjects of this group to recognize the ni yotte phrase in this
type of passive as a G-role transmitted phrase.

In contrast, the subjects of the highly advanced group demonstrated very good

knowledge of passivization. Their only grammaticality judgment which clearly deviated

218

 

from the NS control group was that of the ungrammatical ni indirect passive where
passivization operations took place. In consideration of their remarkably good
performances in judgment of passive sentences where passivization processes were
manipulated, their failure to reject the passivized ni indirect passive indicated that they
had analyzed the ni indirect passive as triggering passivization in its lower clause (see
Washio, 1989/1990 for this line of analysis). This firrther suggested that they analyzed
this type of passive as involving the optionality of accusative case absorption in its lower
clause passivization. This is so because their IL grammar accommodated both
(grammatical) unpassivized and (ungrammatical) passivized ni indirect passive sentences.
Crucially, they only differ with respect to the presence or absence of an accusative o-
marked NP in the lower clause as a result of accusative case absorption not being applied
for the former and applied for the latter. If this reasoning is correct, the optionality
analysis above would be highly difficult for them to releam due to the unavailability of
input contradicting such analysis.

On the whole, as summarized above, the findings of the present study show that
the NNS subjects (both low and highly advanced groups) classified the three syntactic
types of Japanese passives in a single class in which they are characterized as involving
passivization operations. Such syntactic characterization overlaps with that of the NS
subjects with respect to the ni yotte and the ni direct passives, since these two types of
passives share the passivization processes as a syntactic function of their respective
passive morphology rare: However, the NNS subjects’ knowledge representation of the
ni indirect passive departs from that held by the NS subjects who seem to have conceived

of it as not triggering the passivization processes. It was hypothesized above that the

219

subjects of the highly advanced group had accommodated, in their IL grammar, two types
of the ni indirect passive in which accusative case absorption either takes place or does
not, by analyzing this type of passive as optionally triggering it.

This non-correspondence between the NS and the highly advanced NNS subjects’
knowledge representations with respect to the ni indirect passive, however, does not lead
to the conclusion that the latter group constructed an illicit grammar of this type of
passive. Indeed, as shown in Baker et al. (1989), there are languages in which the
thematic object of a passivized verb may appear in either a nominative case form or an
accusative case form, that is to say, accusative case absorption either applies or does not.

Such a case is illustrated below in Ulaainian (Sobin, 1985, cited in Baker et al., 1989, p.

236):
(87) a. cerkv-a bul-a zbudova-n-a v 1640 roc’i
church-NOM/Fem was-Fem built-Pass-Fem in 1640
b. cerkv-u bul-o zbudova-n-o v 1640 roc’i

church-ACC/Fem was-Imp built-Pass-Imp in 1640
‘the church was built in 1640’

In (87a) accusative case absorption has taken place and thus cerkv-a ‘church’, the internal
argument of the verb, is nominative case-marked: In (87b), on the other hand, accusative
case absorption has not applied and, consequently, the thematic object of the verb retains
its accusative case. Interesting in this respect is that optionality of accusative case
absorption does not exist either in Japanese or English: Nevertheless, the subjects of this
group appear to have entertained this line of analysis which is permissible in grammars of
other languages. Thus a natural question arises as to how they arrived at such analysis.

Unavailability of input data in both TL and NL of the NNS subjects strongly suggests

220

 

that they attained it learner-internally, i.e., fi'om properties of Universal Grammar. If so,
this provides an illuminating case in which adult SL learners construct a grammar on
their own, one that is different from that of their TL and NL, yet which appears to be
internally consistent by virtue of being driven by TL input data, a knowledge of NL, and
Universal Grammar.

Turning to the investigations of their semantic knowledge of Japanese passives, of
particular interest was whether or not the NNS subjects’ groups observed the ‘affectee’ 0-
subject versus non-O-subj ect contrasts between the ni passives and the ni yotte passive
which were brought out in various semantic conditions. First, the subjects of the highly
advanced group evinced that they knew that the ni passive, both direct and indirect, might
be read with an adversative connotation, but not with a non-adversative one. Namely,
they made differential judgments of the ni direct as well as the ni indirect passive
depending on its adversative and non-adversative readings. There was further evidence
that they knew that there was a propensity for the ni indirect passive to carry an
adversative connotation in that their judgments of the ni indirect passive improved
considerably when matched with a pronounced adversative reading. No such judgment
improvement occasioned in the case of the ni direct passive with an adversative reading
(see Prediction 4). The subjects of the low advanced group made differential judgments
between the two readings only in the ni direct passive. (It was suggested that they had
not yet learned syntactic and semantic properties of the ni indirect passive.)

Second, in judging the ni direct and the ni yotte passive with either a perfective or
a non-perfective reading, the subjects of the highly advanced group demonstrated the

knowledge that the ni direct passive sentences with an inanimate passive subject must be

221

read with a perfective reading but not with a non-perfective one, whereas their ni yotte
passive counterparts would not impose such a reading restriction. Specifically, they
accepted outright ta- and iru-marked ni direct passive sentences carrying a perfective
reading, while they did not do so with those carrying a non-perfective reading (i.e., a past
reading and a progressive reading for the ta— and iru-marked ones, respectively). They
furthermore equally accepted the ni yotte counterparts irrespective of the two readings,
either perfective or non-perfective. The subjects of the low advanced group showed the
same judgment pattern as the NS control and the highly advanced groups, albeit to a
lesser extent, in rating the ta-marked ni direct passive sentences. The low advanced
group, however, equally accepted the iru-marked ni direct passive sentences not only
with a perfective but also a non-perfective reading. Their mean judgment scores of all
the ni yotte passive sentences stayed very low (see Predictions 5 and 6).

Third, the subjects of the highly advanced group clearly showed differential
judgments in grammaticality between the ni indirect passive and the ni yotte indirect
passive with an adversative reading. This indicates that they knew that the ni yotte
marking was not compatible with an ‘affectee’ 0-role subject status of the indirect
passive. In contrast, they did not evince a clear judgment distinction between the ni
direct and the ni yotte direct passive with an adversative reading. The subjects of the low
advanced group, on the other hand, failed to show any clear contrastive judgments
between ni and ni yotte passives (either direct or indirect) carrying a pronounced
adversative reading. It thus appears that they were not sensitive to the incompatibility
between an affectee O-role subject status and the ni yotte marking (indicating a non-O-role

subject position) (see Prediction 7).

222

 

Fourth, in rating the ni direct and the ni yotte passive with regard to verb-induced
viewpoint differences, the highly advanced and the low advanced groups accepted the ni
direct passive with Japanese native verbs (expressing a personally involved viewpoint) as
well as the ni yotte passive with Sine-Japanese verbs (expressing an objective one).
Furthermore, both groups rejected the ni yotte passive when used with Japanese native
verbs. Thus it appears that the subjects of both groups knew that the ni yotte passive was
consonant with an objective viewpoint, but not with a personal one when the viewpoint
differences were brought out by the Sino-Japanese versus Japanese native verb contrast.
On the other hand, the highly advanced group, but not the low advanced group, made a
contrastive judgment between the ni direct passive and the ni yotte passive where
passivized affective verbs set forth a context in which the passive subject was
psychologically affected by the event described by the passive sentence (see Prediction
8). Hence, these highly advanced group’s judgment patterns indicate that the subjects of
this group knew that the subject position of the ni yotte passive is a non-O-position while
that of the ni direct passive is an ‘affectee’ O-position. By contrast, it was suggested that
the low advanced subjects’ knowledge of the ni direct and the ni yotte passive contrast
with respect to Japanese native versus Sino-Japanese verbs contrast was an isolated case
of knowledge development not attributable to the learning of the critical semantic
difference between the two types of passives.

Overall, throughout the above findings pertaining to subjects’ semantic
knowledge of Japanese passives, it has been convincingly shown that the subjects of the
highly advanced group knew that the ni passive (direct and indirect) and the ni yotte

passive are different in light of the passive subject status: viz., the former carries an

223

 

‘affectee’ 0-subj ect position while the latter a non-O-one. Particularly noteworthy is that
they made differential judgments between perfective and non-perfective readings in
rating the ta- and iru-marked ni direct passive sentences despite the unavailability of
negative input indicating the ungrammatical status of non-perfective readings of these
types of ni direct passive. This strongly suggests that the highly advanced subjects
deduced these ni direct passive grammaticality contrasts from its ‘affectee’ 0-subj ect
requirement such that an ‘affective’ reading for an inanimate passive subject is not
available when the situation is in a non-perfective description, whereas such a reading is
readily available in a perfective description.

There was also evidence suggesting that the low advanced group had been aware
that the ni direct passive requires an ‘affectee’ O-subject. Namely, we found 1) that they
knew that the ni direct passive might be read with an adversative reading, but not with a
non-adversative one; 2) that they had been developing knowledge of the contrastive
grammaticality in the ta-marked ni direct passive with an inanimate subject between
perfective and non-perfective readings; and 3) that they had knowledge that the ni direct
passive was consonant with a personal expression whereas the ni yotte (direct) passive
was appropriate for an objective one. Hence, the present study has clearly established
that as they become highly proficient in the TL, English-speaking learners of Japanese
are able to learn that the at passive, direct and indirect, requires an ‘affectee’ O-subject
whereas the ni yotte passive takes on a non-O-subject (see also F eng, 1993 for successful
learning of part of the ni and ni yotte passive contrasts by Chinese-speaking learners of

Japanese).

224

 

In contrast, the only semantic area in which the highly advanced group failed to
demonstrate the necessary knowledge concerns recognition of the incompatibility
between the ni yotte direct passive and an enhanced adversative reading. The subjects of
this group did not rate the ni yotte direct passive with an adversative reading significantly
differently from the ni yotte passive without such reading. Considering the evidence that
they arrived at the knowledge that the ni passive requires an ‘affectee’ O-subj ect while the
ni yotte passive assumes a non-B-subject position; and, moreover, hypothesizing that the
ni and ni yotte marking contrast served an important role in their semantic interpretation
of passive sentences in such a way that the ni marking was associated with an adversative
reading and the ni yotte marking with a non-adversative one, the following speculation
follows: subjects of the highly advanced group failed to learn that, albeit to a lesser extent
than in the ni indirect passive, the passive verb rare may intrinsically invoke an
adversative connotation in the ni direct passive. Consequently, they did not read
adversity in the ni yotte-marked (not ni-marked) direct passive when a neutral-
connotation verb was passivized (since there was no source indicating an adversative
reading to them in such passive sentence, i.e., the lack of the ni phrase and no negative-
connotation word). If this line of reasoning is correct, it would be this area, i.e., the
property of the passive verb rare invoking an adversative connotation in the ni direct
passive, that is most difficult for English-speaking learners to learn in terms of the
semantics of Japanese passives (see also Tanaka, 1993; Watabe et al., 1991 for findings
of English-speaking learners’ difficulty with use of the adversative ni direct passive).

On the whole, as summarized above, the findings of the present study clearly

show that the NNS subjects (particularly those of the highly advanced group) classified

225

the three types of Japanese passives into two semantic classes in exactly the same way as
the NS subjects did: one class is characterized as carrying an ‘affectee’ 0-subject, i.e., the
ni direct and the ni indirect passives, the other as carrying a non-O-subject, viz. the ni
yotte passive. (Note also that the low advanced group appears to have developed the
knowledge that the ni direct passive carries an ‘affectee’ 0-subject.) This, in turn, leads
to the conclusion that the subjects of the highly advanced group successfully attained
knowledge of properties of the Japanese passive morphology rare with respect to the
presence or absence of external 0-role assignment such that the passive verbs of the ni
passives assign their external ‘affectee’ O-role to their subject, while the passive affix of
the ni yotte passive does not.

In conclusion, together with the findings of their knowledge of passivization
processes involved in the three types of passives, the (highly advanced) NNS subjects’
knowledge representation of properties of Japanese passives would be characterized as

given in table 29 below:

Table 29: NNS subjects’ characterizations of the three types of
Japanese passive morphology rare

 

 

‘Affectee’ 0-role Passivization
Assignment
ni direct passive + +
ni indirect passive + + w/optionality
of case
absorption
ni yotte passive - +

 

 

This representation is almost identical with that given in Table 1 which is assumed to

represent NSs’ characterizations of Japanese passives. This almost perfect

226

correspondence between the highly advanced NNS subjects’ and the NS subjects’
representations comes as no surprise in view of the remarkably good knowledge of
passives the highly advanced NNS subjects acquired. (As for the low advanced NNS
subjects, they seem to have attained knowledge of both properties of the passive verb
rare of the ni direct passive, but not those of the other two.)

