Currie, Gregory - Image and Mind PDF

This is a book about the nature of film: about the nature of
moving images, about the viewer's relation to film and about

the kinds of narrative that film is capable of presenting. It rep-
resents a decisive break with the semiotic and psychoanalytic
theories of film which have dominated discussion over the past
twenty years.
The central thesis is that film is essentially a pictorial medium
and that the movement of film images is real rather than illu-
sory. A general theory of pictorial representation is presented
which insists on the realism of pictures and the impossibility of
assimilating them to language. It criticizes attempts to explain
the psychology of film viewing in terms of the viewer's imagi-
nary occupation of a position within the world of film. On the
contrary, film viewing is nearly always impersonal.
Gregory Currie provides a general theory of narration and its
interpretation in both pictorial and linguistic media, and con-
cludes with an analysis of some ways in which film narrative
and literary narrative differ.
Image and mind
Film, philosophy and cognitive science
GREGORY CURRIE
.... ~····· CAMBRIDGE

::: UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 1RP
40 West 2oth Street, New York, NY 10011-4211, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
© Cambridge University Press 1995
First published 1995
Library of Congress Cataloging-in-Publication Data

Currie, Gregory.
Image and mind : film, philosophy and cognitive science I Gregory
Currie.
P· em.
Includes bibliographical references and index.
ISBN o-521-45356-9 (hardback)
1. Motion pictures- Philosophy. I. Title.
PN1995.C795 1995
791.43'01 - dc20 95-1357
CIP
A catalog record for this book is available from the British Library.
ISBN o-521-45356-9 Hardback
Transferred to digital printing 2004

For Gabriel Christopher, future film watcher
Doctrine and life, colours and light, in one
When they combine and mingle, bring
A strong regard and awe; but speech alone
Doth vanish like a flaring thing,
And in the ear, not conscience, ring
George Herbert, The Windows
Contents
Preface page xi
Acknowledgements xxi
Film, 1895-1995 xxiii
Introduction: the essence of cinema 1

1.1 Film as representation 2
1.2 Film and the visible 3
1. 3 Degrees of inessentialness 6
1.4 The visible and the pictorial 7
1. 5 Fictional pictures 9
1.6 Documentary and fiction 12
Part I Representation in film

Chapter 1 The myth of illusion 19
1.1 Transparency, likeness and illusionism 19
1.2 Cognitive illusions 22
1.3 Film and dreaming 27
1.4 Perceptual illusionism 28
1.5 The reality of cinematic images 30
1.6 On the motion of images 34
1.7 Two ways or three? 42
Chapter 2 The imprint of nature 48
2.1 Representational media and representational
arts 49
vii
Contents
2.2 Presenting and representing 49
2.3 Photographs as evidence 52
2.4 Counterfactual dependence 53
2.5 Inconclusive arguments 56
2.6 The conditions for perception 61
2.7 Putting movement in the picture ~
2.8 The significance of similarity 70
2.9 The aesthetics of photography 72
2.10 Photography, painting and the real 75
Chapter 3 Realism 79
3.1 Depictions 8o
3.2 Natural generativity 88
3·3 Nondepictive representation in film and other
media 90
3·4 Three kinds of temporality 92
3·5 Representing time by means of time 96
3.6 Spatial representation 103
3·7 Realist film style 1o6
3.8 The relativity of likeness 108
Chapter 4 Languages of art and languages of film 113

4.1 Finding the thesis 114
4.2 Cinema language and natural language 117
4·3 The shape of natural language 120
4·4 Objections rejected 124
4·5 Interpretation and utterance meaning 126
4.6 Confusions about convention 128
4·7 Relations between images 134
4.8 A language of vision? 136
Part II Imagination
Chapter 5 Imagination, the general theory 141
5.1 Perspective shifts 142
5.2 Simulation 144
5·3 Fiction and two kinds of simulation 152
viii
Contents
5·4 Consequences 155

5·5 The dangers of imagining 162
Chapter 6 Imagination, personal and impersonal 164
6.1 The classical theory 165
6.2 Cinema, theatre and other visual fictions 168
6.3 Against imagining seeing 170
6.4 Impersonal imagining and film 179
6.5 Perceptual imagining 181
6.6 Clarifications and rebuttals 185
6. 7 The myth of total cinema 191
6.8 Psychologism 193
6.9 Iconic signs 196
Chapter 7 Travels in narrative time 198
7.1 Tense in film 198
7.2 The proper treatment of anachrony 2o6
7·3 Is this revisionism? 216
Appendix: anachrony and ellipsis 219
Part III Interpretation

Chapter 8 The interpretive problem 225
8.1 Intersubjective agreement and interpretive
principles 226
8.2 Terms and conditions 231
8.3 Interpreting behaviour and interpreting works 235
8.4 Interpreting works as interpreting behaviour 239
8.5 Real authors, implied authors 243
8.6 Why we still need intention 247
8. 7 Interpretive deadlock and truth 249
8.8 The evidence for a cinematic interpretation 251
8.9 Implied author and auteur 258
8.10 Structure and function 259
Chapter 9 Narrative and narrators 26o
9.1 Implied authors and narrators 261
9.2 The asymmetry between literature and film 265
ix
Contents
9·3 Kinds of unreliability

9·4 Ambiguous and unreliable narratives
In conclusion 281
Named propositions
Bibliography
Index
X
Preface
Hard is his lot, that here by fortune placed,
Must watch the wild vicissitudes of taste,
With every meteor of caprice must play,
And chase the new-blown bubbles of the day.
Dr Johnson
This book aims to place film in relation to other things: to other

arts of the same and of different kinds; to modes of represen-
tation like pictures and language; to time and the representation
of time; to narrative and the comprehension of narrative; to
imagination and belief; to the real world. To all these things,
film stands in interesting and controversial relation. Getting that
relation right will mean we are on the way to understanding
the medium of film.
This is a philosophical book about film, a claim that I had
better expand on, lest it be written off as a vague gesture in the
direction of depth and subtlety. I mean, quite specifically, that
the method of this book is generalizing, systematizing, argu-
mentative and conceptual. It aims at conclusions of maximum
generality rather than a concentration on particular works,
schools or genres; it aims to integrate what can be said about
film with (what I take to be) our best theory of the rest of the
world; it proceeds in steps, laying out premises and conclusions,
indicating how we get from one to the other; it tries to provide
analyses of opaque, complex and contested notions like lan-
guage, image, representation and belief.
The book continues the project of an earlier work, The Nature
of Fiction, in which I presented a basic framework for the anal-
ysis of fictions of all kinds, though I had little to say there about
film. At various points in this study I draw on the theory pre-
sented earlier, but I try to enable the reader to follow my ar-
xi
Preface
gument without having to turn to the earlier work for help.
Also, at various points in this volume I deviate from the theory
of fiction in the earlier book. I have not in general attempted to
explain or justify the deviations, or even to indicate that this is
what they are. I have assumed that the reader will be less in-
terested in the development of my ideas than in getting a rela-
tively coherent theory. There are references to the previous book
in the footnotes to this one for those who want to follow up
certain issues in more detail.
One aspect of the relations between that book and this one
deserves mention here. At the end of The Nature of Fiction I noted
that almost everything in my analysis of fiction depended on
the concept of make-believe or imagination, a notion which was
itself left unexplained. The present work attempts to fill the gap.
It does not offer a conceptual analysis of imagination; rather, it
postulates a mechanism by which imagining can be understood
to work as part of the ordinary, evolutionarily adaptive func-
tioning of the mind. That mechanism is mental simulation.
Chapter 5 explains the idea in outline, an idea I hope to develop
in more detail in a further work.
While nothing is discussed here unless it is put to use in un-
derstanding film, I provide answers to a number of questions of
general artistic concern: What is a pictorial medium, and what
kinds of pictures can we call realistic? What, if anything, is spe-
cial about the photographic method of producing pictures? Do
pictures represent by convention, or by resemblance? What is
the relation between picturing and language? Can there be a
pictorial language, or a linguistic picture? By what means do we
interpret works of fiction? Is there an account of interpretation
that covers both cinematic and other media? How does imagi-
nation work, and what role does it play in our response to fic-
tions? Is there a distinctively pictorial mode of imagining?
It may be inferred from this budget of questions that the cen-
tral concepts I use are representation and imagination. Both are
general categories with special application to film; by taking
them in a connected way as my leading notions I aim to place
film within a broader framework of theorizing about the fic-
tional. One question I don't ask: what makes a film good or
xii
Preface
valuable? I have nothing to say here about the value of films,
aesthetic or otherwise, either individually and in comparison
with one another, or collectively in comparison with other kinds
of arts. This neglect is not the result of indifference. With film
as with other arts, questions about value are the most important
ones. But they are also the hardest, partly because answering
them depends on answering a lot of other questions first. Some
of those prior questions are answered here.
My decision to approach film by way of a number of other,
more general issues may raise the suspicion that I fail to take
film seriously, or that I use it simply as a convenient peg on
which to hang other concerns. This is not so, as I think I can
briefly indicate.
Film certainly has not been taken seriously by most philoso-
phers of the kind in which I place myself. With philosophers of
the broadly analytical stripe, film is generally assumed to be a
marginal phenomenon within that almost terminally marginal-
ized subdiscipline, aesthetics. As a consequence, film has be-
come almost exclusively the theoretical province of those who
take their inspiration from other schools: semiotics, psychoanal-
ysis and Marxism. Much has been said about the relation of
cinematic "signs" to the signs of language, about the relations
between film, dreaming and illusion, about the role of film in
creating and promoting a false consciousness of self. The result
has been disappointing. But there are indications - few in num-
ber but significant in content - of a growing dissatisfaction with
these sorts of models for the theoretical analysis of film.' My
aim is to develop an alternative, and to connect the analysis of
film with the best work in contemporary metaphysics, philoso-
phy of mind, philosophy of language and cognitive science.
It is primarily because I reject the old framework and seek a
' See Noel Carroll, Mystifying Movies. See also George Wilson's Narration in
Light, for fascinating studies of particular films in which Wilson rejects the
tired categories of realism and illusionism usually trotted out to analyse "the
Hollywood Film". Some deficiencies in the methodology of contemporary
film theory are explored in depth, especially as they apply to the production
of particular film interpretations, in David Bordwell, Making Meaning.
xiii
Preface
new one that I have emphasized the connections between film
and other things; a framework for film that bore no interesting
relations to other arts and representational forms, and could not
claim explanatory success in other areas, would have little to
recommend it. But a framework of any merit must also help in
understanding film itself. The framework I offer does so. It en-
ables us to answer a number of specifically filmic questions:
What are cinematic images? What is the nature of our imagi-
native involvement with film? What is the truth in the claims of
so-called cinema-realism? How is time represented in film, and
what are the limits of this representation? What strategies does
the viewer use to interpret a film, and how do they relate to the
strategies we employ to understand verbal fiction?
For those familiar with contemporary film theory, some of the
answers I give will have an air of wilful implausibility. But our
judgements of plausibility are determined largely by our frame-
work of background assumptions; implausible theories can be
cogent, highly explanatory, even true, and background assump-
tions hopelessly false. I'll say something about how my as-
sumptions differ from those of film theory as recently practised.
The first assumption I reject is a conjunction: that psycho-
analysis, or some version of it, is correct, and that it is capable of
illuminating our experience of film. I happen not to believe this,
since I believe that psychoanalysis is false, not just in the sense of
getting a few things wrong, as relativity theory probably does,
but in the sense of being wildly, deeply and unrescuably false, as
Aristotle's physics is. And even if I believed psychoanalysis or
some version of it to be true, I would be sceptical of recent appli-
cations of it to our experience of the cinema. Of course the expe-
rience of cinema, like that of anything else, is a matter for
psychological investigation, and cannot be understood in a priori
philosophical terms. But the psychology we need is not psychoa-
nalysis - particularly not in the version of Lacan, about which I
shall say a little in a moment. Contemporary empirical psycholo-
gists and philosophers of language and mind have found a way
to pool their resources in the project called cognitive science. The
aim is to build plausible models of the mind and its functions
more detailed and specific than philosophers on their own could
xiv
Preface
devise, and more flexible and abstract than neuropsychology
alone could deliver. In contrast with the psychoanalytic program,
cognitive science combines rigorous and clear argument with a
commitment to the most demanding standards of empirical test-
2
ability we can devise.
Cognitive science is not a doctrine. While its practitioners
share broad assumptions, there is no one theory of mind which
all or a majority of cognitive scientists accept. The view of the
mind which I adopt here as a working hypothesis has certain
features I can lay out very briefly: it treats the mind as a hier-
archically organized structure with levels of more or less intel-
ligent decision making going on in it; it regards some of the
systems of the mind - visual perception, for example - as op-
erating in relative isolation from other systems and consequently
unable to benefit from certain information sources; it regards the
mind's knowledge of other minds as resulting not (or not only)
from the possession and deployment of a theory of mind, but
from the ability to make empathetic contact, an ability I shall
explain in terms of mental simulation. The components of this
view will receive appropriate elaboration as the steps in the ar-
gument require.
One particularly damaging consequence of the psychoanalytic
paradigm has been the tendency to think of film as an essentially
illusory medium, capable of causing the viewer temporarily to
think of the film world as real, and of himself as occupying a
place of observation within that world. Thus film theorists have
expended a great deal of effort in trying to show that the point of
view of the camera is usually understood to represent that of a
perceiving agent- that of a character, a supposed narrator or the
spectator, who is assumed to occupy the camera's position
through a process of identification.3 I shall argue in Part I that
, The contrast I see here between Freudian theory and cognitive science is not
accepted by everyone. See, e.g., Clark Glymore, "Freud's Androids".
3 See, e.g., Kaja Silverman, The Subject of Semiotics, p. 202: "The gaze which
directs our look seems to belong to a fictional character rather than to the
camera." See also Jacques Aumont, "The Point of View", p. 2: "The frame in
narrative cinema is always more or less the representation of a gaze, the
auteur's or the character's."
XV
Preface
films are not standardly illusionistic, and in Part II that it is sim-
ply false that the spectator identifies with the camera.
The second assumption of traditional theorizing about film is
the semiotic assumption: that there is a fundamental common-
ality between pictures and language. This is a belief that goes
along with the rejection of the hopelessly old-fashioned view
that, while words operate by convention, pictures operate by
similarity. On the semiotic view, all representation is conven-
tional, and the idea that pictures might in some sense be like
the things they picture is part of a benighted ideology of realism.
This assumption, unlike the psychoanalytic assumption, has
broad support across the intellectual community. A version of
it has been argued for by Nelson Goodman, and there are hints
of it in the work of the art historian Ernst Gombrich and the
perceptual psychologist Richard Gregory, whose views are con-
nected to Karl Popper's idea that perceptions are "hypotheses".4
The semiotic assumption has seen hard times, as people have
come up against awkward dissimilarities between the structure
of language and the structure of visual images.5 Yet it has shown
a remarkable tendency to persist, particularly in film studies.
Christian Metz, for example, recognized fairly early on that
there is nothing in the cinema corresponding to "a language-
system's characteristics and internal organization" .6 But he has
continued to apply the categories, or at least the terminology, of
linguistic analysis to film; he says, for example, that photo-
graphs lack the "syntactic components of discourse so numerous
in cinema", and he describes optical effects as "clauses of
speech" .7 And while the emphasis in film theory has moved
4
Popper is a self-proclaimed realist, but his views on perception seem to me
to undermine realism.
5
Gombrich, for example, has recently distanced himself from the semiotic rel-
ativism of Nelson Goodman's LAnguages of Art. But Gombrich's view that
pictures can be genuine likenesses sits uncomfortably with his insistence that
pictures are conventional. See his "Image and Code: Scope and Limits of
Conventionalism in Pictorial Representation", p. 12.
• Christian Metz, "On the Notion of Cinematographic Language".
7
Christian Metz, "Trucage and the Film", pp. 158 and 165. See also Jacques
Aumont et al., Aesthetics of Film, chapter 4·
xvi
Preface
away from the straightforwardly linguistic to the psychoana-
lytic, the impetus for this move seems to have come not from a
rejection of the linguistic model, but from the thought that psy-
choanalytic models are themselves language-like. Thus one of
the ideas that seems to put the psychoanalytic thinker Lacan in
favour with the film theorists is his claim that the unconscious
is "structured like a language".
I am of the contrary opinion: that pictures and language are
fundamentally distinct, that there is a sense in which pictures
are able to represent by means of likeness rather than conven-
tion. But it is important not to create a false dichotomy here
between those who think that works in pictorial media are
wholly understandable in terms of perceptual skills universal
across humankind and those who think that pictures require an
act of interpretation which by no means guarantees the same
outcome for everyone.8 If pictures appeal to basic perceptual
skills which are widely shared across communities and, as I be-
lieve, to some extent across species, there is still a good deal of
interpretive work left to be done once perceptual skills are de-
ployed. I take up issues of interpretation in the final part of this
book, where I argue for some fundamental commonalities be-
tween the interpretation of linguistically encoded works and the
interpretation of film and other pictorial media, despite the ex-
istence of the nature/ convention gulf which divides them.
So film theorists have misunderstood the relation between the
symbolic and the pictorial orders, and they have failed to pro-
duce a plausible psychology of the experience of cinema.9 But
the failures of film theory are more than failures of doctrine.
8
As do Norman Bryson et al., in their editorial introduction to Visual Theory.
9
This failure and the ones I discuss later are not exemplified everywhere in
theoretical writing on film, though the better kind of writing tends to be
critical rather than constructive. See, e.g., Brian Henderson, "Two Types of
Film Theory". There is also some spirited resistance to the Lacanian model,
as with Raymond Durgnat: "Film watching is no more phallic, self- (mirror)
centered, or voyeuristic than any other self forgetful activity, like reading or
listening to music" ("Theory of Theory -and Bufiuel the Joker", pp. 32-44).
In a review of Metz's Psychoanalysis and Cinema, Durgnat notes that Metz
"swallows Lacan bones, feathers, fur and all" (p. 6o).
xvii
Preface
They are also failures of style and of method. The failures of
doctrine cannot be fully understood in isolation from these other
difficulties.
It is frequently and truly said that writing in film theory has
a tendency to be obscure. There is also a great deal of unex-
plained jargon which is then used in so unsystematic a way that
no clear meaning for it can be inferred from its use. This failure
10
of style connects with the failure of doctrine: the lack of clarity

of much writing on film functions to protect bad theory from
the light of criticism.
The failure of doctrine is even more closely connected with a
failure of method. Film theorists have used intellectual strategies
that were almost bound to lead to disaster. One of them is the
casual employment of vague analogies. Profound connections
have been claimed between the cinema and Plato's cave, be-
tween the screen and the breast, between the experience of
movie watching and dreaming." Here is an example of the film-
is-like-something-else phenomenon, with Gilles Deleuze discov-
ering some unlikely parallels between film and mathematical
physics:
Crystalline narration will fracture the complementarity of a

lived hodological space and a represented Euclidean
space.... It is in this sense that we can talk about Riemannian
spaces in Bresson, in neo-realism, in the New Wave and in
the New York School, of quantum spaces in Robbe-Grillet, of
probabilistic spaces in Resnais, of crystallized spaces in Her-
zog and Tarkovsky. 12
•o Others see the issue of clarity differently: Vivian Sobchack speaks of the
"sloppy liberal humanism that retrospectively characterized cinema studies
before it was informed by the scientific methods and technically precise vo-
cabularies of structuralism and semiotics"; Sobchack, The Address of the Eye,
p. xiv.
" See also Anne Friedberg's exploration of the relations between cinema and
shopping ("Les Fl4neurs du Mal(l): Cinema and the Postmodem Condition",
pp. 419-431).
,. Gilles Deleuze, Cinema 2: The Time-Image, p. 129. See also the discussion of a
"gravity-free world" in Rick Altman, General introduction to Sound Theory,
Sound Practice, pp. 3-4.
xviii
Preface
Out of context of course, and a translation too, but at least a
prima facie example of wheels idly turning. Not that we should
retreat to a narrow formalism, or insist that nothing useful can
be gained by comparing cinema with other things. But we must
pick our analogies with care and attend to the details of their
justification. We can learn a great deal about cinema, I shall
argue, by comparing the interpretation of cinema with that of
language and of intentional behaviour. But these comparisons
cannot be implemented at the level of vague likenesses that are
really nothing more than metaphors. Metaphors are useful for
certain purposes; they can also be extremely misleading. That is
especially likely when it is forgotten that they are, after all, meta-
phors, or when the investigator has lost a sense of the distinction
between the literal and the metaphorical.
Perhaps the most significant failure of method in film theory
has been the habit of appealing uncritically to controversial, and
sometimes poorly corroborated, theories from other disciplines.
This is evident in the move to connect film with psychoanalysis.
It is standard for film theorists to appeal casually to Lacan's idea
of the "mirror stage" as support for some theory about the
relation between film and the viewer, remarking simply that
Lacan "has shown" that such and such is the case.' 3 When some-
one appeals to a theory as if it were established fact it's natural
to suppose that there is a substantial body of evidence in sup-
port of the theory in question, and that this evidence is so much
a part of our common knowledge that it would be tedious to
explain or even to refer to it. But there is no such well-known
body of evidence in the case of Lacan' s claims about the mirror
stage. So far as I have been able to gather, there is no evidence
for them at all.' 4
' See, e.g., Laura Mulvey: "Jacques Lacan has described how the moment
3
when a child recognizes its own image in the mirror is crucial for the con-
stitution of the ego" ("Visual Pleasure and Narrative Cinema", p. 8o7). See
also Daniel Dayan, "The Tutor-Code of Classical Cinema", p. 441; and Chris-
tian Metz, The Imaginary Signifier, p. 6.
'• Jacques Aumont put it mildly: 'The metapsychological model elaborated by
Metz, Baudry and others around 1975 is not easily supported by empirical
evidence" ("The Point of View", p. 19).
xix
Preface
I could go on. But extended polemic is a tedious thing; con-

structive theory is much more exciting. I suppose many pre-
Copernican astronomers had moments of gloom when they
stood back a little from the vast implausible shambles that was
Ptolemaic astronomy. They probably found consolation in the
idea that it was the only game in town. What is on offer in this
book may not be of Copernican proportions, but it is, at least,
another game.
XX
Acknowledgements
The pleasures of writing this book have been many, and some
of them have to do with people and places. A draft was written
in 1991 during a year's sabbatical from the University of Otago.
It began at the Research School of Social Sciences at the Austra-
lian National University in Canberra; I doubt if there is a better
environment for philosophical research anywhere. The philos-
ophers there who helped straighten out my early ideas on the
subject I thank later. Here I thank Professor Tom Campbell, for
the use of his delightful house and exciting car. The writing
continued in Washington, D.C., and at the nearby College Park
campus of the University of Maryland, where teaching the phi-
losophy of film and talking to the excellent and friendly philos-
ophers helped to give the idea of this book a workable shape.
A first draft was completed in Cambridge, where the staff and
fellows of Clare Hall made us welcome. My special thanks go
to the president, Anthony Low, and to the bursar, John Garrod.
Through that year and on subsequent occasions I gave talks,
based on draft chapters of the book, at institutions as far apart
as Tromse and Sydney, San Diego and Sussex. A great many
people made important contributions to the discussions that fol-
lowed, and I have tried to incorporate their suggestions and to
shore up my position against their criticisms in this book.
Among those whose contributions I recall are John Bigelow, An-
drew Brennan, Neil Cooper, George Couvalis, Martin Davies,
John Haldane, Jane Heal, David Hills, Robert Hopkins, Ian
xxi
Acknowledgements
Hunt, Frank Jackson, Philip Kitcher, David Lewis, Paisely Liv-

ingston, Gary Malinas, Hugh Mellor, Graham Nerlich, Philip
Pettit, Charles Pigden, Anthony Price, Georges Rey, Gideon Ro-
sen, Bob Sharp, Elliott Sober, Ivan Soll, Roger Squires, John
Stokes, Scott Sturgeon, Kenneth Taylor, Michael Tooley, Aubrey
Townsend, Kendall Walton, Peter Wetherall, Timothy William-
son, Jamie Wyte and Steven Yalowitz. To Jerrold Levinson I owe
especial thanks for his careful reading of parts of various chap-
ters and for the discussions we have had in Washington, Cam-
bridge, London and Dunedin.
This book was completed during the first half of 1994 at Flin-
ders University, Adelaide. My colleagues in the Philosophy De-
partment deserve thanks for their warm welcome and their
critical acumen. Marty Davies was especially helpful in reading
the penultimate version, making important suggestions for
changes and subediting the whole thing. Vladimir Popescu
helped prepare the index. Three readers for Cambridge Univer-
sity Press made helpful suggestions that I have incorporated.
In writing this book I have tried to bring together certain aspects

of my work in recent years from different areas- mainly from
film and from philosophy of mind. I have therefore incorporated
ideas, and sometimes text, from articles appearing in a variety
of places: parts of Chapter 1 appeared in "Film, Realism and
Illusion", in D. Bordwell and N. Carroll (eds.), Post-Theory, Mad-
ison, University of Wisconsin Press, 1995; parts of Chapter 2 in
"Painting, Photography and Perception", Journal of Aesthetics and
Art Criticism, 1991; parts of Chapter 4 in ''The Long Goodbye",
British Journal of Aesthetics, 1992; parts of Chapter 5 in "Imagi-
nation and Simulation: Aesthetics Meets Cognitive Science", in
M. Davies and T. Stone (eds.), Mental Simulation, Oxford, Black-
well, 1995; parts of Chapter 6 in "Fictions in Visual Media",
Philosophical Quarterly, 1991; parts of Chapter 7 in "McTaggart
at the Movies", Philosophy, 1992; parts of Chapter 8 in "Inter-
pretation and Objectivity", Mind, 1993; and parts of Chapter 9
in "Unreliability Refigured", Journal of Aesthetics and Art Criti-
cism, 1994. I thank the editors of these books and journals for
permission to republish.
xxii
Film, 1895-1995
Film is an art of recent invention. In fact it is just one hundred

years old. We know more about its early history than about that
of any other medium: the technical inventions and discoveries
involved, the intentions of the early pioneers, the reactions of
audiences, the transitions of style and genre, the social forces
which affected film and the impact of film on society. Film's
development is laid out before us with a completeness that his-
torians and theorists of other cultural phenomena must envy.
Yet I believe that the nature of film is less well understood than
that of any other art. If my arguments are correct, current theory
is based on a serious misunderstanding of the film medium and
its effects on the viewer.
The roots of that misunderstanding go back to early writers
like Miinsterberg, who thought that cinema was preeminently a
medium of subjectivity. That error is as strongly entrenched in
film theory today as at any other time. It might have been oth-
erwise; the work of Bazin contained within it the seeds of a
better view. But Bazin's "realism" was overstated, and the re-
action against it was inevitable.
There is too much in Bazin that is confused or simply wrong
for his work to constitute the basis of a theoretical renewal. The
intellectual roots of the present work would be as foreign to
Bazin as they are to most contemporary film theorists. It owes
much, in spirit at least, to the linguistics of Chomsky, and noth-
ing to Saussure; much to contemporary philosophy of mind and
xxiii
Film, 1895-1995
cognitive science, and very little to Freud and his followers. But
the strongest influence on this work is that almost obsessional
concern with realism so distinctive of the best in Australian phi-
losophy.
xxiv
Image and mind
Introduction: the essence of cinema
Tired of the ponderous prescriptivism that dogged much film

aesthetic (and, until recently, much aesthetics in general), film
theorists frequently tell us they have put aside "essentialism".
Why bring it back? My essentialism is not especially prescrip-
tive: no more so than the claim that water is essentially H 0. 2
Films, like samples of water, have something in common that

makes them films rather than something else. It's more than just
being films, and it's more than just family resemblance. It's an
essence, but knowing what that essence is doesn't help much in
figuring out what films are good, typical or paradigmatic.' Since
the subject of this book is film, I ought to say something about
the essence of film. What I say has some bookkeeping signifi-
cance for what follows, but not much intrinsic interest. This is
the boring part; let's get it over with. Alternatively, you can skip
it and come back if and when you need to.
' Oaims about the aesthetics of film based on some supposed essence of the
media are still made. Sometimes they are very strained. Thus Stanley Cavell:
'The most significant films ... will be found to be those that most signifi-
cantly discover and declare the nature of the medium of film ... a feature of
the medium of film ... is film's power of metamorphosis or transfiguration.
In remarriage comedy, this feature ... is expressed as the woman's suffering
creation, which cinematically means the transfiguration of flesh-and-blood
women into projections of themselves on the screen. Hence the obligation in
those films to find some narrative occasion for revealing ... the woman's
body" ("Ugly Duckling, Funny Butterfly", pp. 222-223).
1
Introduction
I. 1 FILM AS REPRESENTATION
Representations come in a variety of kinds: words written and

spoken, gestures established by convention or made up on the
spur of the moment, pictures drawn and painted, carved
or moulded shapes, smoke signals, flashed headlights, sound
recordings, photographs, the projected images of film. Just
about anything we do or use can be a representation, though
some things are richer than others in their representational capa-
cities; they are more apt to allow the communication of complex
meanings, and what cannot be represented cannot be commu-
nicated.
Film is a representational medium; it is a means by which
representations, themselves distinctively cinematic, are pro-
duced and displayed. What distinguishes cinematic represen-
tations from those of other kinds? One thing separating them
from representations of some other kinds is their pictorial char-
acter. One of the peculiarities of recent film theorizing is that it
has been concerned almost entirely with what the theorists re-
gard as the supposedly hidden or "deep structural" features of
film- the supposed codes by which films are "read"- and has
had nothing of significance to say about the matter film so ob-
viously does trade in - pictures. Perhaps the reason is that it is
assumed by the theorists that the idea of a picture needs no
independent explanation, because picturing is just one kind of
linguistic or semiotic representation. I shall argue later that pic-
tures and linguistic signs are different kinds of things. There is
no language of pictures, and pictures have little of theoretical
interest in common with linguistic items. Nor is their represen-
tational function significantly the product of, or mediated by,
convention. That is as true of cinematic pictures as of any other
kind.
Cinematic pictures are not like the static images of painting,
nor are they made in the way that paintings and drawings are
made. These are differences we need to take account of; in Part
I we shall do so. But these differences must not obscure the fact
that film is a pictorial medium; it gives us- exactly- moving
pictures.
2
The essence of cinema
I. 2 FILM AND THE VISIBLE
Films, like other kinds of pictures, are meant to be seen. The

cinema, by which name I denote the medium to which particular
movies belong/ is a visual medium. A work in the cinematic
medium is an artefact with certain properties accessible to sight,
and no one has made appropriate contact with the work unless
she is visually acquainted with these properties. Movies often
have other properties, accessible through senses other than
sight, and appropriate contact with those movies requires ac-
quaintance, through the right sense, with those properties. Au-
ditory properties are the most obvious and widely used, but
there are other less well explored, and perhaps less rewarding,
possibilities: recall Smellorama and Sensurround. But all these,
including sound, are incidental accretions so far as cinema itself
is concerned, because there can be, and in fact there are, works
in the medium which eschew all sensory engagement except the
visual. Visual properties are what cinematic works have to have.
But the same is true of painting, still photography, sculpture,
mime and musicless dance. What, then, distinguishes cinema
from these other media? Typically, a cinematic work involves
projection onto a surface of an image that is capable of giving
the appearance of movement. (Later I will ask whether this ap-
pearance is illusory or not.) But it is unclear whether any of
these standard conditions are essential to the medium. 3 Imagine
• I have no interest in distinguishing movies and films here, and I use those
terms interchangeably, always to refer to particular works, specified or un-
specified, within the medium of cinema. But a distinction can be made, and
might be useful in other, more aesthetically conscious contexts. See, e.g., Ger-
ald Mast, Film/Cinema/Movie.
3
Attempts to identify the essential features of cinema have not been very suc-
cessful. Gerald Mast's idea that projection is the essential feature of the cine-
matic medium is especially off the mark. Television's failure to achieve the
clarity, luminosity and size of the cinema screen is, as Mast concedes (ibid., p.
267), likely to be a temporary restriction that future technology will overcome
without resorting to projection. Nor is it true that projection ensures the past-
ness of the events we are watching (p. 266). And while projection may empha-
size the artificiality of cinema (p. 270), so would many other salient pieces of
technology devised to deliver an image; see text immediately following.
3
Introduction
a device you load film into, then plug directly into your brain
to give you visual sensations exactly like those you would get
if you were watching a conventionally projected film. Isn't that
cinema without the projection of an apparently moving image
on a surface? If so, what is essential to cinema is the visual
experience, irrespective of how it is delivered. Difficulties also
arise when we try to specify the more distant causes of the ex-
perience- the existence of animated film shows that the cause
need not be photographic.4 What, anyway, constitutes a photo-
graphic method? Creation of a series of images by exposure to
light? Expectant parents can testify to the cinematic, or at least
televisual, quality of images produced by ultrasound. What is
so special about light rays?5
But we need not get involved in a debate about what, exactly,
constitutes allowable methods of production and delivery for
the cinema in the ordinary sense of the term. Instead we can
shape a concept of our own to suit our theoretical purpose.
There is a group of artefacts, interesting and worthy of study
because of the problems to which they give rise, which we may
characterize in the following way. They are produced by pho-
tographic means and delivered onto a surface so as to produce,
or be capable of producing, an apparently moving image. Things
of that kind I am going to call movies, films or works of cinema.
As long as it is remembered that the expressions "movie",
"film" and "cinema" have here this quasi-stipulative use, we
can concentrate on the class of entities they name when so used,
and forget verbal issues. What I shall say about cinema in this
sense will be applicable in some degree to the plug-in, sonic,
and other nonstandard forms, as well as to real-life relatives of
film like television; but I leave it to others or another occasion
to work out precisely what degree that is.
Cinema as I define it is an essentially visual medium. Movies
may or may not present audible speech and other sounds as
well as visual representations, and such auditory accompani-
ments may or may not constitute part of the fictional content
• As Gerald Mast points out (ibid., p. 5).

5 For more on light rays and seeing, see Chapter 2.
4
the movie presents (sound, where it occurs, may be "intra-" or
"extradiegetic", in the favoured lingo). The auditory, as well as
the olfactory and the rest, are optional. Where these things do
accompany the visual component we have a work that is more
than purely cinematic; it has the features a thing must have to
be a cinematic work, and it has others besides.
The claim that the auditory is optional needs careful han-
dling; it is not the claim that we could delete sound from films
that possess it without loss to our understanding or appreciation
of them: an obviously false proposition. Nor is it the claim that
the sound possessed by a given film is inessential to it; we need
to distinguish what is optional for the medium and what is
optional for any particular work in the medium. Thus I take
it to be an essential (that is, an obligatory) feature of a film
with diegetic sound that it has diegetic sound; to run Lawrence
of Arabia without the sound would be to fail, in some degree,
to give access to that film. My claim concerns the medium it-
self, not any particular work that exemplifies it. That's just one
example of the difference between what is essential to a thing
which happens to belong to a kind, and what is essential for
belonging to that kind of thing. On one view, a person's origin
is essential to his or her identity; no one lacking the par-
ents Albert had could be Albert. But having Albert's par-
ents is not a condition for membership in the kind human
being.
I am not claiming that we would make better films if sound
were not available, or that films of the very best kind would not
employ sound because "being of the very best kind" means be-
ing an example of pure cinema, which in tum means having no
optional elements. Nor, finally, am I claiming that visual prop-
erties always or usually contribute more to the value of partic-
ular films than do auditory properties (a somewhat more
plausible claim than the others, but still controversial). Those
are all aesthetic claims, and mine is not of that kind; it is the
claim that a work without visual properties would fail, for that
reason, to count as a cinematic work of whatever value. Audi-
tory properties are not like that; they are optional so far as the
medium is concerned.
5
Introduction
1.3 DEGREES OF INESSENTIALNESS

We can take our analysis a little further than a simple distinction
between the optional and the obligatory. Consider three changes
to cinema that have been controversial at one time or another:
sound, considered as something synchronized with the action
and as part of the diegesis, colour and the wide screen. These
developments are not all on the same footing. Wide screen is
optional in that the screen does not have to be- and once was
not- wide, where ''being wide" means having an aspect ratio
of 1:2.35 or something like it. But the basic technical machinery
of cinema requires that there be some ratio of screen dimensions.
The determinable property, having some ratio of width to height, is
an obligatory property of cinema in the sense that a cinematic
work must have that property; but no determinate of that de-
terminable (say a screen ratio of 1:1.33) is obligatory.6 Diegetic
sound is innovative in a quite different sense; having this com-
bination of diegetic sounds is a determinate property optional for
a cinematic work, but so is the determinable property, having
some combination of diegetic sounds, since films don't have to have
any combination of diegetic sounds. Wide screen is merely a
weak option, while sound is a strong one. An option is weak if
some option of that kind must be chosen, and strong if no option
of that kind need be chosen. (Slightly more technically: an op-
tion is weak if it is an optional determinant of a determinable
that is obligatory, strong if it is a determinant of a determinable
that is not.)
This rather scholastic-sounding distinction gives us theoreti-
cal backing for the intuition that the introduction of sound was
a more dramatic change to cinema than was the introduction of
the wide screen- a more dramatic change because sound is a
strong option and wide screen a weak one. Is colour a weak or
a strong option? That depends on how we conceptualize the
relation between colour and monochrome cinematography. If
black and white and shades of grey are colours, then having
6
The meaning of "determinate of a determinable'' should be clear from the
context. Another example: red is a determinate of the determinable colour.
6
colour is an obligatory determinable, while the precise choice of

colour is optional, and so colour is weakly optional. If black and
white are considered not to be colours, then colour is strongly
optional. But the question whether black and white are colours
in the sense relevant here seems to be dangerously verbal. So
the issue of whether colour is a strong or a weak option also
sounds dangerously verbal. The best we can say is that colour
doesn't fall obviously into either of the categories occupied, re-
spectively, by wide screen and sound. If we think of it as falling
between the two, we again match theory with intuition, since
colour seems less innovative than sound and more innovative
than the wide screen.
1.4 THE VISIBLE AND THE PICTORIAL
It is one thing to say, as I have said, that the cinema is a visual

medium; it is another to say, as I have also said, that it is a
pictorial one, that it is a medium that trades essentially in pic-
torial representations. So far I have ignored this distinction, but
it is theoretically crucial. All pictorial representation is visual,
but not all visual representation is pictorial. Subtitles and those
helpfully orienting dates and place names that used to appear
on the screen (and occasionally still do) are visual representa-
tions and not pictorial ones. The subtitles are visual because they
have to be seen, but they are not pictorial; the name ''Vienna"
is not a picture of Vienna- not, anyway, in the sense of picture
I am going to explain more fully in Chapter 3· When the words
"Paris 1939" appear on the screen over an image of the Eiffel
Tower, they serve to inform us about what is fictional- they tell
us that, fictionally, the scene is now Paris in 1939 - but they do
not convey this information pictorially. The reason, briefly, is
this. I have, let us assume, a visual capacity to recognize Vienna.
That is, there are certain things which, if I see them from certain
points of view, let me recognize that what I am seeing is Vienna,
or parts thereof. A painting, a photograph, or a cinematic image
is a picture of Vienna when it is possible to recognize what it
represents by deploying my visual capacity to recognize Vienna
when I see it. But the word ''Vienna", though I may recognize
7
Introduction
it as referring to Vienna, does not represent Vienna pictorially.

The visual capacity to recognize Vienna alone would not enable
me to recognize that "Vienna" represents Vienna. I need, in ad-
dition, some knowledge of English. More on this in Chapter 3·
On the other hand, if the hero is shown reading a letter, and
the camera closes in to show the letters and words on the page
for us to read, there is a sense in which the visual image here
functions pictorially. That is, it functions in one way pictorially,
as well as in another way nonpictorially. The shapes and other
visual properties of the letters on the page inform us pictorially
about something that is part of the fiction, that the character who
wrote the letter formed words with just those shapes and other
visual properties (this might, or might not, be important to the
development of the fiction, but even if it is unimportant it is still
part of the story). The capacity I deploy to recognize the shapes
displayed on the screen is exactly the capacity I would deploy
to recognize the shapes on the page if I were looking at it. Here
the visual image functions pictorially and nonpictorially; picto-
rially to inform us about what is fictional concerning the shapes
and other visual properties of certain written characters, and
nonpictorially to inform us about what the content of the letter
was - what was actually said in it. For no purely visual recog-
nitional capacity will enable me to know the meanings of words.
The conclusion is, then, that among the ways in which visual
images can function to inform us about what is fictional, three
broad categories can be distinguished. There are images or ele-
ments of images which function only pictorially (as an image of
red wallpaper may function pictorially to tell us that it is fic-
tional that events are taking place in a room wallpapered thus,
while not telling us anything about what is fictional). There are
also visual images which function nonpictorially, as with sub-
titles and scene setters. Finally, there are images that function
in both ways, as with the camera-attracting letter, read by a
character.
There is a corresponding distinction to be made for the way
that sound represents in film. If the sound includes the audible
speech of characters, that sound will function in two ways. It
will represent sonic qualities of the fiction - the tone and volume
8
of a character's voice, for example. The capacities I deploy in
order to recognize the tones and volumes of sounds on the
sound track are just the capacities I use to recognize the tones
and volumes of speech directly heard. But sound here will also
function to convey semantic properties of the fiction, the mean-
ings of words and sentences uttered by the characters. In the
first way it functions as the audible analogue of a picture; in the
second way it is analogous to the visual but nonpictorial use of
the image.
When image and sound function to represent visual and au-
ditory features of the fiction itself, let us say that their function
is a sensory one. Our sensory experience of those images and
sounds is crucial to our understanding what is represented; if
the shapes or colours onscreen were different, we would un-
derstand that the shapes and colours of objects represented were
different. And if the sensed quality of the sounds were different,
we would understand the characters to be speaking in a differ-
ent tone of voice. Literary representation is not like that; it does
not matter, for the purpose of conveying what is true in the
story, what exact shapes or colours go to form the words on the
page. Replacing the author's handwriting with print does not
affect what is true in the story. Literary forms like the novel are
not sensory; we may read by looking at the print, but exactly
how it looks doesn't matter.
1.5 FICTIONAL PICTURES

Suppose we have a visual image that functions pictorially. The
image I see onscreen is a pictorial representation of a man. That
image represents a fictional character who appears in the story
the film presents: Harry Lime. Seeing that image tells me all
sorts of things about what is fictional in this story, for example,
that the man called "Harry Lime" has a certain visual appear-
ance. If the image I see onscreen were different in various ways,
it would tell me different things about what is fictional in the
story: that Harry is slimmer or shorter, or has a different facial
expression. What we seem to have here is pictorial representa-
tion of a fictional character.
9
Introduction
There is a problem. Cinematic images certainly are represen-

tations of the people and objects of which they are photographs;
the photograph (or the sequence of photographs that makes up
the film) of Orson Welles represents Orson Welles. It also pic-
torially represents him; the capacity to recognize that this is a
film image of Welles is the capacity to recognize Welles when
and if I see him. But if our concern is with movies like The Third
Man that present narrative fictions (rather than, say, documen-
taries about film stars) we are surely concerned with the rep-
resenting relations that hold between the cinematic images and
the fiction itself, its characters and events. So far as the fiction
goes, we are concerned with Harry Lime rather than with Orson
Welles. But the problem is that, at least in typical cases, there
are no such things as those characters and events, and so there
are no relations, representing or otherwise, between the movie
and fictional things? And so our image is not, after all, a picture
-or any other sort of representation- of Harry Lime. 8
One way to solve the problem would be to deny the obvious
-to insist that Harry Lime does exist (or that he has some other,
positive grade of being), and that the film images represent him
pictorially. Another would be to say that images can pictorially
represent the nonexistent as well as the existent. I shall not at-
tempt a solution of either kind. 9 On the other hand, we should
not give up too easily; the problem posed by the nonexistence
of fictional characters looks like a technical hitch. If cinematic
images are not, literally, pictorial depictions of (nonexistent) fic-
tional things, then whatever representational function they do
7
Sometimes the characters of fiction are real things, as Hitler is a character in
The Last Days of Hitler. There the cinematic images represent (in fact misrep-
resent) him pictorially. But this is not the typical case. There are also cases
hard to classify. Is Erich von Stroheim a character in Sunset Boulevard, played
by himself, a character of whom it is fictional that (among other things) his
name is Max von Mayerling?
8
This problem is not peculiar to film. In any medium that represents pictori-
ally, there are fictional representations: pictures and sculptures of imaginary
beings, plays that tell us about people who never lived and events that never
happened.
• For reasons, see my Nature of Fiction, chapter 4·
10
have seems, intuitively, to be pictorial. Let us examine this idea

of functioning pictorially to present fiction.
The cinematic images in The Third Man pictorially represent
Orson Welles. They do not pictorially, or in any other way, rep-
resent Harry Lime, since there is no such person for them to
represent. But there is such a thing as The Third Man, a certain
fiction of the cinema. Think of a particular sequence of cinematic
images as giving information about that, rather than as repre-
senting the characters of the fiction. The manner in which the
images convey that information is, I shall argue, a pictorial way.
What do we learn about a fiction from the images we see on
screen? We learn what is fictional in that fiction. In a given 10
fiction, various things are fictional, and different fictions are dis-
tinguished, partly, by the different things that are fictional in
them. One difference between The Third Man and, say, Stagecoach
is that in the former but not in the latter, it is fictional that there
is a man whose name is "Harry Lime" who is engaged in var-
ious criminal projects, who has a certain appearance, who moves
in a certain way and who stands in a certain setting on a certain
occasion. How he looks, moves and stands are among his visible
features - or, more precisely, they are among the visible features
concerning which it is fictional that someone has them in this
film. How do we come to know that they are among these fea-
tures? We know that by looking at the screen, and recognizing
there an image of a man (Orson Welles) who looks, moves and
stands in that way. And the recognitional capacity we exercise
in order to do this is just the recognitional capacity we would
exercise if we saw a man in front of us who looked, moved and
stood that way. That, long-windedly, is the story about how our
knowledge that certain things are fictional in the film depends
upon the visual capacities we have to recognize the visible fea-
tures of things when we see them. And that is what it is for the
film's images to function pictorially in their representation of
the story."
'
0
On being fictional (what is sometimes called "being true in fiction") see ibid.,
chapter 2.
" For a somewhat more precise account of picturing, see this volume, Chapter
3, Section 3.1.
11
Introduction
This does not mean that the image is a pictorial representa-
tion; that would require it to be a pictorial representation of
2
something, which it isn't.' It is not, in particular, a pictorial
representation of fictional characters and things. And it is not a
pictorial representation of the fiction itself. A fiction does not
have the kinds of properties - shape, size, colour - that could
be represented pictorially. Saying that cinematic and other im-
ages function pictorially to present fictions is as close as we can
get to saying that they are pictorial representations without say-
ing something false. That is close enough for our purposes. It
will enable us to make all the distinctions we want to make -
especially that between functioning pictorially to present fiction
and functioning linguistically, which is how literature functions.
This will tum out to be a very significant distinction, as I argue
in Chapter 4·
Despite my scepticism about the idea that cinematic images
are pictorial depictions of fictional things, it will be convenient
to speak as if they are, and hence to blur the distinction between
being a pictorial representation of fictional things and function-
ing pictorially to present fictions. From now on I shall simply
speak as if an image represents Harry Lime, and represents him
pictorially as well as visually. Instead of speaking like that, I
could say that the image functions pictorially to present a fiction
in which a man called "Harry Lime" does such and such. But
that would be tedious.
1.6 DOCUMENTARY AND FICTION
I have said that cinematic images, at least the photographically

made ones we are considering here, represent real people in real
settings and situations. Should we say that a movie is always a
documentary as well as being a piece of fiction - a (highly se-
lective) documentary about the activities of actors and film mak-
ers as they go about making a fictional movie? I say no. There
" Goodman's suggestion that, while there are no representations of Mr Pick-
wick, there are "Pickwick representations" strikes me as a purely verbal
solution to the difficulty posed by representation in fiction (see his Languages
of Art, chapter 1).
12
are two representing functions that cinematic images can per-
form: the photographic and the fictional. Documentary films
perform only the first while fiction films perform both. A fiction
film, like a documentary, is a record of what happened in front
of the camera at the time the film was exposed. But with the
fiction film and not with the documentary, that record is in-
tended for the further purpose of presenting a fictional story. A
rough way to indicate the distinction would be to say that a
documentary has as its primary purpose the representation of
the real, whereas the fiction film uses representations of the real
to represent the unreal or the fictional. What the movie really
represents is determined by the causal processes which result in
the exposure of cinematic film, and not by intention.' 3 A film
may (really) represent something quite other than what its mak-
ers intended it to represent, as when we find evidence concern-
ing crimes or unsuspected historical information in casual
holiday footage. But however hard we look at a film, we shall
not find unsuspected fictional stories represented there; it is fic-
tional - presents a fictional story - only if it was intended to be
fictional.
What intention might a film maker have in giving us repre-
sentations of the unreal? The more or less traditional view is
that we are intended to "suspend disbelief" in the presence of
cinematic (and other) fictions. It has never been very clear what
the suspension of disbelief is supposed to be, but one result of
the use of this phrase has been that writers on cinema have been
able to claim that, in various ways and to various degrees, the
viewer comes to believe in the reality of these fictional events.
In Chapter 5, I argue that this is wrong, and that what is dis-
tinctive of the experience of cinematic and other fictions is not
belief but what I shall call imagining. In general, what is pre-
sented as fictional is what we are intended to imagine, and we
engage in the appropriate way with the fiction when what is
fictional and what we imagine coincide. ' 4
3
' For more on the role of causation in photographic representation, see this
volume, Chapter 2, Section 2.4.
'4 See Currie, Nature of Fiction, chapters 1 and 2; and Kendall Walton, Mimesis
as Make-Believe, chapters 1 and 2.
13
Introduction
It is easy to conflate the two representing functions of cine-

matic images - the photographic and the fictional. The costs of
doing so are high; grand theories can collapse when the confla-
tion is recognized. The claim that the essence of cinema is the
representation of "reality as such" cannot survive the recogni-
tion that, with the fiction film, what is intended to be repre-
sented isn't real at all; what is real is just the means towards
that representation.
There are documentary films, just as there are literary docu-

ments; and there are fictional films and fictional works of liter-
ature. ' 5 Gone with the Wind is fictional cinema based on fictional
literature; Citizen Kane is fictional cinema with no such antece-
dents. It is fictional despite its (partial) basis in actuality. But
there is an asymmetry between the cinematic and the literary
case. There is a threefold categorization possible for film that is
not possible for literature. Consider first the literary case.
Someone wishes to inform us of certain events that have oc-
curred, say the British poll-tax riots of 1990, and he sets down
a written description of them to the best of his knowledge; that
is what you do when you write nonfictional literature.'6 But
what does the film maker do, if his aim is the same? He may
do either of two things (unless one of them is excluded by the
passage of time or the prohibitions of distance). He may first
capture those very events on film and present them in the form
of a documentary. Second, he may capture on film a contrived
representation of those events by the artful manipulation of ac-
tors, extras and settings. Perhaps the aim of informing is better
realized, other things being equal, in the first way than in the
second; but the second does seem to be a way of realizing that
aim, however imperfect it may be. In both cases the film maker
5
' Sometimes people speak and write as if fiction and film were contraries. Thus
the subtitle of a recent work announces its subject as "the rhetoric of narra-
tive in fiction and film", thereby imposing, one must presume, a distinction
between the two (Seymour Chahnan, Coming to Terms). But I take it this is
not so; fiction applies as much within film as it does within literature.
6
' Of course you can write deceptive nonfiction, but I exclude that case as not
relevant here and as potentially distracting.
informs us about what he believes to be the case by giving us
cinematic representations. In neither case do we have fiction as
a result. ' 7 But the representations the film maker gives us in
these two ways are otherwise very different. In the first, docu-
mentary, case, what is represented is a matter to be settled by
facts about causation: it is causal connections between rioters
rioting and the camera that makes it the case that the riots are
represented on film. In the second case there are the same kinds
of causal connections at work, but they do not get us to the
representations we are looking for. They connect the film with
actors and sets instead. In this second case the representational
target- the riots- is arrived at by intention; the film represents
those riots because the filming of actors and sets was intended
to provide a representation of those riots.' 8
Is there a comparable distinction, within the class of works
of written literature, between causal and intentional represen-
tations? No. The writer cannot choose between recording real
events and merely producing verbal representations of them;
the latter is the only option available. And written literature is,
essentially, an intentional form of representation, for the proc-
ess of writing is mediated by belief in a way that photographic
representation need not be. Automatic writing, if there is such
a thing, must be intentionally mediated- via the writer's sub-
conscious intentions, or through the magical agency of another
- or it would be the production of meaningless shapes rather
than writing. An accidental photograph, on the other hand, can
easily occur (there is more on photography and causation in
Chapter 2).
So cinema presents us with a threefold distinction where lit-
7
' Assuming that, in the case in which the events are restaged with actors, the
intention is to convey only that which is factual. Most "dramatic recreations"
on film are not like this and contain much that is invented and which will
be understood as such by the audience. Films of this kind are fictions; at
least, they contain fictional elements.
8
' There might be more to it than intention; we might need a condition speci-
fying that the intention has been carried through in a reasonably successful
way. The point is that intention is relevant in this case but not in the docu-
mentary case; the documentary film maker might have been thoroughly con-
fused about what he was filming.
15
Introduction
erature allows only two possibilities. In cinema we may have
events and characters presented not as fact but as the material
for make-believe, we may have actors and sets artfully contrived
to inform us of actual events or we may have straightforward
documentary film. Within the realm of the literary, we have dis-
tinctions corresponding to only the first two of these three.
So far, we have seen some of the things that are essential to

cinema, and some of the things which are, to varying degrees,
inessential. We have made distinctions between the visual and
the pictorial, between the pictorial and the linguistic, between
documentary and fiction and between the kinds of nonfictional
aspirations which can be realized in film and those which can
be realized in literature. Distinctions are useful, but their use-
fulness is always relative to the theory they serve. What we have
so far are distinctions in search of a theory, and it is time to
begin to construct that theory.
16
Part I
Representation in film
F
ILM is said to be in various ways a realistic medium
of representation. I distinguish three kinds of realism
that might be attributed to film. Only one of them, I
shall argue, is correctly attributable. I also examine, and find
wanting, the idea that filmic representation is language-like
in structure.
Chapter 1
The myth of illusion

Film gives simultaneously the effect of an actual hap-
pening and of a picture.
Rudolf Arnheim
Arnheim is just one of many, now as in his own day, who say
that film creates illusion. Like others, he qualifies the claim, say-
ing the illusion is partial. I say film does no such thing; it creates
no illusion, partial or otherwise. Film has considerable powers
to engage and to persuade, but these powers are not accounted
for in terms of illusion. I'll argue in Part II that they are ac-
counted for in terms of imagination. In this chapter we shall see
that there are different kinds of illusions, and that claims that
film engenders illusion can be more or less plausible depending
on what kind of illusion is in question. I'll distinguish two kinds
of illusions film might be said to engender, concluding that it
engenders neither of them. Before I get to that, I need to distin-
guish claims about the illusory nature of film from other claims
which are sometimes made.
1.1 TRANSPARENCY, LIKENESS AND

ILLUSIONISM
Let us start by distinguishing three doctrines about cinema; all

of them have been called "realism", and each is to some degree
deserving of the name. They are, however, quite distinct, and to
underline their distinctness I shall give each a different title.
There is first the claim that film, where it uses the photo-
graphic method, reproduces rather than merely represents the
real world, because photographs capture objects themselves
19
rather than likenesses or representations of them. This view is
associated most notably with Andre Bazin. Following Kendall
Walton, I shall call this the doctrine of Transparency: film is trans-
parent in that we see "through" it to the real world, as we see
through a window or a lens.' Next is the idea that the experience
of film watching approximates the normal experience of per-
ceiving the real world. I shall call this Likeness; it says the ex-
perience of film is, or can be, like the experience of the real
world. This doctrine has been asserted, again by Bazin, in con-
2
nection with long-take, deep-focus style. But as I shall argue,

this kind of realism is a matter of degree, and long-take style is
merely more realistic in this sense than is, say, montage style.
Finally, there is the claim that film is realistic in its capacity to
engender in the viewer an illusion of the reality and presentness
of fictional characters and events portrayed. I shall call this view
Illusionism. It is, as I said, widely believed. Studio publicity writ-
ers, Marxist critics of the Hollywood film, as well as more con-
ventional figures like Arnheim, constantly assert it.
Much of the history of film and film theory can be recon-
structed as a debate about the relations between these three doc-
trines. Some theorists - I am thinking of the early montagists -
have argued that Transparency requires us to play down the
likeness film makes possible; film's mechanistic commitment to
reality has to be compensated for by visual styles that give us
visual experiences unlike our visual experiences of the real
world. Others, like Bazin, say that Transparency requires the film
maker to exploit the possibilities for reproducing in film our
visual experience of the world; film presents the real world, so
it should do so in a way which approximates as closely as pos-
' See Kendall Walton, "Transparent Pictures". "Transparency" is used in other

senses by other authors. It sometimes refers to a certain kind of supposedly
invisible narration (see, e.g., Edward Branigan, Narrative Comprehension and
Film, p. 84). The reader should be careful to distinguish my (Waltonian) sense
from these others.
• So "Likeness", with an initial capital, is the name of a thesis, namely the thesis
that the experience of film is, or can be, like that of the real world, while
"likeness" is here used in its ordinary senses: (i) the quality of being like
(something); and (ii) something which is like something (as in "a likeness").
20
sible our experience of that world. And many theorists have

agreed that Likeness makes for Illusionism; the closer the ex-
perience of film watching approximates the experience of seeing
the real world, the more effectively film engenders in the viewer
the illusion that he or she is watching reality. What is not agreed
is whether this is a desirable goal. Other theorists have argued
that the very notions of likeness, verisimilitude and realism in
film are suspect or even incoherent.
Whatever disagreements there have been between particular
theorists on these issues, there is general agreement about the
close connection between the three doctrines: Transparency,
Likeness and Illusionism. I, on the other hand, take the three
doctrines to be independent from one another. Adopting any
one of them, we are free to adopt or reject the others. In this
first part of the book I tentatively reject Transparency, firmly
reject Illusionism and accept Likeness.
Transparency and Likeness are doctrines I discuss later.
About Illusionism, which is the subject of this chapter, I argue
two things. First, I shall argue that it is a mistaken doctrine.
Second, I want to distinguish between two versions of illusion-
ism: the view that film creates an illusion of the presentness and
reality of the fictional events it portrays and the view that the
basic mechanism of film creates an illusion of movement. So
there are, as I indicated at the beginning of this chapter, two
doctrines of illusionism. Sorting them out will require a distinc-
tion between cognitive and perceptual illusions.
There is another kind of realism about film which has been his-
torically important and which I shall not discuss: the view that
films can be placed along a dimension of realism according to
whether and to what extent they represent deeply significant
social relations. 3 A fictional film might be said to be realistic in
this sense because it portrays fictional characters as standing in
social relations important for determining the outcomes of in-
terpersonal interactions in real life, and portrays the outcomes
of the characters' interactions as (largely) determined by their
3
Here I am grateful to George Couvalis.
21
standing in those relations. This seems to me a perfectly intel-
ligible sense of realism. The question of whether there are such
relations and, if so, what they are, is an important- perhaps the
most important- question of social theory. I leave that question
to be settled by social theorists, whose skills and knowledge are
quite different from those typical of film theorists and philoso-
phers.
1.2 COGNITIVE ILLUSIONS

