Japanese Discourse and the Process of 
Centering 
Marilyn Walker* 
University of Pennsylvania 
Sharon Cotes 
University of Pennsylvania 
Masayo Iida t 
Stanford University 
This paper has three aims: (1) to generalize a computational account of the discourse process 
called CENTERING, (2) to apply this account to discourse processing in Japanese so that it can be 
used in computational systems for machine translation or language understanding, and (3) to 
provide some insights on the effect of syntactic factors in Japanese on discourse interpretation. We 
argue that while discourse interpretation is an inferential process, syntactic cues constrain this 
process; we demonstrate this argument with respect to the interpretation of ZEROS, unexpressed 
arguments of the verb, in Japanese. The syntactic cues in Japanese discourse that we investigate 
are the morphological markers for grammatical TOPIC, the postposition wa, as well as those for 
grammatical functions such as SUBJECT, ga, OBJECT, o and OBJECT2, ni. In addition, we investigate 
the role of speaker's EMPATHY, which is the viewpoint from which an event is described. This is 
syntactically indicated through the use of verbal compounding, i.e. the auxiliary use of verbs such 
as kureta, kita. Our results are based on a survey of native speakers of their interpretation of 
short discourses, consisting of minimal pairs, varied by one of the above factors. We demonstrate 
that these syntactic cues do indeed affect the interpretation of ZEROS, but that having previously 
been the TOPIC and being realized as a ZERO also contributes to the salience of a discourse entity. 
We propose a discourse rule of ZERO TOPIC ASSIGNMENT, and show that CENTERING provides 
constraints on when a ZERO can be interpreted as the ZERO TOPIC. 
1. Introduction 
1.1 Centering in Japanese Discourse 
Recently there has been an increasing amount of work in computational linguistics 
involving the interpretation of anaphoric elements in Japanese (Yoshimoto 1988; Kuno 
1989; Walker, Iida, and Cote 1990; Nakagawa 1992). These accounts are intended as 
components of computational systems for machine translation between Japanese and 
English or for natural language processing in Japanese alone. This paper has three 
aims: (1) to generalize a computational account of the discourse process called CEN- 
TERING (Sidner 1979; Joshi and Weinstein 1981; Grosz, Joshi, and Weinstein 1983; Grosz, 
Joshi, and Weinstein unpublished), (2) to apply this account to discourse processing 
* University of Pennsylvania, Computer Science Department, Philadelphia PA 19104. E-mail: 
lyn@linc.cis.upenn.edu. 
t CSLI, Stanford University, Stanford CA 94305. E-maih iida@csli.stanford.edu. 
:~ University of Pennsylvania, Linguistics Department, Philadelphia PA 19104. E-mail: 
cote@linc.cis.upenn.edu. 
@ 1994 Association for Computational Linguistics 
Computational Linguistics Volume 20, Number 2 
in Japanese so that it can be used in computational systems, and (3) to provide some 
insights on the effect of syntactic factors in Japanese on discourse interpretation. 
In the computational literature, there are two foci for research on the interpretation 
of anaphoric elements such as pronouns. The first viewpoint focuses on an inferential 
process driven by the underlying semantics and relations in the domain (Hobbs 1985a; 
Hobbs et al. 1987; Hobbs and Martin 1987). A polar focus is to concentrate on the role 
of syntactic information such as what was previously the topic or subject (Hobbs 
1976b; Kameyama 1985; Yoshimoto 1988). We will argue for an intermediate position 
with respect to the interpretation of ZEROS, unexpressed arguments of the verb, in 
Japanese. Our position is that the interpretation of zeros is an inferential process, but 
that syntactic information provides constraints on this inferential process (Joshi and 
Kuhn 1979; Joshi and Weinstein 1981). We will argue that syntactic cues and semantic 
interpretation are mutually constraining (Prince 1981b, 1985; Hudson-D'Zmura 1988). 
The syntactic cues in Japanese discourse that we investigate are the morphological 
markers for grammatical TOPIC, the postposition wa, as well as those for grammatical 
functions such as SUBJECT, ga, OBJECT, o, and OBJECT2, ni. In addition, we investigate the 
role of speaker's EMPATHY, which is the viewpoint from which an event is described. 
This can be syntactically indicated through the use of verbal compounding, i.e. the 
auxiliary use of verbs such as kureta, kita. 
In addition to the argument that a purely inference-based account does not con- 
sider limits on processing time, another argument against a purely inference-based 
account is provided by the minimal pair below. Here, the only difference is whether 
Ziroo is the subject or the object in the second utterance. Note that the interpretation 
of zeros is indicated in parentheses: 
Example 1 
a. Taroo ga 
b. 
C. 
kooen o sanpositeimasita. 
Taroo SUBJ park in walking-was 
Taroo was taking a walk in the park. 
Ziroo ga 0 hunsui no mae de mitukemasita. 
Ziroo SUBJ OBJ fountain of front in found 
Ziroo found (Taroo) in front of the fountain. 
0 0 kinoo no siai no kekka o kikimasita. 
SUBJ OBJ yesterday of game of scores OBJ asked 
(Ziroo) asked (Taroo) the score of yesterday's game. 
Example 2 
a. Taroo ga 
b. 
C. 
kooen o sanpositeimasita. 
Taroo SUBJ park in walking-was 
Taroo was taking a walk in the park. 
0 Ziroo o hunsui no mae de mitukemasita. 
SUBJ Ziroo OBJ fountain of front in found 
(Taroo) found Ziroo in front of the fountain. 
0 0 kinoo no siai no kekka o kikimasita. 
SUBJ OBJ yesterday of game of scores obj asked 
(Taroo) asked (Ziroo) the score of yesterday's game. 
In lb and 2b, the syntactic position in which Ziroo is realized has the effect that lc 
194 
Marilyn Walker et al. Japanese Discourse 
means Ziroo asked Taroo the score of yesterday's game, while 2c means Taroo asked Ziroo the 
score of yesterday's game. On the other hand, some purely syntactic accounts require that 
antecedents for zeros be realized as the grammatical TOPIC, and thus cannot explain 
the above example because Taroo is never explicitly marked as the topic (Yoshimoto 
1988). 
In the literature, ZEROS are known as zero pronouns. We adopt the assumption of 
earlier work that the interpretation of zeros in Japanese is analogous to the interpre- 
tation of overt pronouns in other languages (Kuroda 1965; Martin 1976; Kameyama 
1985). Japanese also has overt pronouns, but the use of the overt pronoun is rare in nor- 
mal speech, and is limited even in written text. This is mainly because overt pronouns 
like kare ('he') and kanozyo ('she') were introduced into Japanese in order to translate 
gender-insistent pronouns in foreign languages (Martin 1976). In this paper, we only 
consider zeros in subcategorized-for argument positions. Since Japanese doesn't have 
subject or object verb agreement, there is no syntactic indication that a zero is present 
in an utterance other than information from subcategorization) 
First, in Section 1.2 we describe the methodology that we applied in this investi- 
gation. In Section 2, we present the theory of centering and some illustrative exam- 
ples. Then, in Section 3, we discuss particular aspects of Japanese discourse context, 
namely grammatical TOPIC and speaker's EMPATHY. We will show how these can easily 
be incorporated into a centering account of Japanese discourse processing, and give 
a number of examples to illustrate the predictions of the theory. We also discuss the 
way in which a discourse center is instantiated in Section 4. 
In Section 5 we propose a discourse rule of ZERO TOPIC ASSIGNMENT, and use the 
centering model to formalize constraints on when a zero may be interpreted as a ZERO 
TOPIC. Our account makes a distinction between two notions of TOPIC--grammatical 
topic and zero topic. The grammatical topic is the wa-marked entity, which is by default 
predicted to be the most salient discourse entity in the following discourse. However, 
there are cases in which it may not be, depending on whether ZERO TOPIC ASSIGNMENT 
applies. This analysis provides support for Shibatani's claim that the interpretation of 
the topic marker, wa, depends on the discourse context (Shibatani 1990). ZERO TOPIC 
ASSIGNMENT actually predicts ambiguities in Japanese discourse interpretation and 
provides a mechanism for deriving interpretations that previous accounts claim would 
be unavailable. 
We delay the review of related research to Section 6 when we can contrast it with 
our account. The two major previous accounts are those of Kuno (Kuno 1972, 1976b, 
1987, 1989) and Kameyama (Kameyama 1985, 1986, 1988). Finally, in Section 7, we 
summarize our results and suggest topics for future research. 
1.2 Methodology 
Most of the examples in this paper are constructed as four utterance discourses that fit 
one of a number of structural paradigms. In all of the paradigms, a discourse entity is 
1 When zero pronouns should be stipulated is still a research issue. For example, Hasegawa (1984) 
described a zero pronoun as a phonetically null element in an argument position. However, as shown 
in the following example, Terazu, Yamanasi, and Inada (1980) assumed that zero pronouns are not 
limited in their distribution and stipulated them in adjunct positions as well (Iida 1993). 
Taroo wa Hanako no kaban o mitukemasita. 
Taroo TOP/SUBJ Hanako GEN bag OBJ found 
Taroo found Hanako" s bag. 
0 0 tanzyoobi no purezento o irernasita. 
birthday GEN present OBJ put 
(Taroo) put a birthday present (in her bag). 
195 
Computational Linguistics Volume 20, Number 2 
introduced in the first utterance, and established by the second utterance as the CENTER, 
what the discourse is about. The manipulations of context occur with the third and 
the fourth utterances. In each case the zero in the third utterance cospecifies the entity 
already established as the center in the second utterance. The fourth utterance consists 
of a potentially ambiguous sentence containing two zeros. The variations in context 
are as shown below: 
Third Utterance Fourth Utterance 
SUBJECT OBJECT(2) SUBJECT OBJECT(2) EXAMPLES 
zero NP(o or ni) zero zero 5 
zero NP(o or ni) zero zero, empathy 36 
NP(ga) zero zero zero 32, 34 
NP(wa) zero zero zero 4, 33 
NP(ga) zero zero zero, empathy 35 
Thus we are manipulating factors such as whether a discourse entity is realized 
in subject or object position in the third utterance, whether a discourse entity realized 
in subject position is ga-marked or wa-marked in the third utterance, and whether a 
discourse entity realized in the fourth utterance in object position is marked as the 
locus of speaker's EMPATHY. 
We collected a group of about 35 native speakers by solicitation on the InterNet to 
provide judgments for most of the examples given in this paper. These native speakers 
were readers of the newsgroups sci.lang.japanese and comp.research.japan. They were 
thus typically well-educated, bilingual engineers. Whenever an example was tested in 
this way, we provide the number of informants who chose each possible interpretation 
to the right of the example. Some examples that are included for expository reasons 
were not tested. 
Participation in our survey was completely voluntary, and the data were collected 
over three surveys. Thus the numbers of subjects varied from one survey to another, 
and this is reflected in the numbers accompanying our examples. This data collection 
was carried out on written examples using electronic mail in a situation in which 
the informants could take as long as they wanted to decide which interpretation they 
preferred. The instructions sent with the surveys are given in Appendix A. 
This paradigm clearly cannot provide information on which interpretation a sub- 
ject might arrive at first and then perhaps change based on other pragmatic factors, and 
thus it contrasts with reaction time studies. However, the judgments given should be 
stable and should reflect the fact that our informants were able to use all the informa- 
tion in the discourse. It is a useful paradigm given that we are exploring the correlation 
of syntactic cues and discourse interpretation. It has been claimed that syntactic cues 
are only used in automatic processing and can be overridden by deeper processing. 
However, Hudson's results suggest that subjects may judge a discourse sequence to 
be nonsensical when it is incoherent according to centering (Hudson-D'Zmura 1988). 
Di Eugenio claims that discourse sequences in Italian that are not discourse-coher- 
ent according to centering theory produce a garden-path effect (Di Eugenio 1990). 
The methods we used allow us to explore the results of these interactions, and yet it 
196 
Marilyn Walker et al. Japanese Discourse 
would be beneficial for these results to be expanded upon by careful psychological 
experimentation. 
For most of the examples reported here, we asked subjects to choose one preferred 
interpretation instead of allowing them to rank interpretations. The motivation for 
doing this was to force differences to come out for slight preferences, with the theory 
being that other variations would come out across subjects. In a few cases we allowed 
subjects to indicate no preference; these examples will be clearly indicated. 
In addition, we used the same gender for multiple discourse entities to prevent any 
tendency for judgments to be influenced by gender stereotypes. We also avoided using 
verbs with causal biases toward one of their arguments, and we used few cue words 
such as but, because, and then, which could result in a bias toward, say, a cause-effect or 
temporal sequence of events interpretation. We also omitted honorific markers, which 
are normally a part of Japanese ambiguity resolution. 2 This was done to isolate the 
effects of the variables that we were exploring in this study, namely topic marking, 
grammatical function, empathy, and realization with a zero or with a full noun phrase. 
2. Centering Theory 
Within a theory of discourse, CENTERING is a computational model of the process by 
which conversants coordinate attention in discourse (Grosz, Joshi, and Weinstein un- 
published). Centering has its computational foundations in the work of Grosz and 
Sidner (Grosz 1977; Sidner 1979; Grosz and Sidner 1986) and was further developed 
by Grosz, Joshi, and Weinstein (1983, unpublished) and Joshi and Weinstein (1981). 
Centering is intended to reflect aspects of ATYENTIONAL STATE in a tripartite view of 
discourse structure that also includes INTENTIONAL STRUCTURE and LINGUISTIC STRUC- 
TURE (Grosz and Sidner 1986). In Grosz and Sidner's theory of discourse structure, 
discourses can be segmented based on intentional structure, and a discourse segment 
exhibits both local and global coherence. Global coherence depends on how each seg- 
ment relates to the overall purpose of the discourse; local coherence depends on as- 
pects such as the syntactic structure of the utterances in that segment, the choice of 
referring expressions, and the use of ellipses. CENTERING models local coherence and 
is formalized as a system of constraints and rules. Our analysis uses an adaptation 
of the Centering algorithm that was developed by Brennan, Friedman, and Pollard, 
based on these constraints and rules (Brennan, Friedman, and Pollard 1987; Walker 
1989). 
The purpose of centering as part of a computational model of discourse interpre- 
tation is to model ATTENTIONAL STATE in discourse in order to control inference (Joshi 
and Kuhn 1979; Joshi and Weinstein 1981). 3 Our approach to modeling attentional 
state is to explore aspects of the correlation between syntax and discourse function. 
This assumes that there are language conventions about discourse salience and that 
conversants attempt to maintain a sense of shared context. 