There is, however, one difference between the two representations. As discussed
above, the highly advanced NNS subjects seem to have opted for analyzing the three
types of passives as having passivization processes in common. That is, with respect to
the ni indirect passive, they appear to have analyzed accusative case absorption as
optional, rather than differentiating it from the other two types of passives, i.e., the ni
direct and the ni yotte passives, by analyzing the former as not triggering passivization
operations (just as NSs do). Why was this the case? Put differently, why did they choose
the first type of analysis over the second when confronted with TL data which had not yet
been incorporated into their IL grammar, notwithstanding apparently equal utilities of the
two analyses in accommodating such data into their IL grammar development? One line
of inquiry would be to explore the possibility that there are some predispositions which
constrain the ways in which SL learners tackle the complex task of SLA just like those
posited for FLA (e. g., the subset principle). In this particular case, it might be
hypothesized that owing to such mechanisms, the highly advanced NNS subjects were
guided to opt for the optionality analysis which admits of a single type of the passive
morphology rare over the one specifying two different types of rare with respect to

passivization.

227

 

 

There is a large body of SLA studies which describ IL developments or changes
in great detail, and which furthermore attempt to account for why such particular changes
took place or did not. To the best of my knowledge, there is as yet no SLA study which
has explored the more basic question of whether there are any guiding mechanisms (as
sketched above) for IL grammar development, and what they are. Our understanding of
the product as well as the process of SLA would surely benefit fiom such an inquiry.

Finally, pedagogical implications of the findings of the present study are
considered. First, the widely held view that the ni indirect passive is most difficult for
English-speaking learners to learn received strong supportive evidence in the present
study. We found no clear evidence suggesting that the low advanced group, let alone the
intermediate group, had learned this type of passive in either syntactic or semantic terms.
More noteworthy is that we successfully identified where such learning difficulties
resided. Namely, it was found that the prime and critical learning difficulty was with its
syntactic properties, more specifically, its embedded clause not being passivized despite
the presence of the passive verb rare and the ni phrase. One pedagogical implication of
this finding would be as follows: since the presence of an accusative o-marked NP is the
only apparent indication that the lower clause is unpassivized, enhancing English-
speaking learners’ attention to its presence may be effective in helping them recognize
this peculiar property of the ni indirect passive. One promising area in future research
would then be to explore ways to draw their attention to the accusative o-marked NP in
the lower clause of the ni indirect passive.

Such instructional intervention strategies may be devised from a wide variety of

 

techniques discussed extensively in the research paradigm of ‘focus on form’ (broadly

228

defined) (see Doughty & Williams, 1998 for thorough discussions of the use of ‘focus on
form’ techniques and its pedagogical implications). Specific proposals of techniques may
include ‘briefing’ (brief, contextualized, yet explicit instructions on the structure of the ni
indirect passive) together with focused repetitions and recasts of the target form (viz. the
o-marked NP) in an intensive fashion.

Next, as for the ni direct passive, it is quite encouraging that we found that both
the highly advanced and the low advanced groups had come to know that the ni direct
passive takes an ‘affectee’ O-subject. A pedagogically and theoretically important
question then arises; viz., what enables English-speaking learners to arrive at this
knowledge which often hinges on the recognition of delicate semantic nuances? The
answer could prove vital in expediting their learning processes in this aspect of the ni
direct passive. (But in order to address this question, we certainly need to achieve a
better theoretical understanding of its semantic properties, particularly in view of the ni
versus the ni yotte marking contrasts in Japanese passives.)

Experimental studies controlling input provided to learners may reveal insightful
information on what leads them to associate O-subj ect status with the ni direct passive.
One good candidate for such experimental input controlling would include the use of
typographical enhancement of input ingrained in the input flood in written modes. In
order to control the characteristics of input in such research, the test sentences used in the
present study will serve as a basis to build upon.

Lastly, given the complementary nature of the ni yotte passive and the ni direct
passive in semantic terms, research along the above lines should also provide clues as to

how best to facilitate the learning of semantic properties of the ni yotte passive. Equally

229

 

important for this type of passive is to explore ways to help learners attend to the
morphological difference between the ni yotte and ni marking. These considerations
should be geared to assisting the English-speaking learners in realizing that the ni yotte

passive is essentially the same as the English passive.

 

230

 

ENDNOTES

231

ENDNOTES

CHAPTER 1

1 This English gloss is modified later in the thesis when a crucial distinction in
Japanese passives is introduced.

2 Roughly speaking, both ni and ni yotte correspond to by in the English passive. Yotte
‘owing’ in ni yotte is an inflected form of the verb yoru ‘owe.’ It forms an adverbial
phrase with the dative marker ni, i.e., John ni yotte, as in (3). This verb may also occur
elsewhere in the language, but regularly appears with ni.

3 Later, Kuroda (1979) rejected a unique semantic characterization of “affectivity”
such as “the passive subject being adversely affected” (p. 310). He instead claimed that
“the semantic concept of “affectivity”.... manifests itself in various forms of semantic
effects, depending on other semantic factors such as the lexical meanings of other
elements in the sentence” (pp. 310-311). Such examples of semantic effects which
clearly bring out the “affectivity” connotation in the ni direct passive are illustrated later
in this section. (See Kuno, 1983; 1986; Kuroda, 1985 for debates over the semantic
theory of Japanese passives.)

4 Basic structures of these two verb classes are contrasted below:

i) John-ga Bill-o mi-ta
-NOM -ACC see-Pst
‘John saw Bill’
ii) J ohn-ni Bill-ga mie-ta
-DAT -NOM visible-Pst

‘Bill was visible to John’

In the first verb category (i.e., transitive verbs), the experiencer (perceiver), John, is
marked with a Nominative case marker and the patient (the object of perception), Bill, is
Accusative case-marked. In the second verb category (e. g. unaccusative verbs), by
contrast, the experiencer is marked with a Dative case marker ni, different from the
passive agentive marker ni, and the patient is Nominative case-marked (See Martin,
1975 for other such pairs of transitive verbs and unaccusative verbs).

5 The tokoro phrase is considered as a kind of temporal adverbial phrase like a
temporal ‘as’ phrase as given in the English gloss.

6 Hoshi assumes the Principles and Parameters approach (see Chomsky, 1981; 1986;
1995).

7 Other linguists such as Kitagawa (1986) and Miyagawa (1989) use “experiencer”, a
more general thematic term, for the 0-role assigned to the ni passive subject.

232

 

8 Lasnik & F iengo (1974) observed this licensing condition on subject oriented
adverbs and alluded to a distinction between be versus get passives in English, much as in
the distionction between ni versus ni yotte passives in Japanese (pp. 553-554).

9 Hoshi’s (1994a; 1994b) analysis adopted Saito’s (1982, cited in Hoshi, 1994a) theory
of case assignment in Japanese: “nominative case go is structurally assigned to an NP
which is immediately dominated by IP; accusative case 0 is assigned to an object; as for
the dative marker ni, it is assigned to an argmnent of a verb which cannot surface with
either nominative case ga or accusative case 0” (p. 23).

10 Hoshi (1994a) adopted Chomsky’s (1981) and Marantz’s (1984) proposal that
passivization is a morphological process triggered by an affix, i.e., -rare- in the case of
Japanese passives.

11 Following Kitagawa (1986) and Kuroda (1965), Hoshi (1994a) assumes that the
passive verb rare assigns an ‘event’ role to the embedded clause without defining it (nor
do Kitagawa and Kuroda offer it). Since the present study does not hinge on a precise
nature of this semantic role, it will be assumed.

12 See Larson (1988) for a discussion of the Larsonian VP shell and Washio
(1989/1990) for an insightfiil application of the VP shell structure to the analysis of
Japanese passives to which Hoshi is indebted.

13 See Hoshi (1994a) for more arguments for the presence and movement of PRO
within VP; in (24). Since the precise nature of the empty category concerned does not
affect discussions of the present study, the structures proposed in (24) will simply be
assumed.

14 The first type corresponds to the ni direct passive and the second to the ni yotte
passive in the terms followed in the present study.

15 Tanaka (1993) maintained, as a generally held view, that the indirect passive is
difficult to learn, especially for learners whose first language (L1) does not have the same
kind of passive construction.

16 Postpositional kara designates a source/ori gin meaning to the NP after which it is
placed; on the other hand, prepositional de assigns a meaning of instrument or locale of
activity.

17 The motivation for inclusion of NNS subjects in Spencer's (1973) study is irrelevant
for the present study; however, it is given in the text for the sake of completeness.

18 Since there is no information on the test sentences available in the paper, it is not

clear on which sentences they agreed or disagreed, nor how many of the accepted
sentences were accepted by both groups.

233

19 Two sentence structures were inadvertently omitted from the booklets for the
absolute judgments condition, resulting in a total of ten structures tested.

20 The Accessibility Hierarchy is given in (i) below:

(i) Subject > Direct Object > Indirect Object > Object of Preposition >
Genitive > Object of Comparative

It is Irnplicational Universal in that if a given language X has, for instance, Indirect
Object relative clauses, it also has all relative clause types higher on the hierarchy,
namely, Direct Object and Subject relative clauses. Languages differ in terms of the
lowest relative clause type they allow. See also Gass & Selinker (2001).

21 The erratic change between Time 1 and Time 2 was operationalized as “a change of
6 points, that is, from +3 to -3 or vice versa” (Gass, 1994, p. 317).

CHAPTER 2

1 Kuno (1983; 1986) plainly rejected some of the semantic differences between the ni
yotte and the ni direct passives Kuroda (197 9) proposed.

Chapter 3

1 Data were collected from a total of 57 N83 for this group. Thirty-six of them did not
specialize in Japanese theoretical linguistics: their concentrations included English
linguistics, teaching Japanese as a second language, psycholinguistics, historical
linguistics, and sociolinguistics. Consequently, they were not included in the present
study.

2 Ten subjects were excluded from this group since they had taken linguistics courses
before.

3 There was one sophomore majoring in English literature in this group. He was
included in the present study for the following considerations: 1) he had not taken any
linguistics courses; and 2) judging from his year in college (the first semester of the
sophomore year), he probably had not taken his major subject matter courses yet.

4 It was decided at the outset of the present study that beginning and low intermediate
learners of Japanese would not be considered due to the complexity of the target
structures under investigation.

5 As many as 26 data sets collected were not included in the present study for one or

more of the following reasons: 1) subjects were not native speakers of English, but were
either Chinese or Korean; 2) the questionnaire, the pre-test, or the background

234

 

 

information sheet was not completed and as a consequence part of the necessary
information was not available; 3) Pre-test scores were too low.

6 There are four levels in the JLPT, e.g., Levels 1, 2, 3, and 4. They are broadly
characterized as the levels attainable after studying Japanese for 900, 600, 300, and 150
hours, respectively.

7 In the testing materials, the term acceptability rather than grammaticality was used to
avoid the possibility of encouraging subjects to rely on explicit knowledge of grammar in
rating test sentences which might have resulted fi'om a potential association of the latter
term with prescriptive grammar. This thesis, however, adopts the latter term which is
more commonly in use to refer to performance data on sentence acceptability
assessments. (See Birdsong, 1989; Schiitze, 1996 for clarification of theoretical
distinction between acceptability judgment and grammaticality judgment.)

8 The following have been consulted to construct the test sentences: Alfonso, 1980;
Hoshi, 1994a; Howard & Niyekawa-Howard, 1976; Inoue, 1976; Jorden, 1987; 1988;
1990; Kitagawa, 1986; Kuno, 1973; 1983; Kuroda, 1965; 1978; 1979; Martin, 1975;
Miyagawa, 1989; Miyagi, 1996; 1999; Shirai & Kurono, 1998; Teramura, 1982; 1984;
Tsujimura, 1996.

9 Determination of the scope of test sentences for syntactic properties was based on
Hoshi’s (1991; 1994a; 1994b) theory and Tanaka’s (1992) data.

10 It is assumed here that the internal argument uta ‘song’ has initially moved to SPEC
of IP and then been topicalized, i.e., has moved to a topic position. (See Miyagawa, 1989
for a summary of various analyses of topic constructions in Japanese.) A detailed
analysis of topicalization in Japanese is irrelevant for the purposes of the present study in
that nothing critical hinges on it.

11 Non-operation of PRO movement was not investigated in the present study. NNS
subjects’ knowledge of PRO movement, proposed in Hoshi (1991; 1994a; 1994b), is very
difficult to test due to its non-phonetic, abstract nature.

12 Furthermore, in (36a), John undergoes topicalization.

13 Determination of the scope of test sentences for semantic properties was based on
Kuroda's (1979) analysis of Japanese passives.

14 Verbs used to test subjects’ knowledge of this semantic property were carefully
selected to include each class of the following verbs; activity, achievement, and
accomplishment verbs. Works consuled include Inoue (1976); Shirai & Andersen
(1995); Shirai & Kurono (1998); Terarnura (1982; 1984); and Tsujirnura (1996). This
was felt necessary to counterbalance any interaction a particular class of verbs may have
with the aspectual interpretation of te iru or to forms.

235

 

15 See Bardovi-Harlig & Reynolds, 1995; Shirai & Andersen, 1995; and Shirai &
Kurono, 1998 for acquisition of tense and aspect at early stages of acquisition. Although
the acquisition of the perfective reading of the te iru form presents some learning
difficulty to learners at early stages, NNS subjects of the present study at their (very) high
proficiency levels were extremely likely to have acquired both ta and te iru forms.

