Sei sulla pagina 1di 320

PERSPECTIVES ON TOPICALIZATION

TYPOLOGICAL STUDIES IN L A N G U A G E (TSL)

A companion series to the journal "STUDIES IN L A N G U A G E "

Honorary Editor: Joseph H. Greenberg


General Editor: T. Givon

Editorial Board:
Alton Becker (Michigan) Paul Hopper (Binghamton)
Wallace Chafe (Santa Barbara) Margaret Langdon (San Diego)
Bernard Comrie (Los Angeles) Charles Li (Santa Barbara)
Scott DeLancey (Oregon) Johanna Nichols (Berkeley)
Gerard Diffloth (Chicago) Andrew Pawley (Auckland)
R.M.W.Dixon (Canberra) Frans Plank (Konstanz)
John Haiman (Winnipeg) Gillian Sankoff (Philadelphia)
Kenneth Hale (Cambridge, Mass.) Dan Slobin (Berkeley)
Bernd Heine (Köln) Sandra Thompson (Santa Barbara)

Volumes in this series will be functionally and typologically oriented, cove­


ring specific topics in language by collecting together data from a wide variety
of languages and language typologies. The orientation of the volumes will be
substantive rather than formal, with the aim of investigating universals of
human language via as broadly defined a data base as possible, leaning
toward cross-linguistic, diachronic, developmental and live-discourse data.
The series is, in spirit as well as in fact, a continuation of the tradition initiated
by C. Li (Word Order and Word Order Change, Subject and Topic,
Mechanisms for Syntactic Change) and continued byT. Givon (Discourse and
Syntax) and P. Hopper (Tense-Aspect: Between Semantics and Pragmatics).

Volume 14

John Hinds, Senko K. Maynard & Shoichi Iwasaki (eds.)

PERSPECTIVES ON TOPICALIZATION
THE CASE OF JAPANESE 'WA'
PERSPECTIVES ON TOPICALIZATION
THE CASE OF JAPANESE WA

edited by

JOHN HINDS
SENKO K. MAYNARD
and
SHOICHIWASAKI

JOHN BENJAMINS PUBLISHING COMPANY


Amsterdam/Philadelphia

1987
Library of Congress Cataloging in Publication Data
Perspectives on topicalization.
(Typological studies in language, ISSN 0167-7373; v. 14)
Includes bibliographies.
1. Wa (The Japanese word) I. Hinds, John. II. Maynard, Senko K. III. Iwasaki, Shoichi.
IV. Series.
PL669.W3P47 1987 495.6'2 87-29982
ISBN 90-272-2885-X (hb.)/90-272-2886-8 (pb.) (European; alk. paper)
ISBN 0-915027-97-6 (hb.)/0-915027-98-4 (pb.) (U.S.; alk. paper)
© Copyright 1987 - John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or
any other means, without written permission from the publisher.
TABLE OF CONTENTS

PREFACE vii

Part I: Wa in Narrative and Expository Discourse


The Use of Wa as a Cohesion Marker in Japanese Oral Narratives
Patricia M. Clancy and Pamela Downing 3
Thematization as a Staging Device in the Japanese Narrative
Senko K. Maynard 57
Thematization, Assumed Familiarity, Staging, and Syntactic Binding in
Japanese
John Hinds 83
Identifiability, Scope-Setting, and the Particle Wa: A Study of
Japanese Spoken Expository Discourse
Shoichi Iwasaki 107
A Study of the So-called Topic Wa in Passages from Tolstoi, Lawrence
and Faulkner (Of course, in Japanese Translation)
S.-Y. Kuroda 143

Part II: Wa and other Syntactic Phenomena


The Role of Wa in Negation
Naomi Hanaoka McGloin 165
Wa and the WH Phrase
Shigeru Miyagawa 185

Part III: Historical Perspectives on Wa


Functions of the Theme Marker Wa from Synchronic and Diachronic
Perspectives
Noriko Fujii Ueno 221
Wa in Diachronic Perspective
Charles M. De Wolf 265
VI CONTENTS

Part IV: Pragmatic Perspectives on Wa


How Relevant is a Functional Notion of Communicative Orientation to
Ga and Wa?
Seiichi Makino 293

ABBREVIATIONS 307
PREFACE

The Japanese topic marker wa has been the research focus of many
scholars both in Japan and overseas. In Japan, Kokugogaku (National Lan­
guage Studies) scholars have identified wa by a variety of terms, among
which kakarijoshi is perhaps one of the most widely used. Different charac­
terizations of wa have been proposed in association with various functions
as well. For example, wa serves the function of tsuite setsumei suru kotogara
o toritatete yuu "mentioning the items to be explained" and kyoochoosuru
"emphasizing" in Yamazaki (1965). Wa appears in handanbun "sentences of
judgments" according to Nagano (1972) and wa expresses fuhensei "univer­
sal phenomenon" as suggested by Miura (1976).
It is through the work of Mikami (1960, 1972) and Kuno (1972, 1973)
however, that the study of wa has become the focus of many scholars
abroad. Mikami characterizes wa as teidai no wa "theme-presenting wa" in
the context of dai-jutsu kankei "theme-predicate correspondence." Mikami
also explores the function of wa across clause and sentence boundaries and
identifies such phenomena as konma-goe "comma crossing" and piriodo-
goe "period crossing." Kuno's study of wa as a theme marker has intro­
duced terms now widely used, such as "anaphoric wa," "wa for generic
NPs" and "NPs of unique reference." These terms characterize "thematic
wa" as opposed to "contrastive wa" and both of these are in contrast to
noun phrases marked by ga in discourse. Both Mikami's and Kuno's works
have influenced almost all subsequent studies of wa abroad as well as in
Japan. Today we find studies of wa within the framework of various
theories and methods in linguistics and related disciplines, some of which
have been developed only recently.
This volume suggests that various approaches to the topic marker wa
are not only possible but are beneficial. We do not expect to have the ulti­
mate answer to persisting questions surrounding wa; rather, we wish to con­
tinue our efforts to explore different pathways and different research
methods based on different philosophical standpoints with the hope that
each perspective will shed additional light on different aspects of Japanese
topicalization. Naturally, this volume offers only some possible approaches.
Vlll PREFACE

In our effort to construct a grammar and a systematic account of language


and language use in its context, the Japanese topic marker wa continues to
offer us a genuine challenge.
The idea for this volume was conceived in March, 1984, when the
editors formed a panel entitled "Functions of wa in Japanese discourse" at
the 1984 Annual Meeting of the Associaton for Asian Studies held in
Washington D.C. At that panel the editors each presented earlier versions
of the papers collected here. Seiichi Makino acted as discussant at the panel
and also presented the broad outline of his paper in this volume.
Stimulated by discussions with the audience and among panelists, we
decided to invite contributions from scholars whose research has dealt with
a variety of issues surrounding wa. In all, eleven scholars have contributed
to this volume.
This volume is organized around four broad subareas of linguistic
analysis at the level of discourse, syntax, studies with a historical perspec­
tive, and research concerned with pragmatic aspects of communication.
A focus on wa in discourse as observed in data consisting of natural
language is the common position shared by all the contributors in Part 1.
Clancy and Downing explore functions of wa as a cohesive device in dis­
course by analyzing data consisting of elicited oral narratives. They distin­
guish the function of wa at two levels of discourse cohesion, namely, global
and local, and propose that it is primarily a locally motivated cohesion that
wa marks. Specifically they characterize the local cohesive function in a con-
trastive context and claim that wa essentially links textual elements of vary­
ing degrees of contrastivity, thus expanding the notion of contrastiveness
associated with wa to a broader context.
Based on data taken from stories written for children, Maynard prop­
oses a "staging" effect (as introduced in Maynard 1980), as one discourse
function of wa in Japanese narrative discourse. By pointing out similar frag­
ments taken from narratives written by the same author where the occur­
rence/non-occurrence of wa and ga cannot be fully accounted for in the
traditional framework of the given/new dichotomy, Maynard argues that
"staging" is one packaging device (realized partly by wa) that the user of
the language has in order to express his/her perspective toward the events
described.
Hinds examines both expository and conversational data from a
number of perspectives and proposes that the choice of wa/ga is motivated
by a host of factors that a speaker/writer faces, namely, "assumed familiar-
PREFACE I

ity" (Prince 1981), "syntactic binding" which realizes the preferred topic-
comment structure, "staging" (Maynard 1980, and in this volume) and
rhetorical organization where discourse connection (particularly in the ten
of the ki-shoo-ten-ketsu organizational principle of Japanese) is
accomplished by thematization.
In his effort to identify the discourse function of wa in oral expository
discourse, Iwasaki focuses on two related issues, the pragmatic status of a
wa-marked noun phrase and the functions of these noun phrases in dis­
course. Based on oral descriptions of living quarter layouts, Iwasaki prop­
oses that "identifiability" characterizes wa-marked noun phrases more
accurately than the often cited "anaphoric" or "generic" category. Addi­
tionally, Iwasaki proposes that the primary discourse function of wa is
"scope-setting," and he adds three derivative functions as well; providing
"multiple predications," indicating "negative scope" and "contrastive ele­
ment" in its discourse context.
Within the framework of his earlier distinction between "categorical"
and "thetic" judgment (Kuroda 1972) — but with a new interpretation —,
Kuroda explores the discourse properties of wa in literature, namely in pas­
sages from Tolstoi, Lawrence and Faulkner in Japanese translation. Specifi­
cally, Kuroda's interest lies in wa/ga markings of noun phrases that are sub­
jects of quotative verbs. His inquiry addresses the literary functions of
wa/ga markings from the perspective of narrative point of view.
In Part 2, grammatical and functional characteristics of wa are revealed
in relation to other syntactic phenomena, namely negation and WH-wa
questions. In the first paper, McGloin's interest lies in the role that wa plays
in aspects of negation. McGloin characterizes negation with ga (as well as
other non-wa marked negation) as "speaker-motivated" since it often rep­
resents a negative statement contrary to common belief. On the other
hand, negation with wa in general negates matter-of-factly, and is referred
to as "discourse-motivated" negation. Furthermore, by incorporating the
contrastive/thematic characterization of wa, McGloin proposes that the
primary function of wa is to determine the scope of negation in Japanese. In
the second paper, Miyagawa proposes the most fundamental property of wa
as being "set-anaphoric" on the basis of evidence observed in WH-wa ques­
tions in Japanese. After identifying ga as a "creator of sets" in contrast to
wa, Miyagawa attempts to characterize various functions of wa and ga
catalogued by Kuno by applying these two broader concepts.
Part 3 contains two diachronic studies of wa. In the first paper Ueno
χ PREFACE

shows through an investigation of wa in Taketori Monogatari and Genji


Monogatari that the functions of wa in present-day Japanese — contrastive-
ness, theme creation, and theme maintenance — can be traced back to his­
torical developments starting in the tenth century. In the second paper,
after tracing the etymology, pronunciation, orthography, and the function
of wa through history, De Wolf concludes that wa in Old Japanese is a local
phenomenon often operating within a single sentence, while wa in Modern
Japanese is primarily a topic marker which interacts with discourse organi­
zation.
Finally, in Part 4, Makino identifies functional differences between wal
ga in terms of a broader communicative context. Makino's claim is that the
inclusion of the concept "Communicative Orientation" (Makino 1983) can
account for many of syntactic and semantic phenomena in Japanese,
including the walga distinction. Makino concludes that the use of wa (both
thematic and contrastive) is listener-oriented, while the use of ga (descrip­
tive, exhaustive listing and objective) signals speaker-orientation.
Although all works included in this volume examine wa, the terminol­
ogy used to refer to wa may differ. Some prefer "theme" or "thematiza-
tion" while others prefer "topic" or "topicalization." In general the ter­
minology selected by each contributor reflects the disciplinary orientation
he/she prefers to take. Thus, although the title of the volume adopts the
term "topicalization" it should be understood to be a broader term encom­
passing both "topic" and "theme."
April, 1986
JH
SKM
SI

REFERENCES

Kuno, Susumu. 1972. Functional Sentence Perspective — A case study


from Japanese and English. Linguistic Inquiry, 3, 269-320.
. 1973. The structure of the Japanese language. Cambridge: The M.I.T.
Press.
Kuroda, S.-Y. 1972. The categorical and the thetic judgment: Evidence
from Japanese syntax. Foundations of Language, 9, 2, 158-185.
Makino, Seiichi. 1983. Speaker/listener-orientation and formality marking
PREFACE XI

in Japanese. Gengo  84, 126-145.


Maynard, Senko K. 1980. Discourse functions of the Japanese theme marker
wa. Dissertation. Northwestern University.
Mikami, Akira. 1960. Zoo wa hana ga nagai. Tokyo: Kuroshio Shuppan.
. 1972. Gendaigohoo josetsu. Tokyo: Kuroshio Shuppan.
Miura, Tsutomu. 1976. Nihongo wa doo yuu kotoba ka. Koodansha
gakujutsubunko. Tokyo: Koodansha.
Nagano, Masaru. 1972. Bunshooron shoosetsu. Tokyo: Asakura Shoten.
Prince, Ellen. 1981. Toward a taxonomy of given-new information. In Rad­
ical pragmatics ed. by Peter Cole, 223-255. New York: Academic Press.
Yamazaki Yoshiyuki. 1965. Nihongo no bunpoo kinoo ni kansuru taikeiteki
kenkyuu. Tokyo: Kazama Shoboo.
PART I: WA IN NARRATIVE AND EXPOSITORY DISCOURSE
THE USE OF WA AS A COHESION MARKER
IN JAPANESE ORAL NARRATIVES

PATRICIA M. CLANCY

University of Southern California

and

PAMELA DOWNING
University of Wisconsin, Milwaukee

In recent years investigation of the notion 'topic' has moved from


attempts to formulate a universal definition at the sentence level (e.g.,
Chafe 1976; Li and Thompson 1976) to the exploration of topicality in con­
nected discourse (e.g., Givón 1983). In this paper we will analyze the use of
the Japanese topic marker wa in references to human participants in
Japanese oral narratives. We have chosen to concentrate on mentions of
human participants in order to build upon the growing literature on refer­
ential progressions in discourse; we have chosen to concentrate on oral nar­
ratives because the planning of oral language is subject to constraints quite
different from those that govern the production of written narratives, which
have been the focus of most previous inquiry.
We have taken as a starting point for our analysis the observation that
all uses of wa can be viewed as serving a cohesive function within the text.
This can be seen quite readily with respect to uses of the so-called "thema­
tic" wa (Kuno 1973), which is used to set up a thematic framework which,
in theory at least, may last for more than a single sentence. In such cases,
the thematic framework established by wa may be seen as providing a
means for internally unifying a section of text and setting it off from sec­
tions of text which do not have the same theme. Uses of wa of the sort that
has been called "contrastive" (Kuno 1973) can also be seen as fulfilling a
cohesive role, in that they mark a relation, i.e., contrast, that necessarily
involves two or more textual elements which might otherwise be presented
4 PATRICIA M. CLANCY AND PAMELA DOWNING

as autonomous.
The scope of the cohesive bond effected by wa of course varies consid­
erably from use to use. A very global exploitation of wďs cohesive power
can be seen in cases where it serves to set up a thematic framework which
persists for the duration of the text in which it appears. In cases where it is
used merely to mark a contrast between elements in adjacent clauses, on
the other hand, the scope of the cohesive bond is extremely local. Although
many cases fall on the continuum which stretches between these extremes,
and although some uses of wa can be seen as working in the interests of
global and local cohesion simultaneously, we have found the local — global
distinction to be a useful one for our analysis, and we have used it as a
means of organizing the discussion which follows. Our primary finding is
that the majority of the was in our narratives can be characterized as locally
motivated; globally motivated thematic was appear quite infrequently by
comparison. This finding is somewhat surprising, given the emphasis on the
thematic function of wa in previous studies; in our conclusions, we specu­
late that the preponderance of locally motivated was in our data may be
linked to the fact that the texts we are examining are oral rather than writ­
ten.
The data for our study came from three collections of oral narratives
elicited from native speakers of Japanese in Tokyo. The first collection
includes 17 narratives based on the "Pear film" (cf. Chafe 1980), a 7-minute
color film with live actors but no dialogue, which was used in a crosslinguis-
tic project on the verbalization of remembered experience. The film was
shown to groups of female university students, who then individually
recounted the plot of the story to a female interviewer. The other two data
collections were elicited from five male and five female university students,
as part of a larger study of the development of narrative structure in
Japanese (Clancy 1980b). For this project each subject was interviewed
individually by two female interviewers, an "elicitor" and a "listener." First
the elicitor showed the subject, one by one, a series of seven hand-drawn
cartoon strips of from five to eight frames each, depicting brief stories
involving four young children. After looking at a strip once, the subject told
the story to the listener while looking at the pictures. Two of these cartoons
are printed in Appendix A. Next, the listener left the room and the elicitor
showed the subject a 7-minute videotape from the popular television car­
toon series Sazaesan. The listener then returned and asked the subject to
tell her the story. To help the reader follow our discussion of these narra-
WA AS A COHESION MARKER 5

tives, a brief summary of each input story is given in Appendix B.

I. WA AS A GLOBAL COHESIVE DEVICE: "THEMATIC" WA-


MARKING
One of the most important aspects of discourse organization is referen­
tial structure, that is, the progression by which discourse participants are
introduced and then maintained or discontinued in subsequent text. This
progression, as well as the linguistic devices used to organize the inter­
relationships among participants in a text, have received much attention in
recent years (Grimes 1975, 1978; Hinds & Hinds 1979; Hinds 1983; DuBois
1980; Clancy 1980a; Givon 1983). Two important claims made by those who
have analyzed the functions of wa in Japanese discourse have been that wa
plays a role in the progression of participant reference, and that it signals a
special status in the discourse for the wa-marked referent. In this section we
will examine these claims in detail, analyzing how wa-status is established,
maintained, and lost in our narratives, and how wa-marked referents differ
from other participants. In this way we hope to clarify the nature of the
"thematic" use of wa, thereby elucidating the function of wa as a global
cohesive device in Japanese narrative.

A. Wa's Position in the Paradigmatic Referential Progression


Prior research on the functions of wa in Japanese discourse has sought
to establish the typical progression of referential forms used to introduce
story characters into narratives, and the place of wa in that progression.
Hinds & Hinds (1979) propose that the introduction of participants in a nar­
rative typically takes place with a "nontopicalizing particle" such as the
nominative marker ga; next, the referent is "highlighted" by a second
explicit nominal reference, this time with wa; then, on subsequent men­
tions, the participant is referred to by ellipsis. Thus the claim is that "to-
picalization precedes ellipsis" in the referential progression introducing
story characters into discourse.
Similarly, Maynard (1980) presents a "paradigmatic model" for "a
maximally explicit way" of creating and maintaining topics. Again, the
progression begins with the introduction of a referent with ga, followed by
the attachment of wa to the NP chosen to be the theme of the paragraph,
and then subsequent maintaining of that theme by NP-wa or elliptical refer­
ences. Maynard's model includes a fourth step, namely, the re-introduction
6 PATRICIA M. CLANCY AND PAMELA DOWNING

of the theme with an explicit NP-wa reference after a certain interval, often
the beginning of a new paragraph. Thus both Hinds & Hinds and Maynard
take the basic progression to be NP-ga --> NP~wa --> 0 , with wa-marking
preceding the use of ellipsis.
In this section we will examine the typical progression of referential
forms and ννα-marking in participant introductions in our data. For this
analysis we will consider all first mentions of story characters, but only
those second mentions which do not follow an intervening reference to a
different character in subject position, since switch reference is a powerful
factor restricting referential choice, often preventing the use of ellipsis in
Japanese (cf. Clancy 1980a). Only the first two, rather than the first three,
mentions of a referent will be included, since it was rare, especially in the
cartoon data, to find three consecutive mentions of the same referent with­
out interference.
Table 1 presents our findings on referential progression for the intro­
duction of new story characters in our three data samples.

Table 1. Referential Progressions for Character Introductions


Cartoons Sazaesan Pear
NP-ga --> Ø 55.4% (46) 62.5% (15) 76.8% (43)
NP-ga --> NP-ga 8.4 (7) 8.3 (2) 10.7 (6)
NP-ga —> NP-wa 6.0 (5) 4.2 (1) 5.4 (3)
NP-wa --> Ø 7.2 (6) - ~
NP* --> Ø 2.4 (2) - 1.8 (1)
NP-oblique --> NP-obl 3.6 (3) - -
NP-obl --> NP-ga 3.6 (3) ~ - ,
NP-ga --> NP-obl 2.4 (2) — —
NP* --> NP-obl 1.2 (1) 4.2 (1) —
NP-obl -> 0 -- 4.2 (1) 1.8 (1)
other 9.6 (8) 16.7 (4) 3.6 (2)
(83) (24) (56)

*NP with no particle indicated refers to cases of subject NPs which


occurred without any postposition.

As the table shows, by far the most frequent sequence (63.8% on the
average) for introductions in each sample was NP-ga --> 0 . An example of
WA AS A COHESION MARKER 7

this progression is presented in 1).1


(1) ano.. sono— chikaku ni, ano san-і no
uh that near LC uh 3-CL (person) LK
otoko-no-ko-tachi ga iru n desu ne?
boys SB exist NOM COP EX
sorede, ano., sore  hirotte ageru wake desu.
then uh that DO pick-up:GER give NOM COP
Nearby, there are three boys. Then (they) pick them up (for
him).
The frequency of this pattern in our data is consistent with recent claims
that NF-wa is an optional step in participant introductions in Japanese
folktales (Hamada 1983; Hinds 1984). However, our results provide a
strong basis for proposing that the prototypical or unmarked referential
progression, at least for oral narratives, is actually NF-ga --> 0 .
Less frequent, but still accounting for 9.2% of all introductions, was
the use of NF-ga on both the first and second mentions of a referent. Most
cases of this progression involved "reruns", in which the narrator either
repeated a clause following a false start or digression, or restated the same
information, as in the following Pear narrative excerpt:
(2) ano— onna-no-ko ga kotchi kara kita wake, chotto
uh girl SB here from came NOM sort-of
kami no nagai ne? onna-no-ko ga kite,
hair SB long EX girl SB come:GER
'A girl came from this direction. Sort of long-haired? A girl
comes,'
An unexpected finding was that, as Table 1 shows, in the cartoon data
wa sometimes occurred on the first mention of a referent. However, since
the four main characters were first introduced to the narrator and then kept
appearing in the seven cartoons, in these cases of NF-wa the narrator was
probably not treating the referents as true first mentions. The fact that NP-
wa on a first mention never occurred with the first cartoon narrated, but
only subsequent ones, supports this conclusion.
In sum, the unmarked introductory referential progression in our data
was clearly NF-ga --> 0 . The pattern with NF-wa was not only optional, it
was quite rare, occurring in only 5.5% of introductions. Moreover, when
NF-wa did occur on second mentions, it was invariably in a local environ-
8 PATRICIA M. CLANCY AND PAMELA DOWNING

ment which tended to elicit wa-marking of participants in any part of the


stories, not just during introductions. These cases of wa-marking thus occur
because of specific local motivations, as will be discussed in Section II
below, rather than because wa-marking is the normal progression for intro­
ducing characters.

B. Correlation Between Wa-Marking and Thematic Status


Since such a small percentage of the referents in our narratives were
marked with wa during their introduction into the storyline, it would
appear that wa-marking is used to confer some special status upon only cer­
tain story characters. In prior research, wa has frequently been attributed
the function of marking the "theme" or "topic". Kuno (1976), for example,
has claimed that one of the two functions of wa is to mark the theme of the
sentence, which he defined as "what the rest of the sentence is about."
More recently, several attempts have been made to analyze the functions of
wa at the discourse level. In this section we will examine the evidence in our
data for the claim that wa marks a discourse-level theme in Japanese narra­
tive.
The position that wa functions to indicate a discourse theme has been
developed most fully by Maynard (1980), who analyzed the use of wa in
Japanese folktales written for children. Maynard's basic claim is that "-wa
functions in the Japanese paragraph primarily to identify the NP which the
writer has chosen as theme," and more specifically in narratives, "to signal
which one is the more prominent, constant figure on our thematic stage."
Since wa indicates that "every piece of information is to be interpreted in
relation to the main character" marked by wa, Maynard claims that wa pro­
vides a framework for the organization and interpretation of the following
text, indicating which propositions are important or subordinate, and how
they are related. Thus in Maynard's approach the "staging" function of wa
is two-fold: it identifies the central figure in the narrative and organizes the
story in terms of that character's point of view.
Hamada (1983), who also based his treatment of wa on a sample of
written Japanese folktales, essentially agrees with this analysis of wa, claim­
ing that "the choice between NP-wa and NP-ga depends on thematicity,"
with wa indicating that the writer has selected a particular participant as the
theme. In contrast, ga is used when the narrator chooses to describe events
objectively. Hamada claims that when one participant is more central or
salient than others, "the narrator is likely to describe the episode from the
WA AS A COHESION MARKER 9

central figure's perspective and to treat him as theme (wa) and the others as
non-theme (ga)." These analyses of the thematic function of wa quite
clearly place it, then, among the inventory of devices available to the
speaker who wishes to manipulate and clarify the global structure of the
narrative being created.
Although this analysis of wa apparently corresponds to native speak­
ers' intuitions, the only evidence in these studies that a particular referent
was, in fact, the theme was whether or not it received wa-marking. In this
paper we have decided, following Iwasaki's (1984) suggestion, to "observe
first what wa is doing in discourse" and then to "ask if this function corre­
lates with a notion of 'topic/theme' which can be independently con­
structed." Accordingly, to avoid circularity, we have sought a definition of
"main character" which does not rely upon wa; presumably, there are other
linguistic and non-linguistic expressions of this status.
For our analysis, we decided on three measures of thematicity or
character importance/prominence/centrality. First, a non-linguistic mea­
sure, available for the cartoon strip data only, is the number of frames in
which a particular character appears, and the number in which s/he appears
alone. These cartoons were deliberately constructed to have characters with
varying degrees of importance and of interference from other, secondary
characters, and it seemed to us that the most prominent characters in the
input could be seen as central to the plots in which they participated. Our
other two measures were linguistic: the persistence of a story character fol­
lowing a mention with NP-wa, and the occurrence of zero switch reference
to a character in a coreferential chain initiated by NP-wa. These linguistic
measures will be discussed in more detail following a consideration of the
results of our non-linguistic measure.

1) Wa and Character Prominence in the Cartoon Input


Table 2 presents our non-linguistic measure of "hero status" in the car­
toons, giving for each of the referents in the seven cartoons the percentage
of frames in which that character appeared or appeared alone. (A look at
the cartoons printed in Appendix A will allow the reader to make a subjec­
tive evaluation of the validity of this measure of character importance.)
For each cartoon the third column gives the percentage of all wa-mark-
ing on story characters that each character received. Clearly, there is an
extremely strong correlation between the strength of a character's hero
status, as indicated by the percentages in the first two columns, and the per-
10 PATRICIA M. CLANCY AND PAMELA DOWNING

;ro Status and Wa -Marking.


% of frames %of
in which % of frames wa-marking
CARTOONS referent app alone on referent
Bike Accident:
Taroo a 100.0% 83.3% 100.0%
*Yuki and Sachi 16.6 0 0
Onigiri:
*Yuki 100.0 71.4 80.0
friends 28.6 0 20.0
Theft:
* Hiroshi 100.0 40.0 82.6
Taroo 60.0 0 17.4
Snow Fight:
*Taroo 100.0 42.9 55.2
Yuki and Sachi 57.1 0 44.8
Exchange:
Yuki 77.7 22.2 57.4
Sachi 66.6 11.1 35.2
*Yuki and Sachi 55.5 33.3 7.4
mothers 22.2 0 0
Dress-up:
Yuki 100.0 0 34.3
Sachi 100.0 0 31.4
*Yuki and Sachi 100.0 57.1 28.5
mothers 42.9 0 5.7
Sand Fight:
Taroo 100.0 0 41.6
Hiroshi 100.0 0 35.4
*Taroo and Hiroshi 100.0 71.4 14.6
teacher 28.6 0 8.3
a
The referent identified as the main character of the story by this mea­
sure is marked by an asterisk. When no one character was sufficiently dom­
inant, in terms of frequency of appearance and appearance alone, the pair
of characters who were most important is marked.
WA AS A COHESION MARKER 11

centage of all wα-marking which that character received. This correlation


between our rather crude measure of hero status in the cartoons and the
frequency of wα-marking is consistent with Maynard's claim that wa marks
the most prominent referent, who figures most constantly in the plot, and it
provides support for the position that the function of wa in Japanese narra­
tives is to identify the main character.
But although the correlation between wa-marking and our measure of
hero status is strong, it is also clear from Table 2 that wa-marking occurred
rather frequently not only on the relatively important subordinate charac­
ters in cartoons like Theft and Snow Fight, but also on very minor charac­
ters, such as the teacher in Sand Fight. Thus a character did not need to be
central to the storyline as a whole in order to receive wa-marking.
This conclusion is supported by the results from the Sazaesan and Pear
narratives, as shown in Table 3.

Table 3. Distribution of Wa-Marking on Referents in the Sazaesan and


Pear Narratives.
Sazaesan Stories Pear Stories
Ikura 19.1% (9) bike boy 34.9% (15)
Wakame and Katsuo 23.4 (11) threesome 32.6 (14)
Wakame 6.4 (3) pear man 11.6 (5)
Ikura's mother 17.0 (8) goat man 9.3 (4)
Ikura's father 2.1 (1) girl 7.0 (3)
adults 12.8 (6) paddleball boy 4.7 (2)
"others"* 12.8 (6)
Sazaesan's father 2.1 (1) (43)
Sazaesan's mother 2.1 (1)
Sazaesan 2.1 (1)

(47)

*This category includes cases in which a narrator specified a certain


character or group of characters and then referred to the remaining people
on the scene or in the story as "others."

In the Sazaesan stories, for example, the primary instigator of the action is
clearly Ikura, but he was marked by wa less frequently than Wakame and
12 PATRICIA M. CLANCY AND PA MELA DOWNING

Katsuo, who were tormented by his mischief. And Ikura's mother, who
appears only sporadically in the story, was marked by wa about as fre­
quently as Ikura. In the Pear narratives, wa-marking similarly occurred not
only on the important characters, such as the bike boy and the threesome,
but also on minor characters, such as the goat man, who was deliberately
included in the film to observe how narrators would deal with a completely
irrelevant character.
In summary, there was a tendency for wa to occur more frequently on
characters having an important role in the overall storyline, especially in
the cartoon narratives, which had more telescoped plots. But wa-marking
occurred frequently enough on minor characters to indicate that a more
local motivation must underlie much of the wa-marking that occurred in
our data. We will take up in detail the nature of these local motivations for
wa-marking in Section II below.
Next, let us turn to our linguistic measures of hero status. By consider­
ing these measures, we hope to avoid the creeping circularity of our non-
linguistic criterion: the more frequently a character appeared in the car­
toons, the more frequently it was mentioned, allowing more opportunities
for wa-marking. We also hope to avoid a purely subjective definition of
theme (on the analyst's part), while allowing for subjectivity in the way the
narrator chooses to present and inter-relate the characters in the story. Our
linguistic measures are based on the assumption that selection of a particu­
lar character as overall or even temporary theme will have linguistic
reflexes other than wa-marking which can provide independent verification
or refutation of claims regarding the correlation between the use of wa and
thematic status. For our analysis we have chosen two linguistic measures
that have been applied cross-linguistically in other studies of referential
choice. Both measures are local in that they focus on ellipsis chains, which
in our data typically did not continue for very long. Thus the measures can
capture cases in which both story heroes and characters of temporary
importance are treated as thematic.

2) Wa and Character Persistence


The first linguistic measure of "hero status" is the persistence of a refe­
rent in the narration following mention with NF-wa. If wa serves to identify
the main character, it seems reasonable to anticipate that this referent will
continue to figure in the subsequent plot, at least for a while, (cf. Givón
1983).
WA AS A COHESION MARKER 13

For our analysis of persistence, we have counted the number of succes­


sive subsequent clauses in which the referent marked by wa continues as an
ellipted semantic argument of the main clause predicate.2 The purpose of this
measure is to discover whether marking with wa confers on a referent
greater durability as topic/theme, allowing for more protracted chains of
elliptical reference subsequent to mention with wa. Another nominal men­
tion of the same referent in subject position with ga (or without any parti­
cle3) would seem to indicate a loss of whatever special thematic status wa
might be capable of conferring. On the other hand, another NP-wa mention
would seem to indicate that the narrator felt a need to re-establish the refe­
rent as the topic. We have assigned a value of 0 in cases where the referent
of a wa-marked mention is not an ellipted argument of the following predi­
cate, and a value of 1 for the first, and then again for each subsequent
clause in which the referent is an ellipted argument in any semantic role.
For comparison, we have also tabulated the persistence of referents men­
tioned with NP-ga using the same procedure. (Again, subsequent nominal
mentions with either NP-ga or NP-wa were treated as breaking the ellipsis
chain). Furthermore, we have calculated the percentage of referents men­
tioned with NP-wa vs. NP-ga that persisted at all (i.e., for at least one sub­
sequent clause), since lengthy persistence in a few cases might substantially
influence the number-of-clauses measure.
Table 4 presents our findings on the persistence of NP-wa vs. -ga refer­
ences.

Table 4. Persistence of Referents Following NP-wa vs. NP-ga Mentions.


NP-wa NP-ga
Average # % Persisting Average # % Persisting
of Clauses at all of Clauses at all
Cartoons 1.5 59.7% 1.3 58.3%
Sazaesan .7 51.4 .8 38.6
Pear 1.9 52.4 2.3 56.4

As the table shows, the average persistence in number of clauses was


extremely similar for NP-wa and NP-ga-initiated chains of ellipsis, indicat­
ing that referents marked by wa do not persist longer in subsequent narra­
tion than referents marked by ga. Moreover, in the cartoon and Pear narra­
tives, the percentage of referents that persisted at all in subsequent narra­
tion was virtually identical for NP-wa and NP-ga. The Sazaesan stories,
14 PATRICIA M. CLANCY AND PAMELA DOWNING

however, show a striking departure from this pattern, with NP-wa introduc­
ing referents that continue to figure in the immediately following narration

Table 5. Relative Persistence of Individual Referents Following NP-W and


NP-ga Mentions.
NP- wa NP -ga
Average # % Persist- Average # % Persist­
CARTOONS of clauses in at all of clauses ing at all
Heroes:
Bike Acc. (Taroo) 2.8 72.7% 4.0 85.7%
Onigiri (Yuki) 2.1 62.5 4.7 90.0
Theft (Hiroshi) 1.2 52.6 1.0 50.0
Snow Fight (Taroo) 2.3 82.4 2.3 90.0
Exchange(Y/S) 1.0/.4 51.6/31.6 1.0/1.5 66.7/72.7
Dress-up (Y/S) 1.5/1.0 63.6/54.5 .5/-* 33.3/--
(Y&S) 1.1 72.7 - --
Sand Fight (T/H) .9/ .7 60.0/33.3 .8/ .9 50.0/50.0
(T&H) .3 33.3 .2 36.4
Subordinate Characters
Bike Acc. (Y & S) — -- .5 40.0
Onigiri (friends) - - .5 47.4
Theft (Taroo) - - 1.0 57.9
Snow F. (Y & S) 1.5 84.6 .7 63.6
Exchange (mothers) - - - -
Dress-up (mothers) -- - 1.0 44.4
Sand F. (teacher) -- - .8 50.0
SAZAESAN STORIES
Ikura .6 57.1 .9 38.0
Wakame & Katsuo 1.2 50.0 1.4 44.4
PEAR STORIES
bike boy 1.3 55.3 1.4 47.2
threesome .9 53.8 1.2 57.1

*For the cartoons, dashes indicate cases of five or fewer mentions,


which are probably unreliable, and so have been omitted from the table. In
the Pear and Sazaesan stories, only the four characters on the table
received more than five mentions with either NF-wa or NP-ga.
WA AS A COHESION MARKER 15

more frequently than NP-ga. Except for this single finding, however, the
persistence data provides little support for the claim that wa marks the main
or thematic character.
To investigate in more detail the relationship between persistence and
wα-marking, and especially the pattern shown on Table 4 for the Sazaesan
stories, we have calculated separately for each referent in the narratives the
relative persistence when that character was mentioned with NP-wa vs. NP-
ga. Table 5 presents the results of this analysis.
Looking first at the cartoon data, we can see from Table 5 that in the
two cartoons with a single important character (Bike Accident and Onigiri),
reference with NP-ga initiated much longer chains of coreferential ellipsis
and a higher percentage of references that persisted at all than did refer­
ences with NP-wa. This is because in relating these cartoons, narrators usu­
ally followed the typical ga —> 0 progression in introducing the heroes, who
then appeared alone in many successive frames, resulting in high persis­
tence figures. Hamada (1983), who also found the progression ga --> 0 in
referent introductions, notes that "if a sentence containing NP-ga precedes
ellipsis, whether the unexpressed referent is theme or not has to be estab­
lished on other grounds." In these cartoon stories the main characters were
presumably identified as such without benefit of wα-marking, by overall fre­
quency of mention, for example, (cf. Givón 1983), and by the nature of the
plot. Sometimes an introductory formula also served this function, only to
be followed by ga-marking on the hero, as in the following:
(3) ... de taroo ... tarookun hito-ri no ohanashi ne. ...
and Taroo Tarookun 1-CL (person) LK story EX
eto ne, ... tarookun ga, sakki wa sanrinsha datta
uh EX Taroo SB before TP tricycle COP:PST
n da kedo,
NOM COP but
And Taroo it's a story about just Taroo. And Taroo had a tricy­
cle before, but ...
Of course, no one has ever claimed that narrators must mark the main
character with wa. But upon reflection, one wonders exactly what wa-mark-
ing with a purely thematic function would contribute to stories other than
those having more than one important character. In such stories, the lis­
tener could be left in doubt as to which character the narrator perceived as
primary, as in certain of the stories Maynard analyzed. Perhaps wa need
not be used to mark the main character unless there is a rather large
16 PATRICIA M. CLANCY AND PAMELA DOWNING

amount of interference (appearance in subject position) by fairly important


subordinate characters.
Turning to the results for the other cartoons, we can see from Table 5
that the longer persistence of NF-ga tends to disappear when the hero
meets with rather strong secondary characters (e.g., Theft, Snow Fight) or
when there are "joint heroes" (Exchange, Dress-up, Sand Fight). In such
stories the high rate of switch reference reduces opportunities for persis­
tence. In many cases the length of persistence is roughly the same for NF-
wa and NF-ga, but there is considerable variation as to whether referents
persist more frequently following NF-ga or NF-wa. In the Pear and
Sazaesan narratives, similar inconsistencies, as well as a general lack of
clear differentiation in persistence values following wa vs. ga, can be seen.
The one striking exception is Ikura, an extremely important character in
the Sazaesan stories, who shows a higher frequency of persistence following
a mention with NF-wa, despite the typical trend toward slightly longer per­
sistence following NF-ga.
Except for this one case, there is virtually no evidence for a higher rate
of persistence following NF-wa references to (potential) main characters.
Both main and clearly subordinate characters may on occasion show greater
persistence with NF-wa, while main characters may show more frequent or
longer persistence with NF-ga. As with our non-linguistic measure of the
distribution of wa-marking on different characters (Tables 2-3), then, the
correlation between the use of NF-wa and this measure of thematic status
for a referent is imperfect at best, and any tendency toward greater persis­
tence with NF-wa is strongly overshadowed by other factors.

3) Wa and Zero Switch Reference


The second linguistic measure of main character or thematic status
which we investigated was the use of zero switch reference. It has fre­
quently been proposed that the hero of a narrative tends to be referred to
with more attenuated linguistic forms than peripheral characters (e.g.,
Ennulat 1978; Perrin 1978; Clancy 1980a). Givón (1983), on the basis of sev­
eral cross-linguistic studies of reference in discourse, concludes that the
more continuous and easy to identify a referent is, the more likely it is to
receive attentuated mentions, such as zero anaphora. Presumably, this is
partly because more important characters tend to persist longer in dis­
course, allowing more opportunity for successive coreference, which tends
to be accomplished with attentuated forms. However, it is probably also the
WA AS A COHESION MARKER 17

case that narrators assume the most important character to be easily identi­
fiable, and therefore treat it as "given information" with attenuated men­
tion (cf. Chafe 1976) even when there is potential competition from other
characters "on stage." In fact, main characters do tend to be mentioned
with attentuated forms even in contexts such as switch reference, which
usually elicit more explicit referential forms (cf. Hinds & Hinds 1979;
Clancy 1980a). If wa functions to establish a story participant as thematic,
then we can predict that NP-wa mentions will be more likely than NF-ga
mentions to initiate coreferential chains that will persist even across men­
tions (in subject position) of other characters. Since elliptical switch refer­
ence is not very frequent, this is an extremely stringent measure of a refe­
rent's status as the main or thematic character. However, there is some evi-

Table 6. Elliptical Switch Reference in Chains Initiated by NF-wa vs. NF-


ga.
Cartoon Stories: NP-wa NP-ga
Main Characters 26 33
Subordinate Characters 1 9
27 (39. 1%) 42 (60.8%)
Sazaesan Stories:
Ikura 1 10
Wakame and Katsuo 8 3
Ikura's father 3 11
Wakame 1 1
Ikura's mother — 5
Sazaesan — 2
Sazaesan and mother 1 —
Sazaesan, Wakame, and
Katsuo — 5
family 2 -
16~(30 .2%) ~37~(69.8%)
Pear Stories:
bike boy 8 15
threesome 2 4
paddleball boy -- 3
pear man -- 2
"Ī0 (29 .4%) ~24~(70.6%)
18 PATRICIA M. CLANCY AND PAMELA DOWNING

dence for the validity of this measure; Hamada (1983), who found cases in
which a central character received elliptical mention following intervention
by a peripheral character, reports that this occurred only when the central
character had already been established with NP-wa before the elliptical
switch reference.
To test the prediction that elliptical switch reference is more common
for main characters established as such with wa-marking, we examined
every case in which a coreferential chain of subjects initiated by NP-wa was
interrupted by a reference in subject position to a different character, fol­
lowed by a zero switch reference upon returning to the character in the wa­
initiated chain. For comparison, we also examined cases of zero switch ref­
erence to characters following ga-initiated chains. 4 Table 6 presents our
findings.
In the cartoon data, 60.8% of all elliptical switch references occurred
on referents whose last nominal mention was with NP-ga. Zero switch ref­
erence occurred much more frequently on main characters in the cartoons
as identified by our non-linguistic measure. However, elliptical switch ref­
erence was clearly more frequent following NP-ga than NP-wa mentions. A
typical example of elliptical switch reference in an NP-ga-initiated chain of
ellipsis is the following:
(4) ... de yukichan ga, daidokoro no naka de
and Yuki SB kitchen LK middle in
issshookenmei, onigiri  tsukutteimasu. ...de kondo
busily onigiri DO is-making and this time
kore  ... — ... iremono ni tsumemashite ne, ... de, ..
this DO sort-of container LC pack:GER EX and
dekakete iku wake desu. ... dekake te iku wake desu.
go-out:GER go NOM COP go-out:GER go NOM COP
... de onigiri o tor- ... toridashite, ... hoobari-
and onigiri DO ta- take out:GERcram-into-mouth-
nagara ... kooen ni ikimasu. ...de kooen ni wa, ...
while park to go and park at TP
hiroshikun to ... satchan ga imashite, ...de 
Hiroshi and Sachi SB be:GER and this
WA AS A COHESION MARKER 19

futa-ri ni onigiri  hitotsu zutsu


2-CL (person)  onigiri DO one-CL (inanimate) each
wakete ageru wake desu.
divide:GER give NOM COP
Yukichan is busily making onigiri in the kitchen. And next (she)
packs them and goes out. (She) goes out. And (she) takes out an
onigiri and, cramming it into (her) mouth, goes to the park. And
in the park are Hiroshi and Sachi, and (she) gives one to each (of
them).
In this example, the nature of the action, as well as the oblique mention of
the intervening prior subject referent (futari ni) certainly make it clear who
gave the onigiri. Although it would be possible for narrators to use elliptical
switch reference for subordinate characters too, relying on contextual infor­
mation to disambiguate as in example 4), this was rarely done. Thus the
results from the cartoon narratives suggest that elliptical switch reference is
a reliable indication of a character's importance in the story, but that it is no
more likely to occur when the referent has been marked by wa than when
it was mentioned with ga.
As Table 6 shows, this conclusion is supported by the results from the
Sazaesan and Pear narratives. In the Pear stories the bike boy and the
threesome were the characters ellipted most frequently at points of switch
reference, but zero switch reference was significantly higher following ga-
initiated chains than ones initiated by wα-marking. In the Sazaesan stories
the percentages of elliptical switch reference following NP-ga was again
more than double the rate following NP-wa. Although Ikura and the pair
Wakame and Katsuo are shown to be important characters by the fre­
quency of zero switch reference, as by certain of the other measures,
Ikura's father now would appear to be even more important, a result which
seems intuitively implausible. And the occurrence of zero switch reference
for relatively minor characters, such as Sazaesan, and the paddleball boy in
the pear narratives, indicates that although there is a tendency for this mea­
sure to correlate with character importance, exceptions are not uncommon
either. It is also interesting to note that there is a less than perfect correla­
tion between the results of this measure and those of the non-linguistic (Ta­
bles 3-4) and persistence measures discussed above.
Thus, as a careful comparison of Tables 2-6 reveals, the results of our
three intended measures of thematic status were not entirely consistent with
20 PATRICIA M. CLANCY AND PAMELA DOWNING

one another. Probably each is subject to somewhat different discourse con­


straints, and while tending to identify main characters, does not have this as
its sole function. This also appears to be true, however, for wa-marking.

 Loss of Thematic Status


Another important issue to examine in analyzing the thematic function
of wα-marking is the question of when and why this status is lost during the
course of narration. If wa marks the main character in a segment of narra­
tive, we would not expect a wa-marked referent to appear subsequently
with ga unless a new thematic character had been established with wa in the
interim, thereby "demoting" the original hero. Similarly, once a referent
has been marked by wa and ellipted, we would not expect the narrator to
return to explicit wa-marked nominal reference unless s/he felt that it was
necessary to re-establish the character in question as thematic for the lis­
tener. An obvious motivation for such re-establishment is the mention of
another character in subject position, causing the narrator to revert to nom­
inal mention of the original thematic participant in order to avoid
ambiguity; wa-marking would then presumably be used to indicate that the
referent in question is still thematic.
However, re-establishment of a referent with wα-marking also occurs
in cases in which no such intervening mention has occurred. In this section
we will consider the reasons for a referent's apparent loss of thematic status
which cannot be explained simply as the result of replacement by a new
thematic participant or the need to re-establish thematic status following
mention of a non-thematic character. We will examine two kinds of evi­
dence: 1) cases in which a participant already established with wa-marking
receives another NP-wa reference, although no other story character has
been mentioned, and 2) cases in which a wa-marked NP appears in sub­
sequent narration as NP-ga with no intervention from another wa-marked
referent.

1) Repeated Wa-Marking without Interference


Starting with the former cases, we find that one of the most important
factors that can disturb the status of the thematic participant within a
coreferential subject ellipsis chain is a discourse boundary. The speaker
may conclude one episode and begin another, for example, or may step
outside the narrative proper to provide background information or com-
WA AS A COHESION MARKER 21

mentary or to indulge in a digression which has no direct bearing on the


plot being developed. As Hinds & Hinds (1979), Clancy (1980a), and Hinds
(1984) have noted with respect to nominal ellipsis, such boundaries,
depending on their importance, may erase the thematic status of partici­
pants, requiring their explicit re-establishment in the new section of text
following the boundary.
Such discourse boundaries can explain the appearance of a number of
was in our narratives which would otherwise seem to be unmotivated, fol­
lowing as they do immediately on the heels of other mentions of the same
participant. In the example below, for instance, the narrator has chosen to
insert an NP-wa reference to Yuki even though she has been the only par­
ticipant in the narrative thus far and a continuation of the chain of zero
anaphora would presumably engender no listener confusion as to who is
going to the park.
(5) yukichan ga, ... onigiri  tsukut-... daidokoro de
Yuki SB onigiri DO ma- kitchen in
onigiri o, tsukuroo-to-shiteimasu. ... (String of zero
onigiri DO trying-to-make:GER
anaphoric references to Yuki as she prepares the onigiri)
sore  bentoo ni shite, yukichan wa sassoo-to
that DO lunch RS make:GER Yuki TP breezily
en ni dekakete ikimasu.
park to go-out:GER go
Yuki is making onigiri, is trying to make onigiri in the kitchen. ...
(Taking) them as her lunch, Yuki breezily sets off for the park.
Although it is Yuki who is involved in both the preceding and the suc-
ceding action, her departure for the park is clearly an important transition
point in the plot, marking the break between her solitary onigiri making in
the kitchen and her interaction with her friends in the park. Less important
action shifts, as when Yuki switches from seasoning the onigiri to shaping
them, never elicited full NP mentions, even though they, like the departure
for the park, were accorded separate frames in the cartoon.
Sometimes, of course, the discourse boundary is in part defined by a
shift to a new main character. In such cases, it is necessary to mark the new
thematic element explicitly in order to avoid confusion.5 When no new par­
ticipant is involved, though, the likelihood that the speaker will choose to
22 PATRICIA M. CLANCY AND PAMELA DOWNING

re-establish the earlier participant as thematic seems to depend on the


importance of the boundary that has been crossed, and different speakers
may make different decisions. If the boundary involves shifts in a number
of the elements that make for cohesion, such as activity, spatial location,
and so on, or if it involves a shift from one discourse mode to another, as in
a shift from narration to commentary, re-establishment of the main charac­
ter with wa is more likely. In most, although not all, of the cases of this sort
in our data, the motivation seems to be dual: both to mark the discourse
boundary and to re-establish wa-status for the main character following the
boundary, (cf. Lee 1984). Perhaps because participant orientation often
changes at a major episode boundary, such points in narration are vulnera­
ble ones for thematic participants, and narrators will use NP-wa to indicate
that the boundary did not cause the referent to lose its special status.

2) Reversion from Wa to Ga
In some cases, a story character may simply lose its thematic status.
The clearest indication of this loss is a reversion to ga-marking for a for­
merly wa-marked referent, although disappearance from the storyline or
appearance only in oblique positions may also reflect such a loss. In our
data we found 20 cases in which a formerly wa-marked referent appeared
with ga, although no other character had been marked by wa in the interim.
These cases fell into two basic types: those in which a clear motivation for
ga-marking was apparent, and those in which wa-marking was lost either at
a discourse boundary or for no obvious reason. In the latter cases it would
appear that the original wa just didn't "take" for more than an extremely
limited bit of narration.
Among the cases in which the loss of thematic status was motivated by
a specific reason for using ga, the majority occurred in embedded or subor­
dinated clauses. It is well-known that in Japanese, thematic wa-marking
does not occur in embedded clauses (cf. Kuno 1973). The other specific
motivation for a shift to ga in our data was the narrator's decision to mark
a referent with "exhaustive listing" ga (cf. Kuno 1973). In 6), for example,
the narrator initially marked the referent with wa, but then, following a dig­
ression hesitating over who performed the action, shifted to 'exhaustive list­
ing' ga to emphasize which of the two referents was in fact the actor.
(6) hiroshikun wa ne, ... boku wa ne, ... wanpaku a!
Hiroshi TP EX I TP EX naughty ah!
WA AS A COHESION MARKER 23

soo ja-nai. a ... yappa soo da. ...


so -NEG ah after-all so COP
hiroshi ga itta n da.
Hiroshi SB said NOM COP
Hiroshi, I, naughty Oh! It wasn't. Oh, right. It's Hiroshi who
said (it).
In such cases the digression apparently caused the speaker to forget that
there was a reason for wa-marking in the prior discourse environment, and/
or provided a new reason for ga-marking.
Our data also included several examples in which thematic status was
lost with no apparent reason for ga-marking. Sometimes the wa --> ga shift
occurred at a major discourse boundary, e.g., the shift from introductory
commentary; in these cases the discourse boundary apparently erased the
referent's thematic status. The transition resulting in the loss of wa-marking
was sometimes extremely minor. In other cases, there was neither a dis­
course boundary nor switch reference, yet wa-status was lost, as in the fol­
lowing example:
(7) (description of Ikura's mischief, throwing a shoe that hits
Wakame)
kanojo wa atama-ni-/kita/. ... de- ... sono
she TP got-angry and that
... yoso no kodomo demo ne, ... warui koto
other-people GN child even EX bad thing
 shita toki ni wa, ... akachan demo okoru beki
DO did time at TP baby even scold should
da tte /kanojo/ ga yuu wake ne.
COP QT she SB say NOM EX
She got angry. And "Even if it's a neighbor's child, when he does
something bad, even if (it's) a baby, (you) should scold (him),"
she says.
The implication of cases like (7) is that the use of wa may sometimes be loc­
ally motivated and thus fail to confer a lasting thematic status on the refe­
rent in question.
24 PATRICIA M. CLANCY AND PAMELA DOWNING

D. Conclusions on Thematic Wa
Having examined a number of linguistic and non-linguistic concomit­
ants of thematic status in narrative and their correlation with the wa-mark-
ing patterns shown in our data, we are ready to address the fundamental
question, "Does wα mark thematic status of participants in oral narra­
tives?" As we have seen, the results of our analysis are somewhat mixed,
but the general conclusion would seem to be that wa-marking is not neces­
sary to establish thematic status, nor does wa-marking, when it appears,
necessarily indicate that the participant in question is thematic, to the extent
that thematicity can be equated with the measures that we have considered,
i.e., frequency of appearance, persistence, or ability to elicit zero switch
reference. Coupled with the evidence from cases where speakers revert
from wa-marking to ga-marking for no apparent reason, these findings
suggest that many cases of wα-marking may be due to extremely local moti­
vations unrelated to the narrator's desire to elevate a particular character to
the role of thematic participant.
One might, of course, object that these conclusions are overly pes­
simistic, since they are based in some cases on the assumption that a thema­
tic participant will function as such for an extended section of text. Our per­
sistence measure, for example, attempts to correlate wa-marking and the
subsequent behavior of the referent of the wa-marked noun phrase; we
would not necessarily expect to find any such correlation if we assumed that
wa could mark sentence-level themes as well as higher level (paragraph or
text level) themes. Ueno (this volume), for instance, takes the view that wa
can mark themes at all three of these different levels. Given an assumption
of this sort, it might be argued that if the theme in question is merely a sen­
tence-level one, we cannot expect to discover the other reflexes of thema­
ticity that we have considered, since they are associated only with the
thematic participants in larger segments of text.
Such an interpretation is in principle consonant with our data; how­
ever, in the face of our negative findings with respect to the correlation
between wα-marking and discourse-level themes, we are reluctant to adopt
such an account without some independent means of verifying the thematic­
ity of a wa-marked participant at the sentence level. And it is not clear to us
how this can be done. Consider example (8).
(8) yukichan no mama to, ... satchan no mama ga odekake
Yuki GN mama and Sachi GN mama SB going-out
WA AS A COHESION MARKER 25

shimashita. ... yukichan to satchan wa, ... ii ko


did Yuki and Sachi TP good child
de asondeita n da kedomo, ...
COP:GER were-playing NOM COP but
shibarakushite, ... otoo-... o- otoosan ya okaasan no
after-a-while fa,- f- father and mother GN
yoofuku ... ga mieta kara, ... sore o ki- ... kiyoo
clothing SB was-visible because that DO wea- wear-will
to yuu koto ni narimashita. ... de, ... yukichan wa, ...
QT say thing RS became and Yuki TP
okaasan no, ... kireina kimono  kite, ... satchan
mother GN beautiful kimono DO put-on:GER Sachi
wa otoosan no sebiro o kimashita. ... nekutai mo
TP father GN suit DO put-on necktie too
shimete ... de ... kuchibeni mo shite, ... chanto shita wa
put-on:GER and lipstick too do:GER nicely did EX
to yuu kanji de, ... iru to, ... okaasan-tachi ga
QT say feeling COP:GER be when mother-PL SB
kaette kimashita. ... wa!... okorareru to
return:GER came uh oh! be-scolded QT
omotta kedo, ... okaasan-tachi wa oowarai  shimashita.
thought but mother-PL TP big-laugh DO did
'Yuki and Sachi's mothers went out. Yuki and Sachi were play­
ing like good children, but after a while, since (they) noticed the
father's and mother's clothing, (they) got involved in putting
them on. And Yuki put on the mother's beautiful kimono, and
Sachi put on the father's suit. (They) put on a necktie, and put
on lipstick, and when (they were there) feeling nicely done up,
the mothers returned. "Uh oh!" (They) thought (they) would get
scolded, but the mothers laughed.'
We can see from this narrative as a whole that from the discourse perspec­
tive, the story is clearly "about" Yuki and Sachi, rather than their mothers.
They appear in a much larger portion of the narrative, and they are men­
tioned much more frequently with both NF-wa and zero anaphors. Even
26 PATRICIA M. CLANCY AND PAMELA DOWNING

within the sentence in which the wa-marked reference to their mothers


appears, in fact, there are elliptical references to Yuki and Sachi. And in
one of these they appear as the subject of a passive, a construction typically
used to promote a thematic underlying object and/or to background a non-
thematic underlying subject (in this case, the mothers). If we are to take the
fall-back position that, in spite of the thematicity of Yuki and Sachi at a dis­
course level, it is the mothers who are thematic in this sentence, what inde­
pendent evidence can we use to corroborate our claim?
It is because we feel there is as yet no satisfactory answer to this funda­
mental question that we have declined to pursue the notion that wa is fulfil­
ling a theme-marking role even when the scope of that thematicity is only a
single sentence. Without independent evidence for thematicity, we are left
with the following dilemma: if thematic status can hold simultaneously at
several different levels of narrative structure (a not implausible claim), and
can in principle apply to any story character at the sentential level, then the
thematic account of wa's function is not only unfalsifiable, it also fails to
provide a principled way of predicting which references to subordinate
characters may be marked by wa.
Instead of assuming sentential themes, then, we have attempted
instead to discern what sorts of local (as opposed to global, thematic) fac­
tors might be triggering the use of wa in cases where it does not confer last­
ing thematic status on the character thus referred to. It is to these locally
motivated wa uses that we will turn in the following section.

IL WA AS A LOCAL COHESIVE DEVICE: "CONTRASTIVE" WA-


MARKING
Most scholars who have worked on wa have noted cases in which the
primary function of wa does not seem to be the marking of a thematic ele­
ment, but rather, the marking of a contrastive relation between the wa-
bearing element and some other element, which may or may not receive
explicit linguistic representation. Kuno (1973), for example, has pointed
out the existence of sentences like the one shown in (9).
(9) oozei no hito wa party ni kimashita ga, omoshiroi
many LK people TP party to came but interesting
hito wa kimasen deshita.
people TP didn't-come
Many people came to the party, but no interesting people came.
WA AS A COHESION MARKER 27

This sentence represents an explicit contrast between the fact that many
people came to the party and the fact that none of them were interesting.
Kuno has argued that the was that appear here are "contrastive", as
opposed to "thematic;" this dichotomy has been widely accepted, and
numerous attempts have been made to describe the contexts in which wa
will carry the contrastive interpretation. Kuno has tried to to link the possi­
bility of a contrastive interpretation to the informational status of the wa-
marked NP, claiming that wa-marked NPs which are neither anaphoric nor
generic can only be interpreted as contrastive. A somewhat different
approach is taken by Maynard (1980), who stresses the importance of con­
text and proposes that the contrastive effect arises when the NP-wa is
embedded in a context where other alternatives (i.e., the other pole(s) of
the contrastive relation) are identifiable. Virtually all researchers have
argued that contrastive uses are identifiable on the basis of a special "con­
trastive" stress pattern, in which the wa-marked NP receives heavy stress.
These claims are relevant here, of course, because they suggest a
means of accounting for the numerous uses of wa in our data which do not
appear to be thematic at the discourse level. Many of our examples do in
fact seem to conform to standard descriptions of contrastive wa, exhibiting
a semantic contrast between a participant or a situation denoted by nearby
(local) linguistic elements. A characteristic example of this sort of wa usage
appears in (10).
(10) de okaasan wa.. wa ne, ... monosugoku ... sono
and mother TP TP EX very-much that
ikurachan no... itte iru koto wakaru n desu
Ikura SB saying is thing understand NOM COP
tte ne, ... hoka no kodomotachi wa, zenzen
QT EX other LK child-PL TP completely
wakannai.
don't-understand
And the mother says she understands very well what Ikura is say­
ing; the other children don't understand at all.
Here the mother's ability to understand the speech of her child Ikura is
being contrasted with the inability of the others to understand, and both the
mother and the others carry was that can be characterized as "contrastive".
In this section of the paper we will examine the conditions under which
28 PATRICIA M. CLANCY AND PAMELA DOWNING

contrastive was appear in our narratives. Before turning to a discussion of


our results, though, a word is in order on how we distinguished contrastive
from thematic uses of wa.
On the basis of examples like those illustrated in (9) and (10), one
might assume this to be a fairly unproblematic task. In these examples, the
contrastive effect emerges quite forcefully, since both poles of the contras­
tive relation are explicitly represented in a highly circumscribed linguistic
locale (i.e., adjacent clauses). But not all cases are so clearcut, for semantic
contrastivity and the contrastive stress pattern said to be characteristic of
such uses are both matters of degree.
The difficulties involved in evaluating stress are amply illustrated by
the literature on contrastive stress in English6 (Bolinger 1972; Schmerling
1976; Ladd 1978), which has yet to resolve the issue of whether there are
intonation patterns that can be characterized as exclusively contrastive.
Although previous treatments of contrastive wa have been in consensus on
the issue of contrastive stress, this consensus is probably due, not to the
easy identifiability of such a stress pattern, but to the fact that most of these
treatments have been embedded in discussions of how thematic wa is used
in written texts.
Let us take the example in (10) as an illustration of the difficulties
involved. From a semantic point of view, both of the was which appear here
would seem to be equally contrastive, marking as they do the two poles of
the contrastive relation in question. But when we review the tape of the
narrative, we find that while the first wa-marked NP, (okaasan), does
receive heavier stress, the second (hoka-no kodomo-tachi) does not. Are
we to assume, then, that the first wa here is contrastive, the second thema­
tic?
Aside from cases of this sort, there were many in which it was impossi­
ble to decide whether the degree of stress accorded a particular wa-marked
NP was truly exceptional (hence contrastive) or not. Our difficulty here was
undoubtedly due in part to our non-native ears (although we did receive
advice from native speakers) but it is certainly consistent with the current
debate over the phonetic properties of "contrastive" stress. It also recalls
the remarks of Mikami (1963), who was one of the first to discuss the differ­
ence between thematic (fumon) and contrastive (taihi) wa, and who argued
that the distinction between the two, like the distinction between contras­
tive and non-contrastive stress patterns, was not clearcut, but rather a mat­
ter of degree. For these reasons, we have reluctantly eliminated the consid-
WA AS A COHESION MARKER 29

eration of stress patterns from analysis here.


Turning to the issue of semantic contrastivity, the difficulties involved
in distinguishing contrastive uses of wa become apparent when we consider
the fact that even thematic was can be seen as reflecting the existence of a
contrast, between the thematic participant being established as such and
other potential thematic participants. Such usages of wa are not associated
with a contrastive feel, we would suggest, because thematicity is a textual
rather than a semantic property, and because the elements with which these
wa-marked NPs are in contrast, i.e., other potential themes, are not neces­
sarily present in the linguistic context adjacent to the thematic wa usage.
In this section we will ignore those wa uses in our data which mark only
this sort of thematic contrast, concentrating instead on those which mark a
"locally contrastive" relation of one of the sorts we will describe below. We
take this approach, not because we feel that the two sorts of wa uses are
unrelated or incompatible, but because we are interested in singling out for
attention the locally cohesive effect of wa which has typically been neg­
lected in favor of the global cohesive effect inherent in thematic uses. Our
data in fact contain a number of examples in which it is most profitable to
view the wa in question as working in the interests of both local and global
cohesion. This is the case, for instance, when a persistence-destined thema­
tic participant is established by means of a wa-marked NP which is in
explicit contrast to another thematic participant which figures in the
immediately preceding context. To the extent that uses of wa are of this
sort, they can be seen as multiply determined. 7
Even if we exclude from consideration those uses of wa which appear
to be sheerly thematic, and confine ourselves to the 75% of wa uses in our
data in which a local relation of semantic contrast can be discerned, we find
wide variation in the strength of the semantic contrast marked by wa. The
force with which the contrastive relation emerges seems to depend on sev­
eral factors, among them:
1) the explicitness with which the participants in the opposition are
represented.
2) the degree of incompatibility between the activities or states which
are opposed.
3) the extent to which the participants can be seen as belonging to a
group within which an opposition can be perceived.
The contrast will be felt more keenly, of course, if both participants in
30 PATRICIA M. CLANCY AND PAMELA DOWNING

the opposition are explicitly expressed. The effect will also be enhanced if
both participants enjoy roughly equal thematic status; an inherently con­
trastive relation between a hero and a relatively unimportant participant
may in fact be treated as a matter of foreground and background activity,
receiving very different linguistic treatment than it would if both partici­
pants were considered to be on a par and equally present in the foreground.
The presence of a relation of the appropriate sort between the participants
in the opposition emerges most clearly when the speaker chooses parallel
labels for them, denoting one as kodomo 'child' and the other as otona
'adulť, for example, or referring to one as X no hoo (literally, 'X's side')
and the other as Y no hoo 'Y's side'.
The nature of the dimension along which the participants in the oppos­
ition are seen to differ will also be of importance. The clearest cases of
incompatibility are, of course, associated with lexical antonyms or negative-
positive verb pairs of the sort illustrated in (9) and (10), but our data also
contained numerous examples of pairs that could be seen as representing
complementary activities, such as kiku 'ask' and iu 'say', or which involved
identical verbs but different complements or adverbial modifiers, e.g.,
sakki 'before' vs. kondo 'this time', or nashi o ageru 'give pears' vs. booshi
o ageru 'give a hať. Context must also be taken into consideration;
although 'walk by' and 'be surprised' would presumably not be analyzed as
semantically complementary lexical items, pairs of precisely this sort were
frequently involved in the locally contrastive wa uses in our data, because
they did constitute complementary actions in the situations our narrators
were describing, as in the Pear film, when the threesome walk by the pear
man at the end of the story, and he reacts with surprise.
Among the various materials which we presented to our subjects, it
was those which could be described in terms that combined the three fac­
tors just discussed that were the most successful in eliciting descriptions
containing locally contrastive was. Particularly striking were the responses
to several of the cartoon strips with joint heroes, where each character per­
formed actions that could be seen as parallel to or in response to those of
the other. All of our subjects, for example, used was in describing the
actions of the two small boys depicted in the Sand Fight cartoon, where
each destroyed the work of the other. Eight of the ten subjects used wa for
mentions of both boys; two used wa on only one of the two participants.
The description in (11) is typical of those we elicited in response to this car­
toon.
WA AS A COHESION MARKER 31

(11) kore ... hi... hiroshikun ya tarookun ga, ...


this Hi Hiroshi and Taroo SB
shoogakkoo no ichinensei gurai na no kashira.
elementary-school LK first-grader about LK NOM perhaps
...  ... maa gakkoo shoogakkoo no kootei
sort-of well school elementary-school LK playground
mitaina kanji de ne. ... eeto basho wa ne
kind-of feeling COP:GER EX uh location TP EX
eeto sunaba ga atte, hiroshikun to tarookun ga
uh sandbox SB exist:GER Hiroshi and Taroo SB
ne, ... nakayoku asonderu no desu ne. ...de- ... tarookun
EX happily playing NOM COP EX and Taroo
no hoo wa ne, ... nanika ... torakku no ... omocha no
GN side TP EX something truck GN toy GN
torakku no nidai e suna o nosete asondeta no.
truck GN back to sand DO put:GER was-playing NOM
...sorede hiroshikun no hoo wa ne, ... ano-- ...suna no
Then Hiroshi GN side TP EX u h - sand LK
oshiro o tsukutte asonderu no.
castle DO make:GER playing NOM
In this one, Hiroshi and Taroo are maybe about first-graders.
The location seems like the playground of an elementary school.
There's a sandbox, and Hiroshi and Taroo are playing happily.
And Taroo was loading sand onto the back of a toy truck. Then
Hiroshi is making a sand castle.
In this description of the initial frame of the cartoon strip, we can observe
all of the factors mentioned above as being conducive to a locally contras­
tive interpretation of wa. First of all, both characters are introduced
together, by means of the single noun phrase Hiroshikun ya Tarookun ga.
This sort of introduction puts both of the participants on stage, in equal
thematic capacities, and it presents them as a pair. Other details reinforce
the identity of the two as a group — they both appear to be in first grade or
so, they are both in the sandbox, and they are both playing. The narrator
then goes on to describe the actions of each boy, and like all the other nar-
32 PATRICIA M. CLANCY AND PAMELA DOWNING

rators who described this cartoon, she does it in a way which continues to
emphasize the parallels between them and their activities, in this case by
using not only the two was, but also parallel forms of reference
{Tarookun no hoo vs. Hiroshikun no hoo) and identical verbs, i.e., asobu
'play'. Because the speaker has chosen this method of description, neither
of the characters emerges as central at the expense of the other, and
because the text in which the N P - W A mentions occur exhibits all three of the
traits mentioned above, both was lend themselves to very robust contrastive
interpretations. 8
It is of interest to note that many of the uses of wa to describe parallel
activities are preceded by precisely the sort of frame-establishing introduc­
tion shown in (11). In many cases, as in (11), the participants in question
are identified as members of the same semantic category, e.g., first graders,
and are assigned as a group to the same temporal and spatial locale. Scene-
setting descriptions of these sorts are highly effective in predisposing the lis­
tener to assign a locally contrastive interpretation to subsequent wa-m-
ing on characters because they set up a limited pool of participants of co­
equal status, so that when one member of the pool is subsequently singled
out for attention, we are left wondering what has become of the other
unmentioned members. 9
A similar effect can be achieved without devoting a section of the text
to scene-setting simply by evoking, by means of some lexical item, a frame
which by necessity involves the participation of more than one character.
Without explicitly mentioning each individual member, for example, the
narrator could with the mention of a family evoke the existence of a
mother, father, etc., subsequently referring to the activities of each with
wa, to contrastive effect. Similarly, as the example in (12) illustrates, the
use of a certain predicate may entail the existence of several participants.
(12) Yoonashi  sono ko-tachi ni, san-і ni
pears DO those child-PL IO 3-CL(person) IO
hito-tsu zutsu ageru wake desu ne. de sorede,
one-CL(inanimate) each give NOM COP EX and then
ano-- sono ko-tachi wa, ano moratte koo ..
uh those child-PL TP uh receive:GER uh
yorokonde kaeru n desu keredomo,
be-glad:GER return NOM COP but
WA AS A COHESION MARKER 33

(He) gives a pear to each of those three children one by one.


