Li Li: Grammar in Cross-Linguistic Perspective

li57 In this collection of papers on syntax, semantics
and pragmatics, linguists specialising in the

li
57 li57 Linguistic Insights
Studies in Language and Communication
Teruhiro Ishiguro & Kang-kwong Luke (eds) • Grammar in Cross-Linguistic Perspective

study of Japanese and Chinese offer fresh ideas
and insights on the theme of grammatical cat-
egories and structure from a comparative per-
spective. Against the background of theoretical
developments in recent years and individual
studies of Japanese, Chinese and English gram-
mar, the papers in this volume are devoted to
new in-depth treatments of distinctive aspects
of Chinese and Japanese grammar informed by
influential theoretical frameworks of the day,
including cognitive grammar, construction gram-
mar, information structure, grammaticalization
theory, and linguistic typology. Topics of investi-
gation include compounding, verb complemen-
tation, tense and aspect, as well as a range of
word order phenomena, such as passive construc-
tions, focus-fronting, and right dislocation.
Teruhiro Ishiguro & Kang-kwong Luke (eds)
Grammar in
Teruhiro Ishiguro is Professor and Dean of the
Faculty of English and American Language and
Cross-Linguistic
Literature, Tokushima Bunri University, Japan, and
Kang-kwong Luke is Professor of Linguistics,
Nanyang Technological University, Singapore.
Perspective
Both of them specialise in the interfaces be-
tween syntax, semantics, and pragmatics, and
have a common interest in grammatical studies
The Syntax, Semantics, and Pragmatics
of Japanese, Chinese and English from a com- of Japanese and Chinese
parative perspective.
ISBN 978-3-03911-445-0
www.peterlang.com Peter Lang

li57 In this collection of papers on syntax, semantics
and pragmatics, linguists specialising in the
li
57 li57 Linguistic Insights
Teruhiro Ishiguro & Kang-kwong Luke (eds) • Grammar in Cross-Linguistic Perspective

study of Japanese and Chinese offer fresh ideas
and insights on the theme of grammatical cat-
egories and structure from a comparative per-
spective. Against the background of theoretical
developments in recent years and individual
studies of Japanese, Chinese and English gram-
mar, the papers in this volume are devoted to
new in-depth treatments of distinctive aspects
of Chinese and Japanese grammar informed by
influential theoretical frameworks of the day,
including cognitive grammar, construction gram-
mar, information structure, grammaticalization
theory, and linguistic typology. Topics of investi-
gation include compounding, verb complemen-
tation, tense and aspect, as well as a range of
word order phenomena, such as passive construc-
tions, focus-fronting, and right dislocation.
Grammar in
Teruhiro Ishiguro is Professor and Dean of the
Faculty of English and American Language and
Cross-Linguistic
Literature, Tokushima Bunri University, Japan, and
Kang-kwong Luke is Professor of Linguistics,
Nanyang Technological University, Singapore.
Perspective
Both of them specialise in the interfaces be-
tween syntax, semantics, and pragmatics, and
have a common interest in grammatical studies
of Japanese, Chinese and English from a com- of Japanese and Chinese
parative perspective.
Peter Lang
Grammar in Cross-Linguistic Perspective
Linguistic Insights
Edited by Maurizio Gotti,

University of Bergamo
Volume 57
Advisory Board
Vijay Bhatia (Hong Kong)
Christopher Candlin (Sydney)
David Crystal (Bangor)
Konrad Ehlich (Berlin / München)
Jan Engberg (Aarhus)
Norman Fairclough (Lancaster)
John Flowerdew (Hong Kong)
Ken Hyland (Hong Kong)
Roger Lass (Cape Town)
Matti Rissanen (Helsinki)
Françoise Salager-Meyer (Mérida, Venezuela)
Srikant Sarangi (Cardiff)
^
Susan Šarcević (Rijeka)
Lawrence Solan (New York)
Peter M. Tiersma (Los Angeles)
PETER LANG
Bern • Berlin • Bruxelles • Frankfurt am Main • New York • Oxford • Wien
Grammar in
Cross-Linguistic
Perspective

of Japanese and Chinese
PETER LANG
Bern • Berlin • Bruxelles • Frankfurt am Main • New York • Oxford • Wien
Bibliographic information published by die Deutsche Nationalbibliothek
Die Deutsche Nationalbibliothek lists this publication in the Deutsche National-
bibliografie; detailed bibliographic data is available on the Internet
at ‹http://dnb.d-nb.de›.
British Library Cataloguing-in-Publication Data: A catalogue record for this

book is available from The British Library, Great Britain
Library of Congress Cataloging-in-Publication Data
Grammar in Cross-Linguistic Perspective : the Syntax, Semantics, and Pragmatics

of Japanese and Chinese / Teruhiro Ishiguro & Kang-kwong Luke (eds).
pages cm. -- (Linguistic Insights-Studies in Language and Communication ; v. 57)
ISBN 978-3-03911-445-0
1. Comparative linguistics. 2. Japanese language--Grammar, Comparative--Chinese.
3. Chinese language–Grammar, Comparative--Japanese. 4. Japanese language--
Grammar. 5. Chinese language–Grammar. 6. English language--Grammar. 7.
Interlanguage (Language learning) 8. Intercultural communication. 9. Speech
acts (Linguistics) 10. Grammaticality (Linguistics) I. Ishiguro, Teruhiro, editor of
compilation. II. Luke, Kang Kwong, editor of compilation.
P207.G73 2013
495.15–dc23
2012045800
ISSN 1424-8689
ISBN 978-3-03911-445-0 pb. ISBN 978-3-0351-0576-6 eBook
© Peter Lang AG, International Academic Publishers, Bern 2012

Hochfeldstrasse 32, CH-3012 Bern, Switzerland
info@peterlang.com, www.peterlang.com, www.peterlang.net
All rights reserved.

All parts of this publication are protected by copyright.
Any utilisation outside the strict limits of the copyright law, without
the permission of the publisher, is forbidden and liable to prosecution.
This applies in particular to reproductions, translations, microfilming,
and storage and processing in electronic retrieval systems.
Printed in Switzerland
Table of Contents
TERUHIRO ISHIGURO and K.K. LUKE

Preface .............................................................................................. 7
YOICHIRO HASEBE
A Cognitive Approach to Compound Kango VNPs
in Japanese ........................................................................................ 9
GAO HUA
Tag-Questions in Mandarin Chinese ..............................................43
NORIFUMI ITO
Information in Discourse and Language as Vehicle
of Thought ......................................................................................65
HIDEMATSU MIURA
Focus-driven Semantic Reflexivity in Japanese ............................89
WINNIE CHOR
From Nonsubjective to Subjective to Intersubjective –
a Pathway of Semantic Change in Grammaticalization:
the Case of faan in Cantonese ......................................................125
YUTAKA SHINODA
Japanese Particle Na as a Marker of the Speaker’s
Subjective Judgment ‘Here and Now’ .........................................149
LEUNG WAI-MUN
Evidential Particles in Cantonese: the Case of wo3
and wo5 ........................................................................................175
6 Table of Contents
HAN YANG
Tense, Aspect and Verbal Morphemes in Chinese, Japanese
and English ...................................................................................205
MASANOBU HORIGUCHI
Japanglish in Katakana and a Comparison with Loanwords
in Cantonese .................................................................................243
WONG PING WAI

Semantic Annotation of Chinese Texts
with Message Structures Based on HowNet ................................271
Notes on Contributors ..................................................................301

7
TERUHIRO ISHIGURO / K.K. LUKE
Preface
The idea of making a collection of papers on the study of syntax,

semantics and pragmatics in Japanese and Chinese was first con-
ceived in 2005, when, as part of a student exchange programme, the
second editor led a delegation of students of the University of Hong
Kong on a visit to Tokushima Bunri University. The two editors had
already known each other for many years, but it was on this occasion
that they first had a serious discussion of the possibility of a book
publication, as a way of adding an element of research and intellec-
tual interflow to the enrichment of students’ educational experiences.
A year later, the two editors were able to construct a firm plan to set
the book project in motion.
The concept of a joint publication was also deemed particularly
feasible and attractive due to a shared interest between the teachers of
the two universities on the empirical study of spoken language and the
exploration of grammatical phenomena from a variety of new theo-
retical perspectives. The discerning reader will no doubt find within
the covers of this book ample evidence of the study of Japanese and
Chinese grammar using new methods and insights from such fields as
Cognitive Grammar, Information Structure, Natural Language Pro-
cessing and Conversation Analysis.
According to the original plan, the book was scheduled to come
out in 2008. However, due to the second editor’s relocation from Hong
Kong to Singapore, the book has taken much longer to materialize
than expected. For this, the second editor would like to express his
sincere apologies to all the authors, and heart-felt gratitude for their
patience and confidence. It is very much hoped that the book will not
only achieve its aim of promoting an exchange of ideas between the
two universities, but will also serve as a memento marking the much
treasured friendship between the two universities.
8 Preface
Last but not least, the two editors would like to thank the two univer-
sities, particularly President Murasaki of TBU, for their unfailing
support of the exchange programme. They are also grateful to their
able and diligent colleagues and graduate students for their valuable
input into this memorable and meaningful process.
9
YOICHIRO HASEBE
A Cognitive Approach to
Compound Kango VNPs in Japanese
1. Introduction
1.1 Phenomenon
This chapter deals with Japanese complex verbal noun phrases of

Chinese-origin (complex kango VNPs).1 The complex kango VNPs
examined are exemplified in (1). These examples and other complex
kango VNPs meet the syntactic and semantic conditions listed in (2).
(1) a. ᝟ሗ཰㞟 (jouhou-shuushuu, ‘information-gathering’)

b. ౯᱁㧗㦐 (kakaku-koutou, ‘price-increase’)
c. ศ๭⤫἞ (bunkatsu-touchi, ‘divide-rule’)
d. ಖᏲ⟶⌮ (hoshu-kanri, ‘maintain-administer’)
(2) a. Single semantically coherent VNPs are composed of two independent kango
words.
b. The second element of the compound has both nominal and verbal charac-
teristics.
c. The VNP as a whole has not completely lexicalized into one word, pre-
serving syntactic and semantic analyzability.
From condition (2a), instances such as those in (3) are excluded.
1 The expression ‘Chinese-origin’ is not always used accurately. Here, it roughly

means that the words in question are conventionally written in Chinese charac-
ters; they may or may not be of Chinese-origin. Thus, I prefer to use the term
kango throughout this chapter.
10 Yoichiro Hasebe
(3) a. ⩦័໬
KŠμ‰»(shuukan-ka,
(shuukan-ka,‘conventionalization’)
‘conventionalization’)
b. ᪥ᮏ〇
“ú–{•» (nihon-sei,
(nihon-sei,‘Japan-made’)
‘Japan-made’)
c. ෌ᣮᡓ
Ä’§•í (sai-chousen,
(sai-chousen, ‘re-try’)
‘re-try’)
d. ึ㣕⾜
‰”ò•s (hatsu-hikou,
(hatsu-hikou,‘first-flight’)
‘first-flight’)
The morphemes ka (‘-ation’), sei (‘-made’), sai (‘re-’), and hatsu

(‘first-’) function as suffixes rather than as independent words. Also
excluded are VNPs composed of two single kanji (Chinese charac-
ter) morphemes as exemplified in (4).
(4) a. 㦵ᢡ
œ • Ü (kossetsu,
(kossetsu,‘bone-break’)
‘bone-break’)
b. ධ㝔
“ü‰@ (nyuu-in,
(nyuu-in, ‘enter-hospital,
‘enter-hospital, hospitalized’)
hospitalized’)
c. ᡂ㛗
¬’· (sei-chou,
(sei-chou,‘become-long,
‘become-long,grow’)
grow’)
d. ど⫈
•‹’® (shi-chou,
(shi-chou,‘view
‘viewand
andlisten’)
listen’)
There are a great number of this kind of Japanese words, and they
account for a large part of the overall verbal vocabulary. Many of
these words, however, are lexicalized, and accordingly their compo-
nent morphemes are neither highly independent nor productive. In
fact, these two-character kango VNPs contribute to the formation of
complex VNPs as component structures. Thus, a brief look at the
structure of these simple kango VNPs is made in Section 2.2. The
complex kango VNPs that meet the first condition in (2) examined in
this chapter consist of two parts, each of which is composed of a
smaller unit.
As (2b) stipulates, the second element of complex kango VNPs
functions either verbally or nominally. For instance, shuushuu of
jouhou-shuushuu in (1a) can be used either as an NP or as a semanti-
cally essential part of a VP. Since complex VNPs inherit this charac-
teristic, they can be part of an argument of a VP as in (5a). On the
other hand, they can also combine with the light-verb suru (roughly
meaning ‘do’) as in (5b) to make up a full-fledged VP construction.
(5) a. ᙼ䛻᝟ሗ཰㞟䜢௵䛫䜛䚹
kare-ni jouhou-shuushuu-o makaseru
he-DAT data-gathering-ACC leave
‘(I) leave data-gathering to him.’
A Cognitive Approach to Compound Kango VNPs in Japanese 11
b. 䜲䞁䝍䞊䝛䝑䝖䛷᝟ሗ཰㞟䛩䜛䚹
intaanetto-de jouhou-shuushuu-suru
Internet-with data-gathering-suru
‘(I) do data-gathering on the Internet.’
Finally (2c) excludes a certain group of compound kango VNPs that

satisfy the first two conditions in (2). These are compound VNPs that
lack transparency and analyzability at least in the most ordinary Japa-
nese speakers’ perception. The reasons for this lack of transparency/
analyzability vary. Possible scenarios are as follows: both the form and
meaning of the complex VNP were directly borrowed from Chinese
and have been frozen since then; the archaic forms have remained
unchanged until now, and as a result many speakers barely recognize
the original meaning of the component elements; or the complex VNP
is always used in a metaphorical sense, making it difficult to evoke the
original meaning. These types of complex kango VNPs are not ad-
dressed, although some of the analyses may be applicable.
1.2 Framework
In the following sections, compound kango VNPs meeting the condi-

tions in (2) are analyzed in the framework of Cognitive Grammar
(Langacker 1990, 1991). Cognitive Grammar (CG) takes an approach
dissimilar to the more formalistic theories of linguistics such as Genera-
tive Grammar, which presupposes the existence of preprogrammed
syntactic and semantic modules inside the brain and subscribes to the
idea that syntax and semantics of language are composed of explicitly
formalized rules and constraints. On the other hand, CG emphasizes
fundamental cognitive abilities that figure not only in language but
also in basic human activities such as category perception, reference-
point construction ability, metaphorical mapping between domains,
and so forth. As opposed to the formalistic theories where gramma-
ticality of a sentence is determined based on a certain set of rules and
constraints of a binary nature, in CG, every possible sequence of
morphemes/words is evaluated according to the construability of the
concept invoked by them.
12 Yoichiro Hasebe
Past attempts to describe word formation and phrase structure

in Japanese often utilized a framework referred to as Lexical Con-
ceptual Structure (LCS).2 In fact, many studies carried out in LCS
have contributed to clarifying the syntactic and semantics structure
of Japanese (cf. Kageyama 1993, 1996). Linguistic phenomena in
Japanese, however, have many aspects that defy formalistic ap-
proaches, as is the case with English and other languages. CG, on the
other hand, abandons the presupposition that there are autonomous
syntactic and semantic modules that work independently from one’s
subjective construal and conceptualization. Thus, it is expected that
CG allows us to treat virtually every possible compound VNP with
various degrees of acceptability in a unified and coherent way, with-
out abstracting subtle but important connotations that even the most
marginal expressions could carry.
1.3 Outline
The outline of this chapter is as follows. In Section 2, I briefly describe

the basic nature of kango VNPs in a bottom-up fashion. First, the
general characteristics of kango, or words of Chinese-origin, are con-
sidered. Although they were only loanwords when first introduced to
Japan, kango have been integrated into the syntactic and semantic
systems of the language over time. Now they occupy a significant part
of Japanese and have acquired structures consistent with the grammar
of Japanese. I discuss some conceivable reasons why kango VNPs
have become so important in Japanese both in number and function.
I also make an informal classification of simple kango VNPs and point
out some notable characteristics of their syntax and semantics.
2 The foundation of LCS theory was laid by Jackendoff (1990). This theory deals
with semantic aspects to language much more than other theories derived from
Generative Grammar. Still, LCS presupposes the existence of so-called se-
mantic primitives and assumes that the meanings of words are generated through
the process of applying semantic functions to those primitives in combination
with other constants and valuables. The latter fact separates the two approaches,
CG and LCS, in a fundamental sense, even though they share a strong affinity
for the semantic aspects of language.
In Section 3, I attempt to classify compound kango VNPs into

four types: V+V, Adv+V, Obj+Vt, and Sbj+Vi types. The patterns of
complex kango VNPs are so varied that the internal structures might
seem arbitrary at first sight. A closer look at their form and meaning,
however, enables us to categorize most of the instances into one of
the four types. I will also analyze each type from a cognitive linguis-
tic perspective and propose a schematization of its conceptual struc-
ture.
In Section 4, the way that complex kango VNPs combine with
the light verb suru is considered. Their internal structure is hard to
access when observed separately from the syntactic and semantic
context. Complex kango VNPs can sometimes directly combine with
suru, and at other times there is a case-marking particle -o interven-
ing between the two. I investigate the conditions in which complex
kango VNPs combine with not only suru but also -o + suru, arguing
that such conditions indicate the conceptual autonomousness of com-
plex kango VNP structures. Although one grammatical construction
seems to refute the validity of the hypothesis, I will demonstrate that
such an example can actually support the hypothesis.
In Section 5, I illustrate that the hypothesis suggested is also
verified in a typological perspective. Many languages show a strong
affinity for one of two natural paths: Action Chain or A/D Layering.
English is considered to be based on the Action Chain event construal,
for instance. Other languages reflect the other natural path, A/D Lay-
ering, if only on some local level of the overall structure. Compound
kango VNPs of the Obj+Vt type and the Sbj+Vi type have an inter-
esting commonality; that is, they both reflect A/D Layering. I sug-
gest that while Japanese is based on Action Chain in the sentence
level, it is based on A/D Layering at the level of phrase-structure of
compound kango VNPs. This supports my argument that compound
kango VNPs are constructed in such a way that a certain level of
conceptual autonomy can always be achieved. Finally, Section 6 con-
cludes the discussion.
14 Yoichiro Hasebe
2. Components of Kango Complex VNP
2.1 Role and Function of Kango
There are several reasons why kango words have become an essen-
tial part of the Japanese language. Sato (1987) gives three sugges-
tions. First, Chinese words (kango) and their written forms were quite
useful in importing and assimilating new concepts from foreign coun-
tries having different cultures. In the old and medieval ages, those
concepts mainly came from China, and the ways in which they were
expressed were directly adopted from China with some necessary
(mostly phonological) modifications. In the modern period, words
of Chinese-origin and characters were again used in translating the
concepts imported from Western culture.
Second, kango made it possible to concisely express complex
ideas and concepts. Japanese grammar uses post-nominal particles
to represent relationships. In Chinese, on the other hand, there is no
need for such extra grammatical elements; only word order matters.
Moreover, since a semantic unit consists of one character and one
syllable in Chinese, kango words tend to be more succinct as com-
pared to their non-kango counterparts.3
Third, the word formation of kango is morphologically simple,
compared to the original Japanese word formation. Because of the
building-block-like characteristics of kanji, a variety of meanings
can be expressed in a semi-nominal form by constructing a sequence
of various character lengths. Because of this characteristic, kango
can nominalize concepts that are complex enough to be alternatively
expressed in a full sentence.4
3 On the other hand, there is a drawback to adopting and coining many kango
words: it has created a great number of homonyms in Japanese. This is because
the Japanese phonological system is simpler than the Chinese system. Chinese
can place four different tones on a single vowel type, whereas Japanese does
not have such differentiation of vowel tones.
4 One marginal but notable fact is that kango words over a certain length are
often shortened into contracted forms without losing much of the original sym-
bolic relationship between the form and meaning.
For the purpose of the present research, the last point is quite
significant. Japanese, originally an agglutinative language that needs
particles placed right after each argument nominal, is equipped also
with a characteristic of Chinese, an isolating language, on the level
of word formation. The next section will introduce some patterns of
kango VNP formation.
2.2 Types of Simple Kango VNs
Simple kango VNs are defined as verbal nouns that are composed of
two kanji and are able to combine with the light verb suru or with the
sequence of accusative case marker o- plus suru. As illustrated in
(6), there are broadly four types of simple kango VNs. Although each
type has its own formal pattern, its instances are rather lexicalized
and thus the patterns are not necessarily highly productive.5
(6) a. sequence of synonymous

synonymous verbal
verbalmorphemes
morphemes
ᩍ⫱
‹³ˆç (kyou-iku, ‘educate-grow’), ㏣ồ
(kyou-iku, ‘educate-grow’), ’Ç‹• (tsui-kyu,
(tsui-kyu, ‘pursue-want’)
‘pursue-want’)
b. the first element
element (adverbial)
(adverbial)modifies
modifiesthe
thesecond
second(verbal)
(verbal)
ᚋ᜼
Œã‰÷ (kou-kai,
(kou-kai, ‘late-regret’), ๓㐍‘O•i
‘late-regret’), (zen-shin,
(zen-shin, ‘forward-proceed’)
‘forward-proceed’)
c. the second element
element (adjectival)
(adjectival)isiscomplement
complementtotothe
thefirst
first(verbal)
(verbal)
ㄝ᫂
à–¾ (setsu-mei, ‘explain-clear’), ᨵၿ
(setsu-mei, ‘explain-clear’), ‰ü‘P(kai-zen,
(kai-zen,‘modify-better’)
‘modify-better’)
d. the second element (nominal) is object of the first element (verbal, transi-
tive)
೵ᡓ (tei-sen, ‘halt-war’), ẅே (satsu-jin, ‘kill-person),
ᖐᅜ (ki-koku, ‘return-country’), ㏥ᐊ (tai-shitsu, ‘leave-room’)
Among simple kango VNs in Japanese, those that contain an argu-

ment structure, i.e. those that belong to (6d), have a VO construction
that reflects the original Chinese syntax (Arakawa 1988). The argu-
ment NP accordingly can take various semantic roles such as theme
(i) a. ⾜ᨻᨵ㠉 (gyousei-kaikaku, administration-reform) → ⾜㠉 (gyou-kaku)

b. ኱Ꮫ༞ᴗ (daigaku-sotsugyou, university-graduate) → ኱༞ (dai-sotsu)
5 The classification in (6) is based on the analysis proposed in Arakawa (1988)
for kango NPs in general.
16 Yoichiro Hasebe
(tei-sen, ‘halt-war’), patient (satsu-jin, ‘kill-person’), goal (ki-koku,

‘return-country’), and source (tai-shitsu, ‘leave-room’).
When these simple transitive kango VNs combine with an ob-
ject NP to make up a compound VNP, the newly generated argument
structure follows the Japanese instead of the Chinese syntax. In other
words, the object and the VNP form an OV construction as seen in
(7) instead of a VO construction.
(7) a. ♫ဨເ㞟 (shain-boshuu, ‘employees-recruit’)

b. グ᠈႙ኻ (kioku-soushitsu, ‘memory-lose’)
c. ỈὋ⚗Ṇ (suiei-kinshi, ‘swimming-ban’)
Virtually all simple intransitive kango VNs are unaccusative in the

sense of Perlmutter (1978).6 Many have a meaning that concerns the
existence, emergence, or disappearance of objects or events. As is
the case with transitive kango VNs, word formation of this type takes
a different order from the one that is most commonly observed in the
sentence-level syntax of Japanese. Consider examples (8) and (9)
below.
(8) a. ⰼ䛜㛤䛟䚹
hana-ga hiraku
flower-NOM bloom
‘A flower comes into bloom.’
b. 㛤ⰼ (kai-ka, ‘bloom-flower’)
(9) a. 㞵䛜㝆䜛䚹
ame-ga furu
rail-NOM fall
‘Rain falls.’
b. 㝆㞵 (kou-u, ‘fall-rain’)
Interestingly, when an unaccusative VN combines with a subject NP

to create a compound kango VNP, the word order is in accordance
with the sentence-level syntax and shows SV construction.
6 See also Levin/Rappaport Hovav (1995) for an extensive discussion about

unnaccusativity.
(10) a. ᬒẼ䛜ᚋ㏥䛧䛶䛔䜛䚹
keiki-ga koutai-siteiru
economy-NOM slowdown-suru-PROG
‘Economy is slowing down.’
b. ᬒẼᚋ㏥ (keiki-koutai, ‘economy-slow down’)
(11) a. ஺㏻䛜῰⁫䛧䛶䛔䜛䚹
koutsuu-ga juutai-siteiru
traffic-NOM congestion-suru-PROG
‘Traffic is congested.’
b. ஺㏻῰⁫ (koutsuu-juutai, ‘traffic-congestion’)
The issue of transitivity in compound kango VNPs is further dis-

cussed in Sections 4 and 5.
2.3 Functions of Kango VN
As mentioned in 2.1, kango words express concepts of various types

and complexity in a concise fashion. Kango VNs can represent rela-
tional concepts in a nominal-like form, and they can be used as NPs
functioning as arguments of the verb. In example (12) below, while
the VN doui in (12a) functions as a part of the predicate VP, in (12b)
the same VN functions as the argument of VP hitsuyou-da.
(12) a. ᙼ䛿⚾䛾᱌䛻ྠព䛧䛯䚹
kare-wa watashi-no an-ni doui-shita
he-TOP I-GEN plan-DAT agree-suru-PAST
‘He agreed to my plan.’
b. 䛣䛾ィ⏬䛾ᐇ⌧䛻䛿䚸ᙼ䛾ྠព䛜ᚲせ䛰䚹
kono-keikaku-no jitsugen-ni-wa kare-no doui-ga hitsuyou-da
this-plan-GEN realization-DAT-TOP he-GEN agree-NOM necessary-AUX
‘His agreement is necessary to carry out this plan.’
Kango VNs are not only alternative expressions of their non-kango

counterparts in Japanese, they are the only choice available in certain
constructions. In general, Japanese non-kango verbs can be converted
to nominals by conjugating their endings. Sometimes, the converted
nominals inherit the argument structure of the original verb as in (13),
18 Yoichiro Hasebe
whereas other times converted nominals do not inherit the argument

structure and are not considered natural expressions as shown in (14).
In the later cases, kango VNs with corresponding semantic contents
often can be used instead as in (14c) (Ito/Sugioka 2002).
(13) a. ୖ₇᫬㛫䜢ၥ䛔ྜ䜟䛫䜛
jouen-jikan-o toiawaseru [verbal]
performance-time-ACC inquire
‘to inquire the performance time (of a play)’.
b. ୖ₇᫬㛫䛾ၥ䛔ྜ䜟䛫
jouen-jikan-no toiawase [nominal]
performance-time-GEN inquiry
‘inquiry of the performance time (of a play)’
(14) a. ㈈ᕸ䜢䛺䛟䛩
saifu-o nakusu [verbal]
wallet-ACC lose
‘to lose (my) wallet’
b. *㈈ᕸ䛾䛺䛟䛧
*saifu-no nakushi [nominal]
wallet-GEN lose
‘loss of (my) wallet’
c. ㈈ᕸ䛾⣮ኻ
saifu-no funshitsu [kango VN]
wallet-GEN disappearance
‘loss of (my) wallet’
An original Japanese verb and its kango alternative do not necessar-

ily have the same meaning. The former tends to refer to particular
events in real life, whereas the latter is more likely to refer to a gen-
eralized, abstract concept of an event. For this reason, original Japa-
nese verbs are preferred in everyday conversations, while kango VNs
are more often observed in books and articles.
3. Cognitive Structure of Compound Kango VNPs
There may seem to be countless ways in which kango words com-

bine with each other to produce a compound VNP, but in fact, most
compound VNPs can be classified into one of four categories to be
described in this section. As shown, Compound VNPs in each type
share a schematic conceptual structure. Although there could be some
specificity to each instance of kango VNP, a common schematic struc-
ture still seems to exist.
3.1 Classification of Compound Kango VNPs
Compound kango VNPs are classified into the four types in (15). Note
that the head of the compound is always the second element of the
word sequence. Although the head elements are all verbal in (15), the
subordinate elements (the first elements) have different characteris-
tics in their part of speech (verbal, adverbial, object nominal, or sub-
ject nominal). In this section, I will examine each of the four types of
compound kango VNPs using the theoretical framework of CG.
(15) a. verb + verb (V+V)

b. adverb + verb (Adv+V)
c. object + transitive verb (Obj+Vt)
d. subject + intransitive verb (Sbj+Vi)
3.2 V + V
(16) a. ಖᏲ⟶⌮
•Û•çŠÇ—• 䇺maintain-administrate䇻)
(hoshu-kanri,
(hoshu-kanri, ‘maintain-administrate’)
b. ㍺ධ㈍኎ (yunyuu-hanbai,䇺import-retail䇻)
—A“ü”Ì”„(yunyuu-hanbai, ‘import-retail’)
c. ෭෾ಖᏑ (reitou-hozon,䇺freeze-keep䇻)
—â“€•Û‘¶(reitou-hozon, ‘freeze-keep’)
d. ᳨ᰝධ㝔
ŒŸ•¸“ü‰@ 䇺checkup-hospitalize䇻)
(kensa-nyuuin,
(kensa-nyuuin, ‘checkup-hospitalize’)
20 Yoichiro Hasebe
Compound VNPs of the V+V type (16) are composed of two kango
words, both having verbal semantic features. There are several cat-
egories among VNPs of this type. First, in VNPs like hoshu-kanri
(‘maintain-administrate’), two similar event concepts are juxtaposed
as equals. Second, in VNPs like yunyuu-hanbai (‘import-retail’), two
related but separate events are arranged in chronological order. Third,
in VNPs like reitou-hozon (‘freeze-keep’), the first element expresses
the way in which the event in the second element is carried out. Fourth,
in VNPs like kensa-nyuuin (‘checkup-hospitalize’) the first element
introduces the purpose of the event expressed in the second element.
Even though the first elements of all of the above subcatego-
ries in this VNP type have some kind of event frame in their concep-
tual structures, the degree of saliency is quite varied. The first ele-
ment of the VNPs in the first type (hoshu-kanri) is relatively high in
its saliency but not as high as that of the second element, which is the
head of the VNP. Still, hoshu in hoshu-kanri is conceptually promi-
nent enough to be regarded appositional to the head kanri. The first
element of a VNP in the second category (yunyuu-hanbai) can repre-
sent an event that is independent of the event depicted by the second
element, as yunyuu (‘import’) can be thought of as a separate event
preceding the next event hanbai (‘retail’). Yet the compound VNP as
a whole does not profile both of the events equally; it is always the
second event that is more salient than the other and thus profiled. For
this reason, yunyuu-hanbai does not refer to a kind of importing but
to a kind of retailing in a default interpretation of the expression.7
The first elements of VNPs in the third (reitou-hozon) and fourth
(kensa-nyuuin) categories are not highly prominent either. Although
events such as reitou (‘freeze’) and kensa (‘checkup’) have concrete
semantic contents, they are regarded only as part of the ground against
which the main events hozon (‘keep’) and nyuuin (‘hospitalize’) are
profiled. Thus, compound kango VNPs of V+V type, which feature
two related events in their conceptual structure, are varied in terms
of the saliency of their first elements. The relationship among the
subtypes is schematically captured in Figure 1.
7 It is possible, however, for yunyuu-hanbai to mean ‘importing AND retailing’

if a tonal segmentation is made between the compound’s component elements.
(a) (b) (c)
Figure 1.
Figure (1a) illustrates the conceptual structure of VNPs like hoshu-

tenken. According to the schematic representation of CG, the arrows
represent relational structures or events of any kind. In Figure (1a),
two arrows drawn in bold lines represent the fact that the two events
are equally given explicit linguistic realization. The circles at one
end of these arrows represent the trajector, or the primary figure, of
an event. Though it is logically supposed that there exists an agent
behind the scenes that actively causes the event to take place, such
an agent is not profiled in the conceptual structure. The circles in
(1a), therefore, are not given bold contours. The overall conceptual
structure is thus composed of two relations that syntactically mani-
fest in the expression. It does not mean, however, that the two events
are totally separated. Rather, it is conceived that they are strongly
related and form a larger, more generic event concept. For that rea-
son, the two circles and two arrows are respectively connected with
dotted lines, and the outermost frame is given a bold contour to indi-
cate that this conceptual structure is profiled as a coherent whole.
Figure (1b) illustrates the conceptual structure of VNPs like
yunyuu-hanbai. The two events are horizontally juxtaposed to one
another, and one extra dotted arrow is added below the outside frame
of the juxtaposed structures, representing the fact that the two events
take place in chronological order. Note that only the box on the right
side is drawn in a bold line. Compound VNPs of this type only pro-
file the second of the two component elements.
Figure (1c) shows the conceptual structure of VNPs like reitou-
hozon and kensa-nyuuin, in which the first element is overshadowed
by the second element. The conceptual overlap is reflected in the
configuration of the boxes and the correspondence lines between the
22 Yoichiro Hasebe
two component structures. Since the profile of the overall structure

is placed on the second element of the pair, the bold contour is given
to the box on the right side.
3.3 Adv + V
(17) a. 㝔ෆឤᰁ
‰@“àŠ´•õ (innai-kansen,
(innai-kansen, ‘inside
䇺inside hospital-infect’)
hospital-infect䇻)
b. ᪂つཧධ
V‹K•Q“ü (shinki-sannyuu,
(shinki-sannyuu,䇺newly-enter䇻)
‘newly-enter’)
c. 㧗㏿ᅇ㌿䇺rapidly-rotate䇻)
(kousoku-kaiten,‘rapidly-rotate’)
‚‘¬‰ñ“] (kousoku-kaiten,
Compound kango VNPs of the Adv+V type (17) are those that are
composed of a head VNP and another preceding kango that modifies
it.8 There are three subtypes in compound VNPs of this type. First,
VNPs like innai-kansen (‘inside-hospital-infect’) have a structure in
which the first element specifies the location in which the event de-
scribed by the second takes place. Second, VNPs like shinki-sannyuu
(‘new-enter [to a business competition]’) have a structure in which
the first element specifies the nature or quality of the event invoked
by the second. Third, VNPs such as kousoku-kaiten (‘rapidly-rotate’)
have a structure in which the first element stipulates the degree of
the speed, rate, duration, etc. of the event described by the second.
In all of these subtypes, the conceptual structure has two com-
ponent structures: one contains an indication of a specific point or
position in a domain or scale of some kind, while the other contains
the event concept that is given the highest degree of prominence
within the overall structure. In innai-kansen, the domain featured is
a spatial/locative one. In shinki-sannyuu, a time scale is featured; in
kousoku-kaiten, the rate/speed scale. The composition process of such
a domain expressed by the first element and an event concept ex-
pressed by the second element is illustrated in Figure 2.
8 Since a kango VNP has both verbal and nominal aspects, it is equally possible
to regard an element modifying it as an adverb or adjective. Since this chapter
focuses on the verbal aspects of kango VNPs, I refer to this type of VNP in
which the first element modifies the second as ‘Adv + V Type’. In a different
context, it would be appropriate to call it ‘Adj+N Type’.
Figure 2.
In Figure 2, the box on the lower left represents the domain/scale and
a point within it invoked by the first element of the compound. This
abstract structure is merged into a rather concrete event structure in
the box on the right, which is invoked by the second element of the
compound. Finally, the composite structure depicted in the third box
on the top is constructed. Note that the composite structure inherits
the basic configuration of the component structure on the right side,
which is the profile determinant and is thus enclosed by a bold rect-
angle.
3.4 Obj + Vt
(18) a. ពぢ஺᥮ (iken-koukan, 䇺opinion-exchange䇻)

b. つไ⦆࿴ (kisei-kanwa, 䇺regulation-relax䇻)
Compound kango VNPs of the Obj+Vt type (18) is composed of a

preceding nominal element and a verbal element, with the former
being the object of the latter. Some compounds of this type can be
paraphrased into a sentence by attaching the accusative marker -o at
the end of the object element, while others are paraphrased similarly
but with the dative marker -ni instead of -o. Examples of these sub-
types and their corresponding sentences are provided in (19) and (20)
respectively.
24 Yoichiro Hasebe
(19) a. ༠ຊ౫㢗 (kyouryhoku-irai, 䇺cooperation-request䇻)

b. ⚾䛿ᙼ䛻༠ຊ䜢౫㢗䛧䛯䚹
watashi-wa
hi kare-ni k kyouryoku-o
i k irai-shita
k
I-TOP he-DAT cooperation-ACC request-suru-PAST
‘I asked him for his help ’
(20) a. ⫋ሙ᚟ᖐ (shokuba-fukki, ‘work place-return’)

b. ᙼዪ䛿᮶᭶⫋ሙ䛻᚟ᖐ䛩䜛䚹
kanojo-wa raigetsu shokuba-ni fukki-suru
she-TOP next-month work-place-DAT return-suru
‘She will return to work next month.’
CG stipulates that even a case particle or suffix has its own concep-
tual structure, and different markers must be treated as such. Pres-
ently, I will not examine further the unique conceptual structures of
each of the case-marking particles -o and -ni. Compound kango VNPs
of the Obj+V type are depicted in Figure 3.
Figure 3.
Figure 3 illustrates the composition of component structures in an

Obj+V kango VNP. The box in the lower right represents the compo-
nent structure invoked by the second element of the compound that
has the characteristic of a transitive verb. In the second element of the
compound, explicit linguistic expression is only given to the event
frame and not to particular participants; thus, only the middle arrow
appears as a thick line. With the event frame as transitive, the concep-
tual structure expects to have the schematic object filled by some
concrete information in another component structure; the cross-

hatched circle indicates this. The cross-hatched object corresponds to
the thing element in the lower-left box; thus, they are connected with
a correspondence line, and the fact that the former is dependent on the
latter structure is indicated by the curvy arrow from right to left.
Finally, the box on top of the two represents the composite structure.
3.5 Sbj + Vi
(21) a. ⤒Ⴀ◚⥢ (keiei-hatan, 䇺management-fail䇻)

b. ஺㏻῰⁫ (koutsuu-juutai, 䇺traffic-congest䇻)
Compound kango VNPs of the Sbj+Vi type (21) have an argument

structure as those of the Obj+Vt type. In VNPs of the Sbj+Vi type,
however, it is the subject NP that is linguistically expressed by the first
element of the compound, and the verbal head of the compound is
intransitive instead of transitive. Since VNPs of the Sbj+Vi type contain
both a subject NP and a verbal element, they can be turned into a
sentence by adding the nominative case-marking particle -ga to the sub-
ject NP and also by adding the light verb suru. Consider (22) and (23).
(22) a. ᅜᐙᔂቯ (kokka-houkai, ‘nation-collapse’)

b. ᅜᐙ䛜ᔂቯ䛧䛯䚹
kokka-ga houkai-shita
nation-NOM collapse-suru-PAST
‘The nation has collapsed.’
(23) a. ᰴ౯ୖ᪼ (kabuka-joushou, ‘stock price-rise’)

b. ᰴ౯䛜ୖ᪼䛧䛶䛔䜛䚹
kabuka-ga joushou-shiteiru
stock-price-NOM rise-suru-PROG
‘Stock prices are going up.’
Only intransitive VNs with unaccusative characteristics are quali-

fied to form a compound kango VNP of this type. An unaccusative
verb is a verb that takes a subject NP that could also function as the
object of a corresponding transitive verb having the same stem (Perl-
26 Yoichiro Hasebe
mutter 1978). In general, the referent of the subject of an unaccusative

verb does not play an active role in the event described by the verb;
rather, it is likely that the referent of the subject comes under the
influence of the event. To put it in another perspective, the event
invoked by an unaccusative verb does not contain an active agent in
an explicit way. The cognitive process behind the formation of com-
pound kango VNPs of this type is shown in Figure 4.
Figure 4.
Figure 4 is somewhat symmetrical to Figure 3, which represents kango

VNPs of the Obj + Vt type. Most of the details of this figure, there-
fore, can be understood accordingly. In the box on the lower right,
however, it is the trajector of the event that is cross-hatched, i.e. the
subject of the event has to be elaborated by a concrete element in the
other component structure. Moreover, the arrow in the lower right
box representing a schematic event does not point to anything; this
means that the event is not recognized as causal but as a more or less
spontaneous occurrence. The role of unaccusative VNs in the forma-
tion of the Sbj+Vi type compound VNP is further examined in Sec-
tion 5.
4. “Kango VNP + Suru” Construction
In Section 3, I classified compound kango VNPs into four types and

analyzed each from a CG perspective. These types of compound kango
VNPs all have more complex structures than the simple VNPs that were
considered in Section 2. What characteristics do such VNPs display when
integrated into a larger linguistic construction? In this section, I will
examine the behavior of compound kango VNPs when combined with the
light verb suru (roughly meaning ‘do’ in English). Then, I will suggest
an important cognitive characteristic that compound kango VNPs have.
4.1 Composition of Simple Kango VN and Suru
Most simple kango VNs can be immediately followed by suru to

become a full-fledged predicate VP. In many cases, it is also possible
for a simple kango VN to take the accusative case marker directly after
it and then suru, forming the construction ‘kango VN + -o + suru’.
Examples of this alternation between suru and -o suru are shown in
(24) and (25).
(24) a. ኴ㑻䛜ᩓṌ䛧䛯䚹
Taro-ga sanpo-shita
Taro-NOM walk-suru-PAST
‘Taro took a walk.’
b. ኴ㑻䛜ᩓṌ䜢䛧䛯䚹
Taro-ga sanpo-o-shita
Taro-NOM walk-ACC-shita
‘Taro took a walk.’
(25) a. ኴ㑻䛜⮬ẅ䛧䛯䚹
Taro-ga jisatsu-shita
Taro-NOM suicide-suru-PAST
‘Taro committed suicide.’
b. ኴ㑻䛜⮬ẅ䜢䛧䛯䚹
Taro-ga jisatsu-o-shita
Taro-NOM suicide-ACC-shita
‘Taro committed suicide.’
28 Yoichiro Hasebe
To a certain group of kango VNs, the construction ‘VNP + -o + suru’

cannot be applied as naturally as to others, if at all possible. Consider
examples (26)-(28).
(26) a. 㛗ዪ䛜ㄌ⏕䛧䛯䚹
Choujo-ga tanjou-shita
first-daughter-NOM born-suru-PAST
‘(Our) first daughter was born.’
b. ? 㛗ዪ䛜ㄌ⏕䜢䛧䛯䚹
? Choujo-ga tanjou-o-shita
first-daughter-NOM born-ACC-suru-PAST
‘(Our) first daughter was born.’
(27) a. ೺ᗣ䜢⥔ᣢ䛩䜛
kenkou-o iji-suru
health-ACC maintain-suru
‘(I) maintain (my) health.’
b. ? ೺ᗣ䛾⥔ᣢ䜢䛩䜛
?kenkou-no iji-o-suru
health-GEN maintain-ACC-suru
‘(I) maintain (my) health.’
(28) a. ᒣ⏣䛿㌴䜢ᡤ᭷䛧䛶䛔䜛
Yamada-wa kuruma-o shoyuu-shiteiru
Yamada-TOP car-ACC own-suru-PROG/STATE
‘Yamada owns a car.’
b. ? ᒣ⏣䛿㌴䛾ᡤ᭷䜢䛧䛶䛔䜛
?Yamada-wa kuruma-o shoyuu-o-shiteiru
Yamada-TOP car-ACC own-ACC-suru-PROG/STATE
‘Yamada owns a car.’
Tanomura (1988) proposes a set of semantic conditions that a kango

VN must meet in order to form a ‘VNP + -o + suru’ construction in
addition to a ‘VNP + suru’ construction.
(29) a. The VNP represents an event that is intentionally carried out by an agent.
b. The VNP represents an event that has specific starting and ending points.
c. The VNP does not represent an event that is emotional/psychological.
(Tanomura 1988, my translation with minor modifications)
Condition (29a) states that a non-intentional event cannot be expressed

by the ‘VN + -o + suru’ construction, and it successfully predicts that
tanjou [be born] does not perfectly fit in this construction. (29b) states
the aspectual requirement of the ‘VNP + -o + suru’ construction; that
is, the event described by the VNP is telic but not atelic. This pre-
dicts that iji [maintain] is not appropriate for the construction. Fi-
nally, (29c) excludes psychological VNs like shoyuu from the group
of words most appropriate for the construction.
Now let us move on to compound kango VNPs and examine
whether Tanomura’s conditions in (29) are applicable. Two examples
from each of the four types from Section 3.1 are combined to suru
and -o-suru respectively, producing the results listed in (30) - (33).
(30) V + V Type a. ಖᏲ⟶⌮䛩䜛 (hoshu-kanri-suru, ‘maintain-administrate’)

ಖᏲ⟶⌮䜢䛩䜛 (hoshu-kanri-o-suru, ‘maintain-administrate’)
b. ෭෾ಖᏑ䛩䜛 (reitou-hozon-suru, ‘freeze-keep’)
෭෾ಖᏑ䜢䛩䜛 (reitou-hozon-o-suru, ‘freeze-keep’)
(31) Adv + V Type a. 㝔ෆឤᰁ䛩䜛 (innai-kansen-suru, ‘inside hospital-infect’)

? 㝔ෆឤᰁ䜢䛩䜛 (innai-kansen-o-suru, ‘inside hospital-infect’)
b. ᪂つཧධ䛩䜛 (shinki-sannyuu-suru, ‘newly-enter’)
᪂つཧධ䜢䛩䜛 (shinki-sannyuu-o-suru, ‘newly-enter’)
(32) Obj + Vt Type a. ពぢ஺᥮䛩䜛 (iken-koukan-suru, ‘opinion-exchange’)

ពぢ஺᥮䜢䛩䜛 (iken-koukan-o-suru, ‘opinion-exchange’)
b. つไ⦆࿴䛩䜛 (kisei-kanwa-suru, ‘regulation-relax’)
つไ⦆࿴䜢䛩䜛 (kisei-kanwa-o-suru, ‘regulation-relax’)
(33) Sbj + Vi Type a. ⤒Ⴀ◚⥢䛩䜛 (keiei-hatan-suru, ‘management-fail’)

? ⤒Ⴀ◚⥢䜢䛩䜛 (keiei-hatan-o-suru, ‘management-fail’)
b. ஺㏻῰⁫䛩䜛 (koutsuu-juutai-suru, ‘traffic-congest’)
? ஺㏻῰⁫䜢䛩䜛 (koutsuu-juutai-o-suru, ‘traffic-congest’)
Among all of the compound kango VNPs in the above examples,

those that express some kind of intentional event show a perfect match
with the ‘VNP + -o + suru’ construction. Other VNPs that express
events having unintentional connotations, however, do not display a
perfect combination with -o-suru, if not impossible (31a, 33a, and
30 Yoichiro Hasebe
33b). Thus, it can be said that the condition regarding the combina-
tion of simple kango VNs and the -o-suru sequence is also valid when
applied to complex VNPs.
It is now clear that compound kango VNPs combine with -suru
and -o-suru just as simple kango VNPs do, as long as the conditions
in (29) are met. This is quite impressive given that compound VNPs
have rather complex structures both syntactically and semantically.
For a VNP to have -suru (roughly meaning ‘do’) immediately follow-
ing it, the VNP must be describing something that the speaker can
conceive of as one single conceptual unit. Moreover, the particle -o is
reasonably considered to evoke a relationship between the vantage
point of a conceptualizer and a view directed at the target object, re-
quiring the VNP preceding it be a ‘coherent whole’. Thus, it is sug-
gested that a well-formed compound VNP has a conceptual auto-
nomousness that gives it a status as a coherent unit. How can a com-
pound VNP, which sometimes is composed of a structure as complex
as an argument structure, be perceived as a single, coherent concep-
tual unit? This problem is further investigated in Section 5.
4.2 Transitivity of “Kango VNP + Suru” Construction
In this section, I introduce a phenomenon that appears to be a counter-

example to the argument made in the preceding section that com-
pound VNPs have a conceptually autonomous structure. However, I
will provide a cognitive explanation of this phenomenon to show that
it does not actually challenge the validity of the previous argument.
Compound kango VNPs of Obj+Vt type and Sbj+Vi type con-
tain an argument structure. In either type, VNPs function in sentences
as intransitive VPs: the Sbj+Vi type obviously has an intransitive
quality to it, and the Obj+Vt type also is regarded as intransitive
because the transitivity of the verbal head is saturated when it com-
bines with the first element, which is conceived as the object of the
event described by the verbal head. As intransitive VPs in general do
not take syntactically explicit object NPs, compound kango VNPs
having an argument structure are not supposed to take external NPs
as their objects.
There exist, however, cases in which a compound kango VNP

takes an external object NP, even though the first element and the
second element are in an argument structure (Kageyama 1993;
Kobayashi 2004). Consider examples (34) from Kobayashi (2004:
219-20):
(34) a. ᩳ㝧䛾㕲㐨⏘ᴗ䜢ᵓ㐀㌿᥮䛩䜛䚹
shayou-no tetsudou-sangyou-o kouzou-tenkan-suru
declining-GEN railroad-industry-ACC structure-transform-suru
‘(They) restructure the declining railroad industry.’
b. ෑ㢌䛾ᩥ䜢ព࿡ゎ㔘䛩䜛䚹
boutou-no bun-o imi-kaishaku-suru
beginning-GEN sentence-ACC meaning-interpret-suru
‘(I) interpret the meaning of the opening sentence.’
In (34a), the compound VNP kouzou-tenkan [structure-transform]

takes an external object tetsudou-sangyou [railroad-industry] even
though the VNP itself contains a relationship between a transi-
tive VN (transform) and its object NP (kouzou). Similarly in (34b),
the verbal head of the compound (kaishaku) takes an object (imi)
internally. The compound VNP, however, further takes an external
object (bun). When a VP takes such an explicit object NP, sup-
posedly it means that the VP does not have conceptually autono-
mous status; it is in need of an additional semantic element to
achieve a coherent semantic structure. Does this disconfirm the ar-
gument that complex VNPs consist of a conceptually autonomous
structure?
In fact, there is a common characteristic shared among virtu-
ally all compound kango VNPs of the Obj+Vt type that take an ex-
ternal object: the external object is always in a hyponymic relation-
ship with the internal object. To put it differently, both object NPs
are in a ‘part-and-whole’, or ‘container-and-contained’ relationship.
In (34a), the internal object kouzou [structure] is understood to be a
part (or aspect) of the external object tetsudou-sangyou [railroad-
industry]. In (34b), the internal object imi [meaning] is in the exter-
nal object bun [sentence].
Thus, the external objects in sentences such as those in (34)
are all in a strong relationship with the internal objects as well as
32 Yoichiro Hasebe
with the verbal head of the compound.9 Those external objects pro-
vide a conceptual setting or domain in which the event expressed by
the predicate VNP can be evoked. The conceptual autonomousness
of the compound VNP is preserved because the two objects, which
have syntactically different realizations, are practically identical. This
is illustrated in Figure 5.
Figure 5.
Note that if the direction of the ‘whole/part’ relationship is incorrect

and the domain is inappropriately set up, the construction with an
external object may become unacceptable as in (35).
(35) a. *ኻᴗ⪅䜢ᙅ⪅ᩆ῭䛩䜛䚹
*shitsugyousha-o jakusha-kyuusai-suru
the-unemployed (people)-ACC the-weak (people)-assist-suru
‘(They) assist the unemployed – weak – people.
b. ?ኴᖹὒ䛾ᓥ䚻䜢⮬↛◚ቯ䛩䜛䚹
?taiheiyou-no shimajima-o shizen-hakai-suru
Pacific-Ocean-GEN islands-ACC nature-destroy-suru
‘(They) destroy the nature of islands of the Pacific Ocean.’
The reason why a question mark, instead of an asterisk, is used in

(35b) is that this sentence could be saved if one interprets it presuppos-
ing that shizen [nature] is one of the aspects of shimajima [islands],
instead of regarding the latter as part of the former.
9 In this regard, the construction with internal and external object NPs is com-
pletely differentiated from syntactically similar-looking phenomena such as
ditransitive constructions. (cf. John gave Mary a flower.)
5. A/D Layering of Compound Kango VNP
In Section 4, I argued that complex kango VNPs have conceptual

autonomy despite their high structural complexity. In this section, I
introduce a linguistic fact that supports this argument in a cognitive/
typological perspective. Compound kango VNPs of the Obj+Vt and
Sbj+Vi types, the group of VNPs that contain an argument structure,
share a common characteristic reflected in their syntax and seman-
tics. This characteristic is derived from one of the natural paths of
our cognition called A/D Layering. Thus, the conceptual autonomous-
ness of complex kango VNPs has a cognitive basis that is observed
universally in many other linguistic phenomena.
5.1 Internal Argument of Compound Kango VNP
The first element of a compound kango containing an argument struc-

ture is either (i) the object of a transitive verb or (ii) the subject of an
intransitive verb. For the latter requirement, there is a further restric-
tion: the intransitive verb must be unaccusative. From a CG view-
point, the trajector of events invoked by kango VNPs of the Obj+Vt
and Sbj+Vi types is always the participant that is affected or acted
upon by something, rather than actively causing an event to happen.
The examples in (36), which are VNPs of the Obj+V type, and (37),
VNPs of the Sbj+V type, illustrate the above facts.
(36) Ẽศ୍᪂ (kibun-isshin, ‘feeling-refresh’),

ᗘㄪᩚ (ondo-chousei, ‘temperature-adjust’)
(37) ᅜᐙᔂቯ (kokka-houkai, ‘nation-collapse’),

ᰴ౯ୗⴠ (kabuka-geraku, ‘stock-price-decline’)
On the other hand, compounds of a transitive VN and the subject are not
considered natural, although they almost make perfect sense in terms of
causal structure of events (38a, 39a). Note that if the same events are coded
differently in accordance with the previously described requirements,
then the resulting expressions are considered appropriate (38b, 39b).
34 Yoichiro Hasebe
(38) a. * ฮ஦㐊ᤕ (keiji-taiho, ‘detective-arrest’)

b. ≢ே㐊ᤕ (hannin-taiho, ‘criminal-arrest’)
cf. keiji-ga hannin-o taiho-suru, ‘A detective arrests the criminal.’
(39) a. * ᩍᖌᣦᑟ (kyoushi-shidou, ‘teacher-teach’)

b. Ꮫ⏕ᣦᑟ (gakusei-shidou, ‘student-teach’)
cf. kyoushi-ga gakusei-o shidou-suru, ‘Teachers teach (their) students.’
The conceptual background behind this apparent unbalance between

the types of compound kango VNPs is investigated from a typologi-
cal viewpoint in the following sections.
5.2 Case Systems and Two Natural Paths
Many languages are categorized into one of two groups: those that are
based on a nominative/accusative case system and those that are based
on an ergative/absolutive case system. The primary difference be-
tween these two systems is as follows. In languages based on a nomi-
native/accusative system, the subject NPs of an intransitive verb and
a transitive verb is given an identical morphological marking (nomi-
native case, NOM), but a different marking is given to the object NP
of a transitive verb (accusative case, ACC). In languages based on an
ergative/absolutive case system, the subject of a transitive verb is
given a certain morphological marking (ergative case, ERG), but the
subject of an intransitive verb and the object of a transitive verb lack
explicit markings of any kind (absolutive case, ABS). This general
contrast between the two systems can be described in matrices (40)
and (41) below. (The symbol ‘=>’ indicates transitivity of verb.)
(40) Absolutive/Nominative System

a. intransitive construction: NOM
b. transitive construction: NOM => ACC
(41) Ergative/Absolutive System

a. intransitive construction: ABS
b. transitive construction: ERG => ABS
Japanese is said to be a nominative/accusative language. In general,

the nominative/absolutive system is more popular than the ergative/
absolutive system. However, many languages reflect the latter system
locally in their morphology and syntax. Researchers observe that lan-
guages that appear to strongly display a nominative/accusative case
system often adopt an ergative/absolutive system in some grammati-
cal aspects. It is also suggested that some languages display so-called
‘split-ergativity’. In such a language, the case system is tactfully cho-
sen according to various aspects of the linguistic environment, such
as context and intention.
Both Langacker (1991) and Williams (1987) point out that
English, a nominative/accusative language, shows ergative/absolutive
characteristics in periphrastic constructions such as “the chanting of
the demonstrators” and “the chanting of the slogans by the demon-
strators.” In the first example, the nominal that the preposition of
introduces (i.e. the demonstrators) is an element that can be placed
in the subject position of the intransitive verb chant. In the second
example, on the other hand, the nominal following of (i.e. the slo-
gans) is an element that can be placed in the object position of the
transitive version of the verb chant; the preposition by introduces a
nominal (the demonstrators) that can be placed in the verb’s subject
position. In sum, of marks the subject of the intransitive verb’s sub-
ject and the transitive verb’s object, and by marks the transitive verb’s
subject. This parallels the ergative/absolutive matrix in (41).
Referring to such examples from English and other languages,
Langacker (1991) argues that the linguistic realizations of a nomina-
tive/accusative system and an ergative/absolutive system are not lim-
ited to case-marking morphology but are observed in a variety of
phenomena that are linked to the argument structure of verbs. In this
connection, Langacker further argues that these two systems should
not be seen as a dichotomy where only one of them is featured in a
single language. In reality, both systems are used freely when the
perceived event structure being described corresponds to one of the
most basic patterns of our cognition called ‘natural paths’.
There are two such natural paths in our cognition. One of them
corresponds to the flow of energy through participants of an event.
Such an ‘Action Chain’ is observed in many linguistic phenomena
36 Yoichiro Hasebe
not only in English but also in Japanese as well as many other lan-
guages. The causal relationship between participants of an event can
be directly reflected in linguistic forms, a prototypical case of which
is a construction containing transitive verbs with a nominative sub-
ject and an accusative object.
The other natural path corresponds to the conceptual autonomy
of the event. In this natural path, based on what Langacker calls ‘Au-
tonomous/Dependent (A/D) Layering’, the primary concern is the
emergence of an event or the state of a thematic object having gone
through some kind of change, not the causal relationship among the
participants that supposedly played important roles in the event. The
term ‘A/D Layering’ reflects the fact that such an event construal
can have multiple layers in its cognitive structure.
To illustrate how these two natural paths work, let us take an
example from Langacker (1991). He suggests that a single event can
often be linguistically expressed either in a transitive construction or
in an intransitive construction: the former is based on Action Chain,
which is considered the default mode of cognition widely observed
in English; the latter is based on A/D Layering, which is linked to
phenomena that display characteristics of ergative/absolutive linguis-
tic coding. Consider the event “Floyd picks up a hammer, and he
breaks a glass.” English allows one to express this event either as
seen in (42a) or (42b)
(42) a. Floyd broke the glass (with a hammer).

b. The glass broke.
(43) a. Action Chain: Floyd hammer glass
b. A/D Layering: glass hammer Floyd
The transitive construction in (42a) and the intransitive construc-

tions in (42b) are reflections of the cognitive process described in
(43a) and (43b) respectively. (43a) displays the sequential order of
the participants in a causal relationship as well as the flow of the
energy exerted by the agent (‘Floyd’) and transmitted from partici-

pant to participant. (43b), on the other hand, has a nesting structure
with the source of the natural path (‘glass’) as its core. This illus-
trates the fact that in constructions of this type based on A/D Layer-
ing, the participant that is ‘acted upon’ or ‘affected by’ another is
given the highest prominence in the whole structure of the event
construal. Note that in either description of natural paths, the source
of the path is printed in bold letters, reflecting that those participants
in the outset are considered the trajector of an event.
5.3 Conceptual Autonomy of Compound Kango VNPs
The first element of a Japanese compound kango VNP containing an

argument structure is an internal argument. In other words, it is com-
posed of either a transitive verbal head with its object or an intransi-
tive (unaccusative) verbal head with its subject. Thus, the syntactic
nature of the first elements of compound kango VNPs parallels the
ergative/absolutive matrix in (41b). It is reasonably assumed, then,
that complex kango VNPs of the Obj+Vt and Sbj+Vi types are based
on A/D Layering, not on the Action Chain.
Let us verify this by examining examples. Consider first com-
pound VNPs of the Obj+Vt type in (44a)
(44) a. ௻ᴗ䛾♫ဨເ㞟
kigyou-no shain-boshuu
company-GEN employee-recruit
‘a company’s recruiting of employees’
b. * ♫ဨ䛾௻ᴗເ㞟
*shain-no kigyou-boshu
employee-GEN company-recruit
‘a company’s recruiting of employees’
c. ௻ᴗ䛜♫ဨ䜢ເ㞟䛧䛶䛔䜛䚹
kigyou-ga shain-o boshuu-shiteiru
company-NOM employee-ACC recruit-suru-PROG/STATE
‘A company is recruiting employees.’
38 Yoichiro Hasebe
(45) a. Action Chain: kigyou boshuu shain
c. A/D Layering: shain boshuu kigyou
The event described in (44c) can be nominalized using a complex

kango VNP as in (44a). This VNP shain-boshuu composed of a tran-
sitive VN boshuu and its object shain is fully legitimate because it is
based on the A/D Layering construal illustrated in (45b). The same
event can be logically conceived of along a causal relationship based
on the Action Chain as shown in (45a). However, conceptual auto-
nomousness, which is argued to be essential to a complex kango VNP,
is only achievable when the construal of the event is based on A/D
Layering. Therefore the VNP kigyou-boshuu (‘company-recruit’) in
(44b), which has a ‘subject + transitive verb’ structure, is not ac-
cepted as a natural expression of the event.
Next, let us consider a VNP example of the Sbj+Vi type.
(46) a. ᨺⅆ䛻䜘䜛ᐙᒇ඲↝
houka-niyoru kaoku-zenshou
arson-by house-burn-down
‘burn-down of a house because of an arson attack’
b. * ᐙᒇ䛾ᨺⅆ඲↝
**kaoku-no
k k houka-zenshou
house-GEN arson-burn-down
‘burn-down of a house because of an arson attack’
c. ᐙᒇ䛜ᨺⅆ䛻䜘䜚඲↝䛧䛯䚹
kaoku-ga houka-niyori zenshou-shita
house-NOM arson-by totally-burn-suru-PAST
‘A house burnt down because of an arson attack.’
(47) a. Action Chain: houka zenshou kaoku
b. A/D Layering: kaoku zenshou houka

The event described in (46c) can be nominalized as in (46a). The

VNP kaoku-zenshou is composed of the unaccusative (intransitive)
VN zenshou and its subject kaoku. This VNP is also based on A/D
Layering as in (45b). The same event can be conceived of along its
causal relationship based on the Action Chain as illustrated in (47a).
However, since conceptual autonomy is not achievable with the Ac-
tion Chain construal, the VNP houka-zenshou [arson-burn-down] in
(46c), which has an ‘external cause + intransitive verb’ structure, is
not acceptable.
6. Conclusion
In this chapter, I have presented an overview of the syntactic and

semantic characteristics of complex kango VNPs in Japanese and
proposed their conceptual structures in the framework of CG. In Sec-
tion 2, the building-blocks of compound kango VNPs were briefly
discussed. I introduced some reasons why kango words are so widely
integrated into the Japanese language, pointing out that kango words
make it possible to express complex concepts in a concise but orga-
nized way. Then, I briefly discussed the basic patterns and functions
of simple kango VNPs, the building-blocks of complex VNPs. In
Section 3, I classified complex kango VNPs into four types and ex-
amined their conceptual structures, proposing a schematic represen-
tation for each type. In Section 4, I examined a few interesting phe-
nomena that are observed when a complex kango VNP is combined
with the light verb suru and argued that compound kango VNPs have
a conceptual autonomousness that makes them work as syntactically
and semantically coherent units. In Section 5, the autonomous na-
ture of complex kango VNPs was further examined from a cogni-
tive/typological point of view. It was suggested that their autonomy
is linked to the fact that languages tactfully codify events based on
one of two natural paths of cognition, Action Chain and A/D Layer-
ing. Based on this assumption, I argued that an event construal based
on A/D layering is reflected in the structure of compound kango VNPs
40 Yoichiro Hasebe
with an argument structure, i.e. those that are classified as Obj+V

type or Sbj+V type.
This research is only preliminary: there are many related phe-
nomena left untouched or overlooked in the preceding discussions.10
Furthermore, the analyses and hypotheses presented have not been
applied to enough data to conclude that they are both theoretically
and practically valid. The present work, however, will hopefully con-
tribute to the literature by: (i) showing that complex kango VNPs are
not just arbitrary, conventionalized sequences of words but contain
highly structured mechanisms; and (ii) such mechanisms are sup-
ported by basic cognitive abilities and general tendencies in perceiv-
ing events.
References
Arakawa, Kiyohide 1988. Fukugou-kango no nicchuu-hikaku (com-

parison of compound words in Japanese and Chinese). Nihongo-
gaku (Japanese Language Studies) 7/5, 56-67.
Ito, Takane / Sugioka, Yoko 2002. Go no Shikumi to Go-keisei (Word
Structure and Word Formation). Tokyo: Kenkyusha.
Jackendoff, Ray 1990. Semantic Structures. Cambridge, Mass.: MIT
Press.
Kageyama, Taro 1993. Bunpou to Go-keisei (Grammar and Word-
formation). Tokyo: Hitsuji-Shobou.
Kageyama, Taro 1996. Doushi-imiron: Gengo to Ninchi no Setten
(Verb Semantics: An Interface of Language and Cognition).
Tokyo: Kuroshio-Shuppan.
10 One of the most important areas for future research should be the comparison
between Japanese complex kango formation and Chinese syntax. As is men-
tioned in Section 2.2, many Japanese kango VNPs inherit the structure of their
original Chinese expressions. This being so, it seems essential to investigate
the grammatical similarities or differences between the two languages as well
as the cognitive basis for them.
Kobayashi, Hideki 2004. Gendai Nihongo no Kango Doumeishi no