16 The distractor sentences with te iru and to forms were also intended for use as an
indicator of NNS subject’s knowledge of the respective form. It turned out, however,
that these sentences were not well-constructed for this purpose. Hence, they were not
used as planned in the analysis. See also endnote 15 above in this chapter.

17 There were cases where participants requested that their corrected proficiency and
vocabulary tests be sent back to them together with a brief explanation of properties of
Japanese passives in question. Their requests were granted.

CHAPTER 4

1 Statistical techniques employed in this chapter were determined through consultation
with a statistician. See Neter, Wasserrnan, & Kutner (1990) for an explication of them.

2 When the mean judgment score of a given test sentence was positive, that judgment
was considered to be grammatical, whereas when it was negative it was regarded as
ungrammatical.

3 There are three possible combinations of the two NS groups’ judgments of any given
test sentence vis-a-vis the theory prediction: 1) both groups agreed with it; 2) one group
agreed but the other did not; and 3) neither group agreed. There were only two cases
where neither group agreed with the predictions (discussed later). There are, thus,
basically two patterns of group judgment combinations to be considered: 1) those where
both groups agreed with the prediction and 2) those in which one group agreed, but the
other did not.

4 See Results and/or Discussion for discussions of these four test sentences on which
linguistically naive NSs did not agree with the predictions.

5 The unaccusative test sentences were included here since they are not passive
sentences consisting of the ni yotte, the ni direct, and the ni indirect passives.

6 Sentence (47b) is supposedly ungrammatical for the following reason: being a
sentient entity (and thus being affected by the event described), the subject must be
assigned an ‘affectee’ O-role which the ni yotte passive does not provide. Sentence (47c)
is expected to be ungrammatical since the sentence-initial adverbial phrase saiwai na koto
ni ‘fortunately’ is not consonant with the adversative reading this ni direct passive
sentence carries.

236

 

7 This particular item (47a) escaped my attention in discussion of the test sentences
with an instructor of Japanese and in pilot testing.

8 The other test sentence of the ni direct passive with an adversative reading was rated
as grammatical by both groups (mean judgment scores of 1.84 for the naive group and of
.48 for the sophisticated group).

9 The numbers given here correspond to those assigned to the test sentences in
Appendix J.

10 It is not entirely clear why the ni yotte passive was judged lower than the ni direct
and ni indirect passives. First, in test-sentence terms, Group 1’s mean ratings of two test
sentences of the ni yotte passive were 1.30 (SD = 1.06) for #47 and 1.07 (SD = 1.31) for
#50. There was no significant judgment difference between them (F(1, 29) = .518, p =
.477), suggesting that they were judged similarly. Second, in individual subjects’ terms,
83.33% of Group 1 (25 out of 30 subjects) rated #47 as grammatical and 77.42% (24 out
of 31 subjects) judged #50 as grammatical (both figures were not alarmingly low).
Furthermore, no single subject of Group 1 judged them both ungrammatical. It is thus
not the case that there was a serious confounding problem with either one of the two test
sentences nor that any small number of subjects’ judgments distorted Group 1’s general
judgment trends.

11 These two types of ungrammatical sentences correspond to those of Ungrammatical
1 and 2 in Table 10. See also Figure 4 for illustration.

12 These two types of ungrammatical sentences correspond to those of Ungrammatical
1 and 2 in Table 11. See also Figure 5 for illustration.

13 For this analysis, the ni indirect passive sentences with intransitive and transitive
verbs were collapsed since there was no judgment difference between the two types for
any group; F(1, 30) = 2.986, p = .094 for Group 1, F(1, 24) = .373, p = .547 for Group 2,
F(l, 35)=1.019,p = .320 for Group 3, and F(1, 19) = .581,p = .455 for Group 4.

14 Group 1’s slightly lower judgment of the ni indirect passive with intransitive verbs
was due to their relatively low judgment of one of the two test sentences, i.e., #13 (which
was taken from Kuroda, 1979). Group 1’s mean judgment score of it was .84 with the
SD of 1.42 (in contrast to 1.61 with the SD of .92 for the other test sentence, viz. #16). In
individual response terms, 71.0% of Group 1 (22 out of 31 subjects) rated it as
grammatical while 29% of them (9 in 31) judged it as ungrammatical.

15 In Group 2 judgments of ni direct passive sentences, there was no significant
difference between the sentences with an adversative reading and those used for the
syntactic analysis (F(1, 24) = .320, p = .577). It is thus not the case that the absence of
judgment differences found between the ni indirect and direct passives in the present
analysis was due to a lowered judgment of the ni direct passive with an adversative
reading.

237

 

16 This difference appears to be due to the extremely high score in judgment of the ni
direct passive with a perfective reading, e. g., the mean judgment score of 1.98 and the SD
of .0898 rather than because of a low rating of the ni yotte passive with a past reading. In
individual response terms, there was only a single occasion of +1 out of 64 responses in
rating the ni direct passive. The remaining 63 were all +2. As for the ni yotte passive
with a past reading, there was no significant judgment difference in all four types of to-
and iru—marked ni yotte passive sentences in question (F (3, 90) = 1.618, p = .191),
suggesting that the ta-marked ni yotte passive with a past reading was not judged
peculiarly low compared to the other types of ni yotte passive sentences.

17 Group 2 did not reject the ni direct passive with a past reading as ungrammatical (a
mean judgment score of .260). An inspection of Group 2 subjects’ individual responses
to the two sentences in question showed that six subjects of a total of 25 judged both
sentences as ungrammatical and eight subjects rated one of the two as ungrammatical; on
the other hand, eleven subjects did not judge either of them as ungrammatical. Note,
however, that in the analysis of judgments of three syntactic types of passives, it was
found that Group 2 subjects rated the ni yotte passive significantly lower than the ni
direct passive. By contrast, the ni direct passive with a past reading in this analysis was
judged significantly lower than the ni yotte passive of both readings. There was no
significant judgment difference in all three types of ni yotte passives above (F(2, 48) =
2.308, p = .110), suggesting that ta-marked ni yotte passive sentences of both readings
were not rated particularly low compared to the ones used in the syntactic analysis. This
in turn strongly indicates that the ni direct passive with a past reading was judged
conspicuously low.

18 The term ‘tendency’ was used in the text since Group 2 as a group did not reject the
ni direct passive with a past reading as ungrammatical.

19 Group 3’s mean judgment score of the ni direct passive with a past reading was .625.
An inspection of Group 3 subjects’ individual responses to the two sentences in question
indicated that eight subjects of a total of 36 judged both sentences as ungrammatical and
six subjects rated one of the two as ungrammatical; on the other hand, twenty-two
subjects did not judge either of them as ungrammatical.

20 Group 2’s mean judgment score of the ni direct passive with a progressive reading
was .48. An inspection of Group 2 subjects’ individual responses to the two sentences in
question showed that five subjects of a total of 25 judged both sentences as
ungrammatical and eight subjects rated one of the two as ungrammatical; on the other
hand, eleven subjects did not judge either of them as ungrammatical and one subject
chose the ‘unsure’ option for both sentences.

21 Groups 1 (and 2) did not reject the ni yotte indirect passive with transitive verbs as

ungrammatical (mean judgment scores of .26 and .0600). This will be considered in the
Discussion.

238

 

 

22 There were no significant judgment differences between the ni direct passive
sentences with a pronounced adversative reading and those without it (used for the
syntactic analysis) for Groups 2 and 3 (F(1, 24) = .320, p = .577 for Group 2; F(1, 35) =
.053,p = .819 for Group 3).

23 Since there is no grammatical ni yotte indirect passive, a comparison between
grammatical and ungrammatical ni yotte indirect passives cannot be performed.
However, note that Group 2, which successfully made an expected grammaticality
distinction, rated the ni yotte indirect passive with intransitive verbs as ungrammatical
(with a mean judgment score of -.60). The results of the ni yotte indirect passive with
transitive verbs will be considered separately in the Discussion.

24 The grammatical test sentences used for the syntactic analysis were contrived to be
as neutral in meaning as possible, i.e., not to import any markedly pronounced effect such
as an adversative reading or viewpoint specification by lexical choice.

25 Specifically, the inspection of responses to ‘korosu’ used in sentences #1 and #2
disqualified four subjects of Group 3 and six of Group 4; for ‘satugai—suru’ used in
sentence #3, four of Group 2, fifteen of Group 3, and twelve of Group 4 were excluded.
Similarly, the examination of vocabulary test responses to ‘homeru’ used in sentences #4
and #5 disqualified six subjects of Group 3 and eight of Group 4; for ‘syoosan-suru’ used
in sentence #6, ten of Group 2, twenty-five of Group 3, and twelve of Group 4 were
excluded.

26 Group l’s non-negative judgment mean score for the ni yotte passive with Japanese
native verbs was due to its acceptance of sentence #2 (= (65b) in the text). Although
Kuroda (1979) discussed sentences such as (65b) as an example of the incompatibility
between the ni yotte passive and Japanese native verbs, it was judged as grammatical by
Group 1 (with a mean judgment score of 1.23). In individual response terms, 77.42% of
Group 1 (24 out of 31 subjects) rated it as grarmnatical while 16.13% of them (5 out of
31) judged it as ungrammatical; 6.45% of them (2 out of 3 1) chose the ‘unsure’ option.
On the other hand, the linguistically sophisticated group judged it as ungrammatical (with
a mean judgment score of -l .05). As for the other test sentence #5, both the linguistically
naive and sophisticated groups rejected it with mean judgment scores of -.77 and -1 .19,
respectively.

27 This process disqualified three, sixteen, and seven subjects from Groups 2, 3, and 4,
respectively.

28 Group 1 rated ni direct passive sentences quite low (a mean judgment score of
.08065). Their judgment means for the two test sentences were .58 for #7 and -.42 for
#10 (F(1, 30) = 8.774, p = .006). Thus, the low judgment mean score of #10 (i.e., (66a)
in the text) was partially responsible for the overall low rating. However, the other
sentence (adopted from Kuroda, 1979) did not receive a high rating, either. The low
rating of the ni direct passive with affective verbs by Group 1 will be considered in the
Discussion.

239

 

 

29 Three and five subjects of Groups 2 and 3, respectively, were excluded fi'om the
analysis.

CHAPTER 5

1 In this respect the present study differed from the previous experimental studies
which addressed the issue of the reliability of linguists’ judgments where subjects
assessed a wide range of types of sentences.

2 The age of 2;3 and 6;2 indicates two years and three months of age and six years and
two months old, respectively.

3 The most advanced children correctly performed nearly 80% of the test sentences for
the two word order types of passive.

4 See Otsu, 1999; Suzuki, 1998 for their methodological concerns about these studies.
Specifically, they criticize these previous studies for not taking care to establish a passive
subject as a discoursal topic, which leads not only to unnaturalness of the sentence, but
also to difficulty with shifting a viewpoint to the one appropriate for the passive sentence.

5 According to Fox & Grodzinsky's (1998) characterization of the non-truncated
passive with ‘actional’ verbs, “by” can directly assign the agent/affector G-role to its
complement noun. This obviated the need for the G-role transmission process to come
into play in this type of passive, leading to children’s good comprehension performance.
They were capable of comprehending the truncated passive with ‘non-actional’ verbs
because there is no by-phrase in this passive. (See also Baker, Johnson, & Roberts, 1989;
Grimshaw, 1990; Jaeggli, 1986 for detailed discussions of the nature of passivization
mechanisms.)

6 Hoshi's (1994a; 1999) analysis of the ni indirect passive as [- Passivization]
processes is primarily motivated by the difference in the reflexive binding phenomena
between the ni direct and the ni indirect passives. Specifically, the ni-marked NP of the
ni direct passive does not serve as an anaphor for reflexive zibun ‘self while that of the
ni indirect passive does, suggesting that the latter NP maintains its subject status (which
in turn implies that it is not passivized). His additional arguments concern passivizability
of ergative verbs as well as the obligatory nature of the ni phrase in the ni indirect
passive. See Hoshi (1994a; 1999) and references cited there for various analyses of these
phenomena.

7 In individual response terms, for one sentence (#20), 48.39% of Group 1 (15 out of
31 subjects) rated it as ungrammatical whereas 51.61% of them (16 in 31 subjects)
accepted it. For the other sentence (#23), 41.94% of them (13 out of 31 subjects) rejected
it while 54.84% of them (17 in 31 subjects) judged it as grammatical (with one subject
who opted for the ‘unsure’ option).

240

 

8 The Double-o Constraint states that a verb can assign accusative case to at most one
NP in Japanese (see Harada, 1973, cited in Hoshi, 1999).

9 Hoshi (1999) presented some evidence for the NP movement analysis of sentence
(80) by drawing on the requirement that a numeral quantifier and its associated NP be
mutually c-commanded (see Miyagawa, 1989 for this requirement).