And then those children get them and happily go on back, but
Here the boy referred to elliptically in the first sentence gives some pears to
three other boys. Although the clause representing the original act of giving
contains no wa and is not presented as contrastive, the use of the verb ageru
'give' entails the existence of another participant, i.e., the recipient of the
act of giving. In the following sentence we see an explicit mention of this
recipient containing a locally contrastive use of wa which is made possible,
not by any advance scene-setting, but by the power of the frame associated
with the verb ageru. Just like the advance enumeration of a pool of partici­
pants, the evocation of a multi-participant frame can leave the listener won­
dering about the fate of any participant who subsequently fails to be men­
tioned explicitly.
Among the locally contrastive uses of wa in our data, we can discern
two major patterns of wa distribution, each with a tendency to represent a
particular sort of contrastive relation, and several minor patterns. Their dis­
tributions in each set of narratives are shown in Table 7 and are described
in the sections that follow.

Table 7. Types of Locally Contrastive Wa Uses.


Cartoons Sazaesan Pear Total
Directly contrastive wa:
Parallel Activities/States 92 17 9 108
Action/State-Reaction 50 18 13 81
Indirectly contrastive wa: 11 4 11 26
Total contrastive/Total wa 153/221 39/46 33/42 225/229
(73%) (85%) (79%) (75%)

A. Parallel Activities/States
Almost half of the locally contrastive was in our sample (108 of 225
(48%)) appear in contexts where the speaker is enumerating the parallel
activities or states of two participants, most often of equal thematic status.
The example in (11) above, in which the speaker describes the parallel
activities of Hiroshi and Taroo in the sandbox, is typical of this usage.
As shown in Table 8, virtually all of the uses of this pattern involve wa-
34 PATRICIA M. CLANCY AND PAMELA DOWNING

marking on both of the participants involved; this property tends to distin­


guish parallel activity/state uses from locally contrastive uses of other sorts.

Table 8. Distribution of Wa-Marking on Locally Contrastive Was of the


Parallel Activity/State Type.
One Member Both Members Represented
Represented
(and marked One member Both members
with wa) wa-marked wa-marked*
Cartoon Stories 2 8 42
Sazaesan Stories 4 8
Pear Stories - - 3 3
Totals 2(3%) 15(21%) 53(76%)
*The figures in this column represent the number of parallel activity/
state relations present in which both members are wa-marked, rather than
the number of wa uses which appear in such double-marked constructions.
Thus except for overlapping chains of wα-marking, each relationship tallied
in this column accounts for two of the wa-marked NPs in our sample, while
those in the preceding column account for only one.

In the fifteen cases in which only one participant is wα-marked, there is no


consistency in the way in which the other participant is marked. In the
example shown in (13), for instance, the wa-marked NP represents the sec­
ond of the two participants, the first of which is the pair indicated by futari-
tomo in the preceding sentence.
(13) futa-ri tomo... moo kontoni minna kankan ni
2-CL(person) both EMPH really all furious RS
natte okotchau no ne. ... demo— ...
become:GER get-angry-end-up NOM EX but
wakame ka na ... kanojo wa zenzen okoranai
Wakame QU EX she TP at-all doesn't-get-angry
no ne.
NOM EX
Both of them get really furious.. But, was it Wakame? she
doesn't get angry at all.
WA AS A COHESION MARKER 35

In others of the fifteen examples, however, the other participant is men­


tioned after the wa-marked NP, and this mention sometimes involves ga-
marked NPs or zero anaphora. In some cases, the lack of wa-marking on
the second participant can be traced to independent factors. In (14), for
example, the speaker is describing a situation in which the baby Ikura's
speech elicits two (parallel) responses from different segments of his audi­
ence, i.e., comprehension and incomprehension.
(14) ikurachan ga kite, okaasan і nі yuu no. ...
Ikura SB come:GER mother IO something say NOM
de futsuu no hito ni wa zenzen wakannai
And ordinary LK person for TP at-all don't-understand
n da kedo  ... gyu-gyu itte ru
NOM COP but sort-of goo-goo saying is
kanji na n da kedo... okaasan ga a! soo
feeling COP NOM COP but mother SB Oh! so
na no.
COP NOM
'Ikura comes, and says something to (his) mother. And although
ordinary people wouldn't be able to understand, it seems like
(he)'s saying "goo-goo", but (his) mother (says), ' O h ! Is that
so."
While many speakers describing this particular episode in the Sazaesan film
in fact chose to use the standard double wa-marking pattern, the speaker in
(14) used an (exhaustive listing) ga instead in referring to the mother's
response, emphasizing the fact that it was only she who was able to under­
stand Ikura's speech.
This example illustrates quite clearly why it is impossible to predict
with certainty whether or not wa will be used in describing a certain situa­
tion. While conditions conducive to wa-marking may be present, as they
are in the case of (14), the situation may also satisfy the conditions for the
use of some competing construction (such as exhaustive listing ga), leading
some speakers to use one construction, some the other.
It is also worth noting that when speakers chose to mark with wa only
one of the two participants in parallel activities or states, they always rein­
forced their description with some other sort of marker of the opposition
involved. Often, as shown in both (13) and (14), the two halves of the
36 PATRICIA M. CLANCY AND PAMELA DOWNING

opposition were connected by an adversative conjunction such as demo or


kedo 'but' ; in other cases they involved parallel labels for the two partici­
pants, e.g., kodomo-tachi 'the children' vs. otona-tachi 'the adults', or
antonymic predicates. These devices can be seen as going at least part way
toward performing the cohesive role associated with double wa-marking.
Just as the speaker may dispense with thematic wa when the identity of the
thematic participant is not in doubt, s/he may apparently omit one of the
was in a parallel activity/state construction if the opposition in question is
represented sufficiently by other linguistic elements.
The speaker may be motivated to use the double-wa pattern in order to
emphasize the unexpected co-occurrence of two opposing states or
activities, or it may be used, after the explicit introduction of a pool of par­
ticipants, to enumerate the states/activities of each member of the pool.
The impetus may arise, then, from a concentration on either the states/
activities involved or on the participants themselves. The example in (14)
seems to be a case of the former type, since there has been no attempt, in
the section of text preceding the excerpt presented here, to present the
mother and the other participants as co-members of a single group. The
example in (11), by contrast, would seem to be of the enumerative type.
The double wa-marked parallel activity/state uses are of special
interest because they require that the speaker have the contrastive nature
of the relation in mind at the time s/he utters the first of the contrasting ele­
ments. Thus the speaker who uses this construction presents not only the 
half as being in contrast to A, but also the A half as being in contrast to B.
This is very different from the way in which action-reaction pairs, which
typically mark only the participant in the reaction (B) with wa, are treated.
Before moving on to this other sort of construction, we should note the
existence of a variant on the parallel activity/state pattern in which only one
half of the opposition is mentioned. The use of this pattern is made possible
by the pre-establishment of a participant pool, as with the scene-setting
device described earlier. In such cases, by mentioning with wa the activities
of only one member, the speaker can rely on the cohesive power of wa to
imply that the other members of the group did not engage in the same activ­
ity. This technique was frequently used, for example, in our subjects'
descriptions of the Snow Fight cartoon, where the activities of the girls,
who gave their snowman an umbrella, were contrasted with those of the
boy, who did not. Many narrators mentioned (with wa) only the action of
the girls, thereby setting up the implication that the boy had not acted in
WA AS A COHESION MARKER 37

the same way. In the description shown in (15), for example, mention of
Yuki and Sachi is marked by what appears to be a locally motivated wa of
this sort; the last preceding NP had referred to the girls and there was thus
no need to re-establish their identity with an explicit NP at this point in the
narrative.
(15) yukichan to satchan wa... atama ga yokute, ...
Yuki and Sachi TP head SB good:GER
kasa , ... jibun-tachi no .. yukidaruma ni
umbrella DO self-PL GN snowman to
sashite-oite agemashita. ... ame ga futtekitara, ...
thrust-put-GER gave rain SB begin-to-fall-when
yukichan to, satchan no yukidaruma wa, ... kichinto
Yuki and Sachi GN snowman TP nicely
shiteita kedo, ... ijiwaru o shita tarookun no
was-done but teasing DO did Taroo GN
yukidaruma wa, .. kowar ete-shimaimashita.
snowman TP be-destroyed-end-up:PST
Yuki and Sachi were smart, and stuck an umbrella on their snow­
man. When it began to rain, Yuki and Sachi's snowman was fine
but the snowman of Taroo, who had been nasty, ended up get­
ting ruined.
Uses of this sort can thus be seen as a variant on the parallel activity type
construction, since the context is strong enough to engender a locally con­
trastive (parallel activity) implication even in the absence of an explicit
mention of the  half of the opposition.

B. Action/State Reaction
In the second pattern that occurs frequently in our data, wa-marking
occurs on only the second member of the opposition. This pattern typically
occurs when the speaker is developing a plot sequence containing succes­
sive activities involving different participants. Most commonly, the A
member of the pair denotes an activity on the part of X to which Y then
reacts, or fails to react, in the  part of the pair. The example shown in (16)
is typical.
38 PATRICIA M. CLANCY AND PAMELA DOWNING

(16) ... mottette yatta no. soshitara sono orei ni,


take-GER gave NOM then that thanks for
... ano- ano ... nashi  nusunjatta tte yuu n
uh uh pear DO stole QT say NOM
ja-nai kedo sa, sono otoko-no-ko wa,
-NEG but EX that boy TP
mit-tsu, ano- nashi ... kurete yatta wake.
3-CL(inanimate) uh pear DO give:GER gave NOM
(hei.) brought it (to him j ). Then in thanks, (hej.) didn't say that
(hej.) had stolen them but, the boy. gave (him.) three pears.
Here the participant in the A member of the opposition is not mentioned
explicitly, and the most recent mention of him was marked with ga rather
than wa. It is thus only when we reach the  member of what we now
realize is a contrastive opposition that we encounter the wa (on sono otoko-
no-ko) which marks its existence.
A speaker may choose to link two events in a narrative by the use of
this pattern when Y's action/state in  is seen as caused by, provoked by, or
even merely enabled by X's action in A. In (17), for example, the existence
of Hiroshi's beautiful sand castle provokes Taroo to kick it over, and the
reference to Taroo includes wa, even though the most recent previous men­
tion of a human participant had also been a ίνα-marked reference to Taroo.
(17) tarookun wa, ...to torakku ni, ... suna — .. nokkete
Taroo TP uh truck to sand DO put:GER
asondete, ... sorede ... hiroshikun no,
playing:GER and Hiroshi GN
eeto o-.. oshiro ga sugoku yoku dekita mon da
uh - .. castle SB very well made NOM COP
kara, ... tarookun wa ... sore o ketobashite kowashichau
because Taroo TP that DO kick:GER smash-end-up
no.
NOM
Taroo is playing, loading sand in a truck, and, because Hiroshi's
castle is so nice, Taroo kicks it over.
To the extent that a text relates a coherent chain of events, it may con-
WA AS A COHESION MARKER 39

tain a chain of was of this type, since the actions of each participant can be
seen as being produced in response to those of the preceding participant. In
such cases, the A member of the opposition contains a wa which marks its
status as a response to the event that preceded it, and the  member con­
tains a wa to mark its opposition to A. This often happens when speakers
are recounting conversations, in which each speech turn occurs in response
to the preceding one, as in (18). Here the participant in the response, i.e.,
the teacher, is a minor character in the overall plot, making it difficult to
interpret the wa which it bears as thematic.
(18) ...  ni natte, ... soshitara sensei ga yattekite
fight RS become:GER then teacher SB come:GER
... futa-ri  shikarimashita. ...de
2-CL(person) DO scolded and
hiroshikun wa... "a! tarookun ga yatta n da yo"
Hiroshi TP A! Taroo SB did NOM COP EX
tte itte, ... iitsukeguchi o shita n da kedomo, ...
QT say:GER tattling DO did NOM COP but
sensei wa, "futa-ri tomo warui n desu yo" tte,
teacher TP 2-CL(person) both bad NOM COP EX QT
... de, ... "futa-ri wa nakayoku ... oyama o
and 2-CL(person) TP nicely mountain DO
tsukurinasai" to itte, ... riari isshoni asobasemashita.
make:IMP QT say:GER forced together play-CAU-PST
(they) started fighting, and then a teacher came, and scolded the
two of them. And Hiroshi tattled and said, "Taroo did it", but
the teacher made them play together, saying, "Both of you are
bad; make a mountain together nicely".
As in uses of the parallel activity/state pattern described above, the speaker
may choose to reinforce the bond between the two activities by using
adverbs or causal or adversative conjunctions as the kara 'because' in (18).
Action/state-reactions are, however, typically not preceded by any intro­
ductory passage in which a pool of participants is introduced; the impetus
for the use of wa seems to arise more from a desire to link the series of
events being described than to exhaustively enumerate the activities of the
participants who are involved. The frame which unites the participants is in
40 PATRICIA M. CLANCY AND PAMELA DOWNING

fact usually evoked by the predicate in the A half of the opposition, rather
than by the use of some advance framing device. Thus the predicate iu 'say'
in the A half of example (18) implies the presence of the listener whose
response is reported in the  half.
Although the two major locally contrastive wa patterns we have
described differ in the ways we have noted, they do have one important
property in common: both typically involve a switch in subject from the A
to the  member of the pair. In fact, of the 299 wa uses in our sample, 227
(76%) occur on switched subjects, although the percentages vary somewhat
depending on the narrative in question, as Table 9 illustrates.

Table 9. Proportion of Wa Uses Involving Switch Subjects.


Cartoons Sazaesan Pear Total
Switch Subjects 163/211(77%) 33/42(79%) 31/46(67%) 227/299(76%)
Other 48/211 (23%) 9/42 (21%) 15/46 (23%) 72/299 (24%)

A switch in subject of course constitutes in itself one kind of local con­


trast between one clause and the next. It is clear, however, that the was we
have been considering are not merely switch subject markers. In the first
place, 24% of wa mentions in the data did not involve switch subjects.
Furthermore, not all switch subjects in the data were marked by wa. In the
cartoon stories, for example, only 33% of switched subjects carried wa, the
remainder eliciting NP-ga, NPs without explicit case marking, and even
ellipses. These distributional patterns demonstrate that wa is not simply an
all-occasion switch subject marker.

 Indirect Contrast Marking (Without Subject Switch)


In addition to the major patterns of locally contrastive wa usage out­
lined above, our data also contained a smaller number of examples in which
the NP referring to a participant bore a wa but was not itself one of the
poles of a contrastive opposition. Instead, the participant served merely as
a bearer for the marker of a contrast involving some other element of the
clause, typically the predicate, in which it appeared. Consider the example
in (19).
(19) yukichan wa, ... suberidai... suberidai de
Yuki TP slide slide with
WA AS A COHESION MARKER 41

asondemashita. ... dakedo, ... yukichan wa, ...


was-playing but Yuki TP
subetteru tochuu-de, ... fuusen hanashite-shimaimashita.
sliding while balloon let-go-end-up:PST
Yuki was playing on the slide. But, while Yuki was sliding, (she)
let go of (her) balloon.
In the section of the narrative that precedes the excerpt presented here,
Yuki and Sachi go with their mothers to the park, where they are happily
playing, Sachi on a swing, and Yuki on the slide. As the adversative con­
junction dakedo suggests, however, this idyllic state of affairs is abruptly
terminated when Yuki lets go of her balloon, and becomes downcast, trig­
gering the remaining events in the plot. Thus although asobu 'play' and
hanasu 'let go' do not obviously contrast as antonymous, within this context
the two verbs represent the contrast between a happy and an unhappy state
of affairs, and the boundary between them constitutes the turning point of
the story. Thus, the NP denoting the participant bears the marker of the
contrast between the predicates, although the participant involved (Yuki)
remains constant throughout. This sort of analysis could probably be
extended to all those cases which have been described as instances where a
discourse boundary triggers the need to re-establish a thematic participant
with NP-Wa.
In a second set of examples, the participant-denoting NP bears a wa
which marks a contrast between the state of affairs represented by the
clause and some other state of affairs which is not explicitly mentioned and
which, like the preceding cases, does not necessarily involve another partic­
ipant in contrast with the wα-marked one. Consider example (20).
(20) tabun sono ko wa zenzen sono ... nashi 
probably that child TP at-all that pear DO
to- totte ru otoko-no-hito to kankeinai
pi- picking is man with not-related
to u  ne?
QT think NOM EX
I think that that child probably wasn't related at all to that man
picking pears.
42 PATRICIA M. CLANCY AND PAMELA DOWNING

In this example, the NP-wa is triggered, we would suggest, by the negative


predicate zenzen kankei nai 'not be at all related", although there is no
explicit mention elsewhere in the text of any positive counterpart to this
predicate, such as kankei aru 'be related'. The locally contrastive wa on
sono ko is motivated, then, not by any explicit contrastive relation, but by
an implicit one attendant on the use of the negative verb. As Tannen (1979)
and Labov (1972) have pointed out, it is the narrator's task to describe what
did happen, not the infinite number of things which did not happen. For
this reason, when the narrator does include in the narration a description of
an event which did not happen, or a state which does not exist, it is because
its non-occurrence or non-existence violates some expectation. In this way,
then, the use of a negative predicate is always linked to its positive counter­
part, whether that counterpart receives explicit linguistic expression or not.
Given this perspective, was of the sort shown in (20) can be seen as
marking an opposition of which only one pole happens to be represented
explicitly, but which is implicitly present in its entirety. In cases of this sort,
wa need not appear on a sentence element which is itself participating in the
contrast.

D. Re-Evaluation of "Thematic" Wa Patterns


Having examined the behavior of the locally contrastive was which
dominate the data we are considering, we are now in a position to return to
several of the observations made earlier regarding loss of wa status and the
behavior of wa in the 'paradigmatic' narrative progression. Turning first to
the issue of the loss of wa-status, we can see, in the light of our discussion
of locally contrastive wa, that in many cases there is nothing to lose, since
the scope of the wa extends only as far as the linguistic elements which rep­
resent the opposition whose presence is marked by wa. In fact, although
most cases of wa-loss were multiply determined, the local scope of the dis­
course context eliciting the original wa was sufficient explanation for every
single case in our data. Thus, for example, the wa which marks a character's
reaction to a preceding action or event may disappear in the very next
clause, when the narrator moves on from the reaction. The was marking
each member of a pair engaged in parallel activities are also subject to
immediate loss in the subsequent narration, as in the following:
(21) ... de— ... de hiroshikun wa ookina, ... suna no
and and Hiroshi TP big sand LK
WA AS A COHESION MARKER 43

oshiro  tsukutteiru wake desu ne. ... de Tarookun wa


castle DO making NOM COP EX and Taroo TP
amari...  tsukutteru  ka wakannai. ...
very something making NOM QU don't-know
s soo-yuu-fuu-ni yatteru wake desu. ...
in-that-way doing NOM COP
de ... tarookun ga sono suna no oshiro o shitto shite,
and Taro SB that sand LK castle DO envy do:GER
And Hiroshi is building a big sand castle. And I don't really
know what Taroo is making. They're doing that. And Taroo gets
jealous of that sand castle ...
As for the "paradigmatic" progression of referent introductions, our
narratives contained only ten cases where first NP-ga and then NF-wa were
used to refer to a single referent without an intervening mention of another
participant. In each of these cases the wa which constitutes the intermediate
stage of the progression appears either after the passage of a discourse
boundary of some sort or in one of the sorts of locally contrastive situations
discussed in the preceding section. Each of these was, in other words, can
be seen as the result of independent motivations, rather than as a link in the
paradigmatic progression for introducing thematic participants.
The effect of discourse boundaries can be seen, for instance, in the fol­
lowing example, taken from one of the Theft cartoon narratives.
(22) eeto ne, ... eeto michi de ne, ... hiroshikun ga
uh EX uh road on EX Hiroshi SB
asondeimasu kedo ne, ... de— ... hiroshikun wa ne soko
playing is but EX and Hiroshi TP EX there
made tabun... sanrinsha ni notte kita n desu
up-to probably tricycle on ride:GER came NOM COP
ne.
EX
And Hiroshi is playing in the street, and Hiroshi must have come
there on a tricycle.
Although this might appear to be a classic case of the paradigmatic progres­
sion from NP-ga to NF-wa, there is actually a subtle but clear discourse
44 PATRICIA M. CLANCY AND PAMELA DOWNING

boundary, with a backtracking in the temporal sequence, at the point where


the speaker shifts to NP-wa. The first mention of Hiroshi appears in a
description of his activities in the first frame of the cartoon, i.e., part of the
narration proper, but the clause with NP-wa presents what is actually an
inference from the presence of the nearby tricycle. Thus the ga --> wa shift
here accompanies a transition from the objectively narrated storyline to the
narrator's presentation of her own opinion. This is precisely the sort of dis­
course boundary-crossing situaton that was highly effective in eliciting was
in textual positions other than introductions. Speakers may, however, be
especially sensitive to the potentially disruptive effect of such boundaries at
points of character introduction, using repeated nominal mentions in an
attempt to "anchor" the new particpant in the listener's mind (cf. Clancy
1980a; Hinds 1984).
The remaining cases of ga --> wa progressions for referent introductions
in the data all involve some kind of local contrast. Perhaps the most obvious
examples of contrast to occur in the participant introductions in our data
involved a transition from a first mention of a pair of characters with a con­
joined ga-marked NP to second mentions of each character individually
with wa, as in example (11) above. Such cases of split reference occurred
frequently in the introductions of characters in the cartoons designed to
have two equally important main characters.
A local contrast of a different sort is illustrated in the following exam­
ple, taken from a Pear narrative, where the speaker contrasts the youth of
the bike boy and the size of his bicycle.
(23) ... kondo wa kotchi no hoo kara koo, ...
now TP this-direction LK side from uum
jitensha ni notta otoko-no-ko ga kuru n desu yo
bicycle on rode boy SB come NOM COP EX
ne ... sono jitensha ttsu no ga koo ... ookii n
EX that bicycle QT:say NOM SB uum big NOM
desu yo sugoku. ... ookii tte yuu ka, otona-yoo no
COP EX very big QT say QU adult-use LK
jitensha na noni, sono ... otoko-no-ko wa ne-, ...
bicycle LK although that boy TP EX
juss sai.. Utenai no ka na?
10-year hasn't-gone NOM QU EX
WA AS A COHESION MARKER 45

Next a boy comes by on a bike. That bike is big, really big. Even
though it's big like an adult bike, that boy must be less than ten.
In sum, our data suggest not only that the NP-ga --> NP-wa --> 0 prog­
ression is not the paradigmatic means for introducing thematic participants
into oral narratives, but that thematicity must be coupled with either the
crossing of a discourse boundary or the presence of a local contrast in order
to trigger the middle, wa-marked stage of this progression.
When we turn our attention, though, to referential progressions
initiated by an NP bearing some case particle other than ga, quite a differ­
ent picture emerges. Consider Table 10.

Table 10. Referential Progressions Initiated with Ga vs. Other Case Parti­
cles
Second mention explicit Second mention wa
Ga-initiated chain 61/285(21%) 14/285 (6%)
Other particle- 69/ 82(84%) 23/ 82(28%)
initiated chain

Here we can see that, of the 82 non-subject NPs which are followed without
interference by another mention of the same character, a full 84% of the
second mentions involve explicit NPs. Of these explicit NPs, most (65%)
bear wa (33%), or ga (32%), so that, overall, a full 28% of second men­
tions in non-ga-initiated chains carry wa. This is a very different profile
from what we get with NP-ga-initiated progressions, where only 21% of the
subsequent mentions are explicit and only 6% involve the particle wa. A
typical example involving a non-subject-initiated chain of the sort we are
considering here appears in (24).
(24) sore  kajiri nagara, ojisan no  
them DO munching while man LK front DO
toorisugite tta no. to ojisan wa, ... boke--tto
go-past:GER went NOM and man TP stupidly
shite,
do: GER
Munching on them, they passed by in front of the man. And the
man was bewildered,
The discrepancy in the behavior of NP-ga and NPs bearing the other case
46 PATRICIA M. CLANCY AND PAMELA DOWNING

particles as referential chain initiators is presumably related to the fact that


ga typically marks the subject, while the other case particles do not. Since
the subject slot is the primary noun slot in the clause, usually occupied by a
highly topical participant if there is one, and by important new participants
otherwise, presentation of a participant in this slot presumably makes it the
most viable (default) candidate for the subject slot in the following sen­
tence. In other words, since the subject participant remains constant in such
cases, it is relatively unproblematic for the speaker to use ellipsis on the sec­
ond mention. When a participant is represented by a non-subject NP, on
the other hand, it is overshadowed by the subject participant, and sub­
sequent elliptical mention may lead to confusion. It is for this reason, we
would suggest, that non-subject-initiated referential progressions typically
contain a second explicit mention (most often involving wa or ga) rather
than proceeding directly to ellipsis. This second mention serves to promote
the participant to subject status, whereupon it can safely be ellipted. Thus
the "paradigm" progression is most likely to apply when the participant is
introduced in a non-subject slot. This line of analysis ties in quite nicely
with the fact that the majority of wa uses in our data, whether thematic or
locally contrastive or both, occurred on switch subjects, i.e., references to
participants who by definition had been non-subjects when last mentioned.
Thus two major results of this study which seemed anomalous in the
light of prior research — the lack of wα-marking in introductory progres­
sions and the seemingly unmotivated loss of wa status — can now be seen
to follow naturally from the basic finding that most instances of wa in our
data were locally motivated.

III. CONCLUSIONS
In summary, our analysis of wα-marking in references to human partic­
ipants in this sample of oral narratives suggests that the primary function of
wa is to serve as a local cohesive device, linking textual elements of varying
degrees of contrastivity. The data provide a basis for expanding earlier
analyses of the contrastive function of wa to include the marking of a broad
range of discourse contrasts, such as those between parallel but contrasting
activities or states, actions or states occurring in reaction to the preceding
event or situation, events or states that are unexpected and hence contrast
with the assumptions set up in the preceding narration, and transitions from
one unit of discourse to the next. The referents participating in these narra­
tive contexts need not be especially important or thematic characters in the
WA AS A COHESION MARKER 47

story, but since important characters generally have a larger role in the
plot, there will be a tendency for wa-marking to occur on thematic partici­
pants. Thus in our data the association between thematicity and wa-mark-
ing of a referent is one consequence of the use of wa as a locally motivated
marker of discourse cohesion.
An obvious question about our conclusions on the function of wa in
narrative is why they differ so much from those of other researchers, who
have emphasized the role of wa as a marker of thematic participants. Part
of the answer, we believe, lies in the differences between our data and
those on which earlier conclusions were based. Prior researchers, such as
Hinds & Hinds (1979), Maynard (1980), and Hamada (1983), have all
analyzed written folktales, whereas our analysis is based on oral narratives.
In the written mode, there is much greater leeway for planning and editing,
and so authors are more likely to use wa as a deliberate staging device to
indicate perspective and plot centrality, as Maynard has proposed, and to
exploit the potential of wa-marking for conveying subtle aspects of text
organization, such as discourse boundaries. Written style tends to be more
explicit than spoken discourse, and so introductory referential progressions
with two explicit mentions of the same character, NP-ga and NP-wa, are
more likely to occur, as Hinds & Hinds have found. Since written discourse
can be more carefully planned, the narrator can choose such a slower pace
for the presentation of especially important thematic characters.
Even within our sample of oral narratives, there were important differ­
ences in wa-usage across our three data collections. In certain respects, our
cartoon data were more similar to the written texts which have been
analyzed by others than were our Sazaesan and Pear narratives. In the car­
toon stories, narrators could take more time for planning because they had
the presence of the cartoon strips, which they could pause and examine, as
an excuse for violating normal expectations on pause length in spoken dis­
course. Moreover, the planning task was surely simplified to a large extent
by the pre-selection and organization of the story into a few important
events, sparing narrators the difficult task of deciding what material to
include while narrating. Also, although the narrators were asked to speak
naturally on the cartoon task, some clearly attempted to mimic a storybook
style, as if reading to a young child. This probably increased narrators' use
of the devices common in the written folktales studied by other researchers,
such as frequent, explicit nominal reference with wa-marking of story
characters. The cartoon data included 211 cases of wa, 64.7% of which
48 PATRICIA M. CLANCY AND PAMELA DOWNING

were on references to story characters, whereas the Sazaesan stories yielded


only 114 cases of wa, 36.8% of which marked story characters, although
these were told by the same ten narrators.
In general, then, the cartoon stories exhibited a significantly higher fre­
quency of wa-marking than the much longer Sazaesan and Pear narratives.
This was probably because the cartoon strips incorporated a high frequency
of the factors that motivate the use of both thematic and locally contrastive
wa. For example, since the cartoons were deliberately designed to have
either one or two primary characters interacting with less important charac­
ters, narrators may often have used wa to clarify the main/subordinate
relationship; this view is consistent with the strong tendency for wa-marked
NPs in the cartoon stories to refer to the hero. Our cartoons were thus simi­
lar in structure, and in the use of thematic wa, to the simple folktales inves­
tigated, for example, by Maynard, which are frequently organized around
one or two main characters.
The cartoon stories also had a higher frequency of locally motivated
was, probably because the cartoon strips had been deliberately designed to
provide opportunities for switch reference. Cartoons with two main charac­
ters or joint heroes, for example, reliably elicited the parallel activity A-wa
X, B-wa Y pattern when there was a single frame in which each of the main
characters was performing a different activity. Thus the physical frames in
the cartoon input functioned as domains for the scope of many locally moti­
vated was. Similarly, the pre-selection of certain key moments to represent
a continuous sequence of events in the cartoons highlighted such relations
between events as Action-Reaction, and by juxtaposing them in
immediately successive frames, increased the likelihood that the narrators
would perceive them as related pairs. In the Sazaesan and Pear narratives,
which were told from memory, the tendency to structure a narrative in
terms of such paired relations was apparently much weaker, contributing to
the lower frequency of wa.
The cartoon stories also had a higher frequency of wa- marking at dis­
course boundaries even when there was no switch reference. Again, the
segmentation of the cartoon input into frames created an obvious physical
source for potential discourse boundaries as the narrators told the story
while going from frame to frame. Even frames showing rather minor trans­
itions sometimes elicited wα-marking, but the fact that ellipsis was by far
the most frequent referential choice for the same referent persisting across
frames indicates that speakers were not mechanically producing NP-wa at
WA AS A COHESION MARKER 49

frame boundaries.
It is thus clear that differences in the type of story and process of narra­
tion can lead to differing usage of wa in Japanese narrative. We feel, how­
ever, that the emphasis in prior research on thematic over locally contras­
tive wa is not totally the result of differences in data base. Rather, it seems
likely that many cases of what we have analyzed as locally motivated wa
have not previously been recognized as such, since the prevailing view of
"contrastive wa" has been restricted semantically to a more limited set of
inter-clausal relationships. Once we recognize a broader range of contexts
for contrastive wa, it is apparent that a large number of the cases of wa-
marking on thematic participants could also be analyzed as locally contras­
tive.
In this light, it is interesting to note that Ueno's (this volume) historical
analysis of wa indicates that the original function of wa was contrastive. The
particle was not, at first, used to create and maintain discourse themes, a
function to which it was later extended. Our findings imply that some sense
of contrast still underlies most uses of wa, and that even the discourse-level
contrasts thus indicated tend to be quite local. The use of wa as a global
thematic device, then, appears to be a derived function, both historically
and in the present sample of data. An obvious basis for the extension of wa
to a larger global function in texts is the fact that the characters most fre­
quently participating in the various types of contrast marked by wa are usu­
ally the more important ones; this provides the potential for the re-interpre­
tation of wa as a global theme marker. This process of re-interpretation, it
would seem, has been carried the furthest in written narrative, where wa
serves as an important literary device in the organization of texts.
Given the variation in the usage of wa which we found across our three
collections of oral narratives and the extensive differences between oral and
written narrative, it is apparent that different cognitive and discourse con­
straints elicit different frequencies and, to some extent, functions of wa-
marking. Obviously, much further research will be necessary for a complete
understanding of the functions of wa in Japanese discourse. On the basis of
the present research, however, we would propose that at least for oral nar­
ratives, the primary function of wα-marking on human participants is to
serve as a cohesive device between locally contrasting elements of text.
50 P A T R I C I A M. C L A N C Y A N D P A M E L A D O W N I N G

NOTES
*We would like to thank each other for all the valuable advice and criticisms we gave one
another while working on this paper and to blame any remaining errors on each other. Thanks
also to Shoichi Iwasaki, for sharing his native intuitions with us, and for helping us in many other
ways, and to Knud Lambrecht, for giving us detailed criticisms of an earlier draft.
1) Our transcription conventions essentially follow those described in Chafe (1980) for the
Pear film narratives. Periods are used for sentence-final falling intonation, question marks for
sentence-final rising intonation, commas for clause-final but not sentence-final intonation con­
tours. Questionable transcriptions are enclosed in slashes, and two dashes indicate lengthened
segments. A single dash marks the cut-off point of a false start. Audible silent pauses are indi­
cated by three dots breaks in timing too short to be heard as silent pauses are indicated by two
dots.
2) In measuring persistence, we counted mentions in both main and subordinate clauses, and
a coreferential embedded subject, e.g., in the complement of a verb thinking like u 'think',
was counted as an elliptical reference to the same subject as the matrix verb. Except for refer­
ences to first or second persons, references to a character within a quote were also allowed to
serve as the starting point for coreferential chains extending beyond the quote, and we counted
"into" quotes if the subject of the quote was coreferential with the character in a preceding ellip­
sis chain. On the other hand, subordinate clauses intervening between matrix arguments and the
matrix predicate were not counted as interrupting an ellipsis chain, nor were embedded clauses,
including relative clauses, complements, or the contents of direct quotes. Clauses with tte yuu
kanji de, however, were.
Referential chains that did not begin with an NP were not tabulated at all. Only the final
version of repaired clauses was counted; repeated clauses with a verb were counted like other
clauses. Cases in which subjects appeared in both preverbal and postverbal position in the same
clause were counted as a single nominal reference in preverbal position. References in subject
position to non-human props were counted as ending a chain of persisting (human) coreference,
but generic NPs and self-references by the narrator were ignored, and persistence chains were
counted "through" them.
In determining the valence of predicates with elliptical arguments, we used the following
principles: Quotations which lacked a verb of saying but had a quotative complementizer (i.e.,
to, tte), were counted as having an ellipted subject at the position of the complementizer; quota­
tions without complementizers were not. "Double subject" constructions such as kiga tsuku were
counted as having human experiences. Indirect objects were attributed to the verbs ageru
'give', kiku 'listen', and request verbs such as tazuneru 'request', but the verb iu 'say' was
counted as having an indirect object only if it was explicit. Subjects of verbs such as wakaru 'un­
derstand' were counted as such even when marked by ni.
3) Although cases of NPs without any particle are sometimes analyzed as instances of NP-WA,
there may be subtle differences between the discourse functions of such "bare" NPs and
topicalized NPs (cf. Lee 1984 for discussion of the differences in Korean). Therefore, we have
treated these NPs, like all other nominal co-references, as ending an ellipsis chain, and accord­
ingly, any special status established by wa-marking.
4) For our tabulation of zero switch reference, we again counted as intervening only mentions
in main and subordinate clauses (e.g., those ending in kara 'because', to 'when, if, etc.), but not
relative or complement clauses. The contents of quotes were ignored, as were main or subordi­
nate clause mentions of non-human referents or the narrator.
WA AS A COHESION MARKER 51

5) Hinds & Hinds (1979) treat such wa-marked switch references as cases of episode bound­
aries marked by a change in participant orientation. However, in our data switch reference with
wa-marking frequently occurred within a single temporal/spatial framework, such as a conversa­
tion, and we have distinguished these much more frequent cases from switch reference at
episode boundaries involving major shifts in time, location, or narrator's point of view.
6) There is unfortunately no comparable literature on contrastive stress in Japanese.
7) A number of investigators have in fact claimed that contrastive uses of wa are simply one
subtype of the thematic uses. See, for example, Maynard (1980) and Hamada (1983).
8) It could also be argued that each of these two NP-was constitutes a thematic usage, result­
ing in a very rapid thematic cutting back and forth as the text proceeds. This interpretation may
be correct, but, as we noted above, in the absence of any clearly agreed upon criteria for recog­
nizing sentence-level thematic participants, we will abstain from any judgments as to the thema-
ticity of the two participants and confine ourselves to discussing the locally contrastive reading
which is more available.
9) Chafe (1976) in fact makes the existence of a limited pool of possible candidates for the
contrastive slot one of the criteria for his definition of contrastive constructions.

REFERENCES

Bolinger, Dwight. 1972. "Accent is predictable (If you're a mind-reader)."


Language 48.633-644.
Chafe, Wallace. 1976. "Giveness, contrastiveness, definiteness, subjects,
topics, and point of view." Subject and Topic, ed. by Charles Li. 25-55.
New York: Academic Press.
Chafe, Wallace, ed. 1980. The pear stories: Cognitive, cultural, and linguis­
tic aspects of narrative production. Norwood, NJ: Ablex.
Clancy, Patricia M. 1980a. "Referential choice in English and Japanese nar­
rative discourse." The pear stories, ed. by Wallace Chafe, 127-202. Nor­
wood, NJ: Ablex.
. 1980b. The acquisition of narrative discourse: A study in Japanese.
Ph.D. Dissertation, University of California at Berkeley.
DuBois, John. 1980. "Beyond definiteness: The trace of identity in dis­
course." The pear stories, ed. by Wallace Chafe, 203-274. Norwood, NJ:
Ablex.
Ennulat, J.H. 1978. "Participant categories in Fali stories." Papers on dis­
course, ed. by Joseph E. Grimes, 143-148. Dallas: Summer Institute of
Linguistics.
Givón, T. 1983. "Topic continuity in discourse: An introduction." Topic
52 PATRICIA M. CLANCY AND PAMELA DOWNING

continuity in discourse: Λ quantitative cross-language study, ed. by T.


Givón, (= Typological studies in language, 1), 1-41. Amsterdam: John
Benjamins.
Grimes, Joseph E. 1975. The thread of discourse. The Hague: Mouton.
Grimes, Joseph E. ed. 1978. Papers on discourse. Dallas: Summer Institute
of Linguistics.
Hamada, Morio. 1983. Referential choices in theme, subject, and ellipsis in
written narrative discourse: A case study of Japanese folktales. Master's
thesis, Cornell University.
Hinds, John. 1983. "Topic continuity in Japanese." Topic continuity in dis­
course: Λ quantitative cross-language study, ed. by T. Givón, (=
Typological studies in language, 1), 43-93. Amsterdam: John Benjamins.
. 1984. "Topic maintenance in Japanese narratives and Japanese con­
versational interaction." Discourse Processes 7. 465-482.
Hinds, John and Wako Hinds. 1979. "Participant identification in Japanese
narrative discourse." Explorations in linguistics: Papers in honor of
Kazuko Inoue, ed. by G. Bedell, E. Kobayashi, and M. Muraki, 201-
212. Tokyo: Kenkyusha.
Iwasaki, Shoichi. 1984. Semantic characteristics and discourse functions of
noun phrase-wa: A study of Japanese spoken expository discourse. Mas­
ter's thesis, University of California at Los Angeles.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge,
MA: MIT Press.
. 1976. "Subject, theme and the speaker's empathy — A reexamination
of relativization phenomena." Subject and topic, ed. by C.N. Li, 417-
444. New York: Academic Press.
Labov, William. 1972. "The transformation of experience in narrative syn­
tax." Language in the inner city, ed. by William Labov, 354-396.
Philadelphia: University of Pennsylvania Press.
Ladd, D. Robert. 1978. The structure of intonational meaning.
Bloomington: Indiana University Press.
Lee, Hyo-Sang. 1984. Discourse presupposition and discourse function of
the topic marker NIN in Korean. Master's thesis, University of Califor­
nia at Los Angeles.
Li, Charles, and Sandra A. Thompson. 1976. "Subject and topic: A new
typology of language." Subject and topic, ed. by Charles Li, New York:
Academic Press.
WA AS A COHESION MARKER 53

Maynard, Senko K. 1980. Discourse functions of the Japanese theme


marker -wa. Ph.D. dissertation, Northwestern University.
McGloin, Naomi. This volume. "The role of wa in negation."
Mikami, Akira. 1963. Nippongo no ronri. Tokyo: Kurosio.
Perrin, Mona. 1978. "Who's who in Mambila folk stories." Papers on dis­
course, ed. by Joseph E. Grimes, 105-118. Dallas: Summer Institute of
Linguistics.
Schmerling, Susan F. 1976. Aspects of English sentence stress. Austin: Uni­
versity of Texas Press.
Tannen, Deborah. 1979. "What's in a frame? Surface evidence for underly­
ing expectations." New directions in discourse processing, ed. by Roy O.
Freedle, 137-181. Norwood, NJ: Ablex.
Ueno, Noriko Fujii. This volume. "Functions of the theme marker wa from
synchronic and diachronic perspectives."

APPENDIX A
Bike Accident Cartoon
54 PATRICIA M. CLANCY AND PAMELA DOWNING

Sand Fight Cartoon


2
WA AS A COHESION MARKER 55

APPENDIX 

Pear Film:
The film opens with a man picking pears in a tree. While he is thus
occupied, a man passes by leading a goat. Next comes a boy on a bicycle,
who rides off with a basket of the pear man's pears. On the road he passes
a girl on a bike; distracted by her, he bumps into a rock and falls from his
bike with the pears. At this point three other boys appear on the scene,
help the bike boy put his pears back, and send him on his way. As this
threesome leaves the scene, they discover the bike boy's hat on the road
where it fell during his accident, and return it to him. In exchange, the bike
boy gives them three pears. The threesome then continue down the road,
one playing with a paddleball, and pass by the pear man. The pear man,
who has in the meantime discovered his loss, watches in puzzlement as the
threesome pass by eating his pears.

Sazaesan Videotape:
The mother of a baby named Ikura becomes ill, and he is left at
Sazaesan's house for babysitting. He empties the shoe rack at the entrance
and ransacks the desk drawers of the school-age siblings of Sazaesan,
Wakame and Katsuo, who demand that he be punished. But the adults of
household — Sazaesan and her mother and father — refuse, pointing out
that Ikura is only a baby. The mischief and the argument between Wakame
and Katsuo and the adults continues until Ikura's father comes to take him
home. The next day Ikura's father arrives alone, and apologizes for the
56 PATRICIA M. CLANCY AND PAMELA DOWNING

baby's mischief. All are amazed, since none of them had complained and
since Ikura, they believe, is too young to talk. In the last scene, however,
Ikura's mother brings him for a visit, and explains that she, as his mother,
can understand his speech, and reveals that he had told her that he had
actually been attempting, with all his apparent mischief, to find medicine
for her.

Cartoon Strips'.
After the narrator was shown a card depicting the four children who
would figure in the stories and was told their names, the cartoon strips were
shown in the fixed order given below.
Theft: Hiroshi is engrossed in playing with his top. He fails to notice
Taroo come and ride off on his bicycle until the last frame, where he stands
amazed.
Sand Fight: Hiroshi and Taroo play happily in a sandbox until Taroo
starts a fight by kicking over Hiroshi's sand castle; a teacher then comes and
has them make up.
Exchange: Yuki and Sachi play in a park; when Yuki loses her bal­
loon, Sachi gives her hers, and in return Yuki gratefully gives Sachi her hat.
Bike Accident: Taroo rides his bicycle past a pond where Yuki and
Sachi are feeding the ducks. He is distracted; his bike hits a rock, and he
falls in the pond.
Snow Fight: Taroo makes a snowman and then throws snowballs at
Yuki and Sachi, who have also made a snowman and given it an umbrella
to protect it against the impending rain. Yuki and Sachi flee indoors under
the barrage of snowballs; when it rains, Taroo's snowman, but not theirs,
melts.
Onigiri: Yuki makes onigiri and then gives them to her friends in the
park, who find them unappetizing.
Dress-up: Yuki and Sachi are left alone by their mothers and dress up
in adults' clothes. When their mothers return, they laugh at the girls'
strange appearance.
THEMATIZATION AS A STAGING DEVICE
IN THE JAPANESE NARRATIVE*

SENKO K. MAYNARD
Rutgers University

It is often said that every art form expresses the creator's point of view,
or what Uspensky refers to as the "viewing position" (Uspensky, 1973:2).
Each writer must take a certain viewing position or positions which deter­
mine what he/she sees and what he/she represents in his/her writing. The
particular perspective chosen by the narrator is an important aspect of story
telling, for it reveals the narrator's personal commitment to the mode of
narrativity.
One powerful device available for expressing the narrator's perspective
in Japanese is thematization and/or the avoidance of it, i.e., non-thematiza-
tion. The Japanese language provides an overt morphological device, wa,
to mark what is widely called the "theme" of the sentence. It is true that the
concept of "theme" has been defined differently by a number of linguists
[see, for instance, Mathesius (Firbas, 1964), Firbas (1964), Halliday (1967),
Hinds (1973), Chafe (1976), Kuno (1976) and Iwasaki (1981)]. Regardless
of the differences in definitions, the implication is clear in all previous
studies that thematization signals crucial information for narrative interpre­
tation.
Consider the fact that thematization is a device for realizing what
Chafe (1976) calls "packaging." Packaging, Chafe states, has to do "primar­
ily with how the message is sent and only secondarily with the message
itself" (1976:28). Chafe points out six packaging phenomena relevant to
nouns within a sentence. These include:
1. given or new
2. focus or contrast
3. definite or indefinite
4. subject of the sentence
58 SENKO K. MAYNARD

5. topic of the sentence


6. speaker's point of view and the speaker's empathy
Although Chafe is primarily concerned with the "packaging" of nouns,
thematization (including non-thematization) on the discourse level serves
to "package" the overall thematic effect of the narrative. In fact, the dis­
course functions of the Japanese theme-marker wa used in narrative are
closely associated with all the aspects of packaging which Chafe proposes.
Closely related to the issue of narrative packaging is the interrelation­
ship between thematization and primary narrative elements. A functional
examination of the linguistic devices in a narrative reveals that thematized
elements such as temporal and spatial framework and participants of the
events represent integral parts of the narrative structure (see Maynard,
1982b). By investigating wa usage as well as wa avoidance in identifying
participants of the narrative, we begin to understand the viewing position
the narrator takes toward the narrative content. Naturally, the thematiza­
tion strategy is not the only device for expressing a viewing position, and
other phenomena must be investigated in order to account fully for the
overall effect of the narrative.
In regard to the functions of wa, the generally accepted observation is
made by a number of linguists [Mikami (1963), Kuno (1972, 1973), Ono
(1973), Chafe (1976) and Inoue (1980)]. Their studies have argued that the
use of the Japanese particles wa and ga depends heavily on the given/new
distinction. Specifically, they have concluded that in the subject position of
matrix sentences, wa marks given information and ga marks new informa­
tion. 1 They also note that in subordinate clauses, ga is used for marking
given information and that if wa appears, the only interpretation is that wa
is contrastive [for example, Kuno (1973)]. The previous accounts of wa,
however, lack a "functional" explanation. That is to say, they fail to address
the question: what functional roles do wa and ga play in the realization of
narrative organization, in the manifestation of the narrator's point of view
and for "packaging" the narrative information? 2
In this study we attempt to answer this question based on an examina­
tion of data drawn from two modern narrative collections: Miyazawa's Kaze
no Matasaburoo and Tsubota's Nihon Mukashibanashishuu. Also, toward
the end of this paper, a report of a preliminary experiment using a para­
graph completion test is presented.
First, let us examine data sets (I) and (II) and contrast participant iden­
tification in each.
THEMATIZATION AS A STAGING DEVICE 59

1. mukashi mukashi aru tokoro ni ojiisan to


"once upon a time" one place LC old-man and
obaasan to ga sundeorimashita.
old-woman and SB lived
Once upon a time there lived an old man and an old woman.
2. tokoroga, natsu no aru hi no koto deshita.
now summer LK one day LK fact COP:PST
Now, it happened one summer day.
3. ojiisan wa yama e shibakari ni
old-man TP mountain to firewood-collecting for
dekakemashita.
went-out
The old man went to the mountain to collect firewood.
4. "itterasshaai."
"see you later"
5. obaasan wa ojiisan o okuridasu to,
old-woman TP old-man DO send-off when
The old woman saw the old man off
6. "dore, dore, watashi wa kawa e senta ni
well well I TP river to washing for
ikimashoo, " to,
go-will QT
and (said) "Well, I'll go to the river to get some washing
done,"
7. tarai  kakaete kawa e sentaku ni
washing-tub DO carry:GER river to washing for
dekakemashita.
went-out
and went out to the river carrying a washing tub.
8. "zabu, zabu, zabu, zabu."
(scrub, scrub, scrub)
Scrub, scrub, scrub.
9. obaasan wa seidashite sentaku  shimashita.
old-woman TP work-hard:GER washing DO did
The old woman worked hard washing clothes.
SENKO K. MAYNARD

10. sukoshisuruto, kawakami kara ukishizumishite nagarete


a-while-later upper-river from float:GER drift:GER
kuru mono ga arimashita.
come thing SB there-was
After a while, something came floating down the stream.
(Tsubota 1975:24)
1. mukashi mukashi aru tokoro ni ojiisan to
"once upon a time" one place LC old man and
obaasan to ga orimashita.
old-woman and SB were
Once upon a time there lived an old man and an old woman.
2. aru hi, obaasan ga kawa e sentaku ni ikimashita.
one day old-woman SB river to washing for went
One day the woman went to the river to do the washing.
3. kawakami kara hako ga futatsu nagarete kimashita.
upper-river from box SB two drift:GER came
From upstream, two boxes approached floating down the
stream
4. puka, puka, puka, puka,
(bobbing, bobbing, bobbing)
They came bobbing down the river.
5. kore o miru to obaasan ga yobimashita.
this DO see when old-woman SB yelled-out
Seeing this, the old woman called out,
6. "mi no aru hako wa kotchi koi, mi no
content SB exist box TP here come content SB
nai hako wa atehi ike."
doesn't-exist box TP there go
"Hey, the box filled with things, come this way! Empty
boxes, go away from me!"
7. mi no aru hako ga otte kimashita.
content SB exist box SB approach:GER came
The box with content approached her.
8. sokode, sore  hirotte uchi e kaerimashita.
so that DO pick-up:GER home to returned
So the old woman picked it up and returned home.
THEMATIZATION AS A STAGING DEVICE 61

9. ban ni ojiisan to futaride akete mitara,


night at old-man and two-of-them open-try-when
naka kara uri ga dete kimashita.
inside from cucumber SB come-out:GER came
That evening when the old woman opened the box with the
old man, a cucumber came out of the box.
(Tsubota 1975:18)
Both (I) and (II) are extracted from the initial discourse segments of
two similar narratives by Tsubota. In both stories, an old man and an old
woman are introduced with ga in (I.1) and (II. 1), since they are introduced
to the narrative world for the first time. In both cases, the story develops
with an episode about the old woman. Note however, that in (1.5) and (1.9),
the old woman is marked by wa, while in (II.2) and (II.5) the old woman
continues to be marked by -ga.
If we identify discourse functions of wa/ga merely by equating them
with given/new, we fail to account for the phenomena observed in (II.2)
and (II.5). Why is it the case that a participant who has been introduced
earlier in discourse and therefore is assumed to comprise the so-called
"given" information continues to be marked by ga as in (II.2) and (II.5)?
We propose that the wa/ga marking strategy, although often coinciding
with the given/new distinction, may be characterized as a manipulation
device made available to the narrator for what we call "staging." "Staging,"
a term first introduced in Maynard (1980) refers to the narrator's manipula­
tion of thematization (including non-thematization) when narrating a story.
Through this "staging" strategy, a narrator accomplishes an organization of
narrative information in accordance with his/her perspective. It is as if the
narrator places participants at different spots on the narrative stage for dif­
ferent durations of time in the consciousness of the narrator and his/her
intended audience, the reader. More specifically, the thematized partici­
pants are expected to remain on stage for a longer period of time and this
serves to provide points of reference for the development of the thematic
flow. What the narrator wishes to accomplish through "staging" is to dis­
criminate the thematized participants from non-thematized ones in such a
way that thematized participants remain activated, evoked, and stored in
the reader's consciousness. In this way these participants serve to provide a
flow of thought to which new and unexpected information may be integ­
rated along the way.
Another aspect of the thematization function involves narrative point
of view. In the sense proposed by Kuno and Kaburaki (1977), the reader
62 SENKO K. MAYNARD

more readily has empathy with a thematized participant than with a non-
thematized one. As will be exemplified later, NPga is often used to describe
the actions as observed from the point of view of the thematized partici­
pant.
For new, unexpected and subordinate information, the avoidance of
wa marking, i.e., the strategy of non-thematizing, is practiced. Most fre­
quently such participants continue to be marked by ga although they repre­
sent "given" information. It is important to note here that ga is attached to
indicate new information not in the sense that it cannot have been previ­
ously mentioned, although it is often the case that it has not been, but in the
sense that the speaker presents it as not being currently activated in the
reader's consciousness. In short, in the case of non-thematized participants
a direct reference point for its integration to the thematic line is missing. In
this sense, non-thematized participants provide subordinate and secondary
information within the structure of the narrative. These participants do not
stay on the stage for a long time; their status is characterized by frequent
appearance, disappearance and/or reappearance at various phases of the
plot development. Non-thematized participants often become the focus of
attention when their action and description are introduced as if constituting
new information from the point of view of thematized participants. Because
of this dramatic impact, the non-thematized participant's action is described
vividly and its movement is often more likely to receive closer attention.
Closely related to this issue of the manner of description is the gen-
shoobun (sentence of immediate description)/handanbun (sentence of judg­
ment) distinction developed by Mio which is discussed in Nagano (1972).3
According to Nagano, Mio points out that genshoobun is syntactically
characterized by NP + ga + VP, where VP includes present progressive
forms such as -teiru, -deiru, -teru, -deru or the past tense -ta, -da forms, and
defines genshoobun as the following. Genshoobuns are sentences that rep­
resent phenomena as they are. These sentences that express phenomena
perceived and reflected emotionally are arrived at without the process of
judgment. There is no gap between the phenomenon and its descriptive
expression. Since there is no subjective view to intrude between the phe­
nomenon and the expression, there is no responsibility on the part of the
user in regard to their contents. 4 In fact, what seems to happen in the narra­
tive is that when ga is used, even when it marks the so-called given informa­
tion, such expression gives the impression of urgency and immediacy that
enjoys the focus of attention.
THEMATIZATION AS A STAGING DEVICE 63

On the other hand, according to Nagano's discussion of Mio, when wa


is used, it often comprises handanbun (sentence of judgment) which is syn­
tactically marked by NP wa NP -da. Handanbun is defined as "sentences
that express logical propositions such as A equals B." 5 While Mio's handan­
bun is more limited (note that only copula da can occur in his handanbun),
it is reasonable to assume that some of the wa sentences reflect the
speaker's judgment and therefore less immediacy of expression is con­
veyed. As a result in its broadest sense we can observe the tendency for ga
sentences to describe action more vividly than wa sentences.
Besides the different contributions thematized NPs and non-
thematized NPs make toward narrative plot development, the staging strat­
egy can serve to manipulate the mode of participant description. That is,
the description of thematized participants are state-oriented in that the
event that happens is viewed as a cumulative chain of changing states
associated with those participants. Here, the world is viewed such that
thematized participants are going through a change of "state." This is
because thematized participants serve as constant reference points or pegs,
and events that happen in the narrative tend to be integrated as changes
happening to them, rather than as independent incidents. On the other
hand, the actions of non-thematized participants are described not as
change of states, but with vivid imagery. As mentioned earlier, in this
case, the participants themselves are focused as new, or as if new, and the
entire movement is captured as an action. The manner of description here
may be described as "action"-oriented as opposed to "state"-oriented in the
case of thematized participants.
Before turning to data analysis and a discussion of "staging" at work,
we need to clarify the terminology "staging." Among all other conceptual
terms and phrases similar to "staging" — expressions such as Kuroda's
"point of view" in literature (Kuroda 1973), Fillmore's "perspective"
(Fillmore 1977) and Hopper's "foregrounding" (Hopper 1979), only
Grimes (1975) uses the identical term "staging." 6 Note, however, that
Grimes' "staging" refers to the function of "theme" which is "to single out
a particular individual or an actor (Grimes 1975:327)." Whereas Grimes'
"staging" refers solely to the function of "theme," we capture staging as an
embracing term that refers to the overall strategy of identifying participants
in a discriminatory manner including both thematization and non-themati-
zation. Note also that as mentioned earlier, non-thematized participants are
often "singled out" as a focus of attention. Thematized participants then
64 SENKO K. MAYNARD

are not so much "singled out" but rather remain in the total conceptual
framework to maintain a flow of thought in the form of evoked, activated
and stored information.
More recently, Brown and Yule (1983) use the term "staging" in an
even broader sense. For them, "staging" includes various rhetorical devices
like lexical selection, rhyme, alliteration, repetition, use of metaphor, mar­
kers of emphasis, etc. In fact, Brown and Yule use "staging" "not as a tech­
nical term, but as a general metaphor to cover the exploitation of these var­
ied phenomena in discourse" (1983:134). In this paper, however, we con­
centrate on the thematization strategy realized by the use and the non-use
of the particle wa in Japanese.
Let us now turn to an examination of data sets (I) and (II) again, this
time in terms of "staging" as proposed here. In data set (I), the old woman,
because it is marked by wa in (1.5) (theme creation), and because it con­
tinues to be marked by wa in (1.9) (theme maintenance), is assumed to
remain stored in the registry of the reader's consciousness. Here the story is
told as if the old woman provided the constant line of thought into which
new information may merge. In (1.9) for example, the old woman functions
to reassure the reader that this discourse segment is still basically a descrip­
tion of the old woman and now, new information which is seidashite sentaku
shimashita "worked hard washing clothes" is added. It is as if the reader
understands the development of the story through successive stative
oriented descriptions of the old woman rather than through the results of
any specific action taking place on the stage. Action, of course, does occur,
but it is packaged in such a manner that the entire scene, and not each act
itself, comes into focus. The narrator prefers this type of plot development
in data set (I), while in data set (II), a non-thematization strategy is
selected. In (II), we find cases where a participant which presumably com­
prises "given" imformation, i.e., obaasan "the old woman," is non-
thematized in a matrix clause as in (II.2) and (II.5). Here, the old woman
as well as her action is in focus. Action is described as a vivid incident which
enjoys the reader's close attention. Sentences (II.2) and (II.5) do not
describe the old woman's action as a change of state as was the case in data
set (I). Rather, unlike the cases of (I.5) and (I.9), the old woman is intro­
duced as if it is new information and thus becomes the focus of temporary
attention. In terms of thematic development then, the old woman in data
set (II) constitutes a secondary subordinate element that must be incorpo­
rated into the thematic line. While the use of the old woman in data sets
(I.5) and (I.9) encourages us to keep her in our consciousness, the use of the
THEMATIZATION AS A STAGING DEVICE 65

old woman in (II.2) and (II.5) encourages the reader to focus on that infor­
mation temporarily. 7 In data set (2), at least in (II.2) and (II.5), obaasan ga
'the old woman' does not serve as strong a function as in (I.5) and (I.9) in
establishing the thematic flow of the narrative. In this sense we can state
that it is as if the ga sentences are subordinate sentences in the total scheme
of narrative organization.
The reader is reminded at this point that it is significant that the ga
which obligatorily replaces noncontrastive wa in subordinate clauses
behaves similarly on the intrasentential level. We can conclude then that on
both intrasentential and intersentential levels, ga marks either new infor­
mation or non-thematized subordinate information.
We have so far examined the contrast in participant identification in
data sets (I) and (II). Now we turn to data sets (III) and (IV) in order to
examine how the staging strategy may be used in a narrative segment. Data
set (III) is taken from Miyazawa's Kaze no Matasaburoo, while data set
(IV) consists of the identical discourse segment but with the three partici­
pants marked with particles different from the original. Namely, in sen­
tences (1) through (19), Kasuke is marked with ga, and in sentences (20)
through (44), Ichiro's brother is marked with wa consistently.8
(III/IV) 1. futo kasuke (III) wa me o hirakimashita.
suddenly Kasuke TP eyes DO opened
(iv) ga
SB
Suddenly, Kasuke woke up and opened his eyes.
2. haiiro no kiri ga hay aku hay aku tondeimasu.
grey LK mist SB fast fast flying
A grey mist was moving in fast.
3. soshite uma ga sugu me no  ni
and horse SB right eye LK in-front LC
nossonto tatteita no desu.
stolidly was-standing NOM COP
And a horse was standing stolidly in front of his eyes.
4. sono me wa kasuke  osorete
those eyes TP Kasuke DO be-afraid-of:GER
yokonohoo  muiteimashita.
sideways DO was-looking
Its eyes were averted in fear of Kasuke.
SENKO K. MAYNARD

5. kasuke (III) wa haneagatte uma no nafuda 


Kasuke TP leap:GER horse GN name-tag DO
(iv) ga
SB
osaemashita.
held
Kasuka leaped up and held the name tag of the horse.
6. sono ushiro kara saburoo ga marude iro no
that behind from Saburo SB totally color SB
nakunatta kuchibiru  musunde kotchi e
lost lips DO close:GER this-way to
dete kimashita.
appear:GER came
From behind the horse, Saburo appeared with his pale lips
tightly closed.
7. kasuke (in) wa buruburu furuemashita.
Kasuke TP (tremblingly) shivered
(iv) ga
SB
Kasuke was shaking.
8. "ooi," kiri no naka kara ichiroo no niisan no
wahoo mist LK inside from Ichiro GN brother SB
 ga shimashita.
voice SB did
"Yahoo!" Ichiro's brother's voice was heard in the mist.
9. kaminari mo gorogoro natteimasu.
thunder also (roaring) roaring
Thunder was resounding making huge noises.
10. "ooi, kasuke, iru ga. kasuke," ichiroo  
Hey Kasuke be QU Kasuke Ichiro GN voice
 shimashita.
also did
"Hello! Kasuke, are you there, Kasuke?" Ichiro's voice
was also heard.
THEMATIZATION AS A STAGING DEVICE 67

11. kasuke (iii) wa yorokonde tobiagarimashita.


