Modes of Discourse - Smith Carlota (2003)

This page intentionally left blank
MODES OF DISCOURSE
In studying discourse, the problem for the linguist is to find a fruitful level
of analysis. Carlota Smith offers a new approach with this study of discourse
passages, units of several sentences or more. She introduces the key idea of
the “Discourse Mode,” identifying five modes: Narrative, Description, Report,
Information, Argument. These are realized at the level of the passage, and cut
across genre lines. Smith shows that the modes, intuitively recognizable as
distinct, have linguistic correlates that differentiate them. She analyzes the
properties that distinguish each mode, focusing on grammatical rather than
lexical information. The book also examines linguistically based features that
appear in passages of all five modes: topic and focus, variation in syntactic
structure, and subjectivity, or point of view. Operating at the interface of syntax,
semantics, and pragmatics, the book will appeal to researchers and graduate
students in linguistics, stylistics, and rhetoric.
c a r l o ta s. s m i t h is Centennial Professor of Linguistics at the University

of Texas. She is the author of The Parameter of Aspect (second edition, 1997).
Her current research is at the interface of syntax, semantics, and pragmatics,
especially in texts.
In this series
60 s a r a h m . b. fag a n The syntax and semantics of middle constructions: a study with
special reference to German
61 anjum p. saleemi Universal grammar and language learnability
62 stephen r. anderson A-Morphous morphology
63 lesley stirling Switch reference and discourse representation
64 henk j. verkuyl A theory of aspectuality: the interaction between temporal and
atemporal structure
65 e v e v. c l a r k The lexicon in acquisition
66 anthony r. warner English auxiliaries: structure and history
67 p. h. matthews Grammatical theory in the United States from Bloomfield to
Chomsky
68 ljiljana progovac Negative and positive polarity: a binding approach
69 r. m. w. dixon Ergativity
70 ya n hua ng The syntax and pragmatics of anaphora
71 k n u d l a m b r e c h t Information structure and sentence form: topic, focus, and the
mental representations of discourse referents
72 luigi burzio Principles of English stress
73 john a. hawkins A performance theory of order and constituency
74 alice c. harris and ly l e c a m p b e l l Historical syntax in cross-linguistic
perspective
75 liliane haegeman The syntax of negation
76 pau l g o r r e l l Syntax and parsing
77 guglielmo cinque Italian syntax and universal grammar
78 h e n r y s m i t h Restrictiveness in case theory
79 d. r o b e r t l a d d Intonational phonology
80 a n d r e a m o r o The raising of predicates: predicative noun phrases and the theory of
clause structure
81 r o g e r l a s s Historical linguistics and language change
82 john m. anderson A notional theory of syntactic categories
83 bernd heine Possession: cognitive sources, forces and grammaticalization
84 nomi erteschik-shir The dynamics of focus structure
85 john coleman Phonological representations: their names, forms and powers
86 christina y. b ethin Slavic prosody: language change and phonological theory
87 barbara dancygier Conditionals and prediction time, knowledge and causation
in conditional constructions
88 c l a i r e l e f e b v r e Creole genesis and the acquisition of grammar: the case of
Haitian Creole
89 heinz giegerich Lexical strata in English: morphological causes, phonological
effects
90 k e r e n r i c e Morpheme order and semantic scope: word formation and the Atha-
paskan verb
91 a . m . s . m c m a h o n Lexical phonology and the history of English
92 m at t h e w y. c h e n Tone sandhi patterns across Chinese dialects
93 g r e g o r y t. s t u m p Inflectional morphology: a theory of paradigm structure
94 j o a n b y b e e Phonology and language use
95 laurie bauer Morphological productivity
96 thomas ernst The syntax of adjuncts
97 e l i z a b e t h c l o s s t r au g o t t and richard b. dasher Regularity in semantic
change
98 maya hickman Children’s discourse: person, space and time across languages
99 d i a n e b l a k e m o r e Relevance and linguistic meaning: the semantics and pragmatics
of discourse markers
100 i a n r o b e r t s and anna roussou Syntactic change: a minimalist approach to
grammaticalization
101 donka minkova Alliteration and sound change in early English
102 m a r k c . b a k e r Lexical categories: verbs, nouns and adjectives
Earlier titles not listed are also available
CAMBRIDG E S T U D I E S I N L I N G U I S T I C S
General editors: p. austin, j. b r e s na n, b. comrie ,
w. dressler, c . j. e w e n, r . l a s s, d. lightfoot ,
i. roberts , s. romaine , n. v. s m i t h
Modes of Discourse
MODES OF
D I S C OURSE
T HE L OCA L STRUCTURE
OF T E X TS
CA RL OTA S . S M ITH
University of Texas
  
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press

The Edinburgh Building, Cambridge  , United Kingdom
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521781695
© Carlota S. Smith 2003
This book is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
First published in print format 2003
-
isbn-13 978-0-511-06906-2 eBook (EBL)
-
isbn-10 0-511-06906-5 eBook (EBL)
-
isbn-13 978-0-521-78169-5 hardback
-
isbn-10 0-521-78169-8 hardback
Cambridge University Press has no responsibility for the persistence or accuracy of

s for external or third-party internet websites referred to in this book, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
To John
Contents
Preface page xiii
Introduction 1
part i discourse structure
1 The study of discourse 7

1.1 Discourse Modes 8
1.2 Approaches to the study of texts 9
1.3 Overview of key ideas 12
1.4 Conclusion 19
2 Introduction to the Discourse Modes 22

2.1 The entities introduced in texts 23
2.2 Text progression in the temporal modes 25
2.3 Text progression in the atemporal modes 31
2.4 Foreground and background in text passages 34
2.5 Rhetorical and linguistic background 38
3 Text representation and understanding 49

3.1 The pragmatic background for discourse interpretation 49
3.2 Types of inference 51
3.3 Mental models and representations 55
3.4 The analysis of text passages in Discourse Representation
Theory 56
part ii linguistic analysis of the discourse

modes
4 Aspectual information: the entities introduced

in discourse 67
4.1 Aspectual categories 67
4.2 General Statives 72
ix
x Contents
4.3 Abstract Entities 74

4.4 Linguistic correlates of situation entities 75
4.5 Aspectual information in Discourse Representation
Structures 82
5 Temporal and spatial progression 92

5.1 Sentences in context: patterns of tense interpretation 92
5.2 Introduction to the temporal system of English 99
5.3 Tense interpretation in Discourse Representation Theory 105
5.4 Spatial information in language 115
5.5 Features of temporal expressions 118
6 Referring expressions in discourse 123

6.1 Atemporal text progression and Primary Referents 123
6.2 Referring expressions 132
6.3 The familiarity status of referring expressions 141
6.4 Referring expressions and Discourse Representation
Theory 149
part iii surface presentational factors
7 Subjectivity in texts 155

7.1 Responsibility for subjectivity 156
7.2 Expressions of communication 159
7.3 Contents of mind 162
7.4 Evaluative and evidential subjectivity 163
7.5 Perception and perspectival sentences 169
7.6 Formalizing the interpretation of subjectivity 175
7.7 Summary and conclusion 183
8 The contribution of surface presentation 185

8.1 Presentational factors 187
8.2 The topic–comment partition 188
8.3 The Focus–Background partition 201
8.4 Dual partitioning and the representation of topic and focus 209
8.5 Conclusion 212
9 Non-canonical structures and presentation 213

9.1 Non-canonical constructions 214
9.2 Argument constructions 215
9.3 Non-argument preposing: Adjuncts 228
Contents xi
9.4 Multi-clause sentences 233

9.5 Paragraphs 236
9.6 Presentational information in Discourse Representation
Structure 238
part iv discourse modes and their context
10 Information in text passages 243

10.1 The Discourse Modes 243
10.2 Multiple analyses of text passages 247
10.3 Formalization in Discourse Representation Structures 254
11 Discourse structure and Discourse Modes 258

11.1 Organizing principles of texts 258
11.2 Discourse relations 260
11.3 Discourse Modes and text structure 265
11.4 Conclusion 266
Appendix A: The Texts 267
Appendix B: Glossary 286
References 294
General index 314
Index of names 318
Preface
This work would not have been possible without many kinds of support that I
have received from individuals and institutions. I am grateful to them all. The
New York Community Trust has supported much of my research on discourse.
It also funded three small conferences on discourse at the University of Texas
which advanced the work. I received a Faculty Research Award from the Uni-
versity of Texas Research Institute in 1994, which enabled me to pursue this
project. During that period I spent some time as a Visiting Scholar at the Maison
Suger of the CNRS in Paris. I was the beneficiary of a Dean’s Fellow award
from the College of Liberal Arts in 1998.
Parts of this work have been presented at conferences and colloquia. I
benefited greatly from the discussions that followed, as well as the presen-
tations themselves. They include three conferences on discourse structure
at the University of Texas; an International Round Table on The Syntax of
Tense and Aspect at Université de Paris, 2000; a symposium on Information
Structure in a Cross-Linguistic Perspective at the University of Oslo, 2000;
a conference on Linguistics in the Next Decade at the Academia Sinica,
Taipei, Taiwan, 2000; a Linguistics Colloquium at the University of Siena
in 1998; an invited lecture series at the City University of Hong Kong in
1998.
I would like to thank the members of a seminar on text structure that I taught
at the University of Texas in the Fall of 2001: Behrad Aghei, Robert M. Brown,
Pascal Denis, Q Wan Kim, Christian Rathmann, Brian Reese, Dong-Rhin Shin,
Cholthicha Sudmuk, and Jiun-Shiung Wu. Their questions and comments on
an earlier version of this manuscript were extremely helpful. I owe a special
debt to Keith Walters, who read the entire manuscript and gave me many valu-
able suggestions on content, presentation, and relevant material. I also thank
Pascal Denis, my research assistant during the critical stages of manuscript
preparation; he asked penetrating questions and provided very useful technical
advice. Finally, I thank the people with whom I have worked at the Cambridge
xiii
xiv Preface
University Press in the publication of this book, especially Neil Smith for
shocking, useful comments on an earlier version of the manuscript; I thank
Leigh Mueller, who did the copy-editing, for her careful work and good humor;
and I thank Jackie Warren for shepherding the book through the publication
process.
Introduction
This book is a partial answer to the question: what can close linguistic analysis
bring to the understanding of discourse? Discourse studies have focused on
pragmatic factors such as genre expectations, discourse coherence relations, and
inference. In part this has been a natural reaction to earlier, rather unsuccessful
attempts to apply the techniques of linguistic analysis beyond the sentence.
The current emphasis also follows from increased understanding of the area of
pragmatics, and of the role of context in language use and interpretation.
It has sometimes seemed, though, that nothing at all is conveyed by linguistic
forms, while everything is due to pragmatics or lexical content. I attempt to
right the balance here, at least in part. I propose a local level of discourse, the
Discourse Mode, which has linguistic properties and discourse meaning. I posit
five modes: Narrative, Report, Descriptive, Information, and Argument.
The Discourse Modes are classes of discourse passages, defined by the enti-
ties they introduce into the universe of discourse and their principle of progres-
sion. The discourse entities are essentially aspectual. They include the familiar
Events and States, and some less-familiar categories. The Discourse Modes
grew out of my work on aspect and tense. In studies of situation types in dis-
course, I noticed interesting differences between passages of different types.
Investigating further, I arrived at the Discourse Modes. If I am right about their
contribution to discourse, they make it clear that temporality is one of the key
sub-systems in language.
I characterize the modes by their linguistic features, that is, grammatical
forms with consistent interpretations. The linguistic features of the modes are
covert categories in the sense of Whorf (1956). They are not overtly marked
but they have characteristic patterns of distribution, and of interpretation. These
properties are subtle, but they are demonstrably part of a person’s knowledge
of language. The emphasis throughout this book is on grammatical rather than
lexical features of discourse.
The modes are, therefore, linguistic categories. I was curious to know whether
they would be related to anything in the field of rhetoric. When I looked at the
1
2 Introduction
literature, I found to my surprise that the Discourse Modes have counterparts

in rhetorical tradition. The Modes correspond to “text types” which have been
recognized as important in discourse but not analyzed before, I believe, in terms
of their linguistic properties. This correspondence is independent validation of
the idea of Discourse Modes, I think.
One major goal of this book, then, is to present and explore the notion of
Discourse Modes as a significant category in discourse.
Another goal is to use grammatical forms as a tool for exploring the com-
plexity of discourse. I wanted to understand and make precise the different
kinds of information that a discourse conveys. It has always intrigued me that
one recognizes immediately whether an example is constructed or “real.” Con-
structed examples seem thin, simplistic. I conjecture that one reason is density
of information, or lack of it. We construct examples to investigate or demon-
strate a particular point – say, anaphoric or tense patterns – and our examples
convey information about that point. The constructed examples have little other
information, however. Natural texts, in contrast, convey information of several
kinds.
Using passages of Discourse Modes as a basis, I study two other kinds of in-
formation conveyed by the sentences of a text: subjectivity and surface structure
presentation.
I argue that we can distinguish “subjective” sentences from others on the
basis of a set of linguistic forms that convey a particular voice. By subjectivity
I include such notions as point of view, perspective, and content of mind. When
we encounter such forms, we ascribe responsibility for them to the author
or another source. I present a “composite” account, stating rules that look at
subjective forms in a sentence and in context. The rules assign the role of
Responsible Source to a participant in the text situation or to the author.
Surface structure presentation concerns how syntactic structures give cues
to the organization of a sentence, and how it affects continuity in the sentences
of a text. I take the notion of sentence topic as the main organizing factor.
The area is a thorny one but I hope to have found a useful synthesis. I use
the notions of topic and strong focus to examine the presentational features of
non-canonical syntactic structures. I bring together facts and theories about the
discourse effects of syntax, although the account is not exhaustive. There are
many studies of single structures, or closely related structures. After looking at
such studies I analyze the syntactic structures in a group of texts, with special
attention to the combinations that appear.
The study is at the interface of syntax, semantics, and pragmatics. It is in-
formed by some of the insights of Cognitive Science, especially the analytic
Introduction 3
stance. I attempt to understand and explain some of the complexity of discourse.

I also attempt to formalize the analysis, using the dynamic framework of Dis-
course Representation Theory, due originally to Kamp (1981) and Heim (1982).
One of the questions that I deal with in the book is what kinds of information
can and cannot be analyzed within this framework.
What is new here is the distinction between grammatical and lexical informa-
tion for texts; the information-based, composite feature approach to presentation
and point of view; the Discourse Modes themselves.
The inquiry was carried out with a group of natural texts that I read, analyzed,
and used for examples. They range in length from books of several hundred
pages to short newspaper articles. I worked with what seemed intuitively to be
good examples of different genres. The core set consists of twenty texts, which
I consulted often. The core was supplemented by other texts that I remembered
or encountered by chance in the course of doing the work. I am aware that this
is a small sample and can only be taken as suggestive, perhaps representative.
Larger-scale studies are needed to supplement this exploratory work.
Most of the texts offer examples of more than one mode, as expected. One
of the arguments that I make in favor of the Discourse Modes is that texts are
quite varied, usually having passages of more than one mode. Some of the texts
that I used are presented in an appendix.
This work is intended for linguists of different stripes, and others interested
in discourse. Since the book includes formalization, some sections are quite
technical, but they are, I hope, made reasonably accessible by the explanations.
I Discourse structure
1 The study of discourse
This book studies discourse passages from a linguistic point of view. Discourse
is made up of sentences, and through linguistic analysis we have learned a good
deal about them. The perspective of linguistics, however, can’t be used directly
to study an entire discourse. Novels, histories, arguments, and other types of
discourse are activities with their own character and conventional structure.
Receivers draw on discourse knowledge to construct interpretations.
The first problem for the linguist interested in close study of discourse, then,
is to find a fruitful level for analysis. Larger units are organized primarily
by convention and expectation. I will work more locally, at the level of the
passage. There are intuitive differences between the passages of a discourse.
People recognize passages of several kinds, namely Narrative, Description,
Report, Information, and Argument. The intuitions are linguistically based: the
passages have a particular force and make different contributions to a text. They
can be identified by characteristic clusters of linguistic features. I shall say that
a passage of text with certain features realizes a particular “Discourse Mode.”
The Discourse Mode is appropriate for close linguistic analysis, because at this
level linguistic forms make a difference. Discourse Modes appear in texts of all
types of activity, or genres. I use the terms “discourse” for spoken and written
material, “text” for written material.
The Discourse Modes constitute an interesting level of text structure. I ana-
lyze them in two ways. I first discuss the differences between text passages of
each mode. I then look at passages in terms of subjectivity and surface structure
presentation, features that the modes have in common. Much of the analysis is
formalized in the framework of Discourse Representation Theory.
Part I of this book discusses the Discourse Modes and lays out the context
for the inquiry. Part II presents the linguistic characterization of the modes,
emphasizing the differences between them. Part III discusses subjectivity and
surface structure presentation across modes. Text passages are thus considered
from complementary points of view in the second and third parts of the book.
The different analyses are brought together in Part IV.
7
Section 1.1 of this chapter introduces the Discourse Modes; 1.2 outlines the
approach to texts and analysis that I take in this book; 1.3 presents the main ideas
to be developed later, with examples of passages analyzed for different kinds
of information that they convey; 1.4 concludes with summary characterization
of the modes and brief comments on the importance of temporality for human
beings.
1.1 Discourse Modes

I recognize five modes: Narrative, Description, Report, Information, and Argu-
ment. This list is not exhaustive, but I think it covers the major modes that appear
in texts. I do not deal with conversation, nor procedural discourse.1 The modes
can be characterized with two features. Each mode introduces certain types of
situation – Event, State, generalization, abstraction – into the universe of dis-
course. The modes also have characteristic principles of progression, temporal
and atemporal. There are linguistic correlates to these features. Knowledge of
one’s language includes knowledge of these forms and meanings, some of them
quite subtle.
The notion of Discourse Mode accounts for the variety that one finds in texts.
Actual texts are usually not monolithic. In narratives, for instance, the significant
unit is the episode: a group of Events and States in sequence that are bound
together by a unifying theme. Narrative episodes, however, rarely consist only
of sequence. There are also descriptive passages, and perhaps argument as well.
Similarly the expository genres often have narrative sequences which support
the main line of argument. Narrative, description, and argument make different
contributions to a text, and have different linguistic features and interpretations.
Each constitutes a distinct Discourse Mode.
The short passages below exemplify the five modes; they are taken from a
group of texts that will be discussed repeatedly throughout this book.2 Sources
for the natural examples are listed at the end of each chapter; some of the texts
are reproduced in Appendix A.
(1) She put on her apron, took a lump of clay from the bin and weighed off enough
for a small vase. The clay was wet. Frowning, she cut the lump in half with a
1. Persuasive discourse is not listed separately. All genres and modes of discourse may have a
persuasive component.
2. The texts were chosen to provide a variety of examples. They include short stories, novels,
books, articles from journals and newspapers. They were analyzed intensively by the author.
Appendix A provides a list of the texts and significant fragments from the ones most often
used.
cheese-wire to check for air bubbles, then slammed the pieces together much
harder than usual. A fleck of clay spun off and hit her forehead, just above her
right eye.
(2) In the passenger car every window was propped open with a stick of kindling
wood. A breeze blew through, hot and then cool, fragrant of the woods and
yellow flowers and of the train. The yellow butterflies flew in at any window,
out at any other.
(3) Near a heavily fortified Jewish settlement in the Gaza Strip, an Israeli soldier
and a Palestinian policeman were wounded as Palestinian protests for the
release of 1,650 prisoners degenerated into confrontations. Israeli military
officials say they are investigating the source of fire that wounded the soldier.
(4) Thanks to advanced new imaging techniques, the internal world of the mind
is becoming more and more visible. Just as X-ray scans reveal our bones, the
latest brain scans reveal the origin of our thoughts, moods, and memories.
Scientists can observe how the brain registers a joke or experiences a painful
memory.
(5) The press has trumpeted the news that crude oil prices are three times higher
than they were a year ago. But it was the $10 or $11 price of February 1999,
not the one today, that really deserved the headlines.
In order, these fragments exemplify the modes of Narrative, Description, Report,

Information, and Argument. Passages of the Discourse Modes are linguistic
units, since they have recognizable linguistic features. They also have rhetorical
significance. In fact the Discourse Modes are text units both linguistically and
notionally. They function as a bridge between the sentences of a text and the
more abstract structures that it evokes. The relations between Discourse Modes
and such abstract structures are discussed in Chapter 11.
In close analysis of a text one considers the linguistic forms, asking what
information is conveyed by the sentences and sentence sequences of a discourse.
Since this study is limited to written texts I do not discuss such matters as stress
and intonation, audience, or specific setting.
The inquiry shows that the information in a text is varied at the local level,
providing multiple meanings. Thus the analysis in this book is a partial expli-
cation of text complexity.
1.2 Approaches to the study of texts

1.2.1 Linguistic features and discourse structure
Discourse is a human activity with language at the center. Types of discourse
are usefully grouped into genres, each genre with its own purpose, structure,
and conventions. Knowing the genre of a discourse provides indispensable cues
to its structure.
The idea of discourse as a type of activity helps us to discard our customary

expectations and to analyze it afresh (Levinson 1979/1992). The idea harks
back at least to Wittgenstein’s “language games.” The term was coined to draw
attention to language as part of action. This passage from Wittgenstein’s Brown
Book outlines a simple language game:
Its function is the communication between a builder A and his man B. B has to
bring A building stones. There are cubes, bricks, slabs, beams, columns. The
language consists of the words “cube,” “brick,” “slab,” “beam,” and “column.”
A calls out one of these words, upon which B brings a stone of a certain shape.
In a different game, calling out the same word would have different force. A and
B might be archaeologists investigating a site, for instance, and A might call
out a word – column, brick – to convey to B what he has found. To interpret A’s
utterances, we have to understand the language game being played: the activity
and the role that language plays in it. Wittgenstein glosses the term “language
game” as referring to “the whole, consisting of language and the actions into
which it is woven.”3
Knowing the language game, or genre, requires knowledge of an activity
as a whole. This knowledge is not conveyed by linguistic forms. The global
structure of a discourse is rarely if ever stated explicitly. People understand
discourse with different kinds of information, including what is conveyed by
linguistic forms. They use general information about genre and principles of
communication, and specific information about a particular case. There are
some differences among genres. Scientific articles and textbooks often lay out
the specific relations between their parts, whereas literary genres tend to be less
explicit.
1.2.2 The linguistic approach

At the level of the passage, close linguistic analysis of discourse can be fruit-
fully pursued. I am interested in working out information that is conveyed
by linguistic forms, directly and indirectly. To interpret text passages, people
draw on syntactic, semantic, and pragmatic knowledge; see Chapter 3 for some
discussion.
I rely on two insights in the analysis. The first is that linguistic meaning is
often due to a group of forms – a composite – rather than to a single form.
3. R. Rhees, in a Preface to The Blue and Brown Books, says that Wittgenstein introduced the
notion of language games “in order to shake off the idea of a necessary form of language . . . He
is insisting that . . . understanding is not one thing: it is as various as the language games
themselves.” The Blue and Brown Books were dictated in 1934–35 and published in 1958.
Whether a sentence expresses an Event or a State, for instance, depends on the

composite of the verb and its arguments, as well as adverbials in the sentence.
Together, these forms express a State or an Event. The composite approach is
used throughout the book.
The second insight is that grammatical terms such as tense and pronouns
often have two different functions in discourse. Besides the direct information
that they code in a sentence, they give cues to local text structure. Maintenance
or change of pronouns, for instance, often indicates continuity or change of
direction. In this way grammatical forms contribute to the pattern of a text.
I work with surface syntactic structures. I assume a generative syntax with
movement rules. The surface syntax makes available the constituent structure of
a sentence, grammatical relations such as subject and object, and the semantic
features associated with particular morphemes and constituents. For specificity,
I use structures roughly following generative theory of the 1990s, somewhat
simplified.4 I use only surface structures in this book: no syntactic rules are
stated.
I take the stance of the receiver of a text. I assume that receivers assem-
ble and interpret the different cues that a text contains. They include lexical
and semantic choices, syntactic and information structure, patterns within and
across sentences, cue words, typography. The preferred interpretation is the
one most compatible with all the information available. I do not attempt to
model the actual processes involved nor the shifts in attention as readers make
their way through a text. The analysis is not a psycholinguistic one but an
idealization, in the tradition of modern linguistics. The interpretation is given
in the form of an ongoing semantic-pragmatic structure, in the framework of
Discourse Representation Theory. This theory is explicitly formulated to deal
with discourse.
Sentences in discourse have a dual nature that has been difficult to understand.
The difficulty is that sentences are self-contained units from a certain point of
view; but for interpretation they depend on linguistic and extra-linguistic con-
text. This dependence cannot be captured simply by making connections be-
tween sentences. The meaning of a sentence often requires information from the
context. The realization that sentence meaning can be elucidated only in con-
text is the leading idea of Discourse Representation Theory (Kamp 1981; Heim
4. The syntactic surface structures that I use are based on such works as Culicover (1997) which
are in the Principles and Parameters generative framework. I do not take a position on types of
movement rules or the mechanisms that trigger movement.
There is no level of Logical Form in this approach: the semantic interpretation is developed
in the Discourse Representation Theory framework.
1982). As discourse is dynamic, so representations must be: new information

is added, familiar entities are referred to, situations change. The representation
is updated as a discourse develops.
Rules of the theory construct a representation from information in the sur-
face structure of sentences. The representation gives the conceptual information
that a receiver grasps in understanding discourse. Text representation consists
of “discourse entities” for individuals, situations, and times; and conditions
that characterize the entities. In some cases a discourse entity is embedded
in a sub-structure and is not available as antecedent for anaphoric reference.
Embeddings represent the scopal effects of operators such as negation, quan-
tifiers, and modality. There is a second, truth-conditional level, at which the
structure is interpreted within a formal model.
Information about Discourse Mode and some aspects of presentational struc-
ture will be encoded in Discourse Representation Structures. I introduce the
theory and its representations in Chapter 3; later chapters formalize the analysis
in the structures of the theory. Given the richness of the information that is con-
veyed in sentences, an interesting question arises: how much of the information
conveyed by a sentence should survive in representations of text meaning? This
question is particularly difficult for those aspects of meaning that are clearly not
truth-conditional. The question will be discussed from time to time throughout
the book.
1.3 Overview of key ideas

I introduce four key ideas explored in later chapters of this book, and then
present multiple analyses of text passages in which all of them are exhibited.
1.3.1 Situation type, text progression, subjectivity, surface

structure presentation
Types of situation: the sentences of a text introduce situations into the universe
of discourse. Events and States are the basic types in most studies of aspect and
discourse. Adding to this tradition, I recognize General Statives and Abstract
Entities as two other types. General Statives are expressed by generic and
generalizing sentences. They invoke patterns of Events and States rather than
particular situations. The complement clauses of certain predicates refer to
facts and propositions, which are Abstract Entities. Situations of all types are
entered in the structures of Discourse Representation Theory as entities, along
with individuals and times. They are known as “situation entities,” discussed
in Chapters 2 and 4.
Text progression: there are several principles of text progression among the
Discourse Modes. In Narrative, situations are related to each other and dynamic
Events advance narrative time. In Reports, situations are related to Speech Time
and time progresses forward and backward from that time. In Description,
time is static and the text progresses in spatial terms through the scene de-
scribed. The Information and Argument modes are atemporal and progress by a
metaphoric path through the domain of the text. Text progression is discussed in
Chapters 5 and 6.
Subjectivity: I distinguish “subjective” sentences from others on the basis of
a set of grammatical forms. All forms of subjectivity convey access to mind –
either the mind of the writer or a text participant – through communication,
mental state, perception, and perspective. For each subjective expression a
Responsible Source must be identified. The main predicate of a sentence may
indicate communication, mental state, evaluation. The arguments of the predi-
cate may indicate perspective with deictic pronouns and reflexives. Modals, ad-
verbials, parentheticals indicate evaluation and evidentiality. Subjective forms
appear in passages of all the Discourse Modes. Subjectivity is discussed in
Chapter 7.
Surface structure presentation: presentational features organize the informa-
tion in a sentence, usually into topic and comment, focus and background.
These features appear in all text passages. I develop an integrative approach,
drawing on current linguistic insights and traditional Prague School views.
Presentational information depends on syntactic surface structure, the linear
and grammatical position of phrases. I will be particularly interested in pre-
sentational progression, which tracks the shifts from one topic to another in
the sentences of a discourse.5 The topic phrase gives the referent that a sen-
tence is about. The main criteria for identifying the topic phrase are salience,
coreferentiality, and continuity.6 See Chapters 8 and 9 for discussion.
Presentational structure is also known as “information structure”; I prefer the
term “presentation” because texts convey other kinds of information besides that
of topic, focus, and associated notions such as familiarity status.
5. The notion of presentational progression is unlike the shifting of attention in reading a text,
studied in psycholinguistics and artificial intelligence. The processes involved in understanding
are beyond the scope of this discussion.
6. Local continuity looks for a topic phrase that is coreferential with the topic phrase immediately
preceding. Global continuity looks for a topic phrase that is coreferential with other phrases in the
context. These factors are recognized in other approaches to local relations between sentences
such as Centering Theory, which ranks local continuity above global continuity (Walker et al.
1998). See Chapter 6 for discussion.
1.3.2 Multiple analysis of text passages

Discourse Mode, subjectivity, and presentational progression are the main con-
cerns of this book. They convey information that complements the lexical and
rhetorical aspects of a text. This section offers passages of three Discourse
Modes, analyzed for these features.
The Discourse Modes differ in the type of situation entities they introduce,
and their principle of text progression. Forms of subjectivity and presentational
progression are found in all the modes of discourse. I discuss these features
of passages in the Narrative, Report, and Information modes. The passages are
given more than once. The first version shows the situation entities and text
progression. Information about subjectivity is added next, and then presenta-
tional progression. The final version displays together the different kinds of
information conveyed by a passage.
The first passage I discuss is in the Narrative mode. Characteristically, a
narrative introduces Events and States into the universe of discourse. The text
progresses as narrative time advances. This advancement is based on sequence:
we interpret the events of a narrative as occurring in sequence, one after another.
Aspectual and temporal linguistic cues in a passage trigger the interpretation
of advancement. Bounded events advance narrative time; temporal adverbials
also advance it. Event clauses with the perfective viewpoint express bounded
events; the progressive expresses ongoing events. The perfective is conveyed
by the simple form of the verb, the progressive by the auxiliary be+ing (called
vs. was calling). These points are discussed in Chapters 2 and 4.
The narrative passage below introduces Events and States. They are marked
with subscripts for each tensed clause; E = bounded event, S = State. Arrows
preceding a clause indicate temporal advancement. Clauses that are not pre-
ceded by arrows do not advance narrative time. When a sentence has more than
one tensed clause the clauses are distinguished by letter (1a, b, etc.).
(6) Narrative a: situations and text progression

1E → A few days later I called on Dr P and his wife at home, with the score of
the Dichterliebe in my briefcase and a variety of odd objects for the testing of
perception. 2aE → Mrs. P showed me into a lofty apartment, bS which recalled
fin-de-siècle Berlin. 3aS A magnificent old Bösendorfer stood in State in the
centre of the room, bS and all around it were music stands, instruments, scores.
4aS There were books, bS there were paintings, cS but the music was central.
5aE → Dr. P came in, a little bowed, bE and→ advanced with outstretched
hand to the grandfather clock, cE but, hearing my voice,→ corrected himself,
dE and→ shook hands with me. 6aE → We exchanged greetings bE and→
chatted a little of current concerts and performances. 7 Diffidently, aE → I
asked him bS if he would sing.
Each E in this passage advances the narrative: the clauses express bounded
Events. State sentences such as 2a, 3a–b, and 4a–b do not advance narrative
time.
Now I add forms of subjectivity. They indicate access to mind, of either the
author or a participant. There is only one such form in this passage, a verb with
an implicit experiencer argument, given in bold.
(7) Narrative b: situations, text progression, subjectivity

fin-de-siècle Berlin. 3aS A magnificent old Bösendorfer stood in state in the
5aE → Dr. P came in, a little bowed, bE and→ ø advanced with outstretched
hand to the grandfather clock, cE but, ø hearing my voice, → corrected himself,
dE and→ ø shook hands with me. 6aE → We exchanged greetings bE and → ø
The verb recall implies an experiencer (recalled to someone). Since the passage
is in the first person, participant and narrator are the same. We infer that the
narrator perceives the room according to the description in sentences 3 and 4.
The first person pronouns woven into the passage do not convey subjectivity in
the intended sense of access to mind.
Finally, information about surface presentational progression is added. The
topic phrases of each clause, in italics, provide the steps of progression through
the passage.
(8) Narrative c: situations, text progression, subjectivity, topic

5aE → Dr. P came in, a little bowed, bE and→ ø advanced with outstretched
hand to the grandfather clock, cE but, ø hearing my voice,→ corrected himself,
dE and→ ø shook hands with me. 6aE → We exchanged greetings bE and→ ø
The topic phrases are subjects in S1, 2b, 3a, 4c, and the clauses of 5, 6, and
7. This is the most common position for topics. Topic phrases are discussed
in Chapters 8 and 10. S3b and 4a–b are non-canonical structures without topic
phrases; see Chapter 9.
The next example is in the Report mode. Reports are similar to Narrative
in the situations they introduce: Events and States, and sometimes General
Statives. They have a different principle of progression, however. In the Report
mode, situations are related to the time of the report, Speech Time, rather than
to each other. The text progresses as time changes. The linguistic cues to change
are tensed verbs, modals, and adverbs that convey temporal information. In (9)
the adverbials and tensed verbs are underlined.
(9) Report a: situations and text progression

1aE1 A week after Ethiopia started an offensive bE2 that it says is aimed at
ending the two-year-old war, cS1 it is now clear dS2 that the whole of Eritrea
could become a battlefield. 2 With hundreds of civilians fleeing the region, aE3
Colonel Kidane said bE4 Ethiopian soldiers continue to skirmish with Eritrean
soldiers on the run here in western Eritrea.
3aE5 Tonight, Ethiopian officials said bE6 planes bombed the main Eritrean
military training center at Sawa, an American-built base 100 miles west of
Asmara, the capital. 4aE7 The officials also said bs they had taken a village,
Maidema, 30 miles from Asmara, on the way from the western front to the
central front along the disputed border. 5aS3 That is where the next round of
fighting, bS4 already heavy, is generally expected.
The time talked about moves back and forth from past to present, with one
modal future (“could become”) and one past perfect (“had taken”). Both modal
and perfect clauses are stative.
Subjective and presentational features are added in (10). The topic phrases
are italicized; subjective features are in bold.
(10) Report b: situations, text progression, subjectivity, topic

1aE1 A week after Ethiopia started an offensive bE2 that it says is aimed at
ending the two-year-old war, cS1 it is now clear dS2 that the whole of Eritrea
could become a battlefield. 2 With hundreds of civilians fleeing the region, aE3
Colonel Kidane said bE4 Ethiopian soldiers continue to skirmish with Eritrean
soldiers on the run here in western Eritrea.
3aE5 Tonight, Ethiopian officials said bE6 planes bombed the main Eritrean
military training center at Sawa, an American-built base 100 miles west of
Asmara, the capital. 4aE7 The officials also said bE8 they had taken a village,
Maidema, 30 miles from Asmara, on the way from the western front to the
central front along the disputed border. 5aS3 That is where the next round of
fighting, bS4 already heavy, is generally expected.
The passage has deictic and evaluative subjective features. The deictics indicate
the time and place of the reporter (“now,” “here,” “tonight”), in addition to their
locating function. The evaluative “clear” implies an evaluator (clear to someone)
and the modal “could” suggests access to a mind. The Responsible Source is
the author, since no plausible text participant is available. The topic phrase is
the subject in all but two clauses of this passage.
The next example is Information, an atemporal mode. The situation entities
introduced include a significant number of facts and propositions, and gener-
alizing statives. They do not involve particular situations located at a time and
place. Therefore text progression in this mode cannot be based on temporal
or spatial location. Passages in the atemporal modes progress by metaphorical
motion through the semantic domain of the text. Motion, or lack of it, depends
on metaphorical changes of location. We track location in this sense by iden-
tifying a Primary Referent in each tensed clause in a passage, and considering
the location of the primary referents.
The Primary Referent is semantically central in the situation expressed. In
Events, the Primary Referent is what moves or changes. In States, the Primary
Referent is located or characterized; or emergent, dependent on the State for
existence. The Primary Referent of a clause usually coincides with the argument
that has the thematic role of Theme/Patient. Criteria for determining Primary
Referents are discussed in Chapter 6.
The fragment in (11) is an Information passage. It introduces Generalizing
Statives, except for S2b which refers to a Fact. These situations are typical
of the Informative mode. In addition, the Primary Referent phrases are shown
with underlining for each tensed clause. In S2 the extraposed clause is Primary
Referent for the main clause, indicated by the underlining of “S” which precedes
the clause. Ge = Generalizing Stative. Within the extraposed clause the Primary
Referent is also underlined.
(11) Information a: situations and Primary Referents
1aGe When people try to get a message from one individual to another in the
party game “telephone,” bGe they usually garble the words beyond recognition.
2aGe It might seem surprising, then, bFact that mere molecules inside our cells
constantly enact their own version of telephone without distorting the relayed
information in the least.
3Ge Actually, no one could survive without such precise signalling in cells.
4aGe The body functions properly only because bGe cells communicate with
one another constantly. 5Ge Pancreatic cells, for instance, release insulin to tell
muscle cells to take up sugar from the blood for energy. 6aGe Cells of the im-
mune system instruct their cousins to attack invaders, bGe and cells of the
nervous system rapidly fire messages to and from the brain. 7aGe Those
messages elicit the right responses only bGe because they are transmitted
accurately far into a recipient cell and to the exact molecules able to
carry out the directives. 8Ge But how do circuits within cells achieve
this high-fidelity transmission?
Metaphorical progression in the first paragraph moves from “a message” to the

smaller unit “the words” to a particular type of message – “their own version
of telephone.” The Primary Referent in the second paragraph changes from
people to “body,” “cells” of several kinds, “insulin,” then back to messages of
a different nature.
I now add subjective forms to the fragment, which begins an article. Sev-
eral appear in S2 and others are scattered throughout. All evoke the author as
Responsible Source. The evaluative adjective implies an experiencer; in this
context it refers to people in general, including the author. The direct question
of S8 also indicates subjectivity: the author directly addresses the audience.
Example (12) presents the fragment with all the types of information noted:
(12) Information b: situations, Primary Referents, subjectivity

party game telephone, bGe they usually garble the words beyond recognition.
one another constantly. 5Ge Pancreatic cells, for instance, release insulin to tell
muscle cells to take up sugar from the blood for energy. 6aGe Cells of the im-
mune system instruct their cousins to attack invaders, bGe and cells of the ner-
vous system rapidly fire messages to and from the brain. 7aGe Those messages
elicit the right responses only bGe because they are transmitted accurately far
into a recipient cell and to the exact molecules able to carry out the directives.
8Ge But how do circuits within cells achieve this high-fidelity transmission?
The subjective forms in S2 are predicative, deictic, modal, adverbial. They are
part of the main sentence, though they do not involve either topic or Primary
Referent phrases. The concentration of subjective forms in this sentence, the
second of the article, conveys subjectivity which can be maintained with fewer
subjective forms later. S8 is in question form, directly invoking author and
reader. This pattern of subjectivity is fairly typical of informative prose. The
author is not a participant but can be glimpsed from time to time.
Next, information about presentational progression is added, with the topic
phrases in italics:
1.4 Conclusion 19
(13) Informative c: situations, Primary Referents, subjectivity, topic

party game telephone, bGe they usually garble the words beyond recognition.
one another constantly. 5Ge Pancreatic cells, for instance, release insulin to
tell muscle cells to take up sugar from the blood for energy. 6aGe Cells of
the immune system instruct their cousins to attack invaders, bGe and cells of
the nervous system rapidly fire messages to and from the brain. 7aGe Those
messages elicit the right responses only bGe because they are transmitted ac-
curately far into a recipient cell and to the exact molecules able to carry out
the directives. 8Ge But how do circuits within cells achieve this high-fidelity
transmission?
The single instance of non-canonical syntax, the extraposed that-clause in S2,

removes the possibility of a topic phrase in subject position.
The two kinds of progression provide a dual patterning in texts. Topic and
Primary Referent phrases appear in two patterns. Primary Referent phrase and
topic phrase may coincide, or interact in counterpoint. In the latter case the
topic phrase is subject and the Primary Referent phrase is in the predicate.
While topic phrases tend to be subjects, Primary Referent phrases tend to appear
in the predicate unless a clause is intransitive. The topic phrase performs its
canonical function as the starting point of the sentence; it serves to introduce
the Primary Referent.
The multiple analyses above give a kind of thick description of text passages.
This demonstration sets the stage for the detailed analysis to follow in the later
parts of the book. In Chapter 10 I return to the multiple approach, this time
with summary and discussion that in effect bring together the main points of
the analyses.
1.4 Conclusion
Summarizing, I give below a brief characterization of the modes, listing the
main properties of each
The Narrative mode

Situations: primarily specific Events and States
Temporality: dynamic, located in time
Progression: advancement in narrative time
The Report mode

Situations: primarily Events, States, General Statives
Temporality: dynamic, located in time
Progression: advancement anchored to Speech Time
The Description mode
Situations: primarily Events and States, and ongoing Events
Temporality: static, located in time
Progression: spatial advancement through the scene or object
The Information mode
Situations: primarily General Statives
Temporality: atemporal
Progression: metaphorical motion through the text domain
The Argument mode
Situations: primarily Facts and Propositions, General Statives
Temporality: atemporal
Progression: metaphorical motion through the text domain
The entities differ in abstractness and temporality. The most specific are situ-
ations which are located in the world at a particular time and place. General
Statives – Generic and Generalizing sentences – are also located, but they
express a pattern of situations rather than a specific situation. Facts and Propo-
sitions, the most abstract entities, are not located in the world. Because of these
differences, information about the domains of time and space, or the absence
of it, is a revealing feature in a text.
The notion of “predominant entity” is flexible. Entities predominate when
there are relatively many of that type, or if they are highlighted in the text by
syntax and/or position in a passage.
Discourse conveys several kinds of information. Underlying a story, historical
account, or argument is information about situations and participants, time
and place, continuity, text progression of two kinds, point of view. Part of the
complexity of a text comes from its multiple linguistic cues to inter-related
meanings, expressed simultaneously. This book explores how some of these
meanings arise.
I have emphasized that temporal factors are important for the modes. The
point is supported by empirical findings which show that people notice tem-
porality in texts. Faigley & Meyer (1983) did an experimental study in which
readers classified texts. In three experiments, subjects were presented with a
varied group of texts and asked to sort them “according to type.” The subjects
were identified as high- and low-knowledge readers, graduate students and
1.4 Conclusion 21
undergraduates respectively. They did not receive special training for the ex-
perimental tasks.
When genre and subject matter were controlled, temporality was the feature
that explained the results. The temporal and aspectual information of a text
correlated with the subjects’ classifications. All subjects put texts into three
classes, identified by Faigley & Meyer as (a) narrative, (b) process-description,
(c) definition–classification. Temporality was recognized on a continuum,
Faigley & Meyer suggest. Passages with many events (narration) are at one
end and passages with many statives are at the other (description, classifica-
tion). In the middle are passages with unspecified time, often with modals such
as should, would, could, etc.
This work confirms the importance of temporality that we have arrived at on
a linguistic basis. The conclusions of Faigley & Meyer go beyond temporality
as such: they claim that there is a cognitive basis for text types if genre is
controlled. Their notion of “text types” is that of traditional rhetoric. Although
traditional text types are not defined in linguistic terms, they are remarkably
close to the Discourse Modes arrived at independently here, as I show in 2.4
below.
Time is one of the key factors that affects behavior, memory, and thinking.
We are only dimly aware of the “biological clocks” in the brain that synchronize
body functions with day and night and track the passage of time. Recent work
has led to understanding of how the body keeps time through circadian rhythms,
or “body time” (Wright 2002). However, we do not yet understand very much
about “mind time.” Mind time deals with the brain mechanisms for organizing
time, and the consciousness and perception of time (Damasio 2002). Time
is currently under study in anthropology, biology, neuroscience, philosophy,
physics. That time plays a role in so many aspects of human life may partly
explain its importance in cognition and discourse.
Example sources in this chapter (page nos. are given only for examples from books):
(1) Peter Robinson, A Necessary End, New York: Avon Books, 1989, p. 182.
(2) Eudora Welty, Delta Wedding, New York: Harcourt, Brace, 1945, p. 1.
(3) Barak fights on many fronts. New York Times, May 20, 2000.
(4) Mapping thoughts and even feelings. New York Times, May 20, 1999.
(5) Robert Mosbacher, Cheap oil’s tough bargains. New York Times, March 13, 2000.
(6–8) Oliver Sacks, The Man who Mistook His Wife for a Hat, New York: Harper &
Row, 1970, p. 11.
(9–10) After a victory, Ethiopia looks toward other fronts. New York Times, May 20,
2000.
(11–13) John Scott & Tony Pawson, Cell communication. Scientific American, June
2000.
2 Introduction to the Discourse
Modes
People intuitively recognize passages of the Discourse Modes, although they

are probably unaware of the linguistic basis for the differences between them.
Each mode – Narrative, Description, Report, Information, Argument – intro-
duces certain entities into the universe of discourse, with a related principle
of discourse progression. The features have linguistic correlates of a temporal
nature. In fact temporality in the larger sense is the key to the discourse modes.
Temporal factors are woven into the fabric of a language and are part of our
tacit knowledge of language structure.
I use the term “passage” for text segments that realize a discourse mode.
Passages must be long enough to establish the linguistic features that determine
a mode. Two sentences suffice to do this. Intuitions are particularly strong
when there is a shift of mode. As an example, consider (1), the beginning of an
article from the National Geographic. The discourse mode shifts twice: from
Information to Narrative and back to Information. The title and paragraphing
follow the original.
(1) Listening to Humpbacks

1 When a big whale dives, currents set in motion by the passage of so many
tons of flesh come eddying back up in a column that smooths the restless
surface of the sea. 2 Naturalists call this lingering spool of glassy water the
whale’s footprint. 3 Out between the Hawaiian islands of Maui and Lanai,
Jim Darling nosed his small boat into a fresh swirl. 4 The whale that had left
it was visible 40 feet below, suspended head down in pure blueness with its
15-foot-long arms, or flippers, flared out to either side like wings. 5 “That’s
the posture humpbacks most often assume when they sing,” Darling said. 6 A
hydrophone dangling under the boat picked up the animal’s voice and fed it
into a tape recorder . . .
7 With the notes building into phrases and the phrases into repeated themes,
the song may be the longest – up to 30 minutes – and the most complex in the
animal kingdom. 8 All the humpbacks in a given region sing the same song,
which is constantly evolving.
22
2.1 The entities introduced in texts 23
The first two sentences express generalizations about humpback whales. At

Sentence 3 there is a shift to Narrative: the sentences express specific Events,
temporally ordered one after the other. The text returns to the Information mode
at sentence 7.
In this chapter I give a general account of the Discourse Modes and the
linguistic features that distinguish them. I then use these notions to discuss
background and foreground in text passages. Finally, I put the notion of Dis-
course Modes in context, discussing related ideas in the field of rhetoric and in
linguistic studies of discourse.
Section 2.1 discusses the types of entities introduced into the universe of
discourse; 2.2 considers text progression in the temporal modes; 2.3 and 2.4
discuss the temporal and atemporal Discourse Modes; 2.5 considers relevant
rhetorical and linguistic work.
2.1 The entities introduced in texts

One way of getting at what a text is about is to ask what sorts of things it
brings into the universe of discourse. NounPhrases (NP) introduce individuals:
people, places, objects, ideas, etc.; tenses and time adverbs introduce times;
clauses introduce situations, e.g. events, and states. This discussion focuses on
situations. The types of situations introduced in a text are the main determining
factor of the Discourse Modes.
2.1.1 Classes of situation entities

People classify situations in the world into categories such as Event and State
through their perceptual and cognitive abilities. Language classifies along the
same lines. In every clause an event, state, or other situation is introduced by
the main verb and its arguments, the “verb constellation.”1 Verb constellations
indirectly classify situations into idealized “situation types,” a group of semantic
concepts. The situations are classified by clusters of their internal temporal
properties.
This sense of the temporal involves internal features of a situation that are
related to the passage of time. The features involve beginnings and endings,
intervals, dynamic stages; they determine the way a situation unfolds in time.
The features have linguistic correlates. Thus the situation types are covert lin-
guistic categories, with distinct distributional and interpretive properties. The
1. The interpretation of a verb constellation may be affected by an adverbial, as in “John read a
book for an hour.” The durative adverbial changes the interpretation of “John read a book” from
telic (conveying that he finished the book) to atelic: “he did some book reading.”
24 Introduction to the Discourse Modes
situation types are aspectual in nature; they are discusssed in detail in Chapter 4.
I recognize three classes of situation entity: situations, General Statives, and
Abstract Entities. Situations consist of Events and States. Events differ from
States in the feature of dynamism. Events are dynamic, with successive stages
that take time, while States are static. Example (2) gives simple sentences that
express situations:
(2) Sentences expressing situations

a We walked to school. Event
b Bill won the race. Event
c Lee is sick. State
d They believe in miracles. State
e We have already eaten dinner. State
Example (2e) is a construction known as the “perfect,” which expresses the

State that results from the occurrence of an Event or the existence of a State.
Perfects are a type of State. Situations of all kinds take place or hold in the world,
and have temporal and spatial coordinates; they are also known as eventualities,
following Bach (1981). The term “situation” is used for both the general class
and the sub-types of Event and State.
General Statives do not express particular events or states, so that they are
relatively abstract and non-dynamic in nature. Generic sentences say some-
thing about a kind, or an abstract individual. Generalizing sentences express
a pattern rather than a particular event or state. Examples of each are given
in (3):
(3) Sentences expressing General Statives

a Dinosaurs are now extinct. Generic
b The lion has a bushy tail. Generic
c Mary speaks French. Generalizing
d John always fed the cats last year. Generalizing
Generalizing sentences do not express specific episodes or isolated facts, but

instead a pattern or regularity (Krifka et al. 1995). They concern objects and
individuals that are located in the world; note the explicit references to time in
(3a) and (3d).
Even more abstract are Facts and Propositions, the class of abstract entities.
Facts are the objects of knowledge, while Propositions are the objects of belief.
The complement clauses of certain predicates refer to Facts and Propositions
and introduce those entities into the universe of discourse. The examples of (4)
illustrate:
(4) Sentences with complements referring to abstract entities

a I know that Mary refused the offer. Fact
b Mary’s refusal of the offer was significant. Fact
c I believe that Mary refused the offer. Proposition
d Mary’s refusing the offer was unlikely. Proposition
The subject or object complement clauses refer to abstract entities. They are
not spatially or temporally located in the world, although the situations they
refer to may be located.
Complement clauses refer to Facts and Propositions. There are, of course,
many sentences that express Facts and Propositions directly. They cannot be
distinguished linguistically, and are therefore beyond the scope of this discus-
sion. This sense of “proposition” should not be confused with the general use
of the term, in which a Proposition is the sense of a sentence, the content that
it expresses.
The class of non-dynamic entities is varied: it includes States, General Sta-
tives, and abstract entities. The Discourse Mode passages in which these entities
predominate are not organized temporally. The feature of dynamism is essential
to the temporal advancement of narrative passages, as I show below.
2.2 Text progression in the temporal modes

As the reader progresses through a text, one part paves the way for the next. On
reaching the conclusion or dénouement one has traversed the full text and is in
a position to structure and understand it fully. This kind of understanding holds
beyond the first reading of a text. Literally, too, the reader progresses through
a text. Since language is sequential, one processes a word, phrase, or sentence
and then another that follows it. Shifts in focus of attention are discussed in
psycholinguistics and artificial intelligence. I will be concerned here with text
progression in the first sense.
The Discourse Modes have principles of progression that differ according
to type of entity and text organization. Texts progress according to changes
in location, temporal or spatial; or to changes in metaphorical location. The
dynamic temporal modes, Report and Narrative, progress as time advances.
Time is static in the temporal mode of Description. Receivers progress by
traversing the space described, mimicking the experience of the narrator or
reporter.
To understand the atemporal modes we need something other than spatial
or temporal progression. I propose that we treat the semantic domain of an
atemporal discourse as terrain to be traversed: a metaphorical space. With the
spatial metaphor we have the possibility of metaphorical motion by a Primary

Referent from one part of the domain to another. We identify a Primary Referent
in each clause. We then look for the metaphorical location of that referent in
the domain of the text, proceeding intuitively across domains. Space is not
unidimensional, like time: it allows travel in many directions. The complexity
of space is needed for metaphorical motion: motion in a text domain can be
hierarchically up or down, lateral, diagonal, etc. The atemporal modes progress
by metaphorical changes of location through the information space of the text.
In the following sections I discuss progression in each of the discourse modes.
I use the term “advancement” for progression that depends on temporal or spatial
location. The supporting linguistic analysis is presented in Part II of this book.
2.2.1 Narrative progression

Narrative presents a sequence of events and states that have the same partici-
pants and/or a causal or other consequential relation (Labov & Waletzky 1966,
Moens 1987). They occur in a certain order, which is crucial for understanding.
Narratives appear in many genres. They predominate in novels, short stories,
histories; and often play a supporting role in essays, newspaper articles, and
other discursive genres. Fictional and non-fictional narratives follow the same
temporal principles.
The key to narrative advancement is the dynamism of events. Recall that
dynamism involves successive stages in time. Events such as winning a race
take only a moment, while durative events have a run-time that varies with
the event. Closing the door has a short run-time; walking to work is longer;
building a house has a long run-time. When an event occurs in a narrative
context, we assume as a default that it takes the standard run-time, unless there
is information to the contrary. Events are bounded when they have reached the
end of their run-time or when an explicit stopping time is given.2
Narrative time advances with bounded events, and explicit temporal adver-
bials; it fails to advance otherwise. This is the basic finding of narrative dis-
course dynamics (Hinrichs 1986, Kamp & Rohrer 1983, Partee 1984). The
generalization ignores other interpretations: simultaneous events, flashbacks,
changes of scale, and changes of order.3 Sentences express bounded events when
2. This oversimplifies somewhat: adverbials can give an explicit bound that is less than the full
run-time of an event.
3. Narrative sequence is the default, in this view. Departures from the default are usually signaled
by adverbials or other cues. Another approach takes sequence as one type of “discourse rela-
tion” and change of scale, causation, etc., as different discourse relations; see Chapter 11 for
discussion.
they have an Event verb constellation and the perfective aspectual viewpoint,
conveyed by the simple verb form, e.g. “Mary walked.”4 Event sentences with
the progressive be+ing, a type of imperfective viewpoint, convey ongoing
Events, e.g. “Mary was walking.” Ongoing events and states are unbounded.
These points are discussed in Chapter 4.
The examples below illustrate passages of the Narrative Discourse Mode;
each has sentences of bounded and unbounded Events. The bounded Events
advance narrative time, marked with arrows →:
(5) Narrative passages
a. 1 → I slipped outside into a shock of cool air and → ran down the pier.
2 Several small boats were rocking lazily to and fro in the water. 3 → I
unfastened the rope to one, → paddled out toward the “Republic,” → then
hauled myself hand over hand up a rope ladder to the topgallant bulwark,
over onto a broad empty deck.
b. 1 → When I got to Harry’s Waldorf Towers apartment, they were winding
up the meeting downstairs. 2 → Harry appeared about a half-hour later, →
greeted me warmly, → went immediately to the telephone.
In (5a) the bounded events of S1 and S3 advance narrative time; in (5b) the
when-clause of S1, and the events of S2, advance the narrative.
Unbounded situations – ongoing events and states, including perfects – do
not move time. In the fragments above, ongoing events are expressed in S2 of
(5a) and the main clause of S1 in (5b). We interpret the unbounded situations
as simultaneous with a time previously established in the text.
The preceding passages are from novels; (6) offers an example of a completely
different type of narrative. The passage is from a book of detailed history and
analysis concerning events in the Middle East during and after the First World
War.
(6) → Townshend became aware in the autumn of 1918 of rising peace sentiment
and, like Newcombe, → he decided to give Events a push.
When Townshend learned that the Talaat ministry had fallen, → he arranged
an interview with the new Grand Vizier, and on → 17 October went to the
Sublime Porte carrying some notes that he had sketched out to indicate the
sort of peace terms that might be asked by Britain. His notes suggested that
Britain would be willing to leave the Ottoman Empire in possession of Syria,
Mesopotamia, and perhaps even the Caucasus, so long as these regions were
allowed local autonomy within a restructured empire that would resemble a
confederation of States.
4. The interpretation of perfective sentences is slightly different for the types of events known as
Activities: they are focused as a bounded unit. The unit need not coincide with the endpoints of
the Event; see Chapter 5.
→ Townshend offered to help Turkey obtain generous terms along these

lines → and offered to make immediate contact with the British authorities.
→ The Grand Vizier told him that it was a crime for the Ottoman Empire
to have made war on Britain, and that it was Enver’s fault. → He accepted
Townshend’s offer of help in securing honorable peace terms without letting
Townshend suspect that he would accept whatever terms he could get.
→ That evening Townshend met with the Minister of the Marine.
The perfective event clauses and time adverbials advance the narrative.
Inference can also advance narrative time. When a narrative changes place,
the reader often infers that there has been a change in time. The fragment in
(7) illustrates, from a novel. Banks’ activities and perceptions are recounted in
the first paragraph. In the second paragraph another person appears in another
place, there. We infer that time and place have moved, and that Banks has
arrived at his destination.
(7) 1 Banks dropped Richmond off at a pub near the West Pier and carried on
along Marine Drive, parking just beyond the closed fun-fair. 2 He buttoned up
his raincoat tight and walked along the road that curved around the headland
between the high cliff and the sea. 3 Signs on the hillside warned of falling
rocks. 4 Waves hit the sea-wall and threw up spray onto the road.
5 Tony Grant was already there, leaning on the railing and staring out to the
point where sea and sky merged in a uniform grey. 6 He wore a navy duffle
coat with the hood down.
The paragraph break at S5 suggests a change of direction, and the adverb already
conveys Banks’ expectation that he was to meet Tony Grant. Readers of this
passage in the novel know that Banks had a destination, which supports the
inference of advancement.
Tense is usually unchanging throughout a narrative; conventionally, past is
the narrative tense. Tense in this mode actually conveys continuity rather than
temporal location. The inferences of temporal advancement are based on as-
pectual information – the dynamism of bounded Events in sequence – and to
explicit time adverbials. These points will be explicated in Chapter 5.
2.2.2 Description: static, with spatial advancement

Descriptive passages tend to focus on specifics: particular objects, people, men-
tal States, as the rhetorician Cairns noted in 1902. Time is static or suspended.
There are no significant changes or advancements. The entities introduced in
descriptions are usually states, ongoing events, atelic events. Description is
predominant in travel writing; it appears in fiction, and most other genres.
Descriptive passages progress spatially through a scene.
(8) Description
a. We were in an impressive and beautiful situation on a rocky plateau. It
was too high for grass, there was very little earth and the place was littered
with boulders, but the whole plateau was covered with a thick carpet of
mauve primulas. There were countless thousands of them, delicate flowers
on thick green stems. Before us was the brilliant green lake, a quarter of a
mile long, and in the shallows and in the streams that spilled over from it
the primulas grew in clumps and perfect circles.
b. An example of this kind of diversity is Eleventh Street between Fifth and
Sixth Avenues in New York, a street admired as both dignified and interest-
ing to walk on. Along its south side it contains, going west, a fourteen-story
apartment house, a church, seven three-story houses . . . a three-story apart-
ment house with a candy and newspaper store at street level. While these
are nearly all residential buildings, they are broken into by instances of ten
other uses. Even the purely residential buildings themselves embrace many
different periods of technology and taste, many different modes and costs
of living. They have an almost fantastic array of matter-of-fact, modestly
stated differences: different heights at first-floor levels, differing arrange-
ments for entrances and sidewalk access.5
These descriptive passages introduces states and events of different kinds. Time
doesn’t advance, at least not significantly.
In passages of the Description mode the reader progresses spatially from
one part of a scene to another. For instance, in the first sentences of (8a) we
tour the rocky plateau from the boulders to the flowers; then to the lake and
its subsidiary streams, following the gaze of the describer. In (8b) the passage
progresses along the buildings of Eleventh Street.
Typically in description there is a locative phrase with scope over the material
that follows, as in these examples. I also assume a tacit durative time adverbial
for descriptive passages. When there is a change of time in description, there is
also a change of scene. The information that tense conveys is anaphoric: the time
of description is that of a time established earlier in the text. The Description
mode is discussed further in Chapter 5.
2.2.3 Report: deictic advancement

Reports give an account of situations from the temporal standpoint of the re-
porter. They are, like narrative, mainly concerned with events and states. The
5. The sentence listing the buildings on Eleventh street is too long to repeat here: it continues for
most of a page along these lines: “a five-story house, thirteen four-story houses, a nine-story
apartment house, five four-story houses with a restaurant and bar at the street level, a five story
apartment, a little graveyard, and a six-story apartment house with a restaurant at street level;
on the north side, again going west . . .”
significant difference between these modes is that, in Reports, the relation

to Speech Time determines temporal advancement. Situations are related to
Speech Time, rather than to each other. This is the deictic pattern of temporal
advancement.
Reports conform to the basic speech situation, in which the speaker is central.
This centrality is signaled by adverbials such as here, now, last week, which
take Speech Time as their orientation, or anchor. Such forms are known as
“deictic.” “Deixis” is the term for linguistic forms that are anchored to the time
of speech. The present tense conveys that a situation holds now, the past tense
conveys that a situation precedes now. In (9a) the adverb here in S1 reinforces
the sense of the deictic center.
(9) Report
a. 1 A week that began in violence ended violently here, with bloody clashes
in the West Bank and Gaza and intensified fighting in Southern Lebanon.
2 Despite the violence, back-channel talks continued in Sweden. 3 Israeli,
Palestinian and American officials have characterized them as a serious and
constructive dialogue on the process itself and on the final status issues.
4 News accounts here say that Israel is offering as much as 90 percent of
the West Bank to the Palestinians, although it is difficult to assess what is
really happening by the bargaining moves that are leaked.
b. 1 At his news conference here, even before he took questions, Schroeder
implicitly challenged the official US explanation for the bombing of the
Chinese embassy in Belgrade – that target analysts relied on a faulty street
map – by renewing his demand for a formal NATO inquiry into the bombing.
2 Diplomats say that Schroeder, who just returned from China, was angry
that a trip he had long planned to herald his chairmanship of the European
Union was transformed into an official apology for the embassy bombing.
Reports progress with changes in temporal and spatial location. Frequent

changes of tense are common. For instance, the situations S1–2 of (9a) are
located in the past, at different spatial locations (“here,” “Sweden”). In S3–4
time advances to the present. The fragment of (9b) has a complex temporal
analysis, beginning in the past, moving to the present (in the second clause of
S1) and then back to the past. The situations are not related to each other.
In passages of the Report mode, the order of events does not determine the
interpretation. Rather, the standpoint of the reporter is the organizing factor, as
Caenepeel observes (1995:231).
Summarizing, the temporal discourse modes are Narrative, Description, and
Report. Passages in these modes progress as temporal or spatial location ad-
vances. Advancement is conveyed by aspectual viewpoint and situation type,
tense and time adverbs, and by spatial information.
2.3 Text progression in the atemporal modes

2.3.1 Progression as metaphorical motion
In passages of the Argument and Information modes, the entities are mainly
General Statives and Abstract Entities. These entities are not dynamic and
progression in such texts is neither temporal nor spatial. There is a connection
with space, however: generalizing from the spatial domain, I shall say that
passages of atemporal modes progress by metaphorical motion.
The atemporal text offers a metaphorical space which readers traverse as they
go through the text. The idea can be implemented with the notion of a Primary
Referent that is semantically central to a situation. Each clause in a passage has
a Primary Referent. When the Primary Referent moves metaphorically from
one part of the text domain to another, one has the intuition of metaphorical
progression.
The Primary Referent of a clause has a central role in the situation expressed
by the clause. In an Event, the Primary Referent is what moves or changes; in a
State, a property is ascribed to the Primary Referent. Example (10) gives simple
clauses with the Primary Referent phrases underlined:
(10) a. Mark opened the door.

b. The door was opened by Mark.
c. Sarah gave the balloon to her sister.
d. Louise lives in Peoria.
The Primary Referent is usually the grammatical object in sentences with tran-
sitive verbs, and subject in sentences with intransitive verbs. Primacy is deter-
mined by event structure, and does not change with shifts of word order and
syntax. So, for instance, the Primary Referent is object in the active sentence
(10a) and subject in the corresponding passive (10b). The Primary Referent
corresponds to the thematic argument role of Theme or Patient in most cases;
see Chapter 4 for discussion.
How a domain is organized in the text can affect progression. There are several
conventional principles for domains: for instance, they may be organized by
hierarchy as in plant and animal taxonomies. Conventional organizing principles
include causal relations, chronology, and geography. Example (11), for instance,
has a geographical organization (cited in Olman 1998):
(11) 1 Cats also spin a story about how creatures adapt to the world they live in.
2 Most landscapes hold wild cats indigenous to every continent except Aus-
tralia and Antarctica. 3 They live in forests, plains, mountains, deserts, snowy
steppes. 4 The margay, a small spotted cat of Central and South America, has
gymnast-like limbs suited to a life of tree climbing in the rain forest. 5 The
gray-white coat of the snow leopard melts perfectly into the rocky highlands
from Siberia through the Himalayas. 6 The slightly webbed front toes of the
fishing cat help it dive for prey in the rivers of tropical Asia. 7 The clouded
leopard, once found only in Asia’s pristine forests, has turned up in habitats
damaged by overgrazing and logging. 8 It has adapted to the careless hand of
man.
Olman points out that the examples distribute themselves across continents and
landscapes, following the rhetorical principle of Comprehensiveness. Under-
standing the way domains are organized is an interesting and difficult problem.6
Linguistic expressions for space and Primary Referents are discussed further
in Chapters 5 and 6.
2.3.2 The Information mode

The Information mode gives information, presenting it as uncontroversial. Infor-
mative passages introduce mainly General Statives – generics and generalizing
sentences – into the universe of discourse. This is the main difference between
the Information and Description modes; the latter focuses on specifics, particu-
lars of a single state of affairs. Informative passages predominate in textbooks,
journals of information, in-depth studies of particular topics – the examples in
this book include discussions of whales, Chinese pottery, and tunnels. Passages
of the Information mode often have a supporting role in manuals, newspaper
and magazine articles, arguments.
The mode is timeless and progression is metaphorical in the text domain.
The text progresses when the metaphorical location of referents changes. The
fragment in (12) illustrates; it is from the same article as (1) above. The sentences
are all stative. The Primary Referent is underlined in each clause; subscripts
indicate that phrases have the same referent:
(12) 1 Humpbacksi are found in every ocean. 2 Together with blue, fin, sei, Bryde’s,
and mink whales, theyi belong to the rorqual family of baleen whales. 3
Fully grown femalesj , whichj are bulkier than the males, can weigh 40 tons
and reach lengths of 50 feet. 4 Humpbacksi tend to favor shallow areas, often
quite close to shore, and theyi are among the most sociable of the great whales
and the most active at the surface, all of which makes themi among the easiest
to observe.
The Primary Referent is the subject NP in all but one clause in the passage:
“humpbacks,” “they,” “fully grown females,” “which,” “humpbacks,” “they.”
The last clause has a transitive verb and its object, the pronoun them, is the
6. This problem is well known in Artificial Intelligence, where knowledge bases must be organized
to allow searches from different points of view. See Porter et al. (1988), Acker & Porter (1994).
Primary Referent. The limited metaphorical motion is reflected in the interpre-

tation of the passage: it presents a collection of facts about humpback whales.
2.3.3 Argument
An argument passage brings something to the attention of the reader, makes
a claim, comment, or argument and supports it in some way. Claims do not
have a particular linguistic form: they appear in all sorts of linguistic structures.
The assertion of something new, surprising, or tendentious may function as a
claim. For instance, one text I examined begins “The American high school is
obsolete and should be abolished.” In the context of current American life, this
is a strong claim. In linguistic terms the sentence is generic, referring to the
class of high schools. Determining when a sentence makes a claim is beyond
the scope of this discussion.
Passages in the Argument mode are concerned with states of affairs, Facts,
and Propositions. They differ in whether the author is strongly present or not:
we can think of the extremes as the argument and commentary poles. Texts near
the argument pole may not directly involve the author or audience; as in (13).
Sentences 1–2 have complement clauses referring to Facts, sentences 3–4 are
Generalizing Statives. The Primary Referents are underlined.
(13) 1 The routine transfer of power may not be the most dramatic feature of Amer-
ican democracy, but it is the most important. 2 It separates us [the United States]
from the majority of countries in the world, which have still not achieved it. 3
Conceding defeat and going home, or staying on in the minority and allowing
the winner to govern – these are not just the elements of good manners and
sportsmanship. 4 They are the core of patriotism.
The metaphorical location changes somewhat in the course of this paragraph.

In S1 it is “the routine transfer of power” in the United States. The first clause
of S2 has the object NP us as primary, suggesting the possibility of a shift. The
next clause returns to the first location, however, with the Primary Referent the
object NP it (‘the routine transfer of power’).
The next example is closer to the commentary pole, partly because of the
implicit subjectivity of the predicates “likely,” “impossible,” etc.; subjective
forms are discussed in Chapter 7. Sentences 1–4 have complements referring to
Propositions, while S5 is an explicit prediction. As before, Primary Referents
are underlined.
(14) 1 It is likely that other new technologies will appear suddenly, leading to
major new industries. 2 What they may be is impossible even to guess at. 3
But it is highly probable – indeed, nearly certain – that they will emerge, and
fairly soon. 4 And it is nearly certain that few of them – and few industries
based on them – will come out of computer and information technology.

5 Like biotechnology and fish farming, each [of them] will emerge from its own
unique and unexpected technology. 6 Of course, these are only predictions.
The first sentences have the same Primary Referent, so there is no progression.
At S6 the Primary Referent changes, with metaphorical motion from “other
new technologies” to the more abstract predictions the author is making (the
referent of “these”). When the primary is a quantified NP, as in S4 (“few of
them”), the relevant class is taken as primary.
This chapter began with an example of a passage that shifted from the Nar-
rative to the Information mode. Another example of a shift in mode, this time
from Argument to Narrative, is given in (15). The fragment is from the same
text as (13) above; the shift occurs at the second paragraph:
(15) I feel reasonably certain of the final verdict on the current impeachment affair
because I think history will see it as the climax of a six-year period marred by
a troubling and deepening failure of the Republican party to play within the
established constitutional rules.
It was on Election Night 1992, not very far into the evening, that the Senate
minority leader, Bob Dole, hinted at the way his party planned to conduct
itself in the months ahead: it would filibuster any significant legislation the
new Democratic President proposed, forcing him to obtain 60 votes for Senate
passage.
In the first paragraph the clauses refer to Facts and Propositions proposed by the
first-person speaker. There is a shift to narration in the second paragraph. The
clauses introduce Events, and the it-cleft construction signals that a narrative
is about to begin. It is a typical way of setting the scene in traditional tales (“It
was on a dark and stormy night that . . .”).
2.4 Foreground and background in text passages

There is a traditional, intuitive distinction between the foreground and the back-
ground of a text: foregrounded information is the most important and back-
ground material is supportive. The distinction has been discussed mainly for
narrative texts. While the idea of narrative foreground is quite well under-
stood, the concept of background is vague. “Background” simply lumps together
everything that is not foreground. The study of Discourse Modes clarifies by
distinguishing different aspects of a text. I use the factors of situation entity
and progression to give an articulated notion of background. Information is
backgrounded if it fails to contribute to progression; or if it involves entities
that are not characteristic of the current Discourse Mode.
I use this strategy to discuss foreground and background in narrative, and

then generalize to the other Discourse Modes.
The foreground of a narrative consists of situations that advance the narra-
tive. The foreground presents the main, sequential events of a narrative, while
the background gives supporting and descriptive information. Foreground sit-
uations are usually events in sequence presented with the perfective viewpoint,
while imperfectives and states are usually background. Hopper (1979) was
one of the first to propose this aspectual characterization of foreground and
background, which has been widely accepted.7 The characterization of back-
ground can be improved, with some of the insights from the study of Discourse
Modes.
I suggest that two kinds of background information can be distinguished:
one involves narrative time and the other, types of situation. Typically states
and ongoing events are backgrounded in narrative. Both types of situation are
related to the time line of the narrative: they are either simultaneous with a
bounded Event, or before or after it. In contrast, General Statives or Abstract
Entities are off the timeline of a narrative. They are part of the background in
another sense.
The generalizing statives and generics of Information discourse, the abstract
entities of Argument, give information of a non-narrative kind. For instance,
Propositions may provide commentary that supports or subverts a narrative.
Shifts in type of entity are also shifts from foreground to background.
The intuition of background and foreground extends to the other modes.
One recognizes important material as foreground, and subsidiary, supportive
material as background. Foregrounding affects progression. Backgrounding
occurs when information does not affect progression, and when entities are not
those of the current Discourse Mode.
To illustrate, I give examples of passages of each mode. The passages have
foregrounded and backgrounded clauses.
Narrative has a clear line of semantic progression, and shifts from this line
are relatively salient. The fragment in (16), for instance, has four background
clauses. Three of them (a, b, d) are on the timeline, though they do not advance
narrative time; the fourth (c) is off the timeline altogether. The relevant clauses
are italicized and lettered.
7. Hopper claims that foreground and background also differ in the familiarity status of subjects,
type of situation, and function. According to Hopper, there is a strong statistical tendency for
foregrounded Events to have familiar, animate subjects, to be punctual rather than durative, to
occur in sequence. In contrast, backgrounded situations tend to have less familiar subjects, to
be durative or stative, not to be sequential (1979:223).
(16) Narrative
a Harry did not want to run away. . . he kept snivelling and burst out, “But I
don’t want to!” “Yes, you do” I informed him . . . I kept my little brother awake
by telling him stories and then, already chilly in the night air, we sneaked from
the house, crept past the lit windows where our parents sat reading and ran off
down the road in the dark, the pillowcase with its loose tins banging against
my legs. b Harry was crying loudly now. c The bush was not then the domestic
bush it has become . . . And then something happened, the two dogs arrived,
to lick our hands and whine and jump around us. d We had not remembered
the dogs . . . we fled into our bedroom and into bed. We giggled and laughed
and shrieked with relief, and the dogs went quietly back to lie in their places
in the lamplight.
Clauses a and b express a State and an ongoing Event respectively, and d is a

past perfect. They are clearly background, since they don’t advance narrative
time. (They are important to the story line, however.) Clause c is a Generalizing
Stative. It makes a generalization about the bush – the story is set in Africa –
which has nothing to do with the timeline or the story.
Passages of the Description mode are static and within the scope of a tacit time
adverbial. Entities that are dynamic, or located at another time, are part of the
background. Entities not typical of this mode are also part of the background.
Example (17) is a descriptive passage; there is one backgrounded clause, in
italics:
(17) Description
The valley rose steeply, flanked by gigantic hills, with little terraced fields of
barley, blue vetch and clover half-hidden among the rocks. Here the irrigation
ditches are of a beautiful complexity and I thought how my children would have
liked them; the water running swift and silent until it reached a place where
the dyke had been deliberately broken by the “Lord of the Waters,” allowing it
to gurgle through into some small property and continue its journey downhill
on a lower level as a subsidiary of the main stream.
The italicized clause is off the static temporal line and involves another type of
entity. The main clause expresses an Event – the author had a particular thought –
and its complement is propositional.
The Report mode relates situations to Speech Time. One would expect that
narrative would be backgrounded in this mode, since it follows a different
pattern of tense interpretation, that of continuity. This expectation is borne out.
The example illustrates: there is a shift to narrative at S4.
(18) Report
1 The paradox is that the leaders of Ethiopia fought side by side with the
Eritreans to oust the military government of Mengistu Haile Mariam in 1991.
2 For the first two years, the two governments were close friends, so close that
they never demarcated the border when Eritrea became independent in 1993. 3
a But tensions grew over personality clashes and economic rivalry, exploding
in May 1998 b when Eritrea claimed the Badme border region based on old
colonial maps. 4 Eritrea moved troops into the area. 5 a Ethiopia said it was
invaded, b and a war ignited that c has defied long peace talks and claimed
tens of thousands of lives on both sides.
The narrative fragment consists of the events of S4 and S5a–b; they are related
to the Event of S3a (“tensions grew”). The Report mode resumes at S5c, with
the present perfect. In this passage the narrative is subsidiary, backgrounded.
The atemporal modes progress metaphorically, and have the deictic pattern
of tense interpretation. Specific situations may be backgrounded. For instance,
the following Information passage consists of generalizing statives, except for
one clause, which is italicized:
(19) Information
The motifs [of the moulded pottery] include a large number of different flowers
and fruits including the chrysanthemum, which is not found in the carved
ware, both the flower and fruit of the pomegranate and melon gourds. To
these are added somewhat abstract foliage scrolls and a debased form of
key-fret, the only geometric motif in Ting decoration. There are proper fish
among aquatic plants and in lotus pools. Finally there are chubby infants,
wholly or partially naked. It seems unlikely that they were incorporated into the
decorative vocabulary much before the late twelfth century. The organization
of the surface [of the pottery] is similar to that of the carved ware.
The italicized sentence has a propositional complement, unlike the Generalizing

Statives of the other sentences. It presents supporting, backgrounded material.
In Argument, specific Events and States may also be backgrounded. In the
passage below, for instance, Sentences 5–6 express a question-and-answer
sequence. The sequence departs from the prevailing Argument mode of the
passage in the two ways identified above:
(20) Argument
1 We in academia must figure out what is really critical to us and what we are
willing to give up.
2 Not all of these choices will be ours alone. 3 Our students, as well as our
governments, have changing expectations. 4 Information economies require
higher levels of education and more frequent education. 5 More of the new
student body may be part time, working and older. 6 I asked some students in
this new breed what relationship they wanted with their colleges. 7 They told
me that it should be like the relationship with a utility company, supermarket or
bank – their emphasis was on convenience, service, quality and affordability.
8 This group is going to gravitate toward online instruction, with education at

home or in the workplace.
The question and answer of S6 and S7 are Events, one following the other. They
are unlike the entities expressed in preceding clauses, which are more abstract
and deictically organized. At S8 the Argument mode resumes with a general
prediction.
The examples show that with changes of entity there is a change from fore-
ground to background in passages of a given mode. The change is, in effect, to
a different mode. However, we tend to take one or two clauses merely as a shift
from foreground to background.
Summarizing, I have shown two ways in which material may be back-
grounded in text passages. They correspond to the distinctive characteristics
of the Discourse Modes. Situations are in the background if they do not ad-
vance the time of a passage in the temporal modes; or if they are of a different
type from those of the current Discourse Mode.
2.5 Rhetorical and linguistic background

The Discourse Modes have counterparts in the field of rhetoric, the art and
study of using language effectively and persuasively.8 Since the nineteenth
century, rhetoricians have recognized a group of “forms of discourse.” They
are similar to Discourse Modes, though without the linguistic focus. The notion
of Discourse Modes was developed independently of the rhetorical approach.
The similarities between the two are striking and, I think, validate each other.
Before discussing the rhetorical forms of discourse, I’ll sketch in the rhetorical
background.
2.5.1 Traditional rhetoric: background

Classical rhetoric is a 2,500-year-old discipline that began with the Greek
sophists and Aristotle; it dealt with persuasion in public discourse. Rhetoric
was seen as an art that could be taught, and Aristotle’s Rhetoric from the mid
fourth century BC was a basic text.
Aristotle says that “rhetoric does not belong to a single defined genus of
subject but is like dialectic . . . its function is not to persuade but to see the
available means of persuasion in each case” (translation of Kennedy 1991:35).
8. I thank Keith Walters for helpful discussion of this section. The definition in the text is from the
American Heritage Dictionary of the English Language, 3rd edition, Boston: Houghton Mifflin
Co., 1992.
Aristotle taught that there are three aspects to persuasion: the truth and log-
ical validity of what is being argued (logos); the speaker’s success in con-
veying to an audience that he or she is to be trusted (ethos); and the emo-
tions that the speaker can awaken in the audience (pathos). In the Rhetoric,
persuasive oratory is organized into categories of political, legal, and ceremo-
nial (praise and censure). Classical rhetoric was further developed by Cicero,
roughly from 103 to 46 BC, and by Quintilian in the first century AD. After the
first century it was taught off and on in Europe until the end of the eighteenth
century.
There was a strong revival of classical rhetoric in nineteenth-century England.
Richard Whately’s 1828 Elements of Rhetoric presents itself as an adaptation
of classical rhetoric. Whately discusses the invention, arrangement, and intro-
duction of propositions and arguments. He teaches students how to find suitable
arguments to prove a point, and how to arrange them skillfully. Whateley gives
a typology of arguments (syllogisms, arguments from example, from testimony,
cause to effect, analogy); the appropriate use of arguments; the best arrange-
ment of arguments. Part II treats Persuasion, “influencing the will.” Whately’s
book was reprinted in 1963 and is still cited in discussions of rhetoric.
Modern rhetoric in the United States emphasizes the social and communica-
tive aspects of persuasive discourse (Lunsford & Ede 1984). Rhetorical texts
teach students the actual processes of planning, speaking, writing, in addition
to classic argumentation. For instance, Edward P. Corbett’s Classical Rhetoric
for the Modern Student, published in 1965 and still reprinted, offers rhetoric
as a system for finding something to say, learning how to select and organize
material, and how to phrase it in the best possible way. Corbett’s main topics
are discovery of arguments, arrangement of material, and style.
The text draws on a categorization of discourse types that is close to the
Discourse Modes of this book. Corbett recognizes “four forms of discourse:
Argumentation, Exposition, Description, Narration” (1965:32). They are not
defined or documented, but assumed as part of the toolkit of the rhetorician.
For instance, Corbett says that, in expository discourse, one can sometimes
organize according to a chronological scheme. But in argumentative discourse,
the speaker has to think about things like where to use the weakest argument
(1965:314). The notion of forms of discourse has a fairly long history in England
and America.
2.5.2 The rhetorical “forms of discourse”

The forms of discourse appear in rhetorics of the late eighteenth century, when
the field of rhetoric had broadened beyond persuasive oratory. Discourse was
classified according to function and content. Scholars attribute the idea to

