Centering: A Framework for Modeling the 
Local Coherence of Discourse 
Barbara J. Grosz* 
Harvard University 
Scott Weinstein ~ 
University of Pennsylvania 
Aravind K. Joshi t 
University of Pennsylvania 
This paper concerns relationships among focus of attention, choice of referring expression, and 
perceived coherence of utterances within a discourse segment. It presents a framework and ini- 
tial theory of centering intended to model the local component of attentional state. The paper 
examines interactions between local coherence and choice of referring expressions; it argues that 
differences in coherence correspond in part to the inference demands made by different types of 
referring expressions, given a particular attentional state. It demonstrates that the attentional 
state properties modeled by centering can account for these differences. 
Preface 
Our original paper (Grosz, Joshi, and Weinstein 1983) on centering claimed that certain 
entities mentioned in an utterance were more central than others and that this property 
imposed constraints on a speaker's use of different types of referring expressions. 
Centering was proposed as a model that accounted for this phenomenon. We argued 
that the coherence of discourse was affected by the compatibility between centering 
properties of an utterance and choice of referring expression. Subsequently, we revised 
and expanded the ideas presented therein. We defined various centering constructs 
and proposed two centering rules in terms of these constructs. A draft manuscript 
describing this elaborated centering framework and presenting some initial theoretical 
claims has been in wide circulation since 1986. This draft (Grosz, Joshi, and Weinstein 
1986) has led to a number of papers by others on this topic and has been extensively 
cited, but has never been published. 1 
We have been urged to publish the more detailed description of the centering 
framework and theory proposed in Grosz, Joshi, and Weinstein (1986) so that an of- 
ficial version would be archivally available. The task of completing and revising this 
draft became more daunting as time passed and more and more papers appeared on 
centering. Many of these papers proposed extensions to or revisions of the theory 
and attempted to answer questions posed in Grosz, Joshi, and Weinstein (1986). It has 
become ever more clear that it would be useful to have a "definitive" statement of 
* Division of Applied Sciences, Harvard University, Cambridge, MA. E-mail: grosz@das.harvard.edu. t Computer and Information Science, University of Pennsylvania, Philadelphia, PA. E-mail: 
joshi@linc.cis.upenn.edu. :~ Department of Philosophy, University of Pennsylvania, Philadelphia, PA. E-mail: 
weinstei@linc.cis.upenn.edu. 1 Early drafts of Grosz, Joshi, and Weinstein (1986) were in circulation from 1983. Some citations to other 
work have dates between 1983 and 1986. This work utilized these earlier drafts. 
@ 1995 Association for Computational Linguistics 
Computational Linguistics Volume 21, Number 2 
the original motivations for centering, the basic definitions underlying the centering 
framework, and the original theoretical claims. This paper attempts to meet that need. 
To accomplish this goal, we have chosen to remove descriptions of many open research 
questions posed in Grosz, Joshi, and Weinstein (1986) as well as solutions that were 
only partially developed. We have also greatly shortened the discussion of criteria for 
and constraints on a possible semantic theory as a foundation for this work. 
1. Introduction 
This paper presents an initial attempt to develop a theory that relates focus of atten- 
tion, choice of referring expression, and perceived coherence of utterances within a 
discourse segment. The research described here is a further development of several 
strands of previous research. It fits within a larger effort to provide an overall theory 
of discourse structure and meaning. In this section we describe the larger research 
context of this work and then briefly discuss the previous work that led to it. 
Centering fits within the theory of discourse structure developed by Grosz and 
Sidner (1986). Grosz and Sidner distinguish among three components of discourse 
structure: a linguistic structure, an intentional structure, and an attentional state. At 
the level of linguistic structure, discourses divide into constituent discourse segments; 
an embedding relationship may hold between two segments. The intentional struc- 
ture comprises intentions and relations among them. The intentions provide the basic 
rationale for the discourse, and the relations represent the connections among these in- 
tentions. Attentional state models the discourse participants' focus of attention at any 
given point in the discourse. Changes in attentional state depend on the intentional 
structure and on properties of the utterances in the linguistic structure. 
Each discourse segment exhibits both local coherence--i.e., coherence among the 
utterances in that segment--and global coherence--i.e., coherence with other segments 
in the discourse. Corresponding to these two levels of coherence are two components of 
attentional state; the local level models changes in attentional state within a discourse 
segment, and the global level models attentional state properties at the intersegmental 
level. 
Grosz and Sidner argue that global coherence depends on the intentional structure. 
They propose that each discourse has an overall communicative purpose, the discourse 
purpose (DP); and each discourse segment has an associated intention, its discourse 
segment purpose (DSP). The DP and DSP are speaker intentions; they are correlates at 
the discourse level of the intentions Grice argued underlay utterance meaning (Grice 
1969). If a discourse is multi-party (e.g., a dialogue), then the DSP for a given segment 
is an intention of the conversational participant who initiates that segment. Lochbaum 
(1994) employs collaborative plans (Grosz and Kraus 1993) to model intentional struc- 
ture, and is thus able to integrate intentions of different participants. Satisfaction of the 
DSPs contributes to the satisfaction of the DP. Relationships between DSPs provide the 
basic structural relationships for the discourse; embeddings in the linguistic structure 
are derived from these relationships. The global coherence of a discourse depends on 
relationships among its DP and DSPs. Grosz and Sidner model the global-level com- 
ponent of the attentional state with a stack; pushes and pops of focus spaces on the 
stack depend on intentional relationships. 
This paper is concerned with local coherence and its relationship to attentional 
state at the local level. Centering is proposed as a model of the local-level component 
of attentional state. We examine the interactions between local coherence and choices 
of referring expressions, and argue that differences in coherence correspond in part to 
the different demands for inference made by different types of referring expressions, 
204 
Barbara J. Grosz et al. Centering 
given a particular attentional state. We describe how the attentional state properties 
modeled by centering can account for these differences. 
Three pieces of previous research provide the background for this work. Grosz 
(1977) defined two levels of focusing in discourse: global and immediate. Participants 
were said to be globally focused on a set of entities relevant to the overall discourse. 
These entities may either have been explicitly introduced into the discourse or suffi- 
ciently closely related to such entities to be considered implicitly in focus (Grosz 1981). 
In contrast, immediate focusing referred to a more local focusing process--one that 
relates to identifying the entity that an individual utterance most centrally concerns. 
Sidner (1979) provided a detailed analysis of immediate focusing, including a 
distinction between the current discourse focus and potential foci. She gave algorithms 
for tracking immediate focus and rules that stated how the immediate focus could be 
used to identify the referents of pronouns and demonstrative noun phrases (e.g., "this 
party, .... that party"). 
Joshi and Kuhn (1979) and Joshi and Weinstein (1981) provided initial results on 
the connection between changes in immediate focus and the complexity of inferences 
required to integrate a representation of the meaning of an individual utterance into 
a representation of the meaning of the discourse of which it was a part. To avoid 
confusion with previous uses of the term "focus" in linguistics, they introduced the 
centering terminology. Their notions of "forward-looking" and "backward-looking" 
centers correspond approximately to Sidner's potential foci and discourse focus. 
In all of this work, focusing, whether global or immediate, was seen to function to 
limit the inferences required for understanding utterances in a discourse. Grosz and 
Sidner were concerned with the inferences needed to interpret anaphoric expressions 
of various sorts (e.g. pronouns, definite descriptions, ellipsis). They used focusing to 
order candidates; as a result the need for search was greatly reduced and the use 
of inference could be restricted to determining whether a particular candidate was 
appropriate given the embedding utterance interpretation. Joshi, Kuhn, and Weinstein 
were concerned with reducing the inferences required to integrate utterance meaning 
into discourse meaning. They used centering to determine an almost monadic predicate 
representation of an utterance in discourse; they then used this representation to reduce 
the complexity of inference. 
In this paper, we generalize and clarify certain of Sidner's results, but adopt the 
"centering" terminology. We also abstract from Sidner's focusing algorithm to specify 
constraints on the centering process. We consider the relationship between coherence 
and inference load and examine how both interact with attentional state and choices 
in linguistic expression. 