2 While native speakers understandably found some of these examples "stilted" or "awkward," they 
were still able to give their judgments based on the information that was provided in the discourses. 3 Recent work in situation theory proposes to control computation with a similar notion of background 
information in terms of constants of the situation that thus are not explicitly realized in an utterance 
(Nakashima 1990). The situation-theoretic work does not as yet distinguish shared knowledge that 
determines discourse salience and derives from the discourse context and the way utterances are 
expressed (Clark and Haviland 1977; Clark and Marshall 1981; Prince 1981b) from shared knowledge 
that is part of general background knowledge such as cultural assumptions (Prince 1978a; Joshi 1982) 
or shared knowledge that might derive from the task context (Grosz 1977). 
197 
Computational Linguistics Volume 20, Number 2 
Section 2.1 presents the centering rules and constraints. Sections 2.2 and 2.3 illus- 
trate the theory and the definitions with a number of examples. Section 2.4 discusses 
the centering algorithm for the resolution of zeros in Japanese. 
2.1 Rules and Constraints 
The centering model is very simple. Each utterance in a discourse segment has two 
structures associated with it. First, each utterance in a discourse has associated with it 
a set of discourse entities called FORWARD-LOOKING CENTERS, Cf. Centers are semantic 
entities that are part of the discourse model. Second, there is a special member of 
this set called the BACKWARD-LOOKING CENTER, Cb. The Cb is the discourse entity that 
the utterance most centrally concerns, what has been elsewhere called the 'theme' 
(Reinhart 1981; Horn 1986). The Cb entity links the current utterance to the previous 
discourse. 
The set of FORWARD-LOOKING CENTERS, Cf, is ranked according to discourse salience. 
We will discuss factors that determine the ranking below. The highest-ranked member 
of the set of forward-looking centers is referred to as the PREFERRED CENTER, Cp. 4 The 
PREFERRED CENTER represents a prediction about the Cb of the following utterance. 
Sometimes the Cp will be what the previous segment of discourse was about, the Cb, 
but this is not necessarily the case. This distinction between looking back to the previ- 
ous discourse with the Cb and projecting preferences for interpretation in subsequent 
discourse with the Cp is a key aspect of centering theory. 
In addition to the structures for centers, Cb and Cf, the theory of centering specifies 
a set of rules and constraints. Constraints are meant to hold strictly whereas rules may 
sometimes be violated. 
• CONSTRAINTS 
For each utterance Ui in a discourse segment UI~... ~ Um: 
. 
2. 
. 
There is precisely one backward-looking center Cb. 
Every element of the forward centers list, Cf(Ui), must be 
realized in Ui. 
The center, Cb(Ui), is the highest-ranked element of Cf(Ui-D 
that is realized in Ui. 5 
Constraint (1) says that there is one central discourse entity that the utterance is 
about, and that is the Cb. The second constraint depends on the definition of realizes. 
An utterance U realizes a center c if c is an element of the situation described by U, 
or c is the semantic interpretation of some subpart of U (Grosz, Joshi, and Weinstein 
unpublished). Thus the relation REALIZE describes zeros, explicitly realized discourse 
entities, and those implicitly realized centers that are entities inferable from the dis- 
course situation (Prince 1978a, 1981b). 
A specialization of the relation REALIZE is the relation DIRECTLY REALIZE. A center is 
directly realized if it corresponds to a phrase in an utterance. We restrict our focus to 
entities realized by noun phrases; however, it is clear that propositions can be centers, 
so we assume that the account given here can be extended to propositional entities as 
well (Webber 1978; Sidner 1979; Prince 1986, 1978b; Ward 1985). 
4 The notion of PREFERRED CENTER corresponds to Sidner's notion of EXPECTED FOCUS (Sidner 1983). 
5 This could possibly be rephrased as: Assume the Cp(Ui-1 is the Cb(Ui) unless there is evidence to the contrary (Carter 1987). 
198 
Marilyn Walker et al. Japanese Discourse 
As we discuss further in Section 3, zeros refer to entities that are already in the 
discourse context. The fact that the current utterance REALIZES one or more zeros 
follows from information specified in the subcategorization frame of the verb. These 
arguments must be interpreted and thus acquire a degree of discourse salience that 
nonsubcategorized-for discourse entities lack. 
Constraint (3) stipulates that the ranking of the forward centers, Cf, determines 
from among the elements that are realized in the next utterance, which of them will 
be the Cb for that utterance. If the PREFERRED CENTER, Cp(Ui), is realized in Ui+l, it 
is predicted to be the Cb(Ui+l). We will use the following forward center ranking for 
Japanese: 6 
(GRAMMATICAL OR ZERO) TOPIC > EMPATHY > SUBJECT > OBJECT2 > OBJECT > OTHERS 
Backward-looking centers, Cbs, are often deleted or pronominalized and some 
transitions between discourse segments are more coherent than others. According to 
the theory of centering, coherence is measured by the hearer's inference load when 
interpreting a discourse sequence (Joshi and Weinstein 1981; Grosz, Joshi, and Wein- 
stein unpublished). For instance, discourse segments that continue centering the same 
entity are more coherent than those that repeatedly shift from one center to another. 
These observations are encapsulated in two rules: 
. 
2. 
RULES 
For each Ui in a discourse segment U1,..., Urn: 
If some element of Cf(Ui-l) is realized as a pronoun in Ui, then 
so is Cb(Ui). 
Transition states are ordered. CONTINUE is preferred to RETAIN is 
preferred to SMOOTH-SHIFT is preferred to ROUGH-SHIFT. 7 
Rule (1) captures the intuition that pronominalization is one way to indicate dis- 
course salience. It follows from Rule (1) that if there are multiple pronouns in an 
utterance, one of these must be the Cb. In addition, if there is only one pronoun, 
then that pronoun must be the Cb. For Japanese, we extend this rule directly to zeros, 
assuming that zeros in Japanese correspond to destressed pronouns in English. 
Rule (2) states that modeling attentional state depends on analyzing adjacent ut- 
terances according to a set of transitions that measure the coherence of the discourse 
segment in which the utterance occurs. Measuring coherence is based on an estimate 
of the hearer's inference load, but this measure must always be relative since there is 
no grammar of discourse. Thus methods for exploring these issues must use compar- 
ative measures of how some discourses are easier to process than others. Centering 
theory models this by stipulating that some transitions are preferred over others. 
The typology of transitions from one utterance, Ui, to the next is based on two 
factors: whether the backward-looking center, Cb, is the same from Ui-1 to Ui, and 
whether this discourse entity is the same as the preferred center, Cp, of Ui. 8 
6 This ranking is consistent with Kuno's Empathy Hierarchies and with Kameyama's Expected Center 
Order (Kuno 1987; Kameyama 1985, 1988). This will be discussed in Section 6. We do not include 
discourse entities for verb phrases or other propositional entities in this ranking since we have not 
studied their contribution (but see Sidner 1979, 1981 and Carter 1987). 
7 Smooth-shift was called shifting-1 by Brennan, Friedman, and Pollard (1987). 8 It is possible that restricting the relation between the Cb(Ui) and the 
Cb(Ui_l) to be coreference 
(equality) may be too strong. Future work should examine the role of shifts to functionally dependent 
entities or entities related by partially ordered set (POSET) relations to the previous Cb. 
199 
Computational Linguistics Volume 20, Number 2 
Cb(Ui) = Cb(Ui-1) 
OR Cb(Ui-1) = \[?1 
Cb(Ui) # Cb(Ui-1) 
Cb(Ui) = Cp(Ui) CONTINUE SMOOTH-SHIFT 
Cb(Ui)   Cp(Ui) RETAIN ROUGH-SHIFT 
Figure 1 
Centering transition states, rule 2. 
KEY 
BACKWARD-LOOKING CENTER = Cb 
PREFERRED CENTER = Cp 
Uninstantiated Cb = \[?\] 
. 
2. 
Cb(Ui) = Cb(Ui_l), or there is no Cb(Ui-1) 
Cb(Ui) = Cp(Ui) 
If both (1) and (2) hold then we are in a CONTINUE transition. The CONTINUE 
transition corresponds to cases where the speaker has been talking about a particular 
entity and indicates an intention to continue talking about that entity. 9 If (1) holds but 
(2) doesn't hold then we are in a RETAIN transition. RETAIN corresponds to a situation 
where the speaker is intending to SHIFT onto a new entity in the next utterance and 
is signalling this by realizing the current center in a lower ranked position on the Cf 
(examples follow below). 
If (1) doesn't hold then we are in one of the SHIFT states depending on whether or 
not (2) holds. This definition of transition states is summarized in Figure 1 (Brennan, 
Friedman, and Pollard 1987). We will use the notation of Cb(Ui-1) = \[?\] for cases 
where there is no Cb(Ui-1). Section 4 will discuss center instantiation. 
The combination of the constraints, rules, and transition states makes a set of 
testable predictions about which interpretations hearers will prefer because they re- 
quire less processing. For example, maximally coherent segments are those that require 
less processing time. A sequence of a CONTINUE followed by another CONTINUE should 
only require the hearer to keep track of one main discourse entity, which is currently 
both the Cb and the Cp. A single pronoun in an utterance is the current Cb (by Rule 
1) and can be interpreted to cospecify the discourse entity realized by Cp(Ui-1) in one 
step (Constraint 3). 
The ordering of the Cf is the main determinant of which transition state holds 
between adjacent utterances. This means that the predictions of the theory are largely 
determined by the ranking of the items on the Cf. But there are many factors that can 
contribute to the salience of a discourse entity; among them are factors that we will 
not examine here such as lexical semantics, intonation, word-order, and tense. 1° In this 
9 A prediction made by the preference for CONTINUE is that intersentential antecedents for pronouns will 
be preferred over intrasentential candidates. This preference is one that distinguishes Centering for 
pronoun interpretation from the proposal made by Hobbs (1976a, 1976b). However, this preference 
needs to be constrained further by the fact that sortal filters may rule out the Cp of the previous 
utterance as the current Cb. In this case the data suggest that perhaps intrasentential candidates should 
be preferred (Walker 1989). Carter explored this in his extension of Sidner's theory of local focusing 
(Carter 1987). 10 See Hudson-D'Zmura (1988) for an examination of the role of lexical semantics in centering. 
200 
Marilyn Walker et al. Japanese Discourse 
paper we explore the influence of various syntactic factors, which we discuss in detail 
in Section 3. We will also examine the relative contribution of pronominalization and 
postposition marking in Section 5. We postulate that the Cf ordering will vary from 
language to language depending on the means the language provides for expressing 
discourse function. However much of this variation can be captured in the ranking of 
the Cf due to the modularity of the theory. 
In Sections 2.2 and 2.3 we will present some simple examples to motivate these 
definitions. In Section 2.4 we will present a slightly modified version of the centering 
algorithm (Brennan, Friedman, and Pollard 1987). In the following discussion we as- 
sume that the centering rules and constraints and the notion of centering transition 
states have some cognitive reality (Brennan submitted; Hudson-D'Zmura 1988; Gor- 
don, Grosz, and Gilliom 1993; Hudson-D'Zmura and Tanenhaus, 1995). However, we 
make no claims about the cognitive reality of the centering algorithm that we discuss 
in Section 2.4. 
2.2 The Distinction between Continue and Retain 
This theory predicts preferences in the interpretation of utterances whose meaning 
depends on parameters from the discourse context. Thus if there are still multiple 
possibilities for interpretation after the application of all constraints and rules, the 
ordering on transitions applies, and CONTINUE interpretations are preferred (Rule 2). 
Indeed, many cases of the preference for one interpretation over another follow directly 
from the distinction between the transition states of CONTINUE and RETAIN. Let us look 
at a simple example. In the discourse segment in 3, the zero in the second sentence is 
understood as referring to Taroo, and not to Hanako. Remember that the interpretation 
of zeros is indicated with parentheses. 
Example 3 
a. Taroo wa 
b. 
Hanako o eiga ni sasoimasita. 
Taroo TOP/SUBJ Hanako OBJ movie to invited 
Taroo invited Hanako to the movie. 
Cb: TAROO 
Cf: \[TAROO, HANAKO\] 
0 itiniti-zyuu nani mo te ni tukimasendesita. 
SUBJ all-day anything even hand to attached-not 
(Taroo) could not do anything all day. 
Cb: TAROO 
Cf: \[TAROO\] 
In example 3, the Cf from 3a contains the discourse entity for Taroo as the first 
element and for Hanako as the second element. When the unexpressed argument is 
interpreted in 3b, the information from this Cf is used. Because the zero subject may 
REALIZE either Taroo or Hanako, both Constraint 3 and Rule 1 would be obeyed with 
either interpretation. 11 However by interpreting the zero as Taroo, Taroo is the Cb, and 
it is possible to get a preferred CONTINUE interpretation Taroo could not do anything all 
day. In this interpretation, Taroo is both the Cb(3b) and the Cp(3b). 
11 The hypothesis that wa in 3a instantiates Taroo as the Cb will be discussed in Section 4. 
201 
Computational Linguistics Volume 20, Number 2 
2.3 The Distinction between Smooth-Shift and Rough-Shift 
In example 4, we illustrate the difference between the transition states of ROUGH- 
SHIFT and SMOOTH-SHIFT. Remember that ROUGH-SHIFT is claimed to be less coherent 
than SMOOTH-SHIFT (Brennan, Friedman, and Pollard 1987). In both cases the speaker 
has shifted the center to a different discourse entity. However, in the SMOOTH-SHIFT 
transition state, the speaker has indicated an intention to continue talking about the 
recently shifted-to entity by realizing that entity in a highly ranked Cf position such 
as subject, whereas no such indication is available with the ROUGH-SHIFT transition. 
The numbers shown to the right of an interpretation correspond to how many native 
speakers preferred that interpretation. 
Example 4 
a. Taroo ga kooen de hon o yondeimasita. 
Taroo SUBJ park at book OBJ reading-was 
Taroo was reading a book in the park. 
Cb: \[?\] 
Cfl: \[TAROO, BOOK\] 
SUBI OBJ 
b. 0 koora o kai ni baiten ni hairimasita. 
SUBJ cola OBJ buy to shop into entered 
(Taroo) entered a shop to buy a cola. 
Cb: TAROO 
Cfl: \[TAROO, COLA\] CONTINUE 
SUBJ OBJ 
c. Ziroo wa 0 sokode guuzen dekuwasimasita. 
Ziroo TOP/SUBJ OBJ there by chance met 
Ziroo met (Taroo) there by chance. 
Cb: TAROO 
Cf: \[ZIROO, TAROO\] RETAIN 
TOP OBJ 
d. 0 0 eiga ni sasoimasita. 
SUBJ OBJ movie to invited. 
(Ziroo) invited (Taroo) to a movie. 