The claim that film causes cognitive illusions is this: film watch-
ing, in some systematic way, and as part of the normal process
of the viewer's engagement, causes the viewer to have the false
belief that the fictional characters and events represented are
real. 4 Strictly speaking, there are weaker and stronger versions
of the thesis. A weak version says merely that the film viewer
comes to believe that what is represented onscreen is real. A
stronger version also asserts that the viewer believes that he or
she is watching real events. On the weak view, what is happen-
ing onscreen is taken for what it really is, namely a represen-
tation, and the illusion comes in when we say the viewer takes
it to be a representation of real events. On this view, film func-
tions in a way not essentially different from, say, the novel. Il-
lusionism about the novel says that the reader takes the events
described in the novel to be real, but no one would seriously
assert that readers think that the words on the pages they read
are those events. As far as I can see, most people who think film
is illusionistic take the stronger view. They think film has the
4
"In cinema it is perfectly possible to believe that a man can fly" Oohn Ellis,
Visible Fictions, p. 40); "One knows that one is watching a film, but one be-
lieves, even so, that it is an imaginary [sic] reality" (Maureen Turim, Flashbacks
in Film, p. 17); "Conditions of screening and narrative conventions give the
spectator an illusion of looking in on a private world" (Laura Mulvey, "Visual
Pleasure and Narrative Cinema", p. 8o6); 'The camera becomes the mecha-
nism for producing an illusion of Renaissance space" (ibid., p. 816). This view
is by now more or less standard; see, e.g., Robert B. Ray, A Certain Tendency
of the Hollywood Cinema, p. 38.
22
capacity to make the viewer think that he or she is actually

watching real events.
Perhaps there is a sense in which the film viewer is watching
real events. If the thesis of Transparency is correct, then when I
watch a film which stars Cary Grant and Ingrid Bergman, I re-
ally see those people; I don't merely see representations of them
as I would if I was looking at a painting or an animated cartoon
which had Bergman and Grant as its subject. But Transparency
is something I take up in the next chapter, since it stands to one
side of my present concern. The strong version of Illusionism
just characterized is not the claim that we see, and believe we
see, real actors on sets and locations acting out parts. If Trans-
parency is correct, there would be no illusion involved in having
that belief, since the belief would be true. The strong version of
Illusionism asserts, rather, that when I see the film starring Berg-
man and Grant, I believe I am watching the fictional events
which the film presents: the activities of, say, U.S. counter-
espionage agents coming to grips with a postwar Nazi plot in
South America.
Illusionism is rarely spelled out in the stark terms I have just
employed. Various qualifications are sometimes made, some of
which I shall consider. But the strong view sits comfortably with
much else film theorists say, particularly concerning the role of
the camera and the viewer's act of identification. The standard
theory seems to be something like this: the illusion peculiar to
film is that the viewer is present at the events of the story,
watching from the position actually occupied by the camera,
which the viewer thinks of as his or her position. Thus Bela
Balazs: "In the cinema the camera carries the spectator into the
film picture itself. We see everything from the inside as it were
and are surrounded by the characters of the film." 5 And Erwin
Panofsky: "Aesthetically [the spectator] is in permanent motion
as his eye identifies itself with the lens of the camera." 6
5 Bela Balazs, Theory of the Film, p. 48.
6
Erwin Panofsky, "Style and Medium in the Motion Pictures", p. 218. See also
Paul Weiss: "A film is completed by the viewers before it; they are trans-
formed by the film into occupants of a world part of which the film makes
visible" (Cinematics, p. 5; quoted in Ian Jarvie, Philosophy of the Film, pp. 131-
23
I do not deny that it is possible for film to engender this sort
of an illusion on the part of a viewer. On a liberal enough view
of possibility, it is possible for anything to create an illusion of
anything else. But this mere possibility is not at issue when peo-
ple claim that film is illusionistic; rather, they claim that the
standard mechanism by which film engages the audience is il-
lusionistic, that the creation of an illusion of reality is a standard
feature of the transaction between film and viewer. That is what
I deny.
There are two serious objections to the idea that film induces
the illusion that fictional events are real and that the viewer is
directly witnessing them. The first is that film viewers simply
do not react in the way that people would react who believed
in the reality of the fictional events the film depicts? You have
only to reflect for a moment on how you would react if you
saw, or thought you saw, a threatening monster, or if you
thought yourself alone in a house with an axe murderer, or if
you thought you were watching someone about to be attacked
by an axe murderer, to see that your behaviour in the cinema is
quite unlike that of someone who really did believe in the reality
of the fiction presented.
There are celebrated cases of viewers reacting to a movie as
if they were in the presence of the thing it depicts, though
whether these cases belong to the history or merely to the folk-
lore of cinema I do not know. It is said, for example, that in 1895
a Parisian audience fled in terror during the showing of a film
by the Lumiere brothers which depicted the arrival of a train. 8
If it really happened, this is to be explained in terms of the
unfamiliarity of the medium to that audience; it sheds no light
on our standard and intended response to cinematic fictions.
132). In Chapter 4 I shall examine a way of taking the statements of Balazs,

Panofsky and others which does not commit them to illusionism.
7
This is well argued in Walton, Mimesis as Make-Believe, chapter 5· Films may
be intended to promote belief at a secondary level, as concerninSs for example,
the rightness of a cause, and they may in fact achieve this result. But it is not
usually the case that they are intended to get the audience to believe in the
reality of the fictional characters and events they depict.
8
See, e.g., Metz, Imaginary Signifier, p. 73·
The case is also not an example of the audience reacting to fic-

tional events as if they were real; the film really showed the
arrival of a train. Examples of this kind cannot be used to sup-
port any general hypothesis about the deceptive powers of cin-
ema, especially when it is fictional cinema at issue.
The reactions I have described as lacking on the part of the
normal film viewer are reactions of behaviour; I said that people
watching movies do not behave like people who believe in the
reality of the fictions they are watching. This might be objected
to on the grounds that this is behaviourism, a discredited doc-
trine; but in fact, the argument does not depend on behaviour-
ism. Behaviourism is the view that mental states like belief are
items of behaviour, or, in a more sophisticated version, that they
are dispositions to behave. You do not have to accept this be-
haviourist identification to think that behaviour is relevant to
the question of whether someone is in a certain mental state.
When I make a claim about a person's mental state it is relevant
to raise questions about behaviour, because behaviour is an im-
portant source of evidence for an hypothesis about a mental
state. We are entitled to be sceptical of the claim that a person
believes in God if the person never engages in any behaviour
we think of as typical or exemplary of a belief in God. Similarly,
there is very little evidence from behaviour that film viewers
believe in the fictions they see. Explanations of our responses to
cinematic fictions in terms of belief work only so long as we do
not take the connection between belief and behaviour seriously.
Of course we do need a psychological explanation of our
sometimes very intense responses to film. In the absence of an
explanation in other terms, an explanation that appeals to belief
can seem attractive, for all its evident drawbacks. In Part II I
shall give an explanation of our responses to fictions in terms
of imagining.
So the first objection to Illusionism cites the evidence from
behaviour. The second objection cites the evidence of introspec-
tion. Illusionism is at odds with much of the experience of film
watching. Consider what would be involved in the film viewer
believing that she was watching real events. The viewer would
have to suppose that her perspective is that of the camera, that
25
she is positioned within and moves through the film space as
the camera is positioned and moved. There have been elaborate
attempts to argue that the viewer does identify with the camera. 9
But identification with the camera would frequently require us
to think of ourselves in peculiar or impossible locations, under-
taking movements out of keeping with the natural limitations
of our bodies, and peculiarly invisible to the characters. None
of this seems to be part of the ordinary experience of film watch-
ing. In the attempt to associate the camera with some observer
within the world of the action with whom the viewer can in
turn identify, film theorists have exaggerated the extent to which
shots within a film can be thought of as point-of-view shots, and
have sometimes postulated, quite ad hoc, an invisible narrator
from whose position the action is displayed and with whom the
viewer may identify. It would be better to acknowledge that
cinematic shots are only rarely from a psychological point of
view, and to abandon the thesis that the viewer identifies with
an intelligence whose point of view is the camera.'o
I have already said that film theorists sometimes try to dis-
tance themselves from the strong version of Illusionism I have
described by adding various qualifications. Sometimes the claim
that the viewer believes the fiction to be real is qualified so as
to suggest that the belief is partial or of lower intensity than
those we get from reallife. This move solves nothing; the film
11
viewer does not behave like one who merely suspects, or be-
lieves to some degree, that he or anyone else is in mortal danger,
or like one within whom belief is in tension with simultaneous
disbelief. If I even vaguely suspected there was a monster on
the loose I would leave the theatre immediately and call the
9
"[The viewer] certainly has to identify ... if he did not the film would be-
come incomprehensible" (Metz, Imaginary Signifier, p. 46). See also Nick
Browne, The Rhetoric of Filmic Narration, chapter 1.
•o For more on this see Chapter 7, Section 7.2.
" Christian Metz, for example, discusses how film creates "a certain degree of
belief in the reality of an imaginary world". He also says that "somewhere
in oneself one believes that [the events of the fiction] are genuinely true"
(Imaginary Signifier, pp. 118 and 72, emphasis in the original). Jean Comolli
says that "we want ... to be both fooled and not fooled [by cinema]" ("Ma-
chines of the Visible", p. 759).
police. Nor is there in the experience of film watching any evi-

dence that we suspect, or partially believe, or are prone to be-
lieve, that we are flying through space or suspended from a
ceiling.
1. 3 FILM AND DREAMING

There is a version of Illusionism which might seem to avoid the
problems I have just now raised. It is the view that the mental
state induced by film is like the state of dreaming. Dreams in-
volve fantastical occurrences which we accept uncritically while
we dream, and which don't provoke much reactive behaviour;
moreover, dreaming seems to have a visual quality in the way
that watching cinema does - though the relation between
dreaming and mental images is in fact more controversial than
many writers on film seem to appreciate. We frequently en-
counter confident statements to the effect that, in watching films,
the viewer approximates the condition of the dreamer, and that
12
it is in fact the aim of cinema to induce this condition. In that
case the camera would correspond to a supposed "inner eye"
by means of which we perceive the images of dreams.
The analogy with dreaming has been a powerful stimulus to
the development of psychoanalytic theories of film and film ex-
perience. In fact, as Noel Carroll has shown, the analogy pro-
ceeds by systematically failing to compare like with like. If film
experience is to be like dreaming, it is what happens in the
dream that matters, and in our dreams we are certainly moti-
vated to avoid danger, though, notoriously, the dream usually
makes this frighteningly difficult. Few of us feel the same des-
perate urge to act when watching film, unless it be simply to
" Thus Sally Flitterman: 'The look ... is an integral part of filmic structure.
The cinematic apparatus is designed to produce the look and to create in the
spectator the sensation that it is she/he who is producing the look, dreaming
these images which appear on the screen" ("Woman, Desire, and the Look:
Feminism and the Enunciative Apparatus in Cinema", p. 24 J). Christian Metz
emphasizes a number of differences between film and dream, but concludes
that "among the different regimes of waking, the filmic state is one of those
least unlike sleep and dreaming, dreamful sleep" (Imaginary Signifier, p. 128).
leave the theatre. Another disanalogy: in dreams the central
character is typically the self, whose acts and sufferings are of
central concern. But film watching is notable for its capacity to
suppress consciousness of the self in favour of the fiction, and
even those who claim that film "puts us in the space of the
action" would not, I suppose, claim that we imagine ourselves
to be active participants in the events portrayed. That both
dreaming and film watching frequently take place in the dark
is another irrelevant consideration; darkness is not typically part
of the experience of dreaming, though it does typically accom-
pany the experience of film watching. Besides, if people were
reduced to the condition of dreamers by film, they would not
be the noisy conversationalists I often find so irritating.
Perhaps in some way film watching is like dreaming; perhaps
everything is in some way like everything else. There does not
seem to be any substantial, systematic likeness between film ex-
perience and dreaming that holds out promise of serious ex-
planatory gains.
1.4 PERCEPTUAL ILLUSIONISM

The version of Illusionism I have been considering so far is very
strong; it commits the Illusionist to saying that film viewers are
systematically caused to have false beliefs. It is interesting to
argue that a strong view is true, but less interesting to argue
that it is false, as I have done. By definition, strong views are
more likely to be false than weak ones. Might there be some
weaker, apparently more plausible, version of Illusionism? If
there is, and if I can show that this weaker version is also wrong,
that would be a result of some interest. I believe there is a
weaker, more plausible version of Illusionism, and I shall argue
against it.
I called the view that film induces false beliefs Cognitive Il-
lusionism. There certainly are cognitive illusions of various
kinds. When you see two lines of equal length, but provided
with "arrowheads" pointing in different directions so that the
lines seem to be of different lengths (as in the so-called Muller-
Lyre illusion), you might believe the lines to be of different
28
lengths. You are then suffering a cognitive illusion. But not all
illusions are cognitive. You may know the two lines are of equal
length and still be subject to the Muller-Lyre illusion: the lines
just look as if they are of different lengths. It is a common feature
of the many kinds of visual illusions that they are, in Zenon
Pylyshyn's phrase, "cognitively impenetrable": belief doesn't
make any difference to the way the illusory phenomenon
looks. 13
An illusion of this kind, which is what I am going to call a
perceptual illusion, occurs when experience represents the world
as being a certain way, when in fact it is not that way and the
subject does not believe it to be that way. My experience of the
two lines may represent the two lines as being of unequal
length, even though I know this experience misrepresents the
relation between the lines.' 4
Someone might claim that cinema is illusionistic in this per-
ceptual sense and not in the cognitive sense I have been consid-
ering up until now. My arguments so far presented against
Illusionism are ineffective against Perceptual Illusionism be-
cause they are designed to show we lack the beliefs necessary
to underwrite the claim of Cognitive Illusionism. I need different
arguments if I am going to oppose Perceptual Illusionism. I shall
'
3
See Zenon Pylyshyn, "Computation and Cognition: Issues in the Foundations
of Cognitive Science", pp. 111-132. See also Jerry Fodor, "Observation Re-
considered". This way of setting up the debate over illusion in film- in terms
of a distinction between cognitive and perceptual illusions - will do for our
purposes. Some writers are sceptical about the distinction, claiming that our
perceptual systems are cognitively penetrable all the way down to the sen-
sory periphery (see, e.g., Paul M. Churchland, "Perceptual Plasticity and The-
oretical Neutrality: A Reply to Jerry Fodor", especially pp. t83-185). The
most such arguments show, I believe, is that there is a continuum between
highly cognitive mental systems within which beliefs are formed on the basis
of other beliefs, and only marginally cognitive systems over which beliefs
can exert some slight influence. What I am here calling the cognitive and
perceptual illusions supposedly created by film would then correspond to
opposite extremes of this spectrum. The arguments I am considering here
are robust under shifts of framework from the dichotomy model to the con-
tinuum model. The continuum model is well argued in W. M. Davies, Ex-
perience and Content: Consequences of a Continuum Theory.
'
4
See Christopher Peacocke, Sense and Content, pp. 5-6.
29
provide some. But bear in mind a point already made: Percep-

tual Illusionism is a distinctly weaker thesis about cinema than
Cognitive Illusionism, and one with quite different conse-
quences. There has been a tendency to assume that the (alleged)
truth of Perceptual Illusionism somehow supports the claim of
Cognitive Illusionism. Perhaps Perceptual Illusionism is the Tro-
jan horse by means of which advocates of Cognitive Illusionism
about film hope to gain their victory.'5 It needs to be said,
therefore, that Perceptual Illusionism, even if true, does not in
itself provide an argument for Cognitive Illusionism, though it
might do so in conjunction with other premises. It is not essen-
tial to my opposition to Cognitive Illusionism that I oppose Per-
ceptual Illusionism as well. Nonetheless, I do oppose Perceptual
Illusionism. I do not claim to be able to refute it; at most I shall
sow the seeds of doubt about it.
The most common version of the claim that cinema induces
a perceptual illusion says that cinema induces a perceptual il-
lusion of movement - that we seem to see movement on the
screen, but really do not. I shall come to this. Before I do, I want
to consider a worry some might have, not just about the reality
of movement within a cinematic image, but about the reality of
the cinematic image itself.
1.5 THE REALITY OF CINEMATIC

IMAGES
I said in the Introduction that film is a pictorial medium.

Someone might question whether the pictures which film em-
ploys are themselves real. There does seem to be a difference
between the substantial pictures we make contact with when we
look at a painting or a photographic print, and the insubstantial
pictures of film - the images on a screen. With painting, the
picture we see is an enduring physical object. The medium of
painting is paint on surface, and when we look at a painting
what we look at is, exactly, paint on a surface. With film, there
5
' See, e.g., Jean-Louis Baudry, "Ideological Effects of the Basic Cinemato-
graphic Apparatus".
30
is a disparity between qualities of the physical material - the
strip of celluloid that passes through the projector - and the
qualities we perceive when watching the movie. In cinema we
watch images. But these images are not the strips of celluloid
themselves. Someone might say, "The celluloid and all the rest
of the apparatus is real, but the image is not; it is a product of
the mind." I want to argue that this dichotomy is false. There is
a sense in which the image we see on the screen is a product of
the mind, but that is not a sense which would justify our saying
that the image is unreal. What we see when we view a film is
a pattern of colour on a surface, usually a screen. That pattern
of colour is, I claim, really there. (For present purposes, I con-
sider black, white and shades of grey to be colours.)
Some philosophers think generally that colours are "not really
there" on a surface, and that when our experience represents an
object or surface as having a certain colour we are thereby sub-
ject to an illusion, because there are no colours to be there.' 6 It
would be a distraction to get deeply involved in this dispute
here; we are after all trying to see whether there are special prob-
lems associated with the idea of cinematic images. We could
proceed simply by ignoring ontological issues about colour al-
together and simply assume that colours, whatever they are, are
real things. However, it will be useful to say something general
about the nature of colour, for what I say about that will be
helpful when we consider the nature of cinematic movement. I
shall argue that cinematic images are real in just the sense col-
ours are real, so we had better establish what this sense is.
I take colours to be properties of surfaces, but properties those
surfaces have in virtue of their being a standard or normal pat-
tern of response to those surfaces on the part of sentient observ-
ers. Colours are, in Mark Johnston's phrase, response-dependent
properties, and in this they differ from properties like being
square, being tall, and being ten miles from Cambridge.' 7 These
6
' See Paul A. Boghossian and J. David Velleman, "Colour as a Secondary Qual-
ity". See also C. L. Hardin, "Color Subjectivism".
'
7
See Mark Johnston, "Dispositional Theories of Value". See also Philip Pettit,
"Realism and Response-Dependence". I am especially indebted to Pettit's
discussion.
31
last three properties are not response dependent because they
are not possessed by an object in virtue of that object's capacity
to elicit a certain psychological response. Of these three prop-
erties, the first is an intrinsic property: to tell whether something
is square you need look no further than at the thing itself. The
last two are extrinsic: whether you are tall depends on the height
of things other than yourself, and how far you are from Cam-
bridge cannot be discerned without taking Cambridge into ac-
count. All response-dependent properties are extrinsic but the
converse is not true, since being tall and being ten miles from
Cambridge are not response-dependent properties, though they
are extrinsic. Being funny is a response-dependent concept and
it is, consequently, extrinsic; things are funny according to how
people react to them, though it is not easy to say exactly what
reactions count as grounds for saying the thing reacted to is
funny.
We can now see how colour properties are different from
many other kinds of properties, while admitting them as real
properties of things. There is some tendency to express the pe-
culiarity of colours by saying they are less real than other prop-
erties, but I find it difficult to make sense of the idea that reality
admits of degrees. Better to put it in terms of a difference be-
tween that which is intrinsic and that which is in various ways
extrinsic. If people grow I shall no longer be tall, though my
height does not change; and if a comet affects our organs of sight
so that things look differently coloured to all of us, they will be
differently coloured though they do not change internally. Being
tall and being red are extrinsic properties of things, whereas
being six feet tall, and having such and such a constitution of
molecules on your surface, are intrinsic properties; changing
those properties would involve a change in you. But extrinsic
properties can be as real as intrinsic ones.
In order to establish the reality of cinematic images, it is not
sufficient to establish the reality of colours in general, for cine-
matic images might constitute a special case where there appear
to be colours but really are not. And there does seem to be a
difference between the image on the cinema screen and that on
32
the painting. The conditions for the presence of colours in the

two cases are rather different, and significantly more precarious
in the case of the screen. The colours of the picture are those it
has for any normally sighted person in conditions roughly ap-
proximating daylight. In our environment, those conditions fre-
quently hold. But what is required for the screen to have that
pattern of colours we identify as a cinematic image is for the
film strip, projector and light source to stand in the right, rather
complex relation to the screen, and that is the case only when
we go to considerable trouble to make it so. These requirements
make the colours of the screen more extrinsic than the colours
of the painting, for they depend not just on our capacity as ob-
servers to respond to them in a certain way, but on the presence
and activation of all the relevant cinematic technology. The
colours on the screen are, we may say, extrinsically sustained,
while those of the painting are intrinsically sustained - what
sustains them is the condition of the canvas itself. Once again,
we should not treat this as grounds for denying that the colours
we see are colours on the screen. All colour is extrinsic to the
extent of being response dependent. There is no such thing as
absolutely intrinsic colour. It is just that the colours on the screen
are extrinsic to a greater degree than are the colours on the can-
vas.
Cinematic images are pictures of a peculiarly transitory and
precarious kind. Their precariousness makes for a further dis-
tinction between them and what we conventionally call pictures.
For a painting or a drawing, we can make a distinction between
the work itself and the intended, standard or ideal viewing con-
ditions for the work. The painting may best be viewed from
about four feet, at eye level, in good diffused light; but the same
picture, with the same colours in the same pattern on its surface,
is still there when viewed from twenty yards in poor light or in
a darkened vault. But screening conditions for a film are to some
extent essential to it. If they are unfavourable to a considerable
degree - as when there is strong ambient light or a very poorly
reflecting surface as a screen- the cinematic image simply is not
there at all, though there is no sharp boundary between accept-
33
able and nonacceptable conditions. My claim is just that, when
those peculiar conditions necessary to display the film hold, the
film image is really there.
1.6 ON THE MOTION OF IMAGES

Let us now introduce the idea of movement, which is so crucial
to understanding the nature and reality of cinematic images.
There is a widely held belief that this movement is illusory.
Speaking of the appearance of movement in film, Hugo Miin-
sterberg said, "We do not see the objective reality, but a product
of our own mind which binds the pictures together."' 8 And
Francis Sparshott writes that "a film is a series of motionless
images projected onto a screen so fast as to create in the mind
of anyone watching the screen an impression of continuous mo-
tion" - an impression Sparshott goes on to call "the basic illu-
sion of motion" .'9 I believe this plausible view to be false. There
is no illusion of movement in cinema; there is real movement,
really perceived.20
Before I argue for this, I want to note that an apparently at-
tractive position here is really not available. We cannot say,
"There really are cinematic images, but there is not really any
movement involved when such an image is projected on the
screen." If you reject the reality of movement, you must reject
the reality of the image as well. Consider an image which lasts
for a few seconds and where there is, apparently, a good deal
8
' Hugo Miinsterberg, "The Means of the Photoplay", p. 332.
9
' Francis Sparshott, "Basic Film Aesthetics", p. 284. See also Haig Khatcha-
dourian, "Remarks on the 'Cinematic/Uncinematic Distinction in Film Art"',
p. 134: "a film ... is necessarily a sequence of visual images that create the
illusion of movement" (emphasis in the original). Jacques Aumont and his
coauthors speak of "the simple illusion of movement", concluding that "psy-
chologically speaking, a film does not exist on film stock or on the screen,
but only within the mind" (Aesthetics of Film, p. 184).
20
My argument here is presented in terms of the case of cinematic represen-
tation of movement, but could easily be altered to apply against those who
deny that there is real motion observed on a television screen or video dis-
play, where the technical mechanism productive of the appearance of move-
mentis different from that in the cinematic case.
34
of movement, perhaps that of a crowd of people running about.
Take away the movement and what is left that could constitute
a real image? If there isn't any motion, the real image must be
static. But what then is it an image of? Nothing, we will assume,
was stationary in the photographed scene. The image can hardly
be of people not moving. Is the image an undifferentiated blank,
with no representational features? Then it is not an image. You
might claim that, when there isn't even any apparent movement
in the image (a fixed closeup of a static object, for example),
then the image is real. But it is hardly credible that when I look
at a static cinematic image I am not subject to an illusion, but
become subject to one as soon as movement appears. Also, since
as a matter of fact few cinematic images are static, there would
be little comfort for the realist in the claim that static cinematic
images are real. To be a realist about cinematic images, you have
to be a realist about cinematic movement.
I am taking the view that cinematic motion is illusory to be
a version of the view that cinema involves a perceptual rather
than a cognitive illusion. Someone might argue that this sup-
posed illusion of cinematic motion is a cognitive, and not merely
a perceptual illusion, because most people who watch films ac-
tually believe they are watching moving images; it is only when
you reflect on the technical mechanisms of cinema that you re-
alize this is not the case. That may be true, but the fact is that
the appearance of cinematic motion does not go away for those
people who convince themselves that it is, indeed, an illusion.
If cinematic motion is illusory, then it is essentially a perceptual
illusion and only incidentally a cognitive one. That is why I shall
treat it simply as a (putative) perceptual illusion.
Before I consider the arguments about the reality of move-
ment in cinematic images, I had better clarify exactly what I
mean when I speak of "moving images". Strictly speaking, the
cinematic image is the whole area of illumination on the screen
during projection. We all agree, I take it, that this does not move,
unless, due to some mechanical failure, the projection equip-
ment starts to shift around. What moves when there is move-
ment of an image is a part or parts of this image; if we are
watching a shot of a man walking along a street, the part of the
35
image which represents the man will move from one side of the
screen to the other. Movement of this kind, which is what I am
concerned with here, needs to be distinguished from the move-
ment which occurs as a result of a continuous change in the
position of the camera during a single shot; the man might be
stationary relative to the background while the camera moves
with respect to him. This latter kind of movement introduces
somewhat complex considerations which I shall not attempt to
deal with here. Also, the movement with which I am concerned
here is not the radically discontinuous movement which might
be said to occur across shots, for example, when we see the
image of the man in one place on the screen in one shot, and in
another place on the screen in the next shot. All I am claiming
here is that there really is movement within a single shot taken
from a fixed perspective. That, obviously, is enough to contra-
dict the claim that movement in film is an illusion produced by
the juxtaposition of static images.
I had also better say a word on metaphysical background.
Arguments about motion, and about change generally,
sometimes raise deep questions about what motion and change
actually are. There are two basic and mutually incompatible
views about this. One, which I shall call three dimensionalism,
says that change takes place when a thing has a property at one
time which it, that very same thing, lacks at another. The other
view, four dimensionalism, says that change occurs when a cer-
tain temporal stage possesses a property and another temporal
stage lacks it, where those stages constitute temporal stages of
the same object. I don't believe there is anything in our common
belief about change which decides one way or the other between
these two theories, and nothing I shall say about cinematic
movement here is intended to prejudge which view is correcC 1
So while I shall speak of our cinematic experience as represent-

ing to us that an image moves from one place to the other, this
is to be taken as neutral between the view that there is one thing
which is in one place at one time and in another place at another
" See Frank Jackson, "Metaphysics by Possible Cases".

time, and the view that there are distinct but suitably related
temporal parts in different places. Both constructs I take to be
inconsistent with the view that there is, literally, no movement
of an image on the screen.
One way to argue that there is real movement of cinematic
images would be to adopt very liberal criteria of reality. In par-
ticular, if we could persuade ourselves that there is no clear
distinction between what it is useful to say and what it is true
to say, it would be easy to establish the reality of cinematic
movement. There is, after all, utility in describing a film by ref-
erence to the movement of the images it presents us with. Daniel
Dennett has recently advocated a kinder, gentler realism that
allows us to say that all sorts of things are real on account of
their usefulness. In Dennett's example, Smith and Jones claim
22
to detect different patterns in the same visual array, the differ-

ences between them being accounted for in the different signal-
to-noise ratios they claim to detect; both do fairly well in
predicting extensions of the array, chalking up their different
patterns of failure to their respective assumptions about noise.
Who is right? Both, says Dennett, and so is anyone else who can
do comparably well by detecting yet another pattern in the ar-
ray. It's no good saying, "Yes, I know they are all making money
out of their systems, but who is right?" That, for Dennett, is
symptomatic of the outmoded, inflexible and unforgiving real-
ism he wants to supersede - we might call it "first-strike real-
ism". While Dennett's argument is complex and fascinating, at
the end of it I still wanted to retain a distinction between use-
fulness and truth in areas where it seemed to me Dennett blurs
the distinction. I shall not, therefore, be arguing for the reality
of the movement of cinematic images simply on the grounds
that it is useful for us to think and speak as if there really were
such movement, though the usefulness of this way of thinking
and speaking is certainly why we are interested in whether this
22
Daniel Dennett, "Real Patterns", pp. 27-51. I am grateful to Jerrold Levinson
for drawing my attention to this work, and for valuable discussion of the
topic of this chapter.
37
kind of movement is real. So I remain a first-strike realist. As

we shall see, even we hawks are capable of ontological gener-
osity.
In fact, I shall not be looking for a positive argument in favour
of the reality of cinematic motion. In debates over whether some
type of experience is illusory, the burden of proof, or perhaps
merely of argument, lies with the party which asserts that the
experience in question is illusory, just as it does with someone
who asserts that a certain belief is false. In both cases -belief
and perception - we have grounds for treating veridical and
nonveridical states asymmetrically, since states of belief and per-
ception are states we have because they tend to be veridical. In
that case, we should hold that cinematic experience of move-
ment is veridical unless there is a significant weight of evidence
and argument against the idea.
However, it is worth pointing out that, from the perspective
of someone who adopts what I have called Transparency, it is
difficult to deny the reality of the cinematic motion we seem to
see.23 The thesis of Transparency- which I shall assess in some
detail in the next chapter - has it that when I see a cinematic
image of Cary Grant, I really and literally see Cary Grant.
Someone who holds this view would find it hard to deny that
the movement we seem to see on the screen is real movement.
If Grant was filmed while moving from one place to another,
and the film represents him as moving, and if I really do see
Grant when I see the film, then surely I really see him moving.
In that case, the appearance of movement cannot be an illusion
of movement. But of course this argument weighs only with
someone who accepts Transparency and not with anyone else,
so I shall not put much weight on it here.
What is the argument for saying the experience of cinema
involves a perceptual illusion of movement? Clear statements of
the argument are hard to come by. I have the impression the
argument sometimes appeals to the fact that there is no move-
ment on the cinematic film roll itself, but rather a sequence of
static images. This is true, but it does nothing to establish the
3
' This point emerged during discussion with Graham Nerlich.
unreality of cinematic movement. After all, when we listen to a
tape recording, there is no sound on the tape itself, but just a
pattern of selective magnetization. We would not conclude from
this that when we listen to a tape recording of music we are
subject to an auditory illusion. The claim of Perceptual Illusion-
ism is that there is no movement on the screen; for this, after all,
is where we seem to see movement.
An argument which might seem to favour Perceptual Illu-
sionism is the following: the supposed movement on the screen
is the product of our perceptual system, and cannot be thought
to exist independently of it. Suppose you described the events
on the screen from the kind of objective viewpoint we try to
occupy in physical science: you exhaustively describe the impact
of particles or waves of light on the screen, and you thereby
describe all the relevant physical events at that surface. But you
do not describe any movement of the kind we claim to observe
there; you do not describe any object as moving from one place
on the screen to another. So there simply isn't any movement,
since the objective description comprehends all the relevant
physical facts but describes no movement. It is only when you
take a subjective point of view and include in the description
the viewer's subjective experience of the screen that movement
enters your story.
This is a poor argument. It is parallel to a class of other ar-
guments that would establish the illusory nature of all our ex-
perience of what are called secondary qualities. Consider again
the case of colour; we describe the object from a physical point
of view exhaustively, including everything about the spectral-
reflectance profile of its surface, but we say nothing about the
way it looks; colour enters our vocabulary only when we include
the observer's subjective point of view in the story. So there are
not really any colours. There are those who welcome this con-
clusion, and say the experience of colour is indeed illusory; ex-
perience represents things as having colour properties when in
fact they do not have them."" But on the whole philosophers
resist such starkly revisionist conclusions, and I go along with
"" Boghossian and Velleman, "Colour as a Secondary Quality".
39
them. What a realist about colour should say is what I have
already said: colours and other secondary properties are real,
response-dependent properties of things. And so it is, I claim,
with the "apparent motion" of projected film; this is real, re-
sponse-dependent motion.
Perhaps the greatest source of resistance to the idea that cin-
ematic images are real is the idea that the apparent motion of
the image is not "tracked'' by any comparable motion of a phys-
ical object. As I have said, no particle or wave, or any physical
thing, moves across the screen as the image of Cary Grant
(seems to) move across it. But normally, when we say that
something moves, we can identify a correlated moving physical
object - or at least we have good reason to believe there is one.
If a person moves, then his body does also; if a car moves, there
is a mass of molecules out of which the car is constructed which
moves. Moreover, the movement of the person or the car su-
pervenes on the movement of the underlying physical object; a
person's movement logically requires the movement of his body
- similarly with the car and its constituent molecules. The move-
ment I have claimed for cinematic images is not like that. But
this is explained in terms of the basic difference between persons
and cars, on the one hand, and cinematic images on the other.
Cars and persons have relatively stable physical constituents,
while images do not. The movement of the image supervenes
on the pattern of light particles striking the screen. But the image
has qualities not possessed by any of the physical things and
events to which we appeal in explaining it: in particular, it has
movement. In just the same way, the colours on a surface are
explainable in terms of the physical properties of those surfaces,
and ultimately in terms of their subatomic constituents. But the
colour has qualities not possessed by any such constituent; the
greenness of the colour is not to be found in any greenness of
its subatomic parts.
So I say that part of the content of cinematic experience is that
there is movement of images, and there really is such move-
ment. We see the cinematic image of a man, and we see that it
is in one place on the screen, and we later see that it is in an-
other; indeed, we see that image move from one place to another
on the screen. That image is not to be identified with some par-

ticular physical object. It is not like the image in a painting,
which consists of a certain conglomeration of physical pigments,
at least relatively stable over time. It is an image sustained by
the continuous impact of light on the surface of the screen, and
no particular light wave or particle is more than minutely con-
stitutive of it. Nonetheless, that image is a particular thing, and
a thing which moves.
To say there is a movement is to say that a certain thing is in
one place at one time and in another place at another time (or,
on the four-dimensionalist view, that a temporal part of the ob-
ject is in one place and a different temporal part is in another).
Motion requires reidentification over time of that which moves.
But what allows us to say that the image in one place on the
screen at one time is the same as the image at some other place
on the screen at another time? 25 One initially plausible answer
is that images get their identity conditions from their causal an-
tecedents: this image now and that one then are both cinematic
images of the same man, so they are the same image. But this
answer is unsatisfactory. First of all, I would make the same
claim for the identity of the image over time in an animated
cartoon. That is, I would claim that we see the real movement
of the image of Mickey Mouse from one side of the screen to
the other. But in this case the argument from sameness of causal
antecedents will not allow us to reidentify images over time,
because there is no Mickey Mouse to be the common causal
ancestor of the image I see in one place and the image I see in
another. We might say that sameness of causal antecedent is a
sufficient but not a necessary condition for identity among im-
ages. The better criterion for the identity of cinematic images
across time is given by their relation to the mental states of the
viewer. This image now is the same as that one then because
both are identified by normal viewers in normal conditions as
being images of the same individual. And that is as true of the
image of Mickey Mouse as of the image of Cary Grant. Here
•s Remember that this question is to be taken as neutral between three and four
dimensionalism.
again, as with colour, the concept we appeal to is response de-
pendent. Identity between images is itself a response-dependent
concept, because questions about how to reidentify images
across time are answered by appeal to facts about the psycho-
logical responses of the viewer to those images. But just as with
colours, this response dependence is compatible with the reality
of the images concerned.
1.7 TWO WAYS OR THREE?
In arguing against Perceptual Illusionism I have been insisting

that cinematic motion is real, using that term to contrast, natu-
rally enough, with "illusory". You may think this taxonomy is
insufficiently refined. After all, we commonly contrast reality
with appearance. My dichotomy will then have us identify that
which belongs to the realm of mere appearances with that which
is illusory. That seems hard on appearances. There ought to be
room for a position which says colours and other secondary
properties belong to the realm of appearances, but which denies
that the experience of colour is illusory.26 (Perhaps in the end
this position will tum out to be incoherent; I just don't want to
rule it out at this stage.) On that view, the real contrasts with
the illusory and with the apparent. Equivalently, we could say
there are two senses of "real": a weak sense which has as the
complement of its extension the illusory, and a stronger sense
which has as the complement of its extension the illusory and
the apparent, which we can then lump together as the unreal.
If we adopt that taxonomy, my view is simply that cinematic
motion is real in the weak sense. I can then agree that in a
strong, metaphysical sense we ought to be antirealists about cin-
ematic motion (thinking of it as unreal), and perhaps about col-
our as well. But we shall need to make a distinction between,
on the one hand, antirealist concepts applicable in the realm of
mere appearances and, on the other, antirealist concepts appli-
cable in the realm of illusion. Whatever your view on colour,
there is surely a difference, for example, between ascribing blue-
6
' Conversation with Brian Medlin suggested this position to me.
ness to a U.S. mailbox, and ascribing greenness to the (actually

white) stripes displayed in a McCullough aftereffect experi-
ment.27 The difference, I submit, is that in the second case but
not in the first, you are subject to an illusion. So, if my parallel
between cinematic motion and colour does not persuade you to
be a strong realist about cinematic motion, it may still be enough
to undermine illusionism about it.
To underline this last point, notice a feature of response-
dependent concepts sometimes thought to be grounds for being
an antirealist about such concepts. Realists sometimes emphasize
the radical fallibility of our beliefs about any domain to which
they claim realism applies; if it's possible, even under episte-
mically ideal circumstances, for us to be mistaken in our beliefs
about that domain, that is a sign we are in the realm of reality.
But with certain kinds of response-dependent concepts, radical
fallibility is ruled ouC8 If a normal observer in normal circum-
stances judges that something is red, it is red; similarly, if a
normally sighted person, sitting in a darkened cinema at the
appropriate distance and attending to the screen as the projector
rolls, judges that the cinematic image is moving, then it is. That,
as I say, might be grounds for rejecting metaphysical realism
about colour, and about cinematic motion. But it cannot be
grounds for thinking cinematic movement is an illusion. Where
there is no possibility of error, there can be no illusion.
If calling cinematic motion "apparent" in a sense which con-
trasts with illusory is to have any significance, we ought to be
able to point to phenomena which are illusory rather than ap-
parent; the appearance-illusion distinction won't be worth
much if the extension of the illusory is empty. Indeed, one com-
mon and natural thought in response to my assertion that cin-
ematic motion is not illusory is that the assertion can be
'
7
The McCullough effect produces an illusion of green or red stripes where
there actually are white stripes. Which colour it seems to produce depends
on the orientation of the bars (horizontal or vertical) on the cards presented
prior to the induction of the illusion. It is therefore called "an orientation-
specific color aftereffect".
28
See Richard Holton, "Intentions, Response-Dependence and Immunity from
Error".
43
sustained only at the cost of reducing, almost to nothing, the
class of phenomena that count as illusory. It is worth seeing that
this is not so.
Someone might claim that, by an argument parallel to the one
I have given for the reality of cinematic motion, we can establish
that the experience induced by the Muller-Lyre phenomenon is
veridical. In those cases of experience singled out as exempli-
fying the Muller-Lyre illusion, we are to say that what experi-
ence represents is the holding, between two lines, of the relation
being longer* than, where length* is not the metrical property of
objects we measure with rigid rods, but rather a response-
dependent length: a length which stands to metrical length as
the response-dependent movement I have been advocating for
cinematic images stands to the movement we measure by track-
ing physical objects across space. In that case there is no illusion
involved in the Muller-Lyre phenomenon, but merely the verid-
ical experience of one line being "longer* than" another.
This objection fails. Our experience of the Muller-Lyre illu-
sion represents the lines as standing in the relation "longer
than", not the relation "longer* than". The visible appearance
of the lines suggests that, were one to measure them in the con-
ventional way, the result would be that one was measurably
longer than the other. That is why this is genuinely a case of an
illusion, rather than a veridical experience of a response-
dependent property. With the experience of screen watching,
however, it is doubtful whether the movement our experience
represents as taking place is of a kind that would be undermined
by independent checks analogous to the measuring check we
can carry out in the case of the Miiller-Lyre illusion. Our ex-
perience of screen watching does not have this as its represen-
tational content: "there are reidentifiable physical objects
moving in front of our eyes" (a content the falsity of which
could be established by independent checks). Rather, its content
is: "there are images of reidentifiable physical objects moving in
front of our eyes." In this respect the experience of cinematic
motion seems not to be undercut by information from other
sources, and therefore to be crucially different from that induced
by the Miiller-Lyre setup.
44
Someone might also object that my appeal to response de-

pendence in the filmic case has the false consequence that there
is no illusion involved in our perception of the movement of a
wave. Our perception of wave phenomena suggests there is
something, namely a wave, which moves outwards from the
centre of disturbance. But if we examine the matter closely we
discover there is no physical object - no single, reidentifiable
body of water- which spreads outwards when wave motion
occurs. There is only the transfer of energy from one molecule
to the next, as can be established by placing a free-floating object
on the wave surface and seeing it remain stationary with respect
to the horizontal axis. But it would seem that my argument for
the reality of the movement of the cinematic image applies
equally to the movement of the wave, in which case we should
have to say that there is, after all, no illusion of movement in
the case of the wave, and that seems wrong.
But again there is a difference between the cinematic image
and the wave. In the case of the wave, but not in that of the
cinematic image, there is a physical object, namely a body of
water, which perception represents to us as moving outward as
the wave "spreads". But our perception of the motion of cine-
matic images does not suggest that there is some particular
physical object which moves when a cinematic image does. That
is why there is a perceptual illusion, or at least a perceptual
error, in the perception of wave motion but not in that of cine-
matic images.
There are a number of other apparent motions which are nor-
mally classed as merely illusory and which retain their status as
illusions on my account. For example, psychologists speak of
induced motion, a phenomenon noticeable when we see clouds
drifting across the face of the moon; if the clouds drift slowly to
the left, the moon appears to be drifting to the right. Similarly,
tall buildings viewed from below against a background of mov-
ing clouds seem to be falling. 29 In these and like cases we have
a perceptual illusion of movement; experience represents
something moving which is not moving, as independent checks
'9 See Stuart Anstis, "Motion Perception in the Frontal Plane: Sensory Aspects".
45
would establish. But these cases are unlike that of cinematic mo-
tion. There are even kinds of motion which cinema sometimes
gives us and which are, or can be, illusory rather than real. For
example, films in 3-D display an illusion of depth; our experi-
ence of watching 3-D is such that objects are perceived as mov-
ing towards or away from the viewer when this is not the case.
The cinematic motion I claim to be real is not of that kind; it
belongs to the kind which psychologists call "motion in the fron-
tal plane".
But even within this restricted class of motion phenomena I
can make distinctions, for not all motion in the frontal plane will
count as real by my lights. It is well known, for example, that
if one looks at a point of light in an otherwise darkened envi-
ronment, the point will seem to shift around when in fact it
remains steady. The explanation seems to be that the appearance
of movement is produced by random eye movements which are
uncompensated for because the viewer has no visible frame of
reference other than the light source itself. 30 Now suppose that,
instead of looking directly at the light source, the viewer looks
at the image of it projected on a screen. Then the projected image
will seem to move around, just as the light source itself would.
And this, I claim, is a case of illusory rather than of real motion
on the screen. For the following is a necessary condition for
there to be genuine movement of an image: that at each place
on the screen occupied by the image as it moves, there should
be illumination at that place (and at the relevant time) on the
screen. But in the case we are considering, there will be only
one fixed and unchanging place on the screen illuminated, and
at many places on the screen where the image seems to be, there
will be no illumination. So here we seem to see movement of
an image where no movement exists. But in the case of what is
conventionally called the moving image of film, at the places to
JO I am grateful here to Sue Feagin and Dan Gilman for helping me to see the
significance of this phenomenon to my theory, and to Michael White, who
explained the details to me. The illusion described is called "autokinetic
movement". See E. L. Brown and K. Deffenbacher, Perception and the Senses,
pp. 412-415·
which the image seems to move, there really is illumination on

the screen.
Cinematic images are unlike those of, say, painting; they are
temporary, response dependent and extrinsic in ways the im-
ages of painting are not. Still, cinematic images are real objects,
reidentifiable across time and occupying different positions at
different times during the viewing of the shot. Or, for those who
insist on a tripartite distinction between the real, the apparent
and the illusory, they are apparent, nonillusory objects. The ba-
sic mechanism of cinema is not, after all, based on illusion. Nor
do I believe there is any other substantial, interesting sense in
which cinema is a medium which creates illusions. Certainly, it
does not typically function to produce the cognitive illusion that
what is represented onscreen is real and present to the viewer.
Realism is a notion film theorists have been uncomfortable with
for a while. I shall argue in Chapter 3 that it has been misun-
derstood, and that it is an indispensable tool for understanding
the nature of film. Illusion, on the other hand, is something we
can well do without.
The conclusions of this chapter go beyond film theory to em-
brace general metaphysics. It is traditional to regard motion as
a paradigmatically primary quality, to be contrasted with those
secondary qualities which are in some sense observer depend-
ent, like colour. If what I have said here about cinematic motion
is correct, we shall have to acknowledge a kind of motion which
takes its place among the secondary qualities.
47
Chapter 2
The imprint of nature

A painting is a world; a photograph is of the world.
Stanley Cavell
There are people who say that the photographic method gives
the cinema the power not merely to represent the world, but to
present it. Exactly what this means may not be clear; I shall try
to make it clear. But note that this claim (I'll call it the Presen-
tation Thesis) and the argument it has engendered predates the
cinema, having arisen first in connection with still photography.
I'll begin by considering the claim just for the case of still pho-
tography. Later, I shall ask whether the claim is more plausible
when applied to the moving images of cinema. Most people who
argue for the Presentation Thesis restrict their claim to photo-
graphic images which have not been subject to substantial ma-
nipulation after exposure. I shall follow them in this.
One source of confusion has to be cleared up right away. In
the Introduction I pointed to the representational duality of film.
When we ask what is represented onscreen, there are two pos-
sible kinds of answers: "Ingrid Bergman and Cary Grant acting
on a set" or "two spies trying to foil a Nazi plot". The claim
that cinematography presents rather than represents the world
must be understood as the claim that it presents the real world
of actors, props, sets and locations, not the unreal world of fic-
tional characters. If the thesis is right, film presents us with In-
grid Bergman and Cary Grant, not with the characters they play
in the movie, for these characters do not exist. Photography may
have special powers, but it does not have the power to turn
nonbeing into being.
2.1 REPRESENTATIONAL MEDIA AND

REPRESENTATIONAL ARTS
Painting is a representational medium; one thing - an object,

scene or event - is represented by another thing, a pattern of
paint. And painting is a representational art in the sense that
the artistic or aesthetic values of a painting can derive, at least
in part, from its representational features.
No one doubts that painting is a representational medium,
and few these days deny that it is a representational art. But
there are those who doubt one or both of the corresponding
theses concerning photography, though which one is sometimes
not clear. I shall argue that photography is representational,
both as medium and as art.
Painting is a representational medium in that works within
that medium can be- though they need not be- representa-
tional. We should not automatically assume, however, that all
representational media allow the creation of works which are
representational works of art; perhaps there are conditions on a
medium in addition to its being representational which have to
be satisfied before it will allow for the creation of works of art.
What those extra conditions are is unclear, partly because the
notion of art is itself unclear. I shall simply assume that there
are two distinct questions to be answered: Is photography a rep-
resentational medium, and is photography a representational
art? I shall argue first that photography is a representational
medium, and second that it is a representational art. That will
leave us with the task of explaining the difference between pho-
tography and painting, which I hope to make clear by the end
of this chapter.
2. 2 PRESENTING AND REPRESENTING
Representations extend our epistemic access to things in the

world; if they are reliable, representations give us information
about things when those things are not directly accessible to us.
And for some purposes a description, a detailed picture or some
other kind of representation can be more informative than a
49
direct perceptual examination of the thing itself; compare the
difficulty of closely examining a furious and hungry tiger with
closely examining a suitably detailed picture of one.
Other devices enhance our perceptual access to things them-
selves. Lenses help us see detail inaccessible to the naked eye.
No one will say, I suppose, that lenses give us representations
of things. They are, rather, aids to vision; they help us to see
things themselves. They present the world to us rather than rep-
resenting it; they just present the world in more detail than the
naked eye can.
Do photographs present or represent? A certain tradition says
they present. But this tradition has not, on the whole, been well
served by its defenders. They have said things which must ei-
ther be taken as so obscurely metaphorical as to be unhelpful,
or as literal statements of what is obviously false. Andre Bazin,
for example, said that "it is false to say that the screen is inca-
pable of putting us 'in the presence of' the actor. It does so in
the same way as a mirror ... but it is a mirror with a delayed
reflection"! But the screen is not literally a mirror, and its action
is not even very much like that of a mirror; nor are we really in
the presence of the actor when we watch the screen.
The Presentation Thesis gets its clearest statement, and its
most careful defence, in the hands of Kendall Walton, whose
version is pleasingly straightforward: To see a photograph of X
is to see X. But to see a painting of X is not to see X. For one
2
thing, paintings can be "of" things that do not and never did
' Andre Bazin, What Is Cinema? vol. 1, p. 97· See also Stanley Cavell, The World
Viewed. For remarks critical of the tradition see Joel Snyder, "Photography
and Ontology".
' "With the assistance of the camera, we can see not only around comers and
what is distant or small; we can also see into the past. We see long-deceased
ancestors when we look at dusty snapshots of them" (Walton, "Transparent
Pictures", p. 251). For similar arguments, see Roger Scruton, "Photography
and Representation". But Scruton, whose concern is mostly with the aesthetics
of photography, does not offer so precise an account of the difference between
photography and painting. See later in this chapter for an analysis of some
of Scruton's arguments. For an examination of the history of this problem,
see Patrick Maynard, "Drawing and Shooting: Causality in Depiction".
exist. You can see a painting of a dragon, but you cannot see a
dragon. Nor can you see a photograph of a dragon.
While Walton's thesis is straightforward, his argument for it
is subtle and depends on some technical notions in philosophy.
Accordingly, I shall spend some time in laying out the argument
and explaining the technical notions. The effort will be worth it;
the argument is a good and an instructive one, even if, as I
believe, it fails to establish its conclusion.
Walton's thesis is not Bazin's, if we take Bazin at his literal
word. Walton's thesis is not that a photograph of X is, or is
part of, X. It is not that, when we are in the presence of a
photograph of X, we are in the presence of X. It is just that
when we see a photograph of X we see X. Photographs are,
Walton says, "Transparent"; we see through them to the
things they are of (I'll call this the Transparency Thesis and
sometimes just "Transparency''). I shall argue that the Trans-
parency Thesis is not susceptible to many of the criticisms lev-
elled against other versions of the Presentation Thesis. But I
shall also argue that Walton has failed to establish a case for
Transparency, and that what is correct in his argument can be
accommodated by the view that photographs are representa-
tional, that seeing a photograph of X is a matter of seeing a
representation of X rather than of seeing X itself. The repre-
sentations photographs give us are certainly very different in
kind from those we get by drawing and painting, and these
differences are the product of differences between the ways
photographs and hand-made images are produced. But a pho-
tographic image is a representation.
A remark on terminology: to avoid lengthy formulations I
shall use certain expressions in a special way. I shall contrast
"ordinary seeing" with "seeing photographs", and sometimes
with "seeing paintings". In literal fact there is no contrast here,
because we see photographs and paintings in just the same way
we see other things - by looking at them. The intended contrast
is between seeing an object or scene in the ordinary way, when
the object is before your eyes, and seeing a photograph or paint-
ing of the object or scene.
51
2.3 PHOTOGRAPHS AS EVIDENCE

In the central and enthralling sequence of his otherwise
somewhat pretentious Blow Up, Antonioni shows a photogra-
pher discovering, by the successive isolation and enlargement
of ever-smaller sections of a photograph, the evidence that a
murder has been committed. His photograph had recorded
something of which he was unaware at the time he took it. Sup-
pose we came to suspect that there was a dead body lying under
a hedge on the day and in the place where Constable had
painted The Haywain. It would be absurd to examine the canvas
in minute detail in order to find evidence of the body there; at
least, it would be absurd if we thought that Constable was un-
aware of the body's presence. Constable could have known
about the body, and painted it in such a way that it would be
very hard to spot unless you knew what you were looking for.
But if he did not know about the body we shall not find evi-
dence of it in the canvas, however hard we look. We cannot, it
seems, discover things in paintings in the same way we can
discover them in photographs.
In Hitchcock's Rear Window, a photographer confined by in-
jury to his apartment discovers evidence that a murder has been
committed in the building opposite. He discovers this, not by
looking at photographs he or anyone else has taken of the events
in that apartment, but by looking in through the window. The
photographer might have taken photographs and found the ev-
idence later by examining enlargements, but that was a twist to
the tale Hitchcock chose to ignore or did not think of. What he
could not have done was discover the evidence by looking at
paintings of the apartment next door unless the painter intended
to place the evidence in his picture.
Blow Up and Rear Window, between which there are obvious
connections of influence, depend on strikingly similar ideas. One
photographer looks at photographs of a scene and discovers ev-
idence of a crime; the other looks at the scene itself and discov-
ers the same thing. They do things by way of photography and
sight that they could not do by way of painting or other hand-
made images.
Perhaps there is here the beginning of an argument that pho-
tographs are transparent. It is an argument which needs some
underpinning.
2.4 COUNTERFACTUAL DEPENDENCE

There are rich patterns of counterfactual dependence between a
photograph and the object photographed. Under certain condi-
tions - "normal conditions" as I shall call them3 - we can expect
a photograph of X to display the visible properties of X in such
a way that if X's visible properties were different, the photo-
graphic image would be correspondingly different. Had the
building photographed been differently shaped, the image of the
building on the photograph would have been correspondingly
differently shaped. That is a case of counterfactual dependence
between the appearance of X and the appearance of the photo-
graph of X. In the same way, my visual experience when seeing
the photograph is counterfactually dependent on the object pho-
tographed: if X's visible properties were different, my visual
experience when seeing the photograph would be correspond-
ingly different.
Ordinary seeing is like that too. When I see the building un-
der normal conditions I have a visual experience which matches
the scene before my eyes, and I do so in virtue of connections
between the scene and the visual experience which are such that,
if the scene were different in various ways, my visual experience
would be correspondingly different.4
Counterfactual dependence alone will not distinguish pho-
tography from painting. Under certain conditions, a painting of
3
"Normal" does not mean "average". Conceivably, a majority of all the pho-
tographs ever taken were so over- or underexposed as to put the relevant
counterfactual connections in doubt. I take "normal conditions" in something
like Ruth Garrett Millikan's sense: a normal condition for photography is a
condition for its proper functioning. (See Millikan, Language, Thought, and
Other Biological Categories.)
• See David Lewis, "Veridical Hallucination and Prosthetic Vision". The con-
dition of counterfactual dependence may not be sufficient for seeing; see Mar-
tin Davies, "Function in Perception".
53
X will display the visible properties of X in such a way that, if
those visible properties were different, the painted image would
be correspondingly different. If the building were differently
shaped the painter would paint a correspondingly different
shape on the canvas. But in the case of the painting, and not in
that of the photograph, counterfactual dependence is mediated
by the beliefs of the artist. The appearance of the painting de-
pends counterfactually on the appearance of the object because
the beliefs of the painter are similarly dependent. If the painter
were having an hallucination, thinking there was a pink ele-
phant in front of him, his painting would display a pink ele-
phant, not the actual scene before his eyes. With the photograph
things are different. Because of the "mechanical" way photo-
graphs are made, it does not matter what the photographer
thinks the object in front of the lens looks like; once the camera
is set up and the film exposed, the camera records the scene
before it. Imagine the same scene successively photographed
and sketched. Now imagine that the scene had looked different
in some way; in that case the photograph and the sketch would
both look different. Now imagine that the scene had looked dif-
ferent and the artist's beliefs about the scene's appearance remained
the same. In that case only the photograph would be different.
Both the photograph and the sketch depend for their appear-
ances on the appearance of the scene, but in the sketch and not
in the photograph, that dependence is mediated by the artist's
beliefs and other mental states.
We can extend this counterfactual dependence between scene
and photograph/sketch to a similar dependence between the
scene and my visual experience of the photograph/ sketch. Had
the scene been different my visual experience on looking at the
photograph/ sketch would have been different. But if the scene
had been different and the artist's mental states the same, my
experience of looking at the photograph would have been dif-
ferent, while my experience of seeing the sketch would have
been the same. In the case of the photograph, the counterfactual
dependence between the scene and the observer's visual expe-
rience is, we may say, independent of belief, but in the case of
the sketch it is not. That is why we might find, in the photo-
54
graph, evidence for something the photographer didn't suspect

was there, while we could not find any such thing in a painting.
Suppose we consider a simpler case: I just look at the scene.
Now there is counterfactual dependence between the scene and
my visual experience. Is this counterfactual dependence like the
photograph case, or like the sketch case? It is like the photo-
graph case. Imagine that the scene I am looking at were differ-
ent; my visual experience on seeing the scene would be different
also, whether or not people's beliefs remain the same. My visual
experience doesn't depend on anyone's belief.5
Call the counterfactual dependence exhibited by ordinary see-
ing and seeing photographs "natural counterfactual depend-
ence", or simply "natural dependence" .6 Call the dependence
exhibited by seeing paintings "intentional counterfactual de-
pendence", or simply "intentional dependence". The argument
for the transparency of photographs is this: what makes ordi-
nary seeing a way of perceiving objects is its natural depend-
ence. Since seeing photographs exhibits natural dependence
also, it too is a way of perceiving objects. But seeing paintings
exhibits intentional dependence, and so is not a way of perceiv-
ing objects? Thus photography is transparent and painting is
not.
5
There are some technical difficulties connected with backtracking arguments
here. Interested readers are invited to consult David Lewis, "Counterfactual
Dependence and Time's Arrow".
6
Note that my definition of "natural dependence" is partly stipulative. To say
that a process exhibits natural dependence carries no implication that the
process is naturally occurring, and many of the examples following in the
text are of processes that are not naturally occurring, as with the working of
thermometers, ammeters and mechanically generated descriptions.
7
I take this to be Walton's central argument for the "sharp break" between
painting and photography. There are occasional hints of other arguments in
his paper. Thus, he emphasizes the "slippery slope" we are on, down from
unaided vision, through seeing with the aid of telescopes and mirrors to see-
ing photographs (see Walton, "Transparent Pictures", p. 252, and idem,
"Looking Again Through Photographs: A Response to Edwin Martin", es-
pecially pp. 8o5-8o6). But, as other critics have noted, we can ask what is to
stop us going further down the slope, to seeing paintings. Presumably the
answer is that we should stop at the point where natural dependence gives
way to intentional dependence. In that case, the slippery-slope argument re-
55
I shall look at this argument in some detail. Before that, I shall
briefly note one other argument for Transparency - and some
arguments against it. These latter arguments are prevalent, but
not persuasive.
2.5 INCONCLUSIVE ARGUMENTS

Certain of our intuitions about and reactions to photographs
seem to be explained by Transparency. We have the sense, when
looking at a photograph, of being in a particularly intimate re-
lation to its subject - more intimate than when we look at a
painting of the same subject. At least, that is so where photo-
graph and painting are of comparable accuracy and definition.
Photographs of atrocities affect us more than paintings of them
do. And one who finds explicit sexuality offensive is more likely
to be offended by a photograph with explicit sexual content than
by a painting with such content (other things being equal). This
is all explained by the Transparency Thesis: when we see pho-
tographs of atrocities or of sexual acts, we see atrocities, we see
sexual acts. This is not the case when we see paintings with
comparable subjects.8
But this argument cuts both ways. While it is true that a pho-
tograph of something distressing or offensive has greater impact
than a comparable painting does, it is also true that seeing that
thing directly (without the aid of a photograph) is more distress-
ing or offending than seeing a photograph of it. The argument
just given fails to put seeing directly and seeing photographs in
one class and seeing pictorial representations in another. At best,
it suggests a continuous scale of psychical effect, with photo-
graphs in the middle, between hand-made images and ordinary,
unmediated seeing. Later, I shall explain this ordering in terms
of the nature of photographic representation.
So one supplementary argument for Transparency fails. But
some arguments against Transparency go wrong also.
lies for its force on the argument from natural dependence, and any weakness
in the latter will be inherited by the former. So I ignore the slippery-slope
argument.
8
See Walton, "Transparent Pictures", pp. 247 and 255.
There seem to be clear cases of seeing photographs of things

that are not cases of seeing the things they are photographs of.
When I see a grossly over- or underexposed or ill-focused pho-
tograph of X, I surely do not see X. 9 The same can be said for
unaided sight, when my visual equipment is impaired. Seeing
an object is a matter of there being rich patterns of counterfac-
tual dependence between my visual sensations and the object
seen. For me properly to see X, my visual experiences must
reflect a good number of the observable properties of the object,
and be such that, if the object's observable properties had been
different, my visual sensations would have been different.
(There is vagueness in all this, but the concept of seeing is vague,
so we can expect a realistic analysis of the concept to be vague
to the same degree.) One whose eyes receive too much or too
little light, or for whom the light is not focused by a lens, does
not see objects in his surroundings. So too with seeing through
photographs. If the photograph is uniformly black, its appear-
ance fails to have the right kinds of counterfactual dependence
on the appearance of the object, and so I do not see the object
when I look at the photograph. But this does nothing to raise
doubts about the transparency of "ordinary", well-focused and
well-lit photos. 10
Another objection to Transparency is that it entails that we
can now see long-dead people because we can see photographs
of them; but how can we now see people who do not now exist?
This objection is countered by noting that with the naked eye
we see stars so far away that the light from them has taken
millions of years to get here- during which many of them will
have ceased to exist.
A further objection to Transparency is that photographs do,
in various ways, depend on mental states." Usually, a photo-
9
See, e.g., Edwin Martin, "On Seeing Walton's Great-Grandfather".
'" See Walton, "Looking Again Through Photographs", p. 804.
" Thus Arnheim: "People who contemptuously refer to the camera as an au-
tomatic recording machine must be made to realize that even in the simplest
photographic reproduction of a perfectly simple object, a feeling for its nature
is required which is quite beyond any mechanical operation" (Film as Art,
p. 19).
57
graph is the result of a deliberate exposure at a scene chosen,
framed and focused on in a certain way, perhaps with the ar-
rangement of objects and the lighting organized for the occasion.
If the scene had not looked the way it did to the photographer,
she would not have chosen to take the shot. Again, this is no
objection to Transparency. For in this sense, ordinary seeing de-
pends as much on intention. That I organize the furnishing and
lighting in my study, together with the fact that I am so aes-
thetically sensitive that I refuse to look at anything disorganized
or badly lit, does not show that I do not really see my study
and its contents when I choose to look at it. Once I focus my
eyes in a given direction it is not, from this point on, up to me
what I see. The same holds for the photograph; once the shutter
is opened and the film exposed, it is not up to me what appears
on the negative.
We can see from this that the argument for Transparency is
not that photography is an entirely mechanical activity. So it is
irrelevant to point out against Transparency that most artistic
human activities are mechanical after a certain point, as with
piano playing: once you have hit the key, everything else in the
sound-making process is done mechanically. 12 The advocate of
Transparency need not be saying that photography is uncrea-
tive.
People who dispute the similarity between seeing photo-
graphs and ordinary seeing sometimes emphasize the relation
between seeing and the path of an uninterrupted light ray.
When we see in the ordinary way, even with the aid of lenses
and mirrors, light emitted by or reflected from the object seen
passes into our eyes. The lens or the mirror merely collects or
deflects the light rays. And when we see those stars that no
longer exist, the light from the stars enters our eyes. But with a
photograph, things are different. The light reflected from the
surface of the photograph into my eye is not the light that trav-
elled from the object to the photographic plate. Should we give
this as a reason for saying that seeing photographs and ordinary
" See Ted Cohen, "What's Special About Photography?" See also Jarvie, Phi-
losophy of the Film, p. 109.
seeing are not much alike? No. First of all, light is not essential
to seeing. It is true that, for our organs of sight to function
unaided, light rays must be reflected from an object into our
eyes. But there could be seeing in which light rays play no part.
Richard Dawkins raises the possibility that bats might have vi-
sual experiences qualitatively similar to our own, but caused by
their very different perceptual systems, which depend on bounc-
ing sound waves off solid objects.'3 I understand this is not likely
to be true of real bats/4 but we can imagine batlike creatures
complex enough for this to be a plausible story. Their visual
sensations, if they exist, are caused by sound waves instead of
by light - an odd idea, but not a confused one. Sonar is capable
of giving information about objects in a creature's local environ-
ment which is sufficiently detailed and up-to-the-minute for the
creature to build up a complex mental representation of its en-
vironment. This representation is informative in the way our
visual representation of the environment is. Super bat sonar is
not functionally dissimilar from our organs of sight; whether the
representations it gives are qualitatively like our visual ones we
cannot know, but there is no reason in principle why they could
not be. Perhaps we shall invent a device which emits high-
frequency sound waves, which can be fitted to humans as a
prosthesis and which gives us visual sensations as a result. Then
the blind would have their sight restored; they would "see" by
means of sound waves.' 5
You may respond that, while seeing need not involve light
rays, it must involve the uninterrupted transmission of
something functionally equivalent (modulo the purposes of see-
ing) to a light wave, and sound waves might do the job. It is
' Richard Dawkins, The Blind Watchmaker.