The remainder of this paper is organized as follows: in Section 2, we briefly de- 
scribe the phenomena motivating the development of centering that this paper aims 
to explain. Section 3 provides the basic definitions of centers and related definitions 
needed to present the theoretical claims of the paper. In Section 4, we state the main 
properties of the centering framework and the major claims of centering theory. In 
Section 5, we discuss several factors that affect centering constraints and govern the 
centering rules given in Section 6. In Section 7, we discuss applications of the rules 
and their ability to explain several discourse coherence phenomena. In Section 8, we 
briefly outline the properties of an underlying semantic framework that are required 
by centering. Finally, in Section 9 we conclude with a brief comparison of center- 
ing with the research that preceded it and a summary of research that expands on 
Grosz, Joshi, and Weinstein (1986). In particular, Section 9 provides references to sub- 
sequent investigations of additional factors that control centering and examinations of 
its crossqinguistic applicability and empirical validity. 
205 
Computational Linguistics Volume 21, Number 2 
2. Phenomena To Be Explained 
Discourses are more than mere sequences of utterances. For a sequence of utterances 
to be a discourse, it must exhibit coherence. In this paper, we investigate linguistic 
and attentional state factors that contribute to coherence among utterances within a 
discourse segment. These factors contribute to the difference in coherence between the 
following two discourse segments: 2 
(1) a. John went to his favorite music store to buy a piano. 
b. He had frequented the store for many years. 
c. He was excited that he could finally buy a piano. 
d. He arrived just as the store was closing for the day. 
(2) a. John went to his favorite music store to buy a piano. 
b. It was a store John had frequented for many years. 
c. He was excited that he could finally buy a piano. 
d. It was closing just as John arrived. 
Discourse (1) is intuitively more coherent than Discourse (2). This difference may 
be seen to arise from different degrees of continuity in what the discourse is about. 
Discourse (1) centers around a single individual, describing various actions he took 
and his reactions to them. In contrast, Discourse (2) seems to flip back and forth among 
several different entities. 
More specifically, the initial utterance (a) in each segment could begin a segment 
about an individual named 'John' or one about John's favorite music store or one 
about the fact that John wants to buy a piano. 
Whereas Discourse (1) is clearly about John, Discourse (2) has no single clear 
center of attention. Utterance (2b) seems to be about the store. If a reader inferred 
that utterance (2a) was about John, then that reader would perceive a change in the 
entity which the discourse seems to be about in going from (2a) to (2b); on the other 
hand, if the reader took (2a) to be about the store then in going to (2b), there is no 
change. In either case, in utterance (2c) John seems to be central, requiring a shift from 
utterance (2b), while the store becomes central again in utterance (2d), requiring yet 
another shift. This changing of 'aboutness' (in fact, flipping it back and forth) makes 
discourse (2) less coherent than discourse (1). 
Discourses (1) and (2) convey the same information, but in different ways. They 
differ not in content or what is said, but in expression or how it is said. The variation in 
'aboutness' they exhibit arises from different choices of the way in which they express 
the same propositional content. The differences can only be explained, however, by 
looking beyond the surface form of the utterances in the discourse; different types of 
referring expressions and different syntactic forms make different inference demands 
on a hearer or reader. These differences in inference load underlie certain differences 
2 This example and the others in this paper are single-speaker texts. However, centering also applies to 
dialogue and multi-party conversations. Issues of the interaction between turn-taking and changes in 
centering status remain to be investigated. 
206 
Barbara J. Grosz et al. Centering 
in coherence. The model of local attentional state described in this paper provides a 
basis for explaining these differences. 
Thus, the focus of our investigation is on interactions among choice of referring 
expression, attentional state, the inferences required to determine the interpretation of 
an utterance in a discourse segment, and coherence. Pronouns and definite descriptions 
are not equivalent with respect to their effect on coherence. We conjecture that this is 
so because they engender different inferences on the part of a hearer or reader. In the 
most pronounced cases, the wrong choice will mislead a hearer and force backtracking 
to a correct interpretation. 3 The following variations of a discourse sequence illustrate 
this problem and provide additional evidence for our conjecture. 
(3) a. Terry really goofs sometimes. 
b. Yesterday was a beautiful day and he was excited about trying out 
his new sailboat. 
c. He wanted Tony to join him on a sailing expedition. 
d. He called him at 6 AM. 
e. He was sick and furious at being woken up so early. 
By using a pronoun to refer to Tony in utterance (e) the speaker may confuse the 
hearer. Through utterance (d) Terry has been the center of attention, and hence is the 
most likely referent of "he" in utterance (e). It is only when one gets to the word 
"sick" that it is clear that it must be Tony and not Terry who is sick, and hence that 
the pronoun in utterance (e) refers to Tony not Terry. A much more natural sequence 
results if "Tony" is used, as the sequence (4a)-(4e) illustrates. 
(4) a. Terry really goofs sometimes. 
b. Yesterday was a beautiful day and he was excited about trying out 
his new sailboat. 
c. He wanted Tony to join him on a sailing expedition. 
d. He called him at 6 AM. 
e. Tony was sick and furious at being woken up so early. 
f. He told Terry to get lost and hung up. 
g. Of course, he hadn't intended to upset Tony. 
In Discourse (4), utterances (f) and (g) exhibit the same kind of misdirection as do 
utterances (3d) and (3e) in Discourse (3). The focus has shifted from Terry to Tony in 
the short subsegment of utterances (e)-(f) so that use of "he" in (g) is confusing. This 
confusion is avoided in the sequence of Discourse (5). 
(5) a. Terry really goofs sometimes. 
b. Yesterday was a beautiful day and he was excited about trying out 
his new sailboat. 
3 We presume utterances are processed in left-to-right order, and that speakers make initial assignments 
of referent and meaning that may have to be retracted if material coming later in the sentence conflicts. 
207 
Computational Linguistics Volume 21, Number 2 
c. He wanted Tony to join him on a sailing expedition. 
d. He called him at 6 AM. 
e. Tony was sick and furious at being woken up so early. 
f. He told Terry to get lost and hung up. 
g. Of course, Terry hadn't intended to upset Tony. 
We conjecture that the form of expression in a discourse substantially affects the re- 
source demands made upon a hearer in discourse processing and through this influ- 
ences the perceived coherence of the discourse. It is well known from the study of 
complexity theory that the manner in which a class of problems is represented can 
significantly affect the time or space resources required by any procedure that solves 
the problem. Here too we conjecture that the manner, i.e., linguistic form, in which 
a discourse represents a particular propositional content can affect the resources re- 
quired by any procedure that processes that discourse. We use the phrase inference load 
placed upon the hearer to refer to the resources required to extract information from a 
discourse because of particular choices of linguistic expression used in the discourse. 
We conjecture that one psychological reflex of this inference load is a difference in 
perceived coherence among discourses that express the same propositional content 
using different linguistic forms. 
One of the tasks a hearer must perform in processing a discourse is to identify 
the referents of noun phrases in the discourse. It is commonly accepted, and is a hy- 
pothesis under which our work on centering proceeds, that a hearer's determination 
of noun phrase reference involves some process of inference. Hence a particular claim 
of centering theory is that the resource demands of this inference process are affected 
by the form of expression of the noun phrase. In Section 7, we discuss the effect on per- 
ceived coherence of the use of pronouns and definite descriptions by relating different 
choices to the inferences they require the hearer or reader to make. 
3. Basic Center Definitions 
We use the term centers of an utterance to refer to those entities serving to link that 
utterance to other utterances in the discourse segment that contains it. It is an utterance 
(i.e., the uttering of a sequence of words at a certain point in the discourse) and 
not a sentence in isolation that has centers. The same sentence uttered in different 
discourse situations may have different centers. Centers are thus discourse constructs. 
Furthermore, centers are semantic objects, not words, phrases, or syntactic forms. 
Each utterance U in a discourse segment (DS) is assigned a set of forward-looking 
centers, Cf(U, DS); each utterance other than the segment initial utterance is assigned 
a single backward-looking center, Cb(U, DS). To simplify notation, when the relevant 
discourse segment is clear, we will drop the associated DS and use CB(U) and Cf(U). 
The backward-looking center of utterance Un+l connects with one of the forward- 
looking centers of utterance Un. The connection between the backward-looking center 
of utterance Un+ 1 and the forward-looking centers of utterance Un may be of several 
types. To describe these types, we need to introduce two new relations, realizes and 
directly realizes, that relate centers to linguistic expressions. 
We will say that 
U directly realizes c 
208 
Barbara J. Grosz et al. Centering 
if U is an utterance of some phrase 4 for which c is the semantic interpretation. Realizes 
is a generalization of directly realizes. This generalization is important for capturing 
certain regularities in the use of definite descriptions and pronouns. 
The precise definition of 
U realizes c 
depends on the semantic theory one adopts, s One feature that distinguishes centering 
from other treatments of related discourse phenomena is that the realization relation 
combines syntactic, semantic, discourse, and intentional factors. That is, the centers of 
an utterance in general, and the backward-looking center specifically, are determined 
on the basis of a combination of properties of the utterance, the discourse segment in 
which it occurs, and various aspects of the cognitive state of the participants of that 
discourse. 