Kenkyuu (A Study of Verbal Nouns of Chinese-origin in Mod-
ern Japanese). Tokyo: Hitsuji-shobou.
Langacker, Ronald W. 1990. Concept, Image, and Symbol: The Cog-
nitive Basis of Grammar. Berlin: Mouton de Gruyter.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar,
Volume 2, Descriptive Application. Stanford: Stanford Univer-
sity Press.
Levin, Beth / Rappaport Hovav, Malka 1995. Unaccusativity: At the
Syntax-Lexical Semantics Interface. Cambridge, Mass.: MIT
Press.
Perlmutter, David M. 1978. Impersonal passives and the unaccusative
hypothesis. BLS 4, 157-189.
Sato, Kiyoji 1987. Nihongo ni okeru kango no za (the position of
words of Chinese-origin in Japanese). Nihongogaku (Japanese
Language Studies) 6/2, 4-12.
Tanomura, Tadaharu 1988. “Heya o souji suru” and “heya no souji o
suru” (“To clean the room” and “to do cleaning of the room”.
Nihongogaku (Japanese Language Studies) 7/11, 70-80.
Williams, Edwin 1987. English as an ergative language: The theta
structure of derived nouns. CLS 23, 366-375.
GAO HUA
Tag-Questions in Mandarin Chinese*
1. Introduction
Tag questions are formed by appending a linguistic ‘mini-question’ –

a so-called ‘question tag’ (henceforth ‘Q-tag’) – to the end of a state-
ment (or rather, a non-question, which is usually declarative but some-
times imperative or exclamatory) to request confirmation or agree-
ment. Such a ‘Statement +Q-tag’ form is said to be a common
question-forming strategy according to Ultan’s (1978) investigation
of 78 languages in the world. However, a tag may vary from lan-
guage to language, and even take different forms in the same lan-
guage.
Unlike English, in Chinese tag questions there is no immedi-
ate structural relationship between the tag and the tagged sentence,
i.e., the tag does not change according to the subject or the verbal
tense/aspect of the relevant part of the sentence. The invariant tag
often takes the form of either a short particle question or a short A-
not-A question meaning ‘right?’, ‘OK?’, such as ᱟੇ (shi ma ‘Is
that so?’), ྭ੗ (hao ba ‘Is that OK?’), ሩнሩ (dui-bu-dui ‘Is that
right?’), ྭнྭ (hao-bu-hao ‘Is that OK?’), ᱟнᱟ (shi-bu-shi ‘Is
that right?’), etc (Shao Jingmin 1996: Chap. 8). Compare the follow-
ing Chinese example and its English translation:
* This chapter is a further study based on my Ph.D thesis A Cognitive-Func-

tional Investigation of Questions in Chinese. I am most grateful to my supervi-
sor, Dr. K.K. Luke, for his continual support and invaluable advice through all
the stages of my study. Many thanks also to the members of my thesis examin-
ing committee, who have offered useful questions and feedback for my revi-
sion. All errors, of course, are my own.
44 Gao Hua
(1) ഋࠔⲴᒤ㓚ᖸ䖫, ྩ᡽ॱҍ኱, ᱟнᱟ? (Shao Jingmin 1996: 128)

sifeng de nianji hen qing, ta cai shijiu sui, shi bu shi?
Sifeng POSS age very small 3F.SG only nineteen year be not be
‘Sifeng is very young; she is only nineteen, isn’t she?’
Although languages may vary in their tag forms, what is interesting

is that all tag questions have their typical function in common, which
is to request confirmation or disconfirmation from the hearer about
what is said in the statement (Chisholm et al. 1984), and the appro-
priate answer to a tag question is usually the affirmation or negation
of the truth of what is stated. Chinese tag questions are no exception.
For example:
(2) TV program host: ᡰԕ֐⧠൘ቡ㿹ᗇ䶎ᑨ⣩䊛, ᱟੇ? 1

suoyi ni xianzai jiu juede feichang youyu, shi ma?
so 2SG now just feel very hesitate right Q
‘So you feel very hesitating now, right?’
TV program guest: ሩ!
dui!
correct
‘Right.’
The tag question in Chinese has attracted attention from many gram-
marians and insights have been gained about its forms and uses.
Nonetheless, there are only a small number of studies that are di-
rectly focused on this apparently ‘special’ question type (Shao Jingmin
1996: 17), which mainly include Niu Baoyi (2001, 2002) and Shao
Jingmin (1996: Chap. 8), while in many others it is only touched
upon in discussions on the grammar of questions in general (see e.g.,
Xu Jie / Zhang Linlin 1993; Zhang Bojiang 1997; Xu Shenghuan
1999). In the literature, the interest on this question type is mainly on
its grammatical status in the grammar of questions in Chinese (i.e.,
whether it should be counted as one independent question type) (e.g.,
1 This study draws examples from a corpus of naturally occurring Putonghua

speech. Thanks to Dr. K.K. Luke, Dr. Zhang Wei and Dr. Zhang Bojiang for
their generosity of allowing me to use their painstakingly collected and tran-
scribed data. The data contains a few television talk shows and conversations
between two or a group of friends in various settings.
Tag-Questions in Mandarin Chinese 45
Lin Yuwen 1985; Shao Jingmin 1996). Relevant to this, attention has
been given to the degree of grammaticalization (䈝⌅ॆ〻ᓖ ) of this
question form (such as when compared with other question forms in
Chinese like the ma-question and the A-not-A question (Zhang
Bojiang 1997)), and its degree of doubt (⯁䰞〻ᓖ ) compared with
other question types (Xu Jie/Zhang Linlin 1993; Xu Shenghuan 1999).
In this study these issues will be re-addressed, and a unified
analysis will be provided of the tag question in Chinese via a cogni-
tive-functional account of the possible motivating factors that are
involved in its formation. Specifically, we try to answer this ques-
tion: if one takes a functional view and considers language as evolved
in response to cognitive-communicative demands (Givón 2001, 2002),
how might the formation of tag questions have been motivated by
these two considerations? It is believed that once an adequate an-
swer to this question is found, a clear picture will come out regarding
the grammatical status of this question type as well as its various
linguistic properties.
2. Previous research on the tag question in Chinese
2.1 The grammatical status of the tag question
The syntactic status of tag questions has been controversial in Chi-

nese, and in other languages (such as English) as well. The point of
contention is whether or not tags should be regarded as reduced ques-
tions in syntax. As briefly described in the previous section, the ques-
tion tag in Chinese often takes an invariant form of either a short yes-
no particle question or a short A-not-A question meaning “is it right?”
or “is it OK?”. Presumably because of this, in works such as Lü Shu-
xiang (1985) and Zhu Dexi (2000), the tag question is not regarded
as a basic grammatical device of forming questions. Lin Yuwen (1985)
even argues that it is not necessary to recognize the tag question as
a distinct question form in Chinese (“ ൘≹䈝Ⲵ⯁䰞ਕѝу・ањ‘
䱴࣐䰞’ᒦнᱟᗵ㾱Ⲵ”). There are two main arguments for this. First,
46 Gao Hua
an apparent tag question like (3) can be treated simply as a sentence

with a subject-predicate (S-P) subject and the A-not-A phrase as the
predicate.
(3) ֐৫ྭнྭ?
ni qu hao bu hao?
2SG go good not good
‘You go, OK?’
Second, there is no structural difference between the tag question as

in example (1) and an A-not-A question such as (4):
(4) ྩᱟнᱟ᡽ॱҍ኱?
ta shi bu shi cai shijiu sui?
3F.SG COP not COP only nineteen year
‘Is she only nineteen?’
However, as Shao Jingmin (1996: Chap. 8) demonstrates, (3) has

different meanings when it contains an S(ubject)-P(predicate) sub-
ject (which means ‘Is it good that you go?’) and when it contains a
tag question (which means ‘You go, OK?’). Second, the two ᱟнᱟ’s
in (1) and (4) are different. For example, the former one may be
replaced by ᱟੇ (shi ma ‘Right?’) while the latter can not, and the
latter can be replaced by ᱟ੖ (shifou ‘Is it true that…?’) while the
former may not.
Apart from Shao Jingmin (1996: Chap. 8), there are other schol-
ars like Li/Thompson (1981), Zhang Bojiang (1997) and Liu Yuehua
et al. (2002), who do recognize the status of the tag question in the
grammar of questions in Chinese.
In Zhang Bojiang (1997), it is argued that the tag question is a
discursive form and thus has a lower degree of grammaticalization
than the structural form ‘A-not-A’ and the morphological ma, though
all of them may be considered as a yes-no question in general, asking
whether a statement holds. In Liu Yuehua et al. (2002), using tags
like ྭੇ (hao ma ‘Is that OK?’) and ሩੇ (dui ma ‘Is that right?’)
is one device of forming a yes-no question, paralleled with other
devices including the ma-particle, the rising intonation and the ba-
particle.
These views may be supported by a fact in Taiwan Min dia-

lect, mentioned in Cheng (1997b: 232), namely that the sentence-
final question particles sio¡ and μ¡ in this dialect are synonymous
with the tag question forms sĶ-m‘-sĶ and kám-m‘-sĶ ‘isn’t it?’, respec-
tively2. The fundamental difference between the two may lie in the
degree of grammaticalization.
A similar conclusion is drawn in Niu Baoyi (2001), which com-
pares tag questions in English and Chinese, and argues that tag ques-
tions in both languages result from grammaticalization of certain
semantic and pragmatic factors, but the English tag question enjoys
a higher degree of grammaticalization than its Chinese counterpart
due to the facts that there is an immediate structural restriction on
the English tag form from the tagged clause while Chinese tags more
rely on lexical and discourse means.
As can be seen, most grammatical research on tag questions
has attached primary importance to the formal property of ‘State-
ment +Q-tag’ as well as different forms the Q-tag may take, such as
ᱟнᱟ (shi-bu-shi ‘Is that correct?’), ᱟੇ (shi ma ‘Is that so?’), and
ᱟ੗ (shi ba ‘Is that right?’). However, such studies have been lim-
ited by their methodology.
First, grammarians have mostly proceeded by looking for writ-
ten examples of tags according to a previewed set of tag forms. As a
consequence, their studies have missed tag types not recorded in
written forms and failed to appreciate the range of contextual mean-
ings of tag questions. For example, an observation of spoken Chi-
nese in use shows that tags in Chinese may be in the form of single
particles such as hã1 (written as the Chinese character ૸), which has
2 Cheng does not say explicitly that the sentence-final sĶ-m“-sĶ and kám-m“-sĶ
‘isn’t it?’ are tags, probably due to two factors. First, there is presumably no
pause between these constituents and their preceding part (therefore no comma
is used in Cheng’s examples, such as I Ĳ-hiáu Ĳng tĶsĶ-μ“-sĶ? ‘Can he use chop-
sticks or not?’). Second, these forms can also appear in split as a higher predi-
cate sĶ ‘be’/kám conjoined with its negative μ“-sĶ ‘not be’. However, consid-
ering the apparent comparability between these forms and Mandarin ᱟнᱟ
(shi-bu-shi ‘Is that right?’) in tag position, there is a strong and good reason
for them to be analyzed as tags when they appear in sentence-final positions.
48 Gao Hua
been widely used in northern Mandarin and now entered Putonghua

(Yin shichao 1999), o2 in Shanghainese, ho2 in Cantonese and ho31
in Chiuchow. All these particle tags3 tend to express a milder ques-
tioning tone than structural tags such as ᱟੇ (shi ma ‘Is that so?’)
and ሩнሩ (dui-bu-dui ‘Is that right?’).
Also, in traditional grammar books, the tag is described as typi-
cally an independent prosodic unit; hence a comma is used in writ-
ing before the tag. However, according to my observation, in oral
exchanges in Mandarin, or other Chinese dialects, or English, there
may be no pause at all between the tag and the preceding stem. In
other words, both may be delivered under the same prosodic con-
tour. Such a point may provide evidence for viewing the tag question
as a tag construction rather than a bi-partite structure.
Actually, in the case of Mandarin Chinese, although the sen-
tence-final question morpheme ma is not grammaticalized from a
question tag, there is an interesting parallelism between the question
tag and the sentence-final question morpheme in that both are ap-
pended as a constituent to a declarative form, and both may be de-
scribed as expressing speaker uncertainty about the accuracy of the
information put in the sentence to which they are attached. Compare:
(5) Ԇ䐏֐а䎧ᶕⲴੇ?
ta gen ni yiqi lai de ma?
3M.SG with 2SG together come PRT Q
‘Has he come with you?’
(6) Ԇ䐏֐а䎧ᶕⲴᱟੇ?
ta gen ni yiqi lai de shi ma?
3M.SG with 2SG together come PRT right Q
‘He has come with you, right?’
Such facts indicate that there may still be room for a more in-depth
investigation of the grammatical status of the tag question in Chi-
nese and its relationship with other question forms particularly from
a cognitive-functional viewpoint.
3 These tag particles are not absolutely bound, and there can be a little pause just
before them.
2.2 Pragmatic characteristics of the tag question
As mentioned earlier, the typical function of tag questions is to elicit

confirmation of or agreement on the statement from the addressee.
But more than that, the tag question in Chinese, like its counterpart
in other languages such as English, expresses the bias of the speaker
toward one answer. That is, it invokes an expected response from the
hearer, either in the affirmative or negative, depending on the con-
text of utterance and the communicative goals of the speaker (Givón
1984; Shao Jingmin 1996: 127; Zhang Bojiang 1997; Xu Shenghuan
1999). For example, comparing examples (5) and (6), it may be seen
that the latter involves a stronger epistemic bias of the speaker than
the former does.
In view of the strong bias of the speaker toward a confirming
or disconfirming answer, the previous research on the tag question in
Chinese describes it as expressing a lower degree of doubt than the
other question types (Xu Jie / Zhang Linlin 1993). In Xu Shenghuan
(1999) the tag question is also described as the typical ‘weak ques-
tioning’ (ᕡਁ䰞) type, which has to do with its syntagmatic relation
that the question tag is added after, thus asks about the truth of, the
statement which already expresses the speaker’s certain presump-
tion in epistemology.
Drawing insights from Cognitive Linguistics that language
structures may be motivated and shaped by human conceptual knowl-
edge, which, in turn, is related to human physical experience (Taylor
2002: 9), Niu Baoyi (2002) characterizes the conceptual basis of the
‘Statement + Q-tag’ form as BELIEF+DOUBT, which reflects an episte-
mological state that one may have a certain belief about a state, ac-
tion or event, but soon feel not sure of it and doubt surfaces immedi-
ately; hence the question tag always follows the statement stem (a
presumably cross-linguistic property of tag questions). Such a char-
acterization indicates that the tag question is iconically motivated
and reflects in some way the structure of experience (Haiman 1983).
When the speaker moves from making an assertion (or direction or
exclamation) which s/he believes, to giving expression of doubt or
uncertainty about what is asserted (or directed or exclaimed), a tag
question is formed. In the following example (7) of tag questions,
50 Gao Hua
ӺཙॱҼਧ [Today is the twelfth] may be viewed as the ‘speaker

belief’, and ሩੇ [Is that so?] the ‘speaker doubt’.
(7) ӺཙॱҼਧሩੇ˛
jintian shi’er hao dui ma?
today twelve date right Q
‘Today is the twelfth, right?’
Unfortunately, Niu’s description does not explain why the tag ques-
tion conveys a low degree of doubt. Comparing (7) and (8), and also
the earlier examples of (5) and (6), it can be noticed that the tag
questions express a stronger speaker assumption about the truth or
falsity of the proposition being asked about, and request confirma-
tion or agreement of that assumption.
(8) ӺཙॱҼਧੇ˛
jintian shi’er hao ma?
today twelve date Q
‘Is it the twelfth today?’
The value of previous studies is still limited in that the investigator

has described contextual senses of tags ranging from questioning,
requesting, being ironic, to threatening and/or challenging (Shao Jing-
min 1996: Chap 8), but failed to give a coherent theoretical account.
In the following sections of this chapter, tag questions in Chinese
will be subjected to a cognitive-functional study, which will reveal
how cognitive and communicative factors motivated the ‘Statement
+Q-tag’ question form and how the form is mapped to its various
pragmatic manifestations.
3. Cognitive-communicative motivations
of the tag construction
I argue that the combination of the statement stem and the Q-tag may
be regarded as a tag construction which has a constructional mean-
ing that expresses both speaker certainty and speaker uncertainty, in
perception of hearer knowledge or attitude, about the truth or accu-
racy of a statement. On the one hand, the form and meaning of the
tag construction are interdependent; on the other hand, the meaning
cannot be derived directly by adding together the meanings of the
statement and the Q-tag (Croft 2001).
Compared with previous research on tag questions in Chinese
which focused on how the tag leads a statement to a question and
relevant speaker belief to doubt, the present understanding of the tag
construction draws attention to the point that such dynamic episte-
mology about the issue concerned is related to how participants in a
conversation view the knowledge of each other. Put in another way,
the tag not only conveys the speaker’s attitude with respect to the topic
or statement at issue, it also relates to the hearer’s knowledge and/or
attitude (i.e. what the speaker thinks the hearer knows or believes). In
a similar vein, Norrick (1995) has argued for the role of tags such as
the English hunh in coding evidentiality in conversation. Cuenca (1997)
has also found that the tag construction is designed to be an interactive
construction that must be interpreted in a conversational exchange.
The interactive nature of the tag construction may be evidenced
clearly by its placement in conversational sequences like the follow-
ing one:
S
1 Sa Uh, (.) նަԆⲴ઒, ަԆ㲭❦␈བྷ㣡ഝᡁᜣਟ㜭ਚঐ
Uh, (.) dan qita de ne, qita suran tao da huayuan wo xiang keneng zhi zhan
but other GEN TOP other though Tao Da garden I think maybe only take
2 (1.1) Ⲯ࠶ѻҼॱӄᐖਣ.
baifenzhiershiwu zuoyou.
twenty-five-percent or-so
‘Uh, (.) but how about the others, among others, although victims in Taoda
Garden I guess may only take (1.1) 25 percent or so.’
3 (.)
52 Gao Hua
4 Jay [ᐞнཊ, Ⲯ࠶ѻйॱ. Ⲯ࠶ѻйॱ.

[chabuduo, beifenzhisanshi. beifenzhisanshi.
not-far-off thirty-percent thirty-percent
‘Not far off, 30 percent. 30 percent.’
5→ Sa [ሩ੗˛
[dui ba?
right PRT
‘right?’
In this excerpt, Sa was talking to her friend Jay about her estimation
of the percentage of SARS victims in a housing estate. After her
making the statement, there was a noticeable pause (line 3) before
she proceeded to check this information with Jay the hearer, with the
question tag ሩ੗ (dui ba ‘Is that right?’). Interestingly, at the same
time the tag was uttered (line 5), Jay took the floor and made a con-
firmation (line 4). Different from examples given in previous sec-
tions of this chapter, the early ‘arrival’ of the confirmation from the
hearer seemed to truncate the speaker’s tag question into two parts
(lines 1-2 and line 5). The declarative part of the tag question in lines
1-2 was virtually made a declarative question (Gao Hua 2003). And
it was possibly due to the lack of an immediate uptake from Jay, as
indicated by the noticeable pause in line 3, that the tag was used,
which may be seen as a further and a more explicit attempt by Sa to
invite verification from Jay of what had been said earlier.
With an understanding of the interactive nature of the tag con-
struction, some general characteristics about tag questions can be
well accounted for. First, tag questions only appear in conversations
where there are addressees actively present, and never in monologues
(monologues in a strict sense, with no ‘explicit’ or ‘implicit’ hearer).
It may be said that the tag question differs from other question types
in that it is not for asking of oneself. Compare (9) and the impossible
case in (10):
(9) (asking oneself) Ԇҏ㾱৫ੇ˛

ta ye yao qu ma?
he also will go Q
‘Is he also going?’
(10) (asking oneself) Ԇҏ㾱৫ˈᱟੇ˛

ta ye yao qu, shi ma?
he also will go right Q
‘He is also going, right?’
Second, the interactive property is also the key difference between

tag questions and utterances of pure speaker speculation (⍻ᓖ). With
the former, the speaker ‘asks a question’ of the hearer, while with
the latter, the speaker expresses his/her guess or estimation but does
not necessarily ‘pose a question’ (Lyons 1977: 755). Compare (11)
and (12):
(11) ཙՊл䴘ˈᱟ੗˛
tian hui xiayu, shi ba?
sky shall rain right PRT
‘It may rain, right?’
(12) ཙՊл䴘੗Ǆ/˛
tian hui xiayu ba./?
sky shall rain PRT
‘It may rain.’ / ‘Is it likely to rain?’
Considering the involvement of speaker epistemology and speaker-

hearer interaction in the formation of the tag construction, it may be
seen that tag questions have a significant place in linguistic theorising.
As Givón has said, grammar codes both propositional semantics and
discourse coherence (2001: 13). The tag construction may most expli-
citly illustrate such a view of grammar. Firstly, it displays syntactico-
semantic properties such that the question is in the form of a mini
question attached to a statement stem, hence conveying uncertainty
about the truth of the statement, which finally leads to requesting
confirmation or disconfirmation. But more than this, this question
form also involves the coding of a broader discourse-pragmatic func-
tion that involves speaker perception of hearer knowledge and/or
attitude in the discourse. Therefore, without taking such a function
into consideration, the true nature of tag questions cannot be fully
understood.
54 Gao Hua
4. Pragmatic characteristics of tag questions in Chinese
In this section, with the cognitive-functional account I have presented

of the tag construction, I will examine and explain some pragmatic
properties of tag questions in Chinese.
4.1 The canonical function of tag questions
As I have argued, the uncertainty about the stated information ex-

pressed by the tag is related to the speaker’s perception of relevant
hearer knowledge or attitude. It follows that the typical function of
the tag question is to check the validity of a statement against the
hearer’s knowledge or attitude. Pragmatically, this amounts to elicit-
ing confirmation or agreement from the hearer. It is this ‘checking’
function that distinguishes the tag question from yes-no questions
such as the ma-question in Chinese.
An examination of naturally occurring data reveals that tag
questions typically appear when the speaker shows an attitude or
understanding with respect to what has been said previously, and
would like to have this attitude or understanding checked and con-
firmed. For example, in the following excerpt, after hearing A’s words
saying articles on folk literature research are easier to get published,
B presented his opinion that this might be related to the fact that
since there were not many people in this field, hence less pressure
with getting their works published: with the tag, his opinion was in
the form of a hypothesis and a confirmation or agreement was re-
quested. Although A did not give a direct reply on this hypothesis,
his words were nonetheless an indirect confirmation of B’s opinion.
(13) A ≁䰤Ⲵᇩ᱃ˈ
minjian de rongyi,
folk GEN easy ‘Folk (literature studies) are easy (to get published).’
→ B ৽↓ҏ⋑ᴹཊቁӪ൘ᩎ૸ˈ
fanzheng ye meiyou duoshao ren zai gao hã
anyway also lack many people ASP do Q-TAG
‘Anyway there are not many people doing (folk literature), right?’
A ≁䰤Ⲵ㘼ф≁䰤᮷ᆖ⹄ウ୺ˈ
minjian de erqie minjian wenxue yanjiu a,
folk GEN moreover folk literature research TOP
䜭ྭࠐᒤ⋑ᴹᤋࡠᆖ⭏Ҷˈ…
dou hao ji nian meiyou zhaodao xuesheng le
even very a-few year no recruit student ASP
‘Folk, moreover, folk literature research has not recruited new postgrads in
the past few years.’
Besides, this canonical function of requesting confirmation is also

reflected by the semantic basis of some tag forms in a fairly explicit
manner. Although tags may take different forms in different languages,
certain tag-forming devices may be found across languages (McGregor
1995; Norrick 1995: 688; Cuenca 1997). Among others, as Cuenca
(1997) points out, invariant tags derived from lexical items related to
the concept of ‘truth’ seem to be used in most, if not all, languages4.
The following are examples of tag questions in some Western lan-
guages examined in this article. The tags in these questions all include
the word ‘truth’ or the like. Compare them with the tag form in Chi-
nese such as ᱟੇ:
(14) He has come with you, right? (English)
(15) Er ist mit dir gekommen, nicht wahr? (German)

3M.SG.NOM be+3SG+PRES with 2SG.DAT come+PAST.PART not right
(16) (No) ha venido contigo, ¿verdad? (Spanish)

(NEG) have+3SG+PRES come+PAST.PART with.you true
(17) Il (n’) est (pas) venu avec toi, pas vrai? (French)
3SG (NEG) be+3SG.PRES (NEG) come+PAST.PART with 2SG.ACC not true
4 Two other common sources summarized in Cuenca (1997) include: i) negative

adverbs (e.g., German nicht?, Spanish ¿no?, French non?); and ii) interjec-
tions focusing on different kinds of subjective meanings (e.g., English huh?/
eh?, German hm?, Spanish ¿eh?, French hein?). As she points out, these in-
variant tag forms are much more common than tags like the inflective tag in
English in the world’s languages.
56 Gao Hua
(18) Ԇ䐏֐а䎧ᶕⲴᱟੇ?
ta gen ni yiqi lai de shi ma?
3M.SG with 2SG together come PRT right Q
4.2 Degree of doubt
As reviewed in Section 2.2, it has been generally agreed that tag

questions convey a lower degree of doubt than other question forms
such as the ma-question or the A-not-A question in Chinese. The
reason for this, as repeated in previous studies, seems to be due to its
syntagmatic relation that the statement to which the Q-tag is attached
already expresses the speaker’s certain presumption in epistemology
(Xu Shenghuan 1999). However, as I have argued, such a contention
is debatable in view of the formal parallelism between the tag ques-
tion and the ma-question. In my opinion, the order of the statement
stem and the Q-tag more reflects the speaker’s epistemological state
of having a certain belief and then being not sure about that. The
comparatively lower degree of doubt as expressed by the tag ques-
tion should also be related to the speaker perception of hearer knowl-
edge and/or attitude in the discourse. Since adding a Q-tag after a
statement may be interactionally motivated, it may signal a lack of
self-confidence or a degree of non-assertiveness on the part of the
speaker; but on the part of the hearer, since the previous statement
has informed him/her of the speaker’s conviction or certain attitude,
such information will remain something of a speaker ‘hypothesis’
even after the question tag is added. It is this impression of speaker
assumption that leads to an impression of a low degree of doubt.
In accordance with this pragmatic feature, the tag typically
follows convictions, sometimes with an indication of assurance or
certainty, as shown by expressions like ᡰԕ (suoyi ‘so’), ᓄ 䈕 (yinggai
‘should’), and ⵏ (zhen ‘really’) in examples (19)-(21), respectively:
(19) ᡰԕ⢩ྭਲ਼ˈ૸˛
suoyi te haoshi, hã?
so very delicious Q-TAG
‘So it’s very delicious, right?’
(20) ᓄ䈕Ҡњ‫˛૸⊐⭥Ⲵ⭥ݵ‬
yinggai mai ge chongdian de dianchi hã?
should buy CL recharge GEN battery Q-TAG
‘(We) should have bought a rechargeable battery, right?’
(21) 䘈ⵏн䭉ˈањҏ⋑┿ˈ૸˛
hai zhen bucuo, yige ye mei lou, hã?
still really not-bad one also not leak Q-TAG
‘(It’s) really not bad, (the dumplings are) still well-stuffed (after the boiling),
right?’
4.3 Other pragmatic functions of the tag question
As I have said, the occurrence of the question tag relies on the

speaker’s view of the hearer knowledge. Logically, a tag is added in
cases the speaker becomes uncertain about the previous statement
when perceiving the hearer probably has more knowledge about the
topic or statement at issue. However, the tag may also be attached to
a statement when both the speaker and the hearer in a conversation
are aware that the hearer does not possess more knowledge. In such
cases, the tag question is not a genuine question checking informa-
tion by requesting confirmation; instead, it may serve various prag-
matic functions, ranging from rhetorical functions such as irony and
challenge to discourse functions of providing ‘discourse-management’
information. In these non-questioning functions, the tag questions
are usually understood as conveying little doubt and being unanswer-
able.
In the following excerpt, the two participants in the conversa-
tion were friends. A’s tag question in the arrowed line may be taken
as a rhetorical device conveying an ironical tone since her statement
“your drawer is really tidy” was apparently against the fact that B’ s
drawer was not tidy. It is such an incongruity that leads to the ironi-
cal sense. The attachment of the Q-tag ૸ actually makes A’s criti-
cism more indirect, thus functions to mitigate, or soften, her ironical
tone and this rhetorical question was finally taken as a mild disap-
proval, to which B expressed his agreement in reply.
58 Gao Hua
(22) → A ֐䘉ᣭቹⵏཏᮤ喀Ⲵ૸ˈ
ni zhe chouti zhen gou zhengqi de hã,
you this drawer really enough tidy PRT Q-TAG
‘This drawer of yours is really tidy, isn’t it?’
B ᱟᴹ⛩ҡˈ...
shi youdian luan, …
empha a-little disorder
‘(It’s) a little untidy indeed, …’
The following example (23) shows another non-questioning function

of the tag question, that of conversation management. When the speaker
A moved from relating an anecdote about her boyfriend to another
(about ‘that letter’; both illustrating the man’s personality), she used
two tags “ᱟੇ” and “૸” together after the indicated topic shift, which
allows the speaker to check that the hearer is following the topic shift
in the conversation and/or is aware of the relevant background infor-
mation. Note that after the tag question, the speaker did not wait for a
full response from the hearer (who actually gave a back channel agree-
ment), but went on with the story-telling. In view of this, the tag question
(in the arrowed line) is not a genuine question, but serves to involve
the hearer in the conversation and facilitate the information exchange.
(23) A …
→ 䛓ཙᡁн੺֐䈤䛓њؑᱟੇ૸.
na tian wo bu gao ni shuo na ge xin shi ma hã.
that day I not tell you say that cl letter right Q Q-TAG
‘Didn’t I tell you that day about that letter?’
[*Ԇ䐏ᡁ䈤ˈ䈤䛓њؑˈᱟ࠶є⅑߉Ⲵˈ...
ta gen wo shuo, shuo na ge xin, shi fen liangci xie de, …
he with I say say that CL letter EMPHA divide two-times write PRT
‘He told me, saying that letter was written twice
B
[୺.
huh
‘Huh.’
The two functions of the tag question in Chinese as exemplified above

happen to fall into the affective category in Holmes’ (1982) distinc-
tion between modal and affective question tags. The affective tags
differ from the modal tags in that they are used not to signal uncer-
tainty on the part of the speaker, but to indicate concern for the ad-
dressee. Moreover, the two Chinese examples above also illustrate

the two subcategories of affective tags respectively: the softener (as
in (22)), which conventionally mitigates the potential face-threaten-
ing force of the previous statement, and the facilitative tag (as in
(23), which invites the listener’s attention to the speaker’s assertion
(either to shift or retain the floor).
Based on my cognitive-communicative account of the tag ques-
tion, the affective functions of the tag question may be understood as
a functional shift of its canonical questioning function. This shift in
function is crucially subject to pragmatic factors such as how partici-
pants in a conversation perceive each other’s epistemology with re-
gards to the statement in question. When both the speaker and the
hearer are aware the hearer owns more knowledge, the tag question
is taken to be a real question inviting the hearer’s opinion. Otherwise
the question may appear to be ‘unreasonable’ and ‘unanswerable’,
hence serving some affective function.
Besides, the pragmatic function of a tag question may also
depend on its placement in the conversational sequence. For example,
the facilitative tags are found to occur frequently in longer turns (e.g.,
B’s turn in the following excerpt) where speakers were developing
ideas, presumably to monitor whether the hearer followed the
speaker’s train of thought and support it. None of the mid-turn tag
questions received a response other than a back-channel (and there
could be eye contact or nodding or smiling).
(24) нᱟԆ䛓њ㫉ᗇᴤহᇣˈ(ㅁ)
A: £º²»ÊÇËûÄÇ¸öÃÉμÃ¸üÀ÷º¦£¬(Ð¦)
ⵏᱟ䘉ṧ‫֐ݯ‬䈤ᴹᰦ‫Ⲵⵏى‬ᱟˈ䈤䎧ᶕ䈱䈤∄䈱ˈᴹᰦ‫ى‬ᜣᜣˈ
B: £ºÕæÊÇÕâÑù¶ùÄãËμÓÐÊ±ºòÕæμÄÊÇ£¬ËμÆðÀ´Ë-Ë
μ±ÈË-£¬ÓÐÊ±ºòÏëÏë£¬
䈦ੰᡁ䘈ᜣ઒∄ྲ䈤Ӻ‫ⴻ⋑ݯ‬Җ૸ ૸ˈ
ÚÀÑ½ÎÒ»¹ÏëÄØ±ÈÈçËμ½ñ¶ùÃ»¿´Êé¹þ£¬
ᜣӺ‫ⴻ⋑ݯ‬ҖӪᇦ⦻㔤ᇦҏн⸕䚃ᒢో઒૸ ૸㛟ᇊⴻҖ઒૸ ૸ˈ
Ïë½ñ¶ùÃ»¿´ÊéÈË¼ÒÍõÎ¬¼ÒÒ²²»ÖªμÀ¸ÉÂïÄØ¹þ¿Ï¶¨¿´ÊéÄØ¹þ£¬
а৫ⴻԆˈ࡛䈤Ӻ‫ⴻ⋑ݯ‬Җ᰾‫ݯ‬ҏ⋑ⴻਾ‫ݯ‬ҏ⋑ⴻ㘱нⴻҖ
Ò»È¥¿´Ëû£¬±ðËμ½ñ¶ùÃ»¿´ÊéÃ÷¶ùÒ²Ã»¿´ºó¶ùÒ²Ã»¿´ÀÏ²»¿´Êé
˄ᑖㅁ˅Ӫᇦ䘉њҏ䇨Ӫᇦᓅᆀྭ੗֐䈤ਚ㜭䘉Ѹᜣˈ
£¨´øÐ¦£©ÈË¼ÒÕâ¸öÒ²ÐíÈË¼Òμ××ÓºÃ°ÉÄãËμÖ»ÄÜÕâÃ´Ïë£¬
նᱟ֐䈤ˈࡠҶ∄ྲ䈤㘱ᐸˈ
μ«ÊÇÄãËμ£¬μ½ÁË±ÈÈçËμÀÏÊ¦£¬
䈤Ӕ֌ъⲴᰦ‫ى‬Ҷ૸ ૸ˈ
Ëμ½»×÷ÒμμÄÊ±ºòÁË¹þ£¬
Ӕь㾯ˈӪᇦ਼ṧᐳ㖞㔉֐‫Ⲵؙ‬ь㾯Ӕˈ
½»¶«Î÷£¬ÈË¼ÒÍ¬Ñù²¼ÖÃ¸øÄãÁ©μÄ¶«Î÷½»£¬
ૡ䘉њ߉ᗇнྭнྭ৽↓ࡂ᣹аㇷӔࠪ৫ҶԆྭˈ
ÔÛÕâ¸öÐ´μÃ²»ºÃ²»ºÃ·´Õý»®À-Ò»Æª½»³öÈ¥ÁËËûºÃ£¬
߉ᗇнྭнྭᴰਾҏӔнࠪ৫
Ð´μÃ²»ºÃ²»ºÃ×îºóÒ²½»²»³öÈ¥
˄ᑖㅁ˅ᩎᗇ㘱ᐸ䜭⋑࣎⌅
£¨´øÐ¦£©¸ãμÃÀÏÊ¦¶¼Ã»°ì·¨
ањањᴸӔⲴь㾯Ԇ䶎ᤆ‫ؙ‬ᴸԘᴸ䘈ӔнҶˈ
60 Gao Hua
A: ᱟੇˈ
B: 䙇ᗇн㹼ˈᴰਾ֐㾱ᱟ䍘䟿⢩儈੗ҏ⋑㿱ᗇ૸
૸⋑㿱Ԇ∄ૡ䘉њྭˈ
к⅑ᡁԜ‫ؙ‬...ˈ
4.4 Functions of tag questions and tag forms
The distinction between modal tags and affective tags in function

seems also to be matched by tag forms in Chinese. As said earlier,
tags in Chinese may either take the form of a mini-question contain-
ing a rather concrete lexical meanings of ‘true’ or ‘right’, such as
нᱟੇ (bu shi ma ‘Isn’t that so?’), ᱟ੗ (shi ba ‘Is that so?’) and
ሩнሩ (dui-bu-dui ‘Is that right?’), or be in the form of a single
particle such as ૸ in Mandarin or ho2 in Cantonese. My observation
of naturally occurring data suggests that there seems to be an inter-
esting division of labor between these two types of tag forms. That
is, when the speaker wishes to check information, he/she tends to use
the mini-question tag, as shown in example (2) of Section 1; when
the speaker does not appear to be in any doubt about the fact or view
expressed in the statement, the particle tag is often used, as shown in
the example of excerpt (24) in Section 4.3. What is interesting is that
when the mini-question tag is used in the affective function, it is
usually said in a reduced form as shown in example (23). Therefore,
the tag forms and their functions seem to illustrate a certain iconicism
in that the question tag in the questioning function is the focus, while
in tag questions of the non-questioning function it is the preceding
statement, not the affective tag, that is the focus of information.
5. Conclusion
In this chapter I have proposed a cognitive-communicative account

of tag questions in Chinese. It is argued that the formation of this
question is not simply a combination of a statement stem plus a ques-
tion tag, but may be regarded as a tag construction which has a con-
structional meaning that expresses not only speaker certainty but also
speaker uncertainty, in perception of hearer knowledge or attitude,
about the truth or accuracy of a statement. On the basis of this, the
pragmatic characteristics of the tag question in Chinese are explained.
It is suggested that whether the tag question functions to check infor-
mation (by requesting confirmation or agreement) or to serve affec-
tive functions (such as mitigating the otherwise impoliteness in the
previous statement or facilitating discourse management) crucially
depends on the interpersonal perception of participants’ subjective
epistemology with regards to the topic or statement in question.
Moreover, this chapter also suggests an iconic form-function map-
ping for tag questions in Chinese. The mini-question tags are related
to uncertainty about ‘truth’, hence typically used to ask about the
truth or accuracy of a statement; while the particle tags are more
often found with the affective, especially the facilitative, function of
the tag question.
References
References in English
Cheng, R.L. 1997. Taiwanese question particles,ǉਠǃॾ䈝Ⲵᰦオǃ

⯁䰞о੖ᇊǊਠे˖䘌⍱ࠪ⡸‫ޜ‬ਨ, 231-271.
Chisholm Jr., W.S., et al. (eds) 1984. Interrogativity: a Colloquium
on the Grammar, Typology, and Pragmatics of Questions in
Seven Diverse Languages. Amsterdam: John Benjamins.
Croft, W. 2001. Radical Construction Grammar: Syntactic Theory in
Typological Perspective. Oxford: Oxford University Press.
Cuenca, J.M. 1997. Form-use mappings for tag questions. In Liebert,
W.-A. et al. (eds) Discourse and Perspective in Cognitive Lin-
guistics. Amsterdam: John Benjamins, 3-19.
Gao, Hua 2003. Declarative questions in Chinese criminal court exami-
nations. Paper presented at the 8th International Pragmatics
Conference, The University of Toronto, July 13-18, 2003.
62 Gao Hua
Givón, T. 1984. The speech-act continuum. In Chisholm Jr., W.S., et

al. (eds) Interrogativity: a Colloquium on the Grammar, Typo-
logy, and Pragmatics of Questions in Seven Diverse Languages.
Amsterdam: John Benjamins, 245-254.
Givón, T. 2001. Syntax: An Introduction. (2 vols.) Amsterdam: John
Benjamins.
Givón, T. 2002. Bio-Linguistics: The Santa Barbara Lectures. Amster-
dam: John Benjamins.
Haiman, J. 1983. Iconic and economic motivation. Language 59: 781-
819.
Holmes, J. 1982. The functions of question tags. English Language
Research Journal 3, 40-65.
Hopper, P. and Traugott, E.C. 1993. Grammaticalization. Cambridge:
Cambridge University Press.
Li, N.C. / Thompson, S. 1981. Mandarin Chinese: a Functional Ref-
erence Grammar. Berkeley: University of California Press.
Lyons, J. 1977. Semantics. (2 vols.) Cambridge: Cambridge Univer-
sity Press.
McGregor, W. 1995. The English ‘tag question’: a new analysis, is(n’t)
it? In Hasan, R. / Fries, P. (eds) On Subject and Theme: a Dis-
course Functional Perspective. Amsterdam: John Benjamins,
91-121.
Niu, Baoyi 2002. Belief and Doubt: A Pragmatic Study of Tag Ques-
tions. Unpublished ph. D thesis. Kai Feng: He Nan University.
Norrick, N.R. 1995. Huhn-tags and evidentiality in conversation,
Journal of Pragmatics 23, 687-692.
Taylor, J.R. 2002. Cognitive Grammar. Oxford: Oxford University
Press.
Ultan, R. 1978. Some general characteristics of interrogative systems.
In Greenberg, J.H. et al. (eds) Universals of Human Language
4: Syntax. Stanford, California: Stanford University Press, 211-
248.
References in Chinese
Lin, Yuwen ᷇㼅᮷ 1985. 䈸⯁䰞ਕˈ ǉѝഭ䈝᮷Ǌ1985 ᒤㅜ 2 ᵏ˖

91-98.
Liu, Yuehua ࡈᴸॾ 1987. ⭘ĀੇāⲴᱟ䶎䰞ਕ઼↓৽䰞ਕ⭘⌅∄
䖳ˈ䖭ѝഭ⽮、䲒䈝䀰⹄ウᡰ⧠ԓ≹䈝⹄ウᇔ㕆ǉਕර઼ࣘ
䇽ǊेӜ˖䈝᮷ࠪ⡸⽮ˈ117-138.
Liu, Yuehua / Pan, Wenyu / Gu, Hua ࡈᴸॾǃ█᮷ၡǃ᭵丑 2001.
ǉᇎ⭘⧠ԓ≹䈝䈝⌅ǊˈेӜ˖ཆ䈝ᮉᆖо⹄ウࠪ⡸⽮.
Lü, Shuxiang ੅਄⒈ 1985. ⯁䰞ǃ੖ᇊǃ㛟ᇊˈ ǉѝഭ䈝᮷Ǌ1985
ᒤㅜ 4 ᵏ˖241-250.
Niu, Baoyi ⢋‫؍‬ѹ 2001. 㤡≹䈝䱴࣐⯁䰞ਕ䈝⌅ॆ∄䖳ˈǉཆഭ
䈝Ǌ2001 ᒤㅜ 2 ᵏ˖49-54.
Shao, Jingmin 䛥ᮜ᭿ 1996.ǉ⧠ԓ≹䈝⯁䰞ਕ⹄ウǊк⎧˖ॾь
ᐸ㤳བྷᆖࠪ⡸⽮.
Xu, Jie / Zhang, Linlin ᗀᶠǃᕐ᷇᷇ 1993. ⯁䰞〻ᓖо⯁䰞ਕᔿˈ
ǉ≹䈝᧿߉䈝⌅ॱ䇪Ǌ䜁ᐎ˖⋣ইᮉ㛢ࠪ⡸⽮ˈㅜ 8 ㄐ
132-145.
Xu, Shenghuan ᗀⴋẃ 1999. ⯁䰞ਕ᧒䈒࣏㜭Ⲵ䗱〫ˈǉѝഭ䈝
᮷Ǌ1999 ᒤㅜ 1 ᵏ˖3-11.
Yin, Shichao ቩц䎵 1999. 䈤䈝≄䇽Ā૸ā઼Ā૸āᆇਕǉᯩ䀰Ǌ
1999 ᒤㅜ 2 ᵏ˖95-103.
Zhang, Bojiang ᕐ՟⊏ 1997. ⯁䰞ਕ࣏㜭⩀䇞ˈ ǉѝഭ䈝᮷Ǌ1997
ᒤㅜ 2 ᵏ˖104-110.
Zhu, Dexi ᵡᗧ⟉ 2000.ǉ䈝⌅䇢ѹǊˈेӜ˖୶࣑ঠҖ侶.
NORIFUMI ITO
Information in Discourse and Language

as the Vehicle of Thought
Discourse is composed of linguistic expressions1 through which we

are able to represent the understanding of symbol systems2 in the
world and are supposed to affect the mental state of others as the
result of acts of communication. So far, in current linguistic theory, it
has been generally assumed that one sentence is a unitary system to
which linguistic analyses, especially syntactic and semantic analy-
ses, are applied and at which we can achieve what we intend to do; in
other words, acts of communication. It has to be borne in mind, how-
ever, that this general assumption as to the linguistic units to which
the theoretical analyses are applied is largely technical and termino-
logical; in other words, it can be generally assumed that in the daily
situation in which we communicate with others, we have a facility
for constructing the cognitive image of reality (or world), represented
in linguistic expressions with the help of the merger of the combina-
tion of word meaning and sentence structure and multiple sources of
information about meaning. As argued by Jaszczolt (2005: 6), “The
representation that is underspecified, or a representation that allows
for an ambiguity of readings, becomes complete after the addition of
information from the surrounding discourse.” In other words, as
Jaszczolt (1999, 2002, 2005, 2006) mentions, in the face of linguis-
1 We use the phrase ‘linguistic expressions’ as the notion including utterances,

written sentences, and propositional content expressed by utterances and writ-
ten sentences. In the following, we differentiate between these notions accord-
ing to the exposition.
2 By the phrase ‘symbol system,’ we mean the mental image as to the world
from the standpoint relativized in space and temporality. For detail, refer to
Heidegger (1953).
66 Norifumi Ito
tic expressions or utterances the meaning of which we cannot make

clear only through the combination of word meaning and sentence
structure, we have to resort to pragmatic inference, social-cultural
knowledge, and so on. Consequently, it is plausible to assume that
we should rely on the merger of information from multiple sources
to get more than the mere propositional meaning of linguistic ex-
pressions or utterances. Therefore, in this research, we aim to define
the function of information structure in linguistic expressions or ut-
terances and the discourse information surrounding linguistic expres-
sions or utterances with reference to R-dislocation construction in
English and Japanese respectively.
The present chapter is organized in the following way. Sec-
tion 1 contains the definitional foundations of information structure
included in what is physically written or uttered. Information struc-
ture is the reflection of the relation which is systematically built into
discourse. In particular, Section 1 assesses a linkage between the
mental state of interlocutors and the information emerging out of the
propositional meaning of linguistic expressions and utterances. Sec-
tion 2 touches on the problem as to the addresser’s attitude towards
propositional meaning; in other words, the intention or attitude of
the addresser reflected in linguistic expressions, and utterances are
brought into investigation with reference to English inversions. It is
evident from the analysis in Section 2 that the adoption of only one
construction as well as one word results in different interpretation of
the reality and affects the mental state of others in the different way
from canonical constructions. In Section 3 and Section 4 we focus
on the role of discourse information involved in the construction of
so-called R-dislocation constructions in English and Japanese, show-
ing that non-canonical constructions have the specific function re-
flected in the information structure included in them, as mentioned
above.
Information in Discourse and Language as the Vehicle of Thought 67
1. Information in Discourse: Setting the Scene
Within the linguistic research carried out so far, there have been con-
tentious proposals as to the discourse function of information repre-
sented in linguistic expressions and utterances. The unified accounts
of information emerging from the surrounding discourse are heavily
burdened with allocating its complicated character to various presup-
posed domains of linguistic research. That is, discussions of informa-
tion in discourse inevitably give rise to the problem of syntax-prag-
matics and semantics-pragmatics interfaces. At this level of linguistic
investigation, that is, at the level of information in discourse, we are
supposed to make an explanation of the mental state of both addresser
and addressee in discourse in addition to that of the structural aspect
of what is physically written and uttered. The complicated links be-
tween pragmatics and syntax or semantics are also reflected in the
usage of terms for representing information in discourse. In the Prague
School, for instance, the term ‘Functional Sentence Perspective’ is
available. In Halliday’s publications (e.g., 1967) , the term ‘Informa-
tion Structure’ or ‘Theme’ is preferable. Chafe (1976) adopted the
term ‘Information Packaging’ and Vallduvi (1990) used the term
‘Informatics.’ In spite of the diversity of term usage mentioned here,
the expression ‘Information Structure’ or ‘Information Packaging’ is
preferable to others for representing information systematically built
into discourse for the specific purpose of expressing the mental state
of addresser and the way of affecting the mental state of addressee.
As a further step for setting the scene for our argumentation,
we should clarify the definition of information in discourse. Accord-
ing to Chafe (1976), information packaging has to do primarily with
‘how the message is sent.’ It strongly suggests that what information
packaging emerging out of linguistic expressions or utterances has
to do at first is to reflect the discourse strategy of addresser and the
way of affecting the mental state of addressee. After the completion
of the primary function of expressing the way of conveying what the
addresser intends to do, information packaging has to do with the
message itself; in other words, as a secondary function, information
packaging fulfills a role of expressing the propositional meaning.
68 Norifumi Ito
(1) I have been using the term packaging to refer to the kind of phenomena at issue
here, with the idea that they have to do primarily with how the message is sent
and only secondarily with the message itself, just as the packaging of tooth-
paste can affect sales in partial independence of the quality of the toothpaste
inside. (Chafe 1976: 28)
In a similar vein, Prince (1981b) has an explanation as to the func-

tion of information packaging. It is assumed that information pack-
aging is a postulate about the addressee’s assumptions and beliefs
and strategies towards the meaning of what is physically written or
uttered. From the point of view of information packaging, what the
addresser should deal with first of all is the hypotheses about the
addressee’s mental state. In other words, information packaging is
the reflection of the addresser’s intention to affect the addressee’s
mental state in some way or other.
(2) That is, information packaging in natural language reflects the sender’s hypo-
theses about the receiver’s assumptions and beliefs and strategies. (Prince
1981b: 224)
Prince (1986) states the function of information more clearly as fol-

lows; nota bene, the notion ‘information’ considered within the analy-
sis (3) in Prince (1986) should be dubbed ‘information packaging’ or
‘information structure.’ In many cases, the addresser attempts to con-
struct a proposition so as to affect and change the mental state of the
addressee by structuring information packaging (or information struc-
ture) in what is physically written or uttered. After the production of
utterance by the addresser, the addressee attempts to interpret the
utterance. At first, she/he attempts to make clear what the addresser
intends to convey by means of only the combination of word mean-
ing and sentence structure, that is, the mere propositional meaning.
If there is more than what the combination of word meaning and
sentence structure conveys, however, the addressee cannot but resort
to information from pragmatic inference, social- and cultural-know-
ledge and so on.
(3) Information in a discourse does not correspond to an unstructured set of pro-

positions; rather, speakers seem to form their utterances so as to structure the
information they are attempting to convey, usually or perhaps always in ac-
cordance with their beliefs about the hearer: what s/he is thought to know, what
s/he is expected to be thinking about. (Prince 1986: 208)
As argued above, we can conclude that information in a discourse is

a structured unit through which the addresser attempts to convey
what she/he intends to do and also affect the mental state of the ad-
dressee, with the result that the addresser attempts to create the new
representation of reality within the mental state of the addressee.
Consequently, information in a discourse can be better dubbed ‘in-
formation structure’ rather than ‘information packaging.’ The notion
of ‘information structure’ is correctly described in Lambrecht (1994)
as follows.
(4) INFORMATION STRUCTURE: That component of sentence grammar in

which propositions as conceptual representations of states of affairs are paired
with lexicogrammatical structures in accordance with the mental states of in-
terlocutors who use and interpret these structures as units of information in
given discourse contexts. (Lambrecht 1994: 5)
In accordance with the definition of discourse information, it can be

emphasized that we have no strong intention to affect and change the
mental state of addressee in the case of availing ourselves of canoni-
cal construction. Indeed, it is almost impossible to clarify the strength
of our intention to do something, but it can be generally assumed that
canonical structured sentences or utterances give rise to the default
interpretation which addressees can reach only through the under-
standing of the combination of word meaning and sentence struc-
ture. Indeed, the default interpretation of linguistic expression or ut-
terance is not capable of having the strength of influencing the
addressee’s mental state as to reality. In order for the mental state of
addressee to be changed to a large degree, however, we avail our-
selves of the sentences or utterances with the reordered constituents,
that is, non-canonical sentences and utterances. As for the proposi-
tional content expressed only through word meaning, there is a dif-
ference between canonical and non-canonical expressions. In the face
of the non-canonical linguistic expressions, the addressee has to in-
terpret what the addresser intends to convey with the help of infor-
mation from pragmatic inference and social- and cultural-knowledge.
70 Norifumi Ito
The procedure of this interpretation inevitably affects and changes

the mental state of the addressee. The relation between information
in a discourse and proposition is given a detailed exposition like the
following quotation from Dahl:
(5) Let us consider one important use of declarative sentences, namely as means to
influence the addressee’s picture of the world. In such cases, the speaker as-
sumes that the addressee has a certain picture – or model — of the world and he
wants to change this model in some way. We might then identify THE OLD or
THE GIVEN with the model that is taken as a point of departure for the speech
act and THE NEW with the change or addition that is made in this model. OLD
will here be equivalent to PRESUPPOSED in one sense of the term. We can say
that the addressee receives ‘new information’ in the sense that he comes to know
or believe more about the world than he did before. What he believes may be true
or false – the information he gets about the world may be correct or incorrect.
If we accept that last statement, it follows that the object of his belief or the new
information must be something which is capable of being true or false – that is
what is usually called a proposition. Let us therefore call this kind of information
PROPOSITIONAL INFORMATION. (Dahl 1976: 38)
2. Discourse Information and the Markedness

of Inversions in English
2.1 Information within English Inversions
According to the analysis by Birner (1998), inversions will be pre-

sented to clarify the discourse function of them in the following ex-
position; five patterns of information that the preposed and postposed
NPs show in inversions are investigated in the following exposition
respectively. First of all, consider the token in which the preposed
constituent represents a discourse-old information while the postposed
constituent represents a discourse-new information.
(6) We have complimentary soft drinks, coffee, Sanka, tea, and milk. Also compli-
mentary is red and white wine. We have cocktails available for $2.00. (Birner
1998: 312)
The preposed constituent ‘complimentary’ in (6) represents infor-

mation evoked in the previous clause; the postposed ‘red and white
wine’ provides information new to the discourse.
Now let us turn our attention to the inversion in which both the
preposed and postposed elements are already evoked in the prior con-
text. The pragmatic constraint, that is ‘recency effects’, is inevitable
for the exposition of constituents’ distribution in inversion (7). The
referent of ‘it’ in (7) is ‘the war between forces within the party and
within the coalition.’ This referent is mentioned more recently than
the referent of the postposed element ‘Jesse Jackson’.
(7) Yes, this is no ordinary general election. ‘Evans is a Democrat; Daley is a

Democrat. Different Democrats have different points of view about the city of
Chicago and its politics,’ Jackson noted. ‘The war between forces within the
party continues, and within our coalition.’ Standing in the middle of it all is
Jesse Jackson. (Birner 1998: 313)
Let us now see inversion (8), in which both the preposed and the
postposed elements are mentioned in the prior discourse. Addition-
ally, we should note that only one phrase in the prior discourse can
evoke both the preposed and postposed constituents in inversion. The
preposed NP ‘authors’ is mentioned in the expression ‘Montagnier
and Barre, whose English was more fluent than Chermann’s,’ that is,
the expression for the larger set of authors. On the contrary, the
postposed NP ‘Barre’ refers to the subset of authors; it is evoked by
the NP ‘Montagnier and Barre, whose English was more fluent than
Chermann’s.’ All in all, the evoking phrase can refer to the superset
directly, while the postposed NP ‘Barre’ can be evoked by a set infer-
ence with the result that the superset is more familiar than the subset:
(8) Over a weekend, Montagnier and Barre, whose English was more fluent than
Chermann’s, hammered out a manuscript reporting the isolation of their new
retrovirus from Frederic Brugiere, who would henceforth be known in the sci-
entific literature as BRU. Listed first among the authors, the position tradition-
ally reserved for the researcher who has made the greatest contribution to the
work, was Francoise Barre. (Birner 1998: 316)
As a further step towards unified treatment of inversions, a prag-

matic factor such as inference should be brought in to look at the
72 Norifumi Ito
property of inversions. Now let us consider the following examples

from Birner (1994). The preposed elements in (9) represent inferable
information. On the basis of the description of the NP ‘labor sav-
ings,’ we can make a reasonable inference that something has been
eliminated.
(9) Labor savings are achieved because the crew is put to better use than cleaning
belts manually; also eliminated is the expense of buying costly chemicals. (Birner
1994: 248)
Now we should investigate the information status of inferable ele-

ments in the familiarity scale provided in Prince (1981b). The scale
can be defined as follows. In the scale, unused information is as-
sumed to be already known to the hearer, but not familiar in the con-
text. Containing-Inferrable is supposed to be included within the in-
ferable NP, and Brand-New-Anchored information includes the
element which is familiar to the discourse.
(10) Evoked > Unused > Inferrable > Containing-Inferrable > Brand-New-Anchored
> Brand-New
As shown in the previous exposition, the more familiar constituent is

supposed to be preposed in inversion. According to scale (10), how-
ever, the distribution of both elements in inversion (11) violates the
pragmatic constraint.
(11) Enclosed Are the GEICO Home Insurance Rates You Requested. (Birner 1994:
250)
The expression ‘Enclosed Are the GEICO Home Insurance Rates

You Requested’ is printed on an envelope. So, the default situation is
that there exists an envelope in front of the reader. Consequently, the
preposed element in (11) can be conceived of as that representing
inferable information. However, the postposed NP is assumed to be
familiar to the reader but to be new to the discourse with the result
that it can be considered to be unused information in the scale (10).
Then we conclude that the analysis by Prince (1981b) cannot give an
exposition as to information status of both preposed and postposed
constituents in (11). However, Birner (1994) suggests that both in-
ferable information and evoked information can be treated as a single

unit of discourse-old in case of determining the felicity of inversions.
However, it is irrational to treat inferable information on an equal
footing with evoked information only with regard to the acceptabil-
ity of inversions.
Here we have to reconsider the scale of information status given
above. It can be assumed that the notion of activation proposed in
Lambrecht (1994) will be useful to explain the distribution of NPs in
English inversions. Lambrecht (1994) proposed that identifiable in-
formation can be defined in terms of the notion of activation and that
it can be categorized into three groups in proportion to the degree of
activation: Active, Accessible and Inactive. Moreover, accessible in-
formation can be grouped into three subclasses: Textually,
Situationally and Inferentially, as follows.
(12) Active > Accessible ( Textually / Situationally / Inferentially ) > Inactive
According to the above criterion, active information is evoked in the

recent flow of time. It cannot keep its cognitive status after another
element appears as a next evoked information. Accessible informa-
tion is inferable information or the equivalent in scale (10). Inactive
information can keep its cognitive status in the hearer’s mind, but it
is not active or accessible in the current discourse. Here, let us con-
sider the problem concerning (11) by means of information scale (12).
The preposed element of (11) ‘enclosed’ is defined as situation-
ally accessible information because of the high accessibility of an
envelope in the discourse. As for the postposed element, its repre-
sentation is stored up in the reader’s mind so that it can be classified
as inactive information. Consequently, the preposed constituent in
(11) is more activated than the postposed constituent; it follows that
inversion (11) is consistent with the pragmatic constraint.3 As op-
posed to the proposal in Birner (1994), we can conclude that even
English inversions have to conform to the single pragmatic constraint.
3 Refer to Ito (1999) for more information as to other English inversions.

74 Norifumi Ito
2.2 Discourse Function of English Inversions

and Mental State of Interlocutors
As argued in the above section, English inversions adhere to the prag-

matic constraint old (or more familiar) information precedes new (or
less familiar) information in the sentence. The next question we have
to ask is the function of inversions in a discourse, or their raison
d’être Based upon the difference of sentences (13), Rochemont/
Culicover (1990) and Bresnan (1994) give an explanation of the in-
versions in terms of presentational focus..
(13) a. In a little white house lived two rabbits.

#It/#The house was the oldest one in the forest, and all the animals worried
that someday it would come crashing down.
b. In a little white house lived two rabbits.
They/The rabbits were named Flopsy and Mopsy, and they spent their days
merrily invading neighborhood gardens.
c. Two rabbits lived in a little white house.
i. It/The house was the oldest one in the forest …
ii. They/The rabbits were named Flopsy and Mopsy … (Birner 1994: 240)
The preposed element ‘the house’ in inversions (13a) and (13b) can-
not play a role as the topic for the following clause, while the
postposed element ‘the rabbits’ becomes the topic of the following
clause. It has to be remembered that canonical sentence (13c) makes
it possible for both NPs to occur as the subject in the following clause.
Rochemont and Culicover (1990) conceive of the difference
of the acceptability of the above sentences as the evidence that the
postposed element in inversion has a special function, that is a pre-
sentational focus. In a similar vein, Bresnan (1994) argues that the
postposed NP plays a role as a presentational focus. The subject NP
is assumed to be inverted to introduce the referent on the scene de-
scribed by the preceding discourse. Here let us see the definition of
the presentational focus by Bresnan (1994).
(14) In presentational focus, a scene is set and a referent is introduced on the scene
to become the new focus of attention. In the core cases, a scene is naturally
expressed as a location, and the referent as something of which location is
predicated – hence, a theme. This imposes a natural selection of the <th loc>
argument structure. (Bresnan 1994: 90)
Pace the above definition in Bresnan (1994), however, we can pro-

vide the following example as counter-evidence. In the example, the
postposed NP ‘Nusseibeh himself ’ is already evoked in the prior dis-
course. It follows that we cannot restrict the function of inversion
only to the presentational focus.
(15) Nusseibeh’s unusual predicament causes concern all around. His friends fear
that Arab hard-liners will turn on Nusseibeh, thinking he is an Israeli ally.
The Israelis, who certainly want to squelch the 17-month-old uprising in
the West Bank and Gaza Strip, are under intense pressure from the United
States not to jail moderates who may figure in their election proposal for the
territories occupied since the 1967 war.
Most immediately affected is Nusseibeh himself. (Birner 1994: 237-238)
In addition to the discourse function to introduce a new referent into

a discourse, inversion is supposed to have another raison d’être that
we have to clarify in the following exposition. It is shown in Green
(1980, 1982) and Schmidt (1980) that English inversions have distinct
meanings and functions; Stein (1995) cites the following example to
capture the association of the reordering of constituents with the dis-
tinct meanings and functions of those constructions. Examples (16)
represent a cline of emotional expression. (16a) can express the most
heart-felt attitude towards the proposition salva veritate and (16c) is
the least; it is assumed that the expression with the unmarked word
order, that is (16c), never has an additional meaning or less so. Con-
sequently, we can conclude that there is an association of the reorder-
ing of constituents with the distinct meanings of those constructions.
(16) a. Bitterly did they repent their decision.
b. Bitterly they repented their decision.
c. They repented their decision bitterly. (Stein 1995: 132)
For the purpose of making clear the generation of additional mean-

ing by means of marked word order, Stein (1995) proposes a more
detailed account of (17).
(17) Him I killed. (Stein 1995: 133)
76 Norifumi Ito
The marked word order in (17) engenders “the evocation of a pre-

suppositional set of candidates from which a choice is made, the
other candidates being discarded” (Stein 1995: 133). In addition, it is
argued that structure (17) has at least two meanings. One of them is
a discourse aspect: “the candidates have been introduced in previous
discourse” (Stein 1995: 133). The second is an affective component:
“it contains an affective component to the extent that the candidate
singled out is in contrast to others or another that had been expected
to be killed” (Stein 1995: 133). Based upon the exposition mentioned
above, it has to be concluded that the sentences with left-shifted ele-
ments have an affective meaning in addition to propositional mean-
ing. So a speaker or writer avails herself/himself of English inver-
sions to imply that “here is something happening that is unexpected
and will surprise you” (Stein 1995: 140).4
3. R-dislocations in English
All in all, we can support the view that the operation of left-shifting
constituents engenders the affective element of evaluation in a propo-
sitional meaning. In this Section we set out to analyse the association
of the right-shifting operation with information in a discourse and
additional meanings generated by that operation. There has been a
lot of research on R-dislocation constructions in English, including ,
Lambrecht (1981,1994), Ziv (1994), Ward and Birner (1996), Michae-
lis and Lambrecht (1996), Grosz and Ziv (1998), Gómez-González
(2001), and Gundel and Fretheim (2004). R-dislocation is illustrated
in example (18), not (19). (19) is so-called ‘afterthoughts’ by which
a speaker amends the referential mistake. It has to be noted that R-
dislocation should hold the coreferential relation between the pro-
noun and the right-shifted NP. On the contrary, the coreferential rela-
tion is not necessarily kept in afterthoughts (19).
4 Refer to Stein (1995) for more information as to the generation of an affective

meaning in English inversions.
(18) a. Hei is here, Jimi.

b. I don’t like themi at all, the copsi.
(19) I met him, your brother, I mean, two weeks ago. (Grosz and Ziv 1998: 296)
As exemplified in quotation (20), the NP in the immediately preced-

ing discourse cannot emerge as the R-dislocated NP. However, we
should specify the influence of the immediately preceding discourse
on the acceptability of R-dislocation later in this Section.
(20) A: Did you see Jack yesterday?

B: #Yes. He is going to Europe, Jack. (Ziv 1994: 641)
Now what we should intend to do is to specify the discourse function

of R-dislocation as well as an additional meaning generated by this
operation. In Lambrecht (1981), it is argued that R-dislocation ren-
ders a discourse what it has to be like. In other words, it is supposed
to reintroduce into a discourse the element regarded as ‘Textually
and Situationally Evoked Entity’ in the Familiarity Scale of Prince
(1981b)5. A more detailed exposition is provided in Grosz/Ziv (1998)
about the discourse function of R-dislocations.
In the extant analyses, it is assumed that the first function of
R-dislocation is to represent the element which is not referred to but
still exists in a discourse. The pronoun ‘he’ in (21) is categorized into
deictic expression due to the absence in the previous utterance. So it
relies heavily on context for its interpretation, but the hearer can eas-
ily recover the referent of this pronoun because she/he is one of the
interlocutors in that discourse. Then this usage of R-dislocation is a
counterexample against the exposition as to (20) in which R-disloca-
tion is judged to be infelicitous due to the fact that the R-dislocated
element is referred to in the immediately preceding discourse. So we
have to assess the first function of R-dislocation again.
(21) He’s terribly confusing, this Chomsky. I can’t figure him out. (Grosz/Ziv 1998:
300)
5 Refer to Ito (2005) for more information as to the Familiarity Scale of Prince
(1981b).
78 Norifumi Ito
Moreover, it has to be remembered that pronouns need not be used

referentially, as argued in Jaszczolt (2002). In (22), the pronoun ‘he’
can be used attributively; ‘the speaker may refer to whoever happened
to paint the picture he/she is currently looking at’ (Jaszczolt 2002: 140).
(22) He must be a famous painter.
Now let us move to the second function; in (23), the R-dislocated NP

is already mentioned in the previous utterance by A. However, there
are five clauses between the antecedent NP and the R-dislocated NP
with the result that its cognitive status decreases to the less familiar
status. So it has been considered so far that the second function is to
reintroduce the referent of the R-dislocated element and render it the
topic for the following context. My first language is Japanese, not
English, so I am not sure whether I can capture the intuition of En-
glish-speaking interlocutors or not. But I am rather skeptical about
the speculation that the referent of the pronoun becomes opaque in a
short dialogue like (23).
(23) A: I asked you to read this article for today.