George Campbell; the approach was developed by Alexander Bain (1877),
John Genung (1900), and others. Bain and Genung distinguished four forms
of discourse: Argumentation, Exposition, Description, Narration. The internal
structure of each form is determined by its purpose: according to Genung, there
is a “natural movement of ideas appropriate to each kind” (1900:475). Bain
and Genung described the main characteristics of each form and tried to teach
students to understand and use them effectively. The four forms are used in
many texts, including modern texts. The term “form” was later replaced by
“mode” in some cases; I will say “rhetorical mode” to avoid confusion with the
linguistic Discourse Modes of this book.
The rhetorical modes are strikingly close to the Discourse Modes, allowing
for differences across fields. The emphasis in rhetoric is on strategy and ef-
fect rather than linguistic features but some of the main insights are the same.
Genung differentiates among the rhetorical modes according to what is talked
about. Narration recounts Events in sequence; Description portrays observed
objects; Exposition explains, classifies, makes clear ideas, terms, or proposi-
tions (the “Information” mode of this book); Argumentation seeks to prove
the truth or falsity of a proposition and deals with issues of conviction
(1900:475). The “kinds of things talked about” that Genung notes are like
the entities distinguished as covert linguistic categories. William Cairns, also in
this tradition, comments: “Description . . . deals with particular objects . . . the
element of time isn’t present . . . [In contrast] catalogues of general characteris-
tics are really Exposition because their object is to explain, define, classify, or
make clear the meaning of a general term or proposition” (1902:114).
People recognized that forms vary throughout a text: “[the modes] are com-
bined in a great many ways, one helping and reinforcing another” (Genung
1900:475). And Brooks & Warren comment that “frequently in discourse which
is primarily intended to explain something or convince us of something, we find
bits of narrative used to dramatize an attitude, to illustrate a point, to bring an
idea home to us” (1958:265).
The rhetorical modes still appear in relatively recent texts, for instance in
Modern Rhetoric by Brooks & Warren, published in 1958 and often reprinted
since. The book recognizes “the traditional four forms of discourse”: Expo-
sition, Argument, Description, Narration. Exposition explains or clarifies by
appealing to the understanding. Argument also appeals to understanding, but
with a different purpose: to convince the reader of the truth or desirability of
something. There is some confusion, mainly terminological, between genre
and rhetorical mode. Brooks & Warren distinguish two types of narrative, for
instance, according to the purpose of the text. “Expository” or “ordinary”
narration aims to give information, to explain (1958:110), while narrative
proper appeals to the imagination; there is a difference in intention, and thus in
method (1958:231). From the point of view of this book, “narrative proper” and
“expository narrative” are both of the narrative type, appearing in different
genres. Brooks & Warren’s comments can be interpreted as a recognition that
rhetorical mode differs from genre.
Many textbooks of composition used the approach of Bain and Genung in the
United States from about 1885 until the 1940s. After that expository writing was
emphasized in many rhetorical texts, to the exclusion of the other forms. There
was also a new emphasis on “process-oriented” teaching that emphasized the
writer’s intentions and methods, beginning in the 1960s (Connors 1997:251).
The rhetorical modes are found less often in texts today, mainly because they
did not help students learn to write, according to Connors; see Larson (1984)
for a different view.
The approach of rhetorical modes has been strongly criticized, especially as
a teaching technique. The pragmatic objection is that it is not effective with
students. Theoretically, people objected that it confuses purpose with mode:
“The categories [are] supposedly based on purpose or intention.Yet narrative
can scarcely be seen as an intention in the same sense as persuasion or exposition
might be. We can perceive widely varying intentions in different narratives such
as a fictional story, a factual report of Events, a scientific account of a sequence
of Events, and a narrative contained within an advertisement” (Britton et al.
975:5). There may be just such a confusion in Brooks & Warren’s treatment of
narrative.
Another point is that different rhetorical modes can be used in pursuit of a
single purpose. Faigley & Meyer comment, “an argument against slums might
well use descriptions of living conditions in slums, personal narratives of life
in slums, and evaluations of housing conditions in slums,” and go on to claim
that this criticism “refutes” the traditional modes (1983:308). But this is too
strong. The problem is rather a confusion between mode and genre. Moreover,
the experiments cited in their own work (mentioned above) give evidence that
the rhetorical modes have a cognitive basis.
There is indeed a certain confusion in discussions of the rhetorical modes.
We are told that they have a particular purpose or intention that determines
their character – which suggests that they function at the level of genre. Yet the
modes can be used to realize different purposes, in different genres. I think that
we can resolve the confusion by clarifying the distinction between genre and
mode. The rhetorical modes do not classify by purpose. Rather, they describe
texts at the more local level identified in this book as the passage.
The rhetorical Discourse Modes correspond nicely to the linguistically based
Discourse Modes of this book. One Discourse Mode, the Report, does not
appear in the traditional classification. The importance of Report as a Discourse
Mode may be due partly to mass communication, with the pervasive use of
newspaper, television, and now internet reports. Moreover, diaries and other
personal genres are taken seriously today and they often use the Report type of
organization.
Actual texts consist of more than one type or mode of discourse: this is an
uncontroversial point on which all agree.
2.5.3 Other discourse classifications

Scholars have classified text types according to purpose and other parameters,
many of them situational. James Kinneavy works with a “communication tri-
angle” that consists of the speaker, the audience, and the subject of discourse.
He posits four general “discourse types or modes” on the basis of their pur-
pose: referential, persuasive, expressive, and literary discourse. For each type,
he undertakes to describe the “distinctive nature, distinctive logic, characteristic
organizational patterns, and the stylistic features peculiar to the particular aim”
(1971:63). Kinneavy notes that actual texts are mixed: “No theory of modes of
discourse ever pretends that the modes do not overlap. In actuality it is impos-
sible to have pure narration, etc. However in a given discourse there will often
be . . . [a] ‘dominant’ mode” (1971:37).
Other parameters lead to slightly different classifications. Using two-valued
parameters of “involvement–detachment” and “integration-fragmentation,”
Chafe (1982) proposes a four-way classification of texts. Longacre (1968) takes
temporality as a primary parameter. Some discourse is primarily time-oriented,
e.g. narrative and procedural, with identification and explanations as peripheral
elements. In other types of discourse, temporal factors are subordinate. In later
work, Longacre (1983–1996) proposes three two-valued parameters – contin-
gent temporal succession, agent orientation, projection – and develops a broad
typology of monologic discourse.
2.5.4 Linguistic features of texts by genre

Perhaps the most basic classification of discourse is according to genre. Dis-
courses of the same genre have similar purpose, setting, and participants. They
have similar global structures, determined by purpose and the other factors.
It would be natural if discourses of the same genre were similar in linguistic

features. But they are not similar in the ways one would expect, as Douglas
Biber has shown.
Biber investigated the occurrence of syntactic and lexical linguistic features in
samples of twenty-three genres, using a sophisticated computational analysis of
a large corpus. He found that the features clustered together into clear categories,
or text types. Strikingly, the text types cut across genre categories rather than
corresponding to them (1989:39).
The corpus analyzed consisted of contemporary British discourse from the
Lund-Oslo-Bergen (LOB) corpus and the London-Lund corpus.9 The investi-
gators catalogued the occurrence and co-occurrence of such linguistic features
as present tense verbs, pronouns, contractions, past tense forms, perfect aspect
verbs, wh-relative clauses, time and place adverbials, infinitives, modals, nom-
inalizations. They found that the linguistic properties they studied clustered
together reliably across texts (Biber 1988, 1989).
Five dimensions of clustering were identified. Biber provides broad inter-
pretive labels for each. The dimensions are: (1) Involved vs. Informational;
(2) Narrative vs. Non-narrative; (3) Elaborated vs. Situation-dependent refer-
ence; (4) overt expression of persuasion; (5) abstract or not. In each dimension,
certain forms have high frequency and others have low frequency. On the dimen-
sion of “involved vs. informational,” for instance, the high-frequency forms are
present tense verbs, private verbs, first- and second-person pronouns, demon-
strative pronouns, contractions, that-deletion. Low-frequency forms are nouns,
prepositional phrases, long words, lexical variation, attributive adjectives.
The analysis results in a set of “text types” that have feature clusters from more
than one dimension. Biber proposes eight “prototypical” text types: intimate
personal interaction, informational interaction, scientific exposition, learned ex-
position, imaginative narrative, general narrative exposition, situated reportage,
and involved persuasion. Text types do not pattern according to genre. Instead,
“texts within particular genres can differ greatly in their linguistic characteris-
tics, e.g. newspaper articles can range from extremely narrative and colloquial in
linguistic form to extremely informational and elaborated in form. On the other
hand, texts of different genres can be quite similar linguistically: newspaper
9. Biber (1989) lists the genres studied as follows – examples of written genres from LOB corpus:
press reportage, editorials, press reviews, religion, skills and hobbies, popular lore, biographies,
official documents, academic prose, general fiction, mystery fiction, science fiction, adven-
ture fiction, romantic fiction, humor, personal letters, professional letters; examples of spoken
genres from London-Lund corpus – face-to-face conversation, telephone conversation, public
conversation, debates, interviews, broadcasts, spontaneous speeches, planned speeches.
articles and popular magazine articles can be nearly identical in form”

(1989:6).
Text types and genres are valid and different typologies, Biber suggests.
Genre distinctions are based on essentially nonlinguistic criteria, reflecting
differences in external format and situation. Other efforts to find consistent
linguistic distinctions that correlate with genres have also been unsuccessful.10
People have also looked for linguistic correlates of genre-based units. For
instance, suppose that a story consists of a Setting and an Episode; and an
Episode consists of an Event and a Reaction (there are stories with such units).
There might be linguistically identifiable stretches of text that correspond to
these sub-units. Implementing this idea, people have written “discourse gram-
mars” which generate stories and other types of discourse. The grammars are
hierarchical: they start with abstract units and in principle end with the sen-
tences of a discourse (van Dijk 1972, Rumelhart 1975, Thorndyke 1977). There
are two crucial assumptions behind these grammars. One is that an orderly
relation holds between the abstract units of the grammar and the linguistic
units – sentences, clauses, phrases, etc. – that make up the text. The other
assumption is that discourse has certain constraints, so that texts, like sen-
tences, are well formed or ill formed. Neither of these ideas holds up under
scrutiny.
It’s plausible that texts would be well and ill formed like sentences; but
in fact constructing a text that is actually ill formed (as opposed to awk-
ward or confusing) is difficult if not impossible. The reason is that almost
any sequence of sentences can be taken as well formed if the receiver supplies
enough inferences.11 Nor is there a clear, predictable correspondence between
10. For instance, Edward Smith (1985) looked for linguistic features correlating with four very
general text types identified by Longacre (1983/1996): Narrative, Procedural, Behavioral,
and Expository texts. He concluded that genre – purpose – isn’t conveyed by linguistic
features.
11. Discourse grammars use phrase structure expansion rules, producing hierarchical structures.
The grammar approach treats a discourse as one big sentence with hierarchically related parts.
Linguists and psychologists developed these grammars in attempting to model the processes of
discourse construction and discourse understanding. In principle, they believed, the units and
relations of a story grammar are psychologically real: “a higher level organization takes
place . . . strings of sentences combine into psychological wholes which account for . . . salient
facts about structure” (Rumelhart 1975:213). In the 1970s, some experimental evidence sup-
porting story grammars was found. Subjects remembered information, for instance, according
to the sub-units posited; but other, more general explanations for these results have been offered
(Garnham 1983, 1985).
According to Garnham, “The problem is to specify the way in which general semantic
definitions of syntactic categories could be used to compute the category of a proposition.
the units of a story grammar and the linguistic units in which a story is real-
ized. For instance, the Reaction part of an Episode may be given by a phrase, a
sentence, several sentences, a paragraph, etc. Recognizing such a sub-part usu-
ally requires inference. But if this is true, we cannot specify a procedure for
going from the abstract “story grammar” to the actual story: no mapping pro-
cedure exists.12
2.5.5 Other linguistic studies of texts

Linguists have looked for text units that are intermediate between sentence
and text. The paragraph is a grammatical unit in the tagmemic framework, in
which sentence and paragraph units are essentially the same. Loriot & Hollen-
bach (1970) consider paragraphs in Shipibo, a Peruvian language. They argue
that there are three sub-systems of paragraph structure in the language; Event–
reference ties and object–reference ties, and the ordering of sentences to form
paragraphs. Their ideas are similar to those of Halliday & Hasan (1976), dis-
cussed below. Paragraphs have been studied extensively in folktales (Dundes
1975). More recently, studies of American Indian and other languages have re-
lated linguistic features to properties of a discourse (Woodbury 1987, Goddard
1990, Thomason 1994).
The field of text-linguistics applies linguistic techniques to text analysis (de
Beaugrande & Dressler 1981). One approach is strongly influenced by gener-
ative semantics and logic and focuses on formal rules (Reiser 1978); another
attempts to integrate the complex components of texts into a full semiotic the-
ory (Petöfi 1978). Text coherence has received a good deal of attention in this
approach: researchers try to understand what underlies the intuition that a text is
coherent, or incoherent. Answers have focused on the notions of text coherence
and cohesion.
Coherence is the network of relations that organize the propositions and
concepts of a text. They involve world knowledge that goes beyond the text.
These relations are not necessarily made explicit in the text but are inferred
by the receiver. The study of coherence has led to procedural accounts with a
strong base in psychology (de Beaugrande & Dressler 1981:Ch. 5). Heyrich
et al. (1989) offers a group of relatively recent studies in the areas of connexity
and coherence.
Story grammarians have never noted this problem, let alone attempted to solve it” (1983:150).
Garnham suggests that psycholinguists’ findings are better explained in terms of referential
continuity and plausibility.
12. Apparent violations of a convention are interpreted with additional inferences that make them
plausible.
Syntactic correlates to discourse phenomena have been identified at the

level of clause and sentence structure. Aspectual viewpoints often convey
foregrounding and backgrounding (Hopper 1979), as noted above. There are
studies of how syntactic subordination relates to discourse (Matthiessen &
Thompson 1988); tense and narrative (Fleischman 1991); grounding and deic-
tic phenomena (Tomlin 1987, Duchan et al. 1995).
“Cohesion,” or connexity, deals with the connections made in a text at the
surface level. Cohesion relates the sentences of a text by a network of gram-
matical and lexical ties. The term is due to Halliday & Hasan (1976). Cohesion
is expressed by coreference; substitution, as when proforms like “do so” stand
in for repeated material; ellipsis; conjunction; and lexical relationships.
Lexical cohesion arises when patterns of related words appear in a text. This
is perhaps the most interesting contribution of Halliday & Hasan. Lexical cohe-
sion involves “reiteration” of a word, either actual repetition or words that are
“systematically” related such as superordinates, synonyms, or near-synonyms.
General nouns can function as cohesive agents, operating anaphorically as a
kind of synonym (1976:278):
(21) Reiteration
a. Accordingly, . . . I took leave, and turned to the ascent of the peak. The
climb was perfectly easy.
b. Henry’s bought himself a Jaguar. He practically lives in that car.
Instances of the same word may be coherent even if the two are not coreferential:
(22) a. Why does this little boy have to wriggle all the time?
b. Other boys don’t wriggle. Boys always wriggle.
c. Good boys don’t wriggle. Boys should be kept out of here.
Lexical items that appear in similar contexts have a cohesive effect. For instance,
in (23) the word “girls” is cohesive:
(23) a. Why does this little boy have to wriggle all the time?
b. Girls don’t wriggle.
“Girls” is related by complementarity to “this little boy” Halliday & Hasan note
that texts may have cohesive “chains” of related words:
candle . . . flame . . . flicker; hair . . . comb . . . curl; sky . . . sunshine . . . cloud . . .
rain.
Cohesive features combine in ways that organize a text, according to Hoey

(1991). Hoey finds that repetition and other types of lexical cohesion form
“networks” that indicate topical sub-units of a text; he focuses primarily on non-

narrative texts. Cohesion across genres is studied by Stoddard (1991). Stoddard
was interested in cohesion as part of the overall patterning of a text that leads
to the perception of “texture.” She did a computational study of thirty-five texts
of different genres, which showed that cohesive elements differed according
to genre. Stoddard interprets this result as evidence that there are definable
cohesive networks in texts.13
The notion of lexical cohesion has been taken up by computational linguists
as a method for identifying units in texts. An algorithm that finds chains of
related terms was developed by Morris & Hirst (1991); the chains are used to
structure texts according to the attentional/intentional theory of Grosz & Sidner
(1986). Another approach, due to Hearst (1997), uses lexical cohesion as the
basis for a computation that discovers “sub-topic” structure in texts by a method
called “Text Tiling.” Text Tiling finds patterns of lexical co-occurrence and
distribution, and uses them as cues to segment a text into units with different
sub-topics. The units of a text are defined as a function of the patterns of
connectivity between its terms (1997:9). Hearst claims that the lexical patterns
are helpful heuristic cues to higher-level structures; lexical cohesion alone is
not responsible for such structures.14
The notion of cohesion is “semantic” and involves the interpretation of one
element as depending on another, according to Halliday & Hasan (1976:4).15
Although intepretation is often triggered by linguistic form, the reader makes
the connection. Nevertheless the term “cohesion” is primarily used for overt
linguistic devices that relate sentences.
13. The contribution of “cohesive factors” to literary texts was studied by Gutwinski (1976). She
examined texts by Henry James and Ernest Hemingway for cohesive ties. James and Heming-
way are considered to have different styles of writing, and Gutwinski found that their choice
of cohesive ties were different. James depended primarily on grammatical cohesion, especially
anaphora, whereas Hemingway used lexical cohesion almost exclusively. Gutwinski’s study
demonstrated that cohesion could be identified and that it contributed to literary
texts.
14. Hearst suggests that one can discover a text’s structure by dividing it up into sentences and seeing
how much word overlap appears among them. The overlap forms a kind of intra-structure;
fully connected graphs might indicate dense discussion of a topic, while long spindly chains
of connectivity might indicate a sequential account.
15. Halliday & Hasan also claim that cohesion distinguishes a text from an incoherent sequence
of sentences. In other words, cohesion is what distinguishes a coherent, well-formed text.
Arguments against this strong claim are given in Brown & Yule (1983), Blakemore (1988). For
discussion of the relation between cohesion and “coherence” see Stoddard (1991:13ff.) and
Sanford & Moxey (1995).
This discussion is intended to provide a context for study of the Discourse

Modes. In the next chapter I turn to the topics of communication and infer-
ence in the understanding of text passages, and representations in Discourse
Representation Theory.
Example sources in this chapter:

(1) and (12) Douglas H. Chadwick, Listening to humpbacks. National Geographic,
July 1999.
(5a) Charles Johnson, Middle Passage, New York: Atheneum, 1990, p. 20.
(5b) Lillian Hellman, Scoundrel Time, Boston: Little, Brown, 1976, p. 55.
(6) David Fromkin A Peace to End All Peace, New York: Henry Holt, 1989, p. 369.
(7) Peter Robinson, A Necessary End, New York: Avon Books, 1989, p. 103.
(8a) and (17) Eric Newby, A Short Walk in the Hindu Kush, Victoria, Australia: Lonely
Planet Publications, 1958/1998, p. 146.
(8b) Jane Jacobs, The Death and Life of Great American Cities, New York: Random
House, 1961, pp. 227–28.
(9a) Barak fights on many fronts. New York Times, May 20, 2000.
(9b) Kosovo strategy splitting NATO. New York Times, May 20, 2000.
(11) Cats. National Geographic, June 1997.
(13) and (15) Alan Ehrenhalt, Hijacking the rulebook. New York Times, December 20,
1998.
(14) Peter Drucker, The information revolution. Atlantic Monthly, October 1999.
(16) Doris Lessing, Under My Skin, New York: Harper Perennial, 1994, p. 107.
(18) After a victory, Ethiopia looks toward other fronts. New York Times, May 20, 1999.
(19) Margaret Medley, The Chinese Potter, London: Phaidon, 1989, p. 114.
(20) Arthur Levine, The soul of a new university. New York Times, March 13, 2000.
3 Text representation and
understanding
Text understanding is a constructive process that results in a mental represen-

tation. In this it is rather like vision. Seeing a tiger, for instance, is the result
of a process of construction. The perceiver’s mind/brain converts information
from a pattern of light and dark on the retina to a representation, an interpreta-
tion of the object – an image of a tiger (Marr 1982). Inference often plays an
important role at the final stages. Similarly, to understand a sentence, one goes
from a sound wave or set of marks to a conceptual representation which brings
together information of different kinds, some of it supplied by inference and
world knowledge. The active nature of understanding informs the approach to
text structure and its representation that I take here.
Studies in cognitive science and psycholinguistics provide background in-
formation about text understanding and mental models. Mental models are
developed in effect with the structures of Discourse Representation Theory.
Section 3.1 discusses the communicative context of language; 3.2 considers
some inferences in language understanding; 3.3 relates language understanding
to the notion of mental models; 3.4 discusses the analysis of text passages in
Discourse Representation Theory, and introduces the rules and structures of the
theory.
3.1 The pragmatic background for discourse interpretation

Participants in discourse have in common several kinds of knowledge. They
may share a general background and specific knowledge about the particular
language activity of a given discourse. The participants also share the informa-
tion that develops in the discourse itself, the common ground or “context set”
(Stalnaker 1978). As a discourse progresses, each new proposition is assessed.
If it is accepted, the proposition is added to the context set. Thus each sentence
as it is processed updates the context.
The dynamic representations of Discourse Representation Theory model
the developing common ground. Other dynamic theories include Dynamic
49
50 Text representation and understanding
Predicate Logic and Situation Semantics. All of these theories have an es-
sential pragmatic thrust: they take context as crucial for interpreting discourse.
They deal with both semantic and pragmatic reasoning and information.
Pragmatics concerns language as communicative action. What a person says
usually underdetermines interpretation, yet most communication is successful.
Pragmatics tries to account for this apparent paradox. Successful communi-
cation results from a tacit negotiation between speaker or writer and receiver,
based on shared assumptions about communication. Speakers try to give enough
information so that receivers can understand their intentions. In order to do
this, receivers must often work out the semantic and pragmatic meanings of a
communication.
The now-classical approach of Grice makes a distinction between the sem-
antic and pragmatic meanings of a sentence. Semantic meaning is conveyed
by linguistic expressions, while pragmatic meaning adds what a speaker im-
plicates and/or intends when uttering a sentence in a particular context. In
principle, sentence meaning is semantic, while speaker’s meaning is prag-
matic. Grice developed a theory of how intended meaning can be derived from
what is said (1975, 1989). He stated a set of basic assumptions, the coop-
erative maxims that underlie communication. Using these maxims and what
they know about the world, people calculate by inference the interpretations
intended in a text. In this view, semantic meaning is the input to pragmatic
meaning.
Pragmatic interpretations, or implicatures, are not logically entailed by what
is said. They can be cancelled, as Grice put it, by additional information. If I
say “John and Mary have two children,” you will probably infer that two is the
total number of children that they have. This is a conversational implicature. I
might go on to say more – possibly “indeed, they have four children” and this
would cause you to cancel your earlier inference. Entailments, truth-conditions,
and the conventional meanings attached to particular words cannot be cancelled
and therefore have a different status.
The division of labor between semantics and pragmatics seemed relatively
clear at one time; it is currently the subject of active debate.1 Complete sepa-
ration of semantics and pragmatics is difficult to maintain, because pragmatic
1. Kent Bach, for instance, proposes that “semantic meaning is the conventional meaning of lin-
guistic expressions, while pragmatic meaning is context-dependent, pertaining to utterances and
facts surrounding them.” In contrast, Levinson argues that conversational implicature plays a
role in assigning truth-condition, so that semantics and pragmatics are interwoven in most cases
(1999:168).
reasoning plays a role at all stages of interpretation. For instance, pragmatic

cues resolve linguistic ambiguity, and determine the referent of a referring
expression and the time and place of an utterance – the indexical coordinates. Yet
this information is part of sentence meaning. Pragmatics intrudes into semantics,
as Levinson puts it, even at the “semantic” level of what is said (2000:164).
Nevertheless semantics and pragmatics are different. Semantic reasoning is
monotonic, pragmatic reasoning is non-monotonic.2
In this book I consider several kinds of information. Conventionally semantic
information, for instance, is conveyed by a verb and its arguments expressing
aspectual entities such as events and states. Other information arguably in-
volves both semantics and pragmatics – for instance, the principles of narrative
text progression, the identification of the Primary Referents of a clause, and the
interpretation of subjective expressions. Information based mainly on pragmatic
reasoning includes the calculation of metaphorical progression and identifica-
tion of sentence topics. Throughout the book I attempt to indicate the bases for
the interpretations that I posit, and to distinguish inferences of different kinds.
Much of the information would have to be included in any account, whether or
not it uses the distinctions I rely on here.3
3.2 Types of inference

Inference plays a role in understanding discourse at all levels. Much of the
reasoning that underlies discourse interpretation requires plausible rather than
logical reasoning. Our understanding of such inferencing patterns is still very
much a work in progress. What is needed is a theory of pragmatics which
draws on information of many kinds, among them lexical semantics, domain
and world knowledge, Gricean principles, and general principles of pragmatic
reasoning. There is no such theory at the moment.
I set out here some well-known types of inference that people use in
understanding sentences and their implicatures. People make inferences con-
stantly and automatically, in the course of linguistic communication. Inferences
2. Non-monotonic reasoning involves defaults, and is defeasible. In defeasible systems, an infer-

ence or argument may be defeated by additional premises. In contrast, deductive systems are
monotonic.
3. Some approaches do not distinguish between semantics and pragmatics, e.g. Relevance Theory,
due to Sperber & Wilson (1986) and other publications. Another approach uses a theory of
commonsense entailment or “glue logic” for discourse, known as “DICE” (Lascarides & Asher
1993, Asher & Lascarides 1998b).
fill out explicit linguistic material with information which is essential for
understanding. Often, in talking and writing, the speaker or writer leaves out
what seems obvious, assuming that the receiver will recover it by inference. In
a detailed account such things must be made explicit.
Interpretation that goes beyond linguistic forms is an everyday part of
language understanding, not an exotic phenomenon. People tend to avoid
redundancy – for instance, relying on the receiver’s ability to interpret reduced
forms. A fully explicit and redundant text is not preferred and may even be
confusing. Consider the examples in (1):
(1) a. John and Mary went to the store and John and Mary bought two large
lobsters.
b. John and Mary went to the store and they bought two large lobsters.
c. John and Mary went to the store and bought two large lobsters.
(1b, c) are more natural and perhaps easier to understand than (1a), the
most explicit. If an unreduced sentence is used it tends to carry a special
meaning.4
Working out the intended referent of a pronoun often requires inference.
There may be more than one possibility. For instance, in the sentence “John
said that he was leaving,” “he” may refer to John or to another person; in a
specific context, one interpretation is usually preferred. In the second sentence
of “John saw Thomas. He was wearing a red hat,” the first natural interpreta-
tion of the pronoun referent is “Thomas”; but other information could override
this. In other cases, pragmatic knowledge about the world makes one interpre-
tation more plausible than another. Compare “Mary lent Jane a bicycle. She
asked her to be careful with it,” and “Mary lent Jane a bicycle. She thanked
her.” In the first sequence “she” would probably refer to Mary, whereas, in
the second, “she” would probably refer to Jane. Normally people use con-
text, inference, and world knowledge to choose the most plausible pronoun
referent.
Inferences are often needed to relate things to each other; one common
relation is part–whole, illustrated in different ways by the examples of (2):
4. One of Grice’s points is that flouting the maxims conveys a special meaning. The maxim of
quantity says that one should make one’s contribution as informative as required, and do not
make it more informative than required for the purposes of the communication. Sentences (1a)
and even (1b) would flout this maxim in most situations.
The additional meanings of unreduced conjunctions are explored by Levinson as examples
of “generalized conversational implicature” (2000:148). Levinson argues that generalized im-
plicature is distinct in kind from the implicatures of a particular utterance in context.
(2) a. i. We checked the picnic supplies.

ii. The beer was warm.
b. i. John and Mary went to India last January.
ii. They flew to Bombay and then took the train to the interior.
c. i. Mary had an accident driving home from work.
ii. The steering-wheel was defective.
In (2aii), we infer that “the beer” is part of “the picnic supplies.” To understand
(2bii) we infer an event, a trip that is implicit in (2bi); the second sentence gives
details of that trip. To understand (2cii) one infers that the steering-wheel is
part of the car that Mary was driving (2ci) and one posits a car. Inferences like
those of (2) are known as “bridging” inferences. Bridging is an obligatory part
of comprehension, since the receiver must identify the intended referents for all
referring expressions. Experiments show that people make bridging inferences
like those above in the course of text understanding (Haviland & Clark 1974).
The inferences involve world knowledge, knowledge of the context, logical
entailments, and plausible reasoning.
Positing an entity that is not explicitly introduced in the discourse is a general
type of inference known as “accommodation.” The receiver infers the existence
of an entity if it is necessary to do so for coherence (Lewis 1979). In accom-
modation one accepts the information conveyed by a linguistic form whether
or not the form has an appropriate antecedent and adds an entity or presuppo-
sition to the common ground.5 For instance, the definite article the pragmati-
cally conveys that the referent of a NP is familiar, either known or identifiable
(see Chapter 6). Encountering a definite NP that is not familiar, the receiver
accommodates by taking the referent of the NP to be a known individ-
ual. Novels often exploit this convention by beginning with a definite NP
whose referent cannot be known to the reader. The experienced reader as-
sumes that the individual is to be added to the universe of discourse, and
waits for more information. The assumption that the referent of the N is
familiar is a pragmatic presupposition. In pragmatic presupposition one takes
information as uncontroversial in the context or common ground. There are
two main uses of the term “presupposition,” one “pragmatic” and the other
“semantic.”
Semantic presupposition is triggered by particular linguistic expressions: they
have the conventional meaning that information is known or presupposed. For
5. We understand that an entity is added to the common ground. Lewis discussed accommodation
in terms of presupposition: “if at time t something is said that requires presupposition p to be
acceptable, and if p is not presupposed just before t then presupposition p comes into existence,
ceteris paribus” (1979:172). See also Heim (1982).
instance, “Tom stopped running” entails “Tom ran.”6 The contexts of semantic
presupposition include the complements of “factive” verbs (know, regret); the
clefted clause of cleft sentences (“It was the butler that stole the jewels”);
temporal adverbial clauses (“John left after calling his mother”). The informa-
tion in such contexts is treated as familiar, whatever its actual status for the
receiver.
The constructive view of language understanding has experimental support
from psycholinguistics. In a classic group of experiments, Bransford et al.
(1972) studied how people understood descriptions of situations that differed
slightly, as in sentences (3a) and (3b) below. The situations have somewhat dif-
ferent consequences. From (3a) a person can infer (3c), using spatial knowledge
of the world. In the situation described, if the fish swam beneath the turtles,
it swam beneath the log. Sentence (3b) does not support such an inference,
however, so that (3d) does not follow from (3b).
(3) a. Three turtles rested on a floating log and a fish swam beneath them.
b. Three turtles rested beside a floating log and a fish swam beneath them.
c. Three turtles rested on a floating log and a fish swam beneath it.
d. Three turtles rested beside a floating log and a fish swam beneath it.
In the experiments subjects were first presented with sentences like (3a) and (3b)
and asked to remember them. Later they heard a group of sentences containing
the originals and new, slightly different sentences – some of them like (3c) and
some like (3d). The subjects’ task was to say whether they had heard a sentence
before.
Subjects tended to “recognize” sentences that were consistent with the orig-
inal descriptions, with high confidence. In other words they tended to confuse
sentences like (3a) with (3c), but not to confuse sentences like (3b) with (3d).
Bransford et al. interpret the result as showing that “recognition was primarily
a function of the complete semantic descriptions constructed rather than a func-
tion of just that information specified by the linguistic input strings” (1972:205).
Results like these strongly support the mental model view of sentence repre-
sentation.
6. There are well-known tests for semantic presupposition. The best-known is negation. Semantic
presupposition is undisturbed by negation, as in (b), the negation of (a):
a. Sam stopped running

b. Sam didn’t stop running.
Both (a) and (b) carry the presupposition that Sam ran. Other tests include embedding under
modality, the antecedent test, the test for constancy under illocutionary force.
3.3 Mental models and representations 55
3.3 Mental models and representations

The Cognitive Science approach has stressed that mental processes are active.
Many cognitive scientists hold that the mind is a symbolic system: in this
hypothesis, people think and reason by manipulating symbolic representations.7
One influential version of this view claims that people “translate” or represent
external events into internal models.
The mental models approach to texts posits a mental representation which
models the state of affairs expressed in the text. The model is constructed from
linguistic information about the world, and from inferences triggered by the text
(Johnson-Laird 1983). The fact that people make bridging inferences is a central
feature of comprehension, and a reason to believe that people construct mental
models, according to Johnson-Laird. He comments that “. . . the explicit content
of a discourse is usually only a blueprint for a state of affairs: it relies on the
reader to flesh out the missing details . . . [the experiments] . . . yield conclusions
of a sort that would be explicit only in models of situations” (Johnson-Laird
1989:471). Mental models are structurally isomorphic to situations, according
to the theory.
The model is constructed incrementally as a text advances. The developing
model functions as context for each sentence, and constrains the interpretation of
a sentence. For instance, the set of entities in the model limits the set of possible
antecedents for an anaphoric expression. Experiments directly positing mental
models are discussed in Glenberg et al. (1987), Garnham & Oakhill (1989),
Oakhill et al. (1989), Tardieu et al. (1992), and others.
Propositional information is emphasized in other approaches to representa-
tion, for instance Graesser (1981), and Kintsch and his colleagues. Van Dijk &
Kintsch (1983) define three levels: a surface form representation, a propo-
sitional text base, and a situation model. The situation model is similar to the
mental model in that it integrates information derived from the text with reader’s
knowledge. Whether some or all types of representations are needed to model
how people understand texts is still under debate. The important point is that
representations are essential.
The semantic representations of Discourse Representation Theory are close
to the mental models posited by Johnson-Laird and his colleagues. The theory
7. Not all agree with this classical view. In the approach of Connectionism, or Parallel Distri-
buted Processing, the mind is not symbolic but rather a network of simple processing units.
Connectionist cognitive modeling tries to understand the mechanisms of human cognition
through the use of simulated networks of simple, neuronlike processing units. This approach
is presented in McClelland & Rumelhart (1986), Bechtel & Abrahamsen (1991), and others.
A. Clark (1989) provides cautionary comments.
states construction rules which take the sentences of a text as input and deliver a
representation as output. It does not try to account for the actual mental process
involved in understanding a text.
3.4 The analysis of text passages in Discourse Representation Theory

The representations of Discourse Representation Theory model the developing
common ground of a discourse. The theory has two levels. At the first level, a
representation of the conceptual meaning of the text is constructed, a Discourse
Representation Structure. The Discourse Representation Structure (DRS) con-
tains discourse entities that represent individuals, situations, and times, together
with conditions that characterize them. The entities and conditions are licensed
by information conveyed in the sentence together with the DRS of the
prior discourse, up to that sentence. The conceptual DRS is evaluated by a
truth-conditional mapping at the second level of the theory.
The framework is flexible enough to allow for semantically and pragmatically
based information. Indeed, both are arguably included in the DRS representa-
tions introduced below: “DRSs in fact incorporate the results of pragmatic
resolution (most obviously, anaphoric linkages) . . . There is a common level
of propositional representation, a slate on which both semantics and prag-
matics can write . . . it is this representation that is assigned a model-theoretic
interpretation” (Levinson 2000:193).
Rules and representations in the Discourse Representation Theory (DR
Theory) framework show how the linguistic forms give rise to discourse mode
interpretation. The heart of the theory at the first level is the construction
rule. The rules interpret linguistic expressions and construct the representation
licensed by them. The predictions of the theory are embodied in the rules and the
structures that they trigger. The input to the rules is the surface structure of sen-
tences; the output is the discourse entities that are entered into the DRS, together
with conditions that characterize them. Thus DRSs embody the traditional
notion that linguistic understanding proceeds from linguistic form to meaning.
The construction rules represent conceptual meanings, which involve semantic
and pragmatic information. I shall assume that both contribute to interpretation.
The second level interprets DRSs within a formal model. This is the truth-
conditional component. It provides an embedding function from the DRS to
a model-theoretic construct. The model is an information structure, a domain
of individuals of various kinds. As expressions of a language can be evaluated
relative to a model, so DRSs are evaluated. To assert a sentence amounts to
asserting that the world, or the model, accords with the information in the DRS.
3.4 Analysis of text passages in DR Theory 57
The model corresponds to the way some state of affairs might be, a mapping
from the DRS to the model and the world. The mapping is stated with rules
of functional application: they assign to sentences a denotation or meaning in
a model. It is impossible to include everything that is in the world in framing
conditions and models. Rather, the information in a sentence focuses on certain
situations, entities, locations, and times; the rest of the world is assumed. Thus
the models specified are partial models. In this book I work with conceptual
representations; the conceptual and truth-conditional levels are discussed in
Kamp (1981), Kamp & Reyle (1993).
3.4.1 The linguistic analysis and Discourse Representation Theory

Much of the analysis of this book is stated in the Discourse Representation
Theory framework. The formal statements give content to my claims about the
linguistic basis of intuitions about discourse mode and subjectivity. Interpreta-
tions that can be stated in construction rules differ significantly from those that
cannot be stated within the theory, as we shall see.
The interpretations of the Discourse Modes depend on composing informa-
tion from several linguistic sources. In interpreting aspectual situation types
we look at information from the verb, its arguments, and adverbials – the latter
sometimes in another sentence. As a composite the semantic features associ-
ated with each form determine the situation type of a clause. The composite
approach is extended in this book to all types of situation entities, and to tempo-
ral advancement as well. Presentational text progression and point of view also
require that various factors be taken into account. The construction rules of the
theory operate on the surface structure of sentences, and license information
in the ongoing Discourse Representation Structure. The rules and structures of
Discourse Representation Theory (DR Theory) allow nicely for the composite
approach, and for subsequent calculations.
Once entered into a representation, information is subject to further interpre-
tation. Rules calculate temporal advancement by relating the current sentence
to the representation already developed. Information of more than one kind is
required: temporal advancement depends on the type of entity involved and
whether it is expressed as bounded or not in the sentence under interpretation.
Both temporal and aspectual information must be integrated to arrive at this
kind of text progression. This is the key factor in dealing with text progression
in the Narrative and Report modes.8
8. Rules for calculating narrative progression in the DR Theory framework are proposed in Kamp &
Reyle (1993); they are somewhat different in detail from the rules given here.
Description is temporally located but static. Text progression is spatial, fol-

lowing through the scene described. Changes in space are inferred from the
lexical information in a passage, supplemented with world knowledge and in-
ference. The static nature of descriptive passages can be accounted for in the
DRS, but spatial information is difficult to state in construction rules. The reason
for the difference is that spatial information is conveyed lexically, rather than
by terms of a closed system (Chapter 5).
In the atemporal modes, text progression is due to changes in the metaphorical
location of Primary Referents. Primary Referents can be identified with the
Theme/Patient argument in a clause; I will assume that such information is
coded and available in surface syntactic structure. Information about Primary
Referents is interpreted by construction rule and entered into the DRS. However,
determining metaphorical motion depends on pragmatic knowledge, lexical
rather than grammatical information. This kind of interpretation cannot be made
in the formal theory.
Surface structure presentation involves subjectivity and topic–focus infor-
mation. Perhaps surprisingly, the forms that convey subjectivity are amenable
to a Discourse Representation Theory treatment. I provide formal rules that
recognize subjective forms in texts and ascribe responsibility for them to the
correct source. The rules are stated for grammatical forms and verb classes that
trigger interpretations of subjectivity; see Chapter 7.
Sentence topic and focus, the other surface presentational factors considered
here, are problematic for construction rules. I suggest that presentational text
progression depends on sentence topics. The principles for identifying the topic
of a sentence are almost entirely pragmatic and are difficult to formalize. I do
not provide construction rules for topic phrases. These matters are discussed in
Chapters 8 and 9.
3.4.2 Discourse Representation construction rules and structures

I now give some construction rules and representations of Discourse Represen-
tation Theory. They introduce aspects of the theory that will be important in the
later discussion. The detailed analysis of Discourse Modes and presentational
structure will be developed in later sections of this book.
The theory determines the meaning of a sentence by rule from its mor-
phemes and surface syntactic structure. I assume a phrase structure syntax
with generative rules. The surface structures are organized on X-bar theory
lines. Inflectional Phrase (IP) is the maximal projection for the sentences in
this book. Tense is treated as the realization of I in IP. I use the functional
projection Aspect Phrase (AspP) for the simple and progressive verb forms.
The subject of a verb is directly dominated by IP; the direct object ap-
pears in the Verbphrase (VP) as sister to the verb. For simplicity, I ignore
higher projections and other projections that might appear in a more complete
structure.
The construction rules of the theory interpret the information in surface
syntactic structure and in the already-constructed DRS. Complex structures
and dependencies are modeled with sub-structures that are within the scope of
operators such as negation, quantifiers, modal subordination. The theory has
had notable successes in analyzing referential, scopal, and other phenomena.
For detailed accounts see the references in the text.
Rules of the theory applies to hierarchical surface structures with syntactic–
semantic features. The structures do not otherwise encode semantic informa-
tion: there is no level of Logical Form nor an associated semantic structure.9
The syntactic–semantic features are stated on the relevant syntactic node and
percolate to higher nodes in the syntactic tree. The features that I will use in
this introduction are NP number, gender, and case; verb class; aspectual view-
point associated with the simple verb form or the progressive; times and their
relations associated with tense morphemes. In discussions later other features
are used.
The rules account for the conventional meanings conveyed by a sentence,
abstracting away from lexical semantics. Structural constraints on interpre-
tation are modeled in a Discourse Representation Structure with embedded
sub-structures. Material within an embedded structure may not be accessible
as antecedent for anaphoric expressions.
As an introductory example, I show how entities for individuals and sit-
uations are represented for a discourse fragment, following the treatment of
Kamp & Reyle (1993). I then add tense and aspectual information. Consider (4):
(4) Mary read Ulysses. It fascinated her.
The figure in (5) gives a simplified syntactic tree representation of the first
sentence of this fragment. In the figure, PN = proper name; ø = zero, associated
9. Logical Form is a semantic level in Principles and Parameters theory and related theories; it is
irrelevant to DR Theory, which provides semantic interpretation.
Syntactic surface structures developed in other syntactic frameworks are also appropriate for
the construction rules of DR Theory. Kamp & Reyle (1993) use the syntactic framework and
representations of Gazdar et al. (1985). The framework is convenient because it is explicit and
has a carefully worked out way of dealing with syntactic–semantic features.
in the tree below with the perfective viewpoint; the E subscript on the verb is a
feature classifying it as an event verb.
(5) Mary read Ulysses.

IP
NP I'
N' I AspP
PN tense Asp' VP
Mary past
Asp V'
Ø V NP
read N'
PN
Ulysses
Construction of the DRS proceeds clause-by-clause; in the system of

Kamp & Reyle (1993) each syntactic constituent within a clause is interpreted
sequentially. The construction rules consist of a triggering configuration and
a statement of operations on the DRS. After a constituent is analyzed that
constituent is erased, so that when a sentence is fully interpreted only the
DRS version of it remains. Entities are introduced into the universe of dis-
course at the top of the DRS, and conditions are given beneath them in the
representation.
In the tree for “Mary read Ulysses” the first constituent below the IP-node
is NP, the sister of “I”; this is the canonical position for the subject of a sen-
tence. The discourse referent x is introduced for Mary, a proper name, with a
condition that identifies x as “Mary.” The construction rule is sketched in (6).
The triggering configurations are given in (6a), the operations in (6b). G e n =
gender, a feature on the NP node.
(6) Construction Rule for Proper Names

a. (i) IP (ii) XP
NPgen I' X NPgen
PN PN
b. (i) Introduce a new discourse referent into the universe of discourse.

(ii) Introduce a condition placing the discourse referent in parentheses
behind the proper name which the PN dominates in the triggering
configuration.
(iii) Replace, in the relevant syntactic structure, the NP-constituent with the
new discourse referent.
For the sentence structure in (5) the rule applies first to the configuration in
(6ai). The third condition of (b) results in an intermediate structure “x read
Ulysses,” not given here.
The next constituent considered here is the VP; I return to the other nodes
later in dealing with aspect and tense information. Rule (6aii) applies to the VP
configuration, licensing the entity y as Ulysses (“reads y”). The highest NP and
the VP together license a situation entity e and a condition that characterizes it;
and the syntactic structure is deleted. The complete rule is not given here; see
Kamp & Reyle (1993) for details.
The resulting partial DRS is given in (7): e introduces the event of Mary’s
reading. The information in the DRS consists of the discourse entities, listed at
the top, and the conditions, displayed below.
(7) Partial DRS for “Mary read Ulysses.”
x y e
1. x = Mary
2. y = Ulysses
3. e: read (x,y)
The figure in (7) follows the standard format for DRSs: a two-tiered box that
encloses the entire structure; embedded structures are represented by sub-DRSs,
which have the form of embedded boxes.
The tree structure for the second sentence of fragment (4), It fascinated
her, is like the first except that the NPs are pronouns. For each pronoun, a
new discourse referent is entered into the top part of the DRS, together with
conditions that identify it. Pronouns may be interpreted as coreferential with
an antecedent discourse referent if they agree in gender and number. I assume
that the construction rules include a mechanism for agreement. In (8), there is
a partial DRS for both sentences of (4); lines 4 and 5 provide that the pronouns
are coreferential with their antecedents, the discourse referents x and y. The
entity at the top is s, the situation of Mary’s being fascinated by Ulysses, a
state.
(8) Partial DRS for “Mary read Ulysses. It fascinated her.”
x y u v e s
1. x = Mary
2. y = Ulysses
3. e: = read (x,y)
4. u=y
5. v=x
6. s: fascinates (u,v)
In (8) the antecedents of the pronouns are available for coreference and there
are no other possibilities, so that the DRS identification is licensed.
In some cases the structure of the DRS makes an NP inaccessible as an an-
tecedent for coreference. When an NP is within the scope of a general quantifier
or negation, for instance, it cannot serve as antecedent to a pronoun. For in-
stance, the sequence “John doesn’t own a car. It is red” is ill formed because
the NP a car is in the scope of negation and cannot function as an antecedent
for the pronoun it. The sequence “The elephant is a gentle beast. He is run-
ning away from danger” is ill formed for similar reasons. The first sentence is
generic, while the second expresses a specific event and the pronoun refers to
a specific elephant. This kind of inaccessibility is represented in a DRS with
embedded structures. NPs are discussed more fully in Part II; the DRT treatment
of coreference is discussed in Kamp & Reyle (1993).
The rules can also be stated with labeled bracketing. The rules can “see” the
relevant configuration, as well as features on NPs, verbs, etc. These features are
needed anyway in a grammar of the language. Example (9), for instance, gives
a labeled bracket version of the rule above.
(9) DRS Construction Rules for Proper Names

a. IP [NP [PN [ . . . ]] I’[a b]] → x = . . .
b. IP [NP I’[a VP [V NP [PN [ . . . ]]] b]] → y = . . .
The structural description appears on the left side of the arrow; it corresponds
to the tree structure in (6). Here only the information relevant to a particular
rule is given; variables a, b represent other material in the sentence. The rule
specifies the relevant phrase structure nodes and, for the NPs in question, the
PN feature; the dots represent the actual proper names. On the right side of
the arrow is the interpretation, in this case entities x and y to be entered into
the DRS with the proper name as specifying condition. Rule (a) recognizes the
subject of a sentence, corresponding to the configuration in (6ai) above. Rule
(b) recognizes the direct object; it corresponds to (6aii) above, though it is more
specific. For the fragment in (7), the dots in rule (9a) represent the proper name
Mary; the dots in rule (9b) represent Ulysses. The rules in (9) result in the same
DRS as those of (6). I will use rules like (9) throughout the book.
Now we add rules that interpret the aspectual and temporal information of
fragment (4), repeated here:
(4) Mary read Ulysses. It fascinated her.
Aspectual information gives the class of the situation entity and the aspectual
viewpoint. The sentence “Mary read Ulysses” licenses an Event entity in the
DRS, because of the E feature on the verb and the proper names in subject
and object position; see Chapter 4 for discussion. The E feature on the verb is
visible to the rule, as in (10):
(10) Preliminary construction rule for an Event entity

b]]] → e = Event
IP [NP ] I’[a VP [VE NP [[PN]]
This rule adds the entity e at the top of the DRS and a condition which specifies
that the entity e is an Event, perhaps of a certain class.
The aspectual viewpoint of the sentence is perfective, because it has the
simple verb form rather than the progressive auxiliary. Example (11) indicates
very roughly how this interpretation is made. The rule focuses on the simple
verb form and the Event verb; it licenses a viewpoint condition in the DRS.
(11) Preliminary viewpoint rule

IP [NP] I’[a [AuxP [ø] VPE ]] b]] → View = Perfective
A more detailed account appears in Chapter 4.

The temporal information in “Mary read Ulysses” comes from the past tense
morpheme on the verb. The past tense indicates that the event occurred at a
time prior to the moment of speech, or Speech Time. This information is given
in terms of two times in the DRS, t1 and t2, with t2 prior to t1. The event
entity and the viewpoint are located at t2 in this preliminary version of tense
interpretation; three times are specified in a more complete account, given in
Chapter 5. The left side of the rule specifies that the past tense occurs on the
main verb, ignoring the possibility of time adverbials. The past tense licenses
the times and conditions as given on the right side; e indicates a situation entity,
event, or state.
(12) Preliminary tense interpretation rule

a VP b]] → t1 = SpT; t2<t1; e, View at t2
IP [[NP] I’[Tns [past]
The rules (10–12) add more information to the DRS, as shown in (13). Discourse
entities for times are added at the top, t1 and t2. Line 4 specifies that the situation
entity is an event, line 5 specifies that the viewpoint is perfective; the next two
lines characterize the times, and line 8 locates the event entity at t2.
(13) Partial DRS for “Mary read Ulysses.”
x y e t1 t2
1. x = Mary
2. y = Ulysses
3. e: read (x,y)
4. e = Event
5. View(e) = Perf
6. t1 = SpT
7. t2 < t1
8. e at 12
This preliminary account is developed further in Part II.

In actual discourse little is sealed off as semantic, as I note at the beginning
of this chapter. Even at the level of reference, semantic information must usu-
ally be supplemented by pragmatic inference. The construction rules of DRT
cannot choose between possible antecedents of a pronoun, nor arrive at other
interpretations based on inference and world knowledge. The simplicity of the
examples used here masks this important point. However, it is equally important
that structural information excludes certain possibilities.
I I Linguistic analysis of the
Discourse Modes
4 Aspectual information: the
entities introduced in discourse
The Discourse Modes introduce situation entities into the universe of discourse,
and have different principles of text progression. The information is conveyed by
forms of aspect and temporal location. The two are complementary. Temporal
location locates a situation in time, while aspect specifies the internal temporal
structure of the situation. This chapter is devoted to aspect; temporal location
is discussed in Chapter 5.
The understanding of aspect is the key to the analysis of Discourse Modes,
because the situation entities are aspectual in nature. The Discourse Modes
characteristically introduce different types of situation entities. Passages of
the Narrative and Report modes primarily involve Events and States. The
Description mode primarily concerns States and ongoing Events. The Infor-
mation mode primarily has General Statives; the Argument mode primarily has
Abstract Entities, Facts and Propositions, and General Statives. After a brief
introduction to aspectual systems, this chapter discusses the concepts of situ-
ation entities and their linguistic correlates, and how aspectual information is
encoded in Discourse Representation Theory.
Section 4.1 introduces the two components of aspectual systems and dis-
cusses Event and State situation types, including shifts of one type to another;
4.2 discusses General Statives; 4.3 discusses Abstract Entities; 4.4 discusses
the linguistic correlates for the three major classes of situation type; 4.5 covers
aspectual information in Discourse Representation Structures.
4.1 Aspectual categories

4.1.1 The two components of aspectual systems
Aspect is a sub-system of language that conveys information about the internal
temporal structure of situations. Aspectual systems have two components, situ-
ation type and viewpoint (Smith 1991). The components interact in sentences.1
1. The two-component theory of aspect is presented in Smith (1997).
67
68 Aspectual information
Both are relevant to the Discourse Modes. Situation entities realize different
situation types, while viewpoints focus all or part of a situation. Together they
determine the boundedness of situations, which is essential for temporal pro-
gression in the Narrative mode. Situation type indirectly classifies situations
into the major categories of Event and State situations, statives, and Abstract
Entities. In previous work I have used the term “situation type” for situations
and General Statives. I now extend it to include abstract entities as a separate
class.
Situation types are idealizations of situations, semantic concepts organized
according to their internal temporal properties. They are conveyed by the com-
posite of the verb and its arguments, the “verb constellation.” Verb constella-
tions of each class have distinctive distributional properties. Because of these
properties, the situation types are covert linguistic categories which are tacitly
available to the speakers of a language.
The aspectual viewpoint of a sentence is like the lens of a camera. It focuses
on all or part of the situation expressed by a sentence, making the focused infor-
mation visible. Only visible information is available for semantic interpretation.
Viewpoint is important for calculating advancement in the temporal Discourse
Modes, because it is one of the factors that determines whether a situation is
bounded or unbounded.
Viewpoint and situation type constitute closed “closed systems,” within
which a choice must be made from a few possibilities. There are two main
viewpoints in English, marked morphologically as in most languages. The
simple verb form is “perfective,” focusing events with bounds (“Mary made
a desk”). States are unbounded in the perfective, however (“John knew the
answer”). The progressive verbal auxiliary is an “imperfective”: it applies only
to events.2 The progressive focuses an open interval, without endpoints: an inter-
nal interval of a durative event (“We are building a sandcastle”) or a preliminary
interval of an instantaneous event (“He is winning the race”). Both are taken as
ongoing events. Progressive viewpoints appear in many languages. Other imper-
fective viewpoints, for instance the French imparfait, apply to states as well as
events.
The information given by a viewpoint depends on the type of situation entity
it focuses. Each class of verb constellation and aspectual viewpoint has an
associated temporal schema, and viewpoints focus all or part of the temporal
schema. The relationship between situation type and viewpoint is modeled by a
2. There is a marked use of the progressive with stative verb constellations, often found in informal
speech, e.g. “I’m loving this walk.” Such sentences present states as events (Smith 1997).
composite of the schema of each component. The aspectual viewpoint schema

focuses all or part of the situation type schema. What follows is a brief account
that deals mainly with material relevant to the Discourse Modes; for detailed
discussion see Smith (1997).
4.1.2 The Event and State situation types

Classifying situations by temporal properties has a long tradition, starting as
we know it with Aristotle, who recognized the distinction between static and
dynamic situations. Vendler (1957) and others developed a finer classification
of States and types of Events, using three temporal properties. Events and States
are typically found in passages of the temporal discourse modes – Narrative,
Report, and Description.
Three two-valued temporal features hold of Events and States: dynamic–
static, telic–atelic, durative–instantaneous. The feature dynamic–static is most
far-reaching, distinguishing states from events. Events take place in time, due
to their dynamism; states are specific situations that hold in time. Dynamic
situations occur at successive stages. Every stage of an event takes time, and
thus involves change of some kind. Instantaneous events consist of a single
stage, a point in principle; events with duration have endpoints, changes to
and from a state of rest. Telicity involves endpoints: telic events have a goal
or natural final endpoint (e.g. walk to school). Atelic events have arbitrary
endpoints: they can end at any time (e.g. walk in the park). In contrast, states
are static, consisting of an undifferentiated interval with no structure (e.g. own
the farm, know the answer).
Clusters of these features distinguish five situation types: States and four types
of Events – Activities, Accomplishments, Semelfactives, and Achievements.
The types are like Vendler’s except that I add the category of Semelfactive,
instantaneous atelic events. Semelfactives are sometimes treated as a subtype of
Achievements. Semelfactives are a separate class because they have distinctive
linguistic properties (Smith 1997).
(1) Situation types:

Events: dynamic
Activities: atelic, durative (laugh, push a cart, stroll in the park)
Accomplishments: telic, durative (build a house, walk to school, learn
Greek)
Semelfactives: atelic, instantaneous (tap, knock, flap a wing)
Achievements: telic, instantaneous (win the race, reach the top)
States: static
States: durative (know the answer, love Mary, be sick, be in the room)
Associated with each situation type is a schema that represents its temporal
properties. The temporal schema of Events includes either a single point or an
initial and final point, depending on the feature of duration. The final endpoint
is natural or arbitrary, depending on telicity. The schema for a State does not
have endpoints: changes into and out of a state are events in their own right
and not part of the state. For simplicity I will sometimes call a verb constel-
lation or sentence by the situation type or temporal feature with which it is
associated – for instance, an “event” or “dynamic” sentence.
4.1.3 Viewpoint and situation type

In an aspectual system, aspectual viewpoint interacts with situation type, deter-
mining the information conveyed by a sentence about the situation expressed.
The examples in (2) illustrate two expressions of a dynamic, durative Event
with natural endpoints (an Accomplishment). Example (2a) has the progressive
viewpoint, (2b) has the perfective. Line (i) of each example gives the situation
type schema: I and Fnat represent initial and final endpoints, the dots represent
internal stages. Line (ii) gives the viewpoint schema, with dots representing the
span of the viewpoint. In the composite schema of line (iii), slashes indicate
the span focused by the viewpoint.
(2) a. Informal temporal schema for They were building a sandcastle
i. [They build a sandcastle] I . . . . . Fnat (Event)
ii. [be+ing] ... (progressive)
iii. They were building a sandcastle I. . . . . . Fnat
//// (composite)
b. Informal temporal schema for They built a sandcastle
i. [They build a sandcastle] I. . . . . Fnat (Event)
ii. [Ø] ........ (perfective)
iii. They built a sandcastle I. . . . . . Fnat
////////////// (composite)
The visible span of the viewpoint inherits the properties of the entire situation
schema.3 This correctly provides that we know the situation type of a sentence,
independent of whether the viewpoint makes visible all or part of the situation.
3. For Activities the precise statement is more complex. Activities are focused by the perfective
viewpoint with implicit temporal bounds which may, but need not, coincide with their endpoints.
In other words, the temporal boundedness of the viewpoint does not require that an activity
terminate.
The perfective viewpoint focuses a segment that is implicitly bounded. If the activity continues
after the implicit bound, the continuation constitutes another temporal segment. The point is
delicate but important for understanding the contribution of Activities to discourse. Assertions
of continuation may felicitously be made, as in the examples below. The second conjunct of
The schemata illustrate the effects of viewpoint. Perfective event sentences

focus Events as bounded, discrete entities, while progressive, imperfective
event sentences focus events as unbounded. The distinction between bounded
and unbounded situations is important for narrative advancement. Telic events
have intrinsic bounds bounds for a situation may also be given with ex-
plicit information. Imperfective events and non-dynamic situations are all un-
bounded, and have certain properties in common.4 Entities that are static include
States, General Statives, and Abstract Entities. Recall their effect in narrative:
perfectives advance narrative time, progressives and states do not.
The interpretations are formalized in Discourse Representation Structures by
introducing situation entities, temporal intervals, and conditions. The conditions
specify the class of a situation entity according to the verb constellation that
introduces it. The “visible” portion of the situation is conveyed according to
the viewpoint and its interaction with the situation type.
4.1.4 Coercion
There is a standard, basic-level set of associations between situations in the
world and idealized situation types. In choices that depart from the standard,
the speaker uses derived situation types (Smith 1995a). They are shifted by rule
from the basic-level types through a mechanism widely known as “coercion,” a
term due to Moens (1987). No additional situation types are needed to account
for coercion: derived situation types fall into the same classes. Shifts in situation
type are triggered by material in the context, often by adverbials. For instance,
“Mary knew the truth” is a state sentence. With the adverbial suddenly, as
in “Suddenly Mary knew the truth,” the sentence expresses an event, that of
Mary coming to know the truth. Another example is the perfect: in the perfect
construction dynamic sentences shift to states, e.g. “We have (already) eaten
each sentence asserts that another segment, or temporal unit, follows the unit presented in the
first conjunct (Smith 1999b).
a. Mary worked and continued working without a break.

b. Sam rehearsed and kept on rehearsing.
c. Alice thought about the math problem while Chris drove her to the station, and she
kept on thinking about it all morning.
4. States and other static situations are homogeneous, with a uniform part structure. They have the
sub-interval property: when a state holds for an interval, it holds equally for any smaller interval
of that interval. Progressives, too, have this property.
In a view that I have characterized as the “strong mereological” approach, states, progressives
and Activities all belong to the one global semantic category because of their homogeneity. For
discussion see Smith (1999) and the references therein.
dinner.”5 Derived situation types include the basic verb constellation and the
triggering adverbial.
Adverbials of duration and completion can also trigger a shift in situation
type. Telic verb constellations become atelic with a simple durative adverbial,
as (3) shows:
(3) a. Mary wrote a letter.
b. Mary wrote a letter for an hour.
The understanding of (3a) is that Mary wrote and completed a letter; that of
(3b) is that she was engaged in the activity of letter-writing, but that she did
not actually complete a letter. The shift in understanding is due to the different
aspectual values of the verb constellation and the adverbial: durative adver-
bials are atelic. Example (3) illustrates a consistent pattern in coercion. If there
is a clash in value between the verb constellation and an adverbial or other
form external to it, the external value overrides. This pattern is formalized in
Section 4.4 as the Principle of External Override.
The progressive is often treated as an instance of coercion. In this approach
the progressive is an operator that applies to events and produces states as a
derived situation type (Vlach 1981, Kamp & Reyle 1993). The treatment has
the advantage of formally unifying progressives and static situations. But there
is a disadvantage: progressives are not entirely like states.6 For discussion see
Smith (1983, 1995b).
I now turn to the two other major classes of entities. They are all static in
nature and thus form a super-class with States.
4.2 General Statives

General Statives are more abstract than the States considered above, because
they do not express particular situations. There are two types, Generic and Gen-
eralizing sentences. Generic sentences hold of kinds; Generalizing sentences
express a pattern of situations, a generalization or regularity. Example (4) illus-
trates, repeated from Chapter 2:
5. The “perfect” construction is formed with the verbal auxiliary have and a participle. In English
there are past, present, and future perfects.
Perfects are semantically states: one piece of evidence is that they do not advance narrative
time. The perfect affects temporal location as well as aspect; see Comrie (1976), Kamp & Reyle
(1993), Michaelis (1994).
6. The differences between them are significant: (a) progressives cannot be taken as inchoatives,
whereas states can be; (b) progressives are dynamic, whereas states are not; (c) progressives and
statives have different co-occurrence properties with adverbials as noted in Glasbey (1998).
4.2 General Statives 73
(4) Sentences expressing statives

a Dinosaurs are now extinct. Generic
b The lion has a bushy tail. Generic
c Mary speaks French. Generalizing
d John always fed the cats last year. Generalizing
Sentences like these are studied in depth in the articles in Carlson & Pelletier
(1995).
Generic sentences refer to kinds rather than individuals: in (4b) the subject
NP denotes the entire class of lions, not a particular lion or lions. Definite NPs
(the lion) and bare plurals (lions) are the main types of NP that are used as kind-
referring. Indefinite NPs, both mass and count, can be used in a “taxonomic”
reading, as in “The World Wildlife Organization decided to protect a (certain)
large cat, namely the Siberian tiger,” an example from Krifka et al. (1995).
Generalizing sentences express regularities, not particular facts. Such sen-
tences often have a frequency adverbial (sometimes, always, never). There are
a number of special forms which lead to the general reading, e.g. used to, the
agentive -er suffix, the middle voice, present simple tense with an Event verb
constellation. Kind-referring NPs can occur in a Generalizing sentence, as in
“Potatoes are served whole or mashed as a cooked vegetable.” The characteri-
zation and examples above are due primarily to Krifka et al. 1995; they use the
term “characterizing” for the class that I call Generalizing sentences.7
Generic and Generalizing sentences are static: they have no dynamism. They
are derived by coercion from verb constellations that express specific situations
at the basic level of categorization. For instance, the verb, object argument,
and adverbial of examples (3c, d) above – “speak French”, “feed the cats” –
standardly express specific events. In context with a definite NP subject, simple
viewpoint, and present tense, however, they have the generalizing interpretation.
Similarly, Generic sentences may have event verb constellations, as in “The lion
eats meat.” Verb constellations that express states at the basic level also appear
in General Statives, as in “have a bushy tail.”
7. Two additional properties of Generic sentences are noted by Krifka et al. (1995): Generic sen-
tences with the indefinite article don’t express accidental properties. For instance, “polyphonic”
is an intrinsic property of “madrigals,” but “popular” is an accidental one. The sentence “The
madrigal is polyphonic” can be taken as generic or particular. But with the indefinite article,
only the ascription of an intrinsic property is acceptable: “A madrigal is polyphonic” is fine, but
#“A madrigal is popular” is odd. This is a slightly different version of the characterization by
Krifka et al.
Kind-denoting NPs in Generic sentences are only plausible with a nominal that refers to a
well-established kind. Thus, “The Coke bottle has a narrow neck” is fine but #“The green bottle
has a narrow neck” is odd (indicated by #), on the generic reading.
General Statives are spatiotemporally located in the world: they hold at par-
ticular intervals and places, though these coordinates may not be specified in a
given sentence.
4.3 Abstract Entities

The class of Abstract Entities consists of Facts and Propositions. Conceptually,
Facts and Propositions can be distinguished from other entities by the factors
of spatiotemporal location and causality. Situations and General Statives are
located in the world and have causal powers. Facts are not so located, yet they
are contingent for truth on situations being a certain way and arguably have
causal powers. Propositions are not located, are not contingent, and do not
have causal powers. The linguistic and philosophical aspects of these entities
are discussed in Vendler (1967, 1972), Asher (1993), Peterson (1997).
Facts are assessments that we make of states of affairs, abstract or concrete.
Facts are not in the world; but they are about the world. They are expressed
linguistically as clausal or deverbal arguments of certain predicates, e.g. matter,
amuse, explain, show, indicate etc., as in these examples:
(5) a. I know that the German war effort collapsed.

b. The collapse of the Germans was surprising.
We point out a fact, regret or rejoice in a fact. If there is a dispute about a

fact, one challenges by asking, “How do you know?” or “Why are you so
sure?” These questions assume that the speaker is in a position to know, or
have satisfactory reasons. Questions of fact are empirical questions, although
facts are not part of the furniture of the world (Herbst 1956). Facts transcend
the limitations of subjectivity: they are “objective,” waiting to be discovered
(Vendler 1972:82). The class of fact predicates includes mental event verbs
(forget, discover, realize); emotives (regret, resent, deplore); evaluatives
(matter, be significant/important/odd/surprising/mysterious); psychological
verbs (surprise, amuse, annoy).
Propositions are the objects of such mental states as beliefs, expectations,
and decisions. They are expressed by clausal and deverbal arguments of verbs
of propositional attitude and other predicates.
(6) a. I believe that Mary refused the offer.

b. Mary’s refusing the offer was unlikely.
Since propositions are the mental states of the particular individuals who hold
them, they have subjective features. Propositions are referentially opaque: they
do not accept substitution of a different expression with the same referent.
This property reveals their subjectivity (Vendler 1972:81). In contrast, facts

allow such substitution. The class of propositional predicates includes mental
verbs (believe, fear, hope); verbs of communication (say, tell, show, indicate,
affirm, deny); epistemic predicates and others (seem, appear, be possible/likely/
unlikely/certain/true/consistent); logical verbs (entail, imply).
Sentences that directly express Facts and Propositions cannot be distin-
guished on linguistic grounds, and so will not be discussed here.
4.4 Linguistic correlates of situation entities

Clauses expressing different types of entities have distinctive linguistic features,
which can be used as tests to determine what type of entity is expressed in a
given clause.
4.4.1 Situations: the temporal properties

The temporal features of the Event and State situation types have characteristic
distributional properties.
The property of dynamism has a close association with the semantic feature
of agency. Dynamic events may have agents as sources of energy and volition;
States do not have agents, at least not directly. Reflecting this, linguistic forms
which are associated with agency occur with event verb constellations but not
with statives (Lee 1971, Ross 1972). Imperative sentences, and complements
of verbs like persuade, command, require event verb constellations. Only an
event that is controllable can be commanded or persuaded. For instance, it’s
natural to say “Wash your car!” or “I persuaded Mary to wash her car”; but odd
to say #“Know Greek!” or #“I persuaded Mary to know Greek.” Also related
to agency are adverbs of manner and instrument.
Adverbs of intention and control are plausible with events, odd with states:
compare “John carefully washed the car” and #“John carefully knew Greek.”
These correlates of agency are semantically based and hold across languages.
There are also two grammatical correlates of dynamism in English: the pro-
gressive viewpoint and the pro-verb do in the pseudo-cleft construction. Both
require event verb constellations, as (7) shows (∗ indicates ill-formedness):
(7) a John was washing the car. (Event)
∗
b Kim was knowing the answer. (State)
c What John did was wash the car. (Event)
d #What John did was know Greek. (State)
In other constructions the pro-verb do appears with both event and state verb
constellations. For instance, in verb-ellipsis constructions such as “John knows
Greek and Mary does too,” “do” is a pro-verb for the State “know Greek.”
The interpretation of present-tense events is another correlate of dynamism.

Events are taken as generalized, or habitual, in sentences with present tense
and the perfective viewpoint. In contrast, similar sentences with state verb
constellations express a particular state. The examples in (8) illustrate:
(8) a Mary swims in the ocean. (Event)

b Ellen is in the room. (State)
Sentence (8a) must be interpreted as generalizing, whereas (8b) expresses a

State that holds at a particular time.
The generalizing interpretation of sentences like (8a) is due to a pragmatic
constraint on bounded events that prevents them from being located in the
Present. In the Present, events must be presented as ongoing, e.g. “John is
talking,” “Mary is drawing a circle.” Event verb constellations are otherwise
taken as conveying a general pattern, like “Tom (often) feeds the cat.” This
constraint, which I call the Bounded Event Constraint, is due to a principle
of communication that holds in language generally: speakers follow the con-
vention that communication is instantaneous. The perspective of the Present is
incompatible with bounded events, because the bounds would go beyond that
perspective; see Chapter 5 for discussion.
Together, these syntactic and semantic properties distinguish sentences with
the temporal feature of dynamism, that is, event sentences. The linguistic corre-
lates of telicity and duration do not pertain directly to the Discourse Modes and
will not be discussed here.8 State verb constellations, which are static, do not
8. Telicity is not expressed directly in language at the level of a general distributional property,
although it is an important conceptual property of Events for human beings. There seem to be
no linguistic correlates of it per se. The syntactic evidence for a telic Event turns on the notion
of completion, which involves the interaction of duration and change of State.
Verbs and adverbials of completion (finish, in an hour) contrast with verbs and adverbials of
simple duration (stop, for an hour). Telic verb constellations are compatible with completion; in
contrast they are odd with forms of simple duration, requiring a special derived interpretation. In
contrast, atelic verb constellations are odd with forms of completion, as the examples illustrate:
a Mary walked to school in an hour. (telic)

b Mary finished walking to school. (telic)
c ? Mary walked in the park in an hour. (atelic)
d ? Mary finished walking in the park. (atelic)
These contrasts are semantically based: the notion of completion is intrinsic to a telic Event,
irrelevant to an atelic Event. Other tests for telicity involve verbs such as stop, finish, verbs of
time take and spend with sentential complements. Take is compatible with telic verb constel-
lations, spend with atelic verb constellations. The property of duration also has adverbial and
verbal linguistic correlates; see Smith (1997).
appear in imperatives, as complements of persuade, with volitional and instru-

mental adverbs, with the progressive viewpoint, or in pseudo-cleft sentences.
In present sentences with the simple, perfective verb form they have a specific
interpretation.
Verb constellations of the different types are composed and interpreted by
rules. The rules are sensitive to features of the verb and its arguments, and to
adverbials. These matters are discussed in 4.5, where some compositional rules
are presented.
4.4.2 The linguistic correlates of General Statives

Generic and Generalizing sentences are not marked by particular morphemes
nor by verb class. Nevertheless, there are a few distributional characteristics
that distinguish them. I summarize them briefly.
Generic sentences: Kind-referring NPs appear in these sentences. If an NP
is felicitous with predicates that refer only to kinds, such as die out, be extinct,
invent, it can function as kind-referring. Definite NPs and bare plurals are most
commonly used to refer to kinds, as in “The dinosaur is extinct,” “Dinosaurs are
extinct.” Strikingly, indefinite NPs cannot appear as subjects of these predicates:
“The lion / lions / will soon become extinct,” but #“A lion will soon become
extinct” is odd (Krifka et al. 1995:9–10).
Generalizing sentences: If a sentence has a frequency adverb such as usually
or typically, or if such an adverb can be felicitously added to a sentence, the
sentence is of the generalizing class. The progressive viewpoint in English is
limited to dynamic situations; it typically excludes generalizing predicates too.
This is not an absolute criterion: one can construct generalizing sentences with
the progressive, e.g. “John is calling his sister more these days.” Sentences
like this often have an explicit or implicit contrast with a different state of
affairs.
When set in the Present time, many Generalizing sentences can be recognized
by a particular combination of aspectual and temporal factors. Such sentences
have an event verb constellation, the perfective viewpoint (the simple verb
form) and present tense, as in “Mary speaks French.” The generalizing stative
interpretation is due to the Bounded Event Constraint. This cue to Generalizing
sentences does not apply to generalizations of states, which have basic state
verb constellations (“Mary always likes Italians”).
Although Generalizing sentences lack the dynamism of particular event sen-
tences, they have some distributional properties of dynamism (Smith 1997).
They can appear with forms associated with agency and control, and with
pseudo-cleft do:
(9) a. John deliberately plays tennis every Friday.

b. I persuaded John to play tennis every Friday.
c. What John did was to play tennis every Friday.
These distributional facts reflect the hybrid nature of Generalizing sentences.

Generalizing sentences are semantically stative. They tend to have dynamic
verb constellations and they imply a series of sub-events which are dynamic.
4.4.3 The linguistic correlates of Abstract Entities

The linguistic properties of abstract entities are both semantic and distributional.
Vendler was perhaps the first to notice the latter: he found distributional differ-
ences between sentences with argument complements that referred to Events
and States, Facts, and Propositions (1967, 1972). The key factors are the form of
the complement and the main predicate. Complements may take one of several
forms in English. Substantives, or deverbal complements, have nominal prop-
erties. They appear with determiners, adjectives, and noun plurals, as in “John’s
clever analysis of the situation.” Clausal that-complements are sentential, with
full verb forms. Gerundives and infinitives are closer to the verbal end of the
continuum, as shown by their ability to appear with adverbials, as in “John’s
cleverly analyzing the situation,” “For John to fully understand the situation.”9
Clausal complements also appear after NPs such as the fact that, the idea that,
the proposition that, etc.
There is a correlation between the form of a complement and its interpreta-
tion. Substantives are associated with situations and that-clauses and gerundive
complements with Abstract Entities, Vendler showed. For instance, take place
and be slow are situation predicates: both allow a substantive complement but
not a gerundive or that-clause complement. In contrast the verb know is a fact
predicate and allows only a that-complement. For the two fairly representative
predicates, the pattern of grammaticality is this:
(10) a. The collapse of the Germans occurred yesterday/was slow.

b. ∗ That the Germans collapsed was slow.
c. ∗ The Germans’ collapsing was slow.
(11) a. I know that the Germans collapsed.
b. ∗ I know the collapse of the Germans.
c. ∗ I know the Germans’ collapsing.
9. The complement forms of English include the substantive with -ing, as in “The collapsing of
the Germans,” “John’s cooking of the dinner,” “Mary’s analyzing of the situation.” These forms
have all the properties of substantives; they are not preferred when a morphological substantive
is available such as collapse, analysis.
The complement of (10a) refers to an Event, that of (11a) refers to a Fact. Thus
in the clearest cases complement form and predicate classes together fall into
distinct distributional patterns.
However, other patterns are less clear. Certain verbs and predicates are
flexible, allowing several kinds of complement. Among the flexible classes
are epistemic predicates such as be unlikely and psychological verbs such as
surprise, horrify. The complements of such predicates may be a substantive,
gerundive, or a that-clause:
(12) a. The collapse of the Germans was unlikely.

b. The Germans’ collapsing was unlikely.
c. That the Germans collapsed was unlikely.
(13) a. The enemy’s destruction of the city horrified us.
b. The enemy’s destroying the city horrified us.
c. That the enemy destroyed the city horrified us.
These sentences are acceptable as paraphrases, as Vendler noted. However,

psychological predicates with a substantive may be ambiguous between an
Event or fact interpretation. For instance, the clausal subject of “John’s singing
of the Marseillaise surprised me” may be interpreted either way.
The distributional facts suggest that we must recognize predicates that are
clearly eventive, factive, or propositional; and predicates that are flexible. For
the latter classes, certain complement types are indeterminate.
To assess the interpretation of a complement, two paraphrase or substitution
tests are offered by Peterson (1997). The first test substitutes That S for a clausal
complement, while the second substitutes an indirect question for a clausal
complement. The question is whether the substitutions preserve grammaticality.
Together, the two tests distinguish clausal complements that refer to Facts,
Propositions, and situations. A clausal complement refers to a Fact if both
substitution tests preserve grammaticality, as in (14):
(14) a. Mary’s having refused the offer was significant.

b. That Mary refused the offer was significant.
c. What Mary refused was significant.
A clausal complement refers to a Proposition if the first substitution test pre-

serves grammaticality and the second does not, as in (15):
(15) a. Mary’s having refused the offer was inconsistent.

b. That Mary refused the offer was inconsistent.
c. ∗ What Mary refused was inconsistent.
A clausal complement refers to an Event or State if both of the substitution tests

destroy grammaticality:
(16) a. Mary’s having refused / refusal of the offer was followed by silence.
b. ∗ That Mary refused the offer was followed by silence.
c. ∗ What Mary refused was followed by silence.
These tests are quite effective. Peterson found that essentially the same proper-
ties support the distinction in such varied languages as Arabic, English, French,
German, Hebrew, Hindi, Kannada, and Marathi. On the basis of this data,
Peterson (1997) argues that the categories are universally linguistic as well as
conceptual.
There are some difficulties in actually applying the tests because not all sen-
tences have the form stipulated. It is often necessary to “normalize” a sentence,
to construct a version of the sentence which fits the substitution template.10
Normalization can be unreliable. Moreover, it’s not always clear when a com-
plex nominal has an underlying clausal structure and when it is just what it
seems to be, a nominal. Abstract nouns, for instance, are arguably often propo-
sitional. In many cases they seem intuitively to involve a proposition, and they
occur with predicates that allow sentential complements.
When are we justified in substituting a clause for an abstract noun or com-
plex nominal? In some cases the substitution requires little change, in other
cases several are needed, changes that are on the borderline toward reinterpre-
tation. The examples of (17) illustrate with sentences from texts in this study.
They require increasingly drastic changes in complement form to arrive at a
normalization.
(17) a The routine transfer of power may not be the most dramatic feature of
American democracy, but it is the most important.
a’ (The fact) that power is routinely transferred may not be the most dramatic
feature of American democracy, but it is the most important.
b The national outpouring after the Littleton shootings has forced us to con-
front something we have suspected for a long time: the American high
school is obsolete and should be abolished.
b’ (The fact) that there was a national outpouring after the Littleton shootings
has forced us to confront something we have suspected for a long time: the
American high school is obsolete and should be abolished.
c But it was the $10 or $11 price of February 1999, not the one today, that
really deserved the headlines.
10. The process that I refer to as “normalization” is called “De-vendlerization” by Peterson

(1997:75). Vendlerization is a “pseudotransformation” deriving a nominal sentence via trans-
formation from underlying full sentence structure. De-vendlerization is the reverse, a procedure
for obtaining the underlying structure.
c’ The $10 or $11 price of February 1999, not the one today, really deserved
the headlines.
c” But that the price was $10 or $11 in February 1999, rather than the one
today, really deserved the headlines.
In (17a, b) normalization turns the subject nominals into verbal clauses. For
(17c), however, more changes are needed. The original is a cleft sentence. The
restatement of (17c’) removes the cleft, and (17c”) gives a that-clause version
which is arguably the same semantically as (17c). Normalizations like this
are often necessary for carrying out the tests proposed by Peterson. Sentence
(17c) illustrates the kind of problem that can arise when trying to produce
sentences of the right form for the substitution tests. One must decide when the
transformation results in a sentence that is appropriately close to the original
sentence, and when the result is too far from the original.
Moreover, some classes of predicates don’t give the right results with these
tests, as Peterson himself points out. The troublesome classes include verbs of
communication and conjecture (e.g. tell, say, show, indicate, guess, predict,
estimate), which are ambiguous between Fact and Proposition readings but
permit indirect question substitution. The substitution test predicts that propo-
sitional complements exclude the substitution of indirect questions, as in (15)
above. Emotive verbs such as regret, resent, deplore are also counter-examples:
they don’t take indirect questions (∗ “I resent who won the race”), but are
expected to do so since semantically their complements denote facts.11
Nevertheless the substitution tests are useful and I will assume that, together
with distributional classes of predicates, they can usually distinguish comple-
ments referring to Facts and Propositions.
There is also a quantification test that distinguishes between types of entities.
Asher shows that quantifiers such as everything, something, can be used in
sentences that express entities of the same class. But they are odd if the entities
are of different classes (1993:33). Consider (18):
(18) a. Everything that happened took an hour.
b. Everything that John believes is true.
c. #Nothing John believes takes an hour.
Example (18a) involves Events, and the sentence is good: the main clause
expresses an Event and the complement clause refers to Events. Similarly, (18b)
expresses and refers to Propositions. But (18c) is odd, even uninterpretable,
11. These verbs are known as “factive” verbs. They presuppose the truth of their complements.
Thus, the truth-conditional status of the complement remains under question and negation, e.g.
“Did you resent John’s winning the race?” “I don’t resent John’s winning the race.” Peterson
extends the term “factive” to include a wider range of Abstract Entity complements.
because it expresses an Event and refers to a belief. The belief context and
temporal location context are not satisfied by the same kind of entity.
4.5 Aspectual information in Discourse Representation Structures

The key to the Discourse Modes is the type of entity introduced in the clauses
of a text passage. In the structures of Discourse Representation Theory, sit-
uations (Events and States), General Statives, and Abstract Entities are rep-
resented as discourse entities of situations, with conditions that characterize
them. Discourse entities can be referred to, hence they are also known as
discourse referents (Karttunen 1976). The aspectual meaning of a sentence
is represented in a Discourse Representation Structure with situation entities,
times, and characterizing conditions.
Every clause introduces a situation entity into the developing DRS. Infor-
mation in the clause licenses a condition giving the situation type of the entity.
I will use [e], [s], and [a] for the classes of situation, general stative, and
abstract entity respectively. Situation type is determined by compositional rules
which take as input the surface structure of a sentence, and produce as output a
situation entity of a given situation type; examples will be given directly. The
temporal properties of the situation types appear in the DRS as intensional con-
ditions on a situation entity. The properties have a procedural force in sentence
interpretation. For instance, if a situation has the property [Telic], the final end-
point made visible by the perfective viewpoint is natural rather than arbitrary.
If a situation has the property [Instantaneous] only a single point is focused by
the perfective.
Aspectual viewpoint makes visible information about a situation, and may
add conceptual meaning; for instance, the progressive has a dynamic sense
(Smith 1997). Viewpoint is independent of situation type.12 To represent this
independence, an interval [I], and times [ti, j], are introduced into a DRS with
each clause. The interval is the locus for viewpoint information. Information
12. This independence is one of the basic ideas of the two-component theory. One argument for it
is that the span of a viewpoint does not necessarily coincide with the span of a situation type.
For instance, the progressive of an instantaneous Event focuses a preliminary interval that is
not part of the Event itself: “He was reaching the top”. There are additional viewpoints in other
language that have similar properties.
Another argument is that the viewpoint of a sentence does not obscure its situation type.
The situation type of a sentence is available to the receiver whatever its viewpoint. Consider
a sentence with the progressive viewpoint “Jane was walking to school.” Receivers of this
sentence know that only part of the Event is visible, and what sort of Event it is. They know
the nature of the final endpoint, although it is not semantically visible and may not occur.
4.5 Aspectual information in DRSs 83
made semantically visible by the viewpoint appears in the DRS as conditions

on the interval [I], given in terms of times and the endpoints or internal stages
of [e].
4.5.1 Interpreting situation type