3
'• See Kathleen A. Akins, "What Is It Like to Be Boring and Myopic?"

'
5
It is worth noting also that there are auditory microscopes - but I admit that
there may be other problems connected with the claim that we see through
such devices. See Ian Hacking, "Do We See Through a Microscope?" Hack-
ing's answer to his question is yes. Also relevant here, as Marty Davies
pointed out to me, is the phenomenon of "tactile vision". See, e.g., G. Guar-
niero, "Experience of Tactile Vision", and P. Bach-y-Rita et al., "Vision Sub-
stitution by Tactile Image Projection".
59
the idea of uninterrupted transmission that is important, not the
particular thing that does the transmitting. That normal seeing
and seeing through lenses and mirrors involves an uninter-
rupted light wave seems important because it is responsible for
some of the informational features of seeing I have described in
connection with bats and to which I shall return: the features of
seeing which enable us to place ourselves in spatiotemporal re-
lations to the things seen. That kind of information is obtained
from sight because the light ray is uninterrupted (though there
are other factors that contribute to this as well). And it is the
maintenance of uninterrupted light transmission which enables
us to track objects over time.
But what is responsible for an essential or quasi-essential
property of X need not itself be an essential property of X: recall
the functionalist's point that neural activity is not essential for
a sentient creature; it is merely what realizes, in us, those func-
tions required for sentience, functions which might be realized
in other ways. So perhaps something other than uninterrupted
transmission could give us the kind of information we normally
get from sight- a simple transducer for instance. Suppose there
is a screen which registers a pattern of light on one side and
emits a qualitatively identical pattern of light from the other-
all this done more or less instantaneously. Looking at the screen
would be just like looking through a window at an object be-
yond. The screen interrupts the light ray, but I think we would
say that when we look at it we see the object on the other side
(assuming the mechanism preserves counterfactual dependence
between our visual sensations and the appearance of the object
beyond). That our way of seeing involves the passage of unin-
terrupted light does not imply that all ways of seeing must.
What is important is that some mechanism make the required
connections. Light propagation (uninterrupted or otherwise) is
one such mechanism; there may be others.
So the fact that there is interruption of the light ray between
the object and my eye when I see a photograph of the object is
not itself an argument against the claim that I see the object
when I see the photograph. However, interruption raises com-
plex issues which will be dealt with later.
6o
2.6 THE CONDITIONS FOR PERCEPTION
The primary argument for the transparency of photographs has

been this: seeing photographs and seeing real things are both
processes that display what I have called natural dependency,
while painting does not display this feature. Seeing real things
is, obviously, transparent: seeing a man, I see a man, not a rep-
resentation of one. Seeing paintings isn't transparent: seeing a
painting of a man, I do not see a man. So it looks as if natural
dependency is the test of transparency. Seeing photographs
passes the test, so seeing photographs is transparent.
But natural dependency is not the test of transparency, be-
cause natural dependence is neither necessary nor sufficient for
transparency.
As things are, or as we presume they are, seeing real things
is a process which displays natural dependence. But that is no
necessary fact about seeing. Consider a case imagined by Ken-
dall Walton: Blind Helen is provided with a prosthesis which
gives her visual experiences matching the scene before her eyes.
But the match is intentionally dependent: her brain is hooked
up to a computer, which is operated by a neurosurgeon. By
operating the computer, the surgeon gives Helen visual expe-
riences which are just those she would have if she saw in the
normal way. Because the dependence here is intentional, Walton
claims this as a case in which Helen merely seems to see; the
surgeon is seeing for her.' 6 I disagree. The case involves an odd
kind of seeing on Helen's part, but still a kind of seeing -as
long as the surgeon's vigilance ensures counterfactual depend-
ence between Helen's visual experience and the scene before her
eyes.'7
6
' Walton, "Transparent Pictures", p. 265.
7
' Perhaps it makes a difference to your intuitions about this case whether you
assume the surgeon gives Helen visual experiences corresponding to what
he (the surgeon) sees, or whether you assume he gives her visual experiences
corresponding to what she would see if her eyes were functioning normally.
As Walton describes the case, the surgeon gives Helen visual experiences
"corresponding to what he [the surgeon] sees" ("Transparent Pictures", p.
265), and I admit at least some doubt about whether Helen could properly
61
If you say Helen is not really seeing in this situation, that
may be because it is difficult to imagine how the surgeon could
control the process to ensure the appropriate degree of counter-
factual dependence. Another case, designed to screen out that
factor, will make my point more clearly. Suppose Malebranche
was right and there are no genuine causal powers in natural
objects. God mediates between the scene and our visual expe-
riences, acting, in his benevolence, to maintain counterfactual
dependence. Under this supposition, we would, as Malebranche
put it, "see all things in God" - but we would still see. We
would see by a process which exhibits intentional dependence.
(If God's omnipotence seems to you a barrier to ascribing to him
actions and intentions in the way we ascribe them to mortals,
think of God as just very powerful, but not as omnipotent.) So
natural dependence is not necessary for perception.
Is natural dependence sufficient for perception? Walton
points out that it is not.' 8 A machine which mechanically gen-
erates descriptions of objects does not enable me to perceive
those objects, even though there is natural dependence between
features of the object and the descriptions generated.'9 So what
makes seeing photographs a way of perceiving objects but read-
ing mechanically generated descriptions not a way of doing so?
Perhaps it is a matter of similarities between investigating things
by examining pictures of them (either photographs or drawings)
20
and investigating them by looking at them directly. As Walton
notes, the discriminations we find it difficult to make in cases
of ordinary seeing and seeing photographs are quite different
from those we find difficult in examining a written description.
A house can easily be mistaken for a bam, and a photograph of
be said to see in that case. But if Helen is given visual experiences corre-
sponding to what she would see if her eyes were functioning, it is much
clearer that she does genuinely see, and under conditions of intentional de-
pendence.
8
' Walton, "Transparent Pictures", p. 271: "A process of discrimination counts
as perceptual only if its structure is thus analogous to the structure of the
world" (my italics).
9
' Ibid., p. 270.
"'Ibid.
62
a house can easily be mistaken for a photograph of a bam. But
it is easier to mistake "house" for "hearse" than for "bam". The
errors of ordinary seeing and seeing photographs are explaina-
ble by real similarities between things, but the errors of reading
21
are not: a house is really more like a bam than like a hearse.
It is these differences between pictures and descriptions which
allow photographs, but not mechanically generated descrip-
22
tions, to give us perceptual access to thingS.
We can now summarize the argument for Transparency. A
mode of access to information about things counts as perceptual
if and only if it (i) exhibits natural dependence and (ii) preserves
real similarity relations. Painting fails to satisfy (i), and the me-
chanically generated description fails to satisfy (ii), so neither is
a perceptual mode of access. Photography satisfies both, so it is
a perceptual mode of access. Seeing photographs enables us to
see the things the photographs are of. So photography is trans-
parent.
I have already argued that (i) is no necessary condition for
perceptual access (recall Malebranchean seeing). Nor are (i) and
(ii) jointly sufficient for it. The length of the mercury column in
a thermometer depends naturally on the amount of ambient
heat. And because small variations in the level of heat corre-
spond to small variations in the length of the column, the dis-
criminatory errors we make when looking at a thermometer are
similar to those we make when we perceive how hot something
is by feeling it on the skin; it is hard to make very small tem-
perature discriminations either by feel or by sight. So thermom-
eters don't "scramble the real similarity relations" between
temperatures. 23 If feeling (a degree of) warmth is a way of per-
" "Descriptions scramble the real similarity relations" between things, while
visual experience, either direct or by means of photographs, preserves them
(ibid., p. 271).
"' Ibid., p. 273.
3
' Most thermometers are gradated, and we judge temperature by matching the
length of the column with a numerical mark. In such cases, reading a ther-
mometer is much like reading a description. But if the gradations on my
thermometer have worn away and I have to judge temperature by height
estimation, we would hardly say that I perceive heat when I see the mercury
column.
ceiving heat, then, by the argument for Transparency, seeing the

length of the column should be a way of perceiving heat also,
which it is not. What I perceive is the length of the column, from
which I infer the level of ambient heat."4
It might be argued in reply that there is a difference between
seeing photographs and seeing thermometers that explains why
photographs enable us to see people but thermometers do not
enable us to see heat. There is, after all, only a patchy and un-
systematic overlap between the errors involved in length esti-
mation and those involved in heat perception. As light fades I
will be prone to more errors in length estimation, but this will
not affect my perception of heat. My perception of heat will be
subject to significant error if I have a fever, but my estimation
of column length will not (not, at least, in the same way). Com-
paring the perception of column length with the perception of
heat, we find the overlap in discriminatory error insufficient to
say that perception of the one is perception of the other. But in
the case of seeing photographs and ordinary seeing, the overlap
is sufficient for us to say that we perceive an object when we
see a photograph of it, just as we do when we see it in the
ordinary way.
There are two reasons why we ought not to accept this ar-
gument. First, no degree of discriminatory overlap, however
great, could force us to the conclusion that seeing a photograph
is a case of perceiving the object itself. There are pairs of objects
A and B for which the following holds: (i) B exhibits natural
dependence on A; (ii) the characteristic pattern of discriminatory
error involved in examining B is exactly as for A; and (iii) per-
ceiving B is not a case of perceiving A. By way of an example,
24
I don't say we could not perceive heat in any way other than via the heat
receptors at the skin. If our eyes were so constructed that we had mono-
chromatic vision, and things looked darker the hotter they were, then per-
haps we would see heat (see Paul M. Churchland, Scientific Realism and the
Plasticity of Mind). Snakes, apparently, see heat (as Marty Davies pointed out
to me; see Peacocke, Sense and Content, p. 90n; and E. Newman and P. Hart-
line, "The Infrared 'Vision' of Snakes". I say only that we do perceive heat
through our skin, and not by looking at thermometers.
consider another Malebranche-inspired case. 25 A and B are two

clocks. The orientation of the hands of A governs that of the
hands of B by means, let us suppose, of radio signals.:W I am
looking at clock B; clock A is out of sight. There is natural de-
pendence between my visual experience of clock Band the ap-
pearance of clock A. If clock A's hands had been in a different
orientation at the moment of my seeing clock B, the orientation
of clock B's hands would have been correspondingly different,
and so would my visual experience. But there is no plausibility
in the claim that I see, or perceive in any way, clock A when I
see clock B. 27 Degree of discriminatory overlap cannot be deci-
sive for whether a naturally dependent mode of access to infor-
mation about an object is a way of perceiving it. So even if the
errors of discrimination characteristic of seeing photographs
were the same as those for ordinary seeing, it would not follow
that seeing a photograph of an object was a way of perceiving
the object itself.
The second reason we should not accept the argument is this.
Part of the argument we are considering is that, from the point
of view of discriminatory error, looking at thermometers is not
sufficiently like sensing heat through the skin for us to say that
looking at thermometers is a way of perceiving heat. But a par-
allel argument will establish that looking at photographs of
things is not a way of perceiving those things. From the point
of view of discriminatory error, how much overlap is there be-
tween ordinary seeing and seeing photographs? Not much; not
25
Malebranche inspired, but really not much like his occasionalism, or the pre-
established harmony to which that doctrine gave rise. Here the one clock
governs the other, rather than both being governed by a common cause. I
adjust the example so as to bring it closer to the case of photography. Object
and photograph are linked not just by a common cause; the object is one of
the causes of the photograph.
26
Assume all this is done by a mechanical process and involves no intentional
states.
27
Other examples meeting the same conditions come readily to mind. The rep-
lication of DNA molecules (by a process that displays natural dependence)
gives rise to two "daughter'' molecules qualitatively identical to the original
one. In examining one of the daughters we are not perceiving the parent.
enough for us to agree that seeing photographs is strikingly
analogous to ordinary seeing. There are kinds of judgements we
make in cases of ordinary seeing - judgements subject to various
failures of discrimination - which have no counterparts in the
case of seeing photographs. 28
With ordinary seeing, we get information about the spatial
and temporal relations between the object seen and ourselves.
We learn not merely that some possible state of affairs is actual,
but that it is actual here and now. Call this kind of information
"egocentric information". That seeing provides us with egocen-
tric information is connected to the fact that seeing is perspec-
tival. I could not place myself in the world if I saw the world
from no particular perspective. And from what perspective I see
things depends on the location of my body - or at least of my
eyes- relative to the things I see.
In seeing, we often make errors about egocentric information.
It's easy to be mistaken about the distance of a star (think of
Ptolemy's estimates), and small mistakes about the distances
and directions of much closer objects are common. Mistakes con-
cerning temporal egocentric information are harder to make; it
is hard to be mistaken about whether some directly visible con-
dition holds now. But for astronomical distances we are prone
to think, mistakenly, that what we see tells us how the seen
object is now. 29 Photographs, on the other hand, do not convey
egocentric information; seeing a photograph does not tell me
anything much about where the object photographed is in re-
lation to me. Since photographs convey no egocentric informa-
tion, there is no question of my failing to discriminate properly
concerning the egocentric information conveyed by one. 30
28
These and other differences are discussed by Nigel Warburton in "Seeing
Through 'Seeing Through Photographs' ".
'9 Unless, as David Lewis suggests, "the stars, as I now see them, are not
straightforwardly past; for lightlike connection has as good a claim as si-
multaneity-in-my-rest-frame to be the legitimate heir to our defunct concept
of absolute simultaneity" ("Veridical Hallucination", p. 2j'7Il).
30
Photographs can serve, along with information from other sources, in an
inference to egocentric information. If I know where and when the shot was
taken, and where I am now (and what the time is now), I may infer that the
scene depicted stands in a certain spatiotemporal relation to my current time
66
Ordinary seeing also enables us to track objects and processes

across time. As long as I am seeing the object, changes in its
position and visible appearance will be reflected in changes in
my visual experience of it. There are discriminatory errors we
are liable to make in processing this kind of information. We
may think it took the egg three minutes to boil when in fact it
took three minutes and four seconds. We can be mistaken about
exactly what happened during a time interval, thinking the ob-
ject changed its shape, colour or position in one way when in
fact it changed in a slightly different way. None of this has any
counterpart in photography, since photographs do not give us
temporally extended information. 3'
When we achieve a proper perspective on the range of dis-
criminatory errors characteristic of vision, we find only a narrow
overlap between ordinary seeing and seeing photographs. The
overlap is no greater than that between the discriminatory errors
characteristic of perceiving heat and seeing thermometers. 32
There are no grounds here for saying perception of a photo-
graph is perception of the object photographed.
In that case photographs really are not like lenses and mir-
rors. We get a sense of the difference between lenses and mirrors
on the one hand and photographs on the other if we compare
the situation of a creature that has, by gift of nature, lenses or
mirrors as aids to sight, with the situation of a creature which
slice. But the egocentric information available in ordinary seeing is not ob-
tained by inference from the visual experience together with other informa-
tion. Clearly, there are all sorts of inferential paths to egocentric information
which do not count as perceptual paths.
3
' Photographs - particularly a series of photographs - can serve to inform us
about temporal change, when we combine them with information from other
sources. But as with egocentric information, the inferential paths are not per-
ceptual paths.
3
' It is instructive to compare the case of prosthetic vision via the surgeon
(Section 2.6) with the following case. The surgeon gives me visual experi-
ences, but he does not ensure a match between my visual experiences and
the scene before my eyes. Rather, he gives me visual experiences correspond-
ing to a collection of photographs he has taken at various times and places,
quite unrelated to the time and place I now occupy. I have no inclination to
say that in this case my sight has been restored.
has visual experiences approximating the condition of one who

sees photographs. Suppose there are creatures which have
evolved little telescopes in front of (or perhaps as part of) their
eyes, or that live in a permanently dark environment but which
have evolved periscopes which reach up to a level where there is
usable light. We would hardly doubt that these creatures could
see. Nor would we doubt that they do, in fact, see under condi-
tions normal for their proper functioning; we would not say they
had the capacity to see but were prevented from doing so by the
presence of their biologically made lenses and mirrors. Suppose,
on the other hand, there are creatures who have evolved cameras
instead of eyes. Instead of receiving a continuous flow of visual
information from their surroundings, they have a device which
takes a snapshot of the local environment, "prints" it on the pho-
tographic analogue of the retina, and sends the visual informa-
tion to the brain, which then - by a process as mysterious in them
as seeing is in us- produces a visual experience, that of a static
image which may last, say, five minutes before it fades. (To make
these creatures' organs more closely analogous with our experi-
ence of photographs we would have to assume that there is nodi-
rect correlation between when and where the snapshot is taken
and when it comes to the creature's consciousness: suppose the
photographic images are jumbled so that the creature's visual ex-
perience now might equally be the result of a shot taken five
minutes ago or five years ago, and that there is nothing in the vi-
sual experience itself to tell the creature which it is. And we must
also suppose that a shot could be a shot of something that is not
now in the creature's local environment but of something in a dis-
tant part of the world.)
Does this second creature see things in the external world? I
think not. It has visual impressions. But the visual impressions
it has do not relate the creature to the world in a way that would
justify our saying that the creature sees the world. It is like a
person who sees only photographs of things and not the things
themselves. Photographic images, for all their sharing of coun-
terfactual dependence with ordinary seeing, belong to a medium
very different from lenses and mirrors; so we have good reason
68
to say lenses and mirrors are transparent, while photographs

are not.
2.7 PUTTING MOVEMENT IN THE

PICTURE
How would the argument I have given against Transparency be

affected if we considered cinematic images instead of still pho-
tographs? The significant difference is that with cinematic im-
ages we do, or can, get transtemporal information; watching a
cinematic image, we are in a position to judge, just by watching,
that a certain process took so long to occur, and that one process
or event occurred before or after another. That is one way in
which seeing cinematic images is like -and seeing still photo-
graphs unlike- ordinary seeing. Nevertheless, cinematic images
are like photographs in failing to give us egocentric information.
Watching cinematic images, we do not get information about
the relation of what is there depicted to our own spatiotemporal
position.
Does the presence, in a cinematic image, of transtemporal in-
formation make such images transparent? I'm not sure. Perhaps
we are now in an area where the concept of perception is vague,
and there simply is no answer to whether we perceive things
when we see cinematic images of them. All the same, I'm in-
clined to say this: the absence of egocentric information in the
cinematic case is a significant factor that dominates the presence
of transtemporal information. The function of seeing is to give
us egocentric information. 33 If it did not do that it would not
contribute to our survival and thence to its own flourishing.
Seeing cinematic images does not retain that crucial ecological
connection with the environment; that justifies us in saying that
cinematic images are not transparent.
What then of mirrors? We are happy to say we see things in
33
David Marr wrote: "What does it mean to see? The plain man's answer (and
Aristotle's too) would be, to know what is where by looking" (Vision, p. J).
It is clear, I think, that "what is where" means in this case "what is where
in relation to the one who sees".
mirrors, but if the mirrors are confusingly iterated, with each

reflection reflected yet again, we can be in a position where we
get little or no idea about the present spatial position of the
object relative to our own position. So mirrors are transparent,
even though, on occasion, they do not convey egocentric infor-
mation. So my case against the transparency of cinematic images
collapses. But do we always see in mirrors? I think not. The rea-
son we are happy to say we see things in mirrors is that we
think of mirrors as functioning "normally" so that they do retain
egocentric spatial information; this is especially the case since
the thing we most often see in a mirror is ourselves! If our nor-
mal experience of mirrors was of confusing iterations that de-
nied us that information, we would be less inclined to say they
were transparent and more inclined to say that what we see in
them are representations.
Earlier I mentioned a widely held view which I said was in-
correct: that seeing requires the existence of an uninterrupted
light ray. We can now see why it is both plausible and incor-
rect. Interruption to the light ray normally results in there be-
ing an unpredictable spatiotemporal gap between one ray
transmission and the next, thereby destroying egocentric infor-
mation. The transducer I described is an exception exactly be-
cause it allows no such gap and hence preserves egocentric
information. That's why it is transparent; by means of it we
see things beyond the screen. Interrupted transmission will
typically result in loss of egocentric information, but not al-
ways.
2.8 THE SIGNIFICANCE OF SIMILARITY
The friends of Transparency might agree with my arguments,

and protest that they do nothing to undermine the substance of
their thesis. For I have agreed that seeing photographs and or-
dinary seeing are alike in sharing natural dependence, and that
both of these are, in this respect, unlike seeing paintings. Isn't
that enough to establish a fundamental commonality between
seeing photographs and ordinary seeing, and a sharp break be-
tween seeing photographs and seeing paintings? It is not. Not
70
every similarity is fundamental, nor every difference a sharp

break. I have argued that, while photography is naturally de-
pendent and painting is not, this difference does not make paint-
ing and photography radically different, the one a form of
representation and the other not.
This is not to say that the natural-intentional dependence
distinction is unimportant, but merely that it does not mark
the divide between perceiving things and perceiving mere rep-
resentations of things. Instead, it marks the distinction be-
tween two kinds of representations. Thermometers give us
signs, or representations, of heat, not perceptual access to heat
itself. They, and other "mechanical" measuring devices like
barometers, ammeters and mechanically generated descrip-
tions, provide us with what we might call natural representa-
tions, since they display natural dependence. 34 Paintings,
intentional uses of language, dumb shows and the like also
give us signs or representations - intentional representations,
since they display intentional dependence. Representations,
both natural and intentional, give us information about things
without giving us perceptual access to them. The two kinds of
representations differ from one another in how that informa-
tion is generated.
Now we see where the argument from natural dependence
to Transparency goes wrong. That photographs are not inten-
tional representations does not mean that they fail to be rep-
resentations altogether. Once we make a distinction within the
class of representations between the natural and the inten-
tional, we can do justice to our intuition that photographs and
paintings are different without having to go to the extreme of
saying photographs give us perceptual contact with the things
they are photographs of. We may say instead that photo-
graphs are natural representations and paintings intentional
ones.
34
They are natural representations because the processes by which they func-
tion exhibit natural dependence. They are representations because they are
used as such by us. I am not suggesting that something can be regarded as
a representation (natural or otherwise) without reference to human intention.
2.9 THE AESTHETICS OF PHOTOGRAPHY
If we do conclude from all this that directly seeing a building

and seeing a photograph or cinematic image of one are signifi-
cantly different activities, rather than just different ways of
seeing the building, certain other arguments concerning
photography and film collapse. Roger Scruton argues that "rep-
resentation can never be achieved by photography alone". He
asks us to compare the photographer with someone who artfully
places a frame in a street so as to give an aesthetically pleasing
view of certain of its buildings and their interrelations. And he
asks, "How could it be argued that what I see in the frame is
not the street itself but a representation of it?" 35 I agree that the
scene inside the frame is not a representation of the street. It
does not follow that a photograph of the street is not a represen-
tation of the street. For that to follow we would need a con-
vincing argument that photography is transparent, and I do not
believe there is one.
Another claim Scruton makes is that the ''brute" dependence
of photographs on what is photographed means they are inca-
pable of expressing thoughts about their subjects. A picture ex-
presses a thought when we can see that thought as having
guided the production of the picture; but the kind of guiding
that is relevant here is absent in photography, because of brute
dependence. 36
The only plausible sense in which photographs are brute de-
pendent on their subjects is the one I have examined in connec-
tion with Walton's argument: photographs are naturally
dependent on their subjects. And that kind of dependence is
consistent with there being, and our recognizing, all kinds of
intentions that have played a role in the production of the pho-
tograph. For example, the photographer exerts intentional con-
trol in her choice of angle, lighting and subject. Scruton is aware
that photography is in various ways under the agent's control,
35
Roger Scruton, "Photography and Representation", p. 120.
36 "With an ideal photograph it is neither necessary nor even possible that the
photographer's intention should enter as a serious factor in determining how
the picture is seen" (ibid., p. 111).
and that the outcome of the photographing process is dependent
on various decisions the photographer makes, but he finds these
facts insignificant when weighed against the (supposed) trans-
parency of photographs. We could, he says, introduce an art of
mirrors, which by their cunning placement, reveal what they
mirror in aesthetically interesting ways. But this would not be
a representational art, nor would photography be one. The an-
swer to this is the same as that to the analogy of the framed
building: the mirror art would not be a representational art
because a mirror is not a representation. Mirrors are genuinely
transparent (so long as they retain egocentric information)
whereas, I have argued, photographs are not. So nothing follows
for the case of photography from that of mirrors.
Scruton also claims that the emotional or aesthetic qualities
of a photograph derive directly from the qualities of what it
represents; a photograph is sad because its subject is sad, touch-
ing because its subject is touching. 37 Not always, and certainly
not always in film. A photograph of people obviously happy
can be sad, if the photograph suggests that these people have
little reason to be happy, or because we know that their hap-
piness is a prelude to sorrow. Scruton might argue that in that
case the subject itself is sad, because the subject is, say, happy
people who will or should be unhappy. 38 But if, as Scruton sup-
poses, the sadness of the photograph has to be derivative on the
sadness of its subject, the photograph can be sad, or express
sadness, only if, as a matter of fact, those people will or should
be unhappy. And a photograph of smiling people can be com-
posed in such a way as to look sad, or to express the photogra-
pher's sadness or the sadness she intends to convey, irrespective
of whether the photographer is right in thinking these people
will or should be happy.
It might be hard, in any particular case of a photograph or
cinematic image, to assign the effect of sadness as between sub-
ject and composition; perhaps there will always be some
grounds for saying that the sadness derives exclusively from the
37
Ibid., p. 115.
38
"Should" in the sense of "having reason to be".
73
subject. But if, as I think we should, we reject the Transparency
Thesis, we undercut the motivation for this move. Someone who
holds Scruton's view ought to ask: apart from the claim that
photography is transparent, is there any reason to insist that the
expressive quality of a photographic image is dependent on its
subject in a way that the expressive quality of a painting is not?
One answer might be that the photographer I cinematographer
lacks control over her medium by comparison with the painter,
and that this lack of control is a barrier to expression. Indeed,
Scruton says that a photographer who aims for aesthetically sig-
nificant representation must also aim to control detail; but detail
is hard to control in a photograph, and if it is controlled there
are "few ways in which such intentions can be revealed in the
photograph" .39 The point seems to be that an audience cannot
tell, in looking at a photograph, which elements of detail were
determined by intention, which by accident. Again, this simply
is not true of all photographs and certainly not of all cinematic
images. A photographer can, under certain circumstances, con-
trol detail very finely, and can make it clear that the detail was
controlled. All that can be said is that detail is easier to control
in painting, where each brush stroke is, ideally, the product of
intention, and that, in painting, the intention to control is easier
to signal. But such an argument cannot establish that photog-
raphy is not a representational art: at most it establishes that it
is less rich in its representational possibilities than painting is.
Perhaps it doesn't even establish that. There are aesthetically
relevant features of works of art that depend for their effect
partly upon our recognition that they are difficult to achieve. In
examples of Chinese calligraphy, we admire the elegance and
smoothness of the brushwork, and we do so partly because we
realize how difficult it is to achieve elegance and smoothness in
a medium that requires the artist to work very fast - any hesi-
tation will result in a blotch of ink. It would be a misunderstand-
ing of aesthetic value to suggest that calligraphy would be a
better art if it employed a medium in which a greater degree of
control could be exercised in the production of a character; it
39
Scruton, "Photography and Representation", p. 117.
74
would eql!-ally be an error to suggest that a synthesizer is a

better musical instrument than a violin because it is capable of
producing a sequence of notes faster than any violinist could.
An improvement in the degree of control possible for an artistic
medium does not automatically mean an improvement in that
medium's expressive power.
2.10 PHOTOGRAPHY, PAINTING AND

THE REAL
A significant fact about photography and cinematography- one

that distinguishes both rather strongly from painting - is said
to be this: just as one can see only that which exists, there can
be photographs only of things that exist. 40 There can be a paint-
ing of a unicorn, but no photograph or cinematic image of one,
since there are no unicorns.
How significant is this difference? Allow me to stipulate that
the phrase "the painting Pis of M" is to mean the same as "M
was the model for the painter of P". This stipulation will bring
our use of "the painting is of M" into line with our use of the
phrase "the photograph is of M", in the sense that the truth of
each will require (i) that M's presence is a significant causal
factor in the production of the photographic or painted image,
and consequently (ii) that M exists. In this sense, no photograph,
and no painting, can be of a nonexistent.
I grant this is not how we normally use the expression "the
painting is of M", since we allow that a painting can be of
something which does not exist, or of something other than the
model, as with biblical subjects. But my point is that there is a
relation in painting, call it what you will, between painting and
model, that is as existentially committed as the relation between
a photograph and what it is of; I have chosen to call it the "of"
relation.
"" Seeing through photographs "is like the ordinary variety [of seeing] in that
only what exists can be seen"; Walton, "Transparent Pictures", p. 254. See
also Scruton, "Photography and Representation", p. 112: "If a photograph is
a photograph of a man, there is some particular man of whom it is a pho-
tograph."
75
My opponent might grant this, and still insist there is an
asymmetry here: as I use the expression "of", being of, and rep-
resenting are relations that come apart in painting, but not in
photography. A painting can represent something other than
what it is of in my sense; and a painting can represent something
without being of (in my sense) anything. But a photograph must
be of something and what it is of is what it represents. A pho-
tograph cannot represent a unicorn, and if I dress someone up
to look like Saint Anne and photograph her, I do not end up
with a representation of Saint Anne, but a representation of (in
other words a photograph of) the model so dressed up; I end
up with a representation of a representation.
Consider the photograph of the model dressed as Saint Anne.
I think it is merely prejudice to say this cannot be a represen-
tation of Saint Anne. An exactly similar prejudice might have
been evident at the time people started using models to paint
pictures of biblical characters; a prejudice that would have
showed itself in the insistence that the painting represents the
model and not Saint Anne. We've now lost that prejudice (if we
ever had it) in painting, through habituation to the conventions
of the medium. The only argument I can think of in the case of
photography is that the relation of natural dependence between
model and photograph somehow precludes the photograph rep-
resenting anything other than the model. But consider an anal-
ogous case: the craft of hand shadows. By setting my hands in
a certain way and having their shadow appear on a wall, I pro-
duce a representation of Abraham Lincoln. 4 ' Here there is
natural dependence, just as in photography, between the
disposition of my hands and the shape of the shadow on the
wall. 42 Following the logic of the argument just given, we should
say that it is merely my hands that are represented, and not
Lincoln at all. That is not very plausible. Nor is it plausible to
deny that in cinema we have representations of the story's char-
acters. Natural dependence ought not to preclude a photograph
4
' I take the example of shadow play from Maynard, "Drawing and Shooting".
42
That is, if we keep fixed the beliefs of the shadow maker, the appearance of
the shadow will depend counterfactually on the disposition of my hands.
or a cinematic image representing something other than what it

is of.
One difference we must allow between photography and cin-
ematography on the one hand, and painting on the other: while
every photograph is of something, not every painting is. But this
is no argument for the transparency of photography, for the
same holds of any natural sign. In the sense in which every
photograph is of something, every pattern of rings on the cross-
section of a tree is also of something: it is of the age of the tree.
But for all that, the rings are still representations of the age. And
when we say a photograph is always of something we mean a
photograph produced in the standard way - it is possible to
draw or scratch on the negative, or to manipulate light so that
it functions much as a paintbrush would, in which case you
have a photograph that represents without being of anything.
So the fact that photographs, as standardly produced, are al-
ways of something cannot be a reason for denying them the
status of representations, unless we are prepared also to say,
implausibly, that the cross-section of the tree does not represent
the tree's age.
So is there nothing special about the photographic method,

nothing distinctive about it as a medium of representation?
There is something special about photography and cinematog-
raphy: natural dependence. Because a photograph is naturally
dependent on its subject, we may feel that its subject is more
intimately connected with the photograph than it would be with
a painting, and more intimately connected with us when we
look at the photograph. But the natural dependence of photo-
graphs and of cinematic images does not take them outside the
realm of representation. Photographs are natural representa-
tions, or natural signs of things, as footprints are natural signs
of the people who make them, and the pattern of rings on the
cross-section of a tree is a natural sign of the age of the tree.
Paintings are nonnatural, intentionally mediated, signs.
In further ways, painting and photography might be repre-
sentationally asymmetric; there might be representational fea-
tures paintings can have but photographs cannot, or vice versa,
77
and there might be representationally dependent values paint-
ings can have but photographs cannot, or vice versa. Painting
and photography may both be representational media, and rep-
resentational arts, even while there are considerable differences
between them from the points of view of representation and of
art. But none of these differences between them entail that pho-
tographic and cinematographic images are transparent.
I haven't established that photographs or cinematic images are
merely representations of things. To prove that we don't see
people when we see photographs of them would require me to
provide necessary and sufficient conditions for perception and
to show that seeing photographs of things doesn't satisfy those
conditions. But all the available candidates for such necessary
and sufficient conditions are too controversial to carry much
weight in the present dispute. My strategy has been defensive.
I start from the position of one who believes photographs and
cinematic images, like paintings, are representations, that we
perceive representations of things when we see photographs,
and not the things themselves. The arguments for Transparency
reviewed here do nothing to undermine this view. In so far as
the arguments point to a difference between photography and
painting, I accommodate the difference by saying that photo-
graphs are natural representations while paintings are inten-
tional ones. And this is no ad hoc move. We commonly
recognize a distinction within the class of representations be-
tween the natural and the intentional. Paintings, gestures and
acts of speech are all intentional representations; thermometers,
barometers and the ringed cross-section of a tree are all natural
representations - of temperature, weather and the tree's age,
respectively.43
43
The tree rings and the thermometer are favoured examples in the develop-
ment of causal theories of representation. See, e.g., Dennis Stampe, "Toward
a Causal Theory of Linguistic Representation"; and Robert Stalnaker, Inquiry,
p. 12.
Chapter 3
Realism
The guiding myth, then, inspiring the invention of
cinema, is ... a re-creation of the world in its own
image.
Andre Bazin
In Chapter 1, I distinguished three theses about cinema: Illu-

sionism, Transparency and Likeness. I have been hard on the
first two. In this chapter I want to defend the thesis of Likeness,
which has been under attack for a while now from those who
reject the notion of likeness or resemblance between images and
the things they are images of, and who stress the artifice, the
conventionality, the "codedness" of cinema. This is one aspect
of their rejection of film realism. I, on the other hand, hold that
Likeness is a defensible version of film realism. I also want to
defend the claim that there are styles of film making which are
especially realistic in the sense explicated by Likeness: long-take,
deep-focus style is a notable example. But I want to avoid a
misunderstanding. My defence of Likeness is metaphysical, not
aesthetic. I am not advocating that film makers adopt styles
which, like long-take, deep-focus style, attempt to exploit the
possibilities for realism in film. I am arguing that Likeness is a
coherent thesis, and that it is possible to achieve a considerable
degree of this kind of realism in film. Whether you think that is
a worthwhile project is another matter.
First I shall offer a general theory of pictorial representation,
or depiction, according to which depictions are like the things
they depict. In this sense depictions are, to various degrees, re-
alistic. The characterization of depiction I shall give will allow
us to say that film depicts space and time; film does, or can,
represent space and time realistically.
79
3.1 DEPICTIONS
Though philosophers have denied it, certain pictures do seem
to be like their subjects; not so like them as to be indistinguish-
able, except under very peculiar circumstances, but like them
nonetheless. Indeed, the denials of philosophers in this regard
have an air of paradox. The likeness we think there is between
some pictures and the things they are of is, of course, a matter
of likeness of appearance; no one is (or should be) claiming that
people and pictures of people belong to the same natural kind,
or have the same essence. So what is claimed is that certain
pictures and their subjects appear alike: that the experience of
looking at the picture is in certain respects like that of looking
at the subject. But how could philosophers be better placed than
ordinary folk when it comes to judging what experiences are
like others? Denying that pictures and their subjects can be alike
in this sense seems well beyond the competence of a philoso-
pher.
The sense in which pictures (or, as I shall sometimes say,
depictions) can be like their subjects is not merely a matter of
being "true to" the subjects. A description of a man can be ac-
curate, without the experience of reading the description being
anything like the experience of looking at the man. There can
be accuracy without likeness, as a description can be accurate
without being like its subject. When we read a description, we
may recognize what is being described, and when we look at a
depiction we may recognize what it depicts, but we do these
things in fundamentally different ways. To comprehend the de-
scription I deploy my linguistic capacities, my understanding of
the semantics and syntax of the language. To see that the picture
is a picture of a horse, I deploy my horse-recognition capacities.
That is, I use the same capacity to recognize the picture of a
horse that I use to recognize a horse.'
We recognize objects when we see them by recognizing cer-
' I think Noel Carroll was the first person to give this characterization of re-
alism; see his "Power of Movies". See also Flint Schier, Deeper into Pictures. I
follow Schier's talk of "triggering recognitional capacities".
So
Realism
tain of their spatial features. I count shape, aspects of shape,

colour, size and position as spatial features, and I count a con-
junction of spatial features as itself a spatial feature. Among
spatial features I also count higher-order features - spatial re-
lations between spatial features, for instance. In this very general
sense of spatial feature, our capacity to recognize an object when
we see it must be a capacity to recognize some of its spatial
2
features. It might be a disjunctive capacity: being able to rec-
ognize an F might consist in associating a certain list of features
a" ... , an with the concept of an F in such a way that detection
of the presence of any one of the a;'s in an object is sufficient to
enable you to recognize that object as an F.
So my visual capacity to recognize a horse is the capacity to
associate some visual feature of what I see with the concept
horse, thereby enabling me to bring what I see under that con-
cept. In that case, when I see that the picture depicts a horse I
must associate some visual feature of what I see, namely a pic-
ture, with the concept horse. What the picture of the horse has
in common with the horse is some spatial feature which triggers
my horse-recognition capacity. It is in that sense that the picture
and the horse are alike. And their being alike in this respect is
consistent with them being very different in other respects.
Some people never tire of telling us, for example, that pictures
are flat and horses are not. So what? The likeness between pic-
ture and horse need not be likeness in any respect that is im-
portant for the horse. It need not be a matter of overall likeness
explicable, say, in terms of sharing a preponderance of features
picked from some favoured class. The likeness might be quite
insignificant or artificial when judged from any perspective
other than that of someone who wants to recognize horses and
pictures of them by looking. But for such a person, the likeness
is a significant one.
I said that the likeness between picture and horse is one of
appearance, and I have said that this likeness of appearance is
a matter of their sharing properties significant for our recogni-
tion of horses. Someone might object that two objects which
, See Jerry Fodor, "Meaning and the World Order".
81
share a property that triggers our horse-recognition capacities

could appear quite different from one another. After all, plenty
of things share certain spatial features without appearing simi-
lar. If that is the case, the likeness between horse and picture
could not be explained entirely in terms of sharing recognition-
ally significant properties, and there would be some missing
ingredient in my account of depiction.
In fact, no ingredient is missing. When we say that horse and
picture look alike, we acknowledge also that they look quite
different. We don't, for example, mistake the one for the other.
Indeed, it can sometimes be very hard for us to identify explic-
itly the respect in which they look the same. There is a sense in
which, while horse and picture trigger the same recognitional
capacity, they look different. Their also looking (in some way)
alike is, I suggest, just a matter of their triggering the same re-
cognitional capacity. In the language of Australian materialism:
saying that horse and picture are alike means just this:
"something is going on in me when I look at the picture like
what goes on in me when I look at a horse." What is going on,
although most of us do not know it, is that our horse-recognition
capacity is being triggered. Sometimes, when we recognize that
the picture is of a horse, we can point to some spatial feature
that horse and picture have in common, and that feature may
indeed be the one which triggers the relevant capacity. But iden-
tifying such a feature is not necessary for us to judge that horse
and picture are alike. Also, recall that the capacity to recognize
an object of a given kind might be sensitive to a disjunction of
properties. So the spatial property of the picture which is re-
sponsible for triggering my horse-recognition capacity need not
be the same feature which would, if I were looking at the horse
itself, trigger that same capacity. All that is required is that both
spatial properties belong to the same favoured disjunction; both
will, in the right circumstances, trigger my capacity to recognize
a horse.
If pictures of horses trigger horse-recognition capacities,
doesn't that mean that pictures of horses fool us into thinking
that we are seeing horses when we see them? Is this not the
discredited view that depictions of things are, by their very na-
82
Realism
tures, devices which create in us the illusion that we are seeing
the things themselves? In reply I say that having your horse-
recognition capacity triggered and judging that there is a horse
in front of you are different things. They differ as to the level at
which these operations are conducted in the mind. Judging that
there is a horse in front of me is something that I do; it is an
operation conducted at the personal level. Having my horse-
recognition capacity triggered happens at a lower, subpersonal,
level of functioning; it is something that happens within me. I
shall explain. 3
The view of the mind prominent in Western philosophy since
Descartes has been something like this: the mind is a unified
and indivisible organ, transparent to itself, and identical with
the person whose mind it is. I am my mind, and my relation to
my bodily states and to other "external" features of the envi-
ronment is incidental. I do not know anything with certainty
about the external world - not even that I have a body - but I
can have certain knowledge of the contents and workings of my
own mind. Contemporary philosophy, at least within the
broadly analytical, Anglo-American tradition, has largely re-
jected this picture. Most contemporary philosophers reject the
idea that the mind is separate and distinct from the body, and
they reject the idea that the subject has infallible access to at
least a large class of his or her own mental states. They think
that what we are learning about the brain tells us important
things about the mind, in particular that the mind is not espe-
cially unified. Different mental processes and functions are lo-
calized at different places in the brain, and the appearance that
the mind is a smoothly operating seamless web quickly breaks
down when we look at cases where there is selective damage at
particular sites of the brain. We know, for example, that the
capacity to recognize objects can continue when the subject has,
through injury or disease, lost the ability to locate objects in
space, that the capacity to recognize certain kinds of objects can
persist when the capacity to recognize others, for example faces,
3
I am indebted here, as elsewhere in this chapter, to the work of Flint Schier.
See his Deeper into Pictures, especially section 9·3·
has gone. We know that severing important connections be-
tween the two halves of the brain can lead to strange situations
in which one half "knows" something which the other half does
not. More and more, philosophers and cognitive scientists are
coming to the view that the mind is a complex system of hier-
archically organized subsystems, where information may regu-
larly fail to pass from one subsystem to the other. Crucially, the
subject herself may know very little about the overall architec-
ture of the mind or about what is going on in it. Indeed, we can
think of the person as constituted by a hierarchy (or complex of
hierarchies) of intelligent creatures or homunculi. The farther
down the hierarchy you go, the less intelligent is the homun-
culus carrying out the operations at that level, until we reach
the ground floor, where intelligence bottoms out into straight-
forward causal interactions where notions of information, rea-
son, evidence and inference play no role, and where everything
that happens is driven by brute causal powers in accordance
with natural law. The person or agent himself occupies the top
level of the hierarchy and is more intelligent than any of the
homunculi that operate at subpersonal levels. Many operations
of the mind are conducted by the person himself; judging that
there is a horse in front of me, or that there is a picture of a
horse in front of me, is something that I do. But a great deal of
mental processing can be thought of as conducted at a level
below that of the person or agent. The primary insight of the
homuncular or hierarchical view is that, when an operation is
conducted below the personal level, we are not driven to de-
scribe that operation in purely causal, nomological terms. We
can describe it as a task carried out for a certain purpose, em-
ploying information of certain kinds, and conducted within cer-
tain constraints of efficiency, reliability and so forth. That way,
we describe it as a task performed by a subpersonal homuncu-
lus.4
Back to horses, and to pictures of horses. When I judge that
there is a horse in front of me, I take into account a great deal
4
See, e.g., Daniel Dennett, Brainstorms; John Haugeland, "The Nature and Plau-
sibility of Cognitivism"; and W. G. Lycan, "Form, Function, and Feel".
Realism
of evidence from various sources, primarily perception, of

course, but also from memory and from the sorts of general
principles I have developed over the years concerning what
sorts of things are likely to be located in what sorts of places
(art galleries are more likely to contain horse paintings than
horses). The bit of judging I do in this case is, as people say, a
top-down process: the sort of thing Jerry Fodor describes as
"slow, deep, global rather than local, ... characterized by com-
putations in which information flows every which way ... the
higher the cognitive process, the more it turns on the integration
of information across superficially dissimilar domains" .5 But na-
ture has endowed me with other, more automatic, less flexible,
less rational capacities as well, among them a horse-recognition
capacity - a quick-and-dirty mechanism which, somewhere
deep in my visual-processing system, identifies a certain input
as a horse. It does so, not on the basis of a detailed, compre-
hensive examination of the visual input in the light of back-
ground belief and all the rest, but on the basis of just a few clues
extracted from the visual input itself. Fodor has described these
perceptual processes as "input driven, very fast, mandatory, su-
perficial, encapsulated from much of the organism's background
knowledge, largely organized around bottom-to-top information
flow". My horse-recognition mechanism is pretty good at de-
tecting horses successfully, and might, in a more natural envi-
ronment than our own, get it right almost all the time. But
because it is quick and dirty, and responds to a few key horse-
identifying features, it is prone to be fooled by donkeys at dusk,
stuffed horses and, in particular, pictures of horses. And that is
how it should be; in the environment in which we have evolved
we have had - and sometimes still do have - a need for mech-
anisms of object detection that work very quickly, especial-
ly where those objects might be predators or prey. The price
of speed is a proneness to false positives, as with the visual
system's identification of a horse picture as a horse. No
matter; there were few horse pictures around at the time we
5
Jerry Fodor, "Precis of The Modularity of Mind", p. 4· For modularity in per-
ception, see also Marr, Vision.
were evolving our present perceptual mechanisms, and we often
could not afford to wait for the slow, deliberative processes
which issue in such judgements as "There's a horse (lion, etc.)
around here".
Perhaps this view of picture recognition constitutes a sort of
illusionism about pictures. Taking seriously the homuncular
model of the mind, it suggests that there is a not-very-intelligent
homunculus in my visual system who is charged with the task
of matching the visual input with a series of stored models of
known objects and who is fairly easy to fool into thinking that
the visual input derives from a horse, or whatever object is the
best match he can find for the input. 6 But this is an illusionism
we can live with. It allows, exactly, that the person seeing can
recognize a picture as representing a horse without him sup-
posing he is actually looking at one. That way, we can be realists
about cinematic and other images, seeing recognition of the con-
tents of images and recognition of the objects they are images
of as fundamentally similar, without being illusionists about
such images?
In basing my account of depiction on the idea of capacities
for object recognition, I do not mean to suggest that recognition
of depictions depends on having a prior visual experience of the
things depicted. Sometimes it is the other way about; if I have
the capacity visually to recognize echidnas, that's because of my
exposure to depictions of them, since I have never seen an
echidna in the flesh. My claim is not that the capacity to rec-
ognize depictions of Fs is a capacity developed in response to
visually presented Fs. It is the claim that my capacity to recog-
nize depictions of Fs and my capacity to recognize Fs are one
and the same capacity, however acquired. But note that this
claim (call it the claim of capacity identity) is compatible with a
6
For more on this process of object recognition, see Chapter 4, Section 4.8.
7
As Daniel Gilman puts it, "If our perception of form, space and pattern is
largely fixed by these fast, automatic early neurological processes, there
is ... no need for illusion here; we typically have a variety of submechanisms
at play and the organism need not be fooled about what it is seeing just
because some of its mechanisms respond alike to picture and subject" (''Pic-
tures in Cognition").
86
Realism
weak version of the idea that really seeing things has what we
might call "recognitional primacy" over seeing depictions of
them. The weak thesis is this: that it is not possible to acquire
the capacity to see something in a picture unless you already
have the capacity to see something else in the picture, and ex-
ercise that latter capacity in coming to acquire the former one.
Let us say that when an F-recognitional capacity is acquired by
looking at pictures of Fs, it is acquired pictorially. Then, com-
bining the capacity-identity claim and the weak thesis, we say:
For every F, it is possible that the F-recognitional capacity is
acquired pictorially, but it is not possible for every F-recogni-
tional capacity to be acquired pictorially.
It is also a consequence of my explanation of depiction that
you cannot tell for certain whether something is a depiction just
by looking at it; something is a picture if it functions in a certain
way, and not otherwise, whatever its appearance. Thus we
might devise a hieroglyphic version of English in which little
icons replace names, thus: ~~~ is tall," where the icon ~ replaces
the name "Fred". The icon may look like Fred, or as much like
Fred as any picture looks like its subject. It may trigger your
Fred-recognition capacity in virtue of its likeness to Fred. But
the icon is not, in this context, a picture; it serves merely to refer
to Fred, and not to make any claim about him - the claim is
made by the sentence in which the icon occurs. And what is
claimed in this sentence about Fred depends not at all on the
spatial properties of the icon, but merely on its being a referring
device. 8
By explaining depiction in terms of visual capacities to rec-
ognize objects, I have confined the notion of depiction to vi-
sual media. That is certainly in line with our common
understanding of the idea of depiction. But having explained
depiction that way, there is a natural generalization of the no-
tion to other media. Thus, to consider a somewhat artificial ex-
ample, we can conceive an art of olfactory representation in
which smells are represented. How are they to be represented?
8
See Jerry Fodor, The Language of Thought, chapter 2. See also Zenon Pylyshyn,
"Imagery and Artificial Intelligence", p. 179.
All sorts of ways are possible, of course, but one natural way
would be simply to rely on the odour-recognition capacities of
the audience; when the artist wants us to recognize that odour
X is represented, he or she gives us a representation with
odour properties which trigger our odour-X recognitional ca-
pacities. That, roughly, is the function of the scratch cards
handed out to members of the audience at some showings of
John Waters's Pink Flamingos. Such a device is of questionable
usefulness or aesthetic merit. But film commonly and profita-
bly uses depictive representation, in my now-generalized
sense, along another dimension of perception: diegetic sound.
When the film represents, by means of diegetic sound, the
character speaking, we recognize what the character is de-
picted as saying by using our capacity to understand the
speech of others when we hear it. So film is realistic along (at
least) two dimensions: visual images and diegetic sound. By
contrast, the use of subtitles ·is not, in my sense, a realistic
mode of representation, a point I made briefly in the Introduc-
tion. When the film represents a character as speaking by
means of subtitles we recognize what is said not by our ca-
pacity to understand speech but by our capacity to read.
J.2 NATURAL GENERATIVITY
My explanation of depiction can now be used to explain

something that would otherwise be puzzling. Flint Schier and
Richard Wollheim have noted that systems of depiction display
a certain feature: that our understanding of them is naturally
generative. 9 With classical Western depictions, for example, you
can recognize that the picture is of, say, a cow, if you can rec-
ognize a cow; and generally you can recognize that the picture
depicts a <I> if you can recognize a <1>. You might need to over-
come your general unfamiliarity with the system by being
shown a few examples of depictions within the system and
9
The term is due to Flint Schier. For a detailed and illuminating discussion see
his Deeper into Pictures. See also Richard Wollheim, Painting as an Art, p. 77;
and Crispin Sartwell, "Natural Generativity and Imitation".
88
Realism
having it pointed out to you what is depicted - imagine yourself

familiar only with classical portraiture and suddenly presented
with impressionist or cubist works. But very soon you become
familiar with the style, and can "go on" on your own, recog-
nizing that this impressionist work depicts a lily, so long as you
know a lily when you see one. Learning a language, by contrast,
is not naturally generative; it's no good being exposed to a few
German colour words correlated with colour chips, expecting
then to be able to guess what colour is named by other German
colour words. You do not have to learn a "vocabulary" of de-
piction in the way that you have to learn the vocabulary of a
language- a significant point I shall take up in the next chapter.
Natural generativity is just what we would expect on the as-
sumption that we use our capacity visually to recognize objects
in the real world in order to understand the depictive content
of pictures. If we recognize that the picture depicts a horse by
having our horse-recognition capacities triggered, we will expect
that anyone with that capacity will be able see that a horse is
depicted, which is just the condition of natural generativity. In-
deed, it is hard to see how we could explain natural generativity
except by appeal to object-recognition capacities. Suppose, con-
trary to what I have been saying, that the capacity to recognize
that a horse is depicted is distinct from the capacity visually to
recognize horses. To explain natural generativity on that sup-
position, we would have to say that acquiring the visual capac-
ity to recognize an object mysteriously brings with it the distinct
capacity to recognize depictions of that object: a more compli-
cated and less plausible theory than the economical proposal
that these capacities are the same. And if words and sentences
are not depictive, as I have claimed they are not, and do not
draw on our natural capacities for object recognition, we have
an explanation of why our understanding of language is not
naturally generative. 10
Where representation is naturally generative it is realistic in

1
° For psychological results concerning the distinctness of pictorial and linguis-
tic recognitional capacities, see Glyn W. Humphreys and Jane M. Riddoch,
"Picture Naming", in their Visual Object Processing: A Cognitive Neuropsy-
chological Approach, and the references therein.
my sense. For naturally generative systems of representation

work by exploiting our visual capacities to recognize the objects
represented, and so the experience of recognizing a picture of a
horse is in an important respect like the experience of recogniz-
ing a horse when you see one. This kind of realism we might
call Perceptual Realism. Conversely, where representation is not
naturally generative, but works by convention, we should expect
little or no perceptual realism. Where convention prevails, we
can represent a given thing by means of anything which con-
venience suggests, and there is no a priori reason it would be
convenient to represent an X by means of a thing which triggers
our capacity visually to recognize Xs, given the other constraints
that have to be satisfied by a system of conventional signs. And
that is what we find with language. The experience of reading
descriptions of objects is quite unlike the experience of seeing
the objects themselves .
. Describing our capacity to recognize what is depicted in a
picture, Richard Wollheim has spoken of seeing-in. 11 We see the
Duke of Wellington in the picture just as we see the face in the
clouds or the figure in the frosted window pane. Seeing-in is a
psychological phenomenon, a mental capacity we contingently
possess. By what mental mechanism, then, does seeing-in op-
erate? Wollheim does not say. I believe the answer is that it
depends on the architecture of our visual object-recognition ca-
pacity. Because that capacity is quick and dirty, apt to be trig-
gered by a few relevant cues and therefore subject to false
positives, the mechanism which recognizes horses can be trig-
gered by a picture of a horse. To see a horse in the picture is to
have your horse-recognition capacity triggered by the picture
and thereby to judge that you are looking at a picture of a horse.
3·3 NONDEPICTIVE REPRESENTATION IN

FILM AND OTHER MEDIA
Realism of the kind I'm proposing is a matter of degree. Suppose

we have a representation, R, of an object, A, and R represents A as
" Richard Wollheim, Art and Its Objects.