B: I know. I tried to very hard, but I was quite busy. I had guests from abroad
who I had to entertain and I had nobody to help me. Besides, it is much too
difficult for me, this article. (Grosz/Ziv 1998: 302)
The third function is exemplified in (24) in which the referent of the

pronoun and R-dislocated NP is not directly coreferential but the
referent of the pronoun is the inferable form the NP referred to in the
previous context. The hearer is forced to interpret the referent of this
pronoun with the help of the linkage between the film with its main
character. Consequently, the NP ‘Charlie Chaplin’ is recognized as a
prime candidate for a topic in the following context.
(24) I saw Modern Times again yesterday. He is amazing, (this) Charlie Chaplin.
(Grosz/Ziv 1998: 302)
And lastly, we should make clear the fourth function of R-disloca-

tion in English. In the exposition about (20), it was argued that the
R-dislocated constituent cannot refer to the antecedent in the im-
mediately preceding utterance. However, we judge utterance (25)

to be acceptable in spite of the fact that the referent of the R-dislo-
cated element is referred to in the immediately preceding utterance.
It is argued in Grosz/Ziv (1998) that “the clause-final NP in the RD
does not merely refer to an entity; instead, it is used to predicate
some property or attribute of the referent” (Grosz/Ziv 1998: 302).
Moreover, this NP is supposed to express emotive content as an ad-
ditional meaning.
(25) I took my dog to the vet yesterday. He is getting unaffordable, the mangy old
beast. (Grosz/Ziv 1998: 302)
Here we have to assess the conclusion in Grosz and Ziv (1998) that
the R-dislocated NP is used attributively but not merely referentially.
As stated in Jaszczolt (2002), a definite description can generate two
interpretations, that is referential or attributive. For instance, sen-
tence (26) “can mean either that there was a particular person, say
Antonio Gaudi, who was insane, or that whoever designed the church
must have been mad. The first reading is referential, the other at-
tributive. Intuitively, the referential reading seems to be more natu-
ral or even unmarked” (Jaszczolt 2002: 133). According to the clari-
fication of the distinction of referential and attributive interpretations
in Jaszczolt (2002) and the analysis in Grosz and Ziv (1998), we
cannot help concluding that the hearer of utterance (25) may not
recover the exact referent of the R-dislocated NP if this NP is used
attributively. As opposed to this speculation, it is generally assumed
that we are easily capable of recovering the referent of this NP ‘this
mangy old beast.’
(26) The architect of this church was mad. (Jaszczolt 2002: 133)
All in all, we can attain a working hypothesis as to the function per-

formed by R-dislocation constructions in English. In Section 2, it
was confirmed that there is an association between left-shifting ma-
terial with the generation of affective meaning in utterances. In other
words, “what comes to mind first is expressed first” (Stein 1995:
133); we are capable of breaking the hearer’s expectation and sur-
prise her/him with the expression involving left-shifting material.
80 Norifumi Ito
However, the category of R-dislocations never comprises expres-

sions capable of being used to break the hearer’s expectation and
surprising her/him due to the involvement of the rightward move-
ment of material in utterance. Indeed, all utterances including R-
dislocated NPs in this Section contain affective elements within them:
‘terribly’ in (21), ‘much too’ in (23), ‘amazing’ in (24), and ‘unafford-
able’ in (25). However, the speaker has to use the expression with the
operation of the left-shifting element like (16a) in order to break the
hearer’s expectation and surprise the hearer. So it can be generally
assumed that a strong emotional feeling of markedness is not ca-
pable of arising within the main clause of English R-dislocations,
that is the clause without the R-disclocated NP. Moreover, it has to
be borne in mind that Grosz and Ziv (1998) analyzes 112 R-dis-
locations in which more than 80% of the pronouns coreferential with
R-dislocated NPs appear as subject of the clauses. As argued in Sec-
tion 1, it was demonstrated that the most accessible elements tend to
appear at subject position and become a topic for the following con-
text. In addition, Gundel et al. (1993) shows that pronouns can be
categorized as ‘in focus’ element, that is the element whose referent
is easily recovered by the addressee. Now we reach a speculation as
to the function of R-dislocations in English as follows. Sentences
containing pronouns have variable truth conditions, so a hearer re-
covers the referent of pronouns with the help of all sources of back-
ground information that are available to them during utterance inter-
pretation. In the main clauses of R-dislocations, that is the clause
without the R-dislocated NP, a speaker conveys an emotional feeling
towards the referent of the pronoun; the emotion a speaker intends to
convey is not so strong as to break a hearer’s expectation and sur-
prise a hearer. Lastly, a speaker uses a definite description as R-dis-
located element in order to confirm the referent of the pronoun which
a hearer has recovered with the help of all sources of background
information. So just when the R-dislocated element emerges at the
sentence end, the referential interpretation of a definite expression is
available.
4. R-dislocations in Japanese
In Section 3, we attempted to provide an exposition as to the function

of English R-dislocations. For the comparison with another linguistic
system, we try to explore R-dislocations in Japanese (so-called
gyakuten-bun). Among researches as to R-dislocations in Japanese,
Fujii (1995) provides one of the most detailed analyses of the con-
struction. Fujii (1995) examines 167 Japanese R-dislocations6 whose
usage motivation and function can vary according to the difference of
the mental state of interlocutors. However, she argued that we are
capable of accounting for them in terms of the general motivation of
pragmatic markedness. According to Payne (1989), pragmatic marked-
ness can be defined as follows: ‘contrast,’ ‘counterexpectation,’ ‘para-
phrase,’ ‘addition of detail,’ ‘question and answer for information,’
‘threat,’ ‘statement against cultural and situational expectation,’ ‘ne-
gation,’ ‘highlight of propositional content,’ and so on. Pragmatically
marked information should be considered to be vital to the success of
conveying what a speaker intends to convey. So it has been argued that
vital information is expressed in the main clause, but the less salient
information is not currently activated in a speaker’s consciousness. In
consequence, the referent is expressed by the R-dislocated element.
Now we should explore example (27) mentioned in Fujii (1995).
(27) 1. kore wa ne,
this TOP FP
2. tachimawari de
sword fight at
3. zubatto kou
completely this way
4. ki rare ta n desu yo,
cut PASS PAST COM COP FP
5. kao o
face ACC
TOP = topic, FP = final particle, PASS = passive, COM = sentential com-
plementiser, COP = copula, ACC = accusative (Fujii 1995: 184)
6 Fujii (1995) investigates 230 examples of so-called Japanese gyakuten-bun of

which 167 are the equivalent of English R-dislocation, 36 are ‘afterthoughts,’
and 27 are R-dislocated adverbs and so on.
82 Norifumi Ito
In (27) one sentence is divided into five segments for the conve-
nience of analysis; this is an interview in a TV programme in Japan.
In the immediately preceding utterance, the speaker mentioned that
he had been struck and kicked. So it is argued in Fujii (1995) that the
proposition ‘X of the body was cut’ is given prominence in this dia-
logue with the result that the information concerning which part of
the body, that is the information in segment 5, is R-dislocated to the
sentence end. However, it has to be remembered that the deictic ex-
pression ‘kore’ [this] is used in segment 1. It refers to the sword cut
on the speaker’s face; it is the most accessible in this context. The
hearer is easily and exactly capable of recovering its referent from
the information in the context. We can assume that the segments
from 1 to 4 provide the descriptive proposition about the sword cut
on the face, and as the final process of completing a sentence the
speaker shows the reference of the deictic expression, that is ‘the
sword cut on this face not on other faces.’
Next let us explain (28) which includes the pragmatic
markedness of ‘contrast’ and ‘negation.’
(28) 1. kodomo mo
children also
2. kika nai n desu yo ne,
listen NEG COM COP FP FP
3. naze ka
why Q
4. souiukoto.
that kind of thing (Fujii 1995: 184)
In the dialogue preceding this sentence, the speaker said ‘Recently

parents attempt never to tell their children a chat over the old days.’
So Fujii (1995) suggests that there exist two contrasts in this utter-
ance, that is the contrasts between ‘parents and children’ and be-
tween ‘tell and listen.’ These two contrasts are given prominence in
the acts of communication of this utterance: the first thing the speaker
intends to convey in this utterance is the existence of these contrasts.
So the expression ‘souiukoto’ [that kind of thing = a chat over the old
days] is R-dislocated to the final position of the sentence. Pace this
suggestion by Fujii (1995), it has to be borne in mind that Japanese
has a general tendency to use zero pronoun when the referent of some-
thing is activated the most in the context. So we restate our earlier
conclusion that the hearer is easily and exactly capable of recovering
the referent of zero pronoun from the information of the context, and
that the segments from 1 to 2 provide the descriptive proposition
about a chat over the old days. And the speaker shows the reference
of zero pronoun, that is ‘a chat over the old days not over anything
else’ as the final process of completing a sentence
In order for our speculation to be confirmed, we will analyze
one more dialogue as follows.
(29) 1. sugoi seisinryoku no tuyoi

very soul Of powerful
2. hito nan desu ne,
person COM COP FP
3. ano hito wa
that person TOP (Fujii 1995: 185)
Zero pronoun is used in this dialogue, too. The hearer recovers the
referent of the zero pronoun from all sources of information in the
context. Segments 1 and 2 provide the description about the referent
the hearer recovered from the context, and after that description the
speaker provides the reference of the zero pronoun.
We have examined only three examples of R-dislocation in
Japanese: the affinity between them is that all of them include affec-
tive elements with themselves, for instance ‘zubatto’ [= completely]
in (27), ‘nai’ [= negation marker] in (28), and ‘sugoi’ [very] in (29).
Just as English equivalents do, Japanese R-dislocations convey an
affective meaning which is not so strong as to break a hearer’s ex-
pectation and surprise a hearer. Moreover, R-dislocated elements
exhibit the characteristic of showing the referent of the most acti-
vated elements in the main clause.
84 Norifumi Ito
5. Conclusion
At a cursory glance, it may seem that interlocutors are not capable of

dealing with information from all sources in the context for the pro-
duction and understanding of linguistics expressions. On closer in-
spection, however, it is evident that speakers avail themselves of vari-
ous linguistic means filled with discourse information according to
the subtle difference of their intention to communicate with others.
In order to respond what is physically written and uttered by speak-
ers, hearers rely on the combination of word meaning and sentence
structure at first. In case of facing utterances with much more infor-
mation conveyed than that conveyed by the combination of word
meaning and sentence structure, hearers have to rely on all sources
of information in the context. So the aim of this investigation was to
explore the role played by information for the interpretation of utter-
ances.
Section 1 has dealt with the notion of information structure in
propositions. An affinity between various definitions as to informa-
tion structure is that it represents the mental state of interlocutors
who use and interpret lexicogrammatical structures to convey their
intention. In Section 2, English inversions have been explored; it is
evident that inversions adhere to the pragmatic constraint that the
more accessible elements should precede the less accessible elements
within the same sentence. So the more accessible elements appear as
preposed constituents while the less accessible elements appear as
postposed constituents. Moreover, inversions give rise to an addi-
tional meaning, that is an affective meaning. They are used to con-
vey the addresser’s message that “here is something happening that
is unexpected and will surprise you” (Stein 1995: 140). Section 3
provides the exposition as to R-dislocations in English. It has been
generally assumed in the existing research that R-dislocations are
used to render prominent the referent of something which has al-
ready exists in a discourse but not so vital to the progress of the
discourse. However, we propose that the main clause of English R-
dislocation gives a description of the referent of an element with an
affective meaning which is not so strong as to break a hearer’s ex-
pectation and surprise a hearer. On the contrary, that referent is re-

ferred to referentially at the R-dislocated position. In a similar vein,
Section 4 gives an explanation as to the function of Japanese R-dis-
locations. Just as English equivalents do, Japanese R-dislocations
play a role as follows: Japanese R-dislocation gives a description of
the referent of an element with an affective meaning which is not so
strong as to break a hearer’s expectation and surprise a hearer, and
the R-dislocated referring expressions are used referentially.
References
Birner, Betty 1994. Information Status and Word Order: An Analysis

of English Inversion. Language 70, 233-259.
Birner, Betty 1996. The Discourse Function of Inversions in English.
New York: Garland.
Birner, Betty 1998. Recency Effects in English Inversion. In Walker,
M. /Joshi, A. / Prince, E. (eds) Centering Theory in Discourse.
Oxford: Clarendon Press, 309-323.
Birner, Betty / Ward, Gregory 1998. Information Status and Non-
canonical Word Order in English. Philadelphia: John Benja-
mins.
Bresnan, J. 1994. Locative Inversion and the Architecture of Univer-
sal Grammar. Language 70, 72-131.
Chafe, Wallece 1976. Givenness, Contrastiveness, Definiteness, Sub-
jects, Topics, and Point of View. In Li, Charles N. (ed.) Subject
and Topic. New York: Academic Press, 25-55.
Chafe, Wallece 1994. Discourse, Consciousness, and Time: The Flow
and Displacement of Conscious Experience in Speaking and
Writing. Chicago: The University of Chicago Press.
Dahl, Östen 1976. What is New Information. Approaches to Word
Order: Reports in Text Linguistics 72.
Fujii, Y. 1995. Nihongo no gojun no gyakuten ni tuite [On the reorder-
ing of Japanese word order], Nichieigo no uhou idou koubun –
sono kouzou to kinou. Takami, K. (ed.). Tokyo: Hitsuji, 167-198.
86 Norifumi Ito
Gómez-González, María Ángeles 2001. The Theme-Topic Interface:

Evidence from English. Amsterdam: John Benjamins.
Green, Georgia M. 1980. Some wherefores of English inversions.
Language 56, 582-601.
Green, Georgia M. 1982. Colloquial and Literary Uses of Inversions.
In Tannen, Deborah (ed.) Spoken and Written Language: Ex-
ploring Orality and Literacy. Norwood, NJ: Ablex, 119-153.
Green, Georgia M. 1985. The description of Inversions in General-
ized Phrase Structure Grammar. BLS 11, 117-146.
Gregory, Michelle L. / Michaelis, Laura A. 2001. Topicalization and
Left-Dislocation: A Functional Opposition Revisited. Journal
of Pragmatics 33, 1665-1706.
Grosz, Barbara. J. / Yael Ziv 1998. Centering, Global Focus, and Right-
Dislocation. In Walker, M. / Joshi, A. / Prince, E. (eds) Center-
ing Theory in Discourse. Oxford: Clarendon Press, 293-307.
Gundel, Jeanette K. et al. 1993. Cognitive Status and the Form of
Referring Expressions in Discourse. Language 69, 274-307.
Gundel, Jeanette K. 1998. Centering Theory and the Givenness Hi-
erarchy: Towards a Synthesis. In Walker, M. / Joshi, A. /
Prince, E. (eds) Centering Theory in Discourse. Oxford:
Clarendon Press, 183-198.
Gundel, Jeanette K. 1999. On Different Kinds of Focus. In Bosch,
P. / van der Sandt, R. (eds) Focus Linguistic, Cognitive, and
Computational Perspective. Cambridge: Cambridge Univer-
sity Press, 293-305.
Gundel, Jeanette K. / Fretheim, Thorstein 2004. Topic and Focus. In
Horn, L.R. / Ward, G. (eds) The Handbook of Pragmatics.
Oxford: Blackwell, 175-196.
Heidegger, M. 1953. Sein und Zeit. Tübingen: Max Niemeyer. Trans.
J. Stambaugh as Being and Time. 1996. Albany: State Univer-
sity of New York Press.
Ito, N. 1996. Information within English Inversions and Sentence
Pattern of English. DLS: Working Papers in Linguistics Spe-
cial Graduate Number 5, 11-24.
Ito, N. 1999. The Language Function of Non-Canonical Sentences
in English. Tokushima Bunri Daigaku bungaku ronsou No. 16,
Tokushima Bunri Daigaku, 21-48.
Ito, N. 2005. Danwa jouhou to eigo koubun kaishaku. Tokyo: Eihou-

sha.
Jaszczolt, K.M. 1999. Disourse, beliefs, and intentions: Semantic
defaults and propositional attitude ascription. Oxford: Elsevier
Science.
Jaszczolt, K.M. 2002. Semantics and pragmatics: Meaning in lan-
guage and discourse. London: Longman.
Jaszczolt, K.M. 2005. Default semantics: Foundations of a composi-
tional theory of acts of communication. Oxford: Oxford Uni-
versity Press.
Jaszczolt, K.M. 2006. Futurity in default semantics. In Von Heusinger,
K. / Turner, K. (eds) Where semantics meets pragmatics. Ox-
ford: Elsevier Science, 471-492.
Lambrecht, Knud 1981. Topic, Anti-Topic and Verb Agreement in Non-
Standard French. Amsterdam/Philadelphia: John Benjamins.
Lambrecht, Knud 1994. Information Structure and Sentence Form.
Cambridge: Cambridge University Press.
Lambrecht, Knud 1995. The Pragmatics of Case: On the Relationship
between Semantic, Grammatical, and Pragmatic Roles in Eng-
lish and French. In Shibatani, Masayoshi / Thompson, Sandra
(eds) Essays in Semantics and Pragmatics. Amsterdam: John
Benjamins, 144-190.
Michaelis, Laura A. / Lambrecht, Knud 1996. Toward a Construc-
tion-Based Theory of Language Function: The Case of Nomi-
nal Extraposition. Language 72, 215-247.
Payne, Doris L. 1989. Meaning and Pragmatics of Order in Selected
South American Indian Languages. MS, University of Oregon.
Prince, Ellen 1981a. Topicalization, Focus Movement, and Yiddish
Movement: A Pragmatic Differentiation. Berkeley Linguistic
Society 7, 249-264.
Prince, Ellen 1981b. Toward a Taxonomy of Given/New Informa-
tion. In Cole, Peter (ed.) Radical Pragmatics, 223-254. New
York: Academic Press.
Prince, Ellen 1984. Topicalization and Left Dislocation: A Functional
Analysis. In White, S.J. / Teller, V. (eds) Discourses in Reading
and Linguistics: Annals of the New York Academy of Sciences
433, 213-225.
88 Norifumi Ito
Prince, Ellen 1997. On the Functions of Left-Dislocation in English

Discourse. In Kamio, Akio (ed.) Directions in Functional Lin-
guistics. Amsterdam: John Benjamins, 117-143.
Schmidt, D.A. 1980. A History of Inversions in English. Ohio State
University. Unpublished PhD dissertation.
Seuren, P. 1986. Presupposition. In Von Stechow, A. / Wunderlich,
D. (eds) Handbook of Semantics. Frankfurt: Athenaum Verlag.
Stein, D. 1995. Subjective meanings and the history of inversions. In
Stein, D. / Wright, S. (eds) Subjectivity and subjectivisation:
Linguistic perspectives. Cambridge: Cambridge University
Press, 129-150.
Takami, Ken-ichi / Kamio, Akio 1996. Topicalization and Subjectivi-
zation in Japanese: Characterization and Identificational In-
formation. Lingua 99, 207-235.
Tanaka, Hidekazu 2001. Right-Dislocation as Scambling. Journal of
Linguistics 37, 551-579.
Vallduvi, E. 1990. The Informational Component. Ph.D. disserta-
tion: University of Pennsylvania.
Vallduvi, E. 1992. The Informational Component. New York: Gar-
land.
Vallduvi, E. / Engdahl, E. 1996. The Linguistic Realization of Infor-
mation Packaging. Linguistics 34, 459-519.
Walker, M. et al. 1994. Japanese Discourse and the Process of Cen-
tering. Computational Linguistics 20, 193-232.
Ward, G. / Birner, Betty 2004. Information Structure and Non-ca-
nonical Syntax. In Horn, L.R. / Ward, G. (eds) The Handbook
of Pragmatics. Oxford: Blackwell, 153-174.
Ziv, Yael 1994. Left and Right Dislocations: Discourse Functions
and Anaphora. Journal of Pragmatics 22, 629-645.
89
HIDEMATSU MIURA
Focus-driven Semantic Reflexivity in Japanese
1. Introduction
Reflexive constructions (RCs) have received much attention since

the beginning of modern theoretical linguistics. Since it has been a
central concern to look for universal syntactic conditions for reflex-
ive binding (e.g. Chomsky 1981), semantic aspects of RCs have been
less discussed in the literature. RCs typically denote events in which
only one entity is involved. It has been pointed out, however, that
RCs do not always denote such a reflexive event where only one
entity is involved. The following example (1), taken from Jackendoff
(1992), cannot make any sense at a first glance, but indeed (1) can be
uttered to denote a non-reflexive event in which Ringo falls on the
‘statue’ of himself in a wax museum, for example.
(1) Ringo fell on himself.
On the other hand, there are RCs that do not allow such non-reflex-
ive interpretation. Suppose there is a statue of Reagan in a wax mu-
seum. Then compare (2a) and (2b). (2a) never allows such a statue
reading whereas (2b) allows the statue reading as well as the normal
reflexive event reading (Lidz 2000).
(2) a. Reagan dressed in the museum

b. Reagan dressed himself in the museum
Lidz (2000, 2001) claims that there are two types of reflexives: pure-
reflexive and near-reflexive. The former requires the reflexive to be
referentially identical to the antecedent whereas the latter does not. He
claims that the difference comes from lexical reflexivity of the verbs.
90 Hidematsu Miura
That is, the verb of (2a) is lexically reflexive while that of (2b) is not.
He concludes that there is a bidirectional relation between lexical
reflexivity and pure-reflexive interpretation. He calls this principle
‘Condition R’. To my knowledge, Lidz’ theory has not been applied
to Japanese. The relevant Japanese data were examined, but it turned
out that some of the Japanese data do not follow Condition R.
The purpose of this chapter is two-fold. First, it will be dem-
onstrated that Japanese lexical RCs show the behaviors expected from
Condition R. This serves as further confirmation of the universality
of the principle. Second, it will be shown, however, that the behav-
iors of syntactic RCs do not follow from it. It will be argued, follow-
ing Liu (2003), that focus structure, as well as lexical reflexivity, can
also yield the (unambiguous) pure-reflexive interpretation. It will be
claimed that, accordingly, the current bidirectional relation of the
lexicon-semantic correlation in Condition R should be weakened to
unidirectional one.
In Section 2, the two types of reflexives and the principle, Con-
dition R, will be further detailed. In Section 3, the Japanese reflex-
ives will be examined in terms of the principle and it will be shown
that there are unexpected behaviors that do not follow from the prin-
ciple. In Section 4, a structure-based account will be given. In Sec-
tion 4, concluding remarks will be made.
2. Two types of reflexives and lexical reflexivity
2.1 Pure-reflexive, Near-reflexive and Condition R (Lidz 2000, 2001)
As briefly mentioned above, it has been observed that reflexives do

not always refer to a reflexive event. This was extensively discussed
in Jackendoff (1992). (1), repeated here as (3), denotes the situation
of Ringo Starr wandering in a wax museum and accidentally falling
on his own statue. The event denoted is not reflexive in the canonical
sense of the word since there are two distinct entities involved, i.e. a
real person and its statue.
Focus-driven Semantic Reflexivity in Japanese 91
(3) Ringo fell on himself.
Lidz (1999, 2001) took the Jackendoff’s insight seriously and claimed
that there are referentially two different types of reflexives: those
that require a complete identity with the antecedent and those that do
not. He terms the former ‘pure-reflexive’ and the latter ‘near-reflex-
ive’1. Even when the referent of the reflexive is different from that of
the antecedent, the entity that the near-reflexive refers to must have
some identifiable association with the antecedent (e.g. statue or por-
trait). The difference between pure- and near-reflexives can be rep-
resented as follows:
(4) a. ¼x [P(x,x)] (Pure-reflexive)

b. ¼x [P(x, f(x))] (Near-reflexive)
Lidz further points out that the pure-reflexive reading is correlated

with lexical reflexivity of predicates. The notion of lexical reflexiv-
ity was introduced in Reinhart and Reuland (1993; R&R hereafter)
to account for the antilocality phenomena observed in some languages
in which reflexives cannot be locally bound. It is argued that, in Dutch,
the verb in (5a), haat [hate], is not lexically reflexive so that the
simple reflexive does not suffice and the complex form must be em-
ployed as in (5b). On the other hand, the verb in (6) is inherently
lexically reflexive so that the simple reflexive is enough.
(5) a. *Max haat zich

Max hates self
‘Max hates himself’
b. Max haat zichzelf
Max hates selfself
‘Max hates himself’
(6) Max wast zich

Max washes self
‘Max washes himself’
1 Safir (2004) calls this type of meaning ‘proxy reading’.

92 Hidematsu Miura
Thus, incorporating both Jackendoff ’s and R & R’s insights, Lidz pro-
poses his own theory of reflexivity in which semantic reflexivity and
lexical reflexivity have bidirectional relation to each other. The fol-
lowing is the principle he calls Condition R:
(7) Condition R: ¼x[P(x, x)] ÅÆ (¼1 =¼2)

semantics theta-grid
The condition on the left is the semantic representation and the one
on the right is the lexical specification. This states that if a predicate
is semantically reflexive (i.e. pure-reflexive), it is lexically reflexive
and vice versa. In other words, what Condition R states is that true
(pure) reflexivity is guaranteed through lexical reflexivity. As a con-
sequence, when this condition is not met, near-reflexive readings
such as statue reading are expected to be available.
Condition R is a semantic condition which does not refer to
any morphosyntactic realization of lexical reflexivity. There are two
realization patterns attested: a morphologically covert type and an
overt type. The Dutch example (6) is an example in which lexical
reflexivity is sublexically expressed (i.e. morphologically covert).
There are languages that have morphologically overt-marking on the
predicate (see Kannada examples below for this morphologically
overt type).
There are two predictions Condition R can make. First, if the
predicate is lexically reflexive, whether it is realized sublexically or
morphologically, only the pure-reflexive interpretation should be avail-
able. This expectation is borne out as in (2), repeated below as (8).
(8) a. Reagan dressed in the museum (Reagan / *statue)

b. Reagan dressed himself in the museum (Reagan / statue)
The assumption here is that the lexical entries of the verbs used in
(8a) and (8b) are lexically reflexive and non-reflexive respectively
(the same logic for the Dutch examples 5 and 6 above). As expected,
the sentence with a lexically reflexive predicate, (8a), does not allow
any near-reflexive (i.e. statue) reading while the one with a lexically
non-reflexive predicate, (8b), allows such reading.
Second, if the predicate is lexically reflexive, only the sloppy

reading should be available in the comparative ellipsis construction2
since the second argument in the semantic representation of Condi-
tion R is a bound variable. Otherwise, both sloppy and strict readings
should be available. This expectation is also borne out as illustrated
in (9a) and (9b).
(9) a. Reagan dresses faster than his nurse does

‘Reagan dresses himself faster than his nurse dresses herself ’ sloppy)
‘*Reagan dresses himself faster than his nurse dresses him’ (strict)
b. Reagan dresses himself faster than his nurse does
‘Reagan dresses himself faster than his nurse dresses herself ’ sloppy)
‘Reagan dresses himself faster than his nurse dresses him’ (strict)
In the case of the languages that do not have overt morphological

marking, it is difficult to see if the predicate is lexically reflexive or
not, but it is possible to use (at least) these two as diagnostic tests for
the lexical reflexivity to some extent3.
2 Lidz (1999, 2001) calls the construction comparative ‘deletion’, but I will call
it comparative ‘ellipsis’ construction following the distinction made in Hoji
(1998). Both ‘deletion’ and ‘ellipsis’ will be used as diagnostic tests below.
The distinction will be further made clear below.
3 Readers should be aware of the circularity in the arguments. Although I do not
pursue this problem any further here, a comment is in order. In order for Con-
dition R to work, we need to know somehow if a predicate is lexically reflexive
or not independently of the two phenomena (the availability of the statue read-
ing and the sloppy reading). When a language has a reflexive morpheme at-
tached on a predicate stem (e.g. Kannada; see below), it is obvious that the
predicate is (lexically) reflexive. But in the case of languages without such a
morpheme (e.g. Dutch), it is hard to see. R&R (1993) suggests a ‘nominalization
test’ for the lexical reflexivity of the Dutch predicates, but the efficacy of the
test is far from convincing. If we use the availability of the statue reading and
the sloppiness as the diagnostic tests for lexical reflexivity, we are led to wrong
conclusions since, as will be discussed later, Chinese and Japanese produce the
unambiguous pure-reflexive interpretation without a lexically reflexive predi-
cate.
94 Hidematsu Miura
2.2 Cross-linguistic variation in lexical reflexivity
Cross-linguistically, three patterns are expected to exist for lexical

reflexivity: (1) all the verbs are (potentially) lexically reflexive (2)
only a subset of the verbs is lexically reflexive (3) there is no lexical
reflexivity. According to Lidz (1999, 2001), Kannada, Dutch and
Malayalam represent each of these. After looking at the three lan-
guages, the lexical reflexivity of Japanese will be introduced (Japa-
nese is a mix of 2 and 3).
Kannada is a language that has a morphologically overt re-
flexive-marking on the predicates (Lidz 2000, 2001). When the predi-
cate is morphologically reflexive-marked, it is lexically reflexive and
therefore only the pure-reflexive interpretation is expected. This ex-
pectation is borne out as shown in (10a).
(10) a. Hari tann-annu nooDi-du-koND-a

Hari self-ACC see-PP-REFL.PST-3SM
‘Hari saw himself ’ (= reflection, *statue)
(10b) and (10c) show that, without this reflexive affix, the predicate
is non-reflexive and the simple reflexive pronoun does not suffice to
form an acceptable sentence. The predicate in (10c), which is same
as (10b), is not lexically reflexive so that the near-reflexive interpre-
tation is available.
b. *Hari tann-annu nooD-id-a

Hari self-ACC see-PST-3SM
‘Hari saw himself ’
c. Hari tann-annu-taane nooD-id-a
Hari self-ACC-self see-PST-3SM
‘Hari saw himself ’ (= reflection or statue)
According to Lidz, Kannada can lexically reflexivize any predicates

by attaching the reflexive morpheme. Thus Kannada represents the
type of the languages in which the predicates are (potentially) all
lexically reflexive.
As we already saw, lexical reflexivity is morphologically co-
vert in Dutch and it can be inferred only through semantic interpreta-
tion and syntactic behaviors. It has been assumed in the literature

(e.g. R&R 1993, Lidz 2001) that some verbs, especially introverted
ones such as shave or wash, have two lexical entries one of which is
lexically reflexive4. The acceptability of (11a) in which the simplex
reflexive is used suggests that the predicate be lexically reflexive. As
expected, the statue reading is not available. On the other hand, the
complex reflexive form employed in (11b) indicates that the under-
lying lexical entry of the predicate is lexically non-reflexive. The
near-reflexive interpretation (i.e. statue reading) is possible.
(11) a. Ringo scheert zich

Ringo shaves self
‘Ringo shaves’ (*Near-reflexive)
b. Ringo scheert zich-zelf
Ringo shaves self-self
‘Ringo shaves himself ’ (ok Near-reflexive)
In the comparative ellipsis construction, the same expectation is borne

out. When the simple reflexive is employed in the construction, only
the sloppy reading is available. If the complex form is used, both
sloppy and strict readings are possible since the lexical entry of the
verb is not lexically reflexive. Compare (12a) and (12b).
(12) a. Zij verdedigde zich beter dan Peter

She defended self better than Peter
‘She defended herself better than Peter defended himself ’ (sloppy)
‘*She defended herself better than Peter defended her’ (*strict)
b. Zij verdedigde zichzelf beter dan Peter
She defended self-self better than Peter
‘She defended herself better than Peter defended himself ’ (sloppy)
‘She defended herself better than Peter defended her’ (strict)
4 Haiman (1983) defines the actions which one generally performs upon one’s
self as ‘introverted’ and the actions which the subject usually performs toward
others as ‘extroverted.’ The following pair represents each type respectively.
Although both of the verbs are transitive, the possibility of the omission of the
reflexive pronoun is an indicator of the distinction.
i. Max washed (himself).
ii. Max kicked himself.
96 Hidematsu Miura
In Dutch, only a subset of all the verbs, especially introverted verbs

such as shave, has two lexical entries, one of which is lexical reflex-
ives. Thus, Dutch represents the type of the languages in which only
a subset of the entire predicates are lexically reflexive.
Malayalam is claimed to have no lexical reflexivity. In Malaya-
lam, the predicates which are lexically reflexive in other languages
are not lexically reflexive as in (13). For example, introverted verbs
such as shave are lexically reflexive in many languages and it is ex-
pected that the simple reflexive suffices for such a lexically reflexive
verb. This expectation is not borne out in Malayalam as in (13a) and
the complex form needs to be employed as in (13b). Furthermore, if
the verb is not lexically reflexive, the near-reflexive interpretation
should be available. This expectation is borne out for (13b).
(13) a. *Raaman tan-ne kshauram ceytu

Raaman self-ACC shaving did
‘Raaman shaved’
b. Raaman tan-ne-tanne kshauram ceytu
Raaman self-ACC-self shaving did
‘Raaman shaved himself ’
Although the data are somewhat limited, Lidz (2000) has concluded that
Malayalam is the language that does not have any lexical reflexivity.
The following basic assumptions of this chapter were laid out
in this section: the distinction between pure- and near-reflexives,
Condition R, the notion of lexical reflexivity and the cross-linguistic
variation of lexical reflexivity.
3. Reflexives in Japanese
Let us turn to our primary concern, RCs in Japanese. Both lexical

and syntactic reflexives in Japanese will now be discussed in that
order. In 3.1, it will be shown that Japanese lexical RCs follow the
principle of Condition R. This will serve as further confirmation of
the universality of the principle. In 3.2, before examining syntactic

reflexives, it will be argued that the Japanese verbs used in syntactic
RCs are not lexically reflexive. In 3.3, it will be demonstrated that
Japanese syntactic RCs produce unambiguous pure-reflexive inter-
pretations despite the non-reflexive nature of the predicate.
3.1 Reflexivity of lexical reflexives
There are lexically reflexive predicates in Japanese; however, they

have been less discussed than syntactic RCs. Tsujimura/Aikawa (1999)
is one of the few previous studies. Lexically reflexive predicates are
morphologically reflexive-marked like Kannada. (14) is an example.
(14) Hanako wa tachiba no juuyoosei o zi-kaku siteiru

TOP position GEN importance ACC self-conscious doing
‘Hanako is aware of the importance of her position’
The process is not productive at all unlike Kannada (somewhat idio-

syncratic) and they are a small subset of the entire predicates like
Dutch. Thus Japanese lexical reflexives are a mixed category in terms
of the classifications given in the previous section.
Although the morphological marking of reflexivity is not pro-
ductive, there are a sizable number of such predicates5. In my cor-
pus, there are two major groups: zi-verbs and ziko-verbs. The initial
morphemes, zi- and ziko-, both mean ‘self’. (15a) and (15b) are ex-
amples of each form.
(15) a. Taroo ga zi-shuu sita

NOM self-learning did
‘Taroo studied by himself ’
b. Taroo ga ziko-hasan sita
NOM self-bankruptcy did
‘Taroo became bankrupt’
5 At my last count, there are 29 zi-verbs and 64 ziko-verbs in my corpus. In

addition to these two major groups, there are idiosyncratic, in terms of the
number of tokens, self-forms such as doku- (doku-gaku ‘self-study’), but these
mean ‘alone’ or ‘by oneself ’.
98 Hidematsu Miura
Zi-verbs are formed by attaching the bound morpheme zi- ‘self’ to

another bound morpheme stem (compare 16a and 16b). As noted in
Tsujimura and Aikawa (1999), zi-verbs can be further grouped into
two types: unaccusative type and inalienable type6. There is a clear
syntactic difference between them. The unacusative type does not
have any linguistic element other than subject and verb. They are
clearly intransitive. This is shown in (16c).
(16) a. Taroo ga zi-ritsu sita

Nom self-stand did
‘Taro established himself.’
b. *Taroo ga ritsu sita
c. *Taroo ga sigoto/kazoku o zi-ritsu sita
Nom career/family Acc self-stand did
‘Taro self-established his career/family.’
On the other hand, the inalienable type does take an object-like ele-
ment as in (17a). The element is semantically bound by the reflexive
morpheme whose original antecedent is the subject. An entity that
has no semantic/associative relation to the binder (subject) cannot
appear. As in (17b), for example, the crime must be the one commit-
ted by the subject, Taroo, or the one Taroo is somehow involved in.
(17) a. Taroo ga hankoo o zi-kyoo sita

NOM crime ACC self-offer did
‘Taroo confessed his crime’
b. *Taroo ga Ken no hankoo o zi-kyoo sita
NOM GEN crime ACC self-offer did
‘Taroo confessed Ken’s crime’
The inalienable type seems to have a transitive structure, but it is not

correct. The unacceptability of (17c) shows that the accusative-marked
element is not a canonical object.
6 The distinction is not relevant to the current discussion, so it will not be de-
tailed here. See Tsujimura/Aikawa (1999) for their observations for the dis-
tinction.
c. *Hankoo ga Taroo ni (yotte) zi-kyoo s-are-ta

crime NOM DAT(by) (by) self-offer do-CAUS-PST
‘The crime was confessed by Taroo’
The intransitive nature of the inalienable type can be further diag-

nosed as follows. The so-called adversative passives in Japanese are
typically formed from an intransitive predicate such as die. If the
predicate is intransitive as a whole, it should be able to form an ad-
versative passive. This expectation is borne out as in (17d). Thus, the
acceptability of the adversative passive in (17d) further shows the
intransitive nature of the entire predicate.
d. Hanako ga Taroo ni hankoo o zi-kyoo s-are-ta
NOM DAT(by) crime ACC self-offer do-CAUS-PST
‘Hanako was affected by Taroo’s confessing his crime’
There is another set of lexically reflexive predicates in which an-

other ‘self’ form, ziko- is attached to form a complex nominal. The
difference from the previous group, zi-verbs, is that, in order to form
the complex reflexive nominal, the morpheme ziko- is attached to a
free stem. Compare (18a) and (18b). The construction can take an
object-like NP in some cases, but it has to have a semantic/associa-
tive relation with the original binder (subject) like the inalienable zi-
verbs above. Compare (18c) and (18d).
(18) a. Taroo ga ziko-bengo sita
NOM self-defense did
‘Taroo self-defended himself ’
b. Taroo ga zibun(-zisin) o bengo sita
NOM self(-self) ACC defense did
‘Taroo defended himself ’
c. Taroo ga sooryoo o ziko-hutan sita
NOM carriage ACC self-charge did
‘Taroo paid the carriage on his own’
d. *Taroo ga Hanako o ziko-bengo sita
NOM ACC self-defense did
‘Taroo self-defended Hanako’
There are many properties common among the different zi(ko)-verbs.

Lexical reflexives are always locally bound and long-distance bind-
100 Hidematsu Miura
ing is not possible as in (19a) and (20a). The examples in (19b) and
(20b) further demonstrate that this locality is seen irrespective of the
juncture type. (19a) and (20a) display a clausal juncture while (19b)
and (20b) show a core juncture7.
(19) a. Hanakoi wa [Tarooj ga zi-ritsu*i/j sita ] to omotteita

TOP Nom self-stand did COMP thought
‘Hanako thought that Taro established himself/*her.’
b. Hanakoi wa Tarooj ni zi-ristu*i/j suru yoo susumeta
TOP DAT self-stand do to recommended
‘Hanako recommended Taroo to establish himself/*her’
(20) a. Hanakoi wa [ Tarooj ga ziko-bengo*i/j sita ] to omotteita

TOP NOM self.defensedid COMP thought
‘Hanako thought that Taroo self-defended himself/*her’
b. Hanakoi wa Tarooj ni ziko-bengo*i/j suru yoo susumeta
TOP DAT self.defensedid to recommended
‘Hanako recommended Taroo to self-defend himself/*her’
Another characteristic is that they do not take a reflexive element as

in (21a) and (21b) although ziko-verbs seem to marginally allow a
reflexive as in (21c). This is a notable difference from the lexical
reflexives in Kannada or Dutch8.
(21) a. *Taroo ga zibun(-zisin) o zi-ritsu sita

NOM self(-self) ACC self-establishment did
‘Taroo self-established himself ’
b. *Taroo ga zibun(-zisin) o zi-kyoo sita
NOM self(-self) ACC self-offer did
‘Taroo self-confessed himself ’
7 The distinction between clause and core is based on Role and Reference Gram-
mar (RRG; Van Valin and LaPolla 1997, Van Valin 2005). A primary indicator
for the distinction between clause and core is whether the juncture is tensed or
not. In (19a) and (20a), past tense is used so that the juncture is clearly clausal
whereas in (19b) and (20b) past tense cannot be used indicating the juncture is
not clausal (i.e. core).
8 I do not pursue this issue here, but this difference may be indicating a funda-
mental difference in self-forms between the two languages (Dutch and Kannada)
and Japanese.
c. ?Taroo ga zibun(-zisin) o ziko-bengo sita

NOM self(-self) ACC self-defense did
‘Taroo self-defended himself ’
Now let us examine the behaviors of the Japanese lexical reflexives

in terms of Condition R. A predicate that denotes physical activity is
generally used to test whether a near-reflexive reading such as statue
reading is available or not, but unfortunately it seems there are no
such predicates in zi(ko)-verbs. All the zi(ko)-verbs denote more or
less abstract events. We will examine the availability of the sloppy
and strict readings in elliptical constructions. In addition to the com-
parative ellipsis construction used in the previous literature, com-
parative deletion construction and null object construction (NOC)
will be employed as additional diagnostic tests.
There are two types of comparative constructions in Japanese
(Hoji 1998): comparative ‘ellipsis’ and comparative ‘deletion’ con-
struction. In the former there appears only one argument (subject) in
the elided subordinate juncture while the latter has the same verb as
well as the subject also in the elided juncture. (22) and (23) are ex-
amples of comparative ellipsis and comparative deletion respectively.
(22) Taroo ga Hanako yori hayaku zibun no tomodachi o suisensita

NOM than soon self GEN friend ACC recommended
a. ‘Taroo recommended his friend sooner than Hanako recommended
her friend’
b. ‘*?Taroo recommended his friend sooner than Hanako recommended
his friend’
(23) Taroo ga Hanako ga suisensuru yori hayaku zibun no tomodachi o

NOM NOM recoomend than soon self GEN friend ACC
suisen sita
recommended
a. ‘*?Taroo recommended his friend sooner than Hanako recommended
her friend’
b. ‘Taroo recommended his friend sooner than Hanako recommended
his friend’
102 Hidematsu Miura
It is observed in Hoji (1998) that the sloppy reading is more readily

available in the comparative ellipsis construction while the strict read-
ing is preferred over the sloppy reading in the comparative deletion
construction. In the examples above, (22a) and (23a) are sloppy read-
ing while (22b) and (23b) are strict reading.
First, comparative ellipsis will be examined. As observed in
Section 2, it has been cross-linguistically attested that lexical reflex-
ives produce only the sloppy reading in comparative constructions9.
It must be remembered that generally the Japanese comparative el-
lipsis allows the sloppy reading more readily than the strict one; there-
fore, with these two (language-specific and cross-linguistic) factors
combined, only the sloppy reading should be available. This expec-
tation is borne out.
(24) Taroo ga Hanako yori hayaku hankoo o zi-kyoo sita

NOM than soon crime ACC self-offer did
‘Taroo confessed his crime sooner than Hanako’
a. ‘Taroo confessed his own crime sooner than Hanako confessed her own
crime.’
b. ‘*Taroo confessed his own crime sooner than Hanako confessed his crime.’
(25) Taroo ga Hanako yori ooku sooryoo o ziko-hutan sita

NOM than more carriage ACC self-charge did
a. ‘Taroo paid more for his carriage than Hanako paid for hers’
b. ‘*Taroo paid more for his carriage than Hanako paid for Taroo’s’
For (24) suppose that Taroo and Hanako committed different crimes
individually and they know about each other’s crimes. That is, (24)
means that Taroo was less tenacious than Hanako in the police sta-
tion. The crime Taroo confessed must be his and the one Hanako
confessed must be hers. There is no interpretive possibility that
Hanako confessed Taroo’s crime. Thus, the strict interpretation, (24b),
is unavailable. For (25) imagine the situation like Taroo and Hanako,
9 The comparative construction used in Lidz’ papers is the comparative ‘ellipsis’

construction in our term although he calls it comparative ‘deletion’.
living in the same area, bought the same item individually from the
same seller and somehow Taroo had to pay more for the carriage due
to the seller’s miscalculation. Again only the sloppy reading is avail-
able10.
Next the comparative deletion construction will be examined.
Hoji (1998: 135) demonstrates that the Japanese comparative dele-
tion construction does not allow the sloppy reading in general and
the strict reading is preferred over the sloppy reading when both are
available. Given the semantic reflexive nature of lexical reflexives,
it is expected that lexical reflexives should produce the sloppy read-
ing despite the general preference for the strict reading. This expec-
tation is borne out. Imagine the same contexts as above for compara-
tive ellipsis construction.
(26) Taroo ga Hanako ga zi-kyoo suru yori hayaku hankoo o

zi-kyoo sita
NOM NOM self-offer do than soon crime ACC
self-offer did
‘Taroo confessed his crime sooner than Hanako’
a. ‘Taroo confessed his own crime sooner than Hanako confessed her own
crime.’ (sloppy)
b. ‘*Taroo confessed his own crime sooner than Hanako confessed his crime.’
(strict)
(27) Taroo ga Hanako ga ziko-hutan suru yori ooku sooryoo o

ziko-hutan sita
NOM NOM self-charge do than more carriage ACC
self-charge did
‘Taroo paid the carriage on his own more than Hanako’
a. ‘Taroo paid more for his carriage than Hanako paid for hers’ (sloppy)
b. ‘*Taroo paid more for his carriage than Hanako paid for Taroo’s’ (strict)
10 It is possible for both (24) and (25) to have the split antecedent interpretation in
which both of them committed the same crime together in (24) or bought the
same item together in (25); however, that is not the reading intended to be
examined here.
104 Hidematsu Miura
The null object construction (NOC) can also be used to show the
reflexivity of zi(ko)-verbs11. (28) is an example of the construction.
As shown in (28), both sloppy and strict readings are available in
some cases (Otani/Whitman 1991).
(28) John wa [zibun no tegami] o suteta.
TOP self GEN letter ACC discarded
‘John discarded self’s letter.’
Mary mo [e] suteta.
also discarded
a. ‘Mary/i also threw out self’s/i letters.’
b. ‘Mary also threw out John’s letters.’
But more generally, the strict reading tends to be preferred over the
sloppy reading just like the comparative deletion construction. Hoji
(1997) points out that in many cases, the NOC allows the strict read-
ing only as in (29b).
(29) a. Johni wa zibun(-zisin)I o nagusameta.

John TOP self(-self) ACC consoled
‘Johni consoled himselfi.’
b. Billj mo [ ] nagusameta.
Bill also consoled
‘Billj consoled himi/*himselfj too.’
Thus NOC provide another robust diagnostic test to show the bound
variable nature of the lexical reflexive construction. (30) and (31)
show that the strict reading is blocked for zi(ko)-verbs, as expected.
(30) a. Taroo ga hankoo o zi-kyoo sita

NOM crime ACC self-offer did
‘Taroo confessed his crime’
11 As Hoji (1998) says, comparative deletion and NOC are expected to show the
same behavior since comparative deletion is a construction which ‘embeds’
NOC in the subordinate (comparative) juncture. In other words, the difference
between comparative deletion and NOC is that the null element of the former
is in intra-sentential position while that of the latter in inter-sentential position.
However, there is a difference also. NOC has a focus particle (mo ‘also’) in the
second sentence which comparative deletion does not have.
b. Hanako mo [ ] zi-kyoo sita

also self-offer did
= ‘Hanako confessed her own crime.’ (sloppy)
= ‘*Hanako confessed Taroo’s crime.’ (strict)
(31) a. Taroo ga sooryoo o ziko-hutan sita

NOM carriage ACC self-charge did
b. Hanako mo [ ] ziko-hutan sita
also self-charge did
= ‘Hanako paid the carriage on her own’ (sloppy)
= ‘*Hanako paid the carriage for Taroo’ (strict)
In this section, it was observed that, with Japanese lexical reflexives,

only the bound variable (i.e. pure-reflexive) interpretation is pos-
sible under any elliptical constructions examined. The fact that Japa-
nese lexical reflexives behave as expected from Lidz’ proposal, Con-
dition R, further supports the universality of the principle.
However, many other regular predicates in Japanese cannot be
morphologically reflexive-marked and do not seem to have any prop-
erties of lexical reflexivity. Nonetheless syntactic reflexives in Japa-
nese (i. e. regular verb + reflexive) produce the pure reflexive inter-
pretation as we will see below. Before looking at the unexpected
behaviors of syntactic reflexives, it will be argued in the next section
that morphologically unmarked regular predicates are neither lexi-
cally reflexive nor do not have an underlying lexical entry for lexical
reflexivity.
3.2 Lexically non-reflexive verbs
In the previous section, it was shown that lexical reflexives in Japa-

nese share the distributional property of Dutch (a subset of the lexi-
con) and the morphological property of Kannada (morphologically
overt marking on the verb). Then, given the cross-linguistic varia-
tion on lexical reflexivity (Section 2), how about other regular verbs
that are not morphologically reflexive-marked? I argue that they have
only one lexical entry which is lexically non-reflexive like Malayalam.
106 Hidematsu Miura
First, I will illustrate how the Japanese introverted verb shave

works in the reflexive context. Second, it will be shown that extro-
verted verbs work exactly in the same way as the introverted ones.
The observation suggests that introvertedness does not constitute one
of the semantic features the organization of the lexicon hinges upon
and that there is only one lexical entry for regular verbs irrespective
of the (non)reflexivity they denote.
The shave verb in Japanese does not take the simple reflexive
zibun ‘self’. This is illustrated in (32). At first glance, this is similar
to the Malayalam case (see 13) and looks like so-called antilocality
in which a reflexive pronoun cannot be locally bound (see 5 for the
Dutch case).
(32) a. *Taroo ga zibun o sotta

NOM self ACC shaved
‘(intended) Taroo shaved himself’
As can be expected from previous data (e.g. Malayalam and Dutch),

the complex form can solve this type of constraint. What is interest-
ing in Japanese is, however, that the complex form zibun-zisin ‘self-
self’ cannot serve either as the locally bound pronoun as demon-
strated in (32b). This is unexpected from the entire discussion above.
Instead, an NP that denotes the exact body-part to be shaved has to
be specified as object as in (32c)12.
b. *Taroo ga zibun-zisin o sotta

NOM self-self ACC shaved
‘(intended) Taroo shaved himself’
c. Taroo ga hige o sotta
NOM beard ACC shaved
‘Taroo shaved (his beard/himself )’
12 As in (i) below, the possessor can be specified by the reflexive pronoun, but it
is generally unexpressed and is pragmatically controlled.
(i) Taroo ga zibun(-zisin) no hige o sotta
NOM self(-self) GEN beard ACC shaved
‘Taroo shaved his own beard’
The following, (33), is another example that shows the relevant point.
The verb, kiru ‘cut’, is generally not regarded as an introverted verb.
(33) Hanako ga daidokoro de *zibun/*zibun-zisin/yubi o kitta

NOM kitchen LOC self/self-self/finger ACC cut
‘Hanako cut her finger/herself in the kitchen’
Thus, the seeming antilocal constraint on reflexive pronouns has no-

thing to do with the introvertedness (i.e. reflexivity) of the event the
verb denotes. Rather it can be hypothesized that the verb has a pecu-
liar selectional restriction in which it simply requires an object NP
that denotes where the event takes place. This prediction can be dem-
onstrated to be true by comparing the above reflexive sentences with
transitive ones.
If the above prediction is correct, the verb should require the
same NP in transitive constructions as well. Indeed, this expectation
is borne out. The sentences in (34) are both transitive constructions
in which two separate entities are involved. As expected, the verb
requires an NP that denotes where the event takes place as in the
reflexive construction (compare 32c and 34b). The contrast between
(34a) and (34b) shows that the seeming antilocality is not due to the
constraint on the reflexive pronouns but is due to the language-spe-
cific selectional restriction of the verbs13.
(34) a. *Taroo ga Ken o sotta

NOM ACC shaved
‘(intended) Taroo shaved Ken’
b. Taroo ga Ken no hige/atama/kao o sotta
NOM GEN beard/head/face ACC shaved
‘Taroo shaved Ken’s beard/head/face’
Given the above observation that the verbs require the same type of
noun as object irrespective of the introvertedness or the transitivity/
reflexivity of the construction, it seems reasonable to assume that
the verbs that do not have any morphological reflexive-marking have
13 The verbs that denote physical activities, especially those that involve change
of state, show this peculiarity in a very consistent manner. See Miura (2002,
2005) for more details on this.
108 Hidematsu Miura
only one lexical entry of the same type (irrespective of the intro-
vertedness or the transitivity/reflexivity). When there is only one lexi-
cal entry, that must be transitive (i.e. lexically non-reflexive) because
it is conceivable to derive reflexives from transitives, but not vice
versa (cf. Sells et al. 1987).
Now we are ready to turn to our primary concern, the behav-
iors of Japanese syntactic reflexives. What we can predict based on
the entire discussion so far is that they should produce both statue
reading and sloppy/strict ambiguity. In the next section, however, it
will be shown this is not borne out.
3.3 Syntactic reflexives
As is well-known, there are three major reflexive forms in Japanese,

zibun ‘self ’, zisin ‘self’, and the combination thereof, zibun-zisin
‘self-self.’ As shown in (35), any of the three forms can be used at the
preverbal object position.
(35) Hanako wa {zibun/zisin/zibun-zisin} o hazita/semeta

TOP self/self/self-self ACC ashamed/blamed
‘Hanako is ashamed of/blamed herself’
It was argued in the previous section that there is good reason to

assume that the Japanese verbs that are not morphologically re-
flexive-marked do not have a lexical entry for lexical reflexivity.
Given the lexically non-reflexive entry of the verb, Condition R ex-
pects reflexives to produce both the pure-reflexive and the near-re-
flexive readings.
First, it will be examined if the so-called ‘statue’ reading is avail-
able or not for syntactic reflexives. Given the lexically non-reflexive
nature of the predicate, Japanese syntactic reflexives are expected to
allow the near-reflexive (e.g. statue) interpretations. Furthermore, the
interpretation should be obtained irrespective of the morphological
complexity of the reflexive forms since Condition R does not make
reference to the morphological complexity of reflexive forms. As
shown in (36), these expectations are not borne out, however.
(36) a. Koizumi shushoo ga zibun ni sawatta (actual person/*?statue)

PM NOM self LOC touched
‘The prime minister Koizumi touched himself’
b. Koizumi shushoo ga zibun-zisin ni sawatta (actual person/*statue)
PM NOM self-self LOC touched
‘The prime minister Koizumi touched himself’
Suppose the same context as the Ringo sentence. The Prime Minister
Koizumi is wandering in a wax museum and finds his statue. Even
though the reflexive in (36a) could be used to refer to the statue of
the prime minister, the acceptability would be highly marginal and
there is some speaker variation on the judgment. The near-reflexive
(statue) interpretation is just not available for the complex form, (36b).
Generally, the Japanese contact verbs such as sawaru ‘touch’, which
do not denote change of state, can take a reflexive pronoun as object
so that the unacceptability cannot be ascribed to the selectional re-
striction discussed in the previous section (although I admit this is
still an arguable point).
There are two unexpected behaviors here. First, given that
Condition R guarantees the ‘unambiguous’ pure-reflexive reading
only through lexical reflexivity, it is unexpected that only the pure-
reflexive interpretation is ‘unambiguously’ available in (36) despite
the absence of lexically reflexive predicate. Second, the difference
in the acceptability between the two reflexive forms is unexpected
since Condition R does not make any reference to morphology, namely
the morphological complexity of reflexive forms.
More compelling data can be given in the comparative con-
structions. Condition R expects that both strict and sloppy interpre-
tations should be available for the comparative ellipsis construction
because the predicate is not lexically reflexive; however, this expec-
tation is not borne out. Here again, only the sloppy (i.e. pure-reflex-
ive) interpretation is ‘unambiguously’ produced without lexically
reflexive predicates as in (37a) (cf. Sells et al. 1987). This is contrary
to what Condition R predicts. (37b) further shows the sloppy reading
is obtained irrespective of the complexity of reflexive forms. Thus
Condition R clearly fails to account for the behaviors of Japanese
syntactic reflexives in this case.
110 Hidematsu Miura
(37) a. Tarooi wa Jirooj yori umaku zibun o bengo sita

TOP than well self ACC defense did
‘Tarooi defended himselfi better than Jirooj defended himselfj’ (sloppy)
‘*Tarooi defended himselfi better than Jirooj defended himi’ (strict)
b. Tarooi wa Jirooj yori umaku zibun-zisin o bengo sita
TOP than good self-self ACC defense did
‘Tarooi defended himselfi better than Jirooj defended himselfj’ (sloppy)
‘*Tarooi defended himselfi better than Jirooj defended himi’ (strict)
It should be noted that Condition R does not make any prediction on

lexically non-reflexive predicates since the principle simply regu-
lates the bidirectional relation between lexical reflexivity and se-
mantic reflexivity. On the other hand, however, there is an implica-
tion drawn from the principle that the ambiguity between the pure-
and near-reflexive readings freely arises in the absence of lexical
reflexivity because there is no system to regulate the interpretive
possibilities of lexically non-reflexive predicates.
In this section, it was demonstrated that Japanese syntactic re-
flexives produce the pure-reflexive interpretation ‘unambiguously’
without lexical reflexivity. This means that there is another way, other
than lexical reflexivity, to achieve semantic reflexivity. In other words,
the pure-reflexive interpretation does not guarantee the predicate is
lexically reflexive. Accordingly, the current bidirectional relation in
Condition R should be weakened to unidirectional one (i.e. lexical
reflexive à pure-reflexive). The unidirectional relation reads: when
the predicate is lexically reflexive, only the pure-reflexive interpre-
tation is unambiguously available, but not vice versa.
There are two questions to be answered: (1) Why do Japanese
syntactic reflexives yield only a pure-reflexive interpretation unam-
biguously despite the absence of a lexically non-reflexive predicate?
(2) Why does the difference in the morphological complexity of the
reflexive forms affect the judgment in some cases?
As we saw, some of the data show acceptability variation. This
implies the phenomena at issue are pragmatic in nature rather than
lexical or semantic. In what follows, it will be argued, in the spirit of
Liu (2003), that the interpretations obtained are a consequence of the
interaction of focus structures.
4. Focus
In the remainder of the chapter, I argue that two types of focus struc-
tures are involved in the phenomena: focus by intensifier and focus
by construction. They will be discussed in turn .
4.1 Focus by intensifier
Reflexives and intensifiers are identical in form in many languages

(König/Siemund 2005)14. Modern English is a good example of this.
(38) is a typical reflexive construction and (39) an intensifier con-
struction. König and Siemund acknowledge two types of intensifier
constructions and call (39a) and (39b) ‘adnominal intensifier’ and
‘adverbial intensifier’ respectively15.
(38) John saw himself in the mirror.
(39) a. The actor himself opened the letter.

b. The actress opened the letter herself.
Although reflexives and intensifiers are morphologically indistin-

guishable in many languages, they are different and intensifiers have
their own morpho-syntactic status in grammar. Syntactically,
(adnominal) intensifiers are adjoined to the host NP as in (40a)
whereas reflexives generally occur as an independent argument as in
14 Their data have been included in WALS (the World Atlas of Language Struc-
tures; Haspelmath et al. 2005) as “(#47) Intensifiers and Reflexive Pronouns”.
Among 168 languages they investigated, 94 languages use identical forms for
both functions and 74 languages differentiated ones. They also have a web site
and all the information about the languages they investigated is available there.
15 Adnominal and adverbial intensifiers are quite different both syntactically and
semantically. I will limit the discussion to the adnominal type and so in what
follows the term ‘intensifier’ means ‘adnominal intensifier’ unless mentioned
otherwise. König and Siemund’s (2005) discussion is based on the cross-lin-
guistic data (around 100 languages), but for the sake of simplicity I will use
English data unless the English data miss the points of their arguments.
112 Hidematsu Miura
(40b). In other words, intensifiers occur in adjunct position while

reflexives in argument position in general.
(40) a. [NP[NP The actor] himself] opened the letter

b. They criticized [NP themselves] in the conference room.
They further claim that although the primary function of intensifiers

is focusing, they are also different from focus particles (e.g. only,
even). Intensifiers inflect for number, gender, person in many lan-
guages whereas focus particles do not (by definition).
The relevant English data are shown in (39). There are invari-
ant intensifiers as well in some languages. For example, German has
a morphologically invariant (‘particle-like’) intensifier, selbst, which
is morphologically different from the reflexive. (41a) is a reflexive
construction and (41b) an intensifier construction. The intensifier can
be adjoined to the reflexive and focus it as in (41c).
(41) a. Johann sah sich im Spiegel.

Johann saw himself in the mirror
‘Johann saw himself in the mirror.’
b. Der Direktor selbst begrüßte uns.
the director himself welcomed us
‘The director himself came to welcome us.’
c. Paul kritisierte sich selbst.
Paul criticized himself himself
‘Paul criticized himself.’
There are prosodic differences between intensifiers and focus par-

ticles. Intensifiers are invariably stressed, whereas focus particles
are not. Compare (46a) and (46b).
(42) a. Even the PRESIDENT came

b. The President HIMSELF came
We can roughly summarize that although intensifiers have some proper-

ties of both reflexives (form) and focus particles (function), they con-
stitute their own category.
The most important for us is the semantic effect of the focus

structure produced by such intensifiers. Focus structure has been ar-
gued in terms of various linguistic orientations (cf. Rooth 1992), but
roughly, it has been the central concern how to formally regulate the
relation between the identity of the focused NP and the alternatives
evoked by the focusing process.
My discussion is based on the basic ideas in Rooth (1985, 1992)
and Eckardt (2001).
Rooth’s (1985, 1992) focus theory, called alternative seman-
tics, has been widely assumed in the literature. He claims, basing his
argument on the analysis of focus particles such as only, that the
general function of focus is to evoke alternatives and focus deter-
mines other focus semantic values. Eckardt (2001) argues that al-
though focus particles and intensifiers are similar to each other in
having a focusing function, they are different. He claims that the
(adnominal) intensifiers are linguistic elements whose function is to
establish identity; namely they have an identity function (ID). (43)
shows how this works. The ID function takes the referent of the fo-
cused NP as an input and exactly the same referent as the input is
cashed out as an output.
(43) [[ the president himself ]] = ID([[ the president ]]) = [[ the president ]]
The semantic operation above looks trivial, but it is not so. Jackendoff
(1992, 1997) observes that in the Ringo sentence, for example, the
subject Ringo must be the very individual and it must not be some
other entity associated with the person (e.g. portrait or statue). Com-
pare (44a) and (44b). This constraint suggests that the ID function is
a necessary operation independently of focus structure, although it is
beyond the scope of this chapter to explore how the ID function works
in the domains other than focus structure.
(44) Ringo fell on himself

a. = Ringo fell on the statue of himself
b. = *The statue of Ringo fell on himself
114 Hidematsu Miura
4.2 Japanese intensifiers
Now it is examined how Japanese intensifiers work. Given the typologi-

cal fact that reflexives and intensifiers are identical in form in many
languages, there are three candidates in Japanese: zibun, zisin and
zibun-zisin. As shown in (45a), only zisin is used as adnominal inten-
sifier while all of them can be used as adverbial intensifier as in (45b).
(45) a. [shushoo *zibun/zisin/*zibun-zisin] ga sore ni dooi sita

prime minister self/self/self-self NOM it DAT agreement did
‘The prime minister himself agreed to it’
b. shushoo ga zeiritsu o zibun/zisin/zibun-zisin de kimeta
prime.minister NOM tax.rate ACC self/self/self-self by decided
‘The prime minister decided the tax rate by himself’
To my knowledge, the intensifier use of zisin has not been detailed in

the literature. It might be of interest to observe some basic facts about
it in passing. As seen above, intensifiers have two classes: the mor-
phologically invariant type (like particles) and the morphologically
variant type that shows some inflections. The adnominal zisin does
not have agreement for gender, number and person unlike European
languages. This is shown in (46)16.
(46) a. Taroo { *?kare-zisin / zisin } ga sore o kimeta

himself / self NOM it ACC decided
‘(intended) Taroo himself decided it’
b. Hanako { *?kanojo-zisin / zisin } ga sore o kimeta
herself / self NOM it ACC decided
‘(intended) Hanako herself decided it’
c. sensee-gata { *?karera-zisin / zisin } ga sore o kimeta
teacher-PL themselves / self NOM it ACC decided
‘(intended) Teachers themselves decided it’
16 The following sentence is fine. It is reasonable to assume that the pronominal

subject is adjoined by the intensifier in the following sentence.
(i) { Kare-zisin / kanojo-zisin / karera-zisin } ga sore o kimeta
he-self / she-self / they-self NOM it ACC decided
‘He himself / She herself / They themselves decided it’
Given the absence of the so-called phi-feature agreement, it seems

reasonable at a glance to conclude that the Japanese intensifier is of
the invariant type (like German). However, Japanese has another
agreement-like system: honorifics. Honorifics have been generally
regarded as agreement of some sort in the literature. Therefore, it is
expected that there is agreement between the host noun and the in-
tensifier. This expectation is borne out as in (47). Japanese focus
particles never have honorific agreement like this; therefore, intensi-
fiers and focus particles are two distinct classes in Japanese17.
(47) a. Ishiguro sensee { go-zisin / ?zisin }ga hoosin o kime-rare-ta

teacher HON-self / self NOM policy ACC decide-HON-PST
‘Dr. Ishiguro himself decided the policy’
b. Taroo { *go-zisin / zisin } ga hoosin o kimeta
HON-self / self NOM policy ACC decided
‘Taroo himself decided the policy’
The default prosodic pattern also supports the distinction. (48) shows
that the same patterns as (42) are observed also in Japanese. Focus
particles (sae ‘even’ here) generally do not receive prosodic peak
whereas intensifiers generally do.
(48) a. ISHIGURO-sensei sae ko-rare-ta

-teacher even came-HON-PST
‘Even Dr. Ishiguro came’
a'. ??Ishigoru-sensei SAE ko-rare-ta
b. Ishiguro-sensei GO-ZISIN ga ko-rare-ta
-teacher HON-self NOM came-HON-PST
‘Dr.Ishiguro himself came’
b'. ??ISHIGURO-sensei go-zisin ga ko-rare-ta
17 Honorific agreement is also observed in the adverbial use of the intensifier as

in (i) as well as the reflexive construction as in (ii) below.
(i) Ishiguro sensee ga {go-zisin/*?zisin} de hoosin o kime-rare-ta
teacher NOM HON-self/self by policy ACC decide-HON-PST
‘Dr. Ishiguro decided the policy by himself’
(ii) Ishiguro sensee ga {go-zisin/*?zisin} o hihans-are-ta
teacher NOM HON-self/self ACC criticize-HON-PST
‘Dr. Ishiguro criticized himself ’
116 Hidematsu Miura
Thus it seems reasonable to assume that zibun is the intensifier in the

language and analyze it on a par with intensifiers in other languages
discussed in the previous section. Namely, zibun has a focusing func-
tion that imposes the ID function on the referent of the NP that it
focuses on.
Now let us turn to the interpretive problems of Japanese re-
flexives. I will begin with the complex form, zibun-zisin. Given the
discussion above, the semantic operation imposed by the intensifier,
i.e. the identification function, is represented in (49).
(49) [[ zibun-zisin ]] = ID ([[ zibun ]])
Zibun is a variable which does not refer to any individual by itself.

The identified referent as the final output of the above operation must
be nothing but the original referent of the antecedent (i.e. subject
argument) since the interpretation of the variable is dependent on the
antecedent. In other words, the ID function applied to a variable makes
the variable into a bound variable. As a natural consequence, the
pure reflexive interpretation is guaranteed. This focus-driven opera-
tion is represented in (50).
(50) ¼x [P(x, [f(x)]F)] Æ ¼x [P(x,x)]
In the comparative ellipsis construction, an argument is missing in

the subordinate juncture. This can be explained by the copy theory
(Kratzer/Heim 1999) that claims that the elided structure has the
same structure as the matrix clause. The bound variable semantic
structure of the matrix juncture is copied onto the subordinate se-
mantic structure replacing the subject argument which serves as the
binder of the bound variable in the subordinate semantic structure.
The sloppy reading is a natural consequence of this operation.
Thus, the interpretations of the complex form, zibun-zisin,
seems to naturally follow from the system hitherto detailed in the
previous literature. On the other hand, however, it must be remem-
bered that the simple (bare) reflexive zibun produces the same inter-
pretations as the complex zibun-zisin.
The bare zibun construction does not have zisin as the intensi-
fier that brings in the ID function. Then we still need to account for
why the same semantic interpretations as the complex reflexive zibun-

zisin result for the simple zibun. I would like to point out that there is
another source for focus in the language that aids the interpretation:
constructional focus.
4.3 Focus by construction
Before jumping onto the issue of the simple zibun, I would like to intro-
duce Chinese data, argued by Liu (2003), that are highly similar to the
Japanese case we are dealing with. Liu claims that the Chinese reflex-
ive, ziji-benshen, produces only the pure-reflexive interpretation un-
ambiguously in the absence of a lexically reflexive predicate. (51) and
(52) are examples that show the (un)availability of the statue reading
and the sloppy reading in the comparative construction respectively.
(51) a. Jiang Jie-Shi/i henhen-de da-le ziji/i yi-xia.

Jiang Jie-Shi furiously hit-Asp self one-Cl
‘Jiang Jie-Shi hit himself furiously.’ (ziji = Jiang Jie-Shi/statue)
b. Jiang Jie-Shi/i henhen-de da-le ziji-benshen/i yi-xia.
Jiang Jie-Shi furiously hit-Asp self-self one-Cl
‘Jiang Jie-Shi hit himself furiously.’ (ziji-benshen = Jiang Jie-Shi/*statue)
(52) a. Zhangsan xianzai bi Lisi guoqu geng guanxin ziji-de liyi.

Zhangsan now compare Lisi past more care-about self-DE benefit
i. ‘Zhangsan/i cares about his/i benefit more than Lisi/j cared about
his/j benefit.’
ii. ‘Zhangsan/i cares about his/i benefit more than Lisi/j cared about
his/i benefit.’
b. Zhangsan xianzai bi Lisi guoqu geng guanxin ziji-benshen-de
liyi.
Zhangsan now compare Lisi past more care-about self-self-DE
benefit
i. ‘Zhangsan/i cares about his/i benefit more than Lisi/j cared about
his/j benefit.’
ii. *’Zhangsan/i cares about his/i benefit more than Lisi/j cared about his
Although the Japanese data (36 and 37) and Chinese data (50 and 51)
look alike, there is a crucial difference between them. The difference
118 Hidematsu Miura
is that the Chinese simple reflexive, ziji, produces pure- and near-
reflexive ambiguity unless used with benshen. On the other hand, as
was shown in (36) and (37), the Japanese simple reflexive, zibun,
shows, without zisin, the pure-reflexive interpretation such as the
unavailability of statue reading or producing the sloppy reading in
elliptical constructions. It is reasonable to assume that Japanese has
another system which imposes the identification (ID) function that
Chinese does not have. I argue that the system is a language-specific,
construction-based focus structure.
It has been pointed out in the literature that preverbal position
is the default focus position for SOV languages (e.g. Kim 1988,
Krifka1998). Japanese is one of such SOV languages. The examples
in (53), from Ishihara (2000), show the canonical SOV word order in
Japanese. The focus is indicated by capitals.
(53) a. Taro ga HON o katta

NOM book ACC bought
‘Taro bought a BOOK’
b. Taro ga kyoo HON o katta
NOM today book ACC bought
‘Taro bought a BOOK today’
Ishihara demonstrates that the preverbal position is indeed the de-

fault focus position in Japanese shown by the following scrambling
data. In (54), the focal NPs in (53) are fronted, losing the positional
focal status, and instead the new preverbal elements receive the po-
sitional focal status of the sentence.
(54) a. hon o TARO ga katta

book ACC NOM bought
‘TARO bought a book.’
b. hon o Taro ga KYOO katta
book ACC NOM today bought
‘Taro bought a book TODAY’
This shows that, everything else being equal, there is a construction-

based inherent focus assignment system. According to Ishihara, the
above observation is acoustically attested as well (e.g. higher F0 on
the focal NP). In sum, this observation means that the object of the
canonical SOV sentences is always in focus18.
Kiss (1998) argues, discussing Hungarian data, that two focus
structures, identificational focus and information focus, must be dis-
tinguished. The former expresses exhaustive identification and the
latter new information. It is claimed that the focus at preverbal posi-
tion in Hungarian, one of SOV languages, is identificational focus.
(55) is an example.
(55) Tegnap este Marinak mutattam be Pétert.

last night Mary.DAT introduced.I PERF Peter.ACC
‘It was to Mary that I introduced Peter last night’
The sentence means, as is clear from the English translation, that,

among the set of individuals present in the domain of discourse, it was
Mary and no one else that the person referred to by ‘I’ introduced Peter
to last night. To put it simply, Peter was introduced only to Mary. The
semantic operation implemented by the exhaustive identification,
which is due to the preverbal identificational focus, is equivalent to
the identification (ID) function (see 43) by Eckardt (2001).
I assume that the identification function is imposed by the focus at
preverbal position also in Japanese. However, the strength of the focus
at preverbal position seems to vary even among the same SOV lan-
guages. According to Kiss (1998), the preverbal focus of Hungarian is
functionally equivalent to it-cleft focus constructions in English. There
is some distributional evidence for this. For example, universal quanti-
fiers cannot occur at preverbal position in Hungarian. Japanese, another
SOV language, does not have such restriction at preverbal position. The
preverbal focus function in Japanese is weaker than that of Hungarian.
18 The constructional focus is, however, not so strong as to always decide the
focus structure of the sentence. As noted in Ishihara, the focus can be easily
overridden by prosodic focus. The sentences in (i) below are a possible prosodic
variation of (56a). Virtually, any element can be assigned a stress.
(i) a. Taro ga kyoo HON O katta
b. Taro ga KYOO hon o katta
c. TARO GA kyoo hon o katta
Taro NOM today book ACC bought
‘Taro bought a book today’
120 Hidematsu Miura
The focus by intensifiers, focus by focus particles or focus by

stress (prosody) is additive focus whereas the constructional focus is
language-specific and structurally inherent focus, so to speak. The
robustness of the constructional focus is not as strong as such addi-
tive-type focus (see footnote 18), but nevertheless the preverbal fo-
cus counts as focus that imposes the ID (identification) function. This
means that since its focus is located at preverbal position, even the
simple reflexive, zibun, is required to have the identity reading even
in the absence of the intensifier, zisin, that lexically assigns the ID
function to the NP it adjoins to.
This analysis makes an interesting prediction. If zibun is
scrambled out of the canonical preverbal focus position, the near-
reflexive interpretation should become available since the ID func-
tion is not imposed on it any longer. We saw that only the sloppy
reading is unambiguously produced in the comparative ellipsis con-
struction (see 37). If zibun is scrambled out of the preverbal position,
it is expected that the strict reading (i.e. near-reflexive reading) should
become easier to get (at least). Although there seems to be speaker
variation, this expectation is borne out as in (56)19,20.
(56) Taroo wa zibun o Jiroo yori umaku bengo sita

TOP self ACC than well defense did
a. ‘Taroo/i defended himself/i better than Jiroo/j defended himself/j’ (sloppy)
b. ‘(?)Taroo/i defended himself better than Jiroo/j defended him/i’ (strict)
It is further expected that the interpretive possibility of the complex

zibun-zisin should not be affected by scrambling since the intensifier
19 Sells et al. (1987) also note the same point. Some of their Japanese informants
judged (56) as acceptable.
20 Scrambling does not seem to change the (un)availability of the statue reading.
There seem to be several options to solve this. One is to assume that the
scrambled element is assigned another focus which outweighs the structural
preverbal focus. The other is to assume that zibun has an inherent [+f(ocus)]
feature (cf. Kiss 1998). I leave this issue open.
(i) *?zibun ni Koizumi shushoo ga sawatta
self DAT PM NOM touched
‘The prime minister Koizumi touched himself ’
(i.e. focus assigner), zisin, is scrambled out together with the reflex-
ive. This expectation is also borne out as in (57)21. This observation
suggests that the focus structure based analysis is on the right track.
(57) Taroo wa zibun-zisin o Jiroo yori umaku bengo sita

TOP self-self ACC than good defense did
a. ‘Taroo/i defended himself/i better than Jiroo/j defended himself/j’ (sloppy)
b. ‘*Taroo/i defended himself/i better than Jiroo/j defended him/i’ (strict)
5. Concluding remarks
In this chapter, I hope I have demonstrated two things. First, Japanese

lexical reflexives (i.e. morphologically reflexive-marked) behave
exactly as expected from the universal principle of Condition R (Lidz
1999, 2001). Lexical reflexives are expected to show only the pure-
reflexive interpretations unambiguously. We could not examine the
availability of the statue reading due to lexical gaps, but we found that
they produce only the sloppy reading unambiguously in the compara-
tive constructions as expected. It can be concluded that the validity of
the principle has been further confirmed by the Japanese data.
Second, more importantly, it was pointed out that the Japanese
syntactic reflexives (i.e. unmarked verb + reflexive) do not show the
expected behaviors. The lexically unmarked predicates were demon-
strated to be lexically non-reflexive. In other words, they are simply
transitive verbs. Given this lexically non-reflexive nature of the predi-
cates, the near-reflexive interpretations are expected to be available.
Contrary to this expectation, however, they consistently produce the
pure-reflexive interpretation. They do not allow the statue reading.
They do not produce the sloppy/strict ambiguity, but they robustly
yield only the sloppy reading unambiguously. It was discussed that the
Japanese pure-reflexivity without lexical reflexivity is due to two types
of focus structure: focus by intensifier and focus by construction.
21 I owe this observation to Takashi Yoshida.