Thus, for a semantic theory to support centering, it must provide an adequate 
basis for computing the realization relation. For example, NP directly realizes c may 
hold in cases where NP is a definite description and c is its denotation, its value-free 
interpretation (discussed in Section 8), or an object related to it by "speaker's reference" 
(Kripke 1977). More importantly, when NP is a pronoun, the principles that determine 
the c's for which it is the case that NP directly realizes c do not derive exclusively from 
syntactic, semantic, or pragmatic factors. They are principles that must be elicited from 
the study of discourse itself. An initial formulation of some such principles is given 
in Section 8. 6 
The forward-looking centers of Un depend only on the expressions that constitute 
that utterance; they are not constrained by features of any previous utterance in the 
segment. The elements of CdUn) are partially ordered to reflect relative prominence 
in Un. In Section 5, we discuss a number of factors that may affect the ordering on the 
elements of Ct. The more highly ranked an element of CdUn ), the more likely it is to 
be Cb(Un+l). The most highly ranked element of Cf(U~) that is realized in Un+I is the 
CB (U~+l). Because Ct(U~) is only partially ordered, some elements may, from Cf(Un) 
information alone, be equally likely to be CB (Un+l). In such cases, additional criteria are 
needed for deciding which single entity is the Cb(Un+l). Some recent psycholinguistic 
evidence suggests that the syntactic role in Un+I may determine this choice (Gordon, 
Grosz, and Gilliom 1993). 
In the remainder of the paper we will use a notation such that the elements of 
Ct are ranked in the order in which they are listed. 7 In particular, for presentational 
4 U need not be a full clause. We use U here to stress again that it is the utterance, not the string of 
words. 
5 In the original manuscript, we defined realize in terms of situation semantics (Barwise and Perry 1983) 
and said the relation held "if either c is an element of the situation described by the utterance U or c is 
directly realized by some subpart of U." We discuss this further in Section 7. 6 In the examples in this paper, we will be concerned with the realization relationship that holds 
between a center and a singular definite noun phrase; i.e., cases where an NP directly realizes a center c. 
Several extensions to the theory presented here are needed to handle plural, quantified noun phrases 
and indefinites. It is also important to note that not all noun phrases in an utterance contribute centers 
to Cf(U) and not only noun phrases do so. More generally, events and other entities that are more 
often directly realized by verb phrases can also be centers, whereas negated noun phrases typically do 
not contribute centers; the study of these issues is, however, beyond the scope of this paper. 7 To simplify the presentation in the remainder of this paper, we will assume in most of the discussion 
that there is a total order with strict ordering between any two elements; at those places where the 
partial ordering makes a significant difference we will discuss that. 
209 
Computational Linguistics Volume 21, Number 2 
purposes, we will use the following schematic to refer to the centers of utterances in 
a sequence: 
For Un: Cb(Un) = a, Cf(Un) -- (el, e2, ...ep), a = ek, for some k. 
For Un+t: Cb(Unq-1) realizes ern and, for all j, j < m, e} is not real- 
ized in Un+l; i.e., em is realized in Un+l, and no higher ranked ej is 
realized in U~+I. 
Finally, we also define three types of transition relations across pairs of utterances. 
. 
. 
. 
CENTER CONTINUATION: Cb(Un+l)= Cb(Un), and this entity is the most 
highly ranked element of Cf(Un+l). In this case, Cb(Un+t) is the most 
likely candidate for Cb(Un+2)) it continues to be Cb in Un+l, and 
continues to be likely to fill that role in Un+2. 
CENTER RETAINING: Cb(Un+l) ~- Cb(Un), but this entity is not the most 
highly ranked element in Cf(Un+l)- In this case, Cb(Un+l) is not the most 
likely candidate for Cb(Un+2); although it is retained as Cb in Un+t, it is 
not likely to fill that role in Un+2. 
CENTER SHIFTING: Cb(Un+l) ~ Cb(Un). 
The coherence of a segment is affected by the kinds of centering transitions engen- 
dered by a speaker's choices of linguistic realizations in the utterances constituting 
the segment. Of particular concern are choices among (1) CONTINUATION of the center 
from one utterance not only to the next, but also to subsequent utterances; (2) RE- 
TENTION of the center from one utterance to the next; (3) SHIFTING the center, if it is 
neither retained nor continued, s 
4. Claims of Centering Theory 
The centering framework described above provides the basis for stating a number of 
specific claims about the relationship between discourse coherence, inference load, and 
choice of referring expression. Underlying these claims is the most fundamental claim 
of centering theory, that to the extent a discourse adheres to centering constraints, its 
coherence will increase and the inference load placed upon the hearer will decrease. 
We briefly list several major claims in this section, and elaborate on the evidence or 
motivation for each in subsequent sections. 
• A unique Cb: Each Un has exactly one backward-looking center. It might 
be thought that a more general definition would allow for multiple 
backward-looking centers as well as multiple forward-looking centers. 
However, this is not the case, as we show in Section 5. 
• Ranking of Ct: The Cf elements are partially ordered according to a 
number of factors. Several of the factors posited to affect this ordering 
are discussed in Section 5, but the full set of factors remains to be 
determined. Ranking of elements in Cf(Un) guides determination of 
8 Shifting of the center does not in itself mark a discourse segment boundary. The center may shift 
within a single segment. 
210 
Barbara J. Grosz et al. Centering 
Cb(Un+l), because Cf(Un) is only partially ordered, additional factors 
may constrain the choice. 9 
• Centering constrains realization possibilities: Rule 1, discussed in 
Section 6, stipulates one constraint centering imposes on realization. We 
expect that other such constraints exist. 
• Preferences among sequences of center transitions: Rule 2, discussed in 
Section 6, hypothesizes a preference among types of transitions. 
• Primacy of partial information: The information needed to compute a 
complete unique interpretation for an utterance may not be available 
until subsequent utterances are produced. Thus, as discussed in 
Section 8, to support centering, a semantic theory must support the 
construction of partial interpretations, in particular for elements of Cf. 
• Locality of Cb(Un): The choice of a backward-looking center for an 
utterance Un is from the set of forward-looking centers of the previous 
utterance Un-> In this sense the Cb is strictly local. Cb(Un) cannot be 
from Cf(U,_2) or other prior sets of forward-looking centers. 1° 
• Centering is controlled by a combination of discourse factors: Center 
determination is not solely a syntactic, semantic, or pragmatic process. 
5. Factors Governing Centering 
Before we can examine the linguistic features that contribute to an entity's being the 
backward-looking center of an utterance, it is necessary to provide support for the 
claim that there is only a single backward-looking center. In the definitions in Section 3, 
there is a basic asymmetry between the Cf, which is a set, and the Cb, which is a 
singleton. Sequences like those in (6) seem to suggest that there might be multiple 
Cb's, analogous to the partially ordered set of Cf's. A priori there is no reason to think 
that either Susan or Betsy alone is the Cb of utterance (6b). 
(6) a. Susan gave Betsy a pet hamster. 
b. She reminded her that such hamsters were quite shy. 
However, if we consider different subsequent utterances, it becomes clear that Susan 
and Betsy do not have an equivalent status in the second utterance. The ranking of 
the Cr's matters. The variants (7)-(10) differ only in their choice of realization of Susan 
and Betsy, in particular in which is pronominalized and which is in subject position. 
(7) a. Susan gave Betsy a pet hamster. 
b. She reminded her that such hamsters were quite shy. 
c. She asked Betsy whether she liked the gift. 
(8) a. Susan gave Betsy a pet hamster. 
9 This point is connected with the discussion of partial ordering in Section 3. 
10 It may appear that Cb(Un ) comes from Cf(Un-2) or prior sets of forward-looking centers, but then it is 
only because it is in Cf(Un-1) also. 
211 
Computational Linguistics Volume 21, Number 2 
(9) 
b. She reminded her that such hamsters were quite shy. 
c. Betsy told her that she really liked the gift. 
a. Susan gave Betsy a pet hamster. 
b. She reminded her that such hamsters were quite shy. 
c. Susan asked her whether she liked the gift. 
(10) a. Susan gave Betsy a pet hamster. 
b. She reminded her that such hamsters were quite shy. 
c. She told Susan that she really liked the gift. 