The situation type of a sentence is determined by the verb and its arguments,
the verb constellation. For instance, “John walked to school” expresses a telic
event, whereas “John walked in the park” expresses an atelic one. The dif-
ference is in the prepositional phrase (PP) complement. The complement is
directional in the first case and locational in the second. In other cases the sub-
ject or object NP affects situation type: “Mary drank a glass of wine” is telic,
“Mary drank wine” is atelic, since no amount of wine is specified. In cases
of coercion, adverbials or other forms outside the verb constellation determine
the derived situation type. These points are uncontroversial (Verkuyl 1972,
Smith 1997).
Compositional rules provide a natural mechanism for interpreting situation
type. The important notion is interpretation. By computing the aspectual value
of the verb and its arguments, the rules arrive at the situation type of a sentence.
They depend on a prior assignment of aspectual values to the constituents of
the constellation. The process of determining the aspectual meaning of a
sentence or clause involves three stages:
A. The surface structure representation of a sentence is scanned for the

features of the verb constellation, the temporal adverbials, and the
viewpoint. The aspectual features of the constituents are assigned in
the lexicon. The features relevant to aspectual value include whether
or not an NP argument is quantized; whether the verb alone is telic
or atelic; the value of complement PPs (locative, directional, etc.) and
adverbials.
B. Compositional rules compute the situation type of the verb constella-
tion and the relevant temporal adverbials, if any. These rules take as
input the surface structure of a sentence and produce aspectual infor-
mation as output. Their output is a situation discourse entity and a
condition that identifies it as a situation, General Stative, or Abstract
Entity.
C. The aspectual meanings are stated in the DRS. Situation type infor-
mation occurs in the DRS in the form of conditions on the situation
entity introduced with every clause. In the case of derived rather than
basic-level categorization, both levels are preserved by the rules.
Verbs are assigned intrinsic aspectual features in the lexicon. The intrin-
sic value of a verb is determined by its value in a simple verb constellation
with obligatory arguments. Simple constellations have the minimum number
of countable arguments that a verb allows. For instance, the verb walk may
appear in intransitive atelic sentences such as “Mary walked,” and in telic sen-
tences such as “Mary walked to school.” Since it is atelic and durative in the
simpler constellation, walk is assigned the intrinsic features [Atelic] and [Dur].
The essential aspectual feature of NP arguments is whether they are count-
able or uncountable. Proper names are countable; pronouns depend on their
antecedents. Common noun NPs are countable when they appear with deter-
miners forming a specific or definite NP, e.g. an apple, those apples, 3 apples.
Common nouns are either count or mass. For instance red and flour are mass
nouns at the basic level, while dog and apple are count nouns at that level.
Most nouns can appear as members of both categories. The notions of standard
and marked choice are relevant to the multiple categorization in nouns, as they
are for verb constellations. The analysis of apple as a mass noun and flour as
a count noun is similar to that of the shifted and derived categories of verb
constellations.
As an introductory example, consider the compositional analysis of an Event
sentence that expresses an Activity. In (10), line (i) gives the sentence, line (ii)
its surface structure feature analysis, and line (iii) the compositional rule. The
rule registers the verb features [Atelic, Durative] and the compatible feature of
the complement. VCon = verb constellation; NP = NounPhrase, V = Verb,
PP = prepositional phrase, Loc = location, Dir = directional.
(19) i Mary strolled in the park

ii VCON [[NP] VP [V [stroll] PP [in the park]]]
iii VCON [NP[+Count] V [Atelic, Durative] PP [Loc]] → VCON [e [Activity]]
The output of the rule is a situation entity of the Activity type, as indicated
on line (iii). The term [Activity] represents the complex of temporal features –
[Dynamic], [Atelic], [Durative] – that characterize the Activity situation type.
I give below a few compositional rules that are slightly more general. They
interpret the situation type of a clause by composing the aspectual feature val-
ues of its constituents. The rules specify only relevant parts of the sentence. For
instance, rule A below composes a simple Activity verb constellation with an
atelic verb and a subject or other argument. Sentences with such verbs are atelic
whether or not the subject NP is quantized: “Susan laughed” and “People
laughed” are both atelic. Therefore the rule does not specify a feature for the
subject NP. The feature of the object NP is slightly more complex, due to the
variation that exists among verbs, but I shall ignore the problem here. Examples
indicate the type of case covered by each rule. Abbreviations: Ct = count, NCt =
non-count, Prt = particle, Adv = adverbial, PP = prepositional phrase, Tel =
telic, Atel = atelic, Dur = durative, Cmp = completive, ( ) = an optional
constituent.
Activity:
A. Atelic verb and compatible (atelic) complement:13
[Susan laugh] [Susan push a cart] [Susan stroll in the park]

VCON [[NP] V [Atel, Dur] ([NP]) (PP [Loc]) (Adv [Dur])] → VCON [e [Activity]]
B. Telic verb and atelic particle:
[Peter read in War and Peace] [Susan work on the report]

VCON [[NP] V [Tel, Dur] Prt [Atel] NP (PP [Loc]) (Adv [Dur])] → VCON [e [Activity]]
C. Telic verb and uncountable argument:
[Peter eat apples] [Guests visit that clinic]

VCON [[NP] V [Tel, Dur] (NP [NCt]) (PP [Loc]) (Adv [Dur])] → VCON [e [Activity]]
Accomplishment:
D. Telic verb and countable argument:
[Mary build a sandcastle (in an hour)] [John arise]

VCON [NP [Ct] V [Tel, Dur] (NP [Ct]) (Adv [Cmp])] → VCON : [e [Accomplishment]]
E. Atelic verb and directional complement:
[They walk to school (in an hour)]

VCON [[NP] V [Atel, Dur] (NP [Ct]) PP [Dir] (Adv [Cmp])] →
VCON [e [Accomplishment]]
The feature stated in this output is entered as a condition on the entity [e]
in the DRS (recall that “Accomplishment” and other situation type labels are
shorthand for clusters of features).14
Derived situation types are also interpreted by compositional rule. Recall that
they are triggered by adverbials or other information in the context. The output
13. There is some variation among atelic verbs.The transitive verb in this example, push, allows
a specific NP – “Susan pushed a/the cart”; but other verbs become telic with a specific NP as
direct object, for instance “walk,” as in “Susan walked the dog.”
14. The features characterize the situation entity by associating it with the class of Accomplish-
ments. The account of situation types with compositional rules can be explicated in terms of
semantic processes, as in Krifka (1989).
of the basic-level compositional rules, together with the feature values of the
adverbial (or other form), provide the input to the derived-level rules.
The trigger for a shift in situation type is usually a clash between feature
values. Typically the temporal feature of an adverbial clashes with the corre-
sponding feature in the verb constellation, as in example (3b) (“Mary wrote
a letter for an hour”). The verb write is telic with a quantized object. In this
sentence the verb constellation has the aspectual feature [+Telic], while the
adverb is a simple durative, with the aspectual feature [−Telic]. The adverb
feature value determines the shifted interpretation by a general principle that
I call the External Override Principle. The coercion rule for interpreting this
sentence is sketched in (20). The verb constellation is already interpreted by
a basic-level compositional rule as telic (only the relevant features are stated).
The rule interprets the combination of the adverbial and verb constellation, with
the output an atelic derived verb constellation (DVCON):
(20) Coercion rule (“Mary wrote a letter for an hour”)

S [VCON [e [+Telic]] + Adv [−Telic]] → DVCON [e [Activity]]
The rule shifts the telic verb constellation, an Accomplishment, to an atelic de-
rived verb constellation, an Activity. The feature value of the adverbial overrides
that of the basic-level verb constellation.
The Principle of External Override holds for many derived situation types.
The principle can be stated as an alpha rule, sketched in (21). The input to
the rule is a verb constellation with a situation type value, and an adverbial
or other form. The situation type value is represented by a cluster of temporal
features (a,b,f); feature f has a given value, symbolized as α. The adverbial
also has feature f, but with the value β. The output of the rule is a derived verb
constellation with the feature value of the adverbial:
(21) Coercion: Alpha rule of External Override

VCON [a,b,fα] + Adv [fβ] → DVCON [a,b,fβ]
The output of the rule is the derived situation type value, which will be entered
in the DRS as a condition on the situation entity.
Information in the context of a sentence may trigger coercion, a shift of
situation type. For instance, a generic or habitual sentence may be within the
scope of an operator in a preceding sentence. The time and space adverbials
of the Description mode are similar. They have scope over the clauses of the
description and may trigger situation type shifts. See Chapter 5, 5.3.5.
Situation type information is represented in a DRS with a situation entity
and conditions that characterize it. For instance, (22) presents a partial DRS for
“Mary read a book.” A compositional rule will interpret the situation entity as
an Event. The type of event, Accomplishment in this case, is not included here;
identifying an event is sufficient for Discourse Mode interpretation. Individual
entities x and y represent Mary and a book; the entity e represents the Event.
The entity e is introduced into the top section of the DRS; the conditions on
lines 1–2 specify further.
(22) Partial DRS for “Mary read a book”
x y e
1. e: = read (x,y)
2. e ∈ {Event}
3. x = Mary
4. y = a book
The curly brackets in line 2 indicate that the entity e belongs to the concept of
an Event with the intensional properties of that concept.15 Partial DRSs will be
developed further, with additional information, in later sections of this book.
4.5.2 General Statives: interpreting situation type

The class of General Statives consists of Generalizing and Generic sentences.
Most are derived by coercion from basic-level situation types, except for Generic
sentences with kind-referring predicates such as extinct.
The strongest linguistic correlates for Generalizing sentences are frequency
adverbs (usually, etc.) and the simple present tense. The frequency adverbial
type is interpreted by a variant of rule (20), which provides that a frequency
adverbial shifts any verb constellation to a General Stative. The input to the rule
is a basic-level verb constellation; the output states that the resulting derived
verb constellation expresses a General Stative, notated as St and General:
(23) Generalizing rule (Frequency)

S [[VCON ] + Adv [Frequency]]
→ DVCON [St [General]]
The label [General] stands for the features [Static, Generalizing, Durative]. All
states are durative. This rule ignores the case of used to, a special verb form
that conveys a past pattern of events in English. The form could be added as
an optional feature of the input verb constellation. The rule in (23) does not
15. The temporal properties of a situation entity concept are intensional; they are realized as the
situation unfolds in time. The DRS account of aspectual information follows Smith (1997); it
differs in certain respects from that of Kamp & Reyle (1993).
interpret Generalizing sentences without a frequency adverbial. The possibility

of adding a frequency adverbial cannot be stated here: construction rules deal
only with the actual forms in a surface structure.
The main type of Generalizing sentence has the simple present tense and
an event verb constellation, such as “John walks to school.” These too involve
coercion, a shift of situation type. The coercion rule recognizes a composite of an
event verb constellation, the present tense, and the simple verb form (conveying
the perfective viewpoint). Together these factors license the interpretation of a
sentence as Generalizing, by the Bounded event Constraint. The input to the
rule is a sentence with an interpreted verb constellation, itself the output of
a compositional rule. Schematically, the rule looks like this: it specifies only
that the basic-level verb constellation is eventive, that tense is present, and that
viewpoint is perfective (the simple verb form). Tns = tense; View = viewpoint:
(24) Generalizing rule (simple present)

S [VCON [Event] + Tns [present] + View [Perf]] → DVCON [st [General]]
I do not State rules for interpreting Generic sentences. The interpretation of

generic sentences on syntactic grounds is notoriously difficult; rules would
look for a kind-referring NP and certain verb classes.
4.5.3 Abstract entities in a Discourse Representation Structure

Abstract Entities are interpreted by compositional rules that interpret informa-
tion about predicate class and complement form. Generally, that-clauses and
gerundive complements refer to Facts or Propositions in the context of fact and
propositional predicates, and flexible predicates. This information suffices for
many clear cases, although it is not always enough to distinguish complements
that refer to Abstract Entities from those that refer to Events; some verb classes
and complement forms are indeterminate (4.4.3 above). The form of a clausal
complement may affect the determination of verb class. For instance, think
is a stative verb of propositional attitude with a that-complement, e.g. “John
thought that the earth was flat.” With an about-PP as complement, however,
the verb is dynamic, and allows a gerundive complement (“John thought about
Mary’s winning the race”). Since neither seems more basic than the other, I will
tentatively assume that there are two verbs think.16
16. Whether there are two lexical entries for think in its different senses, or one is more basic than
the other, is a question about the lexicon and compositional rules generally. The difference in
dynamism between the verb constellations NP thinks that S and NP thinks about NP can’t be
attributed to the complements. That-complements appear in some verb constellations that are
dynamic (“John was saying that he will leave early”) and about-PPs appear in some that
are static (“The book is about the war”).
The first rules interpret sentences with verbs that clearly involve reference to
Facts; if they have an object complement, it is a that-clause. The rule specifies
the class of Fact predicates (Fact); the main sub-classes are noted above in
4.3. The complement may have the form of a that-clause or gerund; other
possibilities are ignored here. comp = complement; Ger = gerund.
(25) Fact rule: object complement

S [[NP] + V [Fact] + COMP [thatS/Ger]] → VCON [[e [State]] + COMP [a [Fact]]]
This rule will interpret sentences, “I know that Mary refused the offer,” and “I
resent Mary’s winning the race.”
The main classes of fact predicates with subject complements are evaluative
and psychological verbs. Both classes are flexible, allowing complements of
all types. That-clause and gerundive complements refer to Facts in the context
of these predicates. The rule specifies that the main verb be of the class of
situations, correctly allowing both State and Event main verb constellations;
psychological verbs like surprise may be dynamic. Eval = evaluative verb.
(26) Fact rule: subject complement

S [COMP [thatS/Ger] + V [Eval] + X] → VCON [e [Situation]] + COMP [a [Fact]]
Other subject complement forms are ambiguous: they also have a situation
(Event or State) interpretation with these predicates, as noted above. The sit-
uation interpretation of these predicates is sketched in (27). Substantive com-
plements are NPs, as provided in the rule. The linguistic expressions after the
predicate of interest are unspecified (X). Subst = substantive complement;
Psych = psychological verb.
(27) Fact/Situation rule: subject complement

+ V [Psych, Eval] + X] → VCON [e [State]] + COMP [[e Event/a Fact]]
S [NP [Subst]
In a general treatment of English, we might dispense with the situation inter-

pretation rule. Instead, there might be a default rule providing that the default
interpretation of a complement is reference to a situation.
Rules for interpreting complements that refer to Propositions are similar.
The rule in (28) is written for verbs of mental state (VME ) and communication
(VCOM ). The rule encodes the association of that-clauses and gerundives with
an Abstract Entity interpretation in the context of such verbs.
(28) Proposition rule: object complement

S [[NP] + V [me/com] + COMP [thatS/Ger]] → VCOM [[e [State]] + COMP [a [Prop]]]
Propositional predicates with subject complements are epistemic predicates,

and logical verbs (entail, imply). The rule for interpreting them is like the others
except for the specification of the verb class, here Epist = epistemological verb,
Log = logophoric verb.
(29) Proposition rule: subject complement

S [COMP [thatS/Ger] + V [Epist, Log] + X] → VCOM [[e [State]] + COMP [a [Prop]]
4.5.4 Viewpoint information in the DRS

The contribution of aspectual viewpoint is modeled in a DRS with a temporal
interval [I] and associated times that are introduced with every clause. The
viewpoint morpheme – simple verb form or progressive auxiliary – licenses the
introduction of [I] in the DRS. Associated with each viewpoint is a formal sem-
antic statement which is represented in the DRS interpretation. The viewpoint
of the clause is located at this interval, and the visible information is specified
as a property of the interval.17
The perfective viewpoint makes visible the endpoints of events. This infor-
mation is given by specifying times within the interval [I] at which the endpoints
occur. [I] consists of instants, the first is [ti ], the last is [tj ]. The initial and final
endpoints of the entity e are indicated by I(e) and F(e) respectively. One view-
point condition states that I(e) occurs at [ti ] and F(e) occurs at [tj ]. The other
viewpoint associates the interval with the concept of a perfective viewpoint.
The structure in (30) adds viewpoint information to the partial DRS for “Mary
read a book.” The sentence has the perfective viewpoint. The entities introduced
include [I] and specified times; the conditions on lines 3–6 specify the visible
information, as licensed by the viewpoint information in the sentence.
(30) Partial DRS for “Mary read a book.”
x y e I ti t j
1. e: = read (x,y)
2. e ∈ {Event}
3. {Viewpoint (I,e) = Perfective}
4. ti,j ∈ I
5. ti = I(e), tj = F(e)
6. t ∈ I, t ≥ ti , t ≤ tj
7. x = Mary
8. y = a book
17. This follows the treatment of viewpoint in Smith (1997). In another DRT approach, viewpoint
might be treated as a trigger for a sub-DRS in which all or part of the situation is visible (Patrick
Caudal p.c.).
The viewpoint concept is set off with curly brackets to indicate its special
status.
The progressive viewpoint focuses an interval of an Event that does not
include its endpoints. This is modeled formally by providing that the interval
[I] includes only times after the initial endpoint of e and before the final endpoint
of e.18 The partial DRS for a progressive sentence is given in (31):
(31) Partial DRS for “Mary was reading a book.”
x y e I ti tj
1. e: = read (x,y)
2. e ∈ {Event}
3. {Viewpoint (I,e) = Imperfective}
4. ti,j ∈ I
5. t ∈ I → t > I(e), t < F(e)
6. x = Mary
7. y = a book
These DRSs are still incomplete because they do not include information about
tense.
The aspectual information that is encoded in a DRS identifies a situation
entity for each clause. This information is essential to determining the enti-
ties introduced in text passages. Viewpoint information specifies whether an
Event is bounded or unbounded, which is essential for calculating temporal
advancement.
18. This statement covers the basic progressive meaning which focuses an internal interval of a
situation. It does not account for cases where the progressive focuses an interval before a single
point, as in “They were reaching the top.”
5 Temporal and spatial progression
Time and space are pervasive in human experience and in language; they are
essential in understanding the Discourse Modes. Text progression depends on
both domains. The temporal modes progress with changes in time and space,
while the atemporal modes progress with metaphorical changes of location
through the text domain. This chapter discusses text progression in the temporal
modes; the atemporal modes are considered in Chapter 6.
The principles for temporal progression are essentially pragmatic, depending
on inference about how situations are related to each other and to times. Tense is
interpreted in three different patterns. After discussing progression in the modes,
I provide the linguistic account that underlies the analysis, and implement it in
Discourse Representation Theory. There is a distinct grammatical sub-system
for talking about time in English, consisting of tense and time adverbials. The
system is a closed one: choice of tense and type of adverbial is limited to a
small set of alternatives. In contrast, spatial information is conveyed by lexical
means that do not constitute a grammatical sub-system, except for a very few
deictic adverbials.
Section 5.1 discusses temporal interpretation in the temporal discourse
modes; 5.2 introduces the sub-system of temporal location in English; 5.3 imple-
ments the interpretations in Discourse Representation Structures; 5.4 discusses
space, and how it differs in linguistic expression from the temporal domain; 5.5
gives the forms and linguistic features of the temporal location sub-system.
5.1 Sentences in context: patterns of tense interpretation

The syntactic domain of tense is the clause: tense appears in some form
in every clause of a language with morphological tense. The interpretation
of tense requires information from context, however. Tense is interpreted
differently according to the Discourse Mode of a passage. There are three
patterns of interpretation: Continuity, Anaphora, and Deixis. Recognizing the
92
5.1 Sentences in context 93
different discourse patterns challenges the traditional idea that tense is deictic;
it also provides a needed supplement to accounts that deal only with narrative
advancement.1
In narrative passages bounded events advance narrative time. I will use the
term Reference Time for the advancing time of a narrative: with each bounded
event, Reference Time advances. The notion of Reference Time is explained
in 5.2 below. Narrative has the Continuity pattern of tense interpretation. In
Description, there is a full Anaphora pattern: time is static, the same as a previ-
ously established Reference Time. Tense is interpreted as Deictic in the other
Discourse Modes – Reports, Argument, Information – and in texts generally.
Deictic tense is oriented to Speech Time; this is the default pattern, as in tradi-
tional accounts.
In this section the different patterns will be demonstrated in some detail, and
principles proposed to account for them.
5.1.1 Continuity: Narrative mode

Narratives advance dynamically. After the first sentence, the Events and States
of a narrative are related to previous events and times in the text, rather than
to Speech Time. The narrative fragment in (1) illustrates; it is familiar from
Chapter 1. The events and states are indicated with subscripts E and S for each
clause:
(1) Narrative: Events and States related to each other
1E1 She put on her apron, E2 took a lump of clay from the bin and E3 weighed
off enough for a small vase. 2S1 The clay was wet. 3 Frowning, E4 she cut the
lump in half with a cheese-wire to check for air bubbles, E5 then slammed the
pieces together much harder than usual. 4E6 A fleck of clay spun off and E7 hit
her forehead, just above her right eye.
In the fragment, narrative time advances with perfective event sentences and
fails to advance otherwise. I consider the example in detail.
The narrative of (1) conveys a series of events in sequence, E1 (put on apron),
E2 (took a lump of clay), etc. There is one State, expressed in sentence 2 (clay
was wet), which we understand to hold at the same time as the preceding event,
and perhaps to extend before and after it. The diagram in (2) gives a time line
for the passage; E indicates an event, S a state; SpT is Speech Time:
1. The need for more than one pattern of tense interpretation is recognized in Caenepeel & Moens
(1994) and Caenepeel (1995). They note that some contexts evoke a narrative interpretation,
while others are deictic.
(2) Time line for (1)

...t1.......t2.......t3.......t4.......t5.......t6.......t7.......<SpT
E1 E2 E3 E4 E5 E6 E7
S1
All these situations are temporally located at a time prior to Speech Time,
following narrative convention. Otherwise the information conveyed by
tense is simple continuity. The past tense is not interpreted deictically: if it
were, the events would be related to Speech Time rather than to each other.
Nor do we interpret the past tense as expressing a series of events successively
prior to one another. The continuity function of tense holds for narratives in the
present or future as well as the past.
The interpretation can be expressed in two general principles that hold for
narrative passages. As a narrative advances it establishes a set of different Refer-
ence Times. In clauses that express bounded events, Reference Time advances.
In clauses of states and ongoing events, the time is the previously established
Reference Time. This is an anaphoric interpretation, limited in narrative to cer-
tain types of situations. The two patterns are set out in (3). When tense is past,
as in many narratives, we take it that Reference Time (RT) is prior to Speech
Time (SpT).
(3) Temporal interpretation in narrative

a. Continuity pattern: bounded Events
E1 ..................E2 ..................E3 ...........
RT1 < SpT RT2 > RT1 RT3 > RT2
b. Limited Anaphoric pattern: States, progressives
E3 ......S1 ...............
RT1 RT2 = RT1
Temporal adverbials also advance narrative time, of course. This brief account
does not consider the relative past tenses, past perfect and future-in-past. They
locate situations before and after the current time established in the text. Similar
principles are stated by Kamp & Reyle (1993) in a slightly different approach.
The continuity pattern also occurs in procedural discourse.
Narrative continuity depends on semantic and pragmatic factors. The inter-
pretation of bounded Events is based on semantic, aspectual information: the
class of a verb constellation and the perfective viewpoint. The past-tense mor-
pheme conveys that a situation is prior to Speech Time (in simple sentences;
for other cases see 5.5 below). The sequential narrative interpretation itself is
pragmatic, and can be cancelled by additional information. Situations may be
taken as simultaneous; they may be related by causation or some other factor;

or we may infer a shift in level of detail.2
Another example of narrative is given below:
(4) Narrative Entities, progression
1S1 Michael White, the mayor of Cleveland, is a Democrat, an African Amer-
ican and the son of a union activist. 2S2 And he is at war with his party.
3E1 → First, he backed an end to forced busing. 4aE2 → Then, he supported
Republican Governor George Voinovich’s radical school choice law, bS2 which
offers students vouchers at parochial as well as private school. 5E3 → Then,
he made city workers compete against private firms for garbage collection,
road maintenance and other contracts, prompting union officials to walk out
of a speech [that] he gave at the Democratic national Convention. 6S3 → Now,
he’s allied with the governor again, backing a bill by two Republican State
legislators to grant him control over Cleveland’s destitute school system and
the authority to get “rid of any people who aren’t directly tied to the direct
education of children.” 7S4 Arrayed against him are “the teachers’ union, the
NAACP and just about every elected Democrat in the city of Cleveland.”
The passage consists of bounded events and states; there are no ongoing events.
The bounded events advance time, here supported by adverbials.
5.1.2 Anaphora: Description mode

In descriptive passages time is static, without dynamism. Tense is anaphoric
to a time in the discourse: all the sentences of a given passage have the same
Reference Time. There is a sense of progression in such passages, spatial in
nature. The text advances as the reader goes from one part of the scene to
another.
Usually a locative adverbial appears at the beginning of a description, with
scope over the material that follows; such a phrase appears in both passages
below.
(5) Description
a. 1 In the passenger car every window was propped open with a stick of
kindling wood. 2 A breeze blew through, hot and then cool, fragrant of the
woods and yellow flowers and of the train. 3 The yellow butterflies flew in
2. Shifts in level of detail are common, as in “We went on a long trip last summer. First we traveled
to India.” The second sentence is at a finer level of detail than the first. The inference of shift is
triggered by the adverb “first” and based in world knowledge. There may be other interpretations
of how sentences are related, more abstract and/or rhetorically based. Such interpretation is
usually triggered by additional information in the context. Discourse relations are discussed
briefly in Chapter 11.
at any window, out at any other . . . 4 Overhead a black lamp in which a

circle of flowers had been cut out swung round and round on a chain as the
car rocked from side to side, sending down dainty drifts of kerosene smell.
b. 1 On the big land below the house a man was ploughing and shouting
admonitions to the oxen who dragged the ploughshares squeaking through
the heavy red soil. 2 On the track to the station the loaded wagon with
its team of sixteen oxen creaked and groaned while the leader cracked his
whip that reached to the horns of the leader oxen and yelled on a note only
they understood. 3 On the telephone wires the birds twittered and sang. 4
The wind sang not only in the wires, but through the grasses, and the wires
vibrated and twanged.
To account for the temporal stability of description, I assume a tacit durative

time adverbial that has scope over the entire passage.
The situations expressed in these fragments include states and ongoing events.
They fit the Anaphoric pattern discussed above. There are events presented with
the simple, perfective viewpoint which do not seem to fit, however. For instance,
several events – “a breeze blew,” “butterflies flew in,” “a . . . lamp . . . swung,”
“the car rocked” – appear in sentences 2–4 of example (5a). The events are
all atelic and durative, of the Activity situation type. In this context they are
interpreted as continuous, iterative: the lamp swung round and round, the car
rocked. This interpretation does not disturb the anaphoric pattern, and is pre-
dicted by the formal analysis of perfective Activities. The viewpoint focuses a
bounded unit which need not coincide with the beginning or end of an Activity
(Smith 1999b). In descriptive contexts such as this, Activity events are taken as
continuing.
To reach a better understanding of description, consider (6), a variant of
(5b). I have changed the example slightly so that the new sentence 3 has a telic
verb constellation (“walk to school”). Strikingly, the Description mode is still
undisturbed: the new sentence 3 does not have telic force in this fragment.
(6) 1 On the big land below the house a man was ploughing and shouting admoni-
tions to the oxen who dragged the ploughshares squeaking through the heavy
red soil. 2 On the track to the station the loaded wagon with its team of sixteen
oxen creaked and groaned . . . 3 A group of children walked to school. 4 On
the telephone wires the birds twittered and sang.
In this context the event of walking to school is taken as atelic. This is an

instance of coercion, a shift in situation type: the potential telicity of the Event
verb constellation is overridden.
The coercion effect can be attributed to the tacit time adverbial of description
posited above. Within the scope of a durative time adverbial telic sentences are
coerced to atelic, undergoing a shift of situation type. This is part of the general
phenomenon of coercion discussed in Chapter 4. Recall that sentences with a
telic verb constellation and a durative time adverbial are atelic. For instance, in
“Mary read a book for an hour” there is no sense that Mary completed the book –
on the contrary. The durative time adverbial overrides the telicity of the verb
constellation.
Similarly, “A group of children walked to school” in (6) is atelic under the
scope of the tacit durative adverbial of description. The actual duration of such
an adverbial is determined by context. The assumed duration is the time in
which the perceiver scans the scene and becomes aware of its properties, or
simply a period during which the situation is expected to hold.
In Description, the anaphoric pattern of tense interpretation holds for all
situations. There are coercion effects due to the tacit adverbial of duration: telic
events shift to atelic. Limited and Full anaphoric patterns of interpretation are
pragmatic, depending on the notion that what is conveyed is a static description
of a scene.
5.1.3 Deictic pattern: Report, Information, Argument

The Deictic pattern of tense interpretation is the default: it holds for the Report,
Information, and Argument modes. The deictic pattern is found in discourse
generally, including conversation. Like the other patterns, the deictic interpre-
tation can be cancelled by information that relates situations in a different way.
I illustrate for the three modes. Reports give an account of situations and their
significance from the temporal standpoint of the reporter. Tense is deictic, with
Speech Time the anchor for the deixis. Passages in the Report mode tend to
have deictic adverbials as well as tense; for instance, next year and in March are
such adverbials, anchored to Speech Time. This point is important in identifying
Report passages, see 5.3.2. A passage in this mode is illustrated here; Events
and States are noted as before.
(7) Report: situations related to Speech Time

1S Downtown Austin will have to live for at least several more months with
the half-finished shell of the Intel Corp design center. 2S Intel has postponed a
decision on what to do about the project until sometime next year, S when the
semiconductor company has a better reading on the strength of the economy.
3E Intel decided in March to halt construction on the 10-story center at Fourth
and San Antonio streets, as well as other projects around the country, to save
money during a chip industry downturn. 4E Intel took heat from critics E who
called the half-finished concrete skeleton, encircled by a chain-link fence, an
eyesore.
The passage begins with a future time, indicated by “will” and the for-adverbial;
at S2 the temporal location shifts to the present (“has postponed”), and then to
another future State. S3 shifts to the past event (“decided”), with a past tense
and time adverbial. The events of S4 (“took heat,” “called”) are also located in
the past relative to Speech Time. They need not occur at the same time as the
event of S3. The diagram in (8) gives a time line for this passage in terms of
Reference Time (RT) and Speech Time (SpT).
(8) Time line for (7)

S1 ..................S2 ...................S3 ..................E1 ..................E2 ...
RT1 >SpT RT2 =SpT RT3 >SpT RT4 <SpT RT5 <SpT
This is the typical pattern of Reports.

Passages in the atemporal Argument and Information modes also have this
pattern of tense interpretation. Tense is deictic, and may change frequently.
The types of entities are noted: F = Fact; P = Proposition; GE = General-
izing Stative. Some of the Abstract Entities posited are due to normalization,
following Peterson (1997) (Chapter 4).3
Atemporal modes with deictic tense interpretation:
(9) Argument
1E The press has trumpeted the news F that crude oil prices are three times
higher than they were a year ago. 2GE But it was F the $10 or $11 price of
February 1999, not the one today, that really deserved the headlines.
3GE When inflation is taken into account, GE that 1999 price was the low-
est in modern history, GE while oil has gone above today’s seemingly high
price several times. 4GE And for the past 14 years, at $17.50, oil has been
one of the real bargains of the modern age. 5GE The P low price has been a
mixed blessing. 6GE In the United States, we have lost over 500,000 jobs in
the oil industry while GE we have grossly increased our dependency on for-
eign oil; we now import 55 percent of what we use. 7 With little incentive for
drillers to find and tap new oil, GE supplies eventually dropped, and in the past
year E the Organization of Petroleum Exporting Countries deliberately
dropped its production.
(10) Information
1GE Surprisingly, tunnels are among the most ancient engineering feats. 2GE
The earliest were very likely extensions of prehistoric cave dwellings. 3E The
Babylonians, in the twenty-second century BC, built a masonry tunnel beneath
the Euphrates River that connected the royal palace with a major temple. 4GE
3. For instance, the subject of S5 is interpreted as a Proposition. The subject nominal “the low
price” is normalized as clausal, That the price of oil was low. The clause then meets the criteria
for a Proposition.
The Egyptians, using copper-bladed saws, excavated long passageways and

intricate rooms inside soft-rock cliffs. 5GE The Romans built an elaborate
network of above- and below-ground acqueducts to carry water. 6GE And
they tunneled through solid rock by repeatedly heating it with fire and then
cooling it with water, causing the rock face to fracture. 7S The greatest of
those acqueduct tunnels, which eventually drained Lake Fucino in central
Italy, stretched more than three miles underground.
The general Deictic pattern of past-tense interpretation is given schematically

in (11):
(11) Deictic tense interpretation

E1 ..................E2 ..................S1 ..................S2 .............
RT1 <SpT RT2 <SpT RT3 <SpT RT4 <SpT
In the Deictic pattern situations and deictics are oriented to Speech Time.
Summarizing, three patterns of tense interpretation have been demonstrated
for non-first clauses of text passages. (I assume that the first sentence of a pas-
sage is interpreted by the default deictic principle.) Tense conveys Continuity,
Anaphora, or Deixis. Each Discourse Mode has a different pattern:
(12) Patterns of tense interpretation in the temporal Discourse Modes

Continuity: Non-first clause, bounded Events, Narrative mode
Anaphora: Non-first clause, unbounded Events and States, Narrative mode
Non-first clause, all situations, Description mode
Deictic: Default – all other cases
As a background for implementing the analysis in Discourse Representation

Theory, I give a brief account of the temporal location system of English.
5.2 Introduction to the temporal system of English

Time is a single unbounded dimension, analogous to space though simpler;
see 5.3 below. Like space, time requires an orientation point or landmark for
location. The speaker is the canonical center of linguistic communication so
that the basic orientation points in language are the speaker, the speaker’s place
(here), and the speaker’s time (now). In sentences out of context, situations are
located with respect to Speech Time, which is always the Present. I use capital
letters to refer to time, lower case for tense.
I will briefly introduce the temporal forms and then discuss their interpreta-
tion. “Tense” is the grammatical category of inflectional verb morphemes that
convey information about time. English has two tenses, past and present. The
modal auxiliary will standardly conveys Future time.4 Tense is obligatory in the
main clause of a sentence. There are complex tenses, the perfect (“Mary has
arrived”) and the embedded future (“Mary would leave soon”). Temporal adver-
bials are optional. Locating adverbials specify times (“Mary called at noon”), as
do clausal time adverbials (“John left when Mary arrived”). Adverbials anchor
either to Speech Time or to a time specified in the linguistic context. The discus-
sion below focuses on tense, although adverbials are included in the summary
of the system in 5.5.
The temporal expressions form a closed system, a limited domain from which
one of a few possibilities must be chosen. Tense in independent sentences is
deictic, anchored to Speech Time. Tense locates the situation in a clause by tac-
itly invoking two times: the time of the situation expressed, which I will call Sit-
uation Time; and Speech Time. Present and past tense, and future will, indicate
that Situation Time is respectively simultaneous with, before, or after Speech
Time. In complex sentences the anchor may be a time other than Speech Time.
There is a third time conveyed by tense, a temporal perspective time. Every
tensed sentence takes a temporal perspective or standpoint, known as “Ref-
erence Time,” following Hans Reichenbach (1947). Reichenbach presents
strong arguments for the notion of Reference Time, which I assume here.5
The centrality of the speaker implies an organizing consciousness that provides
a standpoint “from which the speaker invites his audience to consider the Event”
4. The modal will has present tense, correspondingly the modal would has past tense and has a
future-in-past meaning. Would is also a conditional form, not discussed here. Enç (1991) shows
that will is a modal not a tense. Will with strong stress conveys volition or requirement, as in “I
WILL go”; “Bill WILL go.”
5. There are three main arguments, for the notion of Reference Time, two of them due to Reichen-
bach himself. The first involves perfect sentences. Consider the difference between a simple
past and a present perfect sentence. Both present an Event that takes place before Speech Time;
they have the same truth conditions, yet they contrast in conceptual meaning.
(i) a Mary arrived. past
b Mary has arrived. present perfect
In both sentences Situation Time (the Event of arriving) is Past.
The notion of Reference Time gives us a way of understanding such contrasts. In past
tense sentences, both situation and temporal standpoint are in the past; in present perfect
sentences, the situation is past but the temporal standpoint is present. In the perfect, RT is
equal to Speech Time. The perfect has an additional aspectual component of stativity, noted in
Chapter 2.
Another argument for Reference Time comes from the relations between situations. The con-
text of a clause gives information that locates situations relative to one another. They occur
in sequence, or overlap. Using the notion of Reference Time, we can say that overlapping
(Taylor 1977:203). In sentences with the basic tenses, Reference Time is the
same as Situation Time. The reader will recognize the term Reference Time
from the earlier discussion of text progression.
5.2.1 Simple sentences

I posit a Reichenbachian system in which the tenses and future will involve three
times: Speech Time (SpT), Situation Time (SitT), and Reference Time (RT).
Each tense conveys a relation between SpT and RT, and a relation between RT
and SitT. In the simple tenses, SitT and RT are the same. The perfect tenses all
convey that SitT is anterior to RT; in the past-in-future, SitT follows SpT (14c).6
(13) a John is here. RT = SpT; RT = SitT
b John arrived. RT < SpT; RT = SitT
c John will arrive. RT > SpT; RT = SitT
(14) a John has arrived. RT = SpT; RT > SitT

b John had arrived. RT < SpT; RT > SitT
c John will have arrived. RT > SpT; RT > SitT
d (i) John said that (ii) Mary would arrive soon.
(i) RT1 < SpT; RT1 = SitT
(ii) RT2 = RT1 ; RT2 < SitT
situations share Reference Time and those in sequence do not. This accounts nicely for the
difference between the examples in (ii).
(ii) a Mary was phoning the police when John arrived.

a’ Mary entered the room. John was smiling.
b Mary phoned the police when John arrived.
b’ Mary entered the room. John smiled.
Thus the notion of RT provides a locus for relating situations in a principled manner, explicated
in a DR Theory framework in Hinrichs (1986).
A third argument for RT concerns the phenomenon of shifted deixis. As is well known, some
deictic adverbials such as now, in 3 days, etc., which normally anchor to the moment of speech,
can anchor to a Past (or Future time):
(i) Mary sat down at the desk. Now she was ready to start work.
In such contexts the shifted now suggests Mary’s perspective. The notion of RT is the anchoring
point for this perspective. These arguments show that Reference Time is indispensable to an
understanding of the temporal information conveyed in sentences. Reichenbach and others use
the term Event Time for what I call Situation Time. The system as Reichenbach stated it must
be modified and extended (Smith 1980, Comrie 1985, Hornstein 1990, Kamp & Reyle 1993).
I will assume these adjustments. The account given here is similar but not identical to that of
Kamp & Reyle (1993).
6. There are future perfect sentences in which the situation must precede RT but need not precede
SpT, e.g. in “The Prime Minister will have burned the documents by Thursday,” as noted in
connection with (14) above.
The tenses in (14b–d) are “relative” tenses because they depend on another
clause for interpretation: the main clause provides the anchor RT for the
embedded clause. Temporal adverbials, if they appear in a clause, specify
RT.
This relational information is the semantic meaning of a tense, associated
with the tense morpheme in the lexicon. In multi-clause sentences, one clause
may be dependent temporally on another, as in (14d).7 Formally, the tenses
have two semantic features, A and B, which give the relation between times
associated with a given tense. The features provide what Kamp & Reyle call
the “two-dimensional theory” of tense. The A feature relates Reference Time
to Speech Time; the B feature relates Reference Time to Situation Time.
Tense interpretation interacts with aspectual information. Recall that the per-
fective viewpoint focuses events as bounded, while the progressive focuses
events as unbounded. States are also unbounded. The property of boundedness
is crucial for the way aspect relates to temporal location. Bounded events are
totally included in the Situation Time, whether it be a moment or an interval
(e ⊆ SitT); unbounded events and states overlap or surround it (e 0 SitT). For
instance:
(15) a. Lee built a sandcastle.

b. Mary is working.
We understand the event of (15a) as taking place within SitT, here an interval.8
The ongoing event of (15b) overlaps SitT: it holds at an interval that includes
the SitT interval. For concreteness I give a semi-formal statement for the two
sentences.
(16) Lee built a sandcastle.

E: bounded event
RT < SpT, RT = SitT
E ⊆ SitT
7. There are several possible relations between the temporal information of a main clause and other
clauses in a sentence. Tenseless complements and adjuncts share the time of the matrix clause.
Tensed complements may be dependent on a time established in the main clause for either RT
or SitT. In the former case, an adverbial in the dependent clause may specify the time of SitT.
The past perfect and future-in-past would typically appear in a complement clause, as in (13d).
Relative clauses may be temporally dependent or independent of the main clause. See Smith
(1981), Ogihara (1989).
8. By convention, SpT is an instant. I will assume that RT and SitT may be moments or intervals,
depending on adverbial and situation information in the sentence. Both are needed in a semantic
account of the temporal system (Kamp & Reyle 1993:501).
(17) Mary is working.

E: unbounded event
RT = SpT, RT = SitT
E 0 SitT
The construction of DRS representations is discussed in 5.3 below.

The interaction of temporal location and aspect excludes bounded events
located in the Present, at Speech Time. This is an important, non-accidental,
gap in the paradigm. It is due to a pragmatic constraint on bounded events: they
cannot be located at Speech Time. Events in the Present must be presented as
ongoing, e.g “John is talking,” “Mary is drawing a circle,” or as Generalizing
Statives that involve a general pattern, e.g. “Tom often feeds the cat.”
The constraint barring bounded events in the Present results from the
Bounded Event Constraint, a general principle of communication. The ex-
planation for the constraint is at once pragmatic and semantic. In taking the
temporal perspective of the Present, speakers are limited by a tacit convention
that communication is instantaneous. The perspective of the Present time is
incompatible with a bounded event, because the bounds would go beyond that
perspective. As Kamp & Reyle put it:
A present tense sentence describes an eventuality as occurring at the time at
which the sentence is uttered, and thus at a time at which the thought is be-
ing entertained which the sentence expresses. So the thought must conceive
the eventuality as it appears from the perspective of the time at which it is
going on. A sentence which describes something as going on at a time –
in the sense of not having come to an end when that time is up – cannot
represent something as an event. For the event would have to be entirely
included within the location time and thus would not extend beyond it.
(1993:536–7)
Due to this constraint, all simple present tense sentences express unbounded
situations. Almost the same notion is called the “punctuality constraint” in
Giorgi & Pianesi (1997:163).9 There are well-known exceptions, notably per-
formatives (“I christen this ship the Queen Elizabeth”) and sports-announcer
reports (“Now Jones throws the ball to third base”) and narratives entirely set in
the present. Another type, less well-known, appears in literary and many other
9. I thank Nina Hyams for pointing out the similarity of Giorgi & Pianesi’s constraint to the one
stated here. The Bounded Event Constraint is realized differently across languages. In Russian,
for instance, the perfective present conveys Future time reference. In French the present is
imperfective in meaning. The Bounded Event has consequences for temporal interpretation in
languages without tense; see Smith & Erbaugh (2001).
kinds of commentary, including scientific journals (“Here the author creates an

interesting metaphor”).10
5.2.2 Sentences with complement clauses

Complement clauses function as the subject or object arguments of certain
verbs. We might expect that the temporal advancement principle would apply
to complement clauses as in independent sentences. But it does not. Rather than
advancing time, most complements present situations that overlap those of the
main clause, as in (18):
(18) a. Lewis knew that the race was over.
b. Lewis saw that the race was over.
c. Lewis succeeded in winning the race.
d. That Mary won the race surprised Lewis.
In these cases the lower clause does not advance narrative time. Main and
complement clauses have the same Reference Time. There are two other re-
lations between main and complement clause. The complement may receive
its Reference Time from the main clause, or the complement may relate tem-
porally to Speech Time. The determining factors are the tenses and modal
will of main and complement clause, although adverbials also contribute to
interpretation.
When the main-clause verb expresses communication, the complement
clause indicates a previous event or overlapping open situation. The examples
illustrate: in (19a) an ongoing event is expressed in the complement, (19b) is a
variation in which the complement clause has a bounded Event.
(19) a. When I got to Harry’s Waldorf Towers apartment, his secretary said they
were winding up the meeting downstairs. Harry appeared about a half-hour
later, greeted me warmly, went immediately to the telephone.
b. When I got to Harry’s Waldorf Towers apartment, his secretary said Harry
(had) left. Harry appeared about a half-hour later, greeted me warmly, went
immediately to the telephone.
When the clause of communication expresses a bounded event, the event

precedes the communication itself. The reason lies in the Bounded Event Con-
straint. Since communication is taken as instantaneous, a bounded event re-
ported by a verb of communication must precede the communication itself.
(In some dialects, including mine, the past perfect is needed in such contexts.)
10. The literary examples can be seen either as Generalizing Statives, or perhaps as stage directions
that are invoked afresh each time the work is read. I thank Robert Brown for bringing these
examples to my attention.
5.3 Tense interpretation in DR Theory 105
The contribution of direct quotation is somewhat different. Direct quotations

are often presented in narrative as events, events that form a conversation. I
cannot explore the matter here.
Dependent clauses include relative clauses, condensed clauses, adjunct
clauses. Relative clauses also contribute to narrative advancement (Depraeterre
1996); the other types of dependent clauses do not.
Summarizing, the interpretation of tense for single sentences is deictic, ori-
ented to Speech Time. Tense conveys information about the relation between
three times, Speech Time, Situation Time, and Reference Time. The relations
are given by A and B features associated with each tense in the lexicon. Bounded
events and states are included in the Situation Time interval; unbounded events
(and other static situation entities) overlap the Situation Time interval. Com-
plement clauses may be dependent or independent of the main clauses. With
main verbs of communication, bounded events in the complement precede the
time of the main clause.
5.3 Tense interpretation in Discourse Representation Theory

The construction rules of Discourse Representation convert the information
conveyed by tense into entities and conditions which are encoded in the Dis-
course Representation Structure (DRS). Featural information associated with
the temporal morphemes is given in 5.5.
Linguistic context often provides information that is essential for interpreta-
tion. This is true for interpretation of tense: it depends on the Discourse Mode
of the passage. I will assume that a minimal passage consists of two clauses
and that two clauses can establish a Discourse Mode. Additional conventions
are needed for the different modes; they are given below.
Tense interpretation requires assessment of the entities in the linguistic con-
text. Further, we require a way of adapting the information associated with tense
morphemes to the different patterns of interpretation. The class of situation en-
tity and temporal advancement appear in the DRS; there is no specific notation
of Discourse Mode.
The temporal interpretation of a clause has two stages in Discourse Repre-
sentation Theory. At the first stage, construction rules interpret the information
conveyed by temporal morphemes and encode it in the DRSs as temporal enti-
ties and their relations. The second stage consists of calculating the times and
relations in the context of the developing discourse. The calculation must take
into account the entities already present in the DRS, minimally those that are
introduced in the immediate context.
5.3.1 Principles for tense interpretation

The basic, deictic pattern of tense encodes in a straightforward manner the in-
formation associated with tense and modal will. Tense licenses the introduction
into a DRS of three times, t1–3 . They are associated with notional times in the
conditions of the DRS, on the convention that t1 = Speech Time, t2 = Reference
Time, t3 = Situation Time. Each tense has an A and B feature, which states
the relations between the times that are conveyed by that tense. The features
trigger the appropriate conditions on the times. If there is a temporal adverbial,
it appears as a specifying condition in the DRS.
The deictic pattern relates a sentence to SpT according to the A and B features
associated with it. The principle applies to single sentences and as a default to
other sentences, unless another principle applies. The interpretation of tense for
passages in the Narrative and Description modes requires a departure from the
deictic pattern. The three principles for these modes are Continuity, Limited
Anaphora, and Full Anaphora.
The principles of interpretation need information about the entities encoded
in the DRS in the immediate context of a given clause. DRSs do not distin-
guish between sentences, but the information can be recovered. Each clause
introduces a new cluster of times t into the DRS. I will distinguish the times
associated with independent clauses by subscript, for instance t1–3a ; t1–3b ; t1–3c .
All entities introduced by a clause are grouped together with the subscripted
times. Thus two independent clauses will enter three times differentiated as
ta , tb , into the DRS, ensuring that the entities associated with each set of times
will be recoverable. This information will be stated as a contextual constraint on
the relevant construction rule. The constraint will look at the situation entities
introduced in the sentences that precede a clause under interpretation.
The different principles are discussed and stated in the following sections.
I begin with single sentences, which have the deictic pattern of tense interpre-
tation.
5.3.2 Single sentences

The interpretation of single sentences follows the deictic pattern of tense inter-
pretation, summarized in (20).
(20) Deictic Tense Principle

Applies to present or past tense, or modal future will.
For a sentence S involving three times: t1 = SpT.
Feature A: t2 is related to t1 according to the tense;
Feature B: t3 is related to t2 according to the tense.
In this sketch I provide an example of a surface structure, a construction rule,

and a temporal DRS.
The example interprets a sentence and its surface structure; IP=Infl Phrase;
AspP=Aspect Phrase:
(21) Mara put on her apron

IP: tense, perfective; e, x, y
NP I'
Mara I AspP: perfective (simple verb form)
tense Asp'
Asp VP:e
ø
put on her apron
This sentence has past tense, the perfective viewpoint, the participants Mara,
her apron.
Information arises as follows: the IP is associated with a situation entity e,
and with a tense. The situation entity is interpreted by compositional rules. The
tense licenses the introduction of three times in the DRS, and conditions stating
the A and B features associated with it. The AspP is associated with the perfec-
tive or imperfective viewpoint, depending on whether the verb has the simple
form or an auxiliary; the participants appear in the highest NP and the VP, and
percolate to the top of the tree. This discussion focuses on temporal location
information.
The temporal location construction rule for sentences like (21) is given in
(22). The statement covers only simple tenses but could be extended to the
perfect tenses and the future-in-past. I include an optional temporal-locating
adverbial.
(22) Temporal location construction rule: single sentences

S[ . . . [tense/will] . . . , Adv (TLoc) . . . ] → t1 , t2 , t3 ; A, B;
(condition on t2 )
The times are automatically interpreted by general convention as conditions in

the DRS: t1 = SpT, t2 = RT, t3 = SitT. If realized, the adverbial specifies t2
(RT), as in “Mara put on her apron at noon.” The rule gives rise to the DRS in
(23):
(23) Temporal DRS for “Mara put on her apron at noon.”
t1 t2 t3 x y e d
1. e: put on (x,y)
2. t1 = SpT
3. t2 < t1
4. t2 = d
5. t1 = SpT
6. d = at noon
7. e ⊆ t3
8. x = Mara
9. y = her apron
Tense is deictic in this single sentence. The tense features are on lines 3 and
5: the A feature for past tense is t2 < t1 ; the B feature t2 = t3 . Line 7 specifies
that the event is included in the Situation Time interval, since the viewpoint is
perfective. Other information is omitted, for simplicity.
The DRS above illustrates the deictic interpretation of tense, which as the
default applies to single sentences. I now consider temporal information in text
passages. Temporal interpretation requires construction rules and principles for
temporal advancement. I will concentrate on the latter.
5.3.3 Narrative passages

The Discourse Mode of a passage is largely determined by the types of entity
introduced into the Discourse Representation Structure. Narrative introduces
particular events and states into the universe of discourse. However, the Report
mode tends to introduce the same entities, so that it is necessary to have a
way of distinguishing them. The Report mode can be recognized quite reliably
by the presence of deictic adverbials. Report passages typically have spatial
and temporal deictics that are anchored to Speech Time, whereas narrative
passages do not.11 Simplifying, I will assume that narrative passages do not
have deictics. The narrative interpretation of tense arises in the context of at
least one preceding narrative clause.
The Narrative mode of tense interpretation in DR Theory requires a prin-
ciple of Continuity and a principle for Limited Anaphora. The Continuity
11. Shifted deictics sometimes appear in narrative, e.g. “Mary had been working all afternoon; she
was tired now.” The decitic now is anchored not to Speech Time but to the Past time of the first
clause. See Chapter 9. Shifted deictics are not dealt with here.
principle advances narrative time with bounded events, and/or time adverbials.
I’ll work through the principles with the narrative fragment of (1), repeated
here as (24):
(24) Narrative: Events and States related to each other

1aE1 She put on her apron, bE2 took a lump of clay from the bin and cE3 weighed
off enough for a small vase. 2S1 The clay was wet. 3 Frowning, aE4 she cut the
lump in half with a cheese-wire to check for air bubbles, bE5 then slammed
the pieces together much harder than usual. 4aE6 A fleck of clay spun off b
and E7 hit her forehead, just above her right eye.
The principle applies when a clause and its immediate context have event and
state entities, which will be notated as entities of type e.
The first part of the rule for continuity is a contextual constraint that looks
at a clause and the immediately preceding context for the appropriate en-
tities. The times and entities of a clause are associated by letter subscript;
recall that clauses themselves do not appear in the DRS. The relevant en-
tities are of type e, events and states; the entity of the clause being pro-
cessed is subscripted with “n” for times tn . The entities of preceding clauses
are identified as n−2, n−1, and associated with temporal entities tn−2 and
tn−1 in the DRS. The principle applies to non-first sentences of a narrative
passage:
(25) Contextual condition for Continuity principle

In the context of times tn of the DRS, entity of type en ; the situation entities
are en−2 and/or en−1 and no deictics are associated with entity en .
The condition requires that the clause being processed have an entity of type
e; that preceding clauses have entities of the same type; and that there are no
deictics. Thus in (23), clause 1a is first, so the condition is applicable to clause
1b: it is clause n, with an entity of type e; the preceding clause 1a has times
tn−1 and an entity of type e. Advancing in the passage, clause 1c will be clause
n and clauses 1a and 1b are clauses n−2, n−1 respectively. Both have entities
of type e so the rule applies.
The Continuity Principle provides for an advancement of RT with bounded
events. The principle has the form of a rule which advances the narrative from
one Reference Time, RTx , to a later time, RTy , indicated by the symbol >:
(26) Advancement Principle

With RTx immediately preceding; if e is a bounded event:
RTy > RTx
This rule applies to the passage in (24). The first application is to clause 1b:
the rule says that the RT of “[she] took a lump of clay from the bin,” RTy , is
later than the RT of the first clause: RTy > RTx . RTy is the updated Reference
Time. The rule applies in the same fashion to the next clause, “[she] weighed
off enough for a small vase,” and so on.
The full Continuity Principle also has a clause providing that a temporal
adverbial can advance narrative time, stated in (27):
(27) Continuity Principle for non-first sentences

Condition: In the context of times tn of the DRS, entity of type en ; the
situation entities are en−2 and/or en−1 and no deictics are associated with
entity en .
a. With RTx immediately preceding; if e is a bounded event:
RTy > RTx
b. If associated with entities tn there is a temporal locating adverbial that
indicates a time tx later than the time associated with entities tn−1
RTy advances to time tx .
The Limited Anaphora Principle applies under the same conditions as the pre-
vious principle, but only to State and unbounded Event entities:
(28) Limited Anaphora Principle

In the context of temporal entities tn of the DRS, entities of type en ; if the
situation entities associated with tn−2 and/or tn−1 are of type e; and no deictics
are associated with entities tn . With RTx immediately preceding; if e is a state
or unbounded event: RTy = RTx
These principles are stated for clear cases that do not require inference. Adverbs
of the locating type advance narrative time for all situations.
I illustrate the principles in a DRS for a narrative fragment of three sentences;
the first two introduce Event entities into the universe, the third introduces a
State. The times introduced by the sentences are subscripted a–c. The situations
e are numbered 1, 2, 3; e1 and e2 are Events, e3 is the State. The entities appear
at the top of the DRS.
(29) 1. Mara put on her apron.

2. She took out a lump of clay.
3. The clay was wet.
t1a t2a t3a e1 x y t1b t2b t3b e2 z w t1c t2c t3c e3 u
1. t1a = SpT (Principle of Deixis)

2. t2a < t1a
3. t2a = t3a
4. e1 ⊆ t3a
5. x = Mara
6. y = her apron
7. e1 : put on (x,y)
8. t2a < t2b
9. t2b > t2a (Principle of Continuity)
10. t2b = t3b
11. eb ⊆ t3b
12. z=x
13. w = a lump of clay
14. e2 : take out (z,w)
15. t2c < t1c
16. t2c = t2b (Principle of Limited Anaphora)
17. t2c = t3c
18. s 0 t3c
19. u=w
20. e3 : be wet (u)
The individual and situation entities are introduced by compositional rules like
those stated in Chapter 4. The rules provide the conceptual aspectual infor-
mation that classifies the entities as Events and States, and the viewpoint as
perfective in each clause. This information licenses the conditions on lines 4
and 11 that an event is included in t3 , or SitT (perfective, bounded); or on line
18 that it overlaps t3 , or SitT (unbounded simple State). The tense interpretation
is the result of a calculation within the DRS.
This DRS provides two RTs (t2 ) for the three sentences: the first is specified
as prior to SpT on the deictic interpretation, line 2; the second follows the first
by the principle of advancement, line 9. The third RT is equal to the second
by the Principle of Limited Anaphora, line 16. The events are included in the
intervals of t3 (SitT), lines 4 and 11; the State overlaps t3 , line 18. Again, other
aspectual information is omitted.
5.3.4 Description: Full Anaphora

The hallmarks of descriptive passages are an adverbial of location in the first
clause, with scope over the following material; I posit a tacit time adverbial
as well. The entities in a descriptive passage are states, ongoing events (in the
progressive) or atelic events. The time adverbial triggers coercion of telic events
to atelic events, as in Chapter 4. After the appropriate entities are entered into
the DRS, the principle of Full Anaphora applies.
The compositional construction rule for Description applies to the first sen-
tence of a descriptive passage. The sentence has an initial adverbial of space
which locates the scene. The rule constructs a tacit durative time adverbial.
The adverbial introduces an entity mt into the DRS and triggers a condition
identifying it as a durative amount, here unspecified. The entity mt follows
Kamp & Reyle’s account of durative adverbs (1993). The durative adverbial
serves as a coercion trigger for this sentence and subsequent sentences of the
passage.
(30) Construction rule: first sentence of a Description passage
In a clause n with temporal entities tn : the situation entity in n is unbounded
or atelic, of type e:
S[Adv [Loc] + e + tense] →
t1n , t2n , t3n ; A, B; adv, mt
The locating adverbial introduces an entity adv into the DRS and triggers a
condition locating the situation at that place.
Non-first sentences do not require a special construction rule for temporal
interpretation: they introduce times in the usual way. Coercion rules will apply
as usual. They must have access to the unspecified adverbial of duration in
the context. Non-first sentences of Description fall under the Full Anaphora
Principle of tense interpretation because of the entities they introduce and the
DRS context in which they are interpreted. The DRS must contain a space
adverbial and a durative time adverbial without amount specification, which
have the relevant sentence in their scope. I invoke here a general principle of
discourse continuity: between the initial Descriptive sentence and the sentence
in question, the context must not contain a time or space adverbial, or a bounded
event. These would interrupt the continuing scope of the Description passage.12
The Full Anaphora Principle is stated in (31).
(31) Full Anaphora Principle
In the context of temporal entities tn of the DRS, if the situation entities
associated with tn−2 and/or tn−1 are unbounded or atelic, and locating and
temporal conditions obtain:
12. Recognizing changes in continuity, when scope is interrupted, is discussed further in Chapter
7. Stating the possibilities in detail is quite complex: for instance, description typically has
space adverbials that fall within the scope of the initial space adverbial, so that one does not
want to block every space adverbial in descriptive contexts.
If the situation entities of n are unbounded and atelic entities of type e.

RTx immediately preceding: RTy = RTx
The diagram in (32) presents a DRS for the simplified first three sentences
of the Descriptive passage in (5a) above. The passage occurs in a narrative
context. By the principle of limited anaphora the first RT is the same as the
preceding RT. The times introduced with each clause are notated with subscripts
“a”–“c.”
The first clause introduces a state; the next two situations are perfective
Activities. In this context they are bounded, but the bounds do not coincide
with the initial or final endpoints of the events. The notation of inclusion (lines
11, 17) applies to the bounded units focused by the perfective viewpoint. The
duration of the states is greater than or equal to the amount mt of the tacit time
adverbial (lines 4, 12, 18).
(32) a. In the passenger car the windows were propped open.

b. A breeze blew through.
c. The yellow butterflies flew in at the window.
t1a t2a t2a−1 t3a e1 x adv mt t1b t2b t3b e2 y t1c t2c t3c e3 z w
1. t2a = t2a−1 (Principle of Limited Anaphora)

2. t2a = t3a
3. e1 0 t3a
4. dur (e1 ) ≥ mt
5. x = the windows
6. e1 : prop open (x)
7. adv = in the passenger car
8. e1 at adv
9. t2b = t2a (Principle of Full Anaphora)
10. t2b = t3b
11. e2 ⊆ t3a
12. dur (e2 ) ≥ mt
13. y = a breeze
14. e2 : blew through (y)
15. t2c = t2b (Principle of Full Anaphora)
16. t2c = t3c
17. e3 ⊆ t3c
18. dur (e3 ) ≥ mt
19. z = yellow butterflies
20. w = the window
21. e3 : fly in (z,w)
5.3.5 Report: temporal advancement

In Reports, tense is oriented to Speech Time. It is not obvious how to understand
the deictic interpretation of tense for sequences of sentences. One approach
would posit an advancing Speech Time and a series of communications. With
the convention that communication is instantaneous, at each clausal unit the
text would advance by a moment (ignoring embeddings). Advancement would
be linear. There is something right about this, but it pertains to shifts in the
focus of attention during reading, rather than interpretation of sentences in text
passages.
I take a different approach that emphasizes the function of RT: Reference
Time is the locus for advancement. In the following example, I apply the
principle for deictic interpretation to the Report mode. RT may shift back
and forth from Speech Time with each clause, because RT is calculated at
each clause. To see this, consider the Report passage below, repeated from
Chapter 2.
(33) 1 A week that began in violence ended violently here, with bloody
clashes in the West Bank and Gaza and intensified fighting in Southern
Lebanon. 2 Despite the violence, back-channel talks continued in Swe-
den. 3 Israeli, Palestinian and American officials have characterized them
as a serious and constructive dialogue on the process itself and on the
final status issues. 4 News accounts here say that Israel is offering as
much as 90 percent of the West Bank to Palestinians, although it is dif-
ficult to assess what is really happening by the bargaining moves that are
leaked.
S1 has an RT prior to SpT, indicated by the past tense. S2 is also in the past
tense. The event expressed, of continuing talks, is not related to the event of S1
in any direct way. The present perfect tense of S3 shifts RT from the Past to
SpT. The rest of the passage proceeds in the same fashion.
No additional construction rules or principles for calculating advancement
are needed for the Report mode. The deictic principle is applied anew for each
clause. I analyze formally a simplified version of this rather complex fragment.
The simplification does not substantially affect the temporal advancement of
the text.
(34) Simplified Report fragment

a. A week ended violently.
b. Back-channel talks continued in Sweden.
c. Officials characterized them as a serious dialogue.
d. Israel is offering as much as 90 percent of the West Bank.
t1a t2a t3a e1 x t1b t2b t3b e2 y z t1c t2c t3c e3 u v w t1d t2d t3d e4 s t
1. t1a = SpT
2. t2a < t1a
3. t2a = t3a
4. e1 ⊆ t3a
5. x = week
6. e1 : end (x)
7. t1b = SpT
8. t2b < t1b
9. t2b = t3b
10. e2 ⊆ t3b
11. y = back-channel talks
12. z = in Sweden
13. e2 : continue (y)
14. t1c = SpT
15. t2c < t1c
16. t2c = t3c
17. e3 ⊆ t3c
18. w = officials
19. u=y
20. v = serious dialogue
21. e3 : characterize (w,u,v)
22. t1d = SpT
23. t2d = t1d
24. t2d = t3d
25. e3 O t3c
26. s = Israel
27. t = 90 percent of the West Bank
28. e4 : offer (s,t)
Aspectual information is omitted since it does not affect the calculation of

temporal location in the Report mode.
This completes the sketch of temporal advancement in the Discourse Modes.
Description passages advance spatially, as the description covers different parts
of what is described. The spatial domain, as we will see directly, is not a closed
system. Spatial advancement may be conveyed in several ways and can’t be
calculated by rule as temporal advancement can.
5.4 Spatial information in language

Space is like time in some ways: both require an orientation point to locate
an entity in the domain. Linguistically too space and time are closely related.
Linguistic expressions treat time as metaphorical space. For instance, we talk of

the past as “behind” and the future as “ahead”; and say “the coming year,” “the
fast-disappearing year,” etc. (H. Clark 1973, Fillmore 1975). Here and in many
other cases, similar linguistic expressions are used for time and space. There
are temporal and spatial adjectives, adverbs, and propositions; and temporal
and spatial complements to verbs.
Nevertheless space and time are expressed differently in language. Temporal
location is conveyed by tense and a relatively small set of prepositions, adverbs,
and adjectives. Information about spatial location, on the other hand, is con-
veyed by many verbs, adverbials, and adjectives. These differences in linguistic
expression are due to the structure of the domains themselves, and to humans’
knowledge of them.
I will discuss three ways in which the domains differ for humans: directness of
perception and cognition, relative complexity, and relation to volitional action.
Space can be experienced directly through various channels, as Freksa (1992)
notes. Humans perceive space by visual, tactile, and acoustic means; and in-
directly through other senses as well. According to Freksa: “Physical space
plays a central role in cognition as the domain in which physical events take
place and as reference domain for the interpretation of non-spatial concepts.”
Time, however, is not perceived directly by senses. We are just beginning to
understand its role in cognition (Damasio 2002). Time in Western culture is a
human construct that is conceived on the metaphor of space (Friedman 1993).
The domains differ in complexity. Space has three dimensions and therefore
objects can be located in space in more than one way. Using one dimension, we
locate an object with a reference point; using two dimensions, with a reference
line; using three, with a reference plane. English has linguistic expressions for
all three dimensions. For instance the prepositions at, on, in differ precisely
in the dimensions they express. To say “John is on the grass” treats grass as
two-dimensional, while “John is in the grass” treats it as three-dimensional
(Fillmore 1975).
There is an additional level of complexity for space, that of orientation.
We often think of spatial location in terms of a canonical set of orientations
such as “front” and “back.” The human body provides the basic orientations,
as documented in H. Clark (1973). These orientations are often extended to
objects in the world. By extension concrete objects have backs and fronts (e.g.
a house), tops and bottoms (e.g. a box); and locations involve internal or external
relations such as within, on top of , above, below. With all these resources the
location of a given entity, or describing a spatial configuration, may be quite
complex. Think of an Italian landscape painting with a cathedral atop a hill and
rocks spilling down toward the right side. To explain its spatial structure would
involve all three dimensions and more than one orienting perspective.
In contrast, the temporal domain is one-dimensional. Because of its relative
simplicity, the possible temporal relations between entities can be enumerated
and calculated. In linear time any two events have exactly one out of thirteen
qualitatively distinct relationships to each other, ignoring orientation (Allen
1983). If we add orientation, there are twenty-six possible relations between
objects, Freksa (1992) has calculated. Nothing so simple as this is possible for
space.
People engage directly with space by volitional action, whereas their inter-
action with time is often involuntary.13 The role of volition brings out a crucial
difference between the domains. Events involving space are agentive, usually
following an active choice by the agent. One decides to cross the street, hang
a picture, take out the garbage, deliver a message. The passing of time is often
parasitic on facts about particular events: crossing a narrow lane takes less
time than crossing a wide boulevard. Again, maintaining rather than changing
a given state of affairs is often voluntary; the passage of time is concomitant, a
secondary effect of the decision to maintain the state. The contrast is not perfect:
there are a few cases in which time is directly and volitionally involved in an
event, e.g. one may deliberately walk or hold one’s breath for a long or a short
time.
Languages typically have rich verbal resources for talking about space. Spa-
tial information is encoded directly in many verbs. For instance, in verbs like
hang the spatial factor makes an essential contribution. English has a set of many
prepositions which also convey spatial information. There only a few temporal
verbs (pass the time, take an hour, spend an hour, and the aspectual verbs begin,
start, etc.).
Most verbs with a strand of temporal meaning involve projected situations and
attitudes. Classes of verbs with a temporal component include verbs of Future
Having (advance, guarantee, owe, etc.), verbs of Future Situation (propose,
plan, arrange for, etc.); verbs of Wish and Desire (want, covet, need, yearn,
etc.); verbs of Future Events (predict, expect, foretell, etc.); verbs of Future
Prevention (preclude, prevent, etc.); and verbs of Past Events (resent, realize,
etc.). Levin (1993) provides detailed lists.
Strikingly, the main factor for these verbs is modality, not time. They involve
attitudes, plans, or factive meanings, many with a discrepancy between the
current state of affairs and other states of affairs. They are only indirectly
13. I thank Jocelyn Cohan for helpful comments on volition.

temporal. There are few temporal analogues to verbs with spatial meanings,
which are directly concerned with space – traversing it in various ways, direc-
tionality, spatial configuration.
These points may explain an otherwise odd fact about spatial and temporal
phrases and coercion. In coercion, an adverbial triggers a shift in the aspectual
value of a basic-level situation type. For instance, a basic-level Activity clause
denotes a non-specific atelic event which is not explicitly bounded (35a). When
a prepositional phrase is added, the resulting clause denotes a bounded, specific
event (35b–c):
(35) a. They walked. (no explicit bound)
b. They walked for three miles. (bounded)
c. They walked for an hour. (bounded)
In these examples the spatial and temporal PPs function in the same way,
providing an explicit bound to the event. With other prepositions, however, there
is a difference between the contribution of the spatial and temporal adverbials.
Compare (36a–b): the spatial PP adds the component of telicity, the temporal
PP does not:
(36) a. They walked to school. (telic: change of state)
b. They walked until noon. (bounded)
In (a) the event involves arrival at a new location, a change of state. But we
don’t think of (b) this way: we do not conceive of an event of walking-until-noon
as an arrival, or as involving a new state. There are grammatical correlates to
this conceptual difference. Sentences like (36a) pass tests for telicity, whereas
sentences like (36b) do not.14 The grammatical and conceptual differences show
that a spatial bound constitutes a telos whereas a temporal bound does not.
Although every concrete entity has a spatiotemporal location, we have seen
that there are important differences between the domains of time and space.
These differences give an explanation for the fact that aspect and temporal
location can be expressed linguistically through a closed system, while spatial
location cannot.
5.5 Features of temporal expressions

The temporal value of a clause is conveyed by temporal expressions and the
syntax of the clause. The information is based on features associated with each
14. Tests for telicity include felicitous co-occurrence with verbs and adverbs of completion
(in an hour, finish) and ambiguity with almost.
temporal expression. This section discusses the features needed in a systematic

account of English. The features indicate the relational meaning of a form and
the systematic possibilities for its interpretation. To develop the system, we
need to consider several kinds of information.
There are three types of features: Relational, Orientation, and Role features.
Relational features give the value of a temporal expression, based on its lex-
ical meaning. For example, the past tense, the adverb before, and the verbal
auxiliary have, all have the value of anteriority, <; now and the present tense
have the value of simultaneity, =; after and the modal will have the value of
posteriority, >.
Orientation features give the anchoring possibilities for an expression. Tenses
are more limited than other temporal forms. The tenses orient to Present or
Future, but not to a Past time. In simple sentences, both past and present tenses
orient to Speech Time (SpT), as we have seen. There are other interpreta-
tions for tenses in complement clauses. If the main clause has a non-Past time
both tenses orient to that time, as in (37); each has its consistent relational
value:
(37) a. Mary thinks that she is a candidate.

b. Bill says that he won the election.
c. Bill will say that he won the election.
d. Mary will announce that she is a candidate.
However, (38) shows that neither tense orients to a Past time:
(38) a. Mary said that she is a candidate.

b. Mary said that she was a candidate.
The main clause of (38a) is past; the complement clause, with present tense,
orients to SpT, not a time simultaneous with the main clause. In (38b), main
and complement clause both have past tense; the complement clause indicates
the same time as the main clause, not a time anterior to it.15 The sequences
15. The complement clause may have a relative tense, which conveys a SitT other than RT; the
perfect indicates that SitT precedes RT (a), the future-in-past that SitT follows RT (b):
a. John said that Mary had left.
b. Mary said that John would leave.
Neither of these cases advances narrative time.
A very different view of the relation between main and complement clauses is presented
in S. Thompson (2001), which argues that the main clauses of traditional analysis express
speaker stance toward the content of the clause. The discussion focuses on examples from
conversational English.
of tenses in (37–38) show that both tenses have an Orientation value of

Nonpast.
Most adverbials, and have and will, are flexible in orientation. We need an
Orientation feature to account for them. There is also a class of deictic adverbials
which orient only to Speech Time (SpT); they require a feature that specifies
this. The Orientation features are thus Nonpast, Flexible, SpT.
Locating adverbials contribute to the specification of either Reference Time
or Situation Time. Their Role depends on the context in which they appear: the
combination of temporal expressions, and syntactic structure, determine the
interpretation of a clause.
(39) Types of Locating adverbials

a. Rigid deictics: yesterday (<), tomorrow (>)
b. Flexible deictics: three days ago, in three days, next week
c. Flexible: on Tuesday, before Mary left
d. Overt Dependent: three days earlier
e. Anaphoric: then, at that time
f. Calendar: June 19, 1984.
With appropriate adverbials it is possible to construct a sentence which relates

to a Past anchor time, yet indicates a time in the Future.16
Tense and adverbials may contribute to RT or SitT, as the examples
illustrate:
(40) a. Mary called on Tuesday.

b. Mary called yesterday. She was excited.
c. Last week, Keith comes up to me all of a sudden and laughs.
d. Bill said that Keith was leaving in three days.
16. The sentence below illustrates this possibility. Assume that it is uttered on a Tuesday, which is
two days after last Sunday:
Bill confessed last Sunday that he would resign in a week.
The complement clause has the same RT as the matrix, the Past time of “last Sunday”; the
adverbial in a week specifies a time following, a time that is in the future of the RT and of the
time of speech. The interpretation is shown below:
S1 [Bill confessed last Sunday S2 [that he would resign in a week]]

......./.............../..............SpT.........../......................
S1 : RT1 = SitT1 S2 : RT2 = SitT1
last Sunday SitT2 > RT2
SpT: Tuesday
SitT1 : last Sunday, two days earlier
SitT2 : a week from last Sunday = Saturday
In (40a), the tense indicates RT which precedes SpT; in the second sentence of
(40b), the tense continues the RT which is previously established. In (40c) the
adverbial specifies RT prior to SpT, whereas in the complement of (40d) the
adverbial specifies SitT. These interpretations are made by construction rules
using feature information.
Auxiliary have has the relational value of anteriority with the same flexibility.
It may contribute to either RT or SitT: thus have indicates SitT in (41), and RT
in (42).
(41) a. Mary has arrived.

b. Mary had already arrived.
c. Mary will have arrived.
d. auxiliary have: SitT < RT
(42) a. Last Sunday John left the country. On Friday he had closed his bank account,
and had rented his house; at his office, he had emptied all the drawers and
shelves; then he (had) told his staff to take a holiday, and (had) made his
final arrangements.
b. auxiliary have: RT2 < RT1
In a narrative, a fragment like (42a) might be a flashback.

We now discuss Role features in connection with future will, a modal with
the relational value of posteriority. Unlike the other temporal expressions of
English, will has only one role: it always contributes to RT. A feature associated
with will provides for a single Role. This analysis, or something like it, is
necessary within the Reichenbach framework (see Ogihara 1996 for a different
view). Clauses with will always allow auxiliary have, and thus an additional
time; (43) illustrates:
(43) a. Mary will have already arrived.

b. Mary says that in three days she will have finished her project.
c. Mary said last Sunday that in three days she would have finished her
project.
These clauses involve three different times: t1 , RT > t1 , SitT < RT (the latter
indicated by auxiliary have). Note that sentence (43c) involves four times alto-
gether, since the main clause sets a Past orientation time for the complement. A
Role Feature limits the role of will to RT. The proposed features are summarized
in (44).
(44) Features for Temporal Expressions in English

Relational Orientation Role
< = > Flexible SpT Non-Past
present tense + +
past tense + +
aux have + +
modal will + + RT
Locating adv
at – o’clock + +
– ago + +
yesterday + +
The rules of the construction algorithm automatically have access to these

features. They will be listed in the lexical entries for temporal expressions.
(Since no working theory of the lexicon has been given, I do not provide actual
lexical entries.)
Summarizing, this section provides the basis for the interpretation of English
temporal location in Discourse Representation Theory.
This chapter has outlined the temporal location system of English and shown
how tense is interpreted in the different Discourse Modes.