122 Hidematsu Miura
Overall, it was demonstrated by the Japanese data, as Liu (2003)

demonstrated by Chinese data, that there is a pragmatically driven
way to achieve semantic reflexivity (i.e. pure-reflexive interpreta-
tion in Lidz’ term). As a theoretical consequence, it was claimed that
the current bidirectional relation between lexical reflexivity and se-
mantic reflexivity in Condition R should be weakened to unidirec-
tional one which states that lexical reflexivity necessitates semantic
reflexivity but not vice versa.
Acknowledgement
I am grateful to the following people for their suggestions and encouragements to

write this chapter: Teruhiro Ishiguro, Norifumi Ito, Takashi Yoshida, Atsuko Nishiyama,
Maho Furuya and Kiyoko Toratani. All the remaining errors are mine.
References
Bach, Emmon / Bresnan, Joan / Wasow, Thomas 1974. Sloppy iden-

tity: An unnecessary and insufficient criterion for deletion rules.
Linguistic Inquiry 5, 609-614.
Chomsky, Noam 1981. Lectures on government and binding.
Dordrecht: Foris.
Fauconnier, Gilles 1985. Mental Spaces: Aspects of meaning construc-
tion in natural language. Cambridge, Mass.: MIT Press.
Hoji, Hajime 1998. Null Object and Sloppy Identity in Japanese.
Linguistic Inquiry 28/1, 127-152.
Ishihara, Shinichiro 2000. Scrambling and Its Interaction with Stress
and Focus. MIT Working Papers in Linguistics 38, 95-110.
Jackendoff, Ray 1992. MME. Tussaud meets the binding theory.
Natural Language & Linguistic Theory 10/1, 1-31.
Jackendoff, Ray 1997. The Architecture of the Language Faculty.
Cambridge, Mass.: MIT Press.
Kim, Alan Hyun-Oak 1988. Preverbal focusing and type XXIII lan-
guages. In Studies in Syntactic Typology, ed. Michael Ham-
mond / Moravcsik, Edith / Wirth, Jessica. Amsterdam: John

Benjamins, 147-169.
Kiss, Katalin 1998. Identificational Focus versus Information Focus.
Language 74/2, 245-273
Koenig, Ekkehard / Volker Gast. Focused Assertion of Identity: A
Typology of Intensifiers. Linguistic Typology 10.
Krifka, Manfred 1998. Scope inversion under the rise-fall pattern in
German. Linguistic Inquiry 29/1, 75-112.
Lidz, Jeffrey 2001. Condition R. Linguistic Inquiry 32.1, 123-140.
Lidz, Jeffrey 2000. Anti-antilocality. In P. Cole, G. Hermon and J.
Huang, eds, Syntax and Semantics: Long-distance Reflexives,
New York: Academic Press: NY, 227-254.
Liu, Chen-Sheng Luther 2003. Pure Reflexivity, Pure Identity, Fo-
cus and Chinese Ziji-Benshen. Journal of East Asian Linguis-
tics. 12/1, 19-58.
Miura, Hidematsu 2002. Semantic Reflexive Constructions in Japa-
nese: Another Look at Antilocality. ms. University at Buffalo,
SUNY.
Miura, Hidematsu 2005. nihongo no saikihyoogen ni okeru hankyo-
kushosei ni tsuite (‘On the Antilocality in the Japanese reflexive
constructions’). In Ishiguro, T. / Nobuyuki, Y. (eds) Gengo
kenkyuu no setten. Tokyo: Eihosha.
Otani, K. / Whitman, J. 1991. V-Raising and VP-Ellipsis. Linguistic
Inquiry 22, 345-358.
Reinhart, Tanya / Reuland, Eric 1993. Reflexivity. Linguistic Inquiry.
24/4, 657-720.
Rooth, Mats 1985. Association with focus. Amherst, MA: University
of Massachusetts dissertation?
Rooth, Mats 1992. A theory of focus interpretation. Natural Lan-
guage Semantics. 1, 75-116.
Safir, Ken 2004. The Syntax of Anaphora. Oxford: Oxford Univer-
sity Press.
Tsujimura, Natsuko and Aiakawa, Takako 1999. The Types of Zi-
Verbs in Japanese. The Journal of the Association of Teachers
of Japanese. 33/1, 26-43.
Van Valin, Robert and LaPolla, Randy 1997. Syntax. Cambridge: Cam-
bridge UP.
WINNIE CHOR
From nonsubjective to subjective to intersubjective –

a pathway of semantic change in grammaticalization:
the case of faan in Cantonese
1. Introduction1
Grammaticalization2 has been a prominent topic in contemporary

grammatical research. It was apparently first used by the French lin-
guist Antoine Meillet (1912), who defined the term as “the attribu-
tions of a grammatical character to a previously autonomous word”.
Studies on grammaticalization have flourished in the last few de-
cades (Heine et al. 1991; Traugott/Heine 1991; Lehmann 1995; Hop-
per/Traugott 2003; among others) and this particular way in which
items become more grammatical through time is recognized as fun-
damental to language change.
Grammaticalization often involves the development of a lexi-
cal item (e.g. verbs, nouns, adjectives, etc.) into a grammatical item
1 I wish to express my gratitude to Dr KK Luke for having first awakened my

interest in verbal particles and for the many valuable discussions we had on
faan while I was in HKU. I am very much obliged to Dr Jane Simpson for her
very useful comments and suggestions on an earlier version of this chapter. I
am also grateful to Dr Derek Herforth for his always insightful and helpful
comments on my work on particles. I am also thankful to the audience at the
LSHK 2004 and ALS 2006, in which parts of this chapter have been presented.
Of course, none of them are responsible in any way for the content of this
chapter.
2 As from its derivational pattern, ‘grammaticalization’ means a process to which
something is made ‘grammatical’ (c.f. globalization, mechanization, etc.). In
this chapter, when I describe an item as ‘grammatical’, I mean that it has a
grammatical role (i.e. not lexical) rather than that it is grammatically correct
(i.e. conforming to grammatical rules).
126 Winnie Chor
which has a predominant functional role (e.g. clitics, particles, aux-

iliaries, connectives, etc). Despite the categorical change from lexi-
cal to grammatical, there is also a gradual shift of meaning from
relatively more descriptive/referential to relatively more abstract.
Grammaticalization may also be accompanied by other features such
as the reduction or loss of phonetic substance.
There has been considerable interest in the different processes
involved in grammaticalization. Subjectification, one of the seman-
tic-pragmatic processes involved in grammaticalization, has attracted
special interest in recent years (especially Langacker 1990, 1995,
1999; Traugott 1982, 1989, 1995, 1997, 1999; Stein/Wright 1995;
among others). It is a tendency in semantic change, which concerns
how meanings tend to become increasingly based in the speaker or
writer’s subjective beliefs, or attitudes toward what is being said and
how it is being said. Subjectification in grammaticalization is inter-
preted as “the development of a grammatically identifiable expres-
sion of speaker belief or speaker attitude to what is said” (Traugott
1995: 32).
By extending her previous studies on subjectification, Traugott
has brought in the notion of intersubjectification, in the sense that
“the development of meanings that encode speaker/writers’ atten-
tion to the cognitive stances and social identities of addressees, arises
out of and depends crucially on subjectification” (Traugott 2003: 124).
Traugott proposes that the direction of semantic change is essen-
tially subjective > intersubjective.
In this chapter, I will explore the path of semantic develop-
ment involved in the grammaticalization of the Cantonese3 verbal
particle faan4. Verbal particles are pervasive in Cantonese and they
3 Cantonese belongs to the Yue dialect group. According to Matthews and Yip,
“Yue dialects are spoken primarily in the southern Chinese provinces of
Guangdong and Guangxi, and in the neighbouring territories of Hong Kong
and Macau” (1994: 2). In this chapter, the term ‘Cantonese’ refers to the vari-
ety of Cantonese that is spoken in Hong Kong.
4 JyutPing (literally ‘Cantonese romanization’, proposed by the Linguistic Soci-
ety of Hong Kong (LSHK) in 1997), is adopted throughout the chapter. (For
details of this romanization scheme, please refer to: <http://cpct92.cityu.edu.hk/
Jyutping/Jyutpin.htm>).
From nonsubjective to subjective to intersubjective 127
are responsible for expressing subtle nuances of meaning in the lan-

guage. They occur in a post-verbal position and most of them are
grammaticalized from their verb counterparts through the verb-par-
ticle construction. Although the semantic relationship of the particles
to their source verb varies, the developments follow some recog-
nized paths of grammaticalization. Faan, originally a directional verb
meaning ‘go back/return’, has undergone grammaticalization to be-
come a directional particle, giving similar meanings as its verb coun-
terpart. I will show that upon further development, the particle faan
has moved along a semantic pathway from giving nonsubjective
meanings (directional meanings) to subjective meanings (speaker-
oriented meanings) and then to intersubjective meanings (hearer-ori-
ented meanings). I will also illustrate how the scope of faan is ex-
tended from VP-internal, to clausal, then to sentential in its evolution.
More details will be given in Sections 4 and 5.
While most studies see grammaticalization from a historical
angle, this chapter will look at semantic change in grammaticalization
from a more discourse-pragmatic perspective. It is suggested that
pragmatics is an important motor of grammaticalization (Bybee et al.
1994; Hopper/Traugott 2003) and there is often a pragmatic enrich-
ment in the process of grammaticalization, especially in the later
stage. In this chapter, I will describe the various stages of change that
faan has undergone and suggest possible motivations for its de-
velopment.
2. Subjectivity and subjectification
Subjectivity is a pervasive phenomenon in natural languages. It con-

cerns the ‘expression of self and the representation of a speaker’s
perspective or point of view in discourse – what has been called a
speaker’s ‘imprint’ (Finegan 1995: 1). It has been a central notion in
discourse and pragmatic studies as the speaker’s subjective beliefs
and point of view always have an effect in shaping the ongoing dis-
course. As has often been pointed out by these studies (e.g. Benveniste
128 Winnie Chor
1971; Lyons 1982; Langacker 1985, 1990; Traugott 1995), one can-
not possibly say or write anything without expressing some aspect of
personal beliefs or point of view. Natural languages should fulfill the
fundamental need of speakers to convey and assess their feelings,
moods and attitudes. Subjectivity is deeply embedded in our lan-
guage, it is everywhere! Benveniste even raised the question whether
a language could still be called a language without subjectivity as
languages are marked so deeply by the expression of subjectivity. He
remarked that ‘a language without the expression of a person cannot
be imagined’ (1971: 226). It seems obvious that speakers in a dis-
course must take a perspective and encode their point of view to-
wards what they express; this inevitably involves subjectivity.
While subjectivity is pervasive in language use, not all ele-
ments in natural languages serve to express it explicitly. In fact, only
a subset of elements or constructions is used to make subjectivity
explicit. Most of them come to do so through the process of subjecti-
fication, which refers to ‘the structures and strategies that languages
evolve in the linguistic realization of subjectivity or to the relevant
processes of linguistic evolution themselves’ (Finegan 1995: 1).
Subjectification then is a process which involves “speakers recruiting
forms with appropriate meanings to externalize their subjective point
of view” (Traugott 1999: 189).
Traugott (1989) has identified three general tendencies in se-
mantic change. One of these tendencies suggests that meanings tend
to become increasingly situated in the speaker’s subjective belief
state/attitude towards the proposition. It is considered a strong ten-
dency in semantic change and is supported by Traugott’s subsequent
works (Traugott 1995, 1997, 1999, 2003) as well as other studies on
subjectification and grammaticalization (e.g. König 1991; Carey
1995; Brinton 1996; Suzuki 1998).
By treating subjectification as a pragmatic-semantic process
whereby meanings become increasingly based in speakers’ beliefs
about, or attitudes towards, what they are discussing, Traugott (1995)
illustrates how certain expressions that initially articulate concrete,
lexical, and objective meanings have come to serve abstract, prag-
matic, interpersonal, speaker-based functions. She identifies eight
examples of subjectification in English grammaticalization, suggest-
ing that all of them originated in a form with nonsubjective meaning.

This supports the direction of semantic change from nonsubjective
to subjective. One of the examples cited is let us (Traugott 1995: 36):
(1) Let us go, (will you?)
(2) Let’s go, (shall we?)
Traugott suggests that grammaticalization always involves the strength-

ening of the expression of speaker involvement. She observes that the
perspective in (1-2) is shifting from objective to subjective, taking the
speaker’s perspective into more account. In (1), the subject of let is
you while the subject of go is us (not including you). However in (2),
the subject of let and go has been extended to include both the speaker
and the hearer. In other words, the speaker has included himself as part
of the subject giving permission and thus (2) is considered more sub-
jective. Traugott claims that only the structure let us go appeared in
Old English, not let’s go. The phrase let’s go only started to appear in
Middle English. The development of let us is thus consistent with the
tendency nonsubjective > subjective in semantic change.
3. From subjectification to intersubjectification
Traugott’s notion of intersubjectification (Traugott 2003) parallels

her notion of subjectification. While subjectification is more centred
on the speakers, intersubjectification cares more about the addressees;
it is “the explicit expression of the speaker/writer’s attention to the
‘self’ of addressee/reader in both an epistemic sense (paying atten-
tion to their presumed attitudes to the content of what is said), and in
a more social sense (paying attention to their ‘face’ or ‘image needs’
associated with social stance and identity)” (Traugott 2003: 128). In
other words, moving from subjectification to intersubjectification
involves a shift from paying attention to the ‘self’ of the speaker to
caring about the ‘self’ of the addressee. This is further demonstrated
by let us (Traugott 2003: 130):
130 Winnie Chor
(3)=(1) Let us go, (will you?)
(4)=(2) Let’s go, (shall we?)
(5) Let’s take our pills now, Roger. (to a child or a person in a retirement home)
While the use of let’s in (4) is considered subjective; it has an

intersubjective sense in (5), in which the speaker is aware of the
hearer’s image/face needs. According to Traugott, the development
of let us as in (3-5) has been from ‘allow us’ (imperative) to ‘I pro-
pose’ (hortative) and further to ‘mitigator/marker of “care-giver reg-
ister”’. Let’s in (5) is used to show that the speaker is joining in/
playing a role in the activity of pill-taking. The speaker is giving a
sign of assurance to the hearer that he is not alone; he has a compan-
ion. Note that (5) is still a hortative sentence; but, in addition it func-
tions to express the speaker’s concern to the hearer.
This phenomenon of the speaker’s show of care to the hearer’s
needs is not uncommon and it is especially noticeable in some con-
texts. It is not difficult to find, for instance, in a medical discourse
that the doctor starts a conversation by saying ‘how are we feeling
today’? This is a good example demonstrating what a discourse should
be about – the speaker is aware that the hearer is also a participant in
a speech event so he or she has to pay attention to the hearer’s needs
in order to make communication smooth and possible. By using we
instead of you, the doctor is aware of the patient’s possible fear of
being alone.
While in English, intersubjectivity does not have to be neces-
sarily present in the language, it is formally required by the grammar
of some languages, especially those with honorific systems, where
explicit attention must be paid to intersubjectivity in these languages.
Japanese, for instance, requires language users to select appropriate
honorific forms depending on the relationship between the speaker
and the addressees. Honorifics in Japanese are referred to as keigo,
literally ‘respectful language’. They are used to serve several pur-
poses, such as making oneself humble, being formal, being polite, etc.
In some cases, the honorific form is selected on the basis of the
speaker/writer’s social standing relative to the referents of an ex-
pression (‘referent honorifics’), while in other cases, it is on the basis
of the speaker/writer – addressee/reader relationship independent of

the referents (‘addressee honorifics’) (Brown/Levinson 1987; Dasher
1995, cited in Traugott 2003) (for more examples, see Onodera 2000;
Traugott/Dasher 2002, ch. 4). As Traugott remarks, “in either case
the choice of expression is both subjective (dependent on the point
of view of SP/W) and at the same time intersubjective (dependent on
SP/W’s conceptualization of his or her relationship to AD/R and AD/
R’s image needs at the time of the speech event)” (2003: 129).
Apart from Japanese, some elements in Chinese have also gone
through subjectification and intersubjectification to express subjec-
tive and intersubjective meanings. In fact, there have been quite a
few papers written from these perspectives (Zhang 2005; Li 2005;
Zeng 2005). However, studies on Cantonese in this area are rela-
tively scanty. By using the Cantonese directional verb faan (literally
‘go back/return’) as an illustrative example, I will show how it is
employed to express subjectivity and intersubjectivity explicitly in
the language. Originating as a directional verb, faan has undergone
grammaticalization to become a particle giving concrete and abstract
directional meanings. It has also gone through the semantic change
pathway from giving nonsubjective to subjective and then to
intersubjective meanings. Thus it is consistent with Traugott’s sug-
gestion about the strong tendency in semantic change.
4. The grammaticalization of faan: from verb to particle
Faan as a directional verb means ‘go back/return’. It indicates a move-

ment towards a goal which is usually the source from which the mover
started. The verb faan in Cantonese is comparable to the verb huí
(⚆) in Mandarin, meaning ‘to go back to (the original place)’. As
with other directional verbs, faan usually takes a locative object and
puts emphasis on the return of the subject to this location, which is
the starting point of the movement. For instance:
132 Winnie Chor
(6) 彼⯳ẩġ
faan uk-kei
faan home
‘return home’
The verb faan has undergone grammaticalization to become a par-

ticle, which change (verb > particle) is common in grammaticalization
cross-linguistically. As a directional particle5, faan has gone through
several metaphorical shifts, from giving directional meanings in the
spatial domain to other more abstract domains (sentences 7-10). Again,
this shift from concrete to abstract is typical in grammaticalization.
(7) 崘彼⬠㟉ġ
zau faan hok-haau
run faan school
‘run back to school’ [spatial return]
(8) ⴤ१ԫʕ࢓ࠃʳ
nam faan jat-di wong-si
think faan some past-events
‘think back over some past events’ [cognitive return (to past)]
(9) ‫ش‬१ʕ‫ء‬஼
jung faan go bun syu
use faan that CL book
‘go back to using that book’ [situational return ]
(10) ᝑ१‫إ‬ᇩଡᠲ‫ؾ‬
gong faan zing-waa go tai-muk aa
talk faan previous CL topic PRT
‘return to the previous topic’ [discourse return]
5 Some studies on Chinese have used the term fĪngxiàng bu ï yu

ï ( 㕡⎹墄婆 ),
literally ‘directional complement’, to describe what I refer to as ‘directional
particle’ here. In English, ‘complement’ generally refers to a word or a phrase
which is necessary in a sentence to complete its meaning. For instance, an
object is considered a complement to a verb. In Chinese, however, buyu is
something rather different. It is often understood as a post-verbal element which
describes the state or result as denoted by the verb (Zhu 1982: 125). In order to
avoid potential confusions between the notions buyu in Chinese and ‘comple-
ment’ in English, I use the term ‘directional particle’ as a cover term for post-
verbal elements whose basic function is to denote direction.
As Hopper and Traugott noted, when grammaticalization takes place,

some of the original concrete meaning of the lexical item will be
lost, specifically motion and directionality (2003: 3). As in sentences
(7-10), the ‘return’ as expressed by faan is becoming more abstract
and less likely to involve actual motion. The ‘return’ demonstrated
in (8-10) is no longer an overt external return, but rather an abstract
conceptual return. Nevertheless, this faan still retains a very impor-
tant property of its verb counterpart – the place or state that the theme
subject is returning to is where the movement started in the first place.
Metaphorical extension from more concrete to more abstract
domains is generally accepted to be fundamental in semantic change.
It is also typical in grammaticalization, especially in the early stage
(Heine et al. 1991: 151). While the original concrete meanings are
lost, some new meanings will be added to the item at the same
time; these are “more abstract and speaker-based meanings” (Hop-
per/Traugott 2003: 3).
5. Pathway of semantic change:

nonsubjective > subjective > intersubjective
The particle faan is observed to have undergone a semantic shift

from giving directional meanings to meanings involving an increase
in subjectification and abstraction. It has, over time, acquired a soft-
ening function and behaves as an element which helps to express the
speaker’s positive attitude toward the action named by the verb, as in
this much-cited example:
(11) ‫┊װڣݺ‬१‫נ‬ᚭ೺Μ
ngo dei heoi tai faan ceot hei laa!
I PL go see faan CL movie PRT
‘(Let’s) go and see a movie!’
Here, the speaker is not suggesting that they should ‘return’ to the
seeing of movie. He is making a suggestion that he thinks it would be
134 Winnie Chor
good to enjoy a movie. This appraisal of ‘enjoyment’ is essentially

subjective and personal. The use of faan implies that the movie is
essentially a good one from the speaker’s perspective.
This particular use of faan has been identified in the literature.
Zhan (1958) suggested that this non-directional faan is simply a cliché,
a position also taken by Gao (1980). Others have argued that it is a
grammatical marker without clear content (Yuan 1989). Most pa-
pers, however, have attempted to find out what exactly the semantic
and grammatical status of faan is, as well as any contextual con-
straints on its use. Cheung (1972) and Peng (1999) argued that this
seemingly non-directional faan is still related to ‘return’, but a con-
ceptual/metaphorical return to a deserved state. Other studies have
noticed that this faan is related to ‘for the good of oneself’ (Matthews/
Yip 1994) and ‘enjoyment/ideal’ (Tang 2001). However, few have
addressed ‘how’ and ‘why’ this sense of ‘good’/‘enjoyment’ develops.
Looking at the examples given in all these studies, the most
salient feature in common is the positive connotation associated with
faan. See below for more examples from their studies:
(12) ‫ݺ‬უᝑ१ࠟ‫؁‬Ζ
ngo soeng gong faan loeng geoi
I wish say faan two sentence
‘I wish to say a few words (on something good).’ (Gao 1980)
(13) ᐊ१ේනΖ
se faan di je
write faan some thing
‘(I wish) to write something (which is worth writing on).’ (Yuan 1989)
(14) ‫ݺ‬უ၇१ԫଡΖ
ngo soeng maai faan jat go
I wish buy faan one CL
‘I wish to buy one (which is worth buying).’ (Matthews/Yip 1994)
(15) ቮ१Ǝ़ᓳΖ
taan faan haa hung-tiu
enjoy faan while air-conditioning
‘(I wish) to enjoy the air-conditioning for a while.’ (Peng 1999)
(16) ‫ݺ‬უ堬१ࣦ࠼೽Ζ
ngo soeng jam faan bui gaa-fe
I wish drink faan CL coffee
‘I wish to have a (nice) cup of coffee.’ (Tang 2001)
None of these sentences can be interpreted as associated with nega-

tive meanings. For instance, it would be extremely odd if faan as in
(16) is used with negative connotations:
(17) *່ኰএ堬१ࣦ࠼೽Ζ
zeoi caam hai jam faan bui gaa-fe.
most poor BE drink faan CL coffee
‘The worst thing to do is to have a cup of coffee.’
Sentence (17) is semantically anomalous probably because the se-

mantics of faan [+enjoyable] clashes with the semantics of caam
‘dreadful’ [-enjoyable]. This positive/desirable sense of faan is even
more apparent if we consider (18-19) below:
(18) ൿ༏‫چ‬Εࢳೋ࿗Δक़ෙࠟଡᤪΜ
sou fuk dei mut zek coeng faa zo loeng go zung
sweep CL floor clean CL window spend PERF two CL hour
‘Swept the floor; cleaned the window – it took (me) two hours!’
(19) ൿ१༏‫چ‬Εࢳ१ೋ࿗Δ
sou faan
fuk dei mut faan zek coeng
sweep faan
CL floor clean faan CL window
क़ෙࠟଡᤪΜ
faa zo loeng go zung
spend PERF two CL hour
‘Swept the floor; cleaned the window – it (pleasantly) took (me) two hours!’
Without faan, activities like sweeping the floor or cleaning the win-
dow are not necessarily enjoyable. With the use of faan, these other-
wise dull and annoying activities become pleasurable and worth
doing. One can add remarks like ‘but I enjoyed it!’ or ‘it’s very pleas-
urable!’ to (19), but not (18). Faan here can be considered as having
an expressive function, to express the affective and emotional feeling
of the speaker. As for (18), on the other hand, one might expect com-
ments like ‘it’s very tiring!’ or ‘that’s backbreaking!’. Faan in (19) is
136 Winnie Chor
actually expressing some kind of ‘mood’ (yu ï qì). As one can see from
these examples and many others, the original directional meaning of
faan (i.e. ‘go back/return’) is not there anymore. Faan involves a shift
of meaning from objective (sense of direction) to subjective (the ap-
praisal of ‘being good’). It is recruited by speakers of Cantonese to
‘externalize their subjective point of view’ (Traugott 1999: 189), in
this case the speaker’s assessment of ‘being good’.
As Finegan remarks, “in many other languages, including En-
glish, subjectivity is marked in ways sufficiently subtle to be easily
overlooked, and sufficiently complex to prove challenging to expli-
cate” (1995: 3). How can we get to this sense of goodness from the
sense ‘go back’? Why is the shift of perspectives from impersonal to
speaker-oriented possible? These are the interesting issues which need
to be addressed.
In the first instance, how can the semantics of ‘pleasurable’ be
developed from the seemingly unrelated sense of ‘direction’6? An
explanation for this development can be found by looking again at
the core semantics of faan as a verb: back to the source. This ‘source’
is often interpreted as the original point from which the movement
started, as in:
(20) ‫ࡵݺ‬୮१ᖂீ
ngo ji-gaa faan hok-haau
I now faan school
‘I now go (back) to school.’
Sentence (20) implies that some time in the past the speaker has been
to the school and now he or she is going back to that school. Interest-
ingly, in some situations, faan can be used with a place where the
speaker has never been to:
(21) १խഏ
faan zung-gwok
faan China
‘go (back) to China’
6 I am especially thankful to Dr K.K. Luke for the helpful discussions we had on

the possible development ‘direction’ > ‘good’.
A speaker who has never been to China can still utter (21), but with
the condition that China is his ancestral homeland. If (21) is uttered
by a non-Chinese, faan can only mean a physical return: ‘China’ is a
place which the person left, and is now coming back to. This obser-
vation can be accounted for if we consider again the meaning of
faan: back to source. Metaphorically, this ‘source’ is extended to one’s
native place and can be interpreted as one’s ‘source of origin’ – the
place where one comes from or belongs to. If the speaker of (21) is a
Chinese, then China is considered his ‘source of origin’. Even if he
has never been to China, he can still use faan as he is actually ‘back
to the source’, back to where he belongs to and originates from! Fol-
lowing this direction, we can see why examples (22-24) below are
possible:
(22) ‫ֲᦫݺ‬รԫֲ१ՠΖ
ngo ting-jat dai-jat jat faan gung.
I tomorrow first day faan work
‘Tomorrow is my first day of work.’
(23) ‫ݺ‬ଡՖᦫֲၲࡨ१ؔᆇႼΖ
ngo go neoi ting-jat hoi-ci faan jau-zi-jyun.
I POSS daughter tomorrow start faan kindergarten
‘My daughter will start going to the kindergarten tomorrow.’
(24) ‫ݦݺ‬ඨ‫ڶݺ‬ԫֲ‫אױ‬१֚ഘΖ
ngo hei-mong ngo jau jat jat ho-ji faan tin-tong.
I hope I have one day can faan heaven
‘I hope I can return to heaven one day.’
The ‘sense of belonging’ that faan possesses is more salient in (22-

24). One can probably work out from the contexts that the subjects
have never been to the places where they faan (return to): the work-
place, the kindergarten, and most obviously, heaven! However, faan
can still be used because places of work or study are often closely
associated with one’s place of belonging and where habitual struc-
tured activities are performed. These are places which define a
person’s identity and which a person is emotionally attached to: a
work place, a study place, and, an ‘eternal home’ (from a religious
perspective). This explains why faan cannot be used with any places,
138 Winnie Chor
unless they are related to work or study or where one is emotionally

attached to and has intimacy with. Sentences like 彼⛘䚌 ‘faan build-
ing-site’, 彼℔⚺ ‘faan garden’ and 彼⹩㇨ ‘faan toilet’ are bad un-
less these are one’s places of work:
(25) ‫אݺ‬ছ१‫چ‬ᒌΔᦫֲၲࡨ१ֆႼΖ
ngo ji-cin faan dei-pun ting-jat hoi-ci faan gung-jyun
I past faan building-site tomorrowstart faan garden
‘I used to work in a building-site, will start working in a garden tomorrow.’
Sentence (25) is perfectly acceptable. Faan can be used with any

locative object as long as the speaker considers that place as his ‘home’/
where he belongs to. There is a general, cross-cultural conception of
‘home’ as a place of love and shelter. Returning to this source/where
one belongs to is often associated with a good and positive feeling,
especially in Chinese society. This association of ‘home is good’ can
bridge the gap between the directional sense ‘return’ and the subjec-
tive sense ‘enjoyment’: returning to places where one belongs to often
creates a pleasurable feeling. These [+pleasurable] and [+positive]
features of faan are retained when it is grammaticalized and becomes
a verbal particle. In other words, it is this implication of positiveness
which drives the semantic development of faan from a verb to a par-
ticle, making the semantic path ‘objective direction’ > ‘subjective
good’ possible. When this implicature of ‘goodness’ is semanticised,
the particle faan can be used with something subjectively good, with-
out any association with ‘direction’.
At a more advanced stage, faan is observed to have undergone
further development, from giving speaker-based meanings to hearer-
oriented meanings. This fits well into Traugott’s proposal of the se-
mantic development subjective > intersubjective.
While the subjective meaning of faan has received a lot of
attention, its intersubjective sense is less studied. The pleasurable
feeling of the ‘self’ of the speaker has been extended to the ‘self’ of
the hearer. This development ‘speaker-oriented’ > ‘hearer-oriented’
is attested in Traugott (2003) and it is motivated by the hearers’ face
wants in a conversation. As a marker of intersubjectification, faan is
used as a tone-softener, to create a sense of ‘mood’ alongside the
sense of request and suggestion, with the hearer’s own comfort taken
into more account. For instance, faan is commonly used as a strategy

to change topics politely in a conversation. This is particularly no-
ticeable in forums or phone-in programmes on the radio. In many
occasions, the host can make use of faan to bring the guest back to
the theme of the talk. For instance, if the programme is about China
and the guest is getting away from this topic, the host can utter:
(26) ᝑ१խഏᖵ‫׾‬෎…
gong faan zung-gwok lik-si aa
talk faan China history PRT
‘(Let’s) talk about Chinese history…’
By uttering (26), the host is trying to bring the guest back to the topic
of Chinese history, which may or may not have been mentioned or
talked about earlier during their conversation. In cases where the
topic has not been mentioned previously, the use of faan provides an
opportunity for the hearer to make an inference: the speaker is actu-
ally making a suggestion to introduce a topic in a polite way, as if we
are just returning to a previously discussed topic but not really chang-
ing the topic. Faan is used as a conversational strategy – whatever
topic appearing after it is the one which the speaker intends to talk
about. This strategy is also used frequently among friends. If a friend
has been talking about a single topic for a really long time, the other
can probably change the topic in a polite way by making use of the
structure ‘gong faan XX…’ (X = your topic in mind). In these occa-
sions, the verbal particle faan is used as a tone softener for the hearer’s
own comfort, making the utterance sound more friendly and polite.
This politeness strategy is essentially directed to the hearer (hearer-
oriented).
In fact, this intersubjective faan is also used in many other
occasions. Here are some examples which I have come across:
(27) (A teacher’s instruction to a young pupil)

ෘࡋ৫ᐊ१‫܃‬ଡ‫ټ‬Ζ
hai nei-dou se faan nei go meng.
in here write faan your CL name
‘Please write your name here.’
140 Winnie Chor
(28) (A reporter’s question directed to the Chief Executive of Hong Kong)7

უം१མ٣‫…ࡋس‬G
soeng man faan zang sin-saang ne…
wish ask faan Tsang Mr PRT
‘I’d like to ask Mr Tsang…’
(29) (An instruction given in a TV cookery programme)

‫ף‬१༓ศᤑ▏Δ᧓१༓ઞΖ
gaa faan gei lap mo-gu, gaau faan gei miu.
add faan few CL mushroom stir faan few second
‘Add a few mushrooms (into the bowl) and stir for a few seconds.’
In (27-29), faan has no concrete semantic meanings. It is used solely

as a tone-softener to create a good and comfortable feeling towards
the hearers. For instance, by uttering (27), a teacher is asking a child
to write his name on an assignment sheet. By using faan, the teacher
is not asking the child to ‘return’ to the action of writing his name,
nor is the teacher expressing any kind of positive feeling about the
act of writing towards himself. Rather, faan is used intersubjectively,
to soften the teacher’s tone and is interpreted as a sign of his concern
for the addressee (the young pupil). Sentence (27) would become
more blunt and rude if it were used without faan.
Example (28) is another instance where the intersubjective faan
is exploited. At a government press conference in Hong Kong, a re-
porter uttered (28) as a preface to a question directed at the Chief
Executive. In the actual speech situation, it was obvious that he was
not returning to the act of asking a question, but asking his first ques-
tion. Faan in this situation is used to soften the tone and is inter-
preted as a sign of the reporter’s respect for Mr Tsang, the Chief
Executive. The preface sounds more polite with the use of faan; and,
politeness typically implies a certain degree of respect. It would be
more direct and straightforward if faan were not used.
The intersubjective sense is even more salient if we consider
(29), something typically heard in a cookery television programme.
Faan is simply employed to create a good and comfortable feeling
towards the audience generally.
7 I am grateful to Dr K.K. Luke for drawing my attention to this example, which

initiated me to think about the intersubjective sense of faan.
To conclude this section, faan as a particle is observed to have

undergone, in Traugott’s terms, subjectification and intersubjectifica-
tion in its pathway of semantic change. The development is found to
be comparable with that of let’s in English (as illustrated in (3-5)) –
from objective (impersonal) to subjective (speaker-based) and further
to intersubjective (addressee-oriented). At the latest stage, let’s is solely
used as a discourse marker, which has little orientation to the hearer:
(30) Let’s see, what shall we do now?
In (30), let’s see as a fixed phrase is more like a sentence particle

“serving the textual and metalinguistic function of bracketing a unit
of discourse” (Traugott 1995: 37). As Lewis (2003) observes, it is
not uncommon for a shift from a lexical item to a discourse particle.
There is also a possibility that faan will become more empty and at
the end solely serve as a discourse marker of mood.
Another interesting observation about faan’s development is
that it has illustrated how the scope of a particle can be extended
from VP-internal, to clausal, then to sentential. As Lewis (2003) re-
marks, “Many discourse markers/particles can be seen to have de-
veloped from extant lexemes/phrases. The semantic change involved
is in the direction of greater subjectification, increased discourse func-
tion and increased scope […] many discourse particles can be traced
back to VP-adverbials/other expressions which developed into sen-
tence-adverbs, acquiring proposition-wide scope and taking on new,
speaker-oriented functions”. It is worth noting that changes in func-
tion do not necessarily bring a change in form. At the early stage,
faan as a particle is used to show the directional ‘return’ as desig-
nated by the verb it is attached to. Later on when faan has gained the
subjective and intersubjective senses, its scope is extended and it has
an effect on the whole clause or proposition. The speaker-oriented or
hearer-oriented meanings that faan expresses are not to deal with the
verb only, but the whole proposition. Notice that the position of faan
is not moving to a higher level (such as at the end of a clause/sen-
tence) with the scope extension, but is staying in the same post-ver-
bal position. Faan as an example can show that particles appearing
after verbs do not only deal with verbs.
142 Winnie Chor
6. Motivations for change
As Stubbs noticed, “whenever speakers (or writers) say anything,

they encode their point of view towards it: whether they think it is a
reasonable thing to say, or might be found to be obvious, irrelevant,
impolite, or whatever” (1986: 1). Language is a strategy for speakers
to express what they think or feel explicitly. This is why subjectifica-
tion as well as intersubjectification is motivated.
According to the principle of economy, participants in a dis-
course always try to put in only the amount of effort needed to make
communication possible. As Martinet remarks, “in order to under-
stand how and why a language changes, the linguist must keep in
mind two ever-present and antinomic factors: first, the requirements
of communication, the need for the speaker to convey his message,
and second, the principle of least effort, which makes him restrict his
output of energy, both mental and physical, to the minimum compat-
ible with achieving his ends” (1962: 139). In other words, a speaker
prefers expressions that are simple and short while a hearer prefers
expressions that are informative and easy to understand.
As Heine et al. (1991) commented, “grammaticalization can
be interpreted as the result of a process which has problem-solving
as its main goal”. There are conversational needs for a speaker to
express his subjective feelings and also for him to take care of the
face wants of the hearers – the other participants in the conversation.
A challenge is raised to the language to find a suitable candidate in
the grammatical system to satisfy these needs. Certainly, there are
always different ways in achieving the same goal in language use.
However, the question would be: is there an appropriate and conve-
nient form that a language user can use so that he can convey his
message effectively with the least amount of effort? It is this need
which motivates the development of faan, which at the end leads to
its pragmatic enrichment.
Subjectification and intersubjectification can be considered as
conversational strategies which are used to facilitate communica-
tion. Faan is a grammatical resource that users of Cantonese employ
to express subjectivity and intersubjectivity in a more economical
way. Instead of literally using phrases like it is very enjoyable to… or

I feel pleasurable when I…, speakers can simply make use of faan to
express their subjective evaluation of ideal and enjoyment. In a simi-
lar way, faan is also employed intersubjectively to show the speaker’s
care towards the hearer in an economic way. In order to make com-
munication possible, it is important that the speaker is aware of the
‘self’ of the addressee and take into account his relationship with the
addressee and that person’s face/image needs. This can be done in an
easy and relatively effortless way with faan.
7. Conclusion
The Cantonese verbal particle faan serves as a very good example of

how objective, impersonal meanings can become, over time, subjec-
tive as well as intersubjective meanings. This development is consis-
tent with Traugott’s suggestion of a typical semantic pathway in
grammaticalization: nonsubjective > subjective > intersubjective. The
shift also involves a move from external context (direction), to
speaker’s internal evaluation (whether something is good/pleasur-
able), and further to inter-speakers relationship (an attempt to be polite
in consideration of the speaker-hearer relationship).
This chapter is aiming at a full picture of faan and it has given,
I hope, a more comprehensive account of the semantic change path-
way of faan as well as how and why the various senses develop. The
non-directional senses of faan are resulted from subjectification and
intersubjectification, which processes are pragmatically motivated
and confined to the principle of economy. Speakers prefer putting in
the least effort for an utterance which is informative enough for the
hearers to understand. This is easily achieved by the use of faan, in
which case the speaker can convey the more complicated subjective
and intersubjective meanings by using only one particle. More effort
would be needed if otherwise.
The semantic change pathway nonsubjective > subjective >
intersubjective is accepted to be typical in grammaticalization. This
144 Winnie Chor
is not restricted to English or other European languages, but is also

observed in Cantonese, as well as other Asian languages such as Korean
and Japanese. Although there is a paucity of semantic change studies
on Chinese languages and apparently the shift nonsubjective > sub-
jective > intersubjective has not yet been attested in many Chinese
dialects; more instances of subjectification and intersubjectification
are likely to be uncovered. In Cantonese, faan has gone all the way
from a verb to a particle, from giving objective meanings, to subjec-
tive and further to intersubjective meanings. This evolution of faan is
still ongoing, and it remains to be seen in what directions it will go.
References
Benveniste, Emile 1971. Subjectivity in language. Problems in gen-

eral linguistics, trans. Mary Elizabeth Meek, 223-230. Coral
Gables: FL: University of Miami Press (originally published
as ‘De la subjectivité dans le langage’, in Problémés de linguis-
tique générale. Paris: Gallimard, 1958, 258-266).
Brinton, Laurel J. 1996. Pragmatic markers in English: grammati-
calization and discourse function. Berlin: Mouton de Gruyter.
Brown, Penelope / Levinson, Stephen C. 1987. Politeness: some univer-
sals in language usage. Cambridge: Cambridge University Press.
Bybee, Joan, Revere Perkins / Pagliuca, William 1994. The evolu-
tion of grammar: tense, aspect and modality in the language of
the world. Chicago/London: The University of Chicago Press.
Carey, Kathleen 1995. Subjectification and the English perfect. In
Stein, Dieter / Wright, Susan (eds) Subjectivity and subjectivi-
sation in language. Cambridge: Cambridge University Press,
83-102.
Cheung, hung-nin 1972. Cantonese as spoken in Hong Kong. Hong
Kong: The Chinese University Press.
Dasher, Richard 1995. Grammaticalization in the system of Japa-
nese predicate honorifics, PhD Dissertation. Standford Uni-
versity.
Finegan, Edward 1995. Subjectivity and subjectivisation: an introduc-

tion. In Stein, Dieter / Wright, Susan (eds) Subjectivity and sub-
jectification. Cambridge: Cambridge University Press, 1-15.
Gao, Huanian 1980. Guangzhou fangyan yanjiu (Guangzhou dialects:
a comprehensive study). Hong Kong: HK Commercial Press
Ltd.
Heine, Bernd / Claudi, Ulrike / Hünnemeyer, Friederike 1991. Gram-
maticalization: A conceptual framework. Chicago: University
of Chicago Press.
Hopper, Paul J. / Traugott, Elizabeth C. 2003. Grammaticalization
(2 edn.). Cambridge: Cambridge University Press.
König, Ekkehard 1991. The meaning of focus particles: a compara-
tive perspective. London: Routledge.
Langacker, W. Ronald 1990. Subjectification. Cognitive Linguistics,
1, 5-38.
Langacker, W. Ronald 1995. Raising and transparency. Language,
71, 1-62.
Langacker, W. Ronald 1999. Losing control: grammaticization, sub-
jectification, and transparency. In Blank, Andreas / Peter Koch
(eds) Historical semantics & Cognition. Berlin: Mouton de
Gruyter, 147-175.
Lehmann, Christian 1995. Thoughts on grammaticalization.
Munchen: LINCOM Europe.
Lewis, Diana 2003. Rhetorical motivations for the emergence of dis-
course particles, with special reference to English of course.
Particles, ed. by Belgian Journal of Linguistics 16, 79-91: John
Benjamins Publishing Co.
Li, Jinxin 2005. Hao de yufahua yu zhuguanxing (The grammati-
calization and subjectivity of hao). Shijie hanyu jiaoxue (Chi-
nese teaching in the world), 1, 44-49.
Lyons, John 1982. Deixis and subjectivity. In Jarvella, J. Robert /
Wolfgang Klein (eds) Speech, place, and action: studies in
deixis and related topics. New York: John Wiley, 101-124.
Martinet, André 1962. A functional view of language. Oxford: Ox-
ford University Press.
Matthews, Stephen / Yip, Virginia 1994. Cantonese: a comprehen-
sive grammar. London; New York: Routledge.
146 Winnie Chor
Meillet, Antoine 1912. L’evolution des formes grammaticales.

Scientia (Rivista di Scienza) 12, no. 26, 6. Reprinted in Meillet,
Antoine 1958. Linguistique historique et linguistique générale,
130-148. Paris: Champion.
Onodera, Noriko O. 2000. Development of demo type connectives
and na elements: two elements of Japanese discourse markers.
Journal of Historical Pragmatics, 1, 27-55.
Peng, Xiaochuan 1999. Guangzhou de dongtai zhuci ‘fan’ (The par-
ticle faan in Guangzhou). Fangyan, 1, 64-69.
Stein, Dieter / Write, Susan (ed.) 1995. Subjectivity and subjectivisa-
tion in language. Cambridge: Cambridge University Press.
Stubbs, Michael 1986. ‘A matter of prolonged field work’: notes to-
ward a modal grammar of English. Applied Linguistics, 7, 1-25.
Suzuki, Ryoko 1998. From a lexical noun to an utterance-final prag-
matic particle: wake. In Toshio Ohori (ed.) Studies in Japanese
grammaticalization. Tokyo: Kurosio Publishers, 67-92.
Tang, Szewing 2001. Yueyu zhuci ‘fan’ de yingyong tiaojian (The
Cantonese particle faan: its conditions of use). Zhongguo yuwen
tongxin, 60, 50-55.
Traugott, Elizabeth C. 1982. From propositional to textual and expres-
sive meanings: some semantic-pragmatic aspects of gramma-
ticalization. In Lehmann, Winfred P. / Malkiel, Yakov (eds)
Perspectives on Historical Linguistics. Amsterdam: Benjamins,
245-271.
Traugott, Elizabeth C. 1989. On the rise of epistemic meanings in
English: an example of subjectification in semantic change.
Language, 57, 33-65.
Traugott, Elizabeth C. 1995. Subjectification in grammaticalization.
In Stein, Dieter / Wright, Susan (eds) Subjectivity and subjecti-
fication. Cambridge: Cambridge University Press, 1-15.
Traugott, Elizabeth C. 1997. Subjectification and the development
of epistemic meaning: the case of promise and threaten. In
Swan, Toril / Jansen, Olaf Westvik (eds) Modality in Germanic
languages: historical & comparative perspectives. Berlin: Mou-
ton de Gruyter, 185-210.
Traugott, Elizabeth C. 1999. The rhetoric of counter-expectation in
semantic change: a study in subjectification. In Blank, Andreas /
Koch, Peter (eds) Historical semantics & Cognition. Berlin:

Mouton de Gruyter, 177-196.
Traugott, Elizabeth C. 2003. From Subjectification to intersub-
jectification. In Hickey, Raymond (ed.) Motives for language
change. Cambridge: Cambridge University Press, 124-139.
Traugott, Elizabeth C. / Dasher, Richard B. 2002. Regularity in se-
mantic change. Cambridge: Cambridge University Press.
Traugott, Elizabeth C. / Heine, Bernd 1991. Approaches to gram-
maticalization. Amsterdam/Philadelphia: John Benjamins.
Yuan, Jiahua 1989. Hanyu fangyan gaiyao (Chinese dialects: an over-
view), (2 edn.). Peking: Wenzi Gaige Chubanshe.
Zeng, Liying 2005. Wo-kan yu ni-kan de zhuguanhau (The sub-
jectivisation of wo-kan and ni-kan). Hanyu xuexi (Chinese Lan-
guage Learning), 2, 15-22.
Zhan, Bohui 1958. Yue fangyan zhong de xuci: qin, zhu, fan, mai,
tian (The particles can, zyu, faan, maai, tim in the Yue dialect).
Zhonggua yuwen (Chinese Language), 3, 119-122.
Zhang, Yisheng 2005. Fuci dou de yufahua yu zhuguanhau – jianlun
dou de biaoda gongyong he neibu fenlei (On the grammatica-
lization and subjectivisation of the adverb dou). Journal of Xu-
zhou Normal University, 1, 56-62.
Zhu, Dexi 1982. Yufa Jiangyi (An outline of Chinese grammar). Bei-
jing: Commercial Press Ltd.
YUTAKA SHINODA
Japanese Particle Na as a Marker of the Speaker’s

Subjective Judgment ‘Here and Now’
1. Introduction
The main purpose of this chapter is to explicate the functions of the

Japanese sentence-final particle na.1 We will make a brief mention
of the function of another important particle ne, too, because the two
particles are closely related.
Dictionaries say that na is a sentence-final particle expressing
exclamation, as seen in (1), and ne is usually defined as a particle of
light exclamation and seeking agreement or confirmation, as in (2). It
is widely accepted that ne is the hearer-directed counterpart of na. Ne
is used only in conversation, while na can be used in monologue and
inner speech as well. Basically, na does not presuppose the existence of
the hearer, even when it appears in conversations. Na is sometimes pro-
nounced as naa with a strong emotion. Naa may be regarded as a variant
of na. The same can be said on the relationship between ne and nee.
(1) Ryuu-chan wa kawaii na(a).
Ryu(a boy’s name)-chan Top. cute SFP
‘Little Ryuu is cute!’
(2) Ryuu-chan wa kawaii ne(e).

Ryu(a boy’s name)-chan Top. cute SFP
‘Little Ryu is cute, isn’t he?’
(-Chan is a kind of honorific title for little children. It expresses intimacy rather
than reverence.)
1 There are two kinds of na. One is basically a monologic particle and used by
both male and female speakers. The other has the same function as ne, which is
a dialogic particle, and is used only by male speakers (See Miyazaki (2002)).
The target of our investigation is the former.
150 Yutaka Shinoda
Although what dictionaries say is not incorrect, both na and ne have

other usages, too. Then what are the characteristics common to all
those usages? What are the exact meanings of these particles? Re-
cently, academic papers on ne have been proliferating; still, there is
no agreement among scholars on its nature. On the other hand, na
has received relatively little attention in the linguistic literature, and
little is known about its character, except that it can be an expression
of exclamation. The precise functions of the two particles remain
something of a mystery.
In the present chapter, I propose that na is a marker of the
speaker’s subjective judgment ‘here and now’. For instance, (1) ex-
presses the speaker’s judgment at the time of the utterance that Ryu
is cute. Ne shares the core meaning with na, though its function ex-
tends significantly beyond the core.
Throughout this chapter, I will use constructed sentences, not
natural utterances collected from actual conversations. From actual
conversations we can collect only acceptable utterances, as is often
pointed out. To grasp the complete picture of the functions of the
particles, it is essential to construct unacceptable sentences as well
as acceptable ones and consider how they are different by using na-
tive speakers’ intuition. In addition, na is a monologic particle which
appears in inner speech. The only means to analyze inner speech is
inner reflection.
Section 2 will survey recent studies on na and ne. Section 3
will examine the functions of na in detail. Section 4 will be a brief
discussion of the function of ne. Section 4 is the conclusion.
2. Previous studies
In this section, we will review some previous studies on ne and na.

Perhaps the most influential theory in recent studies on ne is what
Kitano (1993) calls “the shared knowledge theory” (Kamio (1990),
Masuoka (1991), etc.). Many Japanese textbooks adopt this view,
which states that the speaker uses ne when she assumes that the hearer
Japanese Particle Na 151
has the same knowledge as she has.2 Thus, the speaker of (3) as-
sumes that the hearer also thinks that it is fine today, and she seeks
for his agreement. Similarly, the speaker of (4) assumes that the hearer
is Mr. Yamada, and she seeks confirmation from him.
(3) Kyou wa ii tenki desu ne.

today Top. fine weather Cop.(Polite) SFP.
‘It’s a fine day today, isn’t it?’
(4) Anata wa Yamada-san desu ne?

you Top. Yamada-Mr. Cop.(Polite) SFP.
‘You are Mr. Yamada, aren’t you?’
The shared knowledge theory can explain these ‘agreement-seeking’

(example (3)) and ‘confirmation-seeking’ (example (4)) usages, which
are the most typical usages of ne.
Another important theory is what Kitano (1993) calls “the in-
ternal verification theory” (Hasunuma (1988), Takubo/Kinsui (1997,
2000), Katagiri (1995), etc.). In this view, sentence-final particles
reflect mental computation inside the speaker’s brain. For example,
Takubo and Kinsui (2000) claim that ne is a marker which indicates
that the speaker is in the process of verifying the proposition in ques-
tion. Thus, the speaker of (4), who assumes that the hearer is Mr.
Yamada, shows him by using ne that she is in the process of verifying
the assumption, whereby she invites him to acknowledge it.
The following instance from Takubo and Kinsui (1997) is highly
suggestive:
(5) A: Ima nan-ji desu ka.

now what-time Cop. Q.
‘What time is it now?’
B. (Looking at the dial of her wrist watch)
Eet, san-ji desu ne.
well 3-o’clock Cop. SFP.
‘Well, it’s 3 o’clock.’
2 Throughout this chapter, I will refer to the speaker with he, him and his, and the
hearer, she and her, unless otherwise stated.
152 Yutaka Shinoda
Note that the shared knowledge theory does not explain this example.
Here the speaker (= B) cannot assume that the hearer (= A) has the
same knowledge as she, that is, the knowledge that it is 3 o’clock
now. It is because the hearer has no such knowledge that he asks the
question.
The internal verification theory would argue that the speaker
of (5) is in the verification process of the proposition in question, just
as in (4). The difference is that in example (4), the speaker needs the
hearer’s response to complete her verification procedure, while in
(5), she verifies her assumption by looking at her own wrist watch.
Takubo and Kinsui (1997, 2000) calls the latter type of confirmation
‘self-confirmation’.
Compare (5) with the following:
(6) A: Anata no namae wa?

You Poss. name Top.
‘What’s your name?’
B: ? Yamada desu ne.
Yamada Cop. SFP.
‘It’s Yamada.’
The utterance B is unnatural. According to the internal verification

theory, this is because the speaker, as a matter of course, knows her
own name in advance, and therefore does not need to verify the propo-
sition that her name is Yamada.
Consider similar examples (Kamio (2002)):
(7) A: Kono niku, ikura desu ka?

this meat how much Cop. Q.
‘How much is this meat?’
B: a. (Seeing the price list)
250 en desu ne.
250 yen Cop. SFP.
‘It’s 250 yen.’
b. (Instantly and definitely)
250 en desu (? ne).
Kamio (2002) reports that many informants feel that the answer
250 en desu ne would be unnatural if the speaker uttered it in an
instant and definite way. The speaker of (Bb) knows in advance that
the meat costs 250 yen, and this information is already activated on
her mind. Consequently, she needs no mental computation for veri-
fication.
Although the internal verification theory is very persuasive con-
cerning such examples as (5)-(7), it does not seem to account for
more common examples like (3) above. When we say Ii tenki desu
ne (‘It’s a fine day, isn’t it?’) in a normal situation, it should be obvi-
ous to us that it is fine. We can see the clear sky with our own eyes at
the very moment of the utterance. The proposition is so obvious that
it does not need to be verified. Then why does (3) have ne? (3) would
be less natural without ne.
Takubo and Kinsui (2000) claim that even (3) involves a trivial
mental computation. But how trivial is that? I think that it is as trivial
as a computation needed when we identify our own names. The in-
ternal verification theory cannot differentiate (3) from (6). For the
speaker, both her own name and the weather when and where the
utterance is made are too manifest. It is not necessary for her to verify
them in any sense.
Takubo and Kinsui (2000) state that ne indicates a verification
procedure for some judgment. I suggest that ne does not mark a veri-
fication of a judgment but a judgment itself. (Probably the function
of ne extends way beyond that, however. See Section 4.) Suffice it to
say here that their theory cannot deal with such a typical example as
(3) satisfactorily.
As for na, Takubo and Kinsui (2000) mention it only briefly;
they assume that it has almost the same meaning as ne, except that
na can be used in monologic discourse. This means that na also marks
a verification process.
Now let us move to review some of the previous studies on na.
One of the very important studies is Miyazaki (2002). As we
have seen in Section 1, na is often referred to as an exclamatory
particle. For instance, the speaker of (8) is impressed by the fine
weather. The sentence form itself is not an exclamatory sentence, but
its meaning is close to ‘How fine the weather is today!’.
154 Yutaka Shinoda
(8) Kyou wa ii tenki da na(a).

today Top. fine weather Cop.(Plain) SFP.
‘It’s a fine day today!’
However, there are a large number of sentences where na cannot be

treated as an exclamation marker, such as:
(9) Katyou no kao o mite, “tobokeru tsumori da na,”

section-chief Poss. face Acc. see pretend-not-to-know intend Cop. SFP.
to Shyunsuke wa omot-ta.
Quote (a man’s name) Top. think-Past.
‘Seeing the section chief’s face, Shyunsuke thought, “he intends to pretend not
to know.” (From Miyazaki (2002), modified)
Miyazaki (2002) assumes that na is used when the speaker states

what she has thought, felt, noticed on the spot (= at the time of the
utterance). Miyazaki refers to this characteristic of na as “the nature
of recognition being made on the spot”. In other words, it might be
said that na is a marker of the speaker’s recognition on the spot.
Shyunsuke’s inner speech in (9) is an expression of what he has no-
ticed at the moment. Miyazaki argues that his characterization of na
holds not only for the exclamation usage but for other usages.
I suggest that although Miyazaki’s (2002) discussion takes a
fundamentally right direction, the term ‘recognition’ has too broad a
meaning. In a sense, every utterance is an expression of the speaker’s
recognition. Every time she utters something, she activates some rec-
ognition in her mind and puts it into words. Furthermore, we cannot
automatically add na to every sentence which expresses what the
speaker has noticed on the spot. Before discussing it, however, let us
see what Miyazaki says about ne.
Miyazaki (2002) points out that ne has a usage which directly
takes over the nature of na, giving the following instance:
(10) (The male speaker seeing the hearer’s glass)

Gimuretto ka. Ii ne. Ore mo moraou kana.
gimlet Q. good SFP. I also will have wonder
‘You’re drinking gimlet. That’s nice. I think I’ll have one, too.’
In this example, the speaker simply shows the hearer what he has
just felt. He does not necessarily seek for the hearer’s agreement or
confirmation. Even if he said ii na, instead of ii ne, the fundamental

cognitive meaning would not change. The difference is that while na
does not explicitly presuppose the existence of the hearer, ne does.
In contrast, when the speaker and the hearer share their recog-
nition, ne may be used in the ‘agreement-seeking’ or ‘confirmation-
seeking’ sense. Miyazaki (2002) admits, however, that in the ‘confir-
mation-seeking’ usage, there are cases where ne does not show the
speaker’s recognition on the spot. For instance, it is possible that the
speaker of (4) has assumed well in advance that the hearer is Mr.
Yamada, not at the utterance time. Miyazaki argues that in such ex-
amples as (10), the speaker presents her on-the-spot recognition to
the hearer, while in examples like (4), the speaker asks on the spot
for the hearer’s approval for her recognition.
Now let us go on to another insightful study on na. Nihongo-
kijutubunpou-kenkyuukai (2003) (henceforth, NKK (2003)), which
is a descriptive grammar book of Japanese, distinguishes naa from
na, though it acknowledges that there is a strong connection between
them.
According to NKK (2003), naa expresses exclamation, and na
indicates the speaker’s newly acquired recognition. It is true that there
is some advantage in distinguishing them. Since exclamation usu-
ally involves a strong emotion, it tends to be expressed by naa. Con-
versely, naa expresses only exclamation, that is, meanings other than
exclamation are expressed by na, but not by naa. Nevertheless, I
believe that naa should be regarded as a variant of na. Whether this
particle is pronounced as na or naa depends on whether it accompa-
nies a strong emotion or not.3 The particle does tend to be pronounced
as naa when it expresses exclamation. However, even when it is pro-
nounced as na, it does not necessarily prevent itself from indicating
exclamation (see (1) and (8)). The other meanings are expressed by
na, because they are not related with a strong emotion.
3 The iconic principle of quantity, which states that the more form, the more
meaning, accounts for this distinction. The more prolonged pronunciation the
particle has, the more emotion it involves. That is, naa, compared to na, con-
veys a stronger emotion.
156 Yutaka Shinoda
NKK’s (2003) characterization of na, as opposed to naa, is

similar to, and as vague as, Miyazaki’s (2002). However, NKK goes
on to point out an intriguing fact, giving the following example:
(11) (Finding a wallet on the ground by chance)

? A, saifu ga ochiteru na.
ah wallet Nom. have fallen SFP.
‘Ah, there is a wallet on the ground.’
(More literally: ‘Ah, a wallet had fallen (to and is lying on the ground).’)
This utterance is less natural than the same one without na. NKK
(2003) states that na cannot be used in a situation where the speaker
makes a discovery without a prior recognition. (11) might be accept-
able in the situation that the speaker has expected a wallet to be on
the ground and confirms the expectation. (See subsection 3.3 for fur-
ther discussion.)
Returning to Miyazaki’s (2002) analysis, it would predict that
(11) is natural because this example, if it were acceptable, would be
an expression of what the speaker has recognized on the spot. As
I said before, therefore, Miyazaki’s definition of na as signaling the
speaker’s ‘recognition’ on the spot, is too broad. It cannot rule out
unacceptable sentences like (11).
Lastly, I would like to mention Suzuki (1998). He also makes
an important observation. He states that na basically expresses the
speaker’s subjectivity, and that consequently it cannot be used to
state as a fact what is self-evident, the speaker’s experience, and so
on. In my opinion, this is the reason why (11) is odd. (11) expresses
a mere fact which the speaker has just discovered. This observation
by Suzuki that na marks the speaker’s subjectivity is essential for the
discussion later.
Suzuki (1998) goes on to say that na is used not to express
facts or the speaker’s experiences just as they are, but to state the
speaker’s feelings about them. However, I doubt that “to state the
speaker’s feelings” is the right characterization. Recall that ne might
be used when the speaker takes some time to answer a question (see
(5) and (7)). It seems that na could be used in a similar but monologic
discourse:
(12) (The speaker has woken up before dawn)

Ima nan-ji kana….
now what-time I wonder
(And looking at a clock)
eeto, go-ji da na.
‘I wonder what time it is now….well, it’s 5 o’clock.’
Here the utterance go-ji da na [it’s 5 o’clock] is not an expression of

the speaker’s feelings. Rather, it simply expresses what the speaker
has just seen on the dial of the clock. But if this utterance merely
describes a fact, why is it acceptable?
Now it is time to explore the exact function of na.
3. The function of na
3.1 Classification of the usage of na
In this section, we will discuss the function of na. It will be useful, to

begin with, to summarize the classification of usages of na. They
may be divided into four types. (But see the end of Section 3.2.1 for
another possibility.)
An example of the type which I would like to see first is seen
in (13):
(13) (Seeing the light is on in Taro’s room)

Taroo wa heya ni iru na.
Taro Top. room in be SFP.
‘(I’ve found) Taro is in his room.’
Miyazaki’s (2002) example (9) belongs to the same type. Let me call
this type the ‘recognition’ type, for lack of a better term. I argue that
this type is the most fundamental. See subsection 3.2.1 for discussion.
A second type is the ‘exclamation’ type. A typical example
is (8), which is cited again below as (14):
158 Yutaka Shinoda
(14) Kyou wa ii tenki da na.

today Top. fine weather Cop.(Plain) SFP.
‘It’s a fine day today!’
This type may be the most salient in that dictionaries define na as

expressing exclamation above all. In fact, however, na is frequently
used in the first sense as well. Some examples are ambiguous be-
tween these two senses. The distinction is not necessarily clear.
A third type is the ‘confirmation’ type, whose example is as
follows:
(15) (When the speaker makes sure that the windows are locked before going out)
Yoshi, shimatteru na.
OK is locked SFP.
‘OK, (the windows) are locked.’
In this type of examples, the speaker has some prior expectation. The
speaker of (15) has expected that the windows are locked, and is
confirming it. The sentence would sound bizarre if it were uttered
out of the blue, without any prior expectation.
An example of the last type is (12), which we observed at the
end of the last section. Let me call this type the ‘self-confirmation’
type, following Takubo and Kinsui (1997, 2000), to whose theory
I do not necessarily commit myself, however. This kind of na could
be regarded as a marker which indicates that the speaker has found
the answer to a question or a problem. (Actually, this type may be
dissolved into other types. See Section 3.2.1)
Now that we have reviewed the classification of na, we will
consider what is the feature common to all the four types.
3.2 What is na?
My hypothesis is that na is a marker of the speaker’s subjective judg-

ment ‘here and now’. The following subsections will be concerned
with how it is so.
3.2.1 A marker of a judgment
Na marks a judgment. Let us consider the four types of na one by

one, investigating how they express judgments.
First, a sentence with the ‘recognition’ type of na directly ex-
presses a judgment. Compare the following pair:
(16) a. Taroo wa heya ni iru.

b. Taroo wa heya ni iru na. (= (14))
(16a) may describe a fact, but (16b) can be interpreted only as judg-
mental. The latter means that ‘Because the light is on in Taro’s room,
I judge that he is in his room.’
Next, let us examine the ‘exclamation’ type. Consider (14)
again. The sentence may seem to describe a mere fact. Note, how-
ever, that the speaker makes an evaluation of the weather, and that
evaluation is a kind of judgment. The speaker judges whether the
weather is fine or not, and how fine it is. When evaluation provokes
a relatively strong emotion, it is expressed as exclamation.
Compare the following three sentences:
(17) a. Kabe ga kirei da na(a).

wall Nom. clean Cop. SFP.
‘The wall is clean!’
b. Kabe ga shiroi na(a).
wall Nom. white SFP.
‘The wall is white!’
c. ? Kabe ga shiro-kuro da na(a).
wall Nom. white-black Cop. SFP.
‘The wall is black and white!’
(17a) is an ordinary exclamation. The speaker judges that the wall is

very clean. (Other persons might judge otherwise.)
(17b) will be unnatural if it is intended to mean that the wall is
not blue or red but white. White is white, and every person with
normal color vision can recognize it. In that sense, the sentence ex-
presses a mere fact, not an evaluation. In another sense, however,
whiteness is a gradable concept. A white wall may be gravely white,
160 Yutaka Shinoda
brightly white, or even too white. On this reading, the speaker of (17b)
judges that the wall is very white, and feels a rather strong emotion.
She might be surprised, impressed, or disappointed at the whiteness
of the wall.
In contrast, shiro-kuro [black and white] is not a gradable ad-
jective, so (17c) is interpreted as a mere fact, not an evaluation. Hence,
the utterance sounds odd. Na, which is a judgment marker, cannot
occur with sentences which describe facts.
Here some comments may be needed on the distinction be-
tween the ‘exclamation’ type and the ‘recognition’ type. As far as
evaluation is a kind of judgment, every ‘exclamation’ type of na-
sentence is of the ‘recognition’ type at the same time. The former
may be regarded as a special type of the latter. In this sense, the
‘recognition’ type is the most basic of all the usages of na; it is a
direct realization of the function of na. When the speaker’s judgment
accompanies a strong emotion, it is realized as exclamation. In fact,
contrary to the popular view that na expresses exclamation above
all, in actual conversations we find as many examples with the ‘rec-
ognition’ na as those with the exclamatory na, and there are a large
number of ambiguous (or rather vague) examples, too.
Incidentally, what exactly is the difference between na and naa?
(18) a. Ii tenki da na.

b. Ii tenki da naa.
(18a) can be interpreted as either an ‘exclamation’ na-sentence or a

‘recognition’ na-sentence. By contrast, (18b) has only the exclamatory
meaning. It is because naa usually indicates a very strong emotion, as
I stated before, I suppose that naa indicates that the speaker takes
some time to assimilate her judgment emotionally. The speaker of
(18b) has judged that the weather is fine, and has accepted it cognitively.
But she is so moved at the fine weather that she has not fully assimi-
lated it emotionally. Compared with naa, na signals that the speaker
takes less time to accept her judgment emotionally. In any case, how-
ever, the fundamental cognitive meanings of na and naa are the same.
Now let us move to the ‘confirmation’ type of na. Consider
again (15), which is repeated as (19) below:
(19) (When the speaker makes sure that the windows are locked before going out)
Yoshi, shimatteru na.
OK is locked SFP.
‘OK, (the windows) are locked.’
The speaker, who has expected that the windows are locked, is con-
firming the expectation, uttering this sentence. Notice that confirma-
tion is a kind of judgment, just as evaluation is. When we confirm
something, we judge whether the expectation has been met or real-
ized, or not.
Lastly, we will consider the ‘self-confirmation’ type. See the
following example (= (12)):
(20) (The speaker has woken up before dawn)

Ima nan-ji kana….
(And seeing a clock)
eeto, go-ji da na.
The speaker of (20) asks herself what time it is now, and needless to
say, does not know the answer in advance. She has to find the an-
swer. The second utterance shows that she has found it. How can it
be judgmental? I would claim that there are two cases where this
utterance becomes judgmental.
One case is as follows: when the speaker woke up, she thought
perhaps vaguely that it was approximately 5 o’clock then, and now
by looking at the clock she has judged that her guess is correct or at
least not incorrect. Na marks this judgment. In this case, therefore,
the utterance is actually of the ‘confirmation’ type.
The alternative possibility is that the utterance involves an im-
plicit judgment. That is to say, when the speaker knew the time, she
implicitly judged, for instance, that because it was already 5 o’clock,
she had better get up then, or that because it was only 5 o’clock, she
still could sleep for one more hour. In short, the utterance conveys a
judgment as an implicature, as it were, though it is not made in a con-
versation. In this second case, the utterance might be considered to be
a special kind of the ‘recognition’ na-sentence. (See Subsection 3.3)
162 Yutaka Shinoda
If the actual answer is far from her speculation (in the first
case above) or the speaker has no particular implicit judgment in
mind (in the second case), she will use ka rather than na, as in (21).
Although ka is usually a question marker pronounced with a rising
intonation, in (21) it is pronounced with a falling intonation and func-
tions as a marker which signals that the speaker has accepted a piece
of new information.
(21) (The speaker has woken up in the middle of the night)

Ima nan-ji kana….
(And looking at a clock)
eeto, san-ji ka.
well 3-o’clock SFP.
In sum, the ‘self-confirmation’ type of na may be reanalyzed as ei-

ther the ‘recognition’ or ‘confirmation’ type.4 It means that there are
three, not four, types of na in all. In any case, they all mark judg-
ments.
3.2.2 A marker of the speaker’s judgment
Here we will observe that the judgment indicated by na is the speaker’s

judgment, not someone else’s.
(22) a. Watashi no yosoo dewa, ashita wa hare da.