Realism
having properties F and G. R might represent F realistically, and

G in some other way. R is such that you can recognize it as rep-
resenting the F-ness of A in virtue of your visual capacity to rec-
ognize the F-ness of an A when you see one. But you can
recognize R as representing the G-ness of A only in virtue of your
knowledge of some convention of language, or perhaps of
someone's intention. When we say that this representation or
mode of representation is realistic, and that one not, we probably
mean that this one is more realistic than that one: it represents
more features realistically than the other does. It will be impor-
tant to bear this in mind.
Since a depiction can, in my sense, be a depiction accessible to
sensory modalities other than sight, depictions, including film,
may cross sensory boundaries without ceasing to be purely de-
pictive. So a sound film can be depictive in all its visual and au-
ditory aspects. However, it would be rare for any film to be
wholly depictive in its representation. Films, like works in other
media, represent more than they depict. Films often represent
things which could not be depicted, because what is represented
is not a matter of the spatial properties of things, or of properties
accessible through other senses. A cinematic image may repre-
sent the man it depicts as sad or as angry. These qualities are not
depicted, because sadness and anger are not qualities that could
be depicted; they are mental qualities not accessible to vision or
the other senses. But often this nondepictive representation is
strongly tied to what the work does depict. Usually, the represen-
tation of the nonperceptual quality of sadness occurs in virtue of
the depiction of visible qualities; the man is represented as sad in
virtue of being depicted as having a sad expression.
Sadness is the sort of thing which cannot be represented de-
pictively and which, if it is to be represented at all, must be
represented in some other way. Interesting cases arise when the
medium allows a choice; a certain feature can be depicted, and
it can be represented in another way, and the choice between
the ways can have artistic and narrative significance. Nothing
better illustrates this than film's capacity to represent space and
time. Film can depict space and time, but this is not the only
way it can represent them. Let us look at the ways an art can
91
represent space and time. That will lead eventually to a discus-

sion of realistic film style.
3·4 THREE KINDS OF TEMPORALITY

Everything that happens happens in time, and everything is
temporally related to everything else. It takes time to watch a
film, but so also does it to watch a play, read a novel, listen to
a symphony and look at a painting. In this sense every art is a
time art. But in some way, the cinema is very distinctively an
art of time; to see what is distinctively temporal about it will
require some careful handling of distinctions. Understanding the
temporality of cinema will raise a question I shall answer only
in Chapter 6: What role does tense play in cinema?
A word about terminology. Talk of space and time in this
context really refers just to the placing of people and objects in
spatial and temporal relations with one another, or making it be
fictional that characters and fictional objects stand in such rela-
tions. Architectural writers speak portentously of so-and-so's
creation and manipulation of space. In fact architecture is not a
kind of metaphysics, and all the architect can do, irrelevant
questions about relativity theory aside, is to place things in spa-
tial relations. The same is true of the film maker. A substantive
theory of space and time, according to which space and time
are more than just relations between things, might be true. But
film makers are not in the business of representing space and
time themselves, supposing there are such things; their concern
is with spatial and temporal relations between objects and
events. If occasionally I take a verbal shortcut and speak as if
the film maker's business is with space and time themselves,
please do not take this seriously.
There are three basic ways to treat temporality in art: we
can focus on temporal properties of the work, on temporal prop-
erties of the observer's experiences of the work or on temporal
properties of what the work represents.' (Recall my earlier de-
2
,. See the excellent discussion in Jerrold Levinson and Philip Alperson, What
Is a Temporal Art?
92
Realism
cision to speak as if works can represent fictional things.) Later,

we shall see how these basic ways combine to give something
more complex and more interesting. Concentrating for the mo-
ment on temporal properties of the works themselves, we can
say this: while all works have temporal properties, some have
them more distinctively, more intrinsically and more interest-
ingly than others. In particular, works which involve change
over time seem more distinctively temporal than those that do
not. But it is change of a special kind that is relevant here: not
the change in a painting as dirt accumulates and pigments fade,
but what we might call the unfolding of a play, movie or sym-
phony over time. How do we distinguish temporal unfolding
from mere aging?
Characteristic of unfolding is the presence of temporal rela-
tions between constitutive elements of the work. With cinema,
image succeeds image in a temporal order that must be specified
if one is to say what counts as constitutive of that film; when
you watch the film it matters, from the point of view of under-
standing and appreciating it, in what order you see its images.
For plays, considered as performances rather than as scripts, the
same applies. With the painting that ages (rather than unfolds)
over time, there are temporal relations between constitutive fea-
tures - the pattern of colours we somehow identify as canonical
- and other features - the later altered, and possibly degraded,
pattern. But there are no temporal relations, except, trivially, the
relation of co-occurrence, between constitutive elements them-
selves. Nor are there significant temporal relations between tex-
tual elements of the novel; the words and sentences of the novel
are ordered, but not temporally ordered.'3 In this sense cinema,
theatre and music are temporal arts, while literature, painting
and static sculpture are not. (I have already granted that there
are other senses in which all these arts are temporal.)
As well as relations of temporal order between constitutive
elements of the work, the elements of a work that is temporal
'
3
We can order textual elements temporally if we want to, for example, by
time of composition. But neither that nor any other temporal order we could
impose would tell us anything constitutive about the work.
93
in this sense may have (one-place) temporal properties of du-
ration; the incidents that go to make up the fiction occupy a
definite amount of time. When specifying what is constitutive
of the work we may need to specify these temporal properties
and relations in precise or in imprecise terms. It is constitutive
of a certain movie that each shot lasts exactly a certain time.
Specifications of duration for a play are vaguer because theatre
is a performance art, and the duration of speech and incident is
to some extent determinable by the makers of each performance
- though in any actual performance each action or speech lasts
a precise amount of time.
We get the second of this trio of kinds of temporality in art
if we look, not for temporal relations within the work itself, but
for a temporal ordering of the observer's experience of the work.
I said that the constitutive elements of the novel are not tem-
porally ordered. Sometimes this is denied; but what people seem
to have in mind when they say that the text is temporally or-
dered is that the ordering of elements- nontemporal as it hap-
pens- induces a temporal ordering in the reader's experience of
the novel.' 4 And it is not just any old ordering of experience,
because the layout of the text, together with the standard direc-
tion of reading, imposes a specific temporal ordering. So it is
with at least some paintings, in that they require a temporally
ordered looking, with certain elements to be looked at before
others. They differ from literary works in that the preferred or-
der of experience cannot simply be read off from any order of
the visible elements themselves; those elements do not have an
intrinsic order in the way that words and sentences do.
In this experiential sense, works of any kind can be temporal,
though not all are; some paintings and some static sculptures
are not: for them there is no preferred ordering of experience. If
we are interested in whether a kind of art, rather than just a
particular work of that kind, is temporal, there is a distinction
worth making. Temporality in our first sense- temporality of
4
' As Gerard Genette says, "The narrative text, like every other text, has no
other temporality than what it borrows, metanymically, from its own read-
ing" (Narrative Discourse, p. 34).
94
Realism
the work, or "temporalityw" as I shall call it- is such that if it

applies to any works of a given kind, it applies to all works of
that kind. (At least, I cannot think of any art kinds which are
counterexamples to this claim!) In this sense, temporalityw in-
heres strongly in the kinds in which it inheres at all.' 5 But ex-
periential temporality - "temporalitye" as I shall call it - inheres
strongly in some kinds but weakly in others; strongly in litera-
ture and film, since for any literary or filmic work there is a
canonical temporal order of experience, but only weakly in
painting and sculpture because, as we have seen, there is a pre-
ferred temporal order of experience for some works in these
kinds but not for others. So there are three possible relations
between any candidate for temporality and any art form: strong
and weak inherence, and exclusion, when no work of that kind
is temporal in the chosen sense.
I tum now to the third sense of temporality, which is con-
cerned with the temporality of things represented rather than
with the temporality of that which represents. Cinema is a time
art in this sense also because time, or the passage of time, is one
of the things film represents. But this does not yet distinguish
cinema from any other representational art. All representational
arts portray time in the sense that they portray temporal prop-
erties of, and temporal relations between, the events they rep-
resent. Hamlet portrays the deaths of Ophelia and Hamlet, and
in such a way that it portrays the one occurring before the other.
Painting does the same; Poussin's Renaldo and Almeda portrays
Almeda gazing on Renaldo as he sleeps; the gazing and the
sleeping are represented as simultaneous, and simultaneity is a
temporal relation between events. There are less trivial examples
of the representation of time in static pictorial arts. Pictures may
represent time in a variety of ways: by encouraging the viewer
to make an inference from what is explicitly depicted about
what came before and what will come after; by juxtaposition of
distinct static images, as when we are shown a series of tem-
porally related photographs; by transforming temporal proper-
ties into spatial ones, as in Filippo Lippi's tondo in the Pitti
5
' On strong and weak inherence see the Introduction.
95
Palace, wherein events earlier in the life of the Virgin are rep-
resented deeper within the picture space; and by special tech-
niques such as blurring and multiple exposure. Music, if it is a
representational art, can surely represent spatial and temporal
relations.
The capacity to represent the temporal- call it "temporality/'
- is one further good sense in which something can be a tem-
poral art. But as the examples just given show, it is a relatively
undiscriminating sense; it is hard to think of an art form capable
of representation which is not temporal in this sense. Note, how-
ever, that it is not so undiscriminating as to make every repre-
sentational work a work of temporal art. Paintings can lack the
kinds of narrative contents and associations that would render
them tempora~; in terms of our earlier taxonomy, temporalityr
inheres only weakly in painting. Later, I shall ask whether it
inheres weakly or strongly in cinema. But certainly it does in-
here in cinema; cinema makes the representation of temporal
relations possible. In fact the vast majority of films do represent
temporal properties and relations.
3·5 REPRESENTING TIME BY MEANS

OF TIME
So far we have three senses in which cinema is a temporal art;

we can put them together to get another, and the result is the
most significant sense in which cinema is a temporal art. The
new sense I have in mind is not just the sum of the previous
senses; that summative sense, being temporal in all those three
ways, is a kind of temporality; but we get something more in-
teresting by connecting the three more subtly.
Let us begin with this idea: that an art form is temporal in
this connected sense- I shall call that "temporalityc"- if tem-
poral properties of elements of the representation serve to rep-
resent temporal properties of the things represented. Cinema is
temporalc- What is distinctively temporal about film is not its
portrayal of time, but the manner of its portrayal: its portrayal
of time by means of time.'6 This easily generalizes: the represen-
'6 Is literature temporaJ.:? No: literary works take time to read, and the events
Realism
tation of spatial properties by spatial properties, colour prop-

erties by colour properties and so forth. Let us examine the
general principles here.
An idea that Wittgenstein made much of in the Tractatus is
that when two expressions refer to different objects, we can rep-
resent a relation between the objects by means of a relation be-
tween the expressions. And just as a name that represents an
object need not be the same thing as that object, so the relation
between names need not be the same thing as the relation be-
tween things named. One name's being to the right of the other
might represent one of the things named being to the right of
the other thing named, or represent the thing named on the right
being taller than the thing named on the left, or represent some
other relation between those things. Similarly for one-place re-
lations or properties: Fred being red could be represented by
writing an expression which refers to Fred ("Fred" say), and
writing it, say, in green ink. The greenness of the ink refers to
the redness of Fred. We might have arranged it so that the ex-
pression's being written in red represents Fred as red. In that
case we have what I shall call automorphic representation: the
representation's having property P represents the thing repre-
sented having property P. Automorphic representation is a spe-
cial case of what I shall call homomorphic representation, where
a property of some kind represents a property of that same kind.
So the examples of the red and green ink are examples of ho-
momorphic representation, since in both cases a colour repre-
sents a colour, but only the former is an example of automorphic
representation. Of course the notion of homomorphic represen-
tation is subject to any relativity that attaches to the notion of a
property kind. If what we group together as colours have no
similarity in nature, then the green-red case can be no more than
homomorphic for us. But I shall not be concerned here about
they describe take time, but the length of the time necessary to read the work,
or part thereof, is not in general a guide to the length of time taken by the
events narrated. Mieke Bal says that narrative writing is "iconic" because it
resembles its content "since both contain a lapse of time" ("Description as
Narration", in On Story-Telling, p. n6). But this iconicity is very weak com-
pared with the rich temporal correspondence possible in film.
97
whether this is so, for perceiver-relative homomorphicity is as
much as we shall need for present purposes.
Automorphic representation is subject to a somewhat differ-
ent relativity. What constitutes automorphic representation on
one occasion need not do so on another. Grant that the redness
of the name represents the redness of the thing named; but if
we don't care about precise shading, we shall not indicate the
precise shading of Fred by the precise shading of the name; it
is understood that having the name coloured some shade of red
represents Fred's being some (possibly other) shade of red. And
if being some colour or other is all that matters to us, then having
the name written in some colour can represent Fred's being
some colour. In that case, if the name is written in green and
Fred is actually red we still have automorphic representation
because it is not being green but being some colour or other that
does the representing, and it represents itself. The lesson: judge-
ments about whether we are dealing with automorphic repre-
sentation are hasty if they are made in ignorance of the level of
specificity at which we are operating.
Consider an ordinary kind of picture. Our picture contains,
let us suppose, a representation of Fred; that representation has
a part coloured red, and its being so represents Fred's tie being
red. So pictures contain parts that represent things, but the pic-
tures themselves represent states of affairs, though they do not
always represent actual states of affairs. The states of affairs they
represent may have temporal properties that are also repre-
sented, and there may be temporal relations between the states
of affairs that are represented. Those temporal properties can be
represented in various ways; that is, various properties of and
relations between the pictures themselves can represent those
temporal properties and relations. I have noted that it is com-
mon with painting for spatial relations between the picture's
elements to represent temporal relations between the events de-
picted, as with the cartoon's progression from left to right,
which denotes the progression of events from earlier to later.
Within the same frame, on the same canvas, and indeed, as parts
of the same picture, we may have several different events de-
picted, to be understood as occurring at different times, where
98
Realism
spatial relations between the depictions denote temporal rela-
tions between the events depicted (as in Filippo Lippi's tondo).
But with painting, temporal properties of events are not repre-
sented by temporal properties of representations, and the reason
is clear: painting is tempora~ but not tempora~, so it does not
have the capacity to be temporalc-
While painting is not a temporalc art, cinema, as I have al-
ready noted, is; temporal properties of cinematic representations
(images) serve to represent temporal properties of the events
represented. It is temporal properties of the cinematic represen-
tation that we mostly observe and rely upon in order to figure
out what temporal properties of the fictional characters and
events are portrayed. Similarly, it is by detecting spatial prop-
erties of the cinematic representation that we determine the spa-
tial relations between characters and things in the fiction
represented. That way, film depicts space and time. Film's rep-
resentation of time by time can be automorphic or merely ho-
momorphic. The represented fight lasts five minutes, and its
lasting that long is represented by the relevant representation
onscreen lasting just that long. It is the default setting for cine-
matic interpretation that the representation of duration in cin-
ema is automorphic; it is the assumption we start with, and from
which we move only when some aspect of the narration, some
clash with real-world belief or some combination of the two sug-
gests we should. In Passolini' s Gospel According to St Matthew
the representation of the Sermon on the Mount lasts a few mo-
ments. But changes in background and lighting suggest that the
whole performance lasted much longer and took place at vari-
ous locations. Here the context of narration and real-world belief
conspire to shift our understanding: the changes of scene were
meant to indicate something, otherwise they would not have
been made; landscape and lighting are, and are commonly
known to be, by and large locally stable; so what is probably
being suggested here is discontinuous shifting of place and
time.'7
7
' You might argue that what is represented here is not the whole sermon, but
just a few parts of it, and thus we can preserve the default setting. But
99
Automorphic representation is standard for duration. Is it
standard also for temporal relations of precedence, like occurring
ten seconds before? It may seem not. In film, ten seconds between
shots cannot usually be taken to imply that it is fictional that
the events represented in the first shot took place ten seconds
before the events represented in the second; jumps in time be-
tween shots are too common in film for that to be a safe infer-
ence. But remember that judgements about automorphism are
sensitive to the intended specificity of description. If what is
intended to be represented is the relation occurring some time
after, then a representation that occurs ten seconds after another
thereby occurs some time after it, and automorphism is pre-
served. Once again, however, the default setting can be aban-
doned, as when context suggests that the event represented next
actually occurs before the event just represented. Then we have
the flashback, about which there is more in Chapter 7·
Film has the capacity to represent the temporal by the tem-
poral, so it is temporaic. But film can represent the temporal in
other ways; one way is to have a character or narrator say that
this occurred twenty years after that, or for the words "twenty
years later" to come up on the screen. Fades and dissolves func-
tion as qualitative representations of temporality, indicating
some significant distance in time between successively narrated
events, where context may or may not tell us roughly how
much. A complication is that all fades and dissolves take place
in time, so there is a temptation to say that what we have here
is the joining together of temporal and other properties of the
representation to represent temporal properties of the events
represented - a kind of watered-down version of my tempor-
alityc condition. This may happen, but it is not generally so, and
it would be a mistake to suppose that it is so simply because all
cinematic effects take place in time. Suppose the dissolve takes
two seconds, and the effect of the dissolve is to signal "some
to insist on this move in all cases would be, in effect, to identify what is
represented with what is displayed, and that identification is hard to sustain.
There is a similar device in Clouzot's Wages of Fear.
100
Realism
time later". The time taken for the dissolve can be considered
to make a contribution to the representation of this temporal
relation only if it can be thought of as operating differentially.
That is, it must be the case that if the dissolve had taken less
time or more time, then the dissolve would have signalled a
correspondingly smaller or greater gap between the represented
events. In certain cases, especially when there are several quick
dissolves in succession, one can understand temporal represen-
tation to be working like that; a dissolve longer than the norm
for this film, or longer than the last one, might indicate a pro-
portionately greater time lapse. But if the sign saying "twenty
years later" had been on the screen a few seconds longer than
it was, that would have made no difference to the fact that there
is twenty years between the represented events, and time con-
sequently plays no representing role here.
So there are three kinds of cases: pure representation of the
temporal by the temporal; pure representation of the temporal
by nontemporal properties; and mixed (and, I believe, unusual)
cases where the temporal and the nontemporal conspire to-
gether to represent the temporal.
Sometimes when an art form is said to be temporal, what is
meant is that the form in question is especially adapted to the
representation of the temporal, that its capacity for representing
the temporal is peculiarly rich and subtle. The temporalityc of
cinema partly explains why this medium is said to be rich and
subtle in its representation of time.'8 Homomorphic and espe-
cially automorphic representations of the temporal make for
ease and flexibility. In film we get precise and detailed infor-
mation about temporal duration and relations, and only in the
most obscure narratives, like Last Year at Marienbad, does the
quality of temporal information become significantly degraded.
Yet even here there is, by comparison with, say, painting or the
novel, a rich structure of temporal information concerning the
8
' Gerald Mast may have had something like this in mind when he said, "The
cinema is the truest time-art of all, since it most closely parallels the operation
of time itself." See Mast, Film/Cinema/Movie, p. 112. Quoted in Levinson and
Alperson, What Is a Temporal Art?
101
duration of a remark, a gesture or a glance. It is relations of
precedence that are the first victims of subversive narration in
film, and they are the main casualties in Last Year.
This account of temporalityc employs only two of our three
basic notions of temporality; it omits consideration of the sec-
ond, the temporality of the viewer's experience (what I called
temporalitye>. In fact temporalitye is implicit in the account I
gave, for I have been assuming that the temporal prope~ties of
the work - shot duration and temporal precedence between
shots - is a guide for the viewer in working out temporal prop-
erties of what is represented. But that will be so only if there is
coincidence between temporal properties of the work and tem-
poral properties of the viewer's experience of the work. And so,
in general, there are. The viewer's experience of the shot lasts
just as long as the shot - assuming an attentive viewer. The
representation of the temporal by the temporal works because
of this coincidence, and if it started to break down in some sys-
tematic way - if cinematic images began to cause viewers to
have brief, unpredictable and unnoticed periods of unconscious-
ness- that kind of representation would be undermined.
Earlier I raised the question: does temporalityc inhere strongly
or only weakly in cinema? Only weakly if it is possible for there
to be a film in which temporal properties of representations do
not represent temporal properties of what is represented; oth-
erwise strongly. (Recall that this discussion is relativized to fic-
tion films, and that abstract filmic compositions are not in
question here.) I say that temporalityc inheres strongly in film.
Try to imagine a film in which the passage of screen time has
no implications for the relations between events represented.
Even if the temporal relations between what is represented in
successive shots is in all cases obscure (Was that a flashback?),
individual shots, so long as they last, must, in virtue of their
own duration, imply something about the duration of events
depicted, though the implications can be utterly trivial. If the
film image focuses steadily on the Empire State Building for a
period of time, we can infer that the building stood fixed and
unchanging during that time. Shooting in slow motion through-
out the duration of the film would avoid any automorphic
102
Realism
representation, but there would still be homomorphic represen-

tation; temporal relations between events onscreen would be
proportional to, but not identical to, temporal relations between
events depicted. Film is a strongly temporal art; it cannot but
represent time by means of time.
3.6 SPATIAL REPRESENTATION
Having decided on a significant sense in which cinema is a tem-

poral art, we shall naturally want to know whether it is in the
same sense a spatial art. Unfortunately, the answer to this is
somewhat equivocal.
In cinema, spatial properties certainly are represented hom-
omorphically; spatial properties of representations represent
spatial properties of the things represented. It seems that spatial
properties are not, in general, represented automorphically. It is
rare for the screen image representation of a six-foot man to be
six feet high, and the size of the representation varies with dis-
tance from the camera in a way that does not indicate any var-
iation in the size of the character. One might respond by
recalling my strictures on level of specificity: perhaps the rep-
resenting property is not the actual height of the representation,
but its height relative to other representations. And that relative-
height property can be thought of as representing the same rel-
ative-height property of the representee. That way we would
preserve automorphicity of representation.
But this answer misses something. It is not just relative size
that is represented on the screen; if that were the case, then what
appears on the screen should leave open all questions about
absolute size, which it usually does not. We are not in any
doubt, for instance, that the human characters in most cinematic
fictions are of about the same absolute size as real people. Some
films do include characters of exceptional size (The Incredible
Shrinking Man, Attack of the Fifty Foot Woman), but here the rep-
resentation of exceptional size is achieved by contrast with the
size of representations assumed to represent what is normal for
the human population. We must conclude that, at least in stan-
dard cases, certain absolute spatial properties are represented in
103
film, but clearly not automorphically. How, then, are they rep-
resented?
I think that absolute spatial properties are represented by rel-
ative spatial properties of representations, taken in conjunction
with background assumptions about what is normal by real-
world standards. Thus the relative spatial properties of cine-
matic representations function automorphically to represent the
relative spatial properties of what is represented, and nonauto-
morphically to represent absolute spatial properties.
This discussion should make it clear that the representation
of spatial properties and that of temporal ones in film are dif-
ferently implemented, and that the representation of space is
more difficult than that of time. The representation of temporal
properties is, for cinema, relatively straightforward, for two
reasons. First, time's one-dimensionality means that keeping
track of the duration and succession of onscreen events is rela-
tively simple, whereas the representation of objects in three-
dimensional space on a two-dimensional screen leads inevitably
to a degradation of information. Worse, cinema imposes a double
relativity on spatial properties, for it represents them perspec-
tively; it is relative spatial properties as perceived from a certain
point of view (namely, that of the camera) that are represented.
When the actor advances towards the camera against a fixed
background, the result is a representation of change in the ap-
parent relations of size between character and setting that may
not, and usually does not, represent any such actual change.
Something vaguely like this can, and occasionally does, occur,
for the presentation of time in film; I am thinking of Eisenstein's
"stretching out'' of the time it takes the soldiers to clear the steps
at Odessa in Potemkin. Perhaps this is a representation of the
time those events seemed to take for those participating in them:
a time distinct from that which those events actually took.' 9 But
this capacity of film to represent time in a quasi-perspectival
'
9
Perhaps, but actually I doubt it. In an earlier scene there is a similar, but
very brief, stretching of time, when a sailor breaks a plate. That occurrence
is not plausibly understood as representing the sailor's subjective perception
of the time involved. Rather, it is done for dramatic effect; so, I imagine, is
the scene on the steps.
Realism
way is a device that can be used or not used at will. It is not,

like film's depiction of space, built into the very mechanism of
cinematic representation. And this quasi-perspectival represen-
tation of time in film is not really parallel to the perspectival
representation of space, for reasons to do with fundamental dif-
ferences between the ways we perceive time and space. An ob-
ject may recede from us, thereby coming to occupy a smaller
area of the visual field than another nearer but actually smaller
object; nevertheless, perception represents the more distant ob-
ject as larger. Perception itself compensates for the effect of per-
spective; it represents objects as occupying roughly the amount
of space they do really occupy. But if we see events "speeded
up" or "slowed down", as film enables us to see them, we do
not see them as occupying the amount of time they really do
occupy; instead we have to make an intellectual adjustment: an
inference from the (artificially compressed) time we perceive the
events as occupying, to the time they really do occupy. There
is, for beings such as us, no genuinely perspectival representa-
tion of time.
Cinema's capacity to represent space differs in three impor-
tant ways from that of theatre. The first is that theatre represents
absolute spatial properties automorphically; in general, an actor
of a certain size represents a character of just that size. 20 Second,
the representation of spatial properties in theatre is less per-
spectival than that of cinema, and has the capacity at least to be
completely nonperspectival. A description of the representa-
tional content of the film image would have to specify the point
of view of filming, but a description of the theatrical represen-
tation would not always be a description of what is visible from
one particular point of view. It is true that, in conventional the-
atre with a proscenium arch, a class of roughly preferred per-
spectives would have to be specified, and sometimes the
perspectival representation of scenery defines a single preferred
viewing position.2 ' But theatre in the round without perspectival
"' There can be exceptions, as when the entire action takes place on Mount
Olympus and all the characters are gods.
" See Bordwell, Narration in the Fiction Film, p. 6.
105
scenery achieves something like coordinate-free representation.
(Of course any spectator always sees the action from a certain
perspective, as indeed we always see the real world from a cer-
tain perspective, namely our own. But this is not to be confused
with any supposed intrinsically perspectival nature of the me-
dium itself.) The third difference between film and theatre con-
cerns the mobility of the camera, which thereby provides a rich
source of information about the spatial relations between objects
and compensates somewhat for the limitations of perspective
and failure to represent absolute spatial properties automor-
phically.
3·7 REALIST FILM STYLE

From what I have said about film's depiction of space and time, it
follows that film has the capacity for realism not merely in its de-
piction of objects but in its depiction of spatial and temporal re-
lations between those objects. When objects and events are
represented onscreen within a single shot, we are able to judge
what spatial and temporal relations the film represents as holding
between those objects and events, by using our ordinary capaci-
ties to judge the spatial and temporal relations between objects
and events themselves. We judge the spatial relations between
objects represented in the same shot by seeing that they are spa-
tially related thus and so. We judge the temporal properties of,
and relations between, events represented within the shot by not-
ing that this event took (roughly) so long to observe, while that
one was experienced as occurring later than the other one. That is
exactly how we judge the spatial and temporal properties of
things and events as we perceive them in the real world. 22
.. Edward Branigan argues that the relation between the time of viewing and
the time of the fictional events themselves is conventional, on the grounds
that what one notices during the period of screen time varies across persons
and occasions (Narrative Comprehension and Film, p. 149). But this undoubted
fact has no tendency to undermine the claim that within a shot, the standard
relation between screen time and the time of fictional events is identity, and
its being so relies on no convention whatsoever.
lo6
Realism
It is in this way that the style called long-take, deep1ocus style
- a style which writers like Bazin have argued is inherently re-
alistic - extends the possibilities for realism in film; it enhances
our ability to detect spatial and temporal properties of the fiction
by using the capacity we have to detect those properties of
things in the real world. When discontinuous shots are edited
together, and when depth of field is limited within a shot, our
ability to exercise our visual capacity to detect spatiotemporal
relations between objects themselves is correspondingly limited.
As shot length and depth of field increase, that ability is given
greater scope. (Length of take and depth of focus are independ-
ent of one another, and long take and deep focus do not always
go together, as David Bordwell pointed out to me. But if I am
right about their capacity to enhance perceptual realism in film,
the combination of these two features constitutes something like
a stylistic "natural kind". It certainly explains the close historical
connection between them.)
In montage style, on the other hand, where there is quick
cutting between very distinct spatial (and sometimes temporal)
perspectives, these spatial and temporal properties and rela-
tions have, with greater frequency, to be judged by means of
inference from the overall dramatic structure of the film. Of
course, as my earlier remarks were intended to suggest, this is
a matter of degree; long-take style is more realistic than mon-
tage style, and montage style can itself be said to be more re-
alistic than some other modes of representation - certainly
more so than linguistic description. Unqualified claims that
long-take, deep-focus style is realistic should be taken as im-
plicitly relative to the class of artistic styles with which it is
most naturally compared, namely other cinematic styles, just
as the claim that elephants are large is understood as relative
to the class of mammals.
It is often remarked that deep-focus style is unrealistic in
that it presents us with an image in which objects are simul-
taneously in sharp focus when they are at considerably differ-
ent distances from the camera, whereas objects at compar-
able distances from the eye could not be seen in focus to-
107
gether. 23 But this does not seriously detract from the percep-
tual realism of deep-focus style. Deep focus, particularly when
used in conjunction with wide screen, enables us to concen-
trate our attention on one object, then to shift our attention at
will to another object, just as we are able to do when perceiv-
ing the real world. Since we are usually not very conscious of
refocusing our eyes, the similarities between viewing deep-
focus style and perceiving the real world are more striking
than the differences. With montage and narrow-focus styles,
on the other hand, we are severely limited by shot length and
depth of field in our capacity to shift our attention from one
object to another at will- though as I have said, this feature is
not entirely absent in montage style.
To summarize: film is, in one sense, a realistic medium. It is
capable of representing depictively, and thereby enables us to
use our visual capacities to identify objects in order to know
that those very objects are represented onscreen. Further, it is
capable of representing spatiotemporal relations between objects
in the same way; we use our capacities to identify spatiotem-
poral relations between things in order to know that those re-
lations hold between the things represented onscreen. And
long-take, deep-focus style is more realistic than some other
styles in that it allows for more of this kind of representation of
spatiotemporal relations.
3.8 THE RELATIVITY OF LIKENESS
Explicating the idea of film realism in terms of my generalized

notion of depiction helps us avoid an error that has dogged
theorizing about the cinema: that realism in film can be at-
tacked on metaphysical grounds because it postulates an ob-
server-independent world - an idea which is then further
associated by some theorists with a politically conservative
agenda of submission to prevailing conditions. But realism as
I have explicated it here appeals to no such postulate (though
., See, e.g., Patrick Ogle, "Technological and Aesthetic Influences on the De-
velopment of Deep-Focus Cinematography in the United States".
108
Realism
one might argue that such a postulate is both philosophically

respectable and politically neutral). The claim of Perceptual
Realism is not that cinema presents objects and events isomor-
phic to those that exist in an observer-independent world, but
rather that, in crucial respects, film watching is similar to or-
dinary perceptual experience of the world, irrespective of
whether and to what extent that world is independent of our
experience of it.
When I say "film watching is similar to ordinary perceptual
experience of the world", I mean that film watching is similar
to our ordinary perceptual experience of the world. There might
be creatures as intelligent and perceptually discriminating as we
are, but who experience the world differently. They might not
be able to deploy their natural recognitional capacities in order
to grasp what is depicted in film and our other pictorial forms
of representation. Recall that bats might have visual experiences
qualitatively similar to our own, but caused by their very dif-
ferent perceptual systems, which depend on bouncing sound
waves off of solid objects. They wouldn't have any success de-
tecting the spatial properties of objects as they are represented
on a flat screen, and film would be a medium with little appeal
for them. So there is a definite relativity about my conception
of realism; what is realistic for us might not be realistic for other
creatures. My concept of realism is what I have called a re-
sponse-dependent concept; it is applicable to things by virtue of
the responses to it of a certain class of intelligent agents, namely
ourselves.""
Some people will find this relativistic concept of realism jar-
ring, perhaps oxymoronic. Among them are those who object to
realism because realism presupposes- so they think - some sort
of absolutist conception of the world and all its aspects. They
think that realism postulates a world describable without ref-
erence to any subjective point of view. But that is a mistake.
Colours are real, relational properties of things- properties they
have in virtue of our responses to them. For the record, let us
give a tolerably precise characterization of Perceptual Realism:
.., See this volume, Chapter 1, Section 1.5.
109
A representation R is perceptually realistic in its representa-
tion of feature F for creatures of kind C iff
(i) R represents something as having F;
(ii) Cs have a certain perceptual capacity P to recognize in-
stances ofF;
(iii) Cs recognize that R represents something as having F by
deploying capacity P.
So what is realistic for us need not be realistic for intelligent bats
or for Martians with their strange sensory faculties. Perhaps the
Martians would not, even after a period of familiarization, be
able to understand our cinematic narratives. But this admission
of relativity does not support the claims of Christian Metz and
others that there is a deep conventionality and cultural specific-
ity in our cinematic depictions. 25 We have learned to be sceptical
of those traveller's tales according to which humans from other
cultures find our styles of pictorial representation alien and un-
interpretable.26 Indeed, the existence of substantial cultural bar-
riers to the understanding of depictions would be surprising in
the light of evidence which shows that creatures belonging to
other species are capable of understanding the depictive con-
tents of photographs.27 And the undoubted fact that cinematic
25
"What is called the analogy, the 'resemblance' ... really lies within a whole
group of highly elaborate mental and social organizations ... and the appre-
hension of a resemblance implies an entire construction whose modalities
vary notably down through history, or from one society to another. In this
sense the analogy is, itself, codified" (Metz, "Cinematographic Language",
P· 584).
26
As Wollheim points out, people from other cultures would not be able to
make the mistakes they are said to make about perspective in the Western-
style pictures with which they are presented if they could not recognize
representational features of those pictures (Wollheim, Painting as an Art, p.
53). Work by Hudson which purports to show that black southern Africans
do not readily understand the depictive content of Western perspective has
been severely criticized; seeR. K. Jones and M. A. Hagen, "A Perspective on
Cross-Cultural Picture Perception".
27
In an experiment by Premack and Woodruff, a chimp, Sarah, was shown a
film of an actor trying to reach some bananas suspended above him; she was
then asked to choose from various photographs depicting further possible
moves in the attempt. Sarah chose the picture that showed the actor piling
110
Realism
representations are different in various ways from what they

represent - by being, for example, flat while the objects they
represent are solid- does not itself make those representations
conventional.:>S But convention is the subject of the next chapter,
and I shall leave detailed discussion of it until then.
There is another sense in which what I have called Perceptual
Realism can be interpreted relativistically. I have suggested that
representations are perceptually realistic when they share per-
ceptually significant properties with the things they represent.
Perhaps that is too reminiscent of the view which metaphysical
antirealists reject: that the world and the things in it have de-
terminate features independent of our conceptualization, and
that we depict the world right when we give our representations
those same features. I happen to think that this view is not so
very far from the truth, but its truth or falsity is irrelevant to
the issue of Perceptual Realism, just as it was irrelevant to the
issue of Illusionism. We need not state Perceptual Realism in
terms of the sharing of properties. We can instead state it in
terms of the appearance, for us, that things and their represen-
tations share properties. Perhaps, as some scientifically minded
philosophers think, things really have no colours, and all our
colour attributions are false. 29 Perhaps, due to some sceptical
scenario, there isn't a real, mind-independent world out there at
all. Perhaps creatures from radically different cultures would
crates underneath the bananas, thereby displaying her recognition that this
was the correct solution. The central claim of this paper - that chimps have
a theory of mind - must be treated with some scepticism, but the evidence
on which the authors built their theorizing - the ability of chimps to sort
photographs according to whether they represented a solution to the problem
in hand - is not disputed. See D. Premack and G. Woodruff, "Does the
Chimpanzee Have a Theory of Mind?"
211
"The very necessity of depicting a bulky three-dimensional object as a flat,
two-dimensional image testifies to a certain conventionality" (Juri Lotman,
Semiotics of the Cinema, p. 6). A version of this argument can be found in
almost every discussion of film semiotics. See, e.g., Robert Lapsley and Mi-
chael Westlake, Film Theory: An Introduction, p. 45· For more detail on the
notion of a convention, see this volume, Chapter 4·
29
See, e.g., Boghossian and Velleman, "Colour as a Secondary Quality". This
view goes back to Galileo.
111
conceptualize the world in ways that are inconsistent or incom-
mensurable with our own, and there is no rational way of choos-
ing between these rival conceptualizations. But none of these
possibilities- if that is what they are- is in any way inconsistent
with Perceptual Realism. That doctrine need claim only that de-
pictions in general and cinematic images in particular appear to
us as in significant ways the same as the kinds of things and
events they represent, and that it is their so appearing that en-
ables us to identify their representational contents.
112
Chapter 4
Languages of art and languages
of film
... the proper objects of sight are lights and colours,
with their several shades and degrees; all which, be-
ing infinitely diversified and combined, form a lan-
guage wonderfully adapted to suggest and exhibit to
us the distances, figures, situations, dimension, and
various qualities of tangible objects: not by similitude,
nor yet by inference of necessary connection, but by
the arbitrary imposition of providence, just as words
suggest the things signified by them.
George Berkeley
That the medium of cinematic images is in some sense pictorial is

hardly controversial. But many will object to the sharp distinction
I have drawn between the pictorial and linguistic, a distinction
Berkeley apparently seeks to undermine in the quotation above.
That distinction is rejected by almost every consciously theoreti-
cal writer on film, and on art and culture generally. The prevail-
ing view is that picturing is a mode of representation at least very
like linguistic representation, that images in general and cine-
matic images in particular operate by means of codes or conven-
tions that are like the semantic and syntactic rules of a language.
Berkeley's view that the appearances of things form a kind of ar-
bitrary code, and that seeing is therefore a kind of reading, is one
which, shorn of its theological trappings, many contemporary
theorists of cinema would endorse.
This idea of the ubiquity of language is so entrenched that
expressions like "the grammar of quattrocento painting", "the
vocabulary of Gothic architecture" and "the language of film''
do not strike us as peculiar or in need of justification.' Some of
' See also the chimerical "story grammar", aptly criticized in R. Wilensky,
"Story Grammars Revisited".
113
these uses may be unexceptionable shorthand: talk of the vo-
cabulary of an architectural style sometimes means just what-
ever devices are typically employed in that style. But sometimes
the usage suggests possession of a linguistic theory that can tum
casual connoisseurship into a powerful technique of analysis.
The suggestion is spurious. Art, architecture, film and the rest
have little in common with any of the uncontroversial examples
of language that have shaped our recent linguistic theorizing. It
is not likely, therefore, that linguistics will help us explain how
it is we use, interpret and appreciate any of these things. So I
claim, and so I shall argue with respect to the case of film.
This much is negative. But there is something important to
learn about film and the comprehension of it by comparing its
communicative aspect with that of language. That will prepare
the way for a theory of the interpretation of film developed in
Part III. So while this chapter rounds off the discussion of cin-
ematic representation with the thought that it is not linguistic,
quasi-linguistic or even remotely linguistic, much of its sig-
nificance will not be apparent until we compare the interpreta-
tion of texts and that of film in Chapter 8.
4.1 FINDING THE THESIS
The hypothesis that there is a language of film is not the true

but uninteresting claim that the language of Citizen Kane is Eng-
lish and that of Rashomon is Japanese. It is the hypothesis that
there is a specifically cinematic language which can and
sometimes does operate independently of accompanying words
or sounds- that there is a language of cinematic images, their
modifications and juxtapositions.2
The idea that there is such a language, along with comparable
claims for painting, architecture and the rest, is part of the legacy
of structuralism. But while structuralism has led us up one blind
• Sometimes advocates of cinema language seem to be asserting that there are
many different such languages. The arguments I shall bring forward will be
just as effective against that hypothesis as they are against the hypothesis that
there is one such language. So I shall not bother to consider the multiple-
languages hypothesis separately.
Languages of art and languages of film
alley after another, the idea that structure is the key to under-
standing ourselves and our culture has proved surprisingly re-
silient; the grudging recognition one hears from time to time
that cinema is, after all, really rather different from anything we
would normally call a language goes along with an enduring
belief that, at some level, an underlying similarity will be found.
Where did the structuralists go wrong? Enthusiastic for uni-
versal principles in the human sciences, they fastened onto the
success of theoretical linguistics, hoping to explain other human
activities on the same basis. Thus the structuralist Tzvetan To-
dorov: "If we admit the existence of a universal grammar, we
must no longer limit it to language alone. It will have, evidently,
a psychological reality.... This psychological reality makes
plausible the existence of the same structure elsewhere than in
speech". 3 The structuralist's strategy was to explain our com-
petence in nonlinguistic - or at least not obviously linguistic -
domains by appeal to those mental structures that underlie our
linguistic competence, and thereby to show that those other do-
mains are, after all, basically linguistic in structure.4 That strat-
egy derives from a view of the mind as a unified or at least
highly interconnected mechanism in which basically the same
competencies apply across domains. I have already rejected that
view of the mind. I have argued that the mind is a complex, less
than fully integrated institution with relatively autonomous de-
partments dedicated to specific functions. Currently available
evidence suggests that language is a particularly good example
of this "vertical" structuring of the mind. A child's acquisition
of linguistic skills seems to be remarkably insensitive to varia-
tions in her other competencies, including perceptual ones. For
example, the use of personal pronouns emerges in congenitally
3
Tzvetan Todorov, The Poetics of Prose, pp. 1o8-1C)9.
• A kind of mirror image to this approach, deriving from the same universal-
izing conception of the mind, is Piaget's theory that linguistic ability is de-
rived from other more primitive and general human competencies, in
particular from "sensorimotor" abilities. For a lively - if partisan - account
of the collapse of Piaget's program under pressure from Chomsky's linguis-
tics, see M. Piattelli-Palmarini, "Ever Since Language and Learning: After-
thoughts on the Piaget--Chomsky Debate".
115
deaf children who employ sign language at just the same time
as in hearing children, with both groups going through a period
of pronoun reversal, confusing "you'' and "I" .5 There just is no
reason - apart from the initial plausibility and simplicity of the
idea- to suppose that mental structures designed to accomplish
one kind of task will be similar in any interesting way to other
such structures, or that a structure used for one purpose will
tum out to be used for others.6
If the view that there is a language of picturing is to be re-
jected, perhaps "there is another hypothesis to be considered: that
film is not to be analysed as a language, but as an example of
the broader category of semiotic systems. In that case cinema
falls within the scope of a more general theory that includes both
natural languages and a great deal else: everything, in fact, that
is a system of signs. If the hypothesis that cinema is a language
is to be rejected, perhaps we should consider the weaker posi-
tion of the semioticists - weaker because the hypothesis that
cinema is a language entails but is not entailed by the hypothesis
that it is a semiotic system. But I shall not confront the semiotic
hypothesis in any detail. One reason is that the generality aimed
for in semiotics has resulted in a great deal of taxonomizing but
little identifiable as theory. It has also never been made clear
what a "code" is supposed to be, though codes are what semi-
oticists are apparently most interested in finding. Nor is it easy
to read off a characterization from the examples offered. Among
cinematic codes there are, we are told, "the complex system
according to which the cinematic equipment (recording camera,
film strip, projector) 'reproduces movement' '',7 and "the rep-
resentational code of linear perspective" .8 It is hard to think of
a definition of code which would cover these two items and not
cover everything else. So a critical assessment of the semiotic
hypothesis would require us to figure out what the hypothesis
is, exactly. Or, if there is no clear hypothesis available, we shall
5
See L. A. Petitto, "On the Autonomy of Language and Gesture: Evidence from
the Acquisition of Pronouns in American Sign Language".
6
For more on Structuralism see Chapter 9, Section 9+
7
Christian Metz, Language and Cinema, p. 191.
8
Comolli, "Machines of the Visible", p. 135·
116
have to devise a plausible hypothesis as a target for criticism.
That would be a lengthy undertaking, and I should like to avoid
it.
It is not just the vagueness of semiotics that makes me un-
willing to tackle it head-on. At least some of what I say in re-
lation to the hypothesis that there is a language of cinema will
count against semiotics as well. Semioticists seem committed to
the "conventionality" of the sign systems they investigate; this
assumption makes them hopeful they will discover a unified
theory of signs. 9 And in the course of arguing against the idea
of a language of film I argue that cinematic images are not con-
ventional signs. I have no quarrel with the claim that cinematic
images are signs, if that means simply that they are represen-
tations. But the nonconventionality of cinematic signs means
that talk of a system of cinematic signs is vacuous.
4.2 CINEMA LANGUAGE AND NATURAL

LANGUAGE
The language of cinema, if there is one, is in various ways star-

tlingly different from any natural language. It is what we might
call medium specific: it has its existence in cinematic images and
their modifications, so it is conveyed to us through sight alone. 10
No natural language is medium specific in this way; natural

languages can be spoken and they can be written, even if in fact
some of them are not, so they are available to sight and to hear-
ing. Braille makes a natural language available to the sense of
touch, and a coding of English into olfactory sensations is pos-
sible, though it would be unwieldy in practice. Natural lan-
guages are available to us through all our senses.
Cinema language is not just medium specific: you could not
even "translate" cinematic images into distinct visual images by,
say, stretching, distortion or colour modulation and still have
• "Signs are correlated with what they stand for on the basis of a rule or a
convention" (Umberto Eco, "On the Contribution of Film to Semiotics",
p. 196).
0
' On the scrcalled medium-transferability of natural languages see John Lyons,
Language and Linguistics, p. 11.
117
what advocates of cinema language would call "the same lan-
guage", for those transformations would impose a change of
meaning; any change in the appearance of what is on the screen
is potentially significant for our decisions about the content of
the story that is thereby presented. But with natural languages
there is nothing special or privileged about any visually specific
way of representing letters and words, as font and handwriting
differences attest. And there are codes in which you can write
English sentences using permutations of our alphabet, other al-
phabets, numerals or whatever.
Perhaps cinema language is so fundamentally different from
natural languages that no significant properties of the one are
attributable to the other. But if cinema language is so unlike
natural language, what is to be gained from treating it as a lan-
guage at all? True, we cannot conclude immediately from the
fact that there are differences between natural languages and
the supposed language of cinema that there is, after all, no such
thing as a language of cinema. There are differences, some of
them very striking, between the various natural languages. Ide-
ally, we would distinguish accidental from essential differences,
and if we found that cinema "language" differs essentially from
natural language we would conclude that the former is not, after
all, a language. But I do not know of any set of characteristics
generally agreed to be essential for a language. We could try
providing some, in the form of a definition, but that wouldn't
settle anything. We don't decide questions about whether this
or that is a cause by appealing to some definition of "cause",
for we are at the stage with the concept of cause where we are
testing definitions against cases, and not the other way around.
So it is with language.
I suggest another method. Whatever disputes there are about
whether this or that is a language, no one, so far as I know, has
denied that English is a language, or that any of the other things
we call "natural languages" are languages. Also, natural lan-
guages like English have been the objects of the most sustained
and systematic theoretical investigation into language. Linguis-
tics, in so far as it is a developed theory, is about these natural
languages. The hope of those who say there is a language of
118
cinema is that they will be able to make use of a well-developed
theory - linguistics - in an area which is theoretically under-
developed." In that case the cash value of the claim that there
is a language of cinema depends directly on the similarity be-
tween film and natural language. The less like natural language
the cinema is, the less probable it is that linguistics will help us
understand it, just as the difference between atoms and social
customs makes it unlikely that physics will help us understand
society.
I shall argue that, in crucial respects, film is very unlike a
natural language. This will enable us to sidestep the definitional
question. If someone agrees with my conclusions and still wants
to say there is a language of film, I shall not argue with her
choice of words. All that matters to me will have been conceded:
that there is insufficient similarity between this "language" and
any natural language for us to expect progress to be made in
understanding the cinema by applying to it linguistics or any
theory that adopts the principles and techniques of linguistics.
The question I am raising here is about the transferability of a
theory, not about the appropriateness of a word.
While I shall be sceptical throughout this chapter about the
relevance of linguistics to film theory, I don't want to imply any
comparable hostility to philosophy of language as a useful tool
in this area. Much of what we class as philosophy of language
is concerned with arguments and conclusions that are not, after
all, specific to language, but which apply to linguistic and non-
linguistic forms of communication alike. And while cinema is
not a linguistic medium, it is essentially a mode of communi-
cation between a story-teller and an audience; more on this in
Part III. Philosophy of language is not exhausted by any single
theory, and within it there is fundamental disagreement about
the nature of language and about the possibility of nonlinguistic
thought and communication. My application of philosophy of
" Metz seems to have thought of it that way: "The methods of linguis-
tics ... provide ... a constant and precious aid in establishing units that,
though they are still very approximate, are liable over time ... to become
progressively refined" (Christian Metz, "Some Points in the Semiotics of Cin-
ema", p. 176).
119
language to film will be the application of a particular contest-
able theory. It will be important to bear this in mind in assessing
my conclusions.
One important aspect of the argument to follow is that lan-
guage and meaning are by no means coextensive; there can be
meaning which is not linguistic meaning. The fault of those who
find language everywhere is that they infer the presence of lan-
guage from the presence of meaning.
4·3 THE SHAPE OF NATURAL

LANGUAGE
I shall specify some salient features of natural language. My
example will be English. The concept of cinema language has
been able to thrive partly because it is never discussed with any
precision, and I shall try to be as precise as I can be in describing
what I take to be theoretically important aspects of any natural
language. This will have at least the virtue that those who want
to defend the applicability of linguistics to film theory will have
to specify exactly where and why my description is wrong, or
where I go wrong in claiming there is nothing in cinema that
corresponds to the description. The account will be somewhat
compressed, but the concepts appealed to should be familiar,
and compression will help to give an overview; it is the logical
relations between these concepts I want to emphasize. Nothing
turns on the choice of example: trivial reformulations would fit
the description to any natural language.
English displays the features of productivity and conventional-
ity. Productivity means that an unlimited number of sentences
of English can be uttered and comprehended; in fact many of
the sentences we utter and comprehend every day have never
been uttered, and so have never been comprehended, before.
Whatever learning English involves, it does not involve learning
meanings sentence by sentence; otherwise we would need in-
struction every time we heard a new sentence.' 2
" What I am here calling "productivity" is sometimes called "generativity",
which, in the present context, might be confused with the notion of natural
generativity to which I appeal in this and other chapters.
120
English is conventional in that what words and sentences of

English mean is determined, not by relations of natural affinity
between words and meanings, or because the human mind is
specially apt to associate certain words with certain meanings,
but by adventitious uniformities of practice adhered to because
they aid communication. The differences between the various
natural languages humans use (and these differences are small
by comparison with those between all the logically possible lan-
guages) are mostly accounted for as differences between these
uniformities of practice. These uniformities of practice are con-
ventions. A convention, as David Lewis has shown, need not
have its origin in agreement. There is a convention to the effect
that a word has a certain meaning when there is a regularity of
use among members of the speech community; they use that
word intending to mean something, and they do so because they
know others do the same, and desire to continue the regularity
because by doing so they are able to achieve a coordination be-
tween what they mean by it and what hearers will take them to
mean, a coordination necessary for successful communication.
It does not much matter what word we use to express a given
meaning; what matters is that most of us use it most of the time
to express the same meaning. That way, we have some idea
about what meaning others use it to express, and they have
some idea about our use. And that way we are able to coordi-
nate our communicative activities.' 3
Conventionality and productivity combine to set certain fur-
ther requirements on the shape of language: Conventionality
means that language has to be learned; if the meanings of words
were, say, innately known, and speakers of a common language
drew on that knowledge in making and understanding utter-
ances, there would be no conventions necessary to connect
words with their meanings. Productivity precludes the language
'
3
See David K. Lewis, Convention. I have presented Lewis's account of conven-
tion in language as if it were of the conventionality of words. In fact there
are problems in extending his account from whole sentences to sentential
components. But these problems are, I take it, problems of detail rather than
of principle. The best work I know on this is Timothy Irwin's unpublished
M.A. thesis, Meaning and Convention.
121
being learned sentence by sentence, since in a language which
displays productivity, speakers understand sentences to which
they have had no previous exposure. So meanings in our lan-
guage, if they are to be learned at all, must be specified recur-
sively: we start with a set of conventions that assign meanings
to a finite stock of words, and we combine words into further
meaningful units (e.g., sentences) by rules of composition, which
tell us how the meaning of the whole depends on the meanings
of the parts.' 4 Thus our language is molecular: its sentences are
built out of independently meaningful units - what I shall call
meaning atoms- by rules that assign meanings to complexes as
a function of the meanings of the basic parts. Words are our
atoms; they are meaningful, and they contribute to the meanings
of larger units to which they belong, but they themselves have
no meaningful parts.'5
Since the atoms- words in English- are assigned meanings
individually, and since the composition rules make the meaning
of the whole a function of the meanings of the parts, meaning
in our language is acontextual. The meaning of a given word is
determined by its meaning convention- not by the meanings
of other words, and not by anything else- and the meanings of
sentences depend only on the meanings of the words in them.' 6
In sum, our language is productive and conventional, so its
meaning-determining conventions are recursive, so it has mean-
ing atoms, so it is molecular, so it is acontextual. A great deal
in the argument that follows will depend on these entailments.
As I have already stressed, none of this is definitive of Ian-
4
' I don't say we learn our language entirely from the bottom up. No doubt
we start with simple sentences, shake out the word meanings by decompo-
sition, and zigzag back and forth, continually expanding our competence in
a way governed partly by trial and error. What is claimed is simply that if
the meanings of linguistic units could not, in principle, be stated recursively
the language would not be learnable.
'
5
Not every word is an atom, as when a single word has a meaning that is a
function of the meaning of other words and operators, as with "invalid".
6
' This account of meaning in natural language contradicts the structuralist's
claim that linguistic meaning is wholly a matter of a sign's relations to other
signs. That is as it should be; a change in the meaning of one term does not
induce a change in the meanings of all terms.
122
guage; there might be languages that are either not productive,
or not conventional, or neither.' 7 But if there is, for cinema, any-
thing that plays a language-like role, it would surely be pro-
ductive; there is an unlimited number of things that can be
conveyed by cinema images. New films do not simply recycle a
fixed stock of images; they present us with new images which,
by and large, we have no special problem understanding.' 8 And
the people who claim that there is a language of film emphasize
the contingency and conventionality of that language, the degree
of its social and cultural determination. (I shall suggest that
there is some confusion on their parts about what this conven-
tionality amounts to.) So advocates of cinema language will hold
to the idea that this language, or any one of the possibly many
cinematic languages, is both productive and conventional. But
that, as I have argued, means that a language of cinema must
also be recursive, molecular and acontextual.
I shall argue that there is nothing in cinema that satisfies all
these requirements, or anything vaguely like them. One objec-
tion no doubt forming already in the minds of cinema-language
advocates is that my account of meaning in natural language is
too narrowly mechanistic to be plausible for the natural Ian-
7
' Perhaps the language of thought, if there is one, is productive and noncon-
ventional. Perhaps Wittgenstein's "block", "pillar'', "slab", "beam" language
described at the beginning of the Investigations is conventional and nonpro-
ductive. Productivity seems to be more important than conventionality for
characterizing what is special about human languages. There is some evi-
dence that chimpanzees can employ rudimentary systems of arbitrary sym-
bols (see, e.g., D. Premack, "Minds with and without Language"; and S.
Savage-Rumbaugh et al., "Spontaneous Symbol Acquisition and Communi-
cative Use by Pigmy Chimpanzees Pan paniscus"), but their use shows little
sign of productivity (seeM. C. Corballis, "On the Evolution of Language and
Generativity"). For an overview, see David Premack, Gavagai! Or the Future
History of the Animal lAnguage Controversy.
8
' If you doubt this, carry out an experiment recommended by Irving Bieder-
man: "Turn on your TV with the sound off. Now change channels with your
eyes closed. At each new channel, blink quickly. As the picture appears, you
will typically experience little effort and delay ... in interpreting the image,
even though it is one you did not expect and even though you have not
previously seen its precise form." "Higher-Level Vision"; page references are
to the partial reprint, here and in note 29, as "Visual Object Recognition".
123
guage case, and so it cannot be the basis for any significant
attack on the idea of cinema language. So I had better try to
dispel at least some of the more obvious objections to the ac-
count of natural language just given.
4·4 OBJECTIONS REJECTED