Kasuke TP be-pleased:GER jumped-up
(iv) ga
SB
Kasuke jumped with joy.
12. "ooi, iru, iru, ichiroo. ooi."
yahoo I'm here Ichiro yahoo
"Yahoo! Here I am, here I am, Ichiro!"
13. ichiroo no niisan to ichiroo ga totsuzen
Ichiro GN brother and Ichiro SB suddenly
me no  ni tachimashita.
eyes LK in-front LC stood
Ichiro's brother and Ichiro suddenly stood before Kasuke.
14. kasuke (iiі) wa niwakani naki-dashimashita.
Kasuke TP immediately cry-began
(iv) ga
SB
Kasuke began to cry immediately.
15. "sagashita zo. abunagatta zo. sukkari nureda
looked-around EX was-dangerous EX totally got-wet
na. doo."
EX here
"We were looking for you. It was quite dangerous. You
are all wet, aren't you? Here."
16. ichiroo no niisan wa naretatetsuki de uma no
Ichiro GN brother TP usual-mannerwith horse GN
kubi  daite mottekita kutsuwa  subayaku
neck DO hold:GER brought bit DO quickly
uma no kuchi ni hamemashita.
horse GN mouth LC put-on
Ichiro's brother held on to the horse's neck in his usual
confident manner and swiftly put a bit he brought with
him in the horse's mouth.
17. "saa obesa."
let's go
"Let's go."
SENKO K. MAYNARD

18. "matasaburo bikkurishita bea." ichiroo ga


Matasaburo was-surprised QU Ichiroo SB
saburoo ni iimashita.
Saburoo IO said
"Matasaburo, you must have been surprised," Ichiro said
to Saburo.
19. saburoo wa damatte  appari kitto kuchi 
Saburo TP be-silent:GER also tightly mouth DO
musunde unazukimashita.
close:GER nodded
Saburo nodded in silence, still with his mouth closed
tightly.
20. minna wa ichiroo no niisan ni tsuite
everyone TP Ichiroo GN brother DO follow:GER
yurui keisha  futatsu hodo noboriori
gentle slope DO two about climbing-up-and-down
s himas hita.
did
Everyone followed Ichiro's brother over two gradually
sloped hills.
21. sorekara, kuroi ookina michi ni tsuite
then black wide road DO follow:GER
shibaraku arukimashita.
for-a-while walked
They then reached a wide black road and continued to
walk a little further.
22. inabikari ga nido bakari kasukani shiroku
lightening SB twice about slightly white
hiramekimashita.
flashed
Twice lightning flashed slightly.
23. kusa  aku nioi ga shite kiri  
grass DO burn smell SB do:GER mist LK inside
 kemuri ga hotto nagareteimasu.
LC smoke SB (floating) streaming
There was the smell of burning grass and the smoke
floated in the mist.
THEMATIZATION AS A STAGING DEVICE 69

24. Ichiro  no niian (iii) ga sakebimashita.


Ichiro GN brother SB called-out
(iv) wa
TP
Ichiro's brother called out.
25. "ojiisan. ida, ida. minna ida."
grandfather was was everyone was
"Grandfather, they are here! Everyone is here!"
26. ojiisan wa kiri no naka ni tatteite,
grandfather TP mist LK inside LC standing:GER
The grandfather who was standing in the mist said,
27. "aa, shinpaishita. shinpaishita. aa yogatta. oo,
oh, was-worried oh be-relieved oh
kasuke, samu gabe sa, saa, haire." to
Kasuke be-cold probably EX here come-in:IMP QT
iimashita.
said
' O h , was I ever worried! I'm so relieved. Oh, Kasukeyou
must be cold. Come on in.
28. kasuke wa ichiroo to onaji yooni yahari
Kasuke TP Ichiro as same manner all-the-same
kono ojiisan no mago no yoo
this grandfather GN grandchild LK appearance
deshita.
was
Kasuke, like Ichiro, seemed to be a grandson of this old
man.
29. hanbun-ni yaketa ookina kuri no ki no
half burned large chestnut LK tree LK
nemoto ni kusa de tsukutta chiisana kakoi ga
root at grass with made small fence SB
atte, choro choro akai hi ga moeteimashita.
be:GER (flickeringly) red fire SB was-burning
At the foot of a half burned chestnut tree, there was a
small fence made of grass and inside, a red flame flickered.
SENKO K. MAYNARD

30. ichiroo no nasan wa uma o naranoki ni


Ichiro GN brother TP horse DO elm-tree to
tsunagimashita.
tied
Ichiro's brother tied the horse to the elm tree.
31. uma mo hihin to naiteimasu.
horse also (neigh) QT is neighing
The horse neighed.
32. " ( oo, muzo ya na. na, nanbo naida gana.
well poor-children COP EX well a-little cried QU
sono waro wa kanayamabori no waro da na.
that boy TP Kanayamabori GN boy COP EX
saa, saa, minna dango taber  taber. na,
here everyone dumpling eat:IMP eat:IMP here
ima kotchi  agu gara na. zentai doko
now this one DO roast so EX in-the-world where
itteta da. "
have-been COP
' O h well, poor little children! You must have cried a bit.
That boy must be from Kanayamabori. Here, everyone,
eat the dumplings. Have some. I'll roast these now.
Where in the world have you been?"
33. u sasaone no orikuchi da" to ichiroo no niisan
Sasaone LK exit COP QT Ichiroo GN brother
(iii) ga kotaemashita.
SB answered
(iv) wa
TP
"At the exit of Sasaone range," answered Ichiro's brother.
34. "abunaigatta. abunaigatta. mukoosa oritara
was-dangerous over-there went-down-if
uma mo hito mo sorekkiri datta zo. saa,
horse also person also the-end COP:PST EX here
kasuke, dango tabero  waro mo tabero.
Kasuke dumpling eat:IMP this boy also eat
saa, saa, koizu mo tabero."
here here this also eat
THEMATIZATION AS A STAGING DEVICE 71

' O h , it was dangerous! Quite dangerous! If you went


down toward the other side of the range, no person or
horse would be able to return. Here, Kasuke, eat your
dumplings. Here, this boy, you have some, too. Here, eat
some more...:
35. "ojiisan, uma oide kuru ga," to ichiroo
grandfather horse leave:GER come EX QT Ichiro
no niisan (III) ga iimashita.
GN brother SB answered
(iv) wa
TP
"Grandfather, I will take the horse and leave it there."
said Ichiro's brother.
36. "un, un. makifu kuru to mata yagamas hi gara
yes farmer come if again be-troublesome so
na. shitadomo, mosukoshi made, mata hareru. aa,
EX but a-little wait again clear-up ah,
shinpaishita. ore mo torakoyama no shita
was-worried I also Mt. Torakoyama LK under
made itte-mite kita. haa, mantsu yokatta ame
till go-try:GER came well anyway was-good rain
mo hareru. "
also clear-up
"Yes, if the farmers come, maybe there'll be more trou­
ble. But let's wait a while. It will clear up soon. Oh, I was
so worried. I did go to the foot of Mt. Torako. Well, but
now everything is all right. The rain will stop soon."
37. "kesa honto-ni tenki yogatta і ."
this-morning really weather was-good though EX
"It was such good weather this morning."
38. "un, mata yogu naru sa. a, ame motte
yeah again be-good become EX oh, rain leak:GER
kıtana. "
came
"Yeah, it will be better soon. Oh, the roof is leaking."
72 SENKO K. MAYNARD

39. ichiroo no näsan () ga dete ikimashita.


Ichiro GN brother SB go-out:GER went
(iv) wa
TP
Ichiro's brother went out.
40. tenjoo ga gasagasa iimasu.
ceiling SB (rustling) make-noise
The ceiling is making a rustling noise.
41. ojiisan ga warai nagara sore o miagemashita.
grandfather SB smile while that DO looked-up
Grandfather looked up at it smiling.
42. niisan (III) ga mata haitte kimashita.
brother SB again enter:GER came
(iv) wa
TP
The brother came in again.
43. "ojiisan, agarugu natta. amea hareda."
grandfather light became rain:TP cleared-up
"Grandfather, it's light again. The rain has stopped."
44. "un, un soo ga. saa, minna yokku hi ni
yeah so QU well everyone well fire to
atare. ora mata kusa karu gara na."
be-exposed I again grass cut so EX
"That is so. Well everyone, dry yourselves off by the fire.
I'll go cut some more grass."
(data set III from Miyazawa 1961:84-87)
Data sets (III) [and (IV)] may be divided into two related but separate
narrative scenes (scene 1 consisting of sentences 1 through 19, scene 2 con­
sisting of sentences 20 through 44) which may be at work in the writer's
and, consequently, in the reader's mind.
In scene 1, the narrator uses the thematization device in such a way
that Kasuke provides information to define the scene. All through sen­
tences (III.l) and (III. 19), the noun kasuke, which is marked by wa and
never by ga, comprises the constant figure on the stage. In this particular
scene, Saburo, Ichiro's brother, and Ichiro are marked by ga, although
readers are frequently reminded throughout the story that all these partici-
THEMATIZATION AS A STAGING DEVICE 73

pants are major characters (although less so in the case of Ichiro's brother)
and are expected to remain in the reader's consciousness.9 However, in this
scene, Saburo [as in sentence (.6)] suddenly comes into Kasuke's view as
he opens his eyes. From Kasuke's point of view, Saburo indeed represents
"new" information. Likewise, Ichiro's brother and Ichiro comprise new
information as they appear in the scene. In sentence (III. 18), Ichiro is intro­
duced as if new, being marked by ga, to which Saburo (which is marked by
wa) responds in sentence (III. 19). Here, Ichiro's action is in focus while
Saburo's response is taken to be a description of the change of state and
Saburo plays a role in maintaining the reader's line of thought.
In scene 2, grandfather, which is marked by wa in (III.26) now
occupies a stable position in the scene and interacts with Ichiro's brother
who continues to be marked by ga throughout the scene, except in sentence
(III.30). Note that the event takes place at grandfather's place and his
response to the return of the boys serves as the thread of discourse. Thus,
the interaction of participants, specifically Ichiro's brother with his grand­
father becomes such that the former is marked by ga and the latter by wa.10
Ichiro's brother's actions themselves are in focus and are to be incorporated
into the plot development that is primarily maintained by the wa identified
participant, the grandfather. In this sense Ichiro's brother's action is pre­
sented as secondary and subordinate in terms of the degree of direct con­
tribution it makes to the maintenance of the thematic line, as was suggested
earlier.
In contrast with this original narrative, data set (IV), with a slight man­
ipulation of staging, represents a different narrative point of view which is
realized by a different thematic structure. In scene 1, Kasuke now is nonth-
ematized and treated as a character temporarily placed on the stage, as if
anticipating that something else will happen, and in scene 2, Ichiro's
brother is now thematized, presenting a constant figure who occupies
center stage for a longer period of time. This time it is not through Kasuke's
eyes, but through Ichiro's brother's eyes that the narrator tells the story.
We can see in data sets (III) and (IV) then, the thematization strategy func­
tioning for the manifestation of the narrator's choice for staging.11
At this point a report on the preliminary experiment on staging is in
order. In order to test our hypothesis about the "staging" function, we con­
ducted a paragraph completion test. Sixty adult native speakers of Japanese
were asked to create a narrative to continue a given discourse segment con­
sisting of four sentences and two participants as shown in paragraph (Va)
and (Vb).
74 SENKO K. MAYNARD

(V) 1. aru nichiyoobi no gogo, tetsuo-kun to taroo-kun


one Sunday LK afternoon Tetsuo and Taro
ga chikaku no hiroba de kyatchibooru o
SB nearby LK playground in catch DO
shite asondeimashita.
do:GER were-playing
One sunday afternoon Tetsuo and Taro were playing catch in
a nearby playground.
2. futo kigatsuku to hiroba no sumi ni
suddenly notice when playground LK corner at
nanika hikaru mono ga miemasu.
something glitter thing SB can-be-seen
Suddenly they noticed something shiny in the corner of the
playground.
3. sore wa maatarashii gohyakuen kooka deshita.
that TP brand-new 500-yen coin COP:PST
It was a brand new five hundred yen coin.
4. taroo (a) wa suguni sono kirakira hikaru
Taro TP immediately that ONO (glitterling) glitter
(b) ga
SB
gohyakuen kooka  hiroimashita
500-yen coin DO picked-up
Taro immediately picked up that glittering five hundred yen
coin.
Note that in sentence (4a) of (V), Taroo is marked with wa while in
sentence (4b), it is marked with ga. Subjects in this experiment, all college
students in Japan, were divided into two groups of thirty each. One group
was asked to continue the paragraph that contains Taroo-wa, and the other
the paragraph that contains Taroo-ga. Data sets (VI) and (VII) are pro­
vided as sample completed narratives to represent each group.
(VI) 1. taroo-kun wa kinoo otsukai no tochuu okane
Taro TP yesterday errand LK on-the-way money
o nakushite-shimaimashita.
DO lose-end-up:PST
Yesterday Taro lost his money while he was on an errand.
THEMATIZATION AS A STAGING DEVICE 75

2. futo taroo-kun no noori ni kinoo okaasan ni


suddenly Taro GN brain in yesterday mother by
hidoku okorareta koto ga omoiukabimashita.
terribly was-scolded fact SB flashed-in-his-mind
Suddenly Taro remembered that yesterday his mother gave
him a bad scolding.
3. taroo-kun wa uchi ni kaette okaasan ni sono
Taro TP home to returniGER mother IO that
koto  shirasemashitara sono gohyakuen kooka de
fact DO notified-when that 500-yen coin with
ishiyakiimo  kai-ni-iku koto 
roasted-sweet-potato DO go-to-buy fact DO
mei jira remas h ita.
was-ordered
When Taro went home and told his mother, she told him to
go out and buy some roasted sweet potatoes with that five
hundred yen coin.
4. taroo-kun wa runrunkibun-de itta tame ishi ni
Taro TP happily went because stone LC
tsumazuite koronde-shimaimashita.
stumble-over:GER fall-down-end-up:PST
Taro then went out so cheerfully that he stumbled over a
stone and fell.
5. soredemo megezuni tachiagatte
but not-giving-up stand-up:GER
ishiyakiimo  kaimashita.
roasted-sweet-potato DO bought
But he wasn't discouraged; he stood up again and went and
bought some roasted sweet potatoes.
6. soshite daikin o haraoo-to-shimashita ga gohyakuen
and charge DO was-going-to-pay but 500-yen
kooka wa arimasen deshita.
coin TP wasn't
But when he was ready to pay for the potatoes, he disco­
vered that the five hundred yen coin was missing.
SENKO K. MAYNARD

7. soo desu. koronda toki otoshita no desu.


so COP fell when dropped NOM COP
That's right! When he fell, he had dropped the coin.
1. soshite taroo wa tetsuo ni doo-shiyoo to iimashita.
and Taro TP Tetsuo IO what-to-do QT said
Then Taro asked Tetsuo what to do.
2. tetsuo wa soo tazunerareta shunkan
Tetsuo TP so was-asked the-moment
tsukatte-shimaoo to yuu kangae ga okori, moo
spend-end-up-will QT say idea SB occur more
chotto de, sore o kuchinidashite-shimaisoo deshita.
a-little with that DO about-to-utter-end-up COP:PST
At that moment Tetsuo thought of spending it, and he almost
said so.
3. shikashi, itsumo oya kara toka gakkoo kara toka
however always parent from and school from and
hirottara kooban ni todokenasai to iwareteita node
found-if police-box IO report:IMP QT was-told since
"kooban ni todokeyoo." to yuu kotoba ga
police-box IO report-will QT say words SB
dete-shimaimashita.
come-out-end-up:PST
But his teachers and parents had told him that if he found
some money he should take it to the nearby police box, so he
said, "Let's take it to the police box."
4. taroo toshitemo tsukatte-shimaoo ka to yuu kangae
Taro for use-end-up-will QU QT say thought
ga nakatta wake de-wa-nai no desu ga tetsuo wa
SB wasn't case isn't NOM COP but Tetsuo TP
kooban ni todokeru tsumori rashii to  to
police-box IO report intend seem QT think when
sore ni dooishi-nakuchaikenai n daroo-naa to omoi
that IO agree-must NOM probably QT think
kekkyoku futari-de kooban ni todoketa no
after-all twosome police-box  delivered NOM
des hita.
COP:PST
THEM ATIZATION AS A STAGING DEVICE 77

Although the thought of spending it did cross Taro's mind,


when he found out that Tetsuo wanted to take it to the
police box, he thought he should agree; so, after all, they
took the coin to the police box.
5. kore wa chotto mukashi no yoi   ohanashi
this TP a-little old-times LK good child LK story
deshita.
COP:PST
This is a story of two good old-fashioned boys.
Earlier the discourse function of "staging" was identified as keeping
thematic information activated and evoked so that the particular participant
will directly contribute to thematic plot development in the narrative. Our
hypothesis here is that those who continue with Taroo wa are more likely to
conceptualize Taro as a dominant character on the narrative stage than
those who continue with Taroo ga. Specifically we can investigate two
aspects of this tendency. First, we can investigate how each participant is
identified according to the marker in the lead sentence of the first para­
graph that the subjects created. And second, we can examine how domin-
antly each participant is established on one stage by the mention of the par­
ticipant and by which device is used to mark each participant throughout
the subject's narrative. In general we learned that because Taroo was cited
in sentence (5.4), regardless of how it was marked, our subjects most fre­
quently selected Taroo as the theme for their first sentence. This finding is
exemplified in sentences (VI.l) and (VILI) of our responses. Out of 30
narratives that were cued by Taroo wa, 15 subjects used Taroo wa in their
initial sentence, while out of 30 narratives cued by Taroo ga, the figure was
20. One significant finding is that in the Taroo wa narratives, Taroo wa was
ellipted in the intial sentence in two cases, while no such ellipsis was
observed in Taroo ga narratives. This seems to suggest that when Taroo wa
is cued, Taro is already established on the thematic stage and therefore
thematic maintenance is accomplished without a second mention. The fact
that Taroo ga narratives start with Taroo wa 20 out of 30 times may be
explained by the fact that when one participant is chosen over the other,
even when it is marked by ga, that mention itself encourages thematic
establishment. It seems reasonable to assume that the mere lexicalization of
one participant could influence the determination of the staging potential.
Another noteworthy point is that among narratives cued by Taroo wa,
Tetsuo wa was chosen in the initial sentence seven times. In narratives cued
78 SENKO K. MAYNARD

by Taroo ga, however, we find only two such cases. This seems to be caused
by the fact that since in Taroo wa narratives, Taro is already established on
the stage, the other participant in the event, Tetsuo, is introduced as an
opposite interactant or an antagonist. But if Taroo wa is not yet established
as in Taroo ga narratives, Tetsuo is less likely to be in the initial sentence.
It is as if the unfinished business of establishing a potential theme has to be
actualized prior to the introduction of the other individual. Tetsuo wa, how­
ever, may occur if Taro is mentioned within the same sentence since this
strategy would ensure the connection (in terms of reference) with sentence
(V.4). In fact, this is precisely the case in the two examples where Tetsuo
appeared in the intial sentence. In both cases, Tetsuo is "staged" to become
the dominant and constant figure only after a clear association with Taro is
made by establishing Taro in subordinate clauses such as: taroo ga sono
gohyakuen hooka  hirou to tetsuo wa suguni sore wa jibun ga kinoo
chichioya ni moratta no da to iidashita "When Taro picked up that five
hundred yen coin, Tetsuo right away began to insist that he got it from his
father the day before" and taroo ga sono gohyakuen kooka  hirou no o te-
tsuo-kun wa mite imashita ga kizukanai furi o shiteimashita "Tetsuo watched
Taro pick up that five hundred yen coin, but he pretended as if he didn't
notice."
Another significant finding of the staging effect is the decision of over­
all theme-establishment being triggered by an earlier occurrence of NP wa.
In Taroo wa narratives, three narratives [including the sample data set
(VI)] evolve only around Taro, while in Taroo ga narratives we find no such
example. While it is possible that either Taroo or Tetsuo may become
thematized during the course of the narrative, since subjects were free to
choose his/her own thematic development, it is significant that only in
Taroo wa narratives do we witness the case where Taro is chosen as the
main character to maintain plot cohesiveness in the created narrative. If the
character is marked by ga, the subject is likely to have more freedom to
choose either or both of the participants to maintain the thematic line.
Whereas the aforementioned interpretation is moderately supportive
of the staging hypothesis, we must bear in mind the weaknesses of this
experiment. Beyond the fact that data are limited, the empirical testing of
a narrative function — especially an elusive notion such as "staging" —
seems only partially possible, at best. Since a linguistic device functions on
more than one level, and since staging is influenced by factors beyond wa
marking, a multitude of variables are simultaneously involved in such an
T H E M A T I Z A T I O N AS A S T A G I N G D E V I C E 79

experiment. On a cautiously optimistic note, however, we believe that this


preliminary experiment supports, if only to a modest extent, our hypothesis
of staging as a discourse function of thematization.
In conclusion, this paper proposes that the staging strategy which is lin­
guistically manipulated by use and non-use of wa plays an important role in
the manifestation of the narrator's point of view. Specifically, all partici­
pants marked by wa are expected to remain evoked, activated and stored in
the consciousness of the reader so that these participants will sustain a con­
tinuity of narration. On the other hand, participants marked by ga are
expected to come into the consciousness of the reader only briefly; and
while participants marked by ga gather focus of attention, they contribute
to thematic development indirectly only through interaction with
thematized participants. This discrimination ofinformation is based on the
narrator's viewing position. The thematization in Japanese narrative dis­
course, then, is an essential part of the staging strategy that constitutes a
vital element in the narrative mode.
In the future, other devices and linguistic manipulations must be inves­
tigated before we begin to understand the literary effect such as staging
from a linguistic perspective. We hope, however, that the attempt made in
this paper will be a step forward in understanding literary manipulation and
artistic function of language in narrative discourse.

NOTES
•This paper is a slightly revised version of my paper "Thematization as a Staging Device in the
Japanese Narrative" read at the 36th annual meeting of the Association for Asian Studies,
Washington D.C., March, 1984. I would like to thank Seiichi Makino, John Hinds, Shoichi
Iwasaki and participants at the panel for helpful discussions of the issues discussed herein. My
special thanks to John Hinds for his cooperation on the experiment portion. Naturally, the
author is solely responsible for the conclusions, opinions and errors.
1) Definitions of the given and new information vary. Prince (1981) characterizes three types
of "givenness," namely, 1) predictability/recoverability, 2) saliency and 3) shared knowledge.
For further discussion on how these concepts are associated with the walga distinction, see
Maynard (1980).
2) Although as suggested elsewhere (Maynard 1982c) it is likely that wa has multiple func­
tions that span over different levels and aspects of the Japanese language, this paper concen­
trates on the function of wa closely associated with participant identification as observed on the
discourse level of the Japanese narrative.
3) Unfortunately Mio's original work (Mio, Isago, 1948 Kokugohoo Bunshooron, Tokyo:
Sanseidoo) has not been accessible. Discussion of Mio's work is based on Nagano (1972).
80 SENKO K. M A Y N A R D

4) Translation is mine. Mio's definition (according to Nagano (1972)) is given in Japanese as


follows.
Genshoobun wa genshoo o arinomama, sonomama o utsushita mono de aru. Handan no kakoo
 hodokosanaide, kangan o tsuujite kokoro ni utsulta marna  sono mama hyoogenshita bun de
aru. Genjitsu to hyoogen no aida ni nan no sukima mo nai. Genshoo to hyoogen to no aida ni
hanashite no shukan ga mattaku hairikomanai no de aru kara, soko ni wa shukan no sekinin
mondai wa nai.
5) In Mio's words (based on Nagano (1972)), handanbun is defined as: Ronrigaku de yuu
meidai, sunawachi ''Koo wa otsu da," no bun.
6) For a review of these terms related to "staging" and for a detailed discussion on how stag­
ing differs from similar terms, see Maynard (1980), Chapter Two.
7) In data set (II) we observe cases where the thematic NP and wa (i.e., obaasan wa) are
assumed to have been ellipted. Thus we understand that in this narrative the old woman eventu­
ally becomes thematized. However, the difference in staging strategy is clearly observable prior
to (II.8).
8) More specifically, data set (IV) is a result of the manipulation of data set (III) in such a
way that in sentences (1) through (19) all Kasuke wa occurrences are changed to Kasuke ga, and
in sentences (20) through (44), all Ichiroo no niisan ga occurrences are changed to Ichiroo no
niisan wa. All other cases remain the same in (IV) as in the original data set (III).
9) Note also that in (III. 16) Ichiroo no niisan is marked by wa and in (III. 19) Saburoo is
marked by wa as well. Here, the writer chooses to encourage these participants to stay on the
thematic stage toward the end of the scene. Also in (III.30), Ichiroo no niisan is marked by wa
and is temporarily thematized.
10) Although ojiisan is overtly marked with wa only once in (III.26), we assume ojiisan wa
(iimashita) 'the grandfather said' is ellipted in (III.27), (III.32), (III.34), (III.36), (III.38) and
(III.44) where the grandfather's speech is directly quoted. The most assumed participant here is
the grandfather who is staged to provide the thematic flow. Naturally, this does not guarantee
that ojiisan will be continuously thematized. As seen in (III.41) where ojiisan is marked with ga
(and therefore is in temporary focus), we find a manipulative change of participant description
on the part of the author.
11) Characterization discussed here should be interpreted as an overall global narrative effect
observed in two scenes. The actual thematic manipulation is most likely to be more complex in
that multiple participants may be simultaneously thematized and that thematized states of par­
ticipants are not stative, but are always changing. In general, however, the staging notion can
explain a primary function of wa.

REFERENCES

Brown, Gillian & George Yule. 1983. Discourse analysis. London: Cam­
bridge University Press.
Chafe, Wallace. 1976. "Givenness, contrastiveness, definiteness, subjects
topics and point of view." Subject and topic, ed. by Charles N. Li, 27-55.
THEMATIZATION AS A STAGING DEVICE 81

New York: Academic Press.


Fillmore, Charles. 1977. "The case for case reopened." Syntax and seman­
tics, Vol. 8, ed. by Peter Cole and Jerald Sadock, 59-81. New York:
Academic Press.
Firbas, Jan. 1964. "On defining the theme in functional sentence analysis."
Travaux Linguistique de Prague, 1,267-80.
Grimes, Joseph E. 1975. The thread of discourse. The Hague: Mouton.
Halliday, M.A.K. 1967. "Notes on transitivity and theme in English, Part
2." Journal of Linguistics, 3,199-244.
Hinds, John. 1973. Japanse discourse structure: Some discourse constraints
on sentence structure. Dissertation, SUNY at Buffalo.
Hopper, Paul J. 1979. "Aspect and foregrounding in discourse." Syntax and
semantics, Vol. 12, ed. by T. Givón, 213-241. New York: Academic
Press.
Inoue, Kazuko. 1980. Furui joohoo, atarashii joohoo. Gengo, Oct., 22-34.
Iwasaki, Shoichi. 1981. "A study of the NP wa construction in Japanese
expository discourse." Proceedings of the Sixth HATJ-UH Conference
on Japanese Language and Linguistics, 11-20.
Kuno, Susumu. 1972. "Functional sentence perspective — A case study
from Japanese and English." Linguistic Inquiry, 3,269-320.
. 1973. The structure of the Japanese language. Cambridge: The M.I.T.
Press.
. 1976. "Subject, theme, and the speaker's empathy — A reexamination
of relativization phenomena." Subject and topic, ed. by Charles N. Li,
417-444. New York: Academic Press.
Kuno, Susumu and Etsuko Kaburaki. 1977. "Empathy and syntax." Lin­
guistic Inquiry, 8,4,627-672.
Kuroda, S.-Y. 1973. "Where epistemology, style, and grammar meet: A
case study from Japanese." A Festschrift for Morris Halle, ed. by
Stephen R. Anderson and Paul Kiparsky, 377-391. New York: Holt,
Rinehart and Winston.
Maynard, Senko K. 1980. Discourse functions of the Japanese theme
marker wa. Dissertation, Northwestern University.
. 1982a. "The given/new distinction and the analysis of the Japanese
particles wa and ga." Papers in Linguistics, 14,1,109-130.
1982b. "Theme in Japanese and topic in English: A functional com­
parison." Forum Linguisticum, 5,10,235-261.
. 1982c. Multi-functional aspects of thematization in the Japanese nar-
82 SENKO K. MAYNARD

rative. Paper presented at the Association of Teachers of Japanese


Roundtable held in conjunction with the 34th annual meeting of the
Association for Asian Studies held in Chicago.
Mikami, Akira. 1963. Nihongo no ronri — wa to ga. Tokyo: Kuroshio Shup-
pan.
Nagano, Masaru. 1972. Bunshooron shoosetsu. Tokyo: Asakura shoten.
Ono, Hideichi. 1973. Japanese grammar. Tokyo: The Hokuseido Press.
Prince, Ellen. 1981. "Toward a taxonomy of given-new information." Rad­
ical pragmatics, ed. by Peter Cole, 233-255. New York: Academic Press.
Uspensky, Boris. 1973. A poetics of composition. Berkeley: University of
California Press.

Text References
Miyazawa, Kenji. 1961. Kaze no matasaburoo. Shinchoobunko. Tokyo:
Shinchoosha.
Tsubota, Jooji. 1975. Nihon mukashibanashishuu. Shinchoobunko. Tokyo:
Shinchoosha.
THEMATIZATION, ASSUMED FAMILIARITY, STAGING,
AND SYNTACTIC BINDING IN JAPANESE 1

JOHN HINDS
Penn State University

In this paper I will first reexamine the traditional distinction between


given and new information in Japanese syntax as these terms interact with
the postpositional particles wa and ga. Following Prince 1981 I will suggest
that a more complicated taxonomy exists than the simple dichotomy "old"
versus "new". I will also examine relevant examples in a variety of dis­
course genre to demonstrate that Prince's taxonomy alone is insufficient to
account for the distribution of the postpositional particles wa and ga. In
order to account completely for the distribution of these postpositional par­
ticles, I will introduce two additional concepts: "syntactic binding" and
"staging".
In precise writing and speech, 2 the first noun phrase in a Japanese sen­
tence may be marked by either the postpositional particle wa or the postpos­
itional particle ga. This is not to say that there are no semantic differences
in the use of these particles, or that they are freely substitutable for one
another. In fact, it is precisely the attempt to explain the relevant distribu­
tional differences which has occupied the energies of a great number of lin­
guists and language teachers. As well, these two postpositional particles are
frequently taken to be representative of a wider range of constructions
whose function is to distinguish between "given" and "new" information. 3
Martin 1975:60ff describes three situations in which the particle wa is
used. These are:
1) You are asking — or answering — a question about some other part
of the sentence, e.g. ano hito wa dare ga yonda ka? [As for that per­
son, who called (him)?].
2) You are denying something about some other part of the sentence,
e.g. tabako wa nai. [As for cigarettes, there are none.].
84 JOHN HINDS

3) You are supplying information about the points of contrast between


grammatically parallel adjuncts in two sentences, such as between
two subjects or two objects, e.g. kore wa ookii ga, sore wa chiisai.
[This is large, but that is small.].
In more concrete terms, Maynard 1981 has surveyed some of the major
research efforts which have sought to place conditions on the occurrence of
wa and ga. She contends that the major research efforts which deal with this
problem, in particular Mikami 1963, Kuno 1972, 1973, Ono 1973, and
Chafe 1976, have used such notions as "given" and "new" to define the dis-
tinction. Specifically, these studies have claimed that in the subject position
of the matrix sentence, wa marks given information, and ga marks new
information. Maynard has further argued that the differences in distribu­
tional properties between wa and ga cannot be described through reference
to these terms.
Chafe 1976:30 provides a standard definition for the terms given and
new information.
Given (or old) information is that knowledge which the speaker assumes to
be in the consciousness of the addressee at the time of the utterance.
Socalled new information is what the speaker assumes he is introducing
into the addressee's consciousness by what he says.
Kuno's account (1972, 1973) corresponds closely to these definitions.
In Kuno 1973, two specific functions of the postpositional particle wa are
listed. Note that these two functions correspond to Martin's points (1) and
(3).
(1) wa marks the "theme" of a sentence,
taroo wa gakusei desu.
(2) wa marks a contrast,
ame wa futte-imasu ga,
In Kuno 1972:269, he
reexamines uses of wa and ga from a different angle, namely, that of what
is old, predictable information and what is new, hence unpredictable infor­
mation in a given sentence, and presents the hypothesis that ga is a marker
that indicates that the subject represents new, unpredictable information
in the sentence.
This statement elaborates on Chafe's distinction by including the con­
cept of predictability. This term "predictable" interacts with the notions
THEMATIZATION IN JAPANESE 85

"anaphoric" and "specific" in Kuno 1973. There he explains that only ele­
ments which are in the "discourse registry" may be marked by wa. The dis­
course registry is an abstract entity which keeps track of objects and con­
cepts which are introduced into a discourse [see also Hinds 1976]. Kuno
considers that objects and concepts in discourse are recorded in this "dis­
course registry", and that there are two major ways in which items may be
registered. First, they may be items of unique reference such as the sun, the
moon, my wife. Second, they may be objects of specific reference which are
added to the registry the first time they are mentioned in a discourse, such
as a man I saw yesterday.
Kuno 1973:39 states:
Only after this entry in the registry is accomplished can [these objects and
concepts] become themes of sentences [i.e. can be marked by wa (JH)].
Thus, for Kuno, an item of specific reference may not become a theme on
its initial mention in a discourse [see also Kuno 1972]. Moreover, it is not
the case that any specific item which is mentioned in a discourse automati­
cally becomes a theme. Kuno 1973:40 states:
What determines whether a specific noun phrase can become a topic or not
depends on whehter the noun phrase is anaphoric and not whether it is
definite.

Additionally, Kuno 1972:276-7 requires that a noun phrase represent


old, predictable information in order to be marked by wa. Generic noun
phrases are outside of this constraint and may freely be marked by wa.
Thus, to summarize Kuno's position, he claims that wa is used to mark
either thematic or contrastive noun phrases. If the noun phrase is thematic,
it must be either anaphoric or generic, and it must represent old, predicta­
ble information. Kuno's 1973:59 principal claim is that "[n]onanaphoric
nongeneric themes result in ungrammaticality
This account of the uses of wa and ga, as well as others which depend
on the notions given and new, provide a legitimate perspective from which
to view the distribution. There are difficulties with this perspective, how­
ever. One of these difficulties, as Maynard 1981 has shown, is that ga may
be used to mark given, or known, information, even though wa is the
expected postpositional particle.
Another difficulty is brought to light through application of concepts
presented in Prince 1981. Prince provides a more complex taxonomy of
86 JOHN HINDS

given-new information. She abandons the terms "shared knowledge" and


"given" to refer to what the speaker assumes the listener is aware of. She
adopts instead the expression "Assumed Familiarity", and asks (p. 233),
From the point of view of a speaker/writer, what kinds of assumptions
about the hearer/reader have a bearing on the form of the text being pro­
duced, where that form is not uniquely determined by the "objective"
information that the speaker/writer is attempting to convey?
Figure 1 presents Prince's taxonomy, and examples (1) through (7)
illustrate each. Note that Prince includes a new attribute in each noun
phrase.

Figure 1
ASSUMED FAMILIARITY

Brand-new Unused (Noncontaining) Containing (Textually) Situationally


Inferrable Inferrable Evoked Evoked

(Unanchored) Anchored

(1) I bought a beautiful dress. (Brand-new + attribute)


(2) A rich guy I know bought a Cadillac. (Brand-new Anchored +
attribute)
(3) Rotten Rizzo can't have a third term. (Unused + attribute)
(4) I went to the post office and the stupid clerk couldn't find a
stamp. (Inferrable + attribute)
(5) Have you heard the incredible claim that the devil speaks English
backwards?(Containing Inferrable + attribute)
(6) Susie went to visit her grandmother and the sweet lady was mak­
ing Peking Duck. (Evoked 4- attribute)
(7) Lucky me just stepped in something. (Situationally Evoked +
attribute)
For English, although Prince's examples obscure this a bit, there is a
THEMATIZATION IN JAPANESE 87

distinct cleavage between examples (1) and (2) on the one hand, and exam­
ples (3) through (7) on the other. Examples (1) and (2) require their noun
phrases to be marked by the indefinite article a, while (3) through (7)
require the. This may easily be seen by substituting the noun phrase The
Mayor for Rotten Rizzo in (3), and by substituting The lucky bum for Lucky
me in (7).
If we reexamine the distribution of the postpositional particles wa and
ga from the standpoint of this taxonomy, we discover that those items
termed "Inferrable" and "Evoked" are typically marked by wa.Ą This much
is well known to the linguistic public, and is simply a refinement of the term
"given".
In Prince's category "New", however, we find, contrary to expecta­
tion, that ga does not mark each of these noun phrases. We find that noun
phrases which indicate "Unused New" information are marked by wa.
Further, we find that "Brand-New Anchored" information may be marked
by either wa or ga. It is only the category "Brand-New Unanchored" which
obligatorily requires ga marking.
Figure 2 presents this information.

Figure 2
ASSUMED FAMILIARITY

New Inferrable Evoked

Brand-new Unused (Noncontaining) Containing (Textually) Situationally

Inferrable Inferrable Evoked Evoked


[ga/wa] [wa] [wa] [wa]

(Unanchored) Anchored
[ga] [ga/wa]

The effectiveness of Prince's approach may be seen by applying some


of these concepts to a radio news broadcoast. Rowe 1982 examines
Japanese radio news items in order to discuss the distribution of the post-
88 JOHN HINDS

positional particles wa and ga. He points out that certain types of news
items, such as reports of traffic accidents, contain almost totally new infor­
mation. This observation accounts for the distribution of wa and ga he finds
in such new reports. In a nine clause article, he finds that there are only
three instances of the postpositional particle wa. Each of these uses is
explained by principles enunciated in Kuno 1972 and Hinds 1976.
Rowe presents without analysis a second news report which contains
some difficulties in the interpretation of the distribution of wa and ga. This
news report is presented here in its entirety.
tsugi ni, mosukuwa hatsu mainichi shinbun tokuden.
Next, a special Mainichi Shimbun report from Moscow.
sobieto kyoosantoo no kikan shinbun purauda wa, yokka, genzai nippon de
okonawarete iru migu 25 no kaitaichoosasagyoo wa nissoo kankei no kyooka
 samatageyoo to suru hanso seiryoku no sakuboo da to hinan suru ippoo.
ippoo.
On the fourth, Pravda, the Soviet Communist Party Newspaper, accused
the Japanese government of the dismantling operation of its MIG 25 by
saying that it is an anti-Soviet scheme which will hinder the strengthening
of Japanese-Soviet ties.
nippon kokunai і  yoo na ugoki ni hantai shi, sobieto to no zenrin
kankei  kyooka suru  ga takamatte iru to hoojimashita.
They report that within Japan, in opposition to this movement, there are
voices being raised to strengthen the good-neighbor relations with the
Soviet Union.
kore wa, purauda no rassheiefu tookyoo tokuhain no kaita kiji de
chakurikushi de no migukaitai wa sobieto no kitainhenkanyookyuu  mushi
shita ue, amerika no senmonka mo sanka shite tettootetsubi supaimokuteki
de okonawarete iru to shiteki suru to tomo ni, nippon no jieitai no shidoobu
wa sobieto o kasooteki to minasu reisendokutorin no shinpoosha de aru
koto ga bakuro sreta to hinan shite imasu.
This article was written by Rasheiev, a special correspondent for Pravda to
Tokyo. He said that the operation was continuing despite the Soviet's
request for the return of the MIG 25, and that some American experts had
been involved through espionage activity. He also reported that the leader
of the Japanese Self-Defense Forces had been a proponent of the Cold
War Doctrine which describes the Soviet Union as a hypothetical enemy.
Of the five boldfaced noun phrases marked by wa, two cause some
trouble for an analysis which requires wa-arked noun phrases to be
anaphoric, or predictable information. The first noun phrase marked by wa
might be considered to be part of the discourse registry, since the byline is
THEMATIZATION IN JAPANESE 89

from Moscow. The second noun phrase, however, is certainly not predicta­
ble information in any sense of the term, since the story from Moscow could
have involved anything related to Russia, or any other item for that matter.
The third noun phrase marked by wa is clearly anaphoric. The fourth
noun phrase is anaphoric, although it is not predictable. Other concepts
could have occupied this position equally well, including sobieto to no zen-
rinkankei o kyooka shiyoo to shite iru nihonjin. The final noun phrase
marked by wa might conceivably be argued to be introduced by implication
[see Hinds 1976], but again a question arises concerning the predictability
of nippon no jieitai no shidoobu.
The distribution of wa and ga in this news item appears to cast strong
doubt on any analysis which relies exclusively on the notions of anaphoric-
ity and predictable information. Even if the concept of anaphoricity is
extended to include those elements which are implied by overt elements,
there is still a problem with the notion of predictability.
If this report is looked at in terms of Prince's taxonomy, however, the
distribution of wa and ga is explainable. The first two noun phrases repre­
sent New-Unused information. The second two represent Textually
Evoked information, while the final noun phrase represents Containing-
Inferrable information, since it may legitimately be assumed that the jieitai
has a leader. All of these types of information, as pointed out above, are
normally marked by wa.
This same analysis can help to explain the systematic use of wa to mark
the initial noun phrase of magazine articles. As the following example
taken from the January 27, 1984 issue of FOCUS demonstrates, even noun
phrases on first mention can be marked by wa. Clearly, in these cases the
noun phrases are neither anaphoric nor predictable. 5
FUKKATSU SHITA "MUCHIUCHI KEF
A RETURN TO FLOGGING
gookanotoko ga uketa isuramu hoo dentoo keibatsu no nākami
The traditional punishment which a rapist receives
pakisutan no toshi karachi ni aru sakkaajoo ni atsumatta gunshuu wa yaku
3000 і.
A crowd of about 3000 gathered at a soccer field in the Pakistani city of
Karachi.
betsu ni sakkaa geemu  kenbutsu shiyoo to iu no de wa naku, karera ga
katazu  nonde machikamaete iru no wa "isuramu hoo ga gunsei ni shiko
sareru shunkan" de aru.
90 JOHN HINDS

They had not come to see a soccer match. They were waiting anxiously for
"the moment that Islamic Law would be strictly carried out."
In the first sentence, "the crowd" is introduced with the postpositional
particle wa. A more literal translation of this sentence is:
The crowd which gathered at a certain soccer field in the Pakistani city of
Karachi [numbered] about 3000.
The second sentence develops this scenario which has been created by
talking about the crowd's "eager anticipation". Again, this is not predicta­
ble information in any sense of the term. A literal translation of the rele­
vant portion of this sentence is:
The thing which they are anxiously awaiting is "the moment that Islamic
Law is strictly carried out.''
Both of these noun phrases represent Containing-Inferrable Informa­
tion. In the first case, the head noun gunshuu "crowd" is modified by the
relative clause "which gathered at a certain soccer field in the Pakistani
city of Karachi". The specificity of this relative clause has determined the
fact that it is Containing. In the second example, the fact that the anaphoric
karera "they" is in the relative clause establishes the fact that it is Contain­
ing.
It is important here to point out that there are syntactic constructions
in Japanese which "bind" an author to a particular spot on the taxonomy.
As an example of this, consider the fact that Japanese, as discussed by Li
and Thompson 1976, is both a topic-prominent language and a subject-
prominent language. The consequences of this are that both sentences of
type (A) and type (B) are unmarked, where (A) typifies a subject-promi­
nent construction, and (B) typifies a topic-prominent construction. (B) type
sentences may be reduced to the syntactic formula NP wa NP desu.
(A) uchi-no musuko ga/wa tookyoo ni ikimashita.
our son Tokyo to went
Our son went to Tokyo.
tanaka-san ga/wa raishuu sono hon o yomu deshoo.
next-week that book read probably
Mr. Tanaka will read that book next week.
(B) tookyoo ni itta no wa uchi-no musuko desu.
Tokyo to went NOM our son copula
The one who went to Tokyo is our son.
THEMATIZATION IN JAPANESE 91

tanaka-san wa sensei desu.


teacher copula
Mr. Tanaka is a teacher.
The major difference between these two sentence types is that in the
(A) sentences, the first noun phrase can be freely marked by either ga or
wa, depending on the presuppositions involved. Typically, when an ele­
ment is introduced into a discourse with a subject-prominent construction,
the postpositional particle ga is more appropriate, because the noun phrase
represents Brand-New information. Subsequent mention of this element
then favors the postpositional particle wa, since it represents information
which is not Brand-New. The (B) sentences typify cleft-constructions and
equational sentences, respectively.
When elements are introduced into a discourse with a topic-prominent
construction, there is a considerable pressure to mark that element with wa
on its first appearance. Examine again the first sentence in the article about
the rapist, reprinted here for convenience.
pakisutan no toshi karachi ni aru sakkaajoo ni atsumatía gunshuu wa y aku
3000 і.
A crowd of about 3000 gathered at a soccer field in the Pakistani city of
Karachi.
This sentence reduces to the basis syntactic formula NP wa NP (desu).
Since the use of this specific syntactic structure is bound to information
which is not Brand-new, I therefore refer to this process as "syntactic bind­
ing". Thus, the decision to begin the article with this topic-comment syntac­
tic structure, rather than a subject-predicate syntactic structure, allows the
first noun prhase to be marked by wa rather than ga.
This fact may be demonstrated in a straightforward way. If this article
began with a subject-prominent sentence, "the crowd" would be more
likely to be marked by the postpositional particle ga than wa.
yaku 3000 і no gunshuu ga (?wa) pakisutan no toshi karachi ni aru sak­
kaajoo ni atsumatta.
A crowd of about 3000 gathered at a soccer field in the Pakistani city of
Karachi.
Thus one constraint on the use of wa to mark nonpredictable
nonanaphoric information in expository writing is that the noun phrase
must be contained in a sentence which has the NP wa NP desu structure.
Use of this structure is restricted to information which is not Brand-New.
92 JOHN HINDS

A second constraint is related to Maynard's notion of staging. Staging,


according to Maynard, is the conscious choice of an author, or speaker, to
select a single concept to become the focal point, or theme, of the sub­
sequent discourse. We may conceptualize a discourse, with Norman and
Rumelhart 1975, as a scenario in which linguistic and nonlinguistic means
are used to build up a more and more complex scene. Characters, props,
events, concepts, and motivations are all introduced and remain in the
scenario until specifically removed. The removal of characters, for instance,
is usually achieved through scene changes.
The following segment of dialogue illustrates how such scenario crea­
tion often operates in conversation.
H209. anoo, shinu n ja-nai ka to omotta koto  ru?
uh die nom neg QU QT thought fact like be
Urn, have you ever thought that you were going to die?
W210. aru, ichido aru.
be once be
Yes I have, once.
H211. dono toki?
which time
When?
W212. ichido, nido aru.
once twice be
Once, no twice.
H213. nido mo an no?
twice also be nom
Twice?
W214. nido aru.
twice be
Twice.
H215. haa.
ha
Oh.
W216. ichido wa ne, (0.8) itsu datta ka-::, ee, tonikaku,
once TP EX when was wonder uh anyway
noosu-karoraina-karoraina ni iru toki ne?
North-Carolina-Carolina in be when EX
THEMATIZATION IN JAPANESE 93

Well, the first time, when was that? uh, anyway, it was when I
was in North Carolina-Carolina, see?
W217. sugoku chiisa na gakkoo datta no ne,
very small LK school was nom EX
and I was at a really small school.
H218. nn.
nn
I see.
W219. sore-de, machi made iku no-ni,
and town to go to
So, in order to get to the town,
W220. (0.7) gakkoo kara basu ga- minibasu ga dete-ru wake,
school from bus SB mini-bus SB going reason
there's a bus- a minibus that goes from the school.
H221. hn-hn.
un-hum
Uh-huh.
W222. tatoeba, ichinichi ni sankai gurai ne?
e.g. one-day in 3-time about EM
For example, it goes about three times a day, see?
H223. nn.
nn
I see.
W224. de, aru fuyu: datta ka-na,
uh a winter was wonder
So, let's see, it was one winter,
W225. , yuki ga futte-te ne,
like snow SB falling EX
uh, it had been snowing, see?
H226. nn.
nn
I see.
W227. sore-de, betsu-ni, , nantomo omowazu-ni
and special like nothing not-think
94 JOHN HINDS

So, uh, you know, without thinking about anything in particu­


lar,
W228. sono basu ni notte
that bus on board-and
I got on the bus and
W229. dauntaun ni itta wake.
downtown to went reason
went downtown.
H230. nn.
nn
Oh.
W231. sore-de mat-anoo, kaerimichi ni gake kara otchatta no,
and uh return on cliff from fell nom
sonoo, anoo, kuruma ga ne?
uh uh car SB EX
So, do-uh, on the way back to school, it fell off a cliff, uh, the
car right?
W232. gake kara ochichatta no, , untenshu no misu de.
cliff from fell nom like driver LK miss by
we fell right off a cliff, you know, because of something the
driver did.
W216 introduces the scenario: The event to be reported took place
when W was in North Carolina. Subsequently, the listener learns that the
event took place at a small school, that there was a mini-bus which went to
town about three times a day, that it was winter and that it had been snow­
ing, and that W was on the bus. Then in W231 and W232, the listener learns
that the bus fell over a cliff.6 Of relevance here is the fact that the develop­
ment of the scenario takes place without the use of the postpositional parti­
cle wa. In the construction of a scenario, no element is brought into central
focus.
The construction of a scenario can also be visualized as a stage to which
characters, props, events, concepts, and motivations are brought through
linguistic and nonlinguistic devices. The analogy is complete if a spotlight is
employed. Of all the elements on stage, the spotlight can shine on only one,
or only one group. This element, or group of elements, will constitute the
theme of the discourse for as long as the spotlight remains in place.
THEMATIZATION IN JAPANESE 95

The linguistic manifestation of this spotlight in Japanese is the use of


the postpositional particle wa. Those elements marked by wa are "center-
stage". They are what the discourse is about. What this means is that the
author, or narrator, has considerable flexibility in terms of which elements
are brought to the center stage. Such flexibility has been demonstrated effec­
tively for narrative discourse in Maynard 1981, where she shows that the
use of the postpositional particle ga rather than wa to mark a specific noun
phrase subordinates that noun phrase to other noun phrases in the story. 7
There are times, of course, when noun phrases are marked by wa on
initial mention. This is done by exploiting the notion of New-Unused infor­
mation. Prince 1981:235 draws a parallel between this type of information
and following a recipe. New information in general is similar to putting
materials on the counter for preparing. Brand-New information must be
created, parallel to going out to buy some ingredient. New-Unused infor­
mation on the other hand, refers to information about which
the hearer may be assumed to have a corresponding entity in his/her own
model and simply has to place it in (or copy it into) the discourse model,
akin to taking some staple off the shelf when its presence is suddenly taken
for granted in a recipe (e.g., salt).
For dramatic effect, certain noun phrases which cannot possibly be known
or predicted are treated as if they are New-Unused information. This is a
common device used by novelists, short story writers, and reporters. 8 It
may be illustrated in the opening sentences of the novel Kuchibue o Fuku
Toki (When I Whistle) by Shusaku Endo.
"shitsurei desu ga ..."
"Excuse me ..."
Ozu wa yukkuri me o akea.
Ozu slowly opened his eyes.
kono shinkansen no naka de itsu no aida ni ka, nemutte ita.
At some point he had dropped off the sleep on the bullet train.
fuyu no wabishii hizashi ga hamanako no namariiro no suimen ni sashite,
ni, sanseki no kobune ga uite iru.
The cheerless winter sun shone upon the grey surface of Lake Hamana,
where two or three boats floated.
"shitsurei desu ga ..."
"Excuse me ..."
96 JOHN HINDS

  kaheta otoko wa hitonatsukoi, natsukashige na hyoojoo o ukabete,


"ozu san, ja ... arimasen ka"
The man who spoke wore an amiable, kindly expression. "Aren't you ....
Mr. Ozu?"
[English translation by Van C. Gessei, When I Whistle, Charles Tuttle]
Here the first two characters are introduced on first mention with the
particle wa. This is an attempt on the part of the author to draw the audi­
ence in, to make the reader feel as if he or she is already aware of the charac­
ters.
This device of treating unknown information as if it were known is not
only a common literary technique, it provides an answer for why certain
unpredictable noun phrases are marked by wa on first mention in the mid­
dle of a passage. Examine the next passage, taken from the daily column
Tensei Jingo which appears in the Asahi Shimbun [5/28/83]. Numbers in
parentheses indicate paragraph boundaries which were marked in the origi­
nal essay.

Waribashi
(1) shokudoo de waribashi  tsukau.
dining-hall LC waribashi DO use
We use "waribashi" (half-split throw-away chopsticks) to eat.
tsukaistue de-aru
use-throw COP
After use, they are thrown away.
suterareta hashi wa ittai doo naru no
threw-PASS chopsticks TM what how become NOM
dar oo.
PRESUMP
What happens to them after they are thrown away?
mottainai-na, to u no wa senchuuha no
wasteful QT think NOM TM war-generation LK
ijimashisa daroo ka.
stinginess PRESUMP QU
Is it merely the stinginess of those who lived through the war to
feel that it is a waste?
THEMATIZATION IN JAPANESE 97

aru shokudoo no hanashi de wa, mikka de


certain dining-hall LK talk by TM 3-day by
ichimanbon no waribashi  tsukaisuteru to iu.
10,000 LK waribashi DO use-throw QT say
According to the owner of one restaurant, his restaurant uses
and throws away 10,000 pairs of chopsticks every three days.
nihon-zentai de wa, ki no waribashi wa ichinen ni
all-Japan at TM tree LK waribashi TM 1-year in
yaku 100-oku-zen mo tsukawareru soo da.
about 10,000-million-pairs also use-PASS hearsay COP
In Japan as a whole, about 10,000 million pairs of wooden
chopsticks are used each year.
sore dake no ryoo no mokuzai ga ichido tsukawareta
that only LK portion LK wood SB 1-time used-PASS
dake de, suterareru kangaete-mireba zeitaku
only COP-and throw-away-PASS think-try-if luxurious
na hanashi de-aru.
LK story COP
That much wood is used just once and thrown away.
kami no genryoo ni suru tame kaishuu shi,
paper LK raw-material into do purpose collect and
saisei sareru to iu hanashi wa kikanai.
reuse do-PASS QT say story TM hear-NEG
We have never heard about wooden chopsticks being collected
and reused as a raw material to make paper.
kaishuuhiyoo ga kakarisugite saisan ga awanai,
collection-cost SB require-too-much remake SB meet-NEG
to iu koto daroo ka.
QT say fact PRESUMP QU
Is it because it would not be a paying proposition since collecting
the chopsticks would cost too much money?
tabemono  kuchi ni hakobu no ni hashi 
food DO mouth to transport NOM to chopsticks DO
tsukau iwayuru hashibunkaken wa chuugoku,
use that-is chopstick-cultures TM China
JOHN HINDS

choosenhantoo, betonamu, soshite nihon, to natte-iru.


Korean-Peninsula Vietnam and Japan RES becoming
The so-called "chopsticks culture" sphere includes China, the
Korean Peninsula, Vietnam, and Japan.
sono  de mo nihon-igai  kuni wa hashi
that among at also Japan-except LK country TM chopstick
to saji no hiyoo de-aru
and spoon LK necessity COP
But all these other countries use spoons as well as chopsticks.
hashi o kihon to suru nihon-ryoori wa kiwamete
chopsticks DO base as do Japan-cooking TM very
tokui na sonzai na no da.
special LK substance LK NOM COP
Japanese cooking which is based on chopsticks, is a very special
thing.
mukashi wa waribashi to ieba yoshino-san no
before TM waribashi QT say-if Yoshino-made LK
sugi-bashi datta.
sugibashi COP-past
Before the war, "waribashi" were Japanese cedar chopsticks
from Yoshino.
yoshino no akasugi no hashi o te ni toru
Yoshino LK red-cedar LK chopsticks DO hand in take
to, mizumizushii sugi no kaori ga suru.
when fresh cedar LK smell SM do
When you pick up a pair of chopsticks made of red Japanese
cedar from Yoshino, you can smell the fresh odor of the cedar.
karukute, yawarakami ga aru.
light-and softness SB exist
They are light and smooth to touch.
m a s a m e ga sutto tootte-ite sugata ga ii.
grain SB straight goes-and shape SB good
They are straightgrained and look good to the eye.
ichizen no sugibashi ni wa nihonjin no biishiki ga
1-pair LK sugibashi in TM Japanese LK aesthetics SB
THEMATIZATION IN JAPANESE 99

komerarete-iru yoo і u.


concentrate-PASS way RES think
We feel that the aesthetic feelings of the Japanese are concen­
trated in a pair of Japanese cedar chopsticks.
rikyuu wa,   motenasu hi no asa,
Rikyu TM guest DO receive day LK morning
akasugi no hashizai  toridashi, ninzuu ni oojite
red-cedar LK material AM take-out number to according
hashi  kezuri kezuritate no sugi no kaori 
chopsticks DO whittle just-made LK cedar LK smell DO
kyoo shita to iu densetsu ga aru.
present did QT say legend SB exist
On the mornings of those days on which he was expecting vis­
itors, the tea ceremony master Rikyu got out some red Japanese
cedar wood and whittled just enough pairs of chopsticks for the
expected number of visitors. He then presented the guests with
the odor of freshly-cut Japenese cedar.
sore ga rikyuu-bashi no umare da.
that SB Rikyu-chopsticks LK birth COP
This is the origin of the "Rikyu-bashi" (Rikyu chopsticks)
nihonjin wa hashi no atarashisa, kiyorakasa 
Japanese TM chopsticks LK freshness purity DO
motorne ta.
demanded
The Japanese demanded freshness and purity in their chopsticks.
jib un no hashi ga tanin ni tsukawareru koto,
own LK chopsticks SB other by use-PASS fact
hito no hashi  tsukau koto o kirau
person LK chopstick DO use fact DO dislike
no wa, keppekikan dake de wa nai.
NOM TM fastidiousness only COP TM NEG
It is not just out of fastidiousness that the Japanese do not like
others to use their chopsticks and also do not like to use the
chopsticks of others.
mukashi no hito wa hashi ni wa sore  tsukau
former LK person TM chopsticks in TM that DO use
JOHN HINDS

hito no reiryoku ga yadoru to shinjita. dakara


person LK spirit SB live QT believed therefore
jibun no reiryoku ga yadotta hashi o tanin ni
own LK spirit SB lived chopsticks DO other by
tsukawareru koto o kıratta no da, to iu
use-PASS fact DO disliked NOM COP QT say
sets  mo aru (Honda Sooichiro: Hashi no
explanation also exist chopsticks LK
Hon).
book
In ancient times, people believed that the spirit of the person
resided in the chopsticks that he used and it is said that this is
why people hated to have their chopsticks used by others. (Hashi
no Hon — Book of Chopsticks — by Soichiro Honda.
saikin wa hokkaidoo-san no, ezomatsu, kaba, shina
recently TM Hokkaido-made LK ezomatsu kaba shina
nado de tsukurareta waribashi no zensei-jidai de-aru.
others by made-PASS waribashi LK heyday COP
It is now the heyday of "waribashi" made from the silver firs,
birches and Japanese lindens of Hokkaido.
korera no waribashi ga tairyoo ni seisan
these LK waribashi SB large-quantity in produce
sarehajimeta no wa dosei-chooki ni
do-PASS-began NOM TM high-economic-growth-age to
haitte kara da.
enter since COP
The "waribashi" made from Hokkaido wood began to be pro­
duced in large quantities after the age of high economic growth
started.
ima, demawatte-iru waribashi no hanbun wa
now arrived waribashi LK half TM
hokkaidoo-san da to omotte ii.
Hokkaidoo-made COP QT think-if good
Half the "waribashi" used now are produced in Hokkaido.
waribashi no tsukaisute sono mono  hitei suru
waribashi LK use-throw that thing AM criticism do
THEMATIZATION IN JAPANESE 101

tsumori wa nai.
intent TM NEG
We have no intention of condemning the use-and throw-away
system in connection with "waribashi"
shikashi nenhan 100-okuzen-bun no ga sono mama
however yearly 10,000-million LK tree SB that condition
kiete-shimau no wa ika-ni-mo mottainai.
disappear NOM TM really wasteful
But it is very wasteful when trees amounting to 10,000 million
pairs of wooden chopsticks disappear each year.
With the exception of the first paragraph, the initial noun phrase in
each paragraph is marked by wa. There are, of course, other noun phrases
marked by wa, but they will not be considered here. The question to be
asked is why each of these noun phrases can be marked by wa. To answer
this question, it is necessary to understand the rhetorical pattern which
organizes this essay.
The essay is organized according to a pattern known as ki-shoo-ten-
ketsu [see Hinds 1983a, 1984]. This rhetorical style is described by
Takemata 1976:
A. ki First, begin one's argument.
B. shoo Next, develop that.
 ten At the point where this development is finished, turn the
idea to a subtheme where there is a connection, but not a
directly connected association [to the major theme].
D. ketsu Last, bring all of this together and reach a conclusion.
The version of the rhetorical pattern which is used in this essay has a
proliferation of ten. This means that there are a number of tangentially
related subtopics brought up rather upbruptly, with little forewarning. The
function of each paragraph in this essay is listed in Chart 1.
Chart 1
(1) ki (5) ten
(2) shoo (6) ten
(3) ten (7) ten
(4) ten (8) ketsu
In the first paragraph, the initial sentence contains no noun phrase
marked by wa.9 The first sentence of paragraph 2 contains an introductory
102 JOHN HINDS

phrase which is marked by wa. This introductory phrase, aru shokudoo no


hanashi de wa "according to the owner of one restaurant," is marked by wa
because the owner of the restaurant has been identified as a unique indi­
vidual through the implementation of the appropriate "script" [see Schank
and Abelson 1977, Hinds 1982] which specifies that a restaurant typically
has a single owner.
There is an abrupt transition in paragraph 3 to a consideration of
"chopstick cultures". This is unpredictable as a topic shift, yet this noun
phrase is marked by wa.
In paragraph 4, the scene is again shifted to "before the war". The
noun phrase which signals this shift, mukashi "long ago", is marked by wa.
Paragraph 5 introduces a specific person into the essay, a tea ceremony
master named Rikyu. The noun phrase which refers to Rikyu is also
marked by wa.
In paragraph 6 there is a shift from Rikyu to the fact that Japanese dis­
like using another's chopsticks. In paragraph 7 this perspective is again
changed to the present, where the noun phrase which forces this shift,
saikin, is marked by wa.
In the final paragraph, the noun phrase marked by wa precedes a nega­
tive verb. The fact that this sentence indicates a negative statement is
enough to trigger the use of wa [see Kuno 1973, Hinds 1983a].
Each of the sentences which initiate paragraphs 3 through 7 begin with
NP wa, and they each indicate New-Unused information. Of course, the
first noun phrases in (4) and (7) indicate temporal anchoring as well.
Let's examine paragraph 5 a bit more closely. Paragraph 5 begins with
the noun phrase rikyuu wa, a literal translation being "As for Rikyu".
Rikyu is not old, predictable information since there was no mention of him
previously. Rikyu is New-Unused information in that all (possibly most?)
Japanese are aware that he was a tea-ceremony master.
The function of wa in this case is a signal to the reader that the noun
phrase so marked has some kind of connection with the overall theme of the
essay. It informs the reader that there is in fact some connection, and that
the reader should make an effort to place this noun phrase in its proper
perspective in the essay as a whole. It tells the reader, in effect, that this
noun phrase should be treated as if it were old, predictable information,
even though it is not. In this respect, the function of wa when it marks the
first noun phrase of a ten is very similar to English transition statements
THEMATIZATION IN JAPANESE 103

such as, "The following may seem to be unrelated to the major point, but
the connection between Rikyu and chopsticks will become clear in due
time."
I have attempted to document the following points in this paper. First,
the taxonomy of "Assumed Familiarity" introduced in Prince 1981 contri­
butes greatly to the understanding of certain aspects of the distribution of
the postpositional particles wa and ga. In particular, only noun phrases
which indicate Unanchored Brand-New information are required to use ga.
Other types of New information — Anchored Brand-New information and
Unused information — may be marked by the postpositional particle wa.
Second, noun phrases which appear in topic-comment structures, as
opposed to subject-predicate structures, typically mark the initial noun
phrase with wa.10 A reason for this is that such noun phrases, even if they
represent new information, frequently indicate anchored new information
or unused information.
Third is the concept of staging, or marking in a special manner those
noun phrases which are to become the central character rather than
peripheral characters. Specifically, noun phrases which are marked by wa
remain on stage for a longer period of time than those not so marked. This
has been discussed in Hinds 1983b and in Maynard 1984 (and this volume).
Maynard (1984:7) states:
What the narrator wishes to accomplish through "staging" is to discrimi­
nate the thematized participants from non-thematized ones in such a way
that thematized participants remain activated, evoked and stored in the
reader's consciousness.
Fourth, there are dramatic and rhetorical pressures for using wa to
mark noun phrases which indicate Unused information. One such dramatic
pressure is that readers are drawn into novels, short stories, or articles more
quickly when it appears the referent of a noun phrase is known to the
reader, wa functions to inform the reader to treat such initial noun phrases
as items of information they should already know. Another pressure is
rhetorical in that the appearance of a noun phrase marked by wa in the ten
of the ki-shoo-ten-ketsu rhetorical pattern informs the reader that the infor­
mation in the noun phrase so marked has a connection with the overall
theme of the article, even if the connection is not immediately obvious.
104 JOHN HINDS

NOTES
1) I would like to thank Shoichi Iwasaki, Wako Hinds, Pat Clancy, and Senko Maynard for
comments and suggestions. Any errors remain my own.
2) See Hinds 1982, 1983a for a discussion of conversational situations in which no postposi­
tional particles occur.
3) See Hinds 1973 for discussion of theme and rheme in which postpositional particles other
than wa and ga are considered.
4) See Maynard 1981 for instances in which ga marks noun phrases which represent inferrable
or evoked information.
5) The translations of these sentences provided here do not necessarily reflect the structure of
the Japanese sentences.
6) Note that kuruma "car" is marked by ga in W231 despite the fact that it is "postposed" [see
Hinds 1982]. This indicates that at this point in the story W is not treating the car as a thematic
element. That is, this segment of the story is not about the car, but about W [see Maynard 1981].
7) This is consistent with Kuno's 1973 account in which ga is used in subordinate clauses
[see also Hinds and Hinds 1979]. This is also consistent with traditional pedagogical grammars
which state that the noun phrase marked by wa focuses on the rest of the sentence [see Alfonso
1966, Jorden 1964, Martin 1954].
8) In Hinds 1975, I discuss the use of "backwards pronominalization" in English to build sus­
pense. There is a direct parallel between the use of pronouns in English and the use of NP wa in
Japanese.
9) The sentence does contain an instance of ellipsis. The "subject" of the sentence is the
speaker, an item of unique reference, and so it may be freely omitted, even on first mention [see
Hinds 1982].
10) As Martin 1975:867 has shown, however, noun phrases in such constructions can be
marked by ga.
(a) mainichi kaisha o deru no ga oru kuji (da).
Everyday I leave the office at 9:00 pm.
(b) ichiban komatta no ga joyuu-busoku de aru.
The worst part is the shortage of actresses.

REFERENCES

Alfonso, Anthony. 1966. Japanese Language Patterns. Tokyo: Sophia Uni­


versity Press.
Brown, Gillian and George Yule. 1983. Discourse Analysis. Cambridge:
Cambridge University Press.
Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects,
topics and point of view. Subject and Topic, ed. by Charles N. Li. NY:
Academic Press, 27-55.
THEMATIZATION IN JAPANESE 105

Hinds, John. 1973. Theme-rheme in Japanese. Seishin Studies 41:1-13.


. 1975. Passives, pronouns, and themes and rhemes. Glossa 9:79-106.
. 1976. Aspects of Japanese Discourse Structure. Tokyo: Kaitakusha
Publishers.
1979. Organizational patterns in discourse. Syntax and Semantics, Vol­
ume 12, Discourse and Semantics, edited by T. Givón, 135-157. New
York: Academic Press.
1980. Japanese expository prose. Papers in Linguistics 13:117-158.
. 1982. Ellipsis in Japanese. Edmonton: Linguistic Research, Inc.
. 1983a. Case marking in Japanese. Linguistics 20:541-557.
. 1983b. Topic continuity in Japanese. Topic Continuity in Discourse: A
Quantitative Cross-Language Study, ed. by T. Givón, 43-93. Amster­
dam: John Benjamins Publishing Company.
1984. Retention of information using a Japanese style of presentation.
Studies in Language 8:45-69.
Hinds, John and Wako Hinds. 1979. Participant identification in Japanese
narrative discourse. Explorations in Linguistics: Papers in Honor of
Kazuko Inove, ed. by G. Bedell, E. Kobayashi, and M. Muraki. Tokyo:
Keukyuska, 201-212.
Jorden, Eleanor. 1963. Beginning Japanese. New Haven: Yale University
Press.
Kuno, Susumu. 1972. Functional sentence perspective — a case study from
Japanese and English. Linguistic Inquiry 3:269-320.
. 1973. The Structure of the Japanese Language. Cambridge: MIT Press.
Martin, Samuel. 1954. Essential Japanese. Rutland: Tuttle.
. 1975. A Reference Grammar of Japanese. New Haven: Yale Univer­
sity Press.
Maynard, Senko. 1981. The given/new distinction and the analysis of the
Japanese particles -wa and -ga. Papers in Linguistics 14:109-130.
. 1984. Thematization as a staging device in Japanese narrative. Paper
presented at the 1984 Annual Meeting of the Association for Asian
Studies, Washington, D.C. March 23-25, 1984.
Mikami, Akira. 1960. Zoo wa Hana ga Nagai. Tokyo: Kuroshio Shuppan.
. 1963. Nihongo no Ronri — Wa to Ga. Tokyo: Kuroshio Shuppan.
Norman, Donald and David Rumelhart. 1975. Explorations in Cognition.
San Francisco: W.H. Freeman.
Ono, Hideichi. 1973. Japanese Grammar. Tokyo: The Hokuseido Press.
Prince, Ellen. 1978. A comparison of WH-clefts and it-clefts in discourse.
Language 54:883-906.
106 JOHN HINDS

1981. Toward a taxonomy of given-new information. Radical Pragma­


tics, ed. by Peter Cole. New York: Academic Press, 223-255.
Rowe, H.M. 1982. An example of discourse analysis: A Japanese radio
news item. Journal of the Association of Teachers of Japanese 17:7-18.
Schank, Roger and Robert Abelson. 1977. Scripts, Plans, and Knowledge.
Norwood: Ablex Publishing Company.
Takemata, Kazuo. 1976. Genkoo Shippitsu Nyuumon [An Introduction to
Writing Manuscripts]. Natsumesha: Tokyo.
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA:
A STUDY OF JAPANESE SPOKEN EXPOSITORY DISCOURSE

SHOICHI IWASAKI
University of California, Los Angeles

0. INTRODUCTION
This chapter is intended to contribute to the understanding of the
Japanese particle wa by observing its behavior in Japanese spoken exposi­
tory discourse, specifically, descriptions of "apartment lay-outs". It has
become increasingly clear to discourse researchers that the behavior of wa
is different in different types of discourse. Yet the picture of wa that this
chapter presents is broad enough to represent many aspects of the distribu­
tion of this particle. I attempt to describe the particle from two different
perspectives: a description of the discourse pragmatic status of a noun
phrase 1 which is marked with wa and that of the main function that a wa
phrase displays in discourse.
Concerning the status of a noun phrase marked by wa, I will suggest
that we replace the categories "anaphoric" and "generic", which have been
ascribed to a wa marked noun phrase, with the notion "identifiable". As for
the discourse function of wa, I propose that "setting the scope for the pred­
ication" is wa's main function, and I also suggest that one of its derivative
functions can be understood as "indicating multiple predications" along
with the well known functions of indicating a "negative scope", and a "con­
trastive element".