Cb: ZIROO 
Cfl: \[ZIROO, TAROO\] SMOOTH-SHIFT 32 
subj obj 
Cf2: \[TAROO, ZIROO\] ROUGH-SHIFT 2 
SUBJ OBJ 
In example 4, the use of TOPIC marking in the phrase Ziroo wa of utterance (c) 
means that (c) is interpreted as a RETAINJ 2 Ziroo becomes the most highly ranked 
discourse entity for c, although Taroo is the Cb since Taroo was most highly ranked 
for utterance (b) (by Constraint 3). Then when we apply the Centering algorithm in (d), 
there are two candidates for the Cb(d) from the Cf(c), both Ziroo and Taroo. However, 
this time when constraint 3 applies, stipulating that the Cb must be the highest-ranked 
12 It has also been claimed that symmetric verbs such as meet by chance mark EMPATHY on the subject 
(Kuno 1976a). 
202 
Marilyn Walker et al. Japanese Discourse 
element of Cf(c) realized in 4d, Ziroo must be the highest-ranked entity realized, and 
therefore must be the Cb. At this point it is clear that some kind of SHIFT is forced by 
the application of constraint 3. The two candidates are a SMOOTH-SHIFT and a ROUGH- 
SHIFT. The SMOOTH-SHIFT interpretation corresponds to the reading Ziroo invited Taroo 
to a movie whereas the ROUGH-SHIFT interpretation corresponds to the Taroo invited Ziroo 
reading. The SMOOTH-SHIFT interpretation is more highly ranked, thus considered more 
coherent and so is the preferred interpretation (Z = 10.93, p < .001). 
2.4 The Centering Algorithm 
The CENTERING ALGORITHM that was proposed by Brennan, Friedman, and Pollard 
incorporates the centering rules and constraints in addition to contra-indexing con- 
straints on coreference (Reinhart 1976; Brennan, Friedman, and Pollard 1987; Iida 1993). 
These contra-indexing constraints specify that in a sentence such as He likes him, that he 
and him cannot co-specify the same discourse entity. The algorithm applies centering 
theory to the problem of resolving anaphoric reference. Application of the algorithm 
requires three basic steps. 
. 
2. 
. 
GENERATE possible Cb-Cf combinations 
FILTER by constraints, e.g. contra-indexing,• sortal predicates, centering 
rules and constraints 
RANK by transition orderings 
In order to apply this algorithm to Japanese, possible Cb-Cf combinations (GEN- 
ERATE step 1) must be constructed from the surface string and information from the 
subcategorization frame of the verb. First the verb subcategorization is examined, and 
if there are more entities than appear in the surface string, zeros are postulated as 
forward centers. These zeros are then treated just like pronouns in English by the rest 
of the algorithm. We use a different ranking for the Cf for Japanese than for English, 
but this has no effect on the actual algorithm itself since the Cf ranking is a declarative 
parameter. 
The steps of the algorithm given above can be interleaved to improve computa- 
tional efficiency. A simple implementation is to: 
• Never propose a Cf that violates linguistic constraints on 
contra-indexing. (In other words, apply the contra-indexing filter as early 
as possible to avoid Cb-Cf combinations that will be eliminated by that 
filter,) 
• If there are pronouns in an utterance, only propose pronouns as possible 
Cbs. (Collect the pronouns from the proposed Cfs as Cbs, from Rule 1.) 
In addition, it is simple to add additional filters to step (2) of the algorithm. For 
instance, any constraint that is lexically specified such as \[±animacy\] can be easily 
applied as a filter. It is also possible to pursue a 'best first' strategy by interleaving 
steps (1), (2), and (3) so that a CONTINUE will be found without extra processing if one 
exists. 
In example 5, we illustrate in more detail how the steps of the algorithm work 
and the difference between CONTINUE and RETAIN. Each utterance shows what the Cb 
and Cf would be for that utterance. We will mostly be concerned with the process of 
resolving the two zeros in utterance 5c. 
203 
Computational Linguistics Volume 20, Number 2 
Example 5 
a. Taroo wa 
b. 
saisin no konpyuutaa o kaimasita. 
TOP/SUBJ newest of computer OBJ bought 
Taroo bought a new computer. 
Cb: TAROO 
Cf: \[TAROO, COMPUTER\] 
0 John ni sassoku sore o misemasita. 
SUBJ John OBJ2 at once that OBJ showed 
'Taroo) showed it at once to John. 
C. 
Cb: TAROO 
Cf: \[TAROO, JOHN, COMPUTER\] CONTINUE 
0 0 atarasiku sonawatta kinoo o setumeisimasita. 
SUBJ OBJ2 newly equipped function OBJ explained 
(Taroo) explained the newly equipped functions to (John). 
Cb: TAROO 
Cfl: \[TAROO, JOHN\] CONTINUE 27 
SUBJ OBJ 
Cf2: \[JOHN, TAROO\] RETAIN 1 
SUBJ OBJ 
Cf3: \[JOHN, JOHN\] CONTRA-INDEX FILTER 
SUBJ OBJ 
Cf4: \[TAROO, TAROO\] CONTRA-INDEX FILTER 
SUBJ OBJ 
Example 5c has explained as the main verb, which requires an animate subject 
and object2. Since there are two animate zeros in 5c, which are also contra-indexed 
by syntactic constraints, both Ziroo and Taroo must be realized in 5c. Constraint (3) 
restricts the Cb to Taroo as the highest-ranked element from the Cf(Sb). The interpretive 
process must also generate the possible candidates for the Cf. If no constraints applied, 
then all four candidates shown above as Cfl, Cf2, Cf3, and Cf4 would be possible. 
However, the contraindexing filter will rule out Cf3 and Cf4. As mentioned above, 
there is no reason that these filters cannot be applied at the GENERATE phase rather 
than later on. 
The only CONTINUE interpretation available, Taroo explained the newly equipped func- 
tions to John, corresponds to the forward centers Cfl. It is a CONTINUE interpretation 
because Cb(5c) = Cb(5b) and also Cb(5c) = Cp(5c). The RETAIN interpretation is less 
preferred and is defined by the fact that Cb(5c) = Cb(5b), but Cb(5c) ~ Cp(5c). This 
example supports the claim that a CONTINUE is preferred over a RETAIN(Z ~- 13.24, 
p < .001). 
In order to find this preferred continue interpretation in a 'best first' fashion, Taroo 
as the Cp(Ui-1) would be tried first as the Cb(Ui), and as the interpretation for the 
subject. Contraindexing rules out Taroo as the object, so John would be tried next as 
the object. 
In the next section, we examine further the application of centering to the inter- 
pretation of zeros in Japanese. We will examine the ranking of forward centers that 
we have adopted for Japanese and explain how this is partially determined by the 
way the Japanese language allows a speaker to express discourse functions. We will 
204 
Marilyn Walker et al. Japanese Discourse 
also give some examples of the interpretation of zeros in cases involving Japanese 
discourse markers for TOPIC and EMPATHY. 
3. Centering in Japanese 
The theory of centering is a formal specification that is intended to model attentional 
state and is defined by the rules and constraints given in Section 2.1. Attentional 
state in turn constrains the discourse participant's interpretation process; one aspect 
of attentional state is the notion of discourse salience. In the centering model, the 
ordering of the forward centers is an approximation of discourse salience. This in turn 
is the main determinant of discourse interpretation processes such as the resolution of 
zeros in Japanese. A crucial question then is what discourse factors must be considered 
to determine the ordering of the forward centers, Cf, in Japanese discourse. 
Being a subject has been shown to be an important factor for English; this is re- 
flected in a Cf ordering by grammatical function (Prince 1981b; Brennan, Friedman, 
and Pollard 1987; Hudson-D'Zmura 1988; Brennan submitted). Aspects of surface or- 
der may also affect the interpretation (Di Eugenio 1990; Hajicova and Vrbova 1982). 
An interpretation algorithm can also use pronominalization as an indicator of what the 
speaker believes is salient (Grosz, Joshi, and Weinstein unpublished). Furthermore, ze- 
ros in Japanese are not realized syntactically so that there must be a way to distinguish 
zeros from other entities inferred to be part of a discourse situation. Consider: 
Example 6 
Taroo ga 0 aimasita. 
Taroo SUBJ OBJ2 met 
Taroo met (0). 
This sentence is not felicitous unless the addressee has already been given some 
information about the person that Taroo met, either in the current discourse or in 
previous discourses. In contrast, nonsubcategorized-for arguments such as adjuncts 
are not necessarily given a specific interpretation, but rather are given a nonspecific 
one. 
Example 7 
Taroo ga Hanako ni aimasita. 
Taroo SUBJ Hanako OBJ2 met 
Taroo met Hanako. 
The sentence means that Taroo met Hanako at some time in some place: the tem- 
poral location of the meeting situation need not be specified. The speaker can utter this 
sentence even if the addressee does not know where and when Taroo met Hanako. 
Thus, in this work, we only represent obligatorily subcategorized arguments of the 
verb on the Cf, assuming that the salience of discourse entities is partially determined 
by virtue of filling a verb's argument role, and the information from the subcatego- 
rization frame is used to determine that a zero is present in an utterance. 
Zeros are then interpreted with reference to the current context. Prince has pro- 
posed that the current context should be categorized by ASSUMED FAMILIARITY (Prince 
1981b; Horn 1986), with a concomitant goal of determining the correlation between 
the use of certain linguistic forms and the types of assumed familiarity. The first di- 
vision of assumed familiarity is into the subtypes of NEW, INFERABLE, and EVOKED. 
205 
Computational Linguistics Volume 20, Number 2 
NEW can be divided into BRAND-NEW, discourse entities that are both new to the dis- 
course and new to the hearer, and UNUSED, discourse entities old to the hearer but 
new to the discourse. The information status of EVOKED can be further divided into 
TEXTUALLY EVOKED, old in the discourse and therefore old to the hearer as well, and 
SITUATIONALLY EVOKED, entities in the current situation. INFERABLES are technically both 
hearer-new and discourse-new but depend on information that is old to the hearer and 
the discourse, and are often treated by speakers as though they were both hearer-old 
and discourse-old. There is a hierarchy of assumed familiarity in terms of discourse 
salience: 
Assumed Familiarity Hierarchy (Prince 1981b): 
TEXTUALLY EVOKED > SITUATIONALLY EVOKED > INFERABLE > UNUSED > BRAND-NEW 
Zeros typically refer to EVOKED entities, 13 but there is a scale of relative salience 
among the EVOKED entities. In our theory this is modeled with Cf ranking. We repeat 
the proposed ranking of the Cf here and justify it in the following sections: 14 
Cf Ranking for Japanese 
(GRAMMATICAL OR ZERO) TOPIC > EMPATHY > SUBJECT > OBJECT2 > OBJECT > OTHERS 
The relevance of the notions of TOPIC and speaker's EMPATHY to centering is that 
a discourse entity realized as the TOPIC or the EMPATHY LOCUS is more salient and 
should be ranked higher on the Cf. Whenever a discourse entity simultaneously fulfills 
multiple roles, the entity is usually ranked according to the highest ranked role. 
In the following sections we will discuss the motivation for this ranking. Section 3.1 
discusses the role of the grammatical topic marker wa in Japanese. Section 3.2 explains 
the role of EMPATHY in Japanese discourse salience and shows that (GRAMMATICAL 
OR ZERO) TOPIC > EMPATHY and that EMPATHY > SUBJ. Section 3.2.1 shows how the 
centering algorithm handles utterances with empathy loci. Zero topics will not be 
discussed until Section 5. 
3.1 Topic 
Discourse entities that are EVOKED, INFERABLE, or UNUSED can be marked as the TOPIC. 
The speaker cannot mark an entity as the grammatical TOPIC unless the hearer is aware 
of the object that s/he is going to talk about (Prince 1978a; Kuno 1976b). For example: 
Example 8 
Hutari wa paatii ni kimasita. 
two-person TOP/SUBJ party to came 
Speaking of two persons, they came to the party. 
13 Under certain circumstances that we cannot explore here, it appears that zeros can at times be used to 
refer to inferable or unused entities, just as pronouns in English sometimes can be. 
14 This ranking resembles Kuno's Empathy Hierarchy and Kameyama's Expected Center Order, but we 
distinguish two kinds of TOPIC and we posit that OBJECT2 is more salient than OBJECT. We continue 
Kuno's use of the term EMPATHY to represent the EMPATHY LOCUS, whereas Kameyama used the property 
IDENT for EMPATHY (Kameyama 1988). 
206 
Marilyn Walker et al. Japanese Discourse 
Example 8 is felicitous only when hutari ('two persons') is understood as meaning 
the two people under discussion. The sentence never means that the people who came to 
the party numbered two. 
The fact that the wa-marked entity should be discourse-old is also shown by the 
fact that a wh-question cannot be answered with a wa-marked NP. 
Example 9 
a. Dono hito 
b-1. 
b-2. 
ga 
which person SUBJ Ziroo OBJ defended 
Which person defended Ziroo ? 
Taroo ga Ziroo o bengosimasita. 
Taroo SUBJ Ziroo OBJ defended 
Taroo defended Ziroo. 
*Taroo wa Ziroo o bengosimasita. 
Taroo TOP/SUBJ Ziroo OBJ defended 
Taroo defended Ziroo. 
Ziroo o bengosimasita ka. 
Q 
What the question context shows is that even in a simple declarative sentence, the 
use of the topic marker wa contrasts with the subject marker ga in what is understood 
as already in the discourse context. For instance, in a discourse initial utterance, 10a, 
assumes no shared information or that someone defended Ziroo and asserts that the 
someone is Taroo. In 10b, the discourse-old proposition is that Taroo did something and 
what is asserted is that what he did was defend Ziroo. 
Example 10 
a. Taroo ga 
b. 
Ziroo o bengosimasita. 
Taroo SUBJ Ziroo OBJ defended 
Taroo defended Ziroo. 
Taroo wa Ziroo o bengosimasita. 
Taroo TOP/SUBJ Ziroo OBJ defended 
Taroo defended Ziroo. 
While topics are often subjects, subject and grammatical topic need not coincide. 
Any argument can be realized as a topic, as shown in examples 11 and 12. 
Example 11 
Taroo wa Hanako ga bengosita. 
Taroo TOP Hanako SUBJ defended 
As for Taroo, Hanako defended (him). 
Example 12 
Tokyoo e wa Hanako ga itta° 
Tokyo to TOP Hanako SUBJ went 
To Tokyo, Hanako went. 
The assumption that the TOPIC is more salient than the SUBJECT, when the two are 
different, is supported by the fact that an indefinite NP in subject position such as who, 
which, or somebody cannot be regarded as the TOPIC: an indefinite NP is never marked 
by the topic marker wa, but by the subject marker ga. For example: 
207 
Computational Linguistics Volume 20, Number 2 
Example 13 
Dono hito ga Ziroo o bengosimasita ka. 
which person SUBJ Ziroo OBJ defended Q 
Which person defended Ziroo ? 