It would be a mistake to suppose the account just given is in-
consistent with the notable vagueness of linguistic meaning. The
conventions that specify the meanings of many, perhaps all, of
the words of a natural language fail to determine the extensions
of those words precisely; there are times when we know all the
relevant empirical facts, as well as everything relevant about the
meanings of words, yet still cannot decide whether a certain
object is properly called a hill rather than a mountain. When
that is so, a word's meaning is vague, but that is no objection
to the recursiveness of meaning; it indicates just that the recur-
sive rules will import vagueness of meaning into any sentences
that contain vague terms. While vagueness raises problems
which cannot be ignored, it should not be used as an excuse for
abandoning the project of giving a theory, perhaps a rather pre-
cise theory, about meaning, and about the transmission of mean-
ing from words to sentences.
Further, the account I have given is not - and this point can
easily be confused with the last - committed to what I shall call
the determinacy of meaning. Is it literally correct to speak of the
legs of chairs? Perhaps, but there was a time when this was not
literally correct, and a later time when this usage could not prop-
erly be called either literal or figurative. The literal meaning of
"literal meaning" is as vague as that of "hill". But just as vague-
ness is no objection to the claim that there are hills, so it is no
objection to the claim that there are literal meanings. Conse-
quently our theory is not committed to the constancy of literal
meaning. Words have literal meanings at one time that they do
not have at others. But this, again, is no objection to the thesis
that words have literal meanings. Otherwise we should not be
able to say that objects have shapes and colours.
The claim that there is such a thing as the literal, acontextual
124
meaning of words is not the claim that the meanings of words
and sentences are infallibly known to speakers of the language.
We can be mistaken about what the literal meaning of an ex-
pression is, even when we understand the language to which
that expression belongs. Since there is no upper bound to the
potential length of sentences in a natural language like English,
there are presumably many sentences of that language which
are so long the human mind cannot process their meanings. And
length is not the only problem: "No head injury is too trivial to
be ignored" is a celebrated case of a sentence the literal meaning
of which almost everyone misunderstands - until they do some
hard work on figuring out what it means.
Nor is it an objection to the theory of language just outlined
that many words, for example "bank", have more than one
meaning; the sensible response to that is to say that some words
have more than one meaning convention attached to them, and
the rules of composition can be applied when one particular
meaning is chosen.' 9 As long as the rules can be seen to apply
whatever meaning is singled out, ambiguity is no threat to the
theory. There would be a threat to the theory if there were
words that had infinitely many distinct meanings, for then we
could give no recursive specification of meaning for a language
containing such a word. But there is no reason to suppose that
this is the case.
Another objection to the idea of context-free literal meaning
is that what is conveyed by a sentence varies from one context
of utterance to another. That is true, but it shows merely that
what is conveyed by a use of a sentence is not always the same
as the literal meaning of the sentence. What is conveyed by a
use of a sentence (if we confine ourselves, for the sake of sim-
plicity, to indicatives) is what the speaker asserts by uttering it.
And a speaker can use a sentence which means one thing to
assert more or other than what the sentence means. An utterance
of "Harold is a snake", performed while witnessing some par-
'
9
Alternatively we can say that in such cases there are two distinct but iden-
tically spelled words, each with its own (unique) meaning convention at-
tached.
125
ticularly discreditable action of Harold's and as part of a con-
versation about Harold's character, might convey the thought
that Harold is given to scheming self-aggrandizement. But the
sentence "Harold is a snake" does not mean that.
But in that case, why postulate literal meanings? Why not be
content with a single category of meaning: the context-
dependent meaning of the utterance? Some literary theorists
have recently taken that option. But it is not a viable one. It is
20
true that what we understand by an utterance is not a function

solely of the meanings of the words uttered, since the same
words uttered in different contexts can be used to communicate
different things. But to conclude from this that literal meaning
plays no role in explaining understanding would be like con-
cluding that colour plays no role in our response to a picture
because the same colour may look garish in this context and
comfortably warm in that one. Take away the colour and you
take away the response; take away literal meaning, and you take
away understanding. Context matters for understanding, but it
cannot be all that matters - otherwise I should be able to un-
derstand the native utterances of my Japanese friends just by
understanding the contexts in which those utterances are made.
I don't understand them, even when I know the context, because
I don't speak Japanese; I lack knowledge of the literal meanings
of Japanese sentences. You might insist that with the utterances
in Japanese there are aspects of context I fail to grasp, but that
would be just to repackage literal meaning under a new label:
"context of utterance''.
4·5 INTERPRETATION AND UTTERANCE

MEANING
So there is a difference between the literal meaning of a sentence

and the meaning of an utterance of that sentence. The meaning
of a particular utterance depends, of course, in part on the mean-
ing of the sentence uttered, which depends on convention; but
20
See Stanley Fish, Is There a Text in This Class? and Doing What Comes Naturally.
See also my "Text without Context: Some Errors of Stanley Fish".
126
it would be a mistake to say that utterance meaning is itself
conventional. It cannot be supposed that we work out the mean-
ing of an utterance by appealing to conventions of the form, "An
utterance of sentence S in context C means M." There is an un-
limited number of values for C, and contexts are not constructed
recursively from a finite set of context constituents. But knowl-
edge of utterance meaning is productive; we understand utter-
ances in contexts we have not previously encountered. How so?
When we figure out utterance meaning we apply the conven-
tions of literal, linguistic meaning, together with nonconven-
tional rules of rationality; we assume the speaker understands
the language and the relevant features of the speech context, and
is able to act appropriately - to choose appropriate words - so
as to get us to realize what it is he intends to get across. The
best hypothesis about that intention gives us the utterance
meaning. This assumption of rationality, and the various specific
subrules to which it gives rise/' are not conventions, because
they are not assumptions to which there is any realistic alter-
native; they are assumptions the world imposes on us rather
than ones we impose on the world. We cannot choose to regard
a speaker as rational and so interpret his utterance one way, or
choose to regard him as irrational and interpret it in another. If
we do not assume him at least minimally rational we cannot
come up with an interpretation - or rather we have no way of
22
deciding among countlessly many of them.
This account of how language functions suggests that the ad-
vocates of cinema language were too optimistic in assessing
what could be achieved for interpretation by way of convention
alone. Literary interpretation is largely a matter of figuring out
utterance meaning; what story the text has to tell is not just a
matter of the literal meanings of the words on the page, which
often have to be taken nonliterally, and even then can sometimes
be thought of only as hinting at the events of the story. Figuring
out the events of the narrative is a matter of figuring out, on the
" On which see Paul Grice, "Logic and Conversation" .
., This is not Davidsonian charity, which would have us regard the speaker's
utterances as (largely) true - clearly a pointless injunction in the case of a
fictional utterance.
127
basis of your understanding of the language and of the principle

of rationality, what story would most likely have been intended
by someone who had chosen to write those words with those
meanings: so I shall argue in Chapter 8. Even if there is a lan-
guage of cinema, and it functions in the presentation of
cinematic narrative in something like the way language func-
tions in literary forms, film interpretation is still going to be
largely a matter of calculating intentions; understanding the lan-
guage of a text or (if there is one) of a film is just the first step
in getting to an interpretation. Cinema theorists often write as
if cinema language is the key to understanding cinematic nar-
rative. If the analogy with natural language held, discovering
that language would be more like being told which shop the
key was in.
So if there is a language of film, it won't provide an answer
to the problem of interpretation. It might, on the other hand, be
a precursor to an answer, and therefore worth having. It's time
to see that there is no such thing to be had.
4.6 CONFUSIONS ABOUT CONVENTION
I have described some crucial features of the way meaning is

articulated and communicated in natural language. We now
have to decide how far the functioning of cinematic images is
like that of a natural language in these respects. We can say,
first of all, that the function of cinematic images in the fiction
film is to tell a story, just as the function of the words and sen-
tences in a fictional text is to tell a story. So the question we
need to ask is this: Do these images perform their story-telling
function in the way that sentences of the literary text do?
Some who are sceptical of the idea of cinema language have
argued that they do not, on the grounds that cinematic images
function contextually, and therefore in a way different from
words and sentences which function, as we have seen, acontex-
tually. Thus George Wilson argues that the meaning we give a
particular sequence of shots depends on how coherently that
sequence, so interpreted, fits into the rest of the film. Whether,
in The Lady from Shanghai, the juxtaposition of a hand pressing
128
a button and a car crashing is to be taken as signifying a causal

relation between the two depends on whether there is elsewhere
in the film evidence for this peculiar causality. This suggests,
says Wilson, "the holistic character of all interpretive work" .23
Wilson is right that what the cinematic image tells us about
the fiction depends on the surrounding context of other images.
But that is true also of words and sentences in a text, where
there is no dispute about the presence of a language. What kind
of relation between described events is suggested by one bit of
text depends upon the role that bit of text is seen to have in the
context of all the other bits. If the text says, "Her hand pressed
the button just before the car crashed'', it is then a matter of
interpretation, which will have to take account of the rest of the
text, as to whether it is part of the story that the pressing caused
the crash. Even if the text were more explicit, and said, "Her
pressing of the button caused the crash", we would have to
decide, on the basis of our overall assessment of the rest of the
text, whether this should be taken as a reliable guide to the
story, rather than, say, the suspect utterance of an unreliable
narrator. So the context dependence of interpretation applies to
literature as much as to film, and it cannot on its own be an
argument against there being a language of the cinema.
How is the context sensitivity we find in literary interpreta-
tion compatible with the fact that linguistic meaning is acontex-
tual? What words and sentences mean is determined by the
conventions of the language, and these conventions provide, as
we have seen, for the acontextuality of meaning. What a speaker
asserts by uttering words and sentences with those meanings
depends on more than meaning-determining conventions. It de-
pends, exactly, on context, together with assumptions about ra-
tionality. The context sensitivity of interpretation is a further
indication that the target of interpretation is the speaker's mean-
ing rather than the literal meanings of words and sentences.
So perhaps the language of film enthusiast can meet Wilson's
challenge to the idea of cinema language by drawing a distinc-
tion between the literal meanings of cinematic images and the
23
Wilson, Narration in Light, p. 203.
129
meaning we attribute to the act of "uttering" - that is, producing

and displaying- them. He would have to show that the literal
meaning of these images preserves a language-like acontextual-
ity, while appealing to their utterance meaning to explain con-
textual dependence. If he could show that this literal meaning,
whatever it is, is in theoretically important ways analogous to
literal meaning as it appears in natural language, his task would
be complete. I shall argue that this last and most important step
is one he cannot take.
The argument hinges crucially on the following claim: it is
not possible to identify any set of conventions that function to
confer meaning on cinematic images in anything like the way
in which conventions confer (literal) meaning on language. This
is the fundamental disanalogy between language and pictorial
modes of representation. And the argument for this claim is that
the meaning of a cinematic image is nonatomic and so nome-
cursive. There are no atoms of meaning for cinematic images;
every temporal and spatial part of the image is meaningful
down to the limits of visual discriminability. When a spatial part
of the image has distinguishable spatial parts within it, the
viewer who can identify what is represented by the larger spa-
tial parts will be able to identify what is represented by the parts
within it. To see a part of the image as representing a face is
necessarily to be able to recognize parts within that image part
which represent facial features, however indistinctly. And the
principle holds for image parts that have no internal discrimin-
able structure. A part of the image of the tiger represents one of
his stripes, and represents it as uniformly black, in virtue of that
image part itself being uniformly black. Still, once it is acknowl-
edged that this uniformly black image part represents a stripe,
it will be understood that any subpart of that image part rep-
resents a subpart of the stripe. 24
Despite being nonatomic, film images are productive: there is
an unlimited number of visible scenes that cinematic images can
represent, and we generally have no trouble grasping the repre-
""' Virtually the same point is made by Roger Scruton in The Aesthetic Under-
standing, p. 107.
130
sentational properties of cinematic images we have never seen
before. Being without meaning atoms, and therefore nonrecur-
sive, and being at the same time productive, the meaning of cin-
ematic images cannot be conventional. It is natural meaning.
With images, productivity is natural generativity, and we explain
that in terms of natural recognitional capacities. As I argued in
Chapter 3, we understand what is depicted onscreen by employ-
ing our visual capacities to recognize the entities depicted. 25
To say that the meaning of a cinematic image is naturally
rather than conventionally determined is not to say that this
meaning has nothing to do with convention. How things are
placed before the camera will be influenced by social institutions
like styles of dress, composition and decor, as well as by con-
siderations of decorum; these things, or some of them, count as
conventions. But this is not grounds for saying the meaning of
the image is itself conventional in the sense that meaning in nat-
ural language is. To be conventional in the way language is there
would have to be a set of conventions governing the meanings
of all the image atoms, and since there are no image atoms there
are no such conventions.
This nice distinction - between what is conventional and
what is influenced by convention - is overlooked by those who
are content to appeal to a vague, impressionistic and all-purpose
notion of convention to support their claims about the conven-
tionality of images. Umberto Eco gives the example of an image
from Fritz Lang's M, in which the girl's balloon is caught in
overhead wires. This image, he says, "stands for" the capture
of the girl by the murderer, but it does so only by convention;
that is, it does so only because in our culture "wires recall ropes,
ropes capture, and so on". 26
The example isn't helpful to Eco's case, since there really are
no conventions associating wire with ropes and ropes with cap-
ture. Wires and ropes are alike in their function and appearance,
which is nothing to do with convention, and it is by no kind of
,, See Section 3.1 in the previous chapter; and Schier, Deeper into Pictures. See
also Sartwell, "Natural Generativity and Imitation", especially pp. 186-187.
26
Eco, "On the Contribution of Film to Semiotics", p. 207.
131
convention that ropes are useful for capturing things. A better
example would be a case where the image of a red traffic light
is a metaphor (not, I grant, a very imaginative or subtle one) for
the end of a love affair. A red traffic light is suggestive of end-
ings because there is a convention which associates the colour
of the light with the instruction to stop. But still, there is no
conventional meaning possessed by this image of the traffic light,
as it appears in our imaginary film. One thing the image can be
said to mean is what it depicts, which is a traffic light. It does
that in virtue of being a cinematic image of a traffic light, and
not by convention. The image might have meaning in virtue of
being a metaphor for, or suggesting, the end of the affair. That
isn't conventional meaning either. An image, like a word or a
sentence, can suggest or be a metaphor for any number of things
depending on context. No convention associates a real red traffic
light with endings of relationships, and no convention associates
the phrase "a red traffic light" with such endings. For that as-
sociation to be conventional, there would have to be a unifor-
mity of practice in the community to use the one as a sign of
the other, or an agreement to use the one for the other if appro-
priate circumstances arose. I'm not sure I have ever seen a traf-
fic-light image used in film in just the way I have described, but
if I had, I think I would have understood its metaphorical sig-
nificance without drawing on, or making an inference to, any
practice of using traffic lights that way, or any agreement to use
them that way. Metaphors don't work by convention; under-
standing them involves understanding intention, and we see the
traffic lights as suggestive of ending when we see that the image
of the traffic lights has been placed in the film just at this point
with the intention of suggesting an ending. 27 Metaphor has to
be explained in terms of utterance meaning, not in terms of lit-
eral (i.e., conventional) meaning.
In another place, Eco has argued for the conventionality of
images generally on the grounds that our images tend to be,
27
See the comments critical of Eco's conception of convention in Gilbert Har-
man, "Eco Location", in the second edition of Film Theory and Criticism. See
also his "Semiotics and the Cinema: Metz and Wollen", in the same volume.
Both are, unfortunately, omitted from subsequent editions.
132
and perhaps inevitably are, affected by our social practices: a
picture of a lion was once praised for its lifelikeness; we can
now see how much it owes to the conventions of heraldic rep-
resentation.28 Again, this argument simply conflates two kinds
of conventions, those that facilitate meaning and those that de-
termine it. From the fact that the characteristics of a sign are
affected by convention it doesn't follow that the sign is a con-
ventional sign. A sign is conventional only if there is a conven-
tion which determines its meaning.
An analogy may help. Colonel Smith asks Colonel Jones if
Colonel Jones will kindly have the barracks cleared up after last
night's party. Smith's request depends for its satisfaction on the
giving of certain orders by Jones. But Smith's request is not itself
an order; Smith is not in a position to order Jones to do anything.
In general, what depends on Xs for its fulfilment is not itself
thereby an X.
It may sound as if I am winning this argument by stipulation,
insisting that images are not conventional in a quite idiosyn-
cratic sense. Not so. Although I believe I am using the word
"convention'' in at least one of its ordinary senses, I agree that
there are other senses my usage does not cover, and that in some
of these senses, cinematic and other images may be said to be
conventional. You may use "convention" in any of these other
senses, or give to it any sense you like, but please use it only in
one way, and make sure you have in your vocabulary other
words, one for each distinguishable concept we need in this in-
quiry. One of those concepts is what I have called "convention-
ality'', and it applies to words in English because their meanings
are determined by a coordinated practice based on mutual ex-
pectation. That is the target concept, whatever you choose to
call it, for anyone who wants to argue that cinematic images
have much in common with the items studied by linguistics.
This target concept does not, and could not, apply to the relation
between cinematic images and their meanings, and the whole
argument just given could be restated without using the word
"convention'' at all.
28
Eco, A Theory of Semiotics, p. 205.
133
4·7 RELATIONS BETWEEN IMAGES
Perhaps the claim that there is a language of cinematic images

is not the claim that these images themselves have a language-
like structure, but rather that these images enter into language-
like combinations with each other. In films, there are
combinations of images; sometimes these combinations form
identifiable and recurrent patterns, and these combinations can
have a meaning which partly depends on the manner of their
combination. Isn't there something here like the articulated
structure of a sentence? No. As many film theorists have
recognized, the representational content of a cinematic image
cannot be equated with that of a name, predicate or other sub-
sentential part of speech. If these images line up with anything
in language it is with sentences themselves; images, like sen-
tences, represent events, situations and states of affairs. They are
not like names and descriptions, which pick out particular ob-
jects. So we cannot hope to find in the articulation of images
anything like the internal syntactic structure of a sentence. The
most we can hope to do is to latch onto the linguistic model at
the point where sentential connectives and operators are intro-
duced: the familiar truth functions like "and" and "not", to-
gether with intensional operators like ''because", "causes" and
the rest. Even if there were a genuine parallel discoverable at
this point between the articulation of images and that of sen-
tences (I shall argue that there is not one) it would not vindicate
the idea that a theory of language comprehension will explain
the comprehension of cinema. A theory of the sentential oper-
ators tells us how our understanding of sentence combinations
depends on our understanding of the sentences which make
them up. To account for our comprehension of the sentential
components we need a prior theory about how we understand
"atomic" sentences - those which do not have sentences as
parts. This very basic part of linguistic theorizing tells us about
the conventions which assign meanings to individual words,
and about the rules for combining words into significant sen-
tences. This part of the theory has, as I have argued, no coun-
terpart in any theory of how we understand images.
134
Are there, in fact, any significant parallels between the con-

nections we find in cinema between images, and the connections
we find in natural languages between sentences? Note that sen-
tential connections are indicated by means of convention: it is a
convention, with us, that we indicate, by the use of "and" to
join together two sentences, that the resulting combination is
true only if both of its sentential components are true. Consider
a fairly standard and familiar image combination in film: one
image is connected to the next by means of a fade out and in.
Typically, this represents there being a significant lapse of time
between the events depicted in the first image and those in the
second. This may reasonably be called a convention: other ways
of connecting the images would do as well, or nearly as well, to
indicate the desired relation. Perhaps other cinematic tech-
niques, like shot-reverse-shot editing and eyeline matches, are
conventions, though we might need a fairly generous account
of convention to make them so. But still, these conventions, if
that is what they are, do not amount to anything like the sys-
tematic, articulated set of conventions that govern a natural lan-
guage. At most they enrich the meaning of an already
nonconventionally meaningful structure; joining two shots or se-
quences by means of a fade adds meaning to that already pres-
ent in the two shots taken individually; it may even alter that
meaning. But if there were no nonconventional meaning there
already, the fade would add none. With language it is the other
way about; without convention there is no meaning at all, and
two meaningless sentences cannot be joined to give a meaning-
ful combination. Nor do the cinematic conventions I've just de-
scribed form any kind of system. There are no rules, recursive
or otherwise, by which meanings are built up by, say, the inter-
polation of a fade between the members of a shot-reverse-shot
pair and according to which certain such combinations are
meaningful or grammatically correct. All one can say of such a
combination is that it does, or does not, work effectively to in-
dicate some intended relationship between what the images
themselves represent.
In just one way, the conventions of film are like the conven-
tions of language: they set constraints on what the film maker
135
can communicate to her audience about how the film is to be
interpreted. Wittgenstein asked why you can't say "It's hot" and
mean "It's cold". The answer is that in certain circumstances
you can - for example, where you can count on your utterance
being understood as ironic. But these circumstances do not al-
ways prevail, and we simply cannot use words however we like
(or if we do, we won't be understood). Just so with film. The
film maker cannot combine a pair of shots with a fade and mean
anything she likes by it, and a shot-reverse-shot pair combined
by means of a fade is likely to produce puzzlement unless there
are strong contextual cues about what might be going on. What
the film maker can communicate to viewers about how they are
to interpret the images is limited by the conventional use to
which the fade and other devices are put. In no respect other than
this is the combination of film images like a language.
4.8 A LANGUAGE OF VISION?

An enthusiast for the language of images might try one last line
of defence. According to my argument, we understand the de-
pictive content of images by deploying the capacities we have
to recognize the objects depicted. How, in fact, do we recognize
those objects? How, in other words, does ordinary vision work?
The answer is, I'm afraid, that we do not know. But one recent,
plausible and to some extent empirically supported hypothesis
might give hope that we are on the verge of discovering a lan-
guage of vision. The hypothesis in question is that "a given view
of an object is represented as an arrangement of simple primitive
volumes, called geons" .'9 Geons are volumes which belong to a
larger set of volumes called generalized cones; geons include
cubes, cones, pyramids, cylinders and other simple and familiar
geometrical shapes in three dimensions. According to this hy-
pothesis (sometimes called "recognition-by-components") com-
binations of two, or at most three, geons are sufficient to specify
just about any object people typically are able to identify (there
are currently estimated to be about thirty thousand of them).
•• Biederman, ''Visual Object Recognition", pp. 12-13.

One distinct advantage of the geon model is that it explains how
people are able to identify an object when viewed from a variety
of positions, even when they have not previously seen the object
from that position.
Sometimes representing a seen object as a combination of
geons in certain spatial relations is called "parsing", suggesting
an analogy with the decomposition of sentences into their sen-
tential parts. Might the geons and their combinations constitute
something like a language of vision? I say not. The mind's stock
of geons, if it has one, is nothing like the lexical knowledge of
the speaker of a natural language. The geons are not represen-
tational devices conventionally connected with what they rep-
resent. They just are the shapes they are, and an object either
does or does not, as a matter of fact, have the shape of a certain
articulated set of geons. There is nothing conventional about the
relation between geons and the objects they help us to recognize.
Nor is there anything corresponding to a grammar of geons.
While there may be compositions of geons which, contingently,
don't correspond to the shape of anything we see in the world,
there are no combinations of geons which "don't make sense",
or are ungrammatical. Any articulation of geons corresponds to
a perfectly coherent shape; there just might not be anything
which has that shape. The recognition-by-components hypoth-
esis suggests one way that objects in the world can be thought
of as structured. There is no reason to think that their structure
has anything in common with that of a language, and no reason
to think that the skills we employ in analysing that structure
bear any interesting relation to linguistic skills.
137
Part II
Imagination
cinematic and other, engage the imagination.
F
ICTIONS,
I present a theory according to which the imagination
is a mental mechanism with certain uses and charac-
teristics which can themselves shed light on the psychology
of fiction. I develop a theory of visual imagining appropriate
to the cinematic case.
Chapter 5
Imagination, the general theory
Imagination, without inducing the experience I imag-
ine, delivers the fruits of experience.
Richard Wollheim
I have argued that it is a mistake to suppose that fictions, cin-

ematic and other, create an illusion of reality. Fictions, I will
argue, appeal not to belief, but to the faculty of imagination. We
shall see that I mean talk of a faculty quite literally here. The
imagination I believe to be a purpose-built system within the
mind, and one subject to selective damage in the way that vision
and the components of vision are.
Imagination has been low on the agenda of recent analytic
philosophy, perhaps on account of its presumed frivolity. Pos-
sibly by way of a reaction, the philosophical writing we have
on the imagination has often been driven by a concern with
psychopathology, certainly evident in the work of Wollheim,
and in the otherwise very different (and generally less reward-
ing) work of film theorists influenced by Lacan.' We need a
psychopathology of imagination. But if we seek a naturalistic,
biological explanation of imagination, we cannot rest with psy-
chopathology; it is hardly credible that so complex and subtle a
mechanism as the imagination should have developed and been
' See, e.g., Richard Wollheim, "Imagination and Identification", especially the
later part. It is noticeable that when occasionally a writer on film seeks a
theoretical alternative to the Lacanian framework, the alternative is still lo-
cated within the realm of theorizing about the psychopathological - as with
Gaylyn Studlar ("Masochism and the Perverse Pleasures of the Cinema"),
who turns to Gilles Deleuze's Masochism: An Interpretation of Coldness and Cru-
elty.
Imagination
sustained if it served no positive purpose. So we need an ac-
count of what is sometimes called the proper functioning of
imagination: a biologically oriented theory that explains the
adaptive benefit we gain from having the capacity to imagine.
But from this point of view, imagination is puzzling. If we think
of the development of mind in evolutionary terms, we might
expect mechanisms of mental representation to be selected ac-
cording to their reliability, and a mechanism that systematically
misrepresents the world, as the imagination surely does, would
seem to have enormous costs and no obvious benefits; for what
we imagine often bears no systematic relation to what actually
is! As we shall see, the capacity for imagining does serve im-
portant cognitive, information-gathering goals. One effect of this
tum towards science will be, I hope, the long-due recognition
that we cannot answer all the philosophically interesting ques-
tions about imagination by a combination of introspection and
a priori analysis. 3
5.1 PERSPECTIVE SHIFTS

Between roughly the ages of three and a half and four and a
half, children normally undergo an important shift in their un-
derstanding of the relation between themselves and others.
Before the shift, children find it difficult to understand that
someone's beliefs can be different from their own; the shift oc-
curs when they come to understand this. Here's an experiment
which illustrates this rather vividly. 4
Sally, a puppet, is seen to hide a sweet in box A. She leaves
the room. Another puppet is seen to move the sweet to box B.
• This problem was pointed out by Alan Leslie, "Pretence and Representation:
The Origins of 'Theory of Mind'". But I reject Leslie's solution. See my "Imag-
ination and Simulation: Aesthetics Meets Cognitive Science".
3
Occasionally, governments cotton on to the power of imagination. According
to the Guardian (August 2, 1992), "Imagining the president's death is in Kenya
a crime punishable by death."
4
The experiment was devised by Hans Wimmer and Josef Perner. See their
"Beliefs about Beliefs: Representation and Constraining Function of Wrong
Beliefs in Young Children's Understanding of Deception". Their result has
been replicated in a great variety of settings.
Sally returns. Question: where does Sally think her sweet is?
Three-year-olds say ''box B". The child thinks the sweet is in
box B, having seen it moved there. The child then credits Sally
with a belief the same as his or her own. By four and a half,
most children realize that Sally did not see the sweet moved,
and that her belief will consequently differ from the child's own.
Before the shift, the child attributes to others his or her own
perspective on the world. After the shift the child is able to
calculate how a different perspective will lead to different be-
liefs.
Perhaps "calculate" isn't quite the right word here. Calcula-
tion suggests that the child possesses some sort of theory- per-
haps a tacit theory, not consciously understood - and applies
the theory to work out, in a particular case, what the other per-
son's belief will be. For instance, the child might, in responding
to the Sally experiment, apply the principle that if someone does
not see something happen they will generally not know that it
has happened, concluding that Sally does not know that the
sweet has been moved. Many psychologists and philosophers
working in the area of cognitive science think that this is, in fact,
what the child does. They think that the shift in understanding
I have just described occurs when the child acquires a new the-
ory or part of a theory, a theory of mind, or acquires the ability
to access and apply that theory. 5 But that is not the only possible
view. One can suppose instead that the shift comes about
through the acquisition of an ability, not through the acquisition
of a theoretical principle, albeit a rough-and-ready one. On this
view, the shift is a matter of knowing how rather than knowing
that. 6
5
See, e.g., S. Baron-Cohen et al., "Does the Autistic Child Have a 'Theory of
Mind'?"; Leslie, "Pretence and Representation"; and S. Stich and S. Nichols,
"Second Thoughts on Simulation". On some views, Leslie's for example, the
theory in question is presumed to be innate. According to others it is learned.
See, e.g., Alison Gopnik, "What Is It if It Isn't a Theory?"
6
Alvin Goldman, "Interpretation Psychologized"; Robert M. Gordon, "Folk
Psychology as Simulation"; Jane Heal, "Replication and Functionalism"; and
the articles in a double issue of Mind and Language, 7, nos. 1-2 (1992) devoted
to simulation theory.
143
Imagination
5.2 SIMULATION
On the knowing-how view, our basic mode of access to the
minds of others works like this: I imagine myself to be in the
other person's position, receiving the sensory information the
other receives. Having thus projected myself imaginatively into
that situation, I then imagine how I would respond to it: what
beliefs and desires I would have, what decisions I would make
and how I would feel having those perceptions, beliefs and de-
sires, and making those decisions. But, crucially, this process of
"imagining how I would respond" is not a matter of my calcu-
lating how I would respond by appeal to rough-and-ready prin-
ciples of mental functioning. Imagining having certain beliefs
and desires is not a matter of considering the proposition that I
have those beliefs and desires, and then deducing, on the basis
of a theory, what I would do as a consequence. Rather, I simply
observe how I do respond in imagination. To imagine having those
beliefs and desires is to take on, temporarily, those beliefs and
desires; they become, temporarily and with other qualifications
I shall describe in a moment, my own beliefs and desires. Being,
thus temporarily, my own, they work their own effects on my
mental economy, having the sorts of impacts on how I feel and
what I decide to do that my ordinary, real beliefs and desires
have. I let my mental processes run as if I really were in that
situation - except that those processes run "off-line", discon-
nected from their normal sensory inputs and behavioural out-
puts. In that way I use my own mind to simulate the mind of
another.
For example, I might start off my imagining by taking on in
this way the beliefs and desires, and also the perceptions, of
someone who sees a lion rushing towards him. These beliefs and
desires then operate on me through their own natural powers;
I start (if my imagining is vivid enough) to feel the visceral
sensations of fear, and I decide to flee. But I don't flee; these
beliefs and desires -let us call them pretend or imaginary beliefs
and desires - differ from my own real beliefs and desires not
just in being temporary and cancellable. Unlike my real beliefs
and desires, they are run off-line, disconnected from their nor-
144
mal perceptual inputs and behavioural outputs. I start my sim-
ulation without requiring that the causal chain be initiated by
actually seeing a lion, and end it at the point where the decision
is made, but before that decision is translated into action. The
function of the simulation is not to save me from a lion, since I
am not actually threatened by one, but to help me understand
the mental processes of someone so threatened. If my own mind
is a reliable model of the other person's mind, then this process
of mental simulation is a good guide to the mental states of the
other. I can simply note that I formed, in imagination, a certain
belief, desire or decision, then attribute it to the other. Let us
call the hypothesis that we do, at least sometimes, come to a
view about other people's mental states in this way, the Simu-
lation Hypothesis. On this hypothesis, the shift which takes place
when a child is able to comprehend another's belief which is,
by her own lights, false, occurs like this: before the shift, the
child is unable to project herself imaginatively into the position
of the other; the shift occurs when the child becomes able to run
a simulation which takes as initial conditions a perceptual state
different from her own. Prior to that point, that child can attrib-
ute belief only by directly attributing to the other her own belief,
and she is therefore unable to cope with differences of belieF
An analogy may be helpful in understanding what is at stake
between the knowing-that view (what is sometimes called the
"theory theory") and the knowing-how view, which I have
called the Simulation Hypothesis. Suppose you wish to know
whether a bridge will withstand a certain kind of stress. One
way to find out would be to have a theory about the strengths
of materials and the way forces act on them, and to calculate,
7
The problem for the child at three years old is not that she lacks the concept
of a belief different from her own ("false belief", as this is sometimes called
in the literature). For children at this age can, in certain circumstances, exhibit
an understanding of false belief- citing, for example, an agent's false belief
that the cat is in the garage in order to explain the fact that the agent is
searching in the garage instead of under the table, where the cat really is (see
H. Wellman and K. Bartsch, "Young Children's Reasoning about Beliefs'').
The three-year-old's problem seems rather to be that she finds it difficult,
without substantial prompts, to correct for false belief when imaginatively
projecting herself into the agent's position.
145
Imagination
on the basis of the theory, whether the bridge will withstand
the stress. If you don't have such a theory, but instead have a
reliable model of the bridge, you can answer the question by
subjecting the model to just those stresses (assume that there are
no scaling calculations to do). The Simulation Hypothesis says
that we do have a reliable model of the mental processes of
others, namely our own mental processes run off-line. Using the
model, we are able to draw conclusions about other minds with-
out having a theory of how minds, including our own, work.
But what reason is there to think that my own mind is a
reliable model of the minds of others? To be a reliable model it
needs to be similar to the minds of others. How similar are our
minds? Roy Sorensen has argued that if simulation theory is
correct, we may expect agents within a given population to be
mentally similar, because being mentally average will confer a
selective advantage.8 The more like others I am, the more reli-
able will be the process of mental modelling I engage in, the
more I will be able to work out what others believe and desire,
what they believe I believe and desire, and what they desire I
believe and desire. That way, I shall be better able to cooperate
with them when cooperation suits me, and to compete with
them when it does not. That way I have a better chance of sur-
viving and breeding. And as people's minds bunch more closely
around the average, the pressure to be average increases, be-
cause by being average there are more and more people I can
successfully simulate. On this view, we may expect an acceler-
ating tendency towards mental homogeneity.
Sorensen's argument is not in favour of the Simulation Hy-
pothesis. That is, it does not provide a reason for saying that we
are simulators rather than theorists when it comes to under-
standing the mental states of others. Rather, it is a defence of
the Simulation Hypothesis against an objection to it. Sorensen's
point is that, if we are simulationists, we can expect that evo-
lution will have made us mentally similar enough for simula-
tions to be successful a significant proportion of the time, and
8
Roy Sorensen, "Self-Strengthening Empathy", unpublished.
so it is no objection to the Simulation Hypothesis to say that
simulation would not deliver the desired result because of men-
tal diversity.
I believe that the Simulation Hypothesis is very important for
our understanding, not just of how we comprehend other minds,
but of how we engage with fictions, including cinematic fictions.
I say that fictions are devices which encourage and guide the
imagination; fictions tell stories, and the things which are parts of
the stories they tell are things which those fictions authorize us to
imagine.9 Clouds of Witness authorizes us to imagine that Lord Pe-
ter establishes his brother's innocence, for that is part of the story
it tells; conversely, it prohibits us from imagining that his brother
is hanged for murder, for it is part of the story that he is not han-
ged. (Sometimes, of course, our decisions about what we are au-
thorized to imagine are more complex and controversial.) We
can, of course, improvise our own stories, providing ourselves
with spontaneous fictions that encourage us to imagine ourselves
in interesting and colourful situations. But we often enjoy having
our imaginations guided by an external source - the work of an
author whose skill in story construction may be superior to our
own and whose outcome we can't predict in advance. Such an ex-
ternal source, however encoded and through whatever sensory
channels it is received, is a fiction, in my sense.'o Films are en-
coded in visual and auditory depictions, and are received
through the senses of sight and hearing. Novels, while normally
received through the sense of sight (we normally read with our
eyes) are encoded not in visual depictions, but in linguistic sym-
bols accessible to sight. Both are forms of fiction.
"Imagination" has a number of senses, none of them very
clear. But one kind of imagining is, I believe, the process of
running our mental states off-line. That is the kind of imagining
which takes place when we engage with fictions. So far, I have
described off-line simulation only in the context of our attempts
to comprehend the mental states of another. But that is simply
one function of simulation. Simulation itself occurs when, for
• See Walton, Mimesis as Make-Believe, p. 51.

0
' See my Nature of Fiction.
147
Imagination
whatever reason, we run our mental states off-line. One reason
we can run our mental states, in particular our belief-desire sys-
tem, off-line is to engage with a fictional work. That happens
when the fictional story itself provides the inputs to the simu-
lation.
Consider first a nonfictional work: a newspaper article or tel-
evision documentary. If we think the work reliable, we shall
form certain beliefs based on the information the work conveys.
We may also acquire certain desires: documentaries about the
dangers of smoking can make you want to give up, and travel
articles extolling the virtues of an exotic location can make you
want to go there. (So as to simplify things, I'll concentrate on
the case of belief for the moment.)
A fictional work, assuming we know that it's fiction, can have
effects on our mental processes similar in various ways to the
effects of nonfiction. Fictional works can engage our attention,
and they can have what is, on reflection, a surprising capacity
to move us. But we do not acquire from them beliefs in the
straightforward way that we acquire beliefs from nonfiction.
With fictions, our mental processes are engaged off-line, and
what we acquire instead of beliefs is imaginings which simulate
belief. When I work out what Sally believes about the location
of her chocolates, I mentally simulate Sally's mental processes,
her having certain perceptions and thereby acquiring certain be-
liefs. In other words, I imagine being in Sally's situation, re-
sponding as she responds. When I engage with fiction I simulate
the process of acquiring beliefs - the beliefs I would acquire if
I took the work I am engaged with for fact rather than fiction.
Here I imagine myself acquiring factual knowledge. In the Sally
case I imagine being in the situation someone else actually is in;
in the fiction case I imagine being in a situation I could be in
but actually am not in. In the first case the imagining is an in-
strument to a further purpose: to inform me of the mental proc-
esses of someone else. In the second case the imagining has no
such further purpose: the simulation provoked by the fiction I
read is simply something I enjoy."
" Simulations engaged in in the course of reading or watching fiction might

We can think of beliefs and desires as states with two aspects:
a representational and a functional aspect. A belief and a desire
might have identical representational contents. I believe that it
will rain, and I hope that it will. Here, the content is the same:
that it will rain. But despite their sameness of content, the belief
and the desire are distinct kinds of states. While they do not
differ in representational content, they differ functionally, that
is, in terms of causes and effects. Believing that it will rain tends
to be caused by kinds of sensory and other experiences different
from those which tend to cause desiring that it will rain, and
believing that it will rain will tend to have consequences for my
behaviour different from the behavioural consequences of desir-
ing that it will. Belief and desire are different attitudes, and their
being different in this regard is a matter of the functional dif-
ferences between them. 12
I said that what makes something a belief rather than some
other kind of state is its causes and effects. In that case there is
something unsatisfactory in my description of simulation as
"running beliefs off-line" (I'll come to the issue of desires in a
moment). Running off-line is, exactly, a matter of severing the
connections between our mental states and their perceptual
causes and behavioural effects. A belief run "off-line" isn't really
a belief, just as a monarch who has been deposed is no longer
a monarch. Revolutions transmute monarchs into ex-monarchs.
Simulation transmutes beliefs into imaginings. Just as a belief
and a desire may have the same content but differ functionally,
so may a belief and an imagining. Believing that it will rain has
certain connections to perception and to behaviour which, when
they are severed, transmutes that belief into a case of imagining
that it will rain.
Since imaginings are states run "off-line", does that mean that
imaginings completely lack a functional aspect, being, as it were,
pure representations? No. While they lack the connections of
have a further purpose: to help me pass a literature or film-studies exam, for
example. My point is that they need not have any further purpose in order
to seem worthwhile to us.
" For an attempt to account in functional terms for the difference between
belief and desire see Jamie Whyte, "The Normal Rewards of Success".
149
Imagination
belief to the external world via perception and behaviour, they
retain some internal connections. With off-line simulation, states
of imagining function as internal surrogates of beliefs because
they retain belief-like connections to other mental states and to
the body. Imaginings, like beliefs, can lead to decisions, and can
cause certain kinds of bodily sensations. Compare believing that
you are confronted by a dangerous bear with imagining that
you are. Imagining there is a bear in front of me, just like be-
lieving there is one, may cause a decision to flee. But in the
imagining case the decision itself is an imaginary one; it is
something that stands to real decision as imagining stands to
belief. And vividly imagining there is a bear there will cause me
to have some of the unpleasant visceral and other bodily feelings
that believing there is a bear there would, though perhaps less
intensely. Note that in this case what is caused by the imagining
is a real state of feeling and not an imaginary one. Imagining
yourself in danger causes you really to feel disturbed; it does
not merely cause a state of imaginary disturbance. The reason
for that is that feelings are states identified, not in terms of their
function, but in terms of how they feel. A state which feels like
a bodily sensation really is one.
I have been discussing the relation of imagining to belief, and
I said I would bring desire into the picture. If mental simulation
is how we understand the minds of others, there must be states
that stand to desire as imaginings stand to beliefs; for us ade-
quately to simulate the mental states of others, we must simulate
their desires as well as their beliefs. So as well as pretend or
imaginary beliefs- what I have been calling imaginings- we
need pretend or imaginary desires. Fictions provoke imaginary
desires just as they do imaginary beliefs. Reading the story, I
want the hero to succeed. Or do I? After all, I am perfectly aware
there is no hero to be the object of my desire. How can I desire
something for an entity I don't believe in?'3 What's really hap-
pening is that desires are being run off-line, in tandem with
3
' For more on this absence of belief problem, see my Nature of Fiction, chapter
5· See also this chapter, Section 5+
pretend beliefs. I have a pretend desire that the hero succeed,
backed up by a pretend belief in his existence.
Couldn't we be more economical here, supposing that there
is one unified state of imagining, and that what we might neu-
trally call "pretend beliefs" and "pretend desires" are accounted
for by saying that in the one case we imagine that we believe
something, and in the other that we imagine that we desire it?
That way we get by with three categories - belief, desire and
imagining - instead of my four - belief, desire, pretend belief
and pretend desire. But this admirably economical move has
problems. It would not then be possible simply to imagine (in
the intuitive sense) that some state of affairs occurred; I would
have to imagine that I believed that it occurred. Now consider
the case of imagining that something occurred and that no one
knows anything about it (cases of this kind will recur when I
discuss imagining in response to film). That would require me
to imagine that I believe that it happened and that no one else
believed that it did. The content of my imagining would in that
and similar cases be paradoxical. But it does seem that I can
imagine that something happens that no one knows anything
about, without thereby imagining something impossible. So
imagining something cannot, in general, be equivalent to imag-
ining believing it.'4 In that case we really are going to need re-
course to what I have called "pretend desires" as well as to
pretend beliefs, in order to explain important aspects of mental
processing.
It is important to see what sort of claim I am making here about

the relation between imagination and mental simulation. In say-
ing that imagination is simulation I am not, for example, offering
a piece of conceptual analysis that might serve to introduce the
concept of imagining to one who lacked it. On the contrary, it
may well be that, to get a grip on the notion of simulation, one
'
4
Also, of course, I can imagine that no one is doing any imagining, but I trust
there is nothing even apparently paradoxical about what I am imagining in
that case.
Imagination
needs already to possess the concept of imagining. My proposal

has closer methodological affinities to such essentialist identifi-
cations as the claim that water is H,O. This is not to be under-
stood as offering an analysis of the concept of water; the claim
is rather that water has a hidden inner structure that is explan-
atory of its more evident surface properties, and which may
bring with it some surprising consequences - that water is not
really a continuous substance, for example. Likewise, the cash
value of the claim that imagining is simulation lies in its ex-
planatory power. The rest of this chapter is designed to give a
brief indication of that power.
5·3 FICTION AND TWO KINDS OF

SIMULATION
I introduced the idea of imagining-as-simulation by describing

our simulation of the mental states of other real people. I then
suggested that the capacity for simulation is something we ap-
ply to fictional works, and that these two functions of simulation
are quite different. Perhaps they are not always so different.
Might it be, for example, that fictional works sometimes require
or encourage us to simulate the mental states of their characters?
I believe the answer is yes: understanding, appreciating and
learning from a fictional work (it might be a cinematic or other
kind of fictional work) sometimes requires that we simulate the
mental states of a character within the fiction.
I want to make a distinction between two kinds of imaginings
that fictions encourage. There is, first of all, imagining what is
fictional in a story. Part of engaging with a fictional work con-
sists of imagining those things which it makes fictional, as Anna
Karenina makes it fictional that Anna commits suicide.'5 Cases
where we imagine what is fictional, I call primary imaginings.
Secondary imagining occurs when we imagine various things so
as to imagine what is true in the story. Frequently, secondary
imaginings are not required for primary imagining to take place:
5
' On this see my Nature of Fiction, chapter 2; and Walton, Mimesis as Make-
Believe, chapter 2.
the story has it that a certain character walked down a dark
street, and we simply imagine that. Then we have primary imag-
ining without the need for secondary imagining. Primary imag-
ining most notably requires secondary imagining in cases where
what we are primarily to imagine is the experience of a char-
acter.'6 If the dark street hides something threatening, the char-
acter who walks it may have thoughts, anxieties, visual and
auditory experiences and bodily sensations that it would be im-
portant for readers to imagine something about. The author may
indicate, to a greater or lesser degree of specificity, what the
character's experience is. But it is notoriously difficult, and in
some cases perhaps impossible, for us to describe people's men-
tal states precisely. Authors who adopt stream of consciousness
and other subjective styles have failed to do it, and so have film
makers like Hitchcock who try to re-create a character's visual
experiences onscreen. Anyway, the attempt at full specificity
and precision in this regard would usually be regarded as a
stylistic vice, leaving, as we significantly say, "nothing to the
imagination". What the author explicitly says, and what can be
inferred therefrom, will constrain our understanding of the char-
acter's mental state. It will set signposts and boundaries. But if
these are all we have to go by in a fiction, it will seem dull and
lifeless. It is when we are able, in imagination, to feel as the
character feels that fictions of character take hold of us. This
process of empathetic reenactment of the character's situation is
what I call secondary imagining. As a result of putting myself,
in imagination, in the character's position, I come to have imag-
inary versions of the thoughts, feelings and attitudes I would
have were I in that situation. Having identified those thoughts,
6
' I am using the term "experience" here in a very broad sense. In my sense,
what a character experiences includes visual, auditory, bodily and other sen-
sations, and also the beliefs, desires and intentions it is fictional that she has,
in so far as they are accessible to her consciousness. Some people argue that
states with propositional content like beliefs and desires are accessible to
consciousness exactly because they have a characteristic feel in the way that
sensations do (see, e.g., Alvin Goldman, "The Psychology of Folk Psychol-
ogy"). But this is very much a minority opinion, and I do not want to judge
this issue one way or the other here.
153
Imagination
feelings and attitudes ostensively, I am then able to imagine that
the character felt that way. That is how secondary imagining is
a guide to primary imagining.'7
Novels with characters into whose mental lives we readily
project ourselves are not works in which we ourselves are char-
acters. While I may simulate the situation of the character, imag-
ining being in his situation, that piece of imagining does not
correspond to anything which the novel makes fiction. It is, ex-
actly, secondary imagining- imagining undertaken because it
will put me in a position where I will be able to imagine
something the novel makes fictional: that the character has cer-
tain thoughts and feelings in response to his situation. It is that
further imagining which is primary.
Being able to construct fictions with characters whose mental
7
' There are times when secondary imagining fails us, because our minds are
not attuned to the task at hand. Robert Gordon drew my attention to the
following passage from Time magazine (May 9, 1994), by John Skow, entitled
"Harold Brodkey's New Novel Is Erotic- But Not to Everybody": "Were-
spond to stories with astonishing versatility of imagination. The three-year-
old listening to his grandmother momentarily becomes Peter Rabbit; the
geezer reading Patrick O'Brian's sea stories feels scared on the quarterdeck
of a storm-blown frigate. But the distinction between what the reader imag-
ines and what he actually experiences remains solid - the geezer does not
actually get seasick. Over the whole range of literature, only erotica functions
differently. If it works, sexual arousal is real, not imaginary. And if it doesn't
work? The most recent example is Harold Brodkey's novel Profane Friendship
(Farrar, Straus & Giroux; 387 pages; $23). The author tells of a long, intensely
erotic affair between the narrator, an American novelist named Nino, and an
Italian named Onni. The names are anagrams of each other- different stir-
rings of the same ingredients, including the same sex. If the drama is to
succeed, the passion must not merely engage the reader intellectually; it must
arouse him. For this heterosexual male, who has imagined himself to be Moll
Flanders and Jonathan Livingston Seagull, the failure is total. Such a state-
ment will surely be called homophobia, but fear and disapproval are not
operating here. In fact, nothing is operating. The reader's reaction is vague
exasperation. His mind simply does not have the software to induce the
intended physiological response to the author's erotic obsessions, and these
are the essence of the book. Such thoughts, of course, must occur regularly
to gays when they read about hetero sex. You don't have to be a rabbit to
enjoy Beatrix Potter, but you may have to be either gay or straight to appre-
ciate gay or straight erotica."
154
lives readily lend themselves to simulation is not a capacity
every fiction maker possesses or wishes to exploit. Through in-
competence and sometimes through design, the characters of
fiction resist simulation: their responses to situations, their
words and even their thoughts (in so far as the author lets us
know what they are) seem not to be those we would have in
their situations. In these cases, when we engage with the fiction,
trying to guess what will happen, trying to fill in its background
of unstated presuppositions and undescribed events, we may
have to rely more than usual on inferences we make concerning
the author's intentions or about the constraints of form and
genre the work conforms to. A rough guide to the degree of
naturalism in a work of fiction is the extent to which we can let
our own minds model those of the characters. To the extent that
these minds are opaque, not merely in the sense of being un-
derdescribed by the author, but in that of resisting simulation,
we make a work that rejects the standards of naturalism.
Is it just an historical accident that most of our fiction is nat-
uralistic in this sense? Despite frequent calls for, or announce-
ments of, the death of naturalism, I think it unlikely. The drive
to simulate seems to be a very powerful one. And with good
reason; I shall argue in the next section that the disastrous effects
of simulative failure are evident in the aloneness and multiple
incompetencies of autism. The close historical association be-
tween fiction and psychological naturalism suggests that fictions
have thrived exactly because they give us opportunities to ex-
ercise our capacity for simulation. Fiction may not be logically
tied to psychological naturalism any more than painting is log-
ically tied to naturalistic representation, but in both cases the
psychological connection seems to go very deep.
5·4 CONSEQUENCES
Identifying imagination with simulation has interesting conse-
quences; some of these concern features of imagining generally
agreed on, while others require us to attribute to imagining fea-
tures that can seem counterintuitive but which I believe to be
correct. I want to point out some of those features.
155
Imagination
First, given the characteristics of off-line simulation I have
described, we can now explain certain features of our experience
when we are engaged by fictions. While fictions do not cause
us to believe in the reality of the fictional story, they can engage
us to the extent of causing within us the sometimes pleasant
and sometimes unpleasant bodily states we associate with being
emotionally moved by events. If fictions encourage simulations,
and simulated beliefs and desires retain their internal connec-
tions to our bodily states, that is exactly what we would expect.
The anxiety that watching horror movies induces in me does
not cause me to call the police, but it does cause me to feel afraid.
Second, the idea that imagination is simulation explains one
peculiarity of imagining in relation to belief and desire: that
there is no such thing as dispositional imagining. With belief and
desire we have the distinction between dispositional and occur-
rent states. I may believe something I have never thought about
and which plays no role in explaining my past or current be-
haviour. Similarly for desire. In these cases we say that the belief
or desire is dispositional rather than occurrent; the state is pos-
sessed in virtue of the fact that if a relevant situation were to
arise you would be likely to behave in a way that would betray
the belief or the desire. But there is no comparable category of
dispositional imagining, and all imagining is occurrent (though
not necessarily conscious, as I shall argue). And while I may be
disposed to imagine certain things, the possession of such a dis-
position does not constitute my dispositionally imagining them.
But it is easy to see that with simulation there is also no dis-
positional-occurrent distinction to be made. You are either run-
ning a simulation or not, and being disposed to run a certain
simulation does not constitute dispositionally simulating - no
more than being disposed to run a mile means you are dispos-
itionally running one.
Third, in identifying simulation and imagination we vindicate
the intuition that imaginings are somehow secondary or deriv-
ative states. Imagining something may be different from believ-
ing it, but it does seem as if there is an asymmetrical dependence
of imagining on belief. We can imagine a creature which has
beliefs but lacks the capacity to imagine. Indeed, there may ac-
156
tually be cases of this in our population. People suffering from
autism show a pattern of disabilities strikingly suggestive of the
hypothesis that what they lack is, literally, imagination.'8 The
condition is marked by a considerable degree of inflexibility and
isolation in social relations; there is no spontaneous enthusiasm
for games of imagination and pretence, and there is considerable
difficulty in verbal communication, though high functioning au-
tistic individuals often have functional mastery of language. Suf-
ferers, even those few of high ability who hold responsible jobs,
have difficulty understanding that other people have different
knowledge, beliefs, desires and generally a different mental out-
look on the world from themselves; and autistic children typi-
cally fail tests of understanding of false belief, such as the
''Sally'' test already described.' 9 This inability to understand the
mental states of others may explain why autistic subjects fail to
provide relevant background information in their speech, fail to
notice that others are bored by their own favourite topics of
conversation, fail to understand that their own behaviour has
caused offence. Often they seem to lack the ability to deceive,
or even to understand the concept of deception.
This inability to comprehend mental states, together with the
'" Autism was first identified by L. Kanner ("Autistic Disturbances of Affective

Contact"). The condition often goes with a greater or lesser degree of general
mental retardation, but 20% of autistics are in the "normal" range.
9
' In a version of Wimmer and Perner's false-belief experiment conducted by
Baron-Cohen and colleagues ("Does the Autistic Child Have a 'Theory of
Mind'?"), three groups of children were tested: clinically normal four-and-a-
half-year-olds, a group of Down's syndrome children (mean IQ = 64, mean
age = eleven years) and a group of autistic children (mean IQ = 82, mean
age= twelve years). The success rate in the normal and Down's syndrome
groups was high: 85% and 86% respectively, while the success rate in the
autistic group was only 20%. See also D. Roth and A. Leslie, "The Recogni-
tion of Attitude Conveyed by Utterance: A Study of Preschool and Autistic
Children". About 30% of autistic subjects pass the false-belief test and its
variants. When set a task that tests understanding of higher-order belief (be-
lief about belieO which normal six-year-olds pass, these autistic subjects fail
(S. Baron-Cohen, "The Autistic Child's Theory of Mind: A Case of Specific
Developmental Delay''). For a simulative account of autistic problems on so-
called false-belief tests see my "Simulation-Theory, Theory-Theory and the
Evidence from Autism".
157
Imagination
lack of interest in pretence and the fictional, when taken in con-
junction with the idea that we understand other people's mental
states by an act of imaginative projection, strongly suggests that
the basic deficiency of autism is a lack of imagination. But un-
doubtedly, autistic people have beliefs and desires. On the other
hand, it is difficult to imagine a being who had the capacity for
imagining, but not for belief. The Simulation Hypothesis ex-
plains this asymmetry. Since simulation runs the belief-desire
system off-line, imagining is parasitic on these other mental
states.
If imagining is simulation, then I think we have an explana-
tion, or the beginnings of one, for a puzzling phenomenon: our
ability to empathize with fictional characters and their situa-
tions. A familiar formulation: how can we pity Anna Karenina
when we know that she does not exist? There are problems here
of different kinds, one of which is logical: if emotions require
certain kinds of beliefs (not merely causally but constitutively)
how can we have emotions directed at fictional characters when
we lack the relevant beliefs? But putting aside the logical prob-
lem which I have discussed elsewhere/ let us concentrate on
0
the psychological issue: why do we get so involved with, dis-

traught and caring about, people whose existence is merely
imagined? I suggest that it is at least an insight into this problem
to acknowledge, with the simulation theorists, that the basic
mechanism by which we make emotional contact with other
(this time real) people involves imagination. We come to un-
derstand, not merely in propositional terms, but in an emotion-
ally attuned way, their situations; to feel as if we were in their
situations by simulating their situations, or, if I am right in the
basic identification of this chapter, by imagining. Empathizing
with fictional characters would then just be an extension of this
imaginative project; we imagine not merely that we are in
someone else's shoes, but someone in whose shoes we then
imagine ourselves.
Another feature of imagining which needs explaining is the
collapse of iterativity: a puzzle related to a problem that David
.. See Currie, Nature of Fiction, chapter 5·

2
Lewis has called the problem of the flash stockman. Try imag- '
ining someone else imagining something; what tends to happen

is that you end up imagining the something yourself. And with
those fictions, like A Thousand and One Nights, where what is
fictional is that someone is telling a fictional story, we very
quickly move to the position where we are imagining the story
told, not the telling of it. Why does imagining that some-
one imagines P tend to collapse into imagining P, whereas, for
example, believing that someone believes P rarely collapses
into believing P? If imagining is simulation, then imagining
someone's imaginings is a matter of simulating her simulation.
How would that work? Consider first the ordinary case: I sim-
ulate her beliefs and desires rather than her imaginings. Sim-
plifying, suppose she believes P and desires Q; I then give
myself appropriately off-line versions of P and Q. P, for me, is
a case of pretend belief, and Q is a case of pretend desire. In
this case I simulate the having of belief P and desire Q. Now
consider the more complex case: I simulate her imagining, that
is, her pretend beliefs and pretend desires. Suppose she has P
as a pretend belief and Q as a pretend desire; she is, in other
words, running off-line versions of P and Q. To model her men-
tal processes in my own mind (in other words, to simulate her),
what must I do? The answer is that I also must run off-line
versions of P and Q. But we have seen, from an examination of
the simple case, that this - running off-line versions of P and Q
-is what counts for me as having Pas a pretend belief and Q
as a pretend desire. The attempt to simulate her simulation has
resulted in my simply imagining what it is she imagines, and
not her imagining of it.
In this chapter I have simply asserted a claim I have argued
elsewhere- that we engage with a fictional story by imagining
its content, or what is "true in the fiction".= An objection to this
is that a good deal of the content of a fictional story is true, and
known to be so by the audience. Thus what happens in fiction
often happens against a background of reallocations correctly
" See David Lewis, postscript to "Truth in Fiction", Philosophical Papers, vol. 1.
,. See Currie, Nature of Fiction, chapter 2.
159
Imagination
described and often includes historical incidents. But is it not at

least very odd, goes the objection, to say that we imagine things
that we believe to be true?23
In reply I say that, while this may sound odd, it is just what
the identification of imagination with simulation would predict
about our capacity to imagine. If I want to figure out what de-
cision someone else will come to, I shall start by feeding into
the simulation as initial conditions some assumptions about
what the person believes. These (putative) beliefs will then be
run off-line- thereby becoming imaginings -leading to a mod-
elling of the other's decision-making process. But given that a
lot of what other people believe I believe myself, there will be
occasions when I shall have to feed in assumptions about what
the person believes which coincide with what I believe. In such
a case I shall have to run my own beliefs off-line, thereby bring-
ing about a situation in which I imagine what I believe.
There is, in addition, a lot of independent evidence that
people imagine what they believe to be true. Let us assume,
plausibly, that imagining is what drives childhood games of pre-
tence. Indeed, without this assumption, it would be difficult to
see how pretence could be distinguished from mere confusion.
The difference between the child who pretends the rocking
horse is a horse and the child who thinks the rocking horse
really is a horse might well not show up in behaviour. The dif-
ference is that in the first case, the child's behaviour is caused
by her imagining the rocking horse to be a horse, and in the
second case by her believing that it is. But it is clear, both from
casual observation and from controlled experiments, that chil-
dren incorporate a great deal of factual belief into pretence. Thus
it might be part of the pretence that, as a result of someone's
pouring activity, some cups are empty and some are full. In fact,
no real pouring has been done, and so all the cups really are
empty. Yet the child's pretence discriminates between empty
and full cups, and so a component of the pretence is that
something is true which is true: that this cup is empty. But if
the behavioural pretence is driven by imagining, it is difficult to
23
Noel Carroll, review of The Nature of Fiction.
16o
avoid the conclusion that the child is imagining something, part
of which is true.
Could we explain the behaviour by saying instead that it is
the outcome of a combination of imagining and belief? That is,
the child imagines that these two cups (which as it happens are
empty) are full, and believes that these other two cups are
empty. But what is the explanation of the fact that other beliefs
of the child do not influence the pretend behaviour? The child
believes that all four of the cups are empty, but only treats two
of them as if they were. The simplest and most natural expla-
nation seems to be that the belief that these two are empty does
become part of what the child imagines, while the belief that all
four are empty does not. And there are other cases where pre-
tence seems to coincide almost entirely with belief- as with the
case reported by Vygotsky of two sisters pretending to be sis-
ters.24
One final and rather startling implication of the assimilation
of imagining to simulation is that imagining should not be
thought of as always and automatically a conscious or even an
intentional action, since simulation, if it really does help us to
understand the minds of others, must be done unintentionally,
mostly at a subconscious level. For me that is a welcome result.
If, as I have argued, fictions function to drive imagination, they
do so in ways of which the subject is sometimes unaware, and
over which the subject rarely exerts conscious control. But there
are those who have argued that it is a priori of imaginative
pretence that it is engaged in consciously and intentionally.25
This claim seems to me symptomatic of a residual Cartesianism
that has not yet been dislodged from this underdeveloped part
of the philosophy of mind. If hitching the concept of imagination
to simulation theory helps us to abandon it, so much the better.
24
Alan Leslie discusses the issue of pretending what is believed in his "Pre-
tending and Believing: Issues in the Theory of ToMM". I owe the reference
to Vygotsky to Leslie's paper (see L. S. Vygotsky, "Play and Its Role in the
Mental Development of the Child"). For criticism of Leslie's theoretical ap-
proach to the issue of pretence, see my "Imagination and Simulation".
25
See, e.g., Noel Carroll, The Philosophy of Horror: Paradoxes of the Heart; and
Anders Pettersson, "On Walton's and Currie's Analysis of Fiction".
161
Imagination
5·5 THE DANGERS OF IMAGINING
Fictions, especially cinematic ones, are sometimes identified as

the source of violent or otherwise undesirable behaviour, and as
encouraging destructive attitudes. Whether and to what extent
they do any or all of these things is a question to which empir-
ical studies have not given a clear answer. However, in light of
the identification of imagination with simulation, it is not hard
to postulate a mechanism whereby fictions may do moral dam-
age.
Suppose I am right in thinking about imagination in terms of
off-line simulation. Imagination enables us to respond to the
world not merely as it actually is and with the beliefs and de-
sires we actually have, but as it might be, will be or was, and
to experience it as we would with beliefs and desires other than
our own. Imagination does this, according to the simulationist,
by running our belief-desire system off-line, disconnected from
action and behaviour. It must involve, therefore, some inhibiting
mechanism which effects the disconnection. But there is a dan-
ger in this. An inhibiting mechanism might go wrong, or at best
be unreliable; that is frequently how it is with mental mecha-
nisms. That way, imagination would have a tendency to spill
over into belief: what starts as an act of imagining might tum
into a real belief or desire by being brought, inadvertently, on-
line. In small ways this happens all the time: old Mr Harding's
tendency to play a nonexistent cello at times of stress might be
an example from literature, and most of us have a tendency to
reach for the relevant body part when imagining a painful in-
jury. Perhaps this also happens less often in more serious ways.
There are people of whom we want to say that they are living
a fantasy life, that they are in the grip of beliefs and desires
which derive from fiction or fantasy rather than from experience
or from sober reflection. Perhaps their inhibitory mechanisms
are more than usually impaired, and their boundary between
fact and fiction has genuinely started to blur.
In some ways, this blurring has a tendency to be corrected by
the force of external reality. I imagine I am floating pleasantly
in a warm swimming pool, but if such an imagining were to
162
show any signs of breaking down the barriers provided by my
inhibitory system and transmuting itself into a belief, the evi-
dence from all my senses would quickly come to the rescue. No
such belief could long survive its evident falsity. 26 But desiring
is under no such tight ecological control. There is, certainly, a
connection between perception and desire; perception influ-
ences, in various ways, what we desire. But the connection is
much more tenuous than that between our empirical beliefs
about the world and our perception of it, and perception does
not provide the powerful second line of defence against rene-
gade imaginings in the desire case that it does in the belief case.
So, by imagining ourselves in the situation of a character with
destructive, immoral desires, and thereby coming to have, in
imagination, the desires of the character - what I have called
secondary imagining - we may be in danger of really acquiring
those desires through failure of the inhibitory mechanism. 27
This chapter has provided a general theory of imagination. The

next will develop a theory of specifically cinematic imagining .
.,; Perhaps hallucinations, or some of them, are produced by really spectacular

blowouts in the inhibitory system.
27
There is experimental evidence that children can come to believe, or be prone
to believe, things which they have initially been asked merely to imagine
being the case. See C. N. Johnson and Paul L. Harris, "Magic: Special but
Not Excluded"; and Jacqueline D. Woolley and Katrine E. Phelps, "Young
Children's Practical Reasoning about Imagination". For interesting evidence
of leakage the other way - from beliefs and desires to imagination, see Nigel
Harvey, "Wishful Thinking Impairs Belief-Desire Reasoning: A Case of De-
coupling Failure in Adults?" The idea that there might be breakdowns in the
mechanism of inhibition governing an off-line process may be useful in other
areas. Jeannerod has recently suggested that certain kinds of compulsive im-
itative behaviour may be explained by a breakdown in the inhibitory control
which governs motor imagery; in these cases, imagining performing a move-
ment leads inevitably to actually performing it (M. Jeannerod, "The Repre-
senting Brain: Neural Correlates of Motor Intention and Imagery", p. 200).
Chapter 6
Imagination, personal and impersonal
A film makes us feel like eye-witnesses of the events
which it portrays.
Victor Perkins
Cinema is a medium in which fictions are presented pictorially.