10 Hoshi (1999) leaves it for further research to precisely characterize the nature of an
“additional” object/ “affected” argument.

11 Alternatively, Kuno (1983) simply discarded the line of interpretation of the ni direct
passive given by Kuroda (1979), sketched in the text, as a matter of personal preference.

12 The present study found some evidence that the highly advanced group appears not
to have fully learned that neutral-connotation verbs which have been passivized may
carry an adversative reading in the ni direct passive. See the full discussion of the rating
of adversity for the ni direct passive in the Discussion.

13 This is the only study to date in SLA investigating the acquisition of unaccusative
verbs in Japanese to my best knowledge (see Miyamoto, Wexler, Aikawa, & Miyagawa
1998 for this issue in first language acquisition of Japanese).

14 Surface unaccusativity refers to a situation where the sole argument of a verb
remains in the object position without moving out of the VP (see Kageyama, 1993; 1996;
Miyagawa, 1989 for unaccusativity in Japanese).

15 Takezawa (1987) provides the surface condition for the case drop phenomenon as

follows: “[w]hen an NP is adjacent to and c-commanded by V, the Case marker attached
to it can drop” (p. 126).

241

 

APPENDICES

242

 

APPENDIX A

Linguistically sophisticated NS subject background information

243

 

 

Table 30: Linguistically sophisticated NS subject
backgound information

 

Subject ID Specialty Educational Level Age Gender
501 JapLX Doctoral 29 M
502 JapLX Doctoral 29 M
503 JapLX Doctoral 30 M
504 JapLX Doctoral 24 M
505 LX Doctoral 37 M
506 JapLX Doctoral 27 M
507 JapLX Senior 22 F
508 LX Master 23 M
509 JapLX Doctoral 25 F
510 JapLX Doctoral 29 M
51 1 LX Doctoral 26 M
512 JapLX Master 25 F
513 JapLX Master 25 M
514 JapLX Master 27 F
51 5 JapLX Master 23 F
516 LX Instructor 27 M
51 7 JapLX Senior 22 F
518 CogSci PhD 52 M
519 LX PhD 35 M
520 LX Master 23 F
521 LX MA 32 F

 

 

 

Note 1: JapLX, LX, and CogSci stand for Japanese linguistics, linguistics, and cognitive
science, respectively.

Note 2: Doctoral and Master represent subjects in their respective graduate programs
whereas PhD and MA indicate PhD and MA holders.

244

 

APPENDIX B

Linguistically naive NS subject background information

245

 

 

Table 31: Linguistically naive NS subject
background information

 

Academic
Subject Id Major Year Age Gender
101 Socio Junior 21 M
102 Socio Junior 20 F
103 Socio Sophomore 20 F
104 Socio Sophomore 19 F
105 Socio Junior 21 M
106 Socio Junior 22 M
107 Socio Senior 21 F
108 Socio Junior 21 F
109 Socio Senior 21 F
1 10 Socio Senior 21 F
111 Socio Junior 21 M
1 12 Socio Junior 20 F
1 13 Socio Junior 20 F
1 14 Socio Sophomore 21 M
1 15 Socio Senior 21 F
1 16 Socio Junior 20 M
1 1 7 Socio Sophomore 19 F
1 18 Socio Junior 20 F
1 19 Eng Lit Sophomore 20 M
120 Socio Sophomore 19 F
121 Socio Sophomore 19 F
122 Socio Sophomore 19 M
123 Socio Sophomore 20 F
124 Socio Sophomore 19 F
125 Socio Junior 20 F
126 Socio Junior 20 M
127 Socio Junior 21 F
128 . Socio Junior 21 F
129 Socio Senior 21 F
130 Socio Junior 20 F
131 Socio Senior 21 F

 

 

Note: Socio and Eng Lit represent sociology and English literature, respectively.

246

 

 

 

APPENDIX C

NNS subject background information

247

 

Table 32: NNS subject background information
Length of Length of

 

 

Subject a Level in
ID Group JLPT Ieaming Japan visitb education Major Age Gender
1 2 25 5 yrs 1 yr MA Japanese Linguistics 28 F
4w32
2 2 25 mons 4 yrs MA Japanese Literature 27 F
3 2 25 4 yrs Senior Japan Studies 23 M
5yrs 6 9 yrs 6
4 2 25 mons mons Professor 52 M
2 mons 2
5 2 25 4 yrs wks BS Electrical Engineering 23 M
6 2 25 6 yrs 5 yrs MA Japanese 34 M
7 2 25 3 yrs Professor M
Political
8 2 25 1 yr Senior Science/Japanese 19 M
Japanese/lntemational
9 2 24 7 yrs 1 yr 3 mons Senior Studies 23 M
10 2 24 20 yrs 10 yrs Professor Japanese F
11 2 24 11 yrs 6 yrs MA East Asian Studies 30 F
12 2 24 2 yrs Junior Electrical Engineering 20 M
13 2 24 20 yrs 20 yrs MA Engineering 57 M
14 2 24 12 yrs 9 yrs PhD 39 M
15 2 24 36 yrs 11 yrs Professor 55 M
English Lit. 8
Lang/Japanese Lit. &
16 2 23 10 yrs 1 yr Junior Lang. 19 F
17 2 23 7 yrs 6 yrs Post Bac. 29 F
18 2 23 5 yrs 1 yr Junior Electrical Engineering 23 M
19 2 23 8 yrs 2 yrs Japanese 23 M
20 2 23 13 yrs 2 yrs PhD Japanese Literature 31 M
21 2 23 12 yrs 2 wks 19 M
22 2 22 1 yr Freshman Biology 19 M
23 2 22 3 yrs 1 yr Senior History/Japanese 21 M
24 2 22 5 yrs Sophomore Japanese/Anthropology 19 M
25 2 22 7 yrs 1 man 1 wk Senior Computer Science 21 M
26 3 21 4 yrs 3 mons Senior Biology/Japanese 26 M
Management Information
27 3 21 2 yrs Junior Systems 21 F
28 3 21 11 yrs 1 mon 27 M
29 3 21 4 yrs 1 mon Sophomore Journalism 19 F
30 3 21 3 yrs 2 mons Senior Anthoropology 22 M
31 3 21 1yr 6 mons Japanese/Asian Studies 19 M
32 3 21 2 yrs 10 mons Sophomore Japanese 21 M
33 3 21 13 yrs 15 yrs
Management Information
34 3 20 4 yrs 4 yrs Senior Systems 22 F
3 yrs 6
35 3 20 mons Senior Japanese 21 F

 

248

 

 

Table 32 (cont’d)

 

 

 

Subject Length of Length of Level in
ID Group JLPT Ieaming Japan visit education Major Age Gender
36 3 20 7 yrs 5 yrs MA Japanese Pedagogy 29 M
37 3 19.4444 4 yrs 10 mons Senior Asian Studies 22 M
38 3 19 1 yr 8 mons Senior Economics/Japanese 22 M
East Asian Lang 8
39 3 19 3 yrs 9 mons Senior Literatures 21 F
Anthropology/Japan
40 3 19 6 yrs 4 yrs MA Studies 30 F
Computer Science 8
41 3 19 4 yrs 3 mons Senior Japanese 22 M
Japanese/lntemational
42 3 19 7 yrs 10 mons Senior Studies 25 F
43 3 19 4 yrs 10 mons Senior East Asian Studies 21 F
44 3 18 3 yrs 1 mon Junior Engineering/Japanese 27 M
45 3 18 4 yrs 3 mons History/Japanese 22 M
3 yrs 6
46 3 18 mons 3 mons Senior lntemational Studies 22 M
47 3 18 3 yrs 1 yr Senior Anthoropology/Japanese 23 M
48 3 18 4 yrs 2 mons 56 M
2 yrs 6
49 3 18 mons Senior Spanish 21 M
50 3 18 1 yr 6 mons Junior Japanese 22 M
3 yrs 6
51 3 17 mons 6 mons Senior Computer Science 22 M
lntemational Affairs/East
52 3 17 3 yrs 10 mons Sophomore Asian Studies 19 M
53 3 17 3 yrs 3 yrs BA Anthropology/Arts 31 M
54 3 17 2 yrs Sophomore Japanese 19 M
55 3 17 3 yrs 1 yr Junior Japanese/Business 20 M
2 yrs 6 Computer
56 3 16 mons Junior Science/Japanese 20 M
2 yrs 6
57 3 16 mons Senior Business 22 F
4 yrs 6
58 3 16 mons 1 yr MA East Asian Studies 25 F
59 3 16 6 yrs 4 mons Senior lntemational Affairs 22 F
60 3 16 4 yrs Senior lntemational Affairs 21 F
61 3 16 2 yrs Junior Japanese 21 F
1 mon 2
62 4 15 7 yrs wks Sophomore Biology/Japanese 19 F
63 4 14 3 yrs Senior Japanese 22 M
64 4 13 6 yrs 4 mons Sophomore Japanese 19 F
65 4 13 7 yrs 2 wks Junior Joumalism/Japanese 20 F
1 mon 2
66 4 13 5 yrs wks Japanese/Mathmatics 20 F
2 yrs 6
67 4 13 mons Junior Japanese 19 F
1 mon 2
68 4 13 8 yrs wks 18 M

 

 

249

 

Table 32 (cont’d)

 

 

Subject Length of Length of Level in
ID Group JLPT Ieaming Japan visit education Major Age Gender
1 men 2

69 4 12 3 yrs wks Senior Japanese 25 M
3 yrs 6

7O 4 12 mons Japanese 22 F
2ws6

71 4 11 mons 1 mon Junior Japanese 24 M

72 4 11 3 yrs Junior English 20 M

Organizational
73 4 11 3 yrs 1 men 1 mon 1 wk Senior Communication 21 M
74 4 11 1 yr Senior Mathematics 22 M
East Asian Lang &

75 4 11 3 yrs Junior Literatures 23 M

76 4 11 7 yrs 2 mons Junior East Asian Studies 20 F

77 4 9 3 yrs Junior Asian Studies 21 F

78 4 9 3 yrs 1 mon Junior Criminology 20 F
3 yrs 6

79 4 9 mons 2 wks Junior East Asian Studies 21 F

80 4 9 3 yrs Junior Biology 21 F

lntemational
81 4 8 4 yrs 4 mons Senior Affairs/Japanese 21 F

 

 

a., h Empty cells in the Length of learning and the Length of Japan visit colurrms indicate that
relevant information was either unavailable or uninterpretable in the background information

sheets.

250

 

APPENDIX D

NS subject background information questionnaire

251

 

E35”? (El 2555”?) - Nilé‘afiié‘filcor VC
(Background Information Sheet)

TSEVDIEBIC’DD‘T. fifilCBEiT‘él/‘o
1. EEZJ (Name):

2. iii: (University) :
3. @151 (Major) :
4. (a) ”$515 (Year/Level) & (b) 513% (Age) & (c) 1&3” (Gender)

(6!) (b) (C)

3853351155“; (Chi T115115 L7”: (REFS) g3??? (B 211365?) 55%0151 a
Efi$ll1f$§ 3 ‘F “‘6 l, ‘) (Linguistics-related courses taken) :

llllll

5.

6.5’H321133E15E (Chi T111131? Ltfilflgg’i’fifiillfigé _F 3 Ir ‘) (Foreign languages
learned)

a. 9113??? (Foreign languages) :

b. 33% (When) :

c. $55 (How long) :

7. Eé’ifi (fiillk‘g‘é _F 3 l/ ‘) (Study abroad experiences) :

a. @396 (countries visited) :
b. 5%% (When) :
0. $1351 (How long) :

252

APPENDIX E

NNS subject background information questionnaire

253

Background Information Sheet

Please fill out the following items briefly.

1) Name: 2) Gender: 3) Age:
4) Major: 5) Year of Study:
6) Primary Language(s)

About Japanese Learning Experiences
7a) Length of Learning (when and how long):

7b) Most Recent Japanese Language Course You Took — Name and Grade Received:

8) Textbooks Used:

9) Previous visiting experience in Japan (in what capacity, when, and for how long):

10) Outside-of-classroom contact with Japanese (with whom and how frequently):

11) Foreign Language Learning Experience Other Than Japanese (what and how long):

254

 

APPENDIX F

NNS proficiency test

 

 

255

 

Proficiency Measurement (Suggested Time: 20 minutes)

PLEASE ANSWER THE FOLLOWING BY YOURSELF. DO NOT CONSULT A
GRAMMAR REFERENCE BOOK OR A DICTIONARY.

1. Circle the most appropriate word that goes in the parenthesis from the four choices
given below each sentence. (Please DO NOT write the word in the parenthesis.)