Example 14 
*Dono hito wa Ziroo o bengosimasita ka. 
who person TOP/SUBJ Ziroo OBJ defended Q 
Which person defended Ziroo? 
It is clear from these examples that the grammatical topic, wa-marked entity, in 
Japanese, represents assumable shared information in an ongoing conversation. It has 
been taken to be the 'theme' or 'what the sentence is about' (Kuno 1973; Shibatani 
1990). In our framework, this is the role of the Cb. We will provide evidence supporting 
this position in Section 4. However, we claim that this is just a default and that other 
factors can contribute to establishing or continuing an entity as the Cb. Kuno also 
claims that a zero subject is equivalent to a wa-marked entity, and we provide support 
for this claim in Section 5, showing that the property of having previously been the 
Cb, in combination with being realized by a zero, contributes to an entity being the Cp. 
3.2 Empathy 
Kuno (1976b) proposed a notion of EMPATHY in order to present the speaker's position 
or identification in describing a situation. In a hugging situation involving a man named 
Taroo and his son Saburoo, Kuno notes that this situation can be described in various 
ways, some of which are shown in example 15. 
Example 15 
a. Taroo hugged Saburoo. 
b. Taroo hugged his son. 
c. Saburoo's father hugged him. 
These sentences differ from each other with respect to camera angle, the position that 
the speaker takes to observe and describe this situation. In 15a, the speaker is assumed 
to be describing the event objectively: the camera is placed at the same distance from 
both Taroo and Saburoo. On the other hand, the camera may be placed closer to Taroo 
in 15b and closer to Saburoo in 15c. This is shown by the use of relational terms such 
as son and father, respectively. The term EMPATHY is used for this camera angle, which 
indicates the speaker's position among the participants in the event describedY 
15 The speaker's position is not determined by his physical proximity, but rather is measured by the 
emotional or social relationship. In this sense, the term speaker's identification (Kuno 1976b) may be more 
suitable than the term speaker's position. Furthermore, the notion of EMPATHY is different from that of 
perspective (Iida 1993). Empathy is the speaker's identification with a discourse entity, but the speaker 
does not have to take the perspective of the person who he empathizes with. For example, consider the 
following utterance: 
(i) Taroo wa Hanako ni migigawa no hon o totte-kureta. 
Taroo TOP/SUBJ Hanako OBJ2 right GEN book OBJ take-gave 
Taroo did Hanako a flavor in taking a book on his~her right. 
In this example, the speaker empathizes with Hanako as indicated by the empathy verb kureru, yet 
he still can describe the given situation from Taroo's perspective, which is indicated by ambiguity in 
the interpretation of the deictic expression migigawa no ('right of'). 
208 
Marilyn Walker et al. Japanese Discourse 
In Japanese the realization of speaker's empathy is especially important when 
describing an event involving giving or receiving. There is no way to describe a giving 
and receiving situation objectively (Kuno and Kaburaki 1977). In 16, the use of the verb 
kureru indicates the speaker's empathy with Ziroo, the discourse entity realized in 
object position, while in 17, the speaker's empathy with the subject Taroo is indicated 
by the use of the past tense form yatta of the verb yaru. 
Example 16 
Taroo ga Ziroo ni hon o kureta. 
Taroo SUBJ Ziroo OBJ2 book OBJ gave 
Taroo gave Ziroo a book. EMPATHY=OBJ2=ZIROO 
Example 17 
Taroo ga Ziroo ni hon o yatta. 
Taroo SUBJ Ziroo OBJ2 book OBJ gave 
Taroo gave Ziroo a book. EMPATHY=SUB=TAROO 
A verb that is sensitive to the speaker's empathy is an EMPATHY-LOADED verb. The 
EMPATHY LOCUS is the argument position whose referent the speaker automatically 
identifies with. In other words, the verb kureru has the EMPATHY LOCUS on the object, 
while verbs like yaru place the EMPATHY LOCUS on the subject. 
The use of deictic verbs such as kuru ('come'), iku ('go'), okuru ('send to'), and yokosu 
('send in') also encode the speaker's empathy. For example, the speaker indicates 
empathy with Taroo by using the past tense form kita of the verb kuru in the following 
example. 
Example 18 
Hanako wa Taroo no tokoro ni kita. 
Hanako TOP/SUBJ Taroo of place to came 
Hanako came to Taroo's place. 
Many Japanese verbs can be made into empathy-loaded verbs because of a pro- 
ductive verb-compounding operation by which these empathy-loaded verbs are used 
as the auxiliary verb, attaching to the main verb} 6 For example, kureru can be used 
as a suffix, to mark OBJ or OBJ2 as the EMPATHY LOCUS. The attachment of yaru marks 
SUBJECT as the EMPATHY LOCUS. The complex predicate made by this operation inherits 
the EMPATHY LOCUS of the suffixed verb. For example: 
Example 19 
Hanako ga Taroo ni hon o yonde-kureta. 
Hanako SUBJ Taroo OBJ2 book OBJ read-gave 
Hanako did Taroo a favor in reading a book. EMPATHY = OBJ2 = TAROO 
In this case Taroo is interpreted as the EMPATHY LOCUS because of the auxiliary 
kureta attached to the main verb. Similarly in example 20, the speaker indicates empa- 
thy with Hanako by using the past tense form yatta of the verb yaru as an auxiliary 
verb to the main verb tazuneru. 
16 Certain intransitive verbs cannot be made into empathy-loaded verbs since the empathy-loaded versions make no sense, e.g. moreru (leak). 
209 
Computational Linguistics Volume 20, Number 2 
Example 20 
Hanako ga Taroo o tazunete-yatta. 
Hanako SUBI Taroo OBJ visit-gave 
(lit.)Hanako received a favor in visiting Taroo. EMPATHY = SUBJ = HANAKO 
As demonstrated in the following examples, a discourse entity that is realized as 
the EMPATHY LOCUS must be EVOKED. 
Example 21 
Taroo ga Ziroo ni okane o kasite-kureta. 
Taroo SUBJ Ziroo OBJ2 money OBJ lend-gave 
Taroo did Ziroo a favor in lending him some money. 
Example 22 
*Taroo ga dareka ni okane o kasite-kureta. 
Taroo SUBJ somebody OBJ2 money OBJ lend-gave 
Taroo did somebody a favor in lending him some money. 
Example 23 
*Taroo ga misiranu hito ni okane o kasite-kureta. 
Taroo SUBJ unknown person OBJ2 money OBJ lend-gave 
Taroo did a stranger a favor in lending him some money. 
The contrast between 21, 22, and 23 demonstrates that the use of a BRAND-NEW 
entity in the EMPATHY LOCUS position of the verb give is not acceptable. Therefore an 
entity in the EMPATHY LOCUS position is ranked in a higher position on the Cf than the 
subject. 
3.2.1 Empathy and the Centering Algorithm. Using the Centering Algorithm, we 
model EMPATHY as a language-specific discourse factor by adding the EMPATHY-marked 
discourse entity to the Cf ranking. Then preferences for CONTINUE over RETAIN when 
EMPATHY is involved can be demonstrated, as in example 24 below: 17 
Example 24 
a. Hanako wa kuruma ga kowarete komatteimasita. 
Hanako TOP/SUBJ car SUBJ broken at a loss-was 
Her car broken, Hanako was at a loss. 
Cb: HANAKO 
Cf: \[HANAKO, CAR\] 
b. Taroo ga 0 sinsetu-ni te o kasite-kuremasita. 
Taroo SUBJ OBJ2/EMP kindly hand OBJ lend-gave. 
Taroo kindly did (Hanako) a favor in helping her. 
Cb: \[HANAKO\] 
Cf: \[HANAKO, TAROO\] 
EMPATHY SUBJ 
17 The verb form kuremasita in 24b is the polite form of kureta, the past tense form of the verb kureru. 
210 
Marilyn Walker et al. Japanese Discourse 
C. Tugi no hi 0 0 eiga ni sasoimasita. 
next of day SUBJ OBJ movie to invited 
Next day (Hanako) invited (Taroo) to a movie. 
Cb: HANAKO 
C fl: \[HANAKO, TAROO\] CONTINUE 16 
SUBJ OBI 
Cf2: \[TAROO, HANAKO \] RETAIN 2 
SUBJ OBJ 
In 24c, the verb invited requires an animate subject and object, and these must 
be realized by different discourse entities because of the contraindexing constraint. 
Hanako is the most highly ranked entity from 24b that is realized in 24c, and there- 
fore must be the Cb. The preferred interpretation is therefore she invited him to a movie 
(Z = 5.25, p < .001). This corresponds to Cfl, the more highly ranked CONTINUE tran- 
sition, in which Hanako is the preferred center, Cp. This interpretation can be found 
with minimal processing by trying the Cp(24b), Hanako, as the Cb(24c), by interpret- 
ing the subject zero as Hanako. This gives a CONTINUE transition. Then contraindexing 
constraints mean that Hanako cannot fill both argument positions, so the object posi- 
tion is interpreted as Taroo. This interpretation is found with minimal processing by 
interleaving the steps of the Centering algorithm proposed in Brennan et al. (1987). 
Note that nothing special needs to be said about the fact that EMPATHY is the 
discourse factor that made Hanako the Cp in 24b and thus predicted that Hanako 
would be the Cb at 24c (pace Brennan, Friedman, and Pollard 1987). The preference 
in the interpretation follows from the distinction between CONTINUE and RETAIN and 
the ranking of Cf. Thus, the centering framework is easily adapted to handle this 
language-specific feature. 
3.3 Topic and Empathy 
In general the assignment of the EMPATHY relationship is pragmatic. It is determined by 
the speaker's relation to the discourse participants in the discourse. In 24, for example, 
the EMPATHY relationship between the speaker and Hanako and between the speaker 
and Taroo is clear: the use of the empathy verb in the second sentence indicates that 
the speaker is closer to Hanako than to Taroo. 
However, besides cases where the speaker clearly expresses who s/he empathizes 
with, it is also possible for the context to provide some information about the speaker's 
proximity relationship with discourse participants in the given discourse, so that the 
hearer can determine the EMPATHY relation that the speaker has in mind. In this paper, 
we only consider cases where EMPATHY is syntactically marked by the use of empathy- 
loaded verbs. 
Kuno's notion of EMPATHY is more general. For instance, Kuno's EMPATHY HIERAR- 
CHY consists of different scales for EMPATHY that include notions such as TOPIC and 
SPEAKER (Kuno 1987). Kuno's Topic Empathy Hierarchy suggests that the discourse 
entity realized as the TOPIC will often coincide with the EMPATHY LOCUS: 
Topic Empathy Hierarchy: Discourse-Topic > Discourse-Nontopic 
Given an event or state that involves A and B such that A is corefer- 
ential with the topic of the present discourse and B is not, it is easier 
for the speaker to empathize with A than with B. 
In support of Kuno's claim, we have found that when no empathy relation is 
clearly indicated and no topic has been clearly established that it is difficult for a 
211 
Computational Linguistics Volume 20, Number 2 
hearer to determine the empathy relation that the speaker intends. Previous Cbs and 
current Cps can be high on the empathy scale, and yet the discourse entity realized as 
the grammatical TOPIC does not necessarily coincide with the discourse entity realized 
as the EMPATHY LOCUS. A simple sentence to show this point is given in example 25 
below: 
Example 25 
Taroo wa Ziroo ni hon o yonde-kuremasita. 
Taroo TOP/SUBJ Ziroo OBJ2 book OBJ read-gave 
Taroo gave Ziroo a //avor off reading a book. EMPATHY = OBJ2 = ZIROO 
In example 25, Taroo is the TOPIC while Ziroo is the EMPATHY LOCUS. Similarly, a 
zero does not have to be realized as the EMPATHY LOCUS. In 26b the zero in the subject 
position realizes the Cb and refers to Taroo. 
Example 26 
a. Taroo wa syukudai o zenbu yari-oemasita. 
Taroo TOP/SUB homework OBJ all do-finished 
Taroo finished his homework. 
b. 0 Ziroo ni hon o yonde-kuremasita. 
SUBJ Ziroo OBJ2 book OBJ read-gave 
(Taroo) gave Ziroo a //avor o//reading a book. EMPATHY = OBJ2 = ZIROO 
TOPIC is higher than EMPATHY in the Cf ranking. The higher degree of salience 
of TOPIC over EMPATHY is shown by the different interpretation of (b) sentences in 
examples 27 and 28. The only difference in these examples is that Mitiko is wa-marked 
in 27a but is ga-marked in 28a: 
Example 27 
a. Mitiko wa kanai o gityoo ni osite-kuremasita. 
Mitiko TOP/SUBJ wife OBJ/EMP chairman OBJ2 recommend-gave 
Mitiko did my wife a favor in recommending her as chairperson. 
b. 0 asu no kaihyoo-kekka o tanosimi-ni siteim asu. 
SUBJ tomorrow of results OBJ look-forward doing-is 
(Mitiko) is looking forward to tomorrow's results. 
Example 28 
a. Mitiko ga kanai o gityoo ni osite-kuremasita. 
Mitiko SUBJ wife OBJ/EMP chairman OBJ2 recommend-gave 
Mitiko did my wife a //avor in recommending her as chairperson. 
b. 0 asu no kaihyoo-kekka o tanosimi-ni siteimasu. 
SUBJ tomorrow of results OBJ look-forward doing-is 
(Mitiko) is looking forward to tomorrow's results. 
(My wife) is looking forward to tomorrow's results. 
212 
Marilyn Walker et al. Japanese Discourse 
The TOPIC Mitiko is preferred as the unexpressed subject of the (b) sentence in 
example 27. TM On the other hand, the subject Mitiko is not strongly preferred, as shown 
in example 28: the zero in the second sentence in 28 is understood as referring to either 
Mitiko or my wife. That is, the possible interpretation in these examples shows that the 
NP my wife, which is realized as the EMPATHY LOCUS, is not as salient as the TOPIC. 19 
So why is it easier to empathize with a discourse entity that has been the topic 
as Kuno demonstrates? It seems important to keep the notions of TOPIC and EMPATHY 
separate, but in Section 5.1 we will demonstrate an effect where the topic entity is 
interpreted as the empathy locus. We claim that the ranking of the Cf and the poten- 
tial for a CONTINUE interpretation determines whether this effect will hold. In other 
words, the tendency for the topic entity to be interpreted as the empathy locus follows 
from more general discourse processing factors, such as a hearer preferring CONTINUE 
transitions within a given local stretch of discourse. 
3.4 Summary 
To summarize, we have outlined the roles of discourse markers such as those for TOPIC 
and EMPATHY by which Japanese grammaticizes some aspects of discourse function, 
and we have argued that TOPIC and EMPATHY markers can only be used on entities 
that are already in the discourse context. 