1. DATA
The data used for this study consist of nine oral descriptions of lay-outs
of living quarters. Seven native speakers of Japanese gave descriptions of
their apartments, one her house, and one his dormitory room. The seven
apartment lay-out descriptions constitute the core data for the present dis­
cussion. Although the other two, the house and dormitory room descrip-
108 SHOICHI IWASAKI

tions, follow the same general pattern in terms of the distribution of wa,
they are excluded from the initial discussion since they deal with a slightly
different physical pattern from apartment type living quarters.
A peculiarity of expository discourse must be noted here. According to
Longacre (1976:199-204), expository discourse refers to discourse which is
impersonal and atemporal, and which contains an abundance of "equative
and descriptive" clauses. For Japanese, an equative or descriptive clause
usually takes the shape [ N P - W A + Predicate] [See Hinds (this volume)].
Thus, from the outset, we anticipate a predominance of this type of con­
struction in the data, and a concomitant lack of expectation for temporal
clauses.
Using these data I discuss first, in 2.1, the pragmatic characteristics of
noun phrases with wa, and then in 2.2, the discourse function of the particle
wa.

2. DISCUSSION
2.1. Pragmatic Status of the wa marked phrase
2.1.1. Identifiability
Kuno (1972, 1973) claims that the pragmatic status of the wa marked
phrase can be characterized as "anaphoric" or "generic". This section chal­
lenges this characterization, and suggests that it be replaced with the more
appropriate notion "identifiability".
That "anaphoricity" is not an appropriate term to describe wa marked
noun phrases is observed in the following example. (1) is taken from a TV
show, Shinkon-san Irasshai "Welcome, Newlyweds". The parentheses indi­
cate a segment which could not be transcribed accurately. After the hus­
band and wife announce their names, the male interviewer, M, continues.
F is the female interviewer.
(1) M: goshujin nakanaka kanroku ( ) kono megane ga
husband quite dignity these glasses SB
mata yoroshii naa
also good EX
F: uun soo desu nee
mm so COP EX
M: oshigoto wa?
job TP
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 109

M: You look like quite an important person. These glasses are


also nice.
F: Mmm. That's right.
M: What is your job?
In M's second utterance, oshigoto "job" is marked with wa although it
has not been mentioned in the preceding discourse (and so is not
"anaphoric"). Yet, crucially, it is understood without doubt to be the hus­
band's job that M is referring to (and so is therefore "identifiable"). 2 "Iden-
tifiability" can thus be defined as the speaker's assumption of the hearer's
ability to "pick out" the correct referent or concept that the speaker wishes
to communicate by uttering some linguistically coded form (See Chafe
1976:39).3
Although there are many instances in which a referent or concept
becomes "identifiable" because it is "anaphoric", we must recognize there
are some other means by which a referent or concept can become "identifi­
able", and consequently its linguistic representation becomes eligible for wa
marking. Chafe recognizes two such means besides "anaphoricity". They
are "extralinguistic environment" and "entailment of a concept". Wa mark­
ing through extralinguistic environment is observed in (2). K, looking at a
picture, utters kono oshashin wa "this picture" which is not anaphoric in the
given discourse.
(2) (Taken from a TV interview show, Tetsuko no heya "Tetsuko's
room.")
K: kono oshashin wa soshitara okaasama no ...
this picture TP then your mother GN
Then this picture is your mother's ...
Entailment of a concept leading to wa marking occurs very frequently
in the data dealing with apartment descriptions, and will subsequently be
illustrated. This concept is also referred to as a "frame" by Fillmore, 4 and
will be discussed in detail in 2.1.3.
Other concepts often cited as characteristic of wa marked noun
phrases, besides "anaphoric", are "generic" (e.g., "the whale [in general]")
and "unique" (e.g., "the sun", "my wife"). Notice that all of these concepts
can be understood as "identifiable". These concepts are in what Kuno
(1973) calls "registry". He assumes that a referent or concept is introduced
into the temporary section of the registry when it is mentioned for the first
time. From its second mention, the noun phrase that represents it is
110 SHOICHI IWASAKI

anaphoric and eligible for being marked with wa. We also understand,
though Kuno does not discusses this, that a referent and concept can be
registered in the temporary section through a "frame" in the sense of
Fillmore (see 2.1.3.).
Concepts of a generic and unique nature are stored in the perma­
nent section of the registry. They need not be introduced into discourse
before noun phrases representing them can take wa (Hinds 1976:29).
Unique referents are identifiable by definition. Generic concepts are also
identifiable in that they refer to a class which is discriminated from all other
classes of the same level of categorization (Rosch 1978). Note again that
"generic and unique", concepts whose identities can always be found out­
side the discourse (i.e., "homophoric" in Halliday and Hasan's term
(1976:71-3)), are very different in nature from "anaphoricity", a concept
operating discourse internally (i.e., Halliday and Hasan's "endophoric"
1976:33).
One final note about "identifiability" is in order. That is, this notion is
not relevant to nonreferential mentions, such as categorizing predicate
nominals and negative pronouns (Du Bois 1980:209-17). One case of non-
referential mention with wa is found in the present data, and this will sub­
sequently be discussed.
To summarize, in this section it is suggested that the most relevant
pragmatic concept that characterizes a noun phrase marked with wa is
"identifiability". The referent of such noun phrases must be identifiable
before they can take wa marking. The present data contain some noun
phrases with wa which are neither "anaphoric" nor "generic/unique" but
nevertheless are "identifiable" through a different source (i.e., "frame"). In
the following two sections, we will see exactly how a referent becomes iden­
tifiable when it is anaphoric and when it is not anaphoric. Since there are no
generic or unique references with wa in the data, the discussion will concen­
trate on non-generic, non-unique referents.

2.1.2. Identifiability through anaphoricity


Many noun phrases with wa in the data are anaphoric. We can make
further observations if we categorize wa marked phrases according to
their referents.
Table 1 lists all instances of wa phrases in the data. It shows which
phrases are anaphoric and which are not. In the case of anaphoric phrases,
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 111

Table 1. All instances of wa in the data

D
#of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses
I. MAJOR PARTS
a) Apartment Building
sono apaato wa 1 - 3
that apt.bldg. (1)
sono apaato tte 2 NP-wa 4
iu no wa (1)
thing called apt. bldg.
apaato no 1 3
birudingu wa (3)
apt. bldg.
ryoo wa 3 NP-ni 5
dormitory (4) NP in predicate
Hale Wainani wa 2 NP in modifier 1
(name of bldg.) (5)
b) Apartment Unit
watashі  2 [NP-o]
apaato wa (1)
my apt.
sono apaato ni wa 3 [NP-o]
in that apt. (1) NP-wa
watashi-tachi no 4 [NP-o] 0
heya ni wa (1) NP-wa (false start)
in our room NP-wa
watashi wa 2 2
I (=my room) (3) [NP-o]
apaato wa 2
apt. unit (3) NP-wa
sangai no heya wa 2
room on the NP-o
3rd floor (4)
heya wa 2 NP in predicate
room (5)
112 SHOICHI IWASAKI

D
#of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses

soko wa 2 [NP-o] 3
there (9)
(=apt. unit)
c) Living Room
ribingu ruumu 2 NP in predicate 3
tte iu no wa(1)
thing called
living room
ribingu ruumu 2 NP-ni 4
ni wa (2)
in the living room
ribingu ruumu wa 3 NP-ga 2
living room (3) NP in predicate
* ribingu ruumu 3 NP(no-)-ni 1
ni wa* (5) [NP-ni-wa]
in the living room
ribingu ruumu wa 3 NP-ga 2
living room (7) NP-ni
ribingu ruumu 2 NP(to-)-ga 1
ni wa (9) (post posed)
in the living room
d) Bedroom
shinshitsu wa 2 NP(to-) in pred. 2
bedroom (1)
beddoruumu ni wa 2 NP in predicate 1
in bedroom (2)
shinshitsu to 2 NP-ga 2
iu no wa(4)
thing called bedroom
futatsu no NP in predicate 1
beddoruumu wa (5) NP-ga
two bedrooms
beddoruumu wa NP-ga 0
bedroom (7) NP-ga (false start)
NP in modifier
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE W A 113

D
#of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses

beddoruumu wa 1 1
bedroom (9)
beddoruumu wa 5 NP-WA 1
bedroom (9) NP-gfl
NP-ga
NP-ga
e) Bathroom
basuruumu wa 2 NP(to-) in predicate 2
bathroom (1)
basuruumu (to-) 1 1
tte iu no wa
thing called bathroom
f) Kitchen
*kitchin ni wa* 4 NP-ga
in the kitchen (3) NP in predicate
[NP-ni-wa]
*kitchin wa* 3 NP(to-)-ga
kitchen (4) [NP-wa]
*kitchin ni wa* 6 NP(to-)-ga
in the kitchen (4) [NP-wa]
NP-wa
NP in predicate
[NP-ni-wa]
*kitchin wa* NP-gfl
kitchen (5) NP(no-)-ni
NP(no-)-ni
[NP-wa]
daidokoro ni wa 2
in the kitchen (7) NP-gfl
(-to) kitchin tte 2
iu no wa (9) NP-gfl
(-and) thing called
kitchen
114 SHOICHI IWASAKI

 D
# of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses

g) Upper Room
nikai ni wa 1 2
in the upper (4)
room
II. MINOR PARTS AND SUPER-STRUCTURE
h) Minor Parts
1) Wall
kabe ni wa 1 - 2
on the wall (2)
kabe ni wa 1 - 1
on the wall (4)
kabe ni wa 1 - 1
on the wall (7)
Table
teeburu wa 3 NP-to 2
table (3) NP(to-)-ga
The Rest
ato wa 1 - 1
rest (of wall) (2)
ato wa 1 2
rest (of things
in the kitchen) (3)
ato wa 1 - 1
rest (of bedroom) (5)
Deixis
koko ni wa 1 2
here (in the
living room) (1)
soko ni wa 2 NP-ga 1
there (shelf
in kitchen) (7)
sore wa 2 NP-ga 1
that(oven)(7)
sore wa 2 NP-ga 1
that (mat) (7)
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 115

D
# of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses

5) Others
zairyoo wa 1 - 1
material (1) - (post posed)
kabe to kauchi no 1 1
aida wa
between wall
and couch (3)
nagashi ni wa 1 1
at the sink (4)
shinku no ue no 1 1
tokoro wa
place above the sink (5)
isu wa 1 - 1
chair (5)
yuka ni wa 1 1
on the floor (7)
kagu wa 1 3
furniture (7)
i) Super-structure
zentai wa 1 1
whole
(apartment) (7)
heya-kazu wa 1 3
the number of
rooms (7)
III. REFERENTS WHICH DO NOT DIRECTLY BELONG TO APARTMENT
j) Residents
watashi-tachi wa 3 NP in modifier 3
we (1) NP in modifier
watashi wa 1 1
1(4)
watashi-tachi wa 1 1
we (7)
watashi-tachi
116 SHOICHI IWASAKI

D
# of Form of # of Post-
Referent NP-wa Referent Previous NP-wa
(Speaker) Mention Mention Clauses

no baai wa
our case (9)
watashi no
ruumumeeto wa NP-ga
my roommate (7)
kanojo wa NP-ga
she (7) NP-wa
NP in modifier
k) Contrastive Parts
hitotsu wa
one (7)
mieru tokoro wa
visible part (7)
ue no hoo wa
upper part (7)
1) Temporal Adverbials
 wa
previously (4)
ima wa
presently (4)
m) Others
yo-nin gurai
made wa
up to about
four people (1)
shita ni
aru no yori wa
than the one
below (4)
dainingu teeburu
to wa NP-ga
besides the NP(no-)-ni
dining table (5)
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 117

the form of the previous mention of the same referent and the number of
mention are indicated. In particular, Column A shows the referents of wa-
marked phrases, and Column  lists their actual surface form. (A wa
phrase between two asterisks in Column  means that this phrase is first
said by the interviewer, not by the interviewee). Column  shows how
many times the referent of the wa marked phrase has been mentioned. The
number 1 means that the wa marked phrase is the first instance in which the
referent is mentioned. The number 2 means that the referent is mentioned
for the second time in the wa phrase. Types of prior mention are shown in
Column D (A wa phrase between brackets in Column D indicates the inter­
viewer's utterance). Parentheses are employed in this column to abbreviate
the actual stretch of a longer noun phrase and so highlight the noun phrase
in question. That is, the phrase
(3) beddoruumu to kitchin ga arimasu
bedroom and kitchen SB exist
is abbreviated to NP (to-) ga arimasu to highlight the first noun phrase, i.e.,
bedroom, and "(-to) NP ga arimasu" to highlight the second noun phrase,
i.e., kitchen. The particles which appear in this column beside wa are: ga
(subject marker),  (direct object marker), ni (locative marker), and to
("and"). Column E shows how many clauses follow each instance of a wa
marked phrase. Discussion of this column is presented in 2.2.
Three major referent types are posited (I, II, and III). Each referent
type consists of a number of referent headings; seven headings for I (a-g),
two for II (h-i), and four for HI (j-m). Referent type I consists of the major
components of an apartment including the building itself. These are; (a)
apartment building, (b) apartment unit, (c) living room, (d) bedroom, (e)
bathroom, (f) kitchen, and (g) second floor. Referent type II consists of:
(h) minor parts (wall, table, sink, etc.) and (i) super-concepts (number of
rooms, and the shape of a whole apartment). Referent type III is a mixture
of referents which do not directly belong to an apartment: (j) people (resi­
dents), (k) temporal adverbials, (1) contrastive parts, and (m) comparative
phrases.
As shown in Table 1, most noun phrases of type I, i.e., major compo­
nents of the apartment, are anaphoric. A 2 or a larger number in Col­
umn  indicates that the noun phrase in question is anaphoric. Column D
shows the first mention. Table 2 is a summary of the first mentioned form.
118 SHOICHI IWASAKI

Table 2. Types of First Mention for the Referents in I in Table I


-ga (9), in Predicate (5), -o (4), -ni (3),
-wa (2), in Modifier (2), -to (1)

The predominant strategy for introducing referents into discourse in


the present data is through a noun phrase marked by the particle ga,
although a referents may also be introduced through a noun phrase marked
by other particles such as  or ni, or it may be introduced in the predicate
position. 5 In (4), we observe that the referent "bedroom" is introduced with
a ga marked noun phrase and it is immediately marked with wa in the fol­
lowing clause.
(4) 10: "two bedroom", "bedroom" ga futatsu
SB two
na n desu keredomo
LK NOM COP but
11: futatsu no "bedroom" wa, anoo, ribingu ruumu
two LK TP uhm living room
to kitchin no mukoogawa ni narandeimasu
and kitchen LK opposite LC is-lined-up
(Speaker 5)
"Two bedrooms." There are two bedrooms. Those two bed-
rooms are lined up at the opposite side of the kitchen.
It is not always the case, however, that the initial mention of a referent
with ga is followed by wa marking with no intervening clauses. Consider the
next text, where there is a six-clause distance between ga marking and wa
marking on the same referent.
(5) 2: sono juuik-kai no ano, yakyuujoo ni
that 11-story GN uhm baseball-field to
menshiteiru hoo no heya no haji no
is-faced side LK room LK end LK
heya ga watashitachi no heya na n desu
room SB we GN room LK NOM COP
IDENTIFI ABILITY, SCOPE-SETTING, AND THE PARTICLE WA 119

9: ano, heya wa yo-nin no onna-no-ko de


uhm, room TP 4-people LK girls with
"share" s hítete
doing:GER (Speaker 5)
Λ room at the end of the row of the rooms which face the
baseball field is our room.

The room is shared by four girls.


We have investigated the case of anaphoric noun phrases being marked
with wa. It is indeed the case for type I noun phrases, i.e., those referring
to the major components of apartments, that there is a strong tendency for
anaphoricity and wa marking to cooccur. Yet, as we will see in the next sec­
tion, anaphoricity in the strict sense is not always required for wa marking.

2.1.3. Identifiability through "frame"


This section will discuss the mechanism involved in making a non-
anaphoric noun phrase identifiable. The discussion will start with the non-
linguistic and linguistic contrast between the two referent types in Table 1.
Recall that referent type I consists of the major parts of an apartment and
II the minor parts and apartment super-structures. It will be argued that
although the noun phrases in the two groups are different on the level of
anaphoricity (i.e., most type I noun phrases are anaphoric but those in II
are not), they are both identifiable. This section will also briefly touch upon
the notion of "givenness" which tends to be confused with ''identifiability",
and show that "identifiability" is the suitable notion for wa marked noun
phrases.
Before considering the linguistic contrast between types I and II, we
must consider the noun phrase in (e) "bathroom" and (f) "kitchen" of
Table I since they are not brought up to talk about by the speakers as fre­
quently as the other referents are. 6 Only two instances of each referent are
expressed with wa without prompting. Nonetheless they can be grouped
together with other noun phrases in I based on our world knowledge of
apartments. Although it is beyond the scope of this chapter to speculate on
possible reasons for their secondary status, it is interesting to note that
some speakers (3, 4, and 5) used the word kitchin or daidokoro, both of
120 SHOICHI IWASAKI

which mean "kitchen", with other particles or in the predicate on several


occasions; yet they did not mark these words with wa until the interviewer
urged them to give more information about the referents.
At least two kinds of interrelated linguistic contrasts can be observed
for groups I and II. First of all, most of the noun phrases in II are intro­
duced as wa marked phrases while most of the noun phrases in I constitute
second or subsequent mention of a referent. Thus the "anaphoricity" we
observed to be necessary in type I is absent in most of the type II noun
phrases. Secondly, the correct interpretation of noun phrases in II depends
on local context, unlike referents of type I which do not need any local con­
text. For example, a type I nominal like "the living room" is interpreted as
the "living room of the apartment in question" and not a living room in
someone else's house unless some obvious comparison is being made. On
the other hand, a type II nominal like "the wall" does not tell the hearer
which walls in an apartment is meant because there are several walls in an
apartment. It will be shown shortly how a correct interpretation is achieved
in context.
Despite this difference between the noun phrases of types I and II, the
referents of noun phrases of both types are considered to be identifiable. It
is worthwhile considering here the possibility of employing the notion of
"givenness" instead of "identifiability" as a property of wa-marked noun
phrases. Chafe (1976:30) defines "given information" as follows:
Given information is that knowledge which the speaker assumes to be in
the consciousness of the addressee at the time of the utterance.
If a noun phrase is the second mention of a referent, the referent is
potentially identifiable. But the referent is not necessarily in the hearer's
consciousness at a particular moment of speech.
In fact, most of the noun phrases of type II are not "given" according
to the above definition. Consider the next text in which the speaker marks
yuka "floor" with wa. Note that it is impossible to assume that the hearer
has "the floor" in his/her consciousness when it is mentioned even though
yuka can be identified as the "floor of the living room".
(6) 16: sorekara ribingu ruumu wa, eeto,
and living room TP mmm
soo desu ne, "director chair" ga
"let me see" SB
hitotsu atte
one exist: GER
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 121

17: sorekara ato too no isu ga hitotsu arimasu


and then rattan LK chair SB one exist
18: chiisai isu desu kedo
small chair COP but
19: de yuka ni wa, anoo, nan te iu no kashira
and floor LC TP uhm what QT call NOM wonder
20: sore wa "Philippine' kara kita n ja nai ka
that TP from came NOM COP:NEG QU
to u n desu kedo
QT think NOM COP but
21: soo iu no ga shiiteatte
so call NOM SB is-spread:GER (Speaker 7)
The living room, uhm, let me see, has one director's chair
and also a rattan chair,
(it's) a small one.
And on the floor, uhm, what do you call it? a mat which
looks like a straw mat.
I think that came from the Philippines.
That is spread out.
From this example, it is clear that a wa marked noun phrase may be
termed "identifiable", but not "given11. That is, the speaker assumes that
the hearer can pick out the referent, which is the "floor in the living room"
in the above example, at the time of the utterance.
Now we can consider the "frame", the mechanism by which "identifia-
bility" of a non-anaphoric noun phrase can be achieved. It has been noticed
that some noun phrases in English can be marked with a definite article on
first mention contrary to the usual procedure of introducing a referent with
an indefinite article (e.g., Chirstophersen 1939, Chafe 1976, Fillmore 1975,
and DuBois 1980). The "front fender" in the next text is a case in point.
This is taken from DuBois (1980).
(7) He picks up a ... the whole basket of pears, ... and puts it on the
h a n d l c . n o on the ...front fender of his bike
This phenomenon can be described by Fillmore's notion of "frame"
(Fillmore 1975, 1977, 1982). To simplify Fillmore's claim somewhat, we
may say that people have conceptual schemata for categorizing actions,
institutions, and objects of real world "scenes". The term "frame" refers to
122 SHOICHI IWASAKI

the linguistic provisions which name and describe the categories and rela­
tions found in schemata. Once a particular frame is evoked by some linguis­
tic means, for example, by mentioning some nouns or verbs in a sentence,
the speaker and the hearer expect that certain relevant concepts and actions
can be used for further construction and interpretation of text (Fillmore
1977:127).
The Japanese equivalent to initial definite marking of noun phrases in
English is initial wa marking. The next text is an example from Hinds
(1976a:36).
(8) senshuu doitsu-sei no sugoi kuruma katta kedo
last-week German-made LK great car bought but
enjin wa nihon-sei datta yo.
engine TP Japan-made COP EX
Last week I bought a great German car, but the engine is made in
Japan.
The frame for this text, the "car" frame, is evoked by the first clause.
Since all cars include engines, the engine is marked with wa. If an item is
not in the "car1' frame (e.g., a wireless telephone as a special feature in a
car), it must be introduced into the discourse explicitly, for example with ga
before it can be marked with wa.
Apartment lay-out descriptions are based on an "apartment" frame.
Concepts which can be inferred from this frame, expressions such as sono
apaato "that apartment building", beddoruumu "bedroom", can be marked
with wa on first mention. We have observed earlier, however, that despite
the fact that people can do this, these noun phrases are typically introduced
through a noun phrase marked by other particles.
It is not clear exactly how a frame gets activated, and it is beyond the
scope of this paper to discuss this matter in detail. I will simply present sev­
eral cases to illustrate the complexities of frames. In some cases a noun
phrase with wa functions as a frame opener. That is, a noun phrase with wa
activates a frame, and consequently elements in that frame can be realized
on first occurrence as a noun phrase with wa. One such example is pre­
sented in (9).
(9) 16: sorekara ribingu ruumu wa, eeto
and living room TP umm
soo desu ne, "director chair" ga
"let me see" SB
IDENTIFI ABILITY, SCOPE-SETTING, AND THE PARTICLE WA 123

hitotsu atte
one exist:GER

22: teeburu wa
table TP
23: a, watashitachi wa binboo gakusei desu kara ne
oh we TP poor student COP so EX
24: kaenai node
cannot-buy so
25: koo hako o tsumikasanete
this way box DO pile-up:GER
26: soko ni koo mieru tokoro ni wa nihon
there LC this-way visible place LC TP Japan
no chiyogami o hatte
LK "chiyo" paper DO paste:GER
27: de, sono ue no hoo ni wa sudare o,
and that upper LK side LC TP bamboo-blind DO
sudare , koo oite
bamboo-blind DO this-way put:GER
28: teeburu ni shite
table RS make:GER
29: tsukatteiru n desu kedo ne
using NOM COP but EX
(Speaker 7)
And the living room, umm, let me see, has one director chair

The table.
Oh, since we are poor students,
We can't buy one, so
We put boxes on top of each other.
We pasted "chiyo" paper on the visible sides.
And we placed a bamboo blind on the upper side.
As a table,
we use it.
124 SHOICHI IW ASAKI

"Living room" in line 16 is an element of the "apartment" frame, and


it in turn establishes its own frame which contains a table (line 22) as an ele­
ment. "Table", in turn, opens up its own frame with the "visible and upper
sides" of it as elements. 7
When the "table" is mentioned with wa, it becomes possible for the
speaker to utilize a "table" frame. The speaker of (9) decided to use this
frame and marked two different phrases in the frame with wa (i.e., mieru
tokoro ni "visible part" and ue no hoo ni "upper part"). (These two phrases
and watashitachi wa in line 23 will be discussed again in the next section.)
It is, however, necessary to recognize that a frame can be activated in
a more indirect manner. The next text illustrates this point.
(10) 38: kochira gawa ga kabe, burokku no kabe
this side SB wall block LK wall
39: kochira gawa ga heya ga r  de ne
this side SB room SB exist NOM COP:GER EX
40: ki no kabe, ki de tsukutta
wood LK wall wood with made
41: nan te iu n deshoo, zairyoo wa
what QT call NOM COP:TENT material TP
42: anoo, burokku ja-nai zairyoo no, anoo,
uhm block -NEG material LK uhm
kabe ga aru n desu kedo
wall SB exist NOM COP but (Speaker 1)
This side is a wall, a wall made of blocks. Because there is a
room on this side, a wooden wall, a wall made of wood, how
do you call it? the material, uhm, a wall whose material isn't
block uhm, is there.
Zairyoo wa "the material" in line 41 is a noun phrase which has been
postposed from the initial position in the same line, and it clearly refers to
the material of the wall mentioned in line 40. However, there is no clear
frame opening before "the wall." Perhaps the repetition of the word kabe
wall" might help a hearer to secure the identity of the "material" together
with some world knowledge about the relationship between "wall" and
"material". Thus besides paying attention to the previous wa phrase, the
hearer must be attentive to the local context, i.e., before and after the wa
phrase in question.
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 125

Because of the complex nature of such frame establishment


mechanisms, difficulties in processing can occur, especially when one of the
previous wa phrases as well as the local context are possible determiners of
the wa phrase whose identity is at issue. The next text8 illustrates this.
(11) 1: mazu uchi wa toonan no
first-of-all house TP south-east LK
kado ni arimasu
corner LC exist
2: de, desu-kara, hiatari wa
and therefore exposure-to-the-sun TP
wariai ii desu
relatively good COP
3: soshite eru-ji gata ni natta uchi
and L-letter shape RS became house
na n desu
LK NOM COP
4: de, mannaka ni niwa ga arimashite
and middle LC yard SB exist:GER
5: de, watakushi wa eru-ji gata no
and I TP L-letter shape LK
choodo sono toonan no kado no
exactly that south-east LK corner LK
hoo ni, anoo, heya o motteimasu
side LC uhm room DO have
6: "share" shiteru mon desu kara
doing MON COP because
7: hanbun wa tomodachi no hoo to iu
half TP friend GN side QT say
koto na n desu
thing LK NOM COP
(Speaker 8)
First of all, the house is located on the corner facing south
east.
So we have a lot of exposure to the sun.
And it's an L-shaped house.
126 SHOICHI IWASAKI

And there is a yard in the middle,


and I have a room on the south east corner of the L-shape.
Since I share it,
half of it is my friend's.
At issue here is hanbun wa "half" in line 7 whose identity must be
found in order to understand this text properly. Notice that there are two
compelling candidates which could make hanbun "half" identifiable; uchi
"house" in line 1 and heya "room" in line 5. Uchi "house" is marked with
wa but is very distant from where hanbun "half" is uttered, heya "room" is
not marked with wa but is mentioned immediately before hanbun "half". In
this particular case, the first option is correct and hanbun means "half of the
house". 9 However there does not seem to be any easy way to make this
determination, and line 7 is therefore recognized as ambiguous.
In the present data, it cannot be determined which method (i.e., the
use or non-use of wa for frame opening) is preferred: among 17 non-
anaphoric noun phrases of type II, there are eight cases in which a preced­
ing wa opens a frame explicitly for them, and nine cases in which a preced­
ing wa is not used for frame opening. The eight cases of use of wa as a
frame opener further divide into two types. In six cases, immediately after
the frame is activated by a wa phrase, an element of the frame is introduced
with another wa phrase. In two cases, there is one or two intervening wa
phrases outside the frame just activated between the frame opening wa and
the wa phrase containing its element.
An interesting example of the force of the "frame" is observed in a
negative statement. Speaker 5 says (12) when she is describing the living
room.
(12) 23: nazeka isu wa nai n desu ga
somehow chair TP not-exist NOM COP but
Somehow there's no chair but
If there is no chair in the room why does she bother to bring it up?
Since a chair is a usual object in a living room, probably the speaker feels
the necessity to respond to an anticipated question from the hearer. [See
Tannen 1977].10
Thus, (11) provides a case of discourse motivation for a negative sen­
tence and of an example of a noun phrase that is nonreferential. In 2.1.1, it
was noted that "identifiability" is a notion irrelevant to nonreferential men­
tions. Therefore isu "chair" does not have to satisfy the "identifiability"
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 127

condition for wa marking.

2.1.4. Identifiability of type III phrase


Type III phrases are miscellaneous phrases which refer to neither
major nor minor components of an apartment. The four subgroupings of
this type are: (j) personal nouns, (k) contrastive parts, (1) temporal adver­
bials, and (m) contrastive adverbial phrases. The last subtype, (m), directs
us to an important characteristic of wa marked phrases, the "scope setting"
function of wa.
The referents in (j) through (1) are identifiable. The items in (j) are
personal nouns and pronouns. 11 First and second person pronouns are iden­
tifiable by definition. The one instance of a third person pronoun, kanojo
"she" in the data, is anaphoricaly identifiable as "the roommate". The con­
trastive parts in (k) are identifiable since the objects to which these parts
belong have been mentioned in the previous discourse. One case has
already been observed in (9). The other case appears in the following con­
text.
(14) 10: desukara, desukara, beddoruumu futatsu
so so bedroom two
arimasu kedo
exist but
11: hitotsu ookikute
one big:GER
12: hitotsu wa chiisai to iu koto na n desu
one TP small QT call thing LK NOM COP
(Speaker 7)
So, so, there are two bedrooms, but one is big,
and one is small.
Hitotsu "one" appears twice, in lines 11 and 12. Once it appears with
wa and once without. Each of these refers to one of the "two bedrooms"
mentioned in line 10.
Temporal adverbials, such as  "previously" and ima "presently"
listed in Table (1), are always exophorically identifiable in that they can be
located in the universe of discourse relative to the speech time. This
explains why they can take wa.
As we have repeatedly noted, wa can be used if a noun phrase is iden­
tifiable, but we must look into the motivation for the use of wa since obvi-
128 SHOICHI IWASAKI

ously not all identifiable noun phrases are marked with wa. The clue to the
speaker's motivation, or the function of wa, is found in the wa phrases in
(m). The items in (m) present an important characteristic of wa. These are
adverbial phrases of a special kind, i.e., adverbials of comparison which
indicate "range" (up to four people), and "points of reference" (compared
to the one on the first floor; from a dining table). These examples present
problems for identifying the function of wa as "thematic" or "contrastive".
These adverbial phrases cannot be "thematic" in any definition of the term.
They are not contrastive elements either in a strict sense since, while a part
of these adverbial phrases (e.g., shita ni aru no "the one in the downstairs")
might be called contrastive, the whole phrase (e.g., shita ni aru no yori
"rather than the one in the downstairs") cannot. I will suggest, then that the
unifying function of wa can be stated as "setting the scope for predication".

2.1.5. Summary.
This section has investigated the pragmatic properties of wa-marked
phrases and has suggested that "identifiability" should replace the notions
of "anaphoricity", "genericness", and "uniqueness" of a referent for a noun
phrase to be marked with wa. The identifiability condition is not relevant
for a nonreferential noun phrase. Since almost all noun phrases in the data
are specific, the mechanisms by which identifiability is established for such
noun phrase referents were considered: a noun phrase referent will poten­
tially become identifiable by its anaphoricity (i.e., the noun phrase referent
has been mentioned explicitly in discourse), or through a "frame" (i.e., a
noun phrase referent is an element in a frame which has been activated). In
some cases a frame is evoked by a wo-marked phrase, but at other times it
is evoked in a more subtle way. Finally the function of wa was hinted at
when dealing with some adverbial phrases of comparison.

2.2. Function of wa marked phrase


From the previous discussion, we have seen that noun phrases can be
marked with wa when they are characterized as identifiable noun phrases.
This is, however, only half the story. In this section the function of wa
marked phrases, or what they do in discourse, will be explored.
There is one main function of a wa marked phrase, the "scope-setting
function". This function specifies the domain in which the predication or
predications are applicable. The "scope-setting function" is present in every
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 129

case of a wa marked phrase and other functions are understood as deriva­


tive functions or specific realizations.12

2.2.1. The scope-setting function.


The scope setting function of wa has been noted by several Japanese
traditional grammarians [e.g., Yamada 1936:468, Sakuma 1940:211] though
it has never been considered as the unifying force for all derivative func­
tions. This concept is large enough to comprise such sub-functions as show­
ing negative scope (McGloin 1976, this volume) and contrastive elements
(e.g., Kuno 1973, McGloin op.cit.). It can also explain the items listed
under (m) of type III wa phrases. These are shown here with their predi­
cates:
(15) yo-nin gurai made wa sumeru raskii desu ne
4-people about up-to TP can-reside seem COP EX
It seems that up to four people can live there. (Speaker 1)
(16) nikai no sono shinshitsu to iu no wa anoo
2nd-floor LK that bedroom QT call NOM TP uhm
shita ni aru no yori wa daibu hirokute
downstairs LC exist one than TP fairly large:GER
The bedroom on the second floor is fairly spacious compared to
the one in the downstairs. (Speaker 4)
(17) daininguteeburu to wa betsu-ni
dining table from TP separate (Speaker 5)
Besides the dining table...
Before we discuss how these examples explain the scope-setting func­
tion of wa, we should note that these three examples are postpositional
phrases and that the identifiability condition imposed on a wa marked
phrase, as has been implicitly assumed, extends to postpositional phrases as
well as to a noun phrase. In the case of a postpositional phrase, the noun
phrase which is its constituent must be identifiable. Thus while (17) is possi­
ble, (17') is impossible.
(17') *aru daininguteeburu to wa betsu ni
certain
*Besides a certain dining table,...
Recall, however, the identifiability condition is not relevant to a non-
referential noun phrase. Thus, (15) is possible where yo-nin "four people"
130 SHOICHI IWASAKI

is nonreferential (i.e., it does not refer to any specific four people). The
function of scope setting, however, applies equally regardless of the status
of such noun phrases. Thus, the postpositional phrase in (15) indicates
"range" (up to four people), and those in (16) and (17) indicate "points of
reference" (compared to the one on the first floor; from a dining table),
which demarcate the boundaries for the following predication. (12) is
interesting, though not a special case at all, since it contains two instances
of wa phrases. It is understood that: "X is fairly spacious" is true only
within the specification made by the wa marked postpositional phrase,
"compared to the one in downstairs". "X is fairly spacious compared to the
one downstairs" is in turn understood to be true when it is applied to "the
second floor bedroom" which is marked with wa.
I have shown that the scope setting function of wa is useful in explain­
ing variously described functions such as markers for negative scope and
contrastive elements. This function also explains what Chafe (1976) calls
"Topic, Chinese style". He says, "the topic sets a spatial, temporal, or indi­
vidual framework within which the main predication holds."
We can modify this statement and define the "scope setting" function
of wa as follows.
The particle wa sets a scope (or demarcates a domain) to which a predica­
tion or predications are supplied.
This function cuts across the distinction between wa marked noun
phrases which are referential (identifiable or nonidentifiable) and nonrefe­
rential. It also applies to any elements besides noun phrases which come
before wa. Kuno 1973 notes that his "anaphoricity" and "genericness" con­
dition does not apply to a noun phrase that represents a contrastive ele­
ment. This observation can be rephrased in the present framework as fol­
lows. If a noun phrase is non-identifiable, it is unlikely to be predicated
because it is pragmatically undesirable to talk about an entity which a
hearer (and sometimes also the speaker) cannot identify. Therefore the so-
called "thematic" interpretation cannot be obtained for non-identifiable
noun phrases and the other option, the "contrastive" interpretation in
Kuno's framework, must be taken. What we are asserting then is that in
both cases wa sets the scope and whether or not the wa marked phrase is
interpreted as a "topic"-like entity depends on pragmatic considerations.
Once the scope is set, the predication is supplied. Yet the scope so set
might be filled with more than one predication. Using Mikami's locution,
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 131

"At least it is convenient if one is allowed to explicate in detail once the


theme (or our "scope") is presented (with more than one predicate)."
(1960:119) Thus we depart here from Chafe's "topic, Chinese style", which
is still a sentence level notion.

2.2.2. Multi-predication
Once we identify the scope-setting function of wa as the main function,
we can understand its derivative functions. We have noted already the
marking of negative scope, that of contrastive element, and that of point of
reference in a comparison clause, etc. These are all specific instances of
"scope setting".
The speaker can also utilize this function when, for any reason, he/she
wants to elaborate on some referent or concept. In such an instance, the wa
marked material must be a nominal on the higher scale of noun categorial-
ity which is defined as the linguistic material which refers to main partici­
pants of the discourse (Hopper and Thompson 1984). This excludes
nonidentifiable and nonreferential noun phrases. That is, one cannot elabo­
rate on material which does not refer to any object or whose identity is not
(yet) known. This mechanism of scope-setting of the main participants for
the purpose of predication is usually known as the "topic marking" or
"theme marking" function of wa. There is no objection to calling it such,
but it must be emphasized that this is just one of the derivative functions
which are made possible by the main function of "scope-setting". Topicali-
zation or thematization may be characterized with the notion of "target of
predication". If a clause starts with a wa-marked noun phrase which is iden­
tifiable, it is possible to continue talking about it.
Although wα-marked phrases in the present data are lower on the topi­
cality scale if topicality is understood as something that is higher on the ani-
macy hierarchy (Comrie 1981:191-2), or as a semantic agent (Givón
1984:139ff), some still secure the identity of referent as the target of predi­
cation with the force of setting the scope and are predicated with more
than one clause.13 This suggests that the notion topic must be understood as
a product of the process of the making of actual discourses.
While the scope-setting function is considered omnipresent for a wa
marked phrase, the function of multi-predications is certainly limited. The
appearance of multi-predications is largely determined by nonlinguistic
considerations such as the goal of the speaker who produces the discourse.
132 SHOICHI IWASAKI

Though this point is well beyond the scope of the present paper we will see
in what follows that the particle wa but not other particles exhibits the pos­
sibility of chaining of predicates.
Despite the low degree of topicality of wa phrases in the present data
according to the criteria mentioned earlier, some show long persistence in
the ensuing discourse. This is shown in Column E in Table 1. The numbers
in this column indicate how many predicates follow after a wa marked
phrase is mentioned.
Length of persistence of a wa marked phrase is measured as follows.14
When a wa marked phrase appears, its compatibility with the immediately
following clause is tested. If the wa phrase makes a coherent unit with the
immediately following clause, the wa phrase is said to persist to the next
clause, and these two clauses constitute a "joined unit". To show this oper­
ation schematically, see (18) and (19). (18) represents a linear order in
which clauses may appear, and (19) is an analysis of a joined unit.
(18) X-WA clause 1, clause 2, ... clause n.
(19) X-WA clause 1
X-WA clause 2

X-WA clause η (Χ remains constant.)


If clause 1 through n are successfully linked to X-WA, the string start­
ing with the X-WA and ending at the end of clause n is a joined unit.
An actual example is shown below. Speaker 3 started her description
with X wa followed by the predicate:
(20) 1: apaato no tatemonoWA yon-kai-date de
apt. LK building TP 4-floor-build COP:GER
The apartment building is a four-story building and ...
The clause which immediately follows (20) is:
(21) 2: ik-kaime ga paakingu ni natteimasu
1-floor SB parking RS has-become
The first level is for parking.
The noun phrase ik-kai-me "the first floor" in (21) is understood as the
first floor of the apartment building she mentioned in (20). Not only do we
understand this semantic relationship, but we also can test the wellformed-
IDENTIFI ABILITY, SCOPE-SETTING, AND THE PARTICLE WA 133

ness of the structure of (20) with the wa phrase of (20). Observe (22):
(22) 2': (apaato no tatemono WA) ik-kai-me ga paakingu ni
natteimasu
The first level of the apartment building is for parking.
The third clause which follows (20) also can take the same wa phrase as
shown in (23).
(23) 3': (apaato no tatemono WA) і-kai kara ue ga,
2-floor from up SB
anoo, juutaku ni natteimasu
uhm living-space RS has-become
The second floor of the apartment building and above are liv­
ing spaces.
Table 3 presents the arithmetic means of the number of clauses which
follow a wa phrase to form a joined unit. The actual number of both post
wa clauses and wa phrases are indicated in parentheses (there are 35 cases
of wa phrases and 81-post-wa clauses altogether in categories (a) through
(g))·

Table 3. Average Number of Post wa Clauses in Each Category of Referent


Referent Types Average # of Post wa Clauses Overall Average
I(a-g) 2.3 (81/35)
II(h-i) 1.4(28/20) 1.9(128/69)
III (j-m) 1.4(19/14)

It is clear from this table that a wa phrase tends to be followed by more


than a single clause. This tendency is higher for the noun phrases which rep­
resent referents in the initial "apartment" frame, i.e., I (a-g). This charac­
teristic of the wa phrase stands out when compared with the ga phrase since
a ga marked phrase is almost always followed by a single clause (141 out of
150 cases, or 94%). 15
Not only does the ga phrase have a high tendency for having a single
post clause, but it also has a tendency to have a very short unmodified pred­
icate. For example:
(24) 7: sorekara, daidokoro ga betsu-ni atte
and kitchen SB seperately exist:GER
134 SHOICHI IWASAKI

8: sorekara, anoo, ribingu ruumu ga betsu desu


and uhm living room SB separate COP
9: doa o haitte
door DO enter:GER
10: soko ga ribingu ruumu ne
there SB living room EX (Speaker 3)
and there is a separate kitchen
and, uhm, the living room is also separated.
you go through the door
and there you are in the living room, okay?
Since the wa phrase can be followed by a longer stretch of clauses after
it, a ga phrase may appear with its predicate as a part of the joined clauses.
(25) is an example.
(25) 8: san-gai no heya wa naka ga mata "two floor" ni
3-floor LK room TP inside SB also RS
wakareteite
is-separated:GER
9: de, anoo, "kitchen" to sorekara "dining room"
and uhm and and
mitai na chiisai no ga atte
like LK small one SB exist:GER
10: sore-kara, basuruumu to senmenjo mitai no ga
and bathroom and washroom like one SB
tsuitete
attached:GER (Speaker 4)
The rooms on the 3rd floor have split levels inside
and (they) have a kitchen and a small room like a dinin-
groom
and (they) have bathrooms and washroom type of things.
In line 8 "the rooms on the third floor" is presented and followed by
three clauses each of which contain a ga phrase.
So far, we have only considered the cases of wa and ga in terms of
multi-predication in the joined unit. We must also consider phrases with
other particles to show this unique function of wa phrases. The three cases
that will be examined below are phrases marked by , , and zero.
IDRNTIFI ABILITY, SCOPE-SETTING, AND THE PARTICLE WA 135

The particle  usually marks the direct object in a transitive construc­


tion. The particle mo, which is similar to wa in some respect (e.g., a noun
phrase referent with mo must be identifiable, it absorbs ga and o. [see also
Martin 1975:66-9]), adds the meaning of "also" to the preceding element. A
bare noun phrase, i.e., a noun phrase marked with zero particle, is only
looked at if the missing particle is understood to be either wa, ga, o, or mo.
That is, bare noun phrases that appear in the oblique case or as predicate
nominals are excluded.
To facilitate comparison, I will repeat the relevant information about
the wa phrase from Table 3.

Table 4. Mean Length of Joined Unit for Different Particles


Noun Phrase # of predicate within Mean length of
Type a joined unit joined unit*
0 1 2 3 4 more
NP-wa 2 37 13 8 7 2 1.9(128/69)
NP-ga 6 141 3 - - - 1.0(147/150)
NP-o 3 31 - - - - 0.9(31/34)
NP-mo 1 5 - - - - 0.8(5/6)
NP-0 5 4 - - - 1 1.3(13/10)

*The mean length of a joined unit is calculated by dividing the total number
of post predicates for each noun phrase type (indicated by the first figure in
parentheses) by the total number of noun phrases with the particle in ques­
tion (indicated by the second figure in parentheses).

It is clear from this table that the wa marked noun phrase and a bare
noun phrase which includes wa ellipsis are the only cases in which
multi-predications are observed. Thus I conclude that it is wďs distinct
function to organize a bigger chunk of discourse while other particles
only organize a single clause unit. It is interesting that mo does not
appear with multi-predication although its affinity to wa has been suggested
earlier.

2.2.3. Summary.
In this section the function of wa and its effect in the discourse was
considered. The main function of wa is to "set the scope" for predication.
136 SHOICHI IWASAKI

This function is present in every case of a wa phrase whether the noun is


referential (identifiable or nonidentifiable) or non-referential, or whether
the phrase is a noun phrase or an adverbial phrase. All other functions such
as marking of topic, of contrasting elements, and of negative scope can be
seen as derivative functions of the "scope setting" function.
Also discussed was the continuous nature of wa phrases in discourse.
Once a scope is set for predication by wa, it can be followed by multiple
clauses. This is a significant function of wa since a phrase marked with ga or
any other particle does not persist in discourse.

3. CONCLUSION
In this paper instances of wa phrases in an actual discourse were
examined. This approach has certain advantages and shortcomings. One
shortcoming is the potential lack of generality. Since this study concerned
apartment lay-out descriptions, the findings might be challenged by
research on other types of discourse. That is, it might be shown that the
findings in this paper may be generalized across different genres of dis­
course or that they are unique to expository discourse. The main advantage
of this approach is that it avoids any preconception that a researcher might
have. This is very important especially when dealing with such a controver­
sial linguistic phenomenon such as the particle wa. If one starts with the
notion of "topic" or "theme" associated with this particle, he/she is in great
danger of losing sight of other functions that the particle might have, and
consequently miss the generalizations possible among these functions.
The function of "scope-setting", which I suggested as wa's main func­
tion, has already been insightfully identified by Sakuma (1940:210-11).
However, it is a mistake to equate jojutsu no hanì no kakutei "demarcation
of scope for statement" with daimoku no settei "topic setting" because
"topic setting" is but one realization of the "demarcation of scope for state­
ment". The scope might be demarcated for other purposes such as to indi­
cate the range to which negation or comparison is applied.
Only after we recognize that "topic" or "theme" marking is one of the
realizations of wa's main function can we profitably investigate the nature
of "topic" or "theme" in Japanese. Here I only indicate one possible direc­
tion that we might take for better understanding of "topic" or "theme" in
Japanese.
A possible parameter that wa-marked phrases can be categorized with is
that of "aboutness". When wa is used for marking the scope of negation,
IDENTIFI ABILITY, SCOPE-SETTING, A N D T H E P A R T I C L E WA Yìl

there is clearly no notion of "aboutness" for the wα-marked phrase. The


notion "aboutness" is obtained more easily if the wa-marked phrase is more
concrete, more definite, or involves higher categorial nouns as Hopper and
Thompson (1984) call them. Such noun phrases are usually more important
participants in a given discourse; thus they are usually humans but do not
necessarily have to be as in the discourse type we used in this paper. The
"topic" or "theme" is clearly a discourse notion that the speaker brings
about according to his/her need in the creation of discourse.

NOTES
1) Linguistic materials that can be marked with wa include noun phrases, postpositional
phrases, adverbial phrases, adjectives, and verbs. In the data examined for this chapter noun
phrases and locative postpositional phrases with ni appear quite frequently due to the nature of
the discourse.
2) Prince (1981) calls this type of discourse entity ''brand-new unanchored." See also Hinds
(this volume) for a discussion of this type of information status.
3) "Identifiabilit" is an alternative term for "definiteness", Chafe (1976) considers "iden-
tifiability" to be a more appropriate label for the concept, but continues to use "definiteness''
since it is the standard term in the linguistic literature. I chose the term '"identifiable" because of
its mnemonic nature.
4) Similar proposals are made by researchers in cognitive psychology (e.g., Minsky 1975) and
artificial intelligence (Norman 1973, Schank and Abelson 1977, etc.).
5) If a referent is mentioned more than twice before a wu-marked phrase, only the very first
mention is counted since our intention here is to see what strategies are available to introduce a
referent into discourse and not to see how an introduced referent changes before it finally
reaches the given wa phrase. However, the situation is more complicated than this since it is not
always the case that the first mention of a referent triggers the marking of wa on a noun phrase
with the same referent later in discourse when there are one or more intervening mentions of the
referent between the first and the latter mention with wa. If the progression is as follows, for
example:
(Referent.) -ga ... (Referent.) -ga ...
(Referent.) -ni ... (Referent ) -wa ...
It cannot be determined a priori which instance of referent mention is actually responsible
for the final wa marking. Sometimes it seems that an accumulation of mentions finally triggers
wa. However, this is a subjective observation and cannot be documented here. Thus I am con­
tent with the decision to count the initial mentions only. Also, if a speaker uses wa for the same
referent more than once, only one such instance is counted in order to avoid counting the same
initial mention more than once.
6) The status of the referent in (g), "the upper split level" in an apartment, is even more
peripheral since it is not a regular feature of an apartment. Thus, the speaker must explicitly
mention this special feature of the apartment before she marks the noun phrase representing it
138 S H O I C H I IW ASAKI

with wa (see (25), line 8). Thus, nikai ni wa "on the second floor" is very different form the rest
of referents in I, and it could be argued that it belongs to II. Yet since "the upper level" is a
larger spatial unit than those in II, I include it in group I.
7) As it is presented, it could be argued that "director chair" has activated the "table" frame
since there is a strong association between the two concepts. However, between lines 16 and 22,
not shown here, there are two noun phrases that are marked with wa, "floor" and "mat". There­
fore, I take the larger frame of "living room" rather than that of "chair" as the frame activator.
8) The wa phrases in this text are not included in Table 1 since it deals with a "house" frame
which is slightly different from an "apartment" frame. It is interesting to note that hiatari "expo­
sure to the sun" in line 2 is marked with wa on its first mention. This is possible because "how
much light a house can get" is one of the most important concerns for Japanese when they
choose a house.
9) This was confirmed in a follow-up interview with Speaker 8.
10) A similar example can be observed in the next text which is the opening of an interview
session. This is taken from Hinds 1982.
(i) Α-l: , mazu anoo, onamae kara,
um first uhm name from
okikishitai n desu kedomo ne
want-to-ask NOM COP but EX
B-2: yamauchi sumi desu
COP
A-3: yamauchi sumi san ne
Ms EX
A-4: , nenrei wa okiki shinai tte iu
um age TP ask NEG QT say
yakusoku na n de
promise LK NOM COP:GER
A-l: Um, first I"d like to start with your name
B-2: It's Sumi Yamauchi.
A-3: Ms. Sumi Yamauchi, right?
A-4: Well, I promised I wouldn't ask your age so...
Since information about age is considered to be essential in this type of interview, the inter­
viewer must feel it necessary to indicate that he has not forgotten about this important informa­
tion. It should be noted, however, that nenrei "age" is clearly identifiable, that is, it refers to the
interviewee's age.
11) Personal pronouns cannot be easily distinguished from nouns grammatically in Japanese
since they share many of the same structural properties. For example:
Modification: (kono oloko vs. kono watashi)
this man this I
Genitive: (sensei no hon vs. watashi no hon)
teacher GN book I GN book
Plural: (Yamada-san-tachi vs. watashi-tachi)
Mr. and others I and others
I D E N T I F I A B I L I T Y , SCOPE-SETTING, A N D T H E P A R T I C L E WA 139

Yet functionally, nouns are different from pronouns in that the latter are "shifters" (Jakobsen
1971).
12) In Iwasaki 1984, I stated that wa functions to "limit the scope for a predication" and to
"signal multi-predication". But now I consider "scope-limiting" to be the main function and
"signaling of multi-predication" as one of its derivative functions.
13) Securing the identity of a referent and predicating that identity can be stated in terms of
"aboutness". That is, the predicate is "about" the referent which is represented with a noun with
wa. In this way the present study supports with empirical data Kuno's assumption concerning
the "aboutness" of wa. For some readers, it might be easier to accept the term, "aboutness", for
a wa-marked item in a nominal clause, i.e., a clause which ends with a nominal and copula as
(ii), than in a verbal clause, i.e., one which ends with a verb as (iii). Thus between "John's" in
(ii) and (iii) below, the first "John" is easier to be understand as the topic/theme, i.e, this sen­
tence is "about" "John". This is due to the fact that the second "John" is the agent and this
semantic role is competing with the functional notion of topic/theme. Yet I still interpret both
instances of "John" as the functional target of the predication.
(ii) jon wa gakusei da.
John TP student COP
John is a student.
(iii) jon wa hon  yondeiru.
John TP book DO reading
John is reading a book.
14) My use of the term "persistence of a referent in discourse" is slightly differently than
Givón 1983:15 who measures persistence of a topic/participant in terms of "an uninterrupted
presence as a semantic argument of the clause, an argument of whatever role and marked by
whatever grammatical means." "Persistence" as used here is closely related to Mikami's notion
of "period crossing" and its elaboration by Kuno (1979).
15) Post ga clauses are counted in the same way as post wa clauses. That is, when a ga phrase
appears, it is added to the next clause to see if the string of words make sense. If it does, these
two clauses make up a joined unit. This precludes the possibility of a joined unit when the first
clause contains a ga phrase and the second contains the same phrase with any particle. Most cru­
cially when two clauses have noun phrases with the same referent and they are marked with ga
in the first clause and with wa in the next, they do not constitute a joined unit but are two sepa­
rate clauses. Even if the other option is taken (i.e., if we continue to count clauses when the sec­
ond clause has an overt noun phrase which has the same referent of the ga marked phrase), the
result will not be affected much since out of 150 cases of ga phrases, there are only three cases
in which a ga phrase is immediately followed by another coreferential noun phrase in the
immediately following clause.

REFERENCES

Chafe, Wallace L. 1976. "Givenness, contrastiveness, definiteness, sub­


jects, topics, and point of view." Subject and topic ed. by Charles Li, 27-
55. New York: Academic Press.
140 SHOICHI IWASAKI

Comrie, Bernard. 1981. Language universais and linguistic typology.


Chicago: The University of Chicago Press.
Christophersen, P. 1939. The articles: a study of their theory and use in Eng­
lish. Copenhagen: Einar Munksgaard.
Du Bois, John. 1980. Beyond definiteness: The trace of identity in dis­
course. The pear stories, ed. by Wallace Chafe, 203-74, Norwood: Ablex
Publishing Company.
Fillmore, Charles J. 1975. "Topics in lexical semantics." Current issues in
linguistic theory, ed. by Roger Cole, 76-138. Bloomington: Indiana Uni­
versity Press.
. 1977. "An alternative to checklist theories of meaning. BLS 1, 123-131.
. 1982. Frame semantics. Seoul: Hanshin Publishing Company.
Givón, T. 1983. "Topic continuity in discourse: An introduction." Topic
continuity in discourse, ed. by T. Givón, (= Typological studies in lan­
guage, 1), 1-41. Amsterdam: John Benjamins.
. 1984. Syntax: A functional-typological introduction. Vol.1. Amster­
dam: John Benjamins.
Halliday, M.A.K. and Ruqaiya Hasan. 1976. Cohesion in English. London:
Longman.
Hinds, John. 1976. Aspects of Japanese discourse structure. Tokyo:
Kaitaku-sha.
. 1982. Ellipsis in Japanese discourse. Alberta: Linguistic Research Inc.
. this volume. "Thematization, assumed familiarity, staging, and syntac­
tic binding in Japanese.
Hopper, Paul J. and Sandra A. Thompson. 1984. "The discourse basis for
lexical categories, in universal grammar." Language 60, 703-52.
Iwasaki, Shoichi. 1984. Semantic characteristics and discourse functions of
noun phrase-wa: A study of Japanese spoken expository discourse. Mas­
ter's thesis. University of California at Los Angeles.
Jakobson, Roman. (1957) 1971. "Shifters, verbal categories, and the Rus­
sian verb." Selected writings. Vol.2:130-147. The Hague: Mouton.
Kuno, Susumu. 1972. "Functional sentence perspective: A case study from
Japanese and English." Linguistic Inquiry 3(3),269-320.
. 1973. The structure of the Japanese language. Cambridge: MIT Press.
. 1978. Danwa no bunpoo. Tokyo: Taishuu Kan.
Longacre, R.E. 1976. An anatomy of speech notions. Lisse: The Peter de
Ridder Press.
Martin, Samuel. 1975. A reference grammar of Japanese. New Haven and
London: Yale University Press.
IDENTIFIABILITY, SCOPE-SETTING, AND THE PARTICLE WA 141

McGloin, Naomi Hanaoka. 1976. "Negation." Syntax and Semantics, vol.5,


ed. by Shibatani Masayoshi, 371-419 New York: Academic Press.
. this volume. "The role of wa in negation".
Mikami, Akira. 1960.Zoo wa hana ga nagai. Tokyo: Kuroshio Shuppan.
Minsky, M. 1975. A framework for representing knowledge. The psychology
of computer vision, ed. by P. Winston, 211-277. New York McGraw-Hill.
Norman, Donald. 1973. "Memory, knowledge, and the answering of ques­
tions." Contemporary issues in cognitive psychology: The Loyola sym­
posium. 136-65. Washington D.C.: V.W. Winston.
Prince, Ellen. 1981. "Toward a taxonomy of given-new information." Rad­
ical Pragmatics, ed. by Cole, Peter. 223-255 New York: Academic Press.
Rosch, Eleanor. 1978. Principles of categorization. Hillsdale: Lawrence
Erlbaum Associates.
Sakuma, є. 1940. Gendai nihongohoo no uu "A study of mod­
ern Japanese grammar". Tokyo: Koseikaku.
Shank, R. and R. Abelson. 1977. Script, plans and knowledge. Norwood:
Ablex.
Tannen, Deborah. 1977. "Well what did you expect?." BLS 3, 506-515.
Yamada, Yoshio. 1936. Nihon bunpoo-gaku gairon "An outline of
Japanese grammar". Tokyo: Hobun-kan.
A STUDY OF THE SO-CALLED TOPIC WA IN PASSAGES
FROM TOLSTOI, LAWRENCE, AND FAULKNER
(OF COURSE, IN JAPANESE TRANSLATION)

S.-Y. KURODA
University of California, San Diego

The following paper, except for minor additions and stylistic improve­
ments, was written about ten years ago, but has never been published. The
purpose of the paper was, as mentioned in the first section, to present a
case study, on the basis of a very limited, but hence well-defined scope, to
show that the so-called topic wa functions to represent a "double"-judg-
ment. The notion of "double"-judgment is derived from the Brentano-
Marty theory of judgment.
I now believe that the Brentano-Marty conception of double-judg-
ment, which underlay the original paper, must be modified for our pur­
pose. Brentano and Marty claimed that a double judgment (a categorical
judgment) involves the recognition of the subject, which is a thetic judg­
ment, but I now maintain that this analysis is not sufficient to characterize
the categorical judgment. What is essential for the categorical judgment is
the apprehension of substance. I presented this modifed view at the col­
loquium on Anton Marty's philosophy and theory of language held at
Freiburg, Switzerland, in December, 1985.
The analysis presented below in fact fits better with this new interpre­
tation of the notion of double judgment (or, categorical judgment) and I
would like to refer the reader to my colloquium paper, "The categorical
and the thetic judgment reconsidered," which will appear in K. Mulligan
ed. Mind, Meaning and Metaphysics: the Philosophy and Theory of Lan­
guage of Anton Marty.*

1. In lieu of an introduction and a conclusion.


Let me only note here that in previous articles1 I compared the distinc­
tion between Japanese sentences with the so-called topic wa and those
without it with the distinction between the categorical and the thetic judg­
ment in the Brentano-Marty theory of judgment and grammar. What they
talked about in terms of a distinction in judgments, I would regard as a
144 S.-Y. KURODA

distinction in meaning or semantic structure. Let us recall that Brentano


considered a categorical judgment as a "double-judgment", consisting of a
recognition of the subject, and a judgment that relates the subject to the
predicate. My intention in this paper is to try to justify certain uses of wa by
this characteristic of a "double"-judgment, although certainly not exactly in
the same sense as Brentano and Marty. I will take this to be a "semantic"
characterization. It is another matter how this semantic characteristic helps
explain pragmatic or discourse properties of wa, or vice versa.

2. In the following sections I shall study the stylistic effect of wa, or that
of the lack of it, in the subject position of a quotative verb in passages from
Tolstoi, Lawrence, and Faulkner in Japanese translation. By a quotative
verb I mean here a verb whose object complement is a direct quotation.
Since I assume most readers are not familiar with Japanese I will not
quote our examples in Japanese. I will put in square brackets those subjects
of quotative verbs that are translated by wa-phrases in our Japanese texts
and in braces those subjects that are translated by ga-phrases. For example,
if we have '"—," said [John]' in our English texts, we have '"—," to John-
wa itta as the Japanese translation, and if we have '"—," said {John}' we
have '"—," to John ga itta' in Japenese translation.
Let it be remembered that the interpretations of the texts given below
are interpretations of the Japanese texts I use. Whether they are adequate
interpretations of the original texts, or to what extent and in what sense
they are so, is not our present concern.

3. The first passage we analyze is from Leo Tolstoi's War and Peace, Book
II, Chapter 15. Nicolai Rostov, staying in Moscow on leave from his regi­
ment is drawn one day into a card game with Dolokhov, who has recently
proposed marriage to Sonya, Nicolai's cousin, and has been rejected.
Nicolai is losing 43,000 rubles. Sonya loves Nicolai, and he knows it; he
doesn't know how he feels about her. In his desperate mood he has decided
to play until the score reached forty-three thousand, forty-three being the
sum of his and Sonya's ages. The following scene is when Nicolai, fallen to
an unwonted bottom of gloom from a joyous life in Moscow, comes home
to find the usual happy gathering of the family.
To say "Tomorrow" and maintain a dignified tone was not difficult,
but to go home alone, to see his sisters, brother, mother, and father, to
confess and ask for money he had no right to after giving his word of
honor, was terrible.
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 145

At home they had not yet gone to bed. The young people, after com­
ing home from the theater, had had supper and were grouped around the
clavichord. As soon as Nikolai entered the ballroom, he was enveloped in
that poetic atmosphere of love that prevailed in the house that winter and
now, after Dolokhov's proposal and Vogeľs ball, seemed to have grown
heavier around Sonya and Natasha, like the air before a thunderstorm.
Sonya and Natasha, looking pretty and conscious of it, in the light blue
dresses they had worn to the theater, were standing by the clavichord,
happy and smiling. Vera was playing chess with Shinshin in the drawing
room. The old Countess, waiting for her son and her husband to come
home, was playing patience with an old gentlewoman who lived in their
house. Denisov, with sparking eyes and ruffled hair, sat at the clavichord,
one leg flung out behind him, striking chords with his short fingers and rol­
ling his eyes as he sang in a small, husky, but true voice, a poem of his own
composition called "The Enchantress," to which he was trying to fit music.
"Oh, tell me, enchantwess, what power is this,
Dwawing me to my forsaken lyre?
My fingers stwumming the stwings in bliss,
Setting my heart on fire"
he sang in passionate tones, his black agate eyes flashing at the frightened
but delighted Natasha.
"Beautiful! Wonderful!" cried [Natasha].
"Another verse," [she] said, not noticing Nikolai.
"Everything's still the same with them," thought [Nikolai], glancing
into the drawing room, where he saw Vera and his mother and the old
lady.
"Ah! Here's Nikolenka!" exclaimed [Natasha], running to him.
"Is Papa home?" [he] asked.
"I'm so glad you've come!" cried [Natasha], not answering his ques­
tion. "We're having such fun!
Vasily Dmitrich is staying on another day for my sake!
Did you know!"
"No, Papa is not back yet," said {Sonya}.
"Koko, are you back? Come here, darling," called the old {Countess}
from the drawing room.
Nikolai went to his mother, kissed her hand, and without saying a
word sat down at her table and watched her hands as she laid out the cards.
From the ballroom came the sound of laughter and merry voices trying to
persuade Natasha to sing.
"All wight, all wight!" cried [Denisov]. "It's no good making excuses
now! It's your turn to sing the Barcawolle — I entweat you!"
The Countess glanced at her silent son.
"What is the matter?" [she] asked.
"Oh, nothing," [he] replied, as though sick of being continually asked
the same question. "Will Papa be home soon?"
"I expect so."
146 S.-Y. KURODA

"Everything's the same with them. They don't know! What am I to do


with myself?" thought [Nikolai], and he got up and went back to the ball­
room. (Translated by Ann Dunnigan, New American Library)
In our text of the Japanese translation (M. Yonekawa's Iwanami
Bunko edition) the subjects of quotative verbs are all wa-phrases, except
for two, the braced {Sonya} and {Countess} in the text. What are the stylis­
tic effects of this contrast?

3.1. We may assume that the quoted passage as a whole represents


Nicolai's point of view. It is a description of the scene in the ballroom and
the drawing room as reflected in Nicolai's consciousness when he entered
the ballroom and proceeded to the drawing room. As I interpret this pas­
sage from the Japanese translation, Nicolai's consciousness was centered
around Natasha as he walked into the ballroom. Nicolai heard Natasha's
voice "Beautiful! Wonderful!" and the ensuing utterances of hers (up to "—
Did you know?") as his attention came to focus more and more on her, ask­
ing her a question and listening to her nonreply to it. But then came
Sonya's voice — "No, Papa is not back yet" — into Nicolai's consciousness
out of the field of his attention, which was followed by the voice of the old
Countess from the drawing room. Then Nicolai went to his mother and
watched her hands when he heard the merry sound from the ballroom. His
attention was directed back to the ballroom. He saw Denisov as he cried
"all wight--." Then as his attention was drawn back to his mother, she asked
him "What is the matter?"