If both Susan and Betsy were equally likely backward-looking centers in the second 
utterance of these sequences, then all of these variants would be equally good or, 
perhaps, there would be a preference for variants (7) and (9), which exhibit continuity 
of grammatical subject and object. However, this is not the case. There is a marked 
decrease in acceptability from version (7) to version (10), and for many people version 
(10) is completely unacceptable. 
The problem is not merely a change from a pronoun back to a proper name, as 
this happens to the same extent in all four variants. It also cannot be attributed solely 
to a change from grammatical subject to grammatical object position, as variant (8) 
involves such a change and yet is better than variant (9), which does not. Rather, it 
must be the case that Susan is the CB at utterance (b) at each of the variants. Variants (9) 
and (10) can be shown to be worse than (7) and (8) because they violate the centering 
rules presented in the next section. 
This example suggests that pronominalization and subject position are possible 
linguistic mechanisms for establishing and continuing some entity as the CB. In the 
second utterance of these sequences, Susan is realized by a pronoun in subject position; 
'she' is the CB of this utterance. Utterance (7c) continues Susan as Cb, whereas utterance 
(8c) merely retains her. Utterances like (8c) may be used to provide a basis for a shift 
in Cb .11 However, this leaves open questions of the independence of syntactic role and 
pronominalization, and the predominance of either, for controlling centering. 
The fact that being in subject position contributes in and of itself to the likelihood 
an entity will be the highest-ranked Cf (i.e. likely to be the next CB) can be seen by 
contrasting the following two sequences, which differ only in their final utterances: 
(11) a. Susan is a fine friend. 
,b. She gives people the most wonderful presents. 
c. She just gave Betsy a wonderful bottle of wine. 
d. She told her it was quite rare. (Susan told Betsy) 
e. She knows a lot about wine. (Susan knows... ) 
11 The effect of various linguistic constructions on center movement and the interactions of centering 
shifts with global discourse structure are active areas of research. Section 9 provides references to such 
work. 
212 
Barbara J. Grosz et al. Centering 
(12) a. Susan is a fine friend. 
b. She gives people the most wonderful presents. 
c. She just gave Betsy a wonderful bottle of wine. 
d. She told her it was quite rare. (Susan told Betsy) 
e. Wine collecting gives her expertise that's fun to share. (Susan's ex- 
pertise) 
In the (c) utterance of each sequence, Susan is the Cb. Either Susan or Betsy might be 
the referent of the subject pronoun in the fourth utterance; however, there appears to 
be a strong preference for Susan (i.e., for the reading "Susan told Betsy"). 12 Because 
this preference might be attributable to parallelism, the last utterance in (12) provides 
a crucial test. If the Cf ranking depended on pronominalization alone, the fourth 
utterance would allow either Susan or Betsy to be the highest-ranked Cf. Parallelism 
would suggest different preferences for the Cb(12e) in the two sequences. However, 
the preferred reading of the pronoun (respectively, "she" and "her") in utterance (e) of 
both sequences is Susan, who is realized in the subject position of the (d) utterances. 
This preference holds regardless of syntactic position in the (e) utterances. Thus, we 
can establish a preference for subject position. In other circumstances, however, as the 
examples below illustrate, the Cb may be realized in other grammatical roles. 
In the first clause of both utterances (13d) and (14d), the direct object is pronomi- 
nalized; the pronoun "it" refers to the green plastic tugboat. In (13) taking the boat to 
be the highest ranked Cf and hence the most likely referent for "the silly thing" in the 
second clause of utterance (d) yields a coherent and easily comprehensible discourse. 13 
In (14), however, pragmatic information leads to a preference for the bear, not the boat, 
to be the referent of "the silly thing" in the last utterance; this preference is in conflict 
with the boat's being the most likely Cb. That (13) is a more coherent discourse than 
(14) can be explained on the basis of this difference. TM 
(13) a. Have you seen the new toys the kids got this weekend? 
b. Stuffed animals must really be out of fashion. 
c. Susie prefers the green plastic tugboat to the teddy bear. 
d. Tommy likes it better than the bear too, but only because the silly 
thing is bigger. 
12 Sequences in which a similar pronominalization pattern is used but in which the fourth utterance 
implies report of a dialogue (e.g., "She thanked her and told her she appreciated that the wine was 
quite rare.") may lead to interpretations in which the subject pronoun is taken as referring to Betsy; 
accentuation of the subject may also be used to achieve this result. The first of these suggests a strong 
interaction between dialogue verbs and centering, which is also apparent in direct-speech dialogue 
examples. The relationship between this kind of lexical-semantic influence over centering and that of 
so-called 'empathy' verbs, e.g., Kameyama (1985), Walker, Iida, and Cote (1994), remains to be 
determined. The second would appear to provide additional evidence for subject preference in 
centering, based on results of Hirschberg and Ward (1991) showing that accenting served to flip 
preferences (in their study from either strict-to-sloppy or sloppy-to-strict readings for anaphors in the 
antecedent clause in VP-ellipsis constructions). 13 For the sake of this argument, assume that children like bigger things more than smaller things. If this 
is not the case, then the argument merely flips which variants are more acceptable. 14 The discrepancy is even greater if "it" is used in the last utterance clause. However, one might 
attribute this to repetition of the use of "it" and so we have avoided the repeated use of a pronoun. We 
also note that "the silly thing" conveys additional information--roughly, the speaker's attitude toward 
the bear or tugboat (cf. Section 8). 
213 
Computational Linguistics Volume 21, Number 2 
(14) a. Have you seen the new toys the kids got this weekend? 
b. Stuffed animals must really be out of fashion. 
c. Susie prefers the green plastic tugboat to the teddy bear. 
d. Tommy likes it better than the bear too, although the silly thing is 
bigger. 
Thus, the discourses in (11)-(14) suggest that grammatical role is a major determinant 
of the ranking on the Cf, with SUBJECT > OBJECT(S) > OTHER. The effect of factors 
such as word order (especially fronting), clausal subordination, and lexical semantics, 
as well as the interaction among these factors are areas of active investigation; Section 
9 again provides references to such work. 
In summary, these examples provide support for the claim that there is only a sin- 
gle Cb, that grammatical role affects an entity's being more highly ranked in Cf, and 
that lower-ranked elements of the Cf cannot be pronominalized unless higher-ranked 
ones are. Kameyama (1985) was the first to argue that grammatical role, rather than the- 
matic role, which Sidner used, affected the Cf ranking. Psycholinguistic research since 
1986 (Hudson-D'Zmura 1988; Gordon, Grosz, and Gilliom 1993) supports the claims 
that there is a single Cb and that grammatical role plays a determining role in identi- 
fying the CB. It furthermore suggests that neither thematic role nor surface position is 
a determinant of the Cb. In contrast, both grammatical role and surface position were 
shown to affect the Cf ordering. Although there are as yet no psycholinguistic results 
related to the effect of pronominalization on determining Cb(Un_l) , cross-linguistic 
work (Kameyama 1985; Prince and Walker 1995; Walker, Iida, and Cote 1994) argues 
that it plays such a role. Section 9 lists several papers appearing after Grosz, Joshi, 
and Weinstein (1986) that investigate factors affecting the Cf ordering. 
6. Constraints on Center Movement and Realization 
The basic constraint on center realization is given by Rule 1, which is stated in terms 
of the definitions and schematic in Section 3. 
Rule 1 
If any element of Cf(Un) is realized by a pronoun in Unq-1, then the Cb(Un+l) must be 
realized by a pronoun also. 
In particular, this constraint stipulates that no element in an utterance can be realized as 
a pronoun unless the backward-looking center of the utterance is realized as a pronoun 
also. is Rule 1 represents one function of pronominal reference: the use of a pronoun to 
realize the CB signals the hearer that the speaker is continuing to talk about the same 
thing. Note that Rule 1 does not preclude using pronouns for other entities so long as 
the Cb is realized with a pronoun. (This is illustrated in examples 7-10 in Section 5.) 
Psychological research (Gordon, Grosz, and Gilliom 1993; Hudson-D'Zmura 1988) and 
cross-linguistic research (Di Eugenio 1990; Kameyama 1985, 1986, 1988; Walker, Iida, 
and Cote 1990, 1994) have validated that the CB is preferentially realized by a pronoun 
in English and by equivalent forms (i.e., zero pronouns) in other languages. 
The basic constraint on center movement is given by Rule 2. 
15 Rule 1 ignores certain complications that may arise if one of the forward-looking centers of Un+l is 
realized by a deictic pronoun. 
214 
Barbara J. Grosz et al. Centering 
Rule 2 
Sequences of continuation are preferred over sequences of retaining; and sequences of 
retaining are to be preferred over sequences of shifting. 