One factor that hasn't been discussed is the role of pronominalization, but many 
researchers have argued that discourse entities realized by pronouns are more salient 
than other discourse entities (Clark and Haviland 1977; Grosz, Joshi, and Weinstein 
unpublished; Kuno 1976b, 1987). We take zeros in Japanese to be analogous to pro- 
nouns in English in this respect. Since pronominalization can apply at any position in 
the ranking of the Cf, the role of its contribution is particularly interesting when it is 
in conflict with some other factor such as grammatical function or topic marking. This 
will be discussed further in Section 5. 
4. Initial Center Instantiation 
INITIAL CENTER INSTANTIATION is a process by which a discourse entity introduced in 
a segment-initial utterance becomes the Cb. In our framework, this happens as a side 
effect of the Centering Algorithm. Typically, when an interpretation is found for the 
second utterance in a discourse segment, the Cb becomes instantiated. 2° The Cb of an 
initial utterance Ui is treated as a variable that is then unified with whatever Cb is 
assigned to the subsequent utterance Ui+l. 
Typically, a discourse entity is introduced as a ga-marked subject, and then is 
referred to by a zero in a subsequent utterance (Clancy and Downing 1987). Consider 
example 29. 
18 The zero may be interpreted as indirectly referring to the speaker. This interpretation is always possible 
when the verb kureru is used: the use of kureru implies that the speaker is closer to the beneficiary 
argument (i.e. the 0-marked NP in these examples), and the favor given to this person is understood as 
a benefit to the speaker as well. 19 Although it seems as though empathy isn't higher than subject, the conflating factor is that topic 
marking establishes a Cb, whereas in 28 no Cb has been established. This is explained in detail in 
Section 4. 20 In Walker, Iida, and Cote (1990) we called this Center Establishment. Henceforth we will refer to this 
process as Center Instantiation in order to avoid confusion with Kameyama's term center 
establishment, which is a different mechanism in her theory (Kameyama 1985). 
213 
Computational Linguistics Volume 20, Number 2 
Example 29 
a. Taroo ga 
b. 
deeta o konpyuutaa ni utikondeimasita. 
Taroo SUBJ data OBJ computer in was-storing 
Taroo was storing the data in a computer. 
Cb: \[?\] I Cf: \[TAROO, DATA\] 
0 yatto hanbun yari-owarimasita. 
SUBJ finally half do-finished 
Finally (Taroo) was half finished. 
\[ Cb: TAROO \] 
Cf: \[TAROO\] CONTINUE 
Using Taroo as the subject in example 29a is not enough to establish this discourse 
segment as being about Taroo. It is the use of the zero in example 29b that serves to 
instantiate Taroo as the Cb. By our definition of CONTINUE, 29b is a continue transition, 
because Cb(29b) = Cp(29b) and there was no Cb in 29a. However, Kuno argues that 
referring to a discourse entity with a zero is equivalent to marking it as the grammatical 
topic with wa (Kuno 1972). Our interpretation of this argument is that the use of wa in a 
discourse-initial utterance instantiates the wa-marked entity as the Cb in one utterance. 
This claim is supported by the contrast with the GA-WA alternation in examples 30 
and 31, where there is a shift in interpretation depending on whether Taroo is marked 
with wa in the first sentence. 21 
Example 30 
a. Taroo ga 
b. 
Ziroo o min'na no mae de tatakimasita. 
SUBJ OBJ all of front in hit. 
Taroo hit Ziroo in front of all the other people. 
Cb: \[71 \[ 
Cf: \[TAROO, ZIROO\] 
Itiniti-zyuu, kanzen-ni 0 0 musi-simasita. 
all-day completely ignored 
(Ziroo) ignored (Taroo) all day. 
Cb: TAROO 
Cf: \[TAROO, ZIROO\] 3 
Cb: ZIROO 
Cf: \[ZIROO, TAROO\] 8 
In example 30, Taroo is introduced by ga. In this case, it appears that there is a ten- 
dency due to lexical semantics to instantiate Ziroo as the Cb in the second utterance. 22 
By the centering definitions, taking either Taroo or Ziroo to be the Cb can result in 
a CONTINUE interpretation. However, assuming that the Cf ordering at example 30a is 
correct, constraint 3 is violated by the preferred interpretation of 30b. Since both of the 
entities in Cf(30a) are realized, the Cb in example 30b should be the most highly ranked 
one. There are two possible conclusions here: (1) In discourse-initial utterances, when 
21 These examples were tested by asking survey participants to indicate preference rankings. The 
numbers given here are only for those subjects who expressed strong preferences; some subjects 
expressed no preference. 
22 The number of subjects here are too small to test statistically. 
214 
Marilyn Walker et al. Japanese Discourse 
no clear indication of topic is given, the Cf ordering alone is not a strong constraint; 
(2) the ordering of the Cf should be partly determined by lexical semantics or other 
knowledge about the situation being described. However, compare example 30 with 
example 31. 
Example 31 
a. Taroo wa Ziroo o min'na no mae de tatakimasita. 
SUBJ OBJ all of front in hit. 
Taroo hit Ziroo in front of all the other people. 
Cb: \[TARO0\] 
Cf: \[TAROO, ZIRO0\] 
b. Itiniti-zyuu, kanzen-ni 0 0 musi-simasita. 
all-day completely ignored 
(Taroo) ignored (Ziroo) all day. 
Cb: TAROO 
Cf: \[TAROO, ZIROO\] 10 
Cb: ZIROO 
Cf: \[ZIROO, TAROO\] 4 
The use of wa in example 31 seems to override the semantic preference that was 
exhibited in example 30, so that subjects now prefer an interpretation in which Taroo 
is the Cb. 23 This shows that Taroo has not been instantiated as the Cb when it is time 
to interpret the two zeros in example 30b. We explain the contrast by assuming that 
the TOPIC instantiates the Cb when it is first introduced in a discourse-initial utterance, 
as in example 31a. Then the only way to get a CONTINUE interpretation for 31b is for 
Taroo to be the Cb at 31b. 
Furthermore, we can detect no differences in the interpretation of the final utter- 
ance between three utterance sequences in which an entity is introduced by wa, and 
four utterance sequences in which an entity is first introduced by ga and then realized 
by a zero in the second utterance. This provides further support for the claim that the 
status of discourse entities realized as grammatical topics and those realized as zero 
subjects is equivalent. 
4.1 Summary 
In sum, we have argued that the use of wa in a discourse-initial utterance instantiates 
the wa-marked entity as the Cb. Cb instantiation can equivalently be done with a two- 
utterance sequence in which the entity is first introduced as a subject, ga-marked, and 
then established as the Cb in the following utterance with a zero referring to that entity. 
In addition, the fact that the Cb is uninstantiated in discourse initial utterances has 
the effect that the Cf ranking in a discourse initial utterance is not a strong constraint, 
as it is once a Cb is established. 
5. Zero Topic Assignment 
In this section we introduce the notion of a ZERO TOPIC and a rule or assumption that 
can be employed as part of the interpretive process called ZERO TOPIC ASSIGNMENT. 
23 The small number of subjects means that we can't provide statistical support for this claim. 
215 
Computational Linguistics Volume 20, Number 2 
The rule of ZERO TOPIC ASSIGNMENT defines our distinction between grammatical 
topic and zero topic. This rule allows a zero that has just been the Cb to continue 
as the Cp, even when it is not realized in a discourse-salient syntactic position such 
as subject. We will demonstrate this with examples that realize both grammatical 
and zero topics. In these cases, the discourse situation is such that the hearer may 
maintain multiple hypotheses about where the speaker's attention is directed, and 
must determine whether to apply the default that the grammatical topic is usually 
the Cp. 24 
Zero Topic Assignment 
When a zero in Ui+l represents an entity that was the Cb(Ui), and 
when no other CONTINUE transition is available, that zero may be in- 
terpreted as the ZERO TOPIC of Ui+l. 
What this means is that, in certain discourse environments, the entity that was 
previously the Cb is predicted to continue as the Cb. We conjecture that ZTA is appli- 
cable in all free word-order languages with zeros. 25 However, ZERO TOPIC ASSIGNMENT 
is optional; here we have suggested two constraints on when it applies. We will give 
examples below of cases where it doesn't apply. 
The option of ZERO TOPIC ASSIGNMENT (henceforth ZTA) has been overlooked in 
previous treatments of zeros in Japanese. ZTA explains why the discourse entity 
Hanako, which is realized as OBJECT2 in example 32c is interpreted as the SUBJECT 
of example 32d. 
Example 32 
a. Hanako wa siken o oete, kyoositu ni modorimasita. 
Hanako TOP/SUBJ exam OBJ finish classroom to returned 
Hanako returned to the classroom, finishing her exams. 
Cb: HANAKO 
Cf: \[HANAKO, EXAM\] 
b. 0 hon o locker ni simaimasita. 
SUBJ book OBJ locker in took-away 
She put her books in the locker. 
Cb: HANAKO 
Cf: \[HANAKO, BOOK\] CONTINUE 
c. Itumo no yooni Mitiko ga 0 mondai no tokikata o 
always like SUBJ Mitiko OBJ2 problem solve-way OBJ 
setumeisidasimasita. 
explained 
Mitiko, as usual, explained (to Hanako) how to solve the problems. 
Cb: HANAKO 
Cfl: \[HANAKO, MITIKO, SOLUTION\] ZTA CONTINUE 
TOP, SUBJ, OBJ 
Cf2: \[MITIKO, HANAKO, SOLUTION\] RETAIN 
SUBJ, OBJ2, OBJ 
24 While some of the utterance sequences we examine are potentially ambiguous for native speakers, the 
examination of these discourse situations offers considerable insight into those where there is no 
ambiguity. 
25 We only look at object topics here but there may be limits as to how lowly ranked on the Cf an entity 
can be and still be the zero topic, e.g. by-passive agentive. 
216 
Marilyn Walker et al. Japanese Discourse 
d. 0 0 ohiru ni sasoimasita. 
SUBJ OBJ lunch to invited 
(Hanako) invited (Mitiko) to lunch. 
Cbl: 
Cfl: 
HANAKO 
\[HANAKO, LUNCH, MITIKO\] CONTINUE from Cfl(c) 28 
SUBJ, OBJ2, OBJ 
Cb2: MITIKO 
Cf2: \[MITIKO, LUNCH, HANAKO\] SMOOTH-SHIFT from Cf2(c) 6 
SUBJ, OBJ2, OBJ 
The possibility of ambiguity as to the attentional state of the speaker is reflected 
in the fact that there are two possible Cfs for example 32c; Cf2 of 32c is the only 
Cf possible without ZTA, and represents a RETAIN rather than a CONTINUE. By the 
formulation of the ZTA rule above, ZTA is triggered by the fact that no CONTINUE 
transition is available. 
The availability of ZTA means that HANAKO can be the Cp even when MITIKO is 
realized as the subject. This leads to a potential ambiguity in example 32d, because it 
is possible for a hearer to simultaneously entertain both of the Cf(32c). In this case the 
ZTA interpretation is preferred (Z = 4.95, p < .001). The less preferred SMOOTH-SHIFT 
interpretation would result from the algorithm's application to Cf2 of 32c. 26 
ZTA explains the contrast between the discourse segments,in examples 32 above 
and 33 below. The only difference between 32 and 33 is that in 32c, MITIKO is a ga- 
marked subject, whereas in 33c, MITIKO is a wa-marked subject/grammatical topic. 
Utterances 32c and 33c have the same meaning. This minimal pair provides a test to 
see whether ZTA actually characterizes these discourse related effects. 
Example 33 
a. Hanako wa siken o oete, kyoositu ni modorimasita. 
Hanako TOP/SUBJ exam OBJ finish classroom to returned 
Hanako returned to the classroom, finishing her exams. 
Cb: HANAKO \[ 
Cf: \[HANAKO, EXAM\] 
b. 0 hon o locker ni simaimasita. 
SUBJ book OBJ locker in took-away 
(Hanako) put (her) books in the locker. 
Cb: HANAKO \[ 
Cf: \[HANAKO, BOOK\] CONTINUE 
C. Itumo no yooni Mitiko wa 0 mondai no tokikata o 
always like TOP/SUBJ Mitiko OBJ2 problem solve-way OBJ 
settrmeisidasimasita. 
explained 
Mitiko, as usual, started explaining (to Hanako) how to solve the problems. 
Cb: HANAKO 
Cfl: \[HANAKO, MITIKO, SOLUTION\] ZTA CONTINUE 
TOP, SUBJ, OBJ 
Cf2: \[MITIKO, HANAKO, SOLUTION\] RETAIN 
TOP, OBJ2, OBJ 
26 See Section 2 for an example of how a smooth-shift interpretation is calculated. 
217 
Computational Linguistics Volume 20, Number 2 
d. 0 0 ohiru ni sasoimasita. 
SUBJ OBJ lunch to invited 
(Hanako) invited (Mitiko) to lunch. 
(Mitiko) invited (Hanako) to lunch. 
Cbl: HANAKO 
Cf2: \[HANAKO, LUNCH, MIT1KO\] CONTINUE from Cfl(c) 18 
SUBJ, OBJ2, OBJ 
Cb2: MITIKO 
Cf2: \[MITIKO, LUNCH, HANAKO \] SMOOTH-SHIFT from Cf2(c) 16 
SUBJ, OBJ2, OBJ 
The wa marking has the predicted effect. Using the grammatical topic marker wa in 
example 33c dampens ZTA and thus affects the interpretation of example 33d, which is 
now completely ambiguous (Z = 0.34, not significantly different than chance). Because 
the discourse entity realized as the grammatical topic and indicated by the wa-marked 
NP is the Cp by default, ten subjects who previously did so can no longer get an 
interpretation that depends on ZTA. It seems that the situation can be characterized 
as a case of competing defaults, so that in example 33, some hearers apply the default 
that the wa-marked entity is the Cp, and others apply the default that CONTINUE 
interpretations are preferred and that zeros realize discourse entities that are ranked 
highly on the Cf. 
The RETA1N interpretation in example 33c, Cf2, indicates that these hearers expect 
the conversation to shift to being about Mitiko; the fact that Mitiko is the Cp(33c), 
along with constraint 3 will force a shift. Given a SHIFT, the Mitiko invited Hanako to 
lunch interpretation is preferred because it is the more highly ranked SMOOTH-SHIFT 
transition. 27 
These examples clearly show that the wa-rnarked NP is not always the Cp and sup- 
port Shibatani's claim that the interpretation of wa depends on the discourse context 
(Shibatani 1990). The astute reader will have noticed that in the cases where Hanako 
is a zero topic, the wa-marked Mitiko discourse entity is ranked according to gram- 
matical function. We conjecture that an inference of contrast is supported when the 
grammatical topic is not the Cp. 