(Example)
iwbé LEE» r
bicbli fififl fiFsW ) 5353*???
1 fl 2 U) 4 'C“
twkfi tfbw
1 EVEN ) E42118 :rbtbi Lice
1 I: 2 a) 3 “c 4 (E
pg w 33 #1 (E
2 47M; Tux/<4 H: fi<( ). 5131:: 13681 Tan.
1 713 2 am 3 (or: 4 t-‘é
bmfiw 'uhaw b
3 :0) 115511 9E$( ) habitat/1,131,711.
1 m: 2 at) 3 CDT“ 4 (“is
s
4 MW) xrznci—m J
Jib o
r3050) ii :EB( )iééibicol
1 'C“ 2 l: 3 7.5: 4 hi
new
5 EEIFPéA/li aria—wig ( ) page“.
1 11:53” 2 11.1 53%: 3 1:1: 53“? 4 1:1: 5ft:

-Over-

256

 

10

11

12

awn!

Be 0
891E154 )Zbck’c. to‘fl'fi'b flail/.15..

A A A 1r
1 E6 2 ET 3 it: 4 it“)
be 0m (fl
7—71: at {Ewitfinb ::i: ( ) 2‘6)“ few
1 trig/VT: 2 ire/vii 3 trip/<5 4 ire/<1
th O< (t
$1.78 (for: esrva~xc+. 85% 2‘6( ) ram.
0) 0) 0) U)
1 81:8?) 2 fikir 3 $77k 4 filth/C“
(Debbie: ( ) wx'c'fnxol

“010%., :l‘o‘iah‘ib‘bi'i‘o J

y); 75.
1 gnaw 2 Envrwcli
n) 31
3 Envix'fli 4 én>t<fb
tam 3:5 c l\
Battle/Vi: 4‘8 tweet 1:301 vibrant). Zara) ( )ci’.
1 7329’) 2 :l: 3113‘ 4 ob”)
ears
%a> :e11 305 UJEBé/vl: «27:22.1 ( )0
1 raisin“ 2 <7‘iéw 3 twat 4 Liwi'f
:AE weft <K hi
3303 E14074: I“ 1512.3 ( )l: Lil/7:,
1 :3: 2 E0) 3 3917‘ 4 3:15
-Go to the next page-

257

11. Circle the most appropriate word that goes in the parenthesis from the four choices
given below each sentence. (Please DO NOT write the word in the parenthesis.)

(Example)
ihbé LASA ;
brain: as 511%“ ) 332233“.
1 75's 2 co
fitfit w0 t
1 ita Obie: £O( ) EN‘iLtco
1 ck 2 7a? 3 t)
<6; bL
2 $1: 0301( ) wet. 137$ Jub< fruit.
1 L7» 2 73:73”) 3 <“61n
:5Brbfibuw ewe
3 BEE513$( ) ear: [Shani braéhi Lt.
1 755‘ 2 75: 3 l:
Bofinhm:5 A?
4 fir-1321:. 7cd>liL( ) brace. 751: emit.
1 T“ 2 l: 3 a»

an r lié

@5 U
5 5M1: 2133's 71% fiat/up b\7‘:n>6. 44m: (

1 18%< 2 167C<T 3 new»;
u<uw -fibfiw 1‘
6 ”iii: ieiw A331: ( )ifil: Eton: Liz.
< e :
1 3E5 2 SET 3 $141»

258

4 72123:“

4 7536

4i)§

$3
)Eéibt,
4 16%<i'6

a
4 $147”:

-Over-

rum (:1er JP; t5 391,7: 3A,

7 Battle/Va) 331: 3:31;. mafia/Vi: BEE! scm1:( )‘6573.
1 0313623 2 0313725 3 prior: 4 013673735
157:1, 1‘88
8 til: it“ Hit (M53134 ) 3:373 Mia/V.
1 035 2 wot: 3 0350) 4 worm)
iii/UL: M: :51, <75
9 1437) I‘LL/1:0) ( )7575: that Tahiti/ens,
on: on on) on»
1 law 2 (£5 319501 41250)
7‘:
10 Bro}: fiN'C ( ). t'Ci) £b\b7b>of:'6"i”o
1 1,726 2 327:6 3 $510726 4 abate
ink
1 1 7°1/12‘yi'cri‘ta. Limb. Stunt: flit“ 2300774 )0

1 Lil/.15 2 4:03:15 3 finial/.15 4 3:13:15

1 2 bar/3:131: 137M) {it ( ) wiser/V.
1 1172351121 2 1:22:11:
3 1133322111113 4 lifxéfbefli
Fifi that» 72:32
1 3 LIJEII—E/wb§ 5159a ) 2:311? ienszenibt.
1 I: 2 A 3 a 4 1:1

259

APPENDIX G

NNS vocabulary test

260

Vocabulary Test

PLEASE ANSWER THE FOLLOWING BY YOURSELF. DO NOT CONSULT A
GRAMMAR REFERENCE BOOK OR A DICTIONARY.

III. Circle the most appropriate meaning of the word from the four choices given below
each word. (Please only choose one.)

(Example)
1:
3&5
1 drink
:7)
1 ET
1 defame

Lr5éh

2 Efité

1 appeal

a:
3 @Zié

1 be attentive

.7;-
4 E26

1 be visible

5 libé

1 enhance

30min
6 3%35

1 assault

cook

kill

recommend

be audible

SflB

praise

murder

261

save

injure

applaud

listen

witness

crave

harm

damage

regard

hear

be tangible

envy

poison

 

APPENDIX H

Instruction sheet on the grammaticality judgment test

262

Instruction sheet on the grammaticality judgment test

Please read the following carefully.

Sentence Intuition_§
Speakers of a language seem to develop a ‘feel’ for what is an acceptable sentence and
for what is not, even in many cases where they have never been taught any particular
rule.

For example, in English you may feel sentence (1) below sounds like an
acceptable English sentence, while sentence (2) does not.

1) John is likely to win the race.
2) John is probable to win the race.

Although sentences (1) and (2) are of the same structure, one can judge without
depending on any rule that sentence (2) is unacceptable in English. Similarly, you may
feel sentences (3) and (4) both sound unacceptable.

3) Which problem do you wonder how John could solve?
4) How do you wonder which problem John could solve?

In this case, however, you may feel that sentence (3) sounds unacceptable to a lesser
extent than sentence (4).

Likewise, in Japanese, you might feel that the first sentence below sounds like it
is an acceptable Japanese sentence, while the second one does not.

5) J ohn-wa Mary ni nihongo no hon-o yom-ase-ta.
6) J ohn-wa Mary ni nihongo-o deki-sase-ta.

The set of papers provided contain Japanese sentences. There is one sentence on
each sheet of paper. We would like you to tell us for each one whether you think it
sounds acceptable in Japanese or not. Even native speakers have different intuitions
about what is acceptable and what is not. Therefore these sentences cannot serve the
purpose of establishing your level of proficiency in Japanese. We would like you to
concentrate on how you ‘feel’ about these sentences.

_A_n_swering the Oucstionngirg

For the following sentences, each one of which is written on a sheet of paper,
please tell us whether you feel they sound acceptable sentences of Japanese to you, or
whether they sound like unacceptable Japanese sentences to you. There may be
sentences which you feel are more or less unacceptable. Alternatively, there may be
sentences which you feel are more or less acceptable. In these cases, mark somewhat
unacceptable and somewhat acceptable respectively. Finally, there may be sentences
where you have no clear feeling for whether they are acceptable or not. In this case mark

263

 

 

not sure. Answer one question at a time and place the sheet back in the envelope after
you have marked it. Please do not refer back to sheets you have already answered.

There is no time limit: you can spend as much time on the questionnaire as
needed. Read each sentence carefully before you answer. Concentrate on how you feel
about the sentence. Please mark only one answer for each sentence. Make sure you have
answered all 77 questions.

PLEASE COMPLETE THE QUESTIONNAIRE BY YOURSELF. DO NOT
CONSULT A GRAMMAR REFERENCE BOOK OR A DICTIONARY.

[The instruction sheet for Japanese NSs is written in Japanese. It is the same as the
English one except for the following changes. The paragraph concerning the English
examples above is omitted. Likewise, in Japanese (in the paragraph on the Japanese
examples) is replaced with For example, and the additional pair of Japanese examples
below is added to illustrate a differing degree of (un)acceptability:

 

a) Bill-ga yonde ita hon-wa Shakespeare data.
b) Bill-ga yonde iru hon-wa Shakespeare data

Finally, the paragraph discussing level of Japanese proficiency and the last capitalized
one are omitted]

 

264

APPENDIX I

Sample grammaticality judgment sheet

265

 

 

 

UT :. :6

Ell/ii, 73M?» w \Eflla’é’i‘ 5 317:,

Bill-wa kawaii musuko ni korosareta.

Acceptable Unacceptable
Somewhat Acceptable Somewhat Unacceptable
Note:

0+ :

75‘391fllfl : dear, Er?" : son

Not Sure

#1

 

266

 

 

 

APPENDIX J

List of test sentences

267

 

Note: The asterisk attached to a sentence number indicates ungrammaticality status of
that sentence.

1

Ell/ii, nxzbv w \.B¥l:i& 2‘5 tut.
Bill-wa kawaii musuko ni korosareta.
‘Bill was affected by being killed by his own beloved son’

Ell/121:. 33391] W ‘Efl: J: o Til??? 5 217:0
Bill-wa kawaii musuko ni yotte korosareta.
(See #1 above.)

Vay-xixu.TXUW®CIAKioTfi%énto
John Smith-wa America no CIA ni yotte satugai sareta.
‘John Smith was murdered by the CIA’

2‘ 7 U “tat/vii. 1001332 LINER é Alliib 6114730
Mary tyan-wa itu mo kibisii otoosan ni homerareta.
‘Mary was affected by being praised by her always strict father’

xrv~eenu.wobmbwaxsnuroteeent.
Mary tyan-wa itu mo kibisii otoosan ni yotte homerareta.
(See #4 above.)

xru—-vayyyu,94—+yxrurotfifiant,
Mary Johnson-wa journalists ni yotte syoosan sareta.
‘Mary Johnson was applauded by journalists’

FREELTl/‘é c‘: C 5%. It'll/73, fli—IVICE [521730

Nozokimi site iru tokoro-o, Bill-ga Paul ni mirareta.
‘Billi was (personally) affected by being seen by Paul as he; was peeping in (a

room)’

Wéfibfmétléé.E»fl.fi—»Kioffigflto
Nozokimi site iru tokoro-o, Bill-ga Paul ni yotte mirareta.
(See #7 above.)

fi—wm,fiwfifi%ELTWEkléfiEito

268

 

10

11

12

l3

14

15

16

a
17

Paul ni, Bill-ga nozokimi site iru tokoro-ga mieta.
‘Billi was visible to Paul as he; was peeping in (a room)’

iru—embtmé&:53.Vayfi.fi—w7vyhtfimnto
Mary to hanasite iru tokoro-o, J ohn-ga girlfiiend ni kikareta.
‘John was (personally) affected by being heard by his girlfiiend as he was talking to

Mary’

XTU—&%wa6&:5&.Vayfi,fi—w7vykniof%mhto
Mary to hanasite iru tokoro-o, John-ga girlfiiend ni yotte kikareta.
(See #10 above.)

Vaymfi—W7VVFK,VayfiXTU—&%Ltw6&:6fl%:it.
John no girlfiiend ni, John—ga Mary to hanasite iru tokoro-ga kikoeta.
‘John was audible to his girlfriend as he was talking to Mary’

Ell/ii. PEI] \Bfillfiéllfié ant.
Bill-wa warui toki ni tomodati ni korareta.
‘Bill was (adversely) affected by his friend visiting with him at an inconvenient time

9

ku.%wfimfiémrotkento
Bill-wa warui toki ni tomodati ni yotte korareta.
(See #13 above.)

ku.wwfiufiauxent.
Bill-wa ii toki ni tomodati ni korareta.
‘Bill was (adversely) affected by his friend visiting with him at a convenient time’

xru—u,ya¥yrbtwefimmuaent.
Mary-wa, jogging site iru toki ni ame ni hurareta.
‘Mary was (adversely) affected by it raining on her while she was jogging’

XTU—fl,93¥V7LTWEfiflmKioffi6flto
Mary-wa, jogging site iru toki ni ame ni yotte hurareta.
(See #16 above.)

269

 

t
l8

19

20

21

22

23

24

25

fiTU—fi.ya¥yflbfm6mtfibhto
Mary-wa, jogging site kara ame ni hurareta.
‘Mary was (adversely) affected by it raining after she had jogged’

gw:ku,Hem,VI—yKfi—w7vyhmewifiéfiiht.
Warui koto ni, Bill-wa, Jane ni girlfriend kara no tegarni—o yomareta.

‘Unfortunately, Bill was (adversely) affected by Jane reading a letter from his
girlfriend’

fiw:tt.Ewm.VI—yuioTfi—w7vyFm6®$flafiinto
Warui koto ni, Bill-wa, Jane ni yotte girlfiiend kara no tegami-o yomareta.
(See #19 above.)

$m&:tu,Bum,Vx—yKfi—W7vambwifiéfiint,
Saiwai na koto ni, Bill-wa, Jane ni girlfriend kara no tegami-o yomareta.