I Poss. prediction on tomorrow Top. fine Cop.
‘On my prediction, it will be fine tomorrow.’
b. Watashi no yosoo dewa, ashita wa hare da na.
(23) a. Tenki yohoo dewa, ashita wa hare da.

weather forecast according to tomorrow Top. fine Cop.
‘According to the weather forecast, it will be fine tomorrow.’
b. ? Tenki yohoo dewa, ashita wa hare da na.
4 My intuition is not clear as to whether the same can be said on the correspond-
ing ‘self-confirmation’ type of ne.
Both (22a) and (23a) are acceptable. On the other hand, though (22b)
is perfect, (23b) is problematic. Why? The reason is that whereas na
is a marker of the speaker’s judgment, tenki yohoo dewa [according
to the weather forecast] shows that the judgment in question is at-
tributed to the weather forecast. This contradiction makes the utter-
ance unacceptable. It may be possible to say that na functions as a
kind of evidential.
3.2.3 A marker of the speaker’s subjective judgment
Next, we will see that the judgment of na is a subjective judgment, as

opposed to an objective one.
(24) a. Zoo wa ookii.

elephant Top. large
‘Elephants are large.’
b. Zoo wa ookii na.
(25) a. Zoo wa nezumi yori ookii.

elephant Top. mouse than large
‘Elephants are larger than mice.’
b. ? Zoo wa nezumi yori ookii na.
These sentences show the same pattern as seen in (22) and (23). Why
is (25b) unnatural? It is because (24a) is a subjective judgment while
(25a) is an objective judgment in that everyone will acknowledge it
as true. The subjectivity of na and the objective nature of (25a) are
incompatible with each other. Hence, (25b), which is (25a) plus na,
is unacceptable.
3.2.4 A marker of the speaker’s subjective judgment ‘here and now’
Finally, let us examine whether na expresses the judgment ‘here and

now’. As stated in Section 2, Miyazaki (2002) argues that na-sen-
tences express the recognition on the spot. The term ‘here and now’
is intended to mean the same as ‘on the spot’. Now consider the
following sentences:
164 Yutaka Shinoda
(26) a. Watashi wa Tomu ga supai da to omou. (Nakau (1994))

I Top. Tom Nom. spy Cop. Quote think
‘I (now) think that Tom is a spy.’
b. Watashi wa Tomu ga supai da to omou na.
(27) a. Watashi wa Tomu ga supai da to omotteiu. (Nakau (1994))

I Top. Tom Nom. spy Cop. Quote think
‘I (always) think that Tom is a spy.’
b. ?Watashi wa Tomu ga supai da to omotteiu na.
As Nakau (1994) points out, the English sentence I think that Tom is
a spy is ambiguous because it offers two interpretations. In Japanese,
the two meanings are represented by distinct expressions, that is,
(26a) and (27a). (26a) has the instantaneous present meaning; it ex-
presses the speaker’s judgment at the utterance time. (27a) has the
continuous (or ‘unrestrictive’ in Leech’s (1987) term) present mean-
ing; it means that the speaker has had the idea for some period of time.
The meaning of the latter is not in harmony with the function of na as
a marker of the judgment ‘here and now’. Hence, (27b) is unnatural.5
3.3 (Un)acceptability of na-sentences
In this subsection, to understand the function of na well, we will

discuss in what way various sentences can be interpreted when na
occurs with them.
Many declarative sentences could be interpreted at least in two
ways: describing a fact or the truth and expressing the speaker’s judg-
ment. For instance, the following sentence may be taken as a de-
scription of the fact or an expression of the speaker’s opinion:
(28) Taroo wa Hanako o aishiteiru.

Taro Top. Hanako Acc. love
‘Taro loves Hanako.’
5 (27a) and (27b) have no difference in cognitive meaning. Both express the
speaker’s judgment at the utterance time. The difference is that whereas (27a)
can be used in either spoken language or written language, (28b) is used only
in speech.
However, the same sentence, if accompanied by na, admits of only

one interpreation; it expresses the speaker’s judgment.
(29) Taroo wa Hanako o aishiteiru na.

Taro Top. Hanako Acc. love SFP.
‘(Judging from some evidence, I have just concluded that) Taro loves Hanako.’
Some sentences are difficult or impossible to have both interpretations.
(30) a. Taroo wa kaetta kamoshirenai.

Taro Top. went-back may
‘Taro may have gone back.’
b. Taroo wa kaetta kamoshirenai na.
(31) a. Ichi tasu ni wa san da.

1 plus 2 Top. 3 Cop.
‘1 and 2 make 3.’
b. ? Ichi tasu ni wa san da na.
(30a) has only a judgmental meaning because the modal expression

‘kamoshirenai’ represents the speaker’s judgment on the proposition.
Thus, na can be added to (30a), to form (30b). The difference be-
tween (30a) and (30b) is that while the latter expresses the speaker’s
judgment ‘here and now’, the judgment in the former is not necessar-
ily made at the utterance time.
On the other hand, (31a) has only a factual meaning. It is not a
judgment by the speaker but a necessarily true proposition. Na is
incompatible with (31a). Hence, the unnaturalness of (31b).
The sentence (32a) below would normally be taken as describ-
ing a fact. The speaker is describing what she has seen without mak-
ing any subjective judgment. Even if na was attached to (32a), there-
fore, the ‘exclamation’ and the ‘recognition’ readings would be hard
to come by.
(32) (Seeing a train approaching the platform)

a. Densya ga kita.
train Nom. have come
‘A/the train has come.’
b. Densya ga kita na.
166 Yutaka Shinoda
The only possibility of construing (32b) is the ‘confirmation’ reading.

That is, the speaker has a certain train to take, and has been waiting
for it to come. Then she has found her train approaching the platform.
(32b) indicates her confirmation that her train has come as expected.
Recall here example (11), cited again as (33) below:

? A, saifu ga ochiteru na.
‘Ah, there is a wallet on the ground.’
This utterance sounds unnatural because it is very difficult to inter-

pret it as a judgmental sentence in an ordinary situation.
On the other hand, the following sentence has at least three
kinds of judgmental interpretations:
(34) Kono kikai wa anzen da na.

this machine Top. safe Cop. SFP.
‘This machine is safe.’
One is the ‘recognition’ reading. It means that ‘Judging from some

evidence, this machine must be safe.’ Another is the ‘exclamation’
reading, which means that the speaker is impressed how safe this
machine is. And the last is the ‘confirmation’ reading. On this read-
ing, the speaker has a prior expectation that this machine should be
safe, and confirms that it really is.
3.4 Implicit judgments
Lastly but not least importantly, na-sentences may not express judg-
ments explicitly but involve them implicitly. (Recall the discussion
in 3.2.1)
To illustrate, consider the next sentences:
(35) a. Kamome ga tonderu.

sea gull Nom. be flying
‘Sea gulls are flying.’
b. Kamome ga tonderu na.
Imagine that you are driving with no purpose. Suddenly you see sea
gulls flying. Then you are likely to say (35a) but not (35b). In this
situation, (35b) would be unnatural because neither the ‘exclama-
tion’ nor ‘recognition’ reading is easy to obtain. Moreover, since you
have no prior expectation about sea gulls, the ‘confirmation’ reading
is not possible, either.
However, this does not mean that the ‘exclamation’ and ‘rec-
ognition’ readings are impossible. They will become possible if the
fact that sea gulls are flying leads you to make another judgment. For
example, by seeing sea gulls, you may judge that you are approach-
ing the sea. In that case, you can say (35b) (the ‘recognition’ reading)
with the implicit meaning ‘I am approaching the sea’. Or suppose
that you happen to dislike noisy sea gulls. Then you might say (35b)
(the ‘exclamation’ reading), with the intended meaning ‘How noisy
sea gulls are! I hate them.’ Or if you are driving in the mountains,
you might say the sentence, meaning ‘How strange! There can’t be
sea gulls around here.’ (the ‘recognition’ and ‘exclamation’ readings).
There will be other possible candidates for the interpretation of (35b).
The above discussion shows that na-sentences may have impli-
cit judgments with them. In the literal sense they may describe mere
facts, not judgments, but they can be accompanied by implicit judg-
ments, which might be communicated as implicatures in conversations.
4. The function of ne
To investigate the function of ne is not the central purpose of the

present study. But since ne is closely related with na, we will take a
look at several problems on the characterization of ne.
In some examples, as Miyzaki (2002) points out, ne seems to
directly inherit the nature of na, as seen in (36) (= (10)):
(36) (The male speaker seeing the hearer’s glass)

Gimuretto ka. Ii ne. Ore mo moraou kana.
gimlet Q. good SFP. I also will-have wonder
‘You’re drinking gimlet. That’s nice. I think I’ll have one, too.’
168 Yutaka Shinoda
Here the speaker utters simply what he has thought, a judgment about
a glass of gimlet. He obviously directs his words to the hearer; nev-
ertheless, he does not necessarily seek the hearer’s response. Even if
he had used na instead of ne, there would have been little difference
in meaning. In examples like this, therefore, we may say that ne, as
well as na, expresses the speaker’s subjective judgment ‘here and
now’. And that can be regarded as the core function of ne.6
So far, so good. But in fact, there are various examples where
ne diverges from this core meaning.
To begin with, as Miyazaki (2002) goes on to argue, ne does
not show what he calls “the nature of the recognition being made on
the spot”, in its original form, when it is used in the ‘confirmation-
seeking’ sense (and perhaps in the ‘agreement-seeking’ sense, too).
It follows that in the case of ne, the speaker’s judgment is not always
made ‘here and now’.
Secondly and more importantly, in the ‘agreement-seeking’ use
and the ‘confirmation-seeking’ use, the speaker assumes that the
hearer shares the same judgment with her. For instance, consider
(37) (= (3)) again:
(37) Kyou wa ii tenki desu ne.

today Top. fine weather Cop.(Polite) SFP.
‘It’s a fine day today, isn’t it?’
Here both the speaker and the hearer see the same sky. The speaker
sees the clear sky and judges that the weather is very fine. She knows
that the hearer sees the same sky, and assumes that he judges the
same. Ne indicates that the speaker assumes that the hearer’s subjec-
tive judgment is the same as her subjective judgment. To put it an-
other way, the speaker’s judgment is not merely subjective, but rather
intersubjective. Hence, ne is not a marker of the speaker’s subjective
judgment; it can express the speaker’s intersubjective judgment.
6 From this it follows that at its core ne has at least three types of usages, just like
na: the ‘recognition’, ‘exclamation’, and ‘confirmation’ types.
However, ne may not be even a marker of judgment. Compare

(33) with the following:
A, saifu ga ochiteru ne.
‘Ah, there is a wallet on the ground, isn’t there?.’
Here the speaker has found a wallet on the ground by chance and
talks to the hearer, assuming that he has also recognized the wallet at
the same time. The only difference between (33) and (38) is whether
the sentence-final particle used is na or ne. Nevertheless, it seems
that (38) is more natural than (33). If, as was discussed earlier, the
unacceptability of (33) can be attributed to the fact that the sentence
describes a mere fact whereas na is a judgment marker, then the only
explanation for the difference between these two sentences is that
unlike na, ne is not a judgment marker.
In addition, ne can be used with more kinds of sentence than
na. Na occurs with statements only. Since na does not presuppose
the hearer, naturally it is not used with utterances which require the
interaction with the hearer. On the other hand, ne can occur with
offers, requests, and even polite orders. But can we say that offers,
requests and orders are judgments? If not, it means again that what
ne indicates is not judgment. One might claim that judgments under-
lie these kinds of utterances. For instance, the request Please come
here presupposes the speaker’s judgment that the hearer should come.
But this is not the whole story.
Ne can be attached to not only sentences but phrases and words;
that is, it can be used sentence-internally:
(39) Ano ne, Taroo ga ne, issyoni iki tai tte.
filler SIP Taro Nom. SIP together go want Quote
‘Uh, Taro says he wants to go with us.’
In this case, ne functions as a sentence-internal particle with no obvi-

ous meaning. Perhaps it signals ‘affective common ground’ between
the speaker and the hearer, as Cook (1990) argues. It may serve to
establish a cooperative relationship between conversation participants.
Moreover, ne can be used independently.
170 Yutaka Shinoda
(40) Ne(e), kiiteru no?

Interjection be listening Comp. or SFP (Question)
‘Hey, are you listening?’
This ne is an interjection. It functions as an attention-getter.

In (39) and (40), ne is not attached to a proposition. Thus it
cannot possibly be a judgment marker.
Then what is the general function common to all the examples
above? The fundamental difference between na and ne is that ne is
hearer-directed but na is not. In fact, as is obvious from the above
discussion, all the extensions of the function of ne is related with the
hearer-directedness. The existence of the hearer is essential to char-
acterize the nature of ne.
Recently, Moriya (2006) has discussed the sentence-final ne
in terms of joint attention (see Tomasello (1999), Honda (2002),
Ooyabu (2004), for example). Moriya argues that by using ne the
speaker urges the hearer to jointly attend to something, by which a
joint attentional scene (Tomasello 1999) is constructed. Although
Moriya considers only the sentence-final ne, I would suggest that the
same line of argument might be promising for the sentence-internal
or interjectional ne, too. For instance, the sentence-internal ne may
get the hearer to jointly attend to the very utterances that the speaker
is making, rather than their referents. But here we will make no fur-
ther inquiry into this problem.7
7 In Shinoda (2005), I argued that since the speaker’s use of ne tends to invite the
hearer’s backchannels, there is a significant correlation between ne and
backchannels. I proposed that ne has interpersonal functions corresponding to
the functions of backchannels. Brunner (1979: 733) suggests that backchannels
give the speaker feedback on the following three levels:
a. At the first level, backchannels signal the auditor’s involvement and par-
ticipation in the interaction. They indicate that the auditor is attending to
what the speaker says and that a conversation, not a monologue, is occur-
ring.
b. At the second level, backchannels provide information about the auditor’s
level of understanding, allowing the speaker to adjust his or her communi-
cative endeavor so as to get the ideas across efficiently.
c. At the third level, backchannels can signal the auditor’s personal response
to what the speaker has just said. This might mean agreement or disagree-
ment, shock, amusement, scorn, or any number of other reactions.
5. Conclusion
The present study has been concerned with the function of the Japa-
nese sentence-final particle na. Na functions as a marker of the
speaker’s subjective judgment ‘here and now’. The speaker’s judg-
ments expressed by na-sentences may be explicit or implicit.
Na can be used in monolog and inner speech, unlike its closely
related particle ne, which presupposes the hearer. Although ne, being
hearer-directed, extends its function in various ways, it seems to in-
herit its core meaning from na. In that sense, explicating the function
of na will contribute to exploring the function of ne. Uncovering
how ne develops various functions from its core meaning will be left
to future research.
In the introduction, I stated that an analysis of na would re-
quire native speakers’ inner reflection. This does not mean, however,
that it is unimportant to observe actual conversations. Rather, the
present study should be supplemented with research from conversa-
tion analysis or corpus linguistics. Further studies will be necessary
to depict a comprehensive grammar of na.
Brunner (1979) says that “In general, if an action functions on a higher level it
also functions on the ones below it.”
In a similar way, I proposed, ne has the following three-level functions:
a. At the first level, ne signals that the speaker is attending to the existence of
the hearer, urging him to attend to what she says and to participate in the
conversation.
b. At the second level, ne urges the hearer to understand what the speaker
says.
c. At the third level, ne urges the hearer to give his responses, such as agree-
ment and confirmation, to what the speaker says.
Just as in the case of backchannels, if an utterance of ne functions on a higher
level, it also functions on the levels lower than it.
This characterization of ne focuses on the speaker’s interaction with the hearer,
so it takes the same direction with analyses in terms of joint attention.
172 Yutaka Shinoda
References
Aoki, Haruo 1986. Evidentials in Japanese. In Chafe, W. / Nichols, J.

(eds) Evidentiality: The Linguistic Coding of Epistemology.
Norwood, NJ: Ablex, 223-238.
Brunner, J. Lawrence 1979. Smiles can be back channels. Journal of
Personality and Social Psychology, 37/5, 728-734.
Cook, Haruko Minegishi 1990. The sentence-final particle ne as a
tool for cooperation in Japanese conversation. Japanese/Ko-
rean Linguistics 1, 29-44.
Dirven, René / Verspoor, Marjolijn 1998. Cognitive Exploration of
Language and Linguistics. Amsterdam/Philadelphia: John
Benjamins Publishing Company.
Hasunuma, Akiko 1988. Zoku, nihongo wanpointo ressun (Japanese
one point lesson, second). Gengo 17/6, 94-95.
Honda, Akira 2002. Kyoudouchuui no tougoron (A joint-attentional
approach to Japanese syntax). In Yamanashi, M. et al. (eds)
Ninchigengogaku Ronkou (Studies in Cognitive Linguistics)
2, 199-229. Tokyo: Hituzi Shobou.
Ikegami, Yoshihiko 2003. Gengo ni okeru syukansei to syukansei no
gengoteki shihyou (Locutionary subjectivity and its linguistic
indices) (1). In Yamanashi, M. et al. (eds) Ninchigengogaku
Ronkou (Studies in Cognitive Linguistics) 3, 1-45. Tokyo:
Hituzi Shobou.
Ikegami, Yoshihiko 2004. Gengo ni okeru syukansei to syukansei no
gengoteki shihyou (Locutionary subjectivity and its linguistic
indices) (2). In Yamanashi, M. et al. (eds) Ninchigengogaku
Ronkou (Studies in Cognitive Linguistics) 4, 1-60. Tokyo:
Hituzi Shobou.
Izuhara, Hideko. Shyuujoshi na (naa) no ichikousatsu: kikite ni nani
o tsutaeteiruno ka (An observation of na (naa): what it com-
municates to the addressee).
Kamio, Akio 1990. Jouhou no Nawabari Riron (The Theory of Ter-
ritory of Information). Tokyo: Taisyuukan Shoten.
Kamio, Akio 1997. Evidentiality and some discourse characteristics
in Japanese. In Kamio, Akio (ed) Directions in Functional Lin-
guistics. Amsterdam/Philadelphia: John Benjamins Publishing

Company.
Kamio, Akio 2002. Zoku, Jouhou no Nawabari Riron (The Theory of
Territory of Information, Second). Tokyo: Taisyuukan Shoten.
Katagiri, Yasuhiro 1995. Shuujoshi niyoru taiwa choosei. Gengo 24/
11, 38-45.
Kitano, Hiroaki 1993. Nihongo no syuujyoshi ne no motsu kihon-
tekina kinou nituite (On the basic function of the Japanese sen-
tence-final particle ne). Gengogaku Kenkyuu (Linguistic Stud-
ies) 12, 73-88. Linguistic Circle at Tokyo University.
Langacker, Ronald W. 2002. The Cognitive Basis of Grammar. Ber-
lin / New York: Mouton de Gruyter.
Leech, Geoffrey N. 19872. Meaning and the English Verb. London/
New York: Longman.
Masuoka, Takashi 1991. Modaritii no Bunpou (Grammar of Modal-
ity). Tokyo: Kurosio Publishers.
Miyazaki, Kazuhito 2002. Shyuujoshi ne to na (A study of the sentence-
final particles ne and na). Handai Nihongo Kenkyuu (Osaka
University Japanese Language Studies) 14, 1-19. Also in K. Mi-
yazaki. 2005. Gendai Nihongo no Gimon Hyougen: Utagai to
Kakunin Youkyuu (Interrogative Expressions in Modern Japa-
nese: Doubt and Confirmation Seeking). Tokyo: Hitsuji Shobou.
Miyazaki, Kazuhiko et al. 2002. Modaritii (Modality) [Shin Nihongo-
bunpou Sensyo 4]. Tokyo: Kurosio Publishers.
Morita, Yoshiyuki 1973. Kandoushi no hensen (The change of inter-
jections). In K. Suzuki / O. Hayashi (eds) Setsuzokushi-Kan-
doushi (Conjunctions-Interjections) [Hinshibetsu Nihonbunpou
Kouza 6], 177-208. Tokyo: Meiji Shoin.
Moriya, Michiyo 2006. Kyoodoo-chuui to syuujoshi-shiyoo (joint atten-
tion and the use of sentence-final particles). Gengo 35/5, 62-67.
Moriyama, Takurou 1997. “Hitorigoto” o megutte: sikou no gengo
to dentatsu no gengo (On “soliloquy”: language of thought and
language of communication). In Y. Kawabata / Y. Nitta (eds)
Nihongobunpou: Taikei to Houhou (Japanese Grammar: Sys-
tem and Method), 173-188. Tokyo: Hituzi Shobou.
Nakau, Minoru 1994. Ninchiimiron no Genri (Principles of Cogni-
tive Semantics). Tokyo: Taisyuukan Shoten.
174 Yutaka Shinoda
Nihongo-kijutubunpou-kenkyuukai (Japanese Descriptive Grammar),

ed. 2003. Gendai Nihongo Bunpou (Modern Japanese Gram-
mar) 4. Tokyo: Kurosio Publishers.
Nitta, Yoshio 1991. Nohongo no Modaritii to Ninsyou (Modality and
Person in Japanese). Tokyo: Hituzi Shobou.
Onodera, Noriko O. 2004. Japanese Discourse Markers. Amsterdam:
John Benjamins Publishing Company.
Ooyabu, Yasushi 2004. Kyoodoo-chuui (Joint Attention). Tokyo: Ka-
washima Shoten.
Shinoda, Yutaka 2005. Nihongo no syuujoshi ne no taijinteki kinoo
to aizuchi (The interpersonal function of the Japanese particle
ne and backchannels. Hikakubunka Kenkyuujo Nenpoo (An-
nual Report of Research Institute for Comparative Culture)
21, 7-21. Tokushima Bunri University.
Suzuki, Hideo 1998. Gendainihongo no syuujyoshi: na o tyuushin
toshite (Sentence-final particles in modern Japanese: focusing
on na). In Kokugokenkyuuronsyuu Hensyuuiinkai, ed. Toukyou-
daigaku Kokugokenkyuushitsu Sousetsuhyakusyuunenkinen,
Kokugokenkyuuronsyuu (100th Anniversary of National Lan-
guage Laboratory at Tokyo University, Chapters of National
Language Studies), 961-981. Tokyo: Kyuuko Shoin.
Takubo, Yukinori / Kinsui, Satoshi 1996. Taiwa to kyooyuuchisiki:
danwakanri-riron no tachiba kara (Conversation and mutual
knowledge: from a viewpoint of a theory of discourse man-
agement). Gengo 25/1, 30-39.
Takubo, Yukinori / Kinsui, Satoshi 1997. Discourse management in
terms of mental spaces. Journal of Pragmatics 28/6, 741-758.
Takubo, Yukinori / Kinsui, Satoshi 2000. Fukusuu no shinteki ryouiki
niyoru danwakanri (Discourse management in terms of mul-
tiple mental spaces). In S. Sakahara (ed.) Ninchigengogaku no
Hatten (Advances in Cognitive Linguistics). Tokyo: Hituzi Sho-
boo, 251-280.
Tomasello, Michael 1999. The Cognitive Origins of Human Cogni-
tion. Cambridge, Massachusetts: Harvard University Press.
Traugott, Elizabeth Closs / Dasher, Richard B. 2005. Regularity in
Semantic Change. Cambridge: Cambridge University Press.
175
LEUNG WAI-MUN
Evidential Particles in Cantonese:

the case of wo3 and wo5
1. Introduction
When people communicate with each other, they often provide the
source of information and express their own standpoint and attitude.
Evidentiality thus refers to how people indicate the source of know-
ledge, attitude towards knowledge and commitment to knowledge.
Evidentiality is a very important linguistic phenomenon, which is
not only concerned with the source of information, whether directly
seen or heard, indirectly inferred, or obtained from other people, but
also reflects how much the speaker is willing to be responsible for
the information he or she is providing, which is related to whether
the information is reliable or not.
Evidentiality generally appears in every human language as a
linguistic phenomenon (Plungian 2001), but it is expressed by differ-
ent grammatical structures. The grammatical forms of evidentiality
are called ‘evidentials’ or ‘evidential markers’, which include two
types: one is inflectional in terms of prefixes or suffixes, and the
other is lexical, involving modal verbs, adverbs and adjectives etc.
Evidentiality basically adheres to sentence propositions, that is, when
the evidentials or evidential markers in a sentence are removed, the
sentence proposition is not affected.
It is rare that people speak completely objectively without sub-
jective feelings. Whatever degree of certainty that the speaker has
about the credibility of the source of information, it is inevitable that
subjective expressions are involved, conveying the speaker’s stand-
points, emotions and attitudes (Shen 2001). These expressions are
referred to as ‘subjectivity’, a concept which is opposite to objectiv-
176 Leung Wai-mun
ity, and can be said as the speaker’s ‘self-impression’ (Lyons 1982,

Finegan 1995, Shen 2001). In this regard, evidentiality definitely
carries subjective expressions, because the speaker reveals to a greater
or lesser extent what he or she thinks, whether positively or nega-
tively and to what degree, about the situation. Evidentiality and sub-
jectivity are thus two important features of human language which
cannot be overlooked.
The study of evidentiality, which has become an indispensable
part of linguistic studies, has had a rapid development in the past few
decades. On the contrary, studies of evidentiality in Cantonese are
relatively few. The following sections will firstly introduce evidentiality
and its derived concept, mirativity (pragmatic extension of evidentia-
lity), and subjectivity. Then the features of the Cantonese evidential
particles wo3 (mid-level tone), which indicates unexpectedness and
noteworthiness (Luke 1990), and wo5 (low rising tone), which ex-
presses hearsay information (Matthews 1998), will be analyzed, and
a discussion on how a speaker expresses his or her understanding of
the objective world through language will be given.
2. Previous studies and data
2.1 Previous work on evidentiality
2.1.1 Evidentiality
Jakobson (1957) was among the earliest scholars to propose the con-
cept of evidentiality. He was also the first to separate evidentiality
from mood and modality clearly and categorized evidence into four
types: quotative evidence, revelative evidence, presumptive evidence
and memory evidence. Based on Jakobson’s observations, Chafe and
Nichols (1986) further divided evidentiality into five kinds: degree
of reliability, belief, inference, hearsay and general expectation. Their
treating of evidentiality as a separate area of study made a great im-
pact on the research that followed. Aikhenvald and Dixon (2003)
Evidential Particles in Cantonese: the case of wo3 and wo5 177
analyzed evidentiality as the grammatical reference to the informa-

tion source, and, modifying the definition given by Chafe and Nichols
(1986), considered that in a narrow sense only belief and hearsay
should be counted as belonging to evidentiality. Moreover, the
Routledge Dictionary of Language and Linguistics (Bussmann 1996)
has ‘evidentiality’ as a separate entry. Semanticist Saeed (1997) in-
cluded ‘evidentiality’ as an independent term, pointing out that
evidentiality is ‘a term for the ways in which a speaker qualifies a
statement by referring to the source of the information’. It can be
seen that evidentiality has become a research topic in its own right.
In China, Hu (1994, 1995) was among the first Chinese lin-
guists to introduce the concept of evidentiality to Chinese scholars.
Zhang (1997) systematically introduced evidentiality in western lin-
guistics and pointed out that in Chinese evidentiality comes in three
forms: 1. parentheses, mostly for showing the source of information,
2. adverbs, mostly for showing the speaker’s attitude towards the
credibility of the information, 3. sentence-final particles, for show-
ing the speaker’s belief in the reliability of the information. Specific
studies on evidentiality in Chinese include Li et al. (1998), who in-
vestigated into the sentence-final particle de and concluded that it is
an evidential marker which shows subjective attitude of recognition.
Zhang (2000) discussed the evidentiality and modality of Chinese
adverbs. Furthermore, Matthews (1998) discussed the different fea-
tures and relations of the Cantonese particles wo3, wo4 and wo5.
Other studies on evidentiality in Chinese include Zhang and Yu (2003)
and Zhu (2006), which shed the light on the importance of the re-
search on the topic.
2.1.2 SFPs wo3 and wo5
As a sentence-final particle (hereafter SFP) indicating moods and

attitudes, wo3 is used very frequently in modern Cantonese. The ear-
liest study mentioning the SFP wo3 is Ball (1888). In the syntax sec-
tion of the book, there is a table called ‘List of Finals, and their Tonal
Variants’. It defines wo according to three tones: tone 3 (mid-level
tone), tone 4 (low-falling), tone 6 (low departing) (1888: 114). The
low departing one does not exist in modern Cantonese. Instead we
178 Leung Wai-mun
have wo in tone 5 (the low-rising tone) which is not mentioned in

Ball’s book. Ball points out that the function of wo3 is ‘denoting that
the statement preceding it has been made by some one before’. Ever
since Ball’s bringing up of wo3, the archives of the past one hundred
years including the dialectal works of missionaries, Yünshu (rhyme
books) of dialects, Difangzhi (local records), Cantonese teaching
materials and other related academic books and articles etc have oc-
casionally mentioned the SFP wo3, but the explanations are often
too short. Luke (1990) has an entire chapter devoted to the detailed
discussion of wo3. He takes a discourse perspective in his study and
adopts a discourse model in his analysis of SFPs, using the analytical
tools and research procedures of Conversation Analysis. He has con-
cluded the meaning of wo3 as ‘unexpectedness, noteworthiness, re-
markableness, unusualness…etc’. After Luke has noted the new sta-
tus of wo3, there have been scholars noticing and discussing this
SFP continuously in the past decade (Ouyang 1990, Leung 1992,
Matthews/Yip 1994). They point out that the meaning of wo3 is ‘re-
minding, discussing and enlightening’.
As regards SFP wo5, it is commonly known as a hearsay par-
ticle in modern Cantonese. This particle is not found in Ball’s books
(1888, 1924). A good place to start is the pioneer work on Cantonese
grammar by Cheung (1972). In his book, he holds the stand that wo5
is a quotative particle which is a combination of the verb waa6 [speak/
say] and the SFP aa3. Cross-linguistic studies reveal that the ‘say’
verb across languages tends to develop into a quotative marker of a
complementizer (Heine et al. 1991; H. Sohn 1999; Klamer 2000;
Heine/Kuteva 2002; Hopper/Traugott 2003). For example it is worth
noting the clear case of grammaticalization regarding the relation-
ship between the verb ‘say’ and the final particle in Taiwanese
(Chappell to appear, Wu 2000: 158-178). Wu examines the ongoing
creation of a new Taiwanese particle, the element kong which is
grammaticalizing from the verb ‘to say’ as a complementizer-type
particle in an unexpected sentence-final position. Xu and Matthews
(2005) also touch on this topic and give an account on the
grammaticalization of ‘say’ in Taiwanese. In a recent work, Liu and
Gu (2008) propose that the reported speech marker di and the re-
ported evidential marker di in Nuosu Yi language are developed from
the lexical verb ‘say’, and the grammaticalization process is trig-

gered by argument structure change motivated by the force of struc-
tural simplification. Likewise, the quotative particle wo5 in Cantonese
might have possibly derived from the lexical verb waa6. Ball (1888)
leaves us a clue that there was an SFP waa5 denoting the previous
statement made by other people which does not exist in modern
Cantonese. However, Cheung (1972) gives no further explanation
for this. Deng (1991) claims that wo5 is used in reported speech to
express utterance or thought, and can be attached to any word class.
A similar position is taken by Luke (1990) and Matthews (1994),
who think that wo5 serves to quote reported speech and hearsay in-
formation. To further discuss this particle, Matthews (1998) argues
that wo5 used in reported speech (explicit or otherwise) appears to
be a typical ‘hearsay’ evidential.
The above is a brief review on the literature concerning the
study of SFPs wo3 and wo5. It is fair to say that numerous scholars
have noticed the characteristics of the two SFPs. However, the rela-
tion between the two is usually assumed without much comment,
and very little discussion is devoted to explaining the differences
between them. Therefore, it seems reasonable to proceed one step
further by examining wo3 and wo5 in detail.
2.2 Data and Romanization
Three samples have been chosen as the basis of linguistic analysis in

the following sections: the first one relies on three Cantonese movies
produced in the 1990s, namely Fight back to the school I (Fight 1991),
Fight back to the school II (Fight 1992) and God of Cookery (God
1996). The second one is the Hong Kong University Cantonese Cor-
pus (HKUCC) for the collection of language examples in the 1990s.
K.K. Luke of the University of Hong Kong has established a corpus
made up of a large amount of Cantonese speech and conversational
recordings which have been fully transcribed, segmented, tagged and
parsed. In the 150 recordings of this 200,000-word corpus, 67 were
conversations between 2-3 speakers, 51 were radio broadcasts and
32 were stories told by one story-teller. The main characteristic of
180 Leung Wai-mun
the HKUCC is its emphasis on colloquial, everyday language. The

third one is early Cantonese materials for the examples between the
late 19th century and the early 20th century. I have selected Bridgman
(1841), Ball (1888) and Wisner (1906) which are the records of
Cantonese dialects attached with Roman phonetic transcription to
indicate the actual sounds and can possibly stand for the colloquial
forms in those periods of time.
According to the level of naturalness, the first sample (mov-
ies) belongs to the category of ‘Topic-restricted/text preparation
needed/not reading from scripts’ and the second sample (HKUCC)
goes to the category of ‘Topic free / No preparation’. The third sample
(the early Cantonese materials) belongs to the category of ‘Topic-
restricted/text preparation needed/reading from scripts’. The com-
mon characteristic of the above samples is that they are all natural
colloquial language and highly representative. With the movies, our
analysis will no longer be based on a small sample or subjective
feelings.
Concerning the romanization scheme, there are about ten or
more major schemes for Cantonese up till now which is very confus-
ing for users. There is little motivation for local people to learn any
of these systems as they seem to prefer using Chinese characters.
The romanization systems are not taught in schools either in Hong
Kong or in Guangdong. Specialists of the Hong Kong Linguistics
Society have developed a new scheme named ‘Cantonese Romaniza-
tion Scheme’ in 1993, which is used in this chapter.
3. Evidentiality, mirativity and subjectivity
3.1 Evidentiality
The different attitudes people have towards different things can be

expressed by a variety of linguistic forms. Let us take English as an
illustration:
(1) A. It’s raining.

B. It’s probably raining.
C. Maybe it’s raining.
D. It must be raining.
E. It sounds like it’s raining.
F. It’s sort of raining.
G. Actually, it’s raining.
Chafe and Nichols (1986) think that sentence A is stating a fact; B

and C represent probability; D expresses a certain prediction; E and
F express a not-so-certain prediction; and G shows unexpected in-
formation. In other words, the speaker conveys his or her attitude
towards the stated fact or proposition using a specific linguistic form,
like the words probably, maybe, must, sort of and actually, and these
linguistic forms are called evidentials. In Mandarin, evidentiality is
often expressed by projecting clauses, modal auxiliaries, verbs, ad-
verbs, adjectives and nouns (Zhu 2006).
Evidentiality deals with the source of information for the
speaker’s utterance. It allows speakers to specify why they believe a
given statement, i.e. what kind of evidence they have for it. These
kinds of evidence might be divided into four criteria as: (1) Direct
experience (Witness vs. Nonwitness), (2) Visual obsevation (Visual
vs. Nonvisual), (3) Inferential and (4) Hearsay. (Givon 1984: 307-
308, Palmer 1986: 66-67). The following evidential hierarchy is taken
from De Haan (1997). Under this hypothesis, it will universally be
the case that direct evidence (e.g., visual and auditory evidence) is
more believable than indirect evidence (e.g., inference and hearsay).
Direct evidence (Confimative) Indirect evidence (Non-confimative)

visual < auditory < non-visual < inference < quotative
→ More believable → Less believable
A witness evidential, which is in contrast with a nonwitness eviden-

tial, indicates that the information source was witnessed or experi-
enced by the speaker personally. This is usually from visual observa-
tion. A nonwitness evidential indicates that the information was not
witnessed or experienced by the speaker personally but was only
obtained through an indirect source. An indirect evidential is used to
mark information that was not observed by the speaker personally.
182 Leung Wai-mun
This may include inferences or reported information. An inferential

evidential, such as uncertainty or probability, indicates that the in-
formation was inferred from secondhand evidence by the speaker.
Hearsay or quotative function is now accepted as one of the
basic categories of evidentiality (Aikhenvald 2004). Reportative
evidentials express that the information was reported to the speaker
by a third party. It also indicates the information may not be accurate
and open to interpretation (Aikhenvald 2004, De Haan 1997, 1999,
Chafe/Nichols 1986). It is an evidential that signals that someone
else is the source of the statement made. The quotative evidential
might be found as a grammatical category in Cantonese because this
mood is neither optional, nor expressed in diverse ways. Evidentiality
is believed to be a very useful functional analysis of the multiple
functions of the particle such as inference and hearsay.
3.2 Mirativity
According to Chafe and Nichols (1986), the term ‘evidentiality’ can

be described as follows:
(1) source of knowledge: evidence, language, hypothesis, etc.

(2) mode of knowing: belief, hearsay, induction, deduction, etc.
(3) knowledge matched against: verbal resources, expectations,
etc.
Matthews (1998) claims that ‘it is the third parameter, whereby know-
ledge is matched against expectations, that mirativity encodes.’ (3) is
a kind of miratives which indicates new and unexpected information
for which the speaker lacks psychological preparation. DeLancey
(1997: 36) remarks:
[The category mirative] marks both statements based on inference and state-
ments based on direct experience for which the speaker had no psychological
preparation, and in some languages hearsay data as well. What these appar-
ently disparate data sources have in common […] is that the proposition is one
which is new to the speaker, not yet integrated into his overall picture of the
world.
In Cantonese, take the examples shown by Li (1995) as further illus-

trations:
(2) 䇠䇠
ª¨ª¨ ¥s §A⎓§Y¨è ¥hἈ Ø{ ⌛⇣ ⍣ ┶
baa4baa1
baa4baa1 giu3 nei5
nei5zik1 zik1
hak1hak1
heoi3 wo3
heoi3 wo3
father
father askask
2SG immediate
2SG go SFP
immediate go SFP
‘Father
‘Father asked you to go immediately!’
immediately!’
(3) 䇠䇠
ª¨ª¨ ¥s §A⎓§Y¨è ¥h Ἀ? ⌛⇣ ⍣ ┅
baa4baa1 giu3 nei5 zik1hak1
nei5 heoi3 wo5 heoi3
zik1hak1 wo5
father askask
2SG immediate
2SG go SFP
immediate go SFP
‘Father said to me:
me: ‘Ask
‘Ask him/her
him/hertotogo
goimmediately.’
immediately.’‘ ‘
In Example 2 wo3 shows a mirative which is an evidential-like cate-

gory used to mark information showing ‘new or surprising’, while
in Example 3 wo5 indicates that the information was reported by
another person, thus it is apparently a quotative evidential. The ex-
amples help clarify the relationship between mirative and quotative.
Another example:
(4) (Fight 1991)

◑炻 ᾳ 㘗 ⤥ 朂 ┶炰
waa3, go3 ging2 hou2 leng3 wo3
INJ, DEM view INT pretty SFP
‘Wow, the view is very beautiful!’
In Example 4, the speaker is at a friend’s home and expresses his

exclamation as he is watching the beautiful sea view from the living
room. Wo3 serves as a reminder to the friend and intensifies the ex-
clamation. This use of wo3 with exclamatory utterances seems to be
a novel one. The category ‘mirative’ in this example indicates unex-
pected information for which the speaker lacks psychological prepa-
ration. It is the semantic counterpart of pragmatic functions and fits
point (3) ‘knowledge matched against expectations’ (Chafe and
Nichols 1998) that was mentioned at the beginning of this section.
Furthermore, Matthews (1998) shows that some other lan-
guages have the same construction or marker to express both
evidentiality as generally understood (inference and/or hearsay) and
mirativity, for example, in the Hare (Athabascan) language the sen-
tence particle lõ has evidential or mirative value:
184 Leung Wai-mun
(5) Mary e-wé’ ghálayeda lõ

Mary its-hide work.IMPF PRT
‘Mary is working on hides (seemingly or unexpectedly).’
(cited in Matthews 1998)
The particle lõ here in conjunction with an imperfective form of the

verb is a mirative one ‘where the speaker has first-hand knowledge
of unanticipated information.’ (DeLancey 1997: 39, cited in Matthews
1998)
3.3 Subjectivity
In verbal communication, not only should the speaker express the

propositional meaning, but also the speaker’s intention, e.g. the real-
ization of language ‘subjectivity’, which refers to the situation that
the speaker utters a sentence to express his stance, attitude and emo-
tion so as to leave the ‘self-impression’ in the discourse. (Lyons 1982,
Finegan 1995, Shen 2001). If this kind of subjectivity obtains a sub-
jective function through the explicit language structure or the evolu-
tion of a language pattern, it can be called ‘subjectivization’. In addi-
tion, Finegan (1995) thinks that the research on subjectivity has mainly
focused on three points, namely the speaker’s perspective and affec-
tion, and the epistemic status of the propositions. Affection includes
emotions and attitudes, and reflects what the speaker judges on the
state of affairs and its possibility.
In the theory of grammaticalization, Hopper and Traugott
(1993) claim that subjectivization means that such meaning neither
refers to the objective context nor indicates objects, but to the
speaker’s subjective opinions to cater for the need of discourse struc-
ture. Let us see how Traugott and Konnig characterize the change in
meaning associated with grammaticalization:
meanings grounded in more or less objectively identifiable extralinguistic situ-

ations to meanings grounded in text-making (for example connectives, anaphoric
markers etc.) to meanings grounded in the speaker’s attitude to or belief about
what is said… (1991: 189).
It is generally said that speakers’ stance or intention can be implic-

itly expressed by some of the wordings in conversation.
After Lyons’ (1982) claim of subjectivization, Traugott (1989)
tried to bring it within her framework. She points out that
subjectivization refers to the process whereby meaning becomes more
and more embodied in the speaker’s subjective belief and attitude
towards the proposition. It is a process of semantic-pragmatic transi-
tion. Traugott emphasizes that subjectivization is an important mecha-
nism in the process of grammaticalization.
In verbal communication of daily life, not only should the
speakers express the propositional meaning, but also the embodi-
ment of language subjectivity. Linguists’ studies of subjectivization
began in the late 1980s. There are mainly two approaches, namely
synchronic and diachronic; the former, especially Langacker’s works
(1985, 1990, 1999), investigates in what structure speakers reflect
their subjectivity at a particular period from the view of cognitive
linguistics, whereas the latter, especially Traugott’s works (Traugott
and Dasher 2002), examines from a diachronic viewpoint the changes
of the sentence structure which shows subjectivity. Traugott (1989)
traces the development in English of lexical and grammatical items
into markers of epistemic modal meaning. A similar approach to Japa-
nese connectives documents their shift from propositional to dis-
course-based meanings (Matsumoto 1988).
Evidentiality studies how the speaker indicates the source of
information or expresses his or her attitude. The information provided
by the speaker is built to a greater or lesser degree on the knowledge
or evidence that he or she holds, and it is uncommon that the inference
is purely subjective without any objective supporting evidence. Since
people can hardly speak purely objectively, evidentiality inevitably
involves elements of subjectivity, and the elements of subjectivity are
not the same in all circumstances but dependent on the speakers.
Besides, the studies of evidentiality and subjectivity are closely re-
lated to epistemic modality. There are two types of modality which are
said to be epistemic modality and deontic modality. Epistemic modal-
ity is used to express the judgment of the state of affairs and the
possibility of the situation. Matthews and Yip (1994) categorized wo3
and wo5 into ‘epistemic particles’. On the other hand, deontic modal-
186 Leung Wai-mun
ity involves obligation and permission (Frawley 1992). As evidentiality

studies a speaker’s attitude and judgment towards some knowledge,
its content is therefore very similar to that of epistemic modality.
Palmer (1986: 8) calls evidentiality ‘evidential modality’, which, same
as epistemic modality, belongs to propositional modality. Subjectiv-
ity also has a close relationship with epistemic modality. In the pre-
vious section, it is mentioned that Finegan (1995) has concluded three
points in his study of subjectivity, and the third point ‘epistemic status
of the propositions’ is indeed a study of the speaker’s expression of
moods or the recognition of propositions, which are mainly conveyed
by means of verbs, adverbs and adjectives.
4. Cantonese evidential particles
In Cantonese, the sentence-final particle wo5 expresses hearsay

(Kwok 1984, Luke 1990), and Matthews (1998) treats it as evi-
dentiality. The features of wo3 include unexpectedness and note-
worthiness (Luke 1990), and Matthews (1998) regards it as mirativity.
These two particles only differ in lexical tones. The following sub-
sections explain the uses of wo5 and wo3 in modern Cantonese and
give an analysis of them.
4.1 The feature of wo5
Modern Cantonese has a particularly rich inventory of SFPs, which

serve various functions. To express objective attitudes, wo5 is a typi-
cal ‘hearsay’ particle used in reported speech (Kwok 1984, Luke 1990,
Matthews 1998). Wo5 is best translated into English with ‘I have
heard that…’, ‘I hear that…’, ‘s/he told me that…’ or ‘Reportedly’.
According to Li (1995), wo5 is to report what other people say with
a sense of emphasis. Compared with wo3, wo5 emphasizes that the
quoted content was uttered by other people. Let us examine Li’s ex-
amples again:
(2) 䇠䇠
ª¨ª¨ ¥s §A⎓§Y¨è ¥hἈ Ø{ ⌛⇣ ⍣ ┶
baa4baa1 giu3 nei5
nei5zik1 zik1
hak1hak1
heoi3 wo3
heoi3 wo3
father askask 2SG
2SG immediate immediate
go SFP go SFP
‘Father asked you to go immediately!’
immediately!’
(3) 䇠䇠
ª¨ª¨ ¥s §A⎓§Y¨è ¥h Ἀ? ⌛⇣ ⍣ ┅
baa4baa1 giu3
giu3 nei5 zik1hak1
nei5 heoi3 wo5 heoi3
zik1hak1 wo5
father askask
2SG immediate
2SG go SFP
immediate go SFP
‘Father said to
to me:
me: ‘Ask
‘Ask him/her
him/hertotogo
goimmediately.’
immediately.’‘ ‘
Li (1995: 509-510) claims that Example 2 is merely a reported speech

and aims at drawing other’s attention, while the mood expressed by
the latter contains ‘the mood of objection’. The relationship between
wo3 and wo5 is not clear and convincing in Li’s treatment. The use
of the hearsay particle wo5 at the end of Example 3 makes the sen-
tence an objective reported speech, and wo5 itself does not contain
any ‘mood of objection’. The claimed mood ‘of objection’ is consid-
ered to be the result of intonation, rather than the meaning of wo5
itself. Two more examples are shown as follows:
(6) (Fight 1991)

旧 sir¸Ü
ªü sir 娙
-ø ¨Ï Ⓖ°μ ? ἧ ┅
aa3 sir waa6 m4m4 sai2
sai2 zou6
zou6 wo5
wo5
SUF sir say NEGNEG need
need do
do SFP
SFP
‘he said it is not necessary
necessary to
to do.’
do.’
(7) (HKUCC)
␊ ??┚◊
©O Ê\ -ø Ἂ
ª¾ ? Ⓖ 䞍 ┅
nei1di1 je5 keoi5 m4 zi1zi1wo5wo5
DEM things 3SG
3SG NEG
NEG know
knowSFPSFP
‘he said he does not know
know about
about these
thesethings.’
things.’
Hearsay evidential wo5 in the above examples is used to report what

someone else has said. Because of that, it may also be a device for
disclaiming one’s responsibility (Example 6) or even to show one’s
doubt (Example 7).
188 Leung Wai-mun
4.2 The features of wo3
In modern Cantonese, wo3 is an SFP used with a high frequency.

Some recent research has shown that this particle is used to show
‘realization and reminders’ (Rao 1981, Kwok 1984, Ouyang 1990,
Deng 1991, Leung 1992, Fang 2003). Consider some common ex-
amples:
(8) (Fight 1992)

Ṳ㖍 ¤Ñ®ð⣑㯋
¤μ¤é ¦n ¦n Ø{ ⤥ ⤥ ┶
gam1jat6 tin1hei3
tin1hei3hou2 hou2 hou2
hou2 wo3 wo3
weather
today weather verySFP
very good good SFP
‘The weather is
is very
very good
good today!’
today!’
(9) (Fight 1991)

␊⹎ -ø -ã
©O«× Ⓖ -¹·Ï Ø{ Ⅾ 梇䄁 ┶
lei1dou6 m4m4zeon2 sik6 jin1 wo3
zeon2 sik6 jin1 wo3
here NEGNEGallow smoking
allow SFPsmoking SFP
‘Smoking is not allowed
allowed here!’
here!’
(10) (HKUCC)
◑炰 ¯u«Y
¼M¡I 䛇Ὢ ´X ®Þ¤Ò ⸦ Ø{¡I㡛⣓ ┶炰
waa1 zan1hai6
waa1 zan1hai6gei2gei2 so1fu4
so1fu4 wo3 wo3
INJ really
INJ really quite
quite wonderful wonderful
SFP SFP
‘Wow, it’s really wonderful!’
‘Wow, wonderful!’
In Example 8, the speaker expresses surprise or wonder at the unusu-

ally or unexpectedly pleasant weather. Example 9 is to remind the
hearer that smoking is prohibited. Example 10 is an exclamatory sen-
tence expressing the tone of surprise. Wo3 can be used in different
types of sentences expressing various tones of voice in conjunction
with different lexical and intonational devices. The different uses of
wo3 deduced from the data are summarized below:
4.2.1 Realization
In Example 11, Speaker A reminds Speaker B that it is nearly three

o’clock and as a result Speaker B remembers that there is a gather-
ing.
(11) (God 1996)

ᶱ ÂI溆?
A: ¡G¤T ♡
saam1 dim2 laa3 laa3
three o’clock SFPSFP
‘It’s three o’clock already.’
already.’
Ὢ Ø{¡AÊ\
B: ¡G«Y ┶炻Ἂ? ·|⑳ ¾¤ ?㚫湶 ♡
wo3,
hai6 wo3, keoi5dei6wui5
keoi5dei6 wui5lai4 laa3
lai4 laa3
yes SFP, SFP, they come
they AUX AUXSFP come SFP
‘Oh yes. They areare coming
coming soon.’
soon.’
In Example 12, the speaker takes some bones to feed a dog, and then
remembers that in the backyard there is a cat, which might be hungry
then, and so he takes some food to feed the cat also. The one who
‘realizes’ that something has to be done is the speaker himself.
(12) (God 1996)

©ð
㉶ ? °©ÀY
┚ 橐柕 ¥h Áý ? ª¯⍣ ¥ý¡A棝¾î Õi⎻μL 䉿 ? ⃰炻㨓㌪䃉 ◊
ling1 di1 gwat1tau4
gwat1tau4heoi3
heoi3wai3
wai3haa5
haa5gau2
gau2sin1,
sin1,waang4dim1
waang4dim1mou4
mou4je5
je5
Take CL CLbonebone
go feed ADVgo feed dog ADV
first , dog
anyway
first no
, thing
anyway no thing
°μ……
…… «YὪØ{¡A ┶炻«á⼴ -± ¥ò朊 ¦³ °¦ ¿ß¡A
ẚ -ø 㚱ª¾ 晣 ¾j 尻炻 Ⓖ 䞍梻
zou6……hai6 wo3, hau6min6
hau6min6 zung6zung6jau5
jau5zek3
zek3maau1,
maau1, m4m4zi1zi1
ngo6 ngo6
do…. yesyesSFP, backback
SFP, still have still
CL cat, NEG
have CLknow cat,hungry
NEG know hungry
Ⓖ¾j 梻
-ø ©O¡H ␊烎
m4 ngo6 ne1? ne1?
ngo6
NEG hungry SFP SFP
‘Let me take the bones and feed the dog as I have nothing to do now…Oh,
there’s also a cat in the backyard, is it hungry or not?’
4.2.2 Reminding
‘Reminding’ and the above-mentioned ‘Realization’ are two sides of

the same coin. In ‘reminding’, it is the speaker who reminds the hearer
and consequently the hearer remembers something, while in ‘real-
ization’, the speaker suddenly realizes something as a result of the
reminder by someone or some situations, and in this case it is the
speaker who remembers something. The following examples show a
situation where the hearer remembers something as a result of the
reminder of the speaker.
190 Leung Wai-mun
(13) (HKUCC)
A: ¡G¶¼
梚 ¥¤⤞¯ù ¦P勞 ¶¼ ¯ù
⎴³£ 梚 -ø ¦P Ø{¡A-øª¾
勞悥 Ⓖ ÂI ¸Ñ ⎴ ┶炻Ⓖ䞍溆妋
jam2 naai5caa4
naai5caa4tung4
tung4jam2
jam2caa4
caa4dou1
dou1m4
m4tung4 wo3,
tung4 m4m4
wo3, zi1zi1
dim2gaai2
dim2gaai2
drink milk
milktea
teaand drink tea ADV
and drink NEGNEG
tea ADV same SFP,
sameNEG knowknow
SFP, NEG why why
¶¼ ¥¤⤞
梚 ¯ù ·| 勞
«d ?㚫?¡A ¦]⇲¬° Ê\ ┚ ▭炻 ⚈ 䁢 Ἂ
naai5caa4 wui6
jam2 naai5caa4 wui6 soek3
soek3 di1 ge2, jan1wai6
di1 ge2, keoi5
jan1wai6 keoi5
drink milk tea AUX
AUX weaken-stomach
weaken-stomach ASP ASPSFP,
SFP,because
becauseit it
㽫 ±o º¢ §r¡H
¿@ ⼿㺗 ⏨烎
nung4
nung4 dak1 zai6 dak1aa4 zai6 aa4
concentrated COM excessive SFP?
‘Milk tea and plain tea taste different. I don’t know why milk tea weakens
the stomach. Is it because it is too concentrated?’
B: Ὢ ⓲ 㽫 ┚ˤ
hai6 aa3 nung4 di1
COP SFP concentrated ASP.
‘Yes, it’s more concentrated.’
In Example 13, Speakers A and B are discussing Chinese tea and

milk tea. Prior to this exchange, Speaker B said that milk tea cer-
tainly does no harm to the stomach. He always drinks Chinese tea.
Here, Speaker A reminds Speaker B that Chinese tea is different from
milk tea.
(14) (HKUCC)
§A ¤d¬è
Ἀ ⋫䣰-ø ¦n Ⓖ Ý¯ ⤥§Ú °Ú¡A
㏝ §Ú ㆹ ¤w¸g ⓲炻 ¤E ㆹ -Ó project
⶚䴻 °Ú¡A ḅ ᾳ project ⓲炻
nei5 cin1kei4
nei5 cin1kei4m4m4hou2
hou2wan2
wan2ngo5
ngo5 aa3,
aa3, ngo5ngo5 ji5ging1
ji5ging1 gau2
gau2 go3go3project
project aa3,
aa3,
2SG ADV
2SG ADV NEG NEGgood
good find
find 1SG SFP, SFP,1SG
1SGalready
already nine
nine CLCLproject SFP, SFP,
project
§A ¦Û¤v
Ἀ 冒⶙¥h ÁÙ ⍣ °Õ¡A§A
怬 ┎炻Ἀ ¦Û¤v ¶¤ ®I 冒⶙¥h Ø{ 昲 ❳ ⍣ ┶
nei5 zi6gei2
nei5 zi6gei2heoi3
heoi3waan4
waan4laa1,
laa1,nei5
nei5 zi6gei2
zi6gei2 deoi2maai4
deoi2 maai4 heoi3
heoi3 wo3
wo3
2SG self
2SG self go return SFP, SFP,
go return 2SG2SGself handselfin ASP
handgo in SFP.
ASP go SFP.
‘Don’t ask me to do it as I have nine projects at hand already. Why don’t you go
yourself? You just need to drop it.’
In Example 14, someone asks the speaker to return some books to

the library, but the speaker refuses and replies that he is too busy, and
he also reminds the hearer that returning books to the library is very
simple as one needs only to place the books in the bookdrop.
(15) (Fight 1991)

ᶨ昋 Ἀ 天嶇 ⮎ ㆹ ┶炻
jat1zan6 nei5 jiu3 gan1 sat6 ngo5 wo3
later 2SG must follow closely 1SG SFP
⑒ ⍰ ℵ 敾䤵 ⓲
mai5 jau6 zoi3 cong2wo6 aa1
NEG again again make trouble SFP
‘Remember, follow me closely and don’t make trouble again.’
The above example is a dialogue between a leader and his attendant.

‘Follow me closely’ is an order to be followed strictly. The impera-
tive tone-of-voice seems to be expressed by the particle wo3 but ac-
tually wo3 serves no more than expressing a strong sense of reminder
to the hearer in the case. If wo3 is replaced by another particle such
as aa3, the tone-of-voice of the utterance is not changed.
4.2.3 Contrasting
The following example demonstrates another function of wo3 which

is to ‘Show an element of Contrast’:
(16) (Fight 1991)

A: 㚱 ℯ ◊ ⓲烎
jau5 mou5 je5 aa3
have NEG thing SFP?
‘Is there something wrong?’
B: 㝞干 ℯ ◊ ┶ˤ
gaa3 ce1 mou5 je5 wo3
CL car NEG thing SFP.
‘There’s nothing wrong with the car.’
A: ◱ 溆妋㚫㚱倚 ▭烎
gam2 dim2gaai2 wui5 jau5 seng1 ge2
DM why AUX have sound SFP?
‘Then why are there some noises?’
B: ␐⚵ 䛯 ⎻ˤ
zau1wai4 tai2 haa5
Around look ASP.
‘Let’s look around.’
In this example, Speaker A hears some noises in the car park and
feels worried, as there are valuables in the car. He then asks Speaker
B whether everything is fine. Speaker B replies that there is no prob-
lem with the car. In Speaker B’s reply, wo3 appears at the end of the
utterance and conveys the meaning ‘in spite of appearances or what
192 Leung Wai-mun
you think, the car is okay’. Thus, the particle shows that the message
contains an element of ‘contrast’.
(17) (Fight 1992)

A: ㇧ ℍ朊 ℯ Ṣ
fong2 jap6min6
jap6min6mou5mou5 jan4 jan4
room inside
insideno person no person
‘There’s
‘There’s nobody
nobody inin the
the room.’
room.’
B: 旧⪚ ╢ ⹎ ┶
aa3maa4
aa3maa4 hai2 dou6hai2 wo3 dou6 wo3
PRE-grandmother
PRE-grandmother COP COPhere here
SFP SFP
‘Hey Granny
Granny isis here.’
here.’
A: Ⓖ ἧ 䎮 Ἂ ⎋㝞
m4 sai2 sai2 lei5 gaa3 keoi5 gaa3
lei5 keoi5
NEG need need
care care
aboutabout
3SG SFP 3SG SFP
‘No need
need to to care
care about
about her.’
her.’
Similarly, in Example 17, speaker A says that there is no one in the

room, but speaker B discovers that the grandmother is in the room
and reminds speaker A that the actual situation is different from the
one he described, despite the fact that speaker A implies that whether
the grandmother is present or not will not affect them in any way.
Wo3 at the end of speaker B’s utterance again has an element of ‘con-
trast’.
4.2.4 Hearsay
Hearsay is reporting other people’s speech. In Cantonese, there are

two particles which express hearsay, namely wo3 and wo5. In many
situations, these two particles are interchangeable, but there are subtle
differences between the two. These differences will be discussed in
the next section, and here we will concentrate on the particle wo3.
Wo3 appeared eleven times in the selected textbooks of the
late 19th century to the early 20th century. It occurred in reported
speech and its main function was ‘quotative’ at that point, there
seemed to be no SFP wo5. As such, the function of wo3 in the past
was very different from those meanings carried by the modern
Cantonese particle wo3. ‘Quotative’ is one of the main functions of
modern wo3 only and is not the principal one. In contrast, in the late
19th century to the early 20th century, wo3 was mainly used to report
other’s speech, thus having a quotative function, which can be seen
in the following two examples:
(18) (BC 1906)

A: Ἂ ⸦㗪 ♇ ␊烎
keoi5 gei2si4 lai4 ne1
3SG when come SFP?
‘What time will he come?’
B: Ἂ 娙 ᶨ昋攻 ⯙ ♇ ┶
keoi5 waa6 jat1zan6gaan1 zau6 lai4 wo3
3SG say a while ADV come SFP
‘He said he would come after a while.’
(19) (CME 1888)

A: Ἂ 娙 Ḅ◊ ␊烎
keoi5 waa6 mat1je5 ne1
3SG say what SFP?
‘What did he say?’
B: ℯ 戨 ┶炻戨ℑ 䵲 ┶ˤ
mou5 ngan4 wo3, ngan4loeng2 gan2 wo3
NEG money SFP, money pressing • @SFP.
‘No money. – Short on cash, he said.’
The two examples above clearly show that wo3 is used for reporting
other’s speech. The English translations were provided by the text-
book authors, and from the translations ‘he said’ and ‘he says’, we
can judge that they are examples of indirect speech. In Example 18,
speaker A asks speaker B what time ‘he’ will come, and speaker B
uses the structure ‘ Ἂ娙…┶ ’ [He said…wo3] in his reply. In Ex-
ample 19, in speaker B’s reply to speaker A’s question ‘What did he
say?’, the clause ‘ Ἂ娙…’ [He said] is omitted as it is known from
the context, and the particle wo3 alone signals that the answer is not
speaker B’s but a report of the speech of that person they are talking
about. In terms of sentence structure, there have not been any major
changes in the past one hundred years; both ‘ Ἂ娙…┶’ [He/She
said…wo3] and ‘…┶ ’ […wo3] are acceptable.
194 Leung Wai-mun
In modern Cantonese, wo3 still carries the function of hearsay. In

Example 20, the speaker conveys good wishes to the hearer from
someone else who is not present.
(20) (Fight 1991)

Ἂ ¸Ü °Ý-Ô
Ê\ 娙 §A ⓷῁ Ἀ ┶炻䌳 Sir炰
Ø{¡A ¤ýSir¡I
man6hau6 nei5
keoi5 waa6 man6hau6 nei5 wo3,
wo3, wong4
wong4sir sir
3SG saysay
greetgreet 2SG
2SG SFP, SFP,Sir!Wong Sir!
Wong
‘He has asked me to send
send his
his regards
regards to
to you,
you,Mr.
Mr.Wong!’
Wong!’
In Example 21, wo3 also has a quotative function, as the speaker

asks A who plays most wildly and A answers that B plays most wildly,
and the particle wo3 expresses the fact that this assessment is from
someone else other than the speaker. It is added by the speaker to
express his surprise that Mr A gave this response.
(21) (HKUCC)
ㆹ °Ý⓷
§Ú 旧䓚 Ãä-Ó
ªü¥Ò ©O¡A ␊炻 ª± 怲ᾳ 䍑
±o ³Ì Åö ©O¡A ⼿㚨䘚 ␊炻
ngo5 man6 aa3gaap3 ne1, bin1go3 waan2dak1
bin1go3 waan2 dak1 zeoi3
zeoi3 din1
din1 ne1, ne1,
1SG askask PRE-A
PRE-A SFP, play
SFP, who who play wildly
COM most COM SFP,
most wildly SFP,
Ἂ ¸Ü Ê\娙Ø{¡C
Ê\ Ἂ ┶ˤ
keoi5wo3 wo3
keoi5 waa6 keoi5
said3SG3SG
3SG said SFP. SFP.
‘I asked Mr. A who played
played the
the most
most wildly,
wildly,and
andhe
hesaid
saidMr.
Mr.BBdid.’
did.’
4.3 Discussions
In Cantotnese, wo5’s main characteristic is hearsay (see 4.1), a fea-

ture which is included in ‘mode of knowing’ of evidentiality by Chafe
and Nichols (1986). The first three features of wo3, realization, re-
minding, contrast, fit Luke’s (1990) generalization that wo3 is to show
‘unexpectedness, noteworthiness and remarkableness’. Matthews
(1998) concludes that:
[Luke’s observations] fit Delancey’s notion of mirativity as a conceptual cat-

egory within the realm of evidentiality, covering surprise and unexpectedness.
This kind of mirativity can be classified as ‘knowledge against ver-

bal resources or expectations’ in the evidentiality classification scheme
of Chafe and Nichols (1986) (see 3.2). We see that both wo5 and wo3
can be called ‘evidential particles’. With the advancement of evi-
dentiality studies, what we now know about the features of wo5 and
wo3 is much more complicated than previously recognized. These
particles do not simply show the moods of the speaker, but reflect
how they recognize the external world as well.
The fourth feature of wo3 is hearsay which is considered to be
more complex to account for. It is said not to be easy to differentiate
wo3 and wo5 in some situations because wo3 overlaps with wo5 in
hearsay evidentiality. Both are often used in reported speech. Sub-
jectivization, however, gives us a very good explanation of the two
particles. We can prove it from the following minimal pair:
(22) ⩥⩥ ¸Ü -ø ¦n娙Ú» ¹qμø

¶ý¶ý Ⓖ Ø{⤥ 䛯暣夾 ┶
maa4maa1 waa6
waa6m4 hou2
m4 tai2
hou2din6si6tai2
wo3 din6si6 wo3
say good
mother say NEG NEG
watch good
TV SFP watch TV SFP
‘Mama said ‘do
‘do not
not watch
watchTV’
TV’‘‘
(23) ⩥⩥ ¸Ü -ø ¦n娙Ú» ¹qμø

¶ý¶ý Ⓖ ? ⤥ 䛯暣夾 ┅
waa6m4 m4
maa4maa1 waa6 hou2 tai2hou2
din6si6 tai2
wo5 din6si6 wo5
say good
mother say NEG NEG
watch good
TV SFP watch TV SFP
‘Mama said ‘do not
not watch
watch TV’
TV’‘‘
If we analyze this closely, we can see that there is a semantic differ-

ence when the two SFPs are employed to express speaker’s modality.
Although sometimes they can be substituted by each other, the differ-
ence of modality can more or less be seen. Wo3 reflects ‘unexpectedness,
noteworthiness, remarkableness, unusualness…etc’ (Luke 1990), and
wo5 is used when the speaker is reporting other people’s views, and
at the same time shows uncertainty about the reliability of the infor-
mation and unwillingness of the speaker to be held responsible for the
information. If language is regarded as a reflection of human’s cog-
nitive activity, wo5 is the interpretation of the original message for the
first time, i.e. the speaker reports only other people’s point of view,
while wo3 is the interpretation of the original message for the second
time, i.e. the speaker reveals his/her comment when reporting other
people’s point of view. Wo5 is the citation of speech and wo3 is the
interpretation of action, and connotes the speaker’s assessment of the
196 Leung Wai-mun
evidence for his or her statement with his or her own attitudes. The two
examples above show the different attitudes and stances of the speak-
ers. Thus, wo5 and wo3 obviously have different social functions from
the pragmatic point of view.
Moreover, the quotative evidential wo5 used in indirect speech
shows that the speaker is not willing to take the responsibility of the
quoted content, and the SFP wo3 explains others’ motivations be-
hind his/her behavior by the speaker’s own speculation and emotion,
which connotes the speaker’s assessment of the evidence for his or
her statement. From daily conversations, it is found that the hearsay
evidential wo5 is used when the speaker is not involved in the re-
ported content so as to speak objectively, irrespective of personal
opinion and subjective speculation on the issue. Aikhenvald (2004)
points out that evidentiality is neutral as to the speaker’s commit-
ment to the reported information. It can be seen that the speaker does
not show his stance when wo5 is employed, just merely reports other
people’s wording unchangeably and objectively, or explains other
people’s intention. However, wo3 is employed when the speaker is
involved in the reported content, or shows the quotation based on
his/her own speculation or intention. The apparent difference can be
seen in the example below:
(24) (Private 1976)

(An employee at a detective agency)
⮳㖍 ªü ©P¤éÍ
´M¤é 旧 ␐㖍⭴
©O¡A ¨Ï ¤H °e␊炻 Ê\ ¦Ñ±C ἧ Ṣ 復 Ἂ 侩⧮
cam4jat6 aa3 zau1jat6fu3
cam4jat6 zau1jat6fu3ne1, sai2
ne1,jan4
sai2sung3
jan4 keoi5 sung3
lou5po4keoi5 lou5po4
yesterday PRE Chow Yat-fu SFPSFP
yesterday orderorder people
people give give
3SG wife3SG wife
⻝¬Û ?¡A
±i 䚠 ¸Ü Ê\♇炻¤μ¤é 娙 Ἂ
¥h ¬~ÀY¡A Ṳ㖍Ê\ ⍣ 㲿柕炻 Ἂ
zoeng1 soeng1 lai4, waa6
zoeng1 waa6keoi5
keoi5gam1jat6
gam1jat6heoi3
heoi3sai2tau4,
sai2tau4,keoi5
keoi5
CL photo
CL photo come
come say 3SGsaytoday
3SGgo wash
todayhair 3SG
go wash hair 3SG
㆟䔹 Ê\ ¦PἊ
ÃhºÃ ⎴ ?¡AṢÜÌ §Ú¦a
¤H ¦³¸ô 㚱嶗Ý¯ ┅炻 ╴ ㆹ⛘ ㏝
waai4ji4 keoi5
waai4ji4 keoi5 tung4jan4
tung4 jan4jau5jau5
lou6lou6
wo5, wo5, aai3 ngo5dei6
aai3 ngo5dei6 wan2 wan2
suspect 3SG
suspect 3SGwithwith
personperson
have have
affairaffair SFP1PL
SFP ask askfind1PL find
Ṣ ¥h ¸ò ¸ò ⍣
¤H Ê\ Ø{ 嶇嶇 Ἂ ┶
jan4
jan4 heoi3 gan1
heoi3gan1gan1
keoi5 wo3 gan1 keoi5 wo3
person
person go follow
go follow 3SGfollow
follow SFP 3SG SFP
‘Yesterday Mr Chou Yat-fu had his wife’s photo sent to us, saying that she is
going to the barber shop today. Mr Chou suspects that his wife is having an
affair with another man, so he asked us to follow her and to collect some evi-
dence.’
Example 24 is the only example that the mirative wo3 and the quota-
tive wo5 appear in the same utterance in our data. Wo3 and wo5 are
certainly not interchangeable in this example, otherwise the mean-
ings conveyed will be totally different. The sentence with the SFP
wo5 is used to show the speaker does not stand for the opinion of the
reported speech, which is ‘Madam committed adultery’. On the con-
trary, what we discover from wo3 used in the last sentence is when
wo3 is used; the speaker shows his stance explicitly. The reported
content and the speaker are closely related, as the speaker interprets
others’ behavior subjectively. Specification associated with wo3 may
have subsequently become reanalyzed and absorbed directly into the
element wo3 as an inherent restriction on its use. From such a con-
trast, we can say that SFP wo5 focuses on the source of information
and wo3 focuses on the speaker’s reactions (surprise and unexpect-
edness). wo3 can also be used for reporting others’ speech, where the
speaker’s stance is still clearly shown, implying that the speaker does
not report exactly what others mean, or just reports others’ messages
by his/her own understanding, or shows how true the reported speech
is (to share the responsibility with the person being cited):
(25) (God 1996)

ℯ ²z¥Ñ
A: GÉN 䎮䓙¬[¡I㝞炰 ⣏⁣
¤j¶Ì -Ó ²Ó©f ᾳ 䳘⥡-Ó
-Ó ¦Ñ¥À ᾳ 侩㭵 ᾳ
mou5 lei5jau4 gaa3,
gaa3, daai6so4
daai6so4 go3 sai3mui2go3 lou5mou2
go3 sai3mui2 go3 lou5mou2
go3 go3
NEG reason SFP
NEG SFPDaiso
Daiso CL sister
CL younger younger
CL sister
mother CLCL mother CL
⣏ἔ -Ó ¥J ¸Ü
¤j¨Ð ᾳ¦³ ¬[Ṽ Ø{¡I «}娙°Õ¡I㚱§A 㝞 ┶炰 ⑒ ┎炰 Ἀ
daai6lou2
daai6lou2 go3 go3zai2 waa6
zai2 jau5 gaa3
waa6 wo3,gaa3
jau5 maai2 laa3,
wo3, nei5 laa3, nei5
maai2
elder brother CL son
elder sonsay have
saySFP
haveSFPSFP
NEG SFP 2SG NEG SFP 2SG
彼¾¤ 湶
ªð ¥ý °Õ¡I ⃰ ┎炰
faan2 lai4 sin1 laa1
faan2 sin1 laa1
back come
back comefirst SFP
first SFP
‘This is strange! Daiso’s sister’s mother’s brother’s son said he had it. Well,
now you come back first.’
B: 怲㚱 Ⓩ炰
bin1 jau5 zek1
where have SFP
‘He does not have it!’
This example shows that wo3 cannot be substituted by wo5 in many

situations. It is because the SFP wo3 expresses the speaker’s stance
which conceals a complicated meaning. The sentence does not only
198 Leung Wai-mun
reveal what ‘Daiso’s sister’s mother’s brother’s son’’ says, but also
the speaker’s own viewpoint, i. e., he does think ‘it did have’ some-
thing. In addition to this, one of the situations we should be aware of
is, when SFPs wo3 and wo5 are used to interpret others’ motivations
behind their behavior, wo3 is preferred once the behavior is gener-
ally accepted and obligatory, and wo5 is preferred once the inten-
tions of the behavior are unclear.
Based on this difference, wo5 has developed another usage, and
the two cannot be substituted by each other. That they are not interchange-
able is shown in the following minimal pairs, where in Example 26
wo3 shows reminder, and in Example 28 wo5 shows that the informa-
tion gathered is not correct and the tone-of-voice is disapproving:
(26) (Private 1976)

¥X
↢ ? ¦æ ♇ Á¿ 埴 ? ¹D¸q ¦n 嫃Ø{ ⎻ 忻佑 ⤥ ┶
ceot1 lai4lai4
haang4 gong2gong2
haang4 haa5 dou6ji6
haa5hou2 wo3
dou6ji6 hou2 wo3
out come
comewalkwalk
talk ADVtalk
virtue good
ADVSFP virtue good SFP
(27) *¥X
*↢ ? ¦æ ♇ Á¿ 埴 ? ¹D¸q ¦n 嫃? ⎻ 忻佑 ⤥ ┅
ceot1 lai4
lai4haang4 gong2gong2
haang4 haa5 dou6ji6
haa5 hou2 wo5dou6ji6 hou2 wo5
out comecome
walkwalk
talk ADVtalkvirtue ADV
good SFP virtue good SFP
‘Being in the society,
society, won’t
won’t you
you try
try and
and be
be kind
kindand
andloyal?’
loyal?’
(28) (Private 1976)

¥H¬°
ẍ䁢 Ê\ ¯qἊ §A 䙲 ?¡A §Q¥Î
Ἀ §A ┅炻 ©Q¡I ⇑䓐 Ἀ ␳炰
ji5wai4 keoi5
keoi5jik1
jik1nei5nei5
wo5, lei6
wo5,jung6 lei6
nei5jung6
zaa3 nei5 zaa3
think 3SG3SG
benefit 2SG2SG
benefit SFP take
SFPadvantage of 2SG SFP
take advantage of 2SG SFP
(29) *¥H¬° Ê\ Ἂ
*ẍ䁢 ¯q §A䙲Ø{¡A Ἀ §Q¥Î §A ┶炻 ©Q¡I ⇑䓐 Ἀ ␳炰
keoi5 jik1
ji5wai4 keoi5 jik1nei5 wo3,
nei5 lei6jung6
wo3, nei5 zaa3
lei6jung6 nei5 zaa3
think 3SG benefit
3SG 2SG 2SG
benefit SFP take advantage
SFP of 2SG SFP
take advantage of 2SG SFP
will benefit
‘Don’t think he will benefit you,
you, he
heisis just
justtaking
takingadvantage
advantageofofyou!’
you!’
One more point to add, from the pragmatic point of view, is that the
elements in the language that are being used to achieve communica-
tive purposes are worth studying since the combined meaning of the
individual words in a sentence may differ from what the speaker or
writer intends to mean. In pragmatics, a speech act is divided into
three parts: locutionary act, illocutionary act, and perlocutionary act.
Of these, the illocutionary act is the core of the speech act. A locution-
ary act is to express the literal meanings of the sentences. An illo-

cutionary act means to perform some intended action by producing
speech. A perlocutionary act is to use language to bring about some
consequences. Austin thinks that every utterance produced by a speaker
contains all the above three effects or forces. The theory of speech acts
underscores the importance of the distinction between linguistic
meaning and language use, and from this viewpoint Example 3 is
clearly a locutionary act, where the utterance with wo5 merely ex-
presses the fact and is a simple reporting of other’s speech. In contrast,
Example 2 is an illocutionary act, where wo3 is used by the speaker
to achieve an action, that is, the speaker is actually reminding and
requesting the hearer: ‘Father asked you to go immediately! Don’t do
anything else!’ It is clear that wo5 and wo3 differ in that the former
concerns linguistic meaning while the latter concerns language use.
5. Conclusions
Evidentiality refers to how the speaker explains the source of know-

ledge, attitude towards knowledge and commitment to knowledge,
and it reflects whether the information is from personal experience,
inference or other people. Evidentiality is present in every language,
but it exists in different linguistic forms, which may be grammatical
inflections or lexical items, and sentence-final particles in Cantonese.
Wo3 expresses mirativity, while wo5 expresses hearsay. According to
Chafe and Nichols (1986), the categories of mirativity and hearsay
belong to the realm of evidentiality. Matthews’ (1998) observation
that ‘the category of mirativity identified by DeLancey enables wo3
to be related systematically to the hearsay evidential wo5’ brings in a
new insight for the investigation of the relationship between the SPFs
wo3 and wo5.
The final point to emphasize is that there is a complicated situ-
ation in that wo3 and wo5 overlap in the feature of hearsay, and this
makes some people believe that these particles are interchangeable.
But on close inspection it is found that they are apparently distinct, and
200 Leung Wai-mun
their difference can be explained using the concept of subjectivity.