The following section discusses the interaction of ZTA with empathy. Then in 
Section 5.2, we discuss further the ramifications of our distinction between grammatical 
and zero topic. 
5.1 Empathy and Zero Topic Assignment 
This section investigates the interaction of EMPATHY and ZERO TOPIC ASSIGNMENT (ZTA). 
The discourse segment in example 34 is a minimal pair with that in example 35. In 
34d the main verb is setumeisita ('explain') without any EMPATHY marking, whereas 
in 35d, the same sentence occurs with an auxiliary empathy verb as setumeisitekureta. 
Remember that kureta marks the OBJ or OBJ2 as the EMPATHY LOCUS. 
27 If MITIKO could represent a topic object in 33d, there would be another equally ranked SMOOTH-SHIFT 
interpretation for 33d. However, according to the formulation of ZERO TOPIC ASSIGNMENT, MITIKO cannot 
be a zero topic because it was not the Cb of the previous utterance, 33c. 
218 
Marilyn Walker et al. Japanese Discourse 
Example 34 
a. Taroo wa deeta o konpyuutaa ni utikondeimasita. 
Taroo TOP/SUBJ data OBJ computer in was-storing 
Taroo was storing the data in a computer. 
Cb: TAROO 
Cf: \[TAROO, DATA\] 
b. 0 yatto hanbun yari-owarimasita. 
SUBJ finally half do-finished 
Finally (Taroo) was half finished. 
Cb: TAROO 
Cf: \[TAROO\] CONTINUE 
c. Ziroo ga 0 hurui deeta o misemasita. 
Ziroo SUBJ OBJ2 old data OBJ showed 
Ziroo showed (Taroo) some old data. 
d. 
Cb: TAROO 
Cfl: \[TAROO, ZIROO, DATA\] ZTA CONTINUE 
TOP, SUBJ, OBJ 
Cf2: \[ZIROO, TAROO\] RETAIN 
SUBJ, OBJ2, OBJ 
0 0 ikutuka no kuitigai o setumeisimasita. 
SUBJ OBJ2 several of differences OBJ explained 
(Ziroo) explained several differences to (Taroo). 
(Taroo) explained several differences to (Ziroo). 
Cbl: TAROO 
Cfl: \[TAROO, ZIROO, DIFFERENCES\] CONTINUE from Cfl(c) 12 
SUBJ, OBJ2, OBJ 
Cb2: ZTROO 
Cf2: \[ZIROO, TAROO, DIFFERENCES\] SMOOTH-SHIFT from Cf2(c) 22 
SUBJ, OBJ2, OBJ 
The interpretations of example 34d show that it is possible for some subjects to 
interpret Taroo as the zero topic in example 34c. This is possible because Taroo was both 
the Cp and the Cb for 34a and 34b. The two Cfs of 34c reflect multiple possibilities in 
attentional state. 28 The competing defaults consist of the assumption that ZTA applies, 
versus the assumption that subjects are more highly ranked than objects on the Cf. In 
this case no preference between the two interpretations can be demonstrated (Z = 1.79, 
not significant). 
Example 35 is a minimal pair with example 34. In 35d, the speaker provides more 
syntactic information by using the empathy verb kureta to indicate that the discourse 
entity realized as the OBJECT2 is the EMPATHY lOCUS. 
Example 35 
a. Taroo wa deeta o konpyuutaa ni utikondeimasita. 
Taroo TOP/SUBJ data OBJ computer in was-storing 
Taroo was storing the data in a computer. 
28 Although both possibilities have the same semantic interpretation. 
219 
Computational Linguistics Volume 20, Number 2 
Cb: TAROO 
Cf: \[TARO0, DATA l 
b. 0 yatto hanbun yari-owarimasita. 
SUBJ finally half do-finished 
Finally (Taroo) was half finished. 
Cb: TAROO 
Cf: \[TARO0\] CONTINUE 
c. Ziroo ga 0 hurui deeta o misemasita. 
Ziroo SUBJ OBJ2 old data OBJ showed 
Ziroo showed (Taroo) some old data. 
d. 
Cb: TAROO 
Cfl: \[TAROO, ZIROO, DATA\] ZTA CONTINUE 
TOP, SUBJ, OBJ 
Cf2: \[ZIROO, TAROO, DATA\] RETAIN 
SUBJ, OBJ2, OBJ 
0 0 ikutuka no kuitigai o setumeisite-KURE-masita. 
SUBJ OBJ2/EMP several of differences OBJ explained-gave 
(Ziroo) did (Taroo) a favor of explaining several differences. 
Cbl: TAROO 
Cfl: \[TAROO, ZIROO, DIFFERENCES\] CONTINUE from Cfl(c) 33 
EMP-OBJ2, SUBJ, OBJ 
Cb2: ZIROO 
Cf2: \[ZIROO, TAROO, DIFFERENCES\] SMOOTH-SHIFT from Cf2(c) 1 
EMP-OBJ2, SUBJ, OBJ 
Empathy associates with the previous Cb to yield a CONTINUE transition, and 
the interpretation changes so that the utterance is no longer ambiguous (Z = 16.24, 
p < .001). In this case it is possible to interpret both example 35c and example 35d 
as CONTINUES by assuming ZTA at 35c. This example also validates ZTA because 
empathy associates with the zero topic (Kuno 1976b, 1987). Furthermore, this minimal 
pair highlights aspects of the interaction between syntax and inference. The fact that 
the empathy verb in 35d is the only difference between examples 34 and 35 shows that 
the preference in interpretation does not follow from inferences based on information 
about who is likely to explain what to whom, depending on who showed •who the 
data, or whether the data is new or old. 
Example 36 contrasts minimally with example 35 but on another dimension. In this 
case, 36c is a CONTINUE with Taroo realized in subject position, rather than a CONTINUE 
based on ZTA. The Ziroo explained to Taroo interpretation is again clearly preferred here 
as in 35d (Z = 3.638, p < .001). 
Example 36 
a. Taroo wa deeta o konpyuutaa ni utikondeimasita. 
Taroo TOP/SUBJ data OBJ computer in was-storing 
Taroo was storing the data in a computer. 
\[ Cb: \[TAROO\] \[ 
Cf: \[TAROO, data\] 
220 
Marilyn Walker et al. Japanese Discourse 
b. 
C. 
d. 
0 yatto hanbun yari-owarimasita. 
SUBJ finally half do-finished 
Finally (Taroo) was half finished. 
Cb: TAROO \[ 
Cf: \[TAROO\] CONTINUE 
0 Ziroo ni hurui deeta o misemasita. 
SUBJ Ziroo OBJ2 old data OBJ showed 
(Taroo) showed Ziroo some old data. 
Cb: TAROO 
Cfl: \[TAROO, ZIROO, DATA\] CONTINUE 
SUBJ, OBJ2, OBJ 
0 0 ikutuka no kuitigai o setumeisite-KURE-masita. 
SUBJ OBJ2/EMP several of differences OBJ explained-gave 
(Ziroo) did (Taroo) a favor of explaining several differences. 
Cbl: TAROO 
Cfl: \[TAROO, ZIROO, DIFFERENCES\] CONTINUE 26 
EMP-OBJ2, SUBJ, OBJ 
Cf2: \[ZIROO, TAROO, DIFFERENCES\] RETAIN 8 
EMP-OBJ2, SUBJ, OBJ 
In 36 as in 35, EMPATHY associates with the previous Cb, ie. Taroo. This follows 
from the ordering of the Cf and hearers' preferences for a CONTINUE interpretation. 
Note that the interpretation of the last utterance in example 36d remains the same 
as that in example 35d, although in this case it is Taroo that shows Ziroo some old 
data in example 36c; nevertheless Ziroo is the one who does the explaining. It seems 
that inference from world knowledge and domain information alone is unlikely to 
predict which interpretations hearers will prefer. Inferential processes and discourse 
structure are mutually constraining (Joshi and Weinstein 1981; Nadathur and Joshi 
1983; Hudson-D'Zmura 1988). 
5.2 Summary 
We proposed a discourse rule of ZERO TOPIC ASSIGNMENT and showed that ZTA is 
conditioned by the rules and constraints of centering theory: (1) ZTA only applies to 
discourse entities that were previously the Cb; (2) ZTA is constrained to cases where 
the only transition available otherwise would be a RETAIN. 
ZTA arises from the interaction between preferences for CONTINUE transitions (Rule 
2) and the fact that Cbs are often zeros (Rule 1). The interaction of these two factors 
leads to the speculation that when the Cb is realized by a pronoun in a lower ranked 
Cf position, which gives rise to a RETAIN transition state, that this type of transition 
is inherently ambiguous. Since different factors contribute to the salience of discourse 
entities, such as 'subjecthood' and 'pronominalization' (Grosz, Joshi, and Weinstein 
unpublished), conflicting defaults can arise when these are in conflict with one another. 
This may be especially true in Japanese since another factor that should contribute to 
Cf ranking, word order, is not present whenever zeros are involved. 
These examples highlight the relation between centering and global coherence 
in discourse. A RETAIN is proposed as a way for a speaker to mark a coordinated 
transition to a new topic; it predicts a shift (Grosz, Joshi, and Weinstein unpublished; 
Brennan, Friedman, and Pollard 1987). However, the way in which centering SHIFT 
transitions are related to larger structures in discourse has not been specified. If a 
221 
Computational Linguistics Volume 20, Number 2 
shift functions as a boundary between segments (Walker 1993b), then the hearer's 
application of ZTA means that the hearer is assuming that the next utterance will be 
part of the same discourse segment. In contrast, a hearer's assumption that the current 
centering transition is a RETAIN means that the hearer assumes that the next utterance 
will begin a new discourse segment. 
The relationship between segmentation and hearer's preferences for ZTA or RE- 
TAIN interpretations may be affected by other discourse factors. Among these factors, 
intonation may indicate whether the current utterance should be taken as initiating a 
new segment and predicting a SHIFT, or continuing the previous one (Silverman 1987; 
Cahn, 1992; Swerts and Geluykens 1992; Walker and Prince 1994). Another factor may 
be the inferred relationship that holds between adjacent utterances such as whether 
it is possible to interpret (d) as Ziroo's reason for having done (c) (Hobbs 1985b). 
However this is clearly not the only factor, or even necessarily the dominant one, as 
we have demonstrated. Future research must provide additional constraints on when 
ZTA is applicable. 
6. Related Research 
Other researchers working on the interpretation of anaphors have focused on the role 
of inference from world knowledge (Hobbs 1985b, 1979). While it is important to eluci- 
date the information needed for inference and the type of inferential process involved 
in discourse interpretation, it is clear from our examples that syntactic realization has 
a strong effect on the interpretive process and provides processing constraints on in- 
ferential processes. We have focused on the interaction between syntax and inference. 
Our treatment of Japanese discourse phenomena builds on earlier work by Kuno 
(Kuno 1972, 1973, 1987, 1989). Our Cf ranking is consistent with Kuno's Empathy and 
Topic Hierarchies and we incorporate a number of Kuno's observations on the function 
of the grammatical topic marker wa and the role of zeros. We have also incorporated 
Kuno's notion of EMPATHY by using EMPATHY in the Cf ranking (Kuno 1976a; Kuno 
and Kaburaki 1977). 
In recent work, Kuno proposes an algorithmic account of the interpretation of 
zeros. He claims that there are two types of zero pronouns, PSEUDO-ZERO-PRONOUNS 
and REAL-ZERO-PRONOUNS (Kuno 1989). REAL-ZERO-PRONOUNS are supposed to have a 
wa-marked NP or a presentational NP as an antecedent (Yoshimoto 1988). PSEUDO-ZERO- 
PRONOUNS are actually examples of deletion, and must follow the same order and the 
same syntactic function as their source NPs. They must obey constraints on deletion 
such as Kuno's Pecking Order of Deletion Principle: Delete less important information 
first and more important information last. According to Kuno, the position just to the 
left of the verb is the default focus position in Japanese, unless the verb itself is the 
focus. Therefore, since the verb in example 37b is the information focus, the zeros are 
assumed to be PSEUDO-ZERO-PRONOUNS. 
Example 37 
a. Taroo ga Hanako ni nani o sita no desu ka. 
Taroo SUBJ Hanako to what OBJ do COMP COPULA Q 
What did Taroo do to Hanako? 
b. 0 0 kisu o sita no desu. 
kiss OBJ did COMP COPULA 
(lit.) (Taroo) did a kissing (to Hanako). 
222 
Marilyn Walker et al. Japanese Discourse 
The combination of these two types of zeros can explain examples like the follow- 
ing: 
Example 38 
a. Taroo wa 
b. 
C. 
Hanako ga sukida. 
Taroo TOP/SUBJ Hanako fond-of-is 
Taroo likes Hanako. 
Ziroo wa Natuko ga sukida. 
Ziroo TOP/SUBJ Natuko fond-of-is 
Ziroo likes Natuko. 
0 Saburoo mo sukida. 
Saburoo also fond-of-is 
(Ziroo) also likes Saburoo. 
*Saburoo also likes (Natuko). 
Kuno's account treats Ziroo in example 38c as a REAL-ZERO-PRONOUN. In this case 
we would predict the preferred interpretation based on our distinction between CON- 
TINUE and RETAIN. However, consider the following example: 
Example 39 
a. Taroo wa 
b. 
Hanako ga sukida. 
Taroo TOP/SUBJ Hanako fond-of-is 
Taroo likes Hanako. 
Ziroo wa kirai da. 
Ziroo TOP/SUBJ 0 fond-of-is 
(Taroo) dislikes Ziroo. 
Ziroo dislikes (Hanako). 
*Ziroo dislikes (Taroo). 
The Taroo dislikes Ziroo interpretation would be an example of ZTA. However, we 
would predict that the Ziroo dislikes Hanako interpretation would be dispreferred, but 
this does not seem to be the case. Kuno's analysis treats the zero in the second reading 
of example 39b as a PSEUDO-ZERO-PRONOUN, which means that it must be interpreted 
as Hanako since Hanako was the object of the previous utterance. 
The interpretation of 39b that we would predict as possible would be the Ziroo 
dislikes Taroo (RETAIN), which native speakers rarely get. However, Kuno's analysis does 
not block this reading either; the zero in 39b could also be a REAL-ZERO-PRONOUN, with 
Taroo as its antecedent. Kuno says that this interpretation is dispreferred because of a 
preference for parallel interpretation (Grober, Beardsley, and Caramazza 1978; Sidner 
1979; Kameyama 1988; Kuno 1989). We have claimed here and elsewhere (Brennan, 
Friedman, and Pollard 1987; Walker, Iida, and Cote 1990) that the preference for paral- 
lelism is an epiphenomenon of the ordering of the Cf and the preference for CONTINUE 
interpretations. 