‘Fortunately, Bill was (adversely) affected by Jane reading a letter from his
girlfriend’

§®%WB®fl,Vaym.f79—Kfiiwfiéfifibhto

Huyu no samui hi no asa, John-wa Mary ni sinsitu no mado-o akerareta.

‘In the cold winter moming John was (adversely) affected by Mary opening his
bedroom windows’

§®§wawm,vayu.iru—KrotE§®%&%Hehto
Huyu no samui hi no asa, John-wa Mary ni yotte sinsitu no mado-o akerareta.
(See #22 above.)

emetnmmwawn,93yu.xru—uaioaemnent.
Haru no totemo atatakai hi no asa, J ohn-wa Mary ni sinsitu no mado-o akerareta.
‘In the pleasant spring morning John was (adversely) affected by Mary opening his
bedroom windows’

fiTU—fi.%®fifl2fifibfitflfwoto
Mary-we mukasi no kare ni 2-zikan mo matarete komatta.

‘Mary was upset to be affected by having been waited for as many as two hours by
her ex-boyfriend’

270

 

 

%‘f79~fl,gwfifli012fi%bfitflffioto

Mary-wa mukasi no kare ni yotte 2-zikan mo matarete komatta.
(See #25 above.)

N*xrv—m.Ewfikzfimbfitnt5nbmoto

Mary-wa mukasi no kare ni 2-zikan mo matarete uresikatta.

‘Mary was pleased to be affected by having been waited for as many as two hours by

her ex-boyfriend’

a;fifia:em.fifik$m.mtfifiutw6hto

Zannen na koto ni, Tokyo University-wa Tanaka kyoozyu ni yamerareta.
‘Regrettably, Tokyo University was affected by being quit by Prof. Tanaka’

29*ififittir:el:. 1335:4413. warau/31422135372.

Zannen na koto ni, Tokyo University-wa Tanaka kyoozyu ni yotte yamerareta.
(See #28 above.)

mismazeu.§ax$e.aeaaueeenn.

31

32

Saiwai na koto ni, Tokyo University-wa Tanaka kyoozyu ni yamerareta.
“Fortunately, Tokyo University was affected by being quit by Prof. Tanaka’

:wmbwnyfin—a—m7u754m.fi—wurOTWBHTwéo
Kono atarasii computer no program-we Paul ni yotte tukurarete iru.
‘This new computer program is being made by Paul’

Cwfibwnyfin—fl—w7075bm.fi—WKWBnwao
Kono atarasii computer no program-wa Paul ni tukurarete iru.
(See #31 above.)

33 IiFPIEEUfi‘iEQJfl. 5'54 ° 571:4:0’Cfilil‘oi’b'tb‘6o

Han-tyuugoku-seehu-undoo-ga Dalai Lama ni yotte tudukerarete iru.
‘The anti-Chinese government movement is being carried on by the Dalai Lama

9

affiraunaau.#54-5+mfinenrwe.

Han-tyuugoku-seehu-undoo-ga Dalai Lama ni tudukerarete iru.
(See #33 above.)

271

 

at!

35

36

37

38

39

4O

41

42

43

 

 

4 3/5"—* “/ 1‘ ii. lilifi‘iWDAl? lCJ: o‘Cifibi’l/‘Cb ‘60
Internet-wa sekaizyuu no hitobito ni yotte tukawarete iru.
‘The intemet has been (being) used by peeple all over the world’

% V57-7t~ y i 1:. ififirlmiw infinitive.

Intemet-wa sekaizyuu no hitobito ni tukawarete iru.

‘The intemet is under the state affected by having been used by people all over
the world’

fiE®E£GG.EEK;oTi<EfléhTWEO
Mondai no zyuuyoosee-wa kokumin ni yotte yoku rikaisarete iru.
‘The importance of the issue has been fully understood by the people’

fiE®§¥HM.EEKi<EfiéhTWEO

Mondai no zyuuyoosee-wa kokumin ni yoku rikaisarete iru.
‘The importance of the issue is under the state affected by having been fully
understood by the people’

AAVybm.VI—7XETKIOT§mhtO
‘Hamlet’-wa Shakespeare ni yotte kakareta.
‘Hamlet was written by Shakespeare’

AA 1/ ‘7 b 11. ‘73::— 2 X 6711;755:1172,
‘Hamlet’-wa Shakespeare ni kakareta.
(See #39 above.)

£§fi.%§fiiof%b6nto
Kaigi-ga gityoo ni yotte hazimerareta.
‘A meeting was begun by a chairperson’

amfl.$emneent.
Kaigi-ga gityoo ni hazimerareta.
(See #41 above.)

fix:7®fifi,NATO$K;oT&§énto
Bosnia no mati-ga NATO gun ni yotte koogeki sareta.
‘A town in Bosnia has been attacked by NATO’

272

45

46

47

48

49

50

t
51

*
52

m2:7®fifl,NATO$K&¥énto
Bosnia no mati-ga NATO gun ni koogeki sareta.
‘A town in Bosnia is under the state affected by having been attacked by NATO’

aaaam.a@oxxauroraaat.
Kimitu syorui-ga tekikoku no spy ni yotte nusumareta.
‘The secret documents have been stolen by a spy from the enemy country’

fifigfifl.fi@wzfltflfiinto

Kimitu syorui-ga tekikoku no spy ni nusumareta.

‘The secret documents are under the state affected by having been stolen by a spy
from the enemy country’

a—uyfinfifl.e»uroraaantme.
Europe ryokoo-ga Bill ni yotte keekaku sarete iru.
‘A trip to Europe is being planned by Bill’

EWKlOT.3—UVNMfiE%E3nTW5°
Bill ni yotte Europe ryokoo-o keekaku sarete iru.
(See #47 above.)

a—Uyfififid.6wfi%@éhTWEo
Europe ryokoo-wa Bill-ga keekaku sarete iru.
(See #47 above.)

iLW%fl.E~bWXKioTt<éb¢Bnto
Utukusii uta-ga Beatles ni yotte takusan tukurareta.
‘Many beautiful songs were made by the Beatles’

E—bWXKioT.%LW$&t<EAWBfltO
Beatles ni yotte utukusii uta-o takusan tukurareta.
(See #50 above.)

ibwflfl.3~beflt<ébWBhto
Utukusii uta-wa Beatles-ga takusan tukurareta.
(See #50 above.)

273

 

 

w xru—u.yayuwonmasne.
Mary-wa John ni itu mo soodan sareru.
‘Mary is affected by being always consulted by John’

a:
54 2‘ 7’ 9 ~11. ‘73 V7D§L\Oisffiffiéfléo
Mary-wa J ohn-ga itu mo soodan sareru.
(See #53 above.)

55 ‘7‘: — Vii. Ell/1C1") 73> 9 iii 35 317:,
Jane-wa Bill ni sukkari damasareta.
‘Jane was affected by being completely deceived by Bill’

*
% VI—ym.ewfitom0fiiénto
Jane-wa Bill ga sukkari damasareta.
(See #55 above.)

m rau.:33u$oaeaent.
Tom-wa kodomo ni kuruma no mado-o warareta.
‘Tom was affected by children breaking his car windows

,

t
% ram.:33flfiwaafiento
Tom-wa kodomo-ga kuruma no mado-o warareta.
(See #57 above.)

*
w rau.$pam:enuaent.
Tom-wa kuruma no mado-ga kodomo ni warareta.
(See #57 above.)

60 871/11. i/“sn— VIC‘EUDEAE 753%] But,
Bill-wa Jane ni sono himitu-o sirareta.
‘Bill was affected by Jane knowing the secret’

*
m B»d.91—yfi%wflméflgnto
Bill-wa Jane-ga sono himitu-o sirareta.
(See #60 above.)

274

*
62

B»d,%®%fiflVx—Vfifi6hto
Bill-wa sono himitu-ga Jane ni sirareta.
(See #60 above.)

Distractors

63

64

65

66

67

68

69

70

Vayfi.L¥K:VEn—5~EE5.
J ohn-ga zyoozu ni computer-o tukau.
‘John uses a computer skillfully’

VaVK.L$K:VEn—a—3E5o
John ni zyoozu ni computer-o tukau.
(See #63 above.)

Ell/hi. J: < E $33n§537b>60
Bill-ga yoku nihongo-ga wakaru.
‘Bill understands Japanese well’

1:714:11. J: < B rename.
Bill ni wa yoku nihongo-ga wakaru.
(See #65 above.)

1511/75. .1: < B $33éfiin>5o
Bill-ga yoku nihongo-o wakaru.
(See #65 above.)

Ell/(:11. J: < B 71:33 533%..
Bill ni wa yoku nihongo-o wakaru.
(See #65 above.)

EDI/7.“, :1 .‘/ t°1—5“§fl§iéo
Bill-ga computer-o tukaeru.
‘Bill can use a computer’

www.myEJ—a—flfiiéo
Bill-ga computer-ga tukaeru.
(See #69 above.)

275

 

 

71

*
72

73

a:
74

75

t
76

77

*
78

79

t
80

Ewnm,nyfia—5—fifiiéo
Bill ni wa computer-ga tukaeru.
(See #69 above.)

Ewuu,:yfin—2—2Eiéo
Bill ni wa computer-o tukaeru.
(See #69 above.)

flfl.75yxwufififitmo
Watasi-wa France no eega—ga mitai.
‘I want to see a French movie’

flu,7§y20&@flfitmo
Watasi ni France no eega—ga mitai.
(See # 73 above.)

279—m.—Acamaafi33fibt.
Mary-wa hito-ri de sizuka na kooen-o sanpo sita.
‘Mary took a walk in a quiet park by herself

2‘ 7 U r-i‘i. *A’GfiblfoC/AI’C‘fifi 1,7120

Mary-wa hito-ri de sizuka na kooen de sanpo sita.

(See #75 above.)

éfiiis 1 filliifiébo 1‘1 ”60
Kaigi-wa l-zi ni wa owatte iru.
‘The meeting will have been over at 1 o’clock’

éfim.%&1fific%botwéo

Kaigi-wa ato 1-zikan de owatte iru.
‘The meeting will have been over in an hour’

1‘) a Viix %) 5 %@T§%E Lite
J ohn-wa moo sono syukudai-o sita.
‘John has already done the homework’

/ a Vii. 175%0515152 bit...

276

 

.nf‘

J ohn-wa mada sono syukudai-o sita.
‘John has done the homework yet’

81 VI—I/fl. f E ‘b lCZF’Stinfiii‘f:o
Jane-ga kodomo ni hon-o yomaseta.
‘Jane let her child read a book’

i
& V:—yfi.¥£ba$afiiuto
J ane-ga kodomo-o hon-o yomaseta.
(See #81 above.)

m 279~a.msw%emwat.
Mary-ga tiisai otooto-o nakaseta.
‘Mary made her little brother cry’

84* 2 7 9 ~73. Ira magician—137E.
Mary-ga tiisai otooto ni nakaseta.
(See #83 above.)

277

 

 

APPENDIX K

Grammaticality judgment disagreements of linguistically naive NSs with a breakdown of

syntactic and semantic sentence types

278

 

 

 

Table 33: Grammaticality judgment disagreements of linguistically naive N83
with a breakdown of syntactic and semantic sentence types

 

Sentence T ypesa

SYNTAX
°Proper passivization

 

°*- Case absorption w/+ 0-role
suppression

-*- 0-role suppression w/+ case
absorption

 

-*- O-role suppression

 

-*+ Passivization
-*- Passivization

SEMANTICS
'Adversative reading
“Non-adversative reading

 

oProgressive reading w/iru-marking
“Progressive reading w/iru-marking
0Perfective reading w/iru-marking

 

-Past reading w/ta-marking
“Past reading w/ta-marking
-Perfective reading w/ta-marking

 

“Adversative reading w/ni yotte-
marking

 

“Personal viewpoint w/Japanese
native verb

-Personal viewpoint w/Japanese
native verb

°Objective viewpoint w/Sino-
Japanese verb

 

0*Personal viewpoint w/affective
verb
~Personal viewpoint w/affective verb

 

Passive T ypesb

Ni yotte Ni direct Ni indirect

Ni yotte Ni direct Ni indirect

X X (w/tr.)

X

a. . . .
The markings of + and — 1n the syntax and semantics colurrms represent the
presence or absence of a relevant feature, whereas the presence or absence of the

asterisk attached to each sentence type stands for its predicted grammatical and

ungrammatical status.

' Each x indicates a single occurrence of judgment disagreement with the prediction.