(1) and (23) Peter Robinson, A Necessary End, New York: Avon Books, 1989, p. 182.
(4) Peter Beinart, The pride of the cities. New Republic, June 1997.
(5a) Eudora Welty, Delta Wedding, New York: Harcourt, Brace, 1945, p. 1.
(5b) and (6) Doris Lessing, Under My Skin, New York: Harper Perennial, 1994, p. 116.
(7) Intel delays downtown decision. Austin American-Statesman, December 15, 2001.
(9) Robert Mosbacher, Cheap oil’s tough bargains. New York Times, March 13, 2000.
(10) Jim Collins, How it works. US Airways Magazine, Attaché, May 2001.
(19) Lillian Hellman, Scoundrel Time, Boston: Little, Brown, 1976, p. 72.
(33) A year after victory, Barak fights on many fronts, New York Times, May 20, 2000.
6 Referring expressions in discourse
This chapter discusses the contribution of referring expressions to discourse. I

first consider in detail the notion of Primary Referent. The Primary Referent is
central in a situation, and the key to atemporal text progression. The discussion
will complete the linguistically based characterization of the Discourse Modes.
With the Discourse Mode analysis fully developed, I turn to other aspects of
referring expressions in texts.
Referring expressions convey information by their form. This is possible
because the class of referring expressions in a language belong to a closed
system. There are a limited number of terms, of which one must be chosen.
Choice of a term has contrastive force that may go beyond lexical and referential
meaning. Referring expressions give information about the familiarity status of
the referent, and signal either continuity or change of direction in a discourse.
Within the narrower range of simple and reflexive pronouns, the choice of a
pronoun conveys additional, more subtle, notions. These matters are pragmatic
in nature.
Section 6.1 discusses Primary Referents and how to determine the primary
referring expression in a clause; 6.2 considers referring expressions and closed
systems, focusing on pronoun forms and meanings; 6.3 discusses the familiarity
status of referring expressions, and their patterns of use in texts; 6.4 considers
referring expressions and Discourse Representation Theory.
6.1 Atemporal text progression and Primary Referents

6.1.1 Determining the Primary Referent
Progression in text passages of the atemporal modes proceeds by metaphor-
ical motion. A Primary Referent that is semantically central is identified
for each clause of a passage. Each referent is located metaphorically in
the semantic domain of the text. When the Primary Referents in successive
clauses are in different locations, the receiver has the intuition of metaphorical
motion.
123
The idea of a semantically central referent is based on our intuition of what

is salient and most significant in a situation. For events, the entity that moves
or changes is most significant. For states, it is the entity whose location is
maintained or asserted. These intuitions are cognitively based, having to do with
properties of the situation in the world. Motion, stability, and change of state
are important, and psychologists have shown that human children recognize
them very early (Gelman 1990, Leslie 1994). The notion of Primary Referent is
quite close in spirit to Talmy’s notion of Figure, a cognitive–semantic category
borrowed from Gestalt psychology. The Figure of a motion or location event
sentence is “a moving or conceptually movable point . . . that functions as a
variable in the situation” (1978:628). The notion of Primary Referent is also
close to Gruber’s “pivot of the situation,” which he identified as the Theme
(1965:29).
The Primary Referent is determined by the components of a situation, and
does not depend on the surface form of a sentence. For instance, in both “George
broke the glass” and “The glass was broken by George” the Primary Referent
is “the glass,” though it is the direct object in the active, subject in the passive.
The full range of situations can be conceptualized with a set of primitives,
components that provide a basis for analysis. The components organize events
and states in terms of causal structure, spatial location, and motion. All com-
ponents need not be expressed in every clause. Example (1) is based on Talmy
(1985) and Croft (1991):1
(1) Cause, Object, Motion, Ground, Path, Final State;
Means, Manner, Instrument
In a motion event an Object traverses a Path, perhaps toward a Final State;

the Cause may be explicit or implicit. Certain states locate an Object in terms
of a Ground or other components. These notions can be extended to all do-
mains, following the localist principle that spatial notions are fundamental and
applicable generally. The insight, due to Gruber, is that “the formalism for en-
coding spatial location and motion, suitably abstracted, can be generalized to
other semantic fields” (Jackendoff 1990:25). Motion and location, then, may
be literal or metaphorical.
1. These components are sometimes presented in a way that expresses causal order, as in Croft
(1991:185). The components are also useful in understanding the way linguistic expressions
in different languages express events. Patterns of realization across languages are explored in
Talmy (1985). Croft focuses on linguistic universals and syntactic categories.
There is another, somewhat different approach to the basic components of situations, due to
Hale & Keyser (1993), Erteschik-Shir & Rapoport (in press), and others. These scholars propose
a small set of components which are related directly to underlying semantic–syntactic structure.
Elaborating these ideas, I give criteria for determining the Primary Referent
of clauses that express situations of different kinds. Example (2) deals with
event clauses; states are discussed directly below. The example sentences are
taken from an Argument passage in this study; the Primary Referent expressions
are italicized.
(2) Criteria: Primary Referent of Events

The Primary Referent is that entity in an event which
a. Undergoes a change of state
The high school outsider becomes the more successful adult.
b. Is causally affected by another participant
The national outpouring has forced us to confront the situation.
c. Doesn’t exist independently of the event
High school students present and past have come forward with stories about
cliques and an artificial world.
d. Moves or otherwise changes
Young people mature substantially earlier in the late 20th century than they
did when the high school was invented.
There is some overlap among the criteria: a referent which undergoes a change
of state (a) is often causally affected by another participant (b) and changes
in some way (d). However, there are cases for which all four criteria are
needed.
The domain of states requires more discussion. The Primary Referent in state
sentences can often be determined with the localist principle: the entity whose
location is asserted is the Primary Referent. By the localist principle, all cases
in which a property is ascribed to an entity are locational. Positional sentences
are basic, with x is at y the prototypical state. But the principle is not always
easy to put into practice. When two entities are involved, it may not be clear
which one is the Primary Referent. In other cases the metaphorical extension
is delicate. Additional criteria are needed to supplement the general localist
principle.
Criteria for determining the Primary Referent of states are listed in (3). Fol-
lowing the list is a discussion of how they apply to the main classes of states,
and to special problematic cases. The examples are from texts in this study;
Primary Referent expressions are italicized.
(3) Criteria: State Primary Referent

The Primary Referent is that entity which is
a. Literally or metaphorically located
Dragons are usually arranged almost heraldically round a conceptual
center point.
b. Dependent on the situation for existence

The predominant output was the white ware with transparent ivory toned
glaze which made the kilns famous.
c. Figure relative to a Ground
A group of kilns is northeast of Ch’ang-an, the capital city of the T’ang
dynasty
d. Has a property ascribed to it
The most important kilns are those at Tao-chu in Shensi.
More specific criteria are applied before the general. Criterion (d), for instance,
holds of the subject in any state sentence.
I distinguish three classes of state: states of location, including position and
possession; mental states of thought, belief, feeling, perception, communica-
tion; states of composition and identification, and property ascription. The cri-
terion of literal and metaphorical location nicely applies to the first two classes.
Consider first states of location. With positional states, the located entity is the
Primary Referent (underlined): for instance in “The lamppost is on the corner,”
the Primary Referent is “the lamppost.” For states of possession, the possessed
entity or concept is the Primary Referent and its metaphorical location is the pos-
sessor, as in “John has a book,” “Mary has an idea.” Mental states, the second
class, fall under the rubric of extended possession. The contents of mental states
are possessed by the person who holds them, as in “John likes strawberries,”
“Mary thinks that Lee is here.”
Compositional and identificational states can be analyzed with the criteria of
dependence and possession. For example:
(4) a. We are truly a spoiled society.

b. Oil has been one of the real bargains of the modern age.
c. The committee consists of eight people.
In (4a–b) the concepts denoted by the predicative NPs are the Primary Referents;
they depend for their existence on the subject NPs.2 In (4c) the referent of “eight
people” is the primary, by possession.
The third criterion applies in location sentences when one entity is located
in relation to another, known as Figure–Ground sentences. Consider (5):
2. The NPs in these examples are predicative: the sentences have only one referent. They can be
conjoined with adjectives: “Mary considers John competent in semantics and an authority on
unicorns” (Partee 1987:119). In uncertain cases, the possibility of conjunction with an adjectival
predicate is a good test for the function of an NP. By this criterion the post-copula NPs in (3b–c)
are predicative. The subject NP of (3d) is also predicative: there is only one referent in the
sentence, the kilns at Tao-chu in Shensi. The formulation of this criterion is similar to a criterion
for Theme role in Dowty (1991).
(5) a. Harry is near John.

b. Harry resembles Aunt Mabel.
c. The bike is near the house.
d. The house is near the mountain.
The sentences seem symmetrical: if John is near Harry, then surely Harry is
near John. But in fact one entity functions as the Ground for locating the other,
the Figure. To show this we look at different types of cases, following Talmy
(2000). For (5 a–b) either NP is plausible as the subject: “John is near Harry,”
and “Aunt Mabel resembles Harry” are good sentences. Often, however, the
alternative realization is semantically peculiar. The alternatives of (5c–d) are
peculiar (indicated by #):
(6) a. # The house is near the bike.
b. # The mountain is near the house.
Pairs like this show that factors other than location play a role in what makes
an acceptable Figure. (6a–b) are odd because a bike is not a plausible reference
point for locating a house, nor is a house plausible as a reference point for
locating a mountain. Although the relation between the referents is apparently
symmetric, one is usually more appropriate than the other as Ground.
The relative size and stability of the referents determine what is appropriate as
Figure and Ground. These are points about the world, not about the grammar.3
In Figure–Ground sentences the subject phrase gives the Primary Referent. An
alternative structure is inversion (“Near the house is a bike”). Nothing quite
like the active–passive alternation exists for sentences with the copula. The
non-canonical existential there construction has the same constraint, strikingly.
For instance, “There is a bike near the house” is good, whereas # “There is a
house near the bike” is odd indeed. Inversions and existential there sentences
are discussed in Chapter 9. I think that the evidence is strong that the notion
of Figure is semantically based in these cases (pace Dowty 1991). Yet, since it
appears in subject position, the Figure conflates semantic event structure and
pragmatic presentation.
3. Talmy lists the main factors: the Ground is usually (i) more permanently located; (ii) larger;
(iii) geometrically more complex; (iv) more familiar/expected; (v) of lesser concern or rele-
vance; (vi) more immediately perceivable; (vii) more backgrounded once Figure is perceived;
(viii) more independent (Talmy 2000: I, 315).
Almost identical criteria are proposed by Polinsky in a discussion of Figure and Ground in
Tsez, a language of the Caucasus, that belongs to the Nakh-Daghestanian family (1996). She
argues that Figure and Ground involve the interface between conceptual structure and syntactic–
semantic structure. The similarities suggest that the Figure–Ground distinction may be coded
similarly in many languages.
Certain State sentences do not have a semantically Primary Referent. There

are two types: sentences with symmetric predicates such as similar (7a), and
sentences that literally identify a referent (7b–c):
(7) a. Sam is similar to Lee.

b. Clark Kent is Superman.
c. Cicero is Tully.
These appear to be truly symmetrical: neither referent is primary. In such sen-

tences the subject position involves a particular perspective, as Talmy points
out. In these, unlike the Figure–Ground cases, either referent can appear as
subject.
Equational clauses and clauses with symmetric predicates have no referent
that is primary in the semantic sense. In such cases the subject referent functions
as Figure, a referent that is presentationally primary.
The notion of Primary Referent corresponds closely to the concept of Proto-
patient role proposed by Dowty (1991). The Proto-patient thematic Role is like
the traditional thematic Role of Theme, augmented with the Role of Incremen-
tal Theme.4 The syntactic arguments that realize this Role are patients rather
than actors, effects rather than causes. In transitive sentences, the direct object
usually expresses this Role; in sentences with the copula, the subject expresses
it. Dowty takes a prototype approach to the notions of Agent and Patient/Theme.
He assembles a cluster of properties which are characteristic of each. In a given
sentence, the Proto-patient is the argument that has the most Patient properties;
the Agent argument has the most Agent properties. Dowty’s approach is se-
mantic, based on the structure of events and states rather than surface structure.
His discussion focuses on “the Argument-selection problem,” seeking princi-
ples that language uses to determine, for each argument of an n-place relation
denoted by a predicate, which argument is expressed by which grammatical
relation.
The Primary Referent of a clause is usually also its Proto-patient argument.
I will use this fact in writing construction rules for determining the Primary
Referent for Discourse Representation Theory. I assume that information about
Proto-patient, or Theme, is available in the underlying syntactic structure of
a sentence, and in surface structure. In principle then, most clauses have a
4. The Incremental Theme argument of a talic predicate participates in a homomorphican from

the structured theme argument denotations to the structured event (Dowty 1991, Krifka
1989).
Primary Referent. The criteria given have been used successfully to identify
Primary Referents in an informal study.5
There is a syntactic notion of prominent argument that may seem related to
Primary Referent. It is not: the two are quite distinct. The “prominent argu-
ment” relates syntactic structure to argument structure. For instance, Grimshaw
(1990) claims that the subject is prominent syntactically because of its position
as an external argument. The Agent and Cause roles that are typically realized
by subjects are also prominent, according to Grimshaw, in argument struc-
ture and event structure respectively. The subject is also regarded as prominent
in Speas (1990), who makes a different proposal. For Speas, thematic roles
originate in positions in lexical conceptual structure. The syntactic realization
corresponds to a thematic hierarchy, and follows the Universal Thematic As-
sociation Hypothesis (UTAH) of Baker (1988). Prominence in the thematic
hierarchy corresponds to syntactic prominence, Speas suggests; Agents are as-
sociated with the prominent subject position. In contrast, the Primary Referent
proposed here is associated with the outcome or motion of a situation, not the
agent/cause.
In determining the Primary Referents in a text I proceed on a clause-by-
clause basis. For clauses with a sentential complement, the Primary Referent is
decided as usual. If the event or state expressed by the complement is the Primary
Referent, the internal Primary Referent is noted as another level of structure. In
complex sentences each clause is treated separately, see Chapter 9.
6.1.2 Primary Referents in text fragments

I now give a sample analysis of the clauses of an atemporal text passage in
terms of Primary Referents. I then state construction rules for determining the
Primary Referent of a clause, using the notion of Proto-patient/Theme.
The passage in (8) is a fragment in the Information mode. The Primary
Referents are underlined for all tensed clauses, and a brief explanation of each
choice given in (9). Clauses without an underlined phrase have no decidable
Primary Referent. The passage appears at the beginning of a magazine article:
(8) 1 The largest and most complex highway-engineering project in American

history has been making headlines over the past several months. 2 Boston’s
central artery project, known as the Big Dig, has become nightly news fare,
5. An informal study ascertained that the criteria can be applied. The members of a seminar on
this material were given two texts and asked to identify the Primary Referent of each clause.
They agreed in most cases.
thanks to the massive cost overruns and steadily growing complaints of mis-
management and corruption. 3 Overshadowed by scandal, the work itself –
whose scale and ambition are truly mind-boggling – continues almost as a
matter of course.
4 The heart of the project is an eight-to-ten-lane highwayi . 5 Iti is be-
ing built below ground to relieve the legendary traffic pressure caused by
Boston’s two major arteriesi , whichi converge downtown awkwardly near
the harbor, skyscrapers, and historic buildings. 6 Threading 35 lane-miles
of tunnel through Boston’s 3-D maze of subway lines and building founda-
tions, all without drastically disturbing the normal routines of surface life, is
an engineering planj thatj borders on the fantastic. 7 In some places the tunnels
are 120 feet deep. 8 The first two frequently asked questions on the project’s
official Web site are “What are you building?” and “Are you nuts?”
9 Surprisingly, tunnels are among the most ancient engineering feats.
The first two sentences have “headlines” and “news fare” as Primary Referents:
there is little change in metaphorical location from one to the other. At S3
the Primary Referent is “the work itself,” at Sentences 4 and 5 “the highway,”
and, at the relative clause, the relative pronoun referring to “Boston’s major
arteries.” At S6 the fantastic nature of the plan is primary, at S7 “the tunnels” is
Primary Referent; S8 has no distinguishable Primary Referent; at S9 “ancient
engineering feats” is primary. This sequence corresponds quite well to one’s
intuition of progression in the text, I think.
The reason for each choice of Primary Referent is listed very briefly in (9).
For sentences with more than one clause, the clauses are lettered a, b, etc.:
(9) Justification of choices of Primary Referents in (8)
S1) referent dependent on the situation for existence
S3) referent that moves/changes in the Event
S5) a, subject NP, referent come into being
b, relative clause, referent located
S7) referent has a property ascribed to it
S8) equational, no single Primary Referent
Although in S8 there is no semantically Primary Referent, the subject is present-

ationally primary; see Part III for discussion.
6.1.3 Primary Referents and Discourse Representation Theory

Primary Referents can be identified by construction rules on two assump-
tions. The first is that the Primary Referent of a clause coincides with the
Theme/Patient Role. I also assume that the argument realizing this role can
be identified by semantic and syntactic factors that relate event structure and
argument structure. I will not try to state them here, so that this construc-
tion rule is something of a promissory note. The enterprise, however, does not
seem hopeless. There is general agreement as to which argument realizes the
Theme/Patient role, although people do not agree on how to state principles of
argument selection. By hypothesis, then, an argument in the surface structure of
a clause has a semantic–syntactic feature that codes Patient/Theme information
from Event structure.
The construction rule identifies the argument with the Theme/Patient feature
(T/P) as Primary Referent. The information about Primary Referent will appear
as a condition on the entity in the Discourse Representation Structure. I state
the rule below and illustrate for the first sentence of (8), simplified:
(10) Primary Referents in the DRS
a. Construction Rule
[. . . a [+T/P]. . . . . . .] → a = Primary Referent
The rule applies to the sentence “The project made headlines over the past
months,” as shown below. The NP “headlines” is the Primary Referent since it
is the direct object of a transitive change-of-state verb (notated as Vt for telic),
and results in the DRS shown in (11b). In (11a), the input to the rule is a surface
structure with an NP subject, a transitive verb (Vt ), as NP object specified as
[+T/P]. “X” and “Y” indicate irrelevant material.
(11) a. IP [NPVP X[Vt NP [+T/P] Y]] → NP = PriRef
b. The project made headlines over the past months
t1 t2 t3 x y e d
1. e: made (x,y)
2. t1 = Sp T
3. t2 < t1
4. d = over the past months
5. t2 = d
6. t2 = t3
7. e ⊆ t3
8. x = the project
9. y = headlines
10. y = Primary Referent
The sentence is first in the DRS, with the deictic interpretation of the past
tense that locates the situation before SpT. The adverb specifies t2 , RT.
The event is perfective in this version of the sentence and included in t3 ,
SitT.
The Primary Referent is the first step in determining atemporal text progres-
sion. The next step, more daunting, requires additional pragmatic calculation.
Recall that atemporal text progression occurs when metaphorical motion has
taken place. I suggested that metaphorical motion proceeds on analogy with
space, through the semantic terrain of a text. We know from the discussion in
Chapter 5 that spatial information is not conveyed by a grammatically closed
system. Morever, often the metaphorical motion in question will be only indi-
rectly spatial. To fully understand these matters we needed a way of accessing
and organizing many kinds of general information. We do not have a formal
account of how such calculations might be made at present. Perhaps the wide-
ranging approach sketched in Asher & Lascarides (1998) would eventually be
able to handle them.
The factors that determine text progression of atemporal text passages are
not grammatically signalled. Therefore the presentational aspect of such texts,
and the discourse relations that organize them at a more abstract level, take on
more importance. In the Information and Argument modes, discourse relations
may be an important part of atemporal text progression; see Chapter 11 for
some discussion.
6.2 Referring expressions

The next sections of this chapter are concerned with referring expressions as
members of closed systems, and the kind of information that they convey.
I discuss the choices available in English, and the significance of a particular
choice. There are a number of possibilities for referring to people or other
entities: proper names, noun phrases with lexical content, and pronouns of
different types. I will be interested here in contrasts between the grammatical
forms, e.g. pronouns and lexical NPs, ignoring lexical content.
6.2.1 Closed systems

Closed systems in language consist of a relatively small set of forms that contrast
with each other. Besides the semantic information that the forms code directly,
they have a systematic meaning: choice of form A conveys that form B is not
appropriate, or less so. Patterns of one form rather than another can convey
continuity in a text.6
6. Closed systems were extensively studied by linguists of the Prague School. Roman Jakobson,
for instance, discussed the meanings of their terms as conveying positive and negative values,
symmetry, and contrastive information (1957). I have proposed a somewhat different approach
Referring expressions are closed systems, and pronouns are an important sub-
system. There is a very limited set of possibilities in English. Some findings
from other languages are pertinent to understanding the uses of English pro-
nouns. As a background for the discussion I make brief comments about noun
classifiers in Mandarin, proximate–obviative pronoun systems, and switch ref-
erence systems.
The forms of closed systems are often obligatory at the level of the sentence,
so that the grammar would predict that they appear in all relevant syntac-
tic contexts. It turns out, however, that extra-grammatical factors affect their
appearance.
In a study of noun classifiers in Mandarin Chinese, Erbaugh (1986) makes
this point quite dramatically. Classifiers (CLs) are a closed set of morphemes.
They appear in principle in NPs when a noun is preceded by a numeral or a
demonstrative, e.g. yiwei lao xiansheng (an old CL gentleman), yi pi ma (one
CL horse); there are two types, general and sortal classifiers. Erbaugh’s study
of classifiers in spoken Mandarin showed that sortal classifiers did not always
appear in such contexts.
Sortal classifiers were not predictable on a grammatical basis, Erbaugh found:
their use depended on genre, familiarity, and discourse structure.7 In relatively
formal speech situations and narratives, classifiers were most frequent. They
tended to appear in NPs that introduced new scenes or characters at event
boundaries. In casual conversation they introduced new topics or new concrete
objects, but they did not occur often. Erbaugh’s work shows these are used in
discourse according to extra-grammatical factors, although they are in principle
obligatory at the sentence level.
Similarly, certain types of pronouns are used to convey discourse meanings.
Proximal–obviative systems have contrasting pronominal morphemes with a
referential meaning. They signal whether a subject NP is coreferential with
a nearby antecedent. The proximate morpheme indicates coreference with a
nearby NP; the other obviates this reference, indicating as antecedent a less
to closed systems which distinguishes between semantic and pragmatic information (C. S.
Smith 1991). Semantic information is associated with a linguistic form in the grammar of a
language, while contrastive information about that form is part of the pragmatic knowledge of the
speaker.
7. Erbaugh showed films and asked subjects to narrate the story, using the “Pear Stories”. She also
collected examples of informal narratives, conversations, letters. There are additional subtleties
in the use of classifiers: for instance, Chu notes that the general classifier can be used instead of
“a specific one to express a casual attitude” (1983:17).
The Sino-Tibetan languages, including Mandarin, tend to have measure classifiers and special
classifiers. Mandarin also has a general classifier ge which may appear with any noun at all.
proximate NP.8 The NPs are usually third person and animate. Systems of this
kind are found in Algonquian languages such as Fox, Plains Cree, Blackfoot.
The notions proximate and obviative have discourse interpretations. The no-
tion “proximate” is often metaphorical: proximate forms tend to be used for
characters of importance in the discourse. There are several patterns.9 In nar-
rative with a consistent central character, the proximate form is often used to
refer to that character throughout. Proximate forms are also used at the boundary
of a discourse unit, for instance the beginning of a new episode. More generally,
“proximate” is predictable, neutral, unmarked. “Obviative” is marked and con-
veys a referent that is special in some way. The extended, metaphorical distinc-
tion between proximal and obviative explains many uses of English stressed
and unstressed pronouns, and simple and reflexive pronouns as well.
Another type of system, known as switch reference, has a pronominal mor-
pheme on the verb which indicates whether the subjects of main and dependent
clauses are the same or different. The Same Subject morpheme (SS) in principle
indicates that the subjects are the same; the Different Subject (DS) indicates that
they are not. Switch reference morphemes are obligatory at the level of sentence
grammar, like noun classifiers. Also like noun classifiers, the morphemes do
not appear in every relevant context, and they usually have discourse meanings
beyond the referential. Switch reference morphemes are primarily used to con-
vey information about temporal and other relations between the situations in
the main and dependent clauses. Languages with switch reference systems in-
clude West African languages, American Indian languages, Papua New Guinea
languages.
The main use of switch reference is to indicate semantic relations between
situations. Often switch reference morphemes function as temporal connec-
tives. In temporal sentences the SS morpheme conveys that the situations in
dependent and matrix clauses are simultaneous, the DS morpheme that they are
sequential.10 These morphemes may also indicate logical or epistemic relations
8. Obviation refers to a form indicating the less salient of two third persons “to obviate con-
fusion,” according to Voegelin & Voegelin (1975:385). Hopi, Algonquian languages, Hokan
languages, Yup’ik Eskimo, and many others have contrasting forms which indicate coreference
and non-coreference between NPs in certain syntactic environments (Jacobsen 1967, Voegelin
& Voegelin 1975). The term “obviation” was proposed by Cuoq.
9. See for instance Frantz (1966) on Blackfoot, Wolfart (1973) on Plains Cree – all Algonquian
languages. Goddard (1990) and Thomason (1994) discuss Fox. Both Frantz and Wolfart note
that the proximate form has the pragmatic use of indicating a referent which is familiar, the
current topic of the discourse: currently in focus, in the sense of discourse focus.
10. When the referential and non-referential meanings of switch reference morphemes clash, the
non-referential meanings override. This leads to what Stirling calls “aberrant” uses of the
between clauses, such as causality or reason. Discussing Amele, a language of

Papua New Guinea, Roberts (1988) observes that switch reference is used to
indicate changes in all the deictic features of discourse: speaker/hearer, world,
time, and place.
Discourse considerations are thus the key to the distribution of switch ref-
erence morphemes. Though in principle they keep track of reference, their
primary function is to indicate closeness or distance of situations or referents
according to temporal and causal factors. Null pronouns in English have some
of these same functions; see (18)–(19) below.
6.2.2 Pronouns
Across languages pronouns take many forms: there are null pronouns, clitic
pronouns, strong pronouns, stressed and unstressed pronouns, reflexive pro-
nouns, possessive pronouns. Some languages have special pronoun forms with
specific functions. For instance, there are pronouns that convey reference to an
antecedent subject of consciousness, known as “logophoric.” Not all types are
available in all languages. If a language doesn’t have a pronoun form with
a dedicated function, other forms may be pressed into service to fulfill it.
There are no specifically logophoric pronouns in English, yet the subject-of-
consciousness meaning may be conveyed with the English reflexive pronoun. In
what follows I will consider stressed and unstressed pronouns, reflexives, null
morphemes. In aberrant DS cases, the morpheme appears when clauses differ in time, place,
or some other situation parameter, even though they have coreferential agentive subjects. The
examples below illustrate for time. The clauses have the same subject in all three examples. The
SS morpheme is grammatical in (a) but ungrammatical in (c); the DS morpheme is grammatical
in (b).
(a) Magau t-r-va (kani) m-augIn

Magau Fut-sg-come (and) SS-eat
Magau will come and eat
(b) Magau r-n-va (kani) t-r-augIn
Magau 3sg-Pf-come and Fut-3sg-eat (zero DS)
Magau has come and will eat (later)
(c) ∗ Magau r-n-va (kani) t-m-augIn
Magau 3sg-Pf-come and Fut-SS-eat
Example (a) is a typical case of Switch Reference: the second clause has the SS morpheme
and is not differentiated in tense from the first clause. In (b) the clauses have different tenses
and the second clause has the DS morpheme. The ungrammatical (c) is like (b) except that
the second clause has the SS morpheme. What this contrast shows is that having coreferential
subjects is not a sufficient condition for the SS marker in Lenakel: the clauses must have the
same tense as well. Stirling interprets this and similar examples as showing that the SS marker
is primarily used to convey that Events are closely related.
pronouns, and possessive pronouns. Reciprocal and relative pronouns are not
discussed.
When used felicitously, pronouns refer to entities that are already known or
established in a discourse. The antecedent for a pronoun is mainly determined
by pragmatic factors.
6.2.2.1 Stressed and unstressed pronouns

Pronouns refer to an entity in a discourse, or in the immediate context; in the
former case they depend on an antecedent for interpretation. There are some
grammatical constraints on the possible linguistic antecedents for a pronoun. I
will assume the constraints here. I deal only with referential pronouns.11 The
antecedent to a pronoun may be in the same sentence or farther away, within
the constraints of grammaticality and agreement.
The unstressed pronouns of English indicate neutral coreference with a rela-
tively salient antecedent entity. Stressed pronouns convey additional meaning.
The contrast is similar to that between proximate and obviative pronouns. How-
ever, in English the obviative, non-neutral meaning is emphatic or contrastive.
The examples illustrate for sentences with two clauses, each containing an NP
and a pronoun. In (11a–b) the pronoun is unstressed, and understood as coref-
erential with the NP in the main clause. The pronoun in (11c) has strong stress
(indicated by capitals) and suggests a referent other than the NP. Subscripts
give the preferred interpretations:
(11) a. When hei works, Johni doesn’t drink.
b. When Johni works, hei doesn’t drink.
c. When HEj works, Johni doesn’t drink.
The referent of the stressed pronoun in (11c) is presented in contrast with

“John,” the nearest possible antecedent.12 Emphasis and focus are discussed in
Chapter 8.
11. I assume the grammatical constraints on possible pronoun reference as stated in Government
Binding Theory. The main constraint is that a pronoun may not be coreferential with an an-
tecedent that is higher in the structure of the same clause, a c-commanding antecedent (Chomsky
1981).
Non-referring uses of pronouns include quantifier contexts in which pronouns are variable
rather than referring expressions (“Everyone that has a donkey likes it”). These sentences are
known as “donkey sentences.”
12. One can construct examples which have a sentence-internal referent for the stressed pronoun:
Although [hisj colleagues]i think that theyi understand Johnj , HEj doesn’t think that
theyi do.
The pronoun in the main clause must be stressed, because the NP “his colleagues” sets up a
contrast set suggesting the focal interpretation of the pronoun. The example shows that the
Context decides whether a given antecedent is plausible for a pronoun. In

certain contexts, the plausible antecedent appears in a different sentence. For
instance, (12) supplies a context for (11c), in which the pronoun is taken as
coreferential with an antecedent in the preceding sentence:
(12) a. My parents are coming to spend a week with us next month.
b. Dadi is a strict teetotaler; hei doesn’t like to be around alcohol.
c. When hei visits, Johnj doesn’t drink.
In this context the antecedent for “he” in (12c) is “Dad,” the referent of the
pronoun in the preceding sentence. Context may also affect the interpretation
of a stressed pronoun. In the context of (13a) the pronoun in (13b) is taken as
coreferential with “John”:
(13) a. Johni insists that Tomj and Billk abstain from drinking on the job.
b. After all, when HEi works, Johni doesn’t drink.
Here “John” is contrasted with the proximate antecedents in the preceding

sentence. This account does not deal with those contexts in which pronouns are
inaccessible for coreference.
The unstressed pronoun in English indicates neutral coreference with an ante-
cedent. The referent may be nearby, literally proximate; the referent may also
be familiar or predictable, metaphorically proximate. The stressed pronoun is
obviative, indicating contrast or emphasis, with a referent that is not necessarily
proximate. Pronoun antecedents are also discussed in 6.3.2 below.
6.2.2.2 Reflexive pronouns

Reflexive pronouns indicate coreference between two arguments of a verb, as
in “Mary washed herself,” “The doctor asked John about himself.” Reflexives
in English are grammatically constrained. In the Binding Theory the constraint
is a locality condition, requiring that a reflexive have a local c-commanding
antecedent (Chomsky 1981). However, there are many reflexives that do not
obey this condition. They either have an antecedent in the same sentence that
does not c-command, or an antecedent in a sentence that is quite far away. I will
call such reflexives “Locally Free Reflexives” following C. L. Baker (1995);
they are also known as “long distance reflexives.”
Locally Free Reflexives (LFRs) may convey that their antecedent is a subject
of consciousness, in addition to coreference. This is a logophoric meaning. The
term was coined by Claude Hagège for pronouns “which refer to the author
information for focal reference can be presented in more than one way; I shall not explore the
possibilities here.
of a discourse or participant whose thoughts are reported” (1974).13 Reflexive

pronouns with a logophoric function appear with verbs of communication and
mental state when the subject of main and complement clauses is the same. They
attribute the contents of the complement clause to the speaker or another an-
tecedent. Pronouns indicating an internal self and/or responsibility for thought
and communication occur in many languages. The Japanese pronoun zibun is
a well-known example: it indicates an internal self in certain contexts (Kuno
1972, Kuroda 1973).
English does not have a direct counterpart of the logophoric pronoun, since
the reflexive doesn’t appear in the subject position of tensed clauses. Reflexives
that indicate a subject of consciousness, however, appear in other logophoric
contexts, as (14) illustrates:14
(14) a. This paper was written by Ann and myself.

b. But Ruperti was not unduly worried about Peter’s opinion of himselfi .
c. It angered himi that she should have the egotism to try to attract a man like
himselfi .
d. Luciei boasted that the chairman invited her husband and herselfi .
In (14a), the reflexive has no overt antecedent and is taken to refer to the
speaker of the sentence. Here and in other contexts, the reflexive alternates
with the ordinary pronoun form so that use of the reflexive represents a choice.
Reflexives that function similarly appear in other languages, e.g. Dutch zich,
Icelandic sig.
Locally Free Reflexives (LFRs) may also convey “empathy,” suggesting the
viewpoint of the antecedent in the situation expressed, as Kuno points out
(1972, 1987). The empathy meaning arises for LFRs in contexts that do not
directly express communication or mental state, as in (15):
13. Hagège compared the African languages with Latin and Japanese. Typically logophoric pro-
nouns are third person in subject position, but some languages are more flexible (Stirling 1993).
Some languages have a special pronoun form for non-subject cases, including the West African
languages Igbo, Gokana, and Ewe (among others).
The term “logophoric” has a general and a more specific use. Very generally, a logophoric pro-
noun attributes a connection of some kind between a pronoun and its antecedent, a connection
that is not licensed by the grammar, that is, the constraints of the Binding Theory (Reinhart &
Reuland 1993).
14. Cantrall (1969), Ross (1972) noticed that when they alternate with ordinary pronouns, reflexives
convey the consciousness or point of view of their antecedent. Sources of the examples: Ross
1970 (a); (b), is quoted in Zribi-Hertz (1989); c–d, Reinhart & Reuland (1993).
Other cases of LFRs include “My husband and myself would like to invite you to dinner”
(Keith Walters, pc).
(15) a. Max pulled the blanket toward himself.

b. Lucie likes pictures of herself.
According to Kuno, these reflexives indicate the point of view or “camera an-
gle” of their antecedent: the speaker thus empathizes with that person. Neither
example falls under Binding Theory principles. In (15a) the reflexive and its
antecedent are not coarguments of the verb; (15b) is a “picture noun context” –
picture nouns have been problematic for the Binding Theory (Reinhart &
Reuland 1991).
Locally Free Reflexives may also indicate emphasis, contrast, or intensifi-
cation. The clearest cases of emphasis have strong stress, conveyed by capital
letters in this example from Zribi-Hertz (1989):
(16) Joycei is just holding herself together . . . Heri defenses are well inside
HERSELFi , not where mine are, outside in clothes, hair, etc.
Contrastive uses may involve explicit or implicit comparison. There is also an

“intensive” LFR. It has a contrastive meaning and requires that the antecedent
be prominent in the discourse, according to C. L. Baker (1995). This use of the
reflexive is primarily found in British English; Baker’s study is based on the
novels of Jane Austen. Example (17) illustrates.
(17) Sir William Lucas, and his daughter Maria, a good humoured girl, but as
empty-headed as himself, had nothing to say that could be worth hearing, and
were listened to with about as much delight as the rattle of the chaise.
(Pride & Prejudice 188)
Baker observes that the antecedent to the LFR “himself ” in this example, Sir
Willliam Lucas, cannot be explained as a subject of consciousness, empathy,
or contrast; but that the antecedent is prominent in the discourse. Discourse
prominence, according to Baker, is due to a character’s role as Agent or Patient,
of high importance or rank, or the primary topic of concern.15 A similar inten-
sive use also occurs with the [NP+reflexive] form, e.g. “The president himself
opened the door,” found in British and American English.
This section has identified several interpretations of Locally Free Reflex-
ives, in addition to their referential meaning. In the logophoric meaning, the
antecedent of the LFR is responsible for the thoughts, feelings, or sayings ex-
pressed in a clause. The empathy use implies the point of view of the antecedent.
15. Baker’s account is based in part on Ross (1970), Moyne (1971) on reflexives, and a suggestion
of Ferro (1993) that “self” may function as a focus marker. According to Baker, in American
English the difference between intensive and non-intensive contrast has been neutralized.
The emphatic use highlights the antecedent of the reflexive, and the contrast use
involves explicit or implicit comparison with another entity. The intensive use,
typical of British English, both contrasts and highlights. Recall that LFRs can
always be replaced by an ordinary pronoun, since they are not subject to gram-
matical constraints. They represent the choice of the writer and carry special
meanings.
6.2.2.3 Null pronouns

Null pronouns have no morphology and thus the minimal amount of coding
material. Null pronouns appear as subjects in limited contexts. They are possible
in conjunctions, in nominals, in the complements of certain verbs.16 They also
appear in temporal clauses (18a), purpose clauses (18b), and participial adjunct
clauses (18c–d).
(18) a. Before leaving the room with Gabriel she signalled to Mr Browne by frown-
ing and shaking her forefinger on warning to and fro.
b. In order to confront dictators of his kind we must set aside the misleading
lesson of our victory in the cold war, achieved by huge military expenditure
on a deterrent strategy.
c. Punctuated by the most spectacular busts in economic history, the boom
continued in Europe for thirty years.
d. Ethiopian troops are pressing deeper into Eritrea, denying with every step
that they plan any long-term invasion of their neighbor and once-tight ally.
The situations in these sentences are tightly bound by their syntax. I will refer to
such clauses as “condensed clauses,” following Jespersen (1940). Condensed
clauses appear in argument and adjunct positions.
Sentences with condensed clauses present situations as a single syntactic
unit. The semantic meaning mirrors the syntactic presentation, as in (19):
(19) Searching with insatiable curiosity for underlying explanations, both did far
more than discover new facts or solve circumscribed problems, such as the
structure of DNA: they synthesized knowledge from a wide range of fields and
16. There are also lexically determined null subjects, as in the following examples. The possibilities
depend on the verb.
a. Mary wanted ø to win the race.
b. Mary wanted herself to win.
c. Mary wanted John to win.
d. Mary tried ø to win.
e. ∗ Mary tried herself to win.
f. ∗ Mary tried John to win.
Note that there is a close relation between the situations in main and complement clauses.
created new conceptual frameworks, large parts of which are still accepted
today.
This fragment, presented as a unit, contains two condensed clauses and two full
clauses, both with conjoined VPs.
Strikingly, condensed clauses with null pronouns in English appear in almost
the same syntactic environments as the Same Subject morphemes of switch
reference systems. They use the two closed systems that allow anaphoric de-
pendence, tense and referential expressions. Condensed clauses have the effect
of presenting situations as tightly bound, just as the corresponding Switch Ref-
erence cases do. This semantic discourse effect follows from the syntax.
Choices from the closed system of referring expressions also convey infor-
mation about the familiarity or unfamiliarity of the referent.
6.3 The familiarity status of referring expressions

Speakers and writers have a choice among linguistic forms when they refer
to a person or other entity. The possibilities include names and titles (Samuel
Barnes, the Secretary), NPs with the definite article (the cat), NPs with the
indefinite article (a cat), demonstrative (this cat, that cat), NPs with additional
lexical material (adjectives, relative clauses, prepositional phrases), pronouns
(he, they), and anaphors (herself ). The choices reflect assumptions about what
the receiver knows, assumes, and can infer. Following Grice, I assume that the
speaker or writer presents information in a way that will enable the receiver to
arrive at the intended message.
Referring expressions may introduce a new entity or refer to a familiar one. In
a model of an ongoing discourse, a referring expression is taken as an instruction
either to add an entity or to find the relevant antecedent in the Discourse Repre-
sentation Structure. New entities are introduced with an identifying condition; if
the referring expression indicates that an entity is familiar, the DRS is searched
for an antecedent with which it is coreferential.
The distinction between familiar and new entities is reflected in the forms
and use of the articles in referring expressions. Typically in English, the in-
definite article introduces new and unfamiliar entities into a discourse, while
the definite article indicates entities that are familiar, readily available to the
receiver in the discourse context. This was recognized in early work of Discourse
Representation Theory by Kamp (1981) and Heim (1982). The sequence in (20)
illustrates:
(20) I saw a cat. The cat was sitting on a fence, watching a bird.
The NP “a cat” refers to a hitherto-unknown cat, and we understand it as an

instruction to introduce a new entity into the DRS. The definite NP “the cat”
refers to a particular, identifiable cat – the one introduced in the preceding
sentence. If one were constructing a DRS for (20), the definite article would be
taken as an instruction to search for the antecedent entity and establish a link
between the antecedent and the NP in question. I ignore generic reference to
kinds.17
“Familiar” and “New” are both susceptible to further distinctions, as Prince
has stressed. Entities may be new in more than one way. An entity that is
New in a discourse may be “Brand-new” or “Unused.” Brand-new entities are
totally unknown to the receiver, while Unused entities are known but not active
or salient at the time of the utterance. They are in the receiver’s long-term
memory or background knowledge. The subject NPs of (21) illustrate, from
Prince (1981):
(21) a. A guy I work with says he knows your sister.

b. Noam Chomsky went to Penn.
The new entity in (21a) “a guy I work with” is quite different in status from
“Noam Chomsky” in (21b), as Prince points out. The former is Brand-new.
The latter is Unused, familiar – to the linguist at any rate – but not necessarily
active. The proper name can refer successfully only if the receiver is already
somewhat familiar with its referent.
The category of “familiar” entity can also be made more precise, Prince
shows.18 An NP with the definite article may refer to an entity actually available
in the discourse context; or to an entity that is readily inferrable in the context.
Example (22) illustrates:
(22) I got on a bus yesterday. Unfortunately, the driver was drunk.
Although no driver has been introduced, we have no difficulty inferring the

presence of one from the NP “a bus” in the preceding clause. This is a bridging
inference, based on our general knowledge that buses have drivers. Inferred
17. Both types of articles can convey reference to kinds:

The lion is a mammal.
A lion is a mammal.
18. The full set of Prince’s categories of Assumed Familiarity:
New → Brand-new, anchored & unanchored; Unused
Inferrable → Noncontaining; containing (from relevant set)
Evoked → Textually; situationally
entities tend to have the definite article, as in the example. Definite possessives
can also be taken as familiar if their possessor is familiar, as in “A man came
in. His daughter was with him,” from Barker (2000).
The simplest adequate account has three main classes of familiarity status,
classified as New, Evoked, or Inferrable (E. Prince 1981). Inferrable entities
arise from background knowledge and a trigger in the text, as the referent
of “the driver” in (22) is triggered by the mention of “a bus” in the pre-
ceding sentence. Prince’s categories are based on the assumption that writ-
ers choose forms based on their understanding of receivers’ knowledge and
expectations.
Evoked entities are familiar, usually coreferential with an entity already in-
troduced in the text. The distance between the entity and its antecedent may
be great or small. The antecedent may be in the same or a preceding sen-
tence, in which case the entity is already in the focus of attention, present in
the consciousness of the receiver. The antecedent may also be relatively far
away.
The forms of referring expressions differ in amount of coding material
and content. Thus the unstressed pronoun he is shorter and has less content
than a full NP such as the man or the man standing by the window. Expres-
sions that have the least information tend to be used when their antecedents
are nearby and/or salient in the text. For instance, if the antecedent of an
entity is nearby, a pronoun is commonly used to refer to it. When the an-
tecedent is further away, people tend to use referring expressions that are more
informative.
Drawing on these and other observations, Ariel (1990) gives a ranking of
referring expressions that links coding material, or amount of information,
and accessibility. Accessible referents are near, or easily found, in the text.
The rank of an expression correlates with how much information it conveys.
Null forms and unstressed pronouns are at the low end of the scale, with a
minimal amount of coding material. In the middle are demonstrative NPs (this
X, that Y) and stressed pronouns. At the high end of the scale are full names
and long definite descriptions. According to Ariel, each expression signals the
reader to search for an entity with accessibility indicated by its ranking on the
scale.
Evidence from text counts supports Ariel’s ranking. Pronouns are predom-
inantly used for short distances, anaphoric demonstratives are used in cases
of intermediate distance, and definite descriptions refer to antecedents that are
relatively far away (1989:70). Ariel notes that this conforms with the principle
of relevance proposed in Sperber & Wilson (1986). The correlation between
form and accessibility is supported by psycholinguistic research, e.g. Marslen-

Wilson et al. (1982), Garrod & Sanford (1985).
The main factors in a text that make an antecedent more or less accessi-
ble, Ariel claims, are (i) distance from the referring expression; (ii) number
of competitors for the role of antecedent that agree in gender, number, and
other features; (iii) topicality; (iv) whether there is a frame for identifying the
antecedent (1989:65).
The correlation between linguistic form and information is further devel-
oped by Gundel et al. (1993). They propose a ranking of referring expres-
sions according to the “cognitive status” of the receiver. Cognitive status
includes location in the text and in memory, attentional state, and position
on the givenness hierarchy. There are six degrees of status. Referring expres-
sions of English appear at each degree. The ranking is given in (23), with the
different ranks numbered for clarity; the pronoun it stands for all unstressed
pronouns:
(23) Cognitive Status (Gundel et al. 1993)

6 5 4 3 2 1
in focus > activated > familiar > uniquely > referential > type
it that, this, that N identifiable indefinite19 identifiable
this N the N this N aN
The ranks are implicationally related. The implicational ranking relates a form
to all others that are below it on the hierarchy. For instance, if an expression of
rank 4 is licensed, so are expressions of ranks 1–3.
In using a particular form a speaker “signals that the associated cognitive sta-
tus is met and . . . that all lower, less restrictive statuses have been met” (Gundel
et al. 1993:184). The implicational relations may explain how people under-
stand expressions from the lower end of the scale used to refer to relatively
familiar entities. Such reference often appears in discourse when there is a
change of direction (see 6.3.2 below). Gundel et al. give similar rankings for
19. The hierarchy in (22) has a definite and indefinite use of the demonstrative this. In the definite
use, the NP is coreferential with an antecedent, as in the toy example below:
A man walked into the room. This man was wearing a red hat.
The indefinite use introduces a new element into the universe of discourse:
There was this guy I saw yesterday who was wearing a red hat.
The second use is common in speech and informal writing, as Keith Walters has pointed out
to me.
the referring expressions of other languages. The rankings in (23) relate nicely
to Prince’s classes of familiarity status. Expressions from rank 6 to 3 are Evoked
or Inferrable; 1 and 2 are New.
The work reviewed above emphasizes the receiver rather than the text. There
is evidence that a text-based approach may be sufficient, even preferable, to
the receiver-based one. Prince (1992) looked at both approaches in a study
of referring expressions in a single written text. She asked whether hearer-
based or discourse-based categories were the most relevant and best predictor
of pronoun form. Comparing the cognitive status for the receiver and text status
of the entities, she found that the two overlapped in many cases. When they did
not overlap, “discourse-old” and “discourse-new” were better predictors of the
form of referring expressions than the corresponding categories “hearer-old”
and “hearer-new” (Prince 1992). The work of Birner & Ward (1998), discussed
in Chapter 8, also bears out this finding.
6.3.1 Patterns in discourse

There are two main patterns of coreference in discourse: continuity and shift.
Pragmatic factors such as the familiarity status of a referring expression, syn-
tactic structure, and the distance and importance of an antecedent play a role in
maintaining continuity. Shifts are indicated by referring expressions that break
continuity.
6.3.1.1 Continuity
Texts have referential continuity when each sentence refers back to entities
in the sentence that immediately precedes it, according to Sanford & Garrod
(1981). The simplest kind of continuity is conveyed by a chain of pronouns after
an entity is introduced. Example (24) is a constructed example from Kameyama
(1985):
(24) Maxi is a chemist. Hei is 30 years old, and hei lives in San Francisco.
The pronouns keep the entity at the same level of focused attention. If the chain
of pronouns is broken by an intervening sentence, the result can be odd and
difficult to interpret, as in (25), from Sanford & Garrod (1981).
(25) The little puppyi trod on a waspj . The puppyi was very upset.
Itj started to buzz furiously.
Examples like this have led people to seek the conditions for felicitous
continuity.
The interpretation of a pronoun is affected by syntactic structure, espe-

cially parallel structure. In sentences with parallel structures, people interpret as
coreferential NPs that have the same grammatical function. Consider the ex-
amples below:
(26) a. Billi talked to Tomj and hei talked to Mary.

b. Billi talked to Tomj and HEj talked to Mary.
The natural interpretation of (26a) has “Bill” as antecedent for the pronoun,
although both nouns in the first conjunct are possible.20 Strong stress on the
pronoun conveys the other interpretation unambiguously, as in (26b); it also
adds a sense of contrast. The principle of parallelism is very strong, overriding
the factor of proximity. For instance, “Tom” is nearer to the pronoun in (26a),
but the parallel interpretation remains. From a procedural point of view, par-
allel syntactic structures are an instruction to the reader to look for a certain
connection between them; the antecedent–pronoun link serves this function
(Blakemore 1988). Parallelism affects pronoun interpretation similarly in other
languages.
There are other patterns of continuity as well. General principles for conti-
nuity in discourse are stated in Centering Theory, which models connections
between referents in discourse (Walker et al. 1998).21 The theory works with
“centers,” semantic entities in an ongoing discourse model. The referring ex-
pressions of a sentence comprise a set of ranked forward-looking centers (Cf).
The highest ranked is a backward-looking center (Cb), which links the sentence
to previous discourse. The Cfs are potential backward-looking centers for other
utterances.
The notion of grammatical salience is the key to the ranking of centers.
Salience is determined by a hierarchy of grammatical relations: subject >
object > second object. The “preferred center” is the Cb. The Cb represents
20. The effects of parallelism can be increased or mitigated by the lexical properties of the main
verb. For instance, note the differences in the pairs below:
a. Frani won the money from Helenj because shei was a skillful player.
b. Frani won the money from Helenj because shej was a careless player.
c. Heleni punished Cathyj because shej confessed to shoplifting.
d. Heleni punished Cathyj because shei disapproved of shoplifting.
21. Centering Theory developed from a computational model of local attention and coherence,
where coherence was understood in terms of the links between sentences of short discourses
(Joshi & Weinstein 1981, Grosz & Sidner 1986, Walker et al. 1998: Ch. 1). Psychological
studies, e.g. Brennan (1995), Gordon & Chan (1995), give evidence that supports the model.
See also studies in Walker et al. (1998).
a discourse entity that the sentence centrally concerns – a notion similar to

discourse topic (see Chapter 7). Pronouns and grammatical subjects are key
factors in English centering transitions (Kameyama 1998).
The theory focuses on pronouns, following the generally accepted notion that
pronouns are used to refer to entities readily available in a discourse. If there
is a pronoun in a sentence, it is the Cb, the backward-looking center. Indeed,
only if the Cb is realized as a pronoun does it contribute to the coherence of the
discourse (Gordon & Chan 1995).
Centering Theory identifies strategies for the transition from one sentence
to another. For instance, the strategy “Continue” predicts that the same Cb ap-
pears in two utterances. The pronouns in (26) above exemplify this pattern. It is
the most highly preferred, according to the theory, because it maximizes con-
tinuity. Other strategies hold when sentences have different Cbs. For instance,
in “Smooth shift” the pronouns have different referents. Example (27) gives a
constructed fragment which exemplifies this strategy, from Walker et al. (1998).
The subscripts indicate the interpretation.
(27) a. Jeffi helped Dickj wash the car.

b. Hei washed the windows.
c. Hej buffed the hood.
The change in pronoun referent is said to be less preferred. Centering Theory

models expectations of continuity, assuming a simple, linear progression.
I return to the topic of continuity in Chapters 8 and 9.
The distance between a referring expression and antecedent also affects con-
tinuity. When the two are near each other, relatively uninformative expressions
are expected; more informative expressions are predicted when greater distance
is involved.
6.3.1.2 Referring expressions and discourse organization

The organization of a text also affects the form and interpretation of refer-
ring expressions. The form of a referring expression may signal continu-
ity, or a break in continuity. Breaks in continuity may coincide with new
themes or episodes, or, more locally, with changes of direction or point of
view.
The clearest distinction is between full NPs and pronouns. Full NPs are
often used where pronouns would have been possible, and indeed expected in
a simple continuity account. The full NPs occur at points where the discourse
shifts in some way. In a study of four written narratives Fox found many such
examples. The full NPs coincided with “the structural organization” of the text
(1987a:167). An illustrative example from Fox is given in (28). The change from
“she” back to “Susan” demarcates a new unit in the text, indicated graphically
by the paragraph indentation.
(28) The four flights up to her floor seemed longer than usual to Susani . Shei
paused on several occasions, because of a combination of physical fatigue and
mental effort.
Susani tried to remember if Bellows had said succinylcholine was among
the drugs found in the locker.
The choice of form, a full NP (Susan) at the beginning of the second paragraph
signals the functional structure of the narrative. Fox (1987b) documents similar
patterns in discursive prose. Similarly, Clancy (1980) observes that full NPs
are used at episode boundaries in narratives based on the “Pear Story” films, in
both English and Japanese.
These findings have experimental support from work in psycholinguistics. In
studies of production and comprehension, Vonk et al. (1992) looked at pronouns
and full NPs in text fragments. They found that when NPs were more specific
than necessary to recover an antecedent, the NPs had a discourse structuring
function. Pronouns were related to thematic continuity, full NPs triggered or
signalled shifts in theme.22 The authors interpret these findings as showing that
the full NP functioned as a signal to set up a new chain of information in the
discourse representation.
Other referring expressions have the same pattern: the fuller, more informa-
tive expression indicates a shift in the discourse. For instance, the pronouns
it and that differ. The pronoun it is neutral and that is deictic, indicating
non-proximate reference. In a study of oral apartment descriptions, Linde
(1979) found that the two appeared systematically according to whether their
referents were in the same or different discourse unit. It was used when
the referent was in the same unit; that when the referent was in a different
unit.
Another prediction of the linear continuity account is that a long gap between
mentions of a referent triggers a full NP. One might predict that pronouns,
22. These patterns also appear in other languages. In an oral Polish narrative with a main pro-
tagonist, reference was made with null pronouns when continuous and predictable. Null pro-
nouns are the Polish equivalent of unstressed ordinary pronouns. Overt pronouns appeared
when there was narrative discontinuity, changes in speaker perspective, and contrast between
foreground and background (Flashner 1987). Similarly Clancy (1980), in a study of spo-
ken English and Japanese narratives, found that people tended to use full NPs to mark dis-
course boundaries, point-of-view changes, and episode changes. See also the articles in Givón
(1979).
6.4 Referring expressions and DR Theory 149
which have little coding material, are not strong enough to access an antecedent
that is relatively far away. This prediction has been disconfirmed. In a study
of narrative, for instance, Fox found that pronouns are used for a particular
character over relatively long stretches of text “until another character’s goals
and actions are introduced, unless those goals and action are interactive with the
first character’s” (1987a:162). In sum, what matters is discourse continuity. In
texts with one central character, reference to that character may be made at any
time with an ordinary pronoun, even if there have been significant interruptions
(Whitaker & Smith 1985).
Similar findings have been documented for non-narrative English discourse.
Hinds (1977) examined the use of pronouns and full NPs in newspaper articles.
He found that the choice between them depended on the structure of the text.
Nominal and pronominal reference were sometimes used to convey different
degrees of prominence, with nominal reference indicating “semantically promi-
nent” information. Hinds found that nominal expressions tended to appear at
discourse peaks.
Summarizing, the form of a referring expression encodes the familiarity status
of a referent. In discourse, anaphoric expressions convey referential continuity.
Fuller expressions are used to convey discourse shifts, and expressions that
code unfamiliarity tend to appear at discourse boundaries.
6.4 Referring expressions and Discourse Representation Theory

Discourse Representation Theory developed a landmark procedural approach
to referential expressions.23 The referential expressions of an ongoing text are
interpreted as instructions to introduce a discourse entity into the universe of
discourse, with a condition that identifies the entity. Conditions indicate that an
expression is coreferential with an entity in the discourse context, or specify
the entity in the wider context. The bare bones of the approach are laid out in
Chapter 3. I return to the topic here, with comments on how the material of this
chapter relates to the theory. For more complete accounts see Kamp (1981),
Heim (1982), Kamp & Reyle (1993).
An important insight of Discourse Representation Theory is that referring
expressions have different functions according to their forms. One class of re-
ferring expressions introduces new discourse entities into the context, another
23. The approach of Discourse Representation Theory was a significant advance over earlier theo-
ries such as the referential theory of indefinites and an alternative, the quantificational theory.
The discussion below necessarily omits many important points, among them varieties of
pronoun reference and plurals.
introduces discourse referents that are coreferential with entities already there.
These classes correspond, more or less, to the distinction between new and
familiar referents. In some cases a referring expression has the form asso-
ciated with familiar referents, but the context does not contain an appropri-
ate antecedent. The receiver accommodates by assuming or inferring such an
antecedent.
Proper names and indefinite NPs license the introduction of new entities into
a DRS; pronouns and certain other NPs refer to entities that are already in the
context.24 In certain contexts coreference is impossible, as noted in Chapter 3.
The theory recognizes this by stating structural accessibility conditions. (This
notion of accessibility is different from accessibility based on distance in a text.)
In conversation, pronouns may refer to an entity in the context; they refer to
discourse entities in a written text.
Anaphora in the DRS is thus a relation that holds between a discourse entity
and a pronoun or other referring expression. This is the standard Discourse
Representation Theory account. The work on referential expressions discussed
above suggests that the standard account should be augmented slightly. Expres-
sions that refer to familiar entities should include NounPhrases with demonstra-
tive (that, this), and null pronouns which are interpreted by rule. Expressions
that introduce new identities should probably include referential indefinites
(“this N”), although it is not easy to distinguish them from nominals of the
same shape referring to “activated” referents.
Reflexive pronouns in syntactically bound contexts require additional con-
ditions, as Kamp & Reyle point out. Reflexives do not alternate with simple
pronouns in such contexts. To convey a coreferential meaning a reflexive form
must be used. For instance, in “Jane washed her,” “her” cannot be interpreted
as coreferential with the subject “Jane.” Coreference can only be expressed
with the reflexive “Jane washed herself.” These conditions are straightforward
in principle, although actually stating them is complex.
Locally Free Reflexives do alternate with simple pronouns and thus re-
quire a different treatment. Two questions arise, one specific to the class and
the other general to pronouns. The specific question is whether the special
quality of LFR should be encoded in the DRS. The mechanism for such en-
coding exists: a special feature could be introduced that would allow for the
24. Proper names are linked to entities in the world by an external anchor. The external anchor
for a discourse referent x is a function which maps x into some real individual a. The anchor
constrains the mechanism used for truth-conditional verification of a sentence (Kamp & Reyle
1993:248).
6.4 Referring expressions and DR Theory 151
appropriate interpretation. There is also a subjective use of LFRs, which has

a different treatment. I will propose that such reflexives be associated with
a discourse entity as a Responsible Source in the DRS; see Chapter 9 for
details.
Discourse Representation Theory provides a mechanism for interpreting re-
ferring expressions, and states the structural and semantic conditions for one
referential expression to be coreferential with another. Entities are either “new”
or “familiar” and their definition is based on the developing DRS for a text.
New entities have no antecedents in the DRS; familiar entities have antecedents
in the DRS. The current theory does not deal with inferred entities. However,
proposals for bridging and other inferences are put forth in Asher & Lascarides
(1998).
The pragmatics of successful reference are not within the purview of the
theory. When more than one antecedent is structurally possible for a referring
expression, the theory does not have principles for the best or most likely choice.
But there is no incompatibility between the pragmatic material discussed above
and Discourse Representation Theory. Rather, pragmatics provides an indis-
pensable complement for the structural approach of the theory. Finding the
appropriate antecedent for a pronoun or NP depends on several factors, as we
have seen. In addition to relative familiarity status, continuity and discourse
organization play a role. The contrastive meanings associated with closed sys-
tems, and the resulting discourse meanings, also belong to pragmatic inter-
pretive principles that go beyond the structural information of a construction
rule.25
As an example of how referential entities are treated in the theory. I give
a DRS for a simplified version of an example discussed above, given here as
(29). The proper name “Bellows” is at the top level of the DRS because the
reference of proper names is independent, not contingent on information in the
sub-DRS. The if -clause is a sub-DRS that encodes the information after “tried
to remember” in the third sentence; this treatment roughly follows that of Asher
(1993).
(29) 1 The four flights up seemed longer than usual to Susani . 2 Shei paused on
several occasions.
3 Susani tried to remember if Bellows had found drugs in a locker.
25. In a more complete account of the pragmatics of discourse, pragmatic principles for contrastive
meanings might be stated. I proposed such principles for contrastive meanings in aspectual
viewpoint systems in C. S. Smith (1997).
(30) Schematic partial DRS for (29)
x y z u v s1 e1 e2
1. x = four flights
2. y = Susan
3. s1: seemed long (x,y)
4. z = she
5. e1: pause on several occasions (z)
6. z=y
7. u = Susan
8. u=y
9. v = Bellows
10. e2: try to remember (v
if
e3 a b c
1. a = Bellows
2. b = drugs
3. c = locker
4. e3: find (a,b in c)
5. a=v
The discourse entities are licensed by the proper names, pronouns, and Noun-
Phrases. In this case there is only one possible antecedent for the pronoun “she”:
Susan. The second instance of the proper name is taken as coreferential with
the first. This is a matter of pragmatic interpretation, indeterminate in the DRS.

(8) Jim Collins. How it works. US Airways Magazine, Attaché, May 2001.
(18a) James Joyce, “The Dead,” In Dubliners, 1916; reprinted London: Penguin Books,
1982, p. 185.
(18b) Tony Judt, Tyrannized by weaklings. New York Times, April 5, 1999.
(18c) Peter Drucker, The information revolution. Atlantic Monthly, October 1999.
(18d) After a victory, Ethiopia looks toward other fronts. New York Times, May 20,
2000.
(19) Jared Diamond, A tale of two reputations: scientific contributions of Charles
Darwin and Sigmund Freud. Natural History, 2001.
I I I Surface presentational factors
7 Subjectivity in texts
In all Discourse Modes and genres, one finds passages that suggest a particular
voice. They convey a sense of subjectivity, a point of view toward propositional
information. “Point of view” is familiar as a literary term referring to presen-
tation of the mind of a fictional character in narrative. More generally, point
of view is “the perceptual or conceptual position in terms of which narrated
situations and events are presented” (G. Prince 1987:73). Linguists now use
the term for expressions of speech and thought, evidentiality, perspective, and
other indications of an authorial or participant voice. “Point of view” is used
almost interchangeably with “perspective” and “subjectivity.” I shall use the
latter term as more general. Subjectivity is conveyed by grammatical forms and
lexical choices.1
Three traditions come together in the area of subjectivity. One is deixis and
its linguistic expression. Deixis is a general term for the centrality of the here
and now in language. The study of deixis takes as basic the canonical speech
situation with Speaker and Addressee, and explores its linguistic ramifications.
The second tradition involves evidentiality, indications of the source and reli-
ability of information. Evidentiality is a relatively new term for the semantic
field of attitude toward knowledge, a kind of modality. Linguistic resources for
this vary strikingly across languages. Finally, subjectivity conveys the contents
of mind and personal perspective; here linguistic study is complemented by
a strong literary tradition. The area is a vast one and I intend this chapter as
an introduction, by no means a complete account. I will give the grammatical
underpinnings of subjectivity, the bare bones.
Subjectivity arises primarily in discourse contexts. It is expressed by gram-
matical forms at the sentence level: verbs and their complements, adverbials,
tense, modals, aspectual viewpoint, anaphors. Often a subjective form has scope
1. Some of the material in this chapter was presented at a symposium at the University of Oslo in
December 2000: “Information Structure in a Cross-linguistic Perspective.” I would like to thank
the audience for their questions and the discussion; I also thank Cathrine Fabricius-Hansen for
helpful comments.
155
beyond a single sentence. To interpret the forms of subjectivity we determine

their scope and identify the mind responsible: the Responsible Source.
I suggest a composite account of subjectivity in Discourse Representation
Theory, using rules that look at subjective forms in a sentence and in the con-
text. The rules assign the role of Responsible Source to a participant in the
text situation or to the author. The interpretation is encoded in the Discourse
Representation Structure, using the technique of subordination. The Respon-
sible Source is associated with an embedded sub-structure that encodes the
relevant information.
The major categories of subjectivity are communication, contents of mind,
evidentiality/evaluation, and perception and perspective. I do not consider
lexical choice. Particular words in a sentence often suggest the mind of the
speaker/writer or a participant in the text situation. For instance, the following
sentence appeared in a recent newspaper article about the political campaign
being prepared in the United States: “The Republicans and Democrats have
concocted their message for the coming campaign.” The word “concocted” sug-
gests a skeptical attitude of the writer rather than a straightforward recounting
of facts. I will not deal with this kind of subjectivity.
As with the discussion of Discourse Modes, I try to provide an armature for
the complex and subtle notions conveyed by texts.
Section 7.1 discusses subjectivity and objectivity, and the notion of Respon-
sible Source; 7.2 considers communication; 7.3, expressions of the contents
of mind; 7.4, evaluation and evidentials; 7.5, perception and perspective; 7.6
formalizes the interpretation of subjectivity; 7.7 summarizes.
7.1 Responsibility for subjectivity

When a sentence expresses a point of view, or takes a particular perspective, we
need to identify the person/mind responsible. Ascribing responsibility is part
of the interpretation of a subjectivity. There are some systematic ambiguities as
to whether the author or a participant in the text situation is responsible. I will
sketch a general account.
There is a prior question, however, which I take up first. It is reasonable
to ask whether we can recognize certain sentences as subjective and others as
objective. I think that we can. In making the case I draw on insights about the
genre of narrative fiction.
The linguistically oriented study of subjectivity begins, perhaps, with
Kuroda’s (1973) observations about Japanese fiction. The Japanese language
has different forms for subjective and other expressions of mental state. To
7.1 Responsibility for subjectivity 157
express one’s own sadness, for instance, a person uses a subjective adjective
with a first-person pronoun; to report the sadness of another, a verb and a third-
person subject. What Kuroda noticed is that the subjective forms appear in third-
person narrative fiction, where they would not, strictly speaking, be expected. In
these fictional contexts the subjective forms directly render a character’s mental
state; they do not function as linguistic communication.2 Kuroda calls this use
the “non-reportive” style, distinguishing it from the “reportive” style typical of
communication (1973:384). Although lacking forms like these, English too can
convey expressions of mental state. In fact, there are English expressions that
correspond to Kuroda’s reportive and non-reportive styles. Banfield argues that
represented speech and thought, which are typical of fiction, are non-reportive
(1982:12). Subjectivity can also be expressed with deictic pronouns and verbs
of mental state. Lyons notes that English has distinct forms for “experiential”
and “non-experiential” expressions (1982:107).
In fact, English has ways of expressing subjectivity in sentences of all genres
and modes. The linguistic forms that convey subjectivity include verbs of mental
state and perception, modals and evaluative adverbs, and others discussed
below. Sentences that lack these forms are neutral or objective in our conven-
tional understanding, which posits a Responsible Source and a point of view
for every sentence. I do not have space here to explore competing views.3 The
force of the subjective and objective is unique in fictional narrative, however.
The fictional world of a story or novel can be known to the reader only from
the text. The functions of sentences with and without subjective elements reflect
this. Certain sentences of a fictional narrative are objective, in the sense that we
accept them as true information about the fictional world. These are sentences
2. Kuroda’s examples include sentences with the subjective adjective forms and third-person sub-
jects. The following, for instance, has the subjective adjectival form for “sad”:
Yamaderra no kane o kiite, Mary wa kanasikatta.
Hearing the bell of the mountain temple, Mary was sad.
Sentences like this are odd, even ungrammatical in isolation. In a narrative context they are
taken to express a character’s point of view directly.
3. This view has been developed primarily for narrative. For instance, Genette (1980) says that
strictly objective narration is impossible, and that all narrative is subjective, or “focalized,”
although in different ways. He makes an important distinction between the “one who sees,” (the
narrator), and “one who speaks.”
There is surely something right about the idea that the speaker/author is present in all genres,
cf. Palacas (1993). In an account that included expressive lexical items, one might want to posit
a Responsible Source for all cases.
The distinction between subjective and objective sentences may not hold for all languages.
Some languages have obligatory morphemes which code subjective information for evidentiality
in all sentences; see Willett (1988).
with no elements of subjectivity. Subjective sentences filter information through

the lens of a character’s perspective. They give information about the character
as well as about the fictional world. Banfield (1982) makes this point at length;
see also Bruder & Wiebe (1995), Galbraith (1995). To understand fiction, then,
it is important to distinguish subjective and objective sentences.
The convention of subjective and objective sentences is not limited to fic-
tion, however: it holds quite generally in discourse. For “objective sentences” –
those without linguistic forms associated with subjectivity – we assume that
the author is the responsible source. We ask whether such sentences are true
but not, usually, who is responsible for them. What is unique to fiction is
the understanding of “objective” sentences as true. Subjective sentences can
be identified by their linguistic forms; I discuss the forms and their interpreta-
tions in later sections of the chapter.
Let us assume that sentences expressing the contents of a mind, or a particu-
lar perspective, are subjective. For such sentences the question arises: to which
mind do we ascribe the material that is expressed or reported? There are clear
cases. In a belief report such as “Mary believed that John was sick” we ascribe
the belief expressed in the complement clause to Mary. In the sentence “Mary
unfortunately may win the race” we ascribe responsibility for the modal and
the adverbial commentary to the speaker, not to Mary. Indirect speech is sys-
tematically ambiguous: responsibility for the content may rest with the reporter
or with the original speaker.
To interpret expressions of subjectivity we must find the source, the person re-
sponsible. Texts can be seen as creating the basic communication situation with
a Speaker/Author, Addressee, and the speaker’s communication. This situation
provides the possibilities for sources of subjectivity. I use the term Responsible
Source for the mind to which responsibility is ascribed. The possibilities depend
on the type of subjective expression involved. In some sentences the Author is
the Responsible Source. In others the source is in the communication itself, a
participant in the text situation. Usually it is the human referent of a subject NP
or evidential PP. I will refer to this mind as the Subject. The Subject is often
distinct from the Responsible Source, unless only first-person reference is in-
volved. Compositional rules for determining the Responsible Source are given
below. One source of responsibility can account for the varied interpretations
of subjectivity.4
4. The view that only one Responsible Source in needed accords with much recent work. In her
discussion of narrative and represented speech, Banfield (1982) formulates syntactic principles
for attributing expressive elements to a unique subjectivity. Stirling (1993) takes a similar
approach in the Discourse Representation Theory framework. She posits a single Validator role,
In the next few sections I discuss the main types of expression that lead
to the interpretation of subjectivity: communication, mental states, evaluation,
perception and perspective. The examples of text passages come from texts in
this study.
7.2 Expressions of communication

Linguistic communication is by definition due to the speaker/author, a sentient
being. The category includes direct address from writer to reader, and commu-
nication by participants in text situations. The latter comprises quoted speech,
represented speech, and indirect speech. Quoted speech is often referred to as
“direct” speech. I use the term “quoted” here as one of several kinds of direct
communication.
Verbs of communication introduce quoted speech, indirect speech, and repre-
sented speech. Quoted speech reproduces what was said, in principle faithfully.
Indirect speech reports what was said. There is always the possibility that the
report is not entirely faithful. Actually both direct and indirect speech almost
always involve the speaker/author, as Tannen (1989) points out.5 The third type,
represented speech, presents what was said or thought: it is neither a report nor
a reproduction. It is closest of the three to Kuroda’s “non-reportive” speech.
The different types have characteristic patterns of deixis.
Quoted and indirect speech are illustrated first; the examples were chosen to
bring out the role of deictic elements. Verbs of communication are sometimes
optional in fiction.
(1) Quoted speech and thought
a. “I am getting ready for the party this afternoon.”
b. Maryi said “Ii am excited.”
c. Maryi told mej yesterday at the station,“Ii will meet youj here.”
d. Maryi asked “Do Ii have to go?”
(2) Indirect speech and thought
a. Maryi told mej yesterday at the station that shei would meet mej there.
b. Maryi asked whether shei had to go.
which is associated with the appropriate sentient being. Speas and Tenny (2001) posit a single
interpretive role, the Evaluator, in a semantically based syntactic approach. In contrast Sells
(1987) argues that three roles are necessary to account for logophoric expressions; for discussion
see Stirling (1993). Sanders & Redeker (1996) also posit a single Responsible Source, using the
framework of cognitive linguistics.
5. The line between quoted and reported speech is sometimes difficult to draw, in a discussion of
real and constructed dialogue and quotation. Tannen (1989) notes that the context for quotation
and report always informs what is said. She claims that even seemingly “direct” quotation is
really “constructed dialogue,” primarily the creation of the speaker rather than the party quoted
(1989:99). I am grateful to Keith Walters for pointing out this book to me.
Quoted speech typically is introduced by a verb of saying and a direct represen-

tation of what was said, as in (1). Person, tense, and other deictics are anchored
to the first-person speaker and the here and now, as in the original or imagined
utterance.
Indirect speech does not present exactly what was said. In reports of indirect
speech, pronouns, tense, and deictic adverbials must shift away from the deictic
center. Compare for instance (1c) and (2a), from Banfield (1982). In (1c) the
complement reproduces Mary’s utterance, with pronoun I, future will, and de-
ictic here. In (2a) the deictic forms are shifted: the pronoun is she, the deictic
is there. The tense is past, in concord with the tense of the main clause verb.6
Indirect speech is syntactically constrained: constructions that appear only in
main clauses, such as questions and exclamations, are blocked unless they have
other forms. Examples (1d) and (2b) illustrate direct and indirect questions.
Complementizers typically appear, e.g. that and whether in (2a–b).
There is a systematic ambiguity in indirect speech. It may present precisely
what was said, or a recoding by the reporter, as in the classic example of (3):
(3) Oedipus said that his mother was beautiful.
This sentence could be used to report an utterance of Oedipus in which the

speaker identifies a person as his mother and says something about her (the “de
re” reading). It could also be used to report exactly an utterance of Oedipus, “My
mother is beautiful” (the “de dicto” reading). The truth of the sentence depends
on both the interpretation and what was actually said: it might be true on the
de re but not on the de dicto reading. Epithets and evaluative terms are ambigu-
ous in the same way; see example (10) below.
Deictic expressions in the lower clause may be oriented to the reporter, as
in (4):
(4) a. John said that Mary was leaving tomorrow.

b. John said that Louise is pregnant.
The deictics convey that the reporter has recoded all or part of what was said.7
In (4a) “tomorrow” relates the leaving to Speech Time rather than to the past
time of John’s utterance. In (4b) the clauses have different tenses. The sentence
6. Tense concord is also known as “sequence of tense.” Tense and person concord is required in
English, but not in all languages. Japanese and Navajo, for instance, do not have shifted person
markers nor tense concord in the complements of communication verbs; Russian does not have
sequence of tense, Amharic does not have sequence of person (Schlenker 1999).
7. Deictic adverbials vary: some can shift orientation to any deictic center, others are oriented only
to the here and now, as noted in Chapter 5. Tomorrow and yesterday are rigid adverbials that do
not shift.
conveys that John’s utterance about Mary’s pregnancy was made in the past;
and that the reporter relates the pregnancy to the time of speech. This type of
sentence is known as a “double-access” sentence; see Ogihara (1996), Abusch
(1997), Giorgi & Pianesi (1997).8
Verbs of communication are a distinct syntactic class. They allow a direct
object complement which expresses the actual communication and an indirect
object referring to the addressee (“X said Y to Z”). The class includes the verbs
say, ask, request, command, declare, confess, advise, insist, claim, shout, read,
sing, remark, observe, note, swear, promise, announce, pray, and many others.
Manner-of-speaking verbs such as shout, scream, whine, whisper, holler have
special properties (Zwicky 1971).
Given the close relation between them, we might attempt to derive indirect
speech from direct speech. However this approach cannot work. The argument
turns on two points: firstly, not all indirect speech reports have plausible coun-
terparts in expressions of direct speech; and secondly, the ambiguity between de
re and de dicto readings arises only for indirect speech (Partee 1973, Banfield
1982).
Communication in discourse may take the form of represented speech or
thought, which has some features of quoted and some of indirect speech. Rep-
resented speech maintains the syntax of actual, or quoted speech, but tense and
pronouns are shifted as in indirect speech; (5) illustrates:
(5) Represented speech and thought
1 He then took from his waistcoat pocket a little paper and glanced at the
headings he had made for his speech. 2 He was undecided about the lines
from Robert Browning, for he feared they would be above the heads of his
hearers. 3 Some quotation that they would recognize from Shakespeare or
from the Melodies would be better. 4 He would only make himself ridiculous
by quoting poetry to them which they could not understand. 5 They would
think that he was airing his superior education. 6 He would fail with them just
as he had failed with the girl in the pantry. 7 He had taken up a wrong tone. 8
His whole speech was a mistake from first to last, an utter failure.
After S1, the sentences have shifted tense and person. The narrator would
think of himself as “I,” would represent the future with “will,” the past with
“took.” The locutions of direct speech and thought are rendered directly, as in
Sentences 6–7.
8. Cathrine Fabricius-Hansen notes (p.c.) that the reporter is responsible for the temporal descrip-
tion but not for the claim. Example (4b) would not necessarily be false if Louise were not
pregnant at the time of the report, as is shown by the felicity of the following:
(i) John said that Louise is pregnant but in fact she isn’t.
It is the claim that John is reported to have made that would be false.
Represented speech is not syntactically embedded: it renders untouched

syntactic structures which appear only in main clauses. Such structures in-
clude exclamations, direct questions, topicalization, elliptical fragments, and a
few other constructions. Represented speech is also known as “narrated mono-
logue,” style indirect libre, and erlebte rede. Literary studies include Hamburger
(1973), Cohn (1978), Genette (1980).
Represented speech and thought are not limited to fiction, as the passage in
(6) illustrates; it is from an article in the New York Times cited in Wiebe (1991):
(6) Looking at the more severely affected countries, experts are wondering where
the saturation point will be. Where will the infection rate level off as most of
those engaging in riskier behavior fall prey: 30 percent? 40 percent?
The second sentence presents a represented thought or utterance: the direct

questions are due to the experts, not to the writer.
Although the forms of these communications are at the sentence level, they
are not limited to a single sentence. The scope of the subjective form may
extend over several sentences, so long as continuity is maintained. This sense
of continuity is discussed in connection with examples (18) and (19).
Direct address to the reader in the form of questions, imperatives, and deictic
you, also appear in texts. They address a communication to the reader, creating
the canonical deictic situation. The source is the author and the addressee is the
reader; (7) illustrates:
(7) Those are rather windy thoughts, but I have had a hard time escaping them
the last few weeks as I’ve tried to make some sense of the Events surrounding
President Clinton, Congress and impeachment. What will people say about
all this 20 years from now? Will they quiz each other on the minutiae of the
Starr report, as they do on the contents of the Warren report? Or will posterity
simply conclude that one of the two political parties, having lost an election,
saw an opportunity to nullify it and proved too weak to resist the temptation?
You may find that a difficult question. I don’t.
The writer is the source of subjectivity in this text. The reader may also be
addressed directly, as in the last sentences of the fragment above.
7.3 Contents of mind

Expressions of mental states such as thoughts, beliefs, and attitudes convey the
contents of mind. Typically they have verbs like think and believe; complements
express the object of thought or belief. The same deictic conventions hold for
complements of mental verbs as for indirect speech, as (8) illustrates:
(8) a. We thought that Bella was in New York.

b. Maryi believes that shei won the race.
In (8b), which expresses a belief, the main and complement clauses have
coreferential subjects. The coreference relation between subjects in clauses of
thought and communication triggers a logophoric interpretation, discussed in
Chapter 6.
The person who holds a belief is responsible for it. But belief sentences are
also reports by the writer and have the possibility of recoding. The writer’s
presence is sometimes suggested by evaluative expressions and epithets, which
are vague as to source. For instance, epithets such as “that fool,” “that idiot of
a doctor,” “a peach of a girl” can be ascribed to the believer or to the reporter,
as in (9):
(9) a. Sam thinks that his beloved cat should have only fresh fish.
b. Mary believed that that fool Kevin wanted to run the committee.
On one reading, the epithets “beloved cat,” “that fool” are part of the reported
beliefs. There is also a reading in which the reporter designates the cat as
beloved, Kevin as a fool. This reading is salient in (9b) because of the deictic
that.9
Sentences with clauses referring to propositions have an additional dimension
of subjectivity. Recall that propositions are subjective, ascribed to the person
holding them (Chapter 4). Clauses referring to Facts are not subjective in this
way. Nominals of communication and mental state are similar to verbs in sub-
jectivity. For instance, the interpretation of “Mary’s belief that John won the
race” is not significantly different from the corresponding tensed clause.
7.4 Evaluative and evidential subjectivity

Different attitudes and degrees of commitment may be expressed by participants
in the text situation, or to the writer as participant or intruder. The linguistic
9. The interpretation is my own; Banfield claims that epithets in clauses of indirect speech and
thought are attributable only to the speaker (1982:54).
Epithets are lexical items whose full interpretation entails a reference to the speaker. In the
embedded clause of indirect speech, however, they only express the state and attitude of the
reporting speaker, according to Banfield, who suggests two classes:
a. Epithets: fool, bastard, etc.
b. Evaluative adjectives: the poor girl, that damned Faustus, etc.
Banfield claims that these forms constitute a well-defined, “non-classificatory” class that can
be distinguished with a lexical feature.
expressions are evaluative and evidential adverbs and adjectives, modals, and
parenthetical expressions. Attitudes are ascribed to participants in the text sit-
uation when they appear within the scope of subjective expressions.
Adverbs can be divided into classes according to their semantic relationship to
the situation and speaker/writer. One class gives the speaker’s attitude toward
what is said (frankly, honestly); another conveys the speaker’s evaluation of
the text situation (fortunately, surprisingly). Evidential adverbials express the
speaker’s commitment to a proposition (clearly, allegedly, seemingly, probably,
possibly); likely is an adjective of the same type. Adverbs like these can appear
in various positions in a sentence: they are also called “parentheticals.”
(10) a. Surprisingly, tunnels are among the most ancient engineering feats. The
earliest were very likely extensions of prehistoric cave dwellings.
b. Most of us don’t suffer as a result of Darwin’s having eventually attributed
too much scope to the process termed sypatric speciation than it actually
deserves. But a powerful man’s mistaken ideas about women have certainly
caused suffering.
c. The rest of the space was taken up by fancy notepaper, glass paperweights,
fluffy animals and fridge-door magnets shaped like strawberries or Humpty
Dumpty. In the back, though, the setup was very different.
(11) a. The predominant output was the white ware with transparent ivory toned
glaze which made the kilns famous. The other wares were a soft, dark
brown or dense black glazed type, a white ware and a group of lead glazed
earthenwares.
The kilns clearly specialized in the production of the porcellanous ware.
b. Sooner or later, oil prices are likely to drop. But prices at today’s levels
have their advantages.
The adverbs of (10a–b) and parenthetical of (10c) express attitude and evalua-
tion. The adverb and adjective of (11) express evidentiality; they are sometimes
called “speaker-oriented” forms. These expressions are attributed to the speaker
as Responsible Source. The adverbs are classified as Speech Act, Evaluative,
and Evidential adverbs by Cinque (1999). Cinque presents a semantically based
syntactic account of these expressions: he posits a syntactic projection for each
class according to their possible positions in a sentence.
Many evaluative and evidential adverbs have corresponding adjectives. They
differ in the Responsible Source, reflecting a different syntactic status. The
speaker is always the source for adverbs, which are external syntactically and
semantically to the sentence. Evaluative adjectives, however, function as the
main predicate. They are associated with an experiencer, explicit or implicit.
This allows more than one interpretation of responsibility; (12) illustrates:
(12) a. Darwin accepted the postulate of “blending inheritance” (the fusion of a

mother’s and father’s characteristics in their off-spring), even though his
own experiments on pigeons refuted it. Much more surprising are two other
errors.
b. One thing the experts are certain of is that this species, depleted by whaling
and not protected throughout its range until 1966, is showing signs of a
comeback.
c. For a big-city mayor to be so at odds with his party would seem peculiar.
Except that it’s happening everywhere.
In (12a), we infer that the author is responsible for evaluating the situation
as surprising. But in (12b) the experts hold the evaluative attitude. Example
(12c), though vague, includes the author. The modal contributes to the inter-
pretation of subjectivity; modals are discussed below. The verbs seem, appear,
suggest have the same range of interpretation. The speaker is responsible for
the evaluation unless a text participant is explicitly mentioned. For instance, a
variant of (12c) has a sentence-internal interpretation: “For a big-city mayor to
be so at odds with his party seemed peculiar to the voters.” Another difference
between evaluative predicates and adverbials is that verbs and adjectives can
be questioned, whereas adverbs cannot. The related “psychological” verbs, e.g.
surprise, frighten, annoy, etc., are not subjective in the sense developed here:
they do not require that responsibility be ascribed to a mind.
Discourse salience, evidentiality, and “performativity” must all be consid-
ered in understanding evaluative predicates and adverbials, according to Nuyts
(1993). He suggests that evaluatives may be “performative” or “descriptive.”
When used performatively, the speaker/author is an active participant in the
evaluation. They are descriptive when they report evaluation that is part of the
text situation.
Evidentiality is the semantic domain of speaker commitment to what is said.
In English, evidentiality is expressed by adverbs and adjectives, as above; and
by verbs and modal auxiliaries.10 I will sketch here how modals fit into an
account of evidentiality. I cannot offer a full account here of the contribution of
modals to interpretation, nor of the domain of modality; I ignore conditionals
and the modal aspect of generics.
Strictly speaking, modality has to do with necessity and possibility: modal
forms express the necessity or possibility of a proposition. More abstractly, they
express a certain way the world might be, and involve a modal relation and a
10. Some languages have highly developed sub-systems for evidentiality, see Chafe & Nichols
(1986), Willett (1988), Dendale & Tasmowski (2001) for surveys. Kamio (1994) discusses
Japanese.
conversational background for interpretation (Kratzer 1981). There is a useful

distinction among modals between the “epistemic,” expressing possibility and
necessity, and the “deontic,” expressing the will in action in the sense of volition,
permission. In English the same forms, modal auxiliaries, are used for both types
of meaning, leading to systematic ambiguity.
Modals of evidentiality express the position of the speaker toward the propo-
sitional content of the clause. In contrast, modals that pertain to the internal
proposition are not evidential. Certain modals have both the internal and
speaker-related interpretation. Must, for instance, has three interpretations in
(13), an example from Lyons (1977). The first is internal, the other two are
evidential:
(13) Alfred must be unmarried.
On the internal interpretation must functions like the modal operator of logical
necessity on the proposition; it does not express the position of the speaker. The
evidential interpretations can be paraphrased as “I (confidently) infer that Alfred
is unmarried” and “Alfred is obliged to be unmarried.” The two interpretations
correspond to the distinction between epistemic and deontic modality. The
epistemic interpretation ascribes confidence to the speaker. It is “subjective,”
qualifying the speaker’s commitment to the proposition that Alfred is unmar-
ried. The deontic interpretation in this sentence is “objective’: it indicates that,
according to the judgment of the speaker, there is a mathematically computable
chance that the proposition is true.
Subjective and objective modality differ in how the speaker participates in
the evaluation. For instance, on the subjective must reading of (13) the speaker
expresses commitment to an evaulation. The objective interpretation is a factual
assertion expressing the speaker’s judgment about content (Lyons 1982).
Three classes of modals are recognized in Verstraete (2001): internal, epis-
temic, and deontic. Internal modals concern ability and volition and pertain
directly to the proposition expressed. Epistemic modals are evidential and sub-
jective. Deontic modals may be subjective or objective, for instance, “Alfred
may be unmarried” is ambiguous between the speaker’s subjective assessment
and an “objective” assessment of circumstances. Such an assessment, although
objective, is due to the speaker and therefore also evidential.11
11. Subjective and objective evidentials are distinguishable by the criteria of behavior under in-
terrogation and in conditional contexts, and surface ordering, according to Verstraete (2001).
Verstraete may not share my interpretation of objective modal interpretation as evidential. Nuyts
(1993) distinguishes subjective and objective evidentials by their availability for negation, and
other criteria. His work is based on a study of adverbial and adjectival expressions in a corpus
of Dutch discourse. He suggests that the English forms can be explained along similar lines.
The following examples illustrate the classes: (14) has a modal of ability,
internal to the proposition. Sentence (15), repeating (12c), has a subjective,
evidential modal; the modals of (16) are evidential and ambiguous between the
subjective and objective interpretations.
(14) Gabriel went to the stairs and listened over the banisters.
He could hear two persons talking in the pantry.
(15) For a big-city mayor to be so at odds with his party would seem peculiar.
Except that it’s happening everywhere.
(16) a. In Belgrade, Yugoslav President Slobodan Milosevic . . . insisted . . . that the
details must be negotiated directly with the United Nations. That would ap-
pear to fall short of Western conditions for halting the eight-week bombing
campaign.
b. For the body to operate properly, it is crucial that diverse hormones and
receptors produce distinct effects on cells. To achieve such specificity, re-
ceptors must engage in somewhat different behavior.
c. Bass rumbles that could have issued from the lowest octave of a cathedral
pipe organ gave way to plaintive moans and then to glissandos like air
squealing out of a balloon when you stretch the neck taut. With the notes
building into phrases and the phrases into repeated themes, the song may
be the longest – up to 30 minutes – and the most complex in the animal
kingdom.
The preferred reading of (16a) is the subjective, the writer’s assessment.

Example (16b) has only an objective interpretation, (16c) is ambiguous between
subjective and objective assessment. When a modal appears in the complement
clause of a verb of saying, responsibility is ascribed to the subject, as in (17):
(17) Allied divisions over using ground troops in Kosovo burst into the open on
Wednesday when Germany’s chancellor declared that his country would block
NATO from fighting a land war.
Similarly, when a modal is within the scope of a mental State verb, the Respon-
sible Source is the subject of the verb.
Context is often essential in interpreting expressions of subjectivity: the
Responsible Source may be in a preceding sentence. For instance, a mental
state verb holds over subsequent sentences unless it is interrupted by another
subjective indicator or a change of time or place. Example (18) gives a simple
example, a variant of (12a); the evaluative was clear in S2 is in the scope of
believe of S1.
(18) 1. Darwin believed in the postulate of “blending inheritance.”