Our account cannot explain the contrast between examples 38 and 39. It seems that 
what is at issue here is the fact that a set of discourse entities plus an open proposition 
such as X likes Y is what is discourse-old in these examples and not just a discourse 
entity (Prince 1981a, 1986, 1992). Our conclusion is that these enumerated lists and 
question-answer discourse segments may need an account of discourse center that is 
broader than that needed for discourse entities realized as NPS. Ktmo's constraints on 
223 
Computational Linguistics Volume 20, Number 2 
deletion must also be integrated to fully explain when entities or propositions in the 
discourse may be unexpressed. 
Our analysis also builds on an earlier analysis put forth by Kameyama (Kameyama 
1985, 1986, 1988). Although Kameyama uses the centering terminology, her account 
is not based on the constraints and rules of centering theory as developed here and 
presented in (Grosz, Joshi, and Weinstein 1983, unpublished; Brennan, Friedman, and 
Pollard 1987). Kameyama proposed that the interpretation of zeros in Japanese de- 
pends on a default preference hierarchy of syntactic properties to be shared between 
the antecedent and the zero (Grober, Beardsley, and Caramazza 1978). Kameyama's 
account of zero interpretation consists roughly of a PROPERTY-SHARING CONSTRAINT, 
henceforth PS, and an EXPECTED CENTER ORDER, henceforth ECO, which may be para- 
phrased as follows: 
PROPERTY-SHARING CONSTRAINT: Two zero-pronouns in adjacent utter- 
ances, which co-specify the same Cb-encoding discourse entity, should 
share one of the following properties (in descending order of prefer- 
ence): 1) both IDENT and SUBJECT, 2) IDENT alone, 3) SUBJECT-alone, 4) 
both NONIDENT and NONSUBJECT, 5) NONSUBJECT alone, or 6) NONIDENT 
alone. 
EXPECTED CENTER ORDER RULE: In a sentence that contains a center- 
establishing zero, if it is to have a full NP as its antecedent, the default 
preference order among its potential antecedent NPs is: Topic > Ident 
> Subject > Object(2) > Others. 
As noted earlier, we use a modified version of Kameyama's EXPECTED CENTER 
ORDER as the ordering of the Cf, but Kameyama's treatment differs from ours in a 
number of respects. 
First, Kameyama used the property IDENT to describe something similar to Kuno's 
notion of EMPATHY, and has an added assumption of a SUBJECT IDENT default, i.e. 
subjects are considered to be EMPATHY loci by default. This means that her theory also 
includes a neutralization device for cases where this default is not in effect (Kameyama 
1988). In contrast, our theory explains examples covered by the SUBJECT IDENT default 
by including EMPATHY in the ranking of the Cf list and by the distinction between 
CONTINUE and RETAIN, as illustrated in example 24. 
We have also expanded Kameyama's treatment of TOPIC. We have elucidated the 
interaction of topic with subject and empathy markers and supported our claim that 
the topic marker wa functions similarly to pronominalization in instantiating the Cb. 
In addition, ZTA and the distinction that we make between grammatical and zero 
topic is new to our account. 
Furthermore, since Center Instantiation is a side effect of the application of the 
centering algorithm, we treat 40c and 41c with the same mechanism. In Kameyama's 
analysis, the PS constraint applies to example 40, while the ECO applies in example 41. 
Example 40 
a. Hanako wa 
b. 
repooto o kakimasita. 
Hanako TOP/SUBJ report OBJ wrote 
Hanako wrote a report. 
0 Taroo ni aini-ikimasita. 
SUBJ-IDENT Taroo OBJ2 see-went 
She went to see Taroo. 
224 
Marilyn Walker et al. Japanese Discourse 
C. Taroo wa 0 kibisiku hihansimasita. 
Taroo TOP/SUBJ OBJ severely criticized 
Taroo severely criticized her. 
Example 41 
a. Hanako wa Taroo ni aini-kimasita. 
Hanako TOP/SUBJ Taroo OBJ2 see-came 
Hanako came to see Taroo. 
b. Taroo wa 0 hon o yonde-kure-masita. 
Taroo TOP/SUBJ OBJ2 book OBI read-gave 
Taroo did her a flavor of reading a book. 
Note that we annotate example 40b with Kameyama's IDENT property, which cor- 
responds to EMPATHY. Kameyama's account predicts that there are different processes 
going on in the resolution of zeros depending on the environments where the zero 
appears. PS applies in the case of example 40c because the previous utterance has a 
zero, but it doesn't apply in the case of example 41b. PS would seem to predict that 
the zero pronoun in 40c should not be interpreted as Hanako, since the zero carries the 
properties \[SuBJ, IDENT\] in 40b and \[NONSUBJ, NONIDENT\] in 40c. In other words, none 
of the required properties of SUBJ, IDENT, NONSUBJ, NONIDENT, which 'should' be shared 
according to the PS constraint, are shared. But in fact 40c is perfectly acceptable under 
the intended reading of Taroo severely criticized Hanako, and 41b is likewise acceptable 
under the reading Taroo did Hanako a favor of reading a book. 
Also, as pointed out in Kuno (1989), Kameyama's theory makes no predictions 
about the interpretation of some of the zeros in examples such as 5, repeated here for 
convenience as example 42. 
Example 42 
a. Taroo wa 
b. 
C. 
saisin no konpyuutaa o kaimasita. 
TOP/SUBJ newest of computer OBJ bought 
Taroo bought a new model of computer. 
Cb: TAROO 
Cf: \[TAROO, COMPUTER\] 
0 John ni sassoku sore o misemasita. 
SUBJ John OBJ2 at once that OBJ showed 
(Taroo) showed it to John. 
Cb: TAROO 
Cf: \[TAROO, JOHN, COMPUTER\] CONTINUE 
0 0 atarasiku sonawatta kinoo o setumeisimasita. 
SUBJ OBJ2 newly equipped function OBJ explained 
(Taroo) explained the newly equipped functions to (John). 
Cb: TAROO 
Cfl: \[TAROO, JOHN\] CONTINUE 27 
SUBJ OBJ 
Cf2: \[JOHN, TAROO\] RETAIN 1 
SUBJ OBJ 
225 
Computational Linguistics Volume 20, Number 2 
The PS Constraint applies only to two zeros in adjacent sentences, and the ECO 
applies only when a Cb is to be established. Example 42c is not a Cb-establishing 
utterance since the Cb ha s already been established in 42b, so the ECO should not 
apply. The PS constraint does apply and predicts that the subject zero must have the 
subject of 42b as its antecedent, but the theory makes no predictions about the possible 
interpretations for the zero object. 
Many of the examples that are explained in Kameyama's theory by the PS con- 
straint are handled on our account by the distinction between CONTINUE and RETAIN. 
However, there are a number of cases where PS makes different predictions than our 
account. In particular, note that for examples 32 and 35, Kameyama's SUBJECT IDENT 
default makes exactly the opposite prediction. Example 35 is repeated below as exam- 
ple 43 and annotated with the SUBJECT IDENT default feature. 
Example 43 
a. Taroo wa deeta o konpyuutaa ni utikondeimasita. 
Taroo TOP/SUBJ data OBJ computer in was-storing 
Taroo was storing the data in a computer. 
b. 0 yatto hanbun yari-owarimasita. 
SUBJ/IDENT finally half do-finished 
Finally he was half Ji"nished. 
C. Ziroo ga 0 hurui deeta o misemasita. 
Ziroo SUBJ/IDENT OBJ2 old data OBJ showed 
Ziroo showed him some old data. 
d. 0 0 ikutuka no kuitigai o setumeisite-KURE-masita. 
SUBJ OBJ2/IDENT several of differences OBJ explained-gave 
(Ziroo) did (Taroo) a favor of explaining several di~erences. 
According to PS, the interpretation in which the property IDENT is shared is pre- 
ferred to the one with SUBJECT shared, and hence, the interpretation Taroo did Ziroo a 
favor in explaining several d~ferences is preferred. However our survey shows that native 
speakers prefer the Ziroo did Taroo a favor reading; this is explained by our discourse 
rule of ZTA and by including empathy in the ranking of the Cf list. 
7. Conclusion and Future Work 
In this paper, we have attempted to elucidate the interaction of syntactic realization 
and discourse salience in Japanese using the discourse-processing framework of CEN- 
TERING. In our theory discourse salience is operationalized by the ranking of the for- 
ward centers for an utterance. We explored speakers' options for indicating salience 
in Japanese discourse, especially the interaction of discourse markers for TOPIC and 
EMPATHY. We then posited a ranking and used it to explain some facts about the in- 
terpretation of zeros in Japanese. 
While there is clearly a correlation between syntax and discourse function, we 
show that discourse context plays an important role. We proposed a discourse rule of 
ZERO TOPIC ASSIGNMENT (ZTA), which distinguishes grammatical and zero topic. We 
showed that centering allows us to formalize constraints on when ZTA can apply. 
However, future work must determine additional constraints on when ZTA applies, 
and which language families support ZTA. 
226 
Marilyn Walker et al. Japanese Discourse 
The preferred interpretation of zeros and the discourse factors that are responsible 
for each interpretation are summarized below. Remember that in each case the zero 
in the third utterance was established as the Cb by the previous two utterances: 
Third Utterance Fourth Utterance Discourse Factor Example 
SUBJECT OBJECT(2) SUBJECT OBJECT(2) 
zero(i) NP(j) zero(i) zero(j) Continue/Retain 5 
zero(i) NP(j) 
NP(ga)(i) zero(j) 
NP(wa)(i) zero(j) 
NP(ga)(i) zero(j) 
zero(j) zero(i),empathy 
zero(j) 
zero(i) 
zero(j) 
zero(i) 
empathy, Continue/Retain 36 
zero(i) ZTA 32, 34 
33 zero(j) 
zero(i) 
WA-effect 
ZTA 
ZTA and empathy zero(j),empathy 35 
This analysis suggests that centering may be a universal of context-dependent 
processing of language, although so far this theory has only been applied to English, 
German, Turkish, Japanese, and Italian (Brennan, Friedman, and Pollard 1987; Walker 
1989; Walker, Iida, and Cote 1990; Di Eugenio 1990; Cote 1992; Rambow 1993; Nakatani 
1993; Hofhnan 1995; Turan 1995). We proposed that the centering component of a the- 
ory of discourse interpretation can be constructed in a language-independent fashion, 
up to the declaration of a language-specific value for one parameter of the theory, i.e., 
Cf ranking (as in Section 2). This parameter is language-dependent because different 
languages offer different means of expressing discourse function. We conjecture that 
ZTA may apply in any free-word order language with zeros. 
Future work must examine the interaction between centering and discourse seg- 
mentation in both monologue and dialogue (Whittaker and Stenton 1988; Walker and 
Whittaker 1990; Walker 1993b), and the role of deictics, lexical semantics, one anaphora, 
and propositional discourse entities in centering (Webber 1978; Sidner 1979; Walker 
1992, 1993a; Cote 1995). It is also important to examine the interaction of zeros with 
overt pronouns and with deictics and the interaction of pronominalization with ac- 
centing (Terken 1995). In addition, the semantic theory underlying centering must be 
further developed (Roberts 1995). Finally, centering transitions are currently defined 
by an equality relation between discourse entities, but POSET relations and functional 
dependencies often link entities in discourse (Prince 1978b, 1981a; Ward 1985; Grosz, 
Joshi, and Weinstein unpublished). The predictions made here should also be tested 
on a large corpus of naturally occurring Japanese discourse (Hurewitz and Linson 
1995). 
Acknowledgments 
We would like to thank Aravind Joshi and 
Ellen Prince for their insight, useful 
discussions, and support for making it 
possible to have the workshop on Centering 
Theory in Naturally-Occurring Discourse at 
the Institute for Cognitive Science at the 
University of Pennsylvania. In addition, 
discussions with Dave Bernstein, Susan 
Brennan, Hitoshi Isahara, Megumi 
Kameyama, Susumu Kuno, Christine 
Nakatani, Hiday Nakashima, Carl Pollard, 
Owen Rainbow, Peter Sells, Mike 
Tanenhaus, Bonnie Webber, and Steve 
Whittaker contributed to the development 
of this work. We would also like to thank 
the anonymous reviewers who provided 
many helpful suggestions. NSF's Summer 
Science and Engineering Institute in Japan 
made it possible to present this work and 
receive useful feedback at ICOT, JEIDA 
working group on Machine Translation, 
NTT's Basic Research Labs, and ATR's 
Interpreting Telephony Lab. We would also 
227 
Computational Linguistics Volume 20, Number 2 
like to thank the readers of 
comp.research.japanese and sci.lang.japan 
who participated in our survey. 
This research was partially funded by 
NSF Science and Engineering Award for the 
Summer Institute in Japan, by ARO grant 
DAAL03-89-C-0031, DARPA grant 
N00014-90-J-1863, NSF grant IRI90-16592, 
and Ben Franklin grant 91S.3078C-1 at the 
University of Pennsylvania, and by Hewlett 
Packard Laboratories. 
References 
Brennan, Susan E. (submitted). Centering 
attention in discourse. 
Brennan, Susan E.; Friedman \[Walker\], 
Marilyn W.; and Pollard, Carl J. (1987). 
"A centering approach to pronouns." In 
Proceedings, 25th Annual Meeting of the ACL, 
Stanford, 155-162. 
Cahn, Janet (1992). An investigation into the 
correlation of cue phrases, unfilled pauses 
and the structuring of spoken discourse. 
In Workshop on Prosody in Natural Speech, 
19-31. Institute for Research in Cognitive 
Science, University of Pennsylvania, TR 
IRCS-92-37. 
Carter, David M. (1987). Interpreting 
Anaphors in Natural Language Texts. Ellis 
Horwood. 
Clancy, Patricia M., and Downing, Pamela 
(1987). "The use of wa as a cohesion 
marker in Japanese oral narratives." In 
Perspectives in Topicalization: The Case of 
Japanese Wa, 3-56. Academic Press. 
Clark, Herbert H., and Haviland, Susan E. 
(1977). "Comprehension and the 
given-new contract." Discourse Production 
and Comprehension, edited by 
R. O. Freedle, 1-40. Ablex Publishing. 
Clark, Herbert H., and Marshall, Catherine 
R. (1981). "Definite reference and mutual 
knowledge." In Elements of Discourse 
Understanding, edited by Joshi, Webber, 
and Sag, 10-63. Cambridge University 
Press. 
Cote, Sharon (1992). "Discourse functions of 
two types of null objects in English." In 
Linguistic Society of America Annual 
Meeting, 12. 