3.2. Here we should distinguish three different categories of wa-phrases as


the subjects of quotative verbs in our text. First, there are two occurrences
of Nicolai as the subject of the verb think. These are two sentences where
it is lexically explicit that the passage represents Nicolai's point of view. In
cases like this let us say that a wa- phrase is the subject of an inner process
verb.
Next, there are two instances in which the subject of a quotative verb
is Nicolai, or more exactly in our text, the pronoun he referring to Nicolai.
('"Is Papa home?" [he] asked' and '"Oh, nothing," [he] replied —') For
these sentences there are no grammatical clues that they represent Nicolai's
point of view. It is a matter of literary (or stylistic) interpretation whether
we take these as representing Nicolai's point of view.
I assume, to repeat, that the whole passage we are analyzing represents
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 147

Nicolai's point of view. (I grant, however, the possibility that the sentence
T h e Countess glanced at her silent son' can be interpreted as representing
the Countess' point of view, exclusively, or in addition to Nicolai's.) This is
the simplest and most natural point of view structure we can impose on the
text. In the text there is no evidence, grammatical or otherwise, that forces
us to assume that the two sentences with quotative verbs represent, for
example, Natasha's point of view, i.e., a description of the verbal events as
they were reflected on Natasha's consciousness. Nor is it necessary to inter­
pret these sentences as a description given independently of the points of
view of all the protagonists, in particular, of Nicolai's (i.e., description from
the narrator's point of view, if you are willing to use such terminology).
There is no reason to assume that the description of Nicolai's stream of con­
sciousness is interrupted by these sentences. So long as they are interpreted
as representing Nicolai's point of view, then, the subjects of these sentences
(i.e., the subjects of the main verbs, ask and reply) are coreferential with
the subject of the point of view that the sentences represent.
Let us agree to say that a sentence (or, the subject of the main verb) is
reflexive of the point of view if the subject of the sentence is coreferential
with the subject of the point of view that the sentence represents.
The above two cases are, then, those where the subject of a sentence
with a quotative verb is (interpreted as) reflexive of the point of view. The
difference between them is that in the first case the main verb is a verb of
inner process like think, while in the latter the sentence in question repre­
sents an external event.
We shall postpone further discussion of these two cases; let us now
consider the remaining case. All the remaining instances of quotative verbs
in our text are such that the subject of a quotative verb is not Nicolai, that
is, reflexive of the point of view. For the moment, then, we consider only
these instances of wa-phrases nonreflexive of the point of view.

3.3. From the interpretive analysis of the text given earlier, we may assume
that in our text if a quotative verb is accompanied by a wa-phrase nonrefle-
xive of the point of view, the subject of the verb (or, more exactly, its refe­
rent) is an independent element of the content of Nicolai's consciousness,
independent of the event described by the quotative sentence itself.
For example, consider the sentence '"Beautiful! Wonderful!" cried
[Natasha].' There took place an event, Natasha's crying "Beautiful! Won­
derful!" Natasha is a constituent of this event. This event was recorded in
148 S.-Y. KURODA

Nicolai's consciousness; the image of this event, or the understanding of


this event, was an element in Nicolai's stream of consciousness. To the
image, or the understanding, of this event corresponds a meaning which
may be represented by a two-place predicate cry, whose first argument is
Natasha (or, more exactly, a term standing for Natasha). Natasha (or the
term standing for her) was contained as an element of this meaning; she (or
her image, notion) was an element of this image, or this understanding, of
the event. As such Natasha was an element in Nicolai's stream of conscious­
ness. However, when this event took place, Natasha (or her image, notion)
had existed, and continued to exist, in Nicolai's consciousness as a center of
his attention, independently of the occurrence of this event, or indepen­
dently of the creation of the image, or the understanding, of this event in
Nicolai's stream of consciousness. The sentence '"Beautiful! Wonderful!"
cried [Natasha]' with wa attached to Natasha corresponds to this state of
Nicolai's consciousness in which the image (or notion) of Natasha played a
double role.
In contrast, in the case where the subject of a quotative verb is accom­
panied by ga, as in '"No, Papa is not back yet," said {Sonya},' the image or
notion of the subject was an element of the image, or the understanding, of
the event represented by the quotative sentence itself, not independently of
this event. It came into Nicolai's consciousness only as an element of this
event.
The stylistic effect of the ga attached to the braced {Sonya} in the sen­
tence mentioned just now is worthy of a special note. As Nicolai entered
the ballroom, Sonya was standing by the clavichord with Natasha. Thus, it
is not that Nicolai did not really see Sonya. Nicolai, however, was, perhaps
unconsciously avoiding Sonya psychologically. He knew she loved him. But
he must have had complex feelings towards the selfless Sonya. She would
be the last person he would have wanted to face when he came home after
that shameful disaster. In contrast, the simple, happy, cheerful sister,
Natasha —, if only he had to face Natasha! This spontaneous psychological
discrimination between Sonya and Natasha on the part of Nicolai drew his
consciousness towards Natasha as he entered the ballroom. It is this
attitude of Nicolai that is reflected in the different particles attached to
Natasha and Sonya as the subjects of quotative verbs. Nicolai let Sonya into
his attentive consciousness only as an element of perception of her voice:
"No, Papa is not back yet."
One might say that the ga attached to {Sonya} is simply to be
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 149

explained as a focus marker (focus in the grammatical sense, not in the


sense of focus of attention). In other words, the sentence " is not back
yet," said Sonya' can, one might say, be paraphrased here as "It is Sonya
who said (answering Nicolai's question to Natasha) that Papa was not back
yet." But, first, even if this interpretation is correct and ga is functioning
here as a focus marker, the stylistic effect of the contrast between the wa
attached to Natasha and the ga attached to Sonya remains intact. Second, it
is doubtful that this sentence must be interpreted as a focus sentence
answering the implicit question "who replied to Nicolai's question?" There
is here a given sequence of events. Nicolai asked the question of whether
Papa was home, which is followed by Natasha's and Sonya's utterance. The
problem posed to us (putting ourselves in the author's place) is how to
describe this sequence of events. And this problem is in principle indepen­
dent of posing the question "who answered Nicolai?" We are to choose,
then, between Sonya wa and Sonya ga as a felicitous expression to repre­
sent Sonya in the description of this sequence of events and of the way the
events are reflected in Nicolai's consciousness. If the focused meaning
appears to be implied here, it could be accounted for as a derived conse­
quence from the intended meaning of the sentence in the reader's mind.
Let us note, in fact, that Sonya would be followed by wa if the assumed
function of Sonya in Nicolai's consciousness were different. Suppose that
Nicolai had turned his attention equally to Sonya as well. Assume, for
example, when he saw Natasha running to him, he also directed his active
consciousness towards Sonya, who was remaining by the clavichord, or
perhaps was slowly walking towards him, smiling. Then, the sentence '—
exclaimed Natasha, running to him' might be followed by '[Sonya] smiled
to him' (Sonya wa Nikolai ni hohoemi kaketa). With this change in context
(that is, with a change in the situation to be described) the quote 'No, Papa
is not back yet' could be (would be) followed by 'said [Sonya] in place of
Natasha' (Sonya wa Natasha ni kawatte itta).

3.4. Let us now turn to the case in which the subject of a quotative verb is
reflexive of the point of view. We have two instances of this case in our
text: "'Is Papa home?" he asked,' and " O h , nothing," he replied.' (As
remarked earlier, however, the latter may also be interpreted from the
Countess's point of view.)
It is unreasonable to maintain that the same analysis holds here as in
the preceding case of wa-phrases nonreflexive of the point of view. For
150 S.-Y. KURODA

example, consider the first of our two sentences. One would not say that
Nicolai's consciousness had been directed towards himself when he asked
whether Papa was home in the same sense that it had been directed towards
Natasha when he perceived her exclaim "Ah! Here's Nikolenka!" and run
to him.
The wa attached to Nicolai in our present example, then, would have
to be justified on the basis of the fact that Nicolai is the subject of the point
of view. This fact is not itself a component of the world of the story, the
world the story is describing. Rather, it is a feature of the way the story is
told.
Yet, so long as this sentence ("'Is Papa home?" [he] asked.') is inter­
preted as representing Nicolai's point of view, some degree of self-aware­
ness on the part of Nicolai is implied. Wa attached to Nicolai in the sen­
tence may not be replaced by ga so long as this sentence is understood as
reflexive of the point of view, i.e., so long as it is to be interpreted as repre­
senting Nicolai's point of view. For this verbal act to be described from
Nicolai's point of view it seems unnecessary that Nicolai's consciousness
stands in a particular relationship with this event. If so, the meaning of wa
in a sentence reflexive of the point of view may also be justifiable internally
to the world of the story. Let us first consider this line of thought. We shall
return to the possible justification external with the world of the story in
3.5. after we consider the case with an inner process verb think.
At the least, Nicolai's asking whether Papa was home was a willful,
self-controlled act of Nicolai. This may be compared with the case in which
Nicolai was hypnotized and uttered the same question. In a description of
this hypothetical scene the same sentence "'Is Papa home?" [he] asked'
([he] being accompanied by wa) may appear. But in such a case Nicolai is
not the subject of the point of view and the sentence cannot be interpreted
as reflexive of the point of view. Nicolai, naturally, is not in the content of
Nicolai's consciousness. In contrast, with the wa-phrase reflexive of the
point of view the referent of the wa-phrase (Nicolai himself) may be said to
be an independent element of the content of Nicolai's consciousness, inde­
pendent of the external event of his asking the question as reflected in his
consciousness, though not in the same way as in the preceding case of the
subject nonreflexive of the point of view.
In the present case, this independence of the self as an element of the
content of consciousness is purely a matter of, so to speak, the "logical"
structure of the content of consciousness. To the extent that he was aware
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 151

that he asked whether Papa was home, Nicolai was also aware of the self
whose willful act he was aware of. He was aware at least that his act of ask­
ing the question was not incompatible with his will at the moment, and to
the extent he was aware of this he would have to be aware of his self inde­
pendently of his awareness of this act; the self was contained as a compo­
nent of the content of his present consciousness independently of the self as
an element of the event (his asking the question), which itself was a content
of his consciousness.
To elaborate on this point a little further, compare the scene in our text
with the following variation. Assume, instead of our sentence, we have '"Is
Papa home?" [he] asked unintentionally.' (In Japanese, "Otoosan wa ie
kai?" to kare wa omowazu kiita.) Nicolai was looking for Papa, for he had
to ask him for the money he had to pay Dolokhov. But Nicolai did not want
others to learn why he was looking for Papa. But in spite of himself he
asked Natasha: "'Is Papa home?'" This is our present imaginary setting of
the scene. Our sentence can still be interpreted as representing Nicolai's
point of view. He was aware of his own act, asking a question to Natasha.
At the same time, he was aware that this spontaneous act was contrary to
his will. The will of his self and his own act, of which he was aware, is
incompatible, and he was aware of this incompatibility. His awareness of
self, then, must be established independently of his awareness of his act in
the logical structure of his present consciousness.
In our original text, Nicolai's asking the question "Is Papa home?" was
a willful act by Nicolai, and he knew it. He knew that his act which was
externalized and of which he was aware was not incompatible with the will
of his self. I think our sentence '"Is Papa home?" he asked' means this, so
long as it is interpreted as reflexive of the point of view. We must admit,
then, that in the logical structure of the content of Nicolai's consciousness
his self is contained as a component independent of the event of his asking
a question to Natasha, which event is also a component of Nicolai's con­
sciousness and of which Nicolai is contained as an element.

3.5. We now proceed to consider the case where the main verb is think.
This is the case where the fact that we are concerned with Nicolai's point of
view is most explicit. But precisely for that reason the stylistic function of
the wa attached to the subject of the main verb is quite different from the
previous cases. At the moment when this inner process of Nicolai's took
place, no one's consciousness may be said to be focused on Nicolai. One
152 S.-Y. KURODA

cannot even assume that Nicolai was conscious of himself in the same sense,
and to the same extent that we assume he was selfconscious in a situation
described earlier by a sentence reflexive of Nicolai's point of view, like that
discussed earlier ("'Is Papa home?" [he] aksed.') The sentence '"every­
thing's still the same with them" thought [Nicolai]' does not imply that this
psychological act is a willful act of Nicolai. That everything is still the same
with them is the content of Nicolai's consciousness, but our sentence does
not imply that he is reflecting on his own thought (though of course is not
contradictory with the conclusion), nor does it imply that this content of
consciousness has any interaction with other components of Nicolai's con­
sciousness.
A sentence of this kind represents no one's act of judging, and one can­
not justify the use of wa here in a way internal to the world of the story. A
communicational theory of narration would here assume that it is an omnis­
cient narrator who is perceiving and judging on an inner process of Nicolai.
Within such a conception one might justify the use of the wa attached to
Nicolai, the subject of the verb think, in analogy with the wa attached to
Natasha in the sentences that represent Nicolai's point of view. The omnis­
cient narrator's consciousness has been centered around Nicolai. I argued
against setting up such an enigmatic creature as an omniscient narrator in
my earlier papers. 2
In contrast, I assume simply that the function of a sentence such as
'"Everything's still the same with them", Nicolai thought' is to create in the
reader's mind a meaning and simultaneously a section of the imaginary
world with which it is referentially interpreted. With this conception of nar­
ration, only with respect to the reader's consciousness could one justify the
role of wa attached to the subject of a verb of inner process.
It is worthy of note that the two wďs in our text attached to the subject
of think, the complement of which represents the point of view of its refe­
rent, cannot be replaced by ga. For example, it is impossible to replace the
wa attached to the Nicolai in '"Everything's still the same with them",
Nicolai thought' by ga. In contrast, the wďs attached to Natasha as the sub­
jects of quotative verbs, subjects noncoreferential with the point of view,
may be replaced by ga in our texts, of course with certain changes in stylis­
tic effect (i.e., changes in meaning). We shall return to the effect of such
replacement shortly. The impossibility of replacement of the wa attached to
Nicolai as the subject of think in our text may be taken as an indication that
an inner event of Nicolai's thinking, such a process internal to Nicolai's con-
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 153

sciousness, may not enter the reader's consciousness in the way an external
event, for example, Sonya's saying "No, papa is not back yet," may enter
(or may be imagined to enter) a protagonist's consciousness (for example,
Nicolai's), or the reader's consciousness, with the help of imagination.
Perhaps an inner element may not constitute a selfcontained element, as an
image or an understanding in the reader's consciousness in the way an
external event may (or may be imagined to) in a protagonist's, and through
the faculty of imagination, in the reader's, consciousness.
It is indeed a simple fact that the reader cannot imagine perceiving the
inner event of Nicolai's thinking such and such, nor can he imagine identify­
ing himself with anyone perceiving this inner event (except by evoking the
notion of an omniscient narrator who has the superhuman faculty of
observing an inner event directly). Note that to imagine identifying our­
selves with Nicolai will not do. For, with our example, Nicolai himself may
not necessarily be said to have been aware that he was thinking "everything
is still the same with them" in a way he was aware of Natasha's crying
"Beautiful! Wonderful!" or Sonya's saying "No, Papa is not back yet," even
though he was certainly aware that everything was still the same with them.
Nonetheless, our sentence creates in the reader's consciousness the
understanding that an inner event of thinking that everything was still the
same with them took place in Nicolai's consciousness; the reader is made
aware that Nicolai thought so. Representing an inner event of thinking
requires a two-place predicate, one argument representing the subject of
consciousness in which this inner process takes place and the other repre­
senting the content of the consciousness created by this process. Nicolai is
in our consciousness as an argument fulfilling this two-place predicate when
we obtain this understanding. But at the same time Nicolai is in our con­
sciousness as the subject of the point of view for this sentence, as the locus
of this inner process, which process we do not imagine perceiving directly
but whose effect on Nicolai's consciousness we are made aware of, as a
result of our empathy with Nicolai.
Thus, the meaning of wa attached to Nicolai as the subject of the verb
think must be justified externally with respect to the story, or in other
words, with respect to how the reader's consciousness is to be related to the
world of the story.
The same external interpretation may also be assumed for the wa's
which are attached to the subjects reflexive of the point of view. What
makes this case different from the case with the verb think is that in this
154 S.-Y. KURODA

case the subject of the point of view himself (in our example, Nicolai) in
addition perceives that he himself is a component, in fact a willful actor, in
the event described by the sentence in question. Thus, we may also talk
about the relationship of Nicolai as the subject of consciousness that per­
ceives this event and Nicolai as an actor in the same event.

3.6. If our preceding argument is basically correct, then, we have the fol­
lowing generalization. We have three cases of wa-phrases as the subjects of
quotative verbs in our text. First, nonreflexive of the point of view; second,
reflexive of the point of view; and finally, as a locus of consciousness. In
each of these cases the existence of the referent of the wa-phrase can be jus­
tified in the consciousness of the subject of the point of view or of the
reader independently of (as well as in addition to) the fact that it is an argu­
ment of the predicate represented by the main verb and as such is also an
element of the content of the understanding consciousness. In the first case,
one might even be allowed to say that the referent of the wα-phrase is
established as an element of the consciousness as a focal point of the subject
of the point of view's attention before it obtains another qualification of
existence as an argument of the main verb, i.e., as an element of an event
which is perceived by the subject of the point of view. But this presumed
temporal order in which the referent of a wa-phrase comes to get doubly
qualified as a component of a consciousness is not essential for the generali­
zation we are now concerned with. What is crucial is the uniformity in the
"logical" structures, so to speak, of the consciousness whose state the sen­
tence is related to in each case (i.e., in our preceding examples, Nicolai's in
the first and the second case, and the reader's in the third case). The recog­
nition of the existence of the referent of the wa-phrase is independent of the
judgment that the event described by the predicate represented by the main
verb takes place. We would conclude that in all the three cases a sentence
with a wa-phrase conforms to the structural characteristic of the Brentano-
Marty concept of categorical judgment as a "double"-judgment.
It might be instructive to conduct some gedanken experiments by try­
ing to replace the wďs in our text by ga. We can repace all the wďs attached
to the subjects of quotative verbs noncoreferential with the subject of the
point of view, Nicolai. The passage as a whole can still be considered as
representing Nicolai's point of view. But one might say the impression of
the passage is much "drier" than before. Nicolai's consciousness does not
particularly seem to be centered around any person or to be responding to
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 155

his surroundings. It is as though Nicolai's consciousness accepts the sound


and the voices around him simply passively, perhaps somewhat absent-min­
dedly. Only the two occurrences of wa attached to the subjects reflexive of
the point of view ('he asked' and 'he replied') indicate some degree of
active response of Nicolai's consciousness to external events, in these cases,
produced by his willful acts.
As I mentioned earlier, the wďs attached to the subject he, referring to
Nicolai, of the verb think cannot be replaced by ga. The two wďs attached
to Nicolai reflexive of the subject of the point of view also resist replace­
ment by ga so long as the sentences with the verb think invite us to interpret
the passage (or at least the neighborhood of these sentences) as represent­
ing Nicolai's point of view. In contrast, if we eliminate these two sentences
with the verb think, thus erasing lexically explicit clues for Nicolai's point of
view from this passage, then, one can replace those wďs attached to the
Nicolai's which are the subjects of ask and reply. Then, the passage could
quite naturally be taken on the whole as representing no one's point of
view. The description becomes still "drier" and more objective. We can no
longer read from the passage any sign of Nicolai's consciousness reacting to
the reality surrounding him. It is as if we were listening to someone describ­
ing to us a scene s/he saw on the stage.

4. Our second example is from D.H. Lawrence's Sons and Lovers, Part
II, Chapter VII. This scene with a son and a lover should need no introduc­
tion.
She wanted to show him a certain wild-rose bush she had discovered.
She knew it was wonderful. And yet, till he had seen it, she felt it had not
come into her soul. Only he could make it her own, immortal. She was dis­
satisfied.
Dew was already on the paths. In the old oak-wood a mist was rising,
and he hesitated, wondering whether one whiteness were a strand of fog or
only campionflowers pallid in a cloud.
By the time they came to the pine-trees Miriam was getting very eager
and very tense. Her bush might be gone. She might not be able to find it;
and she wanted it so much. Almost passionately she wanted to be with him
when he stood before the flowers. They were going to have a communion
together — something that thrilled her, something holy. He was walking
beside her in silence. They were very near to each other. She trembled,
and he listened, vaguely anxious.
Coming to the edge of the wood, they saw the sky in front, like
mother-of-pearl, and the earth growing dark. Somewhere on the outer-
156 S.-Y. KURODA

most branches of the pinewood the honeysuckle was streaming scent.


"Where?" [he] asked.
"Down the middle path," [she] murmered, quivering.
When they turned the corner of the path she stood still. In the wide
walk between the pines, gazing rather frightened, she could distinguish
nothing for some moments; the greying light robbed things of their colour.
Then she saw her bush.
"Ah!" [she] cried, hastening forward.
It was very still. The tree was tall and straggling. It had thrown its
briers over a hawthornbush, and its long streamers trailed thick, right
down to the grass, splashing the darkness everywhere with great split stars,
pure white. In bosses of ivory and in large splashed stars the roses gleamed
on the darkness of foliage and stems and grass. Paul and Miriam stood
close together, silent, and watched. Point after point the steady roses
shone out to them, seeming to kindle something in their souls. The dusk
came like smoke around, and still did not put out the roses.
Paul looked down into Miriam's eyes. She was pale and expectant
with wonder, her lips were parted, and her dark eyes lay open to him. His
look seemed to travel down into her. Her soul quivered. It was the com­
munion she wanted. He turned aside, as if pained. He turned to the bush.
"They seem as if they walk like butterflies, and shake themselves,"
[he] said.
She looked at her roses. They were white, some incurved and holy,
others expanded in an ecstasy. The tree was dark as a shadow. She lifted
her hand impulsively to the flowers; she went forward and touched them in
worship.
"Let us go," [he] said.
There was a cool scent of ivory roses — a white, virgin scent. Some­
thing made him feel anxious and imprisoned. The two walked in silence.
All the subjects of quotative verbs in this passage are marked by wa in
the Japanese translations I consulted (Akira Honda's Iwanami Bunko edi­
tion). It might appear that we have a simpler structure here than in the pre­
vious example from War and Peace. But Lawrence's prose exhibits an intri­
cate and effectively ambiguous structure of points of view. In this passage,
Paul's, Miriam's and the "objective" point of view are interwoven, often
indistinguishably. In fact, I discussed parts of the above passage from
Lawrence in earlier papers of mine to illustrate this point. I refer the reader
to those papers for details. I think that all the quotative sentences (sentences
with a direct quotation) are, from the point of view of the points of view,
ambiguous. For example, consider the following two lines:
"Where?" [he] asked.
"Down the middle path," [she] murmured, quivering.
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 157

Interpreted from Paul's point of view the first sentence is reflexive of


the point of view and the second sentence is nonreflexive of the point of
view. Interpreted from Miriam's point of view the role of the sentences
switches. And these interpretations are both not only possible but are
simultaneously called for, it seems to me, for an adequate understanding of
our passage. Paul is in Miriam's consciousness and Miriam is in Paul's as
they walk side by side. His voice heard by Miriam does not form an inde­
pendent impression in her consciousness but interfuses with Paul existing in
her. Her voice heard by Paul'does not form an independent impression in
his consciousness but interfuses with Miriam existing in his consciousness.
At the same time, Paul is aware that he asks a question, and Miriam is
aware she murmurs, quivering.
It is not impossible to replace some of the wa's attached to the subjects
of quotative verbs in this passage by ga with, of course, changes in the
stylistic effects, or in the meaning, of the passage. For example, we might
replace wa attached to he following the "Where" quoted above. Then, the
sentence '"Where?" he asked' is, it seems to me, separated from Paul's
point of view. And, furthermore, his voice, it seems to me, enters Miriam's
consciousness somewhat abruptly, as if she is taken back to herself by this
voice. One might note that the adverb totuzen ("suddenly") may quite well
be added to the sentence: "Doko" to kare-ga totuzen tazuneta ("Where?"
he asked suddenly).
Removing the sentence '"Where?" he asked' from Paul's point of view
(which I think is a necessary consequence of the replacement of the wa by
ga) impoverishes, albeit slightly, the intricate effects of interaction between
Paul's and Miriam's consciousness which was to climax in the communion
Miriam wanted. In order to see the full consequence of removing Paul's
point of view one might as well try to rewrite the passage in the first person
from Miriam's point of view, replacing Miriam and she by I and they by we.
This can be done with some necessary deletions. For example, continuing
the two lines I quoted above, one might have, with the bracketed part
deleted,
"Where?" he asked.
"Down the middle path," I murmured, quivering.
When we turned the corner of the path I stood still. In the wide walk
between the pines [gazing rather frightened] I could distinguish nothing for
some moments; the greying light robbed things of their colour. Then I saw
my bush.
"Ah!" I cried, hastening forward.
158 S.-Y. KURODA

The change caused by this rewriting is not simply that we do not know any
more that Paul saw Miriam gazing rather frightened. This phrase, which
exclusively represents Paul's point of view, has the effect of extending a
horizon of Paul's point of view beyond its small stretch in the original text.
Paul was watching Miriam unable to distinguish anything in the greying
light. In a similar vein, replacement by ga of the wa attached to the he after
"Where?" in the Japanese version might have caused a more serious effect
than simply removing this sentence from Paul's point of view.
The effect of wa in the original passage, then, is that Paul and Miriam
existed in each other's consciousness through the passage of time indepen­
dently of their understanding of each other's actions. Thus, for example, he
in the sentence "'where?' he asked" not only functions as a term referring
to the actor of the action of asking (which is the only syntactic information
coded in English), but, with wa attached to it in the Japenese translation,
allows (or, perhaps forces) us to understand that this action of Paul's took
place under circumstances where Paul and Miriam were attentive to them­
selves and to each other. The sentence represents a "double" judgment in
the mind of each of them.

5. Our third and last example is from William Faulkner, The Sound and
the Fury, the first section, April seventh, 1928. This section is a first-person
narration by Benjy, an imbecile, of his experience in the afternoon and in
the evening until he went to bed on April seventh, 1928, his 33rd birthday.
Benjy does not speak and he does not understand language; he is totally
incapacitated mentally. It is Faulkner's ingenious stylistic adventure to have
an imbecile without language competence narrate his own experience (with
flashbacks — according to someone's count there are 127 flashbacks in this
section). Benjy's consciousness simply responds passively to outward
stimuli and stimuli evoked by his memory. It is an ironical contradiction
that an imbecile lacking totally the faculty of reflective consciousness nar­
rates his experience (in the past tense).
"It's too cold out there." {Versh} said. "You don't want to go out
doors."
"What is it now." {Mother} said.
"He want to go out doors." {Versh} said.
"Let him go." Uncle {Maury} said.
"It's too cold." {Mother} said. "He'd better stay in. Benjamin. Stop
that, now."
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 159

"It wont hurt him." Uncle {Maury} said. "You, Benjamin."


{Mother} said. "If you don't be good, you'll have to go to the kitchen."
"Mammy say keep him out the kitchen today." {Versh} said. "She say
she got all that cooking to get done."
"Let him go, Caroline." Uncle {Maury} said. "You'll worry yourself
sick over him."
"I know it." {Mother} said. "It's a judgment on me. I sometimes won­
der."
"I know, I know." Uncle {Maury} said. "You must keep your
strength up. I'll make you a toddy."
"It just upsets me that much more." {Mother} said. "Don't you know
it does."
"You'll feel better." Uncle {Maury} said. "Wrap him up good, boy,
and take him out for a while."
Uncle Maury went away. Versh went away.
"Please hush." {Mother} said. "We're trying to get you out as fast as
we can. I don't want you to get sick."
Versh put my overshoes and overcoat on and we took my cap and
went out. Uncle Maury was putting the bottle away in the sideboard in the
dining-room.
"Keep him out about half an hour, boy." Uncle {Maury} said. "Keep
him in the yard, now"
"Yes, sir." {Versh} said. "We don't never let him get off the place."
We went out doors. The sun was cold and bright.
We note that in the translation of this passage (by Seiji Onoe, Huzanhoo,
Tokyo) all the subjects of quotative verbs are translated in ga-phrases. The
talk by Versh, Mother, Uncle Maury is all simply voices falling on Benjy's
ears. It is as though Benjy is a recording machine. Their actions are
recorded as if they were events passing through Benjy's mind one by one.
Each sentence with a quotative verb represents a "simple" judgment by
Benjy, in which the subject of the verb only functions to refer to the actor
of the verbal action.
It is not the case that the translator avoided using wa completely
throughout the first part of The Sound and the Fury. There are a few
instances where the subject of a quotative verb is marked with wa. It is dif­
ficult to determine what stylistic effects the translater intended by choosing
wa instead of ga in those very rare cases. In comparison, a rough count indi­
cates that the subjects of quotative verbs are translated with ga and wa
about half and half in the fourth part of the book, which is written from an
"omniscient", or "objective", point of view. Again it is difficult to deter­
mine exactly what motivated the translater to choose either of them in each
160 S.-Y. K U R O D A

instance. Nonetheless, it is remarkable that an overwhelming majority of


the subjects of quotative verbs are rendered into ga phrases in the transla­
tion of the first part of the novel.

NOTES

*Final preparation of this paper was done while the author was a fellow at the Center for
Advanced Study in the Behavioral Sciences. I am grateful for financial support provided by the
Center, the Alfred P. Sloan Foundation, and the System Development Foundation. I am also
grateful to Susan Fischer and Leslie Saxon for improvements in the text.
1) "The categorical and the thetic judgment. Evidence from Japenese syntax," Foundations
of Language 9.2, 1972; "Subject" in M. Shibatani ed. Japanese Generative Grammar: Syntax and
Semantics Vol. 5, Academic Press, New York, 1976; both papers are also reprinted in S.-Y.
Kuroda, The (w)hole of the Doughnut:Syntax and its Boundaries, E. Story Scientia, Ghent-
Antwerp, 1979. For the original work of Anton Marty on this matter, see Anton Marty, Unter­
suchungen zur Grundlegung der allgemeinen Grammatik und Sprachphilosophie, Max
Niemeyer, Halle, 1908; Psyche und Sprachstruktur, Francke, Berne, 1925.
2) "Where epistemology, style and grammar meet," in S.R. Anderson and P. Kiparsky eds.,
A Festschrift for Morris Halle, Holt, Rinehart and Winston, Sydney, 1973, pp. 377-391; "On
grammar and narration," in Actes du Colloque Franco-Allemand de Grammaire Transfor­
mationelle, II. Etudes de Semantique et autres, Niemeyer, 1974, pp. 165-173; "Reflections on the
foundations of narrative theory," in T. van Dijk ed. Pragmatics of Language and Literature
North Holland Publ. Co., Amsterdam. The first and third papers are also reprinted in Kuroda
op. cit.

REFERENCES

Anderson, S. and P. Kiparsky. 1973. eds. A festschrift for Morris Halle.


Sydney: Holt, Rinehart and Winston.
Kuroda, S.-Y. 1972. "The categorical and the thetic judgment: Evidence
from Japanese syntax." Foundations of Language. 9.2.
. 1973. "Where epistemology, style and grammar meet." Anderson and
Kiparsky, 377-391.
1974. "On grammar and narration." Rohrer and Ruwet, 165-173.
1976a. "Subject." Shibatani, 1-6.
1776b. "Reflections on the foundations of narrative theory." van Dijk,
108-140.
1979. The (w)hole of the doughnut .syntax and its boundaries, E. Story
Scientia, Ghent-Antwerp.
SO-CALLED TOPIC WA IN TOLSTOI, LAWRENCE AND FAULKNER 161

. To appear. "The categorical and the thetic judgment revisted." Mulli­


gan et al.
Marty, Anton. 1908. Untersuchungen zur Grundlegung der allgemeinen
Grammatik und Sprachphilosophie. Halle: Max Niemeyer.
. 1925. "Psyche und Sprachstruktur." Berne: Francke.
Mulligan, Kevin et al. To appear. Mind, meaning and metaphysics: the
philosophy and theory of language of Anton Marty. The Hague: Nijhoff.
Rohrer, Ch. and N. Ruwet. 1974. eds. Actes du colloque franco-allemand
de grammaire transformationeile. II. études de sémantique et autres.
Tuebingen: Niemeyer.
Shibatani, Masayoshi. 1976. ed. Japanese generative grammar: syntax and
semantics Vol. 5. New York: Academic Press.
van Dijk, Teun. 1976. ed. Pragmatics of language and literature. Amster­
dam: North-Holland.
PART II: WA AND OTHER SYNTACTIC PHENOMENA
THE ROLE OF WA IN NEGATION*

NAOMI HANAOKA McGLOIN


University of Wisconsin, Madison

1. Introduction
It has long been observed that the particle wa figures prominently in
negative sentences. Negative sentences generally contain more was than
their corresponding affirmative sentences, as is exemplified in (1) and (2).
(1) a. tooji no hooritsu de wa onna mo kubunden ga
that-time LK law by women also land
mora-e-ta.
receive-can-PST
By the law of that time, women could also receive an allot­
ment of land,
b. tooji no hooritsu de wa onna wa kubunden wa
that-time LK law by women land
mora-e-na-katta.
receive-can-NEG-PST
By the law of that time, women could not receive an allot­
ment of land.
(2) Q: kesa koohii o nomimashi-ta ka?
this-morning coffee DO drink-PST QU
Did you drink coffee this morning?
A: a) ee, (kesa koohii ) nomimashi-ta.
yes drink-PST
Yes, I did.
b) iie, kesa wa koohii wa nomimas-en-deshita.
no this-morning coffee drink-NEG-PST
No, I did not drink coffee this morning.
The purpose of this paper, then, is to examine the role of wa in nega-
166 NAOMI HANAOKA McGLOIN

tive sentences. I will try to do this by first contrasting wa and other particles
in negative sentences, and secondly, by investigating how the occurrence of
wa affects the interpretation of negative sentences in Japanese.

2. Wa-marked vs. Non-wa-marked Negation


Although negative sentences often mark involved NPs by wa, particles
other than wa also appear in negative sentences. Consider the following
sentences.
(3) a. hon   om-ana-katta.
book DO read-NEG-PST
I did not read a book.
b. hon wa yom-ana-katta.
TP read-NEG-PST
I did not read a book.
In sentence (3a), the object NP is marked with the regular object marker o,
while, in (3b), the object is marked with wa. We will call the (3a)-type of
negation "non-vvtf-marked" negation and the (3b)-type "wa-marked" nega­
tion. Non-wa-marked negative sentences are those that employ regular case
particles such as in the (a)-sentences of (3)-(7), while wa-marked negative
sentences attach wa to these particles, as in the (b) sentences of (3)-(7).
(4) a. kodomo ga i-nai.
child SB be-NEG
The child is not here.
b. kodomo wa i-nai.
child TP be-NEG
The child is not here.
(5) a. basu de ko-na-katta.
bus by come-NEG-PST
b. basu de wa ko-na-katta.
bus by come-NEG-PST
I did not come by bus.
(6) a. uchi de benkyoo-shi-na-katta.
home at study-do-NEG-PST
b. uchi de wa benkyoo-shi-na-katta.
home at study-do-NEG-PST
I did not study at home.
THE ROLE OF WA IN NEGATION 167

(7) a.doyoobi ni ko-na-katta.


Saturday on come-NEG-PST
b. doyoobi ni wa ko-na-katta.
Saturday on come-NEG-PST
I did not come on Saturday.
Both non-wa-marked negative sentences and wa-marked negative sen­
tences potentially have either a non-implicational negation interpretation
(i.e., the denial of an action itself) or an implicational negation interpreta­
tion (i.e., a reading where some other proposition or argument is implied.)
So, sentence (6a) could be either a flat denial of one's studying at home
(non-implicational) as in (8), or a denial of the location of studying (impli­
cational reading) as in (9).
(8) ano hito wa soto de asonde bakari ite,
that person TP outside LC play:GER only be:GER
chittomo uchi de benkyoo shi-nai.
at-all home at study do-NEG
That person is spending all his time playing outside and he does
not study at home at all.
(9) ano hito wa toshokan de bakari benkyoo shite
that person TP library at only study do:GER
ite, uchi de benkyoo shi-nai.
be:GER home at study do-NEG
That person studies only in the library and does not study at
home.
Sentence (6b) could also be implicational or non-implicational, depending
on the nature of wa.1
Non-wa-marked negative sentences and wa-marked negative sen­
tences, however, are uttered in discourse/pragmatically different contexts.
Consider the following sentences.
(10) kusuri  nom-ana-katta.
medicine DO drink-NEG-PST
I did not take the medicine.
(11) kusuri wa nom-ana-katta.
medicine drink-NEG-PST
I did not take the medicine.
168 NAOMI HANAOKA McGLOIN

Sentence (11), but not (10), is an appropriate answer to a question


where the corresponding affirmative is assumed.
(12) A: kesa kusuri o non-da deshoo ne.
this morning medicine DO drink-PST probably EX
You took the medicine this morning, didn't you?
B: kusuri nom-ana-katta.

I didn't take the medicine.


Sentence (11), but not (10), is a natural utterance when the corresponding
affirmative can be inferred (or at least when the speaker believes it can be
inferred) from the previous discourse, as in (13).
(13) kinoo wa atama ga itakute, benkyoo deki--katta.
yesterday TP head SB hurt:GER study can-NEG-PST
demo, kusuri nom-ana-katta.

but medicine drink-NEG-PST


Yesterday, I had a headache and so I could not study. But, I did
not take the medicine.
In (13), the hearer/reader can reasonably be expected to infer that the
speaker/writer might have taken the medicine having heard/read that he
had a headache. The sentence negates this expectation.
Non-wa-marked negative sentences, on the other hand, often occur
discourse-initially without prior mention of the corresponding affirmative.
Observe the following.
(14) a! ikenai. kesa kusuri o nom-ana-katta.
Oh! No! this-morning medicine DO drink-NEG-PST
Oh, no! I didn't take the medicine this morning.
In a situation where one suddenly remembers that he did not take the
medicine,  but not wa is natural.
Now, in many discourses where the corresponding affirmative can be
inferred from the previous discourse, both wa-marked negation and non-
wa-marked negation occur. Observe the following.
(15) sumisu-san wa kyonen no kugatsu ni nihon e
Smith TP last-year LK September in Japan to
ikimashi-ta. tookyoo no daigaku de bekyoo o
go-PST Tokyo LK college at study DO
THE ROLE OF WA IN NEGATION 169

shimashi-ta. mainichi nihongo o benkyoo shimashi-ta.


do-PST everyday Japanese DO study do-PST
demo, amari terebi mimas-en-deshita.

but not-too-much TV see-NEG-PST


Last September, Mr. Smith went to Japan. He studied at a uni­
versity in Tokyo. He studied Japanese everyday, but he didn't
watch TV too much.
If the paragraph is to end right here, wa is the most natural choice. Wa
contrasts the action of studying Japanese and the action of watching TV,
which in turn relates to the topic of what the student did in Japan. When 
is used, on the other hand, it is felt that the speaker/writer is making a cer­
tain evaluation that one should have watched TV, because TV, after all, is
a useful resource for learning a language. Hence, this discourse with 
sounds most natural if it is followed by a remark such as (16).
(16) desukara, nihongo ga amari joozu ni
therefore Japanese SB not-too-much well RS
narimas-en-deshita.
become-NEG-PST
Therefore, his Japanese did not improve too much.
The evaluative nature of nοn-wa-marked negation explains the follow­
ing facts.
(17) a. shikago e it-ta noni,
Chicago to go-PST although
nihon-ryoori tabe-na-katta n desu ka?

Japanese-food eat-NEG-PST NOM COP QU


You went to Chicago, and didn't eat Japanese food?
b. shikago e it-ta keredo, nihon-ryoori
Chicago to to-PST but Japanese-food
tabe-na-katta.

I went to Chicago, but I didn't eat Japanese food.


Both noni and keredo are conjunctions meaning but or although. Noni,
however, imposes a strong sense of "contrary-to-expectation" — i.e., in a
S1 noni S2 construction, what is said in S2 is contrary to what one would
170 NAOMI HANAOKA McGLOIN

expect from S1 With noni, as in (17a), ηοη-wa-marked negation is prefer­


red. Sentence (17a) implies that one is expected to eat Japanese food when
one goes to Chicago, and that the speaker is surprised that the addressee
did not do so. Keredo, on the other hand, does not impose such a strong
value judgment. Here, as in (17b), wa5-marked negation is preferred. In
(17b), the two events are simply contrasted.
Hence, the speaker/writer effects a subtle change in one's evaluation of
the situation by the use of wa-marked vs. non-wa-marked negation.
Observe the following.
(18) sarani ishida-kenji no shitai ga hitatteita mizu no
moreover (name) LK body SB immersed water LK
shita wa dorotsuchi ni natteiru ga, sono dorotsuchi
under TP mud RS has-become but that mud
ni wa shitai no omomi ni-yoru kanbotsu ga
LC body LK weight by depression SB
mi-rare-na- katta.
see-PSS-NEG-PST
Moreover, there was mud underneath the water in which Ishida's
body was immersed. However, the mud did not show any depres­
sion from the weight of the body.
(Ishida Kenji no Kaishi: 17)
By using ga as in kanbotsu ga, the writer indicates that the fact that
there was no depression in the ground was something strange, something
unusual. In other words, the writer is making a certain evaluation that there
should have been a depression if it was indeed the case that he fell to the
ground from the train. This usage of non-wa-marked negation is very fre­
quent in passages where the writer is pointing out what he considers to be
peculiar. Similar examples are:
(19) shitai ni wa shukketsu ga na-katta.
body in blood be:NEG-PST
There was no bleeding in the body.
(20) kutsu-ura no hyoomen ni wa sonshoo ya
sole LK surface LC damage and
konseki ga nai.
trace be:NEG
There was no damage or traces on the soles of his shoes.
THE ROLE OF WA IN NEGATION 171

By using ga, the writer builds up his case to say that Ishida's death was
indeed not accidental, contrary to what was commonly believed to be the
case. If wa were to be used in the above sentences, the facts would be pre­
sented matter-of-factly, and there would be no building of suspense.
In the following example, the use of ga vs. wa reflects a subtle differ­
ence in the participants 1 view of the same fact.
(21) aru hi zettaini junban ga ki-temo "shi-na-i"
one day never turn SB come-though do-NEG
to iiharu ko ga i-ta. sore wa "hanashi wa
QT say-insist child SB be-PST that talk TP
nannimo і!" to yuu otoko-no-ko datta.
nothing be:NEG QT say boy COP:PST
... koochoo-sensei wa sono ko no kara ni
principal TP that child GN empty RS
nat-ta obentoo-bako no not-ta tsukue no 
become-PST lunch-box SB place-PST desk LK front
ni iku to it-ta.
LC go when say-PST
"kimi wa hanashi ga -i no kaa ...."
you TP talk be:NEG NOM QU
"(hanashi wa)2 nannimo nai."
talk nothing be:NEG
One day, a boy would not talk when his turn came. The boy
insisted that he had nothing to say. The principal walked over to
his desk where an empty lunch box was placed and said, "You
don't have anything to say?" "Nothing to say," the boy
answered.
(Kurovanagi, Madogiwa no Tottochan: 144)
Here, the boy uses wa as in "hanashi wa nanimo nai" while the princi­
pal is using ga as in "hanashi ga nai no kď" Hanashi 'talk' has been estab­
lished in this discourse in that everyone was expected to have something to
talk about. Wa is a flat denial of this proposition. The boy is merely stating
a fact. By using ga, however, the teacher again expresses his evaluation —
i.e., he thinks it very strange for anybody not to have anything to talk
about; everybody should have something to talk about.
The above examples illustrate, then, that the use of non-wa-marked
negation is highly evaluative and reflects the speaker/writer's subjective
172 NAOMI HANAOKA McGLOIN

evaluation of the non-occurrence of an action or non-existence of a state as


strange, unusual, or odd. This type of negative sentence strongly implies
that the corresponding affirmative is the norm or what should be the case. 3
The use of wa-marked negation, on the other hand, is objective. Here,
negation is directed without the speaker's subjective evaluation toward a
proposition which has been introduced in the discourse or a proposition
which can be inferred.

3. Wa and the Interpretation of Negative Sentences


The particle wa plays a crucial role in interpreting wa-marked negative
sentences. Two uses of wa can be distinguished: thematic and contrastive.
Briefly stated, thematic wa marks a theme or a topic of a sentence — the
element of the sentence which represents given information. Information
can be considered given by virtue of having been previously mentioned in
the discourse, by being predictable from the discourse, or by being in the
domain of shared knowledge of the speaker and the hearer. Contrastive wa
focuses on an element or a proposition in contrast with another explicit or
implicit element or proposition.
The distinction between thematic wa and contrastive wa, however, is
not always clear.
(22) watashi wa biiru wa nomimasu ga, uisukii wa nomi-masen.
I TP beer drink but whisky drink-NEG
I drink beer, but I don't drink whisky.
(23) Q: biiru wa nomimasu ka?
beer drink QU
Do you drink beer?
A: hiiru wa nomi-masen.
beer drink-NEG
I don't drink beer.
In (22), although "beer" and "whisky" are contrasted, what would pre­
vent us from saying that hiiru wa and uisukii wa are themes of respective
sentences? Biiru wa in (23A) is given information in this discourse. Does
this mean, then, that it is thematic? (23A) could certainly be uttered with
the implication that the speaker drinks other things but not beer.
In fact, the concept of theme itself entails a certain amount of contrast.
Choosing one element as a topic is done in implicit contrast to all the other
elements which could have become a topic. This leads Inoue (1983:46-54)
THE ROLE OF WA IN NEGATION 173

to postulate that the basic function of wa is contrastive and that whether wa


assumes a thematic function or a contrastive function is determined in rela­
tion to the OLD-NEW information structure of a sentence. 4 When wa
occurs in the position of OLD in an OLD-NEW information structure, then
wa assumes a thematic function. Wa assumes a contrastive function when it
occurs in the position of NEW in a sentence.
While it may be the case that there is only one basic function of wa
(contrastive in nature, according to Inoue), it still is useful to talk of thema­
tic wa and contrastive wa since there are some obvious differences between
them.
1. Thematic wa occurs only with anaphoric or generic NPs. 5 There is
no such restriction for contrastive wa. In (24),
(24) oozei no hito wa kimas-en-deshita.
many LK person come-NEG-PST
Not many people came.
since oozei no hito is neither generic nor anaphoric, wa is interpreted only
as contrastive.
2. Contrastive wa can carry emphatic stress, while thematic wa does
not.
3. Thematic wa is attached only to NPs while contrastive wa is
attached both to NPs and non-nominal constituents such as verbs, adverbs
and quantifiers.
4. Thematic wa does not occur in embedded sentences, while contras­
tive wa is generally permitted in embedded sentences.
Now, whether wa is interpreted as thematic or contrastive does influ­
ence the interpretation of wa-marked negation greatly. Thematic wa repre­
sents given information and is therefore outside the domain of NEG. Con­
trastive wa, on the other hand, represents new information, and hence sig­
nals the target of NEG. Consider the following.
(25) a. tabe wa shi-na-katta.
eat do-NEG-PST
I did not eat.
b. kono kyookasho wa atarashiku wa nai.
this textbook TP new NEG
This textbook is not new, but ...
c. suzuki-san wa bijin de wa nai.6
beauty COP NEG
Ms. Suzuki is not beautiful.
174 NAOMI HANAOKA McGLOIN

(26) joozu-ni wa kak-e-nai.


well write-can-NEG
I cannot write well.
(27) zenbu wa deki-na-katta.
all can-NEG-PST
I could not do it all.
In sentences (25)-(27), wa is contrastive since it is attached to a predi­
cate (verb/adjective/copula), an adverb, and a quantifier, which represent
new information in a sentence. Hence, wa marks the target of negation.
Sentence (25a) does not deny one's having done something but denies the
act of eating. Sentence (25b) denies that the book is new but implies that
some other condition holds7 (e.g., "it is a good textbook."). 8 Sentence (25c)
also implies that Ms. Suzuki has some qualities other than being beautiful
(e.g., "She is intelligent.") Sentence (26) does not deny that one can write
but denies the manner in which one can write. Sentence (27) means that
one can do some but not all.
Kuno (1976) proposed that there is a hierarchy for accessibility to
thematic interpretation of noun phrases, which directly corresponds to the
Keenan-Comrie accessibility hierarchy (=(28)). 9
(28) Subject > Dir Obj > Indir Obj > Obj of Prep >
Possess NP > Obj of Comparative Prt
It is easiest to interpret the subject of a sentence as the theme of a sen­
tence while it is most difficult to so interpret the object of a comparative
particle. This leads us to assume that there is also a similar hierarchy for the
accessibility of NPs to contrastive interpretation. This hierarchy will be the
direct opposite of the thematic hierarchy. The more difficult it is to inter­
pret a constituent as thematic, the easier it is to interpret it as contrastive.
Wa-marked negation involving NPs, then, is interpreted along this hierar­
chy. Observe the following sentences.
(29) osaifu wa dokonimo mie-na-katta.
wallet nowhere seen-NEG-PST
The wallet could not be seen anywhere.
(30) sono hon wa yondei-nai.
that book read-NEG
I have not read that book.
THE ROLE OF WA IN NEGATION 175

(31) yooko-san ni wa purezento  age-na-katta.


Yoko  present DO give-NEG-PST
I did not give a present to Yoko.
(32) majison ni wa depaato ga amari nai.
Madison in department-store SB not-too-many be:NEG
In Madison, there are not too many department stores.
(33) shikago e wa ik-ana-katta.
Chicago to go-NEG-PST
I did not go to Chicago.
(34) aiko to wa yar-ana-katta.
Aiko with do-NEG-PST
I did not do (it) with Aiko.
(35) toshokan de wa benkyoo-shi-na-katta.
library in study-do-NEG-PST
I did not study in the library.

(36) suteeki wa ohashi de wa tabe-rare-nai.


steak TP chopsticks with eat-can-NEG
You cannot eat steak with chopsticks.
(37) sankagetsu made wa yubishaburi  yar-ana-katta.
three months until thumb-sucking DO do-NEG-PST
(He) did not start sucking (his) thumb until he was three months
old.
(38) imoolosan wa oneesan hodo wa deki-nai.
younger-sister TP older-sister as can-NEG
The younger sister does not do as well as the older sister.
Sentences (29) and (30) contain a subject NP and a direct object NP,
respectively. It is easiest to interpret wa in (29) as thematic. An unmarked
reading here is one where NP-wa is outside the scope of negation. As we
move from (29) toward (38), it gets increasingly difficult to interpret wa as
thematic. In other words, as we move from (29) to (38), we are more likely
to obtain a reading where NP-wa is the target of negation — i.e., a reading
where NP-wa is inside the scope of negation. Sentences (31) and (32) con­
tain an indirect object NP and locative NP respectively, and here, an
unmarked reading is still the one where NP-wa is outside the scope of nega-
176 NAOMI HANAOKA McGLOIN

tion. Sentences (33), (34), (35) and (36) involve e (direction), to (commita-
tive), de (locative) and de (means), respectively. Sentences (33)-(36) predo­
minantly give a reading where NP-wa is inside the scope of negation,
although the other reading is not entirely impossible. Sentence (36), for
example, is most naturally interpreted as denying the use of chopsticks and
not the eating of steaks. In (37) and (38), which are at the bottom of the
thematic hierarchy, NP-wa is solely interpreted as being inside the scope of
negation. Sentence (38) contains an object of comparative particle. It does
not deny that the younger sister does well but denies that it is to extent of
her older sister.
This is not to say that wa in cases such as (29) and (30) never becomes
the target of negation. However, since they are high in the thematic hierar­
chy, it takes more extra devices for contrastive interpretation to obtain here
— e.g., stress on wa or a context which would force a contrastive interpre­
tation.

4. The Position of Wa and the Interpretation of Negation


We have observed above that when wa receives a contrastive interpre­
tation, the constituent which immediately precedes wa is the target of
NEG. This, however, is not always the case. Consider the following sen­
tences.
(39) kinoo no paatii de wa, osake wa -ana-katta.
yesterday LK party at liquor drink-NEG-PST
Yesterday, at the party, (we) did not drink liquor.
(40) demo, a. koohii o takusan non-da.
but coffee DO much drink-PST
But, (we) had lots of coffee,
b. uta o utatta-ń, odotta-ri shi-ta.
song DO sing-and dance-and do-PST
But, (we) did things like singing songs and dancing.
Sentence (39) can be naturally continued either by (40a) or (40b). Sen­
tence (39) can be said to have an NP-scope reading when (40a) is implied.
Sentence (39) can be said to have a VP-scope reading when (40b) is
implied. In its VP-scope reading, then, sentence (39) denies that people
drank liquor but does not deny that some other action took place. In this
reading, (39) is synonymous to (41).
THE ROLE OF WA IN NEGATION 177

(41) kinoo no paatii de wa, osake  nomi wa shi-na-katta.


liquor DO drink do-NEG-PST
The fact that [NP-wα V-NEG] does not always give a VP-scope reading
can be seen from the following sentences.
(42) a. (taroo wa hanako to) eiga wa mi-na-katta.
TP movie see-NEG-PST
Taroo did not see a movie with Hanako,
b. (taroo wa) hanako to wa (eiga o) mi-na-katta.
Taroo did not see a movie with Hanako.
(43) a. (taroo wa toshokan de) hon wa -ana-katta.
TP library in book read-NEG-PST
In the library, Taroo did not read a book.
b. (taroo wa) toshokan de wa (hon o) yom-ana-kat-ta.
Taroo did not read a book in the library.
(44) a. (taroo wa kinoo) gakkoo e wa ik-ana-katta.
TP yesterday school to go-NEG-PST
Yesterday, Taroo did not go to school.
b. (taroo wa) kinoo wa (gakkoo e) ik-ana-katta.
Taroo did not go to school yesterday.
The (a)-sentences in (42)-(44) have a VP-scope reading as well as an
NP-scope reading, depending on whether the sentences are associated with
a set of NPs or a set of possible actions. The (b)-sentences in (42)-(44),
however, give only an NP-scope reading. Sentence (42b), for example,
means only that Taroo saw a movie with someone other than Hanako.
(43b) implies that Taroo read a book somewhere other than in the library.
Certain NPs are felt to be more closely linked to the meaning of verbs.
Object NPs are felt to be more closely tied to the transitive verbs than loca­
tive or commitative NPs. The directional NP is felt to be more closely linked
to the meaning of motion verbs such as iku 'go' than NPs which indicate
"with whom", "how" or "when" one went. It is when wa immediately fol­
lows an NP which is felt to be most closely linked to the semantic content of
verbs that [NP-wa (Neg)] can potentially give a VP-scope reading.
Subject NP-wa generally does not give a VP-scope reading under nega­
tion, but there are exceptions.
(45) ame wa fur-ana-katta.
rain rain-NEG-PST
It did not rain.
178 NAOMI HANAOKA McGLOIN

(46) atama wa itaku-nai.


head hurt-NEG
I don't have a headache.
When we utter sentence (45), we normally do not intend an NP-scope
reading (i.e., it didn't rain but it snowed, etc.) but rather the implication
that some other condition of the weather (e.g., the wind being strong) was
true. Sentence (46) is often used to imply that some other symptoms of ill­
ness exist (e.g., seki ga deru 'one coughs', netsu ga aru 'one has a fever'.
karada ga darui 'one feels tired', etc.). This seems again to be due to the
fact that in (45)-(46) the subject NP and the predicate form a tight semantic
unit.
As we have observed before, wa can be also inserted after a predicate
(verb, adjective or copula).
(47) make wa shi-na-katta.
lose do-NEG-PST
I did not lose (but,...)
(48) zettai make wa shi-nai.
never lose do-NEG
I will not lose.
This type of wa, which we will call "post-predicate wα" is generally
used to mark a predicate as a target of negation (as in (47)), although some­
times it adds an emotional emphasis and conveys a strong denial10 (as in
(48)). Sentence (47) implies that the speaker did not lose but made a lot of
mistakes, for example. Sentence (48), on the other hand, conveys the
speaker's determination not to lose.
Post-predicate wa, moreover, not only marks a verb/adjective/copula
as a target of negation, but sometimes marks other constituents as a target
of negation. Following are some examples.
(49) uchi no kodomo wa yoku yasai o tabemasu
my family GN child TP well vegetable DO eat
ga, otaku no okosan wa?
but your-family GN child TP
My child eats vegetables well, but how about yours?
a. uchi no kodomo wa mainichi chittomo tabe-masen.
TP everyday not-at-all eat-NEG
My child does not eat them (at all) everyday.
THE ROLE OF WA IN NEGATION 179

b. uchi no kodomo wa mainichi wa tab e-masen (ga.)


My child does not eat them everyday, (but).
c. uchí no kodomo wa mainichi tabe wa shi-masen (ga).
My child does not eat them everyday, (but).
(50) a. minna kimas-en-deshita.
all come-NEG-PST
All (of the people) did not come.
b. minna wa kimas-en-deshita.
Not all came.
c. minna ki wa shi-masen deshi-ta.
Not all came.
In (49) and (50), the occurrence of wa changes the meaning entirely. In
(49a) and (50a), mainichi and minna are outside of the scope of negation.
In (49b-c) and (50b-c) — cases with wa, mainichi and minna are the targets
of negation. Sentences (49b) and (49c) both mean that the child eats vege­
tables but not everyday. Sentences (50b) and (50c) mean that some people
came, but not all. Sentences (49c) and (50c), moreover, indicate that post-
predicate wa does mark a constituent which does not immediately precede
wa as a target of negation. Whether this function of a post-predicate wa
extends to ordinary (i.e., non-quantifier-like) NPs or not, however, is a
question yet to be investigated.11

5. Summary
Givón (1978:109) states:
Negatives are consistently more marked in terms of discourse-pragmatic
presupposition, as compared to affirmatives. ...negatives are uttered in a
context where corresponding affirmatives have already been discussed, or
else where the speaker assumes the hearer's belief in — and thus familiar­
ity with — the corresponding affirmative.
The general preference for wa in negative sentences is no doubt a
direct consequence of this presuppositional nature of negative sentences.
Japanese, however, syntactically distinguishes at least two types of negative
use. Non-wα-marked negation is evaluative, and is used to impose the
speaker/writer's subjective value judgment. Wa-marked negation is more
objective or non-evaluative, and is used when the corresponding affirma­
tives can be reasonably assumed from the discourse.
The particle wa plays a crucial role in interpreting wa-marked negative
180 NAOMI H A N A O K A McGLOIN

sentences. The following principles hold for determining the scope of nega­
tion in wa-marked negative sentences. 12
1. In general, the constituent which represents the NEW information
in a sentence is the target of negation. A constituent marked by the thema­
tic wa represents given information and is outside the scope of NEG. A
constituent marked by the contrastive wa represents new information and
hence signals the target of NEG.
2. Wa which follows a predicate is always contrastive. Whether post-
nominal wa is thematic or contrastive is determined along a hierarchy simi­
lar to the Keenan-Comrie accessibility hierarchy.
3. Generally, a constituent which immediately precedes contrastive wa
is the target of negation.
4. Post-nominal wa can give a VP-scope NEG reading if there is a
strong semantic/syntactic tie between an NP and the verb.

NOTES
*Many of the ideas expressed in this article are also discussed in my forthcoming monograph
entitled Negation in Japanese. This article, however, presents some revision and extension of
parts of the monograph. I am indebted to John Hinds, Shoichi Iwasaki and Senko Maynard for
valuable comments on an earlier version of the paper.
1) If wa is interpreted as thematic, sentence (6b) will have a non-implicational reading, while
a contrastive interpretation of wa gives an. implicational reading. The thematic vs. contrastive
uses of wa will be discussed in some detail in section 3.
2) The phrase hanashi wa is in parentheses because this phrase does not appear in the actual
passage. I have supplied this phrase for the sake of clarity.
3) The examples given so far involve only ga and  vs. wa. The above generalization, how­
ever, holds also for other particles, although actual examples are rather scarce.
(i) a. kokontoko shi go nichi uchi e kaer-anai n desu yo.
these-days four five days home to return-NEG NOM COP EX
He has not returned home for four or five days lately,
b. kokontoko shi go nichi uchi e wa kaetteimas-en.
has-returned-NEG
(ii) a. kuruma de ko-na-katta.
car by come-NEG-PST
I did not come by car.
b. kuruma de wa ko-na-katta.
I did not come by car.
Sentences (ia) and (iia) are still felt to be evaluative. By uttering (ia), the speaker presents the
act of not returning home as unusual/strange. Sentence (iia) is more likely to be uttered when
the norm is to come by car. Sentences (ib) and (iib) lack such an implication.
THE ROLE OF WA IN NEGATION 181

4) Inoue assigns the following information structures to sentences involving thematic wa as in


(iii), exhaustive listing ga as in (iv), and neutral description ga as in (v).
(iii) hanako wa kaze o hii-ta.
OLD NEW
Hanako caught a cold.
(iv) a. Dare ga kaze o hii-ta.
who SB cold DO catch-PST
Who caught a cold?
b. hanako ga kaze o hii-ta.
NEW OLD
(v) a. kinoo nanika at-ta no?
yesterday something happen-PST QU
Did anything happen yesterday?
b. hanako ga tazunete kita.
NEW NEW
Hanako came to visit.
5) For more detail, see Kuno (1973).
6) Wa in de wa nai is contrastive in its function. Martin (1975:373) distinguishes ja nai/de nai
from de wa nai. According to Martin, de wa nai is "focused" while no "nuclear focus" is felt with
ja nai/de nai. This is a legitimate distinction, and can be further supported by the following sen­
tences.
(vi) ano hito ga hon kaw-ana-kereba, watashi mo kai-masen.

that person SB book buy-NEG-if I also buy-NEG


If that person does not buy a book, I will not buy it either.
(vii) kanemochi -kereha, kekkon shi-masen.

rich COP NEG-if marry do-NEG


I won't marry if (he/she) is rich.
The contrastive wa is not acceptable in a conditional sentence such as (vi). Similarly, de wa nai
is unacceptable in (vii) while de nai/ja nai are fine. This illustrates that de wa nai and ja nai/de
nai are not simply stylistic variants.
7) With what Horn (1978) calls scalar predicates, wa can also indicate 'less than'. So, sentence
(25b) can imply that the textbook is less than new, although it is not old.
8) Shoichi Iwasaki (personal communication) has indicated that the implied contrast is not
necessary between two adjectives, as can be seen from the following examples.
(viii) kowaku wa -i kedo, me  tsubutte-shimat-ta.
scary NEG but eye DO clos-end-up-PST
I was not scared, but I closed my eyes.
(ix)  kyookasho wa atarashiku wa -i kedo, yoku ureteinı.
this textbook TP new NEG but well sold
This textbook is not new but is selling well.
182 N A O M I H A N A O K A McGLOIN

In (viii) and (ix), however, the implied contrast still holds between two predications of the same
topic, and not between two independent propositions.
There are, however, rare cases where wa implies a contrast between two propositions, as in
the example cited in Kuno (1973:46).
(χ) ame wa futteimasu ga, kasa wa motteikimas-en.
rain falling but umbrella bring-NEG
It is raining, but I won't take my umbrella with me.
9) [Obj of Prep] is a large category and needs to be broken down further in Japanese. Inoue
(1976:187) presents the following more detailed hierarchy for Japanese.
Subject > Dir obj > Indir obj > Loc NI > Loc  >
Dir E/NI > Loc DE > "jokaku" DE > "kijunkakiŕ DE >
"dakkaku" > Possess > Origin > "zuikaku" > reason >
Obj of Comparative Prt
10) The emphatic particle wa is often contracted to ya when it follows a verbal stem, as in
"make ya shinai."
11) Native speakers' judgment on the following sentences vary greatly.
(xi) ?osake  nomi wa shi-na-katta ga, koohii o takusan non-da.
liquor DO drink do-NEG-PST but coffee DO a-lot drink-PST
We did not drink liquor, but drank lots of coffee.
(xii) ?kinoo wa ginkoo e iki wa shi-na-katta. yunion e it-ta.
yesterday TP bank to go do-NEG-PSTunion to go-PST
Yesterday, I didn't go to a bank, but went to the Union.
Most people find these sentences awkward, and that would indicate that a post-verbal wa does
not extend to an NP in a sentence.
12) Kuno claims that the scope of a negative morpheme nai normally includes only the verbal
(i.e., verbs, adjectives and X dal desu) which immediately precedes nai. He also states that
other constituents can be included in the scope of nai only if it is the focus of a sentence which
has a "multiple-choice" type information structure. For more details on this, see Kuno (1980 and
1983)

REFERENCES

Givón, T. 1978. "Negation in language: pragmatics, function, ontology."


Syntax and semantics 9: pragmatics, ed. by P. Cole, 69-112. New York:
Academic Press.
Horn, Lawrence R. 1978. "Some aspects of negation." Universais of human
la,igi age, ed. by Greenberg, 127-210. Stanford University Press.
Inoue, Kazuko. 1976. Henkei-bunpoo to nihongo. Tokyo: Taishuukan.
(ed.) 1983. Nihongo no kihon-koozoo. Tokyo: Sanseidoo.
Keenan, Edward L. and Bernard Comrie. 1977. "Noun phrase accessibility
and universal grammar." Linguistic Inquiry, 8, 63-99.
THE ROLE OF WA IN NEGATION 183

Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge,


MA: The MIT Press.
1976. "Subject, theme, and the speaker's empathy — A reexamination
of relativization phenomena." Subject and topic, ed. by Charles Li, pp.
417-444. New York: Academic Press.
. 1980. "The scope of the question and negation in some verb-final lan­
guages." CLS 16, 155-169.
1982. "The focus of the question and the focus of the answer." CLS
18, 134-157.
1983. Shin nihon bunpoo uu. Tokyo: Taishuukan.
Kuroda, S.-Y. 1965. Generative grammatical studies in the Japanese lan­
guage. Ph.D. Dissertation, MIT.
Labov, William. 1972. Language in the inner city. Philadelphia: University
of Pennsylvania Press.
McGloin, Naomi Hanaoka. 1972. Some aspects of negation in Japanese.
Ph.D. Dissertation, University of Michigan.
1976. "Negation." Syntax and semantics 5: Japanese generative gram­
mar. ed. by M. Shibatani, 371-419. New York: Academic Press.
. 1986. Negation in Japanese. Edmonton: Linguistic Research, Inc.
Ross, Claudia. 1978. "The rightmost principle of sentence negation." CLS
14, 416-429.
WA AND THE WH PHRASE*

SHIGERU MIYAGAWA
Ohio State University

Kuno (1973) distinguishes between the "thematic" and the "contrastive"


uses of the particle wa in Japanese. The goal of this paper is to formulate a
unified hypothesis of wa from which we can derive these uses of the particle.
We will claim that the one essential property of wa is that it is "set-
anaphoric." By this is meant that wa always anaphorically refers to a con-
textually determinable set of individuals. The argument for this set-
anaphoric property of wa comes from an unexpected source: those con­
structions in which wa attaches to a WH phrase. It is generally assumed that
such a construction is unacceptable. However, a WH wa question is used by
native speakers in certain restricted contexts (e.g. Martin 1975:61). By
identifying the conditions that make a WH wa question appropriate, we
arrive at the essential property of wa as being set-anaphoric.
Another goal of this paper is to formulate a unified hypothesis on the
uses of the particle ga based on our analysis of wa. Our concern is not the
syntax of ga, but the various uses of gα catalogued by Kuno (1973). We will
claim that, while wa is set-anaphoric, thus requiring a predetermined set of
individuals, the essential property of ga is that it creates a set.
Section 1 will discuss the WH wa question. In Section 2, we will analyze
non-WH wa sentences based on the hypothesis put forth in the first section.
In Section 3, we will present an analysis oï ga as a set-creating particle.

1. Wa and the WH Phrase


Kuno (1973:44-49) distinguishes two types of wa, "thematic" and "con­
trastive." A thematic wa phrase is either anaphoric or generic, and it is
never stressed, while a contrastive wa phrase need not be anaphoric or
generic, and it is always stressed. (1) and (2) respectively illustrate thematic
186 SHIGERU MIYAGAWA

and contrastive uses of wa. A discussion of "generic" wa, illustrated in (lb),


is postponed until Section 2.
(1) a. jon wa hon   onda.
John TP book DO read
As for John, he read a book.
b. kujira wa honyuu-doobutsu desu.
whales TP mammals COP
Whales are mammals.
(2) ame wa futteimasu ga, yuki wa futteimasen.
rain TP falling but snow not-falling
It's raining, but it isn't snowing.