In this way, its fictions differ from those of the novel and other
linguistic forms. But to characterize the difference between the
pictorial and the linguistic just at the point of delivery would be
to leave us with something important unexplained: the distinc-
tive experience of pictorial, and in particular cinematic, fictions.
Watching movies is very different from reading novels. It is also
different from watching plays. Any adequate theory of cinema
will have to account for the differences. That theory must tell us
not only what is distinctive about the delivery of cinematic fic-
tions but also about their reception - and about the connection
between the two. It will not be sufficient simply to say that view-
ers of a film are required to imagine various things, and that the
film is constructed so as to guide them in their imaginings, for
that is true of all fictions in whatever media. I have my own
story to tell about what is distinctive about cinematically in-
duced imaginings, but it will appear strange and unmotivated
unless it is arrived at by the elimination of more familiar and
perhaps initially more plausible rivals. So I begin with what I
take to be the main rival account which I distil, by a process of
rational reconstruction, from the writings of film theorists of var-
ious times and persuasions. When we see what is wrong with
that account we shall have reason to look on my own proposal
with some favour.
6.1 THE CLASSICAL THEORY
By the term ''classical film theory'' I mean the kind of theorizing

developed during the period of tumultuous development in the
cinema through the thirties, forties and fifties. Considered as a
period, it is generally regarded as having ended in the early
sixties with the work of Jean Mitry, though there are people
writing today who share significant doctrinal or methodological
assumptions with the classical theorists. Its greatest exponent
was Andre Bazin. The theorists I call classical often disagree
about a great deal, and perhaps it is only by contrast with more
recent developments in semiotics and the psychoanalytic explo-
ration of the cinema that we can see the classical theorists as
having anything much in common. However, there is one sub-
stantial doctrine which most of them hold. We have come across
a version of it already when discussing the doctrine of illusion-
ism. The classical illusionistic theory of cinema claims that the
viewer is made to believe that he or she is a spectator at a real
action, watching from within the space of the action, situated
where the camera is. We have seen plenty of reason to reject
that view. But perhaps we ought to consider whether some
weakened version of it is true. And indeed, there is a weaker
version of it which classical film theorists can be read as sup-
porting, though it is sometimes hard to know whether they re-
ally have a strong or a weak version in mind. This weaker
version substitutes for the notion of belief the notion of imag-
ining. Instead of claiming, implausibly, that the viewer is made
the subject of an illusion and is caused to have false beliefs about
his or her situation, we might say this: that the viewer merely
imagines that he or she is within the space of the action, watching
real events from the position of the camera. Thus when Jean
Mitry says, "I know that I am in the movie theatre, but I feel
that I am in the world offered to my gaze, a world that I ex-
perience 'physically','" we might take the phrase "feeling I am
' Jean Mitry, quoted approvingly both in Wilson, Narration in Light, p. 55, and
in Charles Affron, Cinema and Sentiment, p. 7· The quotation is from Esthetique
et psychologie du cinema, I: les structures, p. 179.
Imagination
in the world offered to my gaze" as meaning imagining that I

am in that world, and imagining yourself in the fictional world
is consistent with knowing you are actually in a movie theatre.
George Wilson, a philosopher/film theorist, is quite explicit
about the role of this kind of imagining in film: "The spectator
knows that he is in the [movie] theatre, but it is make-believe for
him that he is watching from within the space of the story. 112 It
would be appropriately charitable to theorists like Balazs, Bazin
and Panofsky to interpret them as offering a thesis compatible
with Wilson's claim. Unlike the view that we literally believe
ourselves to be within the space of the action - the view I crit-
icized in Chapter 1 - Wilson's view is not obviously wrong.
Nonetheless, I think it is wrong.
In criticizing this view, I am not going to be attacking the
idea that the film viewer imagines various things; I have been
arguing that that is exactly what fictional works, including cin-
ematic fictional works, encourage. What I shall object to is the
idea that cinematic works encourage us to imagine ourselves to
be observers of the fictional events, placed within the world of
the fiction. It will be useful to have some terminology with
which to mark a distinction I am going to be using here. When
I imagine merely that such and such happens, without imagin-
ing that I see (or have other kinds of epistemic contacts with)
what happens, we have a case of impersonal imagining. When
imagining involves the idea that I am seeing the imagined
events, we have a species of personal imagining. (There are other
kinds of personal imaginings: imagining that I am hearing
something, etc.) More specifically, it is a case of imagining seeing.
In the light of my earlier identification of imagination with sim-
ulation, this distinction seems unexceptionable. If imagination
consists of running our beliefs off-line, then any distinction that
holds among beliefs should hold among imaginings, and there
surely is a distinction between believing that an event occurs
and believing that you are seeing that event occur. I shall call
• Wilson, Narrations in Light, p. 56, italics in the original. Wilson uses the phrase
"it is make-believe for him" rather than saying the viewer imagines, but that
is certainly one plausible interpretation of his words, and it was, as Wilson
confirmed in discussion, the intended one.
t66
the view that the imagining appropriate to film is imagining

seeing the Imagined Observer Hypothesis. 3 That is the thesis we
can most charitably attribute to the classical theorists, and that
is the thesis I reject.
If the thesis is right, there is something very special about the
movies as a medium of fictional presentation; they put us, in
imagination, right there with the characters in a way that novels
and certain other media cannot do (I'll say more about plays in
this regard in a while). 4 My view is that, in respect to narrative
structure, movies are much more like novels than the Imagined
Observer Hypothesis would have us believe. Still, there are ob-
vious differences between literature and the movies which no
theory can ignore. I'll try to show that we don't need the Imag-
ined Observer Hypothesis to account for these differences.
On the whole, the classical theorists seem not merely to have
assumed that film is special in the way just described- that it
puts us in the picture- but to have approved and sometimes
celebrated this capacity. More recent theory, or a good deal of
it, can be seen as reacting against this approval and as arguing
that film must strive to overcome this tendency. But new and
classical theorists seem to have agreed that putting us in the
picture is something that film naturally succeeds in doing unless
"distancing" or "alienating" measures are taken by the film
maker to prevent it. Since I shall argue that film does not have
this capacity to any marked degree, I shall argue against clas-
sical and new theorists alike.
I noted kinds of personal imagining besides imagining seeing.
Hearing is the most obvious case, but let us not forget the work
of path-breaking auteurs like John Waters and William Castle,
who gave us, respectively, Odorama and electrified seating.5 If
3
This thesis, or something like it, seems to be implicit in the work of early
theorists like Miinsterberg and Pudovkin.
4
"In cinema it is not possible to speak, in the strict sense, of a narrator. The
film does not narrate, but rather it places the spectator directly without in-
termediaries, in the presence of the facts narrated" (Julio Moreno, "Subjective
Cinema: The Problem of Film in the First Person", p. 354· Quoted in Branigan,
Narrative Comprehension and Film, p. 144).
5
As I noted earlier, viewers of Waters's Pink Flamingos are sometimes provided
Imagination
we can imagine that we see fictional things, presumably we can
imagine that we smell and feel them, that we have sensory con-
tact of all kinds with them. We can imagine other, nonsensuous
modes of contact with them: that we know, like, distrust or ad-
mire them. But I shall confine my remarks to vision: it is perhaps
the most interesting case because of the highly structured and
complex nature of visual experience, and it is naturally the case
most frequently discussed in connection with cinema. What I
say about the case of seeing will be easily applicable to the other
senses.
To repeat: I take it to be a distinctive thesis of classical film
theory that cinema encourages a certain kind of imagining
which I have called imagining seeing: imagining that you are
seeing the fictional events of the film, and seeing them from the
point of view of the camera.
6.2 CINEMA, THEATRE AND OTHER

VISUAL FICTIONS
What of theatre? You may argue that the theatre-goer imagines

that he sees Othello when he sees the actor playing Othello on
the stage. But there's a difference between the theatrical and the
cinematic case that any theory will need to accommodate. The
view I ascribed to writers like Panofsky and Mitry is that we
imagine not only that we see the characters in the film, but that
we do so by occupying a certain position within its "space" or
"world", and the position we occupy is that of the camera. What
makes that a plausible hypothesis is that the role of the camera
gives the movie what we might call an intrinsic point of view,
a point of view that shifts quite independently of the actual
position of the viewer, but which the viewer can imagine him-
self to occupy. A theatrical performance, obviously enough, has
no intrinsic point of view.6 Recognizing this, writers on theatre
with odour-producing scratch cards. Castle arranged for some of the theatres
showing his Tingler to give mild electric shocks to the audience at appropriate
moments during the program.
6
A theatrical performance may implicitly define a point of view by being
168
have tended to emphasize the distance - and the barriers - be-
tween the audience and the fiction presented on stage. Haig
Khatchadourian explicitly contrasts plays and movies in this re-
spect: "It is probably true that we cannot experience the action
of a play from 'within' its space time in the way we can in film." 7
Bernard Williams says that the theatrical audience "see what
they may well describe as, say, Othello in front of a certain pal-
ace in Venice.... But they are not themselves at any specifiable
distance from that palace". As members of the audience, they
"are spectators of a world they are not in".8 I assume Williams
means that spectators imagine they see the palace (they can't
really see a palace which isn't there), but they don't imagine
being any particular distance from it. They do not imagine that
they stand in any perspectival relation to the palace.
Later I shall address the question whether this idea of imag-
ined nonperspectival seeing makes sense. For the moment I
want to ask whether the idea of imagining seeing - perspectival
or not - could be the key to identifying a broad class of visual
fictions: a class which would include film, theatre, painting and
certain other forms. The proposal is this: a fiction is visual, or
belongs to a visual medium, if the kind of imagining it makes
appropriate is imagining seeing. Kendall Walton thinks that
painting is a visual medium in this sense. On his view, it is
central to our interpretation of paintings that we engage in var-
ious visual imaginings with respect to them: we recognize the
picture as a picture of a deer, he says, when we notice ourselves
imagining that we are seeing a deer when we look at the pic-
ture.9
Painting, theatre and film certainly are visual :media. The ques-
tion is whether this is because they encourage imagining seeing.
The idea is not implausible. It offers a neat contrast with another
staged to favour a certain viewing position. But this does not give it an in-
trinsic point of view in the strong sense that a movie has one.
7
Haig Khatchadourian, "Space and Time in Film", p. 176. Italics in original.
8
Bernard Williams, "Imagination and the Self", in Problems of the Self: Philo-
sophical Papers, 1956-72, p. 35·
9
Walton, Mimesis as Make-Believe, p. 294. See also Wollheim, Painting as an Art,
chapter 2.
Imagination
mode of fiction of which the novel is the best example, and a con-
sequent elaboration of the showing-telling distinction. With vi-
sual fictions, we have a showing: we imagine that our act of
looking at the film, stage or picture is an act of looking at the fic-
tional events there portrayed. With poetry and its literary succes-
sors we have not a showing but a telling: we imagine that the
events of the fiction occur, but not that our reading the text is an
act of witnessing those events. That way we have two kinds of
media, divided according to the kinds of representations they
employ, and two kinds of imagining; the proposal is that they
pair off neatly: pictorial fictions with imagining seeing, which is a
species of personal imagining, and (predominantly) linguistic fic-
tions with impersonal imagining. All this, I say, is false.
6.3 AGAINST IMAGINING SEEING
I don't say moviegoers never imagine themselves to be seeing

the characters and events of the fictions portrayed in the movie.
Perhaps there are times when viewers, or some of them, imagine
this, just as readers of novels sometimes imagine they see the
characters and events they read about. People can imagine just
about anything, and the viewer may occasionally imagine that
he sees the characters and events of the fiction. But that need
not be the kind of imagining made appropriate by the film.
What I shall argue is that this kind of imagining is not- except
in a few extraordinary cases- appropriate for, or required by,
the fictions that movies present. In a good many cases this kind
of imagining would be confusing because of its tendency to un-
dermine the structure of the fiction so presented. Further, we
don't need to invoke the idea of imagining seeing to explain
what is distinctive about cinematic fictions.
I said that imagining seeing is not appropriate cinematic
imagining - except in a few extraordinary cases. Here is one.'o
In Hitchcock's Vertigo there is a shot down a stairwell that em-
ploys a dolly-zoom combination to give the effect of a vertigi-
nous experience. To the extent that it is effective, I think this
•o David Hills reminded me of this one.

shot encourages us to imagine that we are having the vertiginous

experience. Is this a significant concession to the Imagined Ob-
server Hypothesis? I think not. The Imagined Observer Hypoth-
esis is a general thesis about the normal mode of cinematic
engagement; all I am conceding here is that it may be true of
some very exceptional shots. The thesis that we are all crazy is
not supported by the craziness of the few - on the contrary, the
salience of their craziness tells against their condition being the
norm. Things are like that with imagining seeing.
For me, perhaps the most striking thing about the view of
Balazs, Panofsky and the others is that it seems to misdescribe
the experience of movie watching. Do I really identify my visual
system, in imagination, with the camera, and imagine myself to
be placed where the camera is? Do I imagine myself on the
battlefield, mysteriously immune to the violence around me, ly-
ing next to the lovers, somehow invisible to them, viewing Earth
from deep space one minute, watching the dinner guests from
the ceiling the next? None of this corresponds to my own ex-
perience of movie watching. George Wilson, whom I have al-
ready quoted in support of the Imagined Observer Hypothesis,
is explicit about the sort of imaginative picture the viewer has
of herself in relation to the film viewed:
Thus, the film viewer's impression of his or her episternic sit-
uation is something such as this. The cinematic time machine
is a ghostly one in that it is both invisible and massless in
relation to anything it does not contain. The spectator sits in
the closed capsule before its large and magnifying window
and observes a slice of the secondary world in which the rna-
chine is situated. Directed by a prescient intelligence that op-
erates the machine for the passive viewer, the capsule has the
capacity to change its position and spatial orientation instan-
taneously so that its window opens onto a series of wider and
narrower prospects, which jointly delineate the action outside.
Being massless, the capsule can even occupy the position of a
person who is in the world outside and thereby present the
precise visual perspective of that person at that time."
" Wilson, Narration in Light, p. 55·

Imagination
Alone among writers on film, Wilson seems to have grasped

the implications of the Imagined Observer Hypothesis, and has,
in this quotation, carefully described what that thesis requires
us to suppose the viewer imagines. But in so carefully drawing
out its consequences, I believe Wilson has effectively sunk the
Imagined Observer Hypothesis, because there simply is no ev-
idence that film viewers imagine their situations to be as Wil-
son describes. And a viewer who did imagine all this would be
so mentally occupied as to be scarcely able to attend to the
film. Further, such imaginings would very often be prohibited
by the film itself. In most films, the fictional world described is
not supposed to be accessible by such magical modes of obser-
vation and inquiry; in most cinematic fictions the world of the
film is pretty much like the real world, where assuredly no
such magical access is possible. Such films implicitly prohibit
our imagining anyone watching the characters from within an
invisible, massless time machine capable of discontinuous
movement.
Many familiar cinematic devices also seem problematic
when taken in conjunction with the Imagined Observer Hy-
pothesis. Sometimes a long shot of two people in a crowd is ac-
companied by a very selective sound track, which highlights
their own conversation and filters out the irrelevant chatter of
other crowd members. One could never in fact see people from
such a distance and hear their conversation with such clarity.
By what means are we supposed to imagine ourselves having
this peculiar combination of visual and auditory access to the
events of the fiction? By magic, or by a kind of causality alien
to this world? Again, the film may be one which in other ways
seems to adopt this-worldly causality and to reject magical in-
terventions. We seem to be discovering hitherto unknown
sources of tension within familiar fictions! But the suspicion
must be that these tensions are not the product of the fictions
themselves but of a false theory about their mode of presenta-
tion.
If we are to imagine ourselves seeing fictional things and
events when we watch a film, we shall have to imagine that our
visual powers are strangely restricted, and that what we see

depends in no way on our own decisions. The camera is often
placed to restrict our view of the action for dramatic purposes,
and in these cases one would often like to see more or see dif-
ferently. But if we are imagining ourselves to be seeing the fic-
tion itself, what are we to imagine concerning the source of this
restriction? Putting aside Wilson's invisible-time-machine hy-
pothesis, the only candidate seems to be this: that we imagine
someone to be filming the action as a documentary, and that we
are seeing the visually restricted result. But to imagine this
12
(something I have in fact never been aware of imagining) would

be to imagine that the fiction contains as a part the assumption
that the action is being filmed by a camera crew and that we
are watching the result. Occasionally, as with Culloden (Peter
Weir, 1964) this would be an appropriate piece of imagining,
but it certainly would not be for most fiction films. 13
In other ways, the assumption of my presence within the fic-
tion would seriously distort the content or genre of the movie.
Suppose I am watching a movie in which the murderer enters
unseen. That the murderer enters unseen is part of the content
of the movie. Do I imagine that I see the unseen murderer?
There are two ways we can take the claim that I do, according
to how much scope we give to the expression "imagine". The
narrow-scope reading is: "There is exactly one thing which is
an unseen murderer and I imagine that I see that thing." There
is nothing intrinsically problematic about a claim of this kind. I
can imagine, of something that is seen, that it is not seen. But
the narrow-scope reading does not properly express my situa-
tion as a movie watcher. For one thing, it commits me to the
existence of an unseen murderer, which is false in the relevant
context since I do not believe there really is any murderer visible
on the screen. For another, it misrepresents the content of my
" Kendall Walton takes this view: when a fictional killing is shown on the
screen, it seems "to the viewer that he is seeing an actual killing via a pho-
tographic film of it" (''Transparent Pictures", p. 258).
3
' For further comments on the idea that the filming process arises from within
the diegesis itself, see Chapter 9, Section 9.2.
173
Imagination
imagining; it ought to be part of my imagining that the murderer
is, exactly, unseen. This reading gives "imagine" insufficient
scope. The other, wide, reading is: "I imagine that there is an
unseen murderer which I see." This does not have either of the
two deficiencies noted of the first formulation, but it does make
the content of my imagining explicitly contradictory. Perhaps it
is possible to perform the feat of imagining something explicitly
contradictory. But it is implausible to suppose that the audience
on such an occasion as I have just described is called on to do
that.
One other response to my scepticism about imagining seeing
is that while the fiction precludes the imagined presence of a
witness "in the world of the fiction", it does not preclude the
possibility that the viewer, who stands outside that world, is a
witness. 14 But most films which can be thought of as excluding
"internal" observers can, with equal or greater plausibility, be
thought of as excluding external ones as well. Naturalistic fic-
tions do not leave open as to whether someone is watching the
scene from another world: such an idea would be wildly at vari-
ance with their conventions.
The intrinsic point of view possessed by a movie raises an-
other difficulty. Occasionally that point of view is one we are
required to imagine occupied by a character: then we have a
subjective shot. In Hitchcock's Spellbound there is a shot, the in-
trinsic perspective of which is defined by the position occupied
by the character John Ballantine, as he raises a glass of milk to
his lips, eyeing the psychologist Dr Bruloff suspiciously as he
does so. How is the viewer to integrate his supposed imagina-
tive occupation of the camera's point of view with imagining
that this point of view is that of a character? Presumably, by
identifying himself with that character. I had better say some-
thing about the idea of identification.
For some writers, identification is the central concept of their
theory of cinema; it is what makes the understanding of cine-
matic narrative possible, and it is to be explained in psychoan-
•• See, e.g., Jerrold Levinson, "Seeing, Imaginarily, at the Movies".
174
alytic terms. 15 I wish to take a somewhat different path. I shall
concentrate on how identification might take its place in a cog-
nitive theory of imagining of the kind I have outlined. On this
view, identification with a character would be a matter of imag-
ining that you are that character. Consider, then, the hypothesis
that the viewer, when he views a subjective shot, imagines that
he is the character whose subjective view is represented there.
What is required in order to engage in that kind of imaginative
identification? At a minimum, the viewer must imagine that
what is (fictionally) happening to that character is happening to
him or her, and that he or she has the most obvious and dra-
matically salient attributes of that character at that time. But
once this is seen to be required as a condition for identification,
it becomes very hard to square the claim of identification with
the experience of watching a subjective shot. In the shot from
Spellbound, salient features of Ballantine are that he is drinking
milk and that his mental state is one of turmoil. Watching this
shot does not incline me in the least to imagine myself doing or
being either of these things. Nor, when watching a celebrated
subjective shot later in the film, in which the murderer points a
gun towards himself and fires, do I imagine that I am killing
myself. Further, identification, if it is a notion with any content
at all, would seem to require the one who identifies to have, or
to imagine having, some concern with and sympathy for the
values and projects of the one with whom she identifies. But, as
Nick Browne has noted in his study of an important scene from
Stagecoach, one of the peculiarities of the setup there is that,
while we see Dallas the prostitute from Lucy's point of view,
we do not, and are not intended to, feel solidarity with Lucy's
attempt to exclude Dallas. 16 That the camera occupies Lucy's
'' " ... [the viewer] certainly has to identify ... if he did not the film would
become incomprehensible." Christian Metz, Psychoanalysis and Cinema, p. 46.
6
' Nick Browne, The Rhetoric of Filmic Narration, chapter 1. So far as I can see,
Browne does not appreciate that identification is the source of the trouble.
He says that we are "asked to see Dallas through Lucy's eyes"; also that we
"identify" with Dallas. Things are made more confusing by Browne's moving
easily between point of view as a literal place of physical observation to point
175
Imagination
point of view has no implications, or only negative ones, for
identification. Nor do subjective shots from the point of view of
a stalking creature or a homicidal maniac make us identify with
the stalker; they typically have the effect of heightening our con-
cern for the potential victim. In all these ways the supposition
of imaginative identification does not square with the experience
of film.
Some of these objections can be made more vivid if we con-
sider an actual film sequence. I have chosen a short scene from
Hitchcock's The Birds, during which Melanie Daniels (Tipi Hed-
ron) crosses Bodega Bay in a hired boat. Compositionally, the
sequence is of some interest, but it does not seem to be an es-
pecially disorienting scene to watch. One can find the repertoire
of shots out of which it is constructed repeated in a great many
scenes in this and other films. The sequence I have in mind
consists of fourteen shots, beginning with three long shots of the
boat: one from behind as it leaves the quay, one from the side,
with the boat travelling across the screen left to right, and one
from more or less front-on. Shot 4 is close in to the boat, and
travels alongside. Shots 5 to 14 constitute a sequence of pairs of
shots, the first of each pair being a point-of-view shot from Me-
lanie's position, and the second a reverse shot to Melanie herself,
with the last shot reversed on her as she ties up the boat and
gets out.
As I say, there is nothing especially remarkable about the ex-
perience of watching this scene. But if the viewer is to imagine
herself seeing the events represented in the sequence from the
position of the camera, the experience of watching it would be
a very peculiar one indeed. The transitions between the first
of view as, in effect, "taking someone's part". This leads to unresolved ox-
ymorons like "identification asks us as spectators to be [sic] two places at
once, where the camera is and "with" the depicted person.... This passage
[from Stagecoach] shows that identification necessarily has a double structure
in the way it implicates the spectator in both the position of the one seeing
and the one seen" (p. 8). Simultaneous identification with distinct and indeed
antagonistic characters is hardly to be credited. But Browne's analysis is more
sophisticated than the inflexible "system of the suture" against which he is
to some extent reacting.
three shots would require her to imagine her position shifted

instantly through ninety degrees twice, around the edge of the
bay. Shot 4 requires her to imagine herself suddenly in the water
by the boat, somehow travelling alongside it at the same (con-
siderable) speed. The transitions between 5 and 14 would then
have her imagine herself shifting back and forth nine times be-
tween Melanie's own position (identifying with Melanie, per-
haps?) and different points on the shore, all within the space of
a minute or two. You have only to describe the sequence in these
terms to realize how clumsy and implausible is the Imagined
Observer Hypothesis: the effort of imagining all this during the
brief scene I have described would be both disorienting and
distracting, and quite unlike the normal experience of watching
the scene. Advocates of the Imagined Observer Hypothesis
never describe "the phenomenology of cinema" in quite these
crass terms. But this is certainly how it ought to be described if
we are to take that thesis literally.
One response to these criticisms would be to say that, while
we do imagine ourselves to be seeing the fictional characters and
events presented onscreen, we do not, and are not expected to,
incorporate into our imagining many of the consequences -
some of them rather awkward from the point of view of the
fiction concerned - of our seeing those things. The argument for
this is that we do not, in general, observe any injunction to imag-
ine all the consequences of what we do imagine (we don't, after
all, believe all the consequences of our beliefs). Indeed, we
couldn't imagine all the consequences of what we do imagine,
because everything nontautological has infinitely many conse-
quences.17
This response might be elaborated by saying that our imag-
ining seeing is a kind of purely visual imagining, unconnected
with any imaginings about where we are seeing from or how it
is that we are able to see. While we are to imagine ourselves
seeing, we are not to imagine ourselves placed anywhere in the
scene, or as undergoing any changes of position. This proposal
is already a step back from the classical view of cinematic imag-
7
' This objection I owe to Kendall Walton.
177
Imagination
ining as expressed by Balazs and Panofsky, which requires us

not merely to imagine that we are seeing the fictional events,
but also to imagine that we are placed where the camera is. It
is more like the view outlined by Bernard Williams when dis-
cussing the theatrical case: we imagine seeing the castle, but do
not imagine seeing it from any position.
Even though this position is weaker than the classical theory
I have been examining so far, it is still objectionable on two
grounds. First, the new proposal does not solve all the difficul-
ties I have described. For example, it does not solve the unseen-
murderer problem. There we are required to imagine that we
are seeing something and to imagine that no one is seeing it, so
in this case contradictory imagining cannot be avoided by being
selective about consequences. Second, I believe that the concept
of seeing and that of occupying a point of view are closer than
this weakening of the classical theory allows. The weakening is
based on the idea that the connection between the two concepts
is, at most, the relation of entailment; the proposition that I am
seeing something entails the proposition that I am placed
somewhere in relation to what I see. The defender of The Imag-
ined Observer Hypothesis simply refuses to have the entailment
be reflected in the act of imagining; I imagine something, but
not what it entails.
Against this I say that the concepts of seeing and of point of
view are linked more intimately than by entailment alone. To
see is to see from a point of view; there is no such thing as
nonperspectival seeing. You cannot imagine, of a certain scene
represented to you onscreen, that you are seeing it, but not that
you are seeing it from any point of view. To imagine seeing it
is to imagine seeing it from the point of view defined by the
perspectival structure of the picture. You can, of course, imagine
seeing it in such a way that your spatial relation to what you
see is rendered unclear- imagining that you see the scene with
the aid of a complex system of mirrors or perhaps even through
the medium of film. But it is not very plausible to suppose that
we imagine either of those things when we are confronted with
cinematic images.
178
One final response: you might be tempted to adopt a mixed

strategy, asserting the Imagined Observer Hypothesis for a re-
stricted domain of cinematic experience: those situations to
which it could apply without gratuitous paradox, or (and this
might be a more restricted thesis still) those where the style of
filming seems, in one way or other, especially to invite partici-
pation. But I'm sceptical of this ecumenical proposal. It seems
to me that there is an enormous psychological difference be-
tween merely imagining something happening, and imagining
yourself there watching it happen. Viewing a movie that con-
stantly took us from one of these kinds of imaginings to the
other and back again would be a roller-coaster experience of
shifting phenomenology. That's just not the sort of experience-
vertiginous shots aside - I have while watching a movie.
6.4 IMPERSONAL IMAGINING AND FILM
By now we have reason enough to abandon the Imagined Ob-

server Hypothesis - reason, at least, to look for a more plausible
alternative. The alternative I suggest is simple. What I imagine
while watching a movie concerns the events of the fiction it
presents, not any perceptual relations between myself and those
events. My imagining is not that I see the characters and the
events of the movie; it is simply that there are these characters
and that these events occur - the same sort of impersonal imag-
ining I engage in when I read a novel.
When we substitute impersonal imagining for imagining see-
ing, our problems start to dissolve. I see displayed on the screen
a man with a knife, and I imagine that there is a murderer,
perhaps even an unseen one. I do not imagine that I see this
unseen murderer. In this case the difference between imagining
seeing and impersonal imagining is captured as a difference of
scope: it's the difference between, respectively, "I imagine that
I see something which is a murderer," and "I see something
which I imagine is a murderer."
What does a subjective point-of-view shot, like the milk-
drinking one in Spellbound, ask us to imagine? We see the shot
179
Imagination
and recognize that its perspective is defined by a certain posi-

tion. We imagine, concerning that position, that it is occupied
by a character (in this case Ballantine). We imagine that he sees
Dr Bruloff as he drains the milk. 18 We imagine that, as he does
so, things look pretty much to him as they look to us on the
screen- or, more plausibly, that what we see onscreen is a rough
guide to what we should imagine his experience is like. That
way, the subjective content of the shot is referred to the char-
acter's experience, not to our own. We are not required to imag-
ine that we see Dr Bruloff from Ballantine's perspective or any
other. In general, subjective shots function to help us imagine
what a character's experience is like, not to imagine ourselves
being that character and having that experience. 19
In the same way, the problems of selective sound and re-
stricted view disappear once we abandon the Imagined Ob-
server Hypothesis. They just become effects of an inevitably
selective narration, and they appear no stranger in film than
they would in literature, which is, of course, selective at every
point in what it presents to the reader.
The proposal that cinema typically encourages impersonal
rather than personal imaginings is simple enough. What needs
to be shown is that it makes the right distinctions between cin-
ematic and other kinds of fictions, including distinctions which
concern our experience of cinematic and other fiction. But
something else needs explaining: if the Imagined Observer Hy-
pothesis is false, how is it that many people believe it? The al-
ternative I propose needs to explain the success of its rival. I
hope to make a start at showing that my alternative can do both
these things.
8
' This shot, like most shots of its kind, is quite unrealistic because it fails to
capture the binocular nature of ordinary vision. Ballantine would have to
have one eye shut to see anything like what is presented on the screen.
9
' The shot may encourage secondary imagining. That is, in order to imagine
appropriately concerning the character's experience, we may have to imagine
having that experience ourselves, thereby engaging in a simulation of the
character's experience (see Section 5.2 on the distinction between primary
and secondary imagining). But this secondary imagining is not a matter of
identifying ourselves with the character.
180
6.5 PERCEPTUAL IMAGINING
Section 6.2 raised the possibility of distinguishing between the

experience of fiction in a visual mode like film or theatre on the
one hand, and written literature on the other. The suggestion
was that the first kind of case involves imagining seeing and the
second, literary, kind does not. Given my rejection of the Imag-
ined Observer Hypothesis, that suggestion is not available to
me, and I must find another way to draw the distinction.
I might try this: the experience of cinema is an essentially
visual one in the sense that what it is appropriate for us to imag-
ine depends on what we see (on what we actually see, that is,
not on anything we imagine we see). Watching Citizen Kane it
is appropriate for me to imagine that a snow globe falls from
the hands of a dying man, because I really see onscreen a sick-
looking man from whose hand a snow globe (apparently) falls.
But will this suffice to put film in the class of visual media
without, unwantedly, including novels as well? I see the words
on the page, and thereby engage in a bit of appropriate imag-
ining. But we do not want to say that the novel is a visual me-
dium. There seems to be an obvious response: it is just a
contingent fact that we normally read with our eyes; novels can
be "read" by using Braille or Morse code symbols, so there is
nothing essentially visual about reading. The cinema, so the re-
sponse goes, is essentially visual, since there is no nonvisual
way to communicate a movie's content. A Braille version of Cit-
izen Kane would be a different work than the movie, in just the
way that the movie Great Expectations is a different work than
the novel. But this explanation really just begs the question: why
count the novel as a work in one medium (i.e., written language),
whether read by sight, Braille or Morse code? Why not count
the novel-as-read-by-sight a work in a visual medium and the
novel-as-read-by-Braille a work in a tactile medium, with a com-
parable category reserved for, say, Morse code? What then
would be the difference from the experiential point of view be-
tween film and the novel-as-read-by-sight? Surely there is a
difference, but the present proposal makes it hard to see what
it is.
181
Imagination
I start with a distinction, which I introduce through the case
of belief. Then I shall apply that distinction to imagination. In
light of the hypothesis that imagining is off-line simulation, that
seems appropriate; distinctions between kinds of imaginings
ought then to have their basis in distinctions between kinds of
beliefs.
What is the difference between seeing a thing and reading a
description of it, both considered as ways to reach beliefs about
its visible properties? In both cases we have belief caused by
perception, but only in the first case do we have what might be
called a genuinely perceptual belief. What makes the first a case
of perceptual belief? First, in the genuinely perceptual case we
have a certain kind of counterfactual dependence between belief
and visible properties of the object; if the thing seen had been a
little bit more red, or larger, or had more closely approximated
squareness, my belief would correspondingly have been that it
was more red, larger or more square. Second, perceptual be-
20
liefs tend to have certain characteristic features of structure and

content. They bunch together in so far as perception tends to
give us beliefs about colour, size and shape as an indissoluble
package with a high degree of specificity. When you see
someone's eyes you get beliefs about their shape as well as
about their colour, and the belief you get about their colour is
that they are exactly that shade of blue. And perception gives
us, as a free bonus, information about the temporal duration of
processes; they take as long as it takes us to watch them occue 1
This is all in contrast to reading a description. There, your

beliefs don't have that counterfactual dependence on the visible
properties of symbols; differences in the shapes of the letters
.., Recall the discussion of what is distinctive about pictorial representation in

Chapter 1.
, It was Frank Jackson, in conversation, who drew my attention to the impor-
tance for my argument of characterizing perceptual beliefs in this way. Jack-
son discusses some of the features I have mentioned in his Perception: A
Representative Theory, p. 4J, but in the context of arguing against those, like
George Pitcher, who would define perception itself as, or as the acquisition
of, a certain kind of belief. Note that Pitcher's definition of perceptual belief
in A Theory of Perception, p. C)O, is quite different from mine.
would not have induced in you different beliefs about the shape
of the thing described.= Nor, with descriptions, do your beliefs
have the kind of bunched, content-specific character we associ-
ate with perceptual beliefs. You can learn about shape from a
description, and learn nothing about colour. And what you can
learn from a description of colour will lack the specificity that
it would have in the case of a perceptual belief. Descriptions
rarely authorize us to believe the eyes of the person described
are exactly that shade of blue. The beliefs we get from descrip-
tions I'll call symbolic beliefs. They contrast, obviously, with per-
ceptual beliefs, but I am not claiming that between them these
two kinds exhaust the field.
This characterization of the difference between perceptual
and symbolic beliefs applies at the level of tokens rather than at
that of types. One cannot, in general, say that a certain belief
type - for example, that this object is round - is perceptual
rather than symbolic. One token of that belief might be percep-
tual and another symbolic. What makes the token perceptual is
the form of its tokening - its tokening within a system of rep-
resentation that gives to the tokens so represented the features
of structure and content I just mentioned. It should, for instance,
be a system in which size, shape and colour come together as a
single unit of representation. Pictures in the head would serve
well in this regard, but I assume we have independent reason
not to believe in them. No matter; other things may do as well,
including connectionist networks with the right kinds of exci-
tatory links. But for purposes of this argument, I am neutral
about what kinds of representations are at work here.
Note that it would not do to characterize perceptual beliefs
by saying that they are the ones you get when you see the thing
itself rather than a description of it. For you can get genuinely
perceptual beliefs, in my sense, by looking at likenesses of
things; likenesses stand in the same or similar relations of coun-
terfactual dependence with your beliefs as do the things of
which they are likenesses, and the beliefs they induce have the
.. Except in so far as a change of shape would constitute a change in the letter's

identity.
Imagination
characteristics of content and structure I just noted. This has
some importance for what follows.
Imagining, on my hypothesis, involves running your belief-
desire system off-line. In that case, the distinction between per-
ceptual and symbolic beliefs ought to carry over into a distinc-
tion between perceptual and symbolic imaginings. And so it
does. Perceptual imagining exhibits the characteristics of coun-
terfactual dependence and structure characteristic of perceptual
belief; when we see the screen we imagine that the character's
eyes are exactly that shape, that colour and that size in relation
to the rest of the character's features. If what we saw on the
screen were shaped or coloured in a slightly different way, what
we would then have imagined about the character's features
would have been correspondingly different. If we had been
reading a novel we might, at a certain point, have read
something that prompted us to image that the character's eyes
were blue. We would be in no position to imagine anything
about their shape, because we are told nothing about shape; nor
would we be in a position to imagine that the eyes are some
specific shade of blue. It would be a matter of indifference to
our imaginings, moreover, whether the text was composed in
this type face (or size, or colour) or that one.
So what makes the experience of cinema, painting and the
other pictorial media an essentially visual one is that it gives
rise to perceptual imaginings. Poetry and the novel, on the other
hand, give rise to symbolic imaginings, whether the poem is
read by sight, Braille or Morse code.
The relation of counterfactual dependence I am pointing to
here does not hold between your imaginings about the charac-
ters of the fiction and the visual properties of the characters
themselves: recall that fictional characters don't exist, so we
certainly don't see them in cinemas or anywhere else. The coun-
terfactual dependence holds between your imaginings and the
visible properties of representations. Does this undermine the par-
allel I have claimed exists between perceptual beliefs and per-
ceptual imaginings? There certainly is a difference between
belief and imagining: our beliefs about Fred, if he is a real per-
son, often derive from perceptual encounters with Fred himself.
184
But if he is a fictional character, our imaginings about him can-
not derive from seeing him. At most they derive, as with film,
from seeing representations of him. But beliefs can be derived
that way also: I may get a belief about Fred's hair colour from
seeing a photograph or painting of him, and the belief I get is
a perceptual belief.
Now we can understand why we so commonly make the mis-
take of saying that we see, or that we imagine that we see, the
characters and events in the film. The imaginings that movies
give rise to are based on the experience of seeing things, or
pictorial representations of them; they are structured very much
as are the beliefs we get from seeing things. It's natural, then,
to say that we see, or imagine ourselves to see, the fictional
events presented on film. But the imaginings that novel-reading
licenses are very unlike what results from seeing things or their
likenesses, and there's little temptation to describe them in terms
of seeing.
I have made two distinctions: one between personal and im-
personal imagining, and the other between perceptual and sym-
bolic imagining. We can now identify the basic confusion which
leads people to assert the Imagined Observer Hypothesis. Sup-
posing, rightly, that there is something distinctly visual about
cinematic fictions, people then suppose that this peculiarly vi-
sual element consists in the fact that cinematic fictions encour-
age us to imagine that we see fictional things. In fact, what is
peculiarly visual about cinema is that it encourages perceptual
imagining. Our two distinctions cut across each other, and while
some perceptual imaginings are personal imaginings, some are
not. In general, film encourages perceptual but impersonal im-
aginings.
6.6 CLARIFICATIONS AND REBUTTALS

In this section I want to make two clarifications to my thesis and
to confront five objections to it.
In making a case for the perceptual nature of cinematic make-
believe, I emphasized the correspondence between the visual
display on the screen, and what we imagine as a result. It's time
185
Imagination
to recognize that the correspondence is patchy. The information

given in perception needs to be interpreted, an issue I shall take
up more fully in Chapter 8. Sometimes the viewer blocks the
inference from perception to imagination by tagging the percep-
tual information, or some part of it, as extradiegetic: when a shot
fades out in the movie we don't imagine that objects slowly and
mysteriously disappear; with slow motion, we don't imagine
that physical processes are taking longer. Sometimes we are re-
quired to extend, revise or discount information from vision so
as to arrive at something that meets overall constraints of co-
herence. We try to construct a narrative, the events of which do
not always correspond directly to what is seen, as with episodes
of fabrication, dreaming and hallucinatory experience. And if
the film presents the hero in Chicago one day and in London
the next, we are usually required to imagine that he travelled
by plane, though no journey by plane was shown. Further, when
what we see onscreen warrants a bit of perceptual imagining, it
rarely warrants only perceptual imagining; our imagining is usu-
ally to be supplemented by, and interpreted in the light of, im-
aginings of a nonperceptual kind. When we see a man onscreen
we are usually supposed to have a perceptual imagining to the
effect that there is, exactly, a man with that appearance doing
such and such. But in Invasion of the Body Snatchers we are to
imagine instead that there is a terrifyingly emotionless alien
creature who has taken on the form of a man. While there is
certainly a difference between these two imaginings, the differ-
ence is not displayed at the level of perceptual imagining. Visual
fictions are those that make it appropriate for us to engage in
some visual imagining; beyond that, visualness is a matter of
degree. So the dependence of imagining on what is seen at the
movies is partial, and the imaginings that movies make appro-
priate are not all perceptual imaginings. I claim only that there
is enough perceptual structuring there to make plausible my ex-
planation of why movies are - and novels are not - works in a
visual medium.
A second clarification. I've argued that engagement with cin-
ematic fiction does not require or license us to imagine that we
see the fictional events and characters it presents. But there are
t86
occasions, I believe, when the film requires us to imagine that

we see something. Let me illustrate how this can arise.
Movies, like works in other media, have styles; in particular,
a movie can have a visual style. Sometimes the viewer, if he is
discerning enough, will recognize that the visual style of the
movie allows or encourages a kind of make-believe. Andre Ba-
zin argued that Renoir's use of irregular panning created an
atmosphere of spontaneity which pervaded his films. 23 Now the
competent viewer will not suppose that Renoir's irregular pan-
ning was genuinely the result of spontaneous composition. He
will know that Renoir's shots were as carefully planned and
unspontaneous as anyone else's. The spontaneity of Renoir's
shots is make-believe; the way those shots are composed makes
it appropriate for the viewer to imagine that those shots were
spontaneous. Does that mean that the viewer imagines himself
to be seeing spontaneously composed shots? I think it does. This
stylistic feature, like other features of visual style, works on us
in the way it does because we imagine ourselves to be seeing
something done in a certain way. The viewer imagines himself
to be seeing the result of spontaneous composition, but that
make-believe isn't part of the make-believe appropriate to the
fictional stories that Renoir's films portray. It's not part of the
story of a film like Rules of the Game that there is a cameraman
following the characters around and recording, in a spontaneous
way, their actions. Watching a movie involves different levels
of imagining, not all of them concerned with the fictional story
the movie presents. My scepticism about the claim that we imag-
ine ourselves to see things in the movies is confined to the base
level of make-believe, that concerning the characters and events
of the fictional story.
Now the objections:
1. Our thought, and sometimes our talk, sometimes seems
best explained on the Imagined Observer Hypothesis. ''Who is
the murderer?" I whisper to my partner at the movie. ''That
one" (said while pointing at the screen). Only an imagining that
23
See Andre Bazin, Jean Renoir; see also Noel Carroll, Philosophical Problems of
Classical Film Theory, chapter J.
Imagination
we are right there could underlie such an explanation. I answer:
we are notorious transferrers of properties from things repre-
sented to representations. I'm told that oscilloscope watchers say
things like "That's a loud one" (pointing at the wave on the
screen). They mean, "That represents a loud noise." That plau-
sibly is what is going on in our imagined conversation at the
cinema.
2. Consider the perceptual imaginings that cinema promotes.
These are in a certain sense perspectival imaginings. I have said
that with perceptual imagining, what we are to imagine de-
pends on what we see; we see a cinematic image of a man with
a certain appearance, and we are to imagine that the character
played by the man has that appearance. But what we see is seen
from a certain perspective: that of the camera. If we imagine
that the character looks like that, we must be imagining him
looking like that from that perspective. There is no such thing as
the way a person looks from no perspective. That is the sense
in which perceptual imagining is essentially perspectival: a spec-
ification of the content of what we imagine must make reference
to a certain perspective. But once we admit that we are dealing
with perceptual imagining that is perspectival, doesn't this
amount to an admission that we are dealing with imagining, the
content of which is of the form "I am seeing such and such"?24
In order to answer the objection, let us once again compare
the case of imagination with that of belief. The general form of
the argument just given is this: if the content of an attitude is
perspectival, then the content of the attitude is of the form "I cp
that I am seeing (or, more generally, perceiving) that P", where
cp is a variable ranging over attitudes, and P a variable ranging
over propositions. But in the case of belief, this is manifestly not
so. I see a painting of Uncle Albert and on the basis of seeing it
form the belief that his appearance is thus and so.25 Of course
what I believe is that he appears thus and so from a certain
perspective. But I do not believe that I am seeing Uncle Albert,
24
Again, this objection was suggested to me, in conversation, by Kendall
Walton.
•s I assume that the pictures in question are nonphotographic, thereby avoiding
an argument with the advocates of Transparency.
188
since I am fully aware that I am seeing a painting. So the ar-

gument is wrong as applied to belief; we can assume that it is
wrong in the imagining case as well.
3· It may appear that my strategy fails to account for the im-
mediacy of cinema, which is capable of generating feelings of
intense concern on our part for the characters. Consider that
paradigm of movie suspense, the scene in North by Northwest
where Roger Thornhill and Eve Kendall (Cary Grant and Eva
Marie Saint) cling to the face of Mount Rushmore, trapped by
their pursuers. Isn't it because I imagine that I am actually watch-
ing them that I respond so intensely to their plight? A compar-
ison with novel reading is instructive here. For many of us, the
feelings generated by reading are as powerful as those generated
by a play or a movie, but these feelings do not arise out of our
imagining that we are present at the action described in the
novel. Our make-believe is simply that the characters act and
suffer as the story says they do. Literary fictions engage our
feelings effectively without placing us imaginatively in direct
perceptual relation to the events they depict. So intensity of feel-
ing is no argument for the Imagined Observer Hypothesis in the
case of cinema.
But let us not go too far. Let us not insist that novels and
movies have equivalent capacities to generate feeling and con-
cern. I doubt if any novelist could give me quite the experiences
Hitchcock gave me in that scene on Mount Rushmore. But the
difference in capacities between the two media can be explained
without supposing that we imagine ourselves to be seeing the
characters in the movie. The difference is due to the structural
differences between film and the novel I have described. The
perceptual structuring of cinematic make-believe provides us
with a wealth of easily assimilated detail. Simply by seeing Cary
Grant's hands as he clings to the rock we are put in a position
to imagine that the character Thornhill's hands are holding the
rock in just that way. A verbal account could hardly provide us
with precisely that information, and extended descriptions of
visual detail make it difficult to sustain imaginative involve-
ment. If visual fictions more easily engage us when it comes to
the presentation of, say, certain kinds of physical danger, that's
189
Imagination
because of the perceptual structuring of their contents, not be-
cause their content involves a perceptual relation between the
action and the audience.
4· A central and highly influential tradition of film making
has been based on the premise that shooting and editing should
produce a result that mimics the experience of an ideally placed
observer, and thereby have the audience "see as the attentive
observer saw". 26 The devices to which this tradition has given
rise, such as eyeline matching and shot-reverse-shot editing, cer-
tainly seem very natural. They constitute part of the repertoire
of what is called "invisible editing", the kind we easily cease to
notice. Why do they seem so natural? If we assume that the
spectator's role in cinema is to imagine himself standing before
the characters, where the camera is, their naturalness is easily
explained: they serve to promote that imagining better than any
other choice of orientation would. But if we reject the Imagined
Observer Hypothesis, then the set of orientations favoured by
invisible editing is just one of an indefinitely large class of ori-
entations that would do equally well. Strange, then, that the
film-making tradition has happened to fix on just this class, the
class which the Imagined Observer Hypothesis selects as privi-
leged!
Part of an answer to the objection questions its presupposition
that standard cinematic orientations mimic perception. Quite a
lot of them don't. In order to have a visual experience of a con-
versation between two people like the experience of watching
shot-reverse-shot editing we would continually have to be
running between the positions defined by the two cameras. Cut-
ting from long shot to close-up just is not - contra Reisz and
Millar - like the process of noticing and focusing on detail. 27
And trains that disappear out of the left visual field rarely reap-
pear from the right. All this suggests that what we find natural
in cinematic editing is not what we would find natural in visual
experience of the real world. There does seem to be something
to the objection, however: Shots from the floor or ceiling strike
•• Pudovkin, quoted in David Bordwell, Narration in the Fiction Film, p. 9·

'
7
See Karel Reisz and Gavin Millar, The Technique of Film Editing.
us as less natural than those taken at roughly eye level. But the
best explanation for this sense of unnaturalness might not be
that we find it unnatural to imagine ourselves in the position
from which the shot is taken. A plausible alternative explanation
is simply that we are unfamiliar with the views of people that
such shots present, together with the fact that facial expressions
are usually so important to the narrative that shots which wil-
fully deny us access to them are very noticeable interventions
on the maker's part. And if unnaturalness were a function of
the extent to which the camera's position deviated from the
norm for a human participant, some shots ought to seem very
unnatural which do not seem unnatural at all. Those views of
the Earth from deep space that we see so often in science-fiction
films do not seem especially unnatural.
5· What, in that case, do we make of the venerable distinction
between diegesis and mimesis, between telling and showing? It
seems that on my view fictions of all kinds are diegetic; all are
experienced as something told. They differ only in the means
used for the telling: words in literature, moving images in cin-
ema. This denial of a universally acknowledged distinction
looks like a problem for my theory.
I reply that my theory still allows for the making of important
distinctions. If all fictions are tellings, there are still fictions
which tell by words, spoken and written, and there are fictions
which tell by showing, as cinema and other visual fictions do.
Also, there are fictions in which the telling is foregrounded, and
is therefore a salient part of the audience's unreflective experi-
ence of the fiction, and fictions in which the artifice of telling is
backgrounded, and easily forgotten. This pair of distinctions will
serve us at least as well as the traditional showing/ telling dis-
tinction.
6.7 THE MYTH OF TOTAL CINEMA
Andre Bazin argued that the history of cinema is the history of

the progressive embodiment of a myth or, better, an ideal, the
possibility of "an integral realism, a re-creation of the world in
191
Imagination
its own image". 28 That way, sound and colour constitute pro-
gress towards the cinematic ideal rather than, as some critics
have claimed, degeneration from the purity of the original, silent
screen.
No doubt Bazin was right about sound and colour being pro-
gressive additions to the cinematic repertoire. But if what I have
been arguing for in this chapter is correct, we shouldn't base
this claim on the idea that progress in cinema is measured by
increments of realism. Realism, in Bazin's sense, is the fashion-
ing of more and more "lifelike" re-creations of the world as we
experience it through our senses. And the farther we go in this
direction the more and more tempting it is for us to participate
imaginatively in the fiction the medium presents; to imagine
that we really are there, seeing fictional things. I have argued
that we do not typically imagine any such thing when we see
films, and that it is just as well that we don't, since the attempt
to integrate our imagined presence within the film world with
the events of the story would produce unnaturalness and even
paradox. But we can imagine a cinematic technology so lifelike
that it would become difficult for us not to imagine ourselves
seeing fictional things from within the fictional world; even the
crudest 3-D techniques, for example, have a startling tendency
to make us imagine ourselves in physical relations to events
onscreen. That threatens to undermine the interest of the die-
gesis, by leaving the viewer absorbed in an attempt to orient
himself in an unfamiliar environment. Such technological ad-
vances as 3-D might be accretions of realism in Bazin's sense/9
but in my view they would not be progress in cinema. Any
medium that wishes to create rich and rewarding fictions must
keep the observer at a distance. Realism and the distancing nec-
essary to discourage imagining seeing are two principles of cin-
ema which are in tension. We might learn a great deal about the
28
Andre Bazin, 'The Myth of Total Cinema", p. 24.
29
The failure of 3-D may, of course, have been a consequence of the failure to
implement it in a technically adequate way. But it is significant that strenuous
efforts were not made within the industry to improve the effect; as ideolog-
ical critics of the cinema so often tell us, what look like purely technical
constraints often have deeper nontechnical explanations.
aesthetics of cinema by examining the various creative attempts
there have been to resolve, or to exploit, the tension. But we
shall not learn about it here~ I am not attempting an aesthetics
of film.
6.8 PSYCHOLOGISM
I have argued that the Imagined Observer Hypothesis is wrong.