279

 

 

 

APPENDIX L

Grammaticality judgment disagreements of linguistically sophisticated N85 with a breakdown of

syntactic and semantic sentence types

280

 

 

Table 34: Grammaticality judgment disagreements of linguistically
sophisticated N85 with a breakdown of syntactic and semantic sentence types

Sentence T ypesa

SYNTAX
-Proper passivization

 

-*- Case absorption w/+ O-role
suppression

-*- O-role suppression w/+ case
absorption

 

°*- 0-role suppression

 

-*+ Passivization
-*- Passivization

SEMANTICS

oAdversative reading
“Non-adversative reading

 

-Progressive reading w/iru-marking
“Progressive reading w/iru-marking
0Perfective reading w/iru-marking

 

°Past reading w/ta-marking
“Past reading w/ta-marking
°Perfective reading w/ta-marking

 

“Adversative reading w/ni yotte-
marking

 

-Personal viewpoint w/Japanese
native verb

“Personal viewpoint w/Japanese
native verb

-Objective viewpoint w/Sino-
Japanese verb

 

0* Personal viewpoint w/affective
verb
-Personal viewpoint w/affective verb

 

Myotte

XX

Ni yotte

Passive T ypesb

Ni direct
x

XX

Ni direct

XX

Ni indirect
x

x
Ni indirect

x X (w/intr.)
X (w/tr.)

X (w/intr.)
x x (w/tr.)

a. . . .
The markings of + and — 1n the syntax and semantics columns represent the
presence or absence of a relevant feature, whereas the presence or absence of the

asterisk attached to each sentence type stands for its predicted grammatical and

ungrammatical status.

' Each x indicates a single occurrence of judgment disagreement with the prediction.

281

 

 

APPENDIX M

Grammaticality judgment inconsistencies of the linguistically naive NS group with a

breakdown of syntactic and semantic sentence types

282

 

Table 35: Grammaticality judgment inconsistencies of the linguistically naive
NS group with a breakdown of syntactic and semantic sentence types

Sentence T ypesa

SYNTAX
~Proper passivization

 

°*- Case absorption w/+ O-role
suppression

-*- 0-role suppression w/+ case
absorption

 

-*- O-role suppression

 

-*+ Passivization
0*- Passivization

SEMANTICS
-Adversative reading
”Non-adversative reading

 

-Progressive reading w/iru-marking
“Progressive reading w/iru-marking
oPerfective reading w/iru-marking

 

oPast reading w/ta-marking
“Past reading w/ta-marking
~Perfective reading w/ta-marking

 

“Adversative reading w/ni yotte-
marking

 

-*Personal viewpoint w/Japanese
native verb

°Personal viewpoint w/Japanese
native verb

-Objective viewpoint w/Sino-
Japanese verb

 

0*Personal viewpoint w/affective
verb
-Personal viewpoint w/affective verb

 

Passive T ypesb

Ni yotte Ni direct Ni indirect

Ni yotte Ni direct Ni indirect

#

 

a. . . .

The markings of + and — 1n the syntax and semantics colurrms represent the
presence or absence of a relevant feature, whereas the presence or absence of the
asterisk attached to each sentence type stands for its predicted grammatical and

ungrammatical status.

' Each # represents an occru'rence of judgment inconsistency in a relevant sentence

type.

283

 

 

APPENDIX N

Grammaticality judgment inconsistencies of the linguistically sophisticated NS group

with a breakdown of syntactic and semantic sentence types

 

284

 

Table 36: Grammaticality judgment inconsistencies of the linguistically
sophisticated NS group with a breakdown of syntactic and semantic sentence

types

Sentence T ypesa

SYNTAX
~Proper passivization

 

-*- Case absorption w/+ O-role
suppression

-*- O-role suppression w/+ case
absorption

 

°*- O-role suppression

 

-*+ Passivization
'*- Passivization

SEMANTICS
-Adversative reading
“Non-adversative reading

 

oProgressive reading w/iru-marking
“Progressive reading w/iru-marking
-Perfective reading w/iru-marking

 

°Past reading w/ta-marking
“Past reading w/ta-marking
-Perfective reading w/ta-marking

 

0*Adversative reading w/ni yotte-
marking

 

-Personal viewpoint w/Japanese
native verb

“Personal viewpoint w/Japanese
native verb

-Objective viewpoint w/Sino-
Japanese verb

 

“Personal viewpoint w/affective
verb
-Persona1 viewpoint w/affective verb

 

Passive T ypesb

Ni yotte Ni direct Ni indirect
#

#
Ni yotte Ni direct Ni indirect

# (w/tr.)

a. . . .
The markings of + and — 1n the syntax and semantics columns represent the
presence or absence of a relevant feature, whereas the presence or absence of the

asterisk attached to each sentence type stands for its predicted grammatical and

ungrammatical status.

' Each # represents an occurrence of j udgment inconsistency in a relevant sentence

type.

285

BIBLIOGRAPHY

286

BIBLIOGRAPHY

Alfonso, A. (1980). Japanese Language Patterns: a structural approach (Vol. I & 11).
Tokyo: Sophia University L.L.: Center of Applied Linguistics.

Baker, M., Johnson, K., & Roberts, 1. (1989). Passive arguments raised. Linguistic
Inquiry, 20(2), 219-251.

Bardovi-Harlig, K., & Reynolds, D. W. (1995). The role of lexical aspect in the
acquisition of tense and aspect. TESOL Quarterly, 29(1), 107-131.

 

Birdsong, D. (1989). Metalinguistics performance and interlanguage competence. New
York: Springer.

Bley-Vroman, R. W., Felix, S. W., & Ioup, G. L. (1988). The accessibility of Universal
Grammar in adult language learning. Second Language Research, 4, 1-32.

Bradac, J. J ., Martin, L. W., Elliott, N. D., & Tardy, C. H. (1980). On the neglected side
of linguistic science: Multivariate studies of sentence judgment. Linguistics,
18(11/12), 967-995.

Carroll, J. M., Bever, T. G., & Pollack, C. R. (1981). The non-uniqueness of linguistic
intuitions. Language, 57, 368-382.

Chaudron, C. (1983). Research on metalinguistic judgments: A review of theory,
methods, and results. Language Learning, 33(3), 343-377.

 

Chomsky, N. (1962). Various discussion sessions. In A. A. Hill (Ed.), Third Texas
conference on problems of linguistic analysis in English (pp. 22-33). Austin:
University of Texas.

Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Holland: Foris
Publications.

287

Chomsky, N. (1986). Knowledge of Language: Its Nature, Origin, and Use. New York:
Praeger.

Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press.

Chomsky, N., & Lasnik, H. (1995). The theory of principles and parameters. In N.
Chomsky (Ed), The Minimalist Program (pp. 13-127). Cambridge, MA: MIT Press.

Christine, K., & Lantolf, J. P. (1992). The ontological status of learner grammaticality
judgments in UG approaches to L2 acquisition. Rassegna Italiana di Linguistica
Applicata, 24(3), 31-57.

Clancy, P. M. (1985). The acquisition of Japanese. In D. I. Slobin (Ed), The
crosslinguistic study of language acquisition (V 01. 1, pp. 373-524). Hillsdale, NJ:
Lawrence Erlbaum Associates, Publishers.

Comrie, B. (1976). Aspect: An introduction to the study of verbal aspect and related
problems: Cambridge University Press.

Comrie, B. (1981). Aspect and Voice: Some reflections on perfect and passive. In P. J.
Tedeschi & A. Zaenen (Eds), Syntax and Semantics (Vol. 14, pp. 65-78). New York:
Academic Press.

Coppieters, R. (1987). Competence differences between native and near-native speakers.
Language, 63(3), 544-573.

Cowan, R., & Hatasa, Y. A. (1994). Investigating the validity and reliability of native
speaker and second-language learner judgments about sentences. In E. E. Tarone, S.
M. Gass, & A. D. Cohen (Eds), Research Methodology in Second-Language
Acquisition (pp. 287-302). Hillsdale, NJ.

Cowart, W. (1997). Experimental syntax: Applying objective methods to sentence
judgments. Thousand Oaks: CA: Sage Publications.

Davies, W. D., & Kaplan, T. I. (1998). Native speaker vs. L2 learner grammaticality
judgments. Applied Linguistics, 19(2), 183-203.

288

Doughty, C. (1991). Second language instruction does make a difference: Evidence from
an empirical study of SL relativization. Studies in Second Language Acquisition, 13,
43 1-469.

Doughty, C., & Williams, J. (1998). Focus on form in classroom second language
acquisition. Cambridge, UK: Cambridge University Press.

Ellis, R. (1990). Grammaticality judgments and learner variability. Proceedings of the
tenth meeting of the Second Language Research Forum (pp. 25-60), Eugene: OR.

Ellis, R. (1991). Grammaticality judgments and second language acquisition. Studies in
Second Language Acquisition, 13, 161-186.

Feng, F. (1993). On influences of the mother tongue - Chinese language - on the learning
processes of the Japanese passive voice. Japanese Journal of Educational
Psychology, 41, 388-398.

Ferreira, F. (1994). Choice of passive voice is affected by verb type and animacy. Journal
of Memory and Language, 33, 715-736.

Fox, D., & Grodzinsky, Y. (1998). Children's passive: A view from the by-phrase.
Linguistic Inquiry, 29(2), 31 1-332.

Fraser, B. (1971). An analysis of "even" in English. In C. J. Fillmore & T. D.
Langendoen (Eds), Studies in linguistic semantics (pp. 151-178). New York: Holt:
Rinehart and Winston.

 

Gass, S. M. (1979). Language transfer and universal grammatical relations. Language
Learning, 29(2), 327-344.

Gass, S. M. (1983). The development of L2 intuitions. TESOL Quarterly, 1 7, 273-291.

Gass, S. M. (1994). The reliability of second language grammaticality judgments. In E. E.
Tarone, S. M. Gass, & A. D. Cohen (Eds), Research Methodology in Second-
Language Acquisition (pp. 303-322). Hillsdale, NJ: Lawrence Erlbaum Associates.

289

 

Gass, S. M., & Selinker, L. (2001). Second language acquisition: An introductory course
(2nd ed). Mahwah, NJ: Lawrence Erlbaum Associates, Publishers.

Givon, T. (1981). Typology and functional domains. Studies in Language, 5(2), 163-193.

Goss, N., Zhang, Y.-H., & Lantolf, J. P. (1994). Two heads may be better than one:
Mental activity in second-language grammaticality judgments. In E. E. Tarone, S. M.
Gass, & A. D. Cohen (Eds), Research Methodology in Second-Language
Acquisition (pp. 263-286). Hillsdale, NJ: Lawrence Erlbaum Associates.

Greenbaum, S. (1977a). Contextual influence on acceptability judgments. International
Journal of Psycholinguistics, 6, 5-11.

Greenbaum, S. (1977b). The linguist as experimenter. In F. R. Eckrnan (Ed), Current
themes in linguistics: Bilingualism, experimental linguistics, and language
typologies (pp. 125-144). Washington: Hemisphere.

Greenbaum, S. (1988). Good English and the grammarian. London: Longman.
Grimshaw, J. (1990). Argument Structure. Cambridge: MA: MIT Press.

Hagiwara, H. (1993). The breakdown of Japanese passives and theta-role assignment
principle by Broca's aphasics. Brain and Language, 45, 318-339.

Hagiwara, H., & Caplan, D. (1990). Syntactic comprehension in Japanese aphasics:
Effects of category and thematic role order. Brain and Language, 38, 159-170.

Hakuta, K. (1982). Interaction between particles and word order in the comprehension
and production of simple sentences in Japanese children. Developmental Psychology,

18(1), 62-76.

Harada, K. I. (1977). The acquisition of passive, causative, and te moraw constructions in
Japanese: A case study of a two-year-old. Annual Reports, 2, 1-16.

290

Harada, S.-I. (1973). Counter equi NP deletion. Annual Bulletin of the Research Institute
of Logopedics and Phoniatrics, 7, 113-147.

Hedgcock, J. (1993). Well-formed vs. ill-formed strings in L2 metalingual tasks:
specifying features of grammaticality judgments. Second Language Research, 9(1),
1-21.

Hirakawa, M. (1999). L2 acquisition of Japanese unaccusative verbs by speakers of
English and Chinese. In K. Kanno (Ed), The Acquisition of Japanese as a Second
Language (pp. 89-113). Amsterdam: John Benjamins Publishing Company.

Hirakawa, M. (2001). L2 acquisition of Japanese unaccusative verbs. Studies in Second
Language Acquisition, 23(2), 221-245.

Hoshi, H. (1991). The generalized projection principle and its implications for passive
constructions. Journal of Japanese Linguistics, 13, 53-89.

Hoshi, H. (1993a). Excorporation in syntax and in LP: The case of Romance causatives
and Japanese passives. MIT Working Papers in Linguistics, 20, 113-127.

Hoshi, H. (1993b). On the syntactic properties of the passive morpheme in Japanese. In
S. Choi (Ed), Japanese/Korean Linguistics (V 01. 3, pp. 137-153). Stanford, CA:
Center for the Study of Language and Information.

Hoshi, H. (1993c). The role of Pro in the ni passive in Japanese. In P. M. Clancy (Ed),
Japanese/Korean linguistics (V 01. 2, pp. 355-3 74). Stanford, CA: Center for the
Study of Language and Information.

Hoshi, H. (1994a). Passive, causative, and light verbs: A study on theta role assignment.
The University of Connecticut.

Hoshi, H. (1994b). Theta-role assignment, passivization, and excorporation. Journal of
East Asian Linguistics, 3, 147-178.