2. It was clear that the mother’s and father’s characteristics fused in their
offspring.
The predicate “be clear” in S2 implies an evaluator, and in the uninterrupted

context of S1, the evaluator is Darwin. Uninterrupted contexts maintain sub-
jective continuity.
So long as continuity is not interrupted, a subjective form may have scope over
later sentences. Interruptions consist of (a) change of time; (b) change of place;
and (c) other candidates for Responsible Source. These criteria can be stated
formally in terms of time and place information, and whether a sentient being
is introduced into the universe of discourse. Modals and some adverbials are
limited in scope to the sentence in which they appear, unlike other subjective
forms. Stretches of text that are uninterrupted in this sense have “subjective
continuity.”
To illustrate these points I present a rather long text fragment. There is one
participant in the text situation, Sir Mark Sykes. The fragment has expressions
indicating communications of Sykes’, his mental state, and evidentials that
suggest the author as Responsible Source. The subjective forms are marked
according to their source: triggers for Sykes as Responsible Source are under-
lined; triggers for the author as Responsible Source are in bold; triggers for
other Responsible Sources are in italics.
(19) 1 Sir Mark Sykes seems to have started worrying about the Syrian problem the
year before in the context of pledges he intended Britain to keep to her allies –
and her allies to keep to her. 2 His concern was that Syrians might not accept
the Sykes–Picot agreement and the terms outlined by Sir Henry McMahon
to the Sherif Hussein. 3 In 1917 he asked the Arab Bureau [of Britain] to
set up a meeting for him with Syrian Arab leaders in Cairo, apparently in
order to arrive at an agreement with them that would be consistent with the
secret accords with France and with the Hejaz – accords whose existence,
however, he could not reveal to them. 4 He claimed he had succeeded; in his
own hand he noted that “The main difficulty was to manoeuvre the delegates
into asking for what we were prepared to give them, without letting them
know that any precise geographical agreement had been come to.” 5 The
“precise geographical agreement” must have meant the Damascus–Homs–
Gana–Aleppo line that was to be the westward frontier of Arab independence
in Syria under the agreement with al-Faruqi in 1915 and with France in 1916.
6 But reports arrived from various quarters that the Ottoman government
might be planning to pre-empt Arab nationalism by granting autonomy to Syria
immediately. 7 That would leave Britain in the awkward position of sponsoring
the claims of King Hussein as against an indigenous Arabic leadership in
Damascus that threatened to be far more popular in the Syrian provinces.
The list in (20) summarizes the interpretation, giving the trigger for subjectivity;
the Responsible Source (RS), and the scope of the subjective expression. The
RS is the Author or the Subject, Sykes unless otherwise stated. I discuss an

issue of scope at the end.
(20) Interpretation of subjectivity in (19)
S1: “seems,” evidential verb: RS Author; scope, S1 or through S7; “worry
about,” mental state: RS Subject; scope, S1 or entire fragment; “intend,”
mental state: RS Subject; scope, rel. clause headed by “pledges”;
S2: “concern,” mental state: RS Subject; scope, that-clause in S2; “might,”
evidential modal: RS Subject; scope, that-clause in S2;
S3: “ask,” communication: RS Subject; scope, subsequent clauses of S3;
“apparently,” attitude adverb: RS Author; scope, subsequent clauses of
S3; “could,” ability modal: no RS; internal to rel. clause headed by
“accords”;
S4: “claim,” “note,” communication: RS Subject; scope, clauses in S4;
S5: “must,” evidential modal: RS Author; scope, rest of S5;
S6: “might,” evidential modal: RS “reports”; scope, S6, possibly S7.
S7: “would,” evidential modal: RS Author, Subject, or “reports”; scope S7.
There are three candidates for the Responsible Source of the modal in the last
sentence. The first is the Author; the second is “reports,” Subject by extension,
though not a sentient being. On this reading the evidential is contained in the
reports and the scope of “reports” is S6 and S7. The third possible source for
“would” in S7 is Sykes, the Subject of the mental State verb “worry about” in
S1 of the fragment.
The entire fragment is concerned with Sykes’ worrying about a complex
situation. The verb “worry about” in S1 has uninterrupted scope through S7
because continuity is maintained. The fragment does not introduce a Respon-
sible Source that competes with Sykes, nor does it contain any deictic shifts
indicating a difference in the situation.
Evaluation may take other forms as well. Sentence connectives such as yet,
anyway, still, but, for instance, convey the author’s view of relations between
situations. There are also idiomatic forms, e.g. after all as in “So he is coming
after all! (despite our expectation to the contrary)” from Cinque (1999). Like
other lexical expressions, cases like this are beyond the scope of this discussion.
Compositional rules for subjective interpretations like those given above are
presented in 7.6.
7.5 Perception and perspectival sentences

We recognize certain sentences as expressing a perspective that filters infor-
mation through a particular mind. Such sentences depart from the objective
stance, in which perspective is transparent. The central examples of perspectival
sentences involve perception, usually by a participant in the text situation. In

other cases, a particular perspective is indicated without reference to percep-
tion. The linguistic forms of perspective include reflexives, deictics, and forms
of direction or location.
7.5.1 Perception
Linguistic presentation of perception may be direct, contextual, or inferred. The
most straightforward cases are reports of direct perception, in which a verb of
seeing, hearing, etc., introduces a complement which expresses the percept, as
in (21):
(21) a. John saw that the sun was shining.
b. John saw Mary walk to school.
c. John saw Mary walking to school.
The examples illustrate the three forms of perception verb complements in

English: propositional, a “bare” or “naked” infinitive, or gerundive. First-person
reports of perception are subjective, expressing the perspective of the reporter as
a participant, including the “unreliable narrator” of fiction. Not all perception
verb sentences convey direct perception. Complements with that convey not
perception but inference from evidence. For instance, if John saw mud on
Mary’s shoes and reasoned that the mud was due to her having taken a walk,
one could say “John saw that Mary (had) walked to school.” Complements with
a different tense from the main clause are also interpreted this way, e.g. “John
sees that Mary walked to school.”12
Less direct but very clear are perception verb sentences that immediately
precede another sentence expressing the percept. Example (22) illustrates:
(22) Gabriel smiled at the three syllables she had given his surname and glanced
at her. She was a slim, growing girl, pale in complexion and with hay-colored
hair.
The subject of the first sentence is the perceiver; the second sentence conveys
what is perceived. Sentences that convey perception in this way are “perspec-
tivally situated” (Caenepeel 1989);13 they tend to occur in narrative contexts.
12. In a full account of direct perception sentences, one might distinguish perceivable situations
from those that can only be inferred, e.g. “Mary saw John thinking about the race.” Such
distinctions would help to separate the extended meanings of see, hear, feel from the meaning
of direct perception. The extended meaning of see would be something like “understand,”
“infer”, while that of hear would be “have been told that”; etc.
13. According to Caenepeel, perceptual reports or perspectivally situated sentences are always
stative (state or progressive event sentences, in our terms). She claims that events – that is,
event verb constellation with the perfective viewpoint – are impossible or awkward as perceptual
The progressive viewpoint is hospitable to subjective interpretation. In tradi-

tional terms, the progressive and other imperfective viewpoints take an “internal
perspective” on a situation, whereas the viewpoint of perfectives is external
(Comrie 1976). More formally, the progressive focuses an interval that is inter-
nal to an ongoing event; thus its formal meaning is compatible with the interpre-
tation of an experiencing mind. Shifted deictics are clear linguistic evidence for
the internal, subjective interpretation. They are good in progressive imperfective
sentences, somewhat limited in perfectives:14
(23) a. 1 Mary had been working hard all day. 2 Now she was ready to stop.
b. 1 Mary had been working hard all day. 2 Now she stopped.
S2 in (23a) suggests Mary’s perspective. The tense of both clauses is past; the
complement clause has the deictic adverb now, which is normally anchored to
Speech Time; the situation is a state. S2 in (247b), however, is indeterminate
between the perspective of Mary and that of the narrator. Oppositions such as
perfective and imperfective often have the pragmatic function of marking what
is traditionally referred to as “point of view” in narrative (Fleischman 1991:26).
There are also contexts of inferred perception, when the situation implies a
percept due to our knowledge of the world. As with indirect perception more
than one sentence is needed. The examples in (24) illustrate:
(24) a. 1 I sipped my drink and nodded. 2 The pulse in his lean grey throat throbbed
visibly and yet so slowly that it was hardly a pulse at all. 3 An old man
two-thirds dead and still determined to believe he could take it.
b. 1 One night in November 1961, Alice went into the tub room to put some
clothes in her old wringer washing machine. 2 a When she turned on the
light, b there was a rat the size of a small cat sitting on the machine.
reports unless there is a contingency relation between the event and the focalizing sentence.
In a contingency relation, the situation of the perspectivally situated sentence is contingent on
the immediately preceding situation.
However, events are acceptable as perceptual reports in extended, continuous situations of
perception:
a. 1 John looked out the window. 2 Birds flew back and forth and the church bell tolled
loudly.
b. John looked out the window. Mary threw the ball to Sue and Bill played in the
sandbox. The neighbor’s dog arrived and trotted around the yard.
Of course, not all sequences with a verb of perception preceding a sentence of an unbounded
situation express a percept. The second sentence may indicate an entirely different situation:
c. John looked out the window. The children were playing quietly in a corner of the
room.
14. This is particularly clear in French: the imparfait past tense allows shifted deictics more freely
than the perfective past tenses (Banfield 1982, Smith 1991).
Both fragments set up a situation with the subject referent of S1 as participant.

In (24a), we take the sentences after S1 to express the percept of the narrator
as he talks to the old man. The adverb visibly supports, but it is not essential.
S2 might be “The pulse in his lean grey throat throbbed so slowly that it was
hardly a pulse at all”; the inference of perception remains. Example (24b)
has a subtle and more complex subjectivity. In S2a, Alice turns on the light.
World knowledge tells us that she is in a position to see the room, and S2b by
implication presents her percept. The possessive her in S1 is oriented toward
the subject, Alice; see the discussion of example (28) below.
Inferred perception is difficult to establish formally since it depends on in-
ference and world knowledge. It is beyond the scope of compositional rules.
Other kinds of perception can be recognized at the level of grammatical form.
There is an additive effect in sentences which suggest but do not require
subjective interpretation. With one such form, the suggestion may be weak;
with two or more, it is stronger. Examples in (25) illustrate. In these examples,
consider whether the reporter or Mary is the source of the question.
(25) a. Mary played in the sandbox. Huge storm clouds covered the sky. Was it
going to rain?
b. Mary was playing in the sandbox. Huge storm clouds were covering the
sky. Was it going to rain?
c. Mary was playing in the sandbox with her brother. Huge storm clouds were
covering the sky. Was it going to rain?
In (25a) the direct question is the only subjective element; in (25b) the pre-
ceding sentence has the progressive, which invites a subjective interpretation.
The example with the strongest subjective interpretation is (25c), which also
has a possessive phrase oriented to Mary. I do not have a formal account
of inferred perception. It depends on pragmatic factors as well as linguistic
form.
7.5.2 Particular standpoints

Perspective may be suggested by forms that indicate the standpoint of a partic-
ipant, or the reporter. The notion of standpoint has a literal basis in the world.
From a particular location, a person sees in a certain way: if I say that the tree is
nearby, it is because of my position in space. If I say that the bank is around the
corner and you say it is across the street, we can both be right if we are standing
in different places (Mitchell 1986:1). The notion of standpoint can be extended
to situations in which one talks as if one were in a location; and metaphorically
to attitudes and views that are not grounded in space.
Perspectival examples have reflexive pronouns, deictics, the progressive, and

other expressions that suggest a particular standpoint. Example (26) illustrates
with reflexives, based on examples from Ross (1970), Cantrall (1969), Kuno
(1987):
(26) a. This paper was written by Ann and myself.

b. Theyi heard the stories about themselvesi .
c. Maryi put the blanket over herselfi .
The perspective of the reflexives’ antecedent is suggested in these sentences.

In (26a) the speaker must be the antecedent of the reflexive, and is plausible
as participant.15 In (26b–c) the subjects “they” and “Mary” are the reflexives’
antecedents; the reflexives locate the stories and blanket from their standpoint or
perspective. The reflexives are Locally Free Reflexives (LFRs), not syntactically
conditioned.16
When they convey a particular perspective, reflexive pronouns represent a
choice between two possible pronouns. To see the contribution of the reflexive,
compare the sentences of (27). Both are grammatical; the choice of pronoun
affects interpretation:
(27) a. Johni pulled the toy toward himi .

b. Johni pulled the toy toward himselfi .
15. Cantrall was perhaps the first to note the perspectival use of the reflexive. Cantrall presents
many examples, among them the following sentences. Cantrall asks us to imagine that they
describe a photograph which portrays a group of standing women who have their backs to the
camera:
a. The womeni were standing in the background, with the children behind themi .
b. The womeni were standing in the background, with the children behind themselvesi .
In (a) the children are located from the perspective of the speaker; in (b) they are located from
the perspective of the women. As Zribi-Hertz notes, the sentences provide empirical evidence
that the reflexive is correlated with an “internal” point of view – that of a discourse protagonist
as opposed to the speaker (1989:704).
Examples like (27a) led Ross (1970) to suggest that all sentences have a higher clause in
underlying structure with a first-person pronoun and a verb of communication in the present
tense. The overt reflexive pronoun would be coindexed with the covert first-person pronoun.
Zellig Harris reached the same conclusion on a different basis (in Harris 1982), as Bruce Nevin
has pointed out to me.
16. Syntactically conditioned reflexives are obligatory in certain contexts; they are defined in
Government Binding Theory with the notions of c-command and locality of domain. The
Binding Theory requires that the antecedent c-command a reflexive if it is within the domain
of the relevant governing category (Chomsky 1981). The Binding Theory as stated has been
the subject of much critical comment. Reinhart & Reuland (1993) offer an extensive revision
in the same general framework.
Example (27b), with the reflexive, suggests the perspective of John as he pulls
the toy, (27a) does not. Other uses of Locally Free Reflexives are discussed in
Chapter 6.
According to Kuno (1987), the reflexives in these examples convey an “empa-
thy perspective” in which the reporter takes the perspective of another person.17
Empathy may be conveyed by other means as well. For instance, consider the
descriptions of the participants in “Accompanied by his son, the informant
went out to investigate.” The writer expresses empathy with the informant by
describing him independently and the son dependently, as Sanders & Redeker
(1996) observe.
Possessive pronouns may also suggest perspective, often of a participant. In
English this is due to the limited resources of the language. Possessive pronouns
have no reflexive alternant. There is only one form, with the potential for a
possessive, reflexive, or perspectival reflexive reading. The examples in (28)
show possessive pronouns suggesting the perspective of the antecedent: (28a)
is from Kuno (1987), (28b) from Hirose (2000):
(28) a. John criticized his brother.
b. Kazuo lost a book that he borrowed from a friend of his.
17. Perspectival LFRs have been identified in many other languages, among them Japanese, Scan-
dinavian languages, and Italian. Kuno (1987) offers a survey of perspectival and logophoric
phenomena across languages.
Hirose gives examples of two uses of zibun, perspectival and logophoric. The logophoric
involves access to consciousness, the perspectival does not; his term for the latter is “point of
view.” Hirose says that in the logophoric example (a) Kazuo is aware that he is shy, because he
says so. On other other hand in example (b), Kazuo does not have to be aware that the book he
lost is the one he borrowed from his friend. This is shown by the fact that (c) is not contradic-
tory (2000:1646). The examples are reproduced with Hirose’s abbreviations and translations.
TOP = topic, COP = copula, QUOT = quotative, STAT = stative, NOM = nominative case,
ACC = Accusative case, NEG = negation.
a. Kazuo wa zibun wa tereya da to itteiru
K. TOP self TOP shy.person COP QUOT say-STAT
Kazuoi says that hei is shy.
b. Kazuo wa zibun ga tomodati karita hon o nakusit
K. TOP self NOM friend from borrowed book ACC lost
Kazuoi lost a book that hei borrowed from a friend.
c. Kazuo wa zibun ga tomodati karita hon o nakusita ga, sono
K. TOP self NOM friend from borrowed book ACC lost but that
hon ga tomodati kara karita mono da to wa kizuite-it-nai. book NOM friend from
borrowed thing COP QUOT TOP realize-STAT-NEG
Kazuoi lost a book that hei borrowed from a friend but he has not realized that the
book is the one he borrowed from a friend.
Hirose says that “zibun” in examples like (b) and (c) conveys “point of view,” whereas in
(a) “zibun” is logophoric.
The perspectival reading of the possessives is relatively weak in these examples.

They can be read as simply giving information about the relationship of the
participants, or as conveying the perspective of the antecedents John and Kazuo.
If they appeared in a narrative context, however, the perspectival readings might
be stronger.
Recognizing differences in strength among the relevant examples, Kuno
posits a continuum of “degrees of empathy.” At the high end of the continuum
the reporter totally identifies with a participant; in the middle the identification
is partial; at the low end the reporter manifests a total lack of empathy with
participants. Perspective may also be suggested by adjectives or epithets that
would be expected from the participant, e.g. “John talked to Mary about his
beloved cat.”
Deictic adverbials can indicate the perspective of a participant. When a deictic
neutrally anchored to Speech Time is reanchored to the past or future, a personal
perspective is suggested strongly. Compare (29a) with (29b–c):
(29) a. Mary lost her watch three weeks ago.

b. Mary had lost her watch three weeks ago.
c. Mary packed her clothes. She would be leaving soon.
In (29a) the deictic “three weeks ago” is anchored to the time of speech; there
is no subjectivity. But (29b–c) suggest the perspective of Mary, the sentence
subject, with deictics anchored to a past time.
The rather delicate interpretation of perspective can be partially accounted
for by compositional rules. Reanchored deictics can be recognized by rules
which look at the tense and deictic adverbial in a clause. If the two are not in
accord – for instance, a past tense and Speech-Time-oriented deictic – and there
is an appropriate Responsible Source, then the deictic is perspectival. Locally
Free Reflexives (LFRs) could in principle be recognized by a rule with access
to the principles of the Binding Theory. If a reflexive has no Binding Theory
antecedent, it is an LFR. However, it need not be perspectival: recall that LFRs
have other functions, noted in Chapter 6. I will not state a compositional rule
for perspectival LFRs.
7.6 Formalizing the interpretation of subjectivity

7.6.1 The approach
More than one linguistic form contributes to subjectivity. This diversity is es-
sential to the composite analysis. Example (30) gives the forms that contribute
to a subjective interpretation in alphabetical order:
(30) Linguistic forms contributing to subjectivity

Communication verbs
Deictic adverbials: place, time
Direction and location PPs
Epithets
Evaluative verbs and adverbs
Evidential adverbials, adjectives, and verbs
Experiencer predicates
Possessive pronouns
Progressive viewpoint
Propositional complements
Situation type: State and activity sentences
The list is not complete but it gives a sense of the many and varied forms
involved. Among the forms ignored here are conditionals and counter-factives;
the subjunctive; conjunctions such as yet, anyway; verbs and PPs with a deictic
component such as come and go, toward and away from. Forms of comparison
and evaluation also indicate subjectivity (Hunston & Thompson 2000); since
they are conveyed lexically they are beyond the scope of this discussion.
The interpretation of subjectivity requires that a Responsible Source be iden-
tified. To account for subjectivity in Discourse Representation Theory, I write a
set of construction rules that introduce the Responsible Source as a conceptual
role. The rules associate this role with the appropriate individual entity, on the
one hand, and with a sub-structure of the representation on the other. The output
of the rule is encoded in the Discourse Representation Structure (DRS).
The construction rules recognize as Responsible Source either the Author
or the Subject, a sentient being in the text situation. When the Author entity
is needed, a special discourse entity is constructed by the rule. The Subject is
usually the subject of a sentence, but may also be the direct object or the object
of a preposition. In clauses with subjective forms the rules look for NPs that
have the feature [+human] in appropriate syntactic relation as candidates for
Subject. Most of the features of subjectivity are needed anyway in a grammar
of English, e.g. the noun feature [+human]; the verb classes of communication,
mental state, and perception; pronoun forms; modals; tense, etc.
Subordinate structures in a DRS model the relation between the Responsible
Source and the relevant material. The information in the scope of a subjective
form appears in a sub-DRS. This correctly identifies subjectively introduced
information, and provides a special status for it. The approach is consistent
with treatments of propositional attitude verbs and modality in Discourse Rep-
resentation Theory since the mid 1980s by Kamp (1985), Roberts (1987), and
Frank & Kamp (1997). I also draw on the DR Theory account of logophoricity
in Stirling (1993). All of these analyses use subordination: material in the scope
of mental state verbs, modal operators, and other forms of subjectivity appears
as a sub-DRS. The mental space theory of Fauconnier (1985) takes essentially
the same approach.
Thus, for instance, a mental state verb will trigger a sub-DRS. The subject
of the verb is identified with the role of Responsible Source. This appears as a
condition in the DRS. The information in the complement of the verb is in a
sub-DRS that is associated with the Responsible Source.
Before introducing the compositional rules, I give two examples of DRSs
with expressions of subjectivity. The first DRS interprets a sentence with the
mental state verb believe and a complement clause: “Mary believes that John
won the race.” By rule, the main clause subject is interpreted as Subject and
Responsible Source; the resulting condition is entered in the DRS (lines 3, 4).
The contents of the belief appear in a sub-DRS associated with the verb. In
a belief sentence like this, individuals are independent of the sub-DRS. The
individual “John” therefore appears at the top of the DRS and as a discourse
referent a in the sub-DRS. In the sub-DRS a is identified with John as a con-
dition (line 6). The DRSs are radically simplified, including only the relevant
information.
(31) Mary believes that John won the race
x y e
1. x = Mary
2. y = John
3. x = Subject
4. x = RS
5. e: believe, x . . .
RS
a b e’
6. a = y
7. b = the race
8. e’: win (a,b)
In some other types of subjectivity all the individuals within the scope of the
subjective form are independently posited and thus appear at the top of the
DRS.
The next DRS interprets a sentence with a modal, for which the Author (A)
is Responsible Source (RS). The Author is introduced as a discourse entity at
the top of the DRS, and is interpreted as RS (lines 2 and 3).
(32) Mary may win the race.
x e A y
1. x = Mary
2. A = Author
3. A = RS
4. y = the race
5. e : Mary may. . .
RS may
a b e’
6. a = x
7. b = y
8. e’: win (a,b)
The material within the scope of the modal is within the sub-DRS, as in the pre-
vious example. However in this case both individuals in the sub-DRS, “Mary”
and “the race,” are posited independent of the modal operator. The formal inter-
pretation of the modal would also be included in a complete DRS. Additional
DRS interpretations of subjectivity will be presented in Chapter 10.
7.6.2 Compositional rules

This section presents compositional rules for interpreting subjectivity that re-
sult in DRSs like those above. The rules are organized into the categories of
subjectivity introduced in this chapter: communication, contents of mind and
evidentiality, perception, and perspective.
7.6.2.1 Rules for subjectivity I: communication

These rules interpret quoted speech, indirect speech, and direct address to the
audience. Represented speech is not included. For quoted speech, the rule recog-
nizes as Responsible Source the sentient being to which the quote is attributed:
Rule 1. For indirect speech, recall that recoding by the Author is always a
possibility. The rule recognizes two Responsible Sources, the Author or the
person whose communication is reported, the Subject. Indirect speech typically
has a verb of communication; a that-complementizer introducing the com-

plement; the past tense in main and complement clause; shifted pronouns
and adverbials. Together, these forms trigger the interpretation of indirect
speech.
I give only the basic rule: Rule 2. The rule does not provide for special cases.
Special interpretation cases include logophoric clauses, where the Subject is
coreferential with the subject of S2; and clauses where the deictic forms in S2 are
proximal rather than distal; and the interpretation of double-access sentences,
with the past tense in the main clause and present tense in S2. Other rules would
be needed for sentences where the Subject is not in subject position.
Direct questions, imperatives, and direct addresses to the audience introduce
the entity Author as Responsible Source. This constitutes the interpretation of
subjectivity. The rules use the abbreviations in (33); they allow for verbs or
nominals of the classes discussed above, for instance, the verb believe or the
nominal belief :
(33) Abbreviations
Animate NP picked out by rule = NPx
Referent of NPx, usually subject or experiencer = Subject
Speaker/Author = Author
Responsible Source = RS
Ability, obligation modal = Modc
Communication verb or nominal = V/Ncom
Evidential evaluative modal, verb, adverb, adjective V/Mod/Adj/Adve
Mental State Verb or nominal = V/Nme
Perception verb or nominal = V/Npe
The rules are listed below. Each rule has 3 parts: (a) a verbal statement of the
surface structure that the rule recognizes, with examples in italics; (b) the formal
rule itself – X, Y indicate optional additional material that is not relevant to a
rule; (c) a statement of the interpretation the rule makes.
Rule 1: Quoted Speech

(a) Subject (NPx) is subject of VCOM , quotes in complement clause, Subject =
RS
Kim said “I’m ready to go home.” Kim said “John went home.”
(b) S1[NPx VCOM [“X”]] → Subject = NPx; Subject = RS (S2)
(c) Rule 1 interprets NPx as Subject, and Subject as the RS for the quoted
material.
Rule 2: Indirect Speech

(a) NPx subj of VCOM , complement clause, Subject and/or Author = RS
Kim said that he was ready to go home.
(b) S1[NPx VCOM tns [past] S2[that [X.tns [past] (Adv[+distal])]]]

→ Author; Author = RS (S) Subject = NPx; Subject = RS (S2)
(c) Rule 2 interprets NPx as Subject, introduces the entity Author, and inter-
prets both Author and Subject as RS.
Rule 3a: Direct Question

(a) Question form not involving a text participant: Author = RS
What will happen in five years time? Will anyone remember this incident?
(b) S [(WH) aux NPx Y Z] → Entity Author; Author = RS (S)
(c) Rule 3a recognizes a direct question, introduces Author, and interprets
Author as RS.
Rule 3b: Direct Imperative

(a) No surface subject, no text participant: Author = RS (S)
Think about this carefully.
(b) S [e tns [e] (aux) V X] → Author; Author = RS (S)
(c) Rule 3a recognizes a direct imperative, introduces Author, and interprets
Author as RS.
Rule 3c: Direct Address

(a) 2nd-person pronoun; Author = RS
You may find that a difficult question. Everybody’s worried about George
Bush except you.
(b) S [X NP [2nd p] Y] → Author; Author = RS (S)
(c) Rule 3c recognizes a sentence containing a 2nd-person pronoun in subject
or another syntactic position, introduces the entity Author, and interprets
Author as RS.
7.6.2.2 Rules for subjectivity II: contents of mind and evidentiality

These rules interpret sentences with verbs of mental state; and sentences con-
taining evaluative, evidential, and modal expressions.
The last rule accounts for cases where a subjective form has no Responsible
Source in the sentence, but there is a relevant RS in the prior context. If continuity
is maintained so that the form is within the scope of V/Nme verb in the preceding
context, the RS for that verb is also the RS for the sentence in question.
Rule 4: Mental state
(a) Subject (NPx) is subject of Vm e, Subj = RS
John believed that the earth is flat.
(b) S1[. . . NPx V/Nme . . . S2[(that) . . . (modal) . . .]]
→ Subject = NPx; Subject = RS (S2)
(c) Rule 4 interprets NPx as Subject, and interprets Subject as the RS for S2,
the clausal complement. If there is a modal in the complement, the Subject
is RS for that modal.
Rule 5: ModalE
(a) V/Modale not in the scope of V/NCOM ; entity Author; Author = RS
The earth may be flat. Mary may be very clever. Mary may think that she
is clever.
(b) S[. . . NP V/Nm . . .] → Author; Author = RS (S)
(c) Rule 5 introduces the entity Author, and interprets Author as RS for the
sentence. The rule doesn’t need to know whether there is a potential Sub-
ject in the clause.
Rule 6: V/Modc
(a) NPx sentient being, subj verb/NP, Modalc, Subject = RS
John could not see the ship on the horizon. RS: John
(b) S[. . . NPx Modc[VP]] → (Modc(VP))
(c) Rule 6 interprets the modal as modifying the VP. Such modals are internal,
with no RS and without intertriggering a sub-DRS.
Rule 7: Evaluative/Evidential + Experiencer
(a) Subject (NPx), PP obj of V/Adje; Subject = RS
It seemed to John that Mary had left. That Mary left was surprising to
John.
It seemed an anomaly to John.
(b) S[NP . . . Adj/Ve. X PP[to NPx]] (Y) → NPx = Subject; RS = Subject
(S)
(c) Rule 7 interprets the NPx of a to-PP as experiencer subject and RS in the
context of an evaluative or evidential predicate.
Rule 8: Evaluative/Evidential
(a) Adj/Vex; no experiencer subject or PP; Author = RS
It is evident that the earth is flat. Evidently, the earth is flat.
(b) S[NP . . . Adj/Ve . . .] → Author; Author = RS (S)
sentence.
Rule 9: Evaluative/Evidential Adverbial
(a) Adverbe; entity Author; Author = RS
Obviously, the earth is flat.
(b) S[{Adve} NP {Adve} vp[X . . .] {Adve}] → Author; Author = RS (S)
sentence. The braces give the three main positions for adverbs of this class:
initial position; before the VP; final position. Only one position is filled,
so the braces mean “choose one.”
Rule 10: Contextually Licensed V/Mode, No Candidate NPx for Subject
(a) In the scope of V/Nm e in the context.
Darwin believed in the postulate of “blending inheritance.” It was
clear that the mother’s and father’s characteristics fused in their
offspring.
(b) [S-[NPx Vme X] Y S[NP V/Advex X]]

Condition: in a context without adverbs of time or place, or a new sentient
being
→ Subject S- = RS (S)
(c) Rule 10 applies if no potential RS appears in a sentence and a sub-
jective form is within the scope of V/Nme in the context, and there is
nothing that interrupts continuity (see the comments on example (19)).
The rule interprets the RS for V/Nme as the RS for the sentence in
question.
7.6.2.3 Rules for subjectivity III: perception and perspective

These rules deal with sentences that present direct and contextually licensed
perception, and particular standpoints or perspectives. I do not include inferred
perception, which is based on pragmatic inference rather than linguistic form;
nor empathy perspectives with LFR reflexives or other forms. LFR reflexives
are excluded because they can express emphasis and contrast as well as sub-
jectivity; the interpretations that I have developed are not sufficiently delicate
to distinguish these cases.
The rules for perception are similar to those for contents of mind. There are
some differences. Perception involves a different class of verbs, and some rules
require that the situation expressed be unbounded (a state, a progressive, or a
perfective Activity). The rules do not interpret this notion; I assume that it would
be available as part of the general principles for interpretation. Rule 11 is like
Rule 4 above, ascribing the contents of the complement clause to the Subject;
but the complement is limited to certain forms that convey direct perception.
Rule 12, like Rule 10a, gives a contextually licensed interpretation, requiring a
situation without endpoints, stated as a disjunction.
Rule 13 interprets sentences with perspectival subjectivity. The rule recog-
nizes shifted deixis: it applies to sentences that have past tense and a deictic that
neutrally anchors to Speech Time. The rule allows for all kinds of situations,
rather than limiting the possibilities as Rule 12 does.
Rule 11: Direct Perception

(a) Subject = NPx subj of Vpe, complement clause, Subject = RS
John saw Mary opening the door
(b) Conditions: S2 does not contain that, nor tense.
S1[. . . NPx Vpe . . . S2[NP V(ing) X]] → Subject = NPx; Subject = RS
(S2)
(c) Rule 11 interprets NPx as Subject, and interprets Subject as the RS for
S2, the clausal complement.
7.7 Summary and conclusion 183
Rule 12: Contextually Licensed Perception

(a) No candidate NPx for Subject; in the scope of Vpe in the context. The
situation must be unbounded.
Gabriel glanced at Mary. She was a slim, growing girl.
(b) Condition: the context has no adverbs of time or place, nor a new sentient
being.
[S-[NPx Vpe X] Y S[e/s unbounded]] → SubjectS = RS (S)
(c) Rule 12 applies if a sentence is within the scope of a Vpe verb in the
context, and there is nothing that interrupts continuity.
Rule 13: Perspective, Shifted Deictic
(a) Sentences indicate the perspective of NPx.
Mary was tired now. John would be leaving in three weeks. Mary had lost
her watch a week ago.
(b) S[NPx tns[past] Deictic[SpT]] → Subject = NPx; Subject = RS (S2)
(c) Rule 13 interprets NPx as Subject, and interprets Subject as the RS for
S2, the clausal complement.
These rules indicate the interpretation of subjectivity in Discourse Representa-

tion Structures.
7.7 Summary and conclusion

Sentences have a dimension of subjectivity when they express a point of view,
or take a perspective on a situation. Communication sentences involve public
events, while contents of mind and evaluation express private events or mental
states. The perspectival category includes perceptual reports and indications
that a situation is viewed from a particular standpoint. Perspective is conveyed
by verbs of perception, indirect cues that perception is involved, reflexives that
are not syntactically conditioned, and deictics.
In interpreting subjectivity we ascribe responsibility for a clause to the Re-
sponsible Source, either the Author or the Subject participant in the text situa-
tion. Perspectival sentences suggest the perspective of a participant in varying
degrees; in the weaker cases, we may ascribe the perspective to the author.
The subjective interpretation may extend over several sentences, so long
as subjective continuity is maintained. Subjective continuity is interrupted by
changes of time or place, or introduction of a new candidate for Responsible
Source.
Compositional rules recognize expressions of subjectivity and ascribe them
to the appropriate Responsible Source. Subjectivity is encoded by embedding
in the DRS.
(5), (14), and (23) James Joyce. “The Dead.” In Dubliners, 1916; reprinted London:
Penguin Books, 1982, pp. 179, 182, 177.
(7) Alan Ehrenhalt, Hijacking the rulebook. New York Times, December 20, 1998.
(10a) Jim Collins. How it works. US Airways Magazine, Attaché, May 2001.
(10b), (12a) and (19) Jared Diamond. A tale of two reputations. Natural History, Febru-
ary 2001.
(10c) Peter Robinson, A Necessary End. New York: Avon Books, 1989, p. 181.
(11a) Margaret Medley, The Chinese Potter. London: Phaidon, 1989, Ch. 5.
(11b) Robert Mosbacher, Cheap oil’s tough bargains. New York Times, March 13, 2000.
(12b) Douglas Chadwick, Listening to humpbacks. National Geographic, June 1999.
(12c) and (15) Peter Beinart, The pride of the cities. New Republic, June 1997.
(16a) and (17) Kosovo strategy splitting NATO. New York Times, May 1999.
(16b) John Scott & Tony Pawson, Cell communication. Scientific American, June 2000.
(19) David Fromkin, A Peace to End All Peace, New York: Henry Holt, 1989, p. 329.
(24a) Raymond Chandler, The Long Goodbye, Boston: Houghton, 1964, Mifflin, p. 64.
(24b) J. Anthony Lukas, Common Ground, New York: Knopf, 1985, p. 149.
8 The contribution of surface
presentation
Sentences that differ in arrangement and sentence accent also differ in meaning,
although their propositional content may be the same. As Bolinger puts it,
discussing the difference between active and passive (1977:9):
The classical case is the passive voice. If truth value were the only criterion,
we would have to say that John ate the spinach and The spinach was eaten
by John are the same. They report the same event in the real world. The same
entities are present, in the same relationship . . . Linguistic meaning covers a
great deal more . . . [it] expresses, sometimes in ways that are hard to ferret
out, such things as what is the central part of the message, what our attitudes
are toward the person we are speaking to, how we feel about the reliability of
our message, how we situation ourselves in the Events we report, and many
other things.
The meanings that Bolinger talks about involve the way a text presents infor-
mation, its presentational structure.
Surface structure presentation instructs the receiver about how to organize
the information in a sentence. The sentences of a text are not undifferentiated
wholes, nor simple linear arrangements of words. I adopt the approach to pre-
sentation originally put forth by Prague School linguists and further developed
in recent years. This approach uses the notions of communicative dynamism,
topic–comment, and focus–background, to understand the internal organization
of sentences and how they are deployed in texts. Although “topic” is notori-
ously difficult, I hope to show that it contributes an important dimension to the
analysis of sentences and texts.
The writer’s choices of how to present material are influenced by assessment
of what is accessible to the receiver. The familiarity status of information is
a key factor. This notion was introduced in Chapter 6, where it was shown
that the forms of referring expressions used in a text tend to correlate with
familiarity of the referent. Familiarity status concerns whether information is
discourse-old or discourse-new, hearer-old or hearer-new. What is new to the
discourse needn’t be new to the hearer; but discourse-old is also hearer-old.
185
186 The contribution of surface presentation
The distinction between discourse-old and discourse-new information is most

useful, as noted in Chapter 6. Familiarity status is intertwined with the notions of
topic and focus, and imposes constraints on an important class of non-canonical
syntactic structures.
Language allows propositional information to be presented in different ways.
For instance, in English one may choose an active sentence such as “Mary drew
a circle” or its passive counterpart, “A circle was drawn by Mary.” The object
may be preposed, as in “A circle Mary drew”; or “extracted” in a pseudo-cleft or
cleft sentence, “What Mary drew was a circle,” “It was a circle that Mary drew”;
etc. These sentences express the same proposition, yet they convey somewhat
different messages. The study of presentation explores the differences. Linear
order, syntax, morphology, and prosody are the linguistic factors that vary in
presentation.
Choices of structure and prosody depend on the speaker’s intent and assess-
ment of the receiver’s current state of knowledge and attention. The main factors
in presentation are the topic–focus structure of a sentence and the familiarity
status of information. They are independent, but coincide in many cases: topics
tend to be familiar, while focused information tends to be new.
In this chapter I discuss the notions of topic and comment, focus and back-
ground for English. They lead to different but overlapping partitions of a sen-
tence; I will argue that both are needed. The discussion of topic is most extensive
since it is key to presentational progression as I develop it. The terms “topic”
and “focus” and the ideas behind them appear in one form or another in many
studies. I do not offer a full account of different theories of presentation, or
information packaging as it is sometimes called. Nor do I present a complete
theory of my own. Accomplishing these goals would require detailed discus-
sion of prosodic information, and of languages of different structural types. Both
are beyond the scope of this book. This chapter discusses canonical sentences,
and contrast and emphasis; non-canonical, or marked, syntactic structures are
discussed in Chapter 9.
There is a sense in which receivers progress through the passages of a text
according to the topic and marked focus of its sentences. This “presentational
progression” complements the text progression of Discourse Modes discussed
in Part II of this book. The duality of progression contributes to the richness of
texts.
Section 8.1 gives introductory comments on presentation; 8.2 discusses
topic–comment; 8.3 considers Focus–Background; and 8.4 presents a com-
bination of the two, in dual partitioning; 8.5 concludes.
8.1 Presentational factors 187
8.1 Presentational factors

Sentences have a dynamism of their own. There is a progression in the sen-
tence from its starting point to the end. The beginning is the point of depar-
ture, from which the receiver continues on to the final phrase, the goal of the
sentence as a communication. This linear development is called ‘communica-
tive dynamism’ by linguists of the Prague School: “The first element has the
lowest degree [of dynamism] . . . the reader progresses from the beginning to
successively more dynamic elements of the sentence” (Daneš 1974).1 The
least dynamic element tends to be information that is familiar. The most dy-
namic element is new, or significant in an unexpected way. It is known as the
focus phrase, and receives the main sentence accent. This is a general pat-
tern across languages. Variations on the basic pattern convey subtly different
messages.
The information in a sentence is organized in terms of a topic – what the sen-
tence is about – and a focus phrase that carries the most significant information.
This organization is reflected in the linear order of most sentences. The linear
order reflects the communicative dynamism of a sentence, going from least to
most dynamic, with topic phrase at the beginning and focus phrase at the end.
The pattern reflects the psychological principles of primacy and recency. The
speaker or writer initiates a sentence with shared information. This enables the
receiver to establish a mental address to which the new information that arrives
later can be directed (Prideaux 1993:54).
The approach assumes a basic systemic order in language. I take it that the
neutral, canonical order in English is [X-Subject-Verb-(Object)-(Y)]. “Object”
denotes subcategorized complements of the verb; “X” denotes parentheticals
and sentence adverbs that appear at the beginning of a sentence (Rizzi 1997,
Cinque 1999); “Y” denotes other adverbs and optional adjuncts. Non-canonical
sentences are marked structures that depart from the neutral. I will assume that
sentences with non-canonical order arise by movement, though I do not state
1. The Prague School is a general term referring to a group of linguists in Central Europe who
worked in the structuralist tradition, first in the 1930s and early 1940s, later in the 1960s
and 1970s. Their ideas are enjoying something of a resurgence; see Hajičová et al. (1995),
Hajičová et al. (1998). The notion of topic and comment can be traced to Henri Weil, a French
classical scholar who published in 1840, to German scholars of the late nineteenth century, and
to Mathesius; see Firbas (1974) for discussion.
The approach presented here owes a great deal to the latter-day Prague School but is not
identical with it. I recognize both topic and focus phrases in most sentences, whereas the
traditional partition consists of focus and background.
movement rules here.2 When a phrase is displaced to the end of the sentence
it becomes the focus phrase and receives the sentence accent, by the “ordering
principle” of Hajicová and Sgall (1987). Alternatively, another phrase may be
highlighted with a contrastive accent.
In the spoken language, prosody is an important cue to the topic and focus
phrases of a sentence. The intonation center, or sentence accent, occurs canon-
ically on the rightmost stressable unit of the sentence, the focus phrase.3 The
scope of the accent is indeterminate: it may be limited to a single phrase or
extend to larger constituents. I assume that prosodic structure is a reflex of
semantic and pragmatic information (cf. Gussenhoven 1983, Erteschik-Shir
1997, Steedman 2000, and many others). Prosody will not be discussed, since
this study is devoted to written texts.
The status of information as more or less familiar plays an important role
in presentation. The familiarity of a referent is often coded by morphology:
pronouns tend to have referents that are familiar and definite NPs to have iden-
tifiable referents. For discourse reasons a definite NP may be used for a referent
that is already familiar, as noted in Chapter 6.
The notions of topic and comment, focus and background, represent different
partitions of a sentence. Topic–comment recognizes a topic referent, what the
sentence is about; the rest of the sentence is comment. The focus–background
partition distinguishes the main contribution of a sentence in a focus phrase,
and the rest of the sentence functions as background for that focus. To cover the
full range of possibilities, both are needed. The apparent contradiction between
them is resolved with dual partitioning. I discuss each of these matters in turn.
8.2 The topic–comment partition

The topic of a sentence is what the sentence is about. In the topic–comment
partition, a referring phrase and its referent are singled out as the topic; the
remaining part of the sentence is comment. Speakers have intuitions about sen-
tence topics, although there are some unclear cases. The intuitive notion of topic
is necessary for languages like English. Some languages have a syntactic topic
2. Movement rules are implied by the notion of a canonical sentence order. Grammatical rules
that move phrases to non-canonical positions are essential to the approach of transformational
generative grammar.
3. The typical pitch contour of sentence accent is high and falling, written H∗ +L in Pierrehumbert’s
(1980) notation (an A accent in Bolinger’s system, 1972). Many sentences have a secondary
intonation peak, typically with a rising pitch contour, written L+H∗ (Bolinger’s B accent).
These notational systems do not bear on the question of how stress and intonation relate to the
syntactic and semantic components of the grammar.
position, e.g. Hungarian and Mandarin Chinese; others have topic morphology,
e.g. Japanese and Korean.4 Topics in such languages may differ from topics in
English.
Subsequent sections discuss the notion of aboutness (8.2.1); the position of
topic phrases in sentences (8.2.2); sentences without topics (8.2.3); different
notions of topic (8.2.4); and determining sentence topic in texts (8.2.5).
The notion of aboutness organizes the information conveyed by a sentence.
We understand that a sentence says something about X, the topic referent. Thus
the topic determines how the truth of a sentence is assessed: we ask “Is it true
of X[topic] that Y?” The topic phrase is canonically the subject of a sentence,
as in (1):
(1) A. Beavers build dams.
B. Sue visited her cousin last week.
People tend to understand (1a) as saying something about the class of beavers
and (1b) as saying something about Sue, unless there is information to the
contrary.
The topic–comment partition mirrors the classic division of a sentence into
“subject” and “predicate,” also known as “theme” and “rheme.” The subject and
predicate are distinct syntactic constituents in traditional accounts. The division
determines interpretation: a sentence makes a comment about the subject ref-
erent, and the comment is expressed by the predicate. The topic of a sentence
indicates how the information it conveys should be added to the developing
mental model of the discourse.
The context of a sentence usually points to a particular topic interpretation.
The pairs in (2) present different contexts for the second sentence; in both
“Mary” is topic, the subject of immediate concern.
(2) a. i. What did Mary wear?
ii. Mary/she was wearing an enormous hat.
b. i. I saw Mary yesterday at the movies.
ii. Mary/she was wearing an enormous hat.
Questions and answers like (2a) are often used to demonstrate sentence topics.
The felicity of this pair contrasts with another in which Mary is not topic of the
second sentence, for instance a cleft sentence like “It was MARY who wore an
enormous hat.”
4. Studies show that languages with syntactic topics differ substantially from those without them;
cf. Li & Thompson (1976), Kiss (1986), Portner & Yabushita (1998), Vallduvı́ & Vilkuna (1998),
etc.
Paraphrase can test and support the intuition that a given phrase is topic. With
paraphrase tests, one recasts a sentence with an explicit introductory phrase
and the putative topic phrase (as for, speaking of ). The result is felicitous if the
phrase in question is indeed the topic. For instance, supporting the intuition that
the (ii) sentences are about Mary, one can say “As for Mary, she was wearing
an enormous hat” and “Speaking of Mary, she was wearing an enormous hat.”
Both are reasonably felicitous. Another paraphrase test uses a verb of saying, as
in “He said about Mary that she was wearing an enormous hat” (Gundel 1974,
Reinhart 1982). There are well-known difficulties with paraphrases as actual
tests for topic; they are both too weak and too strong.5 Nevertheless the tests
can be helpful in illustrating and identifying the topic of a sentence.
8.2.1 Aboutness
The intuition of aboutness is often taken as primitive. Strawson and Reinhart,
however, offer useful discussions.
The notion of aboutness explains how the information of a sentence is un-
derstood and assessed in context, according to Strawson. Aboutness relates a
sentence to a matter of concern in the context: “We do not, except in social des-
peration, direct isolated and unconnected pieces of information at each other,
but . . . intend to give or add information about what is a matter of standing
5. Reinhart gives the following pairs. Only the first is felicitous:

Felix is an obnoxious guy. Even Matilda can’t stand him.
Felix is an obnoxious guy. As for Matilda, even she can’t stand him.
The second pair is odd because the topic that is structurally identified does not correspond to
the intuitive aboutness topic.
The paraphrase tests are too weak because they work for non-topics; too strong, because
they aren’t always appropriate for a topic. Moreover, as for phrases do not always introduce
the topic referent of a sentence, as the fragment below illustrates. Under discussion are various
types of fuel, listed early on as natural gas, nuclear power, and coal. The phrase of interest is
italicized:
Despite current capacity and price problems, clean-burning natural gas should be rec-
ognized as America’s best source of energy for the future . . . All Americans should
support construction of the proposed transcontinental natural gas pipeline, following
the route from Alaska’s north slope to the 48 States.
As for nuclear power, we have not yet solved the intractable problem of where to
store the thousands of metric tons of nuclear waste scattered across the country.
Here “as for” introduces a very general referent (“nuclear power”) that signals a change in the
direction of the discourse. But the introducing phrase is not coreferential with any referent in
the following sentence. It’s more like a discourse or “Chinese-style” topic: “As for fruit, I like
watermelon” (Li & Thompson 1976). The example shows that as for does not always introduce
a sentence topic.
current interest or concern” (1964:97). Statements are assessed as information

about their topics. In determining the truth of an assertion, the topic plays a key
role: one asks whether an assertion is true with respect to its topic. On the one
hand, a topic has an anchor in context. On the other, the topic organizes the
understanding and assessment of a sentence.
The concept of common ground adds another dimension to the notion of
topic, as Reinhart shows. The common ground consists of the “context set” of
propositions, the information of a developing discourse. This can be a large
body of information. If unstructured, it would be difficult for a person to use.
Reinhart suggests that the context set is organized along the lines of a sub-
ject catalogue in which sentence topics are referential entities which classify
propositions. In constructing the context set, “the two procedures, assessing and
storing information, are relativized to topics” (Reinhart (1982):24). The idea
appears in various guises in later work on topic.
In this approach sentences have many potential topics. The propositional
information of a sentence may be relevant to context in different ways, each
corresponding to a different topic interpretation. The sentences of (3) illustrate,
from Reinhart (1982). The first, (a), presents an assertion; (b–c) give different
topic interpretations of (a), each with an introductory topic phrase and additional
reinforcing material:
(3) a. It’s no wonder that Carter is considering withdrawing the American athletes
from the Olympic games.
b. As for the Olympic games, it’s no wonder that Carter is considering
withdrawing the American athletes from them (because they are such a
farce).
c. As for Carter, it’s no wonder that he is considering withdrawing the Amer-
ican athletes from the Olympic games (because he is such a hard-liner).
Reinhart formalizes her proposal for multiple potential topics in terms of pos-
sible pragmatic assertions. In a given context one assertion is selected from the
set.
The topic of a sentence affects truth-conditional semantic interpretation. A
definite NP triggers a presupposition of existence when it is the topic and subject
of the sentence, but not otherwise (Strawson 1964, Hajičová 1971):
(4) a. The king of France is bald.

b. Yesterday Prague was visited by the king of France.
Example (4a) presupposes that there is a king of France. Example (4b) has
no such presupposition, as standard tests such as denial and questioning show.
Another point is that, in sentences with complex quantifiers, the phrase in

topic/subject position has an interpretation not available to other phrases.
(5) a. Everyone in this room knows at least two languages.

b. At least two languages are known by everyone in the room.
The wide-scope reading for “at least two languages” is available for (5b) but
not for (5a). These phenomena are well known, but not always ascribed to
topicality.
Other evidence that sentence topics affect semantic interpretation comes from
work in formal semantics. In if/when sentences with adverbials of quantifica-
tion, the topic phrase determines which indefinites are bound by the adverbial
(Chierchia 1992).6 The notion of sentence topic is part of the interpretation of
quantifiers, according to Partee (1991) and Büring (1999). Topics are modeled
as a presupposed salient set of alternatives, arrived at by constructing implicit
questions by von Fintel (1994) and McNally (1998). In a study of Japanese,
Portner & Yabushita argue that topics affect semantic interpretation in comput-
ing entailments, implicature, and certain scopal phenomena (1998). Their work
is suggestive, though Japanese topics differ from English topics in more than
one way.
Sentence topics are often demonstrated with question–answer pairs. As in
(2a) above, the topic of a sentence is clear in the context of a particular question.7
The observation has been generalized as a basic organizing principle of texts.
Following this principle, one looks for the question that a sentence answers in
a given context. If no question actually appears in a text, the analyst constructs
an implicit question. For instance, van Kuppevelt uses implicit questions to
construct an interpretation “which does not differ in acceptability and coher-
ence” from a text without them (1995:116). He claims that the approach gives
an operational way of characterizing topics.
8.2.2 Topic phrases in sentences

Typically, the topic referent of a sentence is in subject position and represents
familiar information. The subject of a sentence is prominent positionally and
6. Chierchia (1992) integrates the treatment of NPs with the theory of generalized quantifiers. He
outlines the “proportion problem,” which concerns the scope of the quantificational determiner
in “donkey sentences” and conditionals, and the scope of quantificational adverbials in if/when
sentences.
7. The question approach to topic was used by Daneš (1974), as well as Vennemann (1975), and
others. For the latter two the topic is identified with one of the presuppositions defined by a
question.
grammatically. As first element it links directly to what precedes and by exten-

sion to the common ground; it is the starting point for the communication of a
sentence. The subject is usually the perspectival center, the reference point from
which the material of the sentence is presented. All sentences of English have
a subject, usually overt. The last element in the sentence is prominent position-
ally, and tends to carry the sentence accent. There is empirical evidence from
psychology for the prominence of the subject (Brennan 1995), and for the claim
that people pay attention to beginnings and ends of sentences.8
Subjects are salient from a grammatical point of view. Syntactically, sub-
jects have a cluster of well-known properties. They determine nominal case
marking and verbal agreement; they can be omitted under conjunction; they
control coreference in multi-clausal constructions. Semantically, the referent
of the grammatical subject phrase is pragmatically presupposed. Typologically,
languages can be divided into “subject” and “topic” languages (Li & Thompson
1976).
The study of grammatical voice has traditionally focused on subjects. Voice
deals with verbal alternations that indicate participants’ roles in a situation, as
in active, passive, and other constructions. The voice of a sentence conveys the
subject’s status vis-à-vis a situation or action, whether the “principal effects
of action devolve upon the subject or the object” (Klaiman 1991:3).9 In other
words, voice makes prominent the thematic role of a subject. For instance,
when the agent is salient the cause of the event is prominent; when the patient
8. When subjects in experiments hear a series – of digits, syllables, words, a long sentence or
series of sentences – they tend to remember the material that begins and ends the series (Neisser
1967:222).
Brennan (1995) studied people’s choice of referential expressions. She found that speakers
mentioned entities as full NPs in subject position before referring to them with pronouns.
New entities tended to be introduced as full NP objects. Experimental subjects tended to use
pronouns to express the most salient entity in a scene. Brennan’s study was done in the context
of Centering Theory (see Chapter 6). Subjects’ behavior was taken as evidence for a Centering
Theory account of pronoun use, rather than a simple knowledge-based strategy. The latter would
predict that pronouns be used for all familiar entities.
Linguists who worked with this and related pragmatic notions include Firbas (1964), Hockett
(1958), Strawson (1964), Halliday (1967), Kuno (1972), Dahl (1974), Hajičová (1971), Daneš
(1974), Reinhart (1982), Davison (1984), Fries (1983), Vallduvı́ (1992), Lambrecht (1995),
Vallduvı́ & Engdahl (1996), and Erteschik-Shir (1997).
9. Cross-linguistic studies of voice are concerned with variations in such constructions as active
and passive, middle, anti-passive, inverse. Klaiman (1991) identifies other approaches to voice.
“Posttraditional linguistics” takes voice as being concerned with the mapping from logical to
grammatical structure, and not associated with participants or viewpoint. For Philippine and
Mayan languages there are “pragmatic voice” alternations which signal assignment among
nominals of a certain pragmatic status.
is salient, the event or its result is prominent. The active–passive, and other
alternations, change pragmatic perspective, as Givón (1993) puts it. The per-
spective varies according to what aspects of a situation – cause, process, and
result – are prominent in a sentence. This view is reminiscent of Grimshaw’s
(1990) notion of aspectual prominence, and is different from the perspectival
notion of subjectivity discussed in Chapter 7.
There is a strong tendency for topic and subject to coincide in texts as well as
single sentences. The empirical correlation will be discussed and examplified
in Chapter 10. But the topic phrase is not always the sentence subject. To
show this, we need a context that clearly determines a given phrase as topic.
Example (6) illustrates: note the third clause of B’s answer. The example is
from Erteschik-Shir (1997):
(6) A. So tell me about the earth. What do you know about the earth?
B. It’s round, it’s a planet, the moon goes around it.
In the last clause of B’s answer, the topic phrase is the pronoun object of the
preposition.
The topic referent must be an individual of some kind: an entity or a concept.
In the case of sentential subjects, the referent is the event, state, or other entity
expressed by the subject clause. Sentences with quantified subject phrases have
no direct topic referent. In such cases, the topic is the domain of quantification,
the class referred to in the quantified phrase. For instance:
(7) a. All the humpbacks in a given region sing the same song, which is constantly
evolving.
b. No person could survive without precise signalling in cells.
The topic referent in (7a) is the class “humpbacks in a given region,” the topic
referent of (7b) is the class of “persons,” and so on.
Topic referents tend to be familiar or readily inferrable from the discourse
context. The topic referent, however, needn’t be familiar. Example (8) illus-
trates: the referent is identified by his connection to the speaker and is unknown
otherwise. The example is based on Prince (1981):
(8) A man I know had a really surprising experience.
Intuitively, the sentence is “about” the referent of the subject NP.

The examples demonstrate that neither familiarity nor subject position is a
defining property of a sentence topic in English.
8.2.3 Sentences without topics

Not all sentences have a topic. Sentences without topics simply offer a situation
to the receiver. They are unpartitioned wholes, all-focus sentences, rather than
predications about a particular referent, as in (9):
(9) a. There are yellow flowers.

b. God exists.
c. It’s raining.
These sentences are known as “thetic.” In contrast, “categorical” sentences ex-

press a judgment or predication about a referent (Kuroda 1973, Sasse 1987;
“neutral descriptions” in Kuno 1972). Thetic sentences can answer questions
such as “What happened?” or “What is there?” that do not single out any
referent as the locus of aboutness. Lambrecht notes that the construction of
(9a) functions to introduce a new entity into the discourse (1994:39). Thus
(9a) is presentational. (The term “presentational” is used both for the fam-
ily of constructions that function in this way and for the factors of topic and
focus.)
The entities introduced in presentational sentences become part of the uni-
verse of discourse. As such they are relevant to the developing information of
a text.
The thetic–categorical distinction is expressed overtly only by accent in
English. The subject receives the sentence accent in thetic sentences, the H+L∗
contour. Categorical sentences have the H∗ +L contour in the predicate and
may have a secondary accent, the L+H∗ contour, on the subject. The examples
illustrate, from Chafe (1976); small capitals indicate sentence accent, as usual;
italics here indicate the secondary accent. The sentences of (10) are thetic, those
of (11) are categorical:
(10) The british are coming.

The butter melted.
(11) Mary is singing.
The butter melted.
Subject accentuation, as in (10), signals that the situation be taken as a whole.

The dual accentuation of the sentences in (11) distinguishes an element denoting
an individual. The accents on the subjects are secondary, while the verbs have
the stronger sentence accent. In texts, the interpretation of a sentence as thetic
or categorical is determined by contextual cues since the intonational factor is
not present.
8.2.4 Notions of topic

The term “topic” and its meaning are often seen as problematic. One reason
for the difficulty is terminological. The term has been used for distinct notions,
and different terms have been used for the notion of topic. The main ideas are
the initial position in a sentence, familiar information, and aboutness.
“Theme” and “topic” are sometimes used almost interchangeably to refer
to the phrase in initial position in a sentence. The initial phrase is the starting
point of the sentence, and tends to convey familiar information. According
to the Prague School, “The theme is that which is known or at least obvious
in the given situation and from which the speaker proceeds” (Firbas 1964).10
The theme–rheme structuring of a sentence is thus based on the two factors
of position and familiarity status. Theme or topic in this sense is developed
further in the work of Firbas (1974), Gundel (1974), Kuno (1976), and others.
Difficulties arise when the factors of initial position and familiarity status are
combined, as pointed out by Daneš (1974), Chafe (1976) and Fries (1983).
“Theme” refers to the initial constituent of a sentence in the work of Halliday
(1967) and in systemic linguistics. The theme is the point of departure for the
message, and is often what the sentence is about, especially in simple sentences.
The unmarked theme is the subject of a declarative sentence, for instance “John”
in “John saw the play yesterday.” Marked themes are non-subject phrases in
initial position, as in “The play John saw yesterday” and “Yesterday John saw the
play” (1967:212). Halliday’s notion of theme has a relatively weak component
of aboutness. The theme includes some phrases that do not refer, and do not
have the aboutness relation to the clause. “Theme” fails to account for phrases
in other positions that bear the aboutness relation to the sentence; but it was not
Halliday’s intention to characterize them.
The term “topic” is discourse-functional for Givón (1983): it refers to a
dimension that includes “backgroundiness, predictability, continuity.” This
notion of topic is non-discrete, with differences of degree according to the
structure of a sentence. For instance, in a sentence such as “John, we saw him
yesterday,” the initial phrase “John” and the subject “we” both have some degree
of topicality. Givón defines topicality in terms of anaphoric accessibility and
“cataphoric persistence” – whether a referent recurs in a text and its thematic
importance and activation. Highly topical referents tend to have antecedents in
preceding clauses and to recur in subsequent clauses (1983:10).
10. The definition is a translation from Mathesius. Most definitions of theme are derivable from
this one, according to Fries (1983).
The aboutness notion of topic is developed in Lambrecht (1994), and some-

what differently in Erteschik-Shir (1997). Even closely related uses of “topic”
represent the tip of different theoretical icebergs. The difference between these
theories is beyond the scope of this discussion.
The term “topic” is also used at the discourse level. “Discourse topic” is
another kind of aboutness, a supersentential notion. Unlike sentence topics,
discourse topics are not always expressed directly. There is often no linguistic
expression that denotes the theme, or topic, of a discourse or discourse unit.
Often the discourse topic must be inferred or constructed.11
The problematic nature of aboutness has led some linguists to a limited ver-
sion of this notion (Vallduvı́ & Engdahl 1996, Walker et al. 1998).12 Others have
argued that sentence topic be abandoned, notably Sperber & Wilson (1986).
However, I shall rely on the aboutness notion of sentence topic, using the
approach of multiple cues discussed below. In this view, the topic phrase of a
sentence denotes a topic referent; the sentence is about that referent.
8.2.5 Determining the topic phrase of a sentence

The topic phrase is typically the subject in canonical S-V-O sentences, and the
topic referent is typically familiar information. But the topic phrase need not
11. Topics in conversation can be derived, according to Keenan & Schieffelin. They claim that any
discourse has a single proposition which represents the discourse topic (1976:338). The listener
establishes the discourse topic by reconstructing the semantic relations between referents.
Chafe argues that discourse topics are aggregates of semiactive information that segment a
conversation into larger chunks than intonation units (1994:135).
For written texts, van Dijk suggests that a discourse topic proposition can be constructed
for the whole (1972:37). His account is based on an analysis of the propositions that underlie
the text. The result is a complex proposition that is entailed by the joint set of propositions
of the text. Van Dijk’s approach has been influential in text studies that focus on memory and
cognition. It is not relevant here because it departs immediately from the text in seeking the
underlying propositions. Moreover, construction of the propositional analysis is difficult at
best, as Garnham (1983) argues.
Discourse topics for several sentences at once have also been proposed. For instance, Asher &
Lascarides (1998) present the following sentence sequence:
I’ve just arrived. The camel is outside.
To make sense of the pair, Asher & Lascarides suggest that the receiver recovers the discourse
topic notion of “transportation,” which correctly links the sentences.
12. One type of limited topic is the “backward-looking center or Cb” of Centering Theory: see
the discussion in Section 6.4. The backward-looking center is defined in terms of grammatical
relations and pronouns. In many ways the Cb is a return to the two-factor notion of theme
proposed by Mathesius. Another limited topic is the operationally defined “Link” of Vallduvı́ &
Engdahl (1996).
be the subject of a sentence in discourse, and the topic referent may be new.
Given these difficulties, the question arises as to whether the topic of a sentence
can be reliably identified. Both Smith (1991) and Vallduvı́ (1992) argue that
there is no structural or “operational” test for topic. I think that this is correct.
Yet I am committed to the position that intuitions about sentence topic are
mostly reliable, and that sentence topics have semantic and pragmatic effects
in discourse.
How then does one decide on the topic phrase of a sentence in the context
of other sentences? I suggest that the intuition of topic is based on a set of
cues. Subject position is the default; it may be overridden by other cues in the
sentence and its context. There is the possibility of tension between sentential
and discourse factors. Cues may reinforce each other, or one cue may override.
Occasionally cues conflict and the intuition of aboutness is weak. In actual texts
the topic phrase of a sentence is usually the subject phrase, as noted above.
Within a sentence, the subject position is salient because of its linear position,
the grammatical relation of subject, and the typical familiarity status. Also
relevant are accent and morphology; topic phrases do not bear sentence accent
except in the marked case of contrastive topics. Topic referents tend to be
familiar, and familiar referents tend to be coded with pronouns. Thematic role
is another factor. Agent and experiencer arguments tend to appear in subject
position, with some exceptions, e.g. the psychological verbs surprise, amuse,
annoy, which encode the experiencer as object.
Discourse factors may also affect the intuition of aboutness. One factor is
continuity. Topic phrases maximize local or global continuity. Local continuity
identifies as topic the phrase that is coreferential with an immediately preceding
topic phrase.13 Global continuity identifies a topic phrase that is coreferential
with other topic phrases in the context. Syntactic parallels in sentence structure,
and changes of direction in the text, are also relevant.
In context, and in sentences that depart from the canonical S-V-O structure,
these factors may not converge. Therefore they must be considered as potentially
separate cues to the topic of a sentence. Criteria for sentence topics are listed
in (12):
(12) Topic cues

An NP may be the topic phrase of a sentence if it
a. is the subject of a sentence;
b. is a pronoun;
13. Local continuity is most highly ranked in Centering Theory. Other “strategies” are recognized
but considered less preferable, as Haihua Pan has pointed out to me.
c. realizes the agent or experiencer argument of the main verb;

d. is coreferential with the topic phrase in the preceding sentence;
e. is coreferential with a topic phrase in the context;
f. is coreferential with a phrase in the context;
g. is lexically related to other material in the context;
h. is in a parallel grammatical position with an NP in the context.
NPs with sentence accent or other focus cues are not candidates for topic. The
criteria usually function together, but in some cases they clash, as we will see.
When a sentence has more than one clause, the clauses are analyzed sequen-
tially in left-to-right order. This treatment is justified in Kameyama (1998); the
alternative is to take the main clause first.
In applying the criteria, the natural strategy is an additive one: the phrase to
which the largest number of these criteria applies is the topic. The topic phrase
must be integrated into the organization of the whole sentence. Recall that every
sentence has a focus phrase, canonically in the predicate.
The criteria and principles are applied to a text fragment in (13), part of it
already presented in Chapter 5. Topic phrases are italicized. The immediately
preceding sentence concerns a large project in Boston; the definite NP in S1
refers to working tunnels in that project. I give in (14) the cues used for deter-
mining each sentence topic; the letters before each cue refer back to the list of
criteria in (12).
(13) 1 In some places the tunnels are 120 feet deep. 2 In some particularly delicate
places the road-work passes within just a few feet of skyscraper founda-
tions or beneath construction projects. 3 The first two frequently asked ques-
tions on the project’s official Website are “What are you building?” and “Are
you nuts?” 4 Surprisingly, tunnels are among the most ancient engineering
feats. 5 The earliest [tunnels] were very likely extensions of prehistoric
cave dwellings. 6 The Babylonians, in the twenty-second century BC, built a
masonry tunnel beneath the Euphrates River that connected the royal palace
with a major temple. 7 The Egyptians, using copper-bladed saws, excavated
long passageways and intricate rooms inside soft-rock cliffs. 8 The Romans
built an elaborate network of above- and below-ground acqueducts to carry
water.
(14) Criteria used for sentence topics

S1: (a) subject; coferential with earlier NP
S2: (a) subject; (g) lexically related context (the tunnels – road-work);
(h) parallel
S3: (a) subject
S4: (a) subject; (g) lexically related context (the tunnels – tunnels)
S5: (a) subject; (f) coferential with earlier NP
S6: (a) subject; (c) agent; (g) lexically related context (earliest)
S7: (a) subject; (c) agent; (g) lexically related context (earliest); (h) parallel
S8: (a) subject; (c) agent; (g) lexically related context (earliest); (h) parallel
The topic phrases are all subjects; there are no non-canonical syntactic struc-
tures. The object NPs are the focus phrases in the last three sentences:
“a masonry tunnel,” “long passageways and intricate rooms,” “an elaborate
network.”14 These phrases pertain to the theme of the passage and might be
taken as topics. But topic–focus organization precludes such an interpretation:
the object NPs are focus phrases. The semantic notion of Primary Referent
applies to these NPs: they are the Primary Referents in their clauses. Each is
the result of the event expressed in the sentence. There is an interesting tension
in the passage between topical and semantic progression.
Although sentence topics can usually be identified, there are some un-
clear cases. They arise when the criteria for topics result in a clash. For
instance, consider clause 2b in the familiar fragment of (15); the first four
clauses are familiar from previous examples. Clause 2b has pronouns as subject
(“she”) and object (“it”), both are coreferential with material in the immediate
context.
(15) 1a She put on her apron, took a lump of clay from the bin and weighed off
enough for a small vase. 2a The clay was too wet, b so she wedged it with a
flat concrete tray, c which absorbed the excess moisture.
The subject pronoun of clause 2b realizes the agent role, and is in subject
position, so by the additive strategy “she” would be the topic phrase. Now
consider the two continuity principles. By local continuity, we take the object
pronoun “it” as topic because its referent is topic of the previous clause. The
global continuity principle looks at other sentences in the passage. S3 and S4
have “she” as topic phrase, so by global continuity “she” would be topic in S2b.
The sentence organization does not resolve the problem: the verb “wedge” is
the focused phrase. The most satisfactory analysis is simply to say that in clause
2b both pronouns are part of the background; see the next section below. In my
judgment the uncertainty of cases like this does not compromise the many clear
cases of sentence topics and the importance of the notion.
The cues to sentence topic are pragmatic in nature. Coreferentiality and lex-
ical relatedness are probably the most important, and neither can be stated in
a construction rule of Discourse Representation Theory. The other cues are
difficult to formalize as well. Therefore I do not give construction rules for
topic or presentational progression. I have argued that sentence topic is needed
14. I thank Pascal Denis for helpful discussion of these sentences.

for conceptual and truth-conditional calculations within the DRS; and for fur-
ther pragmatic interpretation. I assume that topic information must be part
of enriched Discourse Representation Structures that allow further pragmatic
interpretation.15
8.3 The Focus–Background partition

The focus phrase represents the speaker’s “declared contribution” to the com-
mon ground (Gussenhoven 1983:383). Focus is conveyed with sentence accent
and/or syntactic structure.
Canonically a sentence progresses from the starting point toward the focused
material, which bears the sentence accent. Every sentence has a focus. Focus
in this sense is “informational”: it is distinct from contrastive focus, discussed
below. Focus is robust because it has a clear linguistic correlate of accent in
the spoken language, and often position in the written language. Section 8.3.1
introduces focus phrases; 8.3.2 considers focus and semantic interpretation;
8.3.3 discusses contrastive and emphatic focus; 8.3.4 concludes.
8.3.1 Properties of focus phrases

The focus may be the smallest phrase that bears the sentence accent, or a larger
phrase.16 The scope of sentence accent is often indeterminate since the predicate
accent allows more than one interpretation. Example (16) illustrates for sentence
(1a) discussed above, an example from Sgall et al. (1986). The sentence has
three focus interpretations for successively larger phrases. Sentence (16a) gives
15. Recognizing sentence topics in a DRS does not necessarily provide for sentence topics as an
organizing factor of information in the common ground, or context set. Organizing potential is
one of Reinhart’s arguments for aboutness, and a feature of Vallduvı́ & Engdahl’s procedural
account of Links. I do not wish to make the stringent requirement that the context set be
organized in any particular way. Rather, I suggest that organization by topics is a potential
feature of organization in memory. It may be best not to limit information access, as Portner &
Yabushita (1998) point out. People are able to organize large amounts of information in multiple
ways, according to what question is being answered or what kind of task is involved. For
instance, research on word access shows that people can search their word memory by first
sound, alphabetically by first letter, according to taxonomic meaning, functional meaning, etc.
(Forster 1979). In the current theory the information that makes up the context set is structured
only in DRS terms.
16. There is an extensive literature on focus and sentence accent, especially contrastive focus.
Semantic analyses of focus operators are given in Rooth (1992) and Krifka (1991). Although
they differ in some respect, they agree on the points that I rely on here. See also Horn (1991),
Fauconnier (1985), and Bosch & van der Sandt (1999).
Stress and intonational highlighting mark focus in cases where the topic–focus articulation
cannot be read off the syntax; they also mark strong contrast or exhaustiveness.
the schematic surface structure, (16b–d) the focus interpretations. Sentence

accent will be indicated with small capital letters; it should not be confused
with contrastive stress, which will be written with large capitals (“JOHN saw
Mary” vs “John saw MARY”).
(16) a. Beavers build [dams].

b. [Beavers build] F [dams].
c. [Beavers] F [build dams].
d. F [Beavers build dams].
The accented NP is the focus in (16b); the VP predicate is the focus in (16c);
the entire sentence is focused in (16d). I use the terms “argument focus” and
“predicate focus” for the first two cases and the term “all-focus” for the third
(from Vallduvı́ 1992, Lambrecht 1994). Sentence (16d), which is all-focus, is
a thetic sentence. The notion of focus is a relational one, internal to a given
sentence. On the argument focus interpretation of (16b), for instance, dams is
the focus relative to the other information in the sentence.
Different focus interpretations arise in different contexts. The argument focus
reading of (16b) is natural in a discussion of what beavers build, or in answer
to a question about the building habits of beavers. The predicate focus reading
would be natural if one were discussing what beavers do. For the sentence
focus reading imagine the sentence uttered “out of the blue,” perhaps by a child
imparting a newly learned fact.
The focus of a sentence is its contribution to a discourse. The focus is a
matter of presentation in that it “reflects the speaker’s decision as to where the
main burden of the message lies” (Halliday 1967:204). The speaker presents
the focused material as relating to the background in the sentence. Standard
examples of focus–background structures in the literature are question–answer
pairs, where the question clearly determines a particular interpretation. Example
(17) is from Prince (1986):
(17) a. What did she give to Harry?

b. She gave the shirt to Harry.
The question provides the background for the answer, making it clear that the
focused phrase provides the contribution to the discourse. In the context of
the question the background is shared knowledge. Typically, the background
represents what is salient or inferrable in the discourse at the time of utterance,
or what the speaker assumes to be so.
This notion of focus should not be confused with other meanings of the term.
It is used for “center of attention” in psychology, computational linguistics, and
artificial intelligence. The attentional focus of a sentence is often taken to be

its topic phrase, it has the opposite meaning from the linguistic term “focus.”
Another meaning is focus of contrast, discussed below.
The Focus–Background partitioning of a sentence need not correspond to
its syntactic structure. The background material may be discontinuous. For
instance, the argument–focus partitioning of “Beavers build dams” has the sub-
ject and verb as background [Beavers build –], not a constituent in syntactic
structure. In “She gave the shirt to Harry,” the background is also discon-
tinuous [She gave – to Harry]. Surface structure does not relate directly to the
background–focus partition in such cases.
8.3.2 Focus and semantic interpretation

Formal accounts of the Focus–Background partition treat the relation between
the parts as that of a variable to an open proposition (Prince 1986, Rooth 1992).
The background part of a sentence is an open proposition with a variable, which
is instantiated by the focus. In modelling this partitioning one replaces the focus
phrase with a variable; the result is an open proposition.
The semantic interpretation of a sentence may be affected by focus. Sen-
tences with different focus phrases sometimes have different truth-conditional
meanings, as in the following examples from Gundel (1999):
(18) a. The largest demonstrations took place in prague in November 1989.

b. The largest demonstrations took place in Prague in november 1989.
If the largest demonstrations of November 1989 were in Budapest rather than

Prague, (18a) would be false because “Prague” is focused in the first example.
Under the same circumstances (18b) would not be false.17 “Prague” is part of
the background, not the focus, in the sentence. Another example shows the
interaction of focus with negation, from Hajičová & Sgall (1987):
(19) a. This time our defeat wasn’t caused by Harry.

b. This time Harry didn’t cause our defeat.
In (19a) the defeat has taken place: it is presupposed, part of the background.
Example (19b) has no such presupposition: the sentence is compatible with
a situation in which we were not defeated. The difference follows from their
background–focus structure, according to Hajičová & Sgall (1987). In (a) “our
defeat” is part of the background, whereas in (b) it is the focus phrase. There
17. If one assumes a three-valued logic, (18b) would have no truth value, as Neil Smith points out
(p.c.).
is some difference of opinion about the source of these intuitions of truth and
falsity.
The semantic consequences of background–focus are further adduced by
Partee (1991). Partee relates the distinction to tripartite semantic structures
that represent the contribution of quantifiers and other operators to a sentence.
The tripartite structure consists of a Restrictor, Nuclear Scope, and Operator.
The Operator applies to the Nuclear Scope in the context provided by the
Restrictor. The Restrictor and Nuclear Scope in such structures correspond to
the background–focus distinction. The focused phrase of a sentence functions
as Nuclear Scope, and the background functions as Restrictor. This and other
interactions of focus and semantics are addressed in Hajičová, Partee & Sgall
(1998).
Focused material need not convey information that is new in familiarity status
to the discourse, or to the receiver. What is essential is that the focus add to the
receiver’s apprehension of the information, relative to the rest of the sentence.
The fragment (20) illustrates; consider the status of “sedimentation,” the focus
phrase in the last sentence (Daneš 1974:111):
(20) Sedimentary rocks: Most of historical geology has to do with sedimentary

rocks and their organic remains . . . Events in earth history are recorded mainly
in terms of differing kinds of sedimentation.
The focus phrase “sedimentation,” far from being new, is fully recoverable
from the context, as Daneš notes. It has a unique function in this sentence,
however.
Examples of another kind also show that focused information need not be
new to the discourse; (21) is from Lambrecht (1994), (22) from Gundel (1999).
In each case the relevant example is B’s answer to A’s question:
(21) A. Where did Sami go last night, to the movies or to the restaurant?
B. Hei /Sami went to F [the restaurant ].
(22) A. Who called?
B. Pati said F [she i ] called.
Reference to the focus phrase referent appears in the same sentence (22), or
in the preceding question (21), so that it cannot be regarded as new. The focus
phrases supply information that completes the background open propositions,
which are [– called] and [Sam went to –]. Thus, focus–background is a sentence-
internal relation established for a particular sentence.
The main cues to focus are sentence accent and linear or syntactic posi-
tion. Position is particularly important in writing, where the only other cue is
typographical indication of contrast, usually italics or capital letters. In certain

syntactic structures the focus phrase appears in a special position. Example (23)
illustrates focus preposing, from Hannay (1991):
(23) a. Did you get wet?

b. F [Bloody soaking] I was.
In other structures, the syntax highlights or focuses a constituent. The exam-

ples illustrate: (24a) is a cleft, (24b) a pseudo-cleft, and (24c) a presentational
sentence. The accent falls on the syntactically focused constituent.
(24) a. It was F [t he bu tler] who took the necklace.

b. What the butler took was F [th e n ecklace].
c. There were F [swans] on the lake.
Non-canonical structures are discussed in Chapter 9.
8.3.3 Contrastive and emphatic focus

Contrastive focus and its close relation, emphatic focus, add an additional di-
mension of meaning to the information in a sentence. I will consider the dif-
ference between contrast and emphasis as conveyed in writing with capital or
italic letters. I’ll refer to them both as “strong focus.”
Contrast and emphasis are often put into a single category of “contrast-or-
emphasis,” but they differ semantically and pragmatically. The sentences in
(25) give an idea of the kinds of examples to be discussed:
(25) a. Sue danced with BILL. (imagine that)

b. Sue danced with BILL. (not Tom)
c. Sue even danced with BILL. (of the possible people)
8.3.3.1 The relation between contrast and emphasis

Contrast and emphasis both single out an entity or property from a set of relevant
possibilities. The two differ in how the singling out is presented. In simple sen-
tences out of context, such as “Sue danced with BILL,” the stressed phrase can
be interpreted as emphatic or contrastive. On the emphatic meaning, the danc-
ing is interesting, surprising, unexpected, or noteworthy in some way. The
contrastive meaning involves choice of a more pointed kind.
Contrast depends on a choice among alternatives. But not all choices are
the same. The more specific the alternatives, the more contrastive the ef-
fect. Bolinger comments: “As the alternatives narrow down, we get closer
to the standard idea of contrast” (1961:87). He illustrates with the examples

in (26):
(26) a. Let’s have a PICnic.

b. Where’ll we have it? Let’s have it in the PARK.
c. Can we all go? No, MARY can’t.
d. Bring some hotdogs. I don’t like hotdogs, I like HAMburgers.
The alternatives increase in specificity. Example (26a) suggests a general con-

trast between picnicking and anything else the group might do. Example (26b)
implies other places where a picnic might be held, but does not state them.
Example (26c) offers a clear set of alternatives (we all). Example (26d) states
alternatives and sets up an opposition between them. This is the strongest type
of contrast, where one alternative is chosen and others rejected.
The difference between contrast and emphasis can be explained with a con-
tinuum, a gradient scale of choice and alternatives. At the weakest end of the
scale is emphasis (26a); explicit contrast between alternatives is in the middle
(26c); at the strongest end is “oppositional contrast” (26d).
Emphasis arises when strong stress is used to draw attention to a particular
referent, rather than to the choice of that referent over alternatives. (27) makes
this clear:
(27) Sherlock Holmes to the butler:

The murderer is YOU.
The example is from Lambrecht, who comments that Holmes needn’t have
had alternative suspects in mind to utter the sentence (1994:287). The focus
is emphatic, adding strength to informational focus, but does not change the
interpretation of a sentence.
8.3.3.2 Types of contrastive focus

Contrast is at the middle of the scale or continuum of choice, and oppositional
contrast is at the strong end. Oppositional contrast rejects the non-preferred
alternative. Contrastive focus singles out a particular referent in preference
to alternatives. The relevant alternatives are explicit, or inferrable from the
context. Strong stress is the simplest type of focus operator. It conveys a contrast
between the focused phrase and the relevant set of alternatives. For instance,
consider “Sue danced with BILL” on the contrastive meaning. Contrastively the
sentence conveys that Sue danced with Bill rather than with others. The semantic
interpretation is that focus triggers a partition of a sentence into background and
focus, with a set of alternatives. The focus applies to the background rather than
to the other alternatives (Krifka 1991, Rooth 1992). This type of interpretation
is set out schematically in (28):
(28) Focus structure of Sue danced with BILL
Background: Sue danced with –
Focus: Bill
Focus operator: strong stress
Alternatives: everyone, Mats, Manfred, Bill, etc.
Focalizing particles such as even, only, also, convey contrastive focus. These
particles trigger and contribute to contrastive interpretation; strong stress is not
needed in their presence. Different particles relate focus and background to
their alternatives in a particular way. For instance, only conveys that there is no
relevant alternative to which the background applies. Even conveys that there
are alternatives to the focused choice which are more probable, more expected,
than the one that actually appears.
Focus that involves clear alternatives is often exhaustive, or “identificational,”
according to Kiss: “Identificational focus . . . identifies the exhaustive subset
of the set [of potential candidates] for which the predicate phrase holds”
(1998:245). Kiss contrasts this type of focus with information focus, which
every sentence conveys. Information focus merely marks the nonpresupposed
nature of the information it carries, according to Kiss.18 By this criterion, some
sentences with contrastive meaning do not have identificational focus. For in-
stance, “Sue danced with BILL” does not necessarily mean that Sue danced with
no person other than Bill. In contrastive contexts, however, focus constituents
express identificational focus in English, according to Cohan (2001).
Phrases with strong stress tend to be focus phrases. Topic phrases, however,
may also be contrastive. The alternatives are often explicitly parallel, as in (29)
from Vallduvı́ & Engdahl (1996):
(29) Where can I find the cutlery?
a. the forks are in the CUPBOARD . . .
b. but I left the knives in the DRAWER.
Contrastive topics like these have the aboutness property, rather than the
semantic properties of focus phrases (Lambrecht 1994:291).
18. In Hungarian informational and identificational focus phrases are associated with different
structural positions. Kiss lists six differences between the two kinds of focus. The first is given
in the text. The next two differences are these: (2) the type of constituent that can function
as information focus is not restricted, but certain types of constituents cannot function as
identificational focus; (3) information focus does not, but identificational focus does, take
scope. The other differences are more technical (1998:248).
I now return to the contrast continuum. Contrast and oppositional contrast

differ in how the focused element relates to alternatives. In contrast, a choice is
preferred to its alternatives. In oppositional contrast, a choice is opposed to its
alternatives and the alternatives are rejected. I present below some fragments
from discursive texts with focus operators. The focus phrases are italicized.
(30) Contrast
a. Chimpanzees engage in what might be called “trickle-down” provisioning
of meat to females and their offspring. Only those females and their young
who are in the vicinity of the killsite when the meat is being devoured will
have a chance of gaining access to it.
b. There is another difference between intelligence and other traits. Height
and weight and speed and strength and even conversational fluency are
real things: there’s no doubt about what’s being measured. Intelligence is
a much murkier concept.
In (30a) background knowledge enables the reader to provide reasonable alter-

natives to the focused phrase; in (30b) the alternatives are explicitly given. The
fragments illustrate the main functions of focus particles in texts: only limits a
category or domain, while even extends the range. The fragment in (31) presents
oppositional contrast
(31) Oppositional Contrast

The desirability of segregating dwellings from work has been so dinned into
us that it takes an effort to look at real life . . . In an article about Harlem we
find “We’ve got hills and views of both rivers . . . and it’s the only close-in
area that doesn’t have any industry.”
Only in planning theory does this make Harlem an “attractive piece of real
estate.” From the time of its . . . beginnings, Harlem never was a workable,
economically vigorous residential district of a city.
The contrast is oppositional because, in addition to a chosen alternative, there

is an alternative that is explicitly rejected.
The alternatives to contrastive focus may be nearby, or relatively far away
in the context. The information may appear in the same syntactic phrase, a
neighboring sentence, or in a paragraph segment other than the one in which
the focus-phrase sentence appears. Contrast occurs in text at the level of the
phrase or sentence; it may also function as a discourse relation or a higher-
level organizing principle. There are lexical ways of indicating opposition, for
instance with phrases such as rather than, instead of , and others; see Smith
(1998) for further discussion.
8.4 Dual partitioning: representation of topic and focus 209
8.3.3.3 Summary and comment

The partitioning of a sentence into focus and background has roots in the
work of Daneš, Firbas, and other scholars of the early Prague School, and
Halliday. The ideas are developed further in recent work by Hajičová and
her colleagues. In another tradition, Chomsky (1971), Jackendoff (1972), and
Rochemont (1986) discuss focus and presupposition, and related distinctions.
Prince (1981, 1986) elucidates the notions of background and focus. Much
current work on the interface between different parts of the grammar uses the
notion of focus because it involves phonology, semantics, syntax, and pragmatic
interpretation.
There is more than one type of focus. Informational focus appears in all
sentences: it is the contribution that a sentence makes to a discourse. Strong
focus is emphatic or contrastive; it is indicated by italics or capital letters in
writing, heavy stress in speaking, or by focus operators. Traditional terminology
varies. For Halliday (1967) and Daneš (1974) the distinction presented here as
focus–background is known as theme–rheme. Sgall, Hajičová, and the current
Prague School use the term Topic–Focus Articulation, or TFA. Their term
“topic” is like “background” as presented here; the background is not structured,
so that a particular topic phrase and referent is not always identified.
There is an enormous literature on focus, in phonology, syntax, and semantics.
In this brief introduction I have introduced the key ideas, with comments of my
own on contrast and emphasis. The notion of focus balances that of sentence
topic; it is a necessary prerequisite for the dual partitioning of sentences.
8.4 Dual partitioning and the representation of topic and focus

The approaches of Background–Focus and topic–comment are complementary
ways of dealing with presentational factors. The two partitionings distinguish
different aspects of how information is organized, and they hold simultaneously
for most sentences. The topic–comment and Focus–Background partitions each
select a particular phrase, leaving the rest of the sentence unorganized. In the
topic–comment partition, the comment has no internal structure. Similarly, the
Focus–Background partition gives no internal structure to the background.
Yet many sentences have phrases that are neither topic nor focus. Consider
the simple transitive sentence “The girl broke a vas e,” for example. The subject
is the topic phrase (“the girl”) and the object is the focus phrase (“a vase”). The
verb is neither topic nor focus, on the narrow interpretation of focus. Yet the
verb should be integrated into the organizational structure of the sentence. Dual
partitioning allows a way of doing this (Dahl 1974, Vallduvı́ 1992).
In dual partitioning, the first partition recognizes a “Focus phrase” and an

optional “Ground.” The focus phrase receives the sentence accent, the H∗ +L
contour. The Ground is optional because there are all-focus sentences without
a Ground, as in example (10). The second partition divides the Ground into
“Link” and an optional “Tail.” The Link has the L+H∗ secondary sentence
accent, optional in subject position. The Link corresponds roughly to the notion
of topic.
The differing types of partitioning are most explanatory for different cases.
The topic–comment partition applies to sentences with a topic phrase which is
evidently the locus of aboutness. Focus–background is useful for answers to
questions and for sentences in which a focus phrase is clearly set off from the
background. But, as Vallduvı́ points out, the focus–background partition fails to
account for the “informational split” in a sentence with a preposed phrase such
as “To Harry she gave the SHIRT.” No function is recognized for the preposed
phrase (1992:51).
Dual partitioning allows information of both kinds to be represented. Taking
this approach, Vallduvı́ & Engdahl (1996) give a procedural account of how
presentational structure organizes information. They rely on an adaptation of
the File Change Semantics of Heim (1982). In understanding a sentence the
receiver updates the developing context of a discourse in terms of “file cards”
for referents. The way information is presented gives instructions for the update
in terms of Link, Focus, and Tail. The Link phrase of a sentence indicates a
locus of update, while the Focus adds information to the relevant file card in
the common ground. The Tail of a sentence, if there is one, further specifies the
update.
The dual partitioning scheme is set out in (32) and exemplified in (33), with
a strong accent on the verb.
(32) Sentence = {Focus, (Ground)}

Ground = {Link, Tail}
(33) The president hates chocolate.
{G (The president – chocolate) F (hates)}
{G (LINK (The president) TAIL (chocolate))}
Analyzing this sentence into Focus and Ground, the discontinuous material
[“the president – chocolate”] is the Ground; the Focus phrase is “hates.” Within
the Ground, “the president” is Link, “chocolate” is Tail. The Link functions
as an instruction for updating the common ground, the locus of update “the
president,” with a further condition specified by the Tail (Vallduvı́ & Engdahl
1996:468–70).
8.4 Dual partitioning: representation of topic and focus 211
Links are structurally encoded by position and accent, an operational def-

inition. Vallduvı́ & Engdahl claim that Link phrases appear in subject and
non-subject position. In non-subject position they have the secondary sen-
tence accent. Secondary accent is optional in subject position, as noted above
(1996:473). Vallduvı́ and Engdahl suggest that the operationally defined “link”
should replace aboutness topics. The account is attractive but empirically inad-
equate. An important difficulty is that pronouns do not have a secondary accent,
and most non-subject topic phrases are pronouns. For instance, a non-subject
topic pronoun appears in the third clause of example (6b) above: “It’s [the
earth’s] round, it’s a planet, the moon goes around it.” The example is far from
trivial: pronouns are very common as recurrent topic phrases in text, and are
not accented.19
I adopt the dual-partition approach, but with the aboutness notion of Topic
instead of the operationally defined Link of Vallduvı́ & Engdahl. The Ground
consists of Topic and Tail. The topic phrase referent is the locus of aboutness
for the sentence; the Tail consists of material that is not included in the Focus.
(34) a. Sentence = {Focus, (Ground)}
b. Ground = {Topic, (Tail)}
Simple intransitive sentences (e.g. “The girl fell”) are exhaustively partitioned
by Topic and Focus; others have a Link as well. All-focus, thetic sentences have
no Ground.
How does presentational structure relate to grammar and to the representation
of discourse? It is clear that presentational structure interacts with semantic
as well as pragmatic interpretation. Some theories have a level of syntactic
structure from which the semantic effects of focus are derived, with some
underspecification, as in the dependency grammar of Hajičová & Sgall (1987).
Erteschik-Shir (1997) proposes an informational level of grammar that affects
semantics and phonology. Combinatory Categorial Grammar has a level of
information structure which includes syntactic and prosodic information; in this
architecture there is no distinct level of surface syntax (Steedman 2000:126).
Others argue that a pragmatic component deals with the “information structure”
of topic and focus (Vallduvı́ 1992, Lambrecht 1994).
Since focus is associated with semantic operators, including negation, it is
directly relevant to truth-conditional representation. Some evidence is cited
19. In fact, the treatment of unstressed pronouns is problematic in this account. Vallduvı́ & Engdahl
(1996) assimilate them to null pronouns, treating both types as a single class. Although this
has a certain appeal for a general, cross-linguistic account, I do not think that it is correct for
English.
above; see also Hajičová, Partee, & Sgall (1998). Rules and representations for
focus are beyond the scope of this work.
8.5 Conclusion
The notions of topic and focus organize the information within a sentence. The
topic phrase of a sentence gives the referent that the sentence is about; the
focus phrase gives the main contribution of the sentence to a discourse. Dual
patterning is needed to account for the full range of cases.
The familiarity status of information is a key factor in presentation, both
within and across sentences. Within the sentence, familiar information tends to
appear first; and topic referents tend to be familiar. Focus phrases may convey
informational, emphatic, or contrastive information.
In the next chapter I look at surface presentational factors in non-canonical
sentences, and discuss whether and how to represent them in Discourse Rep-
resentation Theory. The account of presentational progression is developed
further in Chapter 10.