But that thesis is just one example of a fundamentally mistaken
way of thinking about the cinema. The most general formulation
of the mistake might be that the content of the cinematic image
is to be interpreted as the content of someone's visual experience.
Since the distinctively cinematic mode of presentation is the im-
age, this is, therefore, an assumption by which the cinema as a
medium is effectively psychologized - made to be a medium
which can be understood, in respect of its content as well as its
effect, only in psychological terms.
This psychologism is rather unspecific; it does not give us any
indication about whose visual experience the image is supposed
to represent. You may say that the image represents the viewer's
(imagined) visual experience; that the viewer imagines herself
to be viewing the cinematic events from within the space of the
movie itself, imagining herself situated where the camera is. On
that construal we have what I called the Imagined Observer
Hypothesis. But instead of psychologizing the image by way of
the viewer, some theorists prefer to go by way of one of the
characters within the fiction itself. It is at this point that the idea
of the point-of-view shot (POV) becomes an important technical
resource for the theory. On this view, each shot is to be inter-
preted, as far as possible, as a POV shot: a shot which represents
what some character within the film's diegetic structure sees.
Obviously it is going to be extremely difficult to find a plausible
point-of-view interpretation for a great many shots, and one
rather predictable strategy employed to overcome the difficulty
is an exercise in progressive concept stretching; the concept of
point of view is applied to cases where its acceptability is mar-
ginal, and then to cases where its application can really only be
metaphorical, but concerning which claims are made that de-
193
Imagination
pend for their truth on the literal applicability of the concept. In
that way, one gets the impression of a doctrine stoutly defended.
A good example of this technique is given by Edward Branigan,
who begins by asserting that the "glance" which defines the
point of view psychologically must be that of a sentient ob-
server, but not necessarily a human one; he then goes on to
include as examples of POV shots one from the perspective of
a dead man and one from that of a statue, though he calls these
and other examples "metaphorical" POV shots. 30 Here the no-
tion of point of view is being asked to do more work than can
reasonably be expected of it.
Another symptom of the psychologizing of the cinema,
closely associated with the concept of point of view and its
abuse, is the so-called system of the suture, described by one of
its advocates, Dayan, as "to classical cinema what verbal lan-
guage is to literature".31 According to Dayan's interpretation of
this model, the typical pattern of cinematic presentation is a pair
of shots, the first of which raises in the viewer's mind the ques-
tion: what is the source of this image?, and the second of which
gives the answer (a lying and ideologically offensive answer,
according to the advocates of this model) by showing the char-
acter whose point of view this was.
Note first that, if the model were correct, the second shot
would not, after all, produce any psychological equilibrium in
the viewer, for it would simply raise the same question as the
first: whose experience is the basis for this image? The answer
to this second question could not be the same as the answer to
the first, because the source of the image of the subject could
not, except in exceptional circumstances, be the subject herself. 32
But the more important observation about the system of the su-
ture is that it presupposes, without argument, that there is a
strong tendency on the viewer's part to inquire into the psychi-
cal ownership of any given shot: a claim for which I have not
seen any convincing evidence and which only a general pre-
30
Edward Branigan, "The Point-of-View Shot".
3
' Dayan, "Tutor-Code of Oassical Cinema", p. 439·
32
An exceptional circumstance might be where the second shot looks directly
into a mirror.
194
sumption in favour of psychologism would make plausible.
Take away that presumption and the motivation for suture the-
ory collapses.
These and other artificial and sometimes desperate attempts
to psychologize the cinema are somewhat ironic in light of what
should be a rather obvious fact about the basic cinema appara-
tus: that it is an impoverished and unconvincing vehicle for
subjectivity. Viewing the cinema screen is, I have argued, in im-
portant ways rather like viewing the real world; it is not at all
like viewing someone's subjective visual experience of the real
world (a notion that barely makes sense anyway). It is extremely
difficult to make what appears onscreen correspond in a con-
vincing way to the content of anyone's visual experience; the
sharply defined boundary of the screen is never effectively over-
come by blurring effects, for they never correspond to our own
graduated field of vision. The screen cannot mimic the capacity
of vision to concentrate on detail, and selective focusing by the
camera simply draws attention to the artifice. Nor does deep
focus correspond to the visual field. For that reason, episodes of
recollection which appear onscreen as visual images rather than
as verbal accounts have a tendency to become, unintentionally,
episodes of objective recounting. For what appears on the screen
always surpasses in detail and clarity the possible content of
anyone's memory. Thus it was that Hitchcock was able to con-
fuse the audience in Stage Fright by presenting, in visual images,
events as recounted from the character's point of view; these
events, as it turns out, did not occur but were the fabrication of
the character. Why do we take visual images as more authori-
tative than verbal recountings, which in the context of a mystery
are automatically treated with some scepticism? The answer
seems to be that the camera's capacity for capturing detail, and
its failure to resemble subjective experience, means that it is only
for short periods and with considerable effort that we are able
or willing to regard what is onscreen as subjective and therefore
as representing what someone thinks or claims, rather than what
actually is.
The oddness of psychologizing the cinema is evident when
we compare it to other media. I should be surprised to learn,
1 95
Imagination
for example, that anyone had suggested, with respect to the
presentation of fictions on radio, that we are required to under-
stand each sound as, or as referred to, the content of some char-
acter's auditory experience. Perhaps the reason this has not been
suggested is that there is nothing in radio comparable to editing
in film, and that it is the editing of shots that enables the viewer
to understand the images the screen presents as referred now to
this character and now to that. (A sufficiently stereophonic radio
might do this, but to my knowledge this suggestion has not been
taken up.) But that psychologism has not prevailed with respect
to radio is instructive, because it does seem that listeners to ra-
dio drama are able to orient themselves and to understand the
narrative without psychologizing the auditory information they
receive. Why should it be otherwise with the cinema?
It is time, I suggest, to tum the psychologizing paradigm on
its head, that is, to take as the default setting an interpretation
of the cinematic image as the rendering of events objectively,
and to allow a subjective interpretation only when no plausible
objective interpretation is available.
6.9 ICONIC SIGNS
I sum up. At the movies, we do not see, nor do we imagine that

we see, fictional characters or events. Rather, we see signs: pic-
torial or "iconic" signs which tell us what it is appropriate to
imagine. I say these signs are iconic; they are not linguistic signs,
or even signs that bear interesting similarities to linguistic ones,
and talk of a language of cinema is very misleading talk, as I
argued in Chapter 4· Still, movies and novels have more in com-
mon than we might otherwise have thought. As long as we sup-
pose that movies require the spectator to imagine himself seeing
fictional things, cinema seems to present fictions of quite a dif-
ferent kind - and present them in different ways - from the
fictions of literature. Novels do not give us the fictional worlds
they create; they describe them for us, they mediate between the
fiction and the reader. That is why we speak of literary narra-
tion. My argument has been that movies are like that too. They
are narrations carried on by other means - by iconic signs. In
196
Part III we look more closely at cinematic narration and narra-
tive.
There are consequences here that go beyond how we should
think about cinema as a fictional medium. One is the idea that
we need to rethink the basis of some distinctions within the
system of the representational arts and outside it. The showing-
telling distinction needs to give way to a distinction between
telling by showing and telling by other means, with a conse-
quent rethinking of the ancient distinction between mimesis and
diegesis. And we have new grounds for thinking of certain me-
dia as belonging to something like a natural kind: painting,
sculpture and theatre as well as cinema present visual fictions
because they promote perceptual imagining. They do so in var-
ious ways: by means of painted surfaces, shaped objects, people
distributed on a stage and cinematic images. But these differ-
ences among them are less significant than the differences be-
tween all of them and those media like the novel which promote
symbolic imagining.
197
Chapter 7
Travels in narrative time
I had the impression [in Je t'aime, Je t'aime] of a sort
of eternal present. The hero relives his past, but when
he relives it we are with him, the film always takes
place in the present. There are absolutely no flash-
backs or anything like them.
Alain Resnais'
In Chapter 3 I discussed the representation of temporal prop-

erties and relations in film, deciding that film is a distinctively
temporal art in that temporal properties are used to represent
temporal properties. I want now to connect some issues in the
filmic representation of time with the theory of imagining out-
lined in the previous two chapters.
7. 1 TENSE IN FILM
One class of philosophically interesting and controversial tem-
poral properties I did not discuss in Chapter 3 is that of tense
properties. This consists of the properties of being past, present
and future. These are properties of events, or so I shall assume
for simplicity's sake. 2 My opening the door is future at one time,
present at a later time and past later still. I have argued that
cinematic representations - moving images - have temporal
properties, and that their temporal properties typically function
to represent temporal properties of the events those images rep-
resent. Can we say tensed properties are among those which
function representationally in this way?
For that to be the case, cinematic images would have to have
tense properties, and, in virtue of possessing those tense prop-
' Quoted in Turim, Flashbacks in Film, p. 220.

, This is in fact controversial. See, e.g., Hugh Mellor's "Unreality of Tense".
1'}8
erties, would have to represent fictional events as having tense

properties. They would have to represent them, that is, as being
past, present or future. So first we need to see whether cinematic
images have tense properties. A cinematic image, or rather the
showing of a cinematic image at a particular place at a particular
time, is an event, and the image shown is present when it is
shown or projected. So anyone witnessing an image can rightly
say, as he or she watches it, that it is now present, was future
and will be past. In that sense, cinematic images have tense
properties. Does the presentness of that image, on the occasion
of its showing, function to represent the presentness of the
events that the image represents? For that to be the case, the
following would have to hold: that the events of cinematic fic-
tions be typically imagined to be occurring presently for the
viewer who would have to imagine those events happening now.
I shall argue that this is not a feature of cinematic fictions, in
which case the tensed properties of cinematic images cannot be
thought of as representing anything. This does not abolish time
in cinema, for there are temporal relations other than tensed
ones, and nontensed temporal relations between cinematic im-
ages can and do indicate nontensed temporal relations between
fictional events. But the dispensability of tense will require us
to rethink our assumptions about what is sometimes called an-
achrony in cinema: the reordering of story time by narrative, of
which the flashback is the most common example. 3
There is a potential confusion here that we have already en-
countered; the question of tense is raised in connection with the
way cinematic images function as representations of the fictional.
As we have seen, these images function also as representations
of the real in virtue of their photographic aetiology. If we con-
sider tense in respect to this second function, we can agree that
these images do not represent the events they photographically
record- actors performing on sets and locations- as happening
now. Perhaps they represent them as past. But that is not rele-
3
''Anachrony" is Gerard Genette's term. See his influential Narrative Discourse.
Genette's taxonomy is widely applied to cinema: see, e.g., Brian Henderson,
"Tense, Mood and Voice in Film (Notes after Genette)".
199
Imagination
vant to the question before us, which concerns only the rep-
resentation of the fiction itself. The question is whether the
presentness of the image represents the presentness, to the
viewer, of fictional events. There is nothing eccentric about this
decision to concentrate attention on the fictional: discussions of
tense in cinema normally focus on the relation between thenar-
rative order and the temporal order of the fictional events de-
picted. We do not count it as anachrony when shot Y, which
follows shot X in the intended order of viewing, was filmed
before shot X. Rather, we count it as anachrony when shot Y
represents a fictional event whose fictional time of occurrence is
prior to that of X.
Many film theorists have argued that cinematic images have

tense, by which they mean not merely that those images have
the property of being present as they occur (which they do) but
that those images represent the fictional events themselves as
present. Of cinematic and other images, the film theorist Bela
Balazs wrote, "They show only the present - they cannot ex-
press either a past or a future tense." 4 This view is held by
practitioners as well as theoreticians. Alain Robbe-Grillet says
that "on the screen verbs are always in the present tense ... by
its very nature what we see on the screen is in the act of hap-
pening, we are given the gesture itself, not an account of it." 5
Let us call this the Claim of Presentness.
• Balazs, Theory of the Film, p. 120. See also, e.g., Jurij Lotman, Semiotics of the
Cinema, p. 77: "In every art which employs vision and iconic signs there is
only one possible artistic time - the present." Balazs seems to conflate the
view that images have only one tense - the present - with the view that they
''have no tenses" (ibid.). See also R. Stephenson and J. R. Debrix, The Cinema
as Art, p. 115: "Film has no tenses- past, present, or future. When we watch
a film, it is just something that is happening- noul' (emphasis in the original).
Some other advocates of the Claim of Presentness are discussed and criticized
by Jarvie in Philosophy of the Film, pp. 12-19.
5
Introduction to the screenplay of LAst Year at Marienbad, p. 12; quoted in Joan
Dagle, ''Narrative Discourse in Film and Fiction: The Question of the Present
Tense". Similar views are occasionally expressed concerning literature: 'The
reader if he is engrossed in his reading translates all that happens from this
moment of [fictional] time onward into an imaginative present of his own
200
The formulations I have just quoted are somewhat problem-

atic in that they seem to presuppose the reality of what is
represented ("what we see on the screen is in the act of hap-
pening"), which is false in respect to the fictional. Our earlier
discussion suggests a better version: the cinematic image rep-
resents something - an embrace, a bank robbery - which does
not in fact happen. To say that the image is tensed is to say that
it represents that event as happening now. 6 Since what it rep-
resents is intended as fiction, the viewer is to imagine what is
represented. And so the viewer is to imagine not merely the
happening, in some abstract sense, of that bank robbery, but its
happening now.?
The Claim of Presentness is a consequence of the Imagined
Observer Hypothesis. If the viewer imagines that she is in the
world of the fiction, watching the events of the narrative, she
presumably imagines that they are happening as she is watching
them (modulo corrections for the finite speed of light!), that they
are happening now. So this discussion, directed against the
Claim of Presentness, will reinforce the criticisms of the Imag-
ined Observer Hypothesis that were brought forward in Chap-
ter 6. 8
A problem for the Claim of Presentness is provided by the
phenomenon of anachrony: the existence of flashbacks and, less
and yields to the illusion that he is himself participating in the action or

situation" (A. A. Mendilow, Time and the Novel, pp. 96--97); also see Meir
Sternberg, Expositional Modes and Temporal Ordering in Fiction, pp. 21-22, for
criticism.
6
Gilles Deleuze says, "It is not quite right to say that the cimematographic
image is in the present. What is in the present is what the image 'represents' ''
(Cinema 2: The Time-Image, p. xii).
7
The issue of film tense is sometimes confused by those who hold that the
viewer typically comes to believe in the fiction presented by the movie. Thus
Victor Perkins holds that the tense of film images is not, strictly speaking, the
present, because if it were, "cinema managers would have always to protect
their screens against assault by gallant spectators rushing to the aid of em-
battled heroines" (G. Mast and M. Cohen, Film Theory and Criticism, 3d ed.,
p. 48).
8
See also Francis Sparshott, "Basic Film Aesthetics"; and Alexander Sesonske,
"Time and Tense in Cinema".
201
Imagination
commonly, flashforwards, in which the image is to be taken to
represent a past or future state of affairs. It is said, by advocates
of the claim, that in these cases the assumption of presentness
is overridden by contextual cues (dissolve, verbal narration or
general considerations of narrative coherence). The default set-
ting is presentness.
In that case cinematic images have two distinct functions for
the audience, depending on whether they present fictional
events anachronously. Where there is no anachrony, we are to
interpret the visual images onscreen by imagining that we ac-
tually see before us the fictional events they represent. In the
case where they present material out of sequence, we are to
interpret them in some other way, for we cannot think of our-
selves as seeing past or future events now. Perhaps we are to
think of those images as signs which provide us with informa-
tion about what happens in the story at some time earlier or
later than the story time we have been involved with up to the
point of the time shift. This is implausible. When I watch a film
that contains anachronous material, I detect no difference be-
tween my experience of the images when they present material
in standard order and when they deviate from that order. The
theory we are considering postulates a functional discontinuity
for which there seems to be no psychological evidence.9
Another way to defend the hypothesis that cinematic images
are present tensed would be to suppose that, with episodes of
anachrony, the viewer imagines herself to be shifted in time
along with the image. When the image dissolves to reveal what
happened twenty, two hundred or two thousand years before,
she imagines herself to be a time traveller, shifted in time by
just that amount. This has the advantage of allowing us to say
that anachronous and temporally standard images function in
the same way for the audience rather than in different ways, as
9
Of course the evidence of my introspection (and that of those to whom I have
spoken on this subject) may not be decisive in this matter. But so far as I
know, the advocates of the Claim of Presentness have never brought forward
evidence to support the idea that there is a functional discontinuity here.
Perhaps that is because they do not realize that their theory commits them to
the existence of such a discontinuity.
202
the previous theory would have it. All images, on this view, are
experienced as representing present events; it is just that what
constitutes the present for the viewer is imagined to change to
compensate for temporal shifts in the narrative. But in other
ways this theory is just as implausible as the last; I don't think
I imagine my own temporal position to shift when I view out-
of-sequence images on the screen. It might be claimed that this
kind of imagined shift takes place unconsciously. Perhaps. I
don't say there never are unconscious imaginings; on the con-
trary, I have emphasized that my identification of imagination
with simulation allows for unconscious imaginings in just the
same way that we allow for unconscious beliefs. But we must
not allow the category of the unconscious to degenerate into an
automatic let-out clause for otherwise falsified psychological
hypotheses. When it comes to the hypothesis that the viewer of
anachrony imagines himself travelling in time, this is not
something that we consciously imagine, nor is it something that
prompting naturally brings to mind as part of our previously
unconscious imagining. The only evidence for these unconscious
imaginings seems to be that they are required in order to save
the Claim of Presentness from refutation.
Yet another way to defend the Claim of Presentness would
be to argue that flashforwards and flashbacks do not change the
presentness of the image: that the content of the image in an
anachronous sequence is always the content of a character's
present memory of, or premonition concerning, an event located
elsewhere in time. In that case there would be no need for us to
imagine ourselves moving around in time, and no sense of un-
ease or dislocation on confronting such a sequence would be
expected.
I might raise awkward questions here about how it is that we
could imagine ourselves to be seeing the subjective mental states
of other people, which is what this proposal requires us to imag-
ine. Let us put these awkward questions aside. This view
amounts, strictly speaking, to the abolition of anachrony, be-
cause it would not allow us to say that any story events are
presented out of their strict chronological order. What we nor-
mally call a flashback could not be said to present the event
203
Imagination
itself, but rather to present, "from the inside", a character's cur-
rent memory or premonition of that event, presenting it in the
correct chronological sequence.'o In that case the view we are
considering imposes a very strong and implausible limitation
on the kinds of film narrative there can be. It says that there can
never be a case where the events of the narrative are presented
out of their correct temporal order, and that all apparent cases
of anachrony are really cases of memory or premonition. But if
a film maker can present scenes which represent a sequence of
fictional occurrences in their correct order, why can't she, simply
by rearranging the order of showing, present them out of se-
quence? Further, it seems wrong to claim that every apparently
anachronous episode in film is in fact one of memory or pre-
monition. It is true that anachronous episodes in film (of which
the majority are flashbacks) are often associated with a charac-
ter's psychological state; the character begins describing the past
event and the image dissolves to the flashback. But at least some
cases of the flashback, and especially the flashforward, are not
associated with any act of memory or premonition on the part
of a character. At least, there is in some films no evidence for
this association, and to insist that the association is there purely
on the grounds that the Claim of Presentness requires it is man-
ifestly ad hoc.
We might, on behalf of the Claim of Presentness, associate the
anachrony with the mind of the narrator, an imaginary being
who the viewer is to think of as the (veridical) source of the
information which the film provides. This being may not appear,
•o Discussions of anachrony in ffim seem often to presuppose that current mem-
ories and premonitions do constitute the material for anachrony. See, e.g.,
Stephenson and Debrix, Cinema as Art, especially p. uS. David Bordwell goes
even farther in this surely mistaken direction. He counts it as a case of the
narrative reordering of story time when a character in the story recounts a
story event, as long as that act of recounting constitutes our source of infor-
mation about the event. See his Narration in the Fiction Film, p. 78. The re-
counting is itself a story event, presented in the narrative in the conventional
temporal order; that it is a recounting of an earlier story event makes for no
disruption of story time - otherwise there would be anachrony every time a
character gave the date of his birth. (Genette makes the same mistake while
analysing a passage from Jean Santeuil, in Narrative Discourse, p. 38.)
204
or be referred to, or have any evident connection with the events

depicted, and so is not a fictional character in the ordinary sense.
The narrator's existence is inferred, if it is to be inferred at all,
on grounds of narrative coherence.
It is controversial as to whether every movie, or at least every
movie containing anachrony, contains a narrator in this sense;
we shall discuss this issue in Chapter 9· Let us assume that it
does. The crucial move would then be to establish some image
or sequence of images as present from the narrator's perspective
at the time of his act of narration, and then to establish the
direction of anachrony with respect to this reference point. If X
is identified as present for the narrator, andY follows X in view-
ing time but preceeds it in story time, we have a flashback. But
this proposal requires us to think, not merely that there is a
narrator narrating, but that he is narrating the events of the story
at the very time that at least some of those events are occurring
- those that we are to call present. With most cases of narration
this is not so; the narrator, where there is one, is to be thought
of as recounting the events of the story at a time later than the
time of occurrence of any of those events.
This last proposal draws our attention to a potential ambi-
guity in the notion of presentness: there is the presentness the
fictional events may have for the viewer, and there is the pres-
entness of those events "within the fiction" - their being present
for the characters involved in them. We might attempt to explain
anachrony by appeal to a disparity between what is present for
the viewer and what is present for the characters; in that case a
flashback would be a sequence present to the viewer, as all se-
quences are now assumed to be, but past for the characters. But
this intuitively appealing idea is hopeless. For any story event,
whether presented in flashback or not, it is fictional that that
event is present, at the time it happens, for the characters in-
volved, and past for them at any later time. As we watch the
flashback scenes in Crossfire, they are present for us (or so the
Claim of Presentness says). But the events of those scenes are
also present for the characters involved in them at the time they
occur, just as every scene in the film is. This proposal fails to
distinguish any fictional scene from any other.
205
Imagination
None of these proposals I've canvassed helps to reconcile the
Claim of Presentness with the phenomenon of anachrony. So it
is worthwhile considering whether we can dispense altogether
with the idea that cinematic images represent fictional events as
present. I believe we can. But there is a counterargument to
consider.
7.2 THE PROPER TREATMENT OF

ANACHRONY
I have said that the best argument against the Claim of Pres-
entness is the difficulty of explaining anachrony. But it may
seem that the argument can be inverted: how can we explain
anachrony without the Claim of Presentness? Temporal forward-
ness and backwardness seem to be notions that make sense only
in relation to the idea of temporal presence, something usually
incorporated in definitions of the flashback, as in "a juncture
wrought between present and past"."
The challenge is to make sense of the concepts of flashback
and flashforward without recourse to the (tensed) concepts of
presentness, pastness and futurity. It will be helpful at this point
to introduce a distinction between two kinds of temporal rela-
tions. This distinction was drawn by McTaggart, preparatory to
his argument that time is unreal. 12 McTaggart's distinction,
which I am going to use, is independent of his larger sceptical
purpose, and what I shall say will really be independent of the
whole issue of the reality of time. The concepts I shall appeal to
in explaining anachrony are, it will turn out, concepts that
McTaggart thought were perfectly coherent; they are the un-
tensed temporal concepts. Whether or not these concepts on
their own and without the assistance of the notion of tense can
" Turim, Flashbacks in Film, p. 1. Turim also quotes a similarly tensed account
of the flashback from Leslie Halliwell (The Filmgoer's Companion, 3d ed.), who
calls it "a break in chronological narrative during which we are shown events
of past time which bear on the present situation".
" See J. M. McTaggart, The Nature of Existence, vol. 2, chapter 33· McTaggart
argued, roughly, that the A-series involves a contradiction, that without it
there is no change, and without change there is no time. So there is no time.
2o6
constitute a conception of time is the issue raised by Mc-

Taggart's argument, but it is an argument we can avoid here.
McTaggart suggested that there are two ways we can think
about the temporality of events. We can, first of all, think of
events as past, present or future. McTaggart calls the series of
events ordered in this way the A-series, and the relations which
order that series are tensed. Alternatively, we can think of events
as earlier than, contemporaneous with, or later than other
events. These relations are themselves unchanging; the Battle of
Hastings always was, is and will be earlier than the Battle of
Waterloo. In this schema, no event is privileged as present, and
so no event can be called past or future. McTaggart calls this
the B-series. Relations in the B-series are untensed.
The point I have been labouring can be put in McTaggart's
terms: we cannot explain anachrony in terms of the A-series,
because we cannot identify some particular sequence of story
events as present without so identifying all of them. But in de-
nying ourselves the use of tenses, we have not thereby denied
ourselves the use of all temporal notions, as the distinction be-
tween the A- and the B-series makes clear. We have at our dis-
posal the B-series relations of earlier than and later than. Of
course we may employ other, nontemporal notions as well, and
this is what I shall do when I come to explain certain aspects of
anachrony. What is important is that we should not make ex-
plicit or implicit use of the notion of tense.
I start by observing that there is nothing problematic about
saying that story events are related according to the B-series:
that, within the time of the story, one event occurs before, co-
temporaneously with, or after, another. (Normally, their being
so related is represented, automorphically, by their cinematic
representations occurring before, cotemporaneously with, or af-
ter, one another; the exceptions to this are, precisely, the cases
of anachrony.) This does not require the viewer to think of those
events as past, present or future.' 3 Hoping, then, to define an-
3
' As Robin le Poidevin points out, McTaggart himself wondered whether
the time of fiction might be a time constructed from the B-series alone, al-
though he seems also to have rejected this view (Nature of Existence, vol. 2,
207
Imagination
achrony in terms of B-series relations only, I start with this pro-
posal:
(A) Film F contains anachrony iff F contains representations
of fictional events X and Y, where the representation of
X in viewing time is after that of Y, but it is fictional that
the time of the occurrence of X is before that of Y.
This definition is not quite right. I want to consider a number
of problems that it faces. The first, put to me by David Lewis,
concerns the relation between anachrony in cinematic fiction
and those cinematic narratives which involve time travel. In a
time-travel story the narrative might begin in, say, 1982, during
which plans are made for a journey into the past; later in the
narrative we are presented with events which take place at the
time travelled back to, say 1952. Now it is intuitively wrong to
assimilate cinematic (or other) narratives involving time travel
to anachronous narratives; there seems to be a sense in which,
in the time-travel story I described, the events of the story are
presented in the chronologically correct sequence; after all, the
journey back to 1952 takes place after the events in 1982, which
are presented first - otherwise, how would the events of 1982
constitute a preparation for the journey? But then the objection
to my definition (A) is, exactly, that it conflates anachronous
narratives and narratives involving time travel.' 4 For it is true,
concerning the time-travel story just described, that events oc-
curring in 1952 are shown later in viewing time than events
occurring in 1982.
Lewis was good enough to suggest a way out of the difficulty
for me. The objection shows not that there is an error in defi-
P· 16). See Robin le Poidevin, "Time and Truth in Fiction" for an interesting
discussion of tense and fiction.
'• It is possible for there to be a time-travel story that would not count as
anachronous according to definition (A). That would be a narrative in which
the events "travelled back to" in 1952 are presented first in viewing time
and the events leading up to the journey in 1982 are presented later in view-
ing time. So the objection is not that (A) makes all time-travel stories come
out as anachronous, but rather that it makes some of them so appear - and
in fact all the time-travel narratives I know about would come out as anach-
ronous on the definition.
2o8
nition (A), but rather that there is an ambiguity in its statement.
That is, there is an ambiguity in the expression "the time of the
occurrence of X". Is this supposed to refer to objective time, or
to what is sometimes called personal or subjective time? With
time travel, as it occurs in stories and as it might occur in reality
if it ever does, there is a disparity between objective and per-
sonal time. The traveller travels back to a time earlier in objec-
tive time than the time she left; from 1982 to 1952, as it might
be. But for her, the events she encounters in 1952 are later (say,
an hour later) than the events she previously encountered in
1982. Here the time traveller's journey is thirty years into the
objective past, and one hour into her personal future.' 5
Normally, in stories and in reality, objective time and per-
sonal time run in the same direction and at the same rate, and
there is no need to distinguish between them. Time travel occurs
when they come apart. How is that possible? Perhaps in this
way: that the direction of time is the direction of causation - the
direction from causes to effects. That is why we can remember
the past but not the future, and why, more generally, we are
familiar with traces of the past in the present, but never en-
counter traces of the future in the present (unconfirmed reports
of premonitions aside). But suppose that not all causal processes
move in the same direction- that there is a small minority of
causal processes that swim against the tide. In that case we
could say that the predominant direction of causation is the di-
rection of objective time, and that, given this direction, objects
undergoing reversed causation are travelling back in objective
time. But for those involved, if they are sentient creatures ca-
pable of thought and memory, their journey backwards will end
after it began; for them, the reversing of causal processes will
mean that objectively later states of consciousness will affect ob-
jectively earlier states, and the travellers will remember doing
things in 1982 when they get to 1952; if the journey takes a
significant amount of time, they may end the journey hungrier
than they began it, have fuller beards and longer fingernails.
5
' On the distinction between objective and personal time see, e.g., David Lewis,
'The Paradoxes of Time Travel".
Imagination
The journey ends later in their personal time, and earlier in ob-
jective time.
With this distinction between objective and personal time, we
can solve our problem. In a time-travel story, events that occur
earlier in objective time may be recounted after events that occur
later in objective time. But according to the model of time travel
just proposed, the events occurring earlier in objective time are
occurring later in personal time. So if we take "the time of the
occurrence of X", as that expression occurs in (A), to refer to a
character's personal time, time-travel stories will not count as
anachronous according to that definition. Of course it would be
awkward to interpret "the time of the occurrence of X" in (A)
as sometimes referring to objective time and sometimes to per-
sonal time, according to whether the work in question is a time-
travel story or not. But we need not do that. We may simply
say that "the time of the occurrence of X" always refers to per-
sonal time, which, in the case of a story which involves no time
travel (whether it involves anachrony or not) will automatically
coincide with objective time.
It might be objected to this that what I have said depends
upon a complex and highly unobvious metaphysics of time, and
that we ought not to appeal to such things when we are ex-
plaining the basis of distinctions that ordinary people make, and
make in the same way even though they have no knowledge of
that metaphysics. But while I grant that the metaphysics is dif-
ficult to spell out, I claim that the central distinction it involves
- that between objective time and personal time - must in some
way be grasped by the viewer if he or she is to make sense of
the narrative as one involving time travel rather than anachrony.
It is surely part of our understanding of time-travel narratives
that "in some sense" the events travelled back to by the char-
acters occur later for them than do the events from which they
have travelled back; if that were not perceived to be the case, it
would be hard to explain how film viewers could ever distin-
guish between a time-travel narrative and an anachronous nar-
rative - and the whole basis of the objection to (A) above was
that we do indeed make such a distinction. So while the meta-
210
physics I briefly outlined may be unobvious, its central distinc-
tion I take to be common currency.
The next objection will require some revision of (A), though
only a rather minor one; the need for it was pointed out to me,
again, by David Lewis. Fictions, including cinematic ones,
do frequently give us a good deal of information about the
temporal relations between the events they portray. But in this
area, as in others, they rarely give us complete information.
Sometimes they are unspecific about the temporal relations be-
tween events, even when those events are explicitly represented
onscreen in a determinate order in viewing time. And
sometimes it is not possible to make any reasonable inferences
- either from the order in viewing time or from any other source
- about the temporal relations between these events. Typical of
this phenomenon are those "summarizing" sequences which
might concern the arrival of the Martians: Martians are seen
landing in Paris, then seen landing in Washington, and so on.
But there need not be any implication that the order of showing
corresponds to the order of occurrence; we are simply to infer
that these events occurred at roughly the same time, and to-
gether constitute a sort of collective phenomenon.
In cases such as these we have, as we always do, a determi-
nate order of viewing,' 6 but the order of viewing does not cor-
respond to any order of occurrence in the story. I suggest that
this phenomenon deserves to be classed as a kind of anachrony
- "weak anachrony'' we might call it, since it involves a deter-
minate relation in viewing time but no determinate relation in
fictional time, rather than the strongly anachronous presenta-
tions we have been considering until now, which involve a de-
terminate relation in viewing time and a determinate (but
opposite) relation in fictional time. Our definition of anachrony
ought to cover weak anachrony as well as strong, and it can be
made to do so with a slight alteration:
6
' Devices like the split screen can present material simultaneously, but still the
temporal order is determinate - it is co-occurrence.
211
Imagination
(A*) Film F contains anachrony iff F contains representations
of fictional events X and Y, where the representation of
X in viewing time is after that of Y, but it is not fictional
that the time of the occurrence of X is after that of Y.
This differs from (A) in that the final words "it is fictional that
the time of the occurrence of X is before that of Y" are replaced
by the words "it is not fictional that the time of the occurrence
of X is after that of Y". Any case of strong anachrony as defined
by (A) is also a case of weak anachrony as defined by (A*), but
not vice versa.' 7
There may be further objection to (A) on the grounds that it
does not provide a necessary condition for anachrony, since it
fails to cover the "flashforward ending" exemplified by a film
whose temporal structure we might indicate as 1235.'8 Writers
on narrative have urged that such a structure should be classi-
fied as an ellipsis - a mere passing over in viewing time of rel-
evant story events between points 3 and 5 - rather than as an
anachrony.' 9 If that is right, the (A)/(A*) combination provides
a necessary and sufficient condition for anachrony, the only
possible cases of which are the flashback and the "returned"
flashforward (exemplified in the structure 132), which in my
taxonomy is the only flashforward there is. However, I have
some doubts about the standard treatment of ellipses which I
shall discuss in the appendix to this chapter.
7
' There is an interesting exploitation of the difference between weak and
strong anachrony in Black Widow (Bob Rafelson, 1986), which begins, appar-
ently, with two episodes of wives murdering their husbands. At first I took
this to be weak anachrony: sequential representation of roughly simultaneous
events involving distinct pairs of characters. Soon it turns out that it is the
same woman in both and that the murder represented second in fact oc-
curred considerably before the first.
8
' Here a number's actual value denotes its position in story time, while its
position in the sequence denotes its position in the narrative order.
9
' Describing a flashforward as a leaping ahead "to events subsequent to in-
termediate events", Seymour Chatman goes on, 'These intermediate events
must themselves be recounted at some later point, for otherwise the leap
would simply constitute an ellipsis" (Story and Discourse, p. 64). See also
Genette, Narrative Discourse, p. 43· But further, see the appendix to this chap-
ter.
212
But there is a more serious worry concerning the definitions.
It can be explained most easily by concentrating on (A): while
(A) might be adequate as a condition for the presence of (strong)
anachrony, it is inadequate to specify the direction of the anach-
rony in any particular case. Imagine that X and Y are events of
the fiction satisfying condition (A). The onscreen representation
of Y occurs before the representation of X, but it is fictional that
Y occurs after X. Do we have here a flashback or a flashforward?
It might seem that we have, unambiguously, a flashback; X oc-
curs before Yin the story, but is represented after Yon screen,
so X is represented in a flashback. But this is not necessarily the
case. Y might have been represented in a flashforward, and the
transition to the representation of X is the return which com-
pletes that flashforward. In order to decide whether the transi-
tion in question signifies a flashback or a flashforward we shall
have to look at the representation of some other, suitably related,
story event, and decide whether that belongs to a flashback or
flashforward. If we are not to start on a regress, we shall have
to locate some transition that can be identified as a flashback or
flashforward without reference to another such transition.
Which one will that be?
We might hope to start with the fictional event presented first
in the narration, and anchor the rest of the narrative to it by
discovering a principle according to which the direction of an-
achrony can be judged unambiguously from there. But there is
no such principle. The event first presented may itself be a flash-
back or even a flashforward, and which one it is can depend on
its relation to later representations in the narrative. It is probably
true that there is a tendency to give a certain weight to the
hypothesis that a scene is temporally standard if that scene is
represented first, but that weighting can be overcome. Priority
in narration time offers no Archimedean point from which to
judge the temporal structure of the rest of the narrative.
It is tempting at this point to say that we need to establish
some story event or sequence of events P as present, and then to
judge the direction of anachrony in relation to P. But that would
be to appeal to a notion of tense, which I have said I will
not do.
213
Imagination
So the problem is this: while B-series concepts enable us to
identify an episode of anachrony, they do not enable us to say
what is the correct description of it; they do not enable us to say
whether it is a flashback or a flashforward. But most of the time
we find noncollusive agreement between subjects as to which
description is appropriate in particular cases: flashback or flash-
forward. If B-series notions offer no way of deciding between
the two, that agreement cannot be based on our employment of
B-series concepts alone. So any explication of the direction of
anachrony in terms of the B-series alone must be inadequate.
But I have not claimed that the direction of anachrony could
be explained in terms of the B-series alone. I have claimed that
the concept of anachrony itself could be so explained. It is one
thing to decide what constitutes anachrony, another to decide
how it is directed in particular cases. Of course I must give an
explanation of this direction. I shall do that by describing the
sorts of considerations that incline us to say we are dealing with
a flashback, or alternatively with a flashforward. These consid-
erations are not themselves temporal, and so they have nothing
to do with tense. They have to do with simplicity and the em-
phasizing of nontemporal connections between events in the
story.
Suppose we have an episode that is anachronous. We have
to choose between one description of it and another, and our
choosing is partly dictated by the principle that we should im-
pose as simple and straightforward a narrative structure as is
consistent with the film itself. Of course, the simplest structure
we can arrive at may in fact be rather complicated, and the film
may so defy our narrative expectations that all efforts to impose
on it a consistent structure fail. 20 It is just that we start with the
assumption of simplicity and introduce complexity as needed;
any other path would make narrative interpretation impossibly
difficult. Let us see how, in particular cases, the constraint of
simplicity works to dictate, or at least to suggest, one description
of anachrony rather than another.
"' Resnais's Last Year at Mnrienbad and Deren's Meshes of the Afternoon are often
cited as examples of this.
214
If the film's narrative structure is represented in B-series

terms as 41235, we tend to say that the sequence 123 is a flash-
back. The alternative, to say that 4 and 5 are separate flashfor-
wards, would ascribe two anachronous episodes to the film
instead of one, and the rule of simplicity tells against our adopt-
ing that alternative description. If the film's structure is repre-
sented in B-series terms as 12356, it looks as if we have a choice
between saying that 123 is a flashback or that 56 is a flashfor-
ward. Simplicity considerations are not telling here. But the no-
tion of ellipsis will help. We may regard the passage from 3 to
5 as an ellipsis, and so avoid the imputation of anachrony to the
film at this point- a victory for simplicity, assuming that ellipsis
is not itself a species of anachrony (see the appendix to this
chapter).
Simplicity considerations compete for our attention with oth-
ers. We employ descriptions of anachrony sometimes to rein-
force or to cohere with certain aspects of the story or of the
"narrative voice" we associate with it. If the structure is 1342,
we are bound to admit anachrony, and we could say that 1 and
2 are flashbacks, or that 34 is a flashforward. Simplicity favours
the latter, but if 1 and 2 are connected to 34 by some episodes
of memory that occurs during 34, we are likely to see 34 as
"looking back" to 1 and 2, and so to describe them as flashbacks,
even when they present fictional material in an objective way. 2 '
Indeed, in classical film narration, anachrony is so frequently
connected with the inner states of characters that there is some-
thing close to a conversational implicature from the anachrony
to the existence, within the story, of some relevant state of mem-
ory or (less frequently in realistic cinema) premonition.22 And
since premonition is a device hostile to realist assumptions, the
implicature works against the flashforward, nudging us towards
a description in terms of flashback, even at some cost to sim-
plicity. But the implicature, being conversational, is cancellable,
and other aspects of the narrative or the story itself may relieve
" See the remarks on representation "from the inside", text to note 10 this
chapter.
,_, On conversational and other kinds of implicatures see Paul Grice, Studies in
the Way of Words, part 1, chapter 2.
215
Imagination
the anachrony of its psychological associations. The quick cut-

ting between the earlier lovemaking and the later dressing in
Don't Look Now are not suggestive of premonition, though later
anachronies in that narrative can be seen, on reflection, to be so
connected. 23 Sometimes in films with a complex or unusual tem-
poral structure, there seem to be distinct but equally good ways
of describing that temporal structure; there does not seem to be
any uniquely correct answer to the question, "Is this a flashback
or a flashforward?" We might call this scene a flashback; but we
could arrive at an equally coherent overall structure by calling
a later scene a flashforward. That is just what we would expect
on my theory, according to which the decision to label a partic-
ular shot or scene anachronous depends on weighing a number
of factors that may conflict, and which may not point us in one
direction rather than another. 24
A more thorough analysis of our strategies for working out
the direction of anachrony I leave to others. I think I have said
enough to make it plausible that judging the direction of an
anachronous episode is a matter of applying interpretive con-
cepts that are not strictly temporal. 25 In that case it's no objection
to my theory that it does not provide for a definition of the
direction of anachrony in temporal terms alone.
7 ·3 IS THIS REVISIONISM?
I have argued in this chapter, as in the preceding one, that some
of our common ways of speaking about the cinematic experience
3
' Flashforwards unconnected with premonition occur sometimes in credit se-
quences, for example, in Clayton's The Innocents and Wilder's Double Indem-
nity. The Godfather, Part 2 is notable for its use of objective flashbacks (unless
we are meant to view them as flashbacks within flashbacks, filtered through
the recollection of Vito Corleone, whose death occurs before the beginning
of Part 2; I think there is some evidence for this interpretation).
24
See, e.g., the discussion of They Shoot Horses, Don't They? in Bernard F. Dick,
Anatomy of Film, p. 1]8.
5
' Noel Burch, I think, was making a similar point when he asked, "Are not
jumps forward and backward in time really identical at the formal organic
level of a film?" (Theory of Film Practice, p. 8).
216
are fundamentally wrong. Yet my aim has not really been a
revisionary one; I have not been trying to show that we should
approach the cinema in a new and unfamiliar way. My aim in
this chapter has been largely to rescue our judgements about
time and especially about anachrony in film from what I take to
be a false theory about the basis of those judgements. I have not
tried to argue that people are wrong in any straightforward
sense when they describe this or that movement of the film's
narration as a movement from the present to the past rather than
as one from the present to the future. I say only that we should
not take them literally when they employ words like "past",
"present'' and "future". The Reverend Dr Spooner is said to
have ended a puzzling sermon by saying, after a moment of
reflection, "Whenever I said 'Aristotle', I meant 'Saint Paul' ".
We would do well to believe him. It is sensible to reassign mean-
ings to a speaker's words if in doing so we can interpret him as
saying something significantly closer to the truth. At least, it is
sensible to do this if we can reasonably credit the speaker with
a sensitivity to the concepts that our meaning reassignments in-
troduce, and reasonably explain his utterance as the result of his
sensitivity to those concepts. We go along with Dr Spooner's
plea for reinterpretation because we believe him to have had a
grasp of the concept of Saint Paul, and we find it plausible to
think that what he said in the sermon was the outcome, in part,
of that grasp. We do not reinterpret the speech of witch finders
so as to make them early investigators of epilepsy, because it is
not reasonable to credit them with a sensitivity to that concept
- they would come out, by our lights, as very poor epilepsy
discriminators. But it is perfectly reasonable to assume that com-
petent and alert film watchers have a sensitivity to those con-
cepts in terms of which I explain the phenomenon of cinematic
anachrony: temporal but untensed ones, plus the machinery of
narrative interpretation. So we should treat them as Dr Spooner
would have us treat him.
There is one somewhat surprising consequence of the view I
have advanced. It is that the concept of the direction of anach-
rony is on a quite different footing from the concept of anach-
217
Imagination
rony itself. Anachrony is explainable in terms of temporal
concepts. The direction of anachrony is not; it is fundamentally
a dramatic notion.
In Chapter 3, I asked: in what sense, if any, is cinema a temporal

art form? My answer was that it is very centrally a temporal art
form because of its capacity for the automorphic representation
of temporal relations between events in the fictions it presents.
But the argument of this chapter may seem to be in tension with
that optimistic conclusion. The argument of this chapter has
been that there are certain kinds of temporal relations which the
cinema is not apt to represent at all; the tensed temporal prop-
erties of pastness, presentness and futurity. Instead, the cinema
represents events- fictional events, that is- as standing in tense-
less relations of priority and co-occurrence.
But we need to bear in mind here a distinction made earlier,
that between presentness for the viewer and presentness for the
characters of the fiction itself. The failure of cinema to represent
fictional events as tensed is a failure to represent them as tensed
from the perspective of the viewer, not from that of the characters.
When events of the fiction are represented onscreen we are to
imagine that those events are present for the characters at the
time of their occurrence, future for the characters at earlier times
and past for them at later times, just as real events are for us.
So the time that cinema represents - the network of temporal
relations between fictional events - is not a peculiar time to
which no tensed predicates apply. Tense applies only within the
fiction itself. 26 Just so with space. While the events of the fiction
are not, so I have argued, spatially present for the viewer - the
viewer is not to imagine that they are occurring here- they are
spatially present for the characters. Characters, if they could be
6
' Here I agree with Christian Metz: "The film is able to express space and time
relationships of some kind, but only anaphorically, within the film it-
self ... and not between the film and someone or something else" ("The Im-
personal Enunciation, or the Site of Film", p. 756). But I am puzzled by
Metz's talk, here and elsewhere in the article, of anaphora and the related
concept of deixis, which I think he misapplies. At least, I have found his
point hard to follow.
218
bothered, could truly and at any time say "I am here now'', just
as we can. So we can say: the space and time of cinematic fiction
is structurally like the space and time of the real world.
APPENDIX: ANACHRONY AND ELLIPSIS
In the body of this chapter I followed the orthodoxy which dis-

tinguishes sharply between anachronies and the phenomenon of
ellipsis, representable in our simple notation by, for example,
1245. Of course it is true that an ellipsis is not the same thing
as either a flashback or a flashforward. But they, in their tum,
are different from each other. The question is whether there is
some decisive reason for classing the last two together as an-
achronies and categorizing the first, ellipsis, as something fun-
damentally different. Or could it be that all three belong together
in a general theory of anachrony?
Note first that from the point of view of what we might call
"the violation of real time", an ellipsis is in the same boat as a
flashforward or flashback. A time traveller, whose activities sub-
vert the order of objective time, might do any of the following:
travel into the past, returning (or not) to the present, or accel-
erate forward into the future, again returning (or not) to the
present. If we join ellipses to our flashforwards and flashbacks,
we get exactly the class of time-travel possibilities. Ellipsis
would be the case of time travel by acceleration forward without
return, suggesting that ellipsis is as deserving of the title anach-
rony as the other candidates. Why should it be treated differ-
ently from the others?
Perhaps for this reason. In literary narrative, there is no gen-
eral correspondence between the time it takes to read the de-
scription of a fictional event, and the time which, fictionally, that
event took to occur. Since reading speed varies between indi-
viduals and occasions, the most carefully embodied intention
that the narrative take as long to read as the events it describes
could succeed only very approximately. In that sense, all literary
narrative is in violation of the time of the fiction it describes -
or perhaps we should say that its conforming to or violating
that time does not arise. In that case, ellipsis, which is a matter
219
Imagination
of temporal duration rather than temporal order, cannot be seen

as a device that works in opposition to any norm that prevails
in literature; it is just more of the same. But flashes forward and
back (or what are more frequently called prolepses and ana-
lepses in literary discussion) are violations of temporal order,
which can itself be preserved - and typically is preserved - in
literature. So when we consider literature, it is natural to think
of these two as separated from ellipsis. But in the case of film,
that argument does not hold. With film, any continuous shot
constitutes the rendition, in a certain space of time - the same
for all observers - of events in the fiction that took a certain time
to occur, and typically the relation between the time of the shot
and the time of the fictional event is identity. So in the filmic
case, at least within the confines of a single shot (and frequently
across the class of shots that constitute a scene), there is no vi-
olation of the time of the film - neither with respect to order
nor with respect to duration. So ellipsis, when it occurs, consti-
tutes a violation of a cinematic norm in the same way that flash-
forwards and flashbacks do. So when we are considering film,
there is a case for classing all three - ellipsis, flashforward, flash-
back - together as anachronies, and for not simply taking over
the theoretical divisions inherited from literature.
If, as I am urging, we class ellipsis as an anachrony, we can
no longer take definition (A*) as a definition of anachrony; it is
instead to be taken as a definition merely of two kinds of an-
achrony, the flashforward or flashback. To define anachrony
fully we need to supplement (A*) with a definition of ellipsis.
We can say that ellipsis occurs when there are times of the fic-
tion, x andy, which are represented by viewing times R(x) and
R(y), and a time of the fiction, z, lying between x andy, which
is not represented by any viewing time. And we can say in
general that anachrony is any violation of story time where the
kinds of violations possible are just those described in definition
(A*) and in the definition just given of ellipsis.
For the filmic case, defining ellipsis is thus relatively straight-
forward. Not so for the literary case. Because literature is not a
time art in the way that film is, we cannot define ellipsis in
literature in terms of a relation between times of the story and
220
times of the representation of the story; we have seen that the
verbal representations of literature do not, strictly speaking,
have times. And the times of, say, a reading of the text of the
literary work do not correspond, point for point, to the times of
the story, as has been pointed out. Ellipsis in literary fictions
seems not to be definable in terms of strict temporal relations
between story and representation; that is why ellipsis really does
stand apart from the flashforward and the flashback in the lit-
erary case. For literature, the closest analogy of the temporally
definable concept of cinematic ellipsis is probably the concept
of a narrative which omits, or fails to describe, not times but
events of the story. Thus we might say that an ellipsis in litera-
ture occurs whenever the narrative recounts story event A, and
story event C, but fails to recount B which (fictionally) happened
between A and C.
The problem with this definition is that it is almost entirely
unhelpful. The narrative declares, "Albert slept soundly for
eight hours." Is there ellipsis here? Certainly there are constit-
uents of this eight-hour sleeping event which are not encom-
passed by the description: the breaths that Albert took, their
order, duration and magnitude, not to mention the events else-
where in the city as he slept. But this is true of just about any
description, however minute its detailing of events; rarely can a
description convey everything that happened in a finite time,
however brief. The definition forces us to say that every part of
every literary narrative involves ellipsis. We need a more de-
manding characterization of ellipsis; one that does not confuse
ellipsis with mere incompleteness.
We might seek instead to define literary ellipsis in terms of
the amount of time passed over; ellipses proper involve the
passing over of a substantial amount of time. What counts as
substantial might be highly sensitive to context: in a narrative
given over to the minutia of a brief experience, a gap of a few
seconds might count as an ellipsis, while in a novel with pre-
tensions to an historical vision, only a gap of years might count.
The problem here is not the context dependence of substantial-
ity, but that a gap that is small even by the standards of the
relevant context can induce an ellipsis if it leaves out a significant
221
Imagination
event. We then depend on a notion of narrative significance that
is so far undefined - though its application might be relatively
clear in practice. 27 But if we define ellipsis in terms of signifi-
cance - itself suitably defined - then we cannot say, as writers
like Chatman have said, that a leap forward in time that is not
"returned" is always an ellipsis rather than a flashforward, for
now its being an ellipsis would depend on its being a case where
some narratively significant event is passed over. We could say
that in these cases - narratives that end with a temporal leap
forward- what makes them ellipses is their extent, thus return-
ing to the first definition. It looks as if extent explains some of
our judgements that we confront an ellipsis, whereas signifi-
cance explains some others; it might even be that between them,
these two criteria exhaust the field, and that we could say that
ellipsis is to be disjunctively defined as grounded either in ex-
tent or in significance. But the trouble here is that the two
grounding concepts - extent and significance - don't seem to
have much to do with each other, and it is not likely that an
intuitively unified concept like that of the ellipsis is actually the
union of two unconnected ones. The true definition must, it
seems, be found at some deeper level. I am unsure where. Since
our primary concern in this case is with cinematic rather than
with literary narrative, we may leave this difficult problem here.
7
' But see the remarks on narrative unreliability and significant questions in
Chapter 9, Section 9·4·
222
Part III
Interpretation
Part I argued that the depictive content of film is given
I
N
to us courtesy of our capacities for recognizing the objects
depicted. The process of recognizing depictive content
hardly counts as interpretive. But there is more to under-
standing a film than merely recognizing its depictive content.
That is where interpretation starts. In Chapter 8, I develop a
theory of interpretation for literary and filmic narratives. But
there are important differences between the kinds of narra-
tives available to film and to literature. In Chapter 9, I illus-
trate this diversity through the example of narrative unreli-
ability.
Chapter 8
The interpretive problem
Even if we reject the thesis that creative interpretation
aims to discover some actual historical intention, the
concept of intention nevertheless provides the formal
structure for all interpretive claims. I mean that an
interpretation is by nature a report of a purpose.
Ronald Dworkin
The subject of this chapter is a problem and how we solve it.