1.1. Conditions on the occurrence of WH wa


The wα that can attach to a WH phrase 1 must always be contrastive.
This is because the thematic wa phrase (used anaphorically) refers to a defi­
nite individual or a definite set of individuals, and a WH by nature indicates
no such reference.
(3) *dare wa kita no?
who TP came QU
*Speaking of whom, did he/she/they come?
(4) dare wa kite, dare wa konakatta no?
who TP come:GER who TP didn't-come QU
Who came, and who didn't?
WH wa and the regular, contrastive NP wa share one feature. In both,
it must be the case that the knowledge of a set of definite individuals is
shared by the speaker and the hearer in the immediate conversation con­
text. (5) illustrates this for NP wa while (6) illustrates this for WH wa.
(5) Speaker A: taroo to hanako ga saakasu e ikitagatteita
Taro and Hanako SB circus to wanted-to-go
kedo, hitori shika ikenakatta soo da.
but one only couldn't-go hear COP
Taro and Hanako were wanting to go to the circus,
but I heard that only one of them was able to go.
Speaker : taroo wa itta kedo, hanako wa ikanakatta.
Taro TP went but Hanako TP didn't-go
Taro went, but Hanako didn't go.
WA AND THE WH PHRASE 187

(6) Speaker A: (Same as 5)


Speaker : dare wa itte, dare wa ikanakatta no?
who TP go:GER who TP didn't-go QU
Who went, and who didn't?
In both (5) and (6), the ''contrastive" utterance by Speaker  is set up
by the utterance of Speaker A. Specifically, Speaker A introduces into the
conversation a set of definite individuals [Taro, Hanako] to which some
property is associated — that one of them went to the circus while the other
didn't. In (5), Speaker  volunteers the information that Taro went but
Hanako didn't. In (6), Speaker  does not know which of the members of
the set went and which didn't, and is seeking this information with the WH
wa question. If there is no identifable set of definite individuals in the con­
versation, neither NP wa nor WH wa is appropriate. 2 Thus, if Speaker A
utters the following sentence in (7a), neither of the utterances by Speaker 
in (7b) or (7c) is acceptable.
(7) a. futαri saakasu e ikitagatteiru hito ga ita kedo,
two circus to want-to-go person SB existed but
hitori shika ikenakatta soo da.
one only couldn't-go hear COP
There were two people who wanted to go to the circus, but I
heard that only one was able to go.
b. *taroo wa itta kedo, hanako wa ikanakatta yo.
 *dare wa itte, dare wa ikanakatta no?
Unlike the examples in (5) and (6), (7a) gives only the information that
there were two people who wanted to go to the circus without identifying
those two individuals. For speaker  to utter (7b) is unacceptable because
neither Taro nor Hanako has been referred to in the conversation. The
reason for the inappropriateness of WH wa in (7c) is less transparent; we
will discuss it later. For the moment, let us simply stipulate the condition
that WH wa and contrastive NP wa presuppose an identifiable set of defi­
nite individuals that is shared by the speaker and the hearer in the immedi­
ate conversation context.
The use of contrastive wa with WH differs in one important respect
from the use of the same particle with referential NPs. When used with ref­
erential NPs, it is possible to list all of the individuals being contrasted, or
simply list one of them. But with a WH phrase, only the former is possible.
188 SHIGERU MIYAGAWA

(8) a. taroo wa itte, hanako wa ikana katta.


Taro TP go:GER Hanako TP didn't-go
Taro went, and Hanako didn't go.
b. taroo wa itta.
Taro went.
c. taroo wa itta no?
Did Taro go?
(9) a. dare wa itte, dare wa ikanakatta no?
Who went, and who didn't?
b. *dare wa itta no?
In (8a), Taro and Hanako are exhaustively compared; in (8b), Taro is
singled out with the assumption that he is being compared to the other indi­
vidual involved; (8c) is the question form of (8b). The WH example in (9a)
parallels (8a) in exhaustively comparing those individuals involved. (9b),
which, unlike (8c), singles out just one of the individuals, is unacceptable.
The acceptability of (9a) and the unacceptability of (9b) give evidence for
another condition on WH wa: every member of the shared set must be rep­
resented in the WH wa question. We thus have two conditions on the
appropriate use of WH wa, summarized in (10).
(10) Conditions for the Appropriate Use of wa with a WH Phrase
1. The speaker and the hearer share the knowledge of the existence
of an identifiable set of individuals in the immediate conversa­
tional context;
2. Every member of this set must be exhaustively represented in the
Wh wa question.
According to these conditions, the speaker, in uttering WH wa ques­
tion such as (9a), first of all assumes the existence of an identifiable set such
as [Taro, Hanako]; the speaker also assumes that the hearer shares this
knowledge. Also, the speaker knows that one went and the other did not,
so that the question in (9a) covers the entire set by questioning "who went
and who didn't?"
To further illustrate Condition 2, observe the following.
(11) dare wa sanji ni itte, dare wa yoji ni itta
who TP 3 o'clock at go:GER who TP 4 o'clock at went
no?
QU
By Condition 1, we assume that both the speaker and the hearer of this
WA AND THE WH PHRASE 189

question already assume an identifiable set, say, [Taro, John, and Mary]. By
Condition 2, which requires every member of the set to be represented in
the question, it must be the case that some of the members went at 3, the
other(s) at 4. For example, Taro and John went at 3 and Mary at 4. It can­
not be the case that Taro went at 3, John at 4, and Mary at some other time
because this excludes Mary from being represented by the question, and
this would violate Condition 2. If the speaker knows that the three mem­
bers went at three different times, all three must be contrasted.
(12) dare wa sanji ni itte, dare wa yoji ni
who TP 3 o'clock at go:GER who TP 4 o'clock at
itte, dare wa goji ni itta no?
go:GER who TP 5 o'clock at went QU
Who went at 3 o'clock, who went at 4 o'clock, and who went at
5 o'clock?
The following is yet another illustration that Condition 2 must be
adhered to without exception. The utterance by Speaker A sets up a defi­
nite set of individuals [Taro, John and Hanako].
(13) Speaker A: taroo to jon to hanako ga betsubetsu-ni
Taro and John and Hanako SB separately
itta yo. hitori wa sanji ni itte,
went EX one TP 3 o'clock at go:GER
moo hitori wa yoji ni itte,
more one TP 4 o'clock at go:GER
sannin-me wa goji ni itta soo da.
third-one TP 5 o'clock at went hear COP
Taro and John and Hanako went separately. I heard
that one went at 3, another at 4, and the third at 5.
Speaker :* taroo wa sanji ni itta kedo,
Taro TP 3 o'clok at went but
dare wa yoji ni itte,
who TP 4 o'clock at go:GER
dare wa goji itta no?
who TP 5 o'clock at went QU
Even though all three members of the shared set are represented in
190 SHIGERU MIYAGAWA

speaker B's utterance, only two of them are found in the WH question por­
tion because taroo is explicitly excluded from it. This comprises a violation
of Condition 2 because not all members of the shared set are exhaustively
represented in the WH question.

1.2. Wa as "Set Anaphoric"


Why is it that the two conditions given in (10) exist, and why must they
be adhered to so strictly for WH wal Let us first consider Condition 1,
which requires that the knowledge of a set of identifiable individuals be
shared by the speaker and the hearer in the immediate context in which
a WH wa question is uttered. What exactly is the relation between this pre­
determined set of individuals and WH wal Surely, WH wa is making some
kind of reference to the set. Let us assume that this reference is a type of
anaphoric relation that holds between WH wa and the set or the individuals
that make up the set.
To see the type of anaphoricity involved in a WH wa construction, let
us first consider a simple case of anaphoric reference.
(14) taroo wa hon o katta.
Taro TP book DO bought
As for Taro, he bought a book.
Here, taroo is the theme marked by wa. Underlying this use of wa is
the assumption that there is a definite individual, Taro, who is known to
both the speaker and the hearer. Usually, this shared knowledge is
obtained in the conversation. The important point is the definiteness of the
reference; this is required for anaphoric reference to be successful. To take
another example, a simple NP in Japanese is either singular or plural, and
either definite or indefinite. When used anaphorically in a thematic con­
struction, however, it can only be interpreted as definite. The example
below thus has the two possible interpretations shown.
(15) hon wa shimaimashita.
book TP put-away
(i) As for the book, (I) put it away,
(ii) As for the books, (I) put them away.
Returning to WH wa, where do we locate the item responsible for the
anaphoric relation between the sentence and the predetermined set of indi­
viduals? It certainly cannot be the WH phrase itself since a WH phrase by
WA AND THE WH PHRASE 191

nature makes no definite reference. We are therefore forced to identify the


particle wa as the source of this anaphoricity. Now, it cannot be the case
that wa makes reference to the individual members of the set because wa
itself does not mean Taro, Hanako, he/she/they, etc. The only other identi­
fiable entity that can be the "antecedent" of wa is the set itself. This, we
claim, is the correct characterization. Because the particular kind of
anaphoric relation identified for (WH) wa is different in nature from simple
cases of anaphoricity, we will term it "Set Anaphoricity." The requirement
for the referent of a set-anaphoric relation is the same as for a simple case
of anaphoric relation: the set must be identifiable and in no way vague or
incomplete. If an incomplete or a vague set is introduced into the conversa­
tion, WH wa cannot be used.
(16) Speaker A: taroo to hanako to dareka ga saakasu
Taro and Hanako and someone SG circus
ni itta. hitori wa niji ni itte,
to went one TP 2 o'clock at go:GER
moo futαri wa sanji ni itta.
more two TP 3 o'clock at went
Taro and Hanako and someone went to the circus.
One went at 2, and the other two went at 3.
Speaker : *dare wa niji ni itte,
who TP 2 o'clock at go:GER
dare wa sanji ni itta no?
who TP 3 o'clock at went QU
Who went at 2, and who went at 3?
Even though the set [Taro, Hanako, someone] is introduced into the
conversation, the presence of the indefinite NP dareka "someone" keeps
this set from being totally identifiable. Hence, WH wa cannot be used to
set-anaphorically refer to the set.
Let us now turn our attention to Condition 2, which specifies that the
members of the shared set must be exhaustively represented in the WH wa
question for the question to be appropriate. This condition essentially fol­
lows from our characterization of (WH) wa as set-anaphoric. If wa refers
set-anaphorically to the shared set as a whole, it is only reasonable to
assume that each member of the set must somehow be represented. A
192 SHIGERUMIYAGAWA

defective set-anaphoric relation would result if wa refers to the entire set,


but the phrase (WH) to which wa attaches only represents a portion of this
set. We provisionally assume that Condition 2 (as well as Condition 1) fol­
lows from the set-anaphoric property of (WH) wa. However, for the pur­
pose of discussion, we will treat it as an independent condition.
We will now look at how Condition 2 is met in a WH wa question. We
begin with an assumption commonly made in grammatical theory (e.g.,
Chomsky 1981) that a WH phrase is an operator that binds a variable at
some relevant level of representation. A WH question such as (17a) has the
logical form in (17b).
(17) a. Who came?
b. Which person χ, χ came.
(17b) is the syntactic/semantic representation of (17a) before any dis­
course factors come into play. In this representation, the possible value of
the variable is unrestricted. It is only when the sentence is placed in context
that the possibilities for the variable are narrowed down to some assumed
group of people, though even contextual information may not have this
effect if the speaker has absolutely no idea who came.
For WH wa, while the logical form of the sentence is similar to the one
in (17b) — the possible value of the variable is unrestricted — once the set-
anaphoric property of wa is brought into play, the range of possible values
becomes highly restricted. This range is narrowed to those members of the
set set-anaphorically designated by wa. This process occurs at the post-logi­
cal form level. For a WH wa example such as (4), repeated below as (18a),
the logical form representation is (18b) with the set-anaphoricity of wa
incorporated.
(18) a. dare wa kite, dare wa konakatta no?
who TP come:GER who TP didn't-come QU
b. SET: [Taro, Hanako]
Which χ, χ member of [Taro, Hanako], x came; and
which y, y member of [Taro, Hanako], y didn't come.
With regard to Condition 2, we can now state the following: in order
for every member of the set to be exhaustively represented in a WH wa
question, the variables bound by the WH phrase must together range
exhaustively over the members of the set. If the possible values of the vari­
ables are such that their combined range is less, or more, than the set mem-
WA AND THE WH PHRASE 193

bers, the sentence is unacceptable. We saw in (13) an example where the


possible values of the variables were restricted to a subset of the set set-
anaphorically referred to by wa. The logical form of (13) with the set-
anaphoric wa incorporated is as follows:
(19) SET: [Taro, John, Hanako]
*Which χ, χ member of [John, Hanako], χ went; and
which y, y member of [John, Hanako], y went.
By restricting the possible value of the variables to just the members of
the set, the set is "filled out," thus making set-anaphoric reference to the
set appropriate with wa.

1.3. Content of the Shared Set


From Condition 1, which is required by the set-anaphoric property of
wa, it follows that the shared set for a WH wa question must have more
than one member. It does not make sense to ask "which individual x" if
both the speaker and hearer know the individual being talked about, i.e.,
the value of χ ranges over just one individual already presupposed in the
immediate conversational context. To ask a WH wa question where the var­
iable ranges over just one individual therefore is tantamount to asking
about something the speaker already knows, a blatant violation of coopera­
tive principle of conversation (cf. Grice 1975). From the combination
of Conditions 1 and 2, it follows that the members of the set must be
contrasted. If all of the members are associated with the same property,
say, the property of "went," it again does not make sense to ask "which
individual x" because the possible value of χ is already known (by Condi­
tion 1), and Condition 2 requires that every member of the set must be
exhaustively represented in the question. It would again amount to asking
a question the answer to which is already assumed.
Below, we will show that WH wa is possible for any type of phrase for
which a referential phrase + wa is possible so long as the contrast is approp­
riately set up.
The particle wα, whether "thematic" or "contrastive," can attach to
virtually any non-predicate phrase (we will use "contrastive" wa for the
examples). 3
194 SHIGERU MIYAGAWA

(20) Noun Phrase


a. taroo wa hon  onda ga, ...
Taro TP book DO read but
Taro read a book, but...
b. hon wa taroo ga onda ga, ...
book TP Taro SB read but
The book, Taro read it, but...
(21) Postpositional Phrase
a. gakkoo ni wa ikanai ga, ...
school to TP won't-go but
(I) won't go to school, but...
b. ano resutoran de wa itsumo sushi o taberu ga, ...
that restaurant at TP always DO eat but
(I) always eat sushi at that restaurant, but...
 niji kara wa chotto muri da ga, ...
2 o'clock from TP little inconvenient COP but
From 2 o'clock is a little inconvenient, but...
For WH wa, NPs are acceptable without a problem.
(22) a. dare wa hon  yonde, dare wa zasshi 
who TP book DO read.GER who TP magazine DO
onda no?
read QU
Who read a book, and who read a magazine?
b. nani wa katte nani wa kawanakatta no?
what TP buy:GER what TP didn't-buy QU
What did (you) buy, and what didn't (you) buy?
In (22) the contrast is clear cut; the question contrasts those who did
X and those who did Y (=a) and those who did X and those who didn't do
X (=b). The clarity of contrast is essential to the acceptability of a WH wa
sentence. For example, a question such as the following sounds slightly
awkward.
(23) ?Dare wa kite, dare wa hon  yondeiru no?
who TP came who TP book DO reading QU
Who came, and who is reading a book
This awkwardness can be attributed to the difficulty in contrasting
"someone came" and "someone is reading a book." The difficulty arises
WA AND THE WH PHRASE 195

because the predicate kite and hon  yondeiru do not share anything in
common to contrast. However, once it is known that it is presupposed that
some members of the definite set came (say, to a party) while others chose
to be reading a book (instead of coming to the party), the sentence becomes
acceptable. The fewer items that the contrasting phrases have in common,
the more difficult it is to set up an obvious contrast, thus requiring a much
more elaborate presupposition.
(24) ???dare wa niji ni daigaku de hon 
who TP 2 o'clock at university at book DO
karite, dare wa aisukuriimu ga suki  ?
borrow:GER who TP ice cream DO like COP QU
Who will borrow a book at the university at 2 o'clock, and
who likes ice cream?
A WH postpositional phrase likewise is acceptable whether the phrase
to which the postposition attaches is just the WH or a longer phrase con­
taining the WH.
(25) doko ni wa itte, doko ni wa ikanakatta no?
where to TP go:GER where to TP didn't-go QU
Where did (you) go, and where didn't (you) go?
(26) doko de wa hon o utteite,
where at TP book DO is-sold:GER
doko de wa zasshi o utteiru no?
where at TP magazine DO is-sold QU
Where are books sold, and where are magazines sold?
(27) dono honya de wa hon o utteite,
which bookstore at TP book DO is-sold:GER
dono honya de wa shinbun  utteiru no?
which bookstore at TP newspaper DO is-sold QU
Which bookstore is selling books, and which bookstore is selling
newspaper?
(28) dono tomodachi no uchi ni wa itte,
which friend GN house to TP go:GER
dono tomodachi no uchi ni wa ikanakatta no?
which friend GN house to TP didn't-go QU
Which friend's house did (you) go to, and which friend's house
didn't (you) go to?
196 SHIGERU MIYAGAWA

A WH postpositional phrase intuitively appears more natural than a


WH "NP" phrase. This is probably due to the fact that a postpositional
phrase with wa is much more likely to be automatically interpreted as con­
trastive than a NP wa (cf. McGloin (this volume); Saito (forthcoming)).

2. Non-WH Phrase and Wa


We have argued in the previous section that the most fundamental
property of wa in WH wa is that it is set-anaphoric. This property gives rise
to the two conditions stated in (10) which must be met for a WH wa ques­
tion to be appropriate. Now, if this fundamental property of wa as set-
anaphoric is true for WH wa, it is inconceivable that the property should be
limited to just the use of wa with WH.
In this section, we will extend our analysis of wa to non-WH construc­
tions. In doing so, we will adopt the following orientation: we will assume
that set-anaphoricity is the most fundamental property of wa in all its uses,
and investigate the consequences of this for the analysis of non-WH wa con­
structions. The most important consequence, as we will see, is that the
"thematic/contrastive" distinction derives from the stated fundamental
property of wa, in this way making it possible to have a unified hypothesis
of wa. In the next section, we will push our analysis further, and develop
a unified hypothesis of the different uses of the particle ga based on the set-
anaphoric property of wa.

2.1. Contrastive Wa
Let us begin with a simple case of non-WH contrastive wa. Suppose
that the following is uttered in a conversation in which Taro and Hanako
have been the topic, i.e. Taro and Hanako contextually comprise an identi­
fiable set.
(29) taroo wa uchi e kita kedo, hanako wa konakatta.
Taro TP house to came but Hanako TP didn't-come
Taro came to (my) house, but Hanako didn't come.
By our assumption, wa set-anaphorically refers to [Taro, Hanako].
Unlike the WH case, however, the NPs to which wa attaches are referen­
tial, and they exhaustively refer to the members of the set. Now, suppose
that, instead of (29), the speaker utters the following.
(30) taroo wa uchi e kita.
Taro TP house to came
WA AND THE WH PHRASE 197

Again, wa set-anaphorically refers to [Taro, Hanako]. But the sen­


tence has just one referential NP, taroo, whose referent is a member of the
set. Here taroo wa is clearly used contrastively, even though only one of the
members is explicitly picked out. This contrastiveness arises from the com­
bination of set-anaphoricity of wa, which makes reference to the entire set,
[Taro, Hanako], and the referential NP taroo, which only picks out a por­
tion of the set. By picking out one member of the set, taroo is being con­
trasted with the rest of the set. The set-anaphoric property of wa places
every member of the set "on stage," and, by referentially picking out just a
portion of these members, the referent is set in contrast with those mem­
bers which are not referentially picked out.
We can push the point even further. Recall that in our discussion of
WH wa, it was shown that every member of the set must be represented.
Let us assume that this is a general feature of all instances of wa. This is a
reasonable assumption because once wa set-anaphorically picks out an
identifiable set, then somehow all of the members of the set, and not just a
portion, must naturally become involved. This has one important conse­
quence for sentences such as (30), in which only a portion of the set is refe­
rentially picked out. The contrastiveness evident in the interpretation of
this sentence is not a basic semantic feature of wa, but rather, it is simply a
consequence of the combination of set-anaphoricity of wa and the require­
ment that all of the members must somehow be represented. Because only
one member is referentially picked out, the only way for the remaining
member(s) of the set to "become involved" is to have the one picked out con­
trasted (exhaustively) with every other member of the set. In other words,
by picking out a portion of the set referentially, that portion gains the rela­
tion "IS IN CONTRAST TO" with every other set member, thereby mak­
ing it possible for all of the members to be exhaustively represented even
though only a subset is referentially picked out. We thus derive the "con­
trastive" use of wa from a more basic property of wa, namely, set-
anaphoricity, together with the assumption that every member of this set
must be represented. It is the relation, IS IN CONTRAST TO, which
makes it possible to meet the latter requirement. Indeed, it is reasonable to
assume that this relation holds even in a set where all of the members are
referentially picked out, as in (29). The relation "bonds together" the mem­
bers referentially picked out into a single, identifiable set shared in context.
We have so far discussed the "contrastive" wa construction involving
definite NPs. What about those cases that involve indefinite NPs? These
198 SHIGERU MIYAGAWA

are the cases in which the NP itself does not make any reference. Observe
the following example and discussion by Kuno (1973:48).
(31) watakushi ga shitteiru hito wa paatii ni kimasendeshita.
I SB know person TP party to didn't-come
(i) Speaking of the persons I know, they did not come to the
party.
(ii) (People came to the party, but) there was none whom I
know.
Watakushi ga shitteiru hito4 'people whom I know' is ambiguous; it
can mean either one or more persons whom the speaker knows, whom he
has already talked about, or some persons whom the speaker knows,
whom he probably has not talked about. If the first meaning is the correct
one, wa can be regard as thematic wa [(the first interpretation)]...On the
other other hand, if watakushi ga shitteiru hito is taken as nonanaphoric,
wa...cannot be thematic; it must be contrastive [(the second interpreta­
tion)].
We will discuss thematic wa later. Our present point of interest is
Kuno's description of the "nonanaphoric" interpretation given in (ii). We
assume that the use of wa here is also set-anaphoric. But what exactly is the
set being referred to? Certainly, we cannot assume that a predetermined
identifiable set exists contextually (at least, we need not assume this for the
sentence to be appropriate). What we claim is that the set in this case is the
one created linguistically by the indefinite NP watakushi ga shitteiru hito
"persons whom I know." The set thus produced includes all of the people
that the speaker knows. The individual members of the set are indefinite
and nonspecific, but the set itself is determinable, and wa refers set-
anaphorically to this set. Our assumption associates a slightly different lit­
eral interpretation from the one given by Kuno for the nonanaphoric use of
the NP in (31). The literal sense of (31) is "all of the people whom I know
did not come," as opposed to Kuno's "some people whom I know..." Our
interpretation directly reflects the truth condition of the sentence. When
used nonanaphorically, the sentence describes a state of affairs in which none
of the people whom the speaker knows came to the party. If at least one
person whom the speaker knows did come, the sentence is false.
There is a second possibility for the membership of the set set-
anaphorically referred to in (31). The sentence has a contrastive reading, as
indicated by Kuno, between "the people I know" and "the people whom I
don't know." One way to capture this contrastive interpretation is by plac-
WA AND THE WH PHRASE 199

ing both types of individuals — those that the speaker knows and those that
he/she does not know — into the set set-anaphorically referred to by wa. If
this is the case, the indefinite NP represents only a portion of the set —
those the speaker knows — so that a contrastive relation is set up between
this subset (those the speaker knows) and the rest of the set (those the
speaker does not know). But this is inconceivable. If this really were the
case, the literal sense of the sentence would be that every individual whom
the speaker does not know came to the party, literally billions of people.
The sentence certainly does not imply that. Consequently, the fact that the
sentence implies that (some) people whom the speaker does not know came
to the party is an implication inferred from the entire sentence in context. 5
We assume that an indefinite NP such as the one in (31) can be
analyzed by a logical quantifier-variable relation. 6 The two potential quan­
tifiers are the universal quantification (All x, such that χ is a person I know)
and the existential quantification (There exists x, such that χ is a person I
know). The interpretation we assume is the former. We predict in fact that
wa requires universal quantification when attached to an indefinite NP
because of its set-anaphoric property. The quantifier ALL inherently forms
a complete set while the existential quantification does not. This predicts
that an indefinite NP with existential quantification is always ungrammati-
cal with wa. This is borne out in the following unacceptable sentence.
(32) *aru hito wa kita.
some person TP came
A person came.
The variable bound by ALL ranges over all of the members of the set
set-anaphorically referred to by wa. This quantifier-variable structure is
similar, but not the same, as the WH wa case. In the latter, the speaker and
the hearer share the knowledge of the existence of a definite set of individu­
als. The WH question must be set up in such a way that contrast is presup­
posed among the members of this set. Otherwise, the question would be
asking about something already shared by the participants in the conversa­
tion. In the universally-quantified indefinite NP, there is no definite set of
individuals assumed, hence contrast — either in the form of a question or a
statement — need not be set up. In fact, we predict that no contrast can be
set up with such a set beyond the "informal" inference illustrated for (31),
which does not involve contrast within the set. The only way for an indefi­
nite NP to appropriately act as the set set-anaphorically referred to by wa is
200 SHIGERU MIYAGAWA

if it is universally quantified. This is because it is only universal quantifica­


tion that allows the formation of an identifiable set (All x...). This mode of
quantification precludes any possibility of contrast among the members that
the variable ranges over. 7
We will return to the issue of indefinite NPs with wa after discussing
the use of thematic wa.

2.2. Thematic wa
It is well known that a WH phrase cannot occur with thematic wa.
Thematic wa always requires a referential NP because a theme is anaphoric
(or generic) (Kuroda 1965, 1972; Kuno 1973). We will show that this use of
wa can reduce to the general property of wa so far outlined for WH wa and
"contrastive" wa.
Consider the following thematic wa construction.
(33) taroo wa ashita kuru.
Taro TP tomorrow come
As for Taro, he will come tomorrow.
The sentence presupposes that Taro has been a topic of the conversation,
so taroo is the theme of the sentence, hence anaphoric {taroo wa does not
receive prominent intonation). By our analysis, this wa set-anaphorically
refers to a determinable set. But what is this set? Because taroo is contextu-
ally shared, we assume that the set is made up exclusively of one member,
taroo. Thus, wa refers to the set [Taro]. The word taroo is referential, and
it refers exhaustively to the set. Both conditions are thus met. Because the
entire set is referentially represented, there is no need to set up an IS IN
CONTRAST TO relation in order to represent those members not referen­
tially picked out.
In our system, then, the difference between thematic wa and contras­
tive wa boils down to how the requirement is met that the members of the
set must be exhaustively represented. If only a portion of the set is referen­
tially picked out, it is put in contrast to the other members, resulting in a
contrastive reading. If, on the other hand, the members of the set are
exhaustively referentially picked out, and are associated with the same
property, the result is thematic wa.
WA AND THE WH PHRASE 201

2.3. Definite and Indefinite NPs with wa


Our analysis cuts across the distinction traditionally drawn between
thematic and contrastive wa. The issue is not whether wa is used themati-
cally or contrastively, but rather, how the members of the set set-anaphori­
cally referred to by wa are exhaustively represented. The basic assumptions
are: (a) wa is set-anaphoric; (b) every member of the set set-anaphorically
referred to by wa must be represented.
If the NP to which wa attaches is definite, two possibilities exist. If the
members of the set are exhaustively referentially picked out with this NP,
and one property is associated with the set members, there is no contrast,
and we have "thematic" wa. On the other hand, if only a portion of the set
is referentially picked out, a contrast must be set up in order to represent
those members not referentially picked out. (34) and (35) below exemplify
these two possibilities.
(34) taroo wa kita. (No stress on the wa phrase.)
As for Taro, he came.
(35) taroo wa kita. (Stress on the wa phrase.)
Taro came.
In (34), it is presupposed that Taro has been a topic of the conversa­
tion, so that wa set-anaphorically refers to [Taro], and the word taroo
exhaustively referentially picks out the set member. In (35), it is presup­
posed that there is a set that contains Taro and at least one other individual.
Wa set-anaphorically refers to this set, and because taroo only referentially
picks out a portion of the set, it must be contrasted with the remaining
member(s) of the set.
If wa attaches to a nonspecific indefinite NP, none of the members can
be referentially picked out, hence another means of representing the set
members must be at work. Let us recall the WH wa cases. In such a con­
struction, the members are exhaustively represented by the variables that
together range over all and only the members of the identifiable set. Is
there any difference between the WH wa case and non-WH indefinite NPs?
There cannot be any fundamental difference because in both cases no act of
reference is made by a referential phrase. As we have seen, in the non-WH
indefinite case, we must assume universal quantification. Consequently, in
both WH wa and non-WH indefinite NP wa, variable(s) exist that range
exhaustively over the members of the set. While the WH wa case consti-
202 SHIGERU MIYAGAWA

tutes contrastiveness, the non-WH indefinite NPs are noncontrastive.


The discussion of the use of wa with an indefinite NP is relevant to an
issue briefly mentioned earlier in the paper. This is the point regarding the
use of "thematic" wa in generic statements. We can now state that this use
of wa is nondistinct from the use of wa with indefinite NPs in which the
members of the set are exhaustively represented by one variable bound by
the universal quantifier. Thus, in the generic statement in (lb), kujira wa
honyuu-doobutsu desu "whales are mammals," the set created linguistically
by All χ, χ a whale is set-anaphorically referred to by wa. The expression
honyuu-doobutsu desu "are mammals" is predicated of this universally
quantified set.
The following summarizes the possible uses of wa for definite and inde­
finite NPs for sets that include two members. "ICT" stands for the relation
IN CONTRAST TO.

(36) DEFINITE NP INDEFINITE NP


CONTRASTIVE (a) taroo wa itta ga, ... (d) dare wa itte, dare wa
taroo went, but... ikanakatta no?
(One member of the set Who went, and
is referentially picked who didn't go?
out; the relation ICT (Which x, x member of
"involves" the other [Taro, Hanako], x went;
member.) and which y, y member of
[Taro, Hanako], y didn't
(b) taroo wa niji ni itte, go.)
hanako wa sanji ni itta.
Taro went at 2, and Hanako
went at 3.
(The members of the set are
exhaustively picked out and
explicitly put in ICT relation.)

NON- () taroo to hanako wa (e) rakudai shita gakusei wa


CONTRASTIVE itta. konakatta.
Taro and Hanako went. Students who failed
(The members of the set didn't come
are exhaustively picked (For all x, x a student
out and attributed the who failed, x didn't
same property, itta.) come.)
WA AND THE WH PHRASE 203

These four ways (we group (36a) and (36b) together) of using a set
with wa are mutually exclusive, so that one cannot use the same set for
more than one of the four types in one sentence. The possibility of some of
the pairings is automatically ruled out by their inherent nature. (36a/b) and
(36c) cannot be used in the same sentence because (36a/b) requires the rela­
tion IN CONTRAST TO, thus associating contrasting properties to the set
members, while (36c) referentially picks out every member of the set and
associates one property to them. (36c) and (36d) cannot be paired for the
same reason: (36c) associates the same property to all of the members while
(36d) requires that the set be partitioned into contrasting parts. With these
two possible pairings out of the way, we are left with four possibilities: the
pairing of (36e) with each of the other three, and (36a/b) and (36d). (36e)
cannot be paired with any other use because it alone does not presuppose a
set of definite individuals.
(37) a. *boku ga shitteiru hito wa,
I SB know person TP
jon wa nijі ni itta ga, ...
John TP 2 o'clock at went but ((36e), (36a/b))
(Of) all the people whom I know, John went at 2, but...
b. *boku ga shitteiru hito wa,
I SB know person TP
jon wa nijì ni itta.
John TP 2 o'clock at went ((36e), (36c))
*(Of) all the people whom I know, as for John, he went at 2.
 *kimi ga shitteiru hito wa, dare wa kite,
you SB know person TP who TP come:GER
dare wa kona katta no?
who TP didn't-come QU ((36e), (36d))
(Of) all the people whom you know, who came and who
didn't?
In these examples, the use of wa in the first portion of the sentence
does not assume a set of definite individuals because the NP is indefinite,
nonspecific, and non-WH, while such a set must be assumed for the wa in
the second part of the sentence. The exclusion of these three pairings leaves
only the [(36a/b), (36d)] pairing to deal with. The inappropriateness of this
pairing is shown below.
204 SHIGERU MIYAGAWA

(38) *taroo wa sanji ni itta kedo,


Taro TP 3 o'clock at went but
dare wa yoji ni itte,
who TP 4 o'clock at go:GER
dare wa goji ni itta no?
who TP 5 o'clock at went QU
Taro went at 3, but who went at 4, and who went at 5?
The original reason given in the previous section for the inappropriate-
ness of this sentence explains why this particular pairing is out. Both the ref­
erential NP wa in the first segment and the WH wa in the second segment
presuppose a set of definite individuals. By Condition 2, both uses of wa
must represent the entire set of individuals. But referentially picking out
taroo in the first occurrence of wa leaves only a partial set for the variables
of the WH phrases to range over in violation of Condition 2. The effect is
the same if we reverse the order of the two segments.
(39) *dare wa yoji ni itte,
who TP 4 o'clock at go:GER
dare wa goji ni itta ka shiranai kedo,
who TP 5 o'clock at went QU don't-know but
jon wa sanji ni itta yo.
John TP 3 o'clock at went EX
I don't know who went at 4 and who went at 5, but John went at
3.
Here, while the WH wa in the first segment presupposes that the mem­
bers of the set are exhaustively associated with the property of "went at 4
o'clock" or "went at 5 o'clock," the wa in the second segment contradicts
this by picking out a member from the same set and associating with it the
property "went at 3 o'clock."

2.4. Set-Contrast
Let us consider here the term "contrast." We have been using this term
to refer to the type of relation shown in (36a) and (36b); this is the relation
IN CONTRAST TO that holds between a member or members referen­
tially referred to and the rest of the set, or between referentially picked out
members (the sum of which exhaustively make up the set members) that
are associated with contrasting properties. Two types of "contrastiveness"
WA AND THE WH PHRASE 205

are so far excluded: (a) the type of contrast between the values of variables
χ and y in WH wa cases (cf. (36d)); (b) the type of contrast exemplified in
Kuno's sentence given in (31) in which the sentence as a whole appears to
imply that (some) people whom the speaker does not know came to the
party. We would want to exclude this latter "contrast" because it has
nothing to do with the inherent property of the set, but rather, it is simply
implied by the entire sentence in context. Indeed, we saw the absurd conse­
quence of attempting to cast this contrast as one between the members of
the set: such an interpretation implies that every person (in the world)
whom the speaker does not know came to the party.
What about the other case of contrast, the one between the values of
variables bound by WH phrases? This constitutes a bonafide case of con­
trast within a set because the values of the variables together must cover the
entire set as required by Condition 2. We would want to have this contrast
included in our sense of contrast. What is it about this case of contrast that
is the same as the contrast set up by the referential NPs illustrated in (36a)
and (36b)? In both, the contrast arises from partitioning the set into two or
more parts. For the referential NP, the set is partitioned into the subset(s)
referentially referred to that are associated with contrasting properties. For
WH wa, the set is partitioned into as many subsets as there are WH
phrases. Let us refer to both of these "intra-set" contrasts as "set-contrast"
to distinguish it from the informally inferred contrast shown for the "party"
sentence in (31).
(40) Set-contrast
Partitioning of a set into two or more subsets, the member(s) of
one subset being associated with a property that can be con­
trasted with the property explicitly or implicitly associated with
the member(s) of the other subset(s).
The fact that different properties can be associated with a portion of
the set presupposes the existence of a set of definite individuals whether or
not the actual word itself is definite. As we have seen, such a set is presup­
posed in both the WH wa and referential NP wa set-contrasts.
The summary given in (36) and the subsequent discussion show that
the use of wa is either set-contrastive or non-set-contrastive. This, we
claim, is the most basic distinction to be drawn for the use of wa. Any
apparent "degrees" of contrastiveness, for example, are due to other, extra-
linguistic factors.8 The most fundamental property of wa is that it is set-
206 SHIGERU MIYAGAWA

anaphoric, and whether a contrast is set up or not depends on the way in


which the members of the set are represented.
Our analysis of wa does not drawn a distinction between "thematic" wa
and "contrastive" wa. We associate a singular property to all cases of wa:
wa is set-anaphoric. The set to which wa refers may be contextually pre­
sent, as in the case of WH wa and referential NP wa, or it may be linguisti­
cally created by a non-WH indefinite NP (or a generic NP) to which wa
attaches. In either case, wa does not create the set to which it refers, but
rather, the set itself is created by some other means. The contrastive read­
ing apparent in some uses of wa derives from Condition 2. The referential
NP that referentially picks out a portion of the set is the simplest case. In
order to include all of the members of the set, the relation "IS IN CON­
TRAST TO" is established between the referentially picked out subset and
the other subset(s). For the case involving WH wa, the fact that a set-con­
trast must be set up is due to an additional pragmatic factor: because a set
of definite individuals is already contextually shared, it does not make sense
to ask a question with only one WH, since this is tantamount to asking
about something that is assumed in asking the question. For a WH wa ques­
tion to be appropriate, at least two contrasting properties must exist that
can be matched with the defintie individuals of the set, viz., which of the
members are associated with the property X and which of the members are
associated with the property Y? This sets up the set-contrast. It is at this
point that Condition 2 comes into play. The variables bound by the WH
phrases together must exhaustively range over all of the members of the
shared set.
These two cases together with the one where all of the members of the
set (cf. (36b)) are referentially picked out comprise the only cases of set-
contrast. The other two cases, that of exhaustively referring to the set with
a referential NP in order to attribute the same property (thematic) and
those involving a universally quantified indefinite NP, do not partition the
set into parts. In both, all of the members are associated with the same
property. Any apparent contrast evident in sentences of these two types are
strictly "informal," or non-set-contrastive; it must be inferred from the
entire meaning of the sentence and the context in which it occurs.
In the next section, we will discuss the use of ga on the basis of our dis­
cussion of wa.
WA AND THE WH PHRASE 207

3. On ga
Kuno (1973:51) identifies two types of ga, "descriptive" and "exhaus­
tive-listing." Sentences with descriptive ga "present an objectively observ­
able action, existence, or temporary state as a new event," while exhaus­
tive-listing ga exhaustively lists individuals associated with one property.
According to Kuno, "only the subject of action verbs, existential verbs, and
adjectives/nominal adjectives that represent changing stages can be fol­
lowed by the descriptive ga, while there are no such restrictions in the case
of exhaustive-listing ga [fn.]" (pp. 49-50). The following are taken from
Kuno 1973; the sentences in (41) exemplify descriptive ga while (42)
exemplifies exhaustive-listing ga.
(41) a. jon ga asoko ni tatteiru.
John SB over-there at standing
John is standing over there.
b. tsukue no ue ni hon ga aru.
desk LK top at book SB exist
There is a book on the top of the desk.
c. sora ga akai.
sky SB red
The sky is red.
(42) jon ga gakusei desu.
John SB student COP
John is a student.
As we did for "thematic" and "contrastive" wa, we will attempt to
unify these two uses of ga by identifying a fundamental property of ga for
all uses, and deriving any differences from the interaction of this property
with other sentential and contextual factors.

3.1. Ga as Creater of Sets


The phenomenon of WH wa pointed to a fundamental property of wa:
wa is set-anaphoric. What this means is that wa must have an existing set to
refer to. In most cases, the set is contextually shared between the speaker
and the hearer, though we identified two instances in which the NP to
which wa attaches linguistically creates a set. These are the indefinite NP +
wa and the generic NP + wa.
What we will claim about ga is that, unlike wa, which is set-anaphoric,
208 SHIGERU MIYAGAWA

gďs fundamental function is to create a set. The members of this set are
exhaustively associated with the property represented in the predicate por­
tion of the sentence. For example, in (43) below, the set [Taro, Hanako] is
created by ga, and the property "came" is exhaustively associated with the
members of this set.
(43) taroo to hanako ga kita.
Taro and Hanako came.
A reasonable way to look at (43) is that the property "came" exhaus­
tively describes the members of the set [Taro, Hanako] created by ga. The
following also illustrates this.
(44) a. dare ga kita no?
who came?
b. jon ga kita no?
Did John come?
In (44a), the question can be rephrased as "to whom can the property
'came' be associated?"; and (44b) can be rephrased as "is it the case that
the property 'came' is associated with John?" Note that by our analysis,
every instance of ga involves a complete set. This precludes establishment
of set-contrast. The following pair of sentences, one with WH ga and the
other with WH wa, illustrates this.
(45) a. dare ga niji ni itte, dare ga sanji ni
who SB 2 o'clock at go:GER who SB 3 o'clock at
itta no?
went QU
Who went at 2, and who went at 3?
b. dare wa niji ni itte,
who TP 2 o'clock at go:GER
dare wa sanji ni itta no?
who TP 3 o'clock at went QU
Who went at 2, and who went at 3?
The WH wa example in (45b) assumes a shared set of definite individu­
als, and the variables bound by the WH phrases together exhaustively
range over the set members, thus setting up set-contrast. For (45a), it is
presupposed that someone went at 2 and someone at 3, but these together
need not exhaustively represent a shared set (if such a set is indeed
assumed). Unlike (45b), it is possible that there can be other individual(s)
WA AND THE WH PHRASE 209

who went at another time. Hence, each instance of ga creates a set, so that
in (45a), there are two separate sets at work.

3.2. The Source of "Descriptive" and "Exhaustive-Listing" Interpretations


If ga is fundamentally characterized as a creater of a set, where does
the distinction between "descriptive" and "exhaustive-listing" interpreta­
tions come from? There is no doubt that this difference exists. Here, an
observation made by Kuno (1973:51-52) is instructive. Commenting on
exhaustive-listing ga examples such as:
(46) a. jon ga gakusei desu.
John SB student COP
It is John who is a student.
b. saru ga ningen no senzo desu.
monkey SB mankind LK ancestor COP
It is the monkey that is the ancestor of mankind.
c. jon ga nihongo o shitteiru.
John SB Japanese DO know
John (and only John) knows Japanese.
Kuno points out that the sentences are "awkward, if not ungrammatical,
out of context. They require contexts that solicit exhaustive listings such as
the following."
(47) a. dare ga gakusei desu ka?
who SB student COP QU
Who is a student?
b. nani ga ningen no senzo desu ka?
what SB mankind LK ancestor COP QU
What is the ancestor of mankind?
 dare ga nihongo  shitteiru ka?
what SB Japanese DO know QU
Who knows Japanese?
Compare those in (46) with the "neutral description" characteristic of
"descriptive" ga in (48).
(48) A, jon ga kita.
Oh, John came.
This sentence does not require a prior utterance that appropriately sets
it up. The sentence, for example, can be the very first utterance of dis-
210 SHIGERU MIYAGAWA

course, unlike the case of exhaustive-listing ga.


The fundamental difference between those in (46) (exhaustive-listing)
and (48) (descriptive) lies in the presupposition — or the lack of presuppos­
ition — about the property represented in the predicate portion of the sen­
tence (e.g. "is a student"). If the knowledge of the existence of the property
is shared in the immediate discourse between the speaker and the hearer,
the ga phrase receives the "exhaustive-listing" interpretation, but if no such
shared knowledge exists, it receives the "descriptive" interpretation. For
the latter, it makes sense that the utterance must represent "an objectively
observable action, existence, or temporary state" because there are no pre­
suppositions, thus the only way to anchor the utterance in context is to rep­
resent a state of affairs that can be directly observed. The following sum­
marizes the two uses of ga.
(49) a. Descriptive
There exists a property X, which is associated with the set
created by ga;
b. Exhaustive-Listing
The property X, the existence of which is a part of the
shared knowledge in the immediate conversational con­
text, is associated with the set created by ga.
In both, the property X is "exhaustively" associated with the members
of the set. There is, however, a difference in where the focus of the sen­
tence is placed: in (49a), nothing is presupposed, hence no portion of the
sentence is emphasized over the other portions; in (49b), the predicate
(property X) portion is already assumed, thus it is natural to view the ga
portion as receiving the focus because it is the portion that contains "new
information." The "exhaustive-listing" interpretation is thus another way of
saying that the set created by ga is being emphasized. 9

3.3. On wa and ga
As our final discussion of ga, we will compare it with wa to give further
evidence that the fundamental function of ga is to create a set. Recall from
our discussion of WH wa that a sentence such as the following is anomal­
ous.
(50) *jon wa hanako ga suki da kedo,
John TP Hanako DO like COP but
WA AND THE WH PHRASE 211

dare wa yoshiko ga suki de,


who TP Yoshiko DO like COP:GER
dare wa toshiko ga suki  ?
who TP Toshiko DO like COP QU
John likes Hanako, but who likes Yoshiko, and who likes
Toshiko?
The reason for the anomaly is that the variables bound by the WH
phrases together must range exhaustively over the entire shared set. In
(50), by virtue of excluding jon from the WH question portion, only a par­
tial set is represented by the WH wa question. Now, note that the sentence
becomes acceptable if we replace WH wa with WH ga.
(51) jon wa Hanako ga suki da kedo,
John TP Hanako DO like COP but
dare ga yoshiko ga suki de,
who SB Yoshiko DO like COP:GER
dare ga toshiko ga suki na no?
who SB Toshiko DO like COP QU (Same as (50))
Here, there are three sets at play. The first set, which concerns the wa
phrase, can be either [John] or [John, Taro, ...]. If we choose the former,
the interpretation is "thematic" because the set is exhaustively referentially
represented; if we choose the latter, it is set-contrastive.
The other sets are those created by the two instances of ga. Unlike wa,
ga is not required to represent a pre-existing set. Whether we choose the
single-membered set or the multiple-membered set for wa, the referent of
ga is independent of this set in the sense that it simply makes up its own set
to associate a property to.
A more convincing illustration of the set-creating property of ga is
shown in the following examples. 10
(52) dare wa nani ga suki de,
who TP what DO like COP:GER
dare wa nani ga kirai na no?
who TP what DO dislike COP QU
Who likes what, and who hates what?
b. *dare ga nani wa suki de,
dare ga nani wa kirai na no?
Who likes what, and who hates what?
212 SHIGERU MIYAGAWA

For (52a), we assume a shared set of definite individuals some of whom


like something (say, pizza) and the others hate something (say, sushi).
(52a) is fine with this assumption because all of the members of this set are
represented by the variables bound by WH (wa). In (52b), for the WH wa
portion to be acceptable, we must assume a shared set of two definite
objects, one of which is liked (by someone) and the other disliked (by
someone else). This assumption, which is reasonable, does not make the
sentence acceptable. By our analysis of ga, the reason is that the two occur­
rences of ga create two different sets. This causes the two variables bound
by the WH phrases to range over different sets. This is shown below.
(52) b'. Shared set (for WH wa): [pizza, sushi]
Set 1 = dare ga1... Set 2 = dare ga 2 ...
which χ, χ member of which y, y member of
[pizza, sushi], χ is liked [pizza, sushi], y is disliked.
By separating the range of the variables into two sets, the sentence fails
to meet Condition 2, which states that the members of the shared set must
be exhaustively represented. What we see in (52b') is that the shared set
[pizza, sushi] is only partially represented in each of the two sets created
by ga. In other words, ga, by its set-creating property, pulls apart the origi­
nal shared set, thereby making WH wa inappropriate. The difference
between the acceptable (52a) and the unacceptable (52b) is then one of set
inclusion/exclusion. In (52a), because the ga phrase is within the domain of
the WH wa (i.e. wa precedes ga), the sets created by ga are included within
the larger set shared by the speaker and the hearer. The important point for
WH wa is that this larger, shared set must be kept intact so that the variable
can range exhaustively over the entire set. But in (52b), the larger, shared
set is "destroyed" by the fact that the ga phrase precedes the wa phrase.
Because of the ordering of the phrases, in which WH wa is included within
the domain of WH ga, each of the variables bound by the WH can only
range over the set created by the ga that immediately precedes (or c-com­
mands) it. This precludes the two variables from together exhaustively
ranging over the original shared set. Consequently, the set that each vari­
able ranges over excludes a subset of the original shared set which results in
an unacceptable sentence.
We can see the same effect in a sentence with only one set of WH.
(53) *jon ga nani wa suki de, tarooga nani wa kirai
John SB what TP like COP:GER what DO dislike
WA AND THE WH PHRASE 213

 no?
COP QU
What does John like, and what does Taro hate?
A sentence with a noninitial wa phrase tends to be awkward. However,
(53) and (51b) are completely unacceptable because the variables, χ and y,
bound by the two WH phrases cannot together range over the original
shared set.
Note, finally, that the following is awkward.
(54) ??dare wa nani wa suki de,
who TP what TP like COP:GER
dare wa nani wa kirai na no?
who TP what TP dislike COP QU
Who likes what, and who hates what?
The difficulty with this sentence arises not from breaking up an origi­
nally shared set (there are in fact two such sets), but from the complexity
involved in interpreting the sentence. Let us assume two shared sets, [Taro,
Hanako] and [pizza, sushi]. The logical form with the set-anaphoricity of
wa incorporated is (55).
(55) Which χ, χ member of [Taro, Hanako], χ likes j ,
j member of [pizza, sushi]; and
which y, y member of [Taro, Hanako], y hates ,
 member of [pizza, sushi].
It is possible to set up a context in which the two sets become shared,
though admittedly it is a forced context.
(56) kinoo taroo to hanako ga kita node,
yesterday Taro and Hanako SB came because
piza to sushi  dashita.
pizza and DO served
hitori wa, dashita ryoori no hitotsu ga suki da to
one TP served food LK one DO like COP QT
itta kedo,
said but
moo hitori wa, dashita moo hitotsu no ryoori ga
more one TP served more one LK food DO
kirai da to itta.
dislike COP QT said
214 SHIGERU MIYAGAWA

Yesterday, because Taro and Hanako came, I served pizza and


sushi. One (of the guests) said that he/she does not like one of
the foods served, and the other guest said that he/she hates the
other food served.
The difficulty with (54) is that it involves a two-step setting of the val­
ues of the variables. One must first decide the values for j and  [pizza,
sushi], which then become the range for χ and  [Taro, Hanako]. This type
of "embedded range" of variables is very difficult to process, though it is
not impossible. Indeed, (54) is slightly better than (52b) (*WH ga WH
wa...), the latter being completely uninterpretable.

4. Summary and Discussion


We have proposed in this paper that "set-anaphoricity" is the most fun­
damental property of wa. This property, which we arrived at by looking at
WH wa questions, interacts with sentential and contextual information to
give rise to the various uses of wa catalogued by Kuno. We also proposed
that ga has a fundamental property which interacts with sentential and con­
textual information to give rise to its various uses. This property is "set-cre­
ation," by which is meant that ga creates a set. This contrasts with wa
which, by its set-anaphoric property, requires a determinable set to which it
can make reference. Informally, we can diagram the set-anaphoricity and
the set-creation properties of wa and ga as follows.
(57) a. [SHARED SET] + wa --> property or properties (e.g.
"went")
b. Some property (e.g. "is a student") --> [Created Set] + ga
Our proposal relates to the important work by Kuroda (1965, 1972,
1976) regarding the notion of "subject" in a theory of judgments. Accord­
ing to Kuroda, a wa phrase is the "subject" in a predicational judgment (or
"categorical" judgment). It has the structure "If χ is A, A is  (is true),"
where A stands for the phrase to which wa attaches, and is the logical sub­
ject, and  represents the property associated with A. 11 The set-anaphoric
property of wa "assists" the predicational judgment by narrowing the possi­
bility of "x" in "If χ is A..." to those members of the set set-anaphorically
designated.
For ga, what Kuroda states in one sense assumes the set-creating prop­
erty of the particle. He states that the particle does not designate a "sub-
WA A N D T H E W H P H R A S E 215

ject" in the sense used for wa, but rather, a ga-sentence is a "subjectless
description" of an event or an action ("thetic" judgment). Our analysis of
ga as set-creating accords well with this proposal. Because a set is not (or
need not be) presupposed, ga makes it possible to associate a property (an
event or action) exhaustively to some individual(s) by forming a set that
includes the individual(s); the property is associated with the set as a whole.
The fact that ga does not require a predetermined set, but rather creates a
set of its own accord, parallels Kuroda's observation that a ga-sentence is
"subjectless," in the sense that a ga-sentence does not predicate of a pre­
determined set ("subject").

NOTES
*I am grateful to John Hinds, Shoichi Iwasaki, Akio Kamio, and Chisato Kitagawa for useful
discussions regarding the content of this paper. I am also grateful to Fumiko Harada for first
pointing out to me that wa can occur with WH phrases. The research for this paper was partially
supported by a Grant-In-Aid from the College of Humanities at Ohio State University.
1) Henceforth, we will abbreviate "WH word/phrase" as simply "WH."
2) The "definite set" need not necessarily be explicitly introduced into the conversation. It
is also possible that, by uttering a contrastive sentence, it inherently implies an identifiable set
(these are the cases where the contrastive wa phrase needs to be "anaphoric" as noted by Kuno
(1973)). We will see cases of implied identifiable sets for contrastive wa involving indefinite
phrases in Section 2.
3) The particle wa does attach to predicate phrases also. We will not deal with these cases.
See Kuroda 1965 among others for discussion of constructions where wa attaches to predicate
phrases. Also, we will not deal with any aspects of the syntax of wa. For this, see Kuno 1973,
Muraki 1974, Kitagawa 1982, and Saito (to appear) among others.
4) The romanization system used by Kuno is different from the Hepburn system used in this
paper. We will convert all examples taken from other authors into the Hepburn system to be
consistent.
5) Part of this inference depends on the "scope" interpretation of negation. For an interesting
discussion of this, see McGloin (this volume) and also Yoshida 1984.
6) For a study of quantification within a grammatical theory, see May 1977.
7) Shoichi Iwasaki (personal communication) has pointed out the following as a possible
counter-example to the claim above that an indefinite NP with wa always has the universal quan­
tification.
(i) watashi ga shitteiru hito wa sannin dake kita.
I SB know person TP three only came
Only three people that I know came.
216 SHIGERU MIYAGAWA

The possible problem here is that while the wa phrase ranges over all of the people "that 1
know," the predicate sannin dake kita identifies three people "that I know." This is not a prob­
lem if we consider that the logical form of (i) is, informally, "of all the people that I know, only
three of them came" (For all χ, χ people I know, only three χ came). The truth condition of this
logical form matches the sentence in (i).
8) See Inoue 1982 for an interesting discussion of how contextual information interacts with
sentential factors to give "degrees of contrastiveness."
9) See Inoue 1982 and Yoshida 1984 for interesting discussions of focus/emphasis for ga. We
exclude from our discussion what Kuno (1973) calls "ga for object marking." It is presumed that
this instance of ga falls into either the "descriptive" or the "exhaustive-listing" use dealt with in
this paper.
10) I am grateful to Akio Kamio for pointing out the following examples.
11) Mikami (1963) proposes a similar, "conditional" representation for wa, though his claim
is not contrasted with any claims about ga.

REFERENCES

Chomsky, Noam. 1981. Lectures on government and binding. Holland:


Foris Publications.
Grice, H.P. 1975. "Logic and conversation." Syntax and semantics 3: speech
acts, ed. by P. Cole and J. Morgan, 41-58. New York: Academic Press.
Inoue, Kazuko. 1982. "An interface of syntax, semantics, and discourse
structures." Lingua 57:259-300.
Kitagawa, Chisato. 1982. "Topic constructions in Japanese." Lingua
57:175-214.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge,
MA: MIT Press.
Kuroda, S.-Y. 1965. Generative grammatical studies in the Japanese lan­
guage. MIT doctoral dissertation.
. 1972. "The categorical and the thetic judgment." Foundations of lan­
guage 9:153-185.
. 1976. "Subject". Syntax and semantics 5, ed. by Masayoshi Shibatani,
1-16. New York Academic Press.
Martin, Samuel. 1975. A reference grammar of Japanese. Yale University
Press.
May, Robert. 1977. The grammar of quantification. MIT doctoral disserta­
tion.
Mikami, Akira. 1963. Nihongo no ronri: wa to ga. Tokyo: Kuroshio Shup-
pan.
WA AND THE WH PHRASE 217

Muraki, Masatake. 1974. Presupposition and thematization. Tokyo:


Kaitakusha.
Saito, Mamoru. To appear. "Three notes on Japanese grammar." Issues in
Japanese linguistics, ed. by T. Imai and M. Saito. Holland: Foris Publi­
cations.
Yoshida, Tomoyuki. 1984. Focus and presupposition and the scope of nega­
tion. International Christian University M.A. Thesis.
PART III: HISTORICAL PERSPECTIVES ON WA
FUNCTIONS OF THE THEME MARKER WA FROM
SYNCHRONIC AND DIACHRONIC PERSPECTIVES

NORIKO FUJII UENO


University of Oregon

Introduction
The complexity of the functions of the so-called theme marker wa has
occasioned much research and controversy in Japanese linguistics. This
paper examines the functions of the particle from both synchronic and dia-
chronic perspectives, and shows how a diachronic study of wa can contri­
bute to our understanding of the particle's functions in its synchronic stage.
This paper consists of two parts. In part I, I will discuss the relationship
between different levels of themes in a discourse and the use of wa in pre­
sent-day Japanese. Part II deals with the use of the particle in Old
Japanese. An examination of a text from Taketori Monogatari "The Wood­
cutter's Tale" (tenth century) and Geriji Monogatari "The Tale of Genji"
(eleventh century) is presented here. I will then make suggestions about the
process of the development of the particle. Part I and Part II together will
clarify how wa in the synchronic stage contains its original function as well
as extended functions.

PART I: FUNCTIONS OF WA IN ITS SYNCHRONIC STAGE

1. Background discussion
Concerning the function of wa in a sentence like (1), the particle wa
has been classified as a "topic" marker or "theme" marker (the terms
"topic" and "theme" have been used interchangeably by most linguists).
But what is the "topic/theme" that this particle indicates?
(1) jon wa gakusei desu.
student is
"John is a student."
222 NORIKO FUJII UENO

Many linguists have discussed "theme/topic" on the sentential level but


with different understandings (e.g. Kuno 1976; Firbas 1966a, 1966b, 1974;
Halliday 1967). Among various discussions on "theme/topic," Chafe's
notion of "frame" is particularly useful in understanding one of the impor­
tant functions of wa.
Chafe (1976:49-53) argues that there are two kinds of topics: one is
English style, and the other is Chinese style. Topics of English style refer to
noun phrases (NPs) that are placed at the beginning of the sentence for the
sake of contrast, such as "the play" in example (2).
(2) a. The play, John saw yésterday.
b. As for the play, John saw it yésterday.
According to Chafe, the so-called topic in English is simply a focus of con­
trast that for various reasons has been placed in an unusual position at the
beginning of the sentence. However, in Chinese, a topic-prominent lan­
guage, there is no requirement that the topic be contrastive, as observed in
the following examples (Chafe 1976:50).
(3) nèi-xie shùmu shù-shen dà
those tree tree-trunk big
"As for those trees, the trunks are big."
(4) nèi-ga rén yáng mίng George Zhang.
that person foreign name
"As for that person, his foreign name is George Zhang."
Chafe designates topics in the Chinese style as "real topics," and notes
that, in contrast to topics in English, Chinese topics function to limit the
applicability of the main predication to a certain restricted domain. In other
words, Chinese topics indicate the frame within which the sentence holds.
Considering the functions of wa, we can say that Japanese wa indicates
a "real topic" in Chafe's terms. In example (1), wa indicates not only that
John is what the rest of the sentence is about but also that John is the frame
within which the main predication has to be interpreted. Wa sets up a spatial,
temporal, or individual framework within which the main predication
holds. That wa sets a framework for the predication in a sentence can be
illustrated by the following examples from Mikami (1975:326):
(5) a. kare wa naijoo o shitteiru kara
he internal situation DO know because
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 223

shinpaishita.
worried
"He worried because he knew the internal situation."
b. kare ga nijoo o shitteiru kara shinpaishita.
"I (or someone else) worried because he knew the internal
situation."
(6) watashi wa katsuragi-fujin no me-no-mae-de kanojo no
I Mrs. Katsuragi in front of her
otto ate no tegami o kaita koto ni hidoku
husband to letter DO wrote that very
manzokushinagara kodomo ga suteppusuru yooni
satisfied while child SB step like
yuki  tokidoki keri yukkuri kubochi no toori 
snow DO sometimes kick slowly potholes street DO
aruiteitta.
walked on
"I slowly walked on a rough street and like a child stepping, I
sometimes kicked up snow, satisfied with the fact that I wrote a
letter to Mrs. Katsuragi's husband in front of her." (From the
novel Banka by Yasuko Harada. In Mikami 1975:326)
In (5a) the person who worried is kare "he." However, when wa is
replaced by the subject marker ga, the sentence no longer conveys the
meaning that the person who knows the internal situation and the person
who worried are the same. Instead, it means that someone else (not kare
"he") worried since he (kare) knew the internal situation. This is because
wa sets a framework for the sentence, limiting the applicability of the pred­
ication to the NP that is marked by wa. Ga, however, being a subject
marker, does not easily extend its applicability beyond the clause, and does
not have the function of setting a frame for the rest of the sentence.
Similarly, in (6), watashi wa semantically corresponds to the final pred­
icate of each clause and of the sentence as a whole, and thus the sentence
reads as it appears in the translation. This is because of the function of wa
to indicate the frame within which the sentence holds. When we change the
wa to ga, the relationship of watashi ga with the final predicate becomes
very remote, and the relationship of the NP becomes limited to the verb
224 NORIKO FUJII UENO

kaita "wrote" in the embedded clause. Therefore, the sentence conveys that
maybe someone else was satisfied and walked slowly.
Was function of setting a frame in (5) and (6) is illustrated below:

Wa can create a frame at different levels in a discourse. Wa may create


a frame only for a clause. In the following example of contrastive wa, the
wa in each clause indicates a frame for the clause and contrasts with the
other (1973:60).
(8) oozei no hito wa paatii ni kimashita ga,
many people to came but
omoshiroi hito wa kimasen deshita.
interesting people did not come
"Many people came to the party, but interesting people did not
come."
Therefore, contrastive wa and thematic wa are not distinct from each other.
In the following paragraphs, I will discuss how wa indicates a frame at
different levels in a discourse and other concurrent functions.

2. Discourse functions of wa
2.1. Theme/frame on the level of discourse
Recently linguists have made significant progress in understanding the
discourse and pragmatic functions of wa. Going beyond the characteriza­
tion on the sentence level that wa is either thematic or contrastive (Kuno
1973), studies such as those by Maynard (1980, 1984) and Hamada (1983)
have shown how the use of wa in a sentence is governed by the organization
of a paragraph/text: namely, that NPs which are chosen to be the theme of
a paragraph are marked by wa when they become the theme of a sentence
in the paragraph. 1 At the same time, it has become clear that the "given"
and "new" distinction which says that wa marks old information (and
shared information between the speaker and hearer) and ga marks new
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 225

information in a discourse is not a decisive factor in controlling the occur­


rence of wa and ga.2
Maynard (1980) argues that the functions of wa cannot be understood
intrasententially since wa is a discourse-organization device. For example,
in the following often-cited sentence, the reason why wa is used after
zoo "elephant" is better understood when we examine the sentence in
context.
(9) a. zoo wa hana ga nagai.
elephant trunk long
Elephants have long trunks,
b. hifu ga haiiro de, gowagowa shiteiru.
skin gray and coarse is
(their) skin is grayish and coarse.
According to Maynard (1980:75), the NP wa, zoo wa in (9a), provides a
thematic territory within which the predication hana ga nagai "trunks are
long" must be interpreted. At the same time zoo wa "elephant" functions to
place a main character of the discourse segment on the thematic stage so
that information provided across the sentential boundary may be inter­
preted in relation to it. This explains why zoo no "elephant's" instead of
zoo wa is inappropriate in (9a). Zoo no cannot cohesively tie into the argu­
ment presented in (9b).
The following examples from Maynard also illustrate that wa is a
device to indicate the writer/speaker's perspective, and to distinguish which
NP is a thematic NP among all possible candidates (1980:87-89). Para­
graphs (10) and (11) share the same plot and the same semantic content.
The only difference between them is the placement of wa. However, this
one difference is a very significant difference.
(10) a. satsujinhan kimura to tanaka junsabuchoo no
murderer Kimura and Tanaka lieutenant's
shoogakkoo jidai no hanashi dearu.
elementary school days of story is
This is a story of the school days of Kimura, the murderer,
and Lieutenant Tanaka.
b. kimura ga yoku gakkoo o sabotta.
Kimura SB often school DO cut
Kimura often skipped class.
226 NORIKO FUJII UENO

c. himura ga sukoshimo benkyoosezu rakudaiten


Kimura SB at all study-NEG falling grade
bakari totta.
only got
Kimura didn't study at all and failed in every exam.
d. tanaka wa kimura no koto  shinpaishite
Tanaka TP Kimura of matter DO worried
benkyoo  tetsudaitai to omotte ita.
study DO want to help Q was thinking
Tanaka worried about Kimura and wanted to help him.
e. shikashi kimura ga sore  kyozetsushita.
but Kimura SB it DO refused
But Kimura refused any help.
f. tanaka wa nandomo kimura no uchi e itta.
Tanaka TP many times Kimura's house to went
g. shikashi itsumo  ni natteshimau no datta.
but always quarrel as end up NOM -PST
But it always ended in a quarrel.
(11) a. satsujinhan kimura to tanaka junsabuchoo no
murderer Kimura and Tanaka lieutenant of
hanashi dearu.
story is
b. kimura wa yoku gakkoo o sabotta.
Kimura TP often school DO cut
c. kimura wa sukoshimo benkyoosezu
Kimura TP at all study-NEG
rakudaiten bakari totta.
failing grade only got.
d. tanaka ga kimura no koto  shinpaishite
Tanaka SB Kimura of matter DO worried and
benkyoo  tetsudaitai to omotte ita.
study DO want to help Q was thinking
e. shikashi kimura wa sore  kyozetsushita.
But Kimura TP it DO refused
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 227

f. tanaka ga nandomo kimura no uchi e itta.