(13) Jim Collins, How it works. US Airways Magazine, Attaché, May 2001.
(15) Peter Robinson, A Necessary End, New York: Aron Books, 1989, p. 182.
(30a) Terrence Deacon, The Symbolic Species, New York: Norton, 1997, p. 393.
(30b) James Fallows, More Like Us, Boston: Houghton Mifflin, 1989, p. 153.
(31) Jane Jacobs, The Death and Life of Great American Cities, New York: Random
House, 1961, p. 175.
9 Non-canonical structures
and presentation
Canonical sentences furnish the background for variation. I consider here struc-
tures that play off this background with different word orders and syntactic
structures. Non-canonical structures have special force, because of their fea-
tures and because they depart from the basic case. Writers choose structures. I
assume that choice is based on assessment, not necessarily conscious, of how
a structure affects interpretation in a specific context.
Sentence-internally, non-canonical structures highlight or downplay the
material in certain positions. Syntax may enhance connectedness between sen-
tences by placing information that is familiar to a discourse first in a sentence. A
given structure may allow or block a topic relation with the following sentence.
Changes in direction may be conveyed by sentences that lack such connection,
and by breaks in the syntactic pattern. Thus syntactic patterning affects the
organization and progression of discourse passages.
This chapter concentrates on non-canonical structures that affect topic and
sentence connectedness, the main factors of presentational progression. I draw
on discussions in the literature of a variety of constructions. Together they give
a sense of the different tools that the language makes available. I will also look
briefly at multi-clause sentences, and will discuss paragraphs as text units.
The interpretations involve inference. Semantic presuppositions are close to
the linguistic forms: they are triggered by particular structures, such as cleft
sentences and temporal clauses; and by particular forms such as the focus
particles “only” and “even.” Pragmatic presuppositions of familiarity status
and linking inferences depend on context, world knowledge, and convention.
Topics are determined by a combination of cues including syntactic position,
lack of sentence accent, and coreference. Most of these cues are pragmatic in
nature.
Section 9.1 introduces non-canonical constructions; 9.2 discusses non-
canonical constructions with arguments; 9.3 discusses adjunct preposing; 9.4
considers multi-clause sentences; 9.5 discusses paragraphs; 9.6 comments on
presentational information in Discourse Representation Structure.
213
214 Non-canonical structures and presentation
9.1 Non-canonical constructions

These constructions affect the topic potential of a sentence and its connection to
context, the main factors in presentational progression. Some structures affect
the verb and its arguments, others affect adjuncts. In the first class are non-
canonical subject constructions, inversions, preposing, and postposing. The
second class consists of structures in which adjuncts are preposed from their
canonical rightward position. This is not an exhaustive list of non-canonical
structures: it ignores particle shift, the many types of conjunction and ellipsis,
and others.
Phrases that are displaced to the beginning or end of a sentence are in
positions of prominence. Clefts highlight a phrase syntactically, while sentences
with empty subjects highlight a non-subject phrase. Structure-based factors of
presupposition and strong, identificational focus are also relevant. Informational
focus, a feature of all sentences, is not.
The effect of a non-canonical construction is often to place information that
is familiar at the beginning of the sentence, where the canonical structure would
have it later. In fact, the felicity of a construction in context often depends on
the placement of familiar information. At the beginning of a sentence it can
link directly to the preceding text, maintaining continuity. The familiar phrase
also functions as starting point for the sentence.
The notion of a link or connection between sentence and context is intuitively
plausible, but the basis for one’s intuition is not always clear. I draw on a formal
approach to linking developed by Ward & Prince (1991). Discourse entities are
linked to context by their relation to a constructed set of entities. The construct is
a partially ordered set of entities evoked directly or by inference in the context,
a “poset.” The link entity is ranked with other entities in the poset according
to formal relations; they include relations such as part-of, attribute-of, prior-to,
etc.; the standard set relations; the identity relation. The poset must be salient
in the context. (“Link” in this sense should not be confused with that proposed
by Vallduvı́ in connection with the topic–focus partitioning of sentences.)
Discourse entities are linked through the poset construct to the context. For
instance, consider the underlined preposed phrase in a sequence from Prince:
“This I don’t call cooking, when you go in that refrigerator and get some beans
and drop them in a pot. And TV dinners, they go stick them in a pot and
she say she cooked.” The referent “beans” has already been evoked when the
receiver encounters TV dinners. “In the absence of any explicitly evoked set to
support ‘TV dinners,’ the entity evoked by beans is construed as an alternate
member of a set available by inference, the set of ‘fast-food home-cooked
foods’ ” (1997:7). Here the relation between the referents is alternate members
of the poset. Another example: “I walked into the kitchen. On the counter was
a large book.” The first sentence with “kitchen” evokes the poset “elements of
a house,” ordered by the part-of relation; the referent of the preposed phrase
“the counter” is related to this set as part-of the kitchen.
If no salient poset can be inferred, there is no linking relation.1 The non-
canonical structure of preposing depends for felicity on poset linking. If a
sentence does not have a poset link to the context, preposing cannot occur. For
instance, in “I walked into the kitchen. On the jacket was a large book,” there
is no salient or inferrable poset that relates “kitchen” and “jacket,” and the
sequence is infelicitous (Birner & Ward 1998:19–20). The notion of “poset”
linking accounts for many inferred entities in the literature. I will use “poset link-
ing” as a technical term for this type of connectedness between the sentences of
a discourse. These examples show that the term “familiar information” includes
material that is either explicitly mentioned or inferrable in the context.
For each structure considered below, I discuss how it differs from the canon-
ical pattern; its topic potential; structure-based particulars, if any; and the role
of familiarity status in its use. The examples are from texts in this study.
9.2 Argument constructions

9.2.1 Non-canonical subjects
Syntactic structures with non-canonical subjects include the passive, existential
there-sentences, cleft and pseudo-cleft sentences, and extraposition sentences.
The variety is striking and suggests that such structures are useful in discourse.
The passive. This is the classic case of a non-canonical structure. Subject and
object arguments are displaced, and the simple active verb form is replaced by
the passive auxiliary. The surface subject is the logical or underlying object;
the underlying subject appears optionally in a postposed by-phrase.
1. The notion of posets is due to Hirschberg (1991) and has been further developed by Ward &
Prince (1991), Birner & Ward (1998).
A partially ordered set is any set defined by a transitive partial ordering relation R, a relation
which is either reflexive and antisymmetric, or irreflexive and asymmetric (Hirschberg 1991).
Ward & Prince give as an example of the first type the relation “is-as-tall-or-taller-than”; as an
example of the second the relation “is-taller-than.” The discourse entity is related to the poset
by a ranking in which the entity represents a lower or higher value, or alternate value.
In the terminology of Birner & Ward (1998), the linguistic material that is related to the
context is the “link” and the poset that relates link and prior context is the “anchor.” The relation
between link and anchor, the “linking relation,” is always a poset relation.
The derived surface subject is often the topic phrase of a sentence. In fact, the
traditional function of the passive is to present as subject a phrase that would
canonically be in object position. There are two main reasons for using the
passive construction: “a greater interest in the passive subject . . . or to facilitate
the connection of one sentence with another” (Jespersen 1924/1965:168; see
also Creider 1979, Davison 1984).
The full passives in the texts of this study all have the surface subject as topic.
In most of them the topic phrase is linked to the prior context via a related poset.
There are several types of connectedness, as the examples in (1) illustrate.
(1) a. There were protests that Freud was unworthy of even being honored by
an exhibition. A corresponding exhibition on Darwin would have been
protested only by creationists.
b. Dynamite was something new, and it gave engineers their most powerful
tool for tunneling through hillsides that couldn’t be opened by digging from
above.
c. Despite the often cavalier attitude toward teaching in college, at least physi-
cists know their physics, mathematicians know and love their mathematics,
and music is taught by musicians, not by graduates of education schools,
where the disciplines are subordinated to the study of classroom manage-
ment.
In (1a), the surface subject is linked to the indefinite NP “exhibition” in the

preceding sentence by an inferred poset of “exhibitions.” In (1b) the passive
is a relative clause on the head noun “hillsides,” which functions as link and
antecedent by identity. In (1c) the passive is parallel to other clauses in the
sentence and is linked by a poset of “subjects taught” to the context.2
2. A recurrent question is whether the length and complexity of a phrase – known as “heaviness” –
is a factor of importance in non-canonical constructions. The heaviness of the underlying subject
does not appear to be a determining factor in the use of a passive, as the following examples
show:
The peals of laughter which followed Gabriel’s imitation of the incident were interrupted
by a resounding knock at the hall-door.
He would only make himself ridiculous by quoting poetry to them which they could not
understand.
In the first example the surface subject is long and complex; in the second the postposed material
is long and complex. They are from the same text. To determine the importance of the factor of
heaviness would require a large, corpus-based study.
Hawkins (1994) argues that heavy phrases in sentence-final position are relatively easy for
receivers to understand; postponing them to the end of a sentence may be helpful for speakers
as well: cf. Arnold et al. (2000). Thus the heaviness of a phrase might lead to the use of a
non-canonical structure which would move it toward the end of a sentence.
Using the notion of link developed above, Birner & Ward (1998) discuss the
familiarity status of passive surface subjects and by-phrases. They propose a
familiarity status requirement couched in relative terms: the subject referent
of a passive must be at least as familiar, in context, as the referent of the by-
phrase (1998:199). This requirement accounted for all full passives found in
an extensive corpus. It also accounts for most of the passives in this study.3
Birner (1996) compared the familiarity status of the subject and by-phrase NPs
in passives in an earlier corpus-based study. In that study too she found that the
surface subject referent was always as familiar as the by-phrase in the context.
Summarizing, the surface subject of a passive tends to be the topic phrase, and
has information that is relatively familiar in the context.
Psycholinguistic experiments have often used the passive in studies of sen-
tence processing because of its clear relation to the active form and its greater
surface complexity.4 Gordon & Chan studied pronouns and full NPs in passive
and active sentences presented in short constructed contexts. They consider sur-
face structure factors and thematic roles and conclude that the passive allows
for “optimal discourse organization” by maximizing continuity based on sur-
face form and syntactic relations. The semantic factor of thematic role was less
important (1995:229). Their conclusion supports the approach of this book, in
which presentation is separated from semantics.
The short passive, without an agent, is much more common than the full
passive in spoken and written discourse. Corpus studies from 1966, 1979, and
1987 found that 80–83% of the passives were agentless (Svarthvik 1966, Givón
1979, S. Thompson 1987). Birner & Ward (1998) also found that short passives
3. There are some passives which appear to be counter-examples to Birner & Ward’s generalization.
In the example below, the surface subject information may be less familiar than the by-phrase.
The passive appears in the third sentence.
There was only one important exception, one new product, in those first fifty years: the
steamboat, first made practical by Robert Fulton in 1807. It had little impact until thirty
or forty years later. In fact, until almost the end of the nineteenth century more freight
was carried on the world’s oceans by sailing vessels than by steamships.
“Freight” is perhaps evoked in the context by “impact” but this does not seem like a salient
poset. However, “sailing vessels” is a link to the context via a poset “ships” to which both
sailing vessels and steamboats belong.
4. Psycholinguistic experiments showed that reversible passive sentences in isolation took longer
for subjects to process than actives (Gough 1965). This result was taken to mean that passives
are more complex in processing. However, processing time was reduced when passives were
presented in the context of previous mention of the subject referent of the passive (Turner &
Rommetveit 1968, Davison & Lutz 1985). When passives and actives were presented in context,
few processing differences were found by Gordon & Chan (1995).
are more frequent than full passives. In the texts in this study, short passives are
also more frequent than full passives.
Existential there. In this construction the underlying subject is displaced to
the predicate with there as the surface subject and the copula as main verb. The
displaced phrase receives the sentence accent. There-sentences are all-focus,
thetic rather than categorical, and therefore have no topic phrase. Given the
lack of a topic phrase, the displaced-subject entity is the locus for presentational
progression.
Sentences of this type present new entities. As Lambrecht puts it, “the
sentence expressing the thetic proposition introduces a new element into the
discourse without linking . . . to an established topic or to some presupposed
proposition” (1994:144). Example (2) illustrates:
(2) Products made in the new factories differed from traditional products only in
that they were uniform, with fewer defects than existed in products made by
any but the top craftsmen of earlier periods.
There was only one important exception, one new product, in those first fifty
years: the steamboat, first made practical by Robert Fulton in 1807. It had
little impact until thirty or forty years later.
In this example “there” introduces the “one exception,” which is not prepared
for in the preceding text. The entity introduced by the displaced NP is the point
of the message, not the starting point (Hannay 1991:138). In addition to the
verb to be, verbs of appearance or emergence such as appear, come, exist are
found in this construction; some researchers treat sentences with the latter class
of verbs as a distinct type of construction (Birner & Ward 1998).
There is a strong requirement on familiarity status in existential there-
sentences: the entity presented must be hearer-new or discourse-new. Typically
the displaced NP in a there-sentence is indefinite. However, it need not be.
Felicitous examples with definite NPs are often cited, e.g. “There strode into
town the ugliest gunslinger alive.” In an extensive discussion, Birner & Ward
show that definite NPs can appear in there-sentences when they refer to a hearer-
new discourse entity. They suggest that the frequency of indefinites results from
the strong requirement that the referent be new (1998:120ff).
Cleft and pseudo-cleft. These constructions syntactically highlight a phrase
by extraction. The extracted, “clefted phrase” is introduced with the main verb
to be, and followed by a “cleft clause” that has a missing element, similar to
a relative clause. It-clefts have expletive it in subject position; wh-clefts, also
known as “pseudo-clefts,” have a wh-phrase as subject.
Cleft sentences semantically presuppose the material in the cleft clause; they
cannot be directly questioned with felicity and are unaffected by question or
negation of the main clause. These are the standard tests for presupposition.
Clefts suggest a set of alternatives, and in positive sentences introduce an entity
that satisfies the predicate of the cleft clause. Canonical sentences that corre-
spond to clefts are cognitively synonymous, but do not have these semantic
properties. Cleft sentences are derived states due to the main verb to be, and
have the discourse properties of states. Wh-clefts are essentially equative, ac-
cording to Heycock & Kroch (1999). In spite of their complex surface structure,
wh-clefts (also known as pseudo-clefts) behave like canonical sentences, even
in their potential for connection between sentences.
The topic and focus of it-clefts depend on sentence accent and context. If
unaccented, the clefted phrase is often the topic phrase, as in (3a). If the clefted
phrase receives the sentence accent, it is the focus, as in (3b–c):
(3) a. The newly hatched Formica workers . . . forage for nectar and dead arthro-
pods, regurgitate food to colony members, remove wastes and excavate new
chambers. When the population becomes too large for the existing nest, it
is the 3,000 or so Formica slaves that locate another site and physically
transport the 2,000 Polyergus workers, together with eggs, larvae, pupae
and even the queen, to the new nest.
b. we sometimes use the stated intentions of dictators against them while ral-
lying world opinion. But to what end? When Stalin engineered the world-
wide peace movement in the early 1950s, or Slobodan Milosevic promises
to allow “peace-loving” Albanian refugees to return, it is our own inten-
tions and our own public opinion that these dictators are reflecting and
manipulating.
c. The truly revolutionary impact of the Information Revolution is just be-
ginning to be felt. But it is not “information” that fuels this impact. It
is not “artificial intelligence.” It is not the effect of computers and data
processing on decision-making, policymaking, or strategy.
The topic phrases in (3b–c) are these dictators and this impact; the referents of
both phrases are familiar from the context.
Two main types of it-clefts are recognized, according to whether the clefted
phrase is topic or focus. I will call them “comment-clauses” and “topic-clauses,”
following Hedberg (1990). In comment-clause clefts such as (3a), the clefted
phrase is the topic and the clause the comment. The sentence accent falls within
the cleft clause. Comment-clause clefts are often used with a backgrounding
effect in discourse, Prince (1978) notes. In topic-clause clefts like (3b–c) the
clefted phrase is the focus and receives the sentence accent; the topic phrase
appears in the cleft clause. Topic-cleft sentences are natural answers to ques-
tions; they may also indicate contrast. And because the information in the cleft
clause of a topic-cleft is familiar, the cleft clause may often be omitted. Fragment
(3c) exemplifies both contrast and the omission of the cleft clause.
Traditionally, clefting is felicitous when there is a prior basis for the presup-
posed information. Thus cleft sentences are natural answers to questions, with
the sentence accent on the extracted constituent. For instance, “Is it john who
writes poetry? No, it is b i l l who writes poetry” (Chomsky 1971, Jackendoff
1972). Neither point, however, holds generally. Clefts often appear with no
prior basis; and they allow more than one placement of the sentence accent.
In fact, new information often appears in the presupposed, clefted clause;
this is perhaps the most frequent use of clefts according to Prince (1978).
It may seem contradictory that a syntactic structure implying presupposition
is used to present new information. But there is no contradiction. Rather, the
clefted clause conveys pragmatically that the presupposed information is to be
accepted as shared knowledge. Receivers automatically assume the information
in the clefted clause, without engaging in a process of accommodation. Thus
the notion of presupposition that is relevant for clefts does not require mutual
knowledge (Delin 1995).
It-clefts also appear in a third use, where the clefted phrase gives information
that sets the stage for the situation in the cleft clause. In this case both topic and
focus phrase fall within the cleft clause. The example in (4) illustrates, repeated
from Chapter 2. The cleft phrase sets the scene:
(4) I feel reasonably certain of the final verdict on the current impeachment affair
because I think history will see it as the climax of a six-year period marred by
a troubling and deepening failure of the Republican party to play within the
established constitutional rules.
It was on Election Night 1992, not very far into the evening, that the Senate
minority leader, Bob Dole, hinted at the way his party planned to conduct
itself in the months ahead.
The clefted phrase is adverbial here; both topic and focus phrases of the sentence
are in the cleft clause. The cleft sentence suggests a story to experienced readers.
There are fairy tales, for instance, that open in a similar way: “It was a dark
and stormy night when the beautiful princess stole out of the castle.” The sense
of fairy tale is bolstered in this text when later one comes upon the sentence
“The Republicans were already on the road to further adventure” (the full text
is given in Appendix A). Other adverbial functions of it-clefts, especially of a
temporal nature, are discussed in Delin & Oberlander (1995).
In discourse, wh- and it-clefts have different distributions and functions. The
wh-cleft clause represents “given information, assumed to be in the hearer’s
consciousness”; the it-cleft clause tends to “mark the information in the that-
clause . . . as known” (Prince 1981:904). In wh-clefts, the clefted information
is often familiar or inferrable, as in (5a). It may also, however, be new, as
in (5b):
(5) a. Underwater they [humpback whales] can blast out at 170 decibels – louder
than a jet’s roar. They also make all kinds of lower intensity social sounds.
I wish I knew what they were talking about. What I like about humpback
research is the way it lets your imagination roam free.
b. The difference between Giuliani’s and Dinkins’s anticrime policies is not
that Giuliani hired more cops: Dinkins hired 7,000 new police in his final
years in office. What Dinkins would not do, according to celebrated former
New York police chief William Bratton, was to reorient policing around a
universal standard of nonviolent deviancy.
It-cleft sentences are more frequent in the texts in this study than wh-clefts.
Extraposition. In Extraposition a sentential subject is postposed, with ex-
pletive it in subject position. Choice of this structure affects the possible topic
phrases of a sentence. Since the expletive subject is not a denoting expression, it
cannot be the topic phrase. However, the topic phrase may be the direct object,
if there is one; the situation or proposition of the extraposed clause; or a phrase
in the extraposed clause. In some cases Extraposition sentences may not have
a topic phrase (see Chapter 10, ex. 7).
The extraposed clause is pragmatically presupposed, as are subjects gener-
ally. The presupposition attached to the clause allows this construction to be
used to present information as uncontroversial, much as the it-cleft construction
is used.
Familiarity status is the main factor in extraposition. Sentential subjects that
are postposed invariably convey new information, as in (6).
(6) a. When people try to get a message from one individual to another in the
party game “Telephone,” they usually garble the words beyond recognition.
It might seem surprising, then, that mere molecules inside our cells con-
stantly enact their own version of telephone without distorting the relayed
Actually, no one could survive without such precise signalling in cells.
b. “The only intention we have is to weaken the forces that have occupied our
country and to regain our sovereignty. That is all.” A week after Ethiopia
started an offensive that it says is aimed at ending the two-year-old war, it
is now clear that the whole of Eritrea could become a battlefield.
The information in the moved clauses is discourse-new. Plausible topic phrases

in (6a–b) are the referents of clauses and/or their subjects: “mere molecules
inside our cells” and “the whole of Eritrea.” Transitive verbs also occur with
extraposition, as in “It surprised Mary that John left early.” In such sentences
the direct object may be the topic phrase.
Extraposition is a discourse requirement for sentential subjects with
discourse-new information. For sentential subjects with discourse-old infor-
mation, extraposition is possible but not required. The generalizations are due
to Miller (2001), who examined sentential that and infinitival VP clauses in a
large corpus. He found that extraposition was chosen when the predicate of the
sentence was not discourse-new and when the sentential subject provides the
topic of the subsequent sentence in the text. Similar conclusions for inversion
were reported in Birner (1994).
Extraposed clauses tend to be relatively “heavy,” that is with many words
and/or syntactic nodes. However, not all heavy clauses are extraposed if the
information they express is discourse-old or inferrable, Miller (2001) notes.
(7) The kilns clearly specialized in the production of the porcellanous ware. It
is characterized by the extreme whiteness and hardness of the body, which
may be translucent. That the kilns confined themselves very largely to the
production of bowls, basins, dishes and plates, is evident from the immense
waster heaps in the vicinity of the kilns and from the predominance of these
shapes in collections all over the world.
The sentential subject is relatively heavy. Examples like this show that famil-
iarity status rather than heaviness is the determining factor in extraposition.
Summarizing, sentences with non-canonical subjects vary in topic possibilities.

For passives, the surface subject is the topic phrase. Cleft constructions have an
expletive subject; the topic may be the clefted phrase itself, or a phrase in the
clefted clause. Extraposed sentences have a postposed topic phrase. Existential-
there sentences are thetic, with no topic phrase. All have strong requirements
on the familiarity status of the displaced phrase.
9.2.2 Inversion
The subject is displaced to the right of the verb in inversion constructions, and
a phrase from the predicate appears in initial position. There are three main
types. In subject–auxiliary inversion the first auxiliary and subject are inverted;
other verbs follow. Quotation inversion preposes the entire quoted phrase, with
the subject and verb following. In predicate inversion a phrase from the pred-
icate is preposed. The first two are illustrated in (8); I discuss the third type
below.
(8) a. It is true that the filibuster has a long and disreputable Senate history, and
that . . . it has been used more by Democrats than by Republicans. But only
after 1992 did it become the centerpiece of opposition conduct toward an
elected President.
b. The whale . . . was visible 40 feet below, suspended head down in pure
blueness with its 15-foot-long arms, or flippers, flared out to either side
like wings. “That’s the posture humpbacks most often assume when they
sing,” Darling said.
The function shared by all inversions is highlighting of the inverted phrase,

which occurs initially, in the canonical subject position. This is particularly
clear with auxiliary inversion. In (8a) the inverted phrase “after 1992” is a
strong focus phrase, with the focus operator “only”; the topic in (8a) is the
subject, the pronoun (“it,” coreferential with “the filibuster”). Green notes that
quotation inversion tends to put new information first and “gets the non-quote
part of the sentence out of the way of the reader” (1974:591). The topic phrase
of (8b) is the subject of the quoted sentence. In contrast, some inversions put
familiar information in the initial position of a sentence.
Predicate inversion is another inversion construction. It has an argument
phrase from the predicate in initial position, followed by the verb; the subject is
displaced to the right of the verb. Familiarity status is the key factor in this type
of inversion, Birner & Ward (1998) show. The preposed phrase must represent
information that is more familiar, or discourse-old, than the information of
the displaced subject. This is almost the same constraint that they propose for
passives. The difference is that, with predicate inversion, familiar information
appears in the initial position of a sentence rather than as grammatical subject.
(9) a. Darwin accepted the postulate of the fusion of a mother’s and father’s char-
acteristics in their offspring, even though his own experiments on pigeons
refuted it. Much more surprising are two other errors: he failed to acknowl-
edge the reality of species as non-interbreeding sets of populations, and . . .
that new species originate predominantly through geographic isolation.
b. The brown-headed cowbird does the same. Each bird has evolved so that
it produces eggs that match those of its chosen baby-sitter.
Even more varied than these avian parasites are the slave-making ants.
In (9a) the adjective phrase with “surprising” is more familiar in the discourse
context than “two other errors.” The sentence is presentational, with no topic
phrase. The preposed PP in (9b) is linked to the context: “these” is coreferential

with the subjects of the preceding clauses “each bird” and “it.”
Locative inversion is a sub-type of predicate inversion with the same relative
constraint on familiarity status. Example (10) illustrates. The fragment is from a
book chapter in which Europe and the Cro-Magnons are both introduced earlier
and are thus familiar.
(10) The evidence for a localized origin of modern humans, followed by their
spread . . . seems strongest for Europe. Some 40,000 years ago, into Europe
came the Cro-Magnons, with their modern skeletons, superior weapons, and
other advanced cultural traits.
The preposed phrase refers to Europe; mentioned in the prior sentence, it is more
familiar than the post-verbal phrase. The topic structures of such sentences vary
according to context. Here the inverted phrase functions as the topic; Coopmans
calls such phrases “topicalized adverbial PPs” (1989:735). Bolinger notes that
the initial phrase in locative inversions often has an almost visual “staging
effect” (1977:94). Locative inversion may also be presentational, with no topic
phrase. The verb of such sentences is the copula or one of a small set of motion
verbs.
Predicate inversion also applies to existential-there sentences.
(11) No other group of adults young or old is confined to an age-segregated envi-
ronment, much like a gang in which individuals of the same age group define
each other’s world.
In no workplace, not even in colleges or universities, is there such a narrow
segmentation by chronology.
The displaced subject is the focus phrase. The familiarity status requirement
for there-inversion is the same as for such sentences generally: the dislocated
phrase must be discourse-new.
The topics of inversion structures vary. In subject–auxiliary inversion and
quotation inversion, the topic is often the inverted surface subject. In predi-
cate inversion the sentence may be presentational, with no topic phrase; or the
inverted phrase may be the topic phrase.
9.2.3 Argument preposing and postposing

Argument preposing structures have an initial, pre-subject phrase that is dis-
placed from its canonical position in the predicate. In that position, a non-subject
phrase is salient: it is adjacent to the preceding context, first in the sentence,
and highlighted by virtue of its non-canonical position. The main preposing
constructions are Topicalization, Focus Preposing, and Left-dislocation. They
tend to occur in speech rather than written texts. Examples are relatively rare
in the texts in this study; the only ones found were in narrative fiction, which
mimic the patterns of speech. The argument postposing constructions are Right-
dislocation, Dative Alternation, and Heavy NP Shift.
In Topicalization, an argument of the verb or the entire verb phrase is preposed
to initial position. The preposed phrase must be linked to the context in the strong
sense of linking developed above.
(12) “Why, what am I a-thinking of!” said Toby, suddenly recovering a position as
near the perpendicular as it was possible for him to assume. “I shall forget my
own name next. It’s tripe!” Tripe it was; and Meg, in high joy, protested he
should say, in half a minute more, that it was the best tripe ever stewed.
In this example the preposed NP “tripe” is linked by identity to the context.

The preposed phrase need not be the topic of a sentence, so that the term
“Topicalization” is something of a misnomer. However, preposed arguments
are often topics. In topicalizations from the corpus of Birner & Ward (1998),
the preposed phrase was topic in over 75 percent of the examples. Contrast was
also a significant factor, accounting for 49 percent of the examples. (Birner &
Ward 1998 do not consider the relation between these factors.)
Focus Preposing has a preposed argument phrase that receives the sentence
accent and is the focus phrase of the sentence. The preposed phrase presents
new rather than familiar information, as in (13):
(13) It was the voice of this same Richard, who had come upon them unobserved,
and stood before the father and daughter, looking down upon them with a
face as glowing as the iron on which his stout sledge-hammer daily rung. A
handsome, well-made, powerful youngster he was; with eyes that sparkled
like the red-hot droppings from a furnace fire.
Both Topicalization and Focus Preposing structure the propositions they repre-
sent into a focus and focus frame (Partee 1991, Prince 1986). The open proposi-
tion conveys that an entity has a certain attribute; the focus represents the value
of that attribute.
Left-dislocation has an initial phrase followed by a full sentence. The phrase
is coreferential with a pronoun in the sentence and may function as the topic,
as in: “My grandmother, I remember when she used to work” (from Prince
1997). The familiarity status of the preposed phrase varies: it may express in-
formation that is familiar and linked to the context, or discourse-new. In the
latter case the coreferential pronoun tends to appear in a position that is not
favored for discourse-new entities, often an embedded subject as in the grand-
mother example. Prince (1997) suggests a processing account of the structure:
Left-dislocation simplifies the processing of a sentence by moving a discourse-

new referent from a disfavored to a more favorable position.5
Right-dislocation is formally the mirror image of left-dislocation: it has a
pronoun subject and a final unaccented phrase coreferential with the subject.
But the construction differs, predictably, in the familiarity status of the sentence-
final phrase. The right-dislocated phrase tends to express information that is
familiar, as in (14):
(14) Miss Kate and Miss Julia were there, gossiping and laughing and fussing,
walking after each other to the head of the stairs, peering down over the
banisters and calling down to Lily to ask her who had come.
It was always a great affair, the Misses Morkan’s annual dance. Everybody
who knew them came to it.
This construction is used when the topic referent is relatively salient, but not
enough to be identified with a pronoun alone, according to Lambrecht. Right-
dislocation signals to the receiver that the topic referent will be named at the
end of the sentence (1994:203). The two dislocation structures are sometimes
used in speech to maintain reference to a topic and thus maximize discourse
continuity, according to Givón (1983).
Dative Alternation and Heavy NP Shift are rightward-movement construc-
tions that move phrases to the right in the predicate of a sentence. In Dative
Alternation, a direct object phrase is moved over an indirect object and the
preposition is deleted: for instance, “We gave a book to Mary” alternates with
“We gave Mary a book.”6 In Heavy NP Shift, a long and/or complex direct ob-
ject or complement is moved over another phrase in the predicate. The phrase
that is moved is the focus phrase, receiving the sentence accent. In English this
accent may hold for the smallest phrase or for successively larger phrases.
(15) a. Dative Alternation
First, he backed an end to forced busing. Then, he supported Republi-
can Governor George Voinovich’s radical school choice law, which offers
5. Prince discusses Left-dislocation cases in which the pronoun rescues what would otherwise be
an island violation. These constructions occasionally occur in writing; there were none in the
texts in this study – for instance: “There are always guests who I am curious about what t h e y
are going to say.” Prince notes that this marginally acceptable sentence would be impossible
with a gap in the position of the pronoun. Right- and Left-dislocation in French are discussed
in Lambrecht (1994).
6. There are several differences between canonical Dative sentences and the Dative Alternative
construction. The alternation is possible only for sentences with certain verbs; Gropen et al.
(1989) attempt to characterize them on semantic grounds. Green (1974) shows that the two
structures may have different entailments: “I taught Latin to the students” entails that they
learned Latin, whereas “I taught the students Latin” does not have such an entailment.
students vouchers at parochial as well as public schools. Then, he made

city workers compete against private firms for garbage collection, road
maintenance and other contracts.
b. Heavy NP Shift
Reasonable people may differ on the quality of the evidence Mr. Starr
has accumulated . . . But his appointment in the first place is impossible to
defend. Republicans selected as independent counsel a lawyer who was
already involved in consultations with a plaintiff suing the president in a
civil dispute.
Since the rightmost phrase in a sentence receives the accent, these constructions
result in a focus phrase that would canonically be in another position. Thus the
discourse function of postposing is to introduce a marked focus phrase (Hajicová
and Sgall 1987; see also Erteschik-Shir 1979).
Familiarity status is an important factor in non-canonical constructions, as the
foregoing discussion has established. Yet in rightward movement the heaviness
of the moved phrase also seems to be significant, cf. Hawkins (1994). Often,
but not necessarily, the moved phrases in the Dative Alternation are relatively
heavy; the rightmost phrase in (15a), for instance, is longer and more complex
than the one that precedes it. The Heavy NP Shift construction by definition
involves a heavy phrase.
The relative importance of heaviness and familiarity status in Dative Alter-
nation and Heavy NP Shift is studied by Arnold et al. (2000). The authors ex-
amined instances of these constructions in a corpus consisting of parliamentary
debate transcripts from Canada; they also elicited dative sentences experimen-
tally. Heaviness was measured by number of words; familiarity status was coded
as given, inferrable, or new. Arnold et al. asked which factor better accounted
for the sentences in which movement did and did not occur. Both factors were
relevant, they found: the phrases that appeared in rightmost position tended to
provide new information and to be relatively heavy.
Summarizing, non-canonical constructions that involve the verb and its
arguments highlight by position and syntax. Topic possibilities vary. No topic
phrase occurs in existential-there sentences or in some inversion constructions.
The topic phrase is the subject or leftmost phrase in passive, some clefts, some
inversions, some argument preposings. The topic phrase is in rightward position
in extraposition, some clefts, some inversions. To determine the actual patterns
of use would require a large-scale study.
What these constructions have in common is that they change phrases with
familiar information from canonical to non-canonical positions. All but two
constructions position relatively familiar information to the left and relatively
new information to the right. The exceptions are subject–auxiliary inversion

and quotation inversion.
The next two sections are devoted to discussion of non-argument preposing
and multi-clause sentences. These constructions do not involve the topic and
focus organization of a sentence. Rather, they offer a choice about how to deploy
information. Non-argument preposing affects the familiarity connectedness of
sentences in texts. Multi-clause sentences present the situations expressed in
the clauses as closely related.
9.3 Non-argument preposing: Adjuncts

Adjuncts canonically appear after the arguments in a clause. They may be pre-
posed to initial position, or to other non-canonical positions. In initial position,
adjunct phrases are highlighted and function as starting point for a sentence.
Preposed adjuncts do not affect the internal organization of the clause. They
contrast with preposed arguments in their connection to context.
There are no familiarity constraints on preposed adjuncts. Connections of all
kinds, including no connection, are found between a preposed adjunct and the
context. Some preposed adjuncts presuppose the information that they express;
others do not.
Many types of phrases can be preposed, including adverbs ( frankly, unfor-
tunately, carefully), locative PPs (in the garden, at noon), adjunct PPs (with
little difficulty, in response to the inquiry, like fish farming), condensed clauses
with non-finite verbs (carrying his briefcase, backed into a corner), temporal
clauses (when S, until S), purpose clauses (in order to S), and adjunct clauses
(because S, unless S).7 Sentence connectives such as as for, and, yet belong to
a separate category because they appear only at the beginning of a sentence.
Locative and adjunct PPs. Preposed time and place adverbials provide a
location – temporal, spatial, or metaphorical – for the situation expressed in a
sentence. The examples in (16) illustrate:
(16) a. But how do circuits within cells achieve this high-fidelity transmission?
For a long time, biologists had only rudimentary explanations. In the past
15 years, they have made great progress in unlocking the code that cells
use for their internal communications.
7. Quirk et al. (1985) distinguish between adjuncts and disjuncts. Adjuncts pertain to the situation
in the main clause and allow syntactic processes of focusing with clefts, pseudo-clefts, negation,
question, and focusing operators. Disjuncts comment on the style, form, or content of what was
said; they are peripheral to the clause and do not allow syntactic focusing (1985:1070–71).
b. At the turn of the century, city governments often took aim at monopo-
listic private corporations and utilities. Now, the focus is more often on
monopolistic government agencies.
c. Our French teacher, a crusty character named Bertram Bradstock, made it
clear that in studying French, speaking it was an unnecessary luxury.
The first preposed adverb in (16a) is unrelated to the context; the second is
linked to the first, since both express time intervals. The contrast from one
preposed adverb to the other contributes to (16b). These adverbs set the scene
for the situation expressed in the sentence.8 In (16c), the metaphorical location
of studying French is evoked by the prior phrase “our French teacher”; the
example is from Lambrecht (1994).
The fragment in (17) has three preposed adjunct phrases:
(17) 1 The low price has been a mixed blessing . . . 2a With little incentive for
drillers to find and tap new oil, supplies eventually dropped, b and in the past
year the Organization of Petroleum Exporting Countries deliberately dropped
its production as well. 3 In response to the law of supply and demand, prices
have now risen.
In clause 2b the preposed phrase is temporally locative; the preposed phrases

in 2a and 3 indicate more abstract relations wth nominal clauses. Nominals
introduced by a preposition are temporally related to the main clause. None of
these adjuncts is directly related to the prior context, by linking or a weaker
relation.
There is a difference in the status of the information conveyed by locative
and other adjuncts. Locationals are part of the assertion of a sentence, whereas
non-locational PPs are presupposed. To demonstrate this I use the “lie test,”
due to Erteschik-Shir & Lappin (1979). These authors have shown that one can
felicitously deny the truth of an assertion but that denial of presupposition is
infelicitous. To test whether the information in a given clause is asserted or
presupposed, then, one constructs a sentence that denies its truth. If the denial
is felicitous the information is asserted. Example (18) uses this test with the
locative preposing sentence from (16a):
8. There have been suggestions from time to time that scene-setting adverbials are topic phrases.
Halliday’s notion of theme essentially takes this view (1967). In some cases it is not implausible
to claim that the adverbial is the topic referent, the locus for truth-conditional assessment and
for storage of the information in the sentence. For (16a) one can say, “Speaking of the past
fifteen years, biologists have made great progress,” though the other topic paraphrases proposed
in Chapter 7 are less felicitous. But in my judgment this analysis fails to account adequately for
the main clause. Lambrecht calls adverbials like this “secondary topics.” For him, scene-setting
pertains to aboutness but does not have topic status.
(18) In the past 15 years, they [biologists] have made great progress in unlocking
the code that cells use for their internal communications.
a. That’s not true: they haven’t made great progress.
b. That’s not true: it wasn’t in the past 15 years.
Both denials are felicitous, which shows that the information of the locative is
asserted. In contrast, the information of non-locative PPs is presupposed rather
than asserted. Example (19) illustrates, using sentence 2 from the fragment
in (17):
(19) With little incentive for drillers to find and tap new oil, supplies eventually
dropped.
a. That’s not true – they didn’t [drop].
b. #That’s not true – there wasn’t [little incentive].
The denial of the main clause is felicitous but that of the adverbial is
not. The information in many adjunct clauses is also presupposed, as noted
below.
Semantically, preposed material is outside the scope of negation and other
operators in the main clause. For instance, the sentence “George doesn’t eat
chocolate because he wants to be thin” is ambiguous. On one reading, the
because clause is within the scope of negation, on the other it is not. In preposed
position, however, the clause is outside the scope of negation: “Because he wants
to be thin, George doesn’t eat chocolate.” This difference holds for adverbials
generally (Davison 1984). One function of preposing, then, is to remove the
adjunct from the scope of semantic operators in the main clause.
Condensed clauses are sentences with participial or infinitival clause ad-
juncts. They are tightly bound to the main clause, as discussed in Chapter 6.
Condensed clauses usually have null subjects and lack a tense morpheme; (20)
illustrates.9
(20) a. 1 it was the explosive growth of the steam-engine based textile industry
that revived slavery. 2 Considered to be practically dead by the Founders
of the American Republic, slavery roared back to life as the cotton gin –
soon steam-driven – created a huge demand for low-cost labor and made
breeding slaves America’s most profitable industry for some decades.
9. English allows condensed clauses with overt subjects, but they are relatively rare in the modern
language. I found one example only in the texts for this study:
The moulds were made of a slightly greyish, compact stoneware clay with a concave
recess on the underside, the upper surface being ornamented with a design of neatly
carved intaglio.
3 Punctuated by the most spectacular busts in economic history, the boom

continued in Europe for thirty years, until the late 1850s, by which time
most of today’s major railroads had been built.
b. In “The Education of Henry Adams,” describing his college experience
under a curriculum that had not changed in several decades, Adams said
he had received an 18th century education when the world was plunging
into the 20th.
The subject of the condensed clause is coreferential with the main clause subject,
and reinforces its role as topic phrase. Condensed clauses are part of the assertion
of a sentence. The condensed clause in S2 of (20a) is related to the context by
the prior phrase “revived slavery.” The other condensed clauses in (20) express
information that is new to the discourse. The condensed clause is temporally
dependent on the main clause; the situation it expresses precedes or overlaps
that of the main clause.
Temporal clauses with after, as, before, when, while, unless, etc., temporally
locate the situation of the main clause.10 The clauses are complete but depend
temporally on the main clause, with a requirement of tense harmony.
(21) a. At his news conference here, even before he took questions, Schroeder
implicitly challenged the official US explanation for the bombing of the
Chinese embassy in Belgrade – that target analysts relied on a faulty street
map – by renewing his demand for a formal NATO inquiry into the bombing.
b. The 19-member alliance is at a critical crossroads: as time runs out to
prepare hundreds of thousands of ethnic Albanians to return to their homes
before winter, the allies are badly divided over the timing and conditions
for the use of ground troops.
c. Soon after SH2 domains were identified, investigators realized that these
modules are present in well over 100 separate proteins.
The condensed clauses in these examples introduce new information to the dis-
course. The position of a temporal adverbial affects the interpretation of a sen-
tence, as de Swart (1999) points out. In preposed position, temporal adverbials
provide Reference Time for the situation expressed in a sentence. Adverbials
in canonical right position may have other functions.
10. Temporal clauses are discussed in Partee (1984), Smith (1991), de Swart (1993), Kamp &
Reyle (1993), Sandstrom (1993), among others. There are certain idiosyncrasies. For instance,
the Event of a before-clause may not take place, due to the meaning of “before”:
Mary Jane brushed past the others and ran to the staircase, but before she reached it the
singing stopped and the piano was closed abruptly.
We infer that Mary Jane didn’t get to the staircase after all.
Temporal clauses often convey presupposed information. On encountering

when, while, after in a text, the receiver assumes the information and adds it
to the ongoing discourse representation. The passage in (22) gives a striking
example of a when-clause from the story Un coeur simple by Flaubert, the
nineteenth-century French writer. When-clauses in French are presupposed as
in English. The example is from Whitaker & Smith (1985).
(22) The driver [of the mailcoach] . . . in a temper, raised his arm, and gave a full
swing of his big whip, lashing her [Félicité] from the stomach to the nape of
her neck, so that she fell to the ground on her back.
The first thing she did, when she regained consciousness, was to open her
basket.
In this fragment Félicité, the main character of the story, tries to stop a mailcoach.
The when-clause informs us as an accepted event that Félicité came back to
consciousness (and indirectly that she had lost consciousness); the receiver adds
these Events to the ongoing discourse model. This is a compact way to convey
information and gives it a certain authority since no source or justification
is cited. Unless, until, and certain cases of before (see footnote 10) do not
presuppose the situations they express.
Temporal clauses usually involve a consequential relation between situations
as well as a temporal one. In some cases the temporal aspect is less important
than the consequential, as the examples in (23) illustrate:
(23) a. When inflation is taken into account, that 1999 price was the lowest in
modern history, while oil has gone above today’s seemingly high price
several times.
b. When puberty meets education and learning in modern America, the victory
of puberty masquerading as popular culture and the tyranny of peer groups
based on ludicrous values meets little resistance.
Indeed, a sentence asserting a purely temporal relation is odd unless a con-

sequential relation also exists, as Moens & Steedman (1987) emphasize. For
instance, the sentence “When my car broke down the sun set” is odd in iso-
lation because there is no obvious relation between the events. But in certain
contexts the sentence would be felicitous – for instance, if the reader knew that
the narrator was in a difficult situation, far from help.
Discourse relations are sometimes necessary to understand the contribution
of a clause with a temporal connective. For instance, while-clauses may be
temporal or concessive, as Caenepeel (1997) observes. Example (24) illustrates
while-clauses of both types:
(24) a. Brunel’s tunnel shield, a giant iron frame, was forced through soft soil by
screw jacks while miners dug through shuttered openings in the shield’s
forward face.
b. While Darwin came in for severe criticism from other scientists and in
turn often expressed his disagreement with their views, he responded cour-
teously, used scientific arguments, and completely avoided personalizing
disputes.
c. In the United States, we have lost over 500,000 jobs in the oil industry
while we have grossly increased our dependency on foreign oil; we now
import 55 percent of what we use.
In (24a) “while” is temporal; in (24b) it is concessive; in (24c) “while” has both

a temporal and a concessive meaning. Caenepeel argues that both sentence-
internal and discourse factors determine the interpretation of a while-clause as
temporal or concessive. The relevant factors that she mentions are whether a
clause expresses an Event or a State; the position of the clause; parallelism and
contrast in the context; and whether the context is narrative or discursive.
Adjunct clauses introduced by because, although, etc., also appear in pre-
posed position. Like temporal clauses, many of them presuppose the informa-
tion that they express. Also like temporal clauses, they may be linked to the
context, otherwise related to it, or unrelated to it. The connectives that appear
in combined clauses are a large class, including as, despite, whereas, since,
though, if (conditional), even if , etc. Many of them introduce nominals as well
as clauses; several require modals or subjunctives.
9.4 Multi-clause sentences

Sentences with more than one clause are non-canonical. They represent the
choice of the writer to organize information in sentential chunks.
The encoding of a sentence as full or reduced, dependent or independent, is
part of presentation. Sentences with more than one clause may have relative
clauses and adjunct clauses. Conjoined sentences, and sentences with more than
one independent clause, are also in this class. The first two groups have clauses
that are traditionally subordinate. It has been suggested that subordinate clauses
contain given information, especially since many are presupposed (Bever 1970,
Givón 1979). However, empirical studies show that this hypothesis cannot be
maintained. Relative clauses, for instance, often express information that is new
to a discourse (Prideaux 1993, S. Thompson 2001).
Relative clauses. Restrictive and appositive relatives are referentially depen-
dent on a shared NP. Temporally, they may be dependent or independent of the
main clause. Relative clauses typically have the shared NP referent as topic:
they are about their topics, as Kuno (1976) argues. However, relatives of locative
phrases and locative inversion are exceptions to the generalization.11
The function of a restrictive relative clause is to modify the shared Noun-
Phrase. More precisely, the relative contributes to a referring expression by
narrowing down the set of possible referents. Example (25) illustrates restric-
tive relative clauses on subject and object NounPhrases:
(25) Restrictive Relative Clauses

a. The political machines that men like White despised had been created to
bridge the material and cultural distance separating the immigrants from
native society. In cities that offered newcomers few formal services, the
immigrant-driven city machines supplied housing, fuel, charity and – most
importantly – jobs.
b. A female cuckoo, for instance, lays her egg in the nest of another species,
such as a warbler, and leaves it for the host to rear. The brown-headed
cowbird does the same. Each bird has evolved so that it produces eggs that
match those of its chosen babysitter.
The material in a subject restrictive relative like (25a) is pragmatically presup-

posed, since the existence of the subject referent is presupposed. There are no
familiarity status requirements for relative clauses: the information may be
familiar, new, or inferrable from the preceding discourse. Non-subject relative
clauses like (25b) are not pragmatically presupposed, and tend to present new
information.
Appositive relatives give information about their head nouns; they may follow
the head noun or appear elsewhere in the sentence, often as a parenthetical.
(26) Appositive Relative Clauses

a. Dragons, which are relatively uncommon in earlier examples, are usually
arranged almost heraldically round a conceptual center point, so that like
the floral scrolls they appear to rotate.
11. For instance, the following sentence has three relative clauses on NPs after the colon, all
preposed with prepositions, and all underlined.
Between these rival ends ran parallel lines of side-dishes: two little minsters of jelly,
red and yellow; a shallow dish full of blocks of blancmange and red jam, a large green
leaf-shaped dish with a stalk-shaped handle, on which lay bunches of purple raisins
and peeled almonds, a companion dish on which lay a solid rectangle of Smyrna figs, a
dish of custard topped with grated nutmeg, a small bowl full of chocolates and sweets
wrapped in gold and silver papers and a glass vase in which stood some tall celery stalks.
All three are inverted locatives: presentational clauses in which the postverbal NP is the new
information.
b. All the humpbacks in a given region sing the same song, which is constantly
evolving.
Appositive relative clauses in rightmost position may advance narrative time,

as in “She gave the letter to the clerk, who copied it” (Depraeterre 1996:699).
When they appear with non-argument NPs, relative clauses do not contribute
to the topic structure of a sentence. Example (27) illustrates relative clauses on
both argument and non-argument NPs. The first sentence has a relative clause
on the object NP, an argument; the second has a relative clause on the head
noun of the adverbial.
(27) 1 As they [workers] moved on, they removed mud, blasted through rock, and
bolted together the iron rings that would form the lining of the tunnel. 2 On a
good day, they moved about 40 feet. 3 Near the New York shore, where they
ran into a thousand feet of solid rock, progress slowed to less than a foot a
day.
In the S1 relative the topic is “that,” coreferential with “the iron rings.” The S3
relative clause has “the New York shore” as topic; this is the noun of a locative
PP, and does not contribute to the topic structure of the main clause.
Adjuncts. Adverbials, PPs, and adjunct clauses of all types appear in rightmost
position after the verb. In this position the information expressed is not presup-
posed. This presentation emphasizes the relations between clauses or PPs. The
contribution of the different connectives is usually interpreted in terms of the
discourse relations, as Matthiessen & Thompson (1988) argue.
Conjunctions also emphasize the close relations between situations. Con-
junction is the only other context besides the condensed clause that allows null
subject pronouns in English, as in (28):
(28) The Littleton killers felt trapped in the artificiality of the high school world
and believed it to be real.
Discussion of the many types of conjunction is beyond the scope of this chapter.
Complex sentences. In written texts one sometimes encounters complex sen-
tences related by connectives, semicolons, colons, and/or dashes. The examples
in (29) illustrate:
(29) a. In the United States, we have lost over 500,000 jobs in the oil industry
while we have grossly increased our dependency on foreign oil; we now
import 55 percent of what we use.
b. Mr Browne, whose face was once more wrinkling with mirth, poured out
for himself a glass of whisky while Freddy Malins exploded, before he had
well reached the climax of his story, in a kink of high-pitched bronchitic
laughter and, setting down his untasted and overflowing glass, began to
run the knuckles of his left fist backwards and forwards into his left eye,
repeating words of his last phrase as well as his fit of laughter would allow
him.
Such complex sentences are an artifact of the written mode and the subtle
possibilities of punctuation.
9.5 Paragraphs
Paragraphs are text units, integral parts of modern written texts. Paragraphs
organize the text into chunks, each set off with spaces and indentations. The
beginnings and endings of paragraphs are salient positions in the text, as are very
short paragraphs. Paragraphs vary in length and organization. They represent a
writer’s choices of chunking and highlighting information. The presentational
aspects of a text thus include its paragraphs. The authors’ paragraphs have been
retained in the examples in this book.
One question is whether paragraphs coincide significantly with Discourse
Modes; another is whether presentational patterns relate to paragraph units.
More generally, we ask whether paragraphing is significant for understanding
text structure.
Although some regularities can be found in paragraphs, they do not conform
to any single pattern or convention. Linguistic, rhetorical, and psychological
studies agree that paragraphs are flexible and can be used in more than one way.
Some paragraphs are about a particular theme or topic, others are not. Some
paragraphs are long; others are short, drawing attention to a particular point or
direction in a text. There is no one convention but rather a set of possibilities.12
Linguists have discussed the nature of the paragraph as a semantic unit.
According to Giora (1983), semantic unity revolves around a discourse topic,
which tends to hold for a paragraph. She suggests that discourse topics are
analogous to sentence topics and can be recognized with the intuitive notion of
aboutness. Another approach to paragraph unity is taken by Longacre (1979)
and Hinds (1979), who argue that cohesive and pragmatic relations between
sentences unify the sentences of a paragraph. The relations they invoke are those
currently known as discourse relations (see Chapter 11). Of course, discourse
relations are not peculiar to paragraphs. They are relevant at all levels of a text:
sentences, parts of paragraphs, paragraphs, and larger segments. This point is
12. Chafe (1994) suggests that paragraphs in written texts correspond to topic boundaries in speak-
ing. They can appear at different levels, according to the writer’s choice.
9.5 Paragraphs 237
made with particular force by Mann & Thompson (1987, 1992), who use the
same discourse relations to structure texts at successively higher levels. Fries
(1983) argues for the thematic unity of paragraphs as structural units, using
Halliday’s notion of theme as the starting point of a sentence.
Coreference patterns often provide linguistic cues to paragraphs. Shifts from
pronoun to full NPs tend to coincide with paragraph breaks in narrative and
expository prose, according to Fox (1987a,b). The impetus for such shifts is
usually a change of theme, or topic. Longacre (1979) suggests that corefer-
ential expressions indicate thematic unity. Other cues are change of direction
or discontinuity, “theme-marking” with non-canonical syntactic structures, and
repetition of terms within a paragraph (Crothers 1979, Bond & Hayes 1984,
Hoey 1991). In some languages there are linguistic forms that have the para-
graph as their domain.13
Paragraph boundaries can be significant indicators of the linear organization
of texts. In a study of discourse topic continuity and discontinuity, Goutsos
(1997) found that paragraph breaks are powerful signals of a shift from one
discourse topic to another.14 However, there is another important pattern for
the introduction of new discourse topics. In some texts, a new discourse topic
is introduced in the final position of a paragraph and then developed in the
paragraph immediately following (Giora 1983).
There is no single pattern for paragraphs and Discourse Modes. Discourse
Modes are often maintained within a paragraph; but a Discourse Mode may
shift in the middle of a paragraph. Examples given in Chapters 2 and 5 have a
shift of Discourse Mode that coincides with the paragraph break; and shifts that
do not so coincide. In other cases, passages of a Discourse Mode may continue
for more than one paragraph. I conclude that there is no clear relation between
Discourse Modes and paragraphs.
The history of texts shows that they were not always divided into para-
graphs. With the development of printing and industrialization in sixteenth- and
13. Longacre (1979) points to markers for paragraph boundaries in Huichol (Mexico), Shipibo
(Peru), Gurung (Nepal), Sanio-Hiowe (New Guinea). Paez and Ica (Colombia) have anaphoric
devices with the paragraph as their domain. In Ica there is a special verbal suffix for referring to
the most prominent participant in a paragraph (Tracy & Levinsohn 1977). This is reminiscent
of the discourse uses of pronouns in proximal–obviative systems of American Indian languages
discussed briefly in Chapter 6.
14. Goutsos found that paragraphs did not necessarily coincide with discourse topic continuity –
more than one discourse topic may occur in a single paragraph. However, topics rarely continue
across paragraphs. Goutsos worked with three corpora of English expository texts. The para-
graphing for newspaper articles was slightly different than other genres, probably reflecting
journalistic norms and requirements.
seventeenth-century Europe, paragraphs became quite common. Paragraphs as

blocks of type were useful for the printer, and broke up the mass of text on
a page. Conventions about paragraphing sought to balance the demands of
rhetorical structure and logical relationships of a text (Parkes 1993:89). In the
nineteenth century people began to consider the paragraph as a unit in its own
right (McArthur 1992).
The modern rhetorical tradition of the paragraph began with Bain’s (1877)
rules for achieving unified and coherent paragraphs in expository writing. Bain
recommended that the subject of a paragraph be prominently indicated, thus
beginning the tradition of the “topic sentence” (Markels 1984:42). Rhetori-
cians, however, note that, although the topic sentence can be useful, the main
desideratum of a paragraph is coherence; see Christiansen (1965), Winkler &
McCuen (1988). A study of actual paragraphs carried out by Braddock (1974)
found many types of organization.
The psychological reality of the paragraph has been studied experimentally.
Koen et al. (1966) presented people with texts and asked people to put in
paragraphs. Their subjects generally agreed with the original paragraphing of
the texts, and with each other. Koen et al. concluded that paragraphing is a
reliable phenomenon; another study by Stern (1976) found less agreement.
Later studies, however, have also found good agreement between subjects’
divisions and original texts. Stark (1988) asked subjects to paragraph texts.
They were able to paragraph in agreement with the authors and with each other.
Stark looked for the basis of her subjects’ decisions. Neither length, nor changes
in surface subject, were consistent cues for paragraph breaks in the three texts
she studied. Quite good cues were non-canonical structures in which the subject
was not the first element and discontinuities. Stark concludes that people have
multiple conventions at their disposal: there is more than one way to segment
a text into paragraphs.
9.6 Presentational information in Discourse Representation Structure

The structures of Discourse Representation Theory represent the conceptual
meanings of a text. They arise from semantic and pragmatic information. I
have argued that some presentational information should be represented in
Discourse Representation Structure. However, I do not provide a formal ac-
count. Topic referents can’t be identified by compositional rules, for the reasons
given in Chapter 8. Focus information is complex and beyond the scope of this
discussion. I assume that a full text representation includes information about
sentence topic and focus. Such a representation would require a much more
9.6 Presentational information in DRSs 239
powerful set of construction rules than those used here. For another view see
McNally (1988).
Whether other aspects of surface presentation should be included in a DRS is
debatable. Syntactic structures have the potential for a certain rhetorical force,
from general effects such as surprise, emphasis, and contrast, to specifics such
as proposition concession and affirmation.15 Such interpretations depend on
details of the context as well as on a given structure: there is no single rhetor-
ical function for syntactic structures. For instance, proposition affirmation is
conveyed by certain inversions, Ward (1985) argues. In the cases he considers,
an affirmed proposition is evoked but not entailed by the context, as in “It’s
odd that dogs eat cheese, but eat it they do.” Concession/affirmation is a spe-
cial rhetorical meaning of inversions, according to Horn (1991). The speaker
concedes one proposition and affirms another; the second may follow from the
first but contrasts with it rhetorically, as in “They barely made it, but make it
they did.”
Subtleties of presentation such as the difference between independent and
relative clauses, or between full and condensed clauses, belong to the surface of
the text. They contribute to rhetorical interpretation. Certain close connections
between clauses are indirectly preserved in the DRS. For instance, since a
condensed clause has no tense, no independent times for that clause appear in
the DRS.
The rhetorical effects of inversion and other non-canonical constructions are
part of text meaning, however. Levinson (2000) suggests that they are General
Conversational Implicatures. They occupy a level of pragmatics between the
general Gricean maxims and the particular implicatures conveyed by a given
sentence in context. These rhetorical interpretations depend on pragmatic rea-
soning, and require access to surface structure. Such pragmatic interpretation
may take place in parallel with the conceptual. The two are distinct in kind,
perhaps belonging to different modules, Levinson suggests. I shall assume that
rhetorical meanings are treated differently from the conceptual meanings ad-
duced in this book.
There is reason to think that the surface structure of a sentence is not included
in the discourse representations that people actually construct. Evidence from
psycholinguistics shows that people do not remember the surface structures of
sentences. For instance, Sachs (1967) found that people rapidly forget details of
15. The studies of construction grammar emphasize the pragmatic meanings of particular con-
structions (as in Lambrecht 1986, Goldberg 1995, etc.). They tend to focus on internal aspects
of a structure rather than its role in a context.
sentence structure, after eighty syllables or more. Similarly, studies by Johnson-

Laird and his colleagues (1970) showed that people remember the meaning but
not the linguistic encoding of a sentence.16 Further experimental studies have
shown that the linguistic form of a sentence is standardly not remembered after
a short time, though people can remember if asked to do so (Garnham 1985).
16. It was seen that after less than a minute people were not able to recall the linguistic forms of
sentences they had seen. They remembered the meanings conveyed, however.
Example sources
(1a) and (9a) Jared Diamond, A tale of two reputations. Natural History, February
2001.
(1b), (24a), and (27) Jim Collins, How it works. US Airways Magazine, Attaché, May
2001.
(1c), (11), (23b), and (28) Leon Botstein, Let teenagers try adulthood. New York Times,
May 1999.
(2), (3c), and (20a) Peter Drucker, The information revolution, Atlantic Monthly,
October 1999.
(3a), (9b), and (25b) Howard Topoff, Slave-making queens. Scientific American,
November 1999.
(3b) Tony Judt, Tyrannized by weaklings. New York Times, April 5, 1999.
(4), (8a), and (15b) Alan Ehrenhalt, Hijacking the rulebook. New York Times, December
20, 1998.
(5a), (8b), and (26b) Douglas H. Chadwick, Listening to humpbacks. National
Geographic, July 1999.
(5b), (15a), (16b), and (25a) Peter Beinart, The pride of the cities. New Republic, June
1997.
(6a), (16a), and (21c) John D. Scott & Tony Pawson, Cell communication. Scientific
American, June 2000.
(6b) After a victory, Ethiopia looks toward other fronts. New York Times, May 20, 2000.
(7) and (26a) Margaret Medley, The Chinese Potter, London: Phaidon, 1989, pp. 106,
112.
(10) and (24b) Jared Diamond, Guns Germs and Steel. New York: W. W. Norton, 1997,
p. 40.
(12) and (13) Charles Dickens, The Chimes. Oxford: Oxford University Press, pp. 89,
93.
(14) and (29b) James, Joyce, “The Dead.” In Dubliners, 1916: reprinted, London:
Penguin Books, 1982, pp. 175, 185.
(17), (23a), (24c), and (29a) Robert Mosbacher, Cheap oil’s tough bargains. New York
Times, March 13, 2000.
(20b) Arthur Levine, The Soul of a New University. New York Times, March 13, 2000.
(21a) and (21b) Kosovo strategy splitting NATO. New York Times, May 20, 1999.
I V Discourse Modes and their
context
10 Information in text passages
This chapter brings together the main points of Discourse Mode, subjectivity,
and presentational progression; and analyzes passages of text with the tools de-
veloped above. Section 10.1 gives the basic points of each area of analysis, with
a reprise of the criteria for determining Primary Referents and sentence topics.
Section 10.2 analyzes text passages of the five Discourse Modes, according to
temporality and types of entities, I discuss temporal, spatial, and metaphorical
progression. For each passage, I also consider subjectivity and presentational
factors. Section 10.3 presents a temporal and an atemporal DRS, with informa-
tion about Discourse Mode and subjectivity.
10.1 The Discourse Modes

To introduce the discussion I summarize the characteristics of the Discourse
Modes. They fall into two classes, temporal and atemporal, according to the
main types of situation introduced in each mode.
The temporal modes are Narrative, Report, and Description. They introduce
situations that are located in the world. In Narrative, events and states are
related to each other in time; the text progresses with bounded events interpreted
in sequence, and/or time adverbials. In Reports, events, states, and General
Statives are related to Speech Time: texts progress back and forth. The mode
of Description has events and states, and time is static. The text progresses
spatially through a scene. Text progression in Description depends on lexical
information, unlike the two other temporal modes.
The Information and Argument modes are atemporal. General Statives pre-
dominate in the Information mode, while Argument has both General Statives
and abstract entities. Text progression in these modes proceeds by metaphorical
motion through the domain of the text. Motion, or lack of it, depends on the
Primary Referent in a clause; it occurs when metaphorical location changes.
The Primary Referent is semantically central in the situation expressed by a
clause. The Primary Referent of an Event is the entity which moves or changes;
243
in States, the Primary Referent is located, characterized, or emergent, as dis-

cussed in Chapter 6. The criteria for determining the Primary Referent of a
clause are listed below.
(1) Criteria for Primary Referent
Events
a. Undergoes a change of state
b. Causally affected by another participant
c. Doesn’t exist independently of the event
d. Moves or otherwise changes
States
e. Literally or metaphorically located
f. Dependent on the situation for existence
g. Figure relative to a Ground
h. Has a property ascribed to it
These criteria determine the Primary Referents in Informative and Argument

text passages. They are applied to text passages below.
10.1.1 Subjectivity
In texts of all modes there are forms that convey subjectivity, access to a
mind. The major categories are communication, contents of mind, eviden-
tiality/evaluation, and perception and perspective. To interpret the forms of
subjectivity we identify the mind responsible – the Responsible Source – and
its scope.
Verbs of communication introduce quoted speech, represented speech, and
indirect speech. Indirect speech is systematically ambiguous: responsibility for
the content may rest with the reporter or with the original speaker. In reports of
belief and other contents of mind such as “Mary believed that John was sick”
we ascribe the belief expressed in the complement clause to Mary. Perception
and perspective follow the same principles as contents of mind. The speaker is
responsible for modals, evidentials, and evaluative forms unless they are clearly
associated with a text participant.
The scope of a subjective form may extend over several sentences, so long
as subjective continuity is maintained. Subjective continuity is broken with a
change of time or place, or introduction of a new candidate for Responsible
Source.
10.1.2 Presentational progression

The surface structure of sentences determines presentation. The main factors
are the initial phrase, the final phrase, grammatical relations, how information
10.1 The Discourse Modes 245
is organized; and how sentences are connected to each other in a text. Sentences
may be linked by poset linking, in the technical sense presented in Chapter 9,
or connected by weaker relations of familiarity.
Presentational progression depends on sentence topics: texts progress from
one topic phrase to another. The subject phrase is canonically the topic of a
sentence, since the subject is salient grammatically and positionally. In non-
canonical structures other phrases may be topic. If there is no topic the entity
introduced is the locus for progression.
The topic phrase refers to what a sentence is about. The main criteria for
identifying the topic phrase are salience, familiarity, and continuity. The criteria
are listed below, repeated from Chapter 8.
(2) Criteria for Topic Phrase

An NP may be the topic phrase of a sentence if it
a. is the subject of a sentence;
b. is a pronoun;
c. realizes the agent or experiencer argument of the main verb;
d. is coreferential with the topic phrase in the preceding sentence;
e. is coreferential with a topic phrase in the context;
f. is coreferential with a phrase in the context;
g. is lexically related to other material in the context;
h. is in a parallel grammatical position with an NP in the context.
NPs with sentence accent or other focus cues are not candidates for topic.
This approach allows for some variation in the position of topic phrases.
However, there is a striking consistency in the text passages that I have exam-
ined. Topics almost always appear in the subject position of a sentence, whether
the syntactic structure is canonical or not.
10.1.3 Patterns of organization in texts

Three patterns of presentational progression were identified in Daneš (1974);
they are still useful. The topic phrases of sentences relate to each other, and to
focus phrases, in different ways. As Daneš notes, various combinations of these
patterns can appear in a text.
The first pattern, which I call Focus–Topic Chaining, is a simple linear pattern
in which the focus phrase referent of one sentence appears as the topic of the
next; the examples are from Daneš (1974):
(3) Focus–Topic Chaining

1 The first of the antibiotics was discovered by Sir Alexander Flemming in
1982. 2 He was busy at the time investigating a certain species of germ which
is responsible for boils and other troubles.
“Sir Alexander Flemming” is the focused referent of S1 and the topic referent
(“he”) of S2. “A certain species of germ” is focus of S2, and topic of the
relative clause at the end of S2. The chaining relation is allowed by the syntactic
structures of the sentences. The passive S1 ends with a proper name, which is
coreferential with the subject and topic phrase of S2.
The second type has a “continuous theme”: the same topic referent recurs
in a succession of sentences. I call this type Topic Chaining, following Givón
(1983).
(4) Topic Chaining

The Rousseauist especially feels an inner kinship with Prometheus and other
Titans. He is fascinated by any form of insurgency . . . He must show an
elementary energy in his explosion against the established order and at the
same time a boundless sympathy for the victims of it. Further, the Rousseauist
is ever ready to discover beauty of soul in anyone who is under the reprobation
of society.
All sentences in (4) have the same topic, the Rousseauist. Note the shift from
pronoun subject back to full nominal in the last sentence. The shift suggests
a change of direction, perhaps the end of a paragraph, as noted in Chapters 6
and 9.
In the third type, Unchained, each sentence has a different topic and a different
focus referent, as in (5). The sentences of the passage are not unrelated, however:
they all predicate something about New Jersey, as Daneš recognized:
(5) Unchained
New Jersey is flat along the coast and southern portion; the northwestern region
is mountainous. The coastal climate is mild, but there is considerable cold in
the mountain areas during the winter months. Summers are fairly hot. The
leading industrial production includes chemicals, . . . food, coal, petroleum,
metals, and electric equipment.
The topic pattern approach does not give a way to establish the connections we
intuitively make between these sentences.
The connections can be established nicely by the poset linking of Chapter 9.
Recall that poset linking relates a referent to a salient poset that can be inferred
in the context. In (5) each subject phrase after the first sentence refers to a
characteristic of New Jersey. The poset of such characteristics is evoked by the
“New Jersey” in the first sentence. Thus the topic phrases of (5) are in effect
chained at the level of poset linking.
The topic phrases in the examples are all surface structure subjects, although
the syntactic position is not the concern of the topic pattern analysis. This is the
position that recurs strikingly across the text passages examined for this study.
Topic phrases tend to appear at the beginning of a sentence, usually as subject

and occasionally in an inverted or preposed phrase; and they tend to involve
relatively familiar information. The pattern was noticed in work of the Prague
School, in Halliday (1967), and has the status of a truism today. Nevertheless
I know of few detailed studies of texts that demonstrate the pattern and how
it comes about.1 Note that the language and the notion of topic allows other
positions for the topic phrase. In fact, I showed in Chapters 8 and 9 that topic
phrases need not be in subject position. However, topic phrases in non-subject
position are rare in the texts of this informal study.
10.2 Multiple analyses of text passages

This section presents analyzed examples of each mode, as in the first chapter;
we are now in a position to discuss and more fully appreciate the analysis.
I begin with a passage in the Narrative mode, first presented in Chapter 1.
There are two versions: the first shows the situation entities and text progression
in tensed clauses, the second adds subjectivity and topic information.
(6) Narrative a: Situations and Text Progression
5aE → Dr. P came in, a little bowed, bE and → ø advanced with outstretched
hand to the grandfather clock, cE but, ø hearing my voice, → corrected himself,
dE and → ø shook hands with me. 6aE → We exchanged greetings bE and → ø
The clauses with event verb constellations all have the perfective viewpoint,
presenting bounded events. They are taken as sequential, advancing the narra-
tive, on the continuity tense pattern. States are expressed in clauses 2b, 3a–b,
4a–c, and 7b; they are located at the preceding Reference Time on the limited
anaphora tense pattern. Since states overlap or surround Reference Time the
states are not limited to that time but are understood as continuing indefinitely
before and after it. The state sentences constitute a small descriptive passage in
themselves.
I now add information about subjectivity and topic phrases; forms of sub-
jectivity appear in bold, and the topic phrases are italicized. Several clauses
1. This pattern was demonstrated for the essay The Elements of Ethics by Bertrand Russell in
Smith (1971).
are presentational and do not have topics; for them, the phrases introducing
referents are noted with single quotes.
(7) Narrative b: Situations, text progression, subjectivity, Topic

1E → A few days later I called on Dr P and his wife at home, with the score
of the Dichterliebe in my briefcase and a variety of odd objects for the testing
of perception. 2aE → Mrs. P showed me into a lofty apartment, bS which
recalled fin-de-siècle Berlin. 3aS A magnificent old Bösendorfer stood in state
in the centre of the room, bS and all around it were ‘music stands, instruments,
scores.’ 4aS There were ‘books,’ bS there were ‘paintings,’ cS but the music
was central. 5aE → Dr. P came in, a little bowed, bE and → φ advanced with
outstretched hand to the grandfather clock, cE but, φ hearing my voice, →
corrected himself, dE and → φ shook hands with me. 6aE → We exchanged
greetings bE and → φ chatted a little of current concerts and performances. 7
Diffidently, aE → I asked him bS if he would sing.
The topic phrases are subjects in all but one clause. Most have several topic
properties. In 1, and 5–7 the topic phrases are agents. After S1 the dynamic topic
phrases are coreferential with earlier phrases; the clauses in 5–6 have parallel
syntactic structures.
The topic pattern follows the shift in mode from Narrative to Description
and back. The first narrative chunk has two clauses, 1 and 2a; the sentences are
related by Topic Chaining.2 The second chunk, a description, consists of clause
2b, and 3 and 4. 2b is linked to 2a by Focus–Topic Chaining. The other clauses
have no topic phrase. Sentences 4a–b are existential there-sentences; S3b is
a predicate inversion structure. The inverted pronoun is linked to the context,
since it is coreferential with the topic of the preceding clause. The inversion
maximizes continuity. The third chunk, also narrative, has Topic Chaining.
At S6 the focus phrase of 5c is picked up, and the topic shifts from “Dr. P” to
“we”; then to “I.”
The one subjective expression is the verb “recall” which implies an experi-
encer. Since the narration is in the first person, the Responsible Source is both
author and participant.
The next example is a passage in the Descriptive mode. The mode is temporal
and static: time does not change. Text progression is spatial through the scene.
The phrases that indicate spatial location are underlined; phrases that introduce
referents in presentational clauses are in single quotes.
2. The topic phrase of clause 2a is the direct object of a canonical S-V-O sentence. The subject
and object phrases are both possible topics: they meet slightly different criteria. The subject
“Mrs P” is coreferential with “his wife” in S1, and has the agent role. The object is coreferential
with the topic of S1 and a pronoun. I identify the object pronoun as the topic phrase, by the
criteria of local topic continuity.
(8) Description a: Situations and spatial progression

1S The front part of the shop was cluttered with goods for tourists. 2S There
were ‘locally knit sweaters’ on shelves on the walls, ‘tables of pottery’ –
some of which Mara had made – and ‘trays of trinkets’, such as key-rings
bearing the Dales National Park emblem – the black face of a Swaledale
sheep. 3S As if that weren’t enough, the rest of the space was taken up by fancy
notepaper, glass paperweights, fluffy animals and fridge-door magnets shaped
like strawberries or Humpty Dumpty. 4S In the back, though, the setup was
very different. 5aS First, there was a ‘small pottery workshop,’ complete with
wheel and dishes of brown and black metallic oxide glaze, b and beyond that
[there was] ‘a drying room and a small electric kiln.’ 6aS The workshopi was
dusty and messy, crusted with bits of old clay, b and iti suited a part of Mara’sj
personality. 7aS Mostly shej preferred cleanliness and tidiness, bs but there
was ‘something special,’ c she found, about creating beautiful objects in a
chaotic environment.
The verb constellations are all states, though of different types. The first clauses
are locational; 6b and 7 indicate personal experience. The passage begins with
a tour of the front of the shop, then moves to the back with its workshop and
drying room, and to Mara. The spatial progression is due to lexical information
and world knowledge.
This passage occurs in a novel, just after a dialogue between Mara and a
friend; context and the subjective forms indicate the perspectival standpoint of
Mara. The subjective elements are given in bold in (9).
(9) Description b: Situations, spatial progression, subjectivity

were ‘locally knit sweaters’ on shelves on the walls, ‘tables of pottery’ – some
of which Mara had made – and ‘trays of trinkets,’ such as key-rings bearing the
Dales National Park emblem – the black face of a Swaledale sheep. 3S As if that
weren’t enough, the rest of the space was taken up by fancy notepaper, glass
paperweights, fluffy animals and fridge-door magnets shaped like strawberries
or Humpty Dumpty. 4S In the back, though, the setup was very different.
5aS First, there was a ‘small pottery workshop,’ complete with wheel and
dishes of brown and black metallic oxide glaze, bi and beyond that [there
was] ‘a drying room and a small electric kiln.’ 6aS The workshopi was dusty
and messy, crusted with bits of old clay, b and iti suited a part of Mara’sj
personality. 7aS Mostly shej preferred cleanliness and tidiness, bs but there
was ‘something special,’ c she found, about creating beautiful objects in a
chaotic environment.
The introductory phrase of S3 “As if that weren’t enough” is a sentence adverb

or parenthetical; we take it as the voice of Mara, the Responsible Source for all
the subjective elements in the paragraph. The second clause of S7 gives Mara’s
thought: the inversion is typical of represented thought and speech.
The third version of the passage adds topic information. The topic phrases are
italicized; in sentences without topics, phrases in presentational constructions
appear in single quotes.
(10) Description c: Situations, spatial progression, subjectivity, topic

were ‘locally knit sweaters’ on shelves on the walls, ‘tables of pottery’ – some
of which Mara had made – and ‘trays of trinkets,’ such as key-rings bearing the
Dales National Park emblem – the black face of a Swaledale sheep. 3S As if that
weren’t enough, the rest of the space was taken up by fancy notepaper, glass
paperweights, fluffy animals and fridge-door magnets shaped like strawberries
or Humpty Dumpty. 4S In the back, though, the setup was very different. 5aS
First, there was ‘a small pottery workshop,’ complete with wheel and dishes of
brown and black metallic oxide glaze, b and beyond that [there was] ‘a drying
room and a small electric kiln.’ 6aS The workshopi was dusty and messy,
crusted with bits of old clay, b and iti suited a part of Mara’sj personality. 7aS
Mostly shej preferred cleanliness and tidiness, bS but there was ‘something
special,’ c she found, about creating beautiful objects in a chaotic environment.
When spatial location and topic phrase coincide, a phrase has both underlining
and italics.
The subject and topic phrases in all but the last sentence are related by poset
linking. The passive in S3 puts the locational phrase in subject position; the
preposed phrase in S4 also highlights location. Both S2 and 5a–b are presenta-
tional there-sentences which do not have a topic phrase. The quotation inversion
of S7b puts presentational “there” in subject position. Thus the paragraph has
relatively few topic phrases. The effect is to highlight the spatial progression.
Now consider a passage in the Report mode, from Chapter 2; as usual only
the tensed clauses are analyzed.
(11) Report: Situations, temporal progression, subjectivity, topic

1a At his news conference here, aE even before he took questions, bE →
Schroeder implicitly challenged the official US explanation for the bombing
of the Chinese embassy in Belgrade – that cE target analysts relied on a faulty
street map – by renewing his demand for a formal NATO inquiry into the
bombing.
2aE → Diplomats say that Schroeder, bE who just returned from China, cS
was angry that cE a trip dS he had long planned to herald his chairmanship of
the European Union was transformed into an official apology for the embassy
bombing.
The situation entities in this passage are all events and states. The passage is
intricate in syntax and temporal relations. There are two events in S1, related
to each other. The temporal clause has strong focus due to the operator “even.”
The event of the appositive clause is not related to the others except by world
knowledge. S2 returns to the present with a report about the past. The subjective
elements have different sources. For the deictic “here” the author is Responsible
Source. For the complement of the “say” the Responsible Source is the subject
referent, “diplomats.” Presentational progression proceeds by Topic Chaining
until S2c, where a new topic appears.
An Informative passage is presented in (12). The entities are marked as
subscripts on clauses – “Ge” for Generalizing sentence; the Primary Referent
phrases are underlined.
(12) Informative a: Situation entities, Primary Referents
1aGe The Information Revolution is now at the point bGe at which the
Industrial Revolution was in the early 1820s, about forty years cE after
James Watt’s improved steam engine (first installed in 1776) was first ap-
plied, in 1785, to an industrial operation – the spinning of cotton. 2aGe And
the steam engine was to the first Industrial Revolution what bGe the computer
has been to the Information Revolution – its trigger, but above all its
symbol. 3aGe Almost everybody today believes bGe that S [nothing in eco-
nomic history has ever moved as fast as, or had a greater impact than, the
Information Revolution]. 4aGe But the Industrial Revolution moved at least as
fast in the same time span, bGe and φ had probably an equal impact if not a
greater one.
The choices of Primary Referent are justified below. The letters following
each phrase refer to the relevant criterion for Primary Referent in (1).
(13) Justification for Primary Referents of (12)
1. a. “The Information Revolution”: e metaphorical location
b. “Industrial Revolution”: e metaphorical location
c. “an industrial operation”: b causally affected
2. a. “the steam engine”: e metaphorical location
b. “the computer”: e metaphorical location
3. a. S: e metaphorical location, object of belief
b. “the Information Revolution”: d entity that moves (metaphorically)
4. a. “Industrial Revolution”: d entity that moves (metaphorically)
b. “an equal impact”: f dependent on the situation
The Primary Referents are in parallel throughout the passage. The text pro-
gresses from the “Information Revolution” and “Industrial Revolution” to more
detail with “steam engine” and “computer”; in the last two sentences the first
parallel is resumed. There is little metaphorical motion in the passage.
Next consider the presentational aspects of the passage. The topic phrases
are in italics, the Primary Referents are underlined, and the subjective forms
are in bold. The syntax is relatively complex in the predicates of the clauses,
but this does not affect the topic phrases: they are the subject NPs in every
clause.
(14) Informational b: Topic, Primary Referent, subjectivity
1aGe The Information Revolution is now at the point bGe at which the
Industrial Revolution was in the early 1820s, about forty years cE after
James Watt’s improved steam engine (first installed in 1776) was first ap-
plied, in 1785, to an industrial operation – the spinning of cotton. 2aGe And
the steam engine was to the first Industrial Revolution bGe what the computer
has been to the Information Revolution – its trigger, but above all its symbol.
3aGeS Almost everybody today believes bGe that nothing in economic history
has ever moved as fast as, or had a greater impact than, the Information Rev-
olution. 4aGe But the Industrial Revolution moved at least as fast in the same
time span, bGe and φ had probably an equal impact if not a greater one.
Topic and Primary Referent are the same in S1a–b, S2a–b, and S4a. They differ
in clauses expressing events with effects or changes: S3b and S4b. The topic
referent tends to be the cause of a change or an effect in canonical sentences;
the Primary Referent is usually the entity that changes.
Finally I present a short passage in the Argument mode. The first version
indicates the situation entities and the Primary Referents:
(15) Argument a: Situations, Primary Referents
1aS A pretty good argument can be made that bProp the defining moment of
American democracy didn’t occur in 1776 or 1787, as commonly supposed,
but in 1801 – on the day cE that John Adams, having been beaten at the polls,
quietly packed his things and went home. 2aE Only then did we know for sure
b thatFact the system worked as advertised.
3aGe The routine transfer of power may not be the most dramatic feature of
American democracy, b but it is the most important.
(16) Justification for Primary Referents of (15)
1. a. “that S,” clausal complement: f dependent on the situation
b. “defining moment”: e temporally located
c. “John Adams”: d moves
2. a. “that S,” clausal complement: e metaphorically located (possessed
mental State)
b. “the system”: h property ascribed
3. a. “The routine transfer of power”: h property ascribed
b. “it”: h property ascribed
The first two sentences introduce abstract entities. The complement that-clause
in S1b refers to a proposition. The corresponding wh-question form would be
ungrammatical, so it meets the substitution criteria for a proposition, discussed
in Chapter 4. The complement clause of 2b refers to a fact: it is a that-clause,
and would allow substitution of the wh-question form. The entity introduced
by S1a is technically a state because of the modal; S2a introduces an event
because the verb “know” is used in an inchoative sense, meaning “come to

know.” S3 is a generalizing stative: it does not express a particular situation.
Metaphorical progression is from the specific to the general; “defining
moment,” “John Adams,” and then “the system,” “the routine transfer of power.”
Now I add subjective information. The passage has several indications of
subjectivity, all with the author as Responsible Source. The dense subjective
forms partly explain the strong sense of authorial presence.
(17) Argument b: Situations, Primary Referents, subjectivity
b thatFact the system worked as advertised.
Also contributing to the sense of subjectivity is the agentless passive, which in

context suggests the author as maker of the argument, and the lexical item for
sure, arguably an evidential.
The topic phrases are consistently subjects.
(18) Argument c: Situations, Primary Referents, subjectivity, topic
b that SFact the system worked as advertised.
The main clause of S1 is passive, putting the phrase “a pretty good argument”
in subject position. S2 has a preposed adverbial phrase with subject–auxiliary
inversion. The syntax highlights the temporal adverbial by putting it in initial
position; demotes the subject phrase from its canonical position of prominence;
and highlights the complement clause, now in final position. Thus non-canonical
syntax enables the topic phrase pattern.
The topic phrases are not related by chaining, poset linking, or familiarity.
Both presentational and text progression pick out the same phrases in four
clauses, giving the text a strong forward sense. When topic and Primary Referent
phrase are different, as in the other three clauses, the organization is more
complex.
This explication sorts out some of the information conveyed and automati-
cally processed by the reader in the text passage above.
10.3 Formalization in Discourse Representation Structures

The interpretation of text passages has been formalized in rules and represen-
tations of Discourse Representation Theory as much as possible. Where there
is a close connection to the linguistic forms, construction rules can be written.
Thus situation entities, temporal advancement, and subjectivity are recognized
by formal rules and entered into a Discourse Representation Structure.
For atemporal progression, the Primary Referents of a clause can be identified
according to thematic role information. However, the calculation of whether
metaphorical location has changed cannot be treated in the same manner. Such
calculation needs another kind of reasoning, and access to several sources of
information. Similarly, spatial progression and topic progression are beyond
the scope of this kind of discourse representation.
The structures developed here could serve as input to richer interpretation.
As I suggest in Chapter 9, this kind of pragmatic interpretation may be parallel
to the analysis given in the DRSs here.
This section presents a Discourse Representation Structure for one temporal
and one atemporal text passage. For the temporal passage, the DRS gives infor-
mation about situation type, temporal advancement, and subjectivity. The DRS
for the atemporal passage has situation type, Primary Referent, and subjectivity.
Temporal information is included but it does not key text progression for this
passage.
Because of the complexity of a full DRS I give only short, simplified passages.
The DRS below is based on the first two sentences of the narrative fragment in
(6):
(19) 1 A few days later I called on Dr P and his wife at home. 2a Mrs. P showed
me into a lofty apartment, 2b which recalled fin-de-siècle Berlin.
I give some explanatory comments on the DRS beforehand.