Aikhenvald (2004: 209) thinks that quotative evidentials commonly
develop ‘mirative extensions’. Using wo3 as an illustration, we can
postulate that the features of wo3 may have undergone a quotative-to-
mirative movement and may have been shifted from ‘this is the news
I heard’ (hearsay) to ‘this is new to me/you’ (newsworthiness). How-
ever, DeLancey (1997: 37) argues that the reverse derivation is at least
as plausible, that is, the mirative feature rather than the quotative is the
basic one. In modern Cantonese, the main features of mirative wo3 are
realization and reminding, and further diachronic studies are needed
before we can decide on the directionality of the derivation.
References
English
Aikhenvald, A.Y. 2004. Evidentiality. Oxford: Oxford University Press.

Aikhenvald. A. / Dixon, R. (eds) 2003. Studies in Evidentiality.
Amsterdam: Benjamins.
Austin, J.L. 1962. How to Do Things with Words. Oxford: Clarendon
Press.
Ball, J.D. 1888. Cantonese Made Easy (2nd edition). HK: The China
Mail Office.
Ball, J.D. 1924. Cantonese Made Easy (4th edition). HK: Kelly and
Walsh Ltd.
Bridgman E.C. 1841. Chinese Chrestomathy in the Canton Dialect.
Macao: S. Wells Williams.
Bussmann, H. 1996. Routledge Dictionary of Language and Linguis-
tics. (Translated and edited by G. Trauth / K. Kazzazi). Lon-
don: Routledge, 1996.
Chafe, W. / Nichols, J. 1986. Evidentiality: The Coding of Episte-
mology in Language. Norwood, NJ: Ablex.
Chafe, W.L. / Nichols, J. (eds). 1986. Evidentiality: The linguistic
coding of epistemology. Norwood, NJ: Ablex.
Chappell, H. To appear. The grammaticalization of the verb kong

‘say’ in Taiwanese Southern Min: A case of reanalysis. In Li,
C.N. / Peryraube, A. (eds), Morphological Change in Chinese.
Amsterdam: John Benjamins.
De Haan, F. 1997. The Interaction of Modality and Negation: A Typo-
logical Study. New York: Garland.
De Haan, F. 1999. Evidentiality and epistemic modality: Setting
boundaries. Southwest Journal of Linguistics. 18, 83-101.
Finegan, E. 1995. Subjectivity and Subjectivisation: an Introduction.
In Stein, D. / Wright, S. (eds) (1995) Subjectivity and subjec-
tivization: Linguistic perspectives. Cambridge: Cambridge
University Press.
Frawley, W. 1992. Linguistic Semantics. New Jersey: Lawrence
Erlbaum Associates.
Givon, T. 1984. Syntax: A functional-typological introduction. Vol-
ume I. Amsterdam: John Benjamins, 307-308.
Heine/Kuteva 2002. World Lexicon of Grammaticalization. Cam-
bridge: Cambridge University Press.
Heine et al. 1991. Granmaticalization: a conceptual framework. Chi-
cago: University of Chicago Press.
Hopper, P.J. / Traugott, E.C. 1993. Grammaticalization. Cambridge:
Cambridge University Press.
Klamer, M. 2000. How report verbs become quote markers and
complementizers. Lingua. 110, 69-98.
Kwok, H. 1984. Sentence Particles in Cantonese. Hong Kong: Cen-
tre of Asian Studies, University of Hong Kong.
Langacker, R.W. 1985. ‘Observations and Speculations on Subjec-
tivity’. In John Haiman (ed.) Iconicity in Syntax. Amsterdam:
Benjamins.
Langacker, R.W. 1990. ‘Subjectification’. Cognitive Linguistics. 1,
5-38.
Langacker, R.W. 1999. ‘Losing Control: Grammaticalization, Sub-
jectification, and Transparency’. In Blank, A. / Koch, P. (eds)
Historical Semantics and Cognition, 147-176, Berlin and New
York: Mouton de Gruyter.
Luke, K.K. 1990. Utterance Particles in Cantonese Conversation.
Amsterdam: Benjamins.
202 Leung Wai-mun
Lyons, J. 1982. ‘Deixis and Subjectivity’. In Jarvella, R.J., and Klein,

W. (eds) Speech, Place ,and Action: Studies in Deixis and Re-
lated Topics. Chichester and New York: John Wiley.
Matsumoto, Y. 1988. From bound grammatical markers to free dis-
course markers: History of some Japanese connectives. In Ax-
maker, S. / Jaisser, A. / Singmaster, H. (eds) Proceedings of the
14th Annual Meeting of the Berkeley Linguistics Society.
Matthews, S. / Xu, H. 2005. Direct and Indirect Speech in Cantonese
and Chaozhou (Southern Min). Chapter presented at RCLT
local workshop, La Trobe Universtiy.
Matthews, S. / Yip, V. 1994. Cantonese: A Comprehensive Grammar.
London and New York: Routledge.
Matthews, S. 1998. Evidentiality and mirativity in Cantonese: wo3,
wo4, wo5! Proceedings of the International Symposium on
Chinese Languages and Linguistics, Academia Sinica.
O’Melia, T.A. 1954. First Year Cantonese. HK Catholic Truth So-
ciety.
Palmer, F.R. 1986. Mood and Modality. Cambridge, England: Cam-
bridge University, 66-67.
Plungian, Vladimir A. 2001. The Place of Evidentiality within the
Universal Grammatical Space. Journal of Pragmatics. 33, 349-
357
Saeed, J.I. 1997. Semantics. Cambridge: Blackwell Publishers.
Sohn, Ho-Min 1999. The Korean Language. Cambridge: Cambridge
University Press.
Traugott, E.C. / Dasher, R.B. 2002. Regularity in Semantic Change.
Cambridge: Cambridge University Press.
Traugott, E.C. / Konnig E. 1991. The Semantics-Pragmatics of
Grammaticalization Revisited. In Traugott, E.C. / Heine, B.
(eds). Approaches to grammaticalization. Amsterdam: John
Benjamins and Co, 189-218.
Traugott, E.C. 1989. ‘On the Rise of Epistemic Meanings in English:
an Example of Subjectification in Semantic Change’. Language.
65, 31-35.
Traugott, E.C. 1995. ‘The Role of the Development of Discourse
Markers in a History of Grammaticalization’. Chapter presented
at ICHL XII, Manchester 1995.
Wisner, O.F. 1906. Beginning Cantonese. Canton: China Baptist Pub-

lication.
Wu, X.Z. 2000. Grammaticalization and the Development of Func-
tional Categories in Chinese. PhD thesis, University of South-
ern California.
Wu, X.Z. 2004. Grammaticalization and Language Change in Chi-
nese: A Formal View. London and New York: Routledge Curzon.
Chinese
Cheung, H.N. ⻝㳒⸜烉˪楁㷗䱝婆婆㱽䘬䞼䨞˫ˤ楁㷗烉楁㷗ᷕ

㔯⣏⬠炻1972 ⸜ˤ
Deng, S. J. 惏⮹⏃烉
S.J. ˨⺋ⶆ㕡妨ⷠ夳䘬婆㯋娆˩ 炻庱˪㕡妨˫ˤ1991
⸜䫔 2 㛇ˤ
Fong, X.Y. 㕡⮷䅽烉˪⺋ⶆ㕡妨⎍㛓婆㯋≑娆˫ˤ⺋ⶆ烉㙐⋿⣏⬠
↢䇰䣦炻2003 ⸜ˤ
Hu, Z.L. 傉⢗湇烉 ˨㻊婆䘬妨㒂⿏␴婆䭯↮㜸˩ ˤ˪㷾⊿⣏⬠⬠⟙˫
䫔 2 㛇炻1995 ⸜炻枩 13-23ˤ
Hu, Z.L. 傉⢗湇烉 ˨婆妨䘬妨㒂⿏˩ ˤ˪⢾婆㔁⬠冯䞼䨞˫䫔 4 㛇炻
1994 ⸜炻枩 9-15ˤ
Leung, C.S. 㠩ẚ㢖烉˪楁㷗䱝婆婆≑娆䘬䞼䨞˫ˤ楁㷗烉楁㷗䎮
ⶍ⣏⬠炻⒚⬠䡑⢓婾㔯炻2005 ⸜ˤ
Li et al. 㛶䲵䫱烉 ˨⽆娙婆奺⹎婾嫱婆㯋娆’䘬’˩ 炻庱˪ᷕ⚳婆㔯˫ˤ
1998 ⸜䫔 2 㛇ˤ
Li, X.H. 㛶㕘櫩䫱烉 ˪⺋ⶆ㕡妨䞼䨞˫ ˤ⺋ⶆ烉⺋㜙Ṣ㮹↢䇰䣦炻1995
⸜ˤ
Liu / Gu ∱泣≯炘栏春烉 ˨㵤Ⱉ⼅婆䘬⺽婆㧁姀␴䣢嫱㧁姀˩ 炻庱
˪㮹㕷婆㔯˫ ˤ2008 ⸜䫔 2 㛇炻枩 16-23ˤ
Ouyang, J.Y. 㫸春奢Ṇ烉 ˨⺋ⶆ娙䘬婆㯋≑娆˩ 炻庱˪䌳≃⃰䓇䲨⾝
婾㔯普˫ ˤ⊿Ṕ⓮⊁⌘㚠炻1990 ⸜ˤ
Rao, B.C. 棺䥱ㇵ䫱烉˪⺋ⶆ娙㕡妨娆℠˫ˤ楁㷗烉⓮⊁⌘㚠棐炻
1981 ⸜ˤ
Shen, J.X. 㰰⭞䃲烉 ˨婆妨䘬’ᷣ奨⿏’␴’ᷣ奨⊾’ ˩ 炻˪⢾婆㔁⬠冯
䞼䨞˫䫔 4 㛇炻2001 ⸜ˤ
204 Leung Wai-mun
Zhang / Yu ⻝ㆸ䤷炘ἁ⃱㬎烉 ˨婾㻊婆䘬⁛ᾉ堐忼˩ ˤ

˪婆妨䥹⬠˫炻
2003 ⸜ 5 㚰炻枩 50-57ˤ
Zhang, B.J. ⻝ỗ㰇烉 ˨娵嬀奨䘬婆㱽堐䎦˩ ˤ
˪⚳⢾婆妨⬠˫䫔 2 㛇炻
1997 ⸜炻枩 15-19ˤ
Zhu, Y.S. 㛙㯠䓇烉 ˨娎婾䎦ẋ㻊婆䘬妨㒂⿏˩ ˤ
˪䎦ẋ⢾婆(⬋↲) ˫
䫔 29 ⌟ 4 㛇炻2006 ⸜炻枩 331-337ˤ
楁㷗婆妨⬠⬠㚫䱝婆㊤枛⫿堐䶐⮓⮷䳬烉 ˪䱝婆㊤枛⫿堐˫ˤ楁㷗烉
楁㷗婆妨⬠⬠㚫ˤ
205
HAN YANG
Tense, Aspect and verbal morphemes

in Chinese1, Japanese and English
1. Introduction
This chapter compares the way in which time information is encoded

in Chinese, Japanese and English. On the basis of this general com-
parison and the conclusion that verbal morphemes – particles and
inflections – do not obligatorily mark time reference, the behaviour
of the le particle in Chinese is examined in a greater detail. It con-
cludes that the le particle signals relative tense in the sense defined
by Comrie, “where the reference point for location of a situation is
some point in time given by the context, not necessarily the present
moment” (1985: 56).
2. Time Expressions in Languages
It is a commonplace that most, if not all languages have various tem-

poral deictic adverbs or temporal particles (cf. Lyons 1977: 679),
and “Certainly most, and possibly all, languages posses time words
and allow their speakers to communicate regarding temporal fea-
tures of experience. Also, context and paralinguistic features prob-
ably would allow implicit temporal references that might not be al-
ready codified in speech.” (Wessman/Gorman 1977: 45). Chinese,
Japanese and English are no exception.
1 Chinese here refers to Mandarin Chinese widely spoken in northern China and
is the national language of the People’s Republic of China.
206 Han Yang
Table 1 below presents a contrastive list of Chinese, Japanese

and English sentences containing time adverbials.
Chinese Japanese English

Ta jingchang fan cuowu. Kare wa yoku machigai wo okosu. He often makes mistakes.
Ta yuanlai shi laoshi Kare wa mottomo sensei datta. He was originally a teacher.
Ta guoqu shenti buhao, Kare wa mukashi yowakatta, He wasn’t in good health
xianzai hai keyi. ima wa mama dasu. before, he’s alright now.
Dangshi ta mei dong, Kare wa sonotouji wakaranakatta, He didn’t understand it then,
keshi xianzai dong le. demo ima wa wakatteiru. but he understands it now.
Ta zongshi hen mang. Kare wa itsumo isogashii. He is always very busy.
Ta mingtian likai Kare wa ashita Hong Kong He is leaving Hong Kong
Xianggang wo tatsu. tomorrow.
Ta qunian biye Kare wa kyonen sotsugyou shita. He graduated last year.
Ta jinnian biye. Kare wa kotoshi sotsugyo suru. He is graduating this year.
Ta ershinianqian you Kere wa nijuunen mae ni okusan He had a wife twenty years
taitai. ga itta. ago.
Table 1.
It is evident from Table 1 that the time reference of the events may
be lexicalized by means of various time adverbials without even con-
sidering how tense and aspect might be grammaticalised in the verb
phrases in these three languages.
Given that both the so-called ‘tensed’ (Japanese and English)
‘tenseless’ (Chinese) languages employ the lexicalization of the time
reference of the events, it is clear that lexicalization, rather than ver-
bal particles or inflections is more basic to the expression of time
reference. Verbal particles and inflections are on the other hand sec-
ondary modifications.
Tense, Aspect and verbal morphemes in Chinese, Japanese and English 207
3. Tense and Verbal Elements
Tense is a means of making linguistic reference to time. Further,

whether a language has tense can be decided only on the basis of a
grammatical analysis of the particular language (cf. Lyons 1977: 678)
In both Japanese and English, the morphological variation in verbal
constructions may indicate either the time reference is past or non-
past as shown by the following table containing contrastive pairs of
examples:
Japanese English
Past (1) Kare wa kyonen sotsugyoo shita (3) He graduated last year.
Non-past (2) Kare wa rainen sotsugyoo suru (4) He is graduating next year.
Table 2.
The deictic time kyonen and the perfective marker -ta in Japanese
(1) indicate the past time reference, and such sentences are generally
regarded as having ‘past tense’ in Japanese. Likewise, the morpheme
-ed and the deictic time adverbial last year in English (3) indicate
past time reference, and such sentences are said to have ‘past tense’
in English.
In the case of Japanese (2), the verb ending -ru, together with
the deictic time adverbial rainen gives the indication that this sen-
tence is of non-past reference. In the same way as Japanese (2), the
copula is and morpheme -ing in English (4), together with the deictic
time adverbial next year indicate that the time reference is one of
non-past.
Propositions equivalent to those expressed by (1)-(4) in Table
2 may be realised in Chinese as follows:
(5) Past:
Ta qunian biye.
He last year graduate
(He graduated last year.)
208 Han Yang
and
(6) Non-past:
Ta mingnian biye.
He next year graduate
(He will graduate / is graduating next year.)
In both sentences, the identical form of biye [graduate] carries no

grammaticalisation of any time reference. The time references are
signaled solely by the time adverbials qunian [last year] and mingnian
[next year]. Therefore, Chinese does not “obligatorily relate the time
of the situation being described to the time of utterance by any system-
atic variation in the structure of the sentence.” (Lyons 1977: 678-679).
One might argue that verbal particles such as le in Chinese
should surely function as an indicator for past tense as supported by
the following example:
(7) Ta zai Dongjing xue le liangnian Riwen.

He at Tokyo study le two-years Japanese
(He studied Japanese in Tokyo for two years.)
Other particles such as zhe signal progressive as shown by:
(8) Waimian gua zhe dafeng.

Outside blow zhe big-wind
(Strong wind is blowing outside.)
However, the following examples reveal that such verbal particles

are neither necessary nor sufficient as signals of tense. We shall look
at these two morphemes in turn:
Zhe does not necessarily indicate progressive:
(9) Ni zai zher deng zhe.

you at here wait zhe
(You wait here.)
The progressive can be expressed without zhe:
(10) Ta zhengzai shangke.

He in-the-process-of attending-class
(He is teaching (at the moment).)
Le does not necessarily indicate past:

(11) Ta chi le fan jiu zou.
He eat le food then leave
(He is leaving/will leave as soon as he has had some food.)
Past time reference does not have to be indicated by le:

(12) Ta yiqian shi laoshi
He in-the-past be teacher
(He used to be a teacher.)
From the above it is clear that Chinese does not have obligatory tense
marking as in Japanese and English.
This can be further evidenced by the following unacceptability
caused by the conflict between the time adverbials and tense mark-
ers in Japanese and English (examples (13)-(16)), as well as the ac-
ceptability of all time adverbials by an identical verb in Chinese with-
out any verbal particles (examples (17)-(18)):
The ungrammaticality of (13) and (14) below is caused by the
conflict between future time reference and morphemes that signal
non-past time reference:
(13) *Kare wa lainen sotsugyo shita.
He topic-prt next-year graduate past-marker
(14) *He graduated next year.
The following ungrammaticality is caused by the incompatibility

between past time reference and non-future tense markers:
(15) *Kare wa kyonen sotsugyo suru.
He topic-prt last-year graduate inf-marking
(16) *He will graduate /is graduating last year.
By contrast, Chinese does not exhibit this type of conflict at the gram-
matical level:
(17) Future time reference:
Ta mingnian jiu biye le.
He tomorrow then graduate le
210 Han Yang
(18) Past reference:

Ta qunian yijing biye le.
He last-year already graduate le
What then is the function of the verbal particles in Chinese? Accord-

ing to Li (1999: 38) the main function of verbal particles is to indi-
cate notion of time, therefore sentences involving stative verbs such
as xing [surname] cannot co-occur worth verbal particles:
(19) a *Ta xing le Auyang.

He surname le Auyang
b *Ta xing zhe Auyang.
He surname zhe Auyang
Below is another pair of examples supporting the same point involv-

ing the stative verb shi [to be]:
(20) a *Ta shi le Zhongguoren.

He be le Chinese
b *Ta shi zhe Zhongguoren.
He be zhe Chinese
This point is further elaborated in Section 5 on the concept of state

and its relationship with verbal morphemes.
4. Aspect of event2 and verbal morphemes
Events can be described from different viewpoints: some are just

beginning to come into existence, and others coming to an end; some
are momentary, and some continuous. Some situations are always
viewed as being constant, while others are viewed as temporary. The
2 Discussions in this and the following section are based on Unit 3 ‘Tense, as-
pect, reference’ in the course material entitled Introduction to Semantics and
Pragmatics, pp. 21-26 written by the present author and published by the Open
University of Hong Kong (1999) and (2001). Reprinted (2002).
expressions describing any of these viewpoints is called aspect, the

‘internal contour of an event’ (Frawley 1992, 294).
Hofmann (1993: 154) distinguishes four types of aspect –
punctive, durative, completive and iterative – based on the way in
which the English language views the F (finish) point of the event.
Hofmann states that “English seems to concentrate on the finishing
of an event” (1993: 154), as illustrated by the unacceptability of the
second sentence, as contrasted with the first:
(21) I will ring you when I finish watching this TV programme.
(22) *I will ring you when I finished watching this TV programme.
The event of ‘watching this TV programme’ in the first example will

have to precede the event of the speaker’s ‘ringing you’. The comple-
tive meaning, i.e. the F point is already included in the verb ‘finish’
itself. The unacceptability of the second example is caused by the
addition of the past tense marker -ed to the verb ‘finish’. Even though
the event of watching the TV programme may be a past event rela-
tive to the speaker’s ringing, the event of watching the TV programme
has to precede the event of speaker’s ringing. The use of the past
tense marker in this sentence is unacceptable for at least two rea-
sons: (i) the speaker is referring to both events as occurring in some
future time; and (ii) the completive meaning of ‘finish’ is inherent in
the verb itself, and therefore the addition of the past tense marker is
not only superfluous but a cause for ungrammaticality.
By contrast, in Japanese, the perfective marking -ta is compul-
sory, as shown below:
(23) Kono bangumi wo mita atode kare ni denwa suru.

This TV-programme prt watch afterwards he prt telephone inf
(I’ll call him after watching this TV programme.)
This indicates that in Japanese the focus of the verbs such as ‘finish’
may well be on the B (beginning) point. Therefore a marker that
specifically indicates one event that precedes another is needed to
indicate the relative order of events.
212 Han Yang
The verbal morpheme -ta in Japanese can therefore be said to

be a relative tense marker rather than a marker for past tense such as
-ed in English. A similar conclusion is reached in Ogihara (1999:
330), who states that “the morpheme -ta is clearly a tense morpheme
although it is a relative one unlike -ed in English.”
The meaning equivalent of the above examples (21) and (23)
in Chinese is:
(24) Wo kan wan zhege dianshi jiemu zai gei ta da dianhua.

I watch finish this TV programme then give him call
(I’ll give him a call when I finish watching this TV programme.)
The substitution of wan [finish] with le, same as the presence of the
past tense marker -ed in English (22) above, causes ungrammaticality:
(25) *Wo kan le zhege dianshi jiemu zai gei ta dianhua.

I watch le this TV programme then give he call
However, this does not imply that the le particle does the same thing
as the English -ed to mark the past tense. This is supported by the
fact that le can actually co-occur with lexical verbs that signal comple-
tion, as in:
(26) Wo kan wan le zhege dianshi jiemu zai gei ta da dianhua.

I watch finish le this TV programme then give him call
(I’ll give him a call when I finish watching this TV porgramme.)
Therefore, the le particle is unlikely to be a marker of past time ref-

erence or past tense.
5. Event types and verbal morphemes
Neither stative nor activity verbs have an end point F, and so they are
called atelic verbs. Whereas the notion of atelic denotes an event
without an F point, telic concentrates on the end result of an event.
Achievement and accomplishment verbs belong to this category. The
classification of telic and atelic goes back to the time of Aristotle. In

modern studies of event types and verb classes, these terms are often
used to apply to the whole sentence as either expressing an event
with or without an end point F.3
All events have a B-beginning and an F-finishing point, but
the focus on either of these points varies from language to language.
Atelic and telic contrast is made according to the latter making spe-
cific reference to the F point of the event. The former lacks this fo-
cus, but neither makes reference to the B point of the event. There-
fore, achievement and accomplishment verbs belong to telic for their
focus is on the end point F, while activity and stative verbs belong to
atelic for their lack of focus on the end point F.
Given the above reasoning, it has been more straightforwardly
assumed in Chinese that there is a two-way verb classification, namely
“the verb class of non-change, including verbs of states and activi-
ties, and the verb class of change, subsuming accomplishments and
achievements” (Liu 1997: 3)
We now look at some event types in relation to verbal mor-
phemes in Chinese, Japanese and English in more detail.
5.1 State
A state is described by a lexical item that denotes a condition that is

unchanging. Most commonly nouns denote objects such as city, and
adjectives describe qualities such as busy and beautiful.
Both something or someone and the qualities attached to them
are constant, and are therefore without apparent B and F points. The
situation is just the same with verbal items. Verbs denote states that
do not accept the progressive marking in English.
This type of verb is termed ‘stative’, as it denotes situations
that seem to the speakers to be without the B or F point. These typi-
cally include verbs that denote mental states such as ‘like’ and ‘live’.
3 Readers interested in this aspect of the study can refer to Hofmann’s discus-
sion on Limits to Events (1993: 139-158) with special reference to aspect of
events and verb classes.
214 Han Yang
For example:
(27) *I am loving you.
(28) *He is living in Homantin.
Chinese behaves in much the same way as English. Stative verbs

such as ‘love’ and ‘live’ denoting continuous state in Chinese do not
accept progressive marking either, as exemplified by the following
examples containing zhe:
(29) Wo ai ni.
I love you
(I love you.)
(30) *Wo ai zhe ni

I love zhe you
(31) Wo zhu Xianggang.

I live Hong Kong
(I live in Hong Kong.)
(32) *Wo zhu zhe Xianggnag.

I live zhe Hong Kong
The ungrammaticality of the two starred sentences is caused by the

presence of the progressive marker, the zhe particle.
Unlike English and Chinese, verbs expressing continuous state
in Japanese on the other hand not only accept progressive marking
-te iru4, but it necessarily requires such marking, as in:
(33) Anata wo aishite iru.

You obj-prt love-te iru
(I love you.)
(34) Watashi wa Hong Kong ni sunde iru.

I topic-prt Hong Kong loc-prt live-te iru
(I live in Hong Kong.)
4 There is a classical study on verb classification in Japanese using a -te iru test
by Kindaich (1955), where this verbal suffix is characterised as expressing
whether progressive or perfect meaning depending on the lexical meaning of
the main verb to which is it attached.
Even when the sentence does not express the present state, this class
of verbs in Japanese still requires progressive marking as in:
Anata wo aishite ita. (past progressive)

You obj-prt love-te ita (-te iru past progressive)
(I loved you)
Watashi wa Hong Kong ni sunde ita. (past progressive)
I topic-prt Hong Kong loc-prt live-te iru (-te iru past progressive)
(I lived in Hong Kong.)
Whereas English in this case takes simple past, as in:
(37) I loved you
and not past progressive:
(38) *I was loving you.
5.2 Activity
Activity verbs are regularly used to describe activities. Activities occur

throughout a continuous period without indication of whether the activ-
ity is completed. Verbs such as ‘read’, ‘sleep’ belong to this category.
(39) Chinese: Ta kan le sange zhongtou de shu.

He read le three hours nom-prt book
(He read the book for three hours.)
(40) Japanese: Taro wa hon wo sanjikan yon da.

Taro topic-prt book obj-prt three-hours read past-marker
(Taro read the book for three hours.)
(41) English: John read War and Peace for three hours.
All of the above three examples from Chinese, Japanese and English
do not tell us if someone has finished reading the book; it simply
indicates that the activity of reading took place, i.e. someone did
some reading, and that reading lasted for three hours.
216 Han Yang
All activity verbs share the same characteristic, that is, with-
out implication of an end point. It is therefore only natural that all
three languages should accept the progressive marking in the case of
activity verbs. Below are some examples:
(42) Ta kan zhe shu.

She is reading.
Taro wa hon wo yonde iru.
(43) Tamen taolun zhe caizheng wenti.

They are talking about the budget.
Karera wa zaimu mondai ni tsuita hanashite iru.
Also there is no conflict between the past tense marking in Japanese

and English, and the le particle in Chinese with this class of verbs
either:
(44) Tamen taolun le caizheng wenti.

They talked about the budget
Karera wa zaimu mondai nitsuite hanashita.
5.3 Achievement
Achievement verbs focus solely on the end point or completion of an

event, and they are essentially without duration. These include verbs
like ‘kick’, ‘arrive’, ‘open’, etc. These verbs also indicate a change
from one situation to another; thus if one arrived at the airport at
3 pm this would imply that one was not at the airport before 3 pm.
Below are some examples:
(45) Ta sandian dao jichang.

he 3o’clock arrive airport
(He arrives at the airport at 3 o’clock.)
(46) Taro wa sanjini kuukou ni tsuku.

Taro topic-prt 3 o’clock airport loc-prt arrive
(Taro arrives at the airport at 3 o’clock.)
(47) John arrives at the airport at 3pm.

The implication is that ‘he/taro/John was not at the airport before

3 pm’ in all of the above three cases.
Because of these characteristics, achievement verbs denote the
punctive aspect of event in all three languages. The following are a
few more examples:
(48) Ta wudian zou.

he 5 o’clock leave
(He leaves at 5 o’clock.)
(49) Taro wa goji ni tsuku.

Taro topic-prt 5 o’clock temp-prt arrive
(Taro leaves at 5 o’clock.)
(50) John leaves at 5pm.
However, there is a marked difference between these three languages

in relation to the achievement verbs when the progressive marking is
present.
In English, when an achievement verb is combined with the
progressive ending -ing, it indicates that the event denoted by the
verb has not yet been realised, as in:
(51) John is arriving at the airport (in two hours’ time)

(Implication: John is not at the airport now.)
Whereas in Japanese, when the progressive marking is present, it

indicates that the event denoted by the achievement verb has been
realised, as in:
(52) Taro wa kuukou ni tsuite iru.

Taro topic-prt airport loc-prt arrive
(Taro has already arrived at the airport, and he is now at the airport.)
This pair of examples supports the assumption made in Section 4

that English verbs focus on the end point F and Japanese the starting
point B.
218 Han Yang
Schematically, the opposing phenomenon in both languages

can be represented as follows:
Point at which Taro arrives at the airport
AIRPORT
English: John is arriving Japanese: Taro tuite iru

(and has not arrived yet.) (and he is at the airport now)
Figure 1.
Achievement verbs in Chinese on the other hand do not accept pro-

gressive marking:
(53) *Ta dao zhe jichang.

He arrive zhe airport
This is because the presence of the progressive marker zhe is incom-

patible with the achievement meaning of the verb dao [arrive].
In order to signify whether someone has or has not yet arrived
at the airport, other semantic means such as mei [not yet] has to be
employed. As in the following example:
(54) Ta hai mei dao jichang.

He yet not arrive airport
(He has not yet arrived at the airport.)
This sentence shows that the person has not yet arrived at the airport,
where mei is a negative item signalling the non-realisation of an event
in Chinese.
The sentence below is of interest: the le particle in post-verbal
position is optional, whereas the le particle is necessarily required at
the sentence-final position.
(55) Ta yijing dao (le) jichang le.

He already arrive (le) airport le
(He has already arrived at the airport.)
It may thus be said that the le particle here behaves in a similar man-
ner as the progressive marker -te iru in Japanese.
This can be seen clearly by utilising the same Figure 1 above
with the addition of the Chinese equivalents as shown below:
Point at which Taro arrives at the airport
AIRPORT
John is arriving Taro mou tuite iru

(and has not arrived yet.) (and he is at the airport now)
------------------------------------------------------------------
Ta hai mei dao jichang. Ta yijing dao jichang le.
(Implying: He is arriving. (Implying he has arrived
and has not arrived yet.) and he is at the airport now)
Figure 2.
5.4 Accomplishment
Like achievement verbs, accomplishment verbs also include the

F point. But they differ from the achievement verbs in that they do
not signal the punctive aspect of the event but rather depict the event
as occurring over a period. Often they imply that a certain amount of
effort was exerted by the subject during this period which was re-
quired. For example:
(56) Jeff learned to ride the bicycle in three days.
(57) Mary taught her son how to speak Japanese.

220 Han Yang
Learning a skill and teaching a language both require time and ef-
fort, and in this sense, the accomplishment verbs are durative as op-
posed to the punctive characteristic of the achievement verbs.
Similarly, verbs such as narau [learn] in Japanese do not have
the implication of attainment. In order to achieve the attainment mean-
ing of ‘learned how to …’, specific verb dekiru with -ta ending is
employed, as in:
(58) Taro wa mikkakan de jidensha ga dekita.

Taro prt three-days prt bicycle prt know-how-to
(Taro learned to ride a bicycle in three days.)
Likewise, a simple verb oshieru [teach] in Japanese does nto have

the implication of ‘taught someone how to…’, but simply the activ-
ity of teaching. In order to specify that ‘Mary taught her son Japa-
nese and her son is now able to speak Japanese’, we need to con-
struct something like the following:
(59) Mary wa musuko ni Nihongo wo oshieta. Sono kekka, musuko wa

Nihongo ga dekiru you ni natta.
(Mary taught her son Japanese, and, as a result, her son can now speak Japanese.)
Or else, something like the following:
(60) Mary no musuko wa Nihongo ga dekiru. Sore wa Mary ga oshieta kara.

(Mary’s son can speak Japanese. This is because Mary taught him how.)
“Accomplishment verbs in English, when in past or perfect tenses,

necessarily imply an attainment of the goal” (Tai 1984: 291). The
equivalents in Japanese as exemplified by above examples do not
have such an implication.
The achievement meaning in Chinese on the other hand is usu-
ally achieved by using resultative compliment such as hui [able] in
combination with a simple verb such xue [to learn], as in:
(61) Ta xue le sannian Riyu, xianzai xuehui le.

he lean le three-years Japanese, now learn-resultative le
(He learned Japanese for three years, and now he knows it.)
Other examples are zhaodao [to find = look for + have found], shasi
[to murder = kill + have died], tingdong [to understand = hear + have
understood], and so on.
When the achievement verb in Chinese is a simple verb, such
as cheng [become], si [die], the le particle is necessarily required,
as in:
(62) Ta cheng le mingxing.

he become le star
(He became a star)
(63) Ta si le
he die le
(He died)
The achievement meaning and the function of the le particle there-

fore are in accord.
An extended discussion and a unified account on the function
of the le particle are presented below.
6. The le particle5
In the preceding sections we have presented a general contrastive

analysis of the verbal morphemes in Chinese, Japanese and English
in relation to time reference and the aspectual properties of different
verbal categories.
We shall devote the remaining part of this chapter to an alter-
native, unified treatment of the le particle. This particle is found not
only in post-verbal position, but also in post-adjectival, post-nominal,
as well as sentence-final positions.
5 Section 6 is largely a reproduction of the article ‘A case for relative tense in

Chinese’ by the present author published in Shao and Luke (eds) (2005) with
some amendments.
222 Han Yang
6.1 General treatment on le
Le is often treated as an aspect marker together with guo, zai, zhe, and
sometimes, ne, within the study of the temporal system of Chinese
(e.g. Wang 1965, Chao 1968, Rohsenhow 1977, Li/Thompson 1981,
Li/Thompson /Thompson 1982, Chu /Chang 1987, Huang 1988,
Huang/Davis 1989, Smith 1991, He 1992, Campbell 1992, Mangione/
Li 1993, et al.). Le is also treated by some scholars as having the
function of marking both tense and aspect (Li 1998), and certain
occurrences of le (e.g. as in You xiwang le.) have even been excluded
from the examination of the temporal system altogether for being a
‘mood’ particle (Li 1998). Also when le comes after a verb at the end
of the sentence, it is claimed to be “difficult to determine whether it
is the perfective verb suffix -le, or the CRS [currently relevant state]
sentence-final particle le”. (Li/Thompson 1981: 296). Nevertheless,
most scholars agree that whenever there is an involvement of le in a
sentence, a sense of sequential ordering of events is detected.
6.2 The notion of tense
Some scholars have in the past suggested that the le particle, when
occurring in post-verbal position, signals the past tense (Gao 1970),
but this position was challenged by many who claim that the post-
verbal le marks perfective aspect. (e.g. Wang 1965, Chao 1968,
Rohsenhow 1977, Li/Thompson 1981, Li/Thompson/Thompson 1982,
Chu/Chang 1987, Huang 1988, Huang/Davis 1989, Smith 1991, He
1992, Campbell 1992, Mangione/Li 1993, et al.).
It is indeed the case that Chinese does not obligatorily relate
the time of situation being described to the time of utterance by any
systematic variation in the sentence structure (cf. Lyons 1977), and
the le particle can and does occur in sentences where non-past time
reference is clearly indicated:
(64) Tamen chi le fan zou.

They eat le food leave
(They will leave after they’ve eaten.)
Nevertheless, most scholars would agree that whenever there is an

involvement of le in a sentence, a sense of sequential ordering of
events is detected as in the above example, where the event of leav-
ing follows the event of eating.
6.3 The temporal ordering function of le
We shall therefore argue, on the basis of an understanding of the

notion of temporal journey – location, direction, movement and bor-
der crossing (Jessen 1973) – that le signifies ordering of events (whether
explicitly stated or presupposed) irrespective of the syntactic position
of this particle – post-verbal, post nominal, post-adjectival, sentence-
final, combined sentence-final and post-verbal or post-adjectival.
We shall further argue that in a sentence lacking explicit ex-
pressions signalling the event order, le is indispensable for indicat-
ing the ordering of the stated and presupposed events. And when
there is no presupposition of border and border crossing, le would
not be used even though the sentence may be well formed syntacti-
cally. Le, therefore indicates a reference point, and that reference
point for location of a situation is some point in time given by the
context, not necessarily the present moment. This conclusion is in
accordance with Comrie’s (1985) definition of relative tense.
We shall also examine the so called indeterminate cases, that
is, when le comes after a verb at the end of the sentence such as:
(65) Xiao Wang zou le.

Xiao Wang leave le
(Xiao Wang has gone/left.)
and
(66) Xiao Wang shui le.

Xiao Wang sleep le
(Xiao Wang has gone to bed/sleep.)
and provide explanations to the effect that the function of le, when
occurring in this position also signifies a transition in border-cross-
224 Han Yang
ing journey irrespective of the relation between the speech time and
the reference time, i.e. the relative tense.
6.4 The Concept of Temporal Journey
The concept of temporal journey proposed by Jessen (1973) involves

the following notions:
Location
Direction
Movement
Border-crossing
An expression such as Egor crossed the French-Italian border can

be represented schematically as follows:
A B
Figure 3. Border-crossing journey.
A and B are contiguous locations, and the arrow (à) is the component
of directed movement, the process of which is an essential compo-
nent of journey events.
Expressions focussing on only one stage or phase of a journey such
as Egor left France (i.e. X left location A) or Egor reached Italy (X reached
location B) may be represented by Figure 4 and Figure 5 below:
A B
Figure 4. X left location A.
A B
Figure 5. X reached location B.

If the remaining states in Figures 4 and 5 are not specified or retriev-

able from the context, “it is again a case of simple border-crossing
with only the initial or the final state specified, the other state being
implicitly specified as the negation of the other.” (Jessen 1973: 115).
6.5 The le constructions
There are four positions in which le occurs:
a. Post-verbal position:
(67) Xiao Wang chi le fan.

S – Vt – le –O
b. Sentence-final position:
(68) Xiao Wang chi fan le.

S – Vt – O – le
c. Both post-verbal and sentence-final positions:

(69) Xiao Wang chi le fan le.
S – Vt – le – O – le
d. Post-verbal / sentence-final position:

S –Vi – le
We shall examine each type in turn.
6.6 The post-verbal le
This construction accepts only past time reference:
(71) Xiao Wang yijing chi le fan.

Little Wang already eat le food
(Xiao Wang already had some food.)
226 Han Yang
(72) * Xiao Wang xianzai chi le fan.

Little Wang now eat le food
(73) * Xiao Wang kuaiyao chi le fan.

Little Wang about-to eat le food
This indicates that the event described must have taken place at some
point before the speech time and that the stated event no longer ex-
ists at the time of utterance.
The post-verbal le construction can however allow another
clause to follow:
(74) Xiao Wang chi le fan jiu zou.

Little Wang eat le food then leave
(Xiao Wang will have some food before leaving. / Xiao Wang will leave after
some food.)
This indicates that the stated event is to be followed by some other

event, either specified as in the above example, or unspecified as in:
(75) Xiao Wang chi le fan.

Little Wang eat le food
(Xiao Wang had some food.)
There are two things that are apparent: (a) Xiao Wang ate some food
and (b) Xiao Wang is no longer eating.
The post-verbal le construction can therefore fit into Jessen’s
three-state-journey (Figures 4 and 5) in the following manner:
Xiao Wang’s eating
Figure 6. Border-crossing journey and post-verbal le construction.
The difference between Figures 4 and 5 (Jessen’s representations)

and Figure 6 is that in the case of the former, both the initial location
A and the final location B are in focus, whereas in the case of post-
verbal le construction, the event being focussed upon occupies the
intermediate state, indicating a specific state of Xiao Wang’s eating.
The final state of the journey is pragmatically presupposed,

that is, the use of Xiao Wang chi le fan (Xiao Wang ate something) is
based on the speaker’s assumption that Xiao Wang is no longer eat-
ing. If Xiao Wang’s eating becomes false, then Xiao Wang’s not eat-
ing will automatically be true, and this condition must be satisfied if
the speaker is to use the sentence Xiao Wang chi le fan. This relation-
ship can be represented as follows:
Xiao Wang’s eating [XiaoWang’s not eating]
Figure 7. Stated and presupposed states in post-verbal le construction.
We can now use a generalised symbol E to signal the stated E[vent]

of Xiao Wang’s eating, and the negation of E, i.e. [~E] to signal the
presupposed event of Xiao Wang’s not eating. Thus:
E [~E]
Figure 8. Stated and presupposed states.
E in figure 8 is the only state of the journey that is in focus, indicat-

ing a specific state of Xiao Wang’s eating; [~E] is a presupposed final
state of the journey, representing the state of Xiao Wang’s having
stopped eating, that is, Xiao Wang exited from state E. The dividing
line, i.e. the border between E and [~E] is an arbitrary point at which
Xiao Wang stopped eating. The cessation of E implies that Xiao Wang
entered state [~E], i.e. the inception of the state of Xiao Wang’s hav-
ing stopped eating. Thus, we can conclude that the post-verbal le has
a cessative meaning that implies inception.
228 Han Yang
In relation to speech time, Figure 8 may be modified as fol-

lows:
ST
--- --- --- --- --- --- ---

time
E [~E]
Xiao Wang’s eating Xiao Wang’s not eating
Figure 9. Post-verbal le construction and speech time (ST).
Figure 9 represents typical post-verbal le sentences that have unspeci-

fied reference time. When no reference time is explicitly stated as in
the case of Xiao Wang chi le fan, the reference time will be taken to
be present, and because the post-verbal le signals cessation before the
reference time, it will, therefore, in such a case, be taken as past.
This position is supported by the fact that Chinese does not
accept the presence of the present or future time reference in post-
verbal le constructions as exemplified in (72) and (73) above.
This position can be further supported by the unacceptability
of (76) below, as its second clause presupposes a continuation of the
event of Xiao Wang’s eating lunch stated in the first clause:
(76) *Xiao Wang chi le wufan, xianzi haizai chi wufan.

Little Wang eat le lunch, now still eat lunch
Clearly the presence of the cessative meaning contained in the post-

verbal le prevents (76) from achieving grammaticality.
It is therefore clear that whether a post-verbal le construction
has specific reference time, the basic requirement for such construc-
tion is the presence of a supposed border between the state E and the
state [~E], either specified or implied. When there is no implication
of a negative state [~E], the border then would not exist; consequently,
expressions such as (72), (73) and (76) are ungrammatical.
The primary function of the post-verbal le is therefore to indi-
cate the ending of an event that precedes some other event – either
specified by another clause as in (74), or implied as in (67) – that is

separate from the event specified by the post-verbal le clause.
Furthermore, when the speech time (ST) is taken as the refer-
ence point in the absence of specific reference time, the post-verbal
le marks past, as the cessation signalled by the post-verbal le has to
precede the ST.
6.7 The sentence-final le
Unlike the post-verbal le construction, the sentence-final le construc-

tion does not allow a clause that explicitly indicates that some other
event follows the event expressed by the sentence-final le sentence:
(77) *Xiao Wang chi fan le jiu zou.

Little Wang eat food le then leave
This implies that the event stated in the le clause Xiao Wang chi fan
le has not come to an end; therefore, the inception of another event
as stated in the second clause cannot take place. The speaker of (68)
is then saying two things: (a) Xiao Wang began eating, and (b) Xiao
Wang has been eating since, with a presupposed meaning of Xiao
Wang stopped being in the state of not eating.
The acceptance of non-past reference time by the sentence-
final le construction as exemplified by (78) below allows us to inter-
pret that this sentence has a meaning of Xiao Wang will begin eating
with a presupposed meaning of Xiao Wang will stop being in the
state of not eating:
(78) Xiao Wang yao chi fan le.

Little Wang about eat food le
(Xiao Wang is about to have some food.)
If we use the same negative [~E] to represent the presupposed state

of not eating, and a dividing line that signifies the border between
being in the sate of not eating and the inception of the state of eating,
we can represent the sentence-final le construction in the following
manner:
230 Han Yang
[~E] E
Xiao Wang’s not eating Xiao Wang’s eating
Figure 10. Sentence-final le construction and temporal journey.
The interpretation of Figure 10 will be that the event of Xiao Wang’s

eating has taken place after the presupposed event of Xiao Wang’s
not eating. That is, if the negation becomes false, then the explicitly
stated event must have started. And in this sense, the sentence-final
le construction is comparable to the post-verbal le construction in
that they both focus on one state of the journey. The difference is that
the presupposed state in the case of the former is the initial state,
whereas it is the final state in the case of the latter.
Figure 9 shows clearly that like the post-verbal le construc-
tion, the sentence-final le construction is a case of border-crossing,
and whereas the way in which the cessative meaning of the post-
verbal le implies an inception, the inceptive meaning of the sentence-
final le implies cessation. In our present case, it is the cessation of
the state of Xiao Wang’s not eating.
The inceptive meaning of the sentence-final le thus enables us
to interpret sentence (68) to mean both (a) Xiao Wang started eating,
and (b) Xiao Wang is now engaged in the activity of eating. The in-
ception of the state of Xiao Wang’s eating presupposes a state of Xiao
Wang’s not eating. Therefore, we conclude that (68) has the meaning
of Xiao Wang began eating, thus supporting the position that the sen-
tence-final le indicates the change of the state of affairs.
Further, as we shall argue in the next section, the le particle
signals ordering of events, i.e. the relative tense, where the reference
point for location of a situation is some point in time given by the context,
not necessarily the present moment (see Comrie 1985: 56). Therefore,
it is not surprising that the sentence-final le construction accepts not only
past, but also present and future time expressions as illustrated below:
(79) Xiao Wang yijing chi fan le.

Little Wang already eat food le
(Xiao Wang has already begun eating.)
(80) Xiao Wang xianzai chi fan le.

Little Wang now eat food le
(Xiao Wang has begun eating now.)
(81) Xiao Wang yao chi fan le.

Little Wang about eat food le
(Xiao Wang is about to have some food.)
Finally, since an event cannot simultaneously be both beginning and

ending, the inceptive meaning of the sentence-final le is incompatible
with another clause such as a jiu-clause (see example (77) above)
which suggests a cessation of an event that has to take place before the
event specified by that other clause. Compare (77) with (82) below:
(82) Xiao Wang chi fan le, xianzai hai zai chi ne.
Little Wang eat food le, now still at eat ne
(Xiao Wang started eating, and he is now still eating.)
6.8 Post-verbal le and the sentence-final le co-occurring

in the same sentence
This is a construction that contains both the meaning of post-verbal le

and sentence-final le. We shall look at both in turn. As the cessative
meaning of post-verbal le is contained within this construction, the
constraint against future time reference on the post-verbal le construction
also applies to this type as exemplified by the unacceptability below:
(83) *Xiao Wang yao chi le fan le.

Little Wang about eat le food le
The meaning of post-verbal le part is therefore consistently that Xiao

Wang ate something and he is no longer eating.
Sentence-final le, like other sentence-final particles, is a sen-
tence operator, and therefore it covers the entire scope of the sentence.
The interpretation of the meaning of (69) would therefore take the
following form, where the parenthesised material in the scope of the
sentence-final le corresponds to the meaning of the post-verbal le sen-
tence (67): (Xiao Wang ate something, and he is no longer eating) le.
232 Han Yang
The meaning of (69) therefore has two parts: (a) Xiao Wang
ate something, as indicated by the past-verbal le; and (b) the state of
Xiao Wang’s having stopped eating has begun, as indicated by the
sentence-final le, as represented below:
E [~E]
[~E] E
Figure 11. Post-verbal le and sentence-final le construction and temporal journey.
As with other types of the le construction, when there is specific

past-time reference such as yijing [already] present, this sentence
type indicates that the border-crossing took place sometime prior to
the speech time (ST).
6.9 Summary
All three types of the le construction we have looked at so far are

border-crossing journeys. Border-crossing journeys share some of
the characteristics of achievement verbs, which are considered mini-
mal journeys, i.e. border-crossing journeys (see Jessen 1973: 117).
The difference between the border-crossing of achievement
verbs and the border-crossing of the le-constructions is however that
while the former may or may not be ended at some arbitrary point as
illustrated by the following contrastive pair, the function of the le
particle always indicates some arbitrary point of beginning (in the
case of sentence-final le), or some arbitrary point of ending (in the
case of the post-le).
(84) Egor crossed the border.
(85) Egor crossed the border at 1:00pm. (Jessen 1973: 121)

The table below summarises the discussion so far:
Type Post-verbal le Sentence-final le Both post-verbal le and sentence-

final le
Function Cessation Inception Cessation and inception
Border-
crossing E [~E] [~E] E E [~E]
journey [~E] E
Meaning 1. Xiao Wang ate 1. Xiao Wang 1. Xiao Wang began eating;
something; began eating; 2. Xiao Wang is no longer eating
2. Xiao Wang is no 2. Xiao Wang has now.
longer eating now. been eating since.
Trans- Xiao Wang ate Xiao Wang has Xiao Wang began eating, but he
lation something and he is begun eating. is no longer eating now.
no longer eating now.
Table 3. Summary of three types of le construction.
6.10 Le occurs in the position that is both post-verbal

and sentence-final
It has long been accepted that when le occurs in a position that is both
sentence-final and post-verbal, the interpretation of the resulting sen-
tence is ambiguous, as observed by Li and Thompson (1981: 296):
When a le comes after a verb at the end of a sentence, it is difficult to determine

whether it is the perfective verb suffix -le or the CRS [current relevant state]
sentence-final particle -le.
We shall suggest in the following that the function of le in this type

of construction is, like in the above three cases, to signify a transition
in border-crossing journey, irrespective of the relation between the
ST and the reference time.
We shall also argue that the apparent ambiguity of this type of
le construction can be clarified by the notion of temporal journey.
The typical occurrence of the le particle in this position is where
intransitive verbs are involved, as in:
234 Han Yang

Little Wang leave le
(Xiao Wang has gone/left.)
The achievement verb zou [to leave], in combination with the func-
tion of the ‘order of events’ of the le particle indicates that a transi-
tion (i.e. border-crossing) from Xiao Wang’s being at a place to his
not being at that place has taken place at some point prior to the ST.
In other words, up to the point of transition, Xiao Wang was at loca-
tion E, at some point in time prior to the ST a transition took place
and that transition is no longer in the process of happening, and that
resulted in him being not in location E; so he is not there in location E
at the ST.
Graphically, this transition in relation to ST can be represented
as follows:
ST
--- --- --- --- --- --- ---
time
E [~E]
Xiao Wang’s at Xiao Wang’s no

location E longer at location E
Figure 12. Xiao Wang zou le and speech time (ST).
This type of le construction allows both the future and past time ref-
erences as exemplified by (87) and (88), and can be represented by
Figure 13 and Figure 14 below:
Past:
(87) Xiao Wang zuotian zou le.
Little Wang yesterday leave le
(Xiao Wang left yesterday.)
Future:
(88) Xiao Wang mingtian zou le.
Little Wand tomorrow leave le
(Xiao Wang is leaving tomorrow.)
Yesterday Today
time ST
E [~E]

Figure 13. Xiao Wang zuotian zou le and speech time (ST).
Today Tomorrow
time ST
E [~E]

Figure 14. Xiao Wang mingtian zou le and speech time (ST).
Figure 13 shows that the border-crossing took place sometime yes-

terday, and figure 14 shows that the border-crossing will take place
sometime tomorrow.
Therefore, when le occurs in the position that is both post-
intransitive verbal and sentence-final, it is clearly a cassative le.
6.11 The ‘ambiguous’ le
Another case involving le in this position is when the object of a

transitive verb has been preposed to the beginning of the sentence,
exposing the le particle at the sentence-final position, as in
(89) Shu kan le

book read le
236 Han Yang
Does this mean that this is ambiguous? Indeed it has any of the three
possibilities:
(a) kan le shu

read le book
(b) kan shu le
read book le
(c) kan le shu le
read le book le
The third possibility exists as Chinese syntax does not allow the se-
quence of two identical particles in a sentence.
This long accepted syntactic ambiguity is in fact invari-
ably dissolved by a given context as demonstrated below:
(a) When le is post-verbal, it occurs in a context comparable to (90):
(90) Naben shu kan le jiuxing, buyong bei chu lai.

That book read le then fine, no need memorise-resultative
i.e. kan le naben shu juxing, …
read le that book then fine
V le O
It cannot be sentence-final
*kan naben shu le juyxing, …
Read that book le then fine
V O le
(b) When le is sentence-final, it occurs in a context comparable to (91):
(91) Naben shu wo henkuai jiuyao kan le.

That book I very-soon about real le
i.e. Wo henkuai jiuayo kan naben shu le.
I very-soon about read that book le
V O le
It cannot be post-verbal:
* Wo henkuai jiuayo kan le naben shu.
I very-soon about read le that book
V le O
(c) When le is a combination of post-verbal le and sentence-final le,

it occurs in a context comparable to (92):
(92) Naben shu wo yijing kan le.

That book I already read le
i.e. Wo yijing kan le naben shu le.
I already read le that book le
V le O le
And it cannot be post-verbal:
* Wo yijing kan le naben shu.
I already read le that book
Nor can it be sentence-final:
* Wo yijing kan naben shu le.
I already read that book le
This le construction is therefore not as indeterminate as commonly

believed. The meaning of the le particle occurring in a position that
is both post-verbal and sentence-final can also be determined by uti-
lizing the notion of temporal journey.
6.12 Le as a relative tense marker
All the le constructions involve ordering of events irrespective of the

position of le.
Further, in a sentence lacking explicit expressions signalling
the event order, le is indispensable for indicating the ordering of the
stated and the presupposed events. This can be demonstrated clearly
by the following semantically equivalent pair:
(93) Xiao Wang xian chi fan, ranhou qu kan dianying.

Little Wang first eat food, afterwards go see movie
(Xiao Wang will eat first, then go to see a movie.)
(94) Xiao Wang chi le fan qu kan dianying.

Little Wang eat le food go see movie.
(Xiao Wang’s going to see a movie after some food.)
238 Han Yang
While (93) contains adverbials explicitly indicating the ordering of

the two events, (94) contains only the le particle, which is indispens-
able in this sentence. Compare (94) with (95) below:
(95) * Xiao Wang chi fan qu kan dianying.

Little Wang eat food go see movie
An attempt of removing le from (94) can only result in ungrammati-

cality.
Le, therefore signifies ordering of events, the ‘relative tense’
in the sense defined by Comrie: “where the reference point for loca-
tion of a situation is some point in time given by the context, not
necessarily the present moment” (1985: 56), rather than the comple-
tion of an action as stated by numerous grammarians.
Le therefore, is a relative tense marker.
7. Conclusion
In this chapter we have examined the ways in which time informa-

tion is encoded in Chinese, Japanese and English, and concluded
that verbal particles and inflections do not obligatorily mark time
reference. They are, in all three languages, secondary modifications.
Based on an understanding of the notion of temporal journey (loca-
tion, direction, movement and border crossing) (Jessen 1973), the
current study reaffirms the position that le signifies ordering of events
(whether explicitly stated or presupposed) irrespective of the syntac-
tic position of this particle.
It further argued that in a sentence lacking explicit expressions
signalling the event order, le is indispensable for indicating the or-
dering of the stated and presupposed events. When there is no pre-
supposition of border and border crossing, le would not be used, even
though the sentence may be well formed syntactically.
This study departs significantly from the generally accepted
position in previous studies on verbal particles in that it claims that
the le particle signifies relative tense rather than aspect, as it indi-

cates a reference point, and that reference point for location of a
situation is some point in time given by the context, not necessarily
the present moment.
This conclusion is in accordance with Comrie’s definition on
relative tense. “where the reference point for location of a situation
is some point in time given by the context, not necessarily the present
moment” (1985: 56).
References
Chao, Y.R. 1968. A Grammar of Spoken Chinese, Berkeley/Los An-

geles: University of California Press.
Cheng, R.L. / Li, Y.C. / Tang, T.C. (eds) 1977. Proceedings of Sym-
posium on Chinese Linguistics, 1977 Linguistic Institute of the
Linguistic Society of America. Taipei: Student Book Co.
Champell, H. 1992. Towards a typology of aspect in Sinitic languages,
Symposium Series of the Institute of History and Philology:
Chinese languages and linguistics, Vol. 1, Chinese Dialects.
Taipei: Academia Sinica, 67-106.
Chu, C.C. / Chang, W.B.V. 1987. The discourse function of the verbal
suffix -le in Mandarin, Journal of Chinese linguistics. 15/2,
309-334.
Comrie, B. 1976. Aspect. Cambridge: Cambridge University Press.
Comrie, B. 1985. Tense. Cambridge: Cambridge University Press.
Dai, Y.J. 1994. “le” zai biashi weilai yiyi juzi zhong de yongfa, in
Xiandai Yuyanxue. Beijing: Yuwen Chubenshe, 114-122.
Dowty, D.R. 1979. Word Meaning and Montague Grammar. Dordrecht:
Reidel.
Frawley, W. 1992. Linguistic Semantics, London: Lawrence Erlbaum
Associates, Publishers.
Gao, M.K. 1970. Guoyu Yufa, Taipei: Letian Chubanshe.
Han, Y. 1991. An Analysis of Particles in Mandarin. Tokyo: Simpusha,
v + 247.
Han Yang
Han, Y. 2005. A case for relative tense in Chinese. In Shao, J.M. /

Luke, K.K. (eds) Hanyu Yufa Yanjiu de Xintuozhan (2).
Hangzhou: Zhejian Education Publisher.
He, B. 1992. Situation types and aspectual classes of verbs in Man-
darin Chinese, PhD Dissertation, Department of East Asian
Languages and Literatures, Ohio State University.
Hofman, Th.R. 1993. Realms of Meaning: An introduction to seman-
tics, London/New York: Longman.
Hopper, P. 1982. Typological Studies in Language, Vol. 1 – Tense-
Aspect: Between Semantics and Pragmatics. Amsterdam: John
Benjamins Publishing Company.
Huang, L.M. 1988. Aspect: A general system and its manifestation in
Mandarin Chinese. Taipei: Student Book Co.
Huang, L.M. / Davis, P.W. 1989. An aspectual; system in Mandarin
Chinese, Journal of Chinese Linguistics. 17/1, 128-166.
Jessen, M. 1973. Preliminaries to a theory of temporal journeys. Edin-
burgh Working papers in Linguistics, no. 2, April 1973.
Jin, L.X. 1998. Shi lun “le” de shiti tezheng, Yuyan Jiaoxue yu Yanju,
Vol. 1.
Kindaichi, H. 1955. Nihongo ddoshi no tensu to asupekuto. Reprinted
in Kindaichi (1976).
Kindaichi, H. 1976. Nihongo Dooshi no Asupekuto. Tokyo: Mugi
Shobohan.
Li, C.N. / Thompson, S.A. 1981. Mandarin Chinese, A functional
reference grammar. Berkeley: University of California Press.
Li, C.N. / Thompson, S.A. / McMillan, T.R. 1982. The discourse
motivation for the perfect aspect: The mandarin particle LE.
In Hopper, P. (ed.) Typological Studies in Language, Vol. 1 –
Tense-Aspect: Between Semantics and Pragmatics. Amsterdam:
John Benjamins Publishing Company.
Li, T.G. 1998. Xiandai hanyu Shizhi Yanju, Liaoning, Daxue Chuban-
she, Liaoning, 2 + 139.
Li, T.G. 1999. “Le, zhe,guo” chengxian xiangduishi gongneng de ji-
zhong yongfa, 㻊婆⬎佺 1999 no. 2, 11-14.
Liu, X.M. 1997. The verb-copyung construction and imperfectivity.
Journal of Chinese language Teachers Association. 32, 1-13.
Lyons, J. 1977. Semantics, Cambridge: Cambridge University Press.
Tense, Aspect and verbal morphemes in Chinese, Japanese and English
Makino, S. / Tsutsui, M. 1991. A Dictionary of Basic Japanese Gram-

mar. Tokyo: The Japan Times.
Mangione, L. / Li, D. 1993. A compositional analysis of -guo and -le.
Journal of Chinese linguistics. 21/1, 65-77.
Ogihara, T. 1999. Tense and aspect, Tsujimura N (ed.) The handbook
of Japanese Linguistics. Oxford: Blackwell, 326-348.
Rohsenow, J.S. 1977a. Perfect le: Temporal specification in Manda-
rin Chinese. Studies in the Linguistic Sciences. 1/2, 142-164.
Rohsenow, J.S. 1977b. Perfect le: Aspect and Relative Tense in Man-
darin Chinese. In Cheng, R.L. / Li, Y.C. / Tang, T.C. (eds) Pro-
ceedings of Symposium on Chinese Linguistics, 1977 Linguis-
tic Institute of the Linguistic Society of America. Taipei: Student
Book Co.
Shao, J.M. / Luke, K.K. (eds) 2005. Hanyu Yufa Yanjiu de Xintuozhan
(2), Hangzhou: Zhejian Education Publisher.
Smith, C. 1991. The Parameter of Aspect. Dordrecht: Kluwer Aca-
demic Publishers.
Smith, C. 1997. The Parameter of Aspect. Second Edition. Dordrecht:
Kluwer.
Tai, J.H.-Y. 1984. Verbs and times in Chinese: Vendler’s four catego-
ries, in Papers from the Parasession on Lexical Semantics.
Chicago: Chicago Linguistic Society, 289-296.
The Open University of Hong Kong (1999, 2001) reprint (2004).
‘Tense, aspect, reference’ in A300 Introduction to Semantics
and Pragmatics. Hong Kong: OUHK.
Tung, P.C. / Pollard, D.E. 1982. Colloquial Chinese. London: Rout-
ledge and Kegan Paul.
Vendler, Z. 1967. Linguistics in Philiosophy. New York: Cornell Uni-
versity Press.
Verkuyl, H.J. 1993. A Theory of Aspectuality. Cambridge: Cambridge
University Press.
Wang, W.S.-Y. 1965. Two aspect markers in mandarin, Language.
41/3, 457-470.
Yang, G.W. 2001. Hanyui taizhi zhong “fuhetai” de shengcheng,
Zhongguo Yuwen 2001, no. 5, 418-427.
Zhu, D.X. 1982. Yufa Jiangyi, Shanwu Yinshuguan.
MASANOBU HORIGUCHI
Japanglish in Katakana and a Comparison

with Loanwords in Cantonese
Contemporary Japanese language contains many English loanwords,

but what may appear to be an English word, such as ‘pension’ or
‘consent,’ may actually differ from English in usage. Consider the
following mock ‘quiz’, designed for non-Japanglish speakers to il-
lustrate some of these differences.
(1) a. TA RE N TO1 (for ‘talent’):

1. TV personality 2. business performance 3. skill 4. good grade
b. KO N SE N TO (for ‘consent’):
1. labor agreement 2. outlet 3. general meeting 4. compromise
c. SU MA A TO (for ‘smart’):
1. automatic 2. intelligent 3. cunning 4. slim
(Answers: a = 1, b = 2, c = 4)
This shows that English words and phrases are not only always used
with their original meanings but are often modified by Japanese speak-
ers depending on the situations in which they are used. It has been said
1 In this chapter, I represent Japanese Katakana letters (i.e., Japanese phonograms)
by the combination of consonant and vowel such as CV, V, CCV. Each set
represents a mora, or phonologically a half-sized unit of foot in English terms.
Japanese morae cannot end with consonant because they should be open syl-
lables, but the exceptions are those morae ending with syllabic nasal (i.e., N)
and the front half of consonant gemination (various consonants). Borders be-
tween morae are represented by blanks, so that the number of morae in (a) is 6,
and that of (b) is 4 while the number of syllables in (a) and (b) are 3 and 2.
(a) KI K KO O MA N (Kikkoman) (b) KYA N DI I (candy) [Katakana representation]
1 2 3 4 5 6 1 2 3 4 [the number of morae]
KIK KOO MAN KYAN DII [syllable representation]
1 2 3 1 2 [the number of syllables]
244 Masanobu Horiguchi
that “English has become another set of symbols – another linguistic

resource – to be used as Japanese people see fit” (Stanlaw 2004: 171).
The Japanese language has also assimilated loanwords from other
countries (for example, Arakawa 1977 and Lovins 1975):
(2) a. I KU RA (from Russian ‘ikra’): salmon roe

b. PI N SE T TO (from Dutch ‘pincet’): tweezers
c. NO I RO O ZE (from German ‘Neurose’): nervous breakdown
d. KO N KU U RU (from French ‘concours’): competition
e. BI RO O DO (from Portuguese ‘veludo’): velvet
Japanglish speakers are typically indifferent to the origins of these