In particular, a pair continuations across Un and across Un+l, represented as 
Cont(Un,U~+0 and Cont(Un+t,U~+2) respectively, is preferred over a pair of reten- 
tions, Ret(Un,U~+0 and Ret(U,+t,Un+2). The case is analogous for pair of retentions 
and a pair of shifts. 
Rule 2 reflects our intuition that continuation of the center and the use of retentions 
when possible to produce smooth transitions to a new center provides a basis for local 
coherence. In a locally coherent discourse segment, shifts are followed by a sequence 
of continuations characterizing another stretch of locally coherent discourse. Frequent 
shifting leads to a lack of local coherence, as was illustrated by the contrast between 
Discourse (1) and Discourse (2) in Section 2. Thus, Rule 2 provides a constraint on 
speakers, and on natural-language generation systems. They should plan ahead to 
minimize the number of shifts. This rule does not have the same direct implementation 
for interpretation systems; rather it predicts that certain sequences produce a higher 
inference load than others. To empirically test the claim made by Rule 2 requires 
examination of differences in inference load of alternative multi-utterance sequences 
that differentially realize the same content. 
Although several cross-linguistic studies have investigated Rule 2 (see Section 9), 
there are as yet no psycholinguistic results empirically validating it. 
7. Applications of the Rules 
The two centering rules along with the partial ordering on the forward-looking centers 
described in Section 5 constitute the basic framework of center management. These 
rules can explain a range of variations in local coherence. TM 
A violation of Rule 1 occurs if a pronoun is not used for the backward-looking 
center and some other entity is realized by a pronoun. Such a violation occurs in the 
following sequence presumed to be in a longer segment that is currently centered on 
John (cf. also examples (9) and (10) in Section 5): 
(15) a. He has been acting quite odd. \[C b = John = referent("he')\] 
b. He called up Mike yesterday. \[CB = John = referent("he')\] 
c. John wanted to meet him urgently. \[CB = John; referent("him') = 
Mike\] 
The violation of Rule 1 leads to the incoherence of the sequence. The only possible 
interpretation is that the "John" referred to in (15c) is a second person named "John," 
not the one referred to in the preceding utterances in (15); however, even under this 
interpretation the sequence is very odd. The next example illustrates that this effect is 
16 These rules and constraints have also been used by others as the basis for pronoun resolution 
algorithms based on centering. The earliest such attempt (Brennen, Friedman, and Pollard 1987) used 
the uniqueness and locality of Cb constraints and ranked the Cf by grammatical role; it employed a 
variant of Rule 2 in which the stated preferences on transitions were restricted to transitions between 
individual pairs of utterances (rather than the longer sequences in the original formulation) and used to 
decide between possible interpretations of pronouns. Section 9 provides references to other work on 
centering algorithms. 
215 
Computational Linguistics Volume 21, Number 2 
independent of the grammatical position of the C b and also demonstrates that Rule 1 
operates independently of the type of centering transition. 
(16) a. John has been acting quite odd. 
b. He called up Mike yesterday. \[C b = John = referent("he")\] 
c. Mike was studying for his driver's test. \[CB = Mike = referent("his')\] 
d. He was annoyed by John's call. 
Without utterance (16c), this sequence, like the sequence in (15), is unacceptable unless 
it is possible to consider the introduction of a second person named "John." The 
intervening utterance (c) here provides for a shift in center from John to Mike, making 
the full sequence coherent. 17 
It is important to notice that Rule 1 constrains the realization of the most highly 
ranked element of the Cf(Un) that is realized in Un+t given that pronominalization is used. 
Obviously any entities realized in Un that are not realized in Un+t, including the Cb(Un) 
as well as the highest ranked element of CdUn), do not affect the applicability of Rule 1. 
Likewise, if no pronouns are used, then Rule 1 is not applicable. Two particular ways 
in which such situations may hold have been noticed in previous research. Each leads 
to a different type of inference load on the hearer, both of which we believe relate to 
Rule 1; however, neither constitutes a violation of Rule 1. The resulting discourses are 
coherent, but the determination of local coherence (in the first case) or the detection 
of a global shift (in the second case) requires additional inferences. 
The first case concerns realization of the Cb by a nonpronominal expression. Rule 1 
does not preclude using a proper name or definite description for the Cb if there are 
no pronouns in an utterance. However, it appears that such uses are best when the full 
definite noun phrases that realize the centers do more than just refer. They convey some 
additional information, i.e., lead the hearer or reader to draw additional inferences. 
The hearer or reader not only infers that the CB has not changed even though no 
pronoun has been used, but also recognizes that the description holds of the old Cb. 
Sequences (17) and (18) are typical cases. TM 
(17) a. My dog is getting quite obstreperous. 
b. I took him to the vet the other day. 
c. The mangy old beast always hates these visits. 
(18) a. I'm reading The French Lieutenant's Woman. 
b. The book, which is Fowles's best, was a bestseller last year. 
The second case concerns the use of a pronoun to realize an entity not in the CdUn); 
such uses are strongly constrained. The particular cases that have been identified 
involve instances where attention is shifted globally back to a previously centered 
entity (e.g. Grosz \[1977\], Reichman \[1985\]). In such cases additional inferences are 
17 Empirical investigations of these claims of Grosz, Joshi, and Weinstein (1986) suggest they are too 
strong. In particular, the results of Gordon, Grosz, and Gilliom (1993) suggest that (16d) without the 
intervening (c) utterance is not as bad as (15c). 
18 Sequence (17) is an adaptation of one of Sidner's examples (Sidner 1979). 
216 
Barbara J. Grosz et al. Centering 
required to determine that the pronoun does not refer to a member of the current 
forward-looking centers and to identify the context back to which attention is shifting. 
Further investigation is required to determine the linguistic cues (e.g., intonation or cue 
phrases \[Grosz and Hirschberg 1992\]) and intentional information that are required to 
enable such shifts while preserving coherence, as well as the effect on inference load. 
A third complication arises in the application of Rule 1 in sequences in which the 
CB of an utterance is realized but not directly realized in that utterance. This situation 
typically holds when an utterance directly realizes an entity implicitly focused by an 
element of the Cf of the previous utterance. For instance, it arises in utterances con- 
taining noun phrases that express functional relations (e.g., "the door, .... the owner") 
whose arguments have been directly realized in previous utterances (e.g., a house) as 
occurs in the sequence, 
(19) a. The house appeared to have been burgled. 
b. The door was ajar. 
c. The furniture was in disarray. 
In this segment, the house referred to in (19a) is an element of the Cf(19a). This house is 
the Cb(19b); it is realized but not directly realized in (19b). Because the house is the CB, 
the Cf (19b) includes it as well as the door that is directly realized in the utterance. The 
CB(19c) is thus again "house." We assume here that the door ranks above the house in 
Cf (19b). For example, if (19b) is followed by a sentence with 'it' in the subject position, 
then 'it' is more likely to refer to the doorJ 9 This is consistent with the ranking of the 
door ahead of the house in Cf (19b). However, continuity of the house as a potential 
CB for (19c) is reflected in the discourse segment being interpreted to be "about" the 
house and (19c) being interpreted in the same way as (19b) with respect to the house. 
In Grosz, Joshi, and Weinstein (1986) we did not explore this issue further; the general 
issue of the roles of functional dependence and implicit focus in centering remain 
open. 2° 
The use of different types of transitions following the rankings in Rule 2 are 
illustrated by the discourse below. 
(20) a. John has been having a lot of trouble arranging his vacation. 
b. He cannot find anyone to take over his responsibilities. (he = John) 
CB = John; Cf = {John} 
c. He called up Mike yesterday to work out a plan. (he = John) 
C6 -- John; Cf = {John, Mike} (CONTINUE) 
d. Mike has annoyed him a lot recently. 
CB -- John; Cf = {Mike, John} (RETAIN) 
e. He called John at 5 AM on Friday last week. (he = Mike) 
Cb = Mike; Cf = {Mike, John} (SHIFT) 
19 However, it can refer to the house. For example if (b) were followed by "Otherwise from the outside it 
appeared quite normal. Inside was a different story." A pronoun could also be used in other grammatical roles to refer to the door. We use subject position as the test, because there is no prior 
sentential context to bias the interpretation. 
20 See Section 9 for some recent references related to this issue. 
217 
Computational Linguistics Volume 21, Number 2 
Utterance (20b) establishes John both as the C b and as the most highly ranked Cf. In 
utterance (20c) John continues as the CB, but in utterance (20d) he is only retained; 
Mike has become the most highly ranked element of the Cf. Finally, in utterance (20e) 
the backward-looking center shifts to being Mike. Rule 1 is satisfied throughout (20). 