Tanaka SB many times Kimura's house to went
g. shikashi itsumo  ni natteshimau no datta.
but always quarrel as end up NOM -PST
Examples (10) and (11) can be given exactly the same translation. How­
ever, because of the placement of wa, these two paragraphs differ signific­
antly. In (10), tanaka is the theme of the paragraph, and constitutes a
framework within which the event is described. On the other hand, (11)
chooses to have kimura as the theme of the paragraph. Thus, the writer can
choose to make one character be the theme, and in this way indicate his/her
perspective. Based on the observation above, Maynard argues that wa func­
tions as a linguistic device for defining the thematic structure of a para­
graph, and discourse cohesion is accomplished by the use of wa (1980:101-
102). She states:
(12) In the words of Halliday and Hasan (1976:11), the single most critical
meaning relation for creating texture occurs when "one element is inter­
preted by reference to another." This semantic dependency between sen­
tences is the indispensable ingredient for making a cohesive paragraph.
The Japanese theme marker wa is indeed a useful device for this need.
Maynard describes wa's function of establishing a frame for a discourse
as a "staging" device. By "staging" she means "the phenomenon of who
constantly remains on the stage as an important character". 3
Similarly Hamada has demonstrated that the motivation for marking
something as a topic/theme on the sentence level relates to motivations for
topic/theme marking on the level of discourse (1983).
Examples (10) and (11) also illustrate that the choice between wa and
ga in the examples is motivated by the writer's selection of theme rather
than by the distinction between "given" and "new".
It is important to note that an occurrence of wa may carry more than
one function at the same time. Some was indicate a frame/theme on the
paragraph level as well as convey a contrast. For example, ichi no miko wa
"the first prince" in (13), which is from Fumiko Enchi's Genji Monogatari
"The Tale of Genji" (1972), is both +thematic and +contrastive. "+Thema­
tic" here means that the NP is the thematic NP of the sentence as well as of
the paragraph. Example (13) comes right after the description of the birth
of a beautiful baby boy who is later named "Shining Genji." The first prince
is contrasted with Genji. But ichi no miko "the first prince" is also thematic
228 NORIKO FUJII UENO

since this paragraph describes the difference in the way the Emperor treats
the first prince and Genji.
(13) a. ichi no miko wa udaijin no nyoogo ga
first child lady SB
o-umi-ni-natta kata de
Hon-give birth-PST person and
The first prince is a child of Udaijin's daughter,
b. gaiseki no ken'i ga tsuyoku
outside relatives of power SB strong
and his outside relatives' power is strong,
 shitagatte kore-koso machigainaku
therefore this (person)-EMP without fail
tooguu ni tatarem o-kata to
(= crown prince) become-HON HON-person Q
seken de   mite taisetsuni
world in also important think and with much care
o-tsukae-mooshiteiru ga
HUM-is serving but
and therefore thinking, he must be the one who will become
the crown prince, people in the world treat him as an impor­
tant person and are serving him with much care,
d. kono atarashii miko no kagayaku-bakari no
this new child of shining
o-utsukushisa ni wa narabu beku mo nakatta node
RES-beauty to comparable even was not since
but he is not comparable to the shining beauty of this newly
born baby,
e. mikado wa ichi no miko o omotemuki
Emperor first (HON) child DO ostensibly
hitotoori taisetsuni-nasaru keredomo
ordinarily care-HON although
and therefore, although the Emperor ostensibly cares about
the first prince,
f. kono wakamiya wa kakubetsu go-hizoo-ni-nasatte
this young prince specially HON-adore-and
he adores the second prince,
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 229

g. go-jishin de o-sewa-nasaru koto mo hitokata


HON-self by HON-take care NOM also usual
de-wa-nai.
not
and the way he himself takes care of the baby is unusual.
The introduction of the first prince at this particular stage of the story
effectively brings out the contrast between the first prince who has strong
support and the second prince (Genji) who does not have strong support
but is blessed with exceptional beauty. Here, ichi no miko "first child" in
(13a) and kono wakamiya "this young prince" in (13f) have the features of
+thematic, +contrastive, while mikado in (13e) is +thematic, —contras­
tive. Thus, the contrastive function and thematic function of wa are not exc­
lusive of each other.

2.2. Theme creation, theme maintenance and theme retrieval


Maynard points out that wa has the two important functions of theme-
creation (i.e. establishing and signing which NP will be a thematic NP) and
theme-maintenance (i.e. signaling that a NP which has been established
as a thematic NP is still on the thematic stage). After a character is intro­
duced into a story (with the form NP ga if it is the grammatical subject of a
sentence), the character will be marked by wa if it is chosen to be a thema­
tic character of the paragraph. In example (11), wa after kimura in (llb) is
an example of "creative wa," and the instances of wa after kimura in (llc)
and ( l l e ) are examples of "theme maintaining wa."
Not only does wa create and maintain a theme, it also retrieves thema­
tic NPs which were marked by wa earlier in the discourse. Examples (14)
and (15) are segments in a sequence from the Kiritsubo Chapter of Genji
Monogatari by Fumiko Enchi (1972). One of the thematic characters of the
text, the Emperor, is retrieved with wa after his absence from the scene.
The paragraph (14) describes Kiritsubo's parents and how their status
affects Kiritsubo's life at court. The emperor, who was away from the scene
for a while, comes back to the scene with wa in (15).
(14) a. ... hahaoya no kita no kata wa yuisho aru iegara no
mother (= title) good lineage
de no orime tadashii hito de atta.
from good-mannered person was
230 NORIKO FUJII UENO

The mother is a lady of good lineage, and a good-mannered


person.
b. ... kooi jishinwa motoyori
court lady herself (= Kiritsubo) of course
otomo no nyooboo no shoozoku sono hoka і 
accompanying lady of clothes other things also
kokoro-o-kubatte
care
... she gave full attention to Kiritsubo's accompanying ladies'
clothes and other things, let alone Kiritsubo's, and
 kekkoo banji ni sotsunaku
considerably everything without mistake
torimakanatteirareru keredomo
take care-HON although
d. ... shikkarishita kookennin ga nai node
strong supporter there is not since
since there is no strong supporter,
e. korezo to yuu toki ni wa soodan aite 
important occasions consultant even
naku
there is not
in important occasions there is no person to consult
f. kokoroboso soo-de-aru
lonely seem
(and therefore) she seems lonely.
(15) a. zensei kara shukuen ga
previous world from fated relationship SB
kakubetsu fukaku arareta mono dearoo.
particularly deep was (CONJECTURE)
Their relationship must have been strong from the previous
world.
b. kono  no mono to ino omowarenu
this world of creature QT cannot be considered
hodo utsukushii otoko-miko   kooi wa
extent beautiful boy DO this court-lady
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 231

oumininatta.
HON-gave birth
This lady gave a birth to a baby boy who is too beautiful to
be considered to be a creature of this world.
 mikado wa donna osu ka to
the Emperor who the situation is QT
higara no tatsu no o machikanete
days SB pass NOM DO cannot wait
The emperor could not wait to see him, thinking of them and
therefore
d. isoide meshiyosete
in a hurry have (someone) bring (the baby)

has (the baby) brought to him in a hurry."


Wa after mikado "the Emperor" signals that he is the thematic character of
the text by retrieving him, as well as establishing him as the thematic
character of the paragraph.
However, it may not be the case that any kind of NPs which were once
established as thematic NPs are automatically retrieved by wa. It seems that
highly ranked topics or characters, i.e. characters who have prime impor­
tance in the discourse as a whole, are retrieved with wa. However, lower
ranking topics or characters may not be retrieved by wa, and they may go
through the process of theme creation in a new paragraph in which they
once again become a thematic NP.
In the story of Gon gitsune "Gon, the Fox", we observe that although
the hero Gon is always retrieved with wa, another character, Heijuu, is not.4
(16) a. ...dakara, heijuu wa, okkasan ni unagi 
therefore mother eel DO
tabesaseru koto ga dekinakatta.
eat-CAU could not
Therefore, Heijuu could not let his mother eat eel.
b. sonomama okkasan wa shindeshimatta-ni-chigainai ....
in its condition mother died-must
She must have died without eating it.
 heijuu ga ido no tokor de, mugi o toideimashita.
well of place at wheat DO grind
232 NORIKO FUJII UENO

Heijuu was grinding wheat at the well.


d. heijuu wa, ima made, okkasan to futari-kiri de
now until mother with two only
kurashiteita node, moo....
lived since
Heijuu was living with his mother until now, and therefore....
In the new paragraph starting at (16c), Heijuu is first marked with ga
and in (16d) he is marked with wa. The second wa indicates that Heijuu is
now the thematic character of this new paragraph. Here, Heijuu is not
retrieved with wa. The writer goes through the process of establishing the
thematic NP, reintroducing him with ga, and then marking him with wa.
Further considerations and more extensive data analysis are necessary,
however, to make a stronger argument about the relationship between the
rank of participants and the way they are established and retrieved as
thematic NPs.
Besides the reintroduction of a character into a story, there are other
cases in which NPs that have been marked by wa become marked by ga in
the subject position later in the story. The next section examines these
cases.

2.3. Alternation between NP wa and NP ga


NPs which are established as thematic NPs and thus have been marked
by wa may be marked by ga in a particular part of the story. In example
(17) from Genji Monogatari "The Tale of Genji" by Fumiko Enchi, mikado
"the Emperor" has been established as a thematic character and has been
marked by wa but will now be marked by ga. First, consider the example:
(17) a. hahagimi no miyasudokoro mo motomoto
mother (title) originally
mikado no osoba ni ite mi no mawari no
the Emperor's side at be and body surroundings
go-yoo  tasu uwamiya zukae no nyooboo no yoona
take care court service lady like
mibun de wa nakatta.
status was not
(Genji's) mother, Miyasudokoro (= Kiritsubo) was not a
person with the same status as ladies who stay near the
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 233

Emperor and take care of him every day.


b. rekki to shita kooi to shite hoka kara
proper court lady as others from
taterareru hazu deatta-shi
respect-PASS was expect to-and also
She should have been respected as a true court-lady,
c. jishin mo sore ni fusawashiku kijin-rashii
herself also it to appropriate noble
hin'і  sonaeteirareru no de atta ga
grace DO have-HON it was that but
and she herself is a person with dignity.
d. mikado ga go-chooai no amari murini 
the Emperor HON-love because of by force even
soba ni hikitsukete o-oki-ni-nari-tai bakari-ni
near want to keep (her) just because
However, the Emperor, who loves Kiritsubo so much, wants
to keep her at his side all the time,
e. shikarubeki...
some
(and therefore whenever he enjoys music or whatever occa­
sion he invites her before anyone else...)
f. sorenari o-soba kara o-hanashi-
after that HON-his side from HON-do not let
ni-naranakattari mo suru
(her) go even do
and there are times when he doesn't let her go after he sleeps
in.
g. tsumari wa amari yoru mo hiru 
in short very night also day also
o-soba-sarazu ni o-atsukai-ni-natta no
HON-his side leave-not HON-treat NOM
ga gyakuni hata kara wa karugarushiku
SB contrarily others from lightly
mo mieta no de atta ga....
even looked it was that but....
234 NORIKO FUJII UENO

In short, because (the Emperor) treated her in such a way


that she accompanies him all the time, contrary to her real
status, she looked like a person of lower status ...'
The choice of ga after mikado "the Emperor" in (17d) seems to relate
to the fact that the description of the Emperor's treatment of Kiritsubo is
not the central idea of this paragraph, but rather is background information
which feeds into the main idea of the paragraph, i.e. Kiritsubo's status and
how she has been pictured in the minds of others. Even though the
Emperor is a thematic NP in the Kiritsubo chapter, he is not so in (17).
Example (17) is a case in which theme at the paragraph level overrides
theme at the textual level in controlling the use of wa.
The following example (18) from the short story Gon gitsune also shows
that when a character has the features of +text theme, —paragraph theme,
+sentence theme, it is not marked by wa. Here, Heijuu, who was on the
thematic stage, retreats from it in a new paragraph. The writer shifts his/her
perspective to Gon. Consider the shift from heijuu wa to heijuu ga.
(18) a. ... heijuu wa biku no nakani, funa
fish basket inside silver carp
ya unagi o gomi to isshoni buchikomimashita.
and eel DO trash with threw in
... Heijuu threw the silver carp and eel with the trash into the
fish basket.
b. heijuu wa, sorekara, biku  motte kawa
then fish basket DO carry river
kara agari, biku o dote ni oite, kaw akarni
from go out fish basket DO bank on put up-stream
no hoo e ikimashita.
toward went.
Then, Heijuu got out from the river with the fish basket, and
went toward upstream, leaving the fish basket on the bank
c. heijuu ga inakunaru-to, gon wa, pyoito kusa no
disappear when swiftly grass
naka kara tobidashite, biku no soba ni
inside from jumped out fish basket near
kakeyorimashita.
ran to
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 235

When Heijuu left, Gon swiftly jumped out from the grass
and ran to the fish basket.
d. chotto ita zura ga shitakunatta no-desu.
a little mischief want to do it is that
He felt like being mischievous.
e. gon wa, biku no naka no funa o
fish basket inside of silver carp DO
tsukamidashite wa, kawa no   megakete
grab river inside DO aim
ponpon nagekomimashita.
one after another threw in
Gon grabbed the silver carp in the basket and threw them in
one after another.
f. ichiban shimaini, imagi o tsukami-ni-kakarimashiîa
finally eel DO grab began
Lastly, he started grabbing the eel.
g. imagi wa, ... gon no kubi ni makitsukimashita.
eel neck wound round
The eel wound round Gon's neck.
h. sonotoki, heijuu ga, muko kara, uwaa, nusutîo
then over there from gosh thief
gitsune me!" to donarimashita.
fox shouted
Then, Heijuu shouted, Hey! You thief!
i. gon wa, bikkurishite tobiagarimashita.
with surprise jumped up.
Gon jumped up with surprise.
Heijuu, who was the thematic NP in (18a) and (18b), is no longer
marked by wa in (18Һ). Instead, the new paragraph, which starts at (18c),
has Gon as the thematic character. The story is now told with Gon as the
theme, and Heijuu's behavior in (18Һ) serves as background information
which feeds into the description of Gon's behavior in (18i).
Hamada (1983) also notes that a NP of the thematic referent may be
marked with ga later on in the discourse. He points out that this phenome­
non involves changes such as a change of theme, a description from another
236 NORIKO FUJII UENO

participant's perspective or a shift in the narrator's perspective, a change


from the description of the participant's state to a description of his specific
action, and the reappearance of the thematic participant (1983:56).
My analysis of the short story, Gon gitsune shows that although the
hero, Gon, is always mentioned with wa in the subject position in matrix
sentences in non-initial explicit mentions (except once when there is
exhaustive listing connotation), the second ranking character, Heijuu, fluc­
tuates between wa marking and ga marking. There are twenty-one explicit
non-initial mentions of Heijuu in the subject position of matrix sentences.
Out of these occurrences, six are marked by ga instead of wa. Among these
six cases, three involve theme change; Heijuu participates only as a part of
the event described, but not as the theme of the paragraph. One case
involves the reintroduction of Heijuu. In the other two cases, it seems that
the writer wants to vividly convey a rapid progression of events and an
alternation of actions by different participants without using any of the
characters as the frame. In this case even a participant who is established as
the thematic character of the paragraph and is being marked by wa
elsewhere in the paragraph may receive ga marking.
(19) a. ... heijuu wa hirumeshi o tabekakete,
lunch DO began to eat
chawan  motta-marna, bonyarito
rice bowl DO holding absentmindedly
kangaekondeimashita.
lost in reverie
Heijuu began to eat lunch, but lost in reverie with his rice
bowl in his hand.
b. henna koto ni wa, heijuu no hoppeta ni,
strange thing of cheek on
kasurikuzu ga tsuiteimasu.
scratch is there
Strangely, there is a scratch on Heijuu's cheek.
c. ...gon ga omotteiru-to, heijuu ga
is thinking when
hitorigoto  iimashita.
said to himself
When Gon was thinking ..., Heijuu said to himself.
In (19c) it is possible to mark Heijuu with wa. However, the choice of
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 237

ga conveys the progression of the event and the shift of focus more vividly.
The same kind of ga usage has also been observed in other stories I have
examined. Since ga does not present an element as a frame and has more
temporal relevance, as we saw in examples (5) and (6), it is suited to
describe speedy actions, sudden or unexpected actions or the rapid alterna­
tion of actions by different participants. This pragmatic effect of ga is in
accordance with Hamada's.finding that all predicates used with non-initial
NP ga are nonstative predicates (1983:56).

2.4. Themes on the sentence/clause level


In the previous sections, we have examined how a frame on the sen­
tence level can relate to a frame on the paragraph and discourse levels. The
NPs which were marked by wa in earlier examples (10), (11), (13) have the
features +text theme, +paragraph theme, +sentence theme or of - t e x t
theme, +paragraph theme, +sentence theme. Therefore they are continu­
ous and thus important NPs in terms of Givón's topic continuity counting
(Givón 1983; Hinds 1983).5
However, there are also elements marked by wa which are not con­
tinuous in a discourse. These elements have only local importance and
therefore the was that mark them have nothing to do with what Maynard
calls "staging." Although these usages are often treated as marginal, they
reflect the original function of the particle, as I demonstrate in Part II. Not
only NPs but also adverbial phrases and te-clauses are often marked by wa
to designate local importance.
First, consider the following examples from Genji Monogatari by
Fumiko Enchi:
(20) hyoomen wa kono ue naku miyabiyakani
surface this above there is not graceful
mieru kookyuu no nyonin-tachi no aida ni
look back palace of lady-PL among
tachimajitte miyazukae no hibi o okutteita.
mingle and court service of days DO was spending
Kiritsubo was serving at Court among ladies of the back palace
who, on the surface, look the most graceful.
(21) okotae mo manzokuni wa mooshiagerarenai
reply even satisfactorily HUM-cannot say
(Kiritsubo) cannot reply satisfactorily.
238 NORIKO FUJII UENO

(22) mikado no gohigo o tada hitotsu no


Emperor of HON-patronage DO only one
kokage to shite mi o yosete wa iru
tree shadow as body DO shelter be
monono.. .shikkarishita ushirodate no nai
though powerful support SB there is not
no  marude hikagemono no yooni sagesundari
NOM DO as if obscure person like look down
nanika ochido o sagashite-yaroo to
something weaknesses DO look for-VOL QT
shiroi me de mihatteiru osoroshii teki no
critical eyes with watch-is fearful enemy SB
ooi nakani....
many among
Although she (= Kiritsubo) is sheltering under the only shade of
the tree — the Emperor's patronage, among many ladies who
criticize her as if she is an obscure person simply because she
does not have a strong support or who watch her with critical
eyes trying to look for her weaknesses ....
The entities to which wa is attached in (20) and (21) do not have much
importance beyond the sentence. They do not continue as the theme in the
discourse, and they do not serve as the central idea of the paragraph to
which other information should be fed. Notice that these wa-marked
entities have contrastive meaning (although the degree of contrast varies).
hyoomen wa "on the surface" in the relative clause of (20) contrasts with,
or differentiates from, what is in the minds of the other court ladies, which
is hardly so gracious. Wa after manzokuni adds the nuance of "contrary to
the Emperor's/our expectations," so the sentence means "(contrary to the
Emperor's/our expectations) she cannot even respond satisfactorily (=
clearly, strongly)" or "she is responding but not in as satisfactory a way as
she should." In these examples wa establishes a contrastive framework or a
framework to which special attention should be drawn.
A similar thing can be said with (21). Here wa is inserted in the com­
pound verb, yoseteiru "is sheltering." Because the phrase is marked by wa
it depicts the contrast between the way Kirisubno is and how the other
ladies are waiting for opportunities to criticize her. Also wa draws our
attention to the preceding part, and gives the feeling of assertion, convey­
ing the connotation "it is true that ...."
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 239

Wa may indicate that the action expressed by the preceding element


occurs repeatedly. Compare the a-sentences and the b-sentences in the fol­
lowing examples:
(23) a. atari no mura ni detekite itazura bakari
surrounding village to appear mischief only
shiteimashita.
was doing
(Gon) appeared in neighboring villages and only did mis­
chief.
b. atari no mura ni detekite-wa itazura bakari shiteimashita.
(24) a. gon wa...funa o tsukamidashite kawa
silver carp DO take out river
no   megakete ponpon
inside DO aim and one after another
nagekomimashita.
threw in
Gon took out the silver carp, and threw them into the river
one after another."
b. gon wa ... funa  tsukamidashite-wa kawa no   mega­
kete ponpon nagekomimashita.
The only difference between the a-sentences and b-sentences is that wa
is attached to the te-clause in the b-sentences. However, because of the wa,
the b-sentences convey different meanings from the a-sentences. Although
(23a) simply indicates that the two actions occurred sequentially, (23b)
implies that Gon repeatedly appeared in the villages and did mischief, or it
conveys that every time he came to the village he did mischief. Similarly,
while (24a) simply describe a sequential action: "Having taken out the
silver carp from the fish basket, Gon threw them into the river," wa in
(24b) adds a connotation that Gon repeatedly took out the silver carp in the
fish basket, and threw them into the river. In these examples, wa indicates
that the action occurred repeatedly.
Although it is not as obvious as in earlier examples of wa, the wa
usages in examples (20) through (24) also create a frame for the predication
to some extent. In (20), hyoomen wa sets up a framework within which the
predication, kono ue naku miyabiyakani meiru "look the most graceful,"
should be interpreted. In (21) manzokuni wa "satisfactorily," and in (22),
240 NORIKO FUJII UENO

mikado no go-higo o tada hitotsu no kohage to shite mi o yosete wa "shelter­


ing under the only shade of the tree — the emperor's patronage," deter­
mine the framework for the predication. Hence these was also have the
same function of setting a frame and limiting applicability of the predication
as was the case in earlier examples of wa usage. That wa can have more
than one function at the same time is quite important in understanding the
development of the particle, as we will see in Part II.
As pointed out above, when wa marks elements which are not continu­
ous as topics in the discourse, the motivation for the use of the particle is
not "staging," but establishing a new framework which is different from the
preceding one or the one which follows. This is necessary because wa-
marked elements need to be differentiated from, or contrasted with, other
elements and/or the reader's assumptions and preoccupations, or because
the elements need to be asserted or emphasized. Another kind of emphasis
is the indication of repetitous or habitual actions as in (23) and (24). There­
fore, there are various motivations for the use of wa.

2.5. Summary of part I


It is quite important to be aware of different levels of themes and their
interrelationship with the use of wa. The different levels of themes/frames
are indicated below.
(25) Different levels of framework/theme 6
clause -- --- - - -
sentence ---
paragraph
text --------------------------------------------
Some was indicate thematic NPs in the discourse, and by so doing
create cohesion in the discourse (examples (9), (10), (11), and (13)). There­
fore in these examples NP wa constructions have the characteristic of +text
theme, +paragraph theme, +sentence theme or of - t e x t theme, +parag­
raph theme, +sentence theme. Examples (17) and (18) show that if a
character has the feature of +text theme, —paragraph theme, +sentence
theme, it is not likely to be marked by wa. Whether or not an NP is a
thematic NP of the paragraph seems to be a strong factor in determining the
occurrence of wa. Considering the interplay between different levels of
themes and the use of wa, we can theoretically think of at least the following
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 241

four cases for the occurrence of wa.


(26) 1. +text theme, +paragraph theme, +sentence theme
2. +text theme, —paragraph theme, +sentence theme
3. -text theme, +paragraph theme, +sentence theme
4. —text theme, —paragraph theme, +sentence theme
Contrastive meaning can be added to all the features depending on the
context. If an element has type 1, it is very likely that the element is marked
by wa. My data suggest that if an element has type 2, it is unlikely that the
element is marked by wa unless it needs to be contrasted with another ele­
ment or needs to be asserted. If an element has type 3, it is usually marked
by wa. A type 4 element is generally not marked by wa unless that element
is contrastive or assertive.
For example, in Gon gitsune, among 83 occurrences of wa, 52 are rec­
ognized to be type 1. I could not find any occurrences of wa that could be
classified as type 2. Fourteen cases are recognized as type 3, and 7 cases as
type 3', which is characterized as - t e x t theme, -paragraph theme as a
whole, +paragraph's local theme, +sentence theme. There are 10 cases
which belong to type 4. Among these, 3 cases indicate repetitive action, wa
being attached to re-clause. In 4 instances wa conveys contrast and in the
other 3 cases, wa adds the nuance of assertion and emphasis.
In this story the characters who have the feature of +text theme are
Gon and Heijuu. When the first ranking character Gon is the theme of the
paragraph and the subject of the matrix sentences (i.e. when Gon has the
features of type 1), he is marked with wa 97.3 percent of the time (36 out of
37 cases). In Heijuu's case, when Heijuu has the features of type 1, he is
marked by wa about 83 percent of the time (15 out of 17 or 18 cases
depending on the interpretation). 7 Also, the characters who are recognized
as thematic characters of paragraphs are marked by wa in the subject posi­
tion of matrix sentences with the ratio of 92.1 percent. In relation to type 3,
low ranking characters such as Heijuu's mother or a fish vendor are marked
by wa only when they become a paragraph theme or a paragraph's local
theme. The examination of Genji monogatari by Fumiko Enchi shows very
similar phenomena.
Since wa operates on different levels and creates a framework through
all of its various purposes, it sometimes looks as if the particle has con­
tradictory pragmatic functions; in some cases wa seems to defocus the pre­
ceding element and in other cases it emphasizes it.8 Some was establish and
242 NORIKO FUJII UENO

keep a framework for a paragraph or text, and other was try to shift atten­
tion from the on-going theme to something else. When wďs function is the
former, it is likely that the NP will be interpreted as defocused. In this con­
text, what is presented with wa is well understood old information which is
in the discourse registry, and what follows the wa-marked entity is new
information to which the reader/hearer naturally pays more attention. On
the other hand, when the particle marks elements which are not activated in
the reader's/hearer's consciousness for the sake of contrast or differentia­
tion, or when wa is attached to an element for the sake of assertion to draw
the reader/hearer's attention, the reader/hearer is likely to feel that the ele­
ment marked by wa is emphasized. This seems to be natural when we con­
sider that in the latter case, the element is presented with a special contras­
tive intonation in speech.

PART II: HISTORICAL DEVELOPMENT OF THE PARTICLE

The complexity of and multifunctionality of wa become much more


understandable when we look at its historical development. In this section,
I will point out the similarities and differences between wa in present-day
Japanese and its counterpart in Old Japanese (I am transcribing it as ha in
this paper. Its pronunciation was probably a bilabial fricative [De Wolf (this
volume)]. Based on the similarities and differences, I will suggest a process
for the development of the particle.
Functions of ha are often described in a very similar way to that in
which wa in present-day Japanese has been explained.
(27) Ha presents what precedes it as the theme or topic of a sentence and it is
used only in matrix sentences (Konojima 1965:288-293. The translation is
mine).
(28) Ha is attached to various elements in a sentence and conveys the notion of
differentiation and emphasis (Nakata 1971:278-279. The translation is
mine).
The explanation given in (27) is exactly the same as the popular defin­
ition of so-called thematic wa and (28) is similar to the explanation of what
Kuno calls "contrastive wa" (1973). Based on these explanations, it is dif­
ficult to tell how ha is different from wa. However, an examination of the
particle from a discourse perspective clearly tells the difference.
In the following, let us consider the functions of ha in Taketori
Monogatari "The Woodcutter's Tale" (tenth century) and Genji
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 243

Monogatari "The Tale of Genji" (eleventh century). A comparison of sev­


eral versions of the translation (to a colloquial form of the language) of
"The Tale of Genji" also brings out differences between wa in present-day
Japanese and its counterpart in Old Japanese.

1. Examination of Taketori Monogatari "The Woodcutters Tale"


Taketori Monogatari examined here is from Iwanami koten taikei.10 In
this text, there are forty occurrences of ha in the first two hundred clauses
(7 pages). Ha is attached to NPs (33 cases out of 40), the quotative particle
to, and te-clauses. Interestingly, there is not one instance of ha marking
okina "old man" or kaguya hime "Princess Kaguya," who are the most
important characters of this story and who are most often referred to in the
examined part of the story. Instead, ha creates a pragmatic effect which is
only locally important. The majority of NP ha's (about 97 percent) convey
some degree of contrast/differentiation. Therefore ha's function seems to
be to establish a contrastive or differentiating framework. It is also a device
to draw attention to elements that would otherwise be dealt with casually.
Examples (29) and (30) illustrate typical usages of ha in the text.
(29) kono  no hito ha to ha onna ni afu koto
this world of person man woman meet NOM
onna ha otoko ni af koto wo su.
woman man meet NOM DO do
Speaking of people in this world, men meet women, and women
meet men (= men get wives and women get husbands).
(30) a. hi kururu hodo rei no atsumarinu.
sun set when those people SB gathered
In the evening those people gathered.
b. aru hi ha fue wo fuki
one person flute DO play
One of them was playing a tune on a flute, and
c. aru hi ha uta wo utahi
one person song DO sing
another was singing a song, and
d. aru hi ha sooga wo shi
one person sing a melody
another was singing a melody, and
244 NORIKO FUJII UENO

e. aru hi ha uso wo fuki


one person whistle
another was whistling and
f. oogi wo narashi nado suru ni
fan DO take a rhythm and so on do
making a rhythm with a fan and so on ...
Example (29) was said to kaguya-hime "Princess Kaguya" by okina
"old man," who had just found out that she was a person from the moon.
Since kono yo no hito "people in this world" is differentiated from the
people in the moon where the princess is from, the NP is marked by ha.
Otoko "man" in the koto clause is contrasted with onna "woman" and vice
versa.
In example (30), ha differentiates one man from the others and brings
out the contrast between them. At the same time, ha helps to shift focus
from one man to another man.
Ha may emphasize the preceding element; in this case the preceding
element is a quotation.
(31) omo naki koto wo ba haji wo sutsu to ha
shameless thing DO EMP shame DO throw away QT
ihikeru.
say
Shameless things are described as 'throw away shame'.
It may also convey habitual or repetitious action, being attached to the
re-clause, as in the earlier example of wa usage in (23) and (24).
(32) yo no naka ni ookaru hito wo dani sukoshi mo
world of inside many people DO even a little
katachi yoshi to kikite ha
appearance good QT hear
mimahoshiusuru hito-domo narikereba
want to see people-PL be-since
Since they are the people who want to see a woman if they hear
that she is beautiful even if she is the kind of beauty you can find
easily in the world.
Unlike wa in present-day Japanese, ha in this text does not func­
tion to create a theme for the paragraphs and the text and to maintain it.
Observe the introduction and reintroduction of okina "old man."
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 245

(33) a. ima ha mukashi taketori no okina


now long time ago (= name)
to yufu mono arikeri.
QT say person there was
Long time ago, there was a person called Taketori no okina.
b. noyama ni majirite take wo toritsutsu
mountains to go in bamboo DO take-while
yorozu no koto ni tsukahikeri.
various things for use-PST
He went into fields and mountains to take bamboo trees, and
he used them for various things.
 na wo ba sanuki no miyakkomaro to namu ihikeru.
name DO EMP QT EMP say-PST

His name was Sanuki no miyakkomaro.

d. taketori no okina take wo toru ni


bamboo DO take and then
When Taketori no okina got bamboos,
e. kono  wo mitsukete nochi ni take wo
this girl DO find after bamboo DO
toru-ni
take and then
when he got bamboos after he found the girl,
f. fushi wo hedatete yogoto ni
joint DO separate every at
kogane aru take wo mitsukuru koto
gold there is bamboo DO find that
kasanarinu.
happen several times
he continued to find gold between all the joints of the bam­
boos
g. kakute okina yauyau yutakani nariyuku
this way gradually rich become

This way he gradually became rich


246 NORIKO FUJII UENO

h.  ko  katachi  kesoo naru koto


this child of appearance SBJ beautiful that
 і 
world in there is not
This child's appearance is exceptionally beautiful and
i ya no uchi ha kuraki tokoro 
house of inside dark place there is not
inside of the house, there is no place where is dark, and
j. hikari michitari
light full of-PERF
it is filled with light
k. okina kokochi ashiku kurushiki toki mo kono ko
feeling bad painful time this child
wo mireba
DO see-if
even at the time when Okina is not feeling well, once he
looks at the girl,
1. kurushiki koto mo yaminu.
painful thing even stopped
his pain goes away.
Example (33) shows the first few paragraphs of the story. Taketori no
okina is introduced with the form NP 0 in (33a). After a few sentences, this
old man becomes a familiar character to the reader and the story goes on to
tell what happens to him. Hence, it is reasonable to think that he is the
most important character and closely related to the theme of the para­
graphs. Example (33d) indicates that the establishment of okina as a
thematic NP is done with the form NP 0 , and (33g) and (33k) show that the
maintenance of this thematic NP is also accomplished by the form NP 0 .
Ha does not appear to establish a thematic NP nor to signal the mainte­
nance of it. This is in contrast to the use of wa in a translation (to present-
day Japanese) by Asano and Nakano (1957). In their translation, from the
second mention of okina, he is marked with wa. Compare the following
translation with (33).
(34) a. ima de wa moo mukashi no koto da
now already long time ago of incident
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 247

ga taketori no okina to yuu mono ga ita.


but QT say person SB there was
b. no ya yama ni haitte wa take 
field or mountain to enter bamboo DO
toritorishite iroirona doogu 
take and various instruments DO
tsukuru no ni tsukatteita.
make NOM for was using
c. sono na o sanuki no miyakkomaro to itta.
its name DO QT said

d. taketori no okina wa take  toru-to


bamboo DO take and then
e. K   mitsukete-kara nochi wa
this child DO find after
take  toru-to
bamboo DO take and then
f. fushi to fushi no aida goto ni subete
joint and joint of between every at all
oogon no haitteiru take o mitsukeru-
gold SB is found bamboo DO find
koto ga nando to naku tsuzuita.
NOM SB many times continued
g. kooshite okina wa shidaini yuufukuni
this way gradually rich
natte-yuku
is becoming

h.    yooboo ga ...


this child of appearance SB
i. ie no naka wa dokonimo kurai tokoro ga
house of inside anywhere dark place SB
nai hodo ni
there is not to the extent
248 NORIKO FUJII UENO

j. hikari ga michite iru


light SB full
k. taketori no okina wa kibun ga waruku
feeling SB bad
kurushii toki mo kono   miru-to
painful time even this child DO see when
1. sono kurushisa mo nakunatta.
its pain even disappeared.
In (34) the establishment of a thematic NP, taketori no okina, and the
maintenance of the thematic NP are carried out by wa. Examples (33) and
(34) clearly illustrate the difference between wa and ha. Namely, ha is not
used to maintain a thematic NP.
It is interesting to see that even a character who is first introduced with
ha does not receive ha marking later in the story when there is no contras­
tive meaning. Example (35) describes the men who showed persistent
interest in Princess Kaguya. Their first introduction is done with ha because
they need to be differentiated from other men who showed interest in the
princess. But after they are established as familiar characters, they are no
longer presented with ha when there is no need to signal contrast with or
differentiation from other characters.
(35) a. atari wo hanarenu kindachi 
vicinity DO do not leave noble men night
wo akashi hi wo kurasu ookari
DO spend day DO spend many
The noble men who didn't leave near the house spent many
days and nights.
b. orokanaru hito ha yaunaki ariki ha
silly people not-useful walking
yoshinakarikeri to te kozunarinikeri
not good-PST QT and not to come-PST
Silly people (= people who were not serious) left saying that
the unrewarding walking was not good.
c. sono nakani naho ihikeru ha irogonomi
its among still said understand love
to iharuru kagiri gonin omohi yamu toki
QT is said only five people love stop time
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 249

naku yoru hiru kikeri


there is not night day came
Among the noble men (who were interested in Princess
Kaguya) people who kept proposing her were five men who
were said to understand love. Their love did not know the
end, and they came day and night.
d. sono na domo "ishizukurino miko," "kuramochi
its name-PL (= name) (= name)
no miko, " "sadaijin abe no mimuraji" "dainagon
(= name) (= name)
ootomo no miyuki, " "chuunagon isono і 
(= name)
marotari," kono hitobito narikeri.
this people be-PST

Their names were Ishizukuri no miko, Kuramochi no miko,


Sataijin abe no mimuraji, Dainagon ootomo no miyuki and
Chuunagon isono kami no marotari. They were these people.

e. kono hitobito arutoki ha taketori wo yobiidete


this people one time DO call and
These people sometimes called Taketori and
f. "musume wo ware ni tabe" to fushi-ogami
daughter DO I to give QT beg

begged saying, "Give me your daughter."...


g. kakareba kono hitobito uchi ni kaherite
therefore this people house to go back
Therefore these men went home and
h. mono wo omohi
wonder
wondered and
i. inori wo shi
pray DO do
prayed and
250 NORIKO FUJII UENO

j. gan wo tatsu
make a petition to God
made a petition to God.
In (35c) sono  ni naho ihikeru "people who still keep proposing" is
marked by ha because of the contrast with orokanaru hito "silly people" in
(35b). After (35c) the story concentrates on the behavior of those men who
were persistent. Notice that in (35e) and (35g) the maintenance of the
thematic NP kono hitobito "these people" does not take ha. Example (35)
again illustrates the lack of a theme-maintaining function of ha.
Ha in this example conveys contrast and/or differentiation. Besides
contrast between "silly men" and "men who were persistent," we see the
use of ha after yaunaki ariki "unrewarding walking" in (35b). Ha here dif­
ferentiates from and contrasts with "walking which brings reward." Ha
after aru toki "one time" in (35e) differentiates from other times, as well as
indicates a time framework for the predication.

2. The Study of Genji Monogatari "The Tale of Genji"


Similar to Taketori monogatari, we do not observe the ha of theme cre­
ation and theme maintenance in the examined part of Genji monogatari.
(The examined text is from Iwanami koten taikei.) First, let us look at how
Kiritsubo is introduced into the story and reintroduced in later paragraphs.
(36) a. izure no ohon-toki ni ka nyoogo
when HON-period at QU court lady
kooi amata saburahi-
(lower rank) court lady many be (HON)
tamahikeru nakani ito yamugotonaki kiha ni ha
HON PST among very noble status at
aranu ga sugurete tokimekitamafu arikeri
NEG SBJ specially prosper-HON there was
In a certain reign, there was (an incident that) a lady of not
the first rank was loved by the Emperor more than anyone
else.
(37) a. hahagimi hajime ori oshinabete no
Mother from the beginning all
uwamiyazukaeshitamafu beki kiha ni
do service at upper court-HON should status
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 251

ha arazariki.
was not
The mother (= Kiritsubo) is not of the status a person who
should stay near the Emperor all the time and serve him.
b. obohe ito yamugotonaku ...
reputation very good
Her reputation is very good ...
(38) a. sono toshi no natsu miyasudokoro hakanaki
that year of summer vague
kokochi ni wazurahite
conditions suffer and
and thinks of leaving Court,
That summer, the boy's mother feels vaguely unwell,
b. makadenamu to shitamafu wo
leave-VOL Q do-HON however
and thinks of leaving Court,
c. itoma sarani yurusasetamahazu
leave atall allow-HON-HON-NEG.
but (the Emperor) does not allow it at all.
In (36) Kiritsubo is introduced for the first time, and the introduction
is done with NP ( 0 noun NP11) ga, i.e. ito yamugotonaki kiha ni ha aranu
ga "a person of not the first rank." after the introduction, the story pro­
ceeds, telling how the Emperor's infatuation with Kiritsubo disturbed the
others, and how their jealousy and criticism hurt Kiritsubo. Then the
author tells of the birth of a beautiful baby boy which symbolizes the deep
relationship between the Emperor and Kiritsubo. In (37) the theme has
changed to Kiritsubo's status and the Emperor's treatment of her. Notice
that despite the fact that Kiritsubo has been established as a familiar
character and is a thematic NP in the paragraph in which (37) is contained,
the reintroduction of her in (37) is not accompanied by ha. About forty
clauses later, Kiritsubo is reintroduced explicitly. The new paragraph (38)
describes Kiritsubo's falling ill and the Emperor's reaction to it. Here again
Kiritsubo's reintroduction is done with NP 0 . This is in contrast to the fact
that the reintroduction of Kiritsubo in the corresponding part in twentieth-
century versions (Yosano 1936; Tanizaki 1951; Enchi 1972; and Imaizumi
1978) is accompanied by the particle wa or mo (mo indicates an additional
252 NORIKO FUJII UENO

topic, conveying the meaning "also"). The examples from Taketori


monogatari and Genji monogatari cited above suggest that in Old Japanese
cohesion is established in a different way from present-day Japanese.
The various contexts in which ha is used suggest that ha functions to
shift attention from one theme/topic to another. It brings out the notion of
differentiation and/or contrast. Observe some examples of ha.
(39) a. hajime yori ware ha to omohiagaritamaheru
beginning from I QT be conceited-HON
ohon-katagata mezamashiki mono ni
HON-people-(HON) presumptuous person as
otoshime sonemitamafu.
speak ill of feel jealous-HON
Ladies who are conceited thinking, I (should be the one who
will be loved by the Emperor)! from the beginning of their
service think of (Kiritsubo) as a presumptuous person.
b. onaji hodo sore yori geroo no kooi-tachi ha
same degree it than lower lady-PL
mashite yasukarazu.
even more resentful
Other ladies with the same status (as Kiritsubo's) and lesser
ladies are even more resentful.
Ware ha 'T!" in (39b) implies differentiation from the other court
ladies, since what each lady thinks is "I should be the one (and not the
others) who is loved by the Emperor." Thus, ha carries the meaning of dif­
ferentiation from the others and emphasizes the preceding noun ware 'T."
In (39b) ha is attached to onaji hodo sore yori geroo no kooi tachi "ladies
with the same status and lesser ladies." By diverting or shifting attention
from the previous sentence theme, "ladies who are conceited" to the theme
of the succeeding sentence, ha differentiates and points out the contrast
between these themes.
Consider another example.
(40) a. otoshime kizu wo motometamafu
speak ill of weakness DO look for-HON
hito ha ohoku
people many
people who speak ill of (her) and look for (her) weaknesses
are many,
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 253

b. wa ga mi ha kayowaku
I GN body fragile
and self (= she herself) is fragile.
In (40) also, ha shifts attention from people who speak ill of her (= Kirit-
subo) to Kiritsubo herself, and contrasts these two people.
As the examples above have indicated, ha in Old Japanese is a local
phrase marker. Quantitative analysis also supports this observation. I have
examined the use of ha in Geriji monogatari and of wa in several rewritings
of the story in a quantitative study with respect to the following points: 1)
frequency of occurrence, 2) topic continuity, 3) the use of the particles in
marking newly introduced characters, and 4) functions of theme-creation
and theme-maintenance. The versions compared here are Genji
monogatari, eleventh century (Genji 1); a rewriting of it by Hanshichi Taga
(1723) (Genji 2); Nise Murasaki Inaka Genji by Ryuutei Tanehiko (1830)12
(Genji 3); a rewriting by Akiko Yosano (1914) (Genji 4), another rewriting
by Akiko Yosano (1936) (Genji 5); a rewriting by Junichiroo Tanizaki
(1951) (Genji 6); a rewriting by Fumiko Enchi (1972) (Genji 7); and that by
Tadayoshi Imaizumi (1978) (Genji 8). The statistics summarize an examina­
tion of the first four hundred clauses of the Kiritsubo chapter of each Genji
(this amounts to somewhere between 11 and 36 pages depending on the size
of the characters of a text).
First, the frequency counting factor shows that ha is less frequently
used than wa in the twentieth-century Genjis. Table 1 shows the occurrence
of ha and wa as observed in the Genji texts. Although we notice variation
among the versions, ha's occurrence in Genji 1 is considerably lower than
that of wa in the other versions. This relates to the lack of hďs theme-main­
taining function.

Genji 1 2 3 4 5 6 7 8
Number of 65 109 95 149 168 96 128 151
Occurrences

Table 1. The occurrence of ha/wa in Genji 1-8

Second, I have applied continuity counting as developed by Givón


(1983) in order to examine the continuity of elements that are marked by
ha/wa in the discourse. Distance (= referential distance) measures the
degree of continuity of the topic NP in terms of its previous occurrence in
254 NORIKO FUJII UENO

the discourse and its current occurrence in a clause. The gap is expressed by
the number of clauses between the two occurrences. The minimum value
that can be assigned is one clause. Although Givón sets the maximum value
at twenty, I set it at thirty because in Genji 1 very distant NPs can be refer­
red to with 0 anaphora. Persistence measures how continuously an NP is
referred to in subsequent clauses. Therefore, a highly topical NP in the dis­
course has a low value for referential distance and a high value for persis­
tence, and a non-topical NP is characterized by the opposite relation.
Table 2 shows that subject NP ha in Genji 1 has a larger number in
terms of distance and a smaller number in terms of persistence than subject
NP was in other Genjis. This suggests that NPs that are marked by ha in
Genji 1 are in general not as continuous as NPs that are marked by wa in
other versions. This means that NPs which are marked by ha have a more
peripheral role in the discourse as a whole, having temporal importance,
than NPs that are marked by wa. It is interesting to see the big gap between

(D = distance, P = persistence)

Genji NP ha/wa NP  NP o N P ga ø subject


1 D 25.94 24.42 28.22 5.21
P 0.55 0.26 0.41 _ 1.11
2 D 17.06 24.89 24.17 30.00 3.26
P 1.76 0.53 0.78 1.68
3 D 20.2 16.75 23.06 23.29 3.71
P 0.64 0.35 0.33 0.37 0.95
5 D 15.22 26.47 25.55 5.59
P 0.97 0.2 _ 0.4 1.01
6 D 20.76 20.84 25.43 4.57
P 0.82 0.53 _ 0.19 1.2
7 D 19.12 20.7 - 25.3 5.58
P 0.84 0.47 0.1 0.97
8 D 17.55 22.8 - 25.85 4.38
P 1.85 0.8 0.12 1.57

Table 2. Topic continuity of subject NPs which are


marked by ha/wa, mo, ga and ø
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 255

Genji 1 and Genji 2. The use of wa in Genji 2 seems to be quite similar to


that of wa in the twentieth-century versions, and I cannot pin down the dif­
ferences between them.
Third, the examination of the extent to which ha or wa is attached to
characters or elements previously introduced indicates that ha is often
attached to newly introduced characters. In this study, I am including
generic NPs and characters whose identities are easily predictable into the
category of familiar characters/old information. characters or entities which
are not transparently predictable or those which the author may be assum­
ing to be predictable are not being included in the category of familiar
characters. The observation that can be taken from Table 3 is that ha is
often used to introduce new characters bringing the notion of contrast.

Genji NO. New (%) Old (%) ?

1 65 61.2 34.3 4.5


5 168 34.2 65.8 -
7 128 43.5 57.5 -

Table 3. New and old information with respect to


elements which are marked by ha/wa.

Genji NO. Theme Creation (%) Theme


Maintenance (%)
1* 2* 3*
1 65 34.3 46.8 14.0 4.6
2 109 32.1 18.3 16.5 33.0
5 168 21.8 27.8 10.0 40.2
7 128 16.4 31.2 11.7 40.6

1* = theme creation for elements whose referential territory does not


go beyond the clause.
*2 = theme creation for elements whose referential territory goes
beyond the clause.
*3 = setting a temporal or spatial framework.
Table 4. Funtion of ha/wa.
256 NORIKO FUJII UENO

Table 4 clearly shows the lack of the theme-maintaining function of ha.


In Genjis 2, 5, and 7 theme maintenance is one of the important functions
of wa. On the other hand, cases in which ha is used to keep the same
character in focus is rare in Genji 1. Ha is often attached to elements which
have temporal or local importance but do not continue as topics in the dis­
course.
These quantitative analyses clearly show the difference between wa
and ha. Ha does not create a frame on the level of paragraph or text.
Rather it locally differentiates the preceding element from another ele­
ment, or contrasts it with another element or emphasizes it.
In Genji 1, however, we see an indication that ha is beginning to func­
tion in marking a paragraph theme. It seems that the function of ha in (41)
is to signal a shift of theme from the previous one (Kiritsubo's hardship) to
the current one (Kiritsubo's residence).
(41) a. ... nakanaka-naru mono-omoi wo zo shitamafu
very difficult worry DO EMP do-HON
... (Kiritsubo) suffers so much.
b. mitsubone ha kiritsubo nari
HON-residence is
Her residence is Kiritsubo.
 amatano ohon-katagata wo sugis asetamahi...
many HON-people(HON) DO pass-HON-HON
Passing by many other ladies ...
Unlike many instances of ha in the data, in this example there is no ele­
ment overtly contrasting with mitsubone "residence" in the context. What is
differentiated here is a topic of the new episode. Since there is no para­
graph marker in Genji 1, ha''s function here is quite important in signaling
an episode change. In cases like (41), even if the primary function of the
particle is to signal theme change, it is easily deduced that hďs primary
function is to establish a theme for the paragraph.
From the observations above, I have concluded that although the orig­
inal function of ha was to present an element as a new framework which
was different from or contrasted with the previously introduced one or the
succeeding one, over the course of several hundred years hďs framework-
creating function came to be understood more broadly, and the particle in
fact came to be used in contexts where particular differentiation or contrast
WA F R O M S Y N C H R O N I C A N D D I A C H R O N I C P E R S P E C T I V E S 257

was not necessary. At the same time, the particle increased its domain from
sentence level to discourse level, and came to function as a device to estab­
lish cohesion in a dicourse. Wa in present-day Japanese has the original
function as well as the extended function. This is illustrated in the follow­
ing:

Conclusion
This paper shows how a diachronic investigation gives deeper insight
into the understanding of a linguistic element in its synchronic stage. The
functions of wa in present-day Japanese are quite complicated, as wa oper­
ates on different levels in a discourse and has more than one function. The
complexity becomes more apparent when we consider the long history of
the development of the particle. This paper opens up an interesting ques­
tion concerning the relationship between contrastive markers or markers of
emphasis and topic/theme markers in languages.

NOTES
1) "Theme" here, of course, does not mean something which is marked by wa. I am using
the term more generally in the sense of what Jones calls "central idea" (1977) and what we are
talking about.
2) It is surprising that so many studies have used this distinction as the basic pragmatic dis­
tinction between these two particles, despite the fact that it soon becomes apparent that it is not
the case when examining a discourse (e.g. Chafe 1976; Inoue 1983; Haig 1983).
258 NORIKO FUJII U E N O

3) Although both Grimes (1975) and Maynard (1980) use a staging metaphor, they do not
mean the same thing by "staging." Compare the following explanations by Grimes (1) and by
Maynard (2) for the word staging.
(1) Clearly the marking of thematization is related to a semantic factor of promi­
nence. It is as though stage directions were given to the spotlight handler in a the­
ater to single out a particular individual or an actor, or as though one actor was
placed close to the audience and another off to the side. In fact, staging
metaphors apear to be highly appropriate for the marked varieties of a whole
range of linguistic phenomena that have a long history of being hard to handle
(1975:327).
(2) ... "staging" is used differently from Grimes. "Staging" here does not necessarily
mean "singling out" as Grimes states, but it means the phenomenon of who con­
stantly remains on the stage as an important character (1980:106).
Referring to one of her examples, Maynard states that the character who is marked by wa con-
tantly remains on the stage although the spotlight may at times be weaker and at other times
stronger. In contrast, the characters who are not marked by wa appear on the stage only at those
moments when their actions become relevant to the response of the character who is marked by
wa (1980:106).
4) Gon gitsune is from an elementary textbook of Japanese, Shoogaku kokugo 4 (fourth
year) (1965).
5) Givón's topic continuity counting contains the following three discourse measurements:
1) referential distance, 2) potential interference (ambiguity), and 3) persistence. It assesses the
importance of characters in terms of how continuous they are in a discourse.
6) Although (25) shows frameworks/themes only on levels which can be formally recog­
nized, there are also themes in units which may not be formally recognized, such as episodes.
Episode boundaries interact with paragraph boundaries in a complex way. Because of this com­
plexity, I did not include them on a level of episode in the schematicization although I am aware
of the importance of it.
7) In determining who are the thematic characters of the text and the rank of the characters
in Gon gitsune, my ten informants showed agreement. In determining thematic characters of
each paragraph, another native speaker independently identified thematic characters for me.
"Depending on the interpretation" reflects the difference in our interpretation.
8) For arguments relating to this in Japanese, see Martin (1975:55).
9) Concerning the pragmatic effect of ha, Nakata claims that ha emphasizes the element
which precedes it (1971:278) and Konojima states that when ha is attached to an element of a
sentence, the focus is given to what follows but not on the element to which ha is attached
(1965:293).
The same kind of arguments are also observed with respect to the Korean theme marker
(n)in (Hook 1984).
10) One of the problems in a diachronic study is that original versions are often not available.
The text of Taketori monogatari in Iwanami koten taikei is based on mutoo-bon from the six­
teenth century, which is the oldest transcription available. To what extent the language was
changed/unchanged in the course of transcribing is unknown. There is the same problem con­
cerning Genji monogatari. The text in Iwanami koten taikei is based on a transcription from the
fourteenth century.
WA F R O M S Y N C H R O N I C A N D D I A C H R O N I C P E R S P E C T I V E S 259

11) 0 noun is a NP in which the head noun is implicit.


12) Strictly speaking, this is not a rewriting of the story. The author, Tanehiko Ryuutei,
created this work as a popular and commercial book based on The Tale of Genji; it was based on
the same participants and events of the story with alternations in the participants' names and set­
tings. This work was examined along with other rewritings since there is no translation of the
tale produced around the same time.

REFERENCES

Alfonso, Anthony. 1966. Japanese language patterns. Tokyo: Sophia Uni­


versity L.L. Center of Applied Linguistics.
Aoyama, Takashi. 1982. The system of focus in Japanese. Columbia Uni­
versity Working Papers in Linguistics. Vol. 7, 91-106.
Asao, Yoshinosuke and Hiroo, Nakano. 1957. Bunpoo shoosetsu: Taketori
monogatari. Tokyo: Nichieisha.
Becker, Alton Lewis. 1967. A generative description of the English subject
tagmemes. Ph.D. Dissertation: The University of Michigan.
Bernard, Robert. 1980. "Subjecthood and consciousness." The pear stories,
ed. by Wallace L. Chafe, 275-299. Norwood: Ablex Publishing Corpora­
tion.
Chafe, Wallace. 1976. "Givenness, contrastiveness, definiteness, subjects,
topics and point of view." Subject and topic, ed. by Charles Li, 27-55.
New York: Academic Press.
. 1980. "The development of consciousness in the production of a narra­
tive." The pear stories, ed. by Wallace L. Chafe, 9-50. Norwood: Ablex
Publishing Co.
Clancy, Patricia M. 1980. "Referential choice in English and Japanese nar­
rative discourse." The pear stories, ed. by Wallace L. Chafe, 127-202.
Norwood: Ablex Publishing Co.
Downing, Pamela. 1980. "Factors influencing lexical choice in narratives."
The pear stories, ed. by Wallace L. Chafe, 89-126. Norwood: Ablex Pub­
lishing Co.
. 1983. The anaphoric use of classifiers in Japanese. Ms.
DuBois, John W. 1980. "Beyond definiteness: the trace of identity in dis­
course." The Pear Stories, ed. by Wallace L. Chafe, 203-274. Norwood:
Ablex Publishing Co.
Enchi, Fumiko. 1972. Genji monogatari. Tokyo: Shinchoosha.
Fillmore, Charles J. 1968. "The case for case." Universals in linguistic
260 NORIKO FUJII UENO

theory, ed. by Emmon Bach and Robert T. Harms, 1-88 New York:
Holt, Rinehart and Winston.
Fillmore, Charles J. "The case for case reopened." Syntax and semantics,
vol. 8, ed. by Peter Cole and Jerrold Sadock, 59-82. New York:
Academic Press.
Firbas, Jan. 1966a. "On defining the theme in functional sentence
analysis." Travaux linguistique de Prague 1, 267-280.
. 1966b. "Non-thematic subjects in contemporary English." Travaux
linguistique de Prague 2, 239-256.
. 1974. "Some aspects of the Czechoslovak approach to problems of
functional sentence perspective." Papers on functional sentence perspec­
tive, ed. by Frantisek Danes, 11-37. Prague: Academic Publishing House
of the Czechoslovak Academy of Sciences.
Fujii, Noriko. 1985. A diachronic study of grammatical subject in Japanese.
The University of Michigan dissertation.
Givón, T. 1979. On understanding grammar. New York: Academic Press.
. 1979. (ed.) Syntax and semantics: Vol. 12, Discourse and syntax. New
York: Academic Press.
. 1983. "Topic continuity in discourse." Typological studies in language,
vol. 3: Topic continuity in discourse: a quantitative cross-language study,
ed. by T. Givón, 1-41. Amsterdam: John Benjamins.
Grimes, Joseph E. 1975. The thread of discourse. The Hague: Mouton.
Haiman, John. 1978. "Conditionals are topics." Language, Vol. 54, No. 3,
546-589.
Halliday, M.A.K. 1967. "Notes on transitivity and theme in English, part
II." Journal of linguistics 3, 199-244.
Hamada, Morio. 1983. Referential choices in theme, subject, and ellipsis in
written narrative discourse: a case study of Japanese folktales. Cornell
University, M.A. Thesis.
Hinds, John. 1973. "Missing subjects." Papers in Japanese linguistics 2:147-
55.
. 1983. "Topic continuity in Japanese." Typological studies in language,
vol. 3. Topic continuities in discourse: quantitative cross-language studies,
ed. by T. Givón, 45-93. Amsterdam: John Benjamins.
. 1984. "Thematization, assumed familiarity, staging, and syntactic
binding in Japanese expository prose." Paper presented at AAS meet­
ing.
Hockett, Charles F. 1958. Λ Course in modern linguistics. New York: The
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 261

MacMillan Company.
Hook, Peter Edwin. 1984. "Some further observations on Kashmiri word
order." Aspects of Kashmiri linguistics, ed. by O.N. Koul and P.E.
Hook, 145-153, New Delhi: Bahri.
. 1985. A contrastive study of topic in Korean, Chejuan, Japanese and
Okinawan. Ms. The University of Michigan.
Hopper, Paul J. 1979. "Aspect and foregrounding in discourse." Syntax and
semantics, Vol. 12, ed. by T. Givón, 213-241. New York: Academic
Press.
Inoue, Kyoko. 1983. The constitution of Japanese and its English transla­
tion — language of negotiation and negotiation in language. Ms.
Ishigaki, Kenji. 1955. Joshi no rekishi-teki kenkyuu. Tokyo: Iwanami Sno­
ten.
Iwasaki, Shoichi. 1984. "Functions of NP-wa in Japanese expository dis­
course." Paper presented at the annual meeting of AAS.
Jones, Linda Kay. 1977. Theme in English expository discourse. Illinois:
Jupiter Press.
Keenan, Edward L. 1976. "Toward a universal definition of subject." Sub­
ject and topic, ed. by Charles Li, 303-334 New York: Academic Press.
Keenan, Elinor Ochs and Schiefferlin, B. 1976. "Topic as a discourse
notion: a study of topic in conversations of children and adults." Subject
and topic, ed. by Charles Li, 335-384. New York: Academic Press.
Kitagawa, Chisato. 1982. "Topic constructions in Japanese." Lingua
57:175-214.
Kobayashi, Junko. 1984. "A contrastive study between the Japanese, wa
ga, and the Burmese, ha ka, particles." Nihongo kyooiku 54:89-98.
Konojima, Masatoshi. 1966. Kokugo joshi no kenkyuu. Tokyo: Oofuusha.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge:
MIT Press.
. 1976. "Subject, theme and the speaker's empathy: a re-examination of
relativization phenomena." Subject and topic, ed. by Charles Li, 417-
444, New York: Academic Press.
Kuroda, S-Y. 1976. "Subject." Syntax and Semantics vol. 5: Japanese
generative grammar, ed. by Masayoshi Shibatani, 1-16, New York:
Academic Press.
Lehman, Winfred P. 1976. "From topic to subject in Indo-European." Sub­
ject and topic, ed. by Charles N. Li, 445-456, New York: Academic
Press.
262 NORIKO FUJII UENO

Li, Charles N. and Sandra A. Thompson. 1976. "Subject and topic: a new
typology of language." Subject and topic, ed. by Charles N. Li, 459-490,
New York: Academic Press.
Longacre, Robert E. 1976. An anatomy of speech notions. Lisse: Peter de
Ridder Press.
Manaster-Ramer, Alexis. 1983. What's a topic? Ms. The University of
Michigan.
Martin, Samuel E. 1975. A reference grammar of Japanese. New Haven:
Yale University Press.
Maynard, Senko Kumiya. 1980. Discourse functions of the Japanese theme
marker wa. Ph.D. Dissertation: Northwestern University.
1984. "Thematization as a staging device in Japanese narrative."
Paper presented at the annual meeting of AAS.
Mikami, Akira. 1960. Zoo wa hana ga nagai. Tokyo: Kuroshio Shuppan.
. 1975 a. Nihongo no rі - wa to ga. Tokyo: Kuroshio Shuppan.
. 1975b. Mikami Akira ronbun-shuu. Tokyo: Kuroshio Shuppan.
Murasaki-shikibu. "Genji Monogatari." Nihon koten bungaku taikei Series.
Tokyo: Iwanami shoten.
Nakata, Norio. 1971. Kookyuu koten bunpoo. Tokyo: Shintoo-sha.
Prince, Ellen F. 1979. "On the given/new distinction." CLS, 15, 267-278.
Schachter, Paul. 1976. "Subject in Philippinne languages: topic, actor,
actor-topic or none of the above?" Subject and topic, ed. by Charles Li,
491-518. New York: Academic Press.
. 1977. "Reference-related and role-related properties of subject." Syn­
tax and semantics, Vol. 8, Grammatical relations, ed. by Peter Cole and
Jerrold Sadock, 279-305, New York: Academic Press.
Shibamoto, Janet S. 1983. "Subject ellipsis and topic in Japanese." Studies
in Japanese language use, ed. by Chisato Kitagawa and Shigeru
Miyagawa, 233-265, Edmonton: Linguistic Research, Inc.
Strawson, P.E. 1974. Subject and predicate in logic and grammar. London:
Methuen.
Tamagami, Takuya. 1964. Genji monogatari hyooshaku, Vol. 1. Tokyo:
Kadokawa shoten.
Tanizaki, Junichiro. 1959. Genji monogatari. Tokyo: Chuuoo Kooronsha.
Teramura, Hideo. 1975. "Case grammar and the teaching of Japanese."
Nihongo kyooiku, No. 26.
Tokieda, Motoki. 1950. Nihongo bunpoo: koogo-hen. Tokyo: Iwanami
shoten.
WA FROM SYNCHRONIC AND DIACHRONIC PERSPECTIVES 263

Verma. Manindra K. 1976. The notion of subject in South Asian languages.


Madison: University of Wisconsin-Madison.
Yamada, Yoshio. 1953. Nihon bunpoo-gaku yooron. Tokyo: Soogensha.
Yamamoto, Yuuzoo et. al. 1965. Shoogaku kokugo. Tokyo: Nihonsyoseki.
Yosano, Akiko. 1914. Shinyaku genji monogatari. Tokyo: Kanao bunen-
doo.
Yosano, Akiko. 1976. Shin-shin-yaku genji monogatari. Tokyo: Kawade
shoboo shinsha.
WA IN DIACHRONIC PERSPECTIVE

CHARLES M. DE WOLF
Chiba University, Japan

1. Introduction
While undoubtedly the far greater part of the already large body of
research on Japanese wa offers a purely synchronic analysis of the particle's
syntactic function, the concerns of this paper are primarily historical and
descriptive. The following three sections consist of: 1) a brief discussion of wa
in typological and contrastive perspective; 2) a more specifically diachronic
description, including (a) etymology, (b) pronunciation and orthography,
and (c) syntactic function; 3) a summary of similarities and differences in
the use of wa between Old Japanese (OJ) and Modern Japanese (MJ).
From this study, I draw two tentative generalizations: 1) while it is
widely assumed that the syntactic function of wa developed from an original
interjectional usage, the evidence to support this assumption is unclear; in
fact, functional overlap characterizes the earliest appearances of the parti­
cle. 2) In MJ, the major (though not sole) function of wa is that of topic-
marker. For OJ, on the other hand, the particle may be more accurately
described as a general thematic phrase marker.

2. Typological and contrastive considerations


Studies in syntactic typology in recent years have encouraged attempts
at postulating causal links between various grammatical phenomena, with
the goal of predicting, if not explaining, similarities and differences across
languages. While the purpose of this paper is not to offer even a tentative
suggestion for "explaining" the existence of the postposition wa in terms of
the overall syntactic structure of the Japanese language, it is useful to pre­
face this diachronic treatment of the particle from a broad typological and
contrastive perspective.
266 CHARLES M. DE WOLF

That general theoretical interest in wa is in part a reflection of contem­


porary studies in syntactic typology can be seen most clearly in Li and
Thompson's (1976) subject/topic prominent dichotomy. NP-wa + S' con­
structions, for example, offer clear evidence of topic prominent characteris­
tics in Japanese. 1
The existence of topic as a salient characteristic of a language does not,
of course, necessarily imply grammatical marking comparable to wa. In
Chinese, for example, the construction corresponding to a topic-comment
structure comparable to watashi wa okane ga nai 'I do not have money'
does not grammatically mark (or thereby distinguish) either I or money.
(1) wo méi-yôu qián.
I NEG be money
'I don't have any money.'
If topic marking cannot be predicted from topic prominence, we might
next consider other typological means of classifying Japanese, e.g. its
agglutinative morphology, SOV word order, and the postpositional case
markers. Here too, however, evidence for a syntactic link is disappointing.
The use of agglutinative postpositions to indicate grammatical relations is,
for example, a general characteristic of the Altaic languages, but in striking
contrast to Japanese, thematic or topical elements are not morphosyntacti-
cally marked. Furthermore, whereas in Japanese, subjects are typically fol­
lowed by wa or ga, they are again unmarked in the Altaic languages.
(2) Taroo ga/wa hashitta
Taro ran
'I ran.'
cf. (3) Mehmet (0) kadı (kad-ı ) evde (ev de) döv-dü
Mehmet woman DO house in hit PST
'Mehmet hit the woman in the house.' (Turkish)
The one language in which we find an exact structural parallel to
Japanese wa is Korean, which, significantly enough, is typologically identi­
cal to Japanese in respect to subject/topic prominence, word order, and
agglutinative case marking.
(4) kare wa/ga hashitte iki-mashita.
he running go PST
WA IN DIACHRONIC PERSPECTIVE 267

(5) ku- nun/ka tal-lye ka-ss-umnita.


he TM SB run go PST POL
'He ran.'
The phonological dissimilarity between two particles strongly argues
against a common source. Two remaining possibilities are 1) parallel but
independent development and 2) syntactic borrowing. Proving cases of the
latter is, of course, notoriously difficult, made all the more so in this
instance by lack of adequate knowledge regarding the precise linguistic
nature of the Japanese-Korean relationship as a whole. On the other hand,
the problem of demonstrating structural imperatives for the development of
wa/(n)un depends heavily on interesting but fragile assumptions regarding
linguistic universais.
A significant though admittedly far from conclusive piece of evidence
is found in Ramstedt (1979), who postulates that Korean (n)un is related to
"Tungus nun 'if, 'as to'...The same [Korean] particle is also affixed to con-
verbum presentís in the endings -myen (-m + ye + nun) and -ke den ( < -ke
ide nun) of the conditional and hypothetical coverbs." (p.166) Similarly in
Japanese, as noted below, wa and (conditional) ba derive from the same
source. 2 Whether such a postpositonal particle followed a simplex or com­
plex (sentential) NP, its function was the same, i.e. to set off the phrase
from the rest of the sentence. In this sense, it may be compared to the
development of absolute constructions attested in Indo-European lan­
guages. 3
It should be emphasized that the above section aims merely at suggest­
ing possibilities and does not claim to solve what may very well be forever
a matter of refined guesswork. Within the narrower confines of J vs. MJ
wa, on the other hand, more definite claims can and will be made.

3. Origin and development of wa


Investigations into the origin and development of function words and
bound morphemes differ from etymological studies of full lexical items in
regard to both the goals they seek and the potential difficulties they are
likely to encounter. It is one thing, for example, to establish a lexical
relationship between words in presumably cognate languages and quite
another to demonstrate a common functional history. The relationship
between Latin per and English for may serve as an illustration: As lexical
items, the two are clearly related, but the fact that they happen to share
268 CHARLES M. DE WOLF

comparable functions as verb prefixes in such words as forbear and perfero


is of more doubtful significance, particularly if our goal happens to be, let
us say, the reconstruction of word formation principles in Indo-European. 4
Likewise, the fact that English the happens to be cognate with the Greek
definite article (ho, he, to) may be of lexical interest and may further shed
light on common tendencies in the development of articles (i.e from
demonstratives), but this would not, of course, justify positing an article
system for the IE proto-language. 5
The development of articles in the European languages provides a use­
ful parallel for this historical study of Japanese wa. Here too, by distin­
guishing etymological and functional factors, we are free to consider possi­
ble cognates outside of Japanese on purely lexical grounds. At the same
time, as illustrated above, we can also propose tentative explanations for
the development of functionally similar elements in other (related or unre­
lated) languages.