Cote, Sharon (in press). Ranking 
forward-looking centers. In Marilyn 
A. Walker, Aravind K. Joshi, and Ellen 
E Prince, editors, Centering in Discourse. 
Oxford University Press. 
Di Eugenio, Barbara (1990). "Centering 
theory and the italian pronominal 
system." In Proceedings, 13th International 
Conference on Computational Linguistics, 
(COLING-90), 270-275. 
Gordon, Peter C.; Grosz, Barbara J.; and 
Gilliom, Laura A. (1993). "Pronouns, 
names and the centering of attention in 
discourse." Cognitive Science, 17-3, 
311-348. 
Grober, Ellen H.; Beardsley, William; and 
Caramazza, Alfonso (1978). "Parallel 
function strategy in pronoun 
assignment." Cognition, 6, 117-131. 
Grosz, Barbara J. (1977). "The representation 
and use of focus in dialogue 
understanding." Technical Report 151, SRI 
International, 333 Ravenswood Ave, 
Menlo Park CA 94025. 
Grosz, Barbara J., and Sidner, Candace L. 
(1986). "Attentions, intentions and the 
structure of discourse." Computational 
Linguistics, 12, 175-204. 
Grosz, Barbara J.; Joshi, Aravind K.; and 
Weinstein, Scott (1983). "Providing a 
unified account of definite noun phrases 
in discourse." In Proceedings, 21st Annual 
Meeting of the Association for Computational 
Linguistics, 44-50. 
Grosz, Barbara J.; Joshi, Aravind K.; and 
Weinstein, Scott (unpublished). Towards a 
computational theory of discourse 
interpretation. 
Hajicova, Eva, and Vrbova, Jarka (1982). 
"On the role of the hierarchy of activation 
in the process of natural language 
understanding." In Proceedings, 9th 
International Conference on Computational 
Linguistics (COLING-84). Prague, 107-113. 
Hasegawa, Nobuko (1984). "On the 
so-called zero pronouns in Japanese." The 
Linguistic Review, 289-341. 
Hobbs, Jerry R. (1976a). "A computational 
approach to discourse analysis." Technical 
Report 76-2, Department of Computer 
Science, City College, City University of 
New York. 
Hobbs, Jerry R. (1976b). "Pronoun 
resolution." Technical Report 76-1, 
Department of Computer Science, City 
College, City University of New York. 
Hobbs, Jerry R. (1979). "Coherence and 
coreference." Coginitive Science, 3, 67-90. 
Hobbs, Jerry R. (1985a). "The logical 
notation: Ontological promiscuity," 
Chapter 2 of Discourse and Inference. 
Technical report, SRI International, 333 
Ravenswood Ave., Menlo Park CA 94025. 
Hobbs, Jerry R. (1985b). "On the coherence 
and structure of discourse." Technical 
Report CSLI-85-37, Center for the Study 
of Language and Information, Ventura 
Hall, Stanford University. 
Hobbs, Jerry R., and Martin, Paul (1987). 
"Local pragmatics." Technical report, SRI 
International, 333 Ravenswood Ave., 
228 
Marilyn Walker et al. Japanese Discourse 
Menlo Park CA 94025. 
Hobbs, Jerry R.; Croft, William; Davies, 
Todd; Edwards, Douglas; and Laws, 
Kenneth (1987). "The tacitus 
commonsense knowledge base." Technical 
report, SRI International, 333 
Ravenswood Ave., Menlo Park CA 94025. 
Hoffman, Beryl (in press). Word order, 
information structure and centering in 
turkish. In Marilyn A. Walker, Aravind 
K. Joshi, and Ellen F. Prince, editors, 
Centering in Discourse. Oxford University 
Press. 
Horn, Laurence R. (1986). "Presupposition, 
theme and variations." In Chicago 
Linguistic Society, 22, 168-192. 
Hudson-D'Zmura, Susan, and Tanenhaus, 
Michael K. (in press). Assigning 
antecedents to ambiguous pronouns: The 
role of the center of attention as the 
default assignment. In Marilyn A. Walker, 
Aravind K. Joshi, and Ellen F. Prince, 
editors, Centering in Discourse. Oxford 
University Press. 
Hudson-D'Zmura, Susan B. (1988). The 
structure of discourse and anaphor resolution: 
The discourse center and the roles of nouns and 
pronouns. Doctoral dissertation. University 
of Rochester. 
Hurewitz, Felicia, and Linson, Brian (in 
press). A quantitative look at discourse 
coherence. In Marilyn A. Walker, Aravind 
K. Joshi, and Ellen F. Prince, editors, 
Centering in Discourse. Oxford University 
Press. 
Iida, Masayo (1993). Context and binding in 
Japanese. Doctoral dissertation, Linguistics 
Department, Stanford University. 
Joshi, Aravind K. (1982). "Mutual beliefs in 
question-answer systems." In Mutual 
Knowledge, edited by Nell V. Smith, 
181-199. Academic Press. 
Joshi, Aravind K., and Kuhn, Steve (1979). 
"Centered logic: The role of entity 
centered sentence representation in 
natural language inferencing." In 
Proceedings, International Joint Conference on 
Artificial Intelligence, 435-439. 
Joshi, Aravind K., and Weinstein, Scott 
(1981). "Control of inference: Role of some 
aspects of discourse structure centering." 
In Proceeding S International Joint Conference 
on Artificial Intelligence, 385--387. 
Kameyama, Megumi (1985). Zero anaphora: 
the case of Japanese. Doctoral dissertation, 
Linguistics Department, Stanford 
University. 
Kameyama, Megumi (1986). "A 
property-sharing constraint in centering." 
In Proceedings, 24th Annual Meeting of the 
Association for Computational Linguistics, 
New York, 200-206. 
Kameyama, Megumi (1988). "Japanese zero 
pronominal binding, where syntax and 
discourse meet." In Papers from the Second 
International Workshop on Japanese Syntax, 
edited by William Poser, 47-74. Stanford: 
CSLI. 
Kuno, Susumu (1972). "Pronominalization, 
reflexivization, and direct discourse." 
Linguistic Inquiry, 3, 161-195. 
Kuno, Susumu (1973). The Structure of the 
Japanese Language. The MIT Press. 
Kuno, Susumu (1976a). "The speaker's 
empathy and its effects of syntax: A 
re-examination of yaru and kureru in 
Japanese." Journal of the Association of 
Teachers of Japanese, 11, 249-271. 
Kuno, Susumu (1976b). "Subject, theme and 
speaker's empathy: A reexamination of 
relativization phenomena." In Subject and 
Topic, edited by C. Li, 417-444. Academic 
Press. 
Kuno, Susumu (1987). Functional Syntax. 
Chicago University Press. 
Kuno, Susumu (1989). "Identification of 
zero-pronominal reference in Japanese." 
Unpublished manuscript. Presented at 
ATR workshop on Machine Translation. 
Kuno, Susumu, and Kaburaki, Etsuko 
(1977). "Empathy and syntax." Linguistic 
Inquiry, 8, 627--672. 
Kuroda, Sige-Yuki (1965). Generative 
grammatical studies in the Japanese language. 
Doctoral dissertation, MIT. 
Martin, Samuel (1976). A Reference Grammar 
of Japanese. Yale University Press. 
Nadathur, Gopalan, and Joshi, Aravind K. 
(1983). "Mutual beliefs in conversational 
systems: Their role in referring 
expressions." In Proceedings, International 
Joint Conference on Artificial Intelligence, 
Austin, TX, 603-605. 
Nakagawa, Hiroshi (1992). "Zero pronouns 
as experiencer in Japanese discourse." In 
Proceedings, Fourteenth International 
Conference on Computational Linguistics, 
324--330. 
Nakashima, Hideyuki (1990). 
"Commonsense reasoning with multiply 
embedded contexts." In Pacific Rim 
Conference on Artificial Intelligence, 414-419. 
Nakatani, Christine H. (1993). Global and 
local functions of topic marking in 
japanese. In Institute for Research in 
Cognitive Science Workshop on Centering 
Theory in Natural-Occurring Discourse. 
Prince, Ellen F. (1978a). "On the function of 
existential presupposition in discourse." 
In Papers from the 14th Regional Meeting. 
CLS, Chicago IL. 
Prince, Ellen E (1978b). "On the function of 
229 
Computational Linguistics Volume 20, Number 2 
wh-clefts and it-clefts in discourse." 
Language, 54, 883-906. 
Prince, Ellen E (1981a). Topicalization, focus 
movement and Yiddish movement: A 
pragmatic differentiation. Proceedings of the 
Seventh Annual Meeting of the Berkely 
Linguistics Society, 249-264, edited by 
Danny K. Alford, Karen Ann Hunold, 
Monica A. Macaulay, Jenny Walter, 
Claudia Brugman, Paula Chertok, Inese 
Civkulis, and Marta Tobey. 
Prince, Ellen F. (1981b). "Toward a 
taxonomy of given-new information." In 
Radical Pragmatics, 223-255. Academic 
Press. 
Prince, Ellen F. (1985). "Fancy syntax and 
shared knowledge." Journal of Pragmatics, 
65-81. 
Prince, Ellen F. (1986). "On the syntactic 
marking of the presupposed open 
proposition." Proceedings of the 22nd Annual 
Meeting of the Chicago Linguistic Society. 
Prince, Ellen F. (1992). "The ZPG letter: 
Subjects, definiteness and information 
status." In Discourse Description: Diverse 
Analyses of a Fund Raising Text, edited by 
S. Thompson and W. Mann, 295-325. John 
Benjamins B.V. 
Rambow, Owen (1993). Pragmatic aspects of 
scrambling and topicalization in german. 
In Institute for Research in Cognitive Science 
Workshop on Centering Theory in 
Naturally-Occurring Discourse. 
Reinhart, Tanya (1976). The Syntactic Domain 
of Anaphora. Doctoral dissertation, MIT. 
Rein_hart, Tanya (1981). "Pragmatics and 
linguistics, an analysis of sentence 
topics." Philosophica, 27, 53-94. 
Roberts, Craige (in press). Salience, 
centering and anaphora resolution in 
discourse representation theory. In 
Marilyn A. Walker, Aravind K. Joshi, and 
Ellen F. Prince, editors, Centering in 
Discourse. Oxford University Press. 
Shibatani, Masayoshi (1990). The Languages 
of Japan. Cambridge University Press. 
Sidner, Candace L. (1979). "Toward a 
computational theory of definite 
anaphora comprehension in English." 
Technical Report AI-TR-537, MIT. 
Sidner, Candace L. (1981). "Focusing for 
interpretation of pronouns." American 
Journal of Computational Linguistics, 7(4), 
217-231. 
Sidner, Candace L. (1983). "Focusing in the 
comprehension of definite anaphora." In 
Computational Models of Discourse, edited 
by M. Brady and R. C. Berwick. MIT 
Press. 
Silverman, Kim (1987). The Structure and 
Processing of Fundamental Frequency 
Contours. Doctoral Dissertation, 
Cambridge University. 
Swerts, Marc, and Geluykens, Ronald 
(1992). "The prosodic structuring of 
information flow in spoken discourse." In 
Workshop on Prosody in Natural Speech, 
221-230. Institute for Research in 
Cognitive Science, University of 
Pennsylvania. 
Terazu, Noriko; Yamanasi, Masaaki; and 
Inada, Toshiaki (1980). "Anaphora in 
Japanese." In Studies in English Linguistics, 
8, 32-52. 
Terken, Jacques M. (in press). Accessibility, 
prominence, pronouns and accents. In 
Marilyn A. Walker, Aravind K. Joshi, and 
Ellen F. Prince, editors, Centering in 
Discourse. Oxford University Press. 
Turan, Umit Deniz (in press). Prominence in 
turkish discourse and forward looking 
center hierarchy. In Marilyn A. Walker, 
Aravind K. Joshi, and Ellen F. Prince, 
editors, Centering in Discourse. Oxford 
University Press. 
Walker, Marilyn A. (1989). "Evaluating 
discourse processing algorithms." In 
Proceedings, 27th Annual Meeting of the 
Association of Computational Linguistics, 
251-261. 
Walker, Marilyn A. (1992). "Redundancy in 
collaborative dialogue." In Proceedings, 
Fourteenth International Conference on 
Computational Linguistics, 345-351. 
Walker, Marilyn A. (1993a). Informational 
Redundancy and Resource Bounds in 
Dialogue. Ph.D. thesis, University of 
Pennsylvania. 
Walker, Marilyn A. (1993b). Initial contexts 
and shifting centers. In Institute for 
Research in Cognitive Science Workshop on 
Centering Theory in Naturally-Occuring 
Discourse. 
Walker, Marilyn A., and Whittaker, Steve 
(1990). "Mixed initiative in dialogue: An 
investigation into discourse 
segmentation." In Proceedings, 28th Annual 
Meeting of the Association for Computational 
Linguistics, 70-79. 
Walker, Marilyn A.; Iida, Masayo; and Cote, 
Sharon (1990). "Centering in Japanese 
discourse." In Proceedings, 13th 
International Conference on Computational 
Linguistics (COLING-90), Helsinki, 1. 
Ward, Gregory (1985). The syntax and 
semantics of preposing. Doctoral 
dissertation, University of Pennsylvania. 
Webber, Bonnie Lynn (1978). A formal 
approach to discourse anaphora. Doctoral 
dissertation, Harvard University. Garland 
Press. 
Whittaker, Steve, and Stenton, Phil (1988). 
230 
Marilyn Walker et al. Japanese Discourse 
Cues and control in expert client 
dialogues. In Proc. 26th Annual Meeting of 
the ACL, Association off Computational 
Linguistics, pages 123--130. 
Yoshimoto, Kei (1988). "Identifying zero 
pronouns in Japanese dialogue." In 
Proceedings, 12th International Conference on 
Computational Linguistics (COLING-88), 
Budapest, Hungary, 779-784. 
231 
Computational Linguistics Volume 20, Number 2 
Appendix A: Instructions to Survey Participants 
Instructions for Survey I and 2 
What interpretation do you get for the THIRD sentence of each set where there are two 
unexpressed arguments? 0(i) in the second sentence indicates that the unexpressed 
argument in the sentence should be interpreted as referring to the NP of the first 
sentence marked with (i). Please rank your preference: it's okay to have more than 
one equally preferred interpretation. 
Instructions for Survey 3 
Dear Participants. Thank you for serving as subjects for us for this informal experi- 
ment. You can help us most by following the directions here. Please read each sample 
discourse in turn and make your interpretation as rapidly as possible. Do not scroll 
back and forth in the file. Please indicate which interpretation, (a) or (b) you get by 
marking your choice with a 1. It is very important that you choose ,one, interpreta- 
tion only, and the one you choose should be the first one that you think of as you are 
reading the sample discourse. Send us back this file with your choices marked. 
232 