Rule 1 depends only on the ordering of elements of Cf, and not on the notions of 
retaining and continuation. 
8. Requisite Properties of Underlying Semantic Theory 
Different semantic theories make different commitments with respect to the complete- 
ness or definiteness required of an interpretation. Because the information needed to 
compute a unique interpretation for an utterance is not always available at the time the 
utterance occurs in the discourse, the ways in which a theory treats partial information 
affects its computational tractability as the basis for discourse interpretation. It is not 
merely that utterances themselves contain only partial information, but that it may 
only be subsequent to an utterance that sufficient information is available for comput- 
ing a unique interpretation. No matter how rich a model of context one has, it will not 
be possible to fully constrain the interpretation of an utterance when it occurs. This is 
especially true for definite noun phrase interpretation. For example, several interpre- 
tations are possible for the noun phrase "the Vice-President of the United States," in 
the utterance 
(21) The Vice-President of the United States is also President of the Senate. 
One interpretation, namely the individual who is currently Vice-President, provides 
the appropriate basis for the interpretation of "he" in the subsequent utterance given 
in (22): 
(22) Right now, he is the President's key person in negotiations with Congress. 
However, a different interpretation, one which retains some descriptive content, pro- 
vides the appropriate basis for an interpretation of the pronoun "he" in the slightly 
different subsequent utterance 
(23) Historically, he is the President's key person in negotiations with Congress. 
A semantic theory that forces a unique interpretation of utterance (21) will require that 
a computational theory or system either manage several alternatives simultaneously 
or provide some mechanism for retracting one choice and trying another later. On the 
other hand, a theory that allows for a partially specified interpretation must provide for 
refining that interpretation on the basis of subsequent utterances. Additional utterances 
may provide further constraints on an interpretation, and sequences of utterances 
may not be coherent, if they do not allow for a consistent choice of interpretation. 
For example, the utterance in (24) is perfectly fine after (22), but yields an incoherent 
sequence after (23). 21 
21 These examples were first written in 1986 when George Bush was Vice-President. They remain useful 
for illustrating the original points if the time of original writing is taken into account. As we discuss 
later, taken as spoken now they illustrate new points. 
218 
Barbara J. Grosz et al. Centering 
(24) As Ambassador to China, he handled many tricky negotiations, so he 
does well in this job. 
To summarize, given that one purpose of discourse is to increase the information 
shared by speaker and hearer, it is not surprising that individual utterances convey 
only partial information. However, the lack of complete information at the time of 
processing an utterance means that a unique interpretation cannot be definitely de- 
termined. In constructing a computational model, we are then left with three choices: 
compute all possible interpretations and filter out possibilities as more information is 
received; choose (on some basis) a most likely interpretation and provide for "back- 
tracking" and computing others later; compute a partial interpretation. We conjecture 
that this third choice is the appropriate one for noun phrase interpretation. 
Centering theory and the centering framework rely on a certain picture of the 
ways in which utterances function to convey information about the world. One role 
of a semantic theory is to give substance to such a picture. At the time Grosz, Joshi, 
and Weinstein (1986) was written, it struck us that situation semantics (Barwise and 
Perry 1983) provided a particularly convenient setting in which to frame our own 
theory of discourse phenomena, though our account relied only on general features 
of this approach and not on details of the theory as then articulated. The two most 
important features of situation semantics from the standpoint of the theory of discourse 
interpretation we wished to develop were (1) that it allows for the partial interpretation 
of utterances as they occur in discourse, and (2) that it provides a framework in which 
a rich theory of the dependence of interpretation on abstract features of context may 
be elaborated. There is now a large situation semantics literature that contains many 
extensions and refinements of the theory to which we refer the interested reader. 
The original book (Barwise and Perry 1983) may be consulted for an account of the 
distinction between value-free and value-loaded interpretations used below. 
In the discussion and examples in previous sections, the Cb and the elements 
of Cf have all been the denotations of various noun phrases in an utterance. The 
actual situation is more complicated even if we ignore for the moment quantifiers and 
other syntactic complexities (cf. Webber 1978) as well as cases in which the center 
is functionally dependent on, or otherwise implicitly focused by, an element of the 
Cf of the previous utterance (cf. Section 7). A singular definite noun phrase may 
contribute a number of different interpretations to Cf. In particular, not only the value- 
free interpretation, but also various loadings may be contributed. 
For example, in the utterance, "The Vice-President of the United States is also 
President of the Senate," the noun phrase "the Vice-President" contributes both a 
value-loaded and a value-free interpretation. The value-free interpretation is needed 
in the sequence (25a-c), whereas the value-loaded interpretation is needed in (26a-c). 
(25) a. 
b. 
C, 
The Vice-President of the United States is also President of the Senate. 
Historically, he is the President's key man in negotiations with Congress. 
He is required to be 35 years old. 
(26) a. 
b. 
C. 
The Vice-President of the United States is also President of the Senate. 
Right now, he's the president's key person in negotiations with Congress. 
As Ambassador to China, he handled many tricky negotiations, so 
he is well prepared for this job. 
219 
Computational Linguistics Volume 21, Number 2 
The Cb(25b) and the Cb(26b) are both directly realized by the anaphoric element "he." 
But Cb(25b) is the value-free interpretation of the noun phrase, "the Vice-President" (as 
in, "The Vice-President of the United States is the President's key man in negotiations 
with Congress"), whereas Cb(26b) is the value-loaded interpretation (as in "the person 
who now is Vice-President of the United States"). That this is so is demonstrated by 
the fact that (25c) is true in 1994, whereas (26c) is not. Centering accommodates these 
differences by allowing the noun phrase "the Vice-President of the United States" 
potentially to contribute both its value-free interpretation and its value-loading at 
the world type to Cf(25a). Cb(25b) is then the value-flee interpretation, and Cb(26b) 
is the value-loaded one (at the time of the writing of Grosz, Joshi, and Weinstein 
(1986), George Bush, but now \[1995\] A1 Gore). In each sequence, the (a) utterance 
underdetermines what element to add to Cf. This underdetermination may continue 
in a subsequent utterance with the pronoun. For example, that would be the case if 
the introductory adverbials were left off the (b) utterances. 
We conjecture that the correct approach to take in these cases is to add the value- 
free interpretation to Cf and then load it for the interpretation of subsequent utterances 
if this is necessary. This conjecture derives from a belief that this approach will most 
effectively limit the inferences required. These loading situations thus constitute a 
component of the centering constituent of the discourse situation. It remains an open 
question how long to retain these loading situations, although those corresponding to 
elements of Cf that are not carried forward (either as the Cb or as Cfs of the subsequent 
utterance) can, obviously, be dropped. 
It is possible for an utterance to prefer either a value-free (VF) or value-loaded 
(VL) interpretation but not force it. For example, the second utterance in the following 
sequence prefers a VF interpretation but allows for the VL interpretation that is needed 
in the third utterance. 
(27) a. A: The Vice-President of the U.S. is also President of the Senate. 
b. B: I thought he played some important role in the House. 
c. A: He did, but that was before he was the Vice-President. 
In a similar way the second utterance in the following sequence 22 prefers the VL 
interpretation, but allows for the VF. The third utterance requires the VF interpretation. 
(28) a. John thinks that the telephone is a nuisance. 
b. He curses it every day. 
c. He doesn't realize that it is an invention that changed the world. 
In these examples, both value-free and value-loaded interpretations are shown to stem 
from the same full definite noun phrase. 
There appear to be strong constraints on the kinds of transitions that are allowed, 
however. In particular, if a given utterance forces either the VF or the VL interpreta- 
tion, then only this interpretation is possible in the immediately subsequent utterance. 
However, if some utterance only prefers one interpretation (in a given context), but 
allows the other, then the subsequent utterance may pick up on either one. 
22 Christine Nakatani provided this example, which is far more compelling than the one originally in 
Grosz, Joshi, and Weinstein (1986). 
220 
Barbara J. Grosz et al. Centering 
For example, the sequence, 
(29) a. The Vice-President of the United States is also President of the Senate. 
b. He's the President's key man in negotiations with Congress. 
in which "he" may be interpreted either VF, or VL, may be followed by either (30) or 
(31): 
(30) As Ambassador to China, he handled many tricky negotiations. (VL) 
(31) He is required to be at least 35 years old. (VF) 
However, if we change (29b) to force the value-loaded interpretation, as in (26), then 
only the value-loaded interpretation (30) is possible. Similarly, if (29b) is changed to 
force the value-free interpretation, as in (25b), then only the value-free interpretation 
(31) is possible. 