3.1. Etymology
Efforts to trace the etymology of wa are limited both by the antiquity
of its usage (it appears in the earliest records of the language) and by the
lack of definitive evidence for lexically related elements in the Altaic lan­
guages. Nevertheless, within the realm of reasonable possibilities, Martin's
(1975) "etymological speculations" are both interesting and useful. These
include the following:
(1) wa < (bound morpheme) -ba 'place', which in turn is derived
from free form *pa, cf. Korean pa 'thing, circumstance, way'.
(2) ba as a doublet of ma 'interval, room'.
(3) "A more interesting possibility would be to consider wa<ba as a
nounlike derivative from wi-<bi- = wor <bo(-r)- 'be'; that is, wi-a>wa
'being..' The modern Okinawan copula yan would seem to be clearly an
amalgam of the particle ya + the auxiliary a-η (equivalent to Literary
Japanese ari) as we can see from the negative form: kii yan 'it is a tree': kii
ya aran 'it is not a tree'. That ya <wa<ba might earlier have derived from
a verb of existence would be very natural." (p.88) 6
Martin does not explain how such a development would be "natural,"
but this hypothesis is undoubtedly the most interesting of the three, if only
because it suggests a broader relationship to Altaic *-bi 'be'. Murayama
and Obayashi (1973), on the other hand, suggest that wa is closely related
WA IN DIACHRONIC PERSPECTIVE 269

to object marker wo in both origin and function, expanding on a claim


made in Murayama (1957). This in turn makes possible a direct comparison
with Proto-Tungusic *-/ә.7

3.2. Pronunciation and orthography8


The particle which for the sake of clarity is represented throughout this
paper as wa is written in Modern Japanese as hiragana (ha). In Old
Japanese, the initial consonant was probably a bilabial fricative (customar­
ily romanized as F-), going back to a still earlier p. In Man yōgana, Chinese
characters used to represent Japanese sounds, the particle appears both
as (Ancient Chinese pua 'wave') a purely phonetic representation (from
which the hiragana symbol is ultimately derived) and as (Ancient
Chinese ffia), a nominalizing particle in Chinese. 9
Following imperfective and perfective forms of the verb, as well as wo
(see IV), the particle is voiced as [ba], written in Manyogana as 'old
woman' (Ancient Chinese pua, Sino-Japanese ba). In his translation of the
(8th Century) Kojiki, Philippi (1968) represents the participle as pa, on the
basis of "evidence that the" [p] pronunciation was preserved until the early
Heian period." (p.24) Possible counter-evidence for this hypothesis is that
in the Kojiki itself we find an instance of the exclamatory use of the particle
written as (Ancient Chinese \Twa), suggesting that dialectal variation
between Fal-wa (e.g. Fara 'fielď/Fujiwara 'wisteria field') and Fa/-ba (e.g.
Matsubara 'pine field') predates Classical Japanese:
(6) Iza agi/ Furu- galitate oFa- zu wa/nipo-dori
 my lads Furu-kuma wounds bear NEG Nipo bird
nolaFumi no umi ni/kaduki se- wa
Afumi lake plunge EX
'My lads, rather than enduring the wounds inflicted by Furu-
kuma, let dive like the Nipo birds into waters of Lake Afumi
(=Biwa).'
(Chapter 96)

3.3. Function
Whether or not wa is etymologically related to wo, the fact remains
that they share a similar functional development. In Old Japanese, both
appear as exclamatory particles. The following examples are taken from the
Manyōshw.
270 CHARLES M. DE WOLF

(7) Faru-gasumi / tanabiku toki ni /koFi no sigeki wa.


spring mist draw in time love thick
'Whenever the spring haze draws in, oh, the denseness of love!'
(8) Ikeru Fito / tuFi ni mo sinuru /mono ni are-ba /
living person eventually die beings CO pbe- because
kono yo naru ma wa/tanosi-ku wo ara-na.
this world be space pleasantly be IMPE
'As all men must one day die, let us enjoy ourselves while we are
in this world.'
(9) Fito- goto no/sigeki kono koro/ tama naraba/te ni
person-word tick this time jewel if hand
maki-moti-te/ koFi- zaramasi wo
wrap holding love (= amorous longings) NEG-HYP
'Amid all the gossip of late, if you were but a jewel, I would wrap
you round my wrist and no longer feel the pangs of love.'
The question that arises in regard to both wa and wo is whether the dis­
tinction between exclamatory and case-marking functions requires us to
postulate for each the existence of two distinct but homophonous particles.
Miller (1971)'s argument for making such a distinction on the basis of "syn­
tactic occurrence" is applied only to wo, but it can readily be extended to
wa and is thus relevant for discussion here. 10
If by "syntactic occurrence," Miller simply means position in the sen­
tence, the distinction between non-final and final occurrences of the parti­
cles is hardly of crucial significance, since a final NP-wo may be postposed
or imply an (ellipted) verb. (10)-(14) are taken from the Manyōshw.
(10) Kose-yama no/turu-turu tubakil tura-tura ni/mi tutu
Mt. Kose in a row camellia carefully gazing at as
sinoFa-na I kose no haru-no wo
think would spring fields
'Gazing intensely at its host of camellias, I would contemplate
Mt. Kose's spring fields.'
(11) Kawa-no-Fe no/tura-tura tubaki/tura-tura ni/mire-do-mo
riverside gaze even
aka- zu/kose no haru-no wa
grow weary NEG
'I gaze without wearying at the host of camellias along the river­
side, Mt. Kose's spring fields.'
WA IN DIACHRONIC PERSPECTIVE 271

(12) Tarnamo karu/wotome wo sugite/natu-gasu no/


raw seaweed cut Otome passing summer grass
noshima-no-saki niliForisu ware wa
Noshima Cape construct a make-shift hut
'Passing Otome, where they cut the raw seaweed, I build my
make-shift hut on Cape Noshima, rank with summer grass.11
A better argument against the single wo (and single wa) hypothesis is
that the particle following verb forms is not the same as that following NPs.
(13) Nagaki  wo/kimi ni koFi- tutu/ ikera zu wal
long night yearn for continuing living not
sakite tiri- nisi /hana nara- mashi wo.
bloom scatter PST blossom become HYP
'As my lord for whom I go on yearning these long (autumn)
nights is no longer living, it is well for blossoms to (quickly) fall.'
(14) Aratarasi-ki/tosi no hazime ni/toyo no tosi/sirusu to
new year beginning fruitful year promise
narasi/yuki no Fureru wa
seem snow having fallen ATTRI
'On New Year's Day, the (heavy) snowfall seems to promise a
rich harvest.'
Here too, however, the fact that direct nominalization is very common in
Old Japanese, more so than in the modern language, provides a ready
counterargument. Fureru 'having fallen' in (14), for example, is the attribu­
tive or nominal form of the verb (Fure+ri). The presence of genitive/
embedded subject marker no further signals the nominal character of the
phrase yuki no Fureru 'the falling of the snow', which may thus be analyzed
as a postposed topic. 12 cf. (15) from the Kojiki:
(15) kuso nasu wa weFite Faki- tirasu to koso a-ga
faeces seem being drunk vomit scatter as my
nase kaku si-tu rame
dear brother thus do PST
(Ama-terasu-o-mi-kami, sun goddess and ancestor of the impe­
rial family, speaking of her brother Susa-no-wo.)
'That which appears to be faeces must be what my brother has
vomited and strewn about while drunk.' (Philippi's translation)
The semantic as well as the syntactic link between the exclamatory and
272 CHARLES M. DE WOLF

thematic functions of wa can be seen in the vocative use illustrated in (16)-


(17), again from the Man'yōshu:
(16) awa- yuki wa/keFu wa na Furi so/
light (spring) snow today
sirotaFe /sode і
white-clot ("pillow word") sleeve roll (=pillow)
Fosa-mu/* Fito  r (*meaning of this
dry PRESUM person be NEG phrase disputed)
'Spring snow, do not fall today, as I have no one (who lying be­
side me) who will dry my sleeves when I make them my pillow.'
(17) wa-ga seko wa/kari- (i)Fo tukura-su /kaya
my husband temporary shelter make HON grass
 walko matu ga moto no/kusa so kara-sa- ne.
NEG small pine base grass cut HON IMPER
'If you, my husband, have no grass with which to build your tem­
porary shelter, cut some from beneath young pines.'
In his diachronic discussion of wa, Sansom (1928) dismisses etymologi­
cal speculations concerning the lexical source of wa with the following com­
ment: "Obviously wa is one of the earliest elements in the language and it
is idle to conjecture its origin." (p.262) In regard to syntactic function, on
the other hand, he clearly assumes a developmental process:
"... it can be stated that the early language contained a large number
of particles of an exclamatory or emphatic nature, not fully differentiated
from one another, in form, meaning, or function.
"Certain combinations of particles gradually come to assume special
meanings, and to perfom syntactical functions... Ha or wa itself was, to
judge from its semantic development, originally exclamatory or emphatic,
but is now specialized as an isolating particle denoting the subject of prop­
osition, with an extended use by which it acts as a conjunctive. It is worth
noting that wa is still used as an exclamatory particle in the modern col­
loquial..."13
"These considerations are not adduced in support of the interjectional
theory of the origins of language; but they do throw light on the genesis of
grammatical forms, and they show at least that it is dangerous to assume
that all suffixes and flexional endings were once independent and signifi­
cant words." (pp.286-287).
As the date of publication may suggest, Sansom's work is marred by
lingering traces of linguistic evolutionism. The fact of the matter is that
there is no evidence that the exclamatory use of wa predates its more spec-
WA IN DIACHRONIC PERSPECTIVE 273

ifically syntactic function. The best general definition of the role of wa from
the earliest records remains G.W. Aston's (1872): "a separative or distin­
guishing particle."
In this section, I have briefly discussed possible etymologies regarding
wa and their implications, particularly in regard to wo; I have also offered
background information regarding pronunciation and orthography. My
primary concerns, however, have been: 1) whether wa represents a single
particle with extended functions or more than one particle, sharing the
same phonological form; 2) whether a developmental process of wa's syn­
tactic function can be inferred from historical evidence. In the first case, I
have argued for a single particle hypothesis; in the second, I have suggested
that the evidence for an historically older, functionally more basic usage is
far from clear.

4. The distribution and function of wa in MJ and OJ


A. Similarities
For the purposes of diachronic comparison, Martin's (1975) three func­
tional categories for wa as it appears in the modern language provide a use­
ful point of departure:
(1) to introduce an interrogative or to answer a question:
kore wa nan desu ka? (kore wa) hon desu.
this what COP QU this book COP
'What's this?' 'This is a book.'
(2) to highlight a negative
kore wa kinoo no yuukan de wa arimasen.
yesterday evening edition COP be
'This isn't yesterday's evening edition.'
(3) to indicate a contrast
boku wa binboo da kedo nyooboo wa kane mochi da.
I poor COP but my wife money having COP
'I am poor, but my wife is rich.'
Examples of all three categories are readily found in Old Japanese. The fol­
lowing data are taken from the Kojiki (A.D. 712):
(18) a. imasi-tati wa ta zo?
you PL who EX-COP
'Who are you (pl)?' (Kojiki, Chapter 19)
274 CHARLES M. DE WOLF

b. a wa kuni-tu-kami oFo-yama-tu-mi-no-kami no ko
I earthly deity O-yama-tsu-mi-no-kami child
nari.
COP
'I am a child of the earthly deity O-yama-tsu-mi-no-kami.'
(19) a. wa-ga ko wa sina-zu te ari-keri
my child die NEG GER be PERF
'My child is not dead (but rather) is here.' (Chapter 34)
b. oki-tu-tori /kamo doku sima nil
offing birds (= "pillow word") wild ducks land island
wa-ga wi-ne- .silimo wa wasure-zi /yo
my sleep together PERF beloved forget NEG life
koto-goto ni
everything
'As long as I live, I shall not forget my beloved, with whom I
slept on an island where wild birds came to land.' (Chapter
45)
(20) kami tu se wa se haya- si. simo tu se wa se
upper GEN current current swift lower
yowa-si
weak
'The current of the upper stream is a current too swift; the cur­
rent of the lower stream is a current to weak.' (Chapter 11)
Sentences of the form NP-wa + S' [NP- (ga/no) VP] S', e.g. zoo wa
hana ga nagai 'elephants have long noses' may be treated in terms of Mar­
tin's first category.14 Such constructions are found in Old Japanese as well
as in the modern language, though in the former the subject of the embed­
ded clause is more likely to be directly linked to the predicate, i.e. without
intervening ga or no, as can be seen in (20)-(21), cf. (22)-(23) (Man'y ōshu):
(21) waga oFo- kimi /kі no mikoto no /taka-sirasu/
our great lord majesty splendidly rule
Futagi no miya walmomoki nasu yama wal
Futagi palace hundreds (=countless) seem mountain
ko-dakasi...
tree tall
'Futagi Palace, where our divine Sovereign rules in splendor: the
mountains are covered with stately trees...'
WA IN DIACHRONIC PERSPECTIVE 275

(22) ari-gayou/ naniFa no miya lumi


frequented (by our Sovereign Naniwa palace sea
tikami/ ama-wotomera ga noveru Fune miyu.
near fisher maids riding boat be seen
T h e palace of Naniwa, oft visited by our Sovereign, is so near
the sea that the boats on which the fisher maids ride can be seen.'
(23) sakisima no/yamato no kuni wa/koto-dama
("pillow word") Yamato (= Japan) land word-spirit
no /tasukuru kuni zol masakiku are koso
GEN/SUB help land EX-COP may you be blessed!
'The Land of Yamato is a land aided by the word-spirit*, so be
blessed!' (*a kind of word magic).'
B. Differences
While uses of wa in MJ are all found in OJ, the reverse does not hold
true. The more significant differences in the two forms of the language are
summarized as follows:
1) In OJ, as in MJ, wa serves to separate thematic information from
rhematic (predicative) information, thereby — to use Martin's terminology
— "subduing" the former while "highlighting" the latter. In MJ, however,
the wa-marked element is more likely to be the subject of the sentence than
in OJ. Martin cites National Language Research Institute statistics showing
that subjects account for 61% of wa-marked elements. Furthermore, "the
subject is focused and thematized rather more often than not — almost
56% of the time." (p.59) In Old Japanese, subjects, both thematic and non-
thematic, are frequently unmarked, as can be seen in the following excerpt
from the Kojiki:
(24) kare,  oFo-kuni-nusi-no-kami, munakata no oki-tu-miya
then this O-kuni-nushi-no-kami Munakata Okitsumiya
ni masu і takirі-bіі--mikoto wo me-tori-te
dwell god Takiri-bime-no-mikoto wife took
umimaseru ko adi-siki-taka-Fikone-no-kami, tugi ni
give birth child Aji-shiki-taka hikone-no kami next
imo-taka-Fime-no-mikoto, mata no na wa
Imo-taka-hime-no-mikoto also name
sita-teru-pime-no-mikoto.
Shita-teru-hime-no-mikoto
276 CHARLES M. DE WOLF

cf. the MJ translation:


sate,  o-kuni-nushi-no-kami ga munekata no
then this O-kuni-nushi-no-kami SUBJ Munekata
okitsu-miya ni chinza nasaru і 
Okitsu-miya be enshrined do HON god
takiri-bime-no-mikoto wo tsuma to shite o- umi ni
Takiri-bime-no-mikoto wife as HON give birth
natta ko wa aji-shiki-taka-hikone-no-kami, tsugi ni
to child Aji-shiki-taka-hikone-no-kami next
imoto taka-hime-no-mikoto de betsu-mei wa
younger sister Taka-hime-no-mikoto COP other name
shita-teru-hime-no-mikoto to iu і de aru.
Shita-teru-hime-no-mikoto QT god COP (Ogiwara (1983)
Then this O-kuni-nushi-no-kami took as his wife the goddess
Takiri-bime-no-mikoto, who dwells in the Oki-tsu-miya of
Munakata, and the child she gave birth to was Aji-shiki-taka-
hikone-no-kami, followed by a younger sister, Taka-hime-no-
mikoto, also known as Shita-teru-hime-no-mikoto.'
In quoted speech, wa following the subject is somewhat more common.
(25) kare, sara ni mata- kaFeri-kite, sono oFo-kuni-nusi-no-kami
then again again return come that O-kuni-nushi-no kami
ni toFi-tamaFaku, "imasi no ko-domokoto-siro-nusi-kami,
ask HON your children Koto-shiro-nushi-kami
take-mi-na-kata-no-kami no Futa-Fashi-ra no kami wa,
Take-mi-na-kata-no-kami both god
ama-tu-kami no miko no mikoto no manima ni tagaFa-
heaven god child word exactly disobey
zi to mawosi-woFarinu. kare, imasi no kokoro wa
NEG QUOT say then your mind
ikani.
how
'Then (Take-mika-zuchi-no-kami) returned and inquired of O-
kuni-nushi-no kami: "Your sons Koto-shiro-nushi-no-kami and
Take-mi-na-kata no-kami have both said that they will not dis­
obey the commands of the offspring of the heavenly deities.
What is your intention in regard to this.?'" (Philipi's translation)
WA IN DIACHRONIC PERSPECTIVE 277

Sansom suggests that the greater frequency of wa in the modern language is


related to the loss of the distinction between attributive (rentaikei) and pre­
dicative (shūshikei) forms.
"(1) In hi wa atsushi the subject is defined by wa, the predicative by the
special predicative termination shi.
(2) In kawa wa nagaru 'rivers flow', the subject is defined by wa, and the
predicate by the special predicative termination ru. In this, as in the
foregoing case, the colloquial has abandoned the use of the special pre­
dicative terminations and thus given an added importance to wa." (pp.259-
60)
The claim, though not implausible, is of course speculative. A safer
generalization would be that the particle has come to take on a more purely
functional role. 15 In any case, it is interesting to note that in Modern
Japanese, wa is far more likely to be omitted in the spoken than the written
language. Moreover, while the presence of wa in Old Japanese could be
treated as involving insertion, the opposite holds true for Modern Japanese,
in which the absence of wa may be (and generally has been) treated as
involving "ellipsis."16
2) In Old Japanese, wa follows the attributive (rentaikei), imperfective
(mizenkei), and perfective (izenkei) forms of the verb, as well as the con­
junctive (renyōkei) form of both verbs and adjectives. Following the imper­
fective and perfective verb form, the particle appears as ba and, as in the
case of adjectives in the renyōkei (...-ku (wa)), forms a conditional, either
causal or hypothetical. (26)-(29) are taken from the Manyōshu.
(26) yuku sa ni wal Futari waga-misi/  saki wo/
go occasion two our seeing this cape
hitori sugure- (PERF) bal kokoro-ganasi mo
alone pass heart sad
'Now alone, when returning past the cape (Minume), which the
two of us once beheld, my heart is filled with sadness.'
(27) iFe nara- ba/imo ga te -  / kusa-
home be (IMPERF) wife's arm pillow would grass
makura/tabi ni koyaseru / kono tabito aFare
pillow journey lie-HON-ATTR this traveler alas
'If he were home, he would be pillowed in his wife's arms, rather
than a traveler who lies (dead along the road) with grass for his
pillow.'
278 CHARLES M. DE WOLF

(28) (17) wa-ga seko wa/kari- (i)Fo tukura-su /kaya


my husband temporary shelter make HON grass
naku wa/ ko matu ga moto no/kusa so kara-sa-
NEG small pine base grass cut HON
ne.
IMPER
'If you, my husband, have no grass with which to build your
temporary shelter, cut some from beneath young pines.'
Since wa in naku wa (28) has the same function as ba in nara-ba (27),
the relationship between the two variants of the particle is, in terms of a
synchronic description of Old Japanese, morphophonemic and not merely
etymological. In Modern Japanese, on the other hand, in which ((r)e) ba
(as a conditional) is the only form remaining in the colloquial, the relation­
ship is of no synchronic significance.
3) In Old Japanese, as in the modern language, wa may be combined
with other particles, e.g. ni:
(29) iwaro ni wa/asiFu take-domo /sumi yoke wo/tukusi
house reed-fire cook living good Tsukushi
ni itaritelkoFusike mowa mo.
arrive love think
'I may be cooking over a reed fire in my house, but I live well.
Still, since arriving in Tsukushi, I cannot help but have loving
thoughts of home.'
Other combinations lacking in the modern language are also found, includ­
ing wo-ba (<-wo-Fa), wa-ya, and wa-mo.
(30) "Kono imasi no musume wo-ba a ni tatematuramu
this your daughter me present HUM
ya:
QU
'Will you give me your daughter.' (Kojiki, Chapter 19, Philippi's
translation)
(31) "Wotome no / toko no be ni / wa-ga okisi /turugi no
maiden bed my left sword
tatil sono tati wa-ya.
long sword that sword
'Next to the maiden's sleeping-place I left the sabre, the sword —
alas that sword!' (Kojiki, Chapter 87, Philippi's translation)
WA IN DIACHRONIC PERSPECTIVE 279

(32) "Sane-sashi/ sagamu no wo-no ni/moyuru Fi no/Fo-


(pillow word) Sagamu plain burn fire fire
 ni tati-tei toFisi kimi wa-mo.
midst standing spoke Lord
'O you, my lord, alas — you who once, standing among the
flames, of the burning fire, spoke my name on the mountain-sur­
rounded plan of Sagamu.' (Kojiki, Chapter 84, Philippi's transla­
tion)
Both wa-ya and wa-mo are classified as emotive particles. Martin
(1975) states in regard to the latter: "Morishige...says that Old Japanese
had such expressions as o-neesama ['older sister'] WA MO, but I have not
confirmed this, Meikai kogo jiten 812c gives examples of sentence final wa
mo, but the wa is exclamatory." (p.55, footnote) Consider, however, the
following:
(33) Faru- Fi wo kasuga no yama/ taka-
spring day (= pillowword) EX-? Kasuga hills lofty
kura no/mikasa no yama/ asa saraiul
throne Mikasa mountain morning every
kumo-wi tanabiki /kaFo-tori no/ma-naku siba
cloud (motionless) layered ? bird ceaseless often
nakul kumo-wi- nasu / kokoro isayoFi I sono tori
cry cloud (motionless) like heart hesitating that bird
no I kata koFi nomi ni/Firu wa mo/Fi no koto-goto /
one-sided love merely day all day
yoru wa  I yo no koto-goto / tatite wi-te lomoFi so
night all night standing sitting think EXCL
α-ga suru awa-   yuwe ni. (Man'yōshu)
I AUX meet NEG girl because of
'Among the hills of Kasuga, where the spring sun is dimmed in
the haze [an untranslatable pun is contained here] on Mikasa
Mountain, like the crown on a lofty altar, the clouds trail every
morning and the cry of the kao* bird never ceases. Like the
motionless clouds, my heart is hesitant, and like that bird I spend
my days and my nights getting up and sitting down, yearning for
the girl who will not meet me.' (*meaning/species unclear)
Here wa thematizes the adverbs, Firu 'day' and y oru 'night', while mo puts
them in an inclusive relationship with one another. In the modern lan-
280 CHARLES M. DE WOLF

guage, either wa or mo is possible (hiru wa - yoru walhiru  - oru ),


but not both.
Another particle combination lacking in the modern language is i-wa,
where і is a rarely occurring postposition which disappears sometime after
the 9th Century. If Miller (1971) is correct in suggesting that this particle is
a borrowing of Korean nominative postposition -i, it is interesting to note
the compatibility of nominative and wa in contrast to the modern incom­
patibility of ga and wa:
(34) ina to iFe-do/katare katare to Inorase kosolsiFi
no! say speak! speak! QUOT say HON even Shihi
i-wa mawose siFi katari to iFu.
speak force speak say
Though I say I do not want to speak, you command me to, say­
ing when I, Shihi, do that I am forcing it upon you.'
(a poem by an old woman in response to a poem by the Empress)
4) The form of cleft constructions (S' -wa + NP (COP)) differs in OJ
and MJ, in that in the modern form of the language, there is an overt
nominalizer (no), while in OJ, the (partial) distinction between predicative
and non-predicative verb forms was itself a indication. In (35), however,
iFu belongs to the quadrigrade conjugation and is the same in both attribu­
tive and predicative forms. In such cases, wa may be analyzed as carrying
the function of both nominalizer and phrase marker.
(35) kaku ihu wa harima no і no  no, kuraudo
thus say Harima governor son treasury clerk
yori kotosi kauburi etaru nari- keri.
from this year 5th rank attained COP PST
'The one who was speaking, the son of the governor of Harima,
had been a treasury clerk, promoted this year to the fifth rank.'
(Genji Monogatari, Wakamurasaki)
cf. the MJ translation:
kou iu hanashi wo suru no wa, harima no і no ko
thus speaking Harima governor son
de, kuraudo kara kotoshi go- і ni jo-
? treasury clerk from this year five position confer
serareta otoko na no datta.
be man COP NOM -PST
(Abe, Akiyama, Imai, and Suzuki (1983))
WA IN DIACHRONIC PERSPECTIVE 281

In summary, OJ wa serves both as a marker of nouns and nominalized


elements and as a conjunction. In MJ, the obscuring of the relationship
between wa and ba, the restrictions on combinations with other particles,
and the modern use of no to mark clause nominalization have all given wa
its more specialized syntactic function.

5. Conclusion
In this paper, I have attempted to provide a diachronic perspective on
wa. While my purpose has been primarily to offer analytic description, gen­
eral conclusions concerning the development of wa since pre-Classical
usage can also be tentatively drawn. First, the claim that wa originates from
an interjection must be questioned, since, as I have shown, the syntactic
functions found in Modern Japanese are also in evidence from the earliest
times. Furthermore, a clear-cut distinction the two usages based on position
in the sentence is difficult to justify. Secondly, wa in Old Japanese appears
to serve more as a local phrase marker than as a topic marker. Multiple
occurences of wa in a single sentence are clearly more marked in Modern
Japanese (particularly in the written language) than in Old Japanese, while
wa topic marking has become regular if not obligatory. This suggests a
"bleaching" of was semantic content and a strengthening of its syntactic
role at the sentence level.

APPENDIX: Excerpt from Hahakigi ("Broom Tree"), The Tale of Genji


The following text is intended to illustrate the use of wa in Classical
Japanese narrative. Though obviously too brief to support any statistical
generalizations, it may be seen as a useful data source, including as it does
examples of nearly all the observations I have made concerning the func­
tion of wa in OJ:
The following points may be noted:
1) wa does not appear typically as the marker of the subject of the sen­
tence, two out of thirteen instances — in contrast to the 68% figure cited by
Martin for MJ. "nasake wa onodukara... (1.9) and "kimi no mi-kokoro
wa... (1.35).
2) wa is used to mark conditional clauses following the conjunctive (ren-
yōkei) form of adjectives: naku wa (1.3), tuyoku wa (1.9), and nama-
ukabi-nite wa (1.42). Following perfective (izenkei) forms, e.g. susumi-
nure- (1.29), namida-otose- (1.34), hosokere- and kobore-somenure- (1.39),
wa appears as allomorph -ba).
282 CHARLES M. DE WOLF

3) wa is combined with particles wo (→ woba) and ya (→ yawa). katati


woba (1.1) may be analyzed either as a highlighted NP or as a reduced con­
ditional, whereas uti-sohe-taramu woba (1.6) is clearly a full conditional. In
both cases, the use of wo may be understood as involving concessive condi­
tionals. Ikeda (1975) explains yawa (1.51) as follows:
"When wa appears at the end of a sentence it most frequenlty occur in
combinations such [sic] yawa, kawa, zowa, mowa, and its function is to
emphasize the particles ya, ka, zo, mo. This use most probably developed
from the use of wa in inverted word order, and it is usually encountered
only in speech and in waka [Japanese poems]. yawa and kawa express plain
doubt or interrogation as well as irony." (p.219, footnote) Ikeda also gives
an example of yawa as a (simple NP) topic marker:
a. ume no hana/ iro koso mie nul ka yawa
plum blossom color even visible NEG fragrance
kakaruru
hide (ATTRIB)
'Plum blossom: even though its beauty is veiled from our sight,
can its fragrance be concealed?" (Kokin Wakashu)
Note here too the contrast between the distribution of wa here and in
Ikeda's modern translation:
b. ume no hana wa, me ni wa mie nai to shite mo nioi wa,
eye visible NEG even smell
kakureru de aroo ka.
hide TENT QU
4) As in MJ, nouns may function as adverbs, thematized by wa, e.g.
uhe wa (1.9). Adverbial clauses likewise set off by wa, e.g. ...toki wa (1.10),
hodo wa (1.21). It should be noted, however, that ima wa (1.1) is a
lexicalized phrase (= 'well'), which has survived into MJ only as a literary
form. Omohu ni wa (1.21) provides an example of direct nominalization,
found also in (literary) MJ: >omou ni wa.
(I) 1. Ima wa, tada, sina ni mo yorazi, katati woba
well just rank depend not beauty
2. sara ni mo ihazi, ito kuti-osiku nedike-gamashiki
further say not too terribly eccentric
3. oboye-dani naku wa, tada hitohe ni mono-mame-yaka ni
impression not but merely steady
WA IN DIACHRONIC PERSPECTIVE 283

4. sizuka-naru kokoro no omo-muki naramu yorube wozo,


calm personality helper
5. tukino tanomi -dokoro ni wa omohi-oku-bekarikeru. amari
final trust (= wife) decide ought above
6. no yuwe, yoshi, kokoro-bahe uti-sohe-taramu woba
that talent disposition have in addition
7. yorokobi ni omohi, sukosi okuretaru kata aramu womo
joy think little lack side be
8. anagati ni motorne kuhahezi. Usiro yasuku
unreasonably demand add NEG (in) future calm
9. nodoke-ki tokoro dani tuyoku wa, uhabe no nasake wa
gentle place even sure latent tastes
10. onodukara mote-tuketu beki waza woya.
of themselves attach will event
11. Emu ni mono-hadisite, urami ihu-beki koto womo
forbearing resent say
12. misiranu sama ni sinobite uhe wa ture-naku-misa
not know as if endure surface emotionless
13. wo dukuri, kokoro hitotu ni omohi-amaru
personality display heart one think excess
14. toki wa, ihamu kata naku sugoki koto no ha,
time to be said not horrible notes
15. aharenaru uta wo yomi-oki, sinobaru-beki
full of pathos poem leave behind remorse-evoking
16. katami wo to domete, fukaki yama zato,
mementos leave deep mountain settlement
17. yo-banaretaru umi-dura nado ni hahi-kakurenuru wori
remote seashore etc. run off to and hide
18. waraha ni haberi-si toki, nyoubau nado no
childhood time serving women
19. monogatari y omisi wo kikite, ito ahare ni
stories read hear very pathetically
284 CHARLES M. DE WOLF

20. kanasiku kokor-і koto kana to


sad profound thing my how! QT
21. namida wo sake namu otosi-haberi. ima omohu ni wa,
tears even shed now think
22. ito karugaru-siku koto sara-bitaru koto nari.
most frivolous irksome thing
23. kokoro-zashi-fukaramu otoko wo okite, miru me no 
full of love man leave to appearances
24. ni turaki koto aritomo, hito no kokoro wo
painful thing be even person's heart
25. misiranu yau ni nige-kakurete, hito wo madohasi kokoro
not knowing escape hiding person upset heart
26. wo mo mimu to suru hodo ni, nagaki-yo no mono
see TENT do extent long world
27. omohi ni naru,
thing (= lament the end to which one has surely
28. ito adiki naki koto nari, "kokoro hukasi-ya"
come) quite absurd "a determined soul"
29. nado hometaterarete, ahare susumi-nure-ba yagate ama ni
etc. praised pathos deepening finally nun
30. narinukasi. omohi tatu hodo wa ito kokoro sumeru yau nite,
becomes for the time being mind-purifying
31. yo ni kaheri misu-beku mo omoherazu, "ide, ana kanasi,
world return not occur oh, how sad!
32. kaku, hata, obosi nari-ni-keru-yo" nado yau ni,
thus oh what misery! etc. as
33. ahi sireru hito ki to burahi, hita sura ni usi to mo omohi
acquaintances come to see not at all abandon think
34. hanarenu otoko kiki-tsukete namida-otose-ba,
separate NEG man hearing tears shed
35. tsukahu hito, hurugo-tati nado, "kimi no mi-kokoro wa
maid old nurses his lordship
WA IN DIACHRONIC PERSPECTIVE 285

36. ahare nari-keru mono wo, atara o-mi wo" nado ihuni,
manofpity pitiful you etc. say
37. midukara hita-higami wo kaki sagurite, ahenaku kokoro-
herself hair goes to touch alas! grieve
38. hosokere-ba, uti-hisominu-kasi. sinoburedo namida
wears a brave face yet trying tears
39. kobore-some-nure-ba, wori-wori goto ni e-nenzi ezu,
shed all the time cannot pray
40. kuyasiki koto mo ohokameru ni, hotoke mo naka-naka
regret things so many Buddha even rather
41. kokoro-gitanasi to mi-tamahitu-besi. nigori ni simeru hodo
mind sullied see will those-sullied-by-the-
42. yori mo, nama-ukabi-nite wa, kaherite
world even more being a half-hearted nun all the more
43. asiki miti ni mo tada-yohi- beku zo oboyuru. taenu
way-to-hell drifting will EX appear karma
44. suguse asakarade, ama ni 
link between husband and wife not shallow nun not
45. nasade tadune tori taramu,  agate
yet becoming visiting take in the end
46. ahi-sohite, sono omohide, uramesikihusi arazaramu ya
meet accompany that memory resentful of inevitable
47. asiku mo yoku mo ahi-sohite
bad good accompanying
48. to aramu wori-mo kakaramu kizarni womo
no matter what else happens occasions
49. misugu sitaramu  koso, tigiri-hukaku
overlook with-determination-to vow deep
50. ahare-narame, ware mo hito mo usiro-metaku kokoro
realizing self other anxieties
51. okarezi yawa.
not resolved lest
English translation by Seidensticker (1981), pp.25-2617
286 CHARLES M. DE WOLF

"No, let us not worry too much about rank and beauty. Let us be satis­
fied if a woman is not too demanding and eccentric. It is best to settle on a
quiet, steady girl. If she proves to have unusual talent and discrimination —
well, count them an unexpected premium. Do not, on the other hand,
worry too much about remedying defects. If she seems steady and not given
to tantrums, then the charms will emerge of their own accord.
"There are those who display a womanly reticence to the world, as if
they had never heard of complaining. They seem utterly calm. And then
when their thoughts are too much for them they leave behind the most hor­
rendous notes, the most flamboyant poems, the sort of keepsakes certain to
call up dreadful memories, and off they go into the mountains or to some
remote seashore. When I was a child I would hear the women reading
romantic stories, and I would join them in their sniffling and think it all
very sad, all very profound and moving. Now I am afraid that it suggests
certain pretenses.
"It is very stupid, really, to run off and leave a perfectly kind and sym­
pathetic man. He may have been guilty of some minor dereliction, but to
run off with no understanding of his true feelings, with no purpose other
than to attract attention and hope to upset him — it is an unpleasant sort of
memory to have to live with. She gets drunk with admiration for herself and
there she is, a nun. When she enters her convent she is sure that she has
found enlightenment and has no regrets for the vulgar world.
"Her women come to see her. 'How very touching,' they say. 'How
brave of you.'
"But she no longer feels quite as pleased with herself. The man, who
has not lost his affection for her, hears of what has happened and weeps,
and certain of her old attendants pass this intelligence on to her. 'He is a
man of great feeling, you see. What a pity that it should have come to this.'
The woman can only brush aside her newly cropped hair to reveal a face on
the edge of tears. She tries to hold them back and cannot, such are her
regrets for the life she has left behind; and the Buddha is not likely to think
her one who has cleansed her heart of passion. Probably she is in more
danger of brimstone now in this fragile vocation than if she had stayed with
us in our sullied world.
"The bond between husband and wife is a strong one. Suppose the
man had hunted her out and brought her back. The memory of her acts
would still be there, and inevitably, sooner or later, it would be cause for
rancor. When there are crises, incidents, a woman should try to overlook
them, for better or worse, and make the bond into something durable."
WA IN D I A C H R O N I C P E R S P E C T I V E 287

NOTES
I wish to thank the editors of this volume, in particular John Hinds, for suggesting needed revi­
sions of this paper. All errors are of course, my own.
1) Li and Thompson classify Japanese, along with Korean, as both topic and subject promi­
nent. (i) illustrates the former tendency (i.e. (the one who) went to school is who?), (ii) the lat­
ter:
i. Gakkoo ni itta no wa dare ka?
school to went NOM TOP who QU
'Who went to school?'
ii. Dare ga gakkoo ni itta ka?
who school to went QU
'Who went to school?'
2) The relationship between wa and ba is generally accepted by both Japanese and non-
Japanese scholars. In regard to Korean (n)un, Lee Ki-Moon (1975) assumes a common origin
with Mongol third person genitive -ni (and thus, we would presume, Altaic *-n).
3) Structural parallels aside, it is interesting to note that according to Lehmann (1974), varia­
tion in the form of absolute constructions (locative for Sanskrit, genitive for Greek, ablative for
Latin) indicates that they arose independently. He adds, however, that "such independent ori­
gins of parallel constructions are plausible only if we can ascribe them to typological reasons."
(pp.210-211)
4) In addition to the problem of calques (e.g. Late Latin pardonare from OHG forgeben),
there is the fact of productivity: *for- continued to be a formant in the Germanic languages into
relatively recent times.
5) For an interesting discussion of the article in Indo-European languages, see Hewson
(1972).
6) In regard to phonology, Martin says: "Traditional kana spelling indicates that wa comes
from proto-Japanese pa, and that would seem to be amply confirmed by the "nigoried" form
ba..., though we might entertain the possibility that ba could be a doublet-relic of the proto form
of wa. (Modern INITIAL w- comes from proto-Japanese b- just as modern h- comes from P?-,
but the /w/ of the particle is not "initial" since particles attach with no juncture; and -w- can
come from either -p- or -b- of the proto language, though the latter is uncommon.") (p.88)
7) "wa/we, the direct object suffixes in Tungusic are strikingly absent in Mongol and Turkic
language groups, but the fact that we find corresponding wo in Old Japanese is of profound sig­
nificance." (Lee Ki-Moon (1975), p.33, translation by CMD)
8) The orthographical representation here of OJ data differs from that of MJ in the following
respects: Consonants that are palatalized or affricatized in the modern language are not marked
as such, (e.g. si, ti, tu rather than shi, chi, tsu. Modern ha, hi, hu ([øu], he, ho are represented as
Fa, Fi, Fu, Fe, Fo for Nara Period (e.g. Manyōshu) data, ha, hi, hu, he, ho for Heian Period
(e.g. Genji Monogatari) data. OJ wo (e.g. wotome> MJ otome 'maiden') is represented as wo.
Except for what might be inferred from this (largely conventional) representation of the OJ con­
sonant system, I have avoided making any suggestion concerning the controversial issues of OJ
phonology, in particular the vowel system.
288 C H A R L E S M. D E W O L F

9) Sansom comments: "Some etymologists have contended that [wa] originally meant mono,
'a thing', which is also so written [i.e. as ]. But the use of was clearly an imitation of
Chinese practice." (p. 262)
10) Miller attempts to clearly distinguish between "direct object" and "time-place" usage by
citing relevant examples from the Manyoshu, but the reader who looks up the poems whose
numbers are cited may be just as easily convinced of an overlapping relationship, e.g. M. 1934.
11) The construction is, of course, ambiguous, analyzable either as a relative clause + subject
without a verb or as a main clause with a delayed subject.
12) Note the modern translation, in which the nominalization is made clear by the use of par­
ticle no.
konna ni yuki ga furi-tsumotte iru no wa
so snow fall stick NOM
'that so much snow is sticking...'
13) The spelling differs, the emphatic particle being written as wa vs. ha for the thematic par­
ticle, but Japanese language scholars in the Kokugo tradition widely assume that they derive
from the same source.
14) For further discussion of sentential predicates, see De Wolf (1985).
15) A stronger argument might be made for a causal relationship between the loss of atrribu-
tive/predicative distinction and the obligatory use of ga (no) as a subject marker in subordiante
clauses. Nevertheless, the basic question that needs to be asked is why case marking in general,
(e.g. wo) is a more consistent feature of the modern language.
Sansom's suggestion that wa has a more crucial syntactic role in the modern colloquial may
again be compared to the theory in Indo-European linguistics that the development of article
systems is to be traced at least in part to the simplification of case systems. See Hewson (1972).
Yet another interesting (though equally unprovable and arguably less plausible) speculation
concerning Japanese is found in Akiba-Reynolds (]981)'s argument for Japanese as a mixed lan­
guage: "...the fact that Old Japanese had very few case markers found in pre-Japanese corre­
sponds perfectly to the observation that pidgin-creoles have very limited noun inflections."
(p.16)
16) For further discussion, see Hinds (1982). Masayoshi Shibatani (personal communication)
provides convincing evidence of subtle but significant semantic differences between subject-ø...
and subject-wa... in colloquial speech, though my own analysis may differ somewhat from his
here, wa draws attention to the NP so marked. It is omitted when the NP is being specified as a
participant in rather than as purely the referent of the predication. watashi wa shinu wa 'I'm
going to die (kill myself)' is a comment about the fate or the action of watashi. watashi shinu wa
is a statement about an event. Consider also the following excerpt from Tetsuro Miura's
Shinobugawa:
"'Nan no suzu?' Shino wa kiita.
what bell Shino asked
'What's (that) bell?'
'ba-sori no suzu, ' watashi wa kotaeta.
horse sleigh bell I replied
'It's a horse-sleigh bell, I replied.'
WA IN D I A C H R O N I C P E R S P E C T I V E 289

'basori tte, nani?'


QUOT what
'What's a horse sleigh?'
'Uma ga hiku sori no koto yo...
horse pull sleigh
Ά sleigh that is pulled by a horse.'
'atashi, mitai wa, ' to Shino ga itta. '"
I see want said
'I want to see it.'
In the last line, the presence of wa following atashi would be inappropriate. The discourse topic
here is basori. atashi is not the referent of the predication but rather a participant in it.
17) The glosses provided here for the text are meant only as a rough guide.

REFERENCES

Abe Akio, . Akiyama, G. Imai, and H. Suzuki, eds. 1983. Genji


Monogatań I. Shogakukan, Tokyo.
Akiba-Reynolds, Katsue. 1984. ''Internal reconstruction in pre-Japanese
syntax," Historical Syntax, in Trends in Linguistics, Studies and Mono­
graphs 23, edited by Jacek Fisiak, 1-23. Mouton, Berlin, New York,
Amsterdam.
Aoki, Takako, Itaru Ide, Haku Itō, Katsuhiko Shimizu, ed. 1976.
Man'yōshü I-V, Shinchoosha, Tokyo.
Aston, W.G. 1872. A Grammar of the Japanese Written Language, with a
Short Chrestomathy. London.
De Wolf, Charles M. 1981. "Nominative, Genitive and Syntactic Typol­
ogy," Descriptive and Applied Linguistics, Vol XIV, 13-24. International
Christian University, Tokyo.
. 1985. "Sentential Predicates in Japanese," Journal of the Association
of Teachers of Japanese, Vol. 19, No. 22, 183-221. University of
Pittsburgh, Pennsylvania.
Hewson, John. 1972. Article and Noun in English. Mouton and Co., the
Hague.
Hinds, John. 1982. Ellipsis in Japanese. Linguistic Research, Inc., Edmon­
ton, Canada.
Ikeda, Tadashi. 1975. Classical Japanese Texts. The Toho Gakkai, Tokyo.
Kurano Kenji, ed. 1965. Kõhoh Kojiki [Kojiki ms] Gakkai, Tokyo.
290 CHARLES M. DE WOLF

Lee, Ki-Moon. 1975. Kankokugo no Rekishi [A History of the Korean Lan­


guage], editorial supervision of Japanese translation by Shichiro
Murayama, translated by Yukio Fuj imoto, Taishuukan Shoten, Tokyo.
Lehmann, Winifred. 1974. Indo-European Syntax. University of Texas
Press, Austin and London.
Li, Charles N. and Sandra Thompson. 1976. "Subject and topic, a new
typology of language," Subject and Topic, edited by C.N. Li, 459-490.
Academic Press, New York.
Martin, Samuel E. 1975. A Reference Grammar of Japanese. Yale Univer­
sity Press, New Haven and London.
Miller, Roy Andrew. 1971. Japanese and the Other Altaic Languages. Uni­
versity of Chicago Press, Chicago.
Miuru, Tetsurō. 1965. Shinobugawa. Shinchösha, Tokyo.
Murayama, Shichirō. 1957. "Vergleichende Betrachtung der Kasus-Suffixe
im Altj apanischen," Studia Altaica. Otto Harrassowitz, Wiesbaden.
Murayama, Shichirõ and Tarō Ōbayashi. 1973. Nihongo no Kigen [The
Origins of the Japanese Language]. Kobundo, Tokyo.
Ogiwara, Asao. 1983. Kojiki. Shogakukan, Tokyo.
Philippi, Donald L., trans. 1968. Kojiki. University of Tokyo Press, Tokyo.
Ramstedt, G.J. 1979. A Korean Grammar. Humanities Press, Atlantic
Highlands, New Jersey. (1939)
Sansom, George. 1928. An Historical Grammar of Japanese. Oxford Uni­
versity Press, London.
Seidensticker, Edward, trans. The Tale of Genji, by Murasaki Shikibu.
1981. Penguin Books, Middlesex, England.
PART IV: PRAGMATIC PERSPECTIVES ON WA
"HOW RELEVANT IS A FUNCTIONAL NOTION OF
COMMUNICATIVE ORIENTATION TO GA AND WAV

SEIICHI MAKINO
University of Illinois at Champaign- Urbana

This paper is a reexamination of the Japanese particles wa and ga in


terms of a functional concept of communicative orientation. To quote the
major findings about wa and ga from Kuno (1984):1
(a) Wa marks the theme of a sentence.
(b) Wa marks contrasted elements.
(c) Ga marks the subject of a neutral description sentence.
(d) Ga marks an exhaustive listing of items.
(e) Ga marks that the subject conveys new information.
(f) Ga marks the object of a transitive stative verbal.
What I intend to do is not to prove that Kuno's findings are wrong; the
intention is rather to show that a superordinate functional concept of com­
municative orientation which can explain a host of other phenomena can
also explain wa and ga, whereas the functional notion of new/given infor­
mation that underlies Kuno's explanation has fewer applications. In short,
inclusion of communicative orientation in the theory of language will
achieve some explanatory adequacy.
First I will explain what communicative orientation is, and then discuss
how it can explain wa and ga in terms of the functional concept.
Communicative Orientation (= CO)
Communicative Orientation is the direction to which communica­
tion is oriented. Depending on whether the communication is directed
towards the listener/reader (which is the 'unmarked' communicative
orientation) or towards the speaker/writer (which is the 'marked' com­
municative orientation), we have listener/reader-orientation or
speaker/writer-orientation, respectively.
294 SEIICHI MAKINO

This statement requires explanation. It is normal for CO to be directed


towards the listener, because communication is normally listener-oriented. 2
More realistically, however, on-going communication can be at least
sporadically speaker-oriented 3 rather than listener-oriented. A clear-cut
case of the speaker-orientation is the case of the monologue.
Strictly speaking a person engaged in a monologue makes himself/her­
self a listener. By the same token, a diary writer makes himself/herself a
reader. In other words the speaker/writer is communicating with a listener/
reader who happens to be himself/herself. But the fact remains that there is
no physically separate listener/reader in a monologue/diary. Therefore, in
my statement of CO the listener/reader should be understood to mean "a
listener/reader who is not identical with the speaker/writer".
Another obvious case of speaker-orientation is the exclamation.
(1) a. a, itai!
oh painful
Oh, ouch!
b. a, shimatta!
oh made-a-mistake
Oh, shoot!
. aa, ureshii!
oh happy
Gee, I'm happy!
d. aa, komatta na.
oh was-in-trouble EX
Gee, what shall I do?
e. a, mata makechatta naa.
oh again lost EX
Gee, I lost again.
Note that unlike monologues, an exclamation is a speech act that does
not assume the speaker himself/herself to be a listener. An exclamation is
normally a spontaneous, automatic reaction to an immediate event. There
might be bystanders near the person who makes an exclamation, and it is
very likely that the bystanders will be made 'involuntary listeners'. Thus, as
in (la), if a person exclaims A itai! "Oh ouch" because his/her toe is step­
ped on by another passenger on a train, all the passengers nearby are made
'involuntary listeners', especially the one who stepped on the toe. Com­
munication between a speaker and an involuntary listener is hardly an opti-
COMMUNICATIVE ORIENTATION AND WA 295

mal communication, because the listener is usually so passive that he


doesn't care to react to the speaker's speech.
According to the distinction between voluntary and involuntary listen­
ers, the 'listener' in the statement of CO should be understood to be a vol­
untary listener, and listener-orientation is, therefore, meant to be an orien­
tation towards a voluntary listener. It follows, then, that if the speaker-
orientation is high, the listener is either the speaker as in a monologue or an
involuntary listener who happens to be present at the time of the exclamat­
ory utterance.
A high degree of listener-orientation involves cases where both
speaker and listener are there to be engaged in conversation and the
speaker either asks a question or makes a request, an order, or a sugges­
tion.
I will now reanalyze the uses of wa and ga in the functional notion of
CO introduced above. The following are a set of hypotheses which I would
like to prove.
(A) Both thematic and contrastive wa are listener-oriented.
(B) Descriptive ga is speaker-oriented and so are exhaustive listing
ga and objective ga.
Thematic wa marks animate/inanimate objects or concepts that are
found in the (semi-)permanent registry of both the speaker and the listener.
(2) a. sumisu-san wa gakusei desu.
Smith TP student COP
Mr. Smith is a student.
b. kujira wa honyuu-doobutsu da.
whale TP mamal COP
Whales are mammals.
c. kinoo kookoo no toki no tomodachi
yesterday high-school LK time LK friend
ni atta n da. sono ko wa moo
IO met NOM COP that girl TP already
kekkonshiteiru tte kiite bikkurishita yo.
married QT hear:GER was-surprised EX
I met my high school girlfriend. I was surprised when I heard
that she is already married.
In all the sentences of (2), the wa-marked nouns (i.e., sumisu, kujira,
296 SEIICHI MAKINO

sono ko) are registered permanently or temporarily in the mind of both the
speaker and the listener. The speaker has used a wa-marked noun to indi­
cate that there is something to talk about — something the knowledge of
which is shared with the listener. In the sense that the speaker is referring
to an object which the listener already knows, the speaker is being listener-
oriented.
The speaker may use wa where it is not warranted, as illustrated by the
following example quoted from Kuno (1984):
(3) A: watakushi, kondo haabaado no suugaku-ka ni
I anew Harvard LK math-department to
kita yamada desü.
came COP
I am Mr. Yamada. I came to the math department of Har­
vard for the first time.
B: haj imemas hite. doko ni osumai desu ka.
"glad to meet you" where LC living COP QU
Glad to meet you. Where do you live?
A: berumonto desu.
Belmont COP
At Belmont.
B: okosan wa, dochira no gakkoo desu ka.
your-child TP which LK school COP QU
Which school are your children attending?
A: iya, dokushin na n desu.
no single LK NOM COP
No, I am single.
The speaker  has wrongly assumed that the speaker A Yamada has a
child. B's original intention was to be listener-oriented in the sense that he
tried to find a topic of conversation with which  could be involved.
It has long been noted that a wh-word such as dare 'who', doko
'where', nani 'what', itsu 'when' etc. cannot be used with wa as shown in (4)
[but see Miyagawa (this volume)].
(4) *dare wa kimasu ka.
who TP come QU
Who is coming?
COMMUNICATIVE ORIENTATION AND WA 297

b. *nani wa arimasu ka.


what TP exist QU
What is there?
c. *doko wa ii desu ka.
where TP good COP QU
Which place do you like?
The standard explanation is that wa, a marker of old information, con­
flicts with a wh-word, because the latter asks for new information. How can
we account for the ungrammaticality of (4) in our functional concept of
CO? A wh-word is speaker-oriented in nature, because it signals the
speaker's desire to obtain information. If wa can be assumed to signal lis­
tener-orientation, then the ungrammaticality of (4) can be automatically
accounted for by an orientational conflict between a wh-word and wa. Note
that if a question is not a wh-question but a yes-no question, the use of the
thematic wa does not create any problem, as shown in (5):
(5) a. yamada wa besu to eiga ni ikimashita ka.
TP Beth with movie to went QU
Did Yamada go to a movie with Beth?
b. nihonjin wa hontoo-ni yoku hatarakimasu ka.
Japanese TP really hard work QU
Are Japanese really hard workers?
The reason why wa can be used in a yes-no question is that there is no
orientational conflict between X and wa sequence. The principle of orienta­
tional conflict can thus be stated as follows:
Orientational Conflict
When a single sentence represents two conflicting orientations (i.e.,
speaker's orientation and listener's orientation), the sentence becomes
unacceptable.
Thematic wa and contrastive wa were sharply distinguished by Kuno
(1972:296). To quote:
Wa marks either the theme or the contrasted element of the sentence. The
theme must be either anaphoric or generic, while there is no such con­
straint for the contrasted element.
As I have argued in Makino (1982), the difference between thematic
and contrastive wa is not as clear-cut as Kuno's generalization suggests. I
would rather argue that contrastive wa is always explicitly or implicitly
anaphoric.
298 SEIICHI MAKINO

Take (6) and (7) for example:


(6) A: yamada-san mo nakada-san mo sake o yoku
also also sake DO a-lot
nomimasu ka.
drink QU
Do both Mr. Yamada and Mr. Nakada drink a lot of sake?
: yamada-san wa yoku nominasu ga, nakada-san wa
TP but
nomimasen.
doesn't-drink
Mr. Yamada drinks a lot, but Mr. Nakada doesn't.
(7) A: soto no otenki wa? (from Kuno (1984))
outside LK weather TP
How is the weather outside?
B: ame wa futteimasu ga, yuki wa futteimasen
rain TP falling but snow TP falling:NEG
It's raining, but it's not snowing.
The proper names Yamada and Nakada in (6B) are marked by con­
trastive wa but each of them has an explicit anaphor in (6A). Ame 'rain'
and yuki 'snow' in (7B) are not mentioned in (7A), but they are items per­
manently registered in the semantic sphere of 'weather'. In short, ame and
yuki have implicit anaphors. I will therefore argue that there is only thema­
tic wa; the contrastive meaning of wa originates not from wa but from the
disjunctive construction X wa — ga, Y wa —. On the surface, contrastive
wa may not be used in a disjunctive construction; in other words, the sec­
ond conjunct Y wa — is deleted along with ga, as in (8):
(8) boku wa tenisu o shinai (ga Hanako wa tenisu o
I TP tennis DO don't but TP
suru).
do
I don't play tennis (, but Hanako does).
In short, all cases of contrastive wa are disjunctive constructions in
underlying representation, regardless of the specific disjunctive conjunction
used in the sentence. If my argument is correct, then no separate functional
discussion of contrastive wa is necessary.
Terakura (1984) has argued convincingly that X wa can be repeated
COMMUNICATIVE ORIENTATION AND WA 299

only when the flow of information is somehow disconnected.


(9) a. boku wa kyoo ginza de hanako to atta.
I TP today Ginza at with met
(hanako wa) suukagetsu awanai uchi-ni
TP several-months meet:NEG while
zuibun kireini natte ita
very pretty RS has-become (from Terakura 1984)
I met Hanako in Ginza today. She has become very pretty
during the several months I have not seen here.
b. boku wa kyoo ginza de hanako to atta.
TP today LC with meet PST
hanako wa piano ga hikeru rashii. ima made
TP DO can play seem now until
chittomo shiranakatta.
at all know NEG PST
I met Hanako in Ginza today. It appears Hanako can play
piano. I didn't know that until now.
In (10a), the first and the second sentences are naturally intercon­
nected, making a coherent semantic unit; in (9b) the two sentences are not
smoothly interconnected. Some new aspect of Hanako has been revealed to
the speaker during the conversation that follows the meeting in Ginza. If
we adopt our hypothesis that wa is a marker of listener-orientation, the dis-
coursal phenomena of repetition or omission of X wa can be automatically
accounted for. In (9a), it is unnecessary to repeat Hanako, because the
second sentence follows as a natural sequence from the first, and it is need­
less to draw the reader's attention to the fact that it was Hanako who had
become beautiful. For (9b), there is a time lapse between the meeting and
the discovery that Hanako seems to be able to play the piano. Thus, it is
considerate of the writer to draw the reader's attention to the fact that it
was Hanako who seems to be able to play the piano. The speech act of
drawing the reader's (or the listener's) attention is a speech act of reader/
listener-orientation.
Kuno (1973:52-53) noted that descriptive ga is limited to predicates
representing actions, existence, or temporary states, as shown by (10).
(10) a. oya yamada-san ga tenisu o shiteiru.
look SB tennis DO playing
Look, Mr. Yamada is playing tennis.
300 SEIICHI MAKINO

b. are, ashi ga tsutchatta!


Gee foot SB have-cramp-end-up:PST
Gee, I have a cramp in my foot!
c. tsukue no ue ni hon ga aru.
table LK top on book SB be
There is a book on the table.
Primarily through visual or auditory sense the speaker is focusing on an
action, existence, or temporary state of an animate/inanimate object which
is marked by ga. The description is essentially speaker-oriented, because it
depends totally on the speaker's direct perception. In fact, descriptive ga
can indicate an exclamatory expression of the speaker. Observe the follow­
ing examples:
(11) a. a! henna yatsu ga kita!
strange guy SB came
Look! A strange guy has come!
b. a! netsu ga aru!
fever SB be
Gee, you have a fever!
c. a! basu ga kita!
bus SB came
Look! Here comes a bus!
d. a! kizu ga tsuiteiru!
scratch SB attached
Look! There is a scratch!
Note that all the sentences in (11) lack the sentence-final particle yo 'I
tell you', the particle of persuasion which is highly listener-oriented.
Exhaustive listing ga has to be basically speaker-oriented, because a
wh-word + ga sequence is a grammatical sequence in contrast to the
ungrammatical sequence of *wh-word + wa. Consider the following short
discourse:
(12) A: tenisu no shiai wa dare ga kachimashita ka.
tennis LK match TP who SB won QU
Who won the tennis match?
B: yamada-san ga kachimashita.
SB won
Mr. yamada did.
COMMUNICATIVE ORIENTATION AND WA 301

A: soo desu ka. yamada-san wa mainichi renshuu


so COP QU TP every-day practice
shiteimasu kara ne.
doing because EX
Is that right? Mr. Yamada is practicing every day, you know.
A is asking for information as to who has won the tennis match. 
supplies the information. In' this discourse A happens to know the winner
and  may well be aware that A also personally knows the winner; in other
words Mr. Yamada has been registered in the mind of both A and B. But
because the fact that Mr. Yamada won has been registered only in the mind
of  and not of A, ga has to be used instead of wa. In the sense that the reg­
istration of Mr. Yamada as the sole winner is made only in B's mind, the ga
of B's sentence is speaker-oriented. In contrast to descriptive ga, exhaustive
listing ga normally has nothing to do with human sensory organs. To the
extent that exhaustive listing ga is unrelated to directly perceptible
phenomena it is less speaker-oriented than descriptive ga.
Kuno (1973: Chapter 4) has argued that ga in (13) marks the object of
a transitive verbal:
(13) a. boku ga besu ga daisuki da.
I TP Beth DO love COP
I love Beth.
b. watashi wa mizu ga nomitai.
I TP water DO want-to-drink
I want to drink water,
 yamada wa kankokugo ga wakaru.
TP Korean DO understand
Yamada understands Korean.
d. hanako wa furansugo ga hanaseru.
TP French DO can-speak
Hanako can speak French.
e. boku wa inu ga kowai.
I TP dog DO be-afraid-of
I am scared of dogs.
f. watashi wa kyoo kuruma ga iru.
I TP today car DO need
I need a car today.
302 SEIICHI MAKINO

As I have argued in Makino (1977) with regard to Japanese potentials,


ga for object marking has an underlying subject function. Be that as it may,
my hypothesis is that ga in (13) indicates something that spontaneously
comes into existence, primarily involving the speaker, as in (13a, b, e). It
should be noted that objective ga can appear in an exclamatory sentence —
another piece of evidence that objective ga is speaker-oriented.
(14) a. boku wa kimi ga suki da!
I TP you DO like COP
I like you!
b. aa, biiru ga nomitai!
oh beer DO want-to-drink
Gee! Do I want to drink beer!
c. aa, nyoobo no kao ga kowai!
wife GN face DO be-afraid-of
Oh, how I am scared of my wife's face!
d. a! fujisan ga mieta.
Mt. Fuji SB could-be-seen
Wow! I saw Mt. Fuji!
It has long been noted that thematic wa has to be changed into ga in a
subordinate clause. Thus:
(15) a. watashi gal*wa dekakeyoo-to-suru toki ni
I SB TP about-to-leave time at
kare ga yattekita.
he SB came
When I was about to leave, he came to see me.
b. watashi-tachi wa ochiba gal*wa ichimen ni
we ignore TP fallen-leaves SB/ TP all-over LC
chirishiiteiru michi  aruiteita.
scattered path LC walking
We were walking on the path on which were scatttered fallen
leaves all over.
c. anata ga/*wa konakatta no ga ikenakatta.
you SB TP didn't-come NOM SB was-wrong
It was wrong that you didn't show up.
d. umi gal*wa ichiban dayaka na tsuki wa
ocean SB/TP most calm LK month TP
COMMUNICATIVE ORIENTATION AND WA 303

shichigatsu desu.
July COP
The calmest month for the ocean is July.
Independently of wa-into-ga conversion in subordinate clauses, the
subordinate clause seems to be a communicative vehicle for speaker-
oriented information. There are at least two pieces of evidence for my
claim. First, the predicate of a subordinate clause is seldom marked as for­
mal. My assumption here is that formality marking results from listener-
orientation. The informal-into-formai switching of the subordinate pre­
dicate in (16) yields unacceptable sentences:
(16) a. yamada-san ga iku /???ikimasu noni
SB go (informal) / go (formal) even-if
anata wa ikanai n desu ka.
you TP don't-go NOM COP QU
Aren't you going there even if Mr. Yamada is going there?
b. hitori-de kurashikku o kiiteiru
alone classical-music DO listening (informal)
l*kiiteimasu toki ga ichiban tanoshii desu.
listening (formal) time SB most happy COP
It is the happiest time when I am listening to classical music
all by myself.
 watashi wa kanojo ga kurannetto 
I TP she SB clarinet DO
fuiteiru /*fuitemasu no o kikimashita.
playing (informal)/playing (formal) NOM DO listened
I listened to her playing the clarinet.
d. mada nihon e itta /*ikimashita
not yet Japan to went (informal)/went (formal)
koto ga arimasen.
NOM SB isn't
I have never been to Japan.
Secondly, most listener-oriented sentence-final particles such as yo 'I
tell you', ne 'a tag question', cannot be used in a subordinate predicate. In
short, wa cannot be used inside a highly subordinate clause, exactly as pre­
dicted by our principle of Orientational Conflict.
So far I have reexamined wa and ga in Japanese in terms of the func-
304 SEIICHI MAKINO

tional notion of CO. It seems to be the case that our functional notion can
account for uses of wa and ga at least as well as the old/new informational
principle. Our notion, however, has a wider range of application than the
functional notion adopted to explain uses of wa and ga.
Let me give some areas of application which I have previously discus­
sed.
1. Formality-into-informality in a discourse is conditioned by speaker-
orientation. (Makino 1983)
2. The nominalizers no and koto are speaker-oriented and listener-
oriented nominalizers, respectively, because no represents directly
perceptible phenomena. (Makino 1983:138)
3. The non-restrictive and restrictive relative clauses are speaker-
oriented and listener-oriented relative clauses, respectively.
(Makino 1983:134-136)
4. The conjunctions node 'because' and noni 'although' are speaker-
oriented conjunctions, but the corresponding kara 'because' and
keredo 'although', are listener-oriented conjunctions. (Makino
1983:130-133).
5. Phenomena of repetition are accountable in terms of speaker/lis­
tener-orientation. Some of the principles discussed in Makino
(1984) are as follows:
A: Principle of Pecking Order of Repetitions
Repeat more important information first, and less important
information last. But the pecking order may be violated if the
speaker of a triggering sentence is not highly listener-oriented.
B: Ban on Repetition of Highly Speaker-Oriented Information
A highly speaker-oriented piece of information should not be
repeated by itself, unless empathy is at work.
C: Principle of Informational Importance
Repeat only communicatively important pieces of information
in a sentence within a discourse. By a communicatively impor­
tant piece of information is understood a highly listener-
oriented information.
D: Principle of Extended Repetition of a Wh-question
An extended repetition of a highly listener-oriented Wh-ques­
tion is possible only when the repeater wants to deny the speech
act of Wh-questioning for some emotive reason.
C O M M U N I C A T I V E O R I E N T A T I O N A N D WA 305

E: Principle of Extended Repetition of Highly Listener-Oriented


Sentence
It is possible to repeat such highly listener-oriented triggering
sentences as request sentences or question sentences in their
entirety if the speaker intends to deny the very speech act of
requesting or questioning.
6. Exclamatory sentences can be characterized as speaker oriented.
(Makino 1983:140)
7. The structure of a monologue is accountable by the speaker-orien­
tation.
8. Turn-taking does not occur at the juncture where the information is
speaker-oriented.
9. Indirect quotes are normally speaker-oriented but direct quotes are
normally listener-oriented.
In conclusion, the functional notion of communicative orientation has
more generality than the functional notion of given/new information, and
therefore, the former has more explanatory adequacy than the latter.

NOTES
The broad outline of the idea included in this paper was first presented at the meeting of the
Association for Teachers of Japanese at Washington D.C., when I served as the discussant for
the linguistic session called "Functions of -WA in Japanese Discourse" on March 25, 1984. I am
grateful to comments I received from the panelists Professors John Hinds, Senko Maynard, and
Shoichi Iwasaki and the audience. But the errors included in this paper are all mine.
1) The statements (d) and (f) are worded differently from Kuno's original statements.
2) 'Listener-oriented' is used here and elsewhere as a cover term for both 'listener-oriented'
and 'reader-oriented'.
3) 'Speaker-oriented' is used here and elsewhere as a cover term for both 'speaker-oriented'
and 'writer-oriented'.

REFERENCES

Kuno, Susumu. 1972. "Functional sentence perspective: a case study from


Japanese and English", Linguistic inquiry, 3:3, 269-320.
. 1973. The structure of the Japanese language, MIT Press.
. 1884. Wa and Ga in Japanese. (A lecture given at the Japanese
School, Middlebury College, Middlebury, Vermont. July 4, 1984.)
306 SEIICHI MAKINO

Makino, Seiichi. 1977. "On the nature of the Japanese potential construc­
tions." Papers in Japanese linguistics. 4:97-124.
1982. "Japanese grammar and functional grammar." Lingua.
57:125-173.
. 1983. "Speaker/listener-orientation and formality marking in
Japanese." Gengo uu. 84:126-145.
. 1984. "Some principles of repetition in Japanese spoken discourse." A
paper presented at the Nitobe-Ohira memorial conference on Japanese
studies, May 23-25, 1984 at the University of British Columbia, Van­
couver, Canada. (To appear in the Proceedings)
Terakura, Hiroko. 1984. "The deletion of the theme of the sentence and
discourse structure." A paper presented at the Nitobe-Ohira memorial
conference on Japanese studies, May 23-25, 1984 at the University of
British Columbia, Vancouver, Canada. (To appear in the Proceedings)
ABBREVIATIONS

ATTRI Attributive
CAU Causative
CL Classifier
COP Copula
DO Direct object
EX Exclamatory particles
GER Gerund
EMP Emphatic
GN Genitive
HON Honorific
HUM Humble
HYP Hypothetical
IMPE Imperative
IN Instrumental marker
 Indirect object
LC Locative marker
LK Linker
NEG Negative
NOM Nominalizer
PASS Passive
PERF Perfect
PL Plural
PST Past/perfect tense
QT Quotative marker
QU Question marker
RS Resultative
SB Subject marker
TENT Tentative
TP Topic marker
VOL Volition

Potrebbero piacerti anche