Japanese English words and phrases. In addition, not a small part of
the Japanese language is made up of ‘Japanese English,’ so foreign-
ers who study Japanese sometimes have a hard time communicating
with Japanese speakers (especially, with younger generations). To
make matters worse, some Japanese English words are replaced with
new Japanese English or traditional Japanese, which gives the older
generation even more difficulties in understanding such words:
(3) a. CHA A MI N GU (for ‘charming’) → Kawaii (for Japanese ‘cute’)

b. MI DO RU E I JI (for ‘middle-age’) → Chuunen (for Japanese ‘middle-age’)
c. PU RE I GA I DO (for ‘play guide’, meaning ‘box office’)
→ CHI KE T TO (ticket) uriba (for Japanese ‘box office’)
d. RE JA A RA N DO (for ‘leisure land’, meaning amusement park)
→ TE E MA PA A KU (for ‘theme park’)
Taking the above examples (1-3) into consideration, it would be im-

proper to use the word ‘loanwords’, so that I will label these words and
phrases as ‘Japanglish (i.e., Japanese English)’ in this chapter. In the
formation and usage of Japanglish, Katakana letters (i.e., Japanese
phonograms) play an important role in that they can almost keep the
sounds of the loanwords without using alphabetic characters. They
also mask the origins and original meanings of Japanglish. As is well
known, Japanese onomatopoeia and mimesis (such as ‘zigzag’ or
‘dingle-dangle’) have contributed to the formation of Japanglish for
Japanglish in Katakana and a Comparison with Loanwords in Cantonese 245
a long time, and, in cultural dimensions, these Japanglish words and

phrases have been used for TV cartoon names and names of best-
selling products. As one of the aims of this paper is to compare and
contrast Japanglish words with Cantonese loanwords, I will first sur-
vey the use of English loanwords in Hong Kong, where there Chinese
characters rather than roman letters or katakana are used.
1. Loanwords in Cantonese
As ideograms, Chinese characters are versatile in packing various

meanings into a single word, and the Cantonese language seems to
make use of these properties as often as possible. This can be seen
from the following Cantonese examples, which are all transcribed by
Katakana phonograms in Japanese (Most examples are from Ito 2003):
(4) a. ┦ᶵ for camera,㟁Დ for elevator,➼㞳Ꮚ for plasma,㌾௳ for soft-

ware
b. 㤶ཱྀ⭺ for chewing gum,ព⢊ for spaghetti
c. ⱥᅧ for Britain,ἲᅧ for France,ᚨᅧ for Germany,⨾ᅧ for America
d. ၈⪁㬞 for Donald Duck,㮰⺋㉸ே for Ultra Man (Japanese supernatural
giant hero),
Ḉ᱈ᑠ୸Ꮚ for Chibimaruko chan (Japanese girls’ cartoon),
㱟⌔ for Dragon Ball (Japanese boys’ cartoon),
㱟⊧ for Tonarino Totoro (Japanese cartoon)
On the other hand, Chinese characters can also be used as phonograms,

as follows:
(5) a. ⓗኈ for taxi,ᕮኈ for bus,㜿㔛ᕮᕮ for Ali Baba,ဋ for card
b. ᮒྂຊ
•éŒÃ—Íforfor chocolate,ᚷᗮ
chocolate, Šõ—õfor for
cream,ྍཱྀྍᵹ for Cocafor
cream, ‰ÂŒû‰Â•Ù Cola,
Coca Cola,
࿼⡀
?? for for
curry, •z™² for for
curry,ᕸ⏩ pudding, •U—–’n forfor
pudding,ᣆ⹒ᆅ brandy, ˆÐ•mŠõ for
brandy,ጾኈᚷ forwhisky
whisky
c. ព኱฼ for Italy,௨Ⰽิ for Israel
d. ㏔ኈᑽ for Disney,ዲⴊሳ for Hollywood,⡿ወ for Mickey,
㧗㐔 for Gundam (Japanese robot cartoon)
In the following examples, letters from the alphabet are used as well
as Chinese characters as phonograms:
(6) a. ༩ᢼ OK for KaraokH

BB ㌴ for Baby Buggy or stroller
༩ᵹ B for Calbee (Japanese snack’s maker)
b. ከၵ㸿ክ for Doraemon (Japanese cartoon)
,4 ༤ኈ for Doctor Slump Arare chan (Japanese cartoon)
• ᑠ㨱ዪ DoReMi for Ojamajo Doremi (Japanese girls’cartoon)
(6b) ከၵ㸿ክwas formerly represented as ྌ⤈㣬, which adopted

Chinese characters purely as ideograms. Finally, there are many con-
temporary examples consisting solely of roman letters. As can be
seen from (7c), it is very common for people to adopt English names
in Hong Kong.
(7) a. FANCL (for a Japanese cosmetics company)

Qoo (for a brand name of Japanese beverage)
YEBISU Beer (for ᜨẚᑑ%HHBeer: a brand name of the Japanese
SAPPORO beer brewery company)
b. BEYOND, TWINS (for Hong Kong singers)
SMAP, Orange Range (for Japanese singers)
Gackt (for a Japanese singer)
c. Jackie Chan, Samuel Hui (for ᡂ㱟チෙഔ : Hong Kong’s actors)
Teresa Teng (for 㒭㯇ྩ: a Taiwanese singer)
The above (7b) FANCL is an ‘acronym,’ which means that this word
is pronounced as [fænkl], not as [ef ei en si: el]. On the other hand,
in the following examples of ‘initialisms,’words are pronounced let-
ter by letter (Tabata 2004 for the distinction between acronym and
initialism):
(8) a. DHC (for a Japanese cosmetics company)

b. F4 (for four-member Taiwanese singers)
For example, (8a) DHC is pronounced as [di: eit5 si:]. Words in this
category come with more ‘foreign’ impressions.
All in all, as can be seen from (4), the Cantonese language
adopts fewer phonograms in loanword formation than Japanese, and
even if Chinese characters are used phonogramically, those loanwords

represented by Chinese characters can be buried among other Chi-
nese characters. Only those loanwords containing alphabet letters, as
in (5-6), indicate they are ‘foreign’ visually. In Japanese, Katakana
letters work as a marker for loanwords as follows.
2. Loanword-formation and the Role of Katakana Letters

in Japanese
As is well known, the Japanese writing system has three sets of sym-
bols: Chinese characters (ideogram); Hiragana letters (phonogram);
and Katakana letters (phonogram). Combinations of Chinese char-
acters and Hiragana letters are thought to be basic Japanese sentence
structures by Japanese speakers while Katakana letter representa-
tions of loanwords, proper nouns, and onomatopoeia or mimesis are
considered somewhat marked words or phrases (i.e., italicized phrases
in English terms). According to Bunkacho (1997: 38), the usage of
Katakana letters for transcribing loanwords dates back to almost 1500
years ago, when Japanese people tried to import Buddhism. By cit-
ing the examples from Bunkacho (1997: 13), the following Buddhist
terminology was first transcribed by Chinese characters phonogrami-
cally, like in the case of Cantonese in (5), and, then, Katakana letter
transcriptions were added alongside these Chinese character words
because Chinese characters can be pronounced in various ways:
(9) a. Buddha → ௖㝀 → BU D DA

(Sanskrit) (Chinese character) (Katakana letter transcription)
b. Dharma → 㐩☻ → DA RU MA
(Sanskrit) (Chinese character) (Katakana letter transcription)
This transcription system (i.e., first by Chinese characters and, then,

by Katakana letters) was further applied to loanword formation from
European cultures. Since guns were first imported from Portugal in
1543 and, later, a missionary Francisco de Xavier brought Christian-
ity to Japan in 1549, Portuguese loanwords have come into Japanese

(in 10a). During the Edo period, from the 1600s to the 1800s, when
trade with foreign countries was banned except for the Netherlands,
Dutch loanwords multiplied as in (10b) (examples are from Morioka
1993):
(10) ྜ⩚(KA
a. ‡‰H (KA P PA from
from Portuguese
Portuguese‘capa,’
‘capa,’meaning
meaning‘raincoat’)
‘raincoat’)
ኳ㯑⨶(TE
“Vên—… NN
(TE PUPU RARAfrom Portuguese
from ‘têmporas,’
Portuguese or tempura)
‘têmporas,’ or tempura)
↮ⲡ(TA
‰Œ‘• BA
(TA KO
BA KO from
fromPortuguese ‘tabaco,’
Portuguese or or
‘tabaco,’ tobacco)
tobacco)
• 㦵∩(KA RU TA from Portuguese ‘carta,’ meaning ‘Japanese-style card
game’)
⌆⍄KO
b. àÛàè (KO OOHI HIIIfrom
fromDutch
Dutch ‘koffie,’
‘koffie,’ or
or coffee)
coffee)
㘚ຊ(BU
èF—Í (BU RI KI from
from Dutch
Dutch ‘blik,’
‘blik,’meaning
meaning‘tin’),
‘tin’),
㯏㓇(BI
”••ð (BI I IRU
RUfrom
fromDutch
Dutch ‘bier,’
‘bier,’ or
or beer)
beer)
ὒⅉ(RA
—m“” (RANNPUPUfrom
fromDutch
Dutch‘lamp,’
‘lamp,’ororlamp)
lamp)
In the above examples, Portuguese loanwords in (10a) are so old and

familiar to Japanese people that they are no longer represented by
Katakana letters but by Hiragana letters: some Japanese speakers may
think of ‘tempura’ as an original Japanese word. In (10b), some Chi-
nese character words such as ⌆⍄and 㯏㓇still remain, but only in
retrospective or nostalgic usage. In the following scientific terminology
from Dutch, Chinese character transcriptions were soon abolished,
and only Katakana letter representations survived (Morioka 1993):
(11) a. ள∞⟠ඣ(from Dutch ‘alcohol’) Ѝ A RU KO O RU

b. Ᏻἐᑽள(from Dutch ‘ammonia’) Ѝ A N MO NI A
c. ள∞ຍ㔛(from Dutch ‘alkali’) Ѝ A RU KA RI
The usage of Chinese characters in an ideogramic way (cited from

Yazaki 1964: 71, 73) has never come into fashion in Japanese.
Katakara letter transcription has dominated:
(12) a. ⃰⦰ᾮ(from Dutch ‘extract’) Ѝ E KI SU

b. ᑟᒀ⟶(from Dutch ‘katheter,’ meaning ‘catheter’) Ѝ KA TE E TE RU
c. ⨶㔪൤(from Dutch ‘kompas,’ meaning ‘compass’) Ѝ KO N PA SU
Likewise, ideogrammic usage of Chinese characters during the Sec-

ond World War, when English was banned as the enemy’s language,
became out of fashion quickly (Koizumi 1978: 321-322):
(13) a. 㟁㧥for‘permanent wave’) Ѝ PA A MA (or PA A MA NE N TO)

b. ㏱᫂ᯈ(for ‘glass’) Ѝ GA RA SU
As a result, Japanese loanwords represented in Chinese characters

are quite rare, which contrasts with the rich examples in this cat-
egory in Cantonese such as in (4a-d). In contemporary Japanese, this
easy way of tracing loanword sounds by Katakana letters comes to
be applied in various languages:
(14) a. from English: CHO KO RE E TO (chocolate), KU RI I MU (cream),

KO KA KO O RA (Coca Cola), KA RE E (curry), PU RI N (pudding),
BU RA N DE E (brandy), U I SU KI I (whisky), etc.
b. from French: A BE K KU (avec, meaning ‘couple’),
KO N TO (conte, meaning ‘comic skit’)
c. from German: A RU BA I TO (Albeit, meaning ‘part-time job’),
RE N TO GE N (Röntgen, meaning’X-ray’) Röntgen is the person who
discovered X-ray.
d. from Portuguese: PA N (pão, meaning ‘bread’), BU RA N KO (balanco,
meaning ‘swing’)
e. from Dutch: O RU GO O RU (orgel, meaning ‘music box’),
GI BU SU (gips, meaning ‘plaster cast)
f. from Russian: KO N BI NA A TO (kombinat, meaning ‘industrial com-
plex’), NO RU MA (norma, meaning ‘sales quota’)
g. from Italian: DA KA A PO (da capo), A N DA N TE (andante),
KA PU CHI I NO (cappuccino), TI RA MI SU (tiramisu),
h. from Spanish: A RU MA JI RO (armadillo), SE NYO RI I TA (señorita,
meaning Mrs.)
i. from Cantonese: SHU U MA I ( ⇦㈽, steamed meat dumpling),
WA N TA N ( 㞼࿐, meat dumplings in soup)
j. from Korean: KI MU CHI (kimchi),
YU K KE JA N (Yuk ge zhang, meaning Korean beef dish)
It was not until 1902 that general guidelines for adopting Katakana
letters in transcribing loanwords were established by the government’s
language research committee (Bunkacho 1997: 38), but Katakana
letters were considered loanword-specific from much earlier times.
3. Basic Structures of Japanglish
So far, the system of absorbing foreign words into Japanese by using

Katakana letters has been discussed, but only half of the nature of
Japanglish has been explained: we must show how Japanese speak-
ers have manipulated these Katakana words in creative ways (or in
sloppy ways) which has lead to unexpected consequences of the es-
tablishment of just more than loanwords, or Japanglish.
First, the following Japanglish words have almost kept the loan-
words’ sounds, but the meanings differ from those of the original.
The reason may be situational: the actual situations or paragraph
contexts where these words were used for the first time in Japanese
may have affected the usage of these words incorrectly:
(15) a. KA N NI N GU (from English ‘cunning’), meaning cheating, not shrewd

b. SU TO O BU (from English ‘stove’), meaning kerosene heater, not cooker
c. HO CHI KI SU (from an American inventor’s name ‘Hotchkiss’), mean-
ing stapler
Compound word formation involves not only English but also other
foreign languages:
(16) a. ME RO N SHI RO P PU (melon [English] + siroop [Dutch: syrup])

b. JA MU PA N (jam [English] + pão [Portuguese: bread])
c. TE E MA SO N GU (Thema [German: theme] + song [English])
d. KA HU SU BO TA N (cuffs [English] + botão [Portuguese: button])
Chinese characters are also involved in the formation of loanword

compounds:
(17) a. A I DO RUḷᡭ(idol + singer): young popular singer

SU RI RU‶Ⅼ(thrill + perfect score): most thrilling
b. ᯝ≀NA I HU (fruits + knife): fruit knife
ᯈCHO KO (bar + chocolate): chocolate bar
Loanwords used frequently in Japanese are abbreviated in accordance

with Japanese phonological restrictions: stable minimum units are two-
mora words such as in (18a) (Poser 1990); three-mora words are also
used as stable as in (18b) (Mori 2002); and four-mora words are most
frequently used as the most stable sequences as in (18c) (Ito 1990):
(18) a. NE GA (negative, in printing photographs), BI RU (building),

RE JI (register) meaning ‘check-out counter,’
GYA RA (guarantee), meaning ‘performance fee’
b. TE RE BI (television set), BA SU KE (basketball)
DA I YA (diagram) meaning ‘timetable for trains,’ KO N BI (combination)
meaning ‘a pair of two persons such as comedians or singers’
c. E A RO BI (aerobics), A PA A TO (apartment house), HA N KA CHI
(handkerchief), I RA SU TO (illustration), RI HA BI RI (rehabilitation),
I N FU RE (inflation), MA N NE RI (mannerism) meaning too stereo-
typed, A KU SE RU (accelerator, or gas pedal)
In contrast, there are relatively fewer examples for abbreviations leav-

ing the second half of the words: most of them happen to be three-
mora words:
(19) a. BA I TO (Albeit, or ‘part-time job’ from German), SHI SU KO (San Fran-

cisco), HO O MU (platform), NE K KU (bottleneck)
In compound abbreviations, which consist of the most important part

of Japanglish, two sets of two-mora abbreviations form the most stable
four-mora structures (in (20b), even the non-abbreviated compound
words do not make sense in English):
(20) a. A ME HU TO (American + football), E A KO N (air + conditioner),

WA A PU RO (word + processor), RI MO KO N (remote + control)
b. TO RE PA N (training + pants): sweat pants, not meaning diaper
MA ZA KO N (mother + complex): mama’s boy
MA SU KO MI (mass + communication): mass media
E N SU TO (engine + stop): engine trouble
There are also three-mora structures in compound abbreviations:
(21) a. HA I SO (high + society)

b. TE RE KA (telephone + card): a card for paying the charge of using a pay
phone
Since two-mora or three-mora units are stable structures, short En-

glish words such as ‘no,’ ‘free,’ ‘one,’ ‘top,’ or ‘my’ can be used like
prefixes in Japanglish as follows:
(22) a. NO O (no) + something:

NO O TA C CHI (no + touch): having nothing to do with a certain matter
NO O GE E MU (no + game): a game is rained off
b. FU RI I (free) + something:
FU RI I DA I YA RU (free + dial): toll-free telephone number
FU RI I SA I ZU (free + size): one-size-fits-all
c. WA N (one) + something:
WA N TA C CHI (one + touch): easy to handle some machines
WA N PA TA A N (one + pattern): too stereotyped and boring
d. TO P PU (top) + something:
TO P PU NYU U SU (top + news): a lead story of a TV broadcasting program
TO P PU SE E RU SU (top + sales): the largest sales
e. MA I (my) + something:
MA I HO O MU (my + home): one’s own house
MA I PE E SU (my + pace): keeping a ‘going-my-way’ style
There are also suffix examples such as ‘in,’ ‘up,’ or ‘down’ as follows2:
(23) a. something + I N (in):

KYA N PU I N (camp + in): a season when professional baseball players
start practicing
DA I YA RU I N (dial + in): direct dialing system
2 Young Japanese speakers have a tendency to attach typical English suffixes

(such as those found in Marchand 1969) to various words:
(a) A A (from -er):
SHA NE RA A (Chanel [French fashion designer] + -er): those who like Chanel products
A MU RA A (Amuro Namie [ Ᏻᐊዉ⨾ᜨ ] + -er): nickname of a Japanese pop singer
(b) I I (from -ie or -y)
HI K KI I (Utada Hikaru [ Ᏹከ⏣ Hikaru] + -ie): nickname of a Japanese pop singer
JI MO TI I (Jimoto [ ᆅඖ: neighbor] + -ie): local residents
b. something + A P PU (up):
I ME E JI A P PU (image + up): improve one’s reputation
RE BE RU A P PU (level + up): raise the level of something
c. something + DA U N (down):
I ME E JI DA U N (image + down): damage one’s reputation
RE BE RU DA U N (level + up): lower the level of something
Based on the method to form the basic structures shown in (15-24),

Japanese people expanded the vocabulary of Japanglish especially
in the field of best-selling product names and TV cartoon heroes and
heroines.
4. Cultural Dimension 1: Best-Selling Product Names
Japanese products dominated the world market much before the era
of the ‘bubble economy,’ and 60 percent of them consist of Katakana
naming, which may make these products sound suitable for export
(if we count Katakana compounds with Chinese characters, the sum
will rise to 74.4 percent, according to Abe 1990: 191). This is in stark
contrast with the small usage of loanwords in newspapers (i.e., 12
percent according to Ishiwata 1985: 17). In the following examples
(from Kimura 1998, Akiba and Chikuma 2000, and Yasuda 2003) of
Katakana brand names, Katakana phonograms are used as if they
played the role of masking the original meanings of companies or
products (Morioka 1993: 1909):
(24) a. KE N WU D DO (Kenwood: Ken + wood)

SA N TO RI I (Suntory: sun + torii [ 㫽ᒃ: president’s family name])
DA SU KI N (DA SU TO [dust] + zookin [ 㞧ᕵ: cloth] = a brand name of
cleaning chapter)
RI KO O (RIKOH: ri [ ⌮] + koh [ ග] = ⌮◊ᕤᏛᕤᴗ)
b. E ME RO N (EMERON: emerald + lion = a brand name of shampoo)

ME RU SHA N (Mercian: Merci + -an = liquor company)
SA RA N RA P PU (Sarah + Ann + wrap = a brand name of plastic wrap)
KU RI NE K KU SU (Kleenex: clean + exit)
c. JI MU NI I (Jeep + tiny: Jeep + -ny → Jeem + -ny = a brand name of car):

assimilation
DE MI O (de + mio: ‘of mine’ in Spanish): Japanese speakers cannot split this
into two
A RI NA MI N (Allium sativm Linnaeus + Thiamin = nutritious drinking
water)
PA N SHI RO N (pantothenic acid + Cillon [German castle] = drug for
stomachache)
In any of the above examples, brand names are too familiar to Japa-
nese people to take trouble to analyze and reconsider the origins of
the words, and most Japanese people do not know what the above
brand names are made up of. In (24a), Japanese speakers can intu-
itively break down the compound into each part, which corresponds
to the sequences of morae, or Japanese minimum phonological units
(Vance 1987 and Kubozono 1989). In (24b), Japanese speakers may
have some difficulty in reaching the complete analyses of these com-
pounds, which entail the concept of syllables in English terms (Kahn
1976 and Selkirk 1982), not morae: for example, E ME RO N should
be divided into E ME R + O N, and the division of one mora ‘RO’
into two subcomponents ‘R + O’ cannot intuitively be possible in
Japanese. In (24c), the examples are too complex to be analyzed in-
tuitively: assimilation or nasalization of [p] into [m] in Jeemny [Jeep
+ -ny] cannot easily be realized in Japanese; and DE MI O (de: of +
mio: mine) sounds like one word, not a compound for ordinary Japa-
nese speakers (because both components are function words, so that
they rarely make sense to ordinary Japanese speakers, except for those
who know Spanish.
Representing Chinese company names or Chinese product
names by Katakana letters dates back to as early as the pre-Second
World War period:
(25) a. SONY Cassette DE N SU KE ( ఏຓ): cassette tape recorder in 1973

b. SA P PO RO ୍␒( ᮐᖠ: capital city of Hokkaido): noodle in 1966 (୍␒:
No. 1)
c. SU BA RU 360 ( ᫡): car brand name in 1958
d. SHA CHI HA TA ( 㪘᪝): stamp maker in 1929
In (25a), DE N SU KE is from the name of a specific detective who

arrested illegal gamblers, but most Japanese speakers will be reminded
of some traditional Japanese male name or other by the sound of DE
N SU KE because of the common suffix ‘SU KE,’ which is attached
to traditional male names. The usage of a very traditional name for a
cutting-edge product made an eye-catching copy in those days.
In contrast with the above examples, which remind Japanese
speakers of the original Chinese character names, the following ex-
amples may have successfully masked the origins as Chinese characters:
(26) a. KI K KO MA N ( ட⏥୓) in 1940: soy sauce maker

KA SHI O [CASIO] ( ᶔᑿ) in 1957: computer and calculator maker
KO KU YO ( ᅜ㄃) in 1961: stationery maker
BA N DA I (ⴙ௦) in 1961: toy maker
b. KO SHI HI KA RI ( ㉺ග) in 1956: an excellent brand name of rice
SU JA A TA ( ಟ㜒ከ= a heroin in a Buddhist story) in 1976: a brand name
of milk
I think most Japanese speakers regard the above maker names (27a)
and product names (27b) as originally Katakana letters or even alpha-
bet letters.
More complex cases of making Japanese phrases or even sen-
tences sound like Western languages can be found in the examples
by Gennai Hiraga (1728-1779), a natural scientist and drama writer
in the Edo period (Bunkacho 1998: 35):
(27) a. I E U GO O KU (meaning ‘earthquake’)

→ IE + U GO O KU
yeah go oak (pronounced like an English phrase)
‘house’ + ‘shake’ (meaning that a house shakes)
b. MA A SU TO KA A TO RU (meaning ‘spiral shaped mosquito-repellent
incense’)
→ MA A SU TO + KA A TO RU
Marston cart -ol (pronounced like an English phrase)
‘If you turn something’ + ‘mosquitoes are caught’ (meaning a mosquito-
repellent)
In the above, ‘yeahgooak’ and ‘Marstoncartol’ are not included in

Bunkacho (1998) but added by myself for the convenience of Eng-
lish speakers. This kind of effort, namely, to make the western-style

sounds based on pure Japanese words by Gennai Hiraga, along with
Portuguese and Dutch loanword-formation from the 1500s, shows
that Japanese people have long stuck to the ‘Hakurai’ doctrine (Kajima
1994: 175) to the present time: ‘Hakurai’ implies that the imported
items by ship from European or American countries are considered
better than domestic products. During the age of the bubble economy
in Japan, however, this ‘the-West-is-better’ attitude was partly di-
minished in two ways.
First, product names came to adopt Chinese characters as fol-
lows:
(28) a. n⇍KA REE E(curry)
KA RE (curry)(in
(in1995:
1995:Glico):
Glico):aabrand
brandname
namefor
forcurry
curry
ឡጔྕ(in
b. ˆ¤•È•† 1983:Matsushita
(in 1983: Matsushita Electronics
Electronics Corp.):
Corp.):electric
electricwasher
washer
c. ˆê‘¾˜Y
୍ኴ㑻(in 1985:Just
(in 1985: JustSystem):
System): word
word processing software
software
d. êi—Ù’W—í
㯃㯌ῐ㯇(in(in
1998: KIRIN
1998: BEER):
KIRIN a brand
BEER): name
a brand for for
name beerbeer
㟝JDᓠ(in
e. –¶ 1967:
ga •ô (in 1967: MITSUBISHI
MITSUBISHI ElectronicsCorp.):
Electronics Corp.):airairconditioner
conditioner
ᩥ㇦(in
f. •¶•‹ 1981:SHARP):
(in 1981: SHARP):word processing software
wordprocessing
g. –ò—p•˜V—Ñ
⸆⏝୙⪁ᯘ(in (in1982:㈨⏕ᇽ):
1982: •‘¶“°):drug
drugto
to grow
grow hair for bald-headed people
h. YU NN KE
KERU
RU(Junker)
(Junker)㯤ᖇᾮ(in 1984:బ⸨〇⸆):
‰©’é‰t (in 1984: nutritious
²“¡»–ò): nutritious drinking water w
drinking
water (Junker [German]: elite young man)
In the above examples, Chinese characters symbolize ‘familiarity’

(as opposed to ‘foreignness’ or ‘indifference’ of Katakana expressions)
in product names concerning food (28a), home electric appliances
(28b), and software (28c), while they are used as ‘gorgeous’ symbols
for the same categories (28d-f). In pharmaceutical products in (28g-
h), Chinese characters not only imply ‘gorgeous’ and ‘expensive’
brands but also ‘effectiveness’ as something based on the ‘long his-
tory’ of Chinese herbal medicine. The familiarity comes from the
domestic nature of Chinese characters while the gorgeousness may
come from the difficulty of writing and reading Chinese characters
compared to Katakana or Hiragana letters as well as the large num-
bers of strokes composing Chinese characters. The concept of the
deep traditional background from Chinese characters is from China
itself, which is foreign and somehow mystic even to Japanese people.
The second example in which the sticking-to-the-western doc-

trine becomes weaker is in the case of adopting Hiragana:
(29) a. Umakacchan: a brand name for instant noodle (in 1979: House Food Corp.),
meaning ‘tasty’ in Japanese Kyushu dialect
b. Donto: a brand name for a pocket body warmer (in 1982: 㔠㫽Corp.),
meaning ‘at a burst’
c. Ken’onkun: a brand name for electric thermometer (in 1983: OMRON Corp.),
meaning ‘Mr. Measurement Man’
d. Otakkusu: a brand name for fax machine (in 1991: Matsushita Electronics
Corp.), meaning ‘home (Otaku) fax’
In all the above examples, Hiragana letters are used to symbolize

‘unaffectedness,’ ‘warmth,’ and ‘familiarity to women and children’
with all of these three characteristics explained in Yasuda (2003: 195-
196). These characteristics partly come from the fact that Hiragana
letters are to be learned at the earliest stage at elementary schools
before Katakana letters or Chinese characters. In addition, Hiragana
letters are composed more of curved (as opposed to straight) strokes
than Katakana letters and Chinese characters. Using ‘warm’ and ‘fami-
liar’ Hiragana letters for expressionless, inanimate products such as
an electric thermometer (29c) or a fax machine (29d), by denying the
usage of foreign-specific Katakana letters, seemed epoch-making as
brand names. Especially in (29c), which sounds like a friendly boy’s
name with a hypocoristic expression (i.e., -kun) user-friendliness of
a machine is stressed.
The usage of Chinese characters and Hiragana letters, how-
ever, could not be the most common method of producing brand
names. Since 2000, many product names with Chinese characters
have returned to Katakana expressions:
(30) a. Panasonic wide-screen TV set ‘ ⏬⋤’ (in the 1990s: meaning ‘screen king’)
→ Panasonic thin flat panel TV set ‘TA U’ (Tau) (in the 2000s: Greek
alphabet)
b. HITACHI wide-screen TV set ‘ 㠉࿨ඣ’ (in the 1990s: meaning ‘revolutio-
nary boy’)
→ HITACHI plasma TV set ‘U U’ (Wooo) (in the 2000s: onomatopoeia)
The fact that not a small percentage of Katakana expressions consist

of onomatopoeia and mimesis is often ignored in the discussion of
Japanglish although there is a large amount of data on Japanese ono-
matopoeia (such as Amanuma 1974, Hamano 1986, and Tamori 2002).
These previous research projects have focused more on the usage of
onomatopoeia or sound symbolism rather than the matter of whether
they should be written in Katakana letters, Hiragana letters, or Chi-
nese characters. Although there is a general guideline in which voice
onomatopoeia (such as ‘bowwow’ and ‘meow’ in English terms)
should be expressed in Katakana letters (Amanuma 1974: 58-59),
other non-voice onomatopoeia (such as ‘rattle’ or ‘slap’ in English
terms) and mimesis (such as ‘zigzag’ or ‘dingle-dangle’) are also
commonly written in Katakana letters. In product names, the usage
of Katakana letters for onomatopoeia and mimesis, as well as for
loanwords, has been playing an important part in dramatically in-
creasing catchy Japanglish from very old times to the present:
(31) a. I BO KO RO RI (wart [I BO] erasing ointment) in 1919: KO RO RI means

‘rolling down.’
PI P PU E RE KI BA N (cold compress using magnetic power
[E RE KI BA N]) in 1972: PI P PU means ‘flashing.’
GO KI BU RI HO I HO I (cockroach [GO KI BU RI] buster) in 1973:
HO I HO I means ‘easily.’
ZA BU (detergent) in 1960: ZA BU means ‘splashing sound.’
b. ෗RU N DE SU (disposable camera) in 1986: RU N means ‘cheerful
mood.’ (DE SU is a ‘be’ verb.)
PI K KA RI KO NI KA (camera) in 1974: PI K KA RI means ‘flashing.’
(KO NI KA is a company name.)
c. BI K KU RI MA N CHO KO (chocolate) in 1977: BI K KU RI means
‘surprise.’
GU RI KO PO K KI I (Glico Pocky chocolate) in 1966: PO K KI I means
‘cracking sound.’
d. HO N DA RA T TA T TA (HONDA’s motorbike brand) in 1976:
RA T TA T TA means ‘speedy and cheerful movement.’
CHO RO ‘Q’ (miniature car) in 1980: CHO RO means ‘swift movement.’
(‘Q’ is from ‘cute.’)
The product categories range from drugs (31a) to photography (31b),

to snacks (31c), and to cars (31d). Although onomatopoeia and mi-
mesis are purely Japanese, they can be used naturally as components

of Japanglish because they are short enough (i.e., as short as three
morae) to be used like prefixes or suffixes. In addition, onomato-
poeia and mimesis in Katakana letters appeal to our senses quite
directly both visually (by standing out in plain Japanese sentences of
Hiragana letters and Chinese characters) and acoustically (by sound-
ing differently from traditional Japanese nouns or verbs), so that they
have much in common with loanwords expressed in Katakana let-
ters. As can be seen from the above (31a) PI P PU, (31b) PI K KA RI,
(31c) BI K KU RI, PO K KI I, and (31d) TA T TA T TA, consonant
gemination (progressive gemination) is frequently adopted in ono-
matopoeia and mimesis. This is because gemination can create the
impression of ‘swiftness’ and ‘speediness’ (Tamori 2002: 135), which
are favorable factors for making catchy words. Also on this point,
onomatopoeia and mimesis are similar to loanwords, because, as
pointed out by many researchers, geminated consonants not found in
original English words are routinely inserted into Japanese loan-
words3.
3 It is widely accepted that, in phonological terms, English has no consonant

gemination like that of Italian or Japanese although there remains orthographical
gemination such as ‘apple’ or ‘attack.’ It is also taken for granted that, phoneti-
cally, glottal stop or laryngeal constriction in English words is recognized as
consonant gemination by Japanese speakers. Ono (1991) pointed out, how-
ever, that even where neither of them is present as in the case of English words
containing the sequence ‘-ss-’ (like ‘lesson’), Japanese loanwords correspond-
ing to them have geminated consonants (like ‘LE S SU N’). Ono also denies
the possibility of inserting consonant gemination into Japanese loanwords by
the influence from original English spellings, citing the following exceptions
where there occurs no gemination in Japanese:
(a) approach → A PU RO O CHI (b) pattern → PA TA A N (c) pocket → PO KE T TO
(not A P PU RO O CHI) (not PA T TA A N) (not PO K KE T TO)
Furthermore, Ono has shown the tendency, in loanword formation, to gemi-
nate voiced consonants, which can rarely occur in traditional Japanese se-
quences, citing the following examples:
(d) bed → BE D DO (e) bag → BA G GU
In another aspect, Lovins (1975: 87) cites the case of deleting geminated con-
sonants, such as changing ‘PA B BU’ (pub) into ‘PA BU,’ among ‘modern,’
Finally, with the computer industry accelerating production by

replacing old types with new ones quickly, the number of acronyms
(such as 32a) and initialisms (such as 32b) has increased:
(32) a. EOS (Electro Optical System): CANON in 1987 (pronounced as I O SU

= 3 morae)
VAIO (Video Audio Integrated Operation): SONY in 1997 (pronounced as
BA I O = 3 morae)
b. RX – 7 (Rotary engine + X [meaning ‘infinite’] + product No. 7):
MAZDA Motors in 1991
i-mode (tourist information): Nippon Telegraph and Telephone Company
in 1997
They are no longer ‘Katakana’ words since they are all composed of
alphabet letters, and many of them are found among computer item
brand names, so that they will create a more indifferent and artificial
image than Katakana brand names. In the case of acronyms such as
(32a), however, the way Japanese speakers pronounce them reveals
that these acronyms keep the most stable phonological length as Japa-
nese (i.e., three or four morae), which corresponds to three or four
Katakana letters.
young Japanese speakers. As far as I know, there are many similar examples
such as follows:
(f) KI S SU (kiss) → KI SU (g) FI RI P PI N (Philippine) → FI RI PI N
Taking all the above factors into consideration, it may safely be said that the
occurrence of consonant gemination in loanword formation is not caused purely
by the phonetic impression of original English words. Instead, insertion of
geminated consonants will create, in such loanwords, a special impression such
as ‘swiftness’ or ‘speediness,’ which older Japanese speakers recognize as more
foreign and consider to be found also in onomatopoeia and mimesis.
5. Cultural Dimension 2: TV Cartoon Titles

and Character Names
The proliferation of Japanglish in TV cartoons is based on more spiri-

tual factors than that through product names. As I pointed out earlier,
the ‘west-is-better’ doctrine has long existed among Japanese people
since the 1500s. After the surrender of Japan in the Second World
War, the ‘income-doubling national project’ led by Prime Minister
Hayato Ikeda (whose administration was from 1960 to 1964) rein-
forced this favorable attitude toward western culture in the form of
the ‘follow-the-American-lifestyle’ doctrine (Sakaiya 1997). In those
days, anti-America sentiment among Japanese people was so strong
that the students’ anti-American movement was common and many
people supported Rikidozan, or a wrestler who often defeated Ameri-
can counterparts. In order to divert people’s attention from the fact
that Japan had no choice but to follow the United States politically,
Prime Minister Ikeda constantly stressed the richness of American
lifestyle. As a result, most Japanese people came to accept American
items and culture whether they liked the United States or not.
In another respect, Sakaiya (1996) pointed out that Japanese
people have long kept their unique concept of ‘sophistication’ since
the age of Genji Hikaru, or an imaginary heroic character in a novel
Genji Monogatari around the 900s. In this book he was described as
a handsome guy with an attractive personality but without any skills
as a tough and sharp politician. His ‘sophisticated sense’ is in sharp
contrast with western-styled ‘dandyism’ (Yamada 1989) or ‘noble-
ness’ (Sakaiya 1996), with both of them requiring ‘critical spirit’ and
‘physical toughness.’ This unique sense of ‘Japanese sophistication’
was reinforced, accidentally, by the policy of Douglas MacArthur, or
the Supreme Commander of the Allied Powers (SCAP), who con-
trolled ‘occupied Japan’ just after the post-World War II period (from
1945 to 1951) and established Japanese policy afterward. In order to
make Japan a peaceful nation which would never revolt against the
United States, he implanted into the souls of all Japanese people,
quite successfully, the principle of denying any forms of violence
and, even physical power, as a solution of disputes (Sakaiya 1997).

Sakurai (1985) describes, in detail, the process in which the students’
movement in the 1970s was suppressed and critical spirit among stu-
dents in those days faded away. As a result, Japanese people came to
cherish their unique sense of ‘sophistication’ and deny critical spirit
or physical toughness.
The above two concepts of ‘follow-the-American-lifestyle’ and
‘Japanese sophistication’ combined and made an appropriate condi-
tion for creating strange settings in TV cartoons in which their titles
and main character names are dominated by Japanglish with strong
western taste. Apart from our real life settings, the imaginary world
of TV programs could include many people or places with western
names (according to the ‘follow-the-American-lifestyle’). In imagi-
nary settings, the battle scenes by giant fighter robots in children’s
programs did not accord to reality, and all the programs, for boys or
for girls, were filled with handsome and cute youngsters (according
to ‘Japanese sophistication’).
After the period of the ‘income-doubling national project’ in the
1960s, during which most Japanese people came to have TV sets at
home, typical boys’ hero programs and girls’ heroine programs were
established in the 1970s. With so few TV channels in those days, Japa-
nese boys watched the same programs at the same time and were strongly
influenced by them. It was the same for Japanese girls as well. All the
girls’ programs (Toei 2004) used western styled heroine names or titles
(and sometimes the settings were in Europe or America), and most of
the heroines were supposed to have the mystic power of witches:
(33) 㨱ἲno MA KO (Mako) chan: in 1970 (‘Chan’ is a hypocoristic suffix for

names.)
㨱ἲ౑i CHA P PI I (Chappie): in 1972 ( 㨱ἲ: magic skills) ( 㨱ἲ౑i: witch)
Miracle ᑡዪRI MI T TO (Limit) chan in 1973
KYU U TI I HA NI I (Cutie Honey): in 1973
㨱ዪᏊ ME GU (Meg) chan: in 1974
KYA N DI I KYA N DI I (Candy Candy): in 1976
At about the same time, boys’ hero programs (Saito 1998: 16) tended
to take more rigid forms as ‘Katakana robot or battleship names with
a few Chinese characters’ as follows:
(34) ㉸㟁☢RO BO KO N BA TO RA A ‘V’ (Robot Combattler Victory): in 1976

↓ᩛ㗰ேDA I TA A N ‘3’ (Daiturn Three): in 1978
ᶵືᡓኈGA N DA MU (Gundam): in 1979
㉸᫬✵せሰMA KU RO SU (Macross): in 1982
ᡓ㜚 ME KA ZA BU N GU RU (Mechanical Zabungle): in 1982
㔜ᡓᶵE RU GA I MU (L-Gaim): in 1984
According to Saito (1998: 21), a lot of scientific and technological

terminology in the above boys’ fighter robot stories were used to
mask the fake, imaginary nature of the programs. This usage may be
another evidence of ‘Japanese sophistication.’
While the above boys’ and girls’ TV programs contained a lot
of unrealistic western names, settings, or battle scenes, the feelings
of the heroes and heroines in those programs are quite easy for view-
ers to sympathize with (Craig 2000: 13). With this strong sympathy,
Japanese boys and girls in those days are thought to have enjoyed the
TV programs as if they had experienced virtual reality. In this virtual
reality, as in TV cartoon settings, we can find a lot of particular speech
styles that Japanese speakers never adopt in real situations: the hero’s
or heroine’s stylish speech styles as well as the terminology peculiar
to boys’ fighter robot stories or peculiar to girls’ cute witch stories.
Although these speech styles are based on imaginary settings and
never used in daily life, most Japanese speakers, young and old, can
recognize them as such because of the nationwide influence by TV
cartoons during the 1970s. These particular speech styles are the most
typical examples of ‘role-playing speech style’ named by Kinsui
(2003), and they do not belong to a particular ‘speech community’ in
a traditional sense (Murray 1998) in that they are imaginary: if we
are to define them as some particular ‘speech varieties’ (Lass 1984:
294-314; Lost 1990: 54-56) such as those based on setting, age, or
dialect variations, we need special categories such as the imaginary
(but prevalent) speech varieties concerning Japanese TV hero and
heroine variations.
The tradition of TV heroes and heroines for boys and girls was
passed on to the present generations in the 2000s, as the sons and
daughters of the youngsters in the 1970s. The following boys’ hero
programs (from Toei 1996) are revival versions of a popular hero
program in the 1970s:
(35) a. ௬㠃RA I DA A (Rider) 555 [2003]:

௬㠃RA I DA A (Rider) means a masked hero riding a stylish motor-
bike (rider).
Among the main characters is FA I ZU (spelled as 555, or Fives, or «’s).
b. ௬㠃RA I DA A (Rider) ๢Blade [2004]:
Blade (in English), Garren (in French), Chalice (in Old French),
Leangle (in Australian English)
c. ௬㠃RA I DA A (Rider) 㡪㨣[2005]: 㡪㨣ጾ྿㨣㎘㨣᩾㨣
d. ௬㠃RA I DA A (Rider) KA BU TO (ණ : beetle) [2006]: KA BU TO
Japanglish in the above programs or character names has become far

more complicated than before in that it includes Greek, old French,
and Chinese characters as well as English and French. Likewise, girls’
heroine programs (Toei 1996) have come to adopt complex Japanglish:
(36) a. ᫂᪥ no NA A JA (Nadja in the Future) [2004]: Heroine’s name is Nadja

Applefield.
b. Hutari wa PU RI KYU A (The Two Girls are Pretty and Cure) [2004]:
⨾ቚNagisa (Cure Black), 㞷ᇛHonoka (Cure White)
c. Hutari wa PU RI KYU A ‘Max Heart’ [2005]: same characters
d. Hutari wa PU RI KYU A ‘Splash Star’ [2006]: ᪥ྥဏ (Cure Bloom),
⨾⩧⯙(Cure Eaglet)
What is unique to girls’ programs is that Hiragana letters are used for
some heroines’ names (cf. the nature of Hiragana letters as familiar-
ity to women and children in Yasuda 2003: 195-196). As pointed out
by Saito (1998), it is true that the border between hero and heroine
cultures is disappearing, and that the stories or settings of young-
sters’ TV programs are getting more and more complicated, but, pos-
sibly due to the influence of parents now in their 30s, 40s, and 50s,
the new generation of children will keep the hero and heroine speech
style, not to mention a lot of Japanglish used in these programs.
6. Concluding Remarks
The period of rapid increase of Japanglish words and phrases coin-

cides with that of rapid economic growth in Japan during the 1970s.
Many people pointed out that, with the growth of exporting Japanese
products, the chances expanded for Japanese people to experience
foreign culture at hand. However, taking trips abroad in those days
was out of question for ordinary Japanese families, and the general
image of western culture was not based on facts but on some imagi-
nation associated with some imported products or foreign TV pro-
grams. Therefore, the rapid growth of Japanglish, of which the core
has been western loanwords, was triggered by domestic causes, not
by the influence from outside. Comparing the case of Japan with that
of Hong Kong, where the English language has always been at hand
since the ages of British administration, it is ironical that Cantonese
has not produced a large amount of westernized words and phrases
like ‘Cantonglish’ during the 1970s.
In terms of phonological structure, Japanglish is mainly com-
posed of loanwords and onomatopoeia. These two have many fac-
tors in common such as the tendency to use consonant gemination,
which is associated with the general image (‘swiftness’ or ‘speedi-
ness’) represented by loanwords or onomatopoeia. Furthermore, in
terms of spiritual factors, Japanglish gained a lot of popularity through
the mass media, which worked favorably to the ‘west-is-better ’ doc-
trine (Hakurai doctrine) since the 1500s, the ‘follow-the-American-
lifestyle’ doctrine since the 1960s, and the unique sense of ‘Japanese
sophistication’ since as early as the 900s. Both onomatopoeia as a
phonological factor and a unique sense of sophistication as a spiri-
tual factor are purely Japanese elements as old as the Japanese lan-
guage itself, not foreign ones. It was not until the 1970s that the time
was ripe for these two factors to combine, contributing to the prolif-
eration of Japanglish through the usage of Katakana letters in ono-
matopoeia and TV hero or heroine names.
In the discussion of the above Japanglish formation process,
one is reminded of the concept of ‘miniaturization’ peculiar to Japa-
nese culture, of which the most typical examples are shown in Yomota
(2006): Haiku (Japanese-styled short poem) and PURIKURA (print

club). Haiku was established much earlier than the Edo period (the
1600s) on the basis of imported Chinese poems, and it developed as
much shorter forms than the original Chinese ones. While Chinese
poems are composed only of Chinese characters, or ideograms, Japa-
nese Haiku is composed of Chinese characters and Hiragana letters,
or phonograms. Haiku, therefore, seems less effective in packing a
lot of ideas than Chinese poems, but it gained popularity among the
general public by its rhythmical structure. Secondly, PURIKURA is
a booth in which people have their photographs taken against the
backgrounds of their own choices. Since its advent in 1995, it has
quickly become popular among Japanese young girls. According to
Yomota, what is special about PURIKURA is not that it is a minia-
ture-sized photograph, but that it focuses on favorable aspects of re-
ality and shows people as if they were in an attractive, imaginary
world. As in Haiku and PURIKURA, we can discuss the usage of
Katakana words especially in the 1970s in terms of ‘miniaturization’
concept: Japanglish tends to be abbreviated in a stable length (three
or four morae) and has peculiar phonological values, such as rhyth-
mic ones, which are different from usual Japanese words and phrases;
hero and heroine TV programs in the 1970s, devoid of unfavorable
reality of our daily life, were uniquely sophisticated ones, which
worked as good tools for the expansion of the usage of Japanglish.
In terms of the phonetic and phonological properties, Katakana
phonograms and Hiragana phonograms are completely the same in
Japanese. If, however, it had not been for Katakana letters, Japanglish
could not have increased as much as in the present situation, because
Katakana letters had been completely different from Hiragana let-
ters in Japanese orthographical system: Katakana letters are indica-
tors of ‘marked’ words or phrases. In contrast with the Cantonese
language, with its very rigid and effective system of Chinese charac-
ters and alphabet letters, the Japanese language multiplied strange
Japanglish in creative and sloppy ways on the basis of Katakana let-
ters, along with the ‘follow-the-American-lifestyle’ doctrine and the
unique sense of ‘Japanese sophistication.’
References
Abe, Hajime 1990. Katakanago no Jooshiki Hijooshiki (What is True

or Untrue about Japanese English). Tokyo: Tokyoshoseki.
Akiba, Yoshinobu / Chikuma, Tadao 2000. Hittoshoohin Neemingu
no Himitsu (Secrets about the Brand Names for Big Selling
Products). Tokyo: Koodansha.
Amanuma, Yasushi 1974. Giongo Gitaigo Jiten (Dictionary of Ono-
matopoeia and Mimesis). Tokyo: Tokyodoshuppan.
Arakawa, Soobee 1977. Dictionary of Loan-Words: Second Edition
by Kadokawa. Tokyo: Kadokawashoten.
Bunkacho (Agency for Cultural Affairs in Japan) (ed.) 1997. Kotoba
ni Kansuru Mondooshuu [Shin Kotoba Shiriizu 6]: Gairaigo
Hen (Question and Answer on Language [New Series on Lan-
guage, No. 6]: Loanwords). Tokyo: Ookurasho Insatsukyoku
(Printing Office for the Japanese Ministry of Finance).
Bunkacho (Agency for Cultural Affairs in Japan) (ed.) 1998. Kotoba
ni Kansuru Mondooshuu [Shin Kotoba Shiriizu 8]: Gairaigo
Hen (Question and Answer on Language [New Series on Lan-
guage, No. 8]: Loanwords 2). Tokyo: Ookurasho Insatsukyoku
(Printing Office for the Japanese Ministry of Finance).
Craig, Timothy J. (ed.) 2000. Japan Pop!: Inside the World of Japa-
nese Popular Culture. New York: M.E. Sharpe, Inc.
Hamano, Shoko Saito 1986. The Sound-Symbolic System of Japanese.
Ph.D. Dissertation. Gainesville, Florida: University of Florida.
Ishiwata, Toshio 1985. Nihongo no Naka no Gaikokugo (Foreign
Words and Phrases in Japanese). Tokyo: Iwanamishoten.
Ito, Junko 1990. ‘Prosodic Minimality in Japanese,’ Chapters from
the 26th Regional Meeting of the Chicago Linguistic Society:
The Parasession on the Syllable in Phonetics and Phonology.
Karen Deaton, Manuela Noske, and Michael Ziolkowski (eds).
Chicago: Chicago Linguistic Society, 213-239.
Ito, Mami 2003. Tabi no Yubisashi Kaiwacho No. 3: Honkon [Dai Ni
Han] (Travellers’ Conversational Dictionary with Pictures,
No. 3: Hong Kong [Second Edition]). Tokyo: Joohoo Sentaa
Shuppankyoku.
Kahn, Daniel 1976. Syllable-based Generalizations in English Phono-

logy. New York: Garland Publishing, Inc.
Kajima, Shoozoo 1994. Katakana Eigo no Hanashi (Stories on Japa-
nese English). Tokyo: Nan’undo.
Kimura, Kazuhisa 1998. Neemingu Taizen: Hitto Shoohin Mei no
Himitsu o Saguru (Collection of Great Brand Names: Key to
Big Selling Products). Tokyo: Jitsumu Kyouiku Shuppan.
Kinsui, Satoshi 2003. Baacharu Nihongo: Yakuwarigo no Nazo (The
Japanese Language as a Virtual Reality: Mystery about the
Role-Playing Speech Style). Tokyo: Iwanamishoten.
Koizumi, Tamotsu 1978. Nihongo no Seeshohou (Japanese Ortho-
graphy). Tokyo: Taishuukanshoten.
Kubozono, Haruo 1989. ‘The Mora and Syllable Structure in Japa-
nese: Evidence from Speech Errors,’ Language and Speech,
Vol. 32, No. 3, 249-278.
Lass, Roger 1984. Phonology. Cambridge: Cambridge University Press.
Lovins, Julie Beth 1975. Loanwords and the Phonological Structure
of Japanese. Bloomington, Indiana: Indiana University Lin-
guistics Club.
Marchand, Hans 1969. The Categories and Types of Present-Day
English Word-Formation. München: C.H. Beck’sche Verlags-
buchhandlung.
Morioka, Kenji 1993. ‘Nihongo no Goi no Taikei to Rekishi (The
Category and History of the Japanese Vocabulary),’ Shuueisha
Kokugo Jiten (Shueisha Japanese Dictionary). Kenji Morioka,
Tokugawa Munemasa, Kawabata Yoshiaki, Nakamura Akira,
Hoshino Kouichi (eds). Tokyo: Shueisha, 1900-1910.
Mori, Yoko 2002. ‘Formation of Three-mora Compound Abbrevia-
tions,’ Journal of the Phonetic Society of Japan, Vol. 6, No.1,
121-137.
Murray, Stephen O. 1998. American Sociolinguistics: Theorists and
Theory Groups. Philadelphia: John Benjamins Publishing Com-
pany.
Ono, Koji 1991. ‘/Q/-Insertion Rule in English Words in Japanese,’
Journal of the Linguistic Society of Japan, No. 100, 67-88.
Poser, William J. 1990. ‘Evidence for Foot Structure in Japanese,’
Language Vol. 66,No. 1, 78-105.
Rost, Michael 1990. Listening in Language Learning. London: Long-

man.
Saito, Minako 1998. Kou Itten Ron: Anime Tokusatsu Denki no Hiroin
Zou (Theory on the Only Heroine among Male Characters in
Cartoons, Monster Dramas, and Biographies). Tokyo:
Chikumashobo.
Sakaiya, Taichi 1996. Nihon o Tsukutta 12 nin: Zenpen (Twelve Great
Figures who has Established Japan: the First Half Story). To-
kyo: PHP Research Center.
Sakaiya, Taichi 1997. Nihon o Tsukutta 12 nin: Kohen (Twelve Great
Figures who has Established Japan: the Latter Half Story). To-
kyo: PHP Research Center.
Sakurai, Tetsuo 1985. Kotoba o Ushinatta Wakamonotachi (Young
People who have Lost Words). Tokyo: Kodansha.
Selkirk, Elisabeth O. 1982. The Syllable. In Van der Hulst, Harry /
Smith, Norval (eds) The Structure of Phonological Represen-
tations 2. Dordrecht: Foris Publications.
Stanlaw, James 2004. Japanese English: Language and Culture Con-
tact. Hong Kong: Hong Kong University Press.
Tabata, Toshiyuki 2004. ‘Eigo ni Okeru Kashiramojigo Keisei no
On’inteki Sokumen (Phonological Aspect of the Acronym-For-
mation in English),’ Shogengo no On’in Kouzou to On’in Riron
ni Kansuru Sougouteki Kenkyuu [A], No. 1: Kenkyuu Seika
Houkokusho (Comprehensive Study on the Phonological Struc-
ture and Theory on Various Languages [A], No. 1: Research
Results). Shosuke Haraguchi (ed.). Tokyo: Ministry of Educa-
tion, Culture, Sports, Science, and Technology, 73-84.
Tamori, Ikuhiro 2002. Onomatope Gi’on Gitaigo o Tanoshimu (En-
joying Onomatopoeia and Mimesis). Tokyo: Iwanamishoten.
Toei 1996. ‘TOEI TV WEBSITE.’ (<http://www.toei.co.jp/tv/>).
Toei 2004. ‘Toei Animeeshon: Sakuhin Rain Nappu (Introduction of the
Toei Cartoons).’ (<http://www.toei-anim.co.jp/tv/index.html>).
Vance, Timothy J. 1987. An Introduction to Japanese Phonology.
Albany, New York: State University of New York Press.
Yamada, Masaru 1989. Dandizumu: Kizoku Shumi to Kindai Bunmei
Hihan (Dandyism: Nobleness and Criticism of Modern Civili-
zation). Tokyo: Hippon Hoso Shuppan Kyoukai.
Yasuda, Teruo 2003. Ano Neeming wa Sugokatta (Looking back on

the Great Brand Names). Tokyo: Chukeishuppan.
Yomota, Inuhiko 2006. Kawaii Ron (Theory on the Word ‘Kawaii’).
Tokyo: Chikumashobo.
Yazaki, Genkuroo 1964. Nihon no Gairaigo (Loanwords in Japa-
nese). Tokyo: Iwanamishoten.
271
WONG PING WAI
Semantic Annotation of Chinese Texts

with Message Structures Based on HowNet
1. Introduction
Corpus annotation is not just a practical task of incorporating lin-

guistic information to plain texts, it also sheds new light on the na-
ture of language and the most effective means of analyzing it. This
chapter reports on the task of using a knowledge base called HowNet
to annotate Chinese texts with semantic information. The annotation
method is Message Structure, which provides an effective way to
analyze Chinese word senses and semantic dependency between
words.
2. Corpus and Corpus Annotation
A corpus is a collection of texts, usually in an electronic form, which

may be processed by computers for various purposes, such as lin-
guistic research and information technology. A corpus is useful only
if we can extract information from it. However, limited information
can be retrieved directly from a raw corpus since linguistic informa-
tion is always implicit in plain texts. That is why we need to make
such implicit information explicit by building in interpretative, lin-
guistic information to the corpus. This process is called corpus anno-
tation.
272 Wong Ping Wai
3. Annotated Chinese Corpora
Efforts of annotating Chinese corpora began in the 1990s. For ex-

ample, the tokenized corpus, e.g., the PH Corpus (Guo 1993), the
parts-of-speech tagged corpora, e.g., the Sinica Corpus (CKIP 1995)
and the PKU corpus (Yu et al. 2003), the syntactically annotated cor-
pora, e.g., the Sinica Treebank (Huang et al. 2000) and the Penn
Chinese Treebank (Xia 2000). Recent efforts turned to a deeper an-
notation level – semantic annotation of Chinese texts.
3.1 Semantic Annotation
Semantic annotation is the process to incorporate semantic informa-

tion into a corpus by attaching semantic tags to words in the text.
There are two broad types of semantic information (McEnery/Wil-
son, 2001):
• Semantic relationships between items – essentially the anno-
tation of the participants involved in events or states.
• Semantic features of words – the annotation of word senses.
3.1.1 Sense Tagged Chinese Corpora
There were a few attempts to sense tag the Sinica treebank and the
Penn Chinese Treebank (for examples, cf. Dang, Chia, Palmer and
Chiou (2002) and Ker and Chen (2004)), but such resources are not
publicly available for research. Neither is the largest sense-tagged
Chinese corpus (one million words) tagged by Li et al. (2003). The
IR-lab of Harbin Institute of Technology (2005) has released a small
corpus of 1,000 sentences (around 21,000 words) to the public in
2005. Like Li et al. (2003), the sense inventory in this study is the
Tongyici Cilin (Thesaurus) (Mei 1983), which uses sense ID num-
bers to represent the hierarchical classification of concepts. The dis-
advantage is that human beings cannot interpret the meanings from
the ID numbers directly, but must check against the matching tables.
Semantic Annotation of Chinese Texts 273
3.1.1.1 HowNet used in Gan and Tham (1999)

The advantage of the sense tags of HowNet (Dong, Z. / Dong, Q.
1999), however, lies in the design that human beings can get the
meanings by reading the tags (details in Section 4). Gan and Tham
(1999) has annotated a small Chinese corpus with word senses de-
fined in HowNet. The corpus consists of 30,976 words, which is a
subset of the Sinica Corpus (C.K.I.P. 1995). The corpus was released
free for research1 in January 2002.
Based on the semantic features revealed in the word senses of
HowNet, this chapter would further proceed to annotate HowNet
semantic dependency relations between words in the corpus, so as to
enrich this resource for NLP research, e.g., word sense disambigua-
tion and automatic dependency annotation.
3.1.2 Semantic Relations
Another type of semantic annotation is the annotation of semantic

relations. Most of the projects focus on the predicate-argument struc-
ture (Xue/Palmer 2005). For example, FrameNet (Baker/Fillmore/
Lowe 1998), Sinica Treebank (Chen et al. 2004) and PropBank (Xue/
Palmer 2005). These works are limited to the labeling of semantic
roles of the noun phrases in relation to the main verb in a sentence.
3.1.2.1 Semantic Dependency of HowNet

Semantic dependency annotation includes more than semantic role
labeling, because it involves the semantic relations between every
individual word in a sentence. There are two works of annotated
Chinese corpora with semantic dependency based on HowNet.
• a small corpus annotated by Gan and Wong (2000) based on
the HowNet message structures released in 2000. It is quite
small, so we need to further expand the corpus to provide es-
sential resource for NLP research.
1 <http://godel.iis.sinica.edu.tw/CKIP/hk/index.html>.
274 Wong Ping Wai
• Semantic Dependency Net (SDN) – a large corpus of one mil-

lion words annotated by Li et al. (2003) with HowNet seman-
tic dependency.
Both conducted research on semantic annotation and bypassed syn-

tactic annotation, thus parts of speech (POS) and syntactic structures
were not annotated in both works. The main difference between them
lies in the sense inventories.
Gan and Wong (2000) applied the whole framework of HowNet,
including both word senses and semantic dependency. Thus, a con-
sistent HowNet-based analysis of semantic features and dependency
is maintained. Li et al. (2003) only used the part of semantic depen-
dency, but tagged word senses using another framework – Tongyici
Cilin (Thesaurus) (Mei 1983). It is hard to make a linkage between
the two frameworks.
However, the corpus of Li et al. (2003) is not released to the
public. Gan and Wong (2000) is the only available resource about
HowNet’s semantic dependency annotation, on which this chapter is
based.
4. HowNet knowledge base
HowNet is an online common-sense knowledge base, which describes

inter-concept relations and inter-attribute relations of concepts. There
are 65,000 Chinese concepts and 75,000 English equivalents. HowNet
2000 is for free download, but it requires licensing fee for later ver-
sions, HowNet 2002-2005. The current study uses the old version
HowNet 2000, not because it is free, but because the HowNet corpus
by Gan and Tham (1999) has used the sense inventory HowNet 2000.
Message Structures Base is also based on HowNet 2000. For these
reasons, the present chapter also uses the 2000 version of HowNet.
4.1 A sample entry of HowNet 2000
The following shows a sample lexical entry of HowNet:
NO.=040263
W_C=姀侭
G_C=N
E_C=
W_E=journalist
G_E=N
E_E=
DEF=human|Ṣ,#occupation|借ỵ,*gather|㍉普,*compile|䶐廗,#news|㕘倆
Key: # related to * agent of
The first line is the record number. The Chinese word is shown in
“W_C”, with the POS and example shown in “G_C” and “E_C”. The
English word and the corresponding POS and example are shown in
“W_E”, “G_E” and “E_E”. The concept definition is shown in “DEF”.
The approach to meaning representation is based on compo-
nential analysis. The sememe appearing in the first position of “DEF”
– ‘human| Ṣ ’ is called the categorical attribute. It names the hyper-
nym or the superordinate term, which gives a general classification
of the concept. The sememes appearing in other positions: ‘occupa-
tion| 借ỵ ’, ‘gather| ㍉普 ’, ‘compile| 䶐廗 , ‘news| 㕘倆’ are additional
attributes, which provide more specific, distinguishing features.
The definition tells us that journalist is a kind of human, who
is the agent of gathering and compiling news. This concept is related
to occupation.
There are four main categories of content words in the HowNet
dictionary. The example mentioned above is an Entity concept, which
mainly consists of nouns.
Another category is Event, which mainly contains verbs, e.g.,
“ ⬀㫦 ” (savings).
The third category is Attribute, which contains nouns referring
to attributes, e.g., “Ṗ⹎ ” (brightness). Quantity is a subcategory of
Attribute, e.g., “ ᾵ ” (times).
276 Wong Ping Wai
The fourth category is Attribute Value, which mainly includes

adjectives, e.g., “ ⃱Ṗ ” (bright). Quantity Value is a subcategory of
Attribute Value, e.g., “ Ḵ ” (two), “ ᶱ ” (three).
4.2 Dependency Grammar
The lexicon of HowNet has been illustrated in 4.1. The grammar of

HowNet will be illustrated in this section. HowNet adopts the model
called Dependency Grammar (DG). It was developed by Tesnière
(1959) in the 1950s. Different from phrase structure grammars, DG
does not have non-terminal nodes. Structures are labeled by the rela-
tion between a head and its dependents.
4.2.1 Governor and Dependent
According to the convention of Mel’cuk (1988), dependency is rep-

resented by labeled arrows. For example, “I will read the book” is
represented as follows:
(1) read
object
subject auxiliary
I will book
determiner
the
X → Y means (e.g. X=“read”, Y=“I”):
• “Y depends on X” or “X governs Y”, e.g. “I” depends on “read” or “read”
governs “I”
• X - the governor/head of Y, e.g. “read” is the governor of “I”
• Y - the dependent of X, e.g. “I” is the dependent of “read”
The relation is asymmetric such that the label, e.g. “subject” and “auxil-
iary” indicates the relation of dependent to the head. For example, the
label “subject” on the arrow “I ← read” means: “I” is the subject of

“read”. The label “auxiliary” on “will ← read” means “will” is the
auxiliary of “read”.
4.2.2 HowNet’s Semantic Dependency
This chapter adopts the conventions described in Gan and Wong (2000)
for the representation of dependency relations, which are based on,
but also a modification of, Mel’cuk (1988), as shown in the following:
(2)
C2 C2
R1 , R2
C1 C1
HowNet has incorporated an element of Link Grammar (Sleator/

Temperley 1991, 1993), in which dependency relations can be bi-
directional.
In the figure above, C stands for concept whereas R stands for
semantic dependency relations. C2 and C1 are governor and depen-
dent, respectively. The dependency relations are bi-directional. R2
and R1 are distinguished by a comma.
R1 is read as: C1 is the R1 of C2;

R2 is read as: C2 is the R2 of C1.
Another way of representing semantic dependency that is widely used

in Gan and Wong (2000) is shown in the following. This format does
not have any arrows. The R1 and R2 are denoted by the relevant
positions of C1 and C2.
Concept Dependency Relation Concept

C2 R2
R1 C1
278 Wong Ping Wai
In the figure above, C2 is located higher than C1. Between them are
the dependency relations R1, R2 or both. C2 is the governor which
governs the dependent C1. R1 is on the same level as C1 whereas R2
is located on the same level as C2.
5. Methodology of Corpus Creation and Analysis
This chapter follows the strategy of analyzing Chinese corpora imple-

mented by Gan and Tham (1999) and Gan and Wong (2000). They
have chosen the ‘crime’ domain of the Sinica Corpus but have only
annotated 55% of the texts. This chapter will finish the remaining
parts following their methods. The methods consist of four steps:
1. Sentence Breaking ( 㕟⎍ )
2. Concept Group Extraction ( ㉥⍾㤪⾝佌 )
3. Sense Tagging ( 娆佑㧁姣 )
4. Message Structure Identification ( ⺢䩳ᾉ〗䳸㥳 )
5.1 Sentence Breaking
A sentence is broken by certain punctuation marks, namely, a full

stop, a comma, a semicolon, a colon, a question mark or an exclama-
tion mark, as shown in the following:
ˤ炻烊烉烎炰
Other punctuation marks would not be regarded as the end of a sen-

tence:
ˣ( )˪˫ˬ˭
ˮ˯“ ”‘ ’炽ʇʇ……
5.2 Concept Group Extraction
This is the term used in Gan and Wong (2000) and Wang (2002). It is
similar to word segmentation. This process is assumed to be mainly
solved in the Sinica Corpus, in which each word is already segmented
and tagged with POS.
However, there are some differences in the criteria of segmen-
tation between the Sinica Corpus and HowNet. This chapter aims to
find out the conversion rules to make adjustments before the step of
sense tagging.
The following tables (person names and place names) show
some examples of the adjustments made by this chapter to the subset
of the Sinica corpus according to HowNet.
Persons Places
Patterns Sinica HowNet Patterns Sinica HowNet
~⩴ (suspect) 㜿⩴ 㜿 ⩴ ~ⶪ (city) ⎘ᷕⶪ ⎘ᷕ ⶪ
~㝸 (someone) 䌳㝸䌳㝸 ~㷗 (harbour) 檀晬㷗檀晬㷗
~⤛ (girl) 惕⤛ 惕 ⤛ ~䷋ (county) ⼘⊾䷋ ⼘⊾ ䷋
~⨎ (woman) 湫⨎ 湫 ⨎ ~悱 (village) 䤷冰悱䤷冰悱
~⥣ (wife) 㜿⥣ 㜿 ⥣ ~捖 (town) 㔿⋿捖㔿⋿ 捖
~㭵 (mother) 昛㭵昛㭵 ~⋨ (district) ⇵慹⋨ ⇵慹 ⋨
~䇞 (father) 昛䇞昛䇞 ~埿 (street) ⃩㔯埿 ⃩㔯埿
~⭞ (family) 昛⭞ 昛 ⭞ ~慴 (lane) ␴⸛慴 ␴⸛ 慴
Table 1. Different Segmentation Criteria between the Sinica corpus and HowNet.
5.3 Sense Tagging
A word may contain more than one sense. It is necessary to find out
the most likely senses and prune out the unlikely ones during the
process of sense tagging (both manual and automatic). It is achieved
by refrence to relevant clues in the context as well as the hand-crafted
knowledge base of HowNet.
280 Wong Ping Wai
5.4 Message Structures Identification
After each word is tagged with HowNet senses, the next step is to
identify message structures. Message structures extend the research
on the HowNet knowledge base to the research on a specific lan-
guage – Chinese (Z. Dong, 2000). Different languages can have dif-
ferent message structures based on the same knowledge base. Mes-
sage structures are formed by words or phrases (two or more), which
are syntactically and semantically sound and convey a message.
A message structure is specified by the following format:
(sememe) [R2] → [Rl] (sememe)
It is noted that the patterns found in word formation are also found in
phrase level and sentence level in Chinese. Take for example, the
pattern “(time| 㗪攻) [time] ← (event| ḳẞ)” can be found at vari-
ous levels (Gan and Wong, 2000):
(3)
word level: “㘐[time]ɤ忳” (exercise in the morning)
phrase level: “ℓ⸜[time]ɤ㈿㇘” (war against aggression for eight years)
sentence level: “4 㚰 6 㖍㗇㛇㖍[time]ɤ㓦`” (It is a holiday on Sunday,
the 6th of April)
Almost all examples in the HowNet message structures are at word

or phrase level, but the specialty of Chinese is that the word forma-
tion patterns are also applicable to sentence level, so message struc-
tures are used to annotate the Chinese corpus. The following shows
an example sentence:
5.4.1 Word Sense Definitions
(4) 㞍㋽徫䉗⼴ ⌛ 㓄ⷞ 嬎㥵
Catch escaped convict after immediately carry gun
“Carried the gun after catching the escaped convict.”
The following table shows the word sense tagging:
Concepts HowNet definitions

㞍㋽ investigate|㞍,catch|㋱ỷ,police|嬎
徫䉗 human|Ṣ,crime|伒,undesired|午,*escape|徫嵹
⼴ time|㗪攻,future|⮯
⌛ aValue|Ⱄ⿏ῤ,duration|ᷭ㙓,TimeShort|㙓
㓄ⷞ bring|㓄ⷞ
嬎㥵 weapon|㬎☐,*shoot|⮬㑲
炻 punc|㧁溆
Table 2. Sense tagging according to HowNet definitions.
5.4.2 Dependency Relations
The dependency relations among the words in a sentence are shown

below:
(5) 炻
kernel
㓄ⷞ!
time patient
manner
⼴! ⌛! 嬎㥵!
,time
㞍㋽!
content
徫䉗!
282 Wong Ping Wai
5.4.3 Message Conveyed in the Sentence
By analyzing the message structures shown above, the computer can

extract the following message:
• Someone is carrying a “gun” ( 嬎㥵)
• A gun is a weapon for shooting.
• The time of someone “carrying a gun” ( 㓄ⷞ嬎㥵 ) is “after
catching an escaped convict” ( 㞍㋽徫䉗⼴ ).
6. The Subset Chosen from Sinica Corpus
Each text in the Sinica Corpus has the text information at the begin-
ning:
%% 㔯栆=⟙⮶
%% 㔯橼=姀㔀
%% 婆⺷=written
%% ᷣ柴=䉗伒
%% ⨺橼=⟙䳁
%% ⥻⎵=
%% ⿏⇍=䓟⤛
%% ⚳䯵=ᷕ厗㮹⚳
%% 㭵婆=ᷕ㔯
%% ↢䇰╖ỵ=冒䓙㗪⟙
%% ↢䇰⛘=冢䀋
%% ↢䇰㖍㛇=199012
%% 䇰㫉=
Gan and Tham (1999) have chosen the text with the theme “crime”
“ ᷣ柴=䉗伒 ”, and manually annotated 103 texts. I checked the Sinica
corpus again and found that there should be 193 texts in total, 90
texts of which are not tagged yet.
Manual tagging by Gan and Tham (1999) is time-consuming.
Thus, this study has processed the remaining part with the help of an
automatic parser, followed by human proofreading.
6.1 Format of the HowNet Corpus
This format is converted to XML format in Gan and Tham (1999).