Speaker intentions may also enter into the determination of which entities are in 
the Cf. The referential uses of descriptions, of which Donnellan (1966) gives examples, 
demonstrate cases in which the "referential intentions" of the speaker in his use of 
the description play a role in determining CB(U). For example, consider the following 
sequence: 
(32) a. Her husband is kind to her. 
b. No, he isn't. The man you're referring to isn't her husband. 
(33) a. Her husband is kind to her. 
b. He is kind to her but he isn't her husband. 
In these examples, 23 the speaker uses a description to refer to something other 
than the semantic denotation of that description, i.e. the unique thing that satisfies 
the description (if there is one). There are several alternative explanations of such 
examples, involving various accounts of speaker's intentions, mutual belief, and the 
like. A complete discussion of these issues is beyond the scope of this paper. 
The importance of these cases resides in showing that Cf(U) may include more 
than one entity that is realized by a single NP in U. In this case, the noun phrase 
"her husband" contributes two individuals, the husband and the lover, to Cf(32a) and 
Cf(33a). This can be seen by observing that both discourses seem equally appropriate 
and that the backward-looking centers of (32b) and (33b) are respectively the husband 
and the lover, which are realized by their anaphoric elements. 
These examples introduce a number of research issues concerning the represen- 
tation and management of the Cb and Cf discourse entities. The account given here 
depends on a semantic theory that permits minimal commitment in interpretations. 
The open question is which constraints on centers are introduced at which points 
during processing. We must leave this as a topic for future work. 
23 These examples are from Kripke (1977, p. 21). 
221 
Computational Linguistics Volume 21, Number 2 
9. Related Work 
This theory can be contrasted with two previous research efforts that spurred this 
work: Sidner's (1979) original work on immediate focusing and pronouns, and Joshi 
and Weinstein's (1981) subsequent work on centering and inferences. 
The centering theory discussed here is quite close to Sidner's original theory, both 
in attacking local discourse issues and in the general outline of approach. However, 
it differs in several details. In Sidner's theory, each utterance provides an immediate 
discourse focus, an actor focus, and a set of potential foci. The discourse and actor 
foci may coincide, but need not. Her potential foci are roughly analogous to our Cf. 
The Cb for an utterance sometimes coincides with her actor focus and sometimes with 
her discourse focus. She distinguishes these two to handle various cases of multiple 
pronouns. However, as we have shown, utterances do not have multiple Cbs. Further- 
more, utterances can have more than two pronouns, so merely adding a second kind 
of immediate focus is of limited use. The difference between these two theories can 
be seen from the following example (from Sidner \[1979\]): 
(34) a. I haven't seen Jeff for several days. 
b. Carl thinks he's studying for his exams, 
c. but I think he went to the Cape with Linda. 
On Sidner's account, Carl is the actor focus after (34b) and Jeff is the discourse focus. 
Because the actor focus is preferred as the referent of pronominal expressions, Carl is 
the leading candidate for the entity referred to by he in (34c). It is difficult to rule this 
case out without invoking fairly special domain-specific rules. On our account, Jeff is 
the CB at (34b) and there is no problem. The type of example Sidner was concerned 
about would occur if utterance (34c) were replaced by "He thinks he studies too 
much." However, the centering rules would still hold in this case. They provide no 
constraints on additional pronouns so long as the highest ranked Cf is realized by a 
pronoun. However, the rules are incomplete; in particular, as given they do not specify 
which pronoun in a multipronoun utterance refers to the Cb. The center management 
rules are based solely on the Cb and the highest ranked member of the Cf. As a result, 
while there are cases of multiple pronouns for which the theory makes incomplete 
predictions, having both an actor and a discourse focus will not handle these cases in 
general. 
Joshi and Kuhn (1979) and Joshi and Weinstein (1981) presented a preliminary re- 
port on their research regarding the connection between the computational complexity 
of the inferences required to process a discourse and the coherence of that discourse 
as assessed by measures that invoke centering phenomena. However, their basic defi- 
nitions conflate the centers of an utterance with the linguistic expressions that realize 
those centers. In some of their examples it is unclear whether the shift in center or 
the particular expression used to realize the center is responsible for differences in 
coherence and inference load. Our present work has clarified these differences while 
maintaining Joshi and Weinstein's basic focus on the interaction between inference 
load and center management. 
Since Grosz, Joshi, and Weinstein (1986) was first circulated a number of re- 
searchers have tested and developed aspects of the theory presented here. 24 This 
24 Our listing in this section is based on the best information available to us. It is quite possible that we 
have missed some references. We will be grateful if readers could send us missing references. 
222 
Barbara J. Grosz et al. Centering 
follow-on research can be roughly grouped in a few main areas: 
• Cross-linguistic work on centering: 25 (German, Hebrew, Italian, Japanese, 
Korean, and Turkish): Di Eugenio (1990); Hoffman and Turan (1993); 
Kameyama (1985, 1986, 1988); Rambow (1993); Walker, Iida, and Cote 
(1990, 1994); Yongkyoon (1991); Ziv and Grosz (1994). 
• Centering algorithms: Baldwin (1993); Brennan, Friedman, and Pollard 
(1987); Kehler (1993); Walker (1989); Walker, Iida, and Cote (1994). 
• Empirical and psycholinguistic evaluation of centering predictions: 
Brennan (1995); Gordon and Chan (in press); Gordon, Grosz, and Gilliom 
(1993); Gordon and Scearce (1995); Hudson-D'Zmura (1988); Hudson, 
Tanenhaus, and Dell (1986); Walker (1989). 
• Centering and linguistic realizations: 26 Cote (1992, 1993); Hurewitz and 
Linson (1993); Kameyama (1993); Nakatani (1993); Passonneau (in press, 
1991); Prince (1994); Prince and Walker (1995); Suri and McCoy (1993); 
Turan (1995); Walker (1989). 
• Centering, dialogue, and global discourse structure: Brennan'(1995); 
Hurewitz and Linson (1993); Roberts (1993); Sparck Jones (1993); Walker 
(1993); Walker and Whittaker (1990). 
Acknowledgments 
We want to thank Breck Baldwin, Felicia 
Hurewitz, Andy Kehler, Karen Lochbaum, 
Christine Nakatani, Ellen Prince, and Lyn 
Walker for their valuable comments, which 
helped us improve both the content and the 
presentation of our paper. We are also 
grateful to Carolyn Elken for helping us 
keep track of the various drafts of this paper 
and for providing valuable editorial help. 
Partial support for the first author was 
provided by grants NSF IRI-90-09018 and 
IRI-93-08173; the second author was 
partially supported by the ARO Grant 
DAAL03-89-0031 and ARPA Grant 
N00014-90-J-1863. 
References 
Baldwin, Breck (1993). "Anaphora 
resolution with centering." In Workshop on 
Centering Theory in Naturally-Occurring 
Discourse. Philadelphia, PA, May 1993. 
Barwise, J., and Perry, J. (1983). Situations 
and Attitudes. MIT Press. 
Brennan, Susan E. (In press). "Centering 
attention in discourse." Language and 
Cognitive Processes. 
Brennan, Susan E.; Friedman, 
Marilyn Walker; and Pollard, Carl J. 
(1987). "A centering approach to 
pronouns." In Proceedings, 25th Annual 
Meeting of the ACL. Stanford, CA. 155-162. 
Cote, Sharon (1992). "Discourse functions of 
two types of null objects in English." In 
Linguistic Society of America Annual 
Meeting, 12. 
Cote, Sharon (1993). "Ranking and 
forward-looking centers." In Workshop on 
Centering Theory in Naturally-Occurring 
Discourse. Philadelphia, PA, May 1993. 
Di Eugenio, Barbara. (1990). "Centering 
theory and the Italian pronominal 
system." In Proceedings, 13th International 
Conference on Computational Linguistics 
26 Kameyama's dissertation research (Kameyama 1985) was carried out contemporaneously with the 
extensions from our 1983 paper to Grosz, Joshi, and Weinstein (1986). This work was the first to 
establish a correspondence between pronouns in English and zero anaphors in Japanese with respect to 
interactions with centering. 26 Turan's recent work (Turan 1995) is about the realization of subjects in Turkish and not centers in 
particular. She has studied the distribution of full NP, overt pronoun, and zero pronoun subjects in 
Turkish and shows that speakers choose one over the other when all are allowed by the grammar in 
accordance with the salience of the discourse entity represented following a centering theory account 
of salience. 