Hoshi, H. (1999). Passives. In N. Tsujimura (Ed), The Handbook of Japanese Linguistics
(pp. 191-235). Cambridge: MA: Blackwell Publishers.

291

Howard, 1., & Niyekawa-Howard, A. M. (1976). Passivization. In M. Shibatani (Ed),
Syntax and Semantics: Japanese Generative Grammar (Vol. 5, pp. 201-237). New
York: Academic Press, Inc.

Iitaka, K. (1989). Development of passive sentence with case particles: A study on 2 to 4
year old Japanese children. RIEE C Report (Tokyo Gakugei University, Division for
the Education of Speech and Language Handicapped), 38, 29-39.

Inoue, K. (1976). Henkei-bunpoo to Nihon-go (Vol. Zyoo). Tokyo: Taisyuukan.

J acobsen, W. M. (1992). The transitive structure of events in Japanese. Tokyo: Kurosio
Publiehrs.

J aeggli, O. A. (1986). Passive. Linguistic Inquiry, 1 7(4), 587-622.

Japan Foundation, & Association of lntemational Education, J. (1994). Nihongo noryoku
siken shutsudai kijun, Japanese language proficiency test: test content specifications .
Tokyo: Bonj insha.

Japan Foundation, & Association of International Education, J. (1999). The 1998
Japanese language proficiency test . Tokyo: Bonjinsha.

Johnson, J. S., Shenkrnan, K. D., Newport, E. L., & Medin, D. L. (1996). Indeterminacy
in the grammar of adult language learners. Journal of Memory and Language, 35,
335-352.

J orden, E. H. with M. Noda. (1987). Japanese: The Spoken Language (Vol. 1). New
Haven: Yale University Press.

J orden, E. H. with M. Noda. (1988). Japanese: The Spoken Language (V 01. 2). New
Haven: Yale University Press.

J orden, E. H. with M. Noda. (1990). Japanese: The Spoken Language (V 01. 3). New
Haven: Yale University Press.

292

J uffs, A. (1996). Learnability and the lexicon: Theories and second language acquisition
research. Amsterdam: John Benj amins Publishing Company.

Kageyama, T. (1993). Bunpoo to gokeisei (Grammar and Word Formation). Tokyo:
Hitsuji Shoboo.

Kageyarna, T. (1996). Dooshi-imiron (Verb Semantics). Tokyo: Kuroshio.

Keenan, E. L., & Comrie, B. (1977). Noun phrase accessibility and Universal Grammar.
Linguistic Inquiry, 8, 63-99.

Kitagawa, Y. (1986). Subjects in Japanese and English. University of Massachusetts,
Amherst.

Klaiman, M. H. (1987). Aktionsart, semantics, and function in the Japanese "Passive".
Studies in Language, 11(2), 401-434.

Kuno, S. (1973). The structure of the Japanese language. Cambridge, MA: MIT Press.

Kuno, S. (1983). Sin Nihon Bunpoo Kenkyuu (A New Study of Japanese Grammar).
Tokyo: Kuroshio.

Kuno, S. (1986). The meaning of passive sentences: A reply to Kuroda. Nihongogaku,
5(2), 70-87.

Kuroda, S.-Y. (1965). Generative grammatical studies in the Japanese language. , MIT,
Cambridge, MA.

Kuroda, S.-Y. (1978). Case marking, canonical sentence patterns, and counter equi in
Japanese (A preliminary survey). In J. Hinds & 1. Howard (Eds), Problems in
Japanese syntax and semantics . Tokyo: Kaitakusha.

Kuroda, S.-Y. (1979). On Japanese passives, Explorations in linguistics: Papers in honor
of Kazuko Inoue . Tokyo: Kenkyusha.

293

Kuroda, S.-Y. (1985). Ukemi nituite no Kuno-setu o Kaisyaku-suru (A critical review of
Kuno's theory of Japanese passives). Nihongogaku, 10, 69-76.

Lasnik, H., & Fiengo, R. (1974). Complement object deletion. Linguistic Inquiry, 5(4),
535-571.

Leow, R. P. (1996). Grammaticality judgment tasks and second-language development.
Georgetown University Round Table on Languages and Linguistics, 126-139.

Levelt, W. J. M. (1974). Formal grammars in linguistics and psycholinguistics (Vol. 3).
The Hague: Mouton.

Mandell, P. B. (1999). On the reliability of grammaticality judgment tests in second
language acquisition research. Second Language Research, 15(1), 73-99.

Marantz, A. (1984). On the Nature of Grammatical Relations. Cambridge, MA: MIT
Press.

Martin, S. E. (1975). A Reference Grammar of Japanese. New Haven: Yale University
Press.

Masuoka, T. (1982). On the semantics of Japanese passives. Gengo Kenkyu, 82, 48-64.

McCawley, N. A. (1972). On the treatment of Japanese passives. In P. M. Peranteau, J. N.
Levi, & G. C. Phares (Eds), Papers from the Eighth Regional Meeting of the
Chicago Linguistic Society (pp. 256-270). Chicago: IL: University of Chicago,
Department of Linguistics.

Miyagawa, S. (1989). Structure and case marking in Japanese (Vol. 22). San Diego:
Academic Press, Inc.

Miyagi, K. (1996). Passive, aflective, and causative constructions in Japanese. ,
Georgetown University, Washington DC.

294

Miyagi, K. (1999). Ni/ni yotte variation in Japanese direct passives: A syntactic,
pragmatic, and historical account. Paper presented at The ninth J apanese/Korean
Linguistics Conference, The Ohio State University.

Miyamoto, E. T., Wexler, K., Aikawa, T., & Miyagawa, S. (1998). Case-dropping and
unaccusatives in Japanese acquisition. Paper presented at The 23rd Annual Boston
University Conference on Language Development, Boston: MA.

Munnich, E., Flynn, S., & Martohardjono, G. (1994). Elicited imitation and
grammaticality judgment tasks: What they measure and how they relate to each other.
In E. E. Tarone, S. M. Gass, & A. D. Cohen (Eds), Research methodology in
second-language acquisition (pp. 227-243). Hillsdale, NJ: Lawrence Erlbaum
Associates.

Murphy, V., A. (1997). The effect of modality on a grammaticality judgment task.
Second Language Research, 13(1), 34-65.

Nabei, T., & Busch, M. (1999). Reviews: The Japan Foundation and Association of
International Education, Japan: Japanese language proficiency test. Applied
Linguistics, 20(1), 127-131.

Neter, J ., Wasserman, W., & Kutner, M. H. (1990). Applied Linear Statistical Models:
Regression, Analysis of Variance, and Experimental Designs. Homewood, IL: Irwin.

Okuda, Y. (1978a). Asupekuto no kenkyuu o megutte, I (On the study of aspect, I).
Kokugo Kyooiku, 53, 33-44.

Okuda, Y. (1978b). Asupekuto no kenkyuu o megutte, H (On the study of aspect, II).
Kokugo Kyooiku, 54, 14-27.

Otsu, Y. (1999). First language acquisition. In N. Tsujimura (Ed), The Handbook of
Japanese Linguistics (pp. 378-397). Malden: MA: Blackwell Publishers.

Ross, J. R. (1979). Where's English? In C. J. Fillmore, D. Kempler, & W. S.-Y. Wang
(Eds), Individual differences in language ability and language behavior (pp. 127-
163). New York: Academic Press.

295

 

 

Ryuzaki, M., & Ito, T. (1999). Knowledge of the argument and the phrase structure of
passives in children with hearing impairments. T okusyu Kyooikugaku Kenkyuu (The
Journal of Special Education), 36(4), 23-30.

Saito, M. (1982). Case marking in Japanese: A preliminary study. ms.

Sano, K. (1977). An experimental study on the acquisition of Japanese simple and cleft
sentences. Descriptive and applied linguistics(10), 213-233.

Sano, T., Endo, M., & Yamakoshi, K. (2001). Developmental issues in the acquisition of
Japanese unaccusatives and passives. In A. H.-J. Do, L. Dominguez, & A. J ohansen
(Eds), BUCLD 25 Proceedings (Vol. 2, pp. 668-683). Somerville: MA: Cascadilla

Press.

Schachter, J. (1974). An error in error analysis. Language Learning, 24, 205-214.

Schachter, J. (1989). Testing a proposed universal. In S. M. Gass & J. Schachter (Eds),
Linguistic perspectives on second language acquisition (pp. 73-88). Cambridge:
Cambridge University Press.

Schachter, J., & Yip, V. (1990). Grammaticality judgments: Why does anyone object to
subject extraction? Studies in Second Language Acquisition, 12, 379-392.

Schfitze, C. T. (1996). The empirical base of linguistics. Chicago: The University of
Chicago Press.

Shibatani, M. (1990). The language of Japan. Cambridge: Cambridge University Press.

Shirai, Y. (1998). Where the progressive and the resultative meet imperfective aspect in
Japanese, Chinese, Korean, and English. Studies in Language, 22(3), 661-692.

Shirai, Y. (2000). The semantics of the Japanese imperfective -teiru: An integrative
approach. Journal of Pragmatics, 32, 327-361.

296

Shirai, Y., & Andersen, R. W. (1995). The acquisition of tense-aspect morphology: A
prototype account. Language, 71, 743-762.

Shirai, Y., & Kurono, A. (1998). The acquisition of tense-aspect marking in Japanese as a
second language. Language Learning, 48(2), 245-279.

Snow, C., & Meijer, G. (1977). On the secondary nature of syntax intuitions. In S.
Greenbaurn (Ed), Acceptability in language (pp. 163-177). The Hague, The
Netherlands: Mouton Pusblishers.

Sobin, N. (1985). Case assignment in Ukrainian morphological constructions. Linguistic
Inquiry, 16, 649-662.

Sorace, A. (1988). Linguistic intuitions in interlanguage development: The problem of
indeterminacy. In J. Pankhurst, M. Sharwood Smith, & P. Van Buren (Eds),
Learnability and second languages . Dordrecht, Holland: Foris.

Specs, H. (1992). Adversity in Japanese passives: An empirical study of native speakers
of Japanese. Unpublished MA Thesis, the University of South Carolina.

Spencer, J. (1973). Differences between linguists and nonlinguists in intuitions of
grammaticality-acceptability. Journal of Psycholinguistic Research, 2(2), 83-98.

Suzuki, T. (1998). Children's comprehension of Japanese passives: A different
perspective. In E. V. Clark (Ed), The Proceedings of the T wenty-ninth Annual Child
Language Research Forum (V 01. 29, pp. 91-100). Stanford: CA: The Center for the
Study of Language and Information.

Takezawa, K. (1987). A configurational approach to Case marking in Japanese.
Unpublished PhD dissertation, University of Washington.

Tanaka, M. (1992). Examination of failure to use the passive by Japanese language
learners. Language Research Bulletin (lntemational Christian University, Tokyo),
6(1), 152-168.

297

Tanaka, M. (1993). A study of J SL students' Interlanguage: An analysis of the use of the
passive in complex/compound sentences. Language Research Bulletin (International
Christian University, Tokyo), 8, 87-117.

Teramura, H. (1982). Nihongo no sintakusu to imi (Syntax and meaning in Japanese)
(Vol. 1). Tokyo: Kuroshio Publishings.

Teramura, H. (1984). Nihongo no sintakusu to imi (Syntax and meaning in Japanese)
(V 01. 2). Tokyo: Kuroshio Publishers.

Tsujimura, N. (1996). An introduction to Japanese linguistics. Cambridge, MA: Basil
Blackwell.

Valian, V. (1982). Psycholinguistic experiment and linguistic intuition. In T. W. Simon &
R. J. Scholes (Eds), Language, mind, and brain (pp. 179-188). Hillsdale: New
Jersey: Erlbaum.

Washio, R. (1989/1990). The Japanese passive. The Linguistic Review, 6, 227-263.

Watabe, M., Brown, C., & Ueta, Y. (1991). Transfer of discourse function: Passives in
the writings of ESL and J SL learners. International Review of Applied Linguistics in
Language Teaching, 29(2), 115-134.

White, L. (1989). Universal Grammar and Second Language Acquisition. Amsterdam:
John Benj amins Publishing Company.

Wierzbicka, A. (1979). Are grammatical categories vague or polysemous? (The Japanese
'adversative' passive in a typological context). Papers in Linguistics, 12, 111-162.

Yang, J. C., & Akahori, K. (1998). Error analysis in Japanese writing and its
implementation in a computer assisted language learning system on the World Wide
Web. CALICO, 15(1-3), 47-66.

Yoon, J. H.-S. (1990). Theta theory and the grammar of inalienable possession
construction in Korean, Mandarin, and French. Paper presented at the North East
Linguistic Society.

298

Yoshimoto, K. (2000). Tense and aspect in Japanese and English (2., revised edition ed).
Frankfurt am Main: Peter Lang.

299

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII

111111111111!ll'IIWI/llll/lWW

293 O 372 0547

          

3