The times of the different clauses are listed with subscripts a, b, c. Speech
Time is always the default for t1 , it is not entered as a condition in the DRS.
Since this passage occurs in the middle of a narrative, the event of S1, e1 , is
related to a previously established past time. I assume a prior Reference Time
(RT), noted as t2a−1 ; the RT of the first sentence, t2a , follows that time, by the
principle of narrative advancement. The event of S2a, e2 , also advances the
narrative: t2b follows t2a . S2b introduces a state, which has the same RT as the
prior Event, by the principle of limited anaphora.
The coreference judgments are assumed. Coreference is automatic for “I”
and “me,” and for “a lofty apartment” and “which”; but recognizing “Mrs. P”
and “his wife” as coreferential is based on pragmatic knowledge.
10.3 Formalization in DRSs 255
The subjective form is the state verb “recall.” The implicit Experiencer argu-
ment is included with the state condition, and its content given in a sub-DRS,
indicated by dots. The Author is a discourse entity (A) at the top of the DRS, and
is interpreted as Responsible Source, RS. Identifying the RS with the narrator
is automatic for a first-person narrator since the passage has no other candidate
for this role. The referents of the “recall” relation appear in the sub-DRS and
also in the main DRS because their existence is not contingent on the sub-DRS.
(20) 1 A few days later I called on Dr P and his wife at home. 2a Mrs. P showed
me into a lofty apartment, 2b which recalled fin-de-siècle Berlin.
t1a t2a t3a x y z w e1 t1b t2b t3b u v s e2 t1c t2c t3c r q p o A
1. e1: call on (x,y,z)

2. t2a < t1a
3. t2a > t2a-1
4. t2a = a few days later t2a-1
5. t3a = t2a
6. x=I
7. y = Dr P, z = his wife
8. w = later
9. e2: show (u,v into s)
10. t2b < t1b
11. t2b > t2a
12. t3b = t2b
13. u=z
14. v=x
15. s = a lofty apartment
16. s: recall (r, ... to p)
17. t2c < t1c
18. t2c = t2b
19. t3c = t2c
20. r=s
21. o = fin-de-siècle Berlin
22. p=x
23. p=A
24. A = Author
25. A = RS ...
RS ... q p
26. q = r
27. p = o
The next DRS interprets an atemporal text passage. I use a simplified fragment
from a passage presented earlier. The DRS encodes information about Discourse
Mode – situation entities and Primary Referents – and subjectivity.
(21) Within the next fifty years fish farming may change us from hunters and
gatherers into marine pastoralists. It is likely that other new technologies will
appear.
t1a t2a t3a x s l y A t1c t2c t3c s2 q
1. s1: may (...)

2. t1a = t2a = t3a
3. A = Author
4. A = RS
5. x = fish farming
6. y = us
7. ... = Primary Referent
RS ...
t1b t2b t3b w u v s e
1. e: change (u,v into s)

2. t2b > t1b; t3b = t2b
3. t2b = w
4. w = within fifty years
5. u=x
6. v=y
7. v = Primary Referent
8. s = marine pastoralists
8. s2: be likely (...)

9. t1c = t2c = t3c
10. q = Proposition
11. A = RS ...
12. ... = Primary Referent
RS ...
t1d t2d t3d r e
1. e: appear (r)
2. t2d > t1d; t3d = t2d
3. r = other new technologies
4. r = Primary Referent
The situation entities of the fragment consist of states, a proposition, and events
in the sub-DRSs. The complement clause of “likely” refers to a proposition, as
10.3 Formalization in DRSs 257
we know from the class of predicate and the substitution possibilities (“that S”
may appear but not a wh-complement).
Entities of situations, times, and individuals appear at the top of the DRS
if they exist independently. In the first clause, for instance, “fish farming” and
“us” are listed independently; but “marine pastoralists” depends for existence
on the sub-DRS. In the second clause “other new technologies” depends on the
sub-DRS.
The Primary Referents are identified by a condition on the entity. In each case
the complement is Primary Referent for the clause. Within the first complement,
“us” is Primary Referent, the entity that changes; within the second complement,
“other new technologies” is primary, dependent on the situation for existence.
There are forms of subjectivity in both clauses; the Author is the Responsible
Source.

(6–7) Oliver Sacks, The Man who Mistook His Wife for a Hat, New York: Harper &
Row, 1970, p. 11.
(8–10) Peter Robinson, A Necessary End, New York: Avon Books, 1989, p. 103.
(11) German chancellor vows to block use of ground forces in Yugoslavia. New York
Times, May 20, 2000.
(12, 14) Peter Drucker, The information revolution. Atlantic Monthly, October 1999.
(15, 17) Alan Ehrenhalt, Hijacking the rulebook. New York Times, December 20, 1998.
11 Discourse structure and
Discourse Modes
This inquiry into local text structure has stayed close to the linguistic ground,
concentrating on linguistic forms and categories. In this final chapter I widen the
range of discussion to consider some of the organizing principles of discourse. I
begin with hierarchical structure and functional units, and then discuss discourse
relations. Finally, I comment on how the Discourse Modes pertain to discourse
relations and to lexical patterns, which distinguish units at different levels of
text structure.
11.1 Organizing principles of texts

11.1.1 Hierarchical structure
Discourse is organized by its purpose, like many other human activities. To un-
derstand its structure we can turn to work in cognitive science on intentions and
plans generally. Intentions have hierarchical structure, G. Miller, Galanter, &
Pribram argue convincingly in an early study. They propose an abstract unit,
the Test-Operate-Test-Exit or TOTE, to represent such structures. TOTE units
can be used to model human intentional behavior: they represent a top level of
control in planning and executing intentions. Intentions themselves are orga-
nized hierarchically into sub-goals, which in turn may have sub-goals, and so
on. For instance, to hammer something once a person must plan and execute
the two sub-goals of lifting and striking (1960:33):
(1) HAMMER
Lifting Striking
Repeated hammering consists of many such steps, each with the same hier-
archically organized sub-goals. More complex activities have many sub-goals.
Even a relatively simple activity such as going to work has a number of
258
11.1 Organizing principles of texts 259
sub-goals: getting ready, leaving the house, deciding how to get there, choosing
a route, etc.
If we assume that intentions are hierarchical, and that discourse is organized
into functional units of goals and sub-goals, we both predict and explain its
hierarchical nature. The recognition of hierarchical structure has led to studies
of how plans are made, executed, and recognized, as in Allen & Perrault (1980),
Litman (1985). Much of this work applies to written discourse as planned, in-
tentional action. Intentional structure is proposed as a distinct level of discourse
structure by Grosz & Sidner (1986).
The intuition that texts have hierarchical structure is shared by all who work
with them. Hierarchical structure is often modeled in abstract tree structures.1
The units posited may function as part of the global structure of the discourse;
their parts may be topically related;2 or they may realize discourse relations
such as Causation, Evidence, etc., discussed below.
11.1.2 Functional units

Texts consist of functional units that are hierarchically structured, each con-
tributing to the purpose of the whole. The units each realize a goal and may
have smaller units with their own sub-goals.
The functional units and their sub-structure are evident in “scripted” discourse
such as legal trials and task-oriented activities. Dialogues between experts and
novices on dismantling a water-pump were studied in Grosz (1977). Grosz
found that the dialogue structure mirrored the structure of the task. Other types
of discourse are organized implicitly by the nature of the material. Asked to de-
scribe the layouts of their apartments, people consistently came up with clearly
defined journeys through the apartments, treating each room as a separate unit
(Linde & Labov 1975). Scientific reports are structured quite rigidly, according
to recognized convention.
1. Linguists come to this conclusion from many traditions; they include Halliday (1967), Linde
(1979), Grimes (1975), Longacre (1983/1996), Polanyi (1988), Martin (1992), and many others.
The tree structures of linguistics are generated with phrase structure rules, which encode
hierarchical relations. Thus the first rule of a simple sentence grammar might be: S → NP + VP
which would be realized in a tree structure with a topmost node S, dominating the two nodes
NP and VP:
S
NP VP
2. Topically related means, roughly, “about the same thing.” The parts of a topically related segment
may or may not have an explicit shared topic; see Chapter 8 for discussion.
260 Discourse structure and Discourse Modes
Narrative often has a relatively clear functional structure. Gerald Prince in

his 1987 Dictionary of Narratology recognizes three main units: Beginning,
Complicating Action, End.3 Each of these units may have a relatively complex
internal structure. For instance, the Complicating Action may have one or more
episodes. An episode consists of a coherent set of situations that have “some kind
of thematic unity – identical participants, time, location, global Event or action”
(van Dijk 1982:199). There is experimental evidence from psycholinguistics
that people use episodes for encoding story information in memory (Black &
Bower 1980).
The functional units of a text can be determined by analyzing its intentional
structure. For instance, Longacre discusses the structure of a fund-raising letter.
He identifies a main goal of influencing and finds four parts, or “high-level
moves,” of the letter (1992:110): (i) establish authority; (ii) present problem
and solution; (iii) issue command to contribute money; (iv) motivation, threats
and promise. The parts are realized by different sections of the letter. The
functional units for other types of text can be determined by understanding
their intentional structure.
11.2 Discourse relations

This section begins by introducing discourse relations; 11.2.1 gives some back-
ground; 11.2.2 discusses current approaches; in 11.2.3 I suggest three main
classes of discourse relations.
Texts are understood in terms of abstract units that realize discourse relations.
They may consist of situations related by Cause or Result, or by rhetorical
relations such as Elaboration, Evidence, Parallelism, etc. These units do not
consist directly of clauses or sentences, but of the situations and propositions
evoked by a text. They are hierarchical in structure, organized by the relationship
that is realized. The units are also known as “coherence relations” and “discourse
coherence relations.”
People recognize or construct relations between the entities and propositions
introduced in a discourse. They do this by inference if the relation is not
explicitly given. In (2) is a natural example in which the situations expressed
are related tacitly by causation:
3. In a study of spoken personal narratives, Labov & Waletsky (1966) posit similar functional
units: Orientation, Action, and Evaluation, with an optional Preface. These studies owe much
to the work of Propp (1958), which identified the basic units of stories and the ways in which
they combine and vary.
(2) (a) He was in a foul humor. (b) He hadn’t slept well that night. (c) His electric
blanket hadn’t worked.
The event of (2b) is the cause of the state expressed in (2a), and (2c) gives the
cause of the situation in (2b); the example is cited by Hobbs (1985:13). The
urge to find relations is so powerful that we can make sense of almost anything
even if many inferences are needed. Inventing texts which are incoherent is
actually quite difficult; Knott gives this example (1996:3):
(3) Sally decided to take the history course. The ducks on the lake were not eating
the bread.
He comments that the text fragment can be understood if we imagine Sally to

have unusual superstitions about the ducks on the lake.
Discourse relations notionally weave together into relational units clauses or
larger parts of a text – more precisely, the entities, propositions, and concepts
that they express. They segment texts into abstract units of increasing size: the
larger units overlap or coincide with genre-based functional units. I will refer to
the material related by a discourse relation as a “relational segment.” The units
that realize discourse relations are hierarchically structured and hold for spans
of discourse of virtually any size. Within such a unit, or relational segment,
more than one such relation may be realized.
After a brief discussion of background and types of discourse relations,
I propose that three classes of relations be recognized.
11.2.1 Background of discourse relations

Discourse relations appear in different guises in most efforts to understand
text structure. In the mid-1970s linguists discussed “coherence” and “deep
structure” relations which need not be overtly marked on the surface of the
text. Grimes (1975) proposes a set of “rhetorical predicates”: they form ab-
stract propositions that relate material in the text. The list includes Conditional-
ity, Causation, Equivalence, Amplification, Summary, Identification, Example.
Specifically, Grimes distinguished two types of predicates: “paratactic predi-
cates” dominate all of their arguments in coordinate fashion; with “hypotactic
predicates” one argument is the center and the other subordinate. The rela-
tion of Explanation is hypotactic, for instance: one argument is central and the
subordinate argument different in kind (1975:216).
The basis for many of these ideas comes from linguists who translated the
Bible into different languages. According to Grimes, Fuller’s (1959) work on
Bible study was the first attempt to identify and work with discourse relations.
Fuller recognized “increasingly large sections of text according to a small num-
ber of explicit organizing relations” (Grimes 1975:208). The approach is further
developed in Beekman (1970). Along similar lines, Halliday & Hasan noted
four types of “underlying semantic relations” which they saw as conjunctions
that are optionally signaled on the surface: Additive, Adversative, Causal, and
Temporal (1976). Again, Longacre (1983) discusses deep structure relations,
“combinations of predications” into larger units; his list includes such notions
as Conjunction, Contrast, Alternation, Comparison.
11.2.2 Current approaches

Understanding of how discourse works advanced quite dramatically in the 1980s
with the impetus of Artificial Intelligence. Computational study of discourse
structure led to Jerry Hobbs’ (1985) theory of coherence relations. Hobbs was in-
terested in the inferences that people make in understanding a text, and the kinds
of knowledge that are implicit in such inferences. He theorized that understand-
ing of a text proceeds as the receiver constructs “discourse coherence relations”
between its parts. Hobbs posited four classes of relation: (1) Occasion relations,
which concern events and states; (2) Evaluation relations, which relate “what
has been said to a goal of the conversation”; (3) a class that relates segments of
discourse to the listener’s prior knowledge, e.g. Background and Explanation;
and (4) Expansion relations such as Exemplification, Generalization, Contrast,
Violated Expectation.
Another approach is Rhetorical Structure Theory, which seeks relational
propositions in a text (Mann & Thompson 1987, 1992). The theory describes
“relational organization” according to the relations between “spans of text.”
Relations may be indicated by particular words, or be implicit in the text. Mann
& Thompson build on Grimes’ insight about types of relations. They distinguish
symmetric and asymmetric relations, assigning the latter a Nucleus–Satellite
structure. Most relations are asymmetric, including Evidence, Justify, Solution,
Circumstance, Enablement, Elaboration. For instance, in an Elaboration seg-
ment one part is the primary nucleus and the elaborating part is the satellite.
Symmetric relations include Contrast, List, and Sequence. The relations hold
for successively larger spans of a discourse. Mann & Thompson give examples
in which entire texts are analyzed into relational segments with hierarchical
internal structure.
One major insight about discourse relations is that they do not depend on the
presence of words that explicitly signal such relations, words such as because,
and, but, etc. Nevertheless there has been some useful study of such words.
Schiffrin looked at the distribution and interpretation of certain “discourse
markers” in conversation. She found that the scope of these words varies: for
instance, in some contexts because has scope over multiple clauses (Schiffrin
1987:197).4 Knott (1996) collected over 200 such “cue words” and used them
in a substitution task to test for basic discourse relations that are psychologically
real.
Relational words and concepts are not always one-to-one. For instance, the re-
lation “cause” is signaled by several cue words, e.g. so, therefore, consequently,
thus, as a result. Conversely, one word may indicate different relations, e.g.
but can indicate “thesis–antithesis” and/or “concession” (Mann & Thompson
1987:71). Relational words do not necessarily translate directly from one lan-
guage to another: the semantic space covered may differ (Fabricius-Hansen
& Behrens 2000). In computational generation of texts, Hovy uses a large
set of coherence relations. He distinguishes three classes, Semantic, Interper-
sonal, and Presentational (1993). Grosz & Sidner (1986) offer a computationally
based theory of text structure that emphasizes attentional and intentional phe-
nomena. They argue that it is futile to seek a fixed list of discourse purposes,
instead recognizing two general “intentional structural relations,” dominance
and satisfaction–precedence.
Taking a semantic–pragmatic approach, Asher & Lascarides have developed
formal accounts of certain discourse relations (Lascarides & Asher 1993). They
propose a defeasible logic and theory of commonsense entailment to account
for the patterns of inference that people use. Their aim is to “place knowledge
in a logic where its implications can be precisely calculated” (1993:439). For
simple narrative texts they offer a procedure for determining the relations of
Elaboration, Explanation, Narration, and Background in short text segments
(Asher & Lascarides 2000).
The abstract relational structures of discourse are represented in an exten-
sion of DR Theory, Segmented Discourse Representation Theory (SDRT). The
theory, due to Asher (1993), constructs hierarchical structures of text segments
that realize the relational segments of discourse relations. The output of the
SDRT construction rules is not a single DRS but a complex propositional struc-
ture of DRSs. The complex representation is a recursive structure of clausal
4. Schiffrin (1987) posits the following components: exchange, action, and idea structures; infor-
mation State; participation framework. She investigated the expressions oh, well, and, but, or,
so, because, now, then, I mean, y’know. Some of these expressions signal relations in the sense
under discussion here, others do not.
DRSs, linked together by relations such as Narration and Parallel. The construc-
tion procedure requires that new units be attached only to tree nodes which are
“structurally accessible.” The rightmost nodes of each sub-tree or constituent
are taken to be accessible, following Polanyi (1988).
In the extended version of the theory, DRSs or SDRSs are organized to form
discourse segments. The stated goal is to model as precisely as possible the
insights of Hobbs, Mann & Thompson, and others about discourse relations.
The theory infers discourse relations with a formal theory of pragmatics which
draws on multiple information sources. The sources of information are primarily
semantic content, including lexical semantics; domain and world knowledge;
and Gricean principles of orderliness. The type of discourse relation that a
segment realizes is notated in the SDRT representation.
11.2.3 Classes of discourse relations

People disagree as to whether there is a basic set of discourse relations.
Culling different lists, I found that some relations appear often: Cause, Goal
or Purpose, Evidence, Elaboration, Background, Contrast, Parallel. Others are
more idiosyncratic. For instance, Solution appears on one list only. Mann &
Thompson (1987) claim that their list of about twenty relations has been
adequate to account for a wide number and range of texts. But it may be futile to
attempt an exhaustive listing: another relation can always be found in principle.
The most useful approach may be to classify discourse relations by the kind
of contribution they make to a text. I suggest that three such classes of discourse
relations be recognized: Consequential, Organizing, and Rhetorical.
Consequential relations hold between the situations introduced in a text. The
main consequential relations are Cause, Enablement, and Result. Example (2)
above illustrates the Cause relation.
Organizing relations hold between the concepts and propositions of a text,
taking proposition as the content that a sentence expresses. The relations in
this class are abstract and cover a wide range. The class includes Conces-
sion, Contrast, Evaluation, Evidence, Explanation, Elaboration, Generalization,
Purpose, Reason, Solution. Some pertain to building an argument or making a
case of some kind (Evaluation, Evidence), others would also be appropriate in
other contexts as well (Contrast, Elaboration, Purpose, Reason).
The third class of relations, the Rhetorical, reflects how the material of a text is
presented. Relations of this class include Parallel, Restatement, and Summary.
They relate propositions or stretches of text to each other. These relations do
not reflect the way situations or propositions are related, but decisions about
the text that presents them.
11.3 Discourse Modes and text structure 265
11.3 Discourse Modes and text structure

This book has taken perhaps as far as possible an analysis of text passages
based on linguistic form. I have argued that the Discourse Mode passage is a
significant unit of text structure. One natural direction for further study is to
look at whether and how the Discourse Modes pertain to discourse relations;
another direction is to look at how the modes relate lexical patterns in texts.
The level of discourse relations is more abstract than the passage. Discourse
relations are sometimes indicated by particular forms in a text, but they are
most often arrived at by pragmatic inference and world knowledge. Relational
segments and Discourse Mode passages are units of different types.
There is no simple mapping between Discourse Modes and relational seg-
ments. However, the nature of discourse relations suggests certain predictions.
One prediction concerns Consequential relations and types of entities. The class
of Consequential relations – Cause, Enablement, etc. – relates situations to each
other, so that we can predict that relational segments of this class will be realized
by the Discourse Modes that introduce such entities.
The distinction between symmetric and asymmetric relations yields other
predictions. The symmetric relations should involve entities of the same type.
Therefore, we would expect them to be realized by passages of a single
Discourse Mode. For instance, a relational segment realizing the Contrast,
Alternative, or Parallel relations would probably involve passages of a single
Discourse Mode.
Asymmetric relations would be expected to involve different Discourse
Modes, following this line of thinking. For instance, the relations of Evidence,
Explanation, and Elaboration tend to have parts that are different in kind:
they often consist of a generalization and then specifics that illustrate (Grimes
1975:210). Thus, relational segments might have a passage corresponding to the
nucleus of the Information or Argument mode; the passage corresponding to the
satellite might be of the Narrative, Description, or Report mode. The prediction
about the same or different Discourse Modes does not hold generally for the
asymmetric discourse relations, however. For some relations, e.g. Circumstance
and Antithesis, one would expect nucleus and satellite to involve the same type
of discourse entity. We can expect to learn more about discourse relations by
looking carefully at how they are expressed in terms of the Discourse Modes.
The interaction of the Discourse Modes with discourse relations is another topic
for investigation.
Turning to lexical patterns, the lexical items of a text fall into clusters that at
once mirror and contribute to text organization. Patterns of lexical co-occurrence
and distribution can be used to divide a discourse into units, as noted in

Chapter 2, and the units developed on this basis correspond nicely to topic
boundaries. A natural question is how Discourse Mode passages relate to units
like this, which are established on a lexical basis. One might expect them to
coincide. Indeed, it would be surprising if a shift from one Discourse Mode to
another did not have ramifications, lexical and notional. Lexical ramifications
would include changes in the lexical items expressed by passages of a given
Discourse Mode. There might also be shifts at the boundaries of Discourse Mode
passages in types of repetition and coreference. For the notional ramifications
of Discourse Modes I turn to the level of discourse relations.
11.4 Conclusion
The classification of text passages into Discourse Modes is a fairly radical
extension of aspectual and temporal notions, and it has been quite successful in
bringing out important features of texts. The essential point is that the modes
are notional units with linguistic correlates.
This book is devoted to English, but it is likely that Discourse Modes are a
general phenomenon in language. There are bound to be differences in the lin-
guistic correlates to the Discourse Modes. Some language-particular correlates
are already known: for instance, in French the Descriptive mode is convention-
ally realized by the imparfait, a past tense with the imperfective viewpoint. The
linguistic features of abstract entities may vary, since not all languages have
the formal variety in their clausal complements of English. The expression of
temporal progression in tenseless languages must rely on linguistic features
other than tense. The features of the discourse modes provide a useful research
tool for studying similarities and differences between languages.
Appendix A: The texts
Appendix A presents a selection of texts that are found frequently in the chapters.
Short texts are given in their entirety; fragments are excerpted from the longer texts.
Unless otherwise noted, the fragment begins at the beginning of the text. The original
paragraphing is preserved. The texts are given in the order presented in this book.
The texts: (a) from A Necessary End, by Peter Robinson, New York: Avon Books,
1989, p. 182; (b) Barak fights on many fronts, New York Times, May 20, 2000; (c) Cheap
oil’s tough bargains by Robert Mosbacher, New York Times, March 13, 2000; (d) After a
victory. Ethiopia looks toward other fronts, New York Times, May 20, 2000; (e) from Cell
communication, by John Scott & Tony Pawson, Scientific American, June, 2000; (f) from
Listening to humpbacks, by Douglas H. Chadwick, National Geographic, July 1999;
(g) from Slave-making queens, by Howard Topoff, Scientific American, November 1999;
(h) Hijacking the rulebook, by Alan Ehrenhalt, New York Times, December 20, 1998;
(i) The Information Revolution, Peter Drucker, Atlantic Monthly, October 1999; (j) The
Chinese Potter, by Margaret Medley, London: Phaidon, 1989; (k) from The pride of
the cities, by Peter Beinart, New Republic, June 1997; (l) from How it works, by Jim
Collins, US Airways Magazine, Attaché, May 2001; (m) from “The Dead” by James
Joyce. In Dubliners, 1916; reprinted London: Penguin Books, 1982, pp. 177–79; (n) Let
teenagers try adulthood, by Leon Botstein, New York Times, May 1999.
THE TEXTS
(a) Fragment from A Necessary End, by Peter Robinson; pages 81ff.
Mara walked along the street, head down, thinking about her talk with Banks. Like
all policemen, he asked nothing but bloody awkward questions. And Mara was sick of
awkward questions. Why couldn’t things just get back to normal so she could get on
with her life?
“Hello, love,” Elspeth greeted her as she walked into the shop.
“Hello. How’s Dottie?”
“She won’t eat. How she can expect to get better when she refuses to eat, I just don’t
know.
They both knew that Dottie wasn’t going to get better, but nobody said so.
“What’s wrong with you?” Elspeth asked “You’ve got a face as long as next week.”
Mara told her about Paul.
“I don’t want to say I told you so,” Elspeth said, smoothing her dark tweed skirt, “but
I thought that lad was trouble from the start. You’re best rid of him, all of you.”
267
268 Appendix A
“I suppose you’re right.” Mara didn’t agree, but there was no point arguing Paul’s
case against Elspeth. She hadn’t expected any sympathy.
“Go in the back and get the wheel spinning, love,” Elspeth said. “It’ll do you a power
of good.”
The front part of the shop was cluttered with goods for tourists. There were locally
knit sweaters on shelves on the walls, tables of pottery – some of which Mara had made –
and trays of trinkets, such as key-rings bearing the Dales National Park emblem – the
black face of a Swaledale sheep. As if that weren’t enough, the rest of the space was
taken up by fancy notepaper, glass paperweights, fluffy animals and fridge-door magnets
shaped like strawberries or Humpty Dumpty.
In the back, though, the setup was very different. First, there was a small pottery
workshop, complete with wheel and dishes of brown and black metallic oxide glaze,
and beyond that a drying room and a small electric kiln. The workshop was dusty and
messy, crusted with bits of old clay, and it suited a part of Mara’s personality. Mostly
she preferred cleanliness and tidiness, but there was something special, she found, about
creating beautiful objects in a chaotic environment.
She put on her apron, took a lump of clay from the bin and weighed off enough for
a small vase. The clay was too wet, so she wedged it with a flat concrete tray, which
absorbed the excess moisture. As she wedged – pushing hard with the heels of her hands,
then pulling the clay forward with her fingers to get all the air out – she couldn’t seem
to lose herself in the task as usual, but kept thinking about her conversation with Banks.
Frowning, she cut the lump in half with a cheese wire to check for air bubbles, then
slammed the pieces together much harder than usual. A fleck of clay spun off and hit her
forehead, just above her right eye. She put the clay down and took a few deep breaths,
trying to bring her mind to bear only on what she was doing.
No good. It was Banks’s fault, of course. He had introduced her to speculation that
caused nothing but distress. True, she didn’t want Paul to be guilty, but if, as Banks had
said, that meant someone else she knew had killed the policeman, that only made things
worse.
Sighing, she started the wheel with the foot pedal and slammed the clay as close to
the centre as she could. Then she drenched both it and her hands with water from a bowl
by her side. As the wheel spun, clayey water flew off and splashed her apron.
She couldn’t believe that any of her friends had stabbed Gill. Much better if Osmond
or one of the students had done it for political reasons. Tim and Abha seemed nice
enough, if a bit naive and gushing, but Mara had never trusted Osmond; he had always
seemed somehow too oily and opinionated for her taste.
(b) A year after victory, Barak fights on many fronts. New York Times,
May 20, 2000
A week that began in violence ended violently here, with bloody clashes in the West Bank
and Gaza and intensified fighting in Southern Lebanon. On the one-year anniversary of
his election by a sweeping majority, Prime Minister Ehud Barak was trying to put out
fires on many fronts at once. Because of trouble in the north, the south, and within
his political coalition, he was weighing a cancellation of his scheduled departure this
weekend to the United States to confer with President Clinton and meet with Jewish
American leaders.
The texts 269
Despite the violence, back-channel talks continued in Sweden. Israelis, Palestinian

and American officials have characterized them as a serious and constructive dialogue
on the process itself and on the final status issues.
News accounts here say that Israel is offering as much as 90 percent of the West
Bank to Palestinians, although it is difficult to assess what is really happening by the
bargaining moves that are leaked.
Near a heavily fortified Jewish settlement in the Gaza Strip, an Israeli soldier and
a Palestinian policeman were wounded as Palestinian protests for the release of 1,650
prisoners degenerated into confrontations. Israeli military officials say they are inves-
tigating the source of fire that wounded the soldier, but that there was no gun-battle
between troops like the one that took place on Monday near Ramallah in the West Bank.
“The prime minister views the events severely and has ordered the Israeli Defense
Forces to act accordingly to restore calm,” Mr. Barak’s office said in a statement tonight.
The protests today, which began as the Friday prayer services ended, were orga-
nized before Monday’s demonstrations erupted into severe rioting that ended with three
Palestinians dead. Palestinian authorities said they were trying to keep tensions low,
although in Ramallah their effort did not appear to be of sustained high intensity. Many
officers stepped to the side as a march, organized by the militant Islamic group Hamas,
headed toward confrontation with the Israeli forces. In Qalqilya, in contrast, Palestinian
police officers held back demonstrators with nightsticks.
At least 20 Palestinians were injured by rubber-coated bullets. Another “day of rage”
was scheduled for Saturday. It was unclear whether the violence would delay the transfer
of Abu Dis and two other villages near Jerusalem to the Palestinian Authority. Mr. Barak’s
advisers said he would turn over the towns within days, provided that Palestinian officials
calmed the streets and cooperated with their investigation of how things got out of hand
on Monday.
In the north, Israeli warplanes attacked suspected guerrilla positions in southern
Lebanon today. This followed fighting Thursday, the heaviest in two weeks, in which 14
people were wounded, including 4 Lebanese civilians and a United Nations peacekeeper.
(c) Cheap oil’s tough bargains, by Robert Mosbacher, New York Times,
March 13, 2000
The press has trumpeted the news that crude oil prices are three times higher than they
were a year ago. But it was the $10 or $11 price of February 1999, not the one today,
that really deserved the headlines.
When inflation is taken into account, that 1999 price was the lowest in modern history,
while oil has gone above today’s seemingly high price several times. And for the past
14 years, at $17.50, oil has been one of the real bargains of the modern age.
The low price has been a mixed blessing. In the United States, we have lost over
500,000 jobs in the oil industry while we have grossly increased our dependency on
foreign oil; we now import 55 percent of what we use. With little incentive for drillers to
find and tap new oil, supplies eventually dropped, and in the past year the Organization of
Petroleum Exporting Countries deliberately dropped its production as well. In response
to the law of supply and demand, prices have now risen.
A high oil price is not the inflationary threat it once was because with the shift toward
the information and service economy, and away from manufacturing, the United States
270 Appendix A
is less dependent on oil. But try to tell a consumer paying to heat his home, a trucker
moving goods across the country or a commuter who goes by car that oil and gasoline
are not major factors in the economy, and the answer is likely to be, “It feels like I’m
being ripped off.”
American consumers have been lulled into thinking cheap oil is their entitlement.
Syndicated columnist Carl T. Rowan wrote recently: “We are truly a spoiled society!
We insist on driving gas guzzlers and using a grossly disproportionate amount of the
world’s energy, and we believe we should forever be able to do so at bargain rates.”
Sooner or later, oil prices are likely to drop. But prices at today’s level have their
advantages. With the incentive for more production back in place, there will be more
drilling in new places, like the deep water off the coasts of many countries of the world,
including the United States. Off our Gulf Coast, deep-water drilling, developing and
producing are already going on.
Remarkable new technology now allows pipelines to bring oil from miles out to sea,
where the water is as much as 8,000 feet deep. The ubiquitous deep waters of the world
have great potential for additional oil and gas reserves, but drilling in 5,000 to 10,000 feet
of water and then through many thousands of feet of sand, shale and other formations
makes for huge costs. In many cases, this kind of oil production can only be justified by
prices of at least $25 to $35 a barrel.
A high price also encourages development of other new technologies already on the
horizon, including three-dimensional seismology for mapping and horizontal drilling.
And $30 oil also brings attention back to development of synthetic fuel, solar energy,
wind power, gasification of coal and other methods of producing energy.
Some members of Congress have been lobbying for taking oil from the Strategic
Petroleum Reserve to help bring prices down, and President Clinton recently announced
he is not ruling this out. But these reserves are set aside for use in an emergency, and
tapping them is only justified if there is a genuine threat of a supply interruption, as there
might be in a war or political crisis. The reserve holds about 570 million barrels, not
much when we use about 19.4 million barrels of oil products daily. Even releasing all of
it could be only a temporary solution to the price problem. Tapping it would also signal
the main oil-producing countries that we were trying to control world energy prices,
which is inconsistent with our normal free and fair trade policies.
Even if the control of oil prices were in American hands, which it is not (since we
have no means of influencing it other than using our reserves or jawboning OPEC), we
would still face a daunting decision. Would we be better off getting prices back under
$20 a barrel to prevent any oil-influenced inflation today, or should we take the longer
view and let today’s price work to bring about more drilling and new interest in other
sources of energy?
Simple realities argue for the latter course: In the long term, even if bent by cartels,
the law of supply and demand will rule. And in the long term, we will need the new
sources of energy that high prices can bring.
(d) After a victory, Ethiopia looks toward other fronts. New York Times,
May 20, 2000
It took three days for Ethiopian troops to battle their way into this town, and now that
they have it, they are pressing deeper into Eritrea, denying with every step that they plan
The texts 271
any long-term invasion of their neighbor and once-tight ally. “We have no intentions, no
plans, no need to occupy Eritrea,” an Ethiopian commander, Colonel Gabre Kidane, said
tonight as he stood on the old Italian fortress in this hilly town 45 miles inside Eritrea.
“The only intention we have is to weaken the forces that have occupied our country and
to regain our sovereignty. That is all.”
A week after Ethiopia started an offensive that it says is aimed at ending the two-
year-old war, it is now clear that the whole of Eritrea could become a battlefield. With
hundreds of civilians fleeing the region, Colonel Kidane said Ethiopian soldiers continue
to skirmish with Eritrean soldiers on the run here in western Eritrea.
Tonight, Ethiopian officials said planes bombed the main Eritrean military training
center at Sawa, an American-built base 100 miles west of Asmara, the capital. The
officials also said they had taken a village, Maidema, 30 miles from Asmara, on the way
from the western front to the central front along the disputed border. That is where the
next round of fighting, already heavy, is generally expected.
For Eritrea, Ethiopia’s rapid advance out of the border trenches into the countryside
is not merely a military setback. For 30 years, when Eritrea was the northernmost
province of Ethiopia, rebels fought two successive governments in Addis Ababa for
their independence. Those rebels, the Eritrean People’s Liberation Front, are now in
power in Eritrea, and they say this advance is a frightening continuation of Ethiopia’s
attempts at domination.
An Eritrean government statement said today that a bombing on Thursday south of the
port of Massawa killed a civilian. That death, the statement said, was “further evidence
of Ethiopia’s resurgent aim to annex this country and, in so doing, to make targets of
Eritrea’s civilian populations.” Ethiopia said the target was a military installation.
The paradox is that the leaders of Ethiopia fought side by side with the Eritreans to
oust the military government of Mengistu Haile Mariam in 1991. For the first two years,
the two governments were close friends, so close that they never demarcated the border
when Eritrea became independent in 1993. But tensions grew over personality clashes
and economic rivalry, exploding in May 1998 when Eritrea claimed the Badme border
region based on old colonial maps. Eritrea moved troops into the area. Ethiopia said
it was invaded, and a war ignited that has defied long peace talks and claimed tens of
thousands of lives on both sides.
Until Ethiopia began its offensive last Friday, the fighting had been confined to three
fronts along the 620-mile border. But since then, Ethiopia has pressed 65 miles into
Eritrea, displacing 340,000 people, the World Food Program said today.
Officials in neighboring Sudan said today that an additional 50,000 Eritreans had fled
across their border. In both places, relief officials said, people lack food and shelter.
Even without the war, the United Nations says 800,000 Eritreans face food shortages
because of drought. Most civilians fled from around this town, the regional capital and
a strategically important point because the area was a main supply route for Eritrea’s
westernmost front.
The battle has been intense. Starting on Monday, Colonel Kidane said, Ethiopian
troops had pounded areas around the town with bombs, tanks and artillery and then
engaged Eritrean troops along a mountain pass eight miles from town. Along the
heavily mined road from the south, the route of Ethiopia’s advance, huge numbers
of empty mortar shells, captured Eritrean ammunition and destroyed trucks from both
272 Appendix A
sides remain. At spots, the smell of rotting corpses was strong. Colonel Kidane would
not give the number of casualties.
On Thursday morning, Ethiopian troops took the town, the biggest one that they have
captured, and called it a major victory. The Eritreans said they had staged a tactical
withdrawal.
This evening, Ethiopian troops milled around and ate their rations at the hilltop build-
ing that was once a fortress for Italian troops in the colonial period and that, until
Thursday, according to the Ethiopians, was the command center for Eritrean troops in
the region. The Ethiopians joked about marching to Asmara. “It’s nice to take the place
of the invading army,” said an Ethiopian soldier, Seife Yechenju, 21, who has fought in
the army since the war began. “There is no question that I am very happy.”
(e) Fragment from cell communication, by John D. Scott & Tony Pawson,
Scientific American, June 2000
As anyone familiar with the party game “telephone” knows, when people try to get a
message from one individual to another in a line, they usually garble the words beyond
recognition. It might seem surprising, then, that mere molecules inside our cells con-
stantly enact their own version of telephone without distorting the relayed information
in the least.
Actually, no one could survive without such precise signalling in cells. The body func-
tions properly only because cells communicate with one another constantly. Pancreatic
cells, for instance, release insulin to tell muscle cells to take up sugar from the blood for
energy. Cells of the immune system instruct their cousins to attack invaders, and cells of
the nervous system rapidly fire messages to and from the brain. Those messages elicit
the right responses only because they are transmitted accurately far into a recipient cell
and to the exact molecules able to carry out the directives.
But how do circuits within cells achieve this high-fidelity transmission? For a long
time, biologists had only rudimentary explanations. In the past 15 years, though, they
have made great progress in unlocking the code that cells use for their internal commu-
nications. The ongoing advances are suggesting radically new strategies for attacking
diseases that are caused or exacerbated by faulty signaling in cells – among them cancer,
diabetes and disorders of the immune system.
The earliest insights into information transfer in cells emerged in the late 1950s,
when Edwin G. Krebs and Edmond H. Fischer of the University of Washington and the
late Earl W. Sutherland, Jr., of Vanderbilt University identified the first known signal-
relaying molecules in the cytoplasm (the material between the nucleus and a cell’s outer
membrane). All three received Nobel Prizes for their discoveries.
By the early 1980s researchers had gathered many details of how signal transmis-
sion occurs. For instance, it usually begins after a messenger responsible for carrying
information between cells (often with a hormone) docks temporarily, in lock-and-key
fashion, with a specific receptor on a recipient cell. Such receptors, the functional equiv-
alent of antennae, are able to relay a messenger’s command into a cell because they
are physically connected to the cytoplasm. The typical receptor is a protein, a folded
chain of amino acids. It includes at least three domains: an external docking region
for a hormone or other messenger, a component that spans the cell’s outer membrane,
and a “tail” that extends a distance into the cytoplasm. When a messenger binds to the
The texts 273
external site, this linkage induces a change in the shape of the cytoplasmic tail, thereby
facilitating the tail’s interaction with one or more information-relaying molecules in the
cytoplasm. These interactions in turn initiate cascades of further intracellular signalling.
Yet no one had a good explanation for how communiqués reached their destination
without being diverted along the way. At that time, cells were viewed as balloonlike bags
filled with a soupy cytoplasm containing floating proteins and organelles (membrane-
bound compartments, such as the nucleus and mitochondria). It was hard to see how, in
such an unstructured milieu, any given internal messenger molecule could consistently
and quickly find exactly the right tag team needed to convey a directive to the laborers
deep within the cell that could execute the order.
Today’s fuller understanding grew in part from efforts to identify the first cytoplas-
mic proteins that are contacted by activated (messenger-bound) receptors in a large and
important family: the receptor tyrosine kinases. These vital receptors transmit the com-
mands of many hormones that regulate cellular replication, specialization or metabolism.
They are so named because they are kinases – enzymes that add phosphate groups to
(“phosphorylate”) selected amino acids in a protein chain. And, as Tony R. Hunter of the
Salk Institute for Biological Sciences in La Jolla, Calif., demonstrated, they specifically
put phosphates onto the amino acid tyrosense.
In the 1980s work by Joseph Schlessinger of New York University and others indicated
that the binding of hormones to receptor tyrosine kinases at the cell surface causes
the individual receptor molecules to cluster into pairs and to attach phosphates to the
tyrosines on each other’s cytoplasmic tails. In trying to figure out what happens next,
one of us (Pawson) and his colleagues found that the altered receptors interact directly
with proteins that contain a molecule they called an SH2 domain. The term “domain”
or “module” refers to a relatively short sequence of about 100 amino acids that adopts
a defined three-dimensional structure within a protein.
At the time, prevailing wisdom held that messages were transmitted within cells pri-
marily through enzymatic reactions, in which one molecule alters a second without
tightly binding to it and without itself being altered. Surprisingly, though, the phospho-
rylated receptors did not necessarily alter the chemistry of the SH2-containing proteins.
Instead many simply induced the SH2 domains to latch onto the phosphate-decorated
tyrosines, as if the SH2 domains and tyrosenes were Lego blocks being snapped together.
(f) Fragment from Listening to humpbacks, by Douglas Chadwick, National

Geographic, July 1999
When a big whale dives, currents set in motion by the passage of so many tons of
flesh come eddying back up in a column that smooths the restless surface of the sea.
Naturalists call this lingering spool of glassy water the whale’s footprint. Out between
the Hawaiian islands of Maui and Lanai, Jim Darling nosed his small boat into a fresh
swirl. The whale that had left it was visible 40 feet below, suspended head down
in pure blueness with its 15-foot-long arms, or flippers, flared out to either side like
wings.
“That’s the posture humpbacks most often assume when they sing,” Darling said. A
hydrophone dangling under the boat picked up the animal’s voice and fed it into a tape
recorder. We could listen in with headphones but hardly needed them. The music was
reverberating through the hull and rising from the waves. Bass rumbles that could have
274 Appendix A
issued from the lowest octave of a cathedral pipe organ gave way to plaintive moans and
then to glissandos like air squealing out of a balloon when you stretch the neck taut.
With the notes building into phrases and the phrases into repeated themes, the song
may be the longest – up to 30 minutes – and the most complex in the animal kingdom.
All the humpbacks in a given region sing the same song, which is constantly evolving.
Experts have analyzed the frequencies, rhythms, and harmonics and the way themes
change from year to year and vary from one population to the next. Yet no one really
understands what these intricate arias are about.
We do know that humpbacks are found in every ocean. Together with blue, fin, sei,
Bryde’s, and mink whales, they belong to the rorqual family of baleen whales. Fully
grown females, which are bulkier than the males, can weigh 40 tons and reach lengths
of 50 feet.
Humpbacks tend to favor shallow areas, often quite close to shore, and they are among
the most sociable of the great whales and the most active at the surface, all of which
makes them among the easiest to observe. As a result, we know more about them than
about any other large whale. But we still don’t know a lot.
One thing the experts are certain of is that this species, depleted by whaling and
not protected throughout its range until 1966, is showing signs of a comeback. Early
population estimates are unreliable, and recent ones are hard to get, but numbers in the
North Atlantic seem to have rebounded from a few thousand to between 10,000 and
12,000. The North Pacific population was thought to have tumbled from 15,000 to fewer
than 2,000. That group stands at 5,000 to 8,000 today.
Knowing I was eager to absorb what biologists have been discovering about hump-
backs and their ongoing recovery, Darling, director of the West Coast Whale Research
Foundation in Vancouver, British Columbia, brought me along to Hawaii.
Since the singer beneath the boat didn’t seem bothered by our company, Darling asked
his longtime research partner, photographer Flip Nicklin, who often serves as Darling’s
eyes underwater, to slip overboard. I followed.
(g) Fragment from Slave-making queens, by Howard Topoff, Scientific

American, November 1999
In the animal world, both predators and parasites survive at the expense of other species.
Nevertheless, they don’t get the same press. I am besieged with mail containing pleas
for money on behalf of wolves and killer whales, but I have yet to see a T-shirt with the
slogan “Long Live the Hookworm.” The problem is, of course, that humans associate
a parasite lifestyle with disease. Our perception is of a furtive organism that insinuates
itself inside us and, unlike a decent predator, intends to destroys us ever so slowly.
. . . But there exists a form of parasitism considerably less macabre. Social parasitism,
as it is called, has evolved independently in such creatures as ants and birds. A female
cuckoo, for instance, lays her egg in the nest of another species, such as a warbler, and
leaves it for the host to rear. The brown-headed cowbird does the same. Each bird has
evolved so that it produces eggs that match those of its chosen baby-sitter.
Even more varied than these avian parasites are the slave-making ants. The unusual
behavior of the parasitic ant Polyergus breviceps – which I have been studying for
15 years in Arizona at the American Museum of Natural History’s Southwestern
Research Station – offers a perfect example. Like the other four species of Polyergus
The texts 275
found throughout the world, these ants have completely lost the ability to care for them-
selves. The workers do not forage for food, feed the young or the queen, or even clean up
their own nest. To survive, Polyergus ants must get workers from the related ant genus
Formica to do their chores for them. Thus, Polyergus workers periodically undertake
a slave raid in which about 1,500 of them travel up to 150 meters (492 feet), enter a
Formica nest, expel the Formica queen and workers, and capture the pupae.
Back at the Polyergus nest, slaves rear the raided brood until the young emerge.
The newly hatched Formica workers then assume all responsibility for maintaining the
mixed-species nest. They forage for nectar and dead arthropods, regurgitate food to
colony members, remove wastes and excavate new chambers. When the population be-
comes too large for the existing nest, it is the 3,000 or so Formica slaves that locate
another site and physically transport the approximately 2,000 Polyergus workers,
together with eggs, larvae, pupae and even the queen, to the new nest.
(h) Hijacking the rulebook, by Alan Ehrenhalt, New York Times,

December 20, 1998
A pretty good argument can be made that the defining moment of American democracy
didn’t occur in 1776 or 1787, as commonly supposed, but in 1801 – on the day that John
Adams, having been beaten at the polls, quietly packed his things and went home. Only
then did we know for sure that the system worked as advertised.
The routine transfer of power may not be the most dramatic feature of American
democracy, but it is the most important. It separates us from the majority of countries in
the world, which have still not achieved it. Conceding defeat and going home, or staying
on in the minority and allowing the winner to govern – these are not just the elements
of good manners and sportsmanship. They are the core of patriotism.
Those are rather windy thoughts, but I have had a hard time escaping them the last
few weeks as I’ve tried to make some sense of the events surrounding President Clinton,
Congress and impeachment. What will people say about all this 20 years from now?
Will they quiz each other on the minutiae of the Starr report, as they do on the contents
of the Warren report? Will the cigar and the stained dress become icons of American
political history? Or will posterity simply conclude that one of the two political parties,
having lost an election, saw an opportunity to nullify it and proved too weak to resist
the temptation?
You may find that a difficult question. I don’t. I feel reasonably certain of the final
verdict on the current impeachment affair because I think history will see it as the climax
of a six-year period marred by a troubling and deepening failure of the Republican party
to play within the established constitutional rules.
It was on Election Night 1992, not very far into the evening, that the Senate minority
leader, Bob Dole, hinted at the way his party planned to conduct itself in the months
ahead: it would filibuster any significant legislation the new Democratic President pro-
posed, forcing him to obtain 60 votes for Senate passage.
This was a form of scorched-earth partisan warfare unprecedented in modern polit-
ical life. Congress is supposed to operate by majority vote. It is true that the filibuster
has a long and disreputable Senate history and that, over the years, it has been used
more by Democrats than by Republicans. But only after 1992 did it become the cen-
terpiece of opposition conduct toward an elected President. What the Republicans did
276 Appendix A
in the Senate in 1993 amounted to an unreported constitutional usurpation. It should

have been denounced as such at the time, but it wasn’t. The punditocracy chose not to
notice.
In any case, it worked. Little that the President proposed became law in the two
years that he operated with Democratic majorities. There was no health care reform, no
economic stimulus package. On the merits, that is just as well. But the procedural con-
sequences turned out to be grave: Congressional Republicans were tempted by success
into even more dangerous constitutional mischief.
In the fall of 1995, emboldened by new majorities in both the House and the Senate,
they forced the closure of the Federal Government. For all the millions of words that have
been written about this event then and since, the reality of it has rarely been portrayed in
succinct terms. This was not a political showdown – it was an attempted constitutional
coup.
The Founding Fathers provided a mechanism for resolving disputes between Congress
and the White House: Congress passes a bill, the President vetoes it, and if sufficient
votes do not exist to override the veto, Congress lives with the decision. For the Republi-
cans to act as they did in 1995 – attempting to make the President sign legislation against
his will rather than trying to find the votes to override him – was an act of recklessness
so blatant that even an inattentive public understood it. Newt Gingrich backed down, the
Government reopened, and Mr. Clinton was re-elected.
But Republicans were already on the road to further adventure. In August of 1994, they
had orchestrated the dismissal of Robert Fiske, the independent counsel investigating
the Whitewater land deal, and replaced him with Kenneth Starr.
Reasonable people may differ on the quality of the evidence Mr. Starr has accumulated,
on the tactics he has used, on the way he has presented himself to the public. But his
appointment in the first place is impossible to defend.
Republicans selected as independent counsel a lawyer who was already involved in
consultations with a plaintiff suing the President in a civil dispute. No one concerned with
ultimate constitutional fairness could possibly have made such a choice; no legislative
majority interested in the appearance of justice could ever have approved it.
Now it is four years later. Congress is consumed by impeachment, and the majority
party seems genuinely puzzled by the absence of public support for the process. It
shouldn’t be. The American people aren’t suspicious about impeachment because of
their love for Bill Clinton. They are suspicious, in large part, because of the track record
of those bringing the charges. It’s not that there is no legitimate case to be made against
the President. It’s that Republicans, over a six-year career of consistent disrespect for
constitutional rules, have forfeited any right to be taken seriously in making it.
The unpleasant truth is that Congressional Republicans, in the generation before
1994, spent too many years out of power – too many years on the sidelines, uninvolved
in managing the governmental process and free to lob grenades at the institutions that
make it work. Eventually, they became very effective at it: that is one reason they won
the election of 1994.
Habits learned over decades do not fade easily. Having been lifted by the American
electorate into a position of genuine power, they have continued to behave more like a
party of insurgents, probing for cracks in the constitutional structure rather than taking
its rules seriously and looking for ways to make them work.
The texts 277
If Republicans in Congress have a common self-image, it is an image of conservatism.

No doubt every one of the Republicans in the current House would accept “conservative”
as an ideological label. But being a conservative must, in the end, be about something
more than tax cuts or family values. It must be about taking some responsibility for the
fragile procedures and institutions that over 200 years have made an orderly public life
possible.
There is nothing conservative about the way Republicans in Congress have conducted
themselves, either in the current impeachment debate or in most of the important con-
frontations of the past six years. The American people seem to know that, and there is
every sign that they are willing to judge accordingly.
(i) Fragment from The information revolution, by Peter Drucker, Atlantic

Monthly, October 99
The truly revolutionary impact of the Information Revolution is just beginning to be felt.
But it is not “information” that fuels this impact. It is not “artificial intelligence.” It is
not the effect of computers and data processing on decision-making, policymaking, or
strategy. It is something that practically no one foresaw or, indeed, even talked about
ten or fifteen years ago: e-commerce – that is, the explosive emergence of the Internet
as a major, perhaps eventually the major, worldwide distribution channel for goods,
for services, and, surprisingly, for managerial and professional jobs. This is profoundly
changing economies, markets, and industry structures; products and services and their
flow; consumer segmentation, consumer values, and consumer behavior; jobs and labor
markets. But the impact may be even greater on societies and politics and, above all, on
the way we see the world and ourselves in it.
At the same time, new and unexpected industries will no doubt emerge, and fast. One
is already here: biotechnology. And another: fish farming. Within the next fifty years fish
farming may change us from hunters and gatherers on the seas into “marine pastoralists” –
just as a similar innovation some 10,000 years ago changed our ancestors from hunters
and gatherers on the land into agriculturists and pastoralists.
It is likely that other new technologies will appear suddenly, leading to major new
industries. What they may be is impossible even to guess at. But it is highly probable –
indeed, nearly certain – that they will emerge, and fairly soon. And it is nearly certain
that few of them – and few industries based on them – will come out of computer
and information technology. Like biotechnology and fish farming, each will emerge
from its own unique and unexpected technology. Of course, these are only predictions.
But they are made on the assumption that the Information Revolution will evolve as
several earlier technology-based “revolutions” have evolved over the past 500 years,
since Gutenberg’s printing revolution, around 1455. In particular the assumption is that
the Information Revolution will be like the Industrial Revolution of the late eighteenth
and early nineteenth centuries. And that is indeed exactly how the Information Revolution
has been during its first fifty years.
The Railroad
The Information Revolution is now at the point at which the Industrial Revolution was
in the early 1820s, about forty years after James Watt’s improved steam engine (first
installed in 1776) was first applied, in 1785, to an industrial operation – the spinning of
278 Appendix A
cotton. And the steam engine was to the first Industrial Revolution what the computer
has been to the Information Revolution – its trigger, but above all its symbol. Almost
everybody today believes that nothing in economic history has ever moved as fast as,
or had a greater impact than, the Information Revolution. But the Industrial Revolution
moved at least as fast in the same time span, and had probably an equal impact if
not a greater one. In short order it mechanized the great majority of manufacturing
processes, beginning with the production of the most important industrial commodity of
the eighteenth and early nineteenth centuries: textiles. Moore’s Law asserts that the price
of the Information Revolution’s basic element, the microchip, drops by 50 percent every
eighteen months. The same was true of the products whose manufacture was mechanized
by the first Industrial Revolution. The price of cotton textiles fell by 90 percent in the
fifty years spanning the start of the eighteenth century. The production of cotton textiles
increased at least 150-fold in Britain alone in the same period. And although textiles
were the most visible product of its early years, the Industrial Revolution mechanized
the production of practically all other major goods, such as paper, glass, leather, and
bricks. Its impact was by no means confined to consumer goods. The production of
iron and ironware – for example, wire – became mechanized and steam-driven as fast
as did that of textiles, with the same effects on cost, price, and output. By the end of
the Napoleonic Wars the making of guns was steam-driven throughout Europe; cannons
were made ten to twenty times as fast as before, and their cost dropped by more than two
thirds. By that time Eli Whitney had similarly mechanized the manufacture of muskets
in America and had created the first mass-production industry.
(j) Fragment from The Chinese Potter, by Margaret Medley; p. 106

The predominant output was the white ware with transparent ivory toned glaze which
made the kilns famous. The other wares were a soft, dark brown or dense black glazed
type, which is rare, a white ware painted in soft iron brown of rather light tone, again
a rare type, and a group of uncertainly dated monochrome and polychrome lead glazed
earthenwares about which there is at present relatively little information.
The kilns clearly specialized in the production of the porcellanous ware. It is charac-
terized by the extreme whiteness and hardness of the body, which may be translucent,
and the glassy, transparent, and warm ivory tone of the glaze. On some bowls this may
show what the Chinese call “tear stains,” slight thickenings where the glaze has run down
the foot after dipping. That the kilns confined themselves very largely to the production
of bowls, basins, dishes and plates, is evident from the immense waste heaps in the
vicinity of the kilns, and from the predominance of these shapes in collections all over
the world. There is by contrast a very small number of vases. It was good economic sense
to limit production to the most popular and practical forms; for it made the achievement
of high quality and large output easier; moreover it may well have contributed to the
long survival of the kilns in this somewhat remote area.
During the late tenth and early eleventh centuries production was more varied than it
became later. There are, for instance, vases and jars dating from this fairly early stage
with rather heavily carved decoration in which lotus petals played a large part. By the
second half of the eleventh century production was already concentrated, to the virtual
exclusion of other shapes, on the manufacture of open and flat wares with carved and
incised decoration of floral scrolls and sprays (especially lotus), of ducks and geese
The texts 279
among reeds, dragons, and of fish. The carving and incising was a highly developed
skill and must have been carried out by craftsmen specially trained in this work. Were
this not the case it would be almost impossible to account for the consistency in design,
the fluency of line and the almost unequalled high quality.
(k) Fragment from The pride of the cities, by Peter Beinart, New Republic,
June 1997
Michael White, the mayor of Cleveland, is a Democrat, an African American and the son
of a union activist. And he is at war with his party. First, he backed an end to forced busing.
Then, he supported Republican Governor George Voinovich’s radical school choice law,
which offers students vouchers at parochial as well as private schools. Then, he made
city workers compete against private firms for garbage collection, road maintenance
and other contracts, prompting union officials to walk out of a speech he gave at the
Democratic National Convention. Now, he’s allied with the governor again, backing a
bill by two Republican State legislators to grant him control over Cleveland’s desti-
tute school system and the authority to get “rid of any people who aren’t directly tied to
the direct education of children.” Arrayed against him: the teachers’ union, the NAACP
and just about every elected Democrat in the city of Cleveland.
For a big-city mayor to be so at odds with his party would seem peculiar. Except
that it’s happening everywhere. Mike White got the idea to take over the schools from
Richard Daley, the Democratic mayor of Chicago, who battled the municipal unions
and his own party’s nominee for governor, and two years ago won the right to introduce
radical reforms to the management of Chicago’s public schools. And White learned
about the introduction of competition into city services in part from a study trip he
and his staff took to Indianapolis, where Republican Mayor Steven Goldsmith has
reduced city bureaucracy so radically that he’s angered both patronage-minded GOP
officials and the traditional pro-Republican police unions. Goldsmith is also viewed as
a “kindred spirit” by Mayor John Norquist of Milwaukee, a Democrat who, for the sin
of supporting Republican Governor Tommy Thompson’s welfare and school choice re-
forms, has made himself persona non grata with his city’s public employee unions, the
NAACP and the Democratic speaker of the House of Representatives.
It goes on. Los Angeles Mayor Richard Riordan, whose support for affirmative
action, gay rights and Senator Dianne Feinstein has led GOP Californians to dub him
a RNO (Republican in Name Only), journeyed to Indianapolis during his first week in
office and cited Goldsmith in his inaugural address. And Rudolph Giuliani, the Mario
Cuomo-endorsing, illegal-immigrant defending, New York Republican who recently
admitted that ideologically he is a “moderate Democrat,” has met with Goldsmith on
four separate occasions. Giuliani and Riordan also received pre-inauguration seminars
from Philadelphia Mayor Ed Rendell, a Democrat, who, like Mike White, forced city
employees to compete for contracts with outside firms, outraging municipal unions and
most of his party.
Something interesting is happening here. Over the past five years or so, a half dozen
Democratic and Republican mayors have come together in what an aide to one calls
“an informal network.” They speak to each other regularly, they cite each other without
prompting, they copy each other’s initiatives. In almost every case, they represent a
radical break with their predecessors in office, and that break is largely about managing
280 Appendix A
city government efficiently in the public interest rather than using it as a mechanism for
arbitrating competing group interests. The new mayors are hugely popular: Mike White,
Richard Daley, Steven Goldsmith, John Norquist, Richard Riordan and Ed Rendell have
all been re-elected by wide margins, and Rudy Giuliani will almost certainly follow
suit this fall. But they are outsiders in their own parties, viewed with suspicion and
even contempt by the parties’ most powerful constituencies. To explain their success
and their iconoclasm, the press often dubs them pragmatists. But that misses the point.
They have an ideology: that cities can dramatically alleviate seemingly endemic urban
afflictions without a massive redistribution of wealth, that the way to achieve this is by
using competition to make city services radically more efficient, and that cities must
tolerate diverse identities without celebrating them to the detriment of a shared sense
of public interest. These ideas have a coherence and a history. They just don’t have a
party, yet.
The history is progressivism. At the turn of the last century, America’s cities were a
scandal. The squalor, pathology and cultural transformation to which mass immigration
had given rise shocked and terrified the native-born middle class. And, as outraged as
they were by the immigrants themselves, respectable people were even more hostile
to the municipal regimes the immigrants spawned. As Andrew White, the president of
Cornell, wrote in the journal Forum in 1890, “the city governments of the United States
are the worse in Christendom.”
The political machines that men like White despised had been created to bridge the
material and cultural distance separating the immigrants from native society. In cities that
offered newcomers few formal services, and in which the private sector often remained
closed, the immigrant-driven city machines supplied housing, fuel, charity and – most
importantly – jobs. Under machine leadership, America’s cities built roads, sewers,
streetlights and railcars at a furious pace, usually with little attention to the projects’
cost-effectiveness, creating as many jobs as possible and distributing them to appease
different constituencies. As Kenneth Fox notes in Better City Government, his history of
that era, America’s big cities in 1880 spent more than twice as much per capita as did the
federal government. And the machines appealed to the immigrants symbolically as well
as materially, installing ward leaders and aldermen who looked and talked like them,
and defending their traditions against a disapproving native middle class that wanted
to close down the immigrants’ saloons and their dance halls – and bust their corrupt,
powerful machines.
(l) Fragment from How it Works, by Jim Collins, US Airways Magazine,

Attaché, May 2001
The largest and most complex highway-engineering project in American history has
been making headlines over the past several months. Boston’s central artery project,
known as the Big Dig, has become nightly news fare thanks to the massive cost overruns
and steadily growing complaints of mismanagement and corruption. Overshadowed by
scandal, the work itself – whose scale and ambition are truly mind-boggling – continues
almost as matter of course.
The heart of the project is an eight-to-ten-lane highway. It is being built below ground
to relieve the legendary traffic pressure caused by Boston’s two major arteries, which
converge downtown awkwardly near the harbor, skyscrapers, and historic buildings.
The texts 281
Threading 35 lane-miles of tunnel through Boston’s 3-D maze of subway lines and
building foundations, all without drastically disturbing the normal routines of surface
life, is an engineering plan that borders on the fantastic. In some places the tunnels are
120 feet deep. In some particularly delicate places the road-work passes within just a few
feet of skyscraper foundations or beneath construction projects. The first two frequently
asked questions on the project’s official Web site are “What are you building?” and “Are
you nuts?”
Surprisingly, tunnels are among the most ancient engineering feats. The earliest were
very likely extensions of prehistoric cave dwellings. The Babylonians, in the twenty-
second century BC, built a masonry tunnel beneath the Euphrates River that connected
the royal palace with a major temple. The Egyptians, using copper-bladed saws, exca-
vated long passageways and intricate rooms inside soft-rock cliffs. The Romans built
an elaborate network of above- and below-ground acqueducts to carry water. And
they tunneled through solid rock by repeatedly heating it with fire and then cooling
it with water, causing the rock face to fracture. The greatest of those acqueduct tunnels,
which eventually drained Lake Fucino in central Italy, stretched more than three miles
underground.
By the 1700s, tunnels were increasingly being constructed for use with canal systems.
In the history of freight transportation, prior to the invention of the railroad, canals proved
the most logical way to ship material over great distances. Sections of the canals often
need to be buried beneath ground, and this was accomplished with long open trenches
that were dug down from the surface, faced with masonry, then covered over. Canals
of extraordinary length were built in this manner, including the longest in the United
States, the 729-feet Union Canal Tunnel in Pennsylvania.
About the same time, an important hard-rock blasting technique was developed using
explosives. Dynamite was something new, and it gave engineers their most powerful
tool for tunneling through hillsides that couldn’t be opened by digging from above.
One of the most famous – and costly – early examples of a corridor blasted with
dynamite is the Hoosac Tunnel, in the Berkshire Mountains of Western Massachusetts.
Started in 1851, the Hoosac was a desperate attempt to compete with the convenient
Hudson River transportation corridor. It was hoped that direct east–west rail service
would connect and help exploit the burgeoning centers that were springing up in the
Midwest and around the Great Lakes. The planned route had just one major problem:
a small mountain directly in its path, an obstacle that would require an unprecedented
tunnel almost five miles long. The Hoosac was the first commercial project to make
use of the powerful and extremely unstable explosive nitroglycerine. Its human toll
would be immense, as science-and-technology writer Fred Hapgood vividly describes:
“In those days hard-rock tunnels were dig by chiseling a hole, filling it with explosive,
lighting a fuse, running back behind a shield, waiting for the blast, then returning to
the heading to wedge reinforcing timers into place, shatter the “muck” with sledges,
and shovel it into muckcars for excavation. This cycle could be interrupted at any point
by falling rocks or machinery, collapsing timbers, unintended explosions, or a dozen
other species of industrial accident, including fire and floods. In 1867 a wooden house
built over a shaft caught fire and collapsed downward, killing all 13 in the shift working
below. The explosives killed constantly, year after year . . . Nearly 200 were killed on
the job.
282 Appendix A
The Hoosac Tunnel took 22 years and $21 million to complete. As with the Big
Dig, critics railed against the project for overruns, poor planning, and delays in the
construction schedule. But at 4.82 miles it was an engineering marvel and it remained
the longest tunnel in America until 1916. It never came close, though, to approaching
the economic expectations of its promoters.
The blasting technique used in the Hoosac’s construction remains basically the same
to this day. But during the final years of the Hoosac Tunnel’s construction, blasting
was enhanced by the safer technique of drilling, which British inventor George Law
developed in 1865. A hammer tool operated by an air-driven piston was quickly being
used to break up rock, which could then be excavated through horizontal shafts called
drifts, or lifted through vertical shafts that also provided ventilation and exhaust.
By the time of the rock drill and the great railroad tunnels, a Frenchman named Marc
Isambard Brunel had already solved another of tunneling’s great challenges: how to
excavate below water without mud and water seeping in and causing the opening to
collapse. Borrowing an idea from nature, Brunel recreated the action of a tiny marine
borer known as the shipworm, whose shell plates allow it to bore through timber and
push sawdust out behind as it goes along. Brunel’s tunnel shield, a giant iron frame, was
forced through soft soil by screw jacks while miners dug through shuttered openings in
the shield’s forward face. Excavated earth was transported back through the frame as it
slowly advanced. Brunel used his rectangular shield to complete the world’s first true
tunnel in 1843 in London, below the Thames River.
Within a few years such cutting shields would be made smaller, circular, and more
powerful. Many of them even made use of compressed air to keep water out while the
steel linings were being installed.
And it was these developments, taken together, that set the stage for one of the truly
remarkable achievements of the twentieth century, the construction of New York and
New Jersey’s Holland Tunnel.
(m) Fragment from “The Dead” by James Joyce; pp. 177–79

“Is it snowing again, Mr Conroy?” asked Lily.
She had preceded him into the pantry to help him off with his overcoat. Gabriel
smiled at the three syllables she had given his surname and glanced at her. She was a
slim, growing girl, pale in complexion and with hay-coloured hair. The gas in the pantry
made her look still paler. Gabriel had known her when she was a child and used to sit
on the lowest step nursing a rag doll.
“Yes, Lily,” he answered, “and I think we’re in for a night of it.” He looked up at the
pantry ceiling, which was shaking with the stamping and shuffling of feet on the floor
above, listened for a moment to the piano and then glanced at the girl, who was folding
his overcoat carefully at the end of a shelf.
“O, then,” said Gabriel gaily, “I suppose we’ll be going to your wedding one of these
fine days with your young man, eh?” The girl glanced back at him over her shoulder and
said with great bitterness:
“The men that is now is only all palaver and what they can get out of you.”
Gabriel coloured, as if he felt he had made a mistake, and, without looking at her,
kicked off his galoshes and flicked actively with his muffler at his patent-leather shoes.
The texts 283
He was a stout, tallish young man. The high colour of his cheeks pushed upwards even
to his forehead, where it scattered itself in a few formless patches of pale red; and on
his hairless face there scintillated restlessly the polished lenses and the bright gilt rims
of the glasses which screened his delicate and restless eyes. His glossy black hair was
parted in the middle and brushed in a long curve behind his ears where it curled slightly
beneath the groove left by his hat. When he had flicked lustre into his shoes he stood
up and pulled his waistcoat down more tightly on his plump body. Then he took a coin
rapidly from his pocket.
“O Lily,” he said, thrusting it into her hands, “it’s Christmastime, isn’t it? Just . . .
here’s a little . . .” He walked rapidly towards the door.
“O no, sir!” cried the girl, following him. “Really, sir, I wouldn’t take it.”
“Christmas-time! Christmas-time!” said Gabriel, almost trotting to the stairs and
waving his hand to her in deprecation. The girl, seeing that he had gained the stairs,
called out after him: “Well, thank you, sir.”
He waited outside the drawing-room door until the waltz should finish, listening to
the skirts that swept against it and to the shuffling of feet. He was still discomposed by
the girl’s bitter and sudden retort. It had cast a gloom over him which he tried to dispel
by arranging his cuffs and the bows of his tie. He then took from his waistcoat pocket a
little paper and glanced at the headings he had made for his speech. He was undecided
about the lines from Robert Browning, for he feared they would be above the heads of
his hearers. Some quotation that they would recognize from Shakespeare or from the
Melodies would be better. The indelicate clacking of the men’s heels and the shuffling
of their soles reminded him that their grade of culture differed from his. He would only
make himself ridiculous by quoting poetry to them which they could not understand.
They would think that he was airing his superior education. He would fail with them
just as he had failed with the girl in the pantry. He had taken up a wrong tone. His whole
speech was a mistake from first to last, an utter failure.
(n) Let teenagers try adulthood, by Leon Botstein, New York Times, May 1999
The national outpouring after the Littleton shootings has forced us to confront something
we have suspected for a long time: the American high school is obsolete and should be
abolished. In the last month, high school students present and past have come forward
with stories about cliques and the artificial intensity of a world defined by insiders and
outsiders, in which the insiders hold sway because of superficial definitions of good
looks and attractiveness, popularity and sports prowess.
The team sports of high school dominate more than student culture. A community’s
loyalty to the high school system is often based on the extent to which varsity teams
succeed. High school administrators and faculty members are often former coaches, and
the coaches themselves are placed in a separate, untouchable category. The result is that
the culture of the inside elite is not contested by the adults in the school. Individuality
and dissent are discouraged.
But the rules of high school turn out not to be the rules of life. Often the high school
outsider becomes the more successful and admired adult. The definitions of masculinity
and femininity go through sufficient transformations to make the game of popularity in
high school an embarrassment. No other group of adults young or old is confined to
284 Appendix A
an age-segregated environment, much like a gang in which individuals of the same age
group define each other’s world. In no workplace, not even in colleges or universities,
is there such a narrow segmentation by chronology.
Given the poor quality of recruitment and training for high school teachers, it is no
wonder that the curriculum and the enterprise of learning hold so little sway over young
people. When puberty meets education and learning in modern America, the victory
of puberty masquerading as popular culture and the tyranny of peer groups based on
ludicrous values meets little resistance.
By the time those who graduate from high school go on to college and realize what
is really at stake in becoming an adult, too many opportunities have been lost and
too much time has been wasted. Most thoughtful young people suffer the high school
environment in silence and in their junior and senior years mark time waiting for col-
lege to begin. The Littleton killers, above and beyond the psychological demons that
drove them to violence, felt trapped in the artificiality of the high school world and
believed it to be real. They engineered their moment of undivided attention and impor-
tance in the absence of any confidence that life after high school could have a different
meaning.
Adults should face the fact that they don’t like adolescents and that they have used
high school to isolate the pubescent and hormonally active adolescent away from both
the picture-book idealized innocence of childhood and the more accountable world of
adulthood. But the primary reason high school doesn’t work anymore, if it ever did, is
that young people mature substantially earlier in the late 20th century than they did when
the high school was invented. For example, the age of menstruation has dropped at least
two years since the beginning of this century and not surprisingly, the onset of sexual
activity has dropped in proportion. An institution intended for children in transition now
holds young adults back well beyond the developmental point for which high school
was originally designed.
Furthermore, whatever constraints to the presumption of adulthood among young
people may have existed decades ago have now fallen away. Information and images, as
well as the real and virtual freedom of movement we associate with adulthood, are now
accessible to every 15- and 16-year-old.
Secondary education must be re-thought. Elementary school should begin at age 4 or
5 and end with the sixth grade. We should entirely abandon the concept of the middle
school and junior high school. Beginning with the seventh grade, there should be four
years of secondary education that we may call high school. Young people should graduate
at 16 rather than 18.
They could then enter the real world, the world of work or national service, in which
they would take a place of responsibility along with older adults in mixed company. They
could stay at home and attend junior college, or they could go away to college. For all the
faults of college, at least the adults who dominate the world of colleges, the faculty, were
selected precisely because they were exceptional and different, not because they were
popular. Despite the often cavalier attitude toward teaching in college, at least physicists
know their physics, mathematicians know and love their mathematics, and music is
taught by musicians, not by graduates of education schools, where the disciplines are
subordinated to the study of classroom management.
The texts 285
For those 16-year-olds who do not want to do any of the above, we might construct
new kinds of institutions, each dedicated to one activity, from science to dance, to which
adolescents could devote their energies while working together with professionals in
those fields.
At 16, young Americans are prepared to be taken seriously and to develop the moti-
vations and interests that will serve them well in adult life. They need to enter a world
where they are not in a lunchroom only with their peers, estranged from other age groups
and cut off from the game of life as it is really played. There is nothing utopian about this
idea; it is immensely practical and efficient, and its implementation is long overdue. We
need to face biological and cultural facts and not prolong the life of a flawed institution
that is out of date.
Appendix B: Glossary
Abstract entity: The class of situations, and discourse entities, that consists
of Facts and Propositions. They are licensed by clausal
complements of certain predicates. For instance, the clausal
complement of “I know that Mary refused the offer” refers
to a Fact; the clausal complement of “Mary’s refusing the
offer was unlikely” refers to a Proposition.
Accommodation: A type of inference in which the receiver infers the existence
of an entity if it is necessary to do so for coherence. For
instance on encountering the cat if no cat has been
introduced, one infers the existence of a cat.
Accomplishment: An event that is telic, durative, and dynamic, resulting in a
change of state, e.g. “John ate an apple,” “Mary walked to
school.”
Achievement: An event that is telic, instantaneous, and dynamic, resulting
in a change of state, e.g. “Mary won the race,” “John reached
the top.” Instantaneous events consist only of a single stage.
Activity: An event that is atelic, durative, and dynamic, e.g. “John
strolled in the park,” “Mary slept.”
Anaphora: The interpretation of a pronoun or tense that depends on an
antecedent for interpretation. For instance, in “John washed
himself,” “Mary said that she was ready,” the anaphor
himself and the pronoun she depend on the antecedent
proper names for interpretation.
Anaphoric Tense: The pattern of tense interpretation in which a time is
identified with an earlier time in the discourse. It is typical of
passages in the Description mode.
Aspectual viewpoint: Viewpoint makes visible for semantic interpretation all or
part of a situation. It is indicated by the simple and
progressive verb forms. The simple form conveys the
perfective viewpoint, making visible a bounded event or
state, as in “Mary talked to Bill.” “Kim was here.” The
progressive makes visible part of an ongoing event, as in
“Mary was talking to Bill.”
Background: Sentences can be partitioned into background and focus. The
background may consist of a topic phrase and other material,
286
Glossary 287
not necessarily forming a syntactic constituent.