223 
Computational Linguistics Volume 21, Number 2 
(COLING-90). Helsinki, 270-275. 
Donnellan, K. S. (1966). "Reference and 
definite description." The Philosophical 
Review, 75. 
Gordon, Peter C., and Chan, Davina (in 
press). "The effects of referring 
expressions and passivisation on 
processing sentences in coherent 
discourse." In CUNY Sentence Processing 
Conference. Amherst, MA. 
Gordon, Peter C., and Chan, Davina (1995). 
"Pronouns, passives and discourse 
coherence." Journal of Memory and 
Language, 34, to appear. 
Gordon, Peter C., and Scearce, Kimberly A. 
(1995). "Pronominalization and discourse 
coherence, discourse structure and 
pronoun interpretation." Journal of Memory 
and Language, 34, to appear. 
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. 
Grice, H.P. (1969). "Utterer's meaning and 
intentions." Philosophical Review, 68(2), 
147-177. 
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. (1981). "Focusing and 
description in natural language 
dialogues." In Elements of Discourse 
Understanding, edited by A. Joshi, 
B. Webber, and I. Sag, 85-105. Cambridge 
University Press. 
Grosz, Barbara J., and Hirschberg, Julia 
(1992). "Some intonational characteristics 
of discourse structure." In Proceedings, 
International Conference on Spoken Language 
Processing, Volume 1. Banff, Alberta, 
Canada, October 1992, 429-432. 
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 of Computational 
Linguistics. 44-50. 
Grosz, Barbara J.; Joshi, Aravind K.; and 
Weinstein, Scott (1986). "Towards a 
computational theory of discourse 
interpretation." Unpublished ms. 
Grosz, Barbara J., and Kraus, Sarit (1993). 
"Collaborative plans for group activities." 
In Proceedings, IJCAI-93, Volume 1. 
Chambery, France, September 1993, 
367-373. 
Grosz, Barbara J., and Sidner, Candace L. 
(1986). "Attentions, intentions and the 
structure of discourse." Computational 
Linguistics, 12, 175-204. 
Hirschberg, Julia, and Ward, Gregory (1991). 
"Accent and bound anaphora." Cognitive 
Linguistics, 2(2), 101-121. 
Hoffman, Beryl, and Turan, Umit (1993). 
"Zero and overt pronouns in Turkish." In 
Workshop on Centering Theory in 
Naturally-Occurring Discourse. 
Philadelphia, PA, May 1993. 
Hudson, Susan B.; Tanenhaus, Michael K.; 
and Dell, Gary S. (1986). "The effect of 
the Discourse Center on the local 
coherence of a discourse." In Proceedings, 
Eighth Annual Conference of the Cognitive 
Society. 96-101. Lawrence Erlbaum. 
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, Rochester, NY. 
Hurewitz, Felicia, and Linson, Brian (1993). 
"A discourse analysis of raising 
constructions with to seem." In Workshop 
on Centering Theory in Naturally-Occurring 
Discourse. Philadelphia, PA, May 1993. 
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 Proceedings, 
International Joint Conference on Artificial 
Intelligence. 385-387. 
Kameyama, Megumi (1985). Zero anaphora: 
The case of Japanese. Doctoral dissertation, 
Stanford University, Linguistics 
Department, Stanford, CA. 
Kameyama, Megumi (1986). "A 
property-sharing constraint in centering." 
In Proceedings, 24th Annual Meeting of the 
Association for Computational Linguistics. 
New York, NY, 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. CSLI. Also 
available as University of Pennsylvania 
Technical Report MS-CIS-86-60. 
Kameyama, Megumi (1993). "Intersentential 
centering." In Workshop on Centering 
Theory in Naturally-Occurring Discourse. 
Philadelphia, PA, May 1993. 
Kehler, Andrew (1993). "The effect of 
establishing coherence in ellipsis and 
anaphora resolution." In Proceedings, 31st 
224 
Barbara J. Grosz et al. Centering 
Annual Meeting of the Association for 
Computational Linguistics. Columbus, OH, 
June, 1993, 62-69. 
Kripke, S. (1977). "Speaker's reference and 
semantic reference." In Contemporary 
Perspectives in the Philosophy of Language, 
edited by P. French, T. Uehling, and 
H. Wettstein, 255-276. University of 
Minnesota Press. 
Lochbaum, K. E. (1994). Using collaborative 
plans to model the intentional structure of 
discourse. Doctoral dissertation, Harvard 
University, Cambridge, MA. Also 
published as Technical Report No. 
TR-25-94, Center for Research in 
Computing Technology. 
Nakatani, Christine (1993). "Accenting on 
pronouns and proper names in 
spontaneous narrative." In European 
Speech Communication Association Workshop 
on Prosody. Lund, Sweden, September 
1993, 164-167. 
Passonneau, Rebecca J. (1991). "Some facts 
about centers, indexicals and 
demonstratives." In Proceedings, 29th 
Annual Meeting of the Association for 
Computational Linguistics, Berkeley, 63-70. 
Passonneau, Rebecca J. (in press). "Getting 
and keeping the center of attention." In 
Challenges in Natural Language Processing, 
edited by R. Weischedel and M. Bates. 
Cambridge University Press. Also 
available as Technical Report 
CUCS-060-90, Dept. of Computer Science, 
Columbia University, New York, NY. 
Prince, Ellen E (1994). "Subject-prodrop in 
Yiddish." In In Focus and Natural Language 
Processing, Volume 1: Intonation and 
Syntax, edited by P. Bosch, 159-174. 
Working Papers of the IBM Institute for 
Logic and Linguistics 6. 
Prince, Ellen E, and Walker, Marilyn A. 
(1995). "A bilateral approach to 
givenness: a hearer-status algorithm and a 
centering algorithm." In Proceedings, 4th 
In ternational Pragmatics Conference. 
Benjamins, to appear. 
Rambow, Owen (1993). "Pragmatic aspects 
of scrambling and topicalization in 
German." In Workshop on Centering Theory 
in Naturally-Occurring Discourse. 
Philadelphia, PA, May 1993. 
Reichman, Rachel (1985). Getting Computers 
to Talk Like You and Me. MIT Press. 
Roberts, Craige (1993). "Centering and 
anaphora resolution in discourse 
representation theory." In Workshop on 
Centering Theory in Naturally-Occurring 
Discourse. Philadelphia, PA, May 1993. 
Sidner, Candace L. (1979). Towards a 
computational theory of definite anaphora 
comprehension in English discourse. Doctoral 
dissertation, Artificial Intelligence 
Laboratory, Massachusetts Institute of 
Technology, Cambridge, MA. June 1979. 
Technical Report 537. 
Sparck Jones, Karen (1993). "How do I 
center large-scale text structure." In 
Workshop on Centering Theory in 
Naturally-Occurring Discourse. 
Philadelphia, PA, May 1993. 
Suri, Linda, and McCoy, Kathleen (1993). 
"Comparing focusing and centering and 
problems with complex sentences." In 
Workshop on Centering Theory in 
Naturally-Occurring Discourse. 
Philadelphia, PA, May 1993. 
Turan, Umit (1995). Null vs. overt subjects in 
Turkish: A centering approach. Doctoral 
dissertation, University of Pennsylvania, 
Philadelphia, PA. 
Walker, Marilyn A. (1989). "Evaluating 
discourse processing algorithms." In 
Proceedings, 27th Annual Meeting of the 
Association for Computational Linguistics, 
251-261. 
Walker, Marilyn A. (1993). "Initial contexts 
and shifting centers." In Workshop on 
Centering Theory in Naturally-Occurring 
Discourse. Philadelphia, PA, May 1993. 
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. 
Walker, Marilyn A.; Iida, Masayo; and Cote, 
Sharon (1994). "Japanese discourse and 
the process of centering." Computational 
Linguistics, 20(2), 193-232. 
Webber, Bonnie Lynn (1978). A formal 
approach to discourse anaphora. Doctoral 
dissertation, Harvard University, 
Cambridge, MA. Garland Press. 
Yongkyoon, N. (1991). "A centering 
approach to the *\[case\]\[topic\] restriction 
in Korean." Linguistics, 29, 653-668. 
Ziv, Yael, and Grosz, Barbara J. (1994). 
"Right dislocation and attentional state." 
In Papers from the Israel Association of 
Theoretical Linguistics Meetings, edited by 
R. Buchalla and A. Mittwoch, 184-199. 
Akademon Press. 
225 