The POS tag is converted to <syntax> with the corresponding end
tag </syntax>.
A word is surrounded by three kinds of tags – from outermost
to innermost: <syntax>, <semantic>, <relation> – and their corre-
sponding end tags.
The <semantic> tag has the attribute class, which reveals the
sense definition according to HowNet’s dictionary.
The <relation> tag usually has two attributes:
(i) head specifies the head or governor of the word.
(ii) name specifies the semantic dependency relation involved.
The following shows a fragment of the file “crime.f80014c.30.xml”:
(6)
<2051>
<syntax class="PERIODCATEGORY"><semantic class="{㧁溆},"><relation head="0" name="duplicate">ˤ
</relation></semantic></syntax>
<syntax class="VJ"><semantic class="䘤䓇,"><relation head="4" name="restrictive">䘤䓇
<syntax class="Na"><semantic class="ḳね,午,#㭢⭛,#㬣,伒,"><relation head="2" name="existent">┳埨㟰
<syntax class="DE"><semantic class="{㥳≑},"><relation head="8" name="dummy">䘬
<syntax class="Nb"><semantic class="㔯⫿,⥻⎵,⮰,"><relation head="8" name="restrictive">㕘ᶾ凒
<syntax class="Neu"><semantic class="㔠慷ῤ,㫉⸷,䫔,"><relation head="7" name="restrictive">⺧ᶨ
<syntax class="Nf"><semantic class="䫎嘇,"><relation head="8" name="restrictive">嘇
<syntax class="Na"><semantic class="凡,*㋱ỷ,#欂,彚,"><relation head="9" name="dummy">㺩凡
<syntax class="COMMACATEGORY"><semantic class="{㧁溆},"><relation head="0" name="EOS">炻
</2051>
Each word in a sentence has an index number. In the excerpt of the file
“crime.f80014c.30.xml” above, there are nine words in the sentence
(line no. 2051). Each word is represented by an index, beginning with 1.
This hidden information is listed explicitly in the following table.
284 Wong Ping Wai
Word Index Head Relation

ˤ 1 0 duplicate
䘤䓇 2 4 restrictive
┳埨㟰 3 2 existent
䘬 4 8 dummy
㕘ᶾ凒 5 8 restrictive
⺧ᶨ 6 7 restrictive
嘇 7 8 restrictive
㺩凡 8 9 kernel
炻 9 0 EOS
Table 3. Semantic Dependency relations between the various words.
It should be noted that in the Sinica Corpus, each sentence begins

with a duplicate punctuation that is the same as the punctuation at
the end of the previous sentence. Thus, the first word does not have a
head (indicated by head = “0”), and the relation is named ‘dupli-
cate’. The second word “ 䘤䓇 ” has the head word “䘬 ” (Index = 4),
and has a ‘restrictive’ relation to the head word.
7. Results and Analysis
7.1 Manual Tagging of Semantic Dependency
The Message Structures base contains patterns of Chinese words and

phrases. In order to annotate relations between words at a sentence
level, there is a need to add some annotation rules.
7.1.1 Punctuations to mark the EOS
Each sentence ends with a punctuation mark defined in 5.1. The top-
most node is the punctuation mark at the end of the sentence, as
shown in the following. In the XML format, as it is the topmost one,
it does not have a head (indicated by head=“0”), and it marks the
“EOS” – “end of sentence”.
ˤ 炻烊烉烎炰
7.1.2 Roots and Events
The root of the sentence is “ ⛐ ”, as marked by “kernel” and its head

is the “EOS” punctuation mark “炻”.
(7) ⛐ ᶨ 䇯びや倚 ᷕ 炻
PREP one CL pity sound LOC
“In the midst of pity”
炻
kernel ⛐
scope ᷕ
scope 倚
content びや
quantity ᶨ
quantity 䇯
Sentential conjunctions are always the head of a sentence. The next

priority is always an event. In the following example, the root is the
verb “ 溻⊝ ” that denotes the event ‘urge| Ὣἧ . The event role frame
is: {agent, patient, ResultEvent}, which provides essential informa-
tion for linking other elements which act as the participants of the
event.
For example, the patient role is played by “ ⼺䛦 ”. Another
event denoted by “ 崟ḳ ” plays the ResultEvent role of the event
denoted by “ 溻⊝ ”. The agent is left implicit in this case. The par-
ticipants are always dependents of the main event.
286 Wong Ping Wai
(8) 溻⊝ ⼺䛦崟ḳ炻(f79109.4)
encourage people uprise
“Encourage people to uprise”.
炻
kernel 溻⊝
patient ⼺䛦
ResultEvent 崟ḳ
The semantic dependency between words can always be found in the

Message Structures base. For example, the classifier phrase “ ᶨ-䇯”
matches the pattern 4.2.5. Therefore, the dependent “ 䇯” is tagged as
the quantity to the head “ ᶨ ”.
(9) Pattern 4.2.5 (㔠慷ῤ) → [㔠慷] (⎵慷)

e.g. ᶨ → [quantity] 䇯
If there is not an exact matching of the sememes in the Message

Structures base, annotators must resort to the hypernyms of the con-
cepts. For example, there is not an exact matching of the sense tags
of “ びや ” and “ 倚”, but the former is ‘event| ḳẞ ’ whereas the lat-
ter is ‘things | 叔䈑 ’ according to the ontology of HowNet. This
matches the pattern 1.3.3.13, so content is used for tagging.
(10) Pattern 1.3.3.13 (ḳẞ,埴≽) [ℏ⭡] ← (叔䈑)
7.1.3 Unresolved
When the element in the sentence cannot be linked, unresolved is

used for annotation. In (11), the combined concept “ 㚱斄㨇斄” can-
not be linked to the other elements in the sentence. It is because it
plays the experiencer role of the event (denoted by the verb “ 侫ㄖ ”)
which will appear in the next sentence (shown in (12)). Thus, unre-
solved is tagged for the head of “ 㚱斄㨇斄 ” – “ 㨇斄 ”. It will be
linked to “ 侫ㄖ ” in the stage of second parse (which deals with event
relation in a group of sentences) in future research.
(11) 㚱斄ġ 㨇斄 ⛐ 䞗㱣䉗伒 ᶲ 炻 (f79109.1, s12)

Concerned department PREP correct crime LOC
“As for the correction of crime, the concerned department,”
炻
dummy ⛐
scope ᶲ
scope 䞗㱣
patient 䉗伒
unresolved 㨇斄
modifier 㚱斄
(12) ⇯ ㅱ 侫ㄖ曺⮹⸜ 䉗伒埴䁢 ⍲ ⽫䎮炻 (f79109.1, s13)

CONJ should consider juvenile crime behaviour and mentality
“Should consider the behaviour and mentality of juvenile crime.”
7.2 Improving the Sense Tagging of Gan and Tham (1999)
The sense tagging process of Gan and Tham (1999) was done on a
running text, without a concordance tool for further proofreading.
There might be some inconsistencies of sense tags. I have made some
corrections after the examination of word senses with a concordance
tool.
7.2.1 Update Sense Tags of HowNet 1999 to HowNet 2000
The annotation of word senses by Gan and Tham (1999) could only
rely on HowNet 1999, because HowNet 2000 was not available at
that time. It is not clear about the content of HowNet 1999, as it was
no longer available for download since the release of HowNet 2000
in October 2000. I have extracted all the senses found in the corpus
and checked against the HowNet 2000 dictionary. The following table
shows the statistics:
288 Wong Ping Wai
Corpus of Gan and Tham (1999)

Total senses definitions 2447
match HowNet 2000 1976
does not match HowNet 2000 471 219 new entries
252 discrepancies
Table 4. Statistics of word senses in Gan and Tham (1999).
There are 2,447 sense definitions in the corpus, 1,976 of which match
the sense definitions in HowNet 2000, whereas 471 do not. Out of
471, there are 219 new entries that are not found in the HowNet 2000
dictionary. They were added by Gan and Tham (1999). I have exam-
ined each of them carefully, and added them to HowNet 2000 (cus-
tomized dictionary). As for the remaining 252 sense definitions, they
constitute the discrepancies between the HowNet 2000 dictionary
and the corpus. In most cases, the definitions would follow HowNet
2000.
For example, “ 㔁ⷓ ” and “ 侩ⷓ ” are translated to “teachers”
in English, but their usage is different. The former is often associated
with an occupation whereas the latter is used as an address term to a
teacher. The difference is reflected in the sememe ‘#occupation| 借ỵ ’
of sense definitions in HowNet 2000. Perhaps there was not a dis-
tinction in HowNet 1999, and neither was there in the corpus of Gan
and Tham (1999). I accepted the changes of sense definition of
HowNet 2000, and made the changes to the corpus accordingly.
Word Definition in the corpus HowNet 2000 dictionary

“㔁ⷓ” human|Ṣ,*teach|㔁,education| human|Ṣ,#occupation|借ỵ,*teach|
(teacher) 㔁做㔁,education|㔁做
“侩ⷓ” human|Ṣ,*teach|㔁,education| human|Ṣ,*teach|㔁,education|㔁做
(teacher) 㔁做
Table 5. Difference of sense definition between Gan and Tham’s corpus and HowNet
2000.
7.2.2 Reducing Inconsistency of Sense Tags
There were a few inconsistencies of sense tags in Gan and Tham

(1999). For example, the numerals (except one and zero) are tagged
with two kinds of sense tags in the corpus, as shown in the following
table:
No. of Tokens in HowNet

Sense Tags
Corpus (Gan and Tham 1999)
1 qValue|㔠慷ῤ,amount|⣂⮹,mass|䛦,cardinal|➢, 117
2 qValue|㔠慷ῤ,amount|⣂⮹,cardinal|➢,mass|䛦, 286
Table 6. Inconsistent sense tags in Gan and Tham’s corpus.
‘Cardinal| ➢ ’ and ‘mass| 䛦 ’ are Secondary Features in HowNet 2000.

The sequence in the definition above can be reversed without chang-
ing the meaning. But in the HowNet 2000 dictionary only the latter
can be found. Thus, the first sense is converted to the second one in
the corpus.
7.2.3 Correcting Minor Mistakes
Two words in the corpus were not sense tagged. I thus manually
tagged them according to the knowledge acquired from the corpus.
Word Tag Correction File

“⋫↮ᷳᶨ” XX qValue|㔠慷ῤ,amount|⣂⮹,few|⮹, crmie.f79109.3.xml
“暞溆ᶨ暞” XX aValue|㔠慷ῤ,amount|⣂⮹,cardinal| crime.f80014a.5.xml
➢,mass|䛦,
Table 7. Missing Sense Tags in Gan and Tham’s corpus.
Each sememe is followed by a comma in the corpus. Even the

last sememe in a sense definition must end with a comma, e.g.
‘㔠慷ῤ,⣂⮹,⮹, ’. However, in the corpus of Gan and Tham (1999),
there are 136 tokens without a final comma in the sense tags. The
correction was made in this chapter, too.
290 Wong Ping Wai
7.3 Automatic Annotation
Taking advantage of the manual sense tagging by Gan and Tham

(1999), this study expands the corpus by applying an automatic tagger,
followed by human proofreading. It saves much time and efforts in-
volved in manual tagging.
7.3.1 Maximum Entropy Approach
This chapter adopts the Maximum Entropy approach to perform word

sense disambiguation (WSD). This framework aims to integrate in-
formation from various sources for classification, which maximizes
the entropy or ‘uncertainty’ subject to constraints which represent
known information acquired from the training corpus.
The basic idea is: if the entropy were minimized, too much
information constraints would be added without the justification of
empirical evidence from the training corpus. Thus, we maximize the
entropy in order not to go beyond empirical evidence when building
an unbiased model.
This chapter has used the off-the-shelf maximum entropy
(MaxEnt) tagger2 (Ratnaparkhi, 1996) to perform word sense disam-
biguation. The following shows the statistics of the corpus of Gan
and Tham (1999).
(13) number of word tokens: 30,976

number of sentences: 3,178
Average number of words in a sentence: 9.75
Before using the tagger to tag the new corpus, I have conducted an
experiment on Gan and Tham’s (1999) corpus. 2,542 sentences (24,780
words) are allocated as the training corpus, with 636 sentences (6,196
words) as the testing corpus.
The following table shows the preliminary test result:
2 <ftp://ftp.cis.upenn.edu/pub/adwait/jmx/jmx.tar.gz>.
Total Correct Accuracy rate

word accuracy 6,196 4,772 76.99%
sentence accuracy 700 133 19%
Table 8. Accuracy rate of the preliminary test of WSD.
The MaxEnt tagger has a default cut-off value for rare words. For
example, words with frequency lower than five would be regarded
as unknown words. For this small corpus, the data sparseness prob-
lems result in too many unknown words, which have reduced the
accuracy in the preliminary test. As it is not possible to adjust the
default value of this tagger, this chapter follows Wong and Yang
(2002), to concatenate three copies of the training corpus. The size
thus increases to 7,626 sentences (74,340 words). Thus, more rare
words are coped with by the parser, resulting in a better test result:
Total Correct Accuracy rate

word accuracy 6,196 5,589 90.2%
sentence accuracy 700 330 47.14%
Table 9. Accuracy rate of the WSD experiment.
With a good experimental result, the MaxEnt tagger is thus used to

help tag the new texts of the ‘crime’ domain in the Sinica Corpus.
First, the tagger would check against the tagset dictionary, which is a
combination of the HowNet dictionary and the corpus. It is found
that 61.74% of word tokens have one sense only and do not need
human checking. Attention could thus be paid to the polysemous
words and unknown words, which accounts for 39% of word tokens
in the corpus. Much manpower is saved compared with purely manual
annotation applied in Gan and Tham (1999).
Known words Unknown

Total
words with 1 sense polysemous words words
Word tokens 16,385 6,784 3,370 26,539
Percentage 61.74% 25.56% 12.70% 100%
Table 10. Statistics of the corpus newly tagged by this study.

292 Wong Ping Wai
7.3.2 Automatic Dependency Parsing
Tom Lai, Changning Huang and Ming Zhou have conducted research
on parsing Chinese texts using syntactic dependency parsers from
1994 to 2000 (Lai/Huang 1994, 1998, 1999, 2000, Zhou/Huang 1994
and Zhou 2000). The parser applied by Lai and Huang is unification-
based, but it is time consuming to write the lexical rules and gram-
matical rules. They have built some small corpora for testing with-
out public releasing, so such valuable resources are not available for
research in the current study.
This chapter applies the deterministic data-driven parser called
Malt Parser to dependency parsing. The parser (version 0.23) applies
the Memory-Based Learning (MBL) algorithm, which simply stores
experiences in memory, and solves problems by reusing solutions
from similar previously solved problems (Nivre/Scholz 2004).
The parser was designed to analyze the syntactic dependency
between POS tags. This study applies the parser to semantic depen-
dency which requires sense tags. The list of POS is much shorter and
exhaustive, but the list of sense tags is very long. For example, there
are 46 POS tags but 2,447 sense tags in the corpus of Gan and Tham
(1999). The semantic tags are too fine-grained. They need to be sim-
plified (become coarse-grained) so that dependency parsing could
be carried out.
The simplification process is based on the sememes involved
in the Message Structures base. For example, all sense tags of At-
tribute concepts are reduced to categorical attribute (the first sememe)
‘Attribute| Ⱄ⿏ ’. Most concepts of attribute value are simplified to
the categorical attribute ‘aValue| Ⱄ⿏ ’, except some special cases
that requires finer-grained senses. The following table shows some
of these examples.
Sense tags Examples Dependency relations
aValue|Ⱄ⿏ῤ,range|ⷭ⹎ 悥, ℐ Range
aValue|Ⱄ⿏ῤ,degree|䦳⹎ 朆ⷠ, ⼰ Degree
aValue|Ⱄ⿏ῤ,type|栆✳,special|䈡恋, 忁 Restrictive
Table 11. Some special cases of the Attribute Value Concepts.

The experiment is evaluated by two kinds of accuracy, namely:

(1) Head word (HW) accuracy rate, which is defined by:
no. of correct head words
HW accuracy =
total no. of words
(2) Semantic dependency (SD) accuracy rate, which is defined by:

no. of correct head words and semantic dependency relations
SD accuracy =
total no. of words
7.3.2.1 Coarse-grained (2447) VS Fine-grained (766) sense tags

The corpus is divided into two parts. 2,821 sentences (27,331 words)
are the training set, whereas 357 sentences (3645 words) are reserved
for the experimental testing set. Experiment results indicate that
coarse-grained sense tags get a better performance (Table 12).
Accuracy Fine-grained (m4 model) Coarse-grained (m4 model)

Head word 76.52% 79.09%
Semantic Dependency 68.86% 72.54%
Table 12: Coarse-grained sense tags VS Fine-grained sense tags
7.3.2.2 Optimal feature model

Three types of feature model are tested in the experiment. For de-
tails, interested readers are referred to Nivre, Hall and Nilsson (2005).
The following shows a summary of the three models. Experiments
show that m4 is the best one (Table 13).
• m2.mbl: 3 POS features, 4 DEP features, 0 LEX features
Accuracy m2 m3 m4
Head word 74.46% 76.98% 79.09%
Semantic Dependency 67.11% 71.28% 72.54%
Table 13. Experiment result of various feature models.

294 Wong Ping Wai
7.3.2.3 Comparison with similar work

There is a related study that works on semantic dependency parsing
based on HowNet – the Semantic Dependency Net (SDN) by Li et
al. (2005). Table 14 shows the comparison of the parsing results with
hand-corrected word senses. Table 15 shows the comparison of pars-
ing results after processing by automatic sense taggers. Li et al. (2005)
has a large corpus of one million words, which better copes with data
sparseness than this chapter (30,976 words only). Nevertheless, this
chapter performs better. It may be because the corpus applied in this
chapter is a subset of a corpus. The lexicon is simpler in the specific
domain compared with the diversity of domains in Li et al. (2005).
Another possibility is that the corpus of Li et al. was tagged by a
number of human annotators, which may result in the inconsistency
between annotators. The corpus applied in this study is tagged by
one person only. It can avoid the problem of inter-annotator incon-
sistency, thus increasing the accuracy.
Accuracy Li et al. Malt parser (coarse-grained, m4)

Head word 76.87% 79.09%
Table 14. Dependency parsing results with hand-corrected word senses.
Accuracy Li et al. Malt parser (coarse-grained, m4)

Sense tagging 90.85% 90.29%
Head word 75.84% 76.41%
Table 15. Dependency parsing results after automatic sense tagging.

8. Conclusion
Natural language understanding needs the support of a large corpus

annotated with semantic information. It is a crucial resource for de-
signing and testing a computer system that can automatically disam-
biguate senses and find out the relations among the words in a sen-
tence. These enable computers to understand the message conveyed
in the sentence and generate natural language in human-computer
communication. The current study reported the work of semantic
annotation of Chinese texts, which provides such an essential re-
source for natural language understanding.
8.1 Significance of the Current Study
There are two broad types of semantic annotation: (i) Semantic fea-
tures of words – annotation of word senses. (ii) Semantic relations of
items in a sentence – annotation of event roles of participants in-
volved in events. Previous works usually either annotated word senses
or semantic relations in a Chinese corpus. Li et al. (2003) annotated
both, but applying two different knowledge bases that hardly seemed
to be integrated. This chapter adopts HowNet’s approach to incorpo-
rate both types of semantic information when annotating a Chinese
corpus.
This chapter, based on the semantic features revealed by the
sense tags, has annotated Chinese texts with message structures. A
message structure base mainly contains examples at word and phrase
levels. The specialty of Chinese lies in the consistent patterns at word,
phrase and sentence levels. In this study, message structures are used
for annotating semantic dependency relations between the words in
a sentence, using the information of semantic features revealed by
sememes in sense definitions. It is a good proof of the robustness of
HowNet.
This study has applied automatic annotation to help annotate
Chinese texts with semantic information based on HowNet’s frame-
work. It can save much time and effort required in manual annota-
296 Wong Ping Wai
tion as employed in previous studies, e.g. Gan and Tham (1999) and
Gan and Wong (2000). As far as the author is aware, this chapter is
the first study that uses a data-driven parser to annotate semantic
dependency in the corpus manually tagged by Gan and Wong (2000).
Sense tags are made more coarse-grained, so that they can be adapted
to automatic semantic dependency parsing.
8.2 Limitations and Future Work
One of the limitations is the sense inventory. This chapter uses the
old version HowNet 2000, which is a free online resource, instead of
the latest version that requires a high licensing fee – HowNet 2005.
This is because the only available corpus tagged with HowNet’s word
senses is Gan and Tham (1999), which applies the HowNet 1999 and
is also publicly available for free. Having updated the sense defini-
tion to HowNet 2000 in the corpus, this chapter continues their work,
aiming to enrich this indispensable resource with more semantic an-
notation. In future work, I will work on an efficient way to update
the word senses of HowNet 2000 to the latest version (HowNet 2005)
in the annotated corpus.
The corpus in the current study breaks a sentence with a comma,
following the convention of the Sinica Corpus. This results in in-
complete sentences. The unlinked elements are tagged as ‘unresolved’
in the first parse, which is to be resolved in the second parse in future
work, which deals with event relations and event role shifting in a
sentence group.
The automatic parser applied in this study has encountered
problems of data sparseness due to small corpus size. The knowl-
edge bases of HowNet (such as dictionary and Message Structures
base) were not fully utilized yet. It is a direction of further research
to incorporate the advantages of both the knowledge base and the
annotated corpus in the parsing task.
References
Baker, C.F. / Fillmore, C.J. / Lowe, J.B. 1998. The Berkeley FrameNet
Project. Paper presented at the COLING/ACL, Montreal,
Canada.
C.K.I.P. 1995. ㈨埻⟙⏲ 95-02:ᷕ⣖䞼䨞昊⸛堉婆㕁⹓䘬ℏ⭡冯婒㖶
[The Content and Illustration of Sinica Corpus of Academia
Sinica, Technical Report no. 95-02]: Institute of Information
Science, Academia Sinica.
Chen, K.-J. / Huang, C.-R. / Chen, F.-Y. / Luo, C.-C. / Chang, M.-C. /
Chen, C.-J. 2004. Sinica Treebank: Design Criteria, Represen-
tational Issues and Implementation. In A. Abeille (Ed.), Build-
ing and Using Parsed Corpora: Kluwer.
Dang, Hoa Trang / Chia, Ching-Yi / Palmer, Martha / Chiou, Fu-
Dong 2002. Simple Features for Chinese Word Sense Disam-
biguation. The 19 International Conference on Computational
Linguistics (COLING 2002), Taipei, Taiwan, August 24 - Sep-
tember 1, 2002.
Dong, Z. 2000. ᷕ㔯ᾉ〗䳸㥳㧉⺷ [The Pattern of Chinese Mes-
sage Structure]. Retrieved December 2, 2004, from <http://
www.keenage.com>.
Dong, Z. / Dong, Q. 1999. HowNet. from <http://www.keenage.com>.
Gan, K.W. / Tham, W.M. (1999). ➢㕤䞍䵚䘬ⷠ嬀䞍嬀㧁姣 [Gen-
eral Knowledge Annotation Based on HowNet]. Computational
Linguistics and Chinese Language Processing. 4/2, 39-86.
Gan, K.W. / Wong, P.W. 2000. Annotating information structures in
Chinese text using HowNet. Proceedings of the 2nd Chinese
language processing workshop, Association for Computational
Linguistics, 85-92.
Guo, J. 1993. PH – A Chinese Corpus. Communications of COLIPS.
3, 45-48.
Huang, C.-R. / Chen, F.-Y. / Chen, K.-J. / Gao, Z.-M. / Chen, K.-Y.
2000. Sinica Treebank: Design Criteria, Annotation Guidelines,
and On-line Interface. Paper presented at the 2nd Chinese Lan-
guage Processing Workshop, ACL, Hong Kong, October 7,
2000.
298 Wong Ping Wai
IR-Lab of the Harbin Institute of Technology 2005. Tongyici Cilin

with Detailed Examples (Extended Version). Retrieved March
10, 2006, from <http://ir.hit.edu.cn/>.
Ker, S.-J. / Chen, J.-N. 2004. Adaptive Word Sense Tagging on Chi-
nese Corpus. PACLIC 18, December 8-10, 2004, Waseda Uni-
versity, Tokyo.
Lai, T.B.Y. / Huang, C. 1994. Dependency Grammar and the Parsing
of Chinese Sentences. Paper presented at the 1994 Kyoto Con-
ference of Joint ACLIC8 and FACFoCoL2, Aug 10-11, 1994.
Lai, T.B.Y. / Huang, C. 1998. Complements and Adjuncts in Depen-
dency Grammar Parsing Emulated by a Constrained Context-
Free Grammar. Paper presented at the COLING-ACL’98 Work-
shop on Processing of Dependency-Based Grammars, Université
de Montreal.
Lai, T.B.Y. / Huang, C. 1999. Functional Constraints in Dependency
Grammar. Paper presented at the GLDV ’99: Multilinguale
Corpora: Codierung, Strukturierung, Analyse, 11. Jahrestagung
der Gesellschaft für Linguistische DatenVerarbeitung, Frank-
furt a.M., July 8-10, 1999.
Lai, T.B.Y. / Huang, C. 2000. Dependency-based Syntactic Analysis
of Chinese and Annotation of Parsed Corpus. Paper presented
at the 38th Annual Meeting of ACL, Hong Kong.
Li, M. / Juanzi, L. / Dong, Z. / Wang, Z. / Lu, D. 2003. Building a
Large Chinese Corpus Annotated with Semantic Dependency.
Paper presented at the Second SIGHAN Workshop on Chinese
Language Processing, ACL 2003, Sapporo, Japan.
Li, M. / Li, J. / Wang, Z. / Lu, D. 2005. Semantic Analysis and Struc-
tured Language Models. Journal of Software. 16/9, 1523-1533.
McEnery, Anthony / Wilson, Andrew 2001. Corpus Linguistics: An
Introduction. Edinburgh: Edinburgh University Press.
Mei, J. / Yiming Lan, Yunqi Gao / Yongxiang Ying 1983. ⎴佑娆娆㜿
[A Dictionary of Synonyms]: Shanghai Cishu Chubanshe.
Mel’cuk, I.A. (1988). Dependency Syntax: Theory and Practice: State
University of New York Press.
Nivre, J. / Scholz, M. 2004. Deterministic Dependency Parsing of
English Text. Proceedings of COLING 2004, Geneva, Switzer-
land, August 23-27, 2004.
Nivre, J. / Hall, J. / Nilsson, J. 2005. MaltParser: A Data-Driven Parser-

Generator for Dependency Parsing. Proceedings of the Fourth
Workshop on Treebanks and Linguistic Theories, Barcelona,
9-10 December 2005.
Ratnaparkhi, A. 1996. A Maximum Entropy Model for Part-of-speech
Tagging. Paper presented at the 1st Empirical Methods in Natu-
ral Language Processing Conference, Philadelphia, USA.
Tesniere, L. 1959. Elements de syntaxe structurale. Paris: Klinck-
sieck
Wang, C.-Y. 2002. Sense Pruning by HowNet – a Knowledge-Based
Word Sense Disambiguation. Unpublished MPhil thesis, Hong
Kong University of Science and Technology, Hong Kong.
Wong, P.-W. / Yang, Y. 2002. A Maximum Entropy Approach to
HowNet-based Chinese Word Sense Disambiguation. Paper pre-
sented at the SemaNet workshop, COLING 2002, 24th Au-
gust - 1st September, 2002, Taipei, Taiwan.
Xia, Fei 2000. The Part-Of-Speech Tagging Guidelines for the Penn
Chinese Treebank (3.0), from <http://www.upenn.edu/ctb/>.
Xue, N. / Palmer, M. 2005. Automatic Semantic Role Labeling for
Chinese Verbs. Paper presented at the 19th International Joint
Conference on Artificial Intelligence, Edinburgh, Scotland.
Yu, S. / Duan, H. / Zhu, X. / Swen, B. / Chang, B. 2003. ⊿⣏婆㕁⹓≈
ⶍ夷䭬 : ↯↮ , 娆⿏冯㲐枛 [Specification for Corpus Process-
ing at Peking University: Word Segmentation, POS Tagging
and Phonetic Notation]. Journal of Chinese Language and
Computing. 13/2, 121-158.
Zhou, M. 2000. A Block-based Robust Dependency Parser for Un-
restricted Chinese Text. Paper presented at the Second Chi-
nese Language Processing Workshop, ACL2000, Hong Kong,
October 8, 2000.
Zhou, M. / Huang, C. 1994. An Efficient Syntactic Tagging Tool for
Corpora. Paper presented at the COLING 1994, Kyoto, Japan.
Notes on Contributors
TERUHIRO ISHIGURO, PhD, received a doctoral degree in English Lin-

guistics from Doshisha University, Japan. He is currently a professor
emeritus at Doshisha University. His research interests include the
contrastive study of English and Japanese, which he has explored in
a theoretical perspective, integrating it whenever necessary with dis-
course analysis. In addition, translation studies and English as a for-
eign language are among his main research areas. Dr Ishiguro has
published extensively on English education in Japan.
K. K. LUKE is Professor of Linguistics at Nanyang Technological

University, Singapore. Prior to joining Nanyang he was Professor of
Linguistics at the University of Hong Kong, and Head of the Depart-
ment of Linguistics from 1997 to 2006. Professor Luke’s research is
in the areas of Chinese Linguistics and Conversation Analysis. He
has worked on Cantonese phonology and grammar and the interface
between language, cognition, and interaction using Chinese and Eng-
lish data. Among his publications are Utterance Particles in Cantonese
Conversation, Language and Society in Hong Kong, and Telephone
Calls: Unity and Diversity in the Structure of Telephone Conversa-
tions across Languages and Cultures. (Email: kkluke@ntu.edu.sg)
WINNIE CHOR received her PhD in Linguistics from the University

of Sydney. She is now Lecturer in English Linguistics and Applied
Linguistics at the Open University of Hong Kong. Winnie’s research
areas include Conversation Analysis, Semantic Change, Cantonese
Linguistics, and the Theory of Grammaticalization. Her research inter-
ests lie primarily in the cognitive-functional aspects of language, with
a special focus on how (inter)subjectivity, mirativity, and evidentiality
are expressed in English and Cantonese, as well as in other Chinese
dialects and Asian languages. Her current research focuses on how
these various linguistic markers evolved from a diachronic perspec-
tive. (Email: wowchor@ouhk.edu.hk)
302 Notes on Contributors
HUA GAO is Associate Professor of linguistics in the English Depart-

ment, Shenzhen University, where she teaches English and linguistics
courses for English majors. Her research interests span from Cognitive-
Functional Linguistics to Interactional Linguistics with an empirical,
usage-based approach to grammatical structure and grammaticalization.
Her recent publications include “A Cognitive-functional investiga-
tion of the neutral question in Chinese” and “Topic-only questions in
Chinese”. Currently she is working on conversion and clipping as
English word formation processes from the cognitive perspective,
and the grammaticalization of the passive construction in English
from expressing passivity to expressing deontic or epistemic modality.
(Email: sdgaohua@gmail.com)
HAN YANG has a Ph.D. in Linguistics from the University of York,

UK. For over twenty years she has taught at universities in the UK,
Singapore, Hong Kong and Japan. She was an Associate Professor in
the School of Arts and Social Sciences at the Open University of
Hong Kong until 2010, when she left academia to become a free-
lance writer and translator. Dr. Han’s specialization is in syntax and
semantics, but she has also done work in sociolinguistics and second
language learning. Her publications include Language and Society
in Hong Kong and several papers on tense and aspect in Chinese and
Japanese.
YOICHIRO HASEBE is an Associate Professor at the Faculty of Global

Communications at Doshisha University, where he teaches courses
on linguistics as well as various communication-related issues in
today’s globalizing world. His main area of research is cognitive lin-
guistics, with a wide variety of interests ranging from subjectivity/
objectivity problems in language to the mechanisms underlying net-
works of linguistic constructions. With expertise on computer pro-
gramming and natural text processing, he has experience working
with a team building computational systems for measuring the read-
ability of Japanese text for learners and teachers. Currently, he is
starting a project in the new field of “applied cognitive linguistics.”
It aims to develop a method of enabling ESL learners to turn their
existing lexical knowledge into a more robust and flexible network
Notes on Contributors 303
with which they can more easily and effectively draw words and
expressions in spontaneous speech and writing in English. (E-mail:
yhasebe@mail.doshisha.ac.jp)
MASANOBU HORIGUCHI is Professor of English Linguistics and Pre-

sentation-Skill Training at Tokushima Bunri University Junior Col-
lege, Faculty of Language and Communication. His principal research
areas span from phonology to education. He visited Finland in 2008
for a research under the title “Establishing the Method to Develop
Presentation Skills and the Criteria in Evaluating Presentations by
Referring to the Finnish Method of Education,” supported by Grant-
in-Aid for Scientific Research (KAKENHI, No. 20530870) by Japan
Society for the Promotion of Science. Some of his recent publica-
tions include “A study of the English pronunciation of John Manjiro
[1827-1898] and “The essence of the Finnish Method of education”.
(Email: watthau@tokushima.bunri-u.ac.jp)
NORIFUMI ITO is Professor of English Linguistics at the Tokushima

Bunri University, Faculty of Letters. His principal research areas are
pragmatics, information structure, and philosophy of language (Dis-
course Information and Interpretation of English Constructions,
2005). He is currently working on the interaction of semantics and
pragmatics, especially in the analysis of context, with the help of
philosophy of langauge and sociology. He has several publications
on English language education. (Email: nito@kagawa.bunri-u.ac.jp)
WAI-MUN LEUNG is an Assistant Professor at Hong Kong Polytechnic

University. She has a PhD in Linguistics from the University of Hong
Kong. Her main areas of research are in the fields of syntax, text
analysis and sociolinguistics. She has published a book on Cantonese
sentence-final particles, and a number of papers on Chinese Linguistics.
(Email: wai-mun.leung@polyu.edu.hk)
HIDEMATSU MIURA, Ph.D. (University at Buffalo, The State Univer-

sity of New York), MA (Doshisha University, Kyoto), is Associate
Professor of Linguistics at Mukogawa Women’s University, Depart-
ment of English, where he teaches both general and English lin-
304 Notes on Contributors
guistics. His research interests span from theoretical comparative

linguistics, the interface between syntax and semantics in particular,
to applied linguistics such as second language acquisition. His re-
cent work includes “A Comparative Study of Reflexives in English,
Korean and Japanese in terms of Antimetonymy” (2010), “Toward a
Better English (e-)Learning Model” (2009) and Grammatical Rela-
tions, Reflexives and Pseudo-raising in Japanese (2008). (Email:
hmiura@mukogawa-u.ac.jp)
YUTAKA SHINODA is Associate Professor of English Linguistics at

Tokushima Bunri University, Japan. His main research areas are the
analysis of jokes and the investigation of exclamations and sentence-
final particles. His recent publications include “The interpersonal
function of the Japanese particle ne and backchannels” (2005), “The
structure of a weird tale – In comparison with that of a joke – (I): The
function of the punch line” (2010) and “The structure of a weird tale
– In comparison with that of a joke – (II): The reality of weird tales”
(2011). (E-mail: shinoday@kagawa.bunri-u.ac.jp)
PING WAI WONG is Assistant Professor at Tung Wah College, Lan-

guages and General Education Centre. His principal research areas
are the contrastive studies of Chinese and English, theoretical linguis-
tics and corpus linguistics, with a particular interest in syntactic struc-
tures of Cantonese. He has investigated how HowNet, which is an
online common-sense knowledge base of Chinese and English, can
be applied to the semantic annotation of Chinese corpus, so as to help
computers analyze naturally occurring texts. At present his work con-
centrates on using corpora for studying such syntactic constructions as
passive constructions and dative constructions in Chinese and Eng-
lish. (Email: percywong@twc.edu.hk; wongpw21@yahoo.com.hk)
Linguistic Insights
This series aims to promote specialist language studies in the fields of linguistic
theory and applied linguistics, by publishing volumes that focus on specific
aspects of language use in one or several languages and provide valuable
insights into language and communication research. A cross-disciplinary ap-
proach is favoured and most European languages are accepted.
The series includes two types of books:
– Monographs – featuring in-depth studies on special aspects of language

theory, language analysis or language teaching.
– Collected papers – assembling papers from workshops, conferences or
symposia.
Each volume of the series is subjected to a double peer-reviewing process.
Vol. 1 Maurizio Gotti & Marina Dossena (eds)

Modality in Specialized Texts. Selected Papers of the 1st CERLIS Conference.
421 pages. 2001. ISBN 3-906767-10-8 · US-ISBN 0-8204-5340-4
Vol. 2 Giuseppina Cortese & Philip Riley (eds)

Domain-specific English. Textual Practices across Communities
and Classrooms.
Vol. 3 Maurizio Gotti, Dorothee Heller & Marina Dossena (eds)

Conflict and Negotiation in Specialized Texts. Selected Papers
of the 2nd CERLIS Conference.
Vol. 4 Maurizio Gotti, Marina Dossena, Richard Dury, Roberta Facchinetti & Maria Lima
Variation in Central Modals. A Repertoire of Forms and Types of Usage
in Middle English and Early Modern English.
Editorial address:
Prof. Maurizio Gotti Università di Bergamo, Facoltà di Lingue e Letterature Straniere,
Via Salvecchio 19, 24129 Bergamo, Italy
Fax: 0039 035 2052789, E-Mail: m.gotti@unibg.it
Vol. 5 Stefania Nuccorini (ed.)
Phrases and Phraseology. Data and Descriptions.
187 pages. 2002. ISBN 3-906770-08-7 · US-ISBN 0-8204-5933-X
Vol. 6 Vijay Bhatia, Christopher N. Candlin & Maurizio Gotti (eds)

Legal Discourse in Multilingual and Multicultural Contexts.
Arbitration Texts in Europe.
Vol. 7 Marina Dossena & Charles Jones (eds)

Insights into Late Modern English. 2nd edition.
378 pages. 2003, 2007.
ISBN 978-3-03911-257-9 · US-ISBN 978-0-8204-8927-8
Vol. 8 Maurizio Gotti

Specialized Discourse. Linguistic Features and Changing Conventions.
351 pages. 2003, 2005.
ISBN 3-03910-606-6 · US-ISBN 0-8204-7000-7
Vol. 9 Alan Partington, John Morley & Louann Haarman (eds)

Corpora and Discourse.
Vol. 10 Martina Möllering

The Acquisition of German Modal Particles. A Corpus-Based Approach.
Vol. 11 David Hart (ed.)

English Modality in Context. Diachronic Perspectives.
Vol. 12 Wendy Swanson

Modes of Co-reference as an Indicator of Genre.
Vol. 13 Gina Poncini

Discursive Strategies in Multicultural Business Meetings.
2nd edition. 338 pages. 2004, 2007.
ISBN 978-3-03911-296-8 · US-ISBN 978-0-8204-8937-7
Vol. 14 Christopher N. Candlin & Maurizio Gotti (eds)

Intercultural Aspects of Specialized Communication.
2nd edition. 369 pages. 2004, 2007.
ISBN 978-3-03911-258-6 · US-ISBN 978-0-8204-8926-1
Vol. 15 Gabriella Del Lungo Camiciotti & Elena Tognini Bonelli (eds)
Academic Discourse. New Insights into Evaluation.
Vol. 16 Marina Dossena & Roger Lass (eds)

Methods and Data in English Historical Dialectology.
Vol. 17 Judy Noguchi

The Science Review Article. An Opportune Genre in
the Construction of Science.
Vol. 18 Giuseppina Cortese & Anna Duszak (eds)
Identity, Community, Discourse. English in Intercultural Settings.
Vol. 19 Anna Trosborg & Poul Erik Flyvholm Jørgensen (eds)

Business Discourse. Texts and Contexts.
Vol. 20 Christopher Williams

Tradition and Change in Legal English. Verbal Constructions
in Prescriptive Texts.
2nd revised edition. 216 pages. 2005, 2007. ISBN 978-3-03911-444-3.
Vol. 21 Katarzyna Dziubalska-Kolaczyk & Joanna Przedlacka (eds)

English Pronunciation Models: A Changing Scene.
2nd edition. 476 pages. 2005, 2008. ISBN 978-3-03911-682-9.
Vol. 22 Christián Abello-Contesse, Rubén Chacón-Beltrán,

M. Dolores López-Jiménez & M. Mar Torreblanca-López (eds)
Age in L2 Acquisition and Teaching.
Vol. 23 Vijay K. Bhatia, Maurizio Gotti, Jan Engberg & Dorothee Heller (eds)
Vagueness in Normative Texts.
Vol. 24 Paul Gillaerts & Maurizio Gotti (eds)

Genre Variation in Business Letters. 2nd printing.
407 pages. 2008. ISBN 978-3-03911-681-2.
Vol. 25 Ana María Hornero, María José Luzón & Silvia Murillo (eds)
Corpus Linguistics. Applications for the Study of English.
2nd printing. 526 pages. 2006, 2008. ISBN 978-3-03911-726-0
Vol. 26 J. Lachlan Mackenzie & María de los Ángeles Gómez-González (eds)

Studies in Functional Discourse Grammar.
Vol. 27 Debbie G. E. Ho
Classroom Talk. Exploring the Sociocultural Structure of Formal ESL Learning.
2nd edition. 254 pages. 2006, 2007. ISBN 978-3-03911-434-4
Vol. 28 Javier Pérez-Guerra, Dolores González-Álvarez, Jorge L. Bueno-Alonso

& Esperanza Rama-Martínez (eds)
‘Of Varying Language and Opposing Creed’. New Insights into Late Modern English.
455 pages. 2007. ISBN 978-3-03910-788-9
Vol. 29 Francesca Bargiela-Chiappini & Maurizio Gotti (eds)

Asian Business Discourse(s).
Vol. 30 Nicholas Brownlees (ed.)

News Discourse in Early Modern Britain. Selected Papers of CHINED 2004.
Vol. 31 Roberta Facchinetti & Matti Rissanen (eds)

Corpus-based Studies of Diachronic English.
Vol. 32 Marina Dossena & Susan M. Fitzmaurice (eds)
Business and Official Correspondence. Historical Investigations.
Vol. 33 Giuliana Garzone & Srikant Sarangi (eds)

Discourse, Ideology and Specialized Communication.
494 pages. 2007. ISBN 978-3-03910-888-6
Vol. 34 Giuliana Garzone & Cornelia Ilie (eds)

The Use of English in Institutional and Business Settings.
An Intercultural Perspective.
372 pages. 2007. ISBN 978-3-03910-889-3
Vol. 35 Vijay K. Bhatia & Maurizio Gotti (eds)

Explorations in Specialized Genres.
Vol. 36 Heribert Picht (ed.)

Modern Approaches to Terminological Theories and Applications.
Vol. 37 Anne Wagner & Sophie Cacciaguidi-Fahy (eds)

Legal Language and the Search for Clarity / Le langage juridique et la quête de clarté.
Practice and Tools / Pratiques et instruments.
Vol. 38 Juan Carlos Palmer-Silveira, Miguel F. Ruiz-Garrido &

Inmaculada Fortanet-Gómez (eds)
Intercultural and International Business Communication.
Theory, Research and Teaching.
Vol. 39 Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt &

Herbert Schendl (eds)
Syntax, Style and Grammatical Norms. English from 1500–2000.
Vol. 40 Marina Dossena & Irma Taavitsainen (eds)

Diachronic Perspectives on Domain-Specific English.
Vol. 41 John Flowerdew & Maurizio Gotti (eds)

Studies in Specialized Discourse.
293 pages. 2006. ISBN 3-03911-178-7
Vol. 42 Ken Hyland & Marina Bondi (eds)

Academic Discourse Across Disciplines.
Vol. 43 Paul Gillaerts & Philip Shaw (eds)

The Map and the Landscape. Norms and Practices in Genre.
Vol. 44 Maurizio Gotti & Davide Giannoni (eds)

New Trends in Specialized Discourse Analysis.
Vol. 45 Maurizio Gotti & Françoise Salager-Meyer (eds)

Advances in Medical Discourse Analysis. Oral and Written Contexts.
492 pages. 2006. ISBN 3-03911-185-X · US-ISBN 0-8204-8382-6
Vol. 46 Maurizio Gotti & Susan Šarcević (eds)
Insights into Specialized Translation.
Vol. 47 Khurshid Ahmad & Margaret Rogers (eds)

Evidence-based LSP. Translation, Text and Terminology.
584 pages. 2007. ISBN 978-3-03911-187-9
Vol. 48 Hao Sun & Dániel Z. Kádár (eds)

It’s the Dragon’s Turn. Chinese Institutional Discourses.
262 pages. 2008. ISBN 978-3-03911-175-6
Vol. 49 Cristina Suárez-Gómez

Relativization in Early English (950-1250). the Position of Relative Clauses.
Vol. 50 Maria Vittoria Calvi & Luisa Chierichetti (eds)

Nuevas tendencias en el discurso de especialidad.
319 pages. 2006. ISBN 978-3-03911-261-6
Vol. 51 Mari Carmen Campoy & María José Luzón (eds)

Spoken Corpora in Applied Linguistics.
274 pages. 2008. ISBN 978-3-03911-275-3
Vol. 52 Konrad Ehlich & Dorothee Heller (Hrsg.)

Die Wissenschaft und ihre Sprachen.
323 pages. 2006. ISBN 978-3-03911-272-2
Vol. 53 Jingyu Zhang

The Semantic Salience Hierarchy Model. The L2 Acquisition of Psych Predicates
273 pages. 2007. ISBN 978-3-03911-300-2
Vol. 54 Norman Fairclough, Giuseppina Cortese & Patrizia Ardizzone (eds)
Discourse and Contemporary Social Change.
555 pages. 2007. ISBN 978-3-03911-276-0
Vol. 55 Jan Engberg, Marianne Grove Ditlevsen, Peter Kastberg & Martin Stegu (eds)
New Directions in LSP Teaching.
331 pages. 2007. ISBN 978-3-03911-433-7
Vol. 56 Dorothee Heller & Konrad Ehlich (Hrsg.)

Studien zur Rechtskommunikation.
322 pages. 2007. ISBN 978-3-03911-436-8
Vol. 57 Teruhiro Ishiguro & Kang-kwong Luke (eds)

Grammar in Cross-Linguistic Perspective.
The Syntax, Semantics, and Pragmatics of Japanese and Chinese.
304 pages. 2012. ISBN 978-3-03911-445-0
Vol. 58 Carmen Frehner

Email – SMS – MMS
294 pages. 2008. ISBN 978-3-03911-451-1
Vol. 59 Isabel Balteiro

The Directionality of Conversion in English. A Dia-Synchronic Study.
276 pages. 2007. ISBN 978-3-03911-241-8
Vol. 60 Maria Milagros Del Saz Rubio

English Discourse Markers of Reformulation.
237 pages. 2007. ISBN 978-3-03911-196-1
Vol. 61 Sally Burgess & Pedro Martín-Martín (eds)
English as an Additional Language in Research Publication and Communication.
259 pages. 2008. ISBN 978-3-03911-462-7
Vol. 62 Sandrine Onillon

Pratiques et représentations de l’écrit.
458 pages. 2008. ISBN 978-3-03911-464-1
Vol. 63 Hugo Bowles & Paul Seedhouse (eds)

Conversation Analysis and Language for Specific Purposes.
Vol. 64 Vijay K. Bhatia, Christopher N. Candlin & Paola Evangelisti Allori (eds)
Language, Culture and the Law.
The Formulation of Legal Concepts across Systems and Cultures.
342 pages. 2008. ISBN 978-3-03911-470-2
Vol. 65 Jonathan Culpeper & Dániel Z. Kádár (eds)

Historical (Im)politeness.
300 pages. 2010. ISBN 978-3-03911-496-2
Vol. 66 Linda Lombardo (ed.)

Using Corpora to Learn about Language and Discourse.
237 pages. 2009. ISBN 978-3-03911-522-8
Vol. 67 Natsumi Wakamoto

Extroversion/Introversion in Foreign Language Learning.
Interactions with Learner Strategy Use.
159 pages. 2009. ISBN 978-3-03911-596-9
Vol. 68 Eva Alcón-Soler (ed.)

Learning How to Request in an Instructed Language Learning Context.
260 pages. 2008. ISBN 978-3-03911-601-0
Vol. 69 Domenico Pezzini

The Translation of Religious Texts in the Middle Ages.
428 pages. 2008. ISBN 978-3-03911-600-3
Vol. 70 Tomoko Tode

Effects of Frequency in Classroom Second Language Learning.
Quasi-experiment and stimulated-recall analysis.
195 pages. 2008. ISBN 978-3-03911-602-7
Vol. 71 Egor Tsedryk

Fusion symétrique et alternances ditransitives.
211 pages. 2009. ISBN 978-3-03911-609-6
Vol. 72 Cynthia J. Kellett Bidoli & Elana Ochse (eds)

English in International Deaf Communication.
444 pages. 2008. ISBN 978-3-03911-610-2
Vol. 73 Joan C. Beal, Carmela Nocera & Massimo Sturiale (eds)

Perspectives on Prescriptivism.
269 pages. 2008. ISBN 978-3-03911-632-4
Vol. 74 Carol Taylor Torsello, Katherine Ackerley & Erik Castello (eds)
Corpora for University Language Teachers.
308 pages. 2008. ISBN 978-3-03911-639-3
Vol. 75 María Luisa Pérez Cañado (ed.)
English Language Teaching in the European Credit Transfer System.
Facing the Challenge.
251 pages. 2009. ISBN 978-3-03911-654-6
Vol. 76 Marina Dossena & Ingrid Tieken-Boon van Ostade (eds)

Studies in Late Modern English Correspondence. Methodology and Data.
291 pages. 2008. ISBN 978-3-03911-658-4
Vol. 77 Ingrid Tieken-Boon van Ostade & Wim van der Wurff (eds)
Current Issues in Late Modern English.
436 pages. 2009. ISBN 978-3-03911-660-7
Vol. 78 Marta Navarro Coy (ed.)

Practical Approaches to Foreign Language Teaching and Learning.
297 pages. 2009. ISBN 978-3-03911-661-4
Vol. 79 Qing Ma
Second Language Vocabulary Acquisition.
333 pages. 2009. ISBN 978-3-03911-666-9
Vol. 80 Martin Solly, Michelangelo Conoscenti & Sandra Campagna (eds)

Verbal/Visual Narrative Texts in Higher Education.
384 pages. 2008. ISBN 978-3-03911-672-0
Vol. 81 Meiko Matsumoto

From Simple Verbs to Periphrastic Expressions:
The Historical Development of Composite Predicates, Phrasal Verbs,
and Related Constructions in English.
235 pages. 2008. ISBN 978-3-03911-675-1
Vol. 82 Melinda Dooly

Doing Diversity. Teachers’ Construction of Their Classroom Reality.
180 pages. 2009. ISBN 978-3-03911-687-4
Vol. 83 Victoria Guillén-Nieto, Carmen Marimón-Llorca & Chelo Vargas-Sierra (eds)

Intercultural Business Communication and
Simulation and Gaming Methodology.
392 pages. 2009. ISBN 978-3-03911-688-1
Vol. 84 Maria Grazia Guido

English as a Lingua Franca in Cross-cultural Immigration Domains.
285 pages. 2008. ISBN 978-3-03911-689-8
Vol. 85 Erik Castello

Text Complexity and Reading Comprehension Tests.
352 pages. 2008. ISBN 978-3-03911-717-8
Vol. 86 Maria-Lluisa Gea-Valor, Isabel García-Izquierdo & Maria-José Esteve (eds)

Linguistic and Translation Studies in Scientific Communication.
317 pages. 2010. ISBN 978-3-0343-0069-8
Vol. 87 Carmen Navarro, Rosa Mª Rodríguez Abella, Francesca Dalle Pezze

& Renzo Miotti (eds)
La comunicación especializada.
355 pages. 2008. ISBN 978-3-03911-733-8
Vol. 88 Kiriko Sato
The Development from Case-Forms to Prepositional Constructions
in Old English Prose.
231 pages. 2009. ISBN 978-3-03911-763-5
Vol. 89 Dorothee Heller (Hrsg.)

Formulierungsmuster in deutscher und italienischer Fachkommunikation.
Intra- und interlinguale Perspektiven.
315 pages. 2008. ISBN 978-3-03911-778-9
Vol. 90 Henning Bergenholtz, Sandro Nielsen & Sven Tarp (eds)

Lexicography at a Crossroads. Dictionaries and Encyclopedias Today,
Lexicographical Tools Tomorrow.
372 pages. 2009. ISBN 978-3-03911-799-4
Vol. 91 Manouchehr Moshtagh Khorasani

The Development of Controversies. From the Early Modern Period
to Online Discussion Forums.
317 pages. 2009. ISBN 978-3-3911-711-6
Vol. 92 María Luisa Carrió-Pastor (ed.)

Content and Language Integrated Learning. Cultural Diversity.
178 pages. 2009. ISBN 978-3-3911-818-2
Vol. 93 Roger Berry

Terminology in English Language Teaching. Nature and Use.
262 pages. 2010. ISBN 978-3-0343-0013-1
Vol. 94 Roberto Cagliero & Jennifer Jenkins (eds)

Discourses, Communities, and Global Englishes
240 pages. 2010. ISBN 978-3-0343-0012-4
Vol. 95 Facchinetti Roberta, Crystal David, Seidlhofer Barbara (eds)
From International to Local English – And Back Again.
268 pages. 2010. ISBN 978-3-0343-0011-7
Vol. 96 Cesare Gagliardi & Alan Maley (eds)
EIL, ELF, Global English. Teaching and Learning Issues
376 pages. 2010. ISBN 978-3-0343-0010-0
Vol. 97 Sylvie Hancil (ed.)
The Role of Prosody in Affective Speech.
403 pages. 2009. ISBN 978-3-03911-696-6
Vol. 98 Marina Dossena & Roger Lass (eds)
Studies in English and European Historical Dialectology.
257 pages. 2009. ISBN 978-3-0343-0024-7
Vol. 99 Christine Béal
Les interactions quotidiennes en français et en anglais.
De l’approche comparative à l’analyse des situations interculturelles.
424 pages. 2010. ISBN 978-3-0343-0027-8
Vol. 100 Maurizio Gotti (ed.)
Commonality and Individuality in Academic Discourse.
398 pages. 2009. ISBN 978-3-0343-0023-0
Vol. 101 Javier E. Díaz Vera & Rosario Caballero (eds)
Textual Healing. Studies in Medieval English Medical, Scientific and Technical Texts.
213 pages. 2009. ISBN 978-3-03911-822-9
Vol. 102 Nuria Edo Marzá
The Specialised Lexicographical Approach. A Step further in Dictionary-making.
316 pages. 2009. ISBN 978-3-0343-0043-8
Vol. 103 Carlos Prado-Alonso, Lidia Gómez-García, Iria Pastor-Gómez &
David Tizón-Couto (eds)
New Trends and Methodologies in Applied English Language Research.
Diachronic, Diatopic and Contrastive Studies.
348 pages. 2009. ISBN 978-3-0343-0046-9
Vol. 104 Françoise Salager-Meyer & Beverly A. Lewin
Crossed Words. Criticism in Scholarly Writing?
371 pages. 2011. ISBN 978-3-0343-0049-0.
Vol. 105 Javier Ruano-García
Early Modern Northern English Lexis. A Literary Corpus-Based Study.
611 pages. 2010. ISBN 978-3-0343-0058-2
Vol. 106 Rafael Monroy-Casas
Systems for the Phonetic Transcription of English. Theory and Texts.
280 pages. 2011. ISBN 978-3-0343-0059-9
Vol. 107 Nicola T. Owtram
The Pragmatics of Academic Writing.
A Relevance Approach to the Analysis of Research Article Introductions.
311 pages. 2009. ISBN 978-3-0343-0060-5
Vol. 108 Yolanda Ruiz de Zarobe, Juan Manuel Sierra &
Francisco Gallardo del Puerto (eds)
Content and Foreign Language Integrated Learning.
Contributions to Multilingualism in European Contexts
343 pages. 2011. ISBN 978-3-0343-0074-2
Vol. 109 Ángeles Linde López & Rosalía Crespo Jiménez (eds)
Professional English in the European context. The EHEA challenge.
374 pages. 2010. ISBN 978-3-0343-0088-9
Vol. 110 Rosalía Rodríguez-Vázquez
The Rhythm of Speech, Verse and Vocal Music. A New Theory.
394 pages. 2010. ISBN 978-3-0343-0309-5
Vol. 111 Anastasios Tsangalidis & Roberta Facchinetti (eds)
Studies on English Modality. In Honour of Frank Palmer.
392 pages. 2009. ISBN 978-3-0343-0310-1
Vol. 112 Forthcoming.
Vol. 113 Mihhail Lotman & Maria-Kristiina Lotman (eds)
Frontiers in Comparative Prosody. In memoriam: Mikhail Gasparov.
426 pages. 2011. ISBN 978-3-0343-0373-6
Vol. 114 Merja Kytö, John Scahill & Harumi Tanabe (eds)
Language Change and Variation from Old English to Late Modern English.
A Festschrift for Minoji Akimoto
422 pages. 2010. ISBN 978-3-0343-0372-9
Vol. 115 Giuliana Garzone & Paola Catenaccio (eds)
Identities across Media and Modes. Discursive Perspectives.
379 pages. 2009. ISBN 978-3-0343-0386-6
Vol. 116 Elena Landone
Los marcadores del discurso y cortesía verbal en español.
390 pages. 2010. ISBN 978-3-0343-0413-9
Vol. 117 Maurizio Gotti & Christopher Williams (eds)
Legal Discourse across Languages and Cultures.
339 pages. 2010. ISBN 978-3-0343-0425-2
Vol. 118 David Hirsh
Academic Vocabulary in Context.
217 pages. 2010. ISBN 978-3-0343-0426-9
Vol. 119 Yvonne Dröschel
Lingua Franca English. The Role of Simplification and Transfer.
358 pages. 2011. ISBN 978-3-0343-0432-0
Vol. 120 Tengku Sepora Tengku Mahadi, Helia Vaezian & Mahmoud Akbari
Corpora in Translation. A Practical Guide.
135 pages. 2010. ISBN 978-3-0343-0434-4
Vol. 121 Davide Simone Giannoni & Celina Frade (eds)
Researching Language and the Law. Textual Features and Translation Issues.
278 pages. 2010. ISBN 978-3-0343-0443-6
Vol. 122 Daniel Madrid & Stephen Hughes (eds)

Studies in Bilingual Education.
472 pages. 2011. ISBN 978-3-0343-0474-0
Vol. 123 Vijay K. Bhatia, Christopher N. Candlin & Maurizio Gotti (eds)
The Discourses of Dispute Resolution.
290 pages. 2010. ISBN 978-3-0343-0476-4
Vol. 124 Davide Simone Giannoni

Mapping Academic Values in the Disciplines. A Corpus-Based Approach.
288 pages. 2010. ISBN 978-3-0343-0488-7
Vol. 125 Giuliana Garzone & James Archibald (eds)

Discourse, Identities and Roles in Specialized Communication.
419 pages. 2010. ISBN 978-3-0343-0494-8
Vol. 126 Iria Pastor-Gómez

The Status and Development of N+N Sequences in
Contemporary English Noun Phrases.
216 pages. 2011. ISBN 978-3-0343-0534-1
Vol. 127 Carlos Prado-Alonso

Full-verb Inversion in Written and Spoken English.
261 pages. 2011. ISBN 978-3-0343-0535-8
Vol. 128 Tony Harris & María Moreno Jaén (eds)

Corpus Linguistics in Language Teaching.
214 pages. 2010. ISBN 978-3-0343-0524-2
Vol. 129 Tetsuji Oda & Hiroyuki Eto (eds)

Multiple Perspectives on English Philology and History of Linguistics.
A Festschrift for Shoichi Watanabe on his 80th Birthday.
378 pages. 2010. ISBN 978-3-0343-0480-1
Vol. 130 Luisa Chierichetti & Giovanni Garofalo (eds)
Lengua y Derecho. líneas de investigación interdisciplinaria.
283 pages. 2010. 978-3-0343-0463-4
Vol. 131 Paola Evangelisti Allori & Giuliana Garzone (eds)

Discourse, Identities and Genres in Corporate Communication.
Sponsorship, Advertising and Organizational Communication.
324 pages. 2011. 978-3-0343-0591-4
Vol. 132 Leyre Ruiz de Zarobe & Yolanda Ruiz de Zarobe (eds)
Speech Acts and Politeness across Languages and Cultures.
402 pages. 2012. 978-3-0343-0611-9
Vol. 133 Thomas Christiansen

Cohesion. A Discourse Perspective.
387 pages. 2011. 978-3-0343-0619-5
Vol. 134 Giuliana Garzone & Maurizio Gotti

Discourse, Communication and the Enterprise. Genres and Trends.
451 pages. 2011. ISBN 978-3-0343-0620-1
Vol. 135 Zsuzsa Hoffmann

Ways of the World’s Words.
Language Contact in the Age of Globalization.
334 pages 2011. ISBN 978-3-0343-0673-7
Vol. 136 Cecilia Varcasia (ed.)

Becoming Multilingual.
Language Learning and Language Policy between Attitudes and Identities.
213 pages. 2011. ISBN 978-3-0343-0687-5
Vol. 137 Susy Macqueen

The Emergence of Patterns in Second Language Writing.
A Sociocognitive Exploration of Lexical Trails.
325 pages. 2012. ISBN 978-3-0343-1010-9
Vol. 138 Maria Vittoria Calvi & Giovanna Mapelli (eds)

La lengua del turismo. Géneros discursivos y terminología.
365 pages. 2011. ISBN 978-3-0343-1011-6
Vol. 139 Ken Lau

Learning to Become a Professional in a Textually-Mediated World.
A Text-Oriented Study of Placement Practices.
261 pages. 2012. ISBN 978-3-0343-1016-1
Vol. 140 Sandra Campagna, Giuliana Garzone, Cornelia Ilie & Elizabeth Rowley-Jolivet (eds)
Evolving Genres in Web-mediated Communication.
337 pages. 2012. ISBN 978-3-0343-1013-0
Vol. 141 Edith Esch & Martin Solly (eds)

The Sociolinguistics of Language Education in International Contexts.
263 pages. 2012. ISBN 978-3-0343-1009-3

Vol. 143 David Tizón-Couto
Left Dislocation in English. A Functional-Discoursal Approach.
416 pages. 2012. ISBN 978-3-0343-1037-6
Vol. 144 Margrethe Petersen & Jan Engberg (eds)
Current Trends in LSP Research. Aims and Methods.
323 pages. 2011. ISBN 978-3-0343-1054-3
Vol. 145 David Tizón-Couto, Beatriz Tizón-Couto, Iria Pastor-Gómez & Paula Rodríguez-Puente (eds)
New Trends and Methodologies in Applied English Language Research II.
Studies in Language Variation, Meaning and Learning.
283 pages. 2012. ISBN 978-3-0343-1061-1
Vol. 146 Rita Salvi & Hiromasa Tanaka (eds)

Intercultural Interactions in Business and Management.
306 pages. 2011. ISBN 978-3-0343-1039-0
Vol. 147 Francesco Straniero Sergio & Caterina Falbo (eds)

Breaking Ground in Corpus-based Interpreting Studies.
254 pages. 2012. ISBN 978-3-0343-1071-0
Vol. 149 Vijay K. Bhatia & Paola Evangelisti Allori (eds)

Discourse and Identity in the Professions. Legal, Corporate and Institutional Citizenship.
352 pages. 2011. ISBN 978-3-0343-1079-6
Vol. 150 Maurizio Gotti (ed.)

Academic Identity Traits. A Corpus-Based Investigation.
363 pages. 2012. ISBN 978-3-0343-1141-0
Vol. 151 Priscilla Heynderickx, Sylvain Dieltjens, Geert Jacobs, Paul Gillaerts &
Elizabeth de Groot (eds)
The Language Factor in International Business.
New Perspectives on Research, Teaching and Practice.
320 pages. 2012. ISBN 978-3-0343-1090-1
Vol. 152 Paul Gillaerts, Elizabeth de Groot, Sylvain Dieltjens, Priscilla Heynderickx &
Geert Jacobs (eds)
Researching Discourse in Business Genres. Cases and Corpora.
215 pages. 2012. ISBN 978-3-0343-1092-5
Vol. 153 Yongyan Zheng

Dynamic Vocabulary Development in a Foreign Language.
262 pages. 2012. ISBN 978-3-0343-1106-9
Vol. 155 David Hirsh (ed.)

Current Perspectives in Second Language Vocabulary Research.
180 pages. 2012. ISBN 978-3-0343-1108-3
Vol. 156 Seiji Shinkawa

Unhistorical Gender Assignment in Lahamon’s Brut. A Case Study of a Late Stage
in the Development of Grammatical Gender toward its Ultimate Loss.
186 pages. 2012. ISBN 978-3-0343-1124-3
Vol. 158 Bárbara Eizaga Rebollar (ed.)

Studies in Linguistics and Cognition.
301 pages. 2012. ISBN 978-3-0343-1138-0
Vol. 159 Giuliana Garzone, Paola Catenaccio, Chiara Degano (eds)
Genre Change in the Contemporary World. Short-term Diachronic Perspectives.
329 pages. 2012. ISBN 978-3-0343-1214-1
Vol. 160 Carol Berkenkotter, Vijay K. Bhatia & Maurizio Gotti (eds)
Insights into Academic Genres.
468 pages. 2012. ISBN 978-3-0343-1211-0
Vol. 162 Patrizia Anesa

Jury Trials and the Popularization of Legal Language. A Discourse Analytical Approach.
247 pages. 2012. ISBN 978-3-0343-1231-8

Li Li: Grammar in Cross-Linguistic Perspective

Uploaded by

Copyright:

Available Formats

You might also like

Li Li: Grammar in Cross-Linguistic Perspective

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Li Li: Grammar in Cross-Linguistic Perspective

Uploaded by

Copyright:

Available Formats

li57 In this collection of papers on syntax, semantics

and pragmatics, linguists specialising in the

Teruhiro Ishiguro & Kang-kwong Luke (eds) • Grammar in Cross-Linguistic Perspective

Teruhiro Ishiguro & Kang-kwong Luke (eds)

www.peterlang.com Peter Lang

Teruhiro Ishiguro & Kang-kwong Luke (eds) • Grammar in Cross-Linguistic Perspective

Teruhiro Ishiguro & Kang-kwong Luke (eds)

Edited by Maurizio Gotti,

The Syntax, Semantics, and Pragmatics

British Library Cataloguing-in-Publication Data: A catalogue record for this

Library of Congress Cataloging-in-Publication Data

Grammar in Cross-Linguistic Perspective : the Syntax, Semantics, and Pragmatics

© Peter Lang AG, International Academic Publishers, Bern 2012

All rights reserved.

TERUHIRO ISHIGURO and K.K. LUKE

WONG PING WAI

Notes on Contributors ..................................................................301

TERUHIRO ISHIGURO / K.K. LUKE

The idea of making a collection of papers on the study of syntax,

This chapter deals with Japanese complex verbal noun phrases of

(1) a. ᝟ሗ཰㞟 (jouhou-shuushuu, ‘information-gathering’)

From condition (2a), instances such as those in (3) are excluded.

1 The expression ‘Chinese-origin’ is not always used accurately. Here, it roughly

The morphemes ka (‘-ation’), sei (‘-made’), sai (‘re-’), and hatsu

Finally (2c) excludes a certain group of compound kango VNPs that

In the following sections, compound kango VNPs meeting the condi-

Past attempts to describe word formation and phrase structure

The outline of this chapter is as follows. In Section 2, I briefly describe

In Section 3, I attempt to classify compound kango VNPs into

2. Components of Kango Complex VNP

2.1 Role and Function of Kango

2.2 Types of Simple Kango VNs

(6) a. sequence of synonymous

Among simple kango VNs in Japanese, those that contain an argu-

(i) a. ⾜ᨻᨵ㠉 (gyousei-kaikaku, administration-reform) → ⾜㠉 (gyou-kaku)

(tei-sen, ‘halt-war’), patient (satsu-jin, ‘kill-person’), goal (ki-koku,

(7) a. ♫ဨເ㞟 (shain-boshuu, ‘employees-recruit’)

Virtually all simple intransitive kango VNs are unaccusative in the

Interestingly, when an unaccusative VN combines with a subject NP

6 See also Levin/Rappaport Hovav (1995) for an extensive discussion about

The issue of transitivity in compound kango VNPs is further dis-

2.3 Functions of Kango VN

As mentioned in 2.1, kango words express concepts of various types

Kango VNs are not only alternative expressions of their non-kango

whereas other times converted nominals do not inherit the argument

An original Japanese verb and its kango alternative do not necessar-

3. Cognitive Structure of Compound Kango VNPs

There may seem to be countless ways in which kango words com-

3.1 Classification of Compound Kango VNPs

(15) a. verb + verb (V+V)

7 It is possible, however, for yunyuu-hanbai to mean ‘importing AND retailing’

(a) (b) (c)

Figure (1a) illustrates the conceptual structure of VNPs like hoshu-

two component structures. Since the profile of the overall structure

(18) a. ពぢ஺᥮ (iken-koukan, 䇺opinion-exchange䇻)

Compound kango VNPs of the Obj+Vt type (18) is composed of a

(19) a. ༠ຊ౫㢗 (kyouryhoku-irai, 䇺cooperation-request䇻)

(20) a. ⫋ሙ᚟ᖐ (shokuba-fukki, ‘work place-return’)