Typically, background represents what is salient or
inferrable in the discourse. “Background” also
contrasts with “foreground” in that background
presents supportive rather than essential information.
Bounded event: An event that is bounded in some way, conveyed in
sentences by the perfective viewpoint. Telic and
instantaneous events have intrinsic bounds, as in “The
child drew an eggplant,” “The bird flapped a wing.”
States and atelic events may have explicit bounds, as in
“She slept from 3 to 4”; or implicit bounds, as in “She
walked.”
Bounded Event A pragmatic constraint that bars the expression of
Constraint: bounded events in the present. By convention,
communication occurs in a single moment, so that
events can only be expressed as ongoing. The constraint
leads to the understanding of perfective event sentences
with present tense as generalizing statives, e.g. “John
feeds the cat.”
Canonical structure: The basic syntactic structure of English,
Subject-Verb-Object-Adjunct; the subject argument is
agent if there is an agent in the sentence, as in “Mary
planted a rosebush yesterday.” Canonical structures are
defined for assertions: questions and imperatives are
different in some way.
Cleft sentence: A non-canonical syntactic structure in which one
phrase is extracted from a clause while the rest appears
in a relative-like structure: “It was Mary who planted a
rosebush,” “What Mary planted was a rosebush.” The
information in the relative-like clause I (the “clefted”
clause) is presupposed.
Closed system: A limited set of forms of a sub-system, as in tense,
aspectual viewpoint, and type of referring expression.
The choice of a form has contrastive force and can
suggest additional discourse meanings beyond the
lexical content. For instance, using a full NP or a
pronoun to refer to a person (Mary Jones vs she)
suggests that the person is identifiable or already
familiar in the discourse.
Coercion: A shift of situation type from a basic type to another,
derived situation type. Shifts are triggered by adverbials
in the context – for instance, a state verb constellation
shifts in the presence of a dynamic adverbial: compare
“Mary knew the truth,” a state, with “Suddenly Mary
knew the truth,” a dynamic event. Derived situation
288 Appendix B
types include the basic verb constellation and the

triggering adverbial.
Comment: The part of a sentence that says something about the topic.
Sentences may be partitioned into topic phrase and
comment. The comment typically conveys new
information, the topic is typically familiar information.
Composite: A group of linguistic forms that together convey a
particular meaning; they may come from different phrases
in a sentence, or from different sentences. For instance, the
temporal location of a sentence is conveyed by the verb
tense and temporal adverbial, if any; aspectual situation
types are conveyed by a composite of the main verb, its
arguments, and adverbials, if any.
Continuity pattern: Pattern of tense interpretation for narrative non-first
clauses, in which bounded events are taken as sequential,
advancing narrative time.
Deixis, deictic: the term for linguistic expressions that are understood in
relation to the speaker’s time and place, the here and now.
Tenses are anchored to Speech Time: we understand past
tense to locate a time before Speech Time for instance.
Other deictic forms are the pronouns I and you; this and
that; the adverbs here, there.
Derived situation type: A situation type that is derived by shifting from a simpler
type.The shift is triggered by an adverbial or other
information in the context. The derived verb constellation
includes the adverbial as well as the verb and its
arguments, e.g. “Mary read a book” is telic; “Mary read a
book for an hour” is atelic, derived by coercion due to the
durative adverbial “for an hour.”
Discourse entity: Symbols in a Discourse Representation Structure, that
represent the people, objects, times, situations talked about
in the discourse and introduced into the universe of
discourse.
Discourse topic: The subject of a large or small chunk of discourse.
Discourse topics need not be explicitly stated: they are
often inferred or constructed by the receiver.
Dynamism: The semantic property that distinguishes events from
states. Dynamic events occur at successive stages in time.
The linguistic correlates of dynamism include the
progressive viewpoint (“Laura was swimming”), the “do”
of pseudo-cleft sentences (“What Laura did was swim”),
forms implying agency or volition.
Event: The class of situations with the property of dynamism.
Events may be durative or instantaneous, telic or atelic.
They are expressed by the composite of a verb and its
arguments. Event clauses have linguistic properties
Glossary 289
associated with dynamism and other features; they

accept the progressive or perfective aspectual
viewpoint.
Evidential: Linguistic expressions that convey evaluation, or
different degrees of commitment to what is asserted.
Evidentials include adverbials (probably, possibly,
etc.), modals (could, may).
External override principle: This principle applies to most cases of coercion, in
which a verb constellation is shifted to a derived verb
constellation. Adverbials often trigger the shift. When
there is a clash between temporal features of the verb
constellation and the adverbial, the value of the
adverbial feature overrides. For instance, “John feeds
the cat” is an event verb constellation; “John often
feeds the cat” is a Generalizing Stative.
Extraposition: A syntactic construction that is non-canonical, with a
clause or phrase moved to the rightmost edge of the
sentence. The most frequent type has a clausal subject
moved to the right, with a non-referring it in subject
position: “It was surprising that the plane left on
time.”
Familiarity status: Classification of individual entities in discourse
according to whether they are familiar in the
discourse or to the receiver. The main classes of
familiarity status are New, Evoked, or Inferrable. New
entities may be brand-new, or Unused; Evoked entities
are familiar, usually coreferential with an entity
already introduced; Inferrable entities are triggered
from background knowledge by information in the
text.
Figure–Ground: Information organized so that one referent, the Figure,
is located in terms of the other, the Ground. In
sentences the subject is Figure and the predicate is the
Ground, as in “The bicycle is near the house.” The
Figure is salient, usually smaller and less permanent
than the Ground.
Focus: The focus phrase gives the main contribution of a
sentence. Focused information is new to the discourse;
or presented in a new way. The focus phrase is
canonically rightmost in a sentence and receives the
sentence accent. If in another position, it is
highlighted by intonation and/or non-canonical
syntactic structure. “Strong focus” is contrastive or
emphatic; it is conveyed by heavy stress or
underlining, or focus operators such as only, even.
“Informational focus” occurs in all sentences.
290 Appendix B
General Stative: The class of situations, and discourse entities, that

consist of generalizations rather than particulars. The
class consists of Generic sentences that refer to
classes or abstract individuals, e.g. “Elephants are
gentle beasts”; and Generalizing sentences that
express a pattern of events or states, e.g. “Mary often
walks to school.”
Generalizing Stative: Generalization over a pattern of events or states, with
an implicit or explicit frequency adverbial.
Sentences that express them are derived statives,
often with an event verb constellation, present tense,
and the perfective viewpoint, e.g. “John reads the
newspaper in the evening.”
Generic: Sentence that predicates something of a class or
abstract individual, rather than a particular referent.
Such sentences often have subjects with a bare plural
NP or the definite article: “Whales live in ocean
waters”; “The whale lives in ocean waters.”
Indirect speech and thought: Linguistic presentation in which the contents of a
person’s speech or thought are in a complement
clause, as in “Mary said that she was pleased.” The
names and deictics are systematically shifted: actual
speech would have I and am.
Inversion: A type of syntactic structure in which the canonical
order is inverted, often so that a phrase canonically
in the predicate appears at the beginning of the
sentence. In locative inversion, a location phrase is
first: “On the table I saw a small notebook.” In other
inversions a complement clause may appear first, as
in “‘Let’s go,’ he said”; or, subject and predicate
may be reversed: “On the table was a small
notebook”; etc.
Limited anaphora: Pattern of tense interpretation, typical of narrative
non-first clauses, in which states and ongoing events
are located at the most recently established
Reference Time.
Link: An element in the partitioning of a sentence; the
others are Focus and Tail. The Link phrase indicates
a locus of update in the theory of Vallduvı́ &
Engdahl (1996).
Locally free reflexive (LFR): A reflexive not locally bound by an antecedent. The
antecedent may be relatively far away, not in a
position to bind the reflexive by Binding Theory
principles – “She pulled the blanket over herself.”
LFRs convey logophoric meaning; subjective
meanings of empathy; contrast or emphasis.
Glossary 291
Logophoric: Pronouns that indicate coreference to an antecedent

subject of consciousness, as in “Maryi thought that
shei would win.”
Metaphorical progression: The type of text progression found in atemporal
texts. They progress metaphorically through the
information space of the text. A Primary Referent is
identified for each clause; progression occurs when
Primary Referents have different metaphorical
locations.
Non-canonical structure: Syntactic structures that depart from the canonical in
some way. Phrases may not be in their original
position, as in a passive: “The rosebush was planted
by Mary”; there may be special emphasis or
syntactic highlighting, as in the cleft sentence “It was
a rosebush that Mary planted.” Non-canonical
structures may convey special meanings: cleft
sentences presuppose the truth of the that-clause.
Perfect: A tense–aspect form with auxiliary have and past
participle form of the main verb. The perfect
conveys that a situation occurred prior to Reference
Time, for instance “Mary and John have left”; “She
had solved the problem”. Perfect sentences are
stataive although the prior situation may be
dynamic.
Perfective viewpoint: The aspectual viewpoint indicated by the simple
form of a verb, without an auxiliary. The perfective
conveys bounded events, and states (in English; this
viewpoint differs across languages), as in “Mary ate
an apple,” “Kim believes in ghosts.”
Poset linking: A type of connectedness between sentences in
which a discourse entity is linked to the context.
Entities are linked through a poset licensed by
information in the context. If a salient poset can be
inferred, there is a linking relation. The link entity is
ranked with other entities in the poset according to
formal relations such as part-of, attribute-of, prior-to,
identity. Entities may be inferred through poset
linking.
Presentational constructions: Syntactic structures that introduce a referent, usually
unfamiliar, to a discourse. For instance, “There was a
unicorn in the garden,” and certain inversions: “In
the garden was a unicorn.” These structures have a
focus phrase (“the unicorn” in each example) but no
topic phrase.
Presentational progression: Text progression from one sentence topic to another.
It is presentational because the sentence topic
292 Appendix B
depends in part on the syntactic surface structure of a

sentence.
Presupposition: In “semantic presupposition,” particular linguistic
expressions conventionally establish that information is
shared knowledge: “Tom stopped running” entails that
“Tom ran.” In “pragmatic presupposition,” the receiver
infers that an entity or a proposition is an already-
established, uncontroversial part of the common ground.
The sequence “Mary had an accident driving home from
work. The steering-wheel was defective.” pragmatically
presupposes the car that Mary was driving.
Primary Referent: The semantically central entity in a situation expressed
by a clause. In events, the Primary Referent moves or
changes; in states, a property is ascribed to the Primary
Referent. The Primary Referent corresponds to the
Theme/Patient argument of a clause. The Primary Referent
is the locus for metaphorical motion in atemporal text
passages.
Progressive: The aspectual viewpoint indicated by the progressive verb
auxiliary be+ing. The progressive is a type of imperfective
viewpoint. It conveys open, unbounded events, as in “Mary
was eating an apple.” The progressive does not occur
neutrally with states.
Reference Time (RT): One of three times introduced by a tense: the temporal
standpoint time. In present tense, RT = Speech Time; in
simple past tense, RT precedes Speech Time; in future, RT
follows Speech Time. RT = t2 in Discourse Representation
Structure.
Responsible Source: The mind to which responsibility for subjective
interpretation is ascribed.
Semelfactive: A type of event with the features dynamic, atelic,
instantaneous, for instance “The bird flapped a wing,” “Bill
coughed.” Instantaneous events consist only of a single
stage.
Situation: The term is used generally, for all types of aspectual
entities; it also refers to the sub-type of particular events and
states.
Situation time (SitT): One of three times introduced by tense: the time at which a
situation occurs. With the simple tenses, Situation Time is
the same as Reference Time. With the perfect tenses, SitT
precedes RT; with the future-in-past, SitT follows RT.
SitT = t3 in Discourse Representation Structure.
Situation type: Situations classified according to their internal temporal
properties, e.g. Events and States, Generalizing Statives,
Facts and Propositions. The situation types are semantic
concepts, idealizations of situations according to internal
Glossary 293
temporal properties. They have linguistic correlates of

distribution and interpretation.
Telic–Atelic: A feature of internal temporal structure that distinguishes
types of events by whether they have a goal (a “telos”) or
natural final endpoint. Telic events have a natural final
endpoint (draw a circle, walk to school); atelic events can
stop at any time, and thus have arbitrary final endpoints (walk
in the park, sleep).
Thetic sentence: The term refers to a sentence that is understood as an
unpartitioned whole rather than a predication about a
particular referent, as “There are yellow flowers” or “It’s
raining.” Thetic sentences can answer questions such as
“What happened?” or “What is there?” that do not single out
any referent. Thetic sentences have no topic phrase.
Topic of a sentence: The referent that a sentence is about. Topic phrases are tracked
in the presentational progression of a text. The topic phrase of
a sentence is usually the subject, and the topic referent is
often familiar in a text; but neither is a defining property.
Verb constellation: The verb and its arguments. Verb constellations express the
situation type of a sentence. In derived situation types adverbs
are part of a derived verb constellation.
References
Abusch, Dorit, 1997. Sequence of tense & temporal de re. Linguistics & Philosophy 20:
1–50.
Acker, Liane, & Bruce Porter, 1994. Extracting viewpoints from knowledge bases.
Proceedings of the 12th National Conference on Artificial Intelligence. Menlo Park,
Calif.: 547–52.
Allen, James, 1983. Maintaining knowledge about temporal intervals. Communications
of the Association for Computing Machinery 26 (11): 832–43.
Allen, James, & C. Raymond Perrault, 1980. Analyzing intentions in utterances. Com-
putational Linguistics, 15: 143–78.
Ariel, Mira, 1989. Referring and accessibility. Journal of Linguistics 24: 65–87.
1990. Accessing Noun Phrase Antecedents. London: Routledge.
Arnold, Jennifer, Thomas Wasow, Anthony Losonoco, & Ryan Ginstrom, 2000.
Heaviness vs newness: The effects of structural complexity and discourse status on
constituent ordering. Language 76: 28–55.
Asher, Nicholas, 1993. Reference to Abstract Objects in Discourse. Dordrecht: Kluwer
Academic Publishers.
Asher, Nicholas, & Alex Lascarides, 1993. Temporal interpretation, discourse relations
and commonsense entailment. Linguistics & Philosophy 16: 437–94.
1998a. Bridging. Journal of Semantics 15: 83–113.
1998b. The semantics and pragmatics of Presupposition. Journal of Semantics 15:
239–99.
Bach, Emmon, 1981. On time, tense, and aspect: An essay in English metaphysics. In
P. Cole (ed.), Radical Pragmatics. New York: Academic Press.
Bach, Kent, 1999. The semantics–pragmatics distinction: What it is and why it matters.
In K. Turner (ed.), The Semantics–Pragmatics Interface from Different Points of
View. Oxford: Elsevier, pp. 65–84.
Bain, Alexander, 1877. English Composition and Rhetoric. London: Longmans.
Baker, Carl Lee, 1995. Contrast, discourse prominence, and identification, with special
reference to locally free reflexives in British English. Language 71: 63–101.
Baker, Mark, 1988. Incorporation. Chicago, Ill.: University of Chicago Press.
Bal, Mieke, 1991. On Story-telling. Sonoma, Calif.: Polebridge Press.
Banfield, Ann, 1982. Unspeakable Sentences. Boston: Routledge & Kegan
Paul.
Barker, Chris, 2000. Definite possessives and discourse novelty. Theoretical Linguistics
26: 211–27.
294
References 295
Bartsch, R., 1976. Topik-Fokus-Struktur und Kategoriale Syntax. In V. Ehrich & P. Find
(eds.), Grammatik und Pragmatik. Kronberg: Scriptor Verlag.
Bechtel, William, & Arthur Abrahamsen, 1991. Connectionism and the Mind: An Intro-
duction to Parallel Processing in Networks. Cambridge, Mass.: Blackwell.
Beekman, John, 1970. Propositions and theory relations within a discourse. Notes on
Translation 37: 6–23.
Benson, James, & William Greaves, 1992. Collocation and field of discourse. In
W. Mann & S. Thompson (eds.), Discourse Description: Diverse Linguistic
Analyses of a Fund-raising Text. Amsterdam: John Benjamins.
Bever, Thomas, 1970. The cognitive basis for linguistic structures. In J. R. Hayes (ed.),
Cognition and the Development of Language. New York: John Wiley.
Biber, Douglas, 1988. Variation across Speech and Writing. Cambridge: Cambridge
University Press.
1989. A typology of English texts. Linguistics 27: 3–43.
Biber, Douglas, & Edward Finegan, 1989. Drift and the evolution of English style: A
history of three genres. Language 65: 487–517.
Birner, Betty, 1994. Information status and word order: An analysis of English inversion.
Language 70: 233–59.
1996. Form and Function in English By-phrase Passives. Proceedings of the Chicago
23rd Annual Meeting, Chicago Linguistic Society. Chicago, Ill.: Chicago Linguistic
Society.
Birner, Betty, & Gregory Ward, 1998. Information Status and Non-Canonical Word
Order in English. Amsterdam: John Benjamins.
Black, J. B., & Gordon Bower, 1980. Story understanding as problem solving. Poetics
9: 224–350.
Blakemore, Diane, 1988. The organization of discourse. In F. Newmeyer (ed.),
Linguistics: The Cambridge Survey, Vol. IV. Cambridge: Cambridge University
Press.
1989. Denial and contrast: A relevance theoretic account of BUT. Linguistics &
Philosophy 12: 15–38.
Blass, Regina, 1990. Relevance Relations in Discourse. Cambridge: Cambridge
University Press.
Bolinger, Dwight, 1961. Contrastive accent and contrastive stress. Language 37: 83–96.
1972. Accent is predictable (if you’re a mind-reader). Language 48: 639–44.
1977. Meaning and Form. London: Longman.
Bond, S. J., & John Hayes, 1984. Cues people use to paragraph text. Research in the
Teaching of English 18: 147–67.
Bosch, P., & R. van der Sandt (eds.), Focus: Linguistic, Cognitive, and Computational
Perspectives. Cambridge: Cambridge University Press.
Bower, G. H., J. B. Black, & T. J. Turner, 1979. Scripts in memory for text. Cognitive
Psychology, 2: 177–220.
Braddock, Richard, 1974. The frequency and placement of topic sentences in expository
prose. Research in the Teaching of English 8: 287–302.
Bransford, John, J. Richard Barclay, & Jeffery Franks, 1972. Sentence memory: A
constructive vs interpretive approach. Cognitive Psychology 3: 193–209.
296 References
Brennan, Susan E., 1995. Centering attention in discourse. Language & Cognitive
Processes, 10: 137–67.
Britton, James, Tony Burgess, Nancy Martin, Alex McLeod, & Harold Rosen, 1975.
The Development of Writing Abilities (11–18). London: Macmillan.
Brooks, Cleanth, & Robert Penn Warren, 1958. Modern Rhetoric. New York: Harcourt,
Brace & Company.
Brown, Gillian, & George Yule, 1983. Discourse Analysis. Cambridge: Cambridge
University Press.
Brown, Penelope, & Colin Fraser, 1982. Social markers in speech. In K. Scherer &
H. Giles (eds.), Advances in the Social Psychology of Language. Cambridge:
Cambridge University Press.
Bruder, Gail, & Janyce Wiebe, 1995. Recognizing subjectivity and identifying sub-
jective characters in third-person fictional narrative. In J. Duchan, G. Bruder, &
L. Hewitt (eds.), Deixis in Narrative: A Cognitive Science Approach. Hillsdale,
N.J.: Lawrence Erlbaum Associates.
Büring, Daniel, 1999. Topic. In P. Bosch & R. van der Sandt (eds.), Focus: Linguistic,
Cognitive, and Computational Perspectives. Cambridge: Cambridge University
Press.
Caenepeel, Mimo, 1989. Aspect, temporal ordering and perspective in narrative fiction.
Ph.D. thesis, University of Edinburgh.
1995. Aspect and text structure. Linguistics 33: 213–53.
1997. Putting while in context. Human Communication Research Centre, #RP-85.
Edinburgh: University of Edinburgh.
Caenepeel, Mimo, & Mark Moens, 1994. Temporal structure and discourse structure.
In C. Vet & C. Vetters (eds.), Tense and Aspect in Discourse. Berlin: Mouton de
Gruyter.
Caenepeel, Mimo & Görel Sandström, 1992. A discourse-level approach to the past
perfect in narrative. In M. Aurnague, A. Borillo, M. Borillo, and M. Bras (eds.),
Semantics of Time, Space, Movement and Spatio-temporal Reasoning. Toulouse:
Université Paul Sabatier.
Cairns, William B., 1902. The Forms of Discourse. Boston: Ginn & Co.
Cantrall, William, 1969. On the nature of the reflexive in English. Ph.D. dissertation,
University of Illinois.
Carlson, Greg, 1977. Reference to kinds in English. Ph.D. dissertation, University of
Massachusetts.
Carlson, Greg, & Geoffrey Pelletier, 1995. The Generic Book. Chicago: University of
Chicago Press.
Chafe, Wallace, 1976. Given, contrastiveness, definiteness, subject, topics and point of
view. In Charles Li (ed.), Subject and Topic. New York: Academic Press.
1982. Integration and involvement in speaking, writing, and oral literature. In
D. Tannen (ed.), Spoken and Written Language: Exploring Orality and Literacy.
Norwood, N.J.: Ablex Publishing Company.
1987. Cognitive constraints on information flow. In R. Tomlin (ed.), Coherence and
Grounding in Discourse. Amsterdam: John Benjamins.
1994. Discourse, Consciousness, and Time. Chicago: University of Chicago Press.
References 297
Chafe, Wallace, & Joanna Nichols (eds.), 1986. Evidentiality: The Linguistic Coding of
Epistemology. Norwood, N.J.: Ablex Publishing Corporation.
Chierchia, Gennaro, 1992. Anaphora and dynamic binding. Linguistics & Philosophy
15: 111–83.
Chomsky, Noam, 1970. Remarks on nominalization. In R. Jacobs & P. Rosenbaum
(eds.), Readings in English Transformational Grammar. Waltham, Mass.: Ginn &
Co.
1971. Deep structure, surface structure and semantic interpretation. In D. Steinberg &
L. Jakobovits (eds.), Semantics: An Inter-Disciplinary Reader. Cambridge:
1981. Lectures on Government and Binding. Dordrecht: Foris.
Christiansen, Francis. 1965. A generative rhetoric of the paragraph. College Composition
and Communication 15: 144–57.
Chu, Chauncy, 1983. A Reference Grammar of Mandarin for English Speakers. New
York: Peter Lang.
Cinque, G. 1999. Adverbs and Functional Heads: A Cross-linguistic Perspective.
Oxford: Oxford University Press.
Clancy, Patricia, 1980. Referential choice in English and Japanese narrative discourse.
In W. Chafe (ed.), The Pear Stories. Norwood, N.J.: Ablex.
Clark, Andy, 1989. Microcognition. Cambridge, Mass.: MIT Press.
Clark, Herbert, 1973. Space, time, semantics, and the child. In T. Moore (ed.), Cognitive
Development and the Acquisition of Language. New York: Academic Press.
1996. Using Language. Cambridge: Cambridge University Press.
Cohan, Jocelyn, 2001. Reconsidering identificational focus. In M. Kim & U. Strauss
(eds.), Proceedings of the New England Linguistic Society 31. Amherst, Mass.:
GLSA, University of Massachusetts.
Cohn, Dorritt, 1978. Transparent Minds. Princeton, N.J.: Princeton University Press.
Comrie, Bernard, 1976. Aspect. Cambridge: Cambridge University Press.
1983. Switch reference in Huichol: A typological study. In J. Haiman & P. Munro
(eds.), Switch Reference and Universal Grammar. Amsterdam: John Benjamins.
1985. Tense. Cambridge: Cambridge University Press.
Connors, Robert J., 1997. Composition–Rhetoric. Pittsburgh, Pa.: University of
Pittsburgh Press.
Corbett, Edward P., 1965. Classical Rhetoric for the Modern Student. New York: Oxford
University Press.
Couper-Kuhlen, Elizabeth, 1989. Foregrounding and temporal relations in narrative
discourse. In A. Schopf (ed.), Essays on Tensing in English, Vol. II. Tübingen:
Niemeyer.
Creider, Chet, 1979. On the explanation of transformations. In T. Givón (ed.), Discourse
and Syntax, Syntax & Semantics, 12. New York: Academic Press.
Croft, William, 1991. Syntactic Categories and Grammatical Relations: The Cognitive
Organization of Information. Chicago: University of Chicago Press.
Crothers, Edward, 1979. Paragraph Structure Inference. Norwood, N.J.: Ablex.
Culicover, Peter, 1997. Principles and Parameters: An Introduction to Syntactic Theory.
Oxford: Oxford University Press.
298 References
Dahl, Oesten, 1974. Topic and comment structure revisited. In O. Dahl (ed.), Topic &
Comment, Contextual Boundedness and Focus, Papers in Text Linguistics, 6.
Hamburg: Buske.
Danasio, Antonio, 2002. Remembering when. Scientific American 287: 66–73.
Daneš, Frederik, 1974. Functional sentence perspective and the organization of the text.
In F. Daneš (ed.), Papers on Functional Sentence Perspective. Prague: Academia.
Davison, Alice, 1984. Syntactic markedness and the notion of sentence topic. Language
60: 797–846.
Davison, Alice, & Richard Lutz, 1985. Measuring syntactic complexity relative to
discourse. In D. Dowty (ed.), Natural Language Processing: Psycholinguistic,
Computational and Theoretical Perspectives. Cambridge: Cambridge University
Press.
de Beaugrande, Robert-Alain, & Wolfgang Dressler, 1981. Introduction to Text Linguis-
tics. London: Longman.
Delin, Judy, 1995. Presupposition and shared knowledge in It-clefts. Language and
Cognitive Processes 10: 97–120.
Delin, Judy, & Jon Oberlander, 1995. Syntactic constraints on discourse structure: the
case of It-clefts. Linguistics 33: 465–500.
Dendale, Patrick, & Liliane Tasmowski (eds.), 2001. Evidentiality and related notions:
Special issue. Journal of Pragmatics 33: 339–48.
de Swart, Henriette, 1993. Adverbs of Quantification: A Generalized Quantifier
Approach. New York: Garland Publications.
1998. Introduction to Natural Language Semantics. Stanford, Calif.: Center for the
Study of Language and Information Publications.
1999. Position and meaning: Time adverbials in context. In P. Bosch & R. van der Sandt
(eds.), Focus: Linguistics, Cognitive, and Computational Perspectives. Cambridge:
van Dijk, Teun, 1972. Some Aspects of Text Grammars. The Hague: Mouton.
1977. Text and Context. London: Longman.
1982. Episodes as units of discourse analysis. In D. Tannen (ed.), Georgetown Uni-
versity Roundtable on Languages & Linguistics. Washington, D.C.: Georgetown
University Press.
van Dijk, Teun, & Walter Kintsch, 1983. Strategies of Discourse Comprehension. New
York: Academic Press.
Dowty, David, 1979. Word Meaning and Montague Grammar. Dordrecht: Kluwer.
1986. The effects of aspectual class on the temporal structure of discourse: Semantics
or pragmatics? Linguistics & Philosophy 9: 37–62.
1991. Thematic proto-roles and argument selection. Language 67: 547–619.
Dowty, David, Robert Wall, & Stanley Peters, 1981. Introduction to Montague Seman-
tics. Dordrecht: Reidel.
Duchan, Judith, Gail Bruder, & Lynne Hewitt (eds.), 1995. Deixis in Narrative: A
Cognitive Science Perspective. Hillsdale, N.J.: Erlbaum.
Dundes, Alan, 1975. Analytic Essays in Folklare. The Hague: Mouton.
Eggins, Suzanne, 1994. Introduction to Systemic Functional Linguistics. London: Pinter
Publishers.
References 299
Enç, Murvet, 1986. Towards a referential analysis of temporal expressions. Linguistics &
1987. Anchoring conditions for tense. Linguistic Inquiry 18: 633–57.
1991. Tense and modality. In S. Lappin (ed.), The Handbook of Contemporary
Semantic Theory. Oxford: Blackwell.
Erbaugh, Mary, 1986. Taking stock: The development of Chinese noun classifiers. In
C. Craig (ed.), Noun Classes and Categorization. Philadelphia: Benjamins.
Erteschik-Shir, Nomi, 1979. Discourse constraints on dative movement. In T. Givón
(ed.), Discourse and Syntax, Syntax and Semantics, 12. New York: Academic
Press.
1997. The Dynamics of Focus Structure. Cambridge: Cambridge University Press.
Erteschik-Shir, Nomi, & Sheldon Lappin, 1979. Dominance and the functional expla-
nation of island phenomena. Theoretical Linguistics 6: 41–85.
Erteschik-Shir, Nomi, & Tova Rapoport, in press. Bare aspect: A theory of syntactic
projection. In J. Guéron and J. Lacarme (eds.), The Syntax of Time. Cambridge,
Mass.: MIT Press.
Fabricius-Hansen, Cathrine, & Bergljot Behrens, 2000. Elaboration and related discourse
relations viewed from an interlingual perspective. In Proceedings of the Third
Workshop on Text Structure. Austin, Tex.: Department of Linguistics, University
of Texas.
Faigley, Lester, & Paul Meyer, 1983. Rhetorical theory and readers’ classification of
text types. Text 3: 305–25.
Fauconnier, Gilles, 1985. Mental Spaces: Aspects of Meaning Construction in Natural
Language. Cambridge, Mass.: MIT Press. Reprinted 1994, Cambridge: Cambridge
University Press.
Fauconnier, Gilles, & Eve Sweetser (eds.), 1996. Spaces, World, and Grammar. Chicago,
Ill.: University of Chicago Press.
Ferro, Lisa, 1993. On “self ” as a focus marker. In Michael Bernstein (ed.), Proceedings of
the Ninth Eastern States Conference on Linguistics. Ithaca, N.Y.: Cornell University
Department of Modern Languages and Linguistics.
Fillmore, Charles, 1975. Santa Cruz Lectures on Deixis. Bloomington, Ind.: Indiana
University Linguistics Club.
von Fintel, Kai, 1994. Restrictions on quantifier domains. Ph.D. dissertation, Mas-
sachusetts Institute of Technology.
Firbas, Jan, 1964. On defining the theme in functional sentence analysis. Travaux de
Circle Linguistique de Prague 1: 267–80.
1974. Some aspects of the Czechoslovak approach to functional sentence perspective.
In F. Daneš (ed.), Papers on Functional Sentence Perspective. Prague: Academia.
1992. Functional Sentence Perspective in Written and Spoken Communication.
Cambridge: Cambridge University Press.
Flashner, Vanessa, 1987. The grammatical marking of theme in oral Polish narra-
tive. In R. Tomlin (ed.), Coherence and Grounding in Discourse. Amsterdam:
Benjamins.
Fleischman, Suzanne, 1991. Verb tense and point of view in narrative. In S. Fleischman &
L. Waugh (eds.), Discourse-Pragmatics and the Verb. London: Routledge.
300 References
Forster, Kenneth, 1979. Levels of processing and the structure of the language processor.
In W. E. Cooper & E. C. T. Walker (eds.), Sentence Processing: Psycholinguistic
Studies Presented to Merrill Garrett. Hillsdale, N.J.: Lawrence Erlbaum.
Fox, Barbara, 1987a. Anaphora in popular English written narratives. In R. Tomlin (ed.),
Coherence and Grounding in Discourse. Amsterdam: Benjamins.
1987b. Discourse Structure and Anaphora in Written and Conversational English.
Frank, Anette, & Hans Kamp, 1997. On context dependence in modal constructions. In
A. Lawson (ed.), SALT VII. Ithaca, N.Y.: Cornell University.
Frantz, Donald, 1966. Person indexing in Blackfoot. International Journal of American
Linguistics, 32: 50–58.
Freksa, C., 1992. Using orientation information for qualitative spatial reasoning. In
M. Aurnague et al. (eds.), Semantics of Time, Space, Movement and Spatio-temporal
Reasoning. Toulouse: Université Paul Sabatier.
Friedman, William, 1993. About Time. Cambridge, Mass.: MIT Press.
Fries, Peter, 1983. On the status of theme in English: Arguments from discourse. In
J. Petöfi & E. Sözer (eds.), Micro and Macro Connexity in Texts. Hamburg: Helmut
Buske.
Fuller, Daniel, 1959. The Inductive Method of Bible Study, 3rd edition. Pasadena: Fuller
Theological Seminary.
Galbraith, Mary, 1995. Deictic shift theory and the poetics of involvement in narrative. In
J. Duchan, G. Bruder, & L. Hewitt (eds.), Deixis in Narrative: A Cognitive Science
Approach. Hillsdale, N.J.: Lawrence Erlbaum Associates.
Garnham, Alan, 1983. What’s wrong with story grammars? Cognition 15: 145–54.
1985. Psycholinguistics. London: Methuen.
1987. Mental Models as Representation of Discourse and Text. Chichester: Ellis
Horwood.
Garnham, Alan, & Jane Oakhill, 1989. The everyday use of anaphoric expressions:
Implications for the “mental models” theory of text comprehension. In N. E.
Sharkey (ed.), Modelling Cognition: An Annual Review of Cognitive Science, Vol. I.
Norwood, N.J.: Ablex.
Garrod, Simon, & Anthony Sanford, 1985. On the real-time character of interpretation
during reading. Language & Cognitive Processes 1: 43–61.
1994. Resolving sentences in a discourse context. In M. A. Gernsbacher (ed.), Hand-
book of Psycholinguistics. New York: Academic Press.
Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, & Ivan Sag, 1985. Generalized Phrase
Structure Grammar. Oxford: Blackwell.
Gelman, Rachel, 1990. First principles organize attention to and learning about rele-
vant data: Number and the animate/inanimate distinction. Cognitive Science 14:
79–106.
Genette, Gérard, 1980. Narrative Discourse: An Essay in Method. Ithaca, N.Y.: Cornell
University Press.
Genung, John F., 1900. The Working Principles of Rhetoric. Boston: Ginn & Co.
Giora, Rachel, 1983. Functional sentence perspective. In J. Petöfi & E. Sözer (eds.),
Micro and Macro Connexity in Texts. Hamburg: Helmut Buske.
References 301
Giorgi, Alessandra, & Fabio Pianesi, 1991. Towards a syntax of temporal representations.
Probus 2: 187–213.
1997. Tense and Aspect. Oxford: Oxford University Press.
Givón, Talmy, 1979. From discourse to syntax: Grammar as a processing strategy. In
T. Givón (ed.), Discourse and Syntax, Syntax and Semantics, 12. New York:
Academic Press.
1983. Topic continuity in discourse: An introduction. In T. Givón (ed.), Topic Conti-
nuity in Discourse: A Quantitative Cross-Language Study. Amsterdam: Benjamins.
1993. The pragmatics of de-transitive voice: Functional and typological aspects of
inversion. In T. Givón (ed.), Voice and Inversion. Amsterdam: Benjamins.
Glenberg, A. M., M. Mayer, & K. Linden, 1987. Mental models contribute to fore-
grounding during text interpretation. Journal of Memory & Language 26: 69–83.
Goddard, Ives, 1990. Aspects of the topic structure of Fox narratives: Proximate shifts
and the use of overt and inflectional NPs. International Journal of American Lin-
guistics 56: 317–40.
Goldberg, Adele, 1995. Constructions. Chicago, Ill.: University of Chicago Press.
Gordon, Peter, & Davina Chan, 1995. Pronouns, passives, and discourse coherence.
Journal of Memory & Language 34: 216–31.
Gough, Philip, 1965. Grammatical transformations and speed of understanding. Journal
of Verbal Learning & Verbal Behavior 5: 107–11.
Goutsos, Dionysis, 1997. Modeling Discourse Topic: Sequential Relations and Strate-
gies in Expository Texts. Norwood, N.J.: Ablex.
Graesser, Arthur, 1981. Prose Comprehension Beyond the Word. New York: Springer.
Green, Georgia, 1974. Semantics and Syntactic Regularity. Bloomington, Ind.: Indiana
University Press.
Grice, H. P., 1975. Logic and conversation. In P. Cole & J. Morgan (eds.), Speech Acts,
Syntax and Semantics, 3. New York: Academic Press.
1989. Studies in the Way of Words. Cambridge, Mass.: Harvard University Press.
Grimes, Joseph, 1975. The Thread of Discourse. The Hague: Mouton.
Grimshaw, Jane, 1990. Argument Structure. Cambridge, Mass.: MIT Press.
Groenendijk, Jeroen, & Martin Stokhof, 1991. Dynamic predicate logic. Linguistics &
Gropen, J., S. Pinker, S. Hollander, R. Goldberg, & R. Wilson, 1989. The learnability
and acquisition of the dative alternation in English. Language 65: 203–55.
Grosz, Barbara, 1977. The representation and use of focus in dialogue understanding.
Tech Note, 15. Menlo Park, Calif.: Artificial Intelligence Center, SRI International.
Grosz, Barbara, & Candace Sidner, 1986. Attention, intensions, and the structure of
discourse. Computational Linguistics 12: 175–204.
Gruber, Jeffrey, 1965. Studies in Lexical Relations. Ph.D. dissertation, MIT. Reprinted
Bloomington, Ind.: Indiana University Linguistics Club.
Gundel, Jeannette, 1974. The role of topic and comment in linguistic theory. Ph.D. disser-
tation, University of Texas. Reprinted 1977, Bloomington, Ind.: Indiana University
Linguistics Club.
1998. Centering theory and a givenness hierarchy. In M. Walker, A. Joshi, & E. Prince
(eds.), Centering Theory in Discourse. Oxford: Clarendon Press.
302 References
1999. On different kinds of focus. In P. Bosch & R. van der Sandt (eds.), Focus:
Linguistic, Cognitive, and Computational Perspectives. Cambridge: Cambridge
University Press.
Gundel, Jeannette, Nancy Hedberg, & Ron Zacharski, 1993. Cognitive status and the
form of referring expressions in discourse. Language 69: 274–307.
Gussenhoven, Carlos, 1983. Focus, mode and the nucleus. Journal of Linguistics 19:
377–417.
Gutwinski, Waldemar, 1976. Cohesion in Literary Texts. The Hague: Mouton.
Hagège, Claude, 1974. Les pronoms logophoriques. Bulletin de la Société de Linguis-
tique de Paris 69: 297–310.
Haiman, John, & Pamela Munro (eds.), Switch Reference and Universal Grammar.
Amsterdam: John Benjamins.
Hajičová, Eva, 1971. Some remarks on presupposition. Prague Bulletin of Mathematical
Linguistics 17: 11–23.
Hajičová, Eva, & Petr Sgall, 1987. The ordering principle. Journal of Pragmatics 11:
435–54.
Hajičová, Eva, M. Cervenka, O. Leška, & Petr Sgall (eds.), 1995. Travaux du Cercle
Linguistique de Prague, nouvelle série, Vol. I. Amsterdam: John Benjamins.
Hajičová, Eva, Barbara Partee, & Petr Sgall, 1998. Topic-Focus Articulation, Tripartite
Structures, and Semantic Content. Dordrecht: Kluwer Academic Publishers.
Hale, Kenneth, & Samuel Jay Keyser, 1993. On argument stucture and the lexical expres-
sion of syntactic relations. In K. Hale & S. J. Kayser (eds.), The View from Building
20: Essays in Honor of Sylvan Bromberger. Cambridge, Mass.: MIT Press.
Halliday, Michael A. K., 1967. Notes on transitivity and theme in English, Part II. Journal
of Linguistics 3: 199–244.
1992. Some lexicogrammatical features of the Aero Population Growth Text. In
W. Mann & S. Thompson (eds.), Discourse Description: Diverse Linguistic
Analyses of a Fund-raising Text. Amsterdam: John Benjamins.
Halliday, Michael A. K., & Ruqaiya Hasan, 1976. Cohesion in English. London:
Longman.
Hamburger, Käte, 1973. The Logic of Literature. Trans. Marilynn Rose. Bloomington,
Ind.: Indiana University Press.
Hannay, Mike, 1991. Pragmatic function assignment and word order variation in a
functional grammar of English. Journal of Pragmatics 16: 131–55.
Harris, Zellig, 1982. A Grammar of English on Mathematical Principles. New York:
Methuen.
Haviland, Susan, & Herbert Clark, 1974. What’s new? Acquiring new information as
a process in comprehension. Journal of Verbal Learning & Verbal Behavior 13:
512–21.
Hawkins, John, 1994. A Performance Theory of Order and Constituency. Cambridge:
Hearst, Marti, 1997. TextTiling: Segmenting text into multi-paragraph subtopic passages.
Computational Linguistics 23: 33–64.
Hedberg, Nancy, 1990. Discourse pragmatics and cleft sentences in English. Ph.D.
dissertation, University of Minnesota.
References 303
Heim, Irene, 1982. The semantics of definite and indefinite Nounphrases. Ph.D. disser-
tation, University of Massachusetts.
Herbst, Peter, 1956. The nature of facts. In A. Flew (ed.), Essays in Conceptual Analysis.
New York: MacMillan.
Heycock, Caroline, & Anthony Kroch. 1999. Pseudo-cleft connectedness: Implications
for the LF interface level. Linguistic Inquiry 30: 365–98.
Heyrich, Wolfgang, Fritz Neubauer, Janos Petöfi, & Erich Sözer (eds.), 1989. Connexity
and Coherence. Berlin: De Gruyter.
Hinds, John, 1977. Paragraph structure and pronominalization. Papers in Linguistics 10:
77–99.
1979. Organizational patterns in discourse. In T. Givón (ed.), Discourse and Syntax,
Syntax & Semantics, 12. New York: Academic Press.
Hinrichs, Erhard, 1986. Temporal anaphora in discourses of English. Linguistics &
Hirose, Yukio, 2000. Public and private self as two aspects of the speaker: A contrastive
study of Japanese and English. Journal of Pragmatics 32: 1623–56.
Hirschberg, Julia, 1991. A Theory of Scalar Implicature. New York: Garland Press.
Hobbs, Jerry, 1985. On the Coherence and Structure of Discourse. Report No. CSLI-
85–37. Stanford, Calif.: Center for the Study of Language and Information.
Hockett, Charles, 1958. A Course in Modern Linguistics. New York: Macmillan.
Hoey, Michael, 1991. Patterns of Lexis in Text. Oxford: Oxford University Press.
Hopper, Paul, 1979. Aspect and foregrounding in discourse. In T. Givón (ed.), Discourse
and Syntax, Syntax & Semantics, 12. New York: Academic Press.
Horn, Laurence, 1988. Pragmatic theory. In F. Newmeyer (ed.), Linguistic Theory:
Foundations. Cambridge: Cambridge University Press.
1991. Given as new: When redundant information isn’t. Journal of Pragmatics 15:
305–28.
Hornstein, Norbert, 1990. As Time Goes By. Cambridge, Mass.: MIT Press.
Hovy, Eduard, 1993. Automated discourse generation using discourse structure relations.
Artificial Intelligence 63: 341–85.
Hunston, S., & G. Thompson (eds.), 2000. Evaluation in Text. Oxford: Oxford University
Press.
Jackendoff, Ray, 1972. Semantic Interpretation in Generative Grammar. Cambridge,
Mass.: MIT Press.
1987. The status of thematic relations in linguistic theory. Linguistic Inquiry 18:
369–411.
1990. Semantic Structures. Cambridge, Mass. MIT Press.
Jacobsen, William, 1967. Switch-reference in Hokan-Coahuiltecan. In D. Hymes &
W. Biddle (eds.), Studies in Southwestern Ethnolinguistics. The Hague: Mouton.
Jakobson, Roman, 1957. Shifters, verbal categories, and the Russian verb. In L. Waugh &
M. Halle (eds.), Russian and Slavic Grammar: Studies 1931–1981. Berlin: Mouton,
1984.
Jesperson, Otto, 1940. A Modern English Grammar on Historical Principles. London:
George Allen & Unwin.
1924/1965. The Philosophy of Grammar. New York: W. W. Norton.
304 References
Johnson-Laird, Philip, 1983. Mental Models: Toward a Cognitive Science of Language,

Inference, and Consciousness. Cambridge: Cambridge University Press.
1988. The Computer and the Mind. Cambridge, Mass.: Harvard University Press.
1989. Mental models. In M. Posner (ed.), Foundations of Cognitive Science.
Cambridge, Mass.: MIT Press.
Johnson-Laird, Philip, & Rosemary Stevenson, 1970. Memory for syntax. Nature 227:
412.
Joshi, Aravind, & Scott Weinstein, 1981. Control of inference: Role of some aspects of
discourse structure – centering. Proceedings of the International Joint Conference
on Artificial Intelligence, Vancouver.
Kameyama, Megumi, 1985. Zero anaphora: The case of Japanese. Ph.D. dissertation,
Stanford University.
1998. Intrasentential centering: A case study. In M. Walker, A. Joshi, & E. Prince
(eds.), Centering Theory in Discourse. Oxford: Oxford University Press.
Kamio, Aki, 1994. The theory of territory of information: The case of Japanese. Journal
of Pragmatics 27: 67–100.
Kamp, Hans, 1981. A theory of truth and semantic interpretation. In J. Groenendijk,
T. Janssen, & M. Stockhof (eds.), Formal Methods in the Study of Language.
Amsterdam: Mathematisch Centrum Tracts. Also in J. Groenendijk et al. (eds.),
Truth, Interpretation, and Information. Dordrecht: Foris.
1985. Context, thought and communication. Proceedings of the Aristotelian Society
85: 239–61.
Kamp, Hans, & Uwe Reyle, 1993. From Discourse to Logic. Dordrecht: Kluwer
Academic Publishers.
Kamp, Hans, & Christian Rohrer, 1983. Tense in texts. In R. Bauerle, R. Schwarze, &
A. von Stechow (eds.), Meaning, Use and Interpretation of Language. Berlin: de
Gruyter.
Karttunen, Lauri, 1976. Discourse referents. In J. McCawley (ed.), Notes from the Lin-
guistic Underground, Syntax and Semantics, 7. New York: Academic Press.
Keenan, Elinor, & Bambi Schieffelin, 1976. Topic as a discourse notion. In Charles Li
(ed.), Subject and Topic. New York: Academic Press.
Kennedy, George, 1991. Aristotle on Rhetoric. Translation and Introduction. Oxford:
Oxford University Press.
Kinneavy, James, 1971. A Theory of Discourse. Englewood Cliffs, N.J.: Prentice-Hall.
Kintsch, Walter, 1988. The role of knowledge in discourse comprehension: A
construction–integration model. Psychological Review 95: 163–82.
Kiss, Katalin, 1986. The order and scope of operators in the Hungarian sentence. In
W. Abraham & S. de Meij (eds.), Topic, Focus and Configurationality. Amsterdam:
John Benjamins.
1998. Identificational focus versus information focus. Language 74: 245–73.
Klaiman, M. H., 1991. Grammatical Voice. Cambridge: Cambridge University Press.
Knott, Alistair, 1996. A data-driven methodology for motivating a set of coherence
relations. Ph.D. dissertation, University of Edinburgh.
Koen, Frank, Alton Becker, & Richard Young, 1966. The psychological reality of the
paragraph. In E. Zale (ed.), Proceedings of the Conference on Language and Lan-
guage Behavior. New York: Appleton.
References 305
Kratzer, Angelika, 1981. The notional category of modality. In H. Eikmeyer & H. Rieser
(eds.), Words, Worlds, and Contexts. Berlin: De Gruyter.
Kratzer, Angelika, 1995. Stage-level and individual-level predicates. In G. Carlson &
F. J. Pelletier (eds.), The Generic Book. Chicago, Ill.: University of Chicago
Press.
Krifka, Manfred, 1989. Nominal reference, temporal constitution and quantification in
event semantics. In R. Bartsch et al. (eds.), Semantics and Contextual Expressions.
Dordrecht: Foris.
1991. A compositional semantics for multiple Focus constructions. In Proceedings
from Semantics and Linguistic Theory 1, Working Papers in Linguistics, 10. Ithaca,
N.Y.: Cornell University Press.
Krifka, Manfred, Francis J. Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro
Chierchia, & Godehard Link, 1995. Genericity: An introduction. In G. Carlson &
F. J. Pelletier (eds.), The Generic Book. Chicago, Ill.: University of Chicago
Press.
Kuno, Susumo, 1972. Functional sentence perspective. Linguistic Inquiry 3: 161–95.
1976. Subject, theme, and speaker’s empathy: A re-examination of relativization
phenomena. In C. Li (ed.), Subject and Topic. New York: Academic Press.
Kuno, Susumo, 1987. Functional Syntax: Anaphora, Discourse, and Empathy. Chicago,
Ill.: University of Chicago Press.
van Kuppevelt, Jan, 1995. Disourse structure, topicality, and questioning. Journal of
1996. Directionality in discourse: Prominence difference in subordination relations.
Journal of Semantics 13: 363–95.
Kuroda, Yuki, 1973. Where epistemology, grammar and style meet: A case study from
Japanese. In S. Anderson & P. Kiparsky (eds.), A Festschrift for Morris Halle. New
York: Holt, Rinehart Winston.
Labov, William, & Joshua Waletzky, 1966. Narrative analysis: Oral versions of personal
experience. In Essays on the Verbal and Visual Arts: Proceedings of the 1996
Annual Meeting, American Ethnological Society. Seattle: University of Washington
Press.
Lambrecht, Knud, 1986. Pragmatically motivated syntax: Presentational cleft construc-
tions in spoken French. In Proceedings of the 13th Annual Meeting, Chicago
Linguistic Society. Chicago, Ill.: Chicago Linguistic Society.
1994. Information Structure and Sentence Form. Cambridge: Cambridge University
Press.
Larson, Richard, 1984. Classifying discourse. In R. J. Connors, L. S. Ede, & A. Lunsford
(eds.), Essays on Classical Rhetoric and Modern Discourse. Carbondale: Southern
Illinois Press.
Lascarides, Alex, & Nicholas Asher, 1993. Temporal interpretation, discourse relations,
and commonsense entailment. Linguistics & Philosophy 16: 437–93.
Lee, Gregory, 1971. Subjects and Agents II. Working Papers in Linguistics, 7. Columbus,
Ohio: Department of Linguistics, Ohio State University.
Leška, Oldrich, 1995. Prague school teaching of the classical period and beyond. In
E. Hajičová, M. Cervenka, O. Leška, & P. Sgall (eds.), Travaux du Cercle
Linguistique de Prague, nouvelle série, Vol. I. Amsterdam: Benjamins.
306 References
Leslie, A. M., 1994. ToMM, ToBy, and Agency: Core architecture and domain speci-
ficity. In L. Hirschfield & S. Gelman (eds.), Mapping the Mind: Domain Specificity
in Cognition and Culture. Cambridge: Cambridge University Press.
Levin, Beth, 1993. English Verb Classes and Alternations. Chicago, Ill.: University of
Chicago Press.
Levinson, Stephen, 1979/1992: Activity types in language. Linguistics 17: 356–99.
2000. Presumptive Meanings. Cambridge, Mass.: MIT Press.
Lewis, David, 1979. Scorekeeping in a language game. In R. Bauerle, U. Egli, & A. von
Stchow (eds.), Semantics from Different Points of View. Berlin: De Gruyter.
Li, Charles, & Sandra Thompson, 1976. Subject and topic: A new typology of language.
In C. Li (ed.), Subject and Topic. New York: Academic Press.
Linde, Charlotte, 1979. Focus of attention and choice of pronouns in discourse. In
T. Givón (ed.), Discourse and Syntax, Syntax & Semantics, 12. New York:
Academic Press.
1993. Life Stories: The Creation of Coherence. Oxford: Oxford University Press.
Litman, David, 1985. Plan recognition and discourse analysis: An integrated approach
for understanding dialogues. Ph.D. dissertation, University of Rochester.
Longacre, Robert, 1968. Discourse, Paragraph, and Sentence Structure in Selected
Philippine Languages. Santa Ana, Calif.: Summer Institute of Linguistics.
1979. The paragraph as a grammatical unit. In. T. Givón (ed.), Discourse and Syntax,
Syntax and Semantics, 12. New York: Academic Press.
1983/1996. The Grammar of Discourse. New York: Plenum Press.
1992. The discourse strategy of an appeals letter. In W. Mann & S. Thompson
(eds.), Discourse Description: Diverse Linguistic Analyses of a Fund-raising Text.
Loriot, James, & B. Hollenbach, 1970. Shipibo paragraph structure. Foundations of
Language 6: 43–66.
Lunsford, Andrea, & Lisa Ede, 1984. On distinctions between classical and modern
rhetoric. In R. J. Connors, L. Ede, & A. Lunsford (eds.), Essays on Classical
Rhetoric and Modern Discourse. Carbondale: Southern Illinois University
Press.
Lyons, John, 1977. Semantics. Cambridge: Cambridge University Press.
1982. Deixis and subjectivity: Loquor, ergo sum? In R. J. Jarvella & W. Klein (eds.),
Speech, Place, and Action. New York: John Wiley & Sons.
Mann, William, & Sandra Thompson, 1987. Rhetorical Structure Theory: A Theory
of Text Organization. ISI Reprint Series, ISI /RS-87–90. Marina Del Rey, Calif.:
Information Sciences Institute.
1992. Rhetorical Structure theory and text analysis. In W. Mann & S. Thompson
(eds.), Discourse Description: Diverse Linguistic Analyses of a Fund-raising Text.
2000. Two Views of Rhetorical Structure Theory. Lyon, France: Society for Text and
Discourse.
Markels, Robin Bell, 1984. A New Perspective on Cohesion in Expository Paragraphs.
Carbondale: Southern Illinois University Press.
Marr, David, 1982. Vision. San Francisco, Calif.: Freeman.
References 307
Marslen-Wilson, William, E. Levy, & Lorraine Tyler, 1982. Producing interpretable

discourse: The establishment and maintenance of reference. In R. J. Jarvella &
W. Klein (eds.), Speech, Place and Action. Chichester: Wiley.
Martin, J. R., 1992. English Text: System and Structure. Amsterdam: Benjamins.
Mathesius, Vilém, 1928/1964. On linguistic characterology with illustrations from mod-
ern English. Reprinted in J. Vachek (ed.), A Prague School Reader in Linguistics.
Bloomington, Ind.: Indiana University Press.
Matthiessen, Christian, & Sandra Thompson, 1988. The structure of discourse and
“subordination.” In J. Haiman & S. Thompson (eds.), Clause Combining in Gram-
mar and Discourse. Amsterdam: John Benjamins.
McArthur, Tom (ed.), 1992. The Oxford Companion to the English Language. Oxford:
Oxford University Press.
McClelland, James, & David Rumelhart, 1986. PDP: Explorations in the Microstructure
of Cognition, Vol. II. Cambridge, Mass.: MIT Press.
McNally, Louise, 1998. On recent formal analyses of topic. In J. Ginzburg,
Z. Khasidashvili, C. Vogel, J. Lévy, & E. Vallduvı́ (eds.), The Tbilisi Symposium
on Logic, Language, and Computation: Selected Papers. Stanford, Calif.: Center
for the Study of Language and Information.
Michaelis, Laura, 1994. The ambiguity of the English present perfect. Journal of
Miller, George, Eugene Galanter, & Karl Pribram, 1960. Plans and the Structure of
Behavior. New York: Holt, Rinehart & Winston.
Miller, Philip, 2001. Discourse constraints on (non)-extraposition from subject in
English. Linguistics 39: 683–701.
Milner, Jean-Claude, 1973. De la syntaxe à l’interprétation: Quantités, Insultes, Excla-
mations. Paris: Seuil.
Milsark, Gary, 1977. Toward an explanation of certain peculiarities of the existential
construction in English. Linguistic Analysis 3: 1–30.
Mitchell, Jonathan, 1986. The formal semantics of point of view. Ph.D. dissertation,
The University of Massachusetts.
Moens, Mark, 1987. Tense, aspect, and temporal reference. Ph.D. thesis, Centre for
Cognitive Science, University of Edinburgh.
Moens, Mark, & Mark Steedman. 1987. Temporal ontology in natural language. Pro-
ceedings of the 25th Annual Conference of the Association for Computational
Linguistics, Stanford, California.
Morgan, Jerry L., 1975. Some remarks on the nature of sentences. In Papers from the
Parasession on Functionalism, Chicago Linguistic Society. Chicago, Ill.: Chicago
Linguistic Society.
Morris, Jane, & Graeme Hirst, 1991. Lexical cohesion computed by thesaural re-
lations as an indicator of the structure of text. Computational Linguistics 17:
21–48.
Moyne, John, 1971. Reflexive and Emphatic. Language 47: 141–62.
Neisser, Ulric, 1967. Cognitive Psychology. New York: Appleton Century Croft.
Nuyts, Jan, 1993. Modality and the layered representation of conceptual and linguistic
structures. Linguistics 31: 933–69.
308 References
Oakhill, Jane, Alan Garnham, & Willem Vonk, 1989. The online construction of
discourse models. Language & Cognitive Processes 4: SI 263–86.
Ogihara, Toshiyuki, 1996. Tense, Attitude, and Scope. Dordrecht: Kluwer.
Olman, Lynda, 1998. Evidence for iconicity: The instance relation in informational
exposition. MA thesis, University of Texas at Austin.
Palacas, Arthur, 1993. Attribution semantics: Linguistic worlds and point of view.
Discourse Processes 16: 239–77.
Palmer, Frank, 1983. Semantic explanations for the syntax of English modals. In
F. Heny & B. Richards (eds.), Linguistic Categories: Auxiliaries and Related
Puzzles, Vol. II. Dordrecht: Reidel.
1986. Mood and Modality. Cambridge: Cambridge University Press.
Parkes, M. B., 1993. Pause and Effect. Berkeley, Calif.: University of California
Press.
Parsons, Terence, 1990. Events in the Semantics of English. Cambridge, Mass.: MIT
Press.
Partee, Barbara, 1973. The syntax and semantics of quotation. In S. Anderson &
P. Kiparsky (eds.), A Festchrift for Morris Halle. New York: Holt, Rinehart &
Winston.
1984. Nominal and temporal anaphora. Linguistics & Philosophy 7: 243–86.
1987. Nounphrase interpretation and type-shifting principles. In J. Groenenjik, D. de
Jongh, & M. Stockhof (eds.), Studies in Discourse Representation theory and the
Theory of Generalized Quantifiers. Dordrecht: Foris.
1991. Topic, focus, and quantification. In S. Moore & A. Wyner (eds.), Proceedings
of SALT I. Ithaca, N.Y.: Department of Linguistics, Cornell University.
Penhallurick, John, 1984. Full-verb inversion in English. Australian Journal of
Peterson, Philip, 1997. Fact, Proposition, Event. Dordrecht: Kluwer Academic Pub-
lishers.
Petöfi, J. S., 1978. A few comments on the methodology of text theoretical research.
Journal of Pragmatics 2: 365–72.
Pierrehumbert, Janet, 1980. The phonology and phonetics of English intonation. Ph.D.
dissertation, MIT.
Polanyi, Livia, 1988. A formal model of the structure of discourse structure and
discourse. Journal of Pragmatics 12: 601–38.
Polinsky, Maria, 1996. Situation perspective: on the relations of thematic roles, discourse
categories, and grammatical relations to figure and ground. In A. Goldberg (ed.),
Conceptual Structure, Discourse, and Language. Stanford, Calif.: Center for the
Study of Language and Information Publications.
Porter, Bruce, et al., 1988. AI Research in the Context of a Multifunctional Knowledge
Base. AI Laboratory Report, 88–88. Austin, Tex.: University of Texas.
Portner, Paul, & Katsuhiko Yabushita, 1998. The semantics and pragmatics of topic
phrases. Linguistics & Philosophy 21: 117–57.
Prideaux, Gary, 1993. Subordination and information distribution in oral and written
narratives. Pragmatics & Cognition 1: 51–69.
Prince, Ellen, 1978a. Comparison of Wh-clefts and It-clefts in discourse. Language 54:
883–906.
References 309
1981. Toward a taxonomy of given–new information. In Peter Cole (ed.), Radical

Pragmatics. New York: Academic Press.
1986. On the syntactic marking of presupposed open-propositions. In Papers from the
Parasession on Pragmatics and Grammatical Theory, Chicago Linguistic Society.
Chicago, Ill.: Chicago Linguistic Society.
1992. The ZPG Letter: Subjects, definiteness, and information status. In W. Mann
& S. Thompson (eds.), Discourse Description: Diverse Linguistic Analyses of a
Fund-Raising Text. Amsterdam: John Benjamins.
1997. On the functions of Left-dislocation in English discourse. In A. Kamio (ed.),
Directions in Functional Linguistics. Amsterdam: John Benjamins.
Prince, Gerald, 1987. A Dictionary of Narratology. Lincoln, Nebr.: University of
Nebraska Press.
Propp, Vladimir, 1958. Morphology of the Folktale. Bloomington, Ind.:Indiana Univer-
sity Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, & Jan Svartvik, 1985. A Com-
prehensive Grammar of the English Language. London: Longman.
Reichenbach, Hans, 1947. Elements of Symbolic Logic. London: Macmillan.
Reinhart, Tanya, 1980. Conditions for text coherence. Poetics Today 1: 161–80.
1982. Pragmatics and Linguistics: An Analysis of Sentence Topics. Indiana University
Linguistics Club.
Reinhart, Tanya, & Eric Reuland, 1991. Anaphors and logophors: An argument structure
perspective. In J. Koster & E. Reuland (eds.), Long Distance Anaphora. Cambridge:
Reinhart, T., & E. Reuland, 1993. Reflexivity. Linguistic Inquiry 24: 657–720.
Rieser, Hannes, 1978. On the development of text grammar. In W. Dressler (ed.), Current
Trends in Textlinguistics. Berlin: De Gruyter.
Rizzi, Luigi, 1997. The fine structure of the left periphery. In L. Haegeman (ed.), Elements
of Grammar. Dordrecht: Kluwer.
Roberts, John, 1988. Arnele switch-reference and the theory of grammar. Linguistic
Inquiry 19: 45–63.
Rochemont, Michael, 1986. Focus in Generative Grammar. Amsterdam: John
Benjamins.
Rooth, M., 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–
116.
Ross, John, 1970. On declarative sentences. In R. Jacobs and P. Rosenbaum (eds.),
Readings in English Transformational Grammar. Waltham, Mass.: Ginn & Co.
1972. Act. In D. Davison and G. Harmon (eds.), Semantics of Natural Languages.
Dordrecht, Holland: Reidel.
Rumelhart, David, 1975. Notes on a schema for stories. In D. G. Bobrow & A. M.
Collins (eds.), Representation and Understanding: Studies in Cognitive Science.
New York: Academic Press.
Sachs, Jacqueline, 1967. Recognition memory for syntactic and semantic aspects of
connected discourse. Perception and Psychophysics 2: 437–42.
Sanders, José, & Gisela Redecker, 1996. Perspective and representation of speech in
narrative discourse. In G. Fauconnier & E. Sweetser (eds.), Spaces, World, and
Grammar. Chicago, Ill.: University of Chicago Press.
310 References
Sandstrom, Görel, 1993. When-clauses and the Temporal Interpretation of Narrative

Discourse. Umea, Sweden: University of Umea.
Sanford, A. J., & Simon Garrod, 1981. Understanding Written Language. Chicester:
Wiley.
Sanford, Anthony, & Linda Moxey, 1995. Aspects of coherence in written language:
A psychological perspective. In M. Gernsbacher & T. Givón (eds.), Coherence in
Spontanteous Text. Amsterdam: John Benjamins.
Sasse, Hans-Jürgen, 1987. The thetic–categorical distinction revisited. Linguistics 25:
511–80.
Schank, Roger, & R. P. Abelson, 1977. Scripts, Plans, Goal, and Understanding: An
Inquiry into Human Knowledge Structures. Hillsdale, N.J.: Erlbaum.
Schiffrin, Deborah, 1987. Discourse Markers. Cambridge: Cambridge University
Press.
Schlenker, Phillipe, 1999. Propositional attitudes and indexicality: A cross-categorial
approach. Ph.D. dissertation, MIT.
Sells, Peter, 1987. Aspects of logophoricity. Linguistic Inquiry 18: 445–79.
Sgall, Petr, 1995. Formal and computational linguistics in Prague. In E. Hajičová,
M. Cervenka, O. Leška, & P. Sgall (eds.), Travaux du Cercle Linguistique de
Prague, Nouvelle Série, Vol. I. Amsterdam: John Benjamins.
Sgall, Petr, Eva Hajičová, & E. Burànová, 1986. The Meaning of the Sentence in its
Semantic and Pragmatic Aspects. Prague: Academia.
Smith, Carlota S., 1971. Sentences in discourse: An analysis of a discourse by Bertrand
Russell. Journal of Linguistics 7: 213–35.
1980. Temporal structures in discourse. In N. Rohrer (ed.), Time, Tense and Quanti-
fiers. Tübingen: Niemeyer.
1991. A valediction for sentence topic. In C. Georgopoulos & R. Ishihara (eds.),
Interdisciplinary Approaches to Language. Dordrecht: Kluwer.
1995a. The range of aspectual situation types: Shifts and a bounding paradox. In
P. Bertinetto et al. (eds.), Temporal Reference: Aspect and Actionality. Turin:
Rosenberg & Sellier.
1995b. The relation between aspectual viewpoint and situation type. Address,
Linguistic Society of America. Published electronically, Eric database.
1995c. Aspect and temporal location in Discourse Representation Theory.
Unpublished ms., University of Texas.
1997. The Parameter of Aspect, 2nd edition. Dordrecht: Kluwer.
1999a. Closed systems in texts. In Proceedings of the Second Workshop on Text
Structure. Austin, Tex.: Department of Linguistics, University of Texas.
1999b. Activities: States or Events? Linguistics and Philosophy 22: 479–508.
Smith, Carlota S., & Mary Erbaugh, 2001. Temporal information in sentences of
Mandarin. In X. Liejiong and S. Jingmin (eds.), New Views in Chinese Syntactic
Research: International Symposium on Chinese Grammar for the New Millennium.
Hangzhou: Zhejiang Jiaoya Chuban she.
Speas, Margaret, 1990. Phrase Structure in Natural Language. Dordrecht: Kluwer.
Speas, Margaret, & Carol Tenny, in press. Configurational properties of point of view
roles. In A. di Scuillo (ed.), Proceedings of UQAM Asymmetry Workshop. Montreal:
University of Québec at Montreal.
References 311
Sperber, Dan, & Deirdre Wilson, 1986. Relevance: Communication and Cognition.
Oxford: Blackwell.
Stalnaker, Richard, 1978. Assertion. In P. Cole (ed.), Pragmatics, Syntax and Semantics,
9. New York: Academic Press.
Stark, Heather, 1988. What do paragraph markings do? Discourse Processes 11:
275–303.
Steedman, Mark, 2000. The Syntactic Process. Cambridge, Mass.: MIT Press.
Stern, Arthur, 1976. When is a paragraph? College Composition and Communication
36: 253–57.
Stirling, Leslie, 1993. Switch-Reference and Discourse Representation. Cambridge:
Stoddard, Sally, 1991. Text and Texture: Patterns of Cohesion. Norwood, N.J.:
Ablex.
Stowell, Tim, 1996. The phrase structure of tense. In J. Rooryck & L. Zaring (eds.),
Phrase Structure and the Lexicon. Dordrecht: Kluwer.
Strawson, Peter, 1964. Identifying reference and truth value. Theoreia 30: 96–118.
Reprinted in D. Steinberg & L. Jakobovits (eds.), 1971. Semantics. Cambridge:
Svarthvik, Jan, 1966. On Voice in the English Verb. The Hague: Mouton.
Swales, John, 1990. Genre Analysis: English in Academic and Research Settings.
Talmy, Leonard, 1978. Figure and ground in complex sentences. In J. Greenberg,
C. Ferguson, & E. Moravcsik (eds.), Universals of Human Language: Syntax.
Stanford, Calif.: Stanford University Press.
1985. Lexicalization patterns: Semantic structure in lexical forms. In T. Shopen (ed.),
Language Typology and Syntactic Description, Vol. III: Grammatical Categories
and the Lexicon. Cambridge: Cambridge University Press.
2000. Towards a Cognitive Semantics. Cambridge, Mass.: MIT Press.
Tannen, Deborah, 1989. Talking Voices. Cambridge: Cambridge University Press.
Tardieu, Hubert, Marie-France Ehrlich & Valérie Gyselinkck, 1992. Levels of repre-
sentation and domain-specific knowledge in comprehension of scientific texts.
Language & Cognitive Processes 7: 335–51.
Taylor, Barry, 1977. Tense and continuity. Linguistics & Philosophy 1: 199–220.
Thomason, Lucy, 1994. The assignment of proximate and obviative in informal Fox
narrative. Ms., The University of Texas at Austin.
Thompson, Ellen, 1999. The temporal structure of discourse: The syntax and semantics
of temporal then. Natural Language & Linguistic Theory 17: 123–60.
Thompson, Geoff, & Susan Hunston, 2000. Evaluation: An introduction. In
S. Hunston & G. Thompson (eds.), Evaluation in Text: Authorial Stance and the
Construction of Discourse. Oxford: Oxford University Press.
Thompson, Sandra, 1987. The passive in English: A discourse perspective. In
R. Chaman & L. Shockey (eds.), In Honor of I1se Lehiste. Dordrecht: Foris.
2001. “Object complements” and conversation: Towards a realistic account. To appear
in Studies in Language.
Thorndyke, P. W., 1977. Cognitive structures in comprehension and memory of narrative
discourse. Cognitive Psychology 9: 77–110.
312 References
Tomlin, Russell (ed.), 1987. Coherence and Grounding in Discourse. Amsterdam: John
Benjamins.
Tracy, H. P., & S. H. Levinsohn, 1977. Participant reference in Ica expository discourse.
In R. Longacre (ed.), Discourse Grammar, Part III. Dallas, Tex.: Summer Institute
of Linguistics.
Turner, Elizabeth, & Ragnar Rommetveit, 1968. Focus of attention in recall of active
and passive sentences. Journal of Verbal Learning and Verbal Behavior 7: 543–48.
Vallduvı́, Enric, 1992. The Informational Component. New York: Garland Press.
Vallduvı́, Enric, & Elisabet Engdahl, 1996. The linguistic realization of information
packaging. Linguistics 34: 459–519.
Vallduvı́, Enric, & Maria Vilkuna, 1998. On rheme and kontrast. In P. Culicover &
L. McNally (eds.), The Limits of Syntax, Syntax and Semantics, 26. New York:
Academic Press.
Vendler, Zeno, 1957. Verbs and times. Philosophical Review 66: 143–60. Reprinted
in Vendler, 1967. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University
Press.
1967. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press.
1972. Res Cogitans: An Essay in Rational Psychology. Ithaca, N.Y.: Cornell University
Press.
Venneman, Theodor, 1975. Topic, sentence accent, ellipsis: A proposal for their formal
treatment. In E. Keenan (ed.), Formal Semantics of Natural Language. Cambridge:
Verkuyl, Henk, 1972. On the Compositional Nature of the Aspects. Foundations of
Language Supplementary Series, Vol. XV. Dordrecht: Reidel.
Verstraete, Jean-Christophe, 2001. Subjective and objective modality: Interpersonal and
ideational functions in the English modal auxiliary system. Journal of Pragmatics
33: 1505–28.
Vlach, Frank, 1981. The semantics of the progressive. In P. Tedeschi & A. Zaenen (eds.),
Tense and Aspect, Syntax and Semantics, 14. Academic Press, New York.
Voegelin, C. F., & F. M. Voegelin, 1975. Hopi (-qa). International Journal of American
Vonk, Wietske, Lettica Hustinx, & Wim Simons, 1992. The use of referential expressions
in structuring discourse. Language & Cognitive Processes 7: 302–33.
Walker, Marilyn, Aravind Joshi, & Ellen Prince (eds.), 1998. Centering Theory in
Discourse. Oxford: Clarendon Press.
Ward, Gregory, & Ellen Prince, 1991. On the topicalization of indefinite NPs. Journal
of Pragmatics 16: 167–77.
Whately, Richard, 1828/1963. Elements of Rhetoric. Carbondale, Ill.: Southern Illinois
Press.
Whorf, Benjamin L., 1956. Language, Thought and Reality: Selected Writings of
Benjamin Lee Whorf . New York: Wiley and Son.
Whitaker, Jeanne, & Carlota S. Smith, 1985. Some significant omissions: Ellipsis in
Flaubert’s “Un Coeur Simple.” Language, and Style 14: 251–92.
Wiebe, Janyce, 1991. Tracking point of view in narrative. Computational Linguistics
20: 233–87.
References 313
Willett, Thomas, 1988. A crosslinguistic survey of the grammaticalization of eviden-

tiality. Studies in Language 12: 51–97.
Winkler, Anthony, & Jo Ray McCuen, 1988. Rhetoric Made Plain. New York: Harcourt
Brace Jovanovich.
Wittgenstein, Ludwig, 1958. The Blue and Brown Books. Oxford: Basil Blackwell.
Woodbury, Anthony, 1987. Native American Discourse: Poetics and Rhetoric.
Wright, Karen, 2002. Times of our lives. Scientific American 287: 58–65.
Zagona, Karen, 1990. Times as temporal argument structure. Unpublished ms., Univer-
sity of Washington, Seattle.
Zribi-Hertz, Ann, 1989. Anaphor binding and narrative point of view: English reflexive
pronouns. Language 65: 695–727.
Zwicky, Arnold, 1971. In a manner of speaking. Linguistic Inquiry 11: 222–33.
General index
Abstract Entities 24, 78, 88 common ground, context set 49, 56, 191,
Facts 24, 74, 89 237
Propositions 24, 74, 89, 163 communication, expressions of 159
accent, accentuation 188, 193–95, 204, direct address 149, 162, 178
219 direct speech 149, 159, 178
accommodation, see inference indirect speech 160, 169–77, 211
adjunct preposing 228, 232, 233 quoted speech 159, 178
locative PPs 228 represented speech 161
temporal clauses 231 verbs of 161
advancement, temporal 14, 26–28, 30, 57, communicative dynamism 185, 187
93 complement clauses and abstract entities 78,
adverbials 104
durative adverbial, tacit 80, 96, 112, 212 communication verbs 104
locating 120 temporal interpretation 27, 69, 119, 120
anaphora 62, 150 composite 10, 57, 175
argument mode 20, 33–34, 37, 98, 252 compositional rules in DR Theory 82, 83, 84,
argument preposing 224 88, 123
aspectual systems 67, 70 condensed clauses 140, 230
attention 13, 25, 202 construction rules, see Discourse
Representation Theory
Background, see Foreground and Background Context 11, 105, 108
Bounded Event Constraint 76, 88, 102–3 continuity 112, 145, 149, 178
exceptions to 103 global continuity 13, 20, 198
Boundedness, bounded events 14, 26, 71, local continuity 13, 198
102–3 subjective 168
continuity tense interpretation, see tense
canonical sentences 13, 148, 187, 213 contrast and emphasis 205, 208
categorical sentences 195–96 alternatives 205
Centering Theory 13, 146, 193 oppositional contrast 198–208
Cleft constructions 24, 220
it-cleft 34, 219, 221 dative alternation 226
wh-cleft 219, 221 Deictic tense, see tense
closed system 13, 69, 100, 118, 132 deixis, deictic 17, 30, 128, 155
coercion 71, 86, 96, 118, 123 adverbs 101, 108, 160
coherence 45, 47 derived situation types, see situation type
cohesion 46 Description mode 20, 28, 36, 95, 248
lexical 46, 47, 265 discourse continuity 149, 178
commentary 33 discourse grammar 44–45
314
General index 315
Discourse Modes 39, 265 focus phrase 187, 201, 202, 203, 207, 212, 227
atemporal 25, 31, 58, 243 all-focus 202, 218
temporal 25, 30 argument focus 202
discourse relations 132, 253, 260 focus–background partition 185, 203,
formalization 263 209–10
inference 260 identificational 207
relational segment 261 informational 201, 207
SDRT 263 predicate focus 202
Discourse Representation Structure (DRS) 25, Focus Preposing 225
60–64, 86, 107, 112, 151 Foreground and Background 34
description fragment 112, 113 in narrative 35
narrative fragment 110 shifts 33, 35
subjectivity 169–77, 211 forms of discourse 38–39, 218
temporal 107 functional units 259
Discourse Representation Theory narrative 259
(DR Theory) 11, 55–57, 149
construction rules 57–58, 60, 107, 254 General Stative 24
discourse topic 197, 236–37 Generalizing 24, 72–73, 77, 87
dislocation 225, 226 Generic 24, 72, 77
duality 19, 80, 96, 186, 212 genre 9, 10, 41, 42
dynamism, dynamic/static 26, 27, 69, 75, 76, global continuity 13, 20, 198
120 grammatical terms 11
emphasis, see Contrast and Emphasis Heavy NP Shift 226

entities, in a DRS 59, 150 hierarchical structure 258
episode 8, 44, 260
Evaluative, see Evidential and Evaluative Inference 25, 28, 51, 262
subjectivity accommodation 53–59, 96
Events 24, 29, 72 bridging 53, 55
Activity 53–59, 96 Information mode 17, 20, 32–33, 37, 98, 129,
Evidential and Evaluative subjectivity 17, 155, 251
164, 165, 169–77, 211 information structure 13, 169–77, 211
adjective 164 inversion 222, 239
compositional rules 158–80 locative inversion 224, 234
modality 165 predicate inversion 223
existential there 39, 218, 224 quotation inversion 222
definite NPs 218 subject–aux inversion 222
External Override Principle 86 topics 224
extraposition 221
heaviness 222 lexical patterns, see Cohesion
linguistic category, covert 23, 92
Facts, see Abstract entities linking
familiarity status 141–42, 143, 145, 185, 214, sentence 213, 214
217, 218, 221, 223, 226, 227 poset 214, 215, 246, 250
features, linguistic 43, 59, 84 localist principle 82, 124
temporal 69, 75, 119 logophoric, see Pronouns
orientation 119
relational 119 mental model 55
role 30, 121 mental state, mind, expressions of 155, 180,
Figure–Ground sentences 126, 127 185
316 General index
metaphorical motion, location 17, 18, 26, 31, Prague School 185, 188, 196, 209
32, 58, 123, 254 predominant entity
multi-clause sentences 233 presentation 13, 37, 212, 233
complex sentences 235 and grammar 5ßß
conjunction 235 in DRS 238
relative clauses 233 progression 13, 213, 244
patterns 245
Narrative mode 14, 19, 27, 33, 108, presupposition 24, 209, 213, 219, 231
247 pragmatic 53, 136, 221
advancement 26–28, 93, 235 semantic 53, 54
as background 36 Primary Referent 17, 19, 31, 123, 129
tense in narrative 26, 108 construction rules 130
non-canonical sentences 213, 215, 239 criteria 125–29
heaviness 216, 222, 227 states 125–27
normalization, complements 80–81, 96, 98, progression, see Presentation, progression; text
212 progression
noun classifiers 133 progressive 14, 27, 68, 69, 71, 72, 91, 120,
171
obviation, see pronoun prominent argument 129
order 187, 188 pronouns 11, 52, 132, 136
in a DRS 61
paragraphs 45, 56, 236–37 Locally Free Reflexives 137–40, 150
and discourse relations 236 empathy 138
partitioning 209 emphasis 139
dual 209, 211 intensive 139
focus–background 210 logophoric 135, 138
Link 210 null 140
passage 22, 105 parallel structure 146
passive 215, 217, 234 proximal–obviative 133, 134, 197
perception and subjectivity 27, 169, 170, reflexive 150, 161, 173
182 stressed 136–37
inferred perception 171, 182 switch reference 134, 141
perfect 24, 72, 100, 101 Propositions, see Abstract entities
perfective viewpoint 27, 68, 71, 91
Activity sentences 70, 96, 113 quantification test, for Abstract Entities 81
and narrative 93
perspectival subjectivity 172, 175, 182 Reference Time 93, 94, 95, 100, 104, 106,
deictic adverbials and 175 114
empathy 174 referring expressions 141, 147, 148, 149,
reflexives 173, 174 169–77, 187, 211
point of view 155, 171 accessibility 143, 150
poset, see linking cognitive status 144
position 192, 204, 214, 223, 227, 235 in DRT 149
pragmatics 50, 51, 58, 64, 94, 97, 123, 132, reflexives, see pronouns
136, 145, 158–80 Report mode 16, 20, 30, 36, 42, 97, 108,
and semantic meaning 50, 56, 127 250
and sentence topic 200 representations
constraint 103 in DR Theory 12
in DRS 56, 151 mental 55
General index 317
Responsible Source 13, 156, 158 modeled with subordination 156, 176
rhetoric objective sentences 157, 158
classical 38 scope 162, 168
modern 39, 238 syntax, syntactic 11, 53–59, 96, 239
rhetorical modes 40–42
Rhetorical Structure Theory 262 telic/atelic 76, 118
role 121 temporality 12
tense 26, 30, 92, 95, 99, 102, 105, 160
salience 146, 198–208 anaphora 97, 112
sentence Topic 13, 15, 19, 58, 187, 192, 197, continuity 28, 94–95, 108–10
207, 212, 216, 221, 225, 229, 234, 246, deictic pattern 95, 106, 113, 114
252, 253, 260 present 76
aboutness 189, 190 text
and grammatical voice 193 linguistics 45
identifying cues 198 units 9, 45
paraphrase tests 190 well- and ill-formed 44
sentences without topic 195 text progression 12
terminology 195–96 atemporal 132
topic–comment partition 185, 188, 209 presentational 19
unclear cases 200 temporal, see advancement
shifts, discourse mode 22, 34, 35, 38 text type 43–44
situation entities 12, 24, 67, 82 thematic role 31, 128
Situation Time 100, 102, 106 Experiencer 15
situation type 68, 69, 82, 83, 86 Theme/Patient 17, 58, 128, 130
derived 71, 85 Theme 189, 196
shifts, see coercion thetic sentences 195, 202
situations – classified 23–25, 69 time, temporality 12, 99
space, spatial 26, 58, 115 Topic, see sentence topic, discourse topic
spatial progression 26–28, 29 topicalization 225
specific situations 22, 32, 37
Speech Time 30, 99, 106, 121 unbounded situations 27, 71, 102–3, 125–29
State 71
static, see dynamism, dynamic/static variety in texts 8, 40
sub-interval property 71 verb constellation 23, 68, 72–73, 83
subjectivity 13, 58, 156, 162, 167, 244 viewpoint
construction, compositional rules 176, aspectual 68–69, 70, 82
178 in DRS 90
in fiction 157, 158–80 visible information 68, 71, 82, 124
lexical choice 156
linguistic forms 175 will auxiliary 30, 100, 121
Index of names
Ariel 143 Delin 219–20

Aristotle 38, 69, 119 Dowty 128
Arnold et al. 216, 227
Asher 81 Enç 100
Erbaugh 129–33
Bach, Kent 50 Erteschik-Shir 194
Baker 139 Erteschik-Shir & Lappin 229
Banfield 157, 160
Barker 143 Faigley & Meyer 41
Biber 43 Fauconnier 177
Birner 217 Fillmore 116
Birner & Ward 215, 217, 218, 223, Firbas 196
225 Flashner 148
Blakemore 146 Fleischman 171
Bolinger 174, 185, 205, 224 Fox 147
Braddock 238 Freksa 116, 117
Bransford et al. 54 Friedman 116
Brennan 193
Garnham 44
Britton 41
Genette 157
Brooks & Warren 40
Genung 40
Giora 236
Caenepeel 30, 101, 155, 170, 232
Givón 194, 196
Cairns 28, 40
Gordon & Chan 147, 217
Cantrall 138, 173
Goutsos 237
Chafe 42, 195, 197, 236
Green 223
Chierchia 192
Grice 50, 141
Chu 133
Grimes 261
Cinque 56, 164
Grimshaw 129
Clancy 148
Gruber 124
Clark 116
Gundel 203
Comrie 171
Gundel et al. 144
Connors 41
Gussenhoven 201
Corbett 39
Gutwinski 47
Culicover 11
Hagège 137, 138
Damasio 21, 116 Hajičová & Sgall 188, 203,
Daneš 187, 204, 245 227
318
Index of names 319
Halliday 196, 202 Mann & Thompson 237, 262

Halliday & Hasan 46, 47 Markels 238
Hannay 205 McArthur 238
Haviland & Clark 53 Miller, G., et al. 258
Hawkins 216 Miller, P. 222
Hearst 47 Mitchell 172
Hedberg 219–20 Moens 26
Heim 11, 210 Moens & Steedman 232
Herbst 74
Heycock & Kroch 219 Nuyts 165, 166
Hinds 149
Hinrichs 101, 155 Olman 32
Hirose 174, 185
Hirschberg 215 Parkes 238
Hobbs 262 Partee 126, 192, 204
Hoey 46 Peterson 79, 80
Hopper 46 Polinsky 127
Horn 239 Portner & Yabushita 192, 201
Hunston & Thompson 176 Prideaux 187
Prince, E. 142–43, 145, 214, 219, 221, 225,
Jackendoff 124 226
Jakobson 132 Prince, G. 101, 155
Jespersen 140, 216
Johnson-Laird 55 Quirk et al. 228
Kameyama 145 Reichenbach 100

Kamp 11 Reinhart 191
Kamp & Reyle 30, 59, 112 Reinhart & Reuland 138
Keenan & Schieffelin 197 Ross 138, 173
Kinneavy 42 Rumelhart 44
Kiss 207
Klaiman 193 Sanford & Garrod 145
Koen et al. 238 Sgall et al. 201
Kratzer 166 Speas 129–33
Krifka 85 Stalnaker 49
Krifka et al. 24, 73, 77 Stark 238
Kuno 139, 174 Stirling 134
Kuroda 156, 157 Stoddard 47
Strawson 190
Labov & Waletzky 26, 260
Lambrecht 206, 218, 226 Talmy 124, 127
Lascarides & Asher 263 Tannen 159
Levin 117 Taylor 101
Levinson 10, 50, 51, 52, 239 Thompson 38, 119
Lewis 53
Linde 148 Vallduvı́ & Engdahl 207, 210
Longacre 42, 237, 260 Van Dijk 197
Loriot & Hollenbach 45 Van Dijk & Kintsch 55
Lyons 157, 166 Van Kuppevelt 192
320 Index of names
Vendler 69, 74, 112 Ward & Prince 214

Verstraete 166 Whately 39
Vonk et al. 148 Wittgenstein 10
Wright 21
Walker et al. 147
Ward 239 Zribi-Herz 139

Modes of Discourse - Smith Carlota (2003)

Uploaded by

Copyright:

Available Formats

Modes of Discourse - Smith Carlota (2003)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modes of Discourse - Smith Carlota (2003)

Uploaded by

Copyright:

Available Formats

What are the five discourse modes identified in the text?

What are the five discourse modes identified in the text?

What linguistic features does the author analyze to distinguish each discourse mode?

What linguistic features does the author analyze to distinguish each discourse mode?

This page intentionally left blank

c a r l o ta s. s m i t h is Centennial Professor of Linguistics at the University

Cambridge University Press

© Carlota S. Smith 2003

This book is in copyright. Subject to statutory exception and to the provision of

First published in print format 2003

Cambridge University Press has no responsibility for the persistence or accuracy of

Preface page xiii

part i discourse structure

1 The study of discourse 7

2 Introduction to the Discourse Modes 22

3 Text representation and understanding 49

part ii linguistic analysis of the discourse

4 Aspectual information: the entities introduced

4.3 Abstract Entities 74

5 Temporal and spatial progression 92

6 Referring expressions in discourse 123

part iii surface presentational factors

7 Subjectivity in texts 155

8 The contribution of surface presentation 185

9 Non-canonical structures and presentation 213

9.4 Multi-clause sentences 233

part iv discourse modes and their context

10 Information in text passages 243

11 Discourse structure and Discourse Modes 258

Appendix A: The Texts 267

Appendix B: Glossary 286

literature, I found to my surprise that the Discourse Modes have counterparts

stance. I attempt to understand and explain some of the complexity of discourse.

1.1 Discourse Modes

In order, these fragments exemplify the modes of Narrative, Description, Report,

1.2 Approaches to the study of texts

The idea of discourse as a type of activity helps us to discard our customary

1.2.2 The linguistic approach

Whether a sentence expresses an Event or a State, for instance, depends on the

1982). As discourse is dynamic, so representations must be: new information

1.3 Overview of key ideas

1.3.1 Situation type, text progression, subjectivity, surface

1.3.2 Multiple analysis of text passages

(6) Narrative a: situations and text progression

(7) Narrative b: situations, text progression, subjectivity

(8) Narrative c: situations, text progression, subjectivity, topic

(9) Report a: situations and text progression

(10) Report b: situations, text progression, subjectivity, topic

Metaphorical progression in the ﬁrst paragraph moves from “a message” to the

(12) Information b: situations, Primary Referents, subjectivity

(13) Informative c: situations, Primary Referents, subjectivity, topic

The single instance of non-canonical syntax, the extraposed that-clause in S2,

The Narrative mode

The Report mode

People intuitively recognize passages of the Discourse Modes, although they

(1) Listening to Humpbacks

The ﬁrst two sentences express generalizations about humpback whales. At

2.1 The entities introduced in texts

2.1.1 Classes of situation entities

(2) Sentences expressing situations