The problem I shall call the interpretive problem, and it is this: an
interpreter starts with something and ends with something else.
The interpreter of a novel starts with a text - a sequence of
words and sentences - and ends with a story: the story which,
as the reader I interpreter sees it, is told by that text. The inter-
preter of a film starts with a sequence of cinematic images and
their auditory accompaniments, and ends with a story: the story
which, as the viewer I interpreter sees it, is told by those images
and sounds. The interpretive problem is how to get from the
one to the other. The philosopher's problem is to give an abstract
characterization of the principles and methodological rules in
accordance with which interpreters solve the interpretive prob-
lem. In solving the philosopher's problem, we give a theory of
interpretation.
The interpretive problem looks so forbiddingly difficult it can
seem astonishing it ever gets solved at all. In the filmic case,
images succeed one another quickly, the film cutting across large
chunks of space and time. Sometimes the narrative is reshuffled,
as with flashback. Usually this goes on without any explicit com-
mentary or other onscreen direction to indicate the relation be-
tween successive images. With literature such commentary is
more common, but it never makes fully explicit more than a
fraction of what the reader is intended to imagine. Yet somehow,
and frequently with little or no conscious thought or effort,
225
Interpretation
viewers construct a story from it all. "Interpreter'' suggests
someone with specialized skills and a reflective, self-conscious
approach. But I shall use that term for anyone who gets, or at-
tempts to get, a coherent story out of the bewildering array of
sights and sounds: for anyone, that is, who attempts to solve an
interpretive problem.
Briefly stated, my theory of interpretation is this: to interpret
is to hypothesize about the intentional causes of whatever it is
that is being interpreted- a temple, a text, a practice, a picture,
the sequence of moving pictures and accompanying sounds that
constitute a movie. Interpretation is intentional explanation, and
it proceeds according to the methodological canons that govern
explanation in general: we count one interpretation as better
than another when it is simpler, more plausible, better sup-
ported by the evidence and in general more explanatory than
the other. In this, textual and filmic interpretation do not differ;
where they differ is in respect of what interpretation is to be an
explanation of - a text in the one case and cinematic images in
the other.
Until late in this chapter, and with the exception of a few
asides, I shall concentrate on the literary case as I build up a
theory of how we solve the interpretive problem. Towards the
end I shall say something about what we need if the model is
to apply to film.
8.1 INTERSUBJECTIVE AGREEMENT AND

INTERPRETIVE PRINCIPLES
The idea that there might be generally valid principles of inter-
pretation applicable across media, genres, times and communi-
ties is not likely to receive much support from literary and
cinema theorists. The most that is generally allowed is that there
are certain principles that have a local, community-based valid-
ity, or which are promoted and fostered by certain powerful,
historically specific interests, and that the functioning of these
principles to guide interpretation is susceptible to a sociological
explanation, but not in any sense to a philosophical justification.
Closely related to this is the view that different genres or styles
226
of film making are characterized by their own local interpretive
principles, and that successful interpretation is largely a matter
of correctly identifying the genre of the work so as to be able to
apply to it the appropriate interpretive principles.
An argument for supposing that the rules of interpretation
are not universal is that people for whom our culture is alien
would interpret works, ours and theirs, in ways different from
our own, and would come up with different interpretive hy-
potheses. I agree; in fact I shall make quite a lot of this point
later on. But this does not show that there are no universal prin-
ciples of interpretation. That Martians might- and humans from
other cultures occasionally do - interpret works in ways differ-
ent from ours need not be because they operate with different
interpretive principles. For the principles are and must be em-
ployed always in the light of culturally specific assumptions;
interpretive divergence may result not from the application of
different principles, but from the application of the same prin-
ciples against different, culturally determined backgrounds.
Two scientists may come to entirely different conclusions about
the likely outcome of a certain experiment, but this need not
show that they disagree about what laws of nature (the analogue
of our principles) are operating, though of course this is one
possible source of their disagreement. It is possible instead that
their conflict is the product of a disagreement about the initial
conditions of the experiment. If one of them thinks that the sam-
ple of uranium has a mass of one gram, and the other thinks it
has a mass of one ton, they are likely to predict very different
outcomes, even though they both reason in the same way from
the same general theory of matter.
So it is, I believe, with the case of interpretation. We may be
able to accommodate interpretive differences between groups,
or between the same group at different times, or even between
subgroups and individuals within the same community, while
holding interpretive principles constant across groups, sub-
groups, times and individuals. We may all operate on the prin-
ciple that the work is to be interpreted by figuring out the
intentions behind its production, but come up with different in-
terpretations because we operate with different assumptions
227
Interpretation
about what sort of intentions the maker is likely to have. An
intention that it would be natural for us to attribute to a maker
might be for others highly unnatural, because of its deployment
of concepts that are alien to their culture, that lack salience in
any reasonable schedule of their concerns, or that are judged by
them to be forbidden or shocking. And some of the intentions
we naturally attribute to makers get their salience for us from
their appearance in other, related works of art that belong to our
culture. Our background of exposure to works which conform
to certain generic norms means that the maker will have to work
hard to get certain nonstandard intentions recognized, while
others that conform to the expectations generated by those other
works will be recognized easily and without explicit signalling.
People without exposure to that same body of work will arrive
at the interpretive task with a system for weighting the evidence
quite different from ours. All these differences can be assumed
to lead to different interpretive results, even where everyone
follows the same general interpretive method.
But this talk of disagreement should not obscure an important
fact: that there is usually a good deal of agreement between
readers and film viewers about how works are to be interpreted,
if we confine ourselves, as I do here, to interpretation of the
story, or what I am going to call narrative interpretation. Narrative
interpretation is the kind of interpretation that even the least
ambitious of us engage in when we read: working out what is
going on in the story. Some of what is going on in the story that
a fictional work has to tell may be obvious and uncontroversial.
That Mr Dombey is the owner of a shipping firm, that Carkar
dies under a railway train, that Florence leaves her father- these
things are generally agreed to be true in Dombey and Son. In
coming to such conclusions about the work we do not think of
ourselves as identifying the fictional events by any process that
deserves the name "interpretation". But the obvious is merely
that which is given by a kind of interpretation of zero degree;
interpretation under default assumptions of literal speech, nar-
rative reliability, rationality and evidential relevance. When we
fail, under those assumptions, to generate a narrative that is
fully coherent we look for nonliteral readings and evidence of
228
narrative idiosyncrasy. At that point we are happy to say we
are interpreting; in fact, we are just moving up an interpretive
gear.
Narrative interpreting is not, of course, all the interpreting
there is, even when we limit the act of interpretation to the fic-
tional. We may agree on what happens in Dickens's novel but
disagree on whether the opening chapter maintains a tension
between comedy and tragedy, whether the railway serves as a
symbol of social change, whether the novel bears the marks of
the onerous constraints of serial production under which it was
written. All these may fairly be called interpretive questions, but
they are not the kinds of questions I have anything to say about
here. Similarly, viewers often disagree about things connected
with the film outside the realm of narrative interpretation: what
kind of mood was conveyed by a particular shot, what moral,
if any, the story was supposed to drive home, how successful
overall the film's project was. These sorts of disagreements are
not, in my sense, about what story the film presented. A disa-
greement of that kind would be, for example, about who com-
mitted the crime, whether the couple who married at the end
are going to live happily thereafter, or what the bride's moti-
vation in marrying was. Disagreements over matters like that
do occur, but against a background of overwhelming agreement
about other matters concerning the story. It is only infrequently,
and only in the case of very eccentric narratives, that there is
fundamental disagreement about what happened in the story.
And in the cases of those eccentric narratives, there is more
likely to be a shared bewilderment about what happened than
a clash of definite opinion. Seen in this light, it is agreement that
is central, and dispute that is marginal.'
While narrative interpreting is a small part of what we nor-
mally call interpreting, I make no apology for concentrating
attention on it. It is, after all, absolutely fundamental to inter-
pretation more broadly conceived. If we cannot agree about
what is going on in the story, we shall hardly agree about the
' Disagreement and its implications I discuss in "Interpretation and Objectiv-

ity".
229
Interpretation
story's significance, message, or relation to other stories. Con-
versely, an argument about symbols or genres that is not re-
flected in the least disagreement at the level of what is fictional
seems an excellent candidate for being a spurious or at least
unresolvable dispute.
You might hope to explain agreement at the level of narrative
interpretation without appeal to universal principles of inter-
pretation by arguing that genre-specific principles would be suf-
ficient to generate agreement. This cannot be right. Deciding
what genre a work belongs to is itself an interpretive matter/
one of deciding to give certain elements in the work a signifi-
cance that depends on their relations to all the other elements
of the work. There is no checklist of features that determines
whether the work is a screwball comedy or a film noire. But as
with other interpretive judgements, we do find a broad measure
of agreement between agents concerning what genre this or that
work belongs to, and so the argument for their being interpre-
tive principles applies as much here as elsewhere. It sometimes
happens that the viewer knows what genre the work belongs to
by some independent route; it might be advertised in such a
way as to make its genre clear. But this is not how it always is,
and we usually are capable of deciding, in an unprompted way,
on the genre of a work by seeing it.
I am not claiming that there are no local principles to be de-
ployed in a media-, genre-, or style-dependent way. There may
well be such principles, but they are not the only ones. Nor am
I claiming that the interpretive method which I regard as uni-
versal always or even usually enables us to arrive at one single
and uncontroversially correct interpretation of the story. Inter-
pretive principles are never mechanical rules that can be applied
as algorithms, and there is often room for rational disagreement
as to whether the rule has been correctly applied in this case;
such disagreement is one source of the plurality of interpreta-
tions. Another is the fact that the rules, even when there is no
disagreement concerning their application, can sometimes lead
• Interpretive in a broad sense, not in the narrow sense I shall be concerned

with here - narrative interpretation.
230
to a tie between competing interpretations: we shall see that the
rules do not guarantee that one interpretation will always be
preferable to the rest.
Before I develop the theory of interpretation I shall need some
distinctions, and some theses about the relations between the
things thus distinguished. That will occupy the next two sec-
tions.
8.2 TERMS AND CONDITIONS
"Interpretation" and "narration", like "assertion" and "utter-

ance", are subject to an ambiguity. Do we mean to refer to the
act of interpreting/narrating/asserting/uttering, or to the results
of doing those things, where the result of interpreting is an in-
terpretation, as the result of asserting is an assertion? My con-
cern with interpretation and narration is with both: with actions
performed and with their results. In some cases we can say the
same things about both - the activity of interpreting is objec-
tively evaluable just in so far as its result is. But in general the
act of interpreting has features rather different from its results:
the act takes time, it is done by someone and it is (or may be)
a rule-governed activity. Similar features distinguish narrating
and its results, and we had better not confuse them. We can
clarify talk of narration by speaking of a narrative when we mean
the result, reserving "narration" for the act. But we cannot (at
least, I will not) speak in the same way of an "interpretive", so
in this case context or explicit disclaimer will have to distinguish
the action and its results.
I have said that my talk of narrative and of interpretation will
be a good deal more restrictive than much current usage would
allow. What I shall say here about narrative interpretation cer-
tainly invites generalization, not all of it desirable. For example,
the interpretive problem as I have described it is somewhat like
that induced 'by the "poverty of the stimulus" argument con-
cerning perception - an argument sometimes taken to show that
perception is itself an interpretive process. The perceiving agent
has a rich visual experience of the world around her, yet the
input to the eyes consists of mere patterns of retinal stimulation.
231
Interpretation
Some people have concluded that perception itself- the process
by which the relatively impoverished stimulus is converted into
our visual experience of the world - must be a richly cognitive
one, that visual experience is the joint product of the low-level
stimulation and high-level "hypotheses". On this view, to have
visual experience of the world is already to interpret the world,
to bring to it assumptions and expectations which materially
alter the quality of experience. If these assumptions and expec-
tations were different we would experience the world in differ-
ent ways. In the same way, narrative interpretation starts with
a relatively impoverished stimulus - a literary text or cinematic
image sequence- and gives rise to a vastly richer experience of
a story, many elements of which are simply not available, or
even hinted at, in the text or image sequence itself.
We certainly can use the term "interpretation" in so broad a
sense as to cover both perception and narrative interpretation,
and indeed to cover any process whereby we get more from
less. But this is a very undiscriminating sense of interpretation.
I put my money on the following hypothesis: while both per-
ception and narrative interpretation are processes by which we
get more from less, the first process gets its enrichment from
biology, and the second from culture. The first process is rela-
tively impervious to belief, the second richly informed by it.
Perception builds up a picture of the external world richer than
the perceptual input, and it does so by applying to the retinal
stimulation rules which have been selected by evolution, given
the sort of physical world in which we live. Human beings,
pretty much independently of their cultural background, will
have largely the same visual experience on the basis of the
same retinal input. They will see edges, volumes and depth re-
lations in very much the same way. But what story they derive
from a text depends crucially on assumptions and expectations
which are culturally determined. We, or some of us, read The
Turn of the Screw as ambiguous between a psychological and a
supernatural tale; for people whose culture makes no such dis-
tinction such a reading is not available. If their culture is un-
critical of supernatural explanations, they will not see the
governess's narrative as unreliable. We read Ring Lardner's
232
"Haircut'' as unreliable, in the sense that that narrator's moral

vision is, and is clearly intended to be seen as, deficient. For
people with a moral outlook sufficiently different from ours, no
such unreliability would be suggested.
This is not a "their reading is as good as ours" argument. It
may be that the correct way to interpret is to do so in light of
assumptions from the work's home culture, which might be dif-
ferent from your own, and that to fail to see ambiguity and
unreliability in the works just mentioned is to interpret them
wrongly. 3 My claim is just that narrative interpretation depends,
as a matter of fact, on cultural assumptions in a way that per-
ception does not.
Ambiguity and unreliability in narrative are topics I shall take
up in more detail in the final chapter. But the brief remarks
already made about them suggest something important about
the relation between three concepts: text, narrative and story -
concepts which are often not sufficiently distinguished. A text
is a sequence of words and sentences, something which can be
uttered by someone - an author perhaps - on a particular oc-
casion.4 A story is a set of propositions to the effect that this
happened, and then that happened, that this happening caused
that to happen, that character A was involved in such and such
a way in initiating the causation, that character B was affected
thus and so. Sentences in the text may, and usually do, express
certain propositions, but the poverty of the stimulus argument
shows that the correspondence between propositions thus ex-
pressed and propositions in the story is very unsystematic. Cer-
tainly, text and story are not the same thing. What, then, of
narrative? I offer the following slogan to encapsulate the relation
between text, narrative and story. A text is narrative in virtue of
the story it tells. That is, a text is a narrative, in the same sense
in which a person- Elizabeth Windsor, say- is a monarch. She
is queen in virtue of her relations to other people and institu-
tions (to previous monarchs, to the Constitution), and her being
3
The rightness or wrongness of an interpretation is a difficult issue I take up
in "Interpretation and Objectivity".
4
For more on the individuation of texts and their relation to authorship see
my "Work and Text".
233
Interpretation
queen is a contingent fact about her. Someone else might have

had those relations and thereby have been queen in her stead.
Yet there are not two things here, E.W. and the queen; E.W. is
the queen. Similarly, those words and sentences authored by
Henry James constitute a narrative by virtue of the story they
tell. The text itself is not essentially a narrative; it tells that story
only by virtue of contingent facts about James's activities and
the culture in which he acted. But by virtue of those contingent
facts, the text is a narrative, and of a certain kind. Perhaps it is
an unreliable narrative. If that is so, it is because the story it tells
is different in crucial ways from the story it appears to tell. (Just
assume for present purposes and in line with certain interpretive
views, that the text appears to tell a ghost story, but actually
tells a story of delusion.) Change your assumptions about what
story it tells, and the unreliability goes away.
So the concepts of a text and of a narrative are different, just
as the concepts of E.W. and of the queen are different. But still,
it is the text which is the narrative, just as E.W. is the queen.
Text and story, on the other hand, are substantially different; no
text is a story, and no story is a text. For a text is composed of
sentences and a story is not.
I said that the task of narrative interpretation is to find out
what is going on in the story. But that is not quite right. I in-
dividuate stories by what is going on in them: different events,
different stories. Speaking more carefully, I say: the task of in-
terpretation is to discover what story this work has to tell. In-
terpretive disputes can be large or small, but where there is even
a minute difference of opinion as to what the story is, I say we
have different narrative interpretations. Not all interpretive dis-
agreements about narrative are significant. Some are.
We now know what an interpretation is: it is an hypothesis
about what is true in the story. Recalling an ambiguity earlier
noted, I say that an interpretation in this sense is what results
from the act of interpretation; an act of interpretation results in
your acquiring an interpretive hypothesis. Is any act which re-
sults in the acquisition of such a hypothesis an act of interpre-
tation? No. To interpret is to acquire a certain kind of
understanding of the work, and to acquire it in a certain way.
234
Not every way of acquiring understanding of the story counts
as interpreting. Someone else who has interpreted the work can
save you trouble by telling you what she knows. That way you
acquire an interpretation, but you do so without interpreting.
To interpret is to acquire this understanding by a certain kind
of engagement with the work. So far I have merely been offer-
ing definitions; saying what kind of engagement I have in mind
here - what acts we have to perform if we are to count as inter-
preting- will take us to the more interesting level of theory.
I have spoken about our knowledge of an interpretation; and
knowledge requires, at a minimum, truth, and belief in the truth.
Should we say, then, that interpretations are things which are
true or false, and that when there is a clash of interpretations at
least one of the rivals is false? Not always, in my view. But that
is an issue I try to sort out later in this chapter. Prior to that
refinement, it will do no great harm to speak of knowledge of
interpretations in an unqualified way.
Now I pass to the question: how does interpreting proceed?
8.3 INTERPRETING BEHAVIOUR AND

INTERPRETING WORKS
I begin with an apparently unrelated issue: how do we interpret

a person's behaviour? We interpret behaviour when we offer an
hypothesis about the mental states that caused it. Forming such
an hypothesis requires us to attribute beliefs and desires to the
agent. It is an hypothesis which says that the subject believed
so and so, and desired so and so, and that his so believing and
desiring was the cause of the behaviour we wish to explain.
The relation between behaviour and mental states is recip-
rocal: while mental states explain behaviour, behaviour is the
evidence for an hypothesis about mental states, and the mental
states explain the behaviour just to the extent that the behaviour
is evidence for the mental states. When we ascribe beliefs to
people, we do so on the evidence of their behaviour, including
their verbal behaviour. But only an extreme and indefensible
behaviourism would tell us that beliefs can be read from behav-
iour in a determinate way. Behaviour is evidence for belief; it is
235
Interpretation
not constitutive of it in any way that would allow us to pair
individual items of behaviour with individual beliefs. 5 Here, as
with most other explanatory projects, we have a choice between
competing hypotheses: there are a variety of combinations of
beliefs and desires that would probably have resulted in that
behaviour; which one actually did?
We can rule out quite a lot of competing hypotheses by at-
tending to the theoretical virtues of simplicity, coherence, pre-
dictive power and rationality. Those virtues might be expressed
in the following injunctions. Choose, where you can, explana-
tions that make the connection between mental states and be-
haviour as direct as possible; avoid, where you can, explanations
that attribute mental states without direct behavioural correlates
(assume the subject uttered the words "Please pass the salt"
because he wanted to salt his food rather than because he be-
lieves in the power of salt to magically transform his surround-
ings). Choose, where you can, explanations that enable you to
predict the subject's future behaviour (assume the subject has a
relatively stable desire for salt in certain circumstances rather
than that this is a one-off choice). Choose, where you can, ex-
planations that make present behaviour cohere with past behav-
iour (consider what a previously salt-avoiding agent might hope
to get from his utterance other than that you pass him the salt).
Choose, where you can, explanations that attribute to the agent
beliefs and desires roughly in line with those you would expect
to have in his situation (be sceptical of the hypothesis that he
wants to salt his dessert). As the examples show, these injunc-
tions can pull in different directions. You may have to trade off
some against the others, and some kinds of behaviour may find
their best explanation in an hypothesis that sacrifices all three
to some extent; suddenly acquired, hidden, or strange beliefs are
not an impossibility. It is just that deviation from any of the
injunctions is an explanatory sacrifice, and should never be un-
dertaken without motivation.
5
Complex dispositions to behave may be constitutive of belief, but that is an-
other matter.
When we explain behaviour by attributing beliefs and desires

we have to assume an enabling background of other beliefs and
desires. Explaining the utterance in our example by postulating
a desire for salt makes sense only if we assume the agent be-
lieves that salt will be of some benefit, that there is salt in the
shaker, that "salt" refers to salt, and so on. So in constructing
an explanation of any particular behaviour we have to beware
of postulating a background that conflicts with the explanation
of other behaviour. That way, we end up ascribing a system of
beliefs to the subject, and the reasonableness of ascribing one of
the beliefs in the system will depend partly on its place in that
very system. For part of what makes it reasonable to attribute a
belief to someone is that the belief coheres well with other beliefs
we attribute to that person, perhaps on the basis of more direct
behavioural evidence.
Explanatory purpose, the search for coherent pattern, the ho-
lism of confirmation: these features of the interpretation of be-
haviour are mirrored within the project of literary interpretation.
An interpretation is an attempt to explain the text, as the attri-
bution of belief is an attempt to explain behaviour. As with be-
lief attribution, textual interpretation may serve a predictive
role, in a som~what extended sense of prediction. An interpre-
tative hypothesis formed on the basis of a reading of an initial
segment of the text may suggest ways in which the text is likely
to develop; we may find its confirmation or disconfirmation
later in the text. But because interpretation is explanatory, there
are constraints on interpretation that go well beyond mere con-
sistency with the text. It may be possible to interpret the Sher-
lock Holmes stories as the deluded ramblings of Doctor Watson,
who never in fact met any detective called "Holmes". Such an
interpretation would, let us suppose, be consistent with the ev-
idence of the text taken as a whole, but it would not be explan-
atory of it any more than if I tried to explain the behaviour of
an apparently rational person by supposing her to be in the grip
of massive delusions that compensate for one another so as to
produce rational-seeming behaviour. We sometimes accept un-
reliable narrator interpretations, but only, as with the likes of
237
Interpretation
Nabokov's Pale Fire, when they make plausible and economical

sense of the text. 6
We decide what is true in the story according to how well
our interpretation explains the text, but we do not proceed cu-
mulatively, pairing off sentences of text with propositions true
in the story. As I have noted, some of what is written in a story
ought not to be taken literally. Fictional descriptions may con-
tain metaphor, irony or some other nonliteral device, or we may
be in the presence of an unreliable narrator. What we take lit-
erally and what we don't depends on the overall impression the
text makes on us. And much that is true in the story may not
be stated at all, forming a kind of implicit background. What's
to be taken as background and what not depends, once again,
on the overall character of the text?
As with claims about belief, claims about what is part of the
story are not to be thought of as corroborated individually by
the text, but only in so far as they belong to a system of such
claims. The reasonableness of one assumption about what is part
of the story depends on its place in a system of such assump-
tions which make good overall sense of the text. A single claim
might be poorly corroborated within a given set of interpretative
hypotheses, yet well corroborated within another. Consider the
claim that it's true in The Turn of the Screw that the governess is
mad. Whether we regard this as well corroborated depends very
much on what other interpretative assumptions we are prepared
to make - that there are no ghosts, that the children are innocent,
and so forth. Some interpretative claims will not be well corrob-
orated whatever other assumptions we make: that the governess
is a figment of Mrs Gross's imagination might be an example.
Others will form part of any acceptable interpretation - that
Miles and Flora are children rather than adult dwarves, for ex-
ample.
I summarize: literary interpretation aims to figure out what
is true in the story - a project not unlike, from both an evidential
and a constitutive point of view, that of figuring out the mental
6
For more on unreliable narrators see Chapter 9·
7
For more on this point see my Nature of Fiction, chapters 2 and 3·
states responsible for a person's behaviour. The next section
brings these two projects more closely together.
8.4 INTERPRETING WORKS AS

INTERPRETING BEHAVIOUR
I've argued that interpreting behaviour is a matter of explaining

it by intentional causes, and that interpreting works is con-
strained and guided by considerations similar to those that con-
strain and guide the interpretation of behaviour. But that merely
sets the scene for the central claim of this chapter: that inter-
preting literary works is a species of the interpretation of behaviour.
Put these ideas together and we get: interpreting literary works
is a matter of explaining them in causal and intentional terms.
How can that be, given that behaviour is one thing and text
definitely another? Start by observing that a text is always the
outcome of behaviour; it is a trace left on the world by an agent.
Another such trace might be a room full of broken-up furniture.
Why is the furniture broken up? Because Smith, angered at the
actions of its owner, broke it up. Explain Smith's behaviour (in
terms of his desire for revenge, his belief that the loss of the
furniture would be grievous) and we explain the outcome as
well. Just so with the text; we explain it when we explain the
behaviour of which it is the outcome. We interpret it when we
formulate a certain kind of hypothesis about its cause.
There are hypotheses we could formulate about the text's
cause that would not count as interpreting. We might, if we
had the knowledge, causally explain a text by citing the de-
tailed pattern of microphysical events in the author's brain,
body and near environment that resulted in the formation of
marks on a surface which constitute the original inscription of
that text. But at that level of analysis we could not hope to re-
cover intentions in a systematic way. My claim is not that in-
terpreting is the only kind of causal explaining of texts there is,
but that it is one kind.
We need to exclude more than just brute physical causality.
There are hypotheses we could formulate about the text's inten-
tional causes that would not count as interpreting the text. The
239
Interpretation
behaviour that generates the text might be the product of vari-

ous causes, some of them overdetermining. The text might have
been produced with the intention of paying the rent or acquir-
ing a literary reputation. Those intentions, causally efficacious
though they might be, won't help us to our interpretive goal,
which is to know what story the text has to tell. We as inter-
preters are concerned with story-telling intentions; the author
intends to tell a story in which such and such happens. So in-
tending, he chooses words to implement that intention: words
which, so he believes, will suggest that story to the reader. So
interpretation is explanation by reference to causally efficacious,
story-telling intentions. 8
Earlier I remarked that there are certain things which we are
relatively sure are part of the story of James's The Turn of the
Screw, for example, that Miles and Flora are children. We can
now say why we are as sure of this as we are. Not, I take it,
because the text says they are children; it might tum out that
an examination of the text does not reveal any such explicit
statement, and even if it does I doubt whether our assurance on
this point is due to the fact that we have noticed such a state-
ment. Anyway, an explicit statement would not by itself clinch
the matter. Lots of things explicitly said in novels need to be
taken nonliterally, or are to be ascribed to an unreliable narrator.
The reason we are relatively sure about the status of Miles and
Flora is that what James does say, taking the story as a whole,
overwhelmingly suggests that his intention was to tell a story
part of which is that Miles and Flora are children. Conversely,
an interpretation according to which the governess is a figment
of Mrs Gross's imagination would not have much plausibility,
because the text as a whole makes this a very unlikely hypoth-
esis about how the author intended his story to be taken. And
8
This argument is a little too quick, at least for those of us who think that
mental events are actually causally inert; we think that they mimic relations
of causal dependence by supervening on causally efficacious states of the
brain. But information about mental states can count as causally relevant in-
formation even when those mental states are causally inefficacious. See Frank
Jackson and Philip Pettit, "Functionalism and Broad Content".
240
where there is room for doubt, concerning, say, the reliability of

the governess's account of the ghosts, there is corresponding
doubt about how the author would have us take his story at
this point.
I have said that, in coming to a view about interpretation, no
single piece of text, considered on its own, can be decisive one
way or another. But this should not be taken to imply that the
text is in any sense unimportant for interpretation, that it can be
selectively ignored or impressionistically treated. The point is,
rather, that it is always the whole text that functions as evidence
for an interpretive hypothesis. And what would, if taken in iso-
lation, seem to be evidence for one interpretive hypothesis might
turn out to be evidence for a contrary hypothesis when seen in
context.
The text as a whole is our evidential basis for an interpreta-
tion, and it is the text as meaningful linguistic item that is in ques-
tion here - the text as a sequence of words and sentences with
certain literal meanings. Our question is not "Why did the au-
thor produce this sequence of marks on paper?" but "Why did
the author produce this sequence of words and sentences with
these meanings?" If we began by asking the first question, we
could never hope to arrive at an interpretive hypothesis- unless
we went for a two-stage interpretation that interpolated literal
meanings between marks and story-telling intentions. But then
we might as well start with the literal meanings as part of the
evidence itself. And we saw in Chapter 4 that literal meanings
are things we may ascribe to the words and sentences that
speakers utter at a stage prior to that of interpreting their utter-
ances.
So interpretation is explanation, and the thing to be explained
is the text, a meaningful sequence of words. The text is a trace
left on the world by an intending agent in the act of telling a
story, the story we are trying to reconstruct. We explain that text
- we say something about its causal history - when we say
something about the author's story-telling intentions, for those
intentions were among the causes of its coming into being. The
hypothesis we arrive at, if it is correct, specifies the content of
241
Interpretation
the author's intention, and the content of his intention is the
content of the story.9
On this view, narrative interpretation turns out to be just a
special case of ordinary, conversational interpretation, the inter-
pretation of the casual remarks of speakers we encounter every
day. We figure out the meaning of a speaker's utterance when
we figure out what the speaker meant by uttering it - when we
figure out, in other words, what intentions the speaker had that
caused him to make that utterance. The difference between the
conversational and the literary case is just a matter of the length
and complexity of the text concerned; of its having, in the lit-
erary case, generally a written rather than a spoken form; and
of the consequent leisure that the relative permanence of writing
affords us for reflection in our interpretive efforts.
In Part II I argued that fictions are things we typically respond
to by engaging in simulations; fictions provide the inputs, or
some of them, for off-line simulations. But in order to know
what inputs are appropriate - should I give myself the pretend
belief that the governess sees ghosts, or the pretend belief that
she is mad?- I need to do some interpreting. A decision about
what the story is determines the contents of my pretend inputs.
But this suggests that fictions may encourage simulation in an-
other way: I need to know what the story is, and to do that I
need to know what story-telling intentions the author most
probably had. The best way to carry out the latter task may be
to do some simulating; I put myself in the author's position,
thinking of myself as the person who produced the text, and
ask myself, ''What story-telling intentions would have led me
to write that text?" If I can run a simulation in which having
those story-telling intentions leads to the (off-line) decision to
write this very text, then I can say that it is at least initially
plausible that the author did have those intentions. That will not
be the end of the matter; reflection on the historical and cultural
differences between the author and myself may convince me
that my simulation did not do enough to correct for differences
• This, of course, is the interpretive ideal, which acts of interpretation never

more than approximate.
in our perspectives. But still, simulation is a way - and in some
cases it may be the best way - to generate an interpretive hy-
pothesis. In that case, fictions encourage different kinds of sim-
ulations: a simulation in which I try to put myself in the position
of the author, and one in which I put myself in the position of
one who is learning facts from a reliable source, rather than one
who is reading or viewing a fictional work. Add to this another
kind of simulation I said was often required by fiction: simula-
tion of a character's state of mind. Fictions can be vehicles for
complex, multilayered simulations, and parallels and tensions
between the layers might be an interesting source of aesthetic
experience.
8.5 REAL AUTHORS, IMPLIED AUTHORS
Let us call the idea that the interpreter's task is to discover the
author's intended meaning for the text "Real Author Intention-
alism", or RAI. It will soon be apparent why I choose this name.
RAI is scarcely a popular option among theorists of criticism,
and some of the arguments against it are good ones. For one
thing, RAI denies the gap between aspiration and performance,
making it impossible for the author to fail to tell the story he
intends to tell. The story he does tell is, according to RAI, just
the one he intended to tell, and it is consequently the job of the
interpreter to discover what his intentions were. But authors
can, and sometimes do, fail to tell the stories they intend to tell.
The author can think that enough has been said to indicate a
certain development of plot or character, and be mistaken, be-
cause readers cannot reasonably be expected to grasp that de-
velopment on the basis of the text he gives them. In that case
the story he intends to tell is not, in at least some respect, the
story he tells.
Defenders of RAI sometimes claim that the author's inten-
tions to which we may legitimately appeal are not merely pri-
vate mental occurrences, but embodied intentions: intentions
which could be understood by a suitably receptive reader on the
basis of a reading of the text, and without access to information
Interpretation
concerning the author's private opinions. So readings privately

10
intended but not embodied will not count, for interpretive pur-
poses, as intended at all. But the problem is not merely how to
exclude interpretations that might have been in the author's
mind but never got into the work- the problem the embodiment
condition is supposed to solve - but how to allow, on realist
grounds, for interpretations which never were in the author's
mind at all. For example, some readers of The Possessed have
found Stepan Trofimovich to be a redeemed character, but Dos-
toyevsky' s notebooks do not indicate that he intended him to
be seen in this light. Yet there are some intuitive grounds for
saying that readers who interpret the story that way are onto
something. How could the intention for us to see Stepan as
11
redeemed get embodied in the work if it was never in Dostoy-

evsky's mind? If you say, as enthusiasts for RAI sometimes do,
that intention is not a matter of what is in the mind, and that
the test of what is really intended is whether it can be found in
the work itself, then you have given up RAI in all but name.
Intentions no longer drive the system; they are now constructed,
and not merely inferred, entities that play no role in figuring
out the right interpretation. Rather, we arrive, somehow, at the
right interpretation, and announce that it is the intended - be-
cause embodied - one.
But while I insist on the psychological reality of intentions, I
do not insist on their being conscious states of mind. The best
current philosophical thinking about intentions emphasizes their
being states capable of causing, and thereby explaining, be-
haviour, but is far from insisting that these states always be
transparent to the agent. So it is a distracting side issue that
supporters of RAI raise when they sometimes point out that the
relevant intentions need not be conscious. The problem for RAI
is not that we sometimes endorse an interpretation of which the
author was not consciously aware at the time of writing, but
rather that we sometimes endorse interpretations which may
'" See, e.g., Hirsch, Validity in Interpretation, especially p. 31, the definition of
"verbal meaning".
11
See Seymour Chatman, Coming to Terms, pp. 97--99, for an instructive dis-
cussion of this example.
very well not have been intended at all, as with Dostoyevsky
and Stepan the redeemed. To insist that these interpretations
were unconsciously intended merely because they seem in other
ways right is once again to operate a mechanism with the wheel
marked "intention" running idle.' 2
If we abandon RAI, as I think we should, need we abandon

the idea that interpretation aims at an hypothesis about the in-
tentions behind the text? No. It need not matter, from the point
of view of explanatory adequacy, that the intentions in terms of
which we explain the text were not those of its author. We may
explain the text, and explain it adequately, by reference to a
personality that seems to have produced it, as Dostoyevsky's
story seems, in my example, to have been the product of
someone with intentions different from Dostoyevsky's own.
This personality corresponds to what some critics have called
the "implied", "apparent" or "postulated" author.' 3 Thus I re-
place RAI with Implied Author Intentionalism, or IAI.
So our interpretation records our decisions about what story-
telling intentions can reasonably be supposed to be among the
causes of the text. These intentions we attribute to the implied
author. But the implied author is not some shadowy entity with
a grade of existence somehow lower than that of the real author.
The implied author is a heuristic device that no one need believe
in, and reference to him is easily eliminated; that the implied
author intends P to be fictional means just that the text can rea-
sonably be thought of as produced by someone intending the
reader to recognize that P is fictional.' 4
Notice that RAI faces a problem about the relation of the text
" For a defence of realism by appeal to unconscious intentions see P. D. Juhl,
Interpretation, section 6.2.
'
3
See Booth, Rhetoric of Fiction. For applications see Alexander Nehamas, "The
Postulated Author: Critical Monism as a Regulative Ideal"; Jenefer Robinson,
"Style and Personality in the Literary Work"; and Bruce Vermazen, "Ex-
pression as Expression".
'
4
Seymour Chatman (Coming to Terms, p. 81) says that "the text is itself the
implied author". Taken literally, this claim is incomprehensible. But I do not
know in what nonliteral way to take it, and Chatman does not tell us. Per-
haps the account in the passage above this note is one such way, in which
case I can agree with Chatman.
245
Interpretation
to the story. If we take our goal to be finding the real author's
narrative intentions, we have to acknowledge that authorial er-
ror can make the text not the best guide to what those intentions
are, and interest may shift from the text to psychobiography.
But no amount of deconstructive rhetoric should make us forget
the centrality of the text. In interpretation, we make inferences
to narrative intentions so as to illuminate and render coherent
the text we have before us, not some other text the author might
have written, one perhaps more in tune with her actual inten-
tions. It is possible for an author to be moved by certain nar-
rative intentions, but fail to give proper effect to them in her
text-tokening activity: she can be wrong in thinking that enough
has been said to indicate a development of plot or character,
and a reading made most plausible by her text can be one she
fails to comprehend. When that happens (as in small ways it
commonly does) the path recommended by RAI- to infer, from
her diaries and acquaintances, the narrative intentions the au-
thor actually had - will not, by assumption of authorial error,
illuminate and make coherent the text we have. The interpreter
who takes that path has ceased to be an interpreter of the work,
and his intentional hypothesizing serves rather to interpret an-
other, hypothetical work- the one that would have been written
had the author's narrative intentions gone well. By contrast, IAI
can guarantee the centrality of the text, irrespective of authorial
error. According to IAI, the task is to hypothesize narrative in-
tentions which make as good sense of the text we have as can
be made of it, irrespective of whether they were the real author's
intentions or not.' 5 Since IAI has us look for the interpretation
which makes best sense of the given text regardless of the psy-
chological facts, it gives the text a properly constitutive as well
as an evidential role.
5
' "As can be made of it" because it is not always possible to avoid imputing
some degree of authorial incompetence and textual incoherence. The dis-
crepant reporting of Natasha's age in War and Peace could be reconciled by
assuming Tolstoy intended us to infer that she is a time traveller, but that
would clash with the conclusion, well-supported by other aspects of the
work, that no fantasy or science-fiction elements are intended in this story.
IAI demands only that we maximize coherence.
246
8.6 WHY WE STILL NEED INTENTION
Having admitted that the real author's intentions are not deci-
sive for interpretation, why insist that the interpreter's task is to
construct another, this time hypothetical, author? Why not say
instead that the interpretive problem is simply to go from what
is explicit in the text to what is implicit, where what is implicit
might not correspond to anything anyone intended? But there
can be no meaning implicit in the words and sentences of the
text itself; the meaning that there is in the text itself is all explicit,
literal meaning. What we call implicit in the text is what we
think of as intended (perhaps, indeed, unconsciously intended)
by an agent who hopes to convey by his words more than the
words themselves literally mean. A text can encourage in the
reader's mind beliefs about what those intentions were even
when, as a matter of fact, the real author had no such intentions.
IAI says that it is what the text makes it reasonable to believe
was intended, and not what actually was intended, that deter-
mines the content of the story.
There is another reason we can't ignore intention. Narratives
obstruct, mislead and manipulate us by their selective presen-
tation of events. But no text, no sequence of visual images, can
do those things; such things require agency. Most of the ways
we describe narrative make no sense if we cut narrative off from
the agency which produced it. Writers on narrative sometimes
ignore this. Thus David Bordwell, speaking of film:
As for the implied author, this construct adds nothing to our
understanding of filmic narration. No trait we could assign to
an implied author of a film could not more simply be ascribed
to the narration itself: it sometimes suppresses information, it
often restricts our knowledge, it generates curiosity, it creates
a tone, and so on. To give every film a narrator or implied
author is to indulge in an anthropomorphic fiction ....
[Filmic] narration is better understood as the organization of
a set of cues for the construction of a story. This presupposes
a perceiver, but not any sender, of a message.' 6
6
' Bordwell, Narration in the Fiction Film, p. 62
247
Interpretation
Bordwell speaks of a narration which "suppresses information".
There is no literal sense in the idea of an unnarrated narration
which suppresses information. Without recourse to the idea of
intention you can speak of a system that fails to deliver all the
information you want, but not of a system that suppresses infor-
mation. And the idea of suppression (rather than just of infor-
mational incompleteness) is essential to Bordwell's description
of filmic narration and our reaction to it; we feel, in some cases,
that we are being deliberately deprived of information (as when
we see only the hands of the murderer), that we are being de-
prived of it for some dramatic or emotional purpose, that our
expectations are being played with. None of this would make
sense unless we understood the story as told.
If Bordwell's claim has any plausibility for the filmic case, I
believe it derives ultimately from what I have called the Imag-
ined Observer Hypothesis. If we think of the viewer's imagi-
native involvement in the movie as requiring him to imagine
that he is viewing the action from within the space of the film,
then of course it would be out of place for him to suppose that
the images he receives are being presented to him by a control-
ling intelligence that presents them in a certain way and in a
certain order to tell a certain story by their means. But if we
reject this view and think of the viewer's imaginative engage-
ment with the film as characterized by perceptual imagining
rather than by imagining seeing, there is no such objection to
the idea that cinematic fictions are experienced as mediated; in
fact we are then obliged to think of them in that way.
Those who announced an intentional fallacy were not so far

wrong. They were right to think that it is not authorial intention
that determines the meaning of the work. They were wrong to
think that interpretation could dispense entirely with the con-
cept of intention.'7 All interpretation is intentionalistic; the issue
7
' Some writers on the psychology of narrative have emphasized the impor-
tance of so-called schemata for interpretation: knowledge structures which
encapsulate a set of standard conditions pertaining to a given situation (air
travel, restaurant dining, etc.) and which we use to fill in the details of the
story from the sketchy outline of the text (see, e.g., J. M. Mandler, Stories,
248
is whether we should concern ourselves with the real intentions
behind the work, or with the intentions which seem to have
been productive of the work.' 8 I say our concern is with the
latter.
8.7 INTERPRETIVE DEADLOCK AND

TRUTH
An interpretation of the text is a hypothesis about the intentions
which seem to have been productive of the text. So the best
hypothesis about what those intentions are is the best interpre-
tation of the work. But two or more hypotheses may be equally
good, and none better (they are, as I shall say, maximal, rather
than best, explanations; there can be only one best explanation).
That gives us a plurality of interpretations which can, with equal
plausibility, be thought of as intended, and nothing else more
plausible. If that happens we have interpretive indeterminacy-
rival interpretations with nothing to help us choose between
them. To suppose that we can choose between them on the
grounds that evidence from the author's diary or letters suggests
that one of these hypotheses corresponds to his intentions and
the other doesn't is once again to embrace RAI, which I have
rejected.
Will such indeterminacy actually occur? We are familiar, of
course, with interpretive disputes. But it is always hard to say
of any such dispute that it really is a case of indeterminacy; an
unresolved dispute is not automatically an unresolvable one. I
think there can be conflicting maximal interpretations of a text -
interpretations which can with equal plausibility be thought of
as intended. A single brief remark can be genuinely ambiguous;
without further interrogation, we can't choose between two or
Scripts and Scenes: Aspects of Schema Theory). Such schemata are important for
interpretation, but their importance is primarily in providing us with evi-
dence about how the author would probably have intended the story to be
filled out.
8
' A point made by Ronald Dworkin in Law's Empire. Some of my conclusions
in this essay have been influenced by Dworkin's work on interpretation (see
also his Matter of Principle).
249
Interpretation
more hypotheses about what the utterer meant by it.' 9 You

might argue that such cases would be the product of textual
brevity, that plausible alternatives would be eliminated as the
text in question gets larger (and we rarely want to engage in
narrative interpretation on a one-line text). But textual increase
does not generally correlate with a reduction in interpretive un-
certainty. Ambiguity can be removed, or reduced, if the textual
additions simply tell us more about the subject of the original
remark. But where text functions to convey narrative, that is not
in general the case. More text, more story to be interpreted and
a textual addition that's narratively significant may close an in-
terpretive option at one end, but will very likely open one at the
other.
So I claim that IAI allows for unresolvable hard cases of in-
terpretation. It does not say anything about how frequent they
are in theory or in practice. It does not identify any of the cel-
ebrated interpretive disputes there have been as examples of
interpretive indeterminacy; perhaps these disputes are all the
product of confusion, ill will or, less culpably, the uncertainty
that attends any judgement of a large body of evidence. Still, it
does caution us against writing off all interpretive deadlock in
those ways.
Earlier I drew attention to the fact that I was speaking of
interpretations as capable of being true or false. But if there are
to be clashes, unresolvable in principle, between interpretations,
we can hardly say that one or other of the rivals is true and the
other false. This is not to say that interpretive claims never have
truth value. Interpretations consist of large numbers of interpre-
tive claims about what this character did and when, what that
character thought about it, and so on, and not all of them are
equally controversial. In that case a truth-valueless interpreta-
tion may have some truth-valuable constituents. Fred and Freida
are deadlocked about how to interpret The Turn of the Screw;
having exhaustively examined the evidence that both regard as
9
' This can be true even when the remark is semantically unambiguous; context
can make the hypothesis that the literal meaning of the sentence was the
intended one equiprobable with some hypothesis that the intended meaning
was something else.
relevant to deciding the issue, they can find nothing that tips
the balance either way. But still, Fred and Freida should not be
represented as disagreeing about everything in the interpreta-
tion of that story. Their rival interpretations may, and probably
will, have a significant common part. Things in that common
part, or some of them, may be uncontroversial; they will belong
to any interpretation of the story which hopes to withstand crit-
ical scrutiny. Thus they agree that Flora and Miles are children,
that Mrs Gross is real and not a figment of the governess's imag-
ination. (Or rather, they agree that all these things are fictional.)
And perhaps every well-informed and reflective critic will agree
with them. Those things which constitute the common core of
any acceptable interpretation ought to count as true if anything
does. In that case it will be straightforwardly true that it is fic-
tional that Miles and Flora are children. Similarly, claims which
no well-informed, reflective interpreter would offer - that Mrs
Gross is imaginary - will count as false. In general, an interpre-
tive claim is true if it belongs to every maximal interpretation
of the text, and false if it belongs to no such interpretation. That
way the uncontroversial and the hopeless interpretive claims
(Hamlet was human, Hamlet was Venusian) come out true and
false respectively, while the controversial claims lack truth
value, for they belong to some maximal interpretations and not
to otherS. And where there is- if there ever is- one maximal
20
(and therefore best) interpretation, it comes out true, since all its
constituents come out true.
8.8 THE EVIDENCE FOR A CINEMATIC

INTERPRETATION
Interpreting the literary work begins with the text - a sequence

of meaningful words and sentences. The interpreter's task is to
work out what story the text has to tell. That is to be accom-
plished by discovering what story-telling intentions can most
plausibly be thought of as productive of the text. Text is, there-
fore, the basis of, and the evidence for, a literary interpretation.
"" For more detail see my "Interpreting Fiction".

Interpretation
Interpretation starts with the text and succeeds to the extent that
it explains the text as the outcome of story-telling intentions. It
seems, then, that the model here proposed fails for cinema
which, as I have argued, lacks a language.
But the nonexistence of a language of film should not lead us
to abandon our basic model of interpretation as causal-inten-
tional explanation. Conventional, textual meaning is where we
start in interpreting the literary work, not where we end. We
end with a system of inferred intentions. Intentions can be in-
ferred from artefacts of all kinds, not just from linguistic ones.
Anthropologists infer intentions from the artefacts they discover;
that is, they infer the uses intended for them by their makers.
A film is an artefact, consisting of cinematic images and their
verbal and other accompaniments - a highly complex, struc-
tured. artefact. Usually, its context of production is familiar to
us; we live in, or near, the culture that produces the films. So
making a stab at the relevant intentions based on a viewing of
the artefactual material itself ought not to be so very difficult.
In fact, I claim, we do it all the time. We may do it in ways
rather different from the ways we interpret literary works, but
the basic strategy is the same; to work out the story the film
presents by inferring story-telling intentions from the images or
other clues that the film gives us.
What we need is something that will play the same, or a
similar, role for the interpretation of film that the literal mean-
ings of words and sentences play for literary interpretation. That
thing, whatever it turns out to be, will be the thing which needs
explaining in causal-intentional terms if we are to interpret the
film. It will be our evidential base, as the text is our evidential
base for literary interpretation. Of course we want, if possible,
for our base to be richly structured and have fine gradations of
texture. After all, our task is to infer a large and complex set of
very precisely specified intentions; the more complex the evi-
dential base, the more likely it is that we shall be successful
in this. Ideally, we should like to be able to say that our eviden-
tial base consists of elements which have meaning, or content,
of some kind: if we start with meaning we shall be more like-
ly to end up with meaning. With texts, we start with literal
252
meaning and end up, via our process of causal-intentional ex-

planation, with utterance meaning. Since there is no language
for cinema, we cannot, in the cinematic case, begin with lin-
guistic meaning. But perhaps there is something we can identify
as the cinematic equivalent of linguistic meaning: equivalent,
that is, from the point of view of an evidential role. Let us call
this kind of meaning, whatever it is, evidential meaning. The task
before us is to say what that kind of meaning is. To simplify
matters I shall ignore sound and concentrate on the evidential
meaning provided by the film image; what I shall say about
meaning in connection with images can be generalized to the
sound case without much difficulty.
What is the best candidate for the evidential meaning of the
cinematic image? One natural proposal is that it is just what the
image records. What the image records is actors performing
actions among props on a set, or on location. If this "photo-
graphic" meaning is what we are to call the evidential meaning
of cinematic images, how much like the literal meanings of
words and sentences in natural language is it? (Recall the char-
acterization of literal meaning in Chapter 4.) It is like literal
meaning at least in this: it is acontextual. The photographic
meaning of a cinematic image does not depend on its relation
to other images, because that meaning is entirely determined by
what was in front of the camera at the time of the take. And by
juxtaposing images one simply gets an accretion of meaning: if
the meaning of image A is M(A) and that of image B is M(B),
then image A followed by image B just means M(A) & M(B),
where the order of juxtaposition is irrelevant to meaning; show-
ing B after A does not mean, in the sense of meaning at issue
here, that the events that A records occurred before the events
that B records. The meaning, in this sense, of a complex of im-
ages is just the sum of the meanings of its constituent images.
And that's how it is with linguistically formulated assertions;
saying A and saying B just commits you to the truth of A and
B, irrespective of the order of saying. 2 '
" In both cases - the filmic and the linguistic - the order of production can
have consequences for what is implicated, in Grice's sense.
Interpretation
This promising beginning notwithstanding, I want to argue
that photographic meaning cannot play the same kind of evi-
dential role in cinema that the literal meanings of words and
sentences can play in interpreting literary fictions. There are two
reasons for this.
The first is that photographic meaning does not bear on the
story in cinema in the same way that linguistic meaning does
in literature. In literature the text consists of words and sen-
tences which describe events within the fictional world; they can
be thought of as doing that in virtue, partly, of the literal mean-
ings they have. Thus if the text says, "Holmes and Watson spent
the evening at Baker Street", we may decide to take that sen-
tence as a literal description of the activities of the characters on
a particular occasion. We might decide instead to take that state-
ment nonliterally, as metaphor or irony, or even as the words
of an unreliable speaker. In that case we would not take the
literal meaning of the sentence as a reliable guide to what hap-
pened in the fiction. But the literal meaning of the text can be,
and sometimes is, a reliable guide, and we always need to know
what the literal meaning is.
But with cinema it is different. We cannot ever take the pho-
tographic meaning of an image to coincide with what happens
in the story at that point. For the photographic meaning is al-
ways something to do with real people (actors, mostly) perform-
ing real actions in the real world. There cannot be a photograph
of something that does not exist, or of an act no one ever per-
formed. But the characters of the fictional film are, at least typ-
ically, characters that do not exist, and the events of the fictional
story are, at least typically, events that do not occur. There is
always a disparity between photographic meaning and what is
true in the fictional story.
The second reason photographic meaning cannot be regarded
as the evidential analogue of linguistic meaning is that viewers
are sometimes unaware of, or mistaken about, the photographic
meaning of a cinematic image. But their being unaware of it or
mistaken about it does not automatically count as evidence that
they lack some piece of knowledge vital to working out what is
going on in the story. A shot may be a trick shot; it may seem
254
to show a man falling a distance that no man involved in the
filming ever did fall. It may seem to show a fantastical creature,
or a man walking on water. Or it may be a shot that seems to
be a distant or blurred view of the main actor but is actually a
cinematic image of a stand-in. In such cases the members of the
audience will sometimes know that there is some trick involved,
but they will seldom have any idea what the trick is; they will
have little idea, that is, about what the camera actually records.
They will, in those cases, therefore have little or no idea about
the photographic meaning of the shot in question. But if there
is a doubt in the minds of the audience about what is happening
in the story - or a disagreement about it - it is implausible to
suppose that it will be resolved by teaching them more about
the cinematographer's tricks of the trade. By learning those
tricks they might learn something about cinema in general, but
they would not be put in a better position to work out what is
happening in the story. In the case of literary interpretation, on
the other hand, it is always true that a reader who lacks knowl-
edge of the literal meaning of a passage in the text will be helped
in forming a reasoned opinion about what is going on in the
story by learning what that literal meaning is. To have the text
and not to know its literal meaning is to be like a monolingual
Chinese reader trying to make sense of a novel in English. So
what I have called the photographic meaning of cinematic im-
ages cannot play the role of evidence for an interpretation in the
way that the literal meaning of words and sentences can.
However, while the photographic meaning of the image may
be unclear to the viewer, it is at least usually possible for the
viewer, given the right background of cultural knowledge, to
say that this image appears to represent a man falling off a build-
ing, or a creature of fantastical appearance, or a man walking
on water; you can make a decision about what the image ap-
pears to represent without knowing a great deal about the con-
text of surrounding images that occur in the film. It may be that
a certain image functions within the diegesis to represent a
dream, a lie or an hallucination, and that it does so can be dis-
covered only by seeing and interpreting the whole film. Still,
one can say of the image, pretty much on the basis of viewing
255
Interpretation
it alone, that it seems to represent a woman running down a
corridor while arms protrude from the walls on either side, or
whatever. It would be possible, if anyone could be bothered to
do it, to establish meaning at this level of "seeming" or appear-
ance for any particular shot by finding a viewer who (i) had no
knowledge of the rest of the film, and (ii) could not identify the
actors in the film and generally knew nothing about its construc-
tion, and then getting this person to describe what the image
represents. The answer given, assuming the viewer is alert and
cooperative, would correspond roughly to what I shall here call
"appearance meaning". There would no doubt be some varia-
tion in the detail of each individual's description, reflecting dif-
ferent levels of attention and visual acuity, but there would in
general be a large measure of agreement between viewers about
what the image appears to represent. It is this possibility of in-
tersubjective agreement that justifies our saying that here we
have another kind of meaning for the image - appearance mean-
ing.
It is appearance meaning, and not photographic meaning,
which plays the role in the interpretation of film that literal
meaning plays in the interpretation of literature. It is the data
to be explained. Appearance meaning is evidence meaning. The
shot shows what appears to be children being attacked by a
flock of birds (though in fact it may be a trick shot that does
not record any such event). The question the interpreter needs
to ask, then, is: why is that shot, with just that appearance, in-
serted in the film just at this point, between these other shots?
The answer will be: because the insertion of that shot was in-
tended to tell us something about what is happening in the
story, and how that event stands in relation to others in the
story. But while appearance meaning is, for film, the analogue
of literary, textual meaning, it should be evident from what
was said in Chapter 4 that appearance meaning is not under-
stood in anything like the way that linguistic meaning is. There
are no atoms of appearance meaning out of which are built
more complex units of meaning, for every visible part of the
image is meaningful. Yet we are able to understand what is
represented in images we have not seen before. Appearance
256
meaning displays the natural generativity possessed by all pic-
torial kinds of meaning.
When we see films we are generally not conscious of what I
have called appearance meaning, but this is not grounds for
thinking it plays no role in our interpretive strategies. We are
rarely conscious, after all, of the literal meaning of the text we
read. If we could not identify each shot, as we see it, at least
roughly at the level of appearance meaning, we would find it
impossible to interpret the film: we would not be able to rei-
dentify characters and locations from one shot to another; we
would not be able to say what kind of action was being under-
taken or what kind of event was happening; and we would have
no expectations about what would follow, and so no sense of
the rightness or disparity of succeeding images. To see a film
and not be able to specify its meaning at the level of appearance
would be like looking at the pages of a book written in an un-
known language; in neither case would it be possible to interpret
the work.
Sometimes appearance meaning is unclear. If the scene is very
dark, or if the location or action undertaken is very unfamiliar,
we may not be able to judge the appearance meaning of the shot
as we see it, or we may be heavily dependent on the context of
surrounding shots and the interpretation of them we have so
far built up in order to judge it. For example, the novice viewer
of The Searchers will not immediately understand that the uni-
formly dark screen at the beginning has any appearance mean-
ing at all; it is only when the door opens to reveal an exterior
scene that she will realize that the featureless screen represented
a darkened interior. But the same can happen with a text. We
do not understand every sentence we read, and some texts con-
tain sentences that deliberately challenge or defy understanding.
But this has no tendency to show that textual meaning is irrel-
evant to understanding. If the text entirely lacked meaning, or
belonged to a language you didn't understand, you wouldn't be
able to interpret it. If no image in the film had any appearance
meaning, you wouldn't be able to interpret that either.
So filmic interpretation is like literary interpretation in this: it
takes as input certain data, and gives as output a hypothesis
257
Interpretation
concerning what story a rational agent would most probably
have intended to communicate by means of that data. Filmic
interpretation is subject to the same vaguenesses, ambiguities
and indeterminacies that attend literary interpretations because
in both cases the relation between data, hypotheses and the prin-
ciples that govern inferences to intention are in various ways
vague, ambiguous and indeterminate. But in neither the literary
nor the filmic case is every interpretation as good as every other
one. It is just that there is not always an interpretation which is
uniquely best.
Filmic interpretation differs from literary interpretation both
in the nature of its data inputs and the competencies required
for processing them. In the literary case the data is linguistic
meaning and the required competence is with the syntax and
semantics of the language. In the filmic case the inputs are cin-
ematic images correctly juxtaposed; the required competence is
the ability to understand the appearance meaning of those im-
ages, a competence based on the possession of visual capacities
to recognize objects, their properties and relations.
8.9 IMPLIED AUTHOR AND AUTEUR
On this model of interpretation, the idea that the cinematic work

is created by an intelligence is central; interpretation is the assign-
ment of intentions to the implied author. There are writers on film
who are fond of telling us that films are the product of many peo-
ple with different roles and skills, that the resulting product
rarely if ever reflects the artistic or narrative conception of one
person, that the constraints of budget, technology, studio control
and prevailing ideology all play their part in determining what
the product is like. Usually this is said in response to the so-called
auteur theory, according to which we should judge films by the
extent that they reflect the individual vision of their directors.
What I have argued in this chapter is in no sense a version
of the auteur theory. It is possible to accept what I have said
here about interpretation and also to think that the most useful
groupings among films for purposes of classification, analysis
and criticism are groupings by genre, period or place of origin.
258
It does not follow from the view I have been advocating that
we should look for, still less that we should expect to find, a
common implied author for all of Hitchcock's films, another for
all of Douglas Sirk's. (Also, it doesn't follow that we shouldn't;
my point is that these issues are independent.) It does not even
follow that we should expect to find, for every film, an inter-
pretive hypothesis that attributes story-telling intentions to a
single implied maker; a film, and indeed a fictional work of any
kind, can be such that it seems to be the product of more than
one agent, in which case we must speak of "implied makers".
The film may be such that these implied makers seem to have
shared or even communal intentions,22 or it may be that they
seem to have intentions that to some extent diverge and even
conflict. Intentionalism, as long as it is Implied Author Inten-
tionalism, is neutral on issues raised by the auteur theory.
8.10 STRUCTURE AND FUNCTION

In Chapter 4 I argued that cinematic and other images, and their
modes of combination, are quite unlike a language. In this chap-
ter I have argued that film images play the same role for us in
our interpretation of film that the linguistic text plays in our
interpretation of literature. Is there a tension between these two
claims? I say no. We need something like a structure/function
distinction here. Text and image are unlike in structure; images
lack the articulated, convention-based structure of complex lin-
guistic units. Text and image are alike in function - or rather
there is one common function they have, along with many func-
tional differences. They both constitute the evidence for the in-
terpretation of a work in their respective mediums. And the
existence of structurally distinct but functionally similar entities
is, I take it, uncontroversial: the current Tay Bridge has the same
function as its collapsed predecessor but - I trust - a different
structure.
22
An intention is communal rather than merely shared when it is shared and
when its being so is common knowledge between those who share it.
259
Chapter 9
Narrative and narrators
For films as for novels, we would do well to distin-
guish between a presenter of the story, the narrator
(who is a component of the discourse), and the creator
of both the story and the discourse (including the nar-
rator): that is, the implied author.
Seymour Chatman
I have been arguing that our basic interpretive strategy, for film
as for literature, is to look for an intentional explanation of what
is put before us - image or text. I have said nothing yet about
the role of a narrator in all this. Here we shall find some im-
portant differences between the roles of narrators in literature
and in film; film offers much less scope for the narrator than
does literature. Also, film requires us to acknowledge that there
can be unreliable narration without there being any narrator to
whom we can ascribe the unreliability. This will require a sub-
stantial revision of accepted theory. This chapter will also lend
weight to the claims of Chapter 8 concerning the implied author;
it will specify in some detail how appeal to the implied author
helps us make sense of the text or image. It will also show how
the concept of the implied author is more significant for under-
standing narrative than is that of the narrator.
There is one other issue of general significance on which this
chapter will bear. Recent writing on the theory of literature and
film has tended to focus on ways in which the act of interpre-
tation resembles or exemplifies pathological behaviour. There
have been attempts to characterize the reader's, and sometimes
the writer's, relation to the work as having an obsessive, pos-
sessive, violent and quasi-sexual nature. No doubt these things
find their place in any human activity. But they are very far from
being the whole story when it comes to interpretation. In op-
position to this tendency I hold that interpretation is a largely
26o
rational enterprise governed by the standards of evidence and
probability that apply in other areas. One thing I hope to do
here is to show how this approach can help us understand why
certain kinds of narrative effects are common while others are
not.
Briefly, my argument will be this: standard accounts of nar-
rative unreliability require there to be a narrator - a fictional
being who belongs within the world of the story and who
claims, falsely, to be recounting events of which he or she has
knowledge. But I shall argue that there are narratives which are
unreliable even when there is no narrator. In these cases, unre-
liability is not the product of a disparity between two conflicting
viewpoints, one internal (the narrator's) and one external. Rather
it is the product of a single, external viewpoint which has, as
we shall see, a rather complex structure. This, I say, enables us
to make sense of unreliability in film, without committing us to
implausible claims about the kinds of narrators which films em-
body. Finally I shall show how unreliability in film and other
media connects with another narrative trope- that of ambiguity.
Ambiguity, I shall argue, facilitates unreliability in ways that tell
us something about the scope and limits of narrative itself.
9.1 IMPLIED AUTHORS AND NARRATORS
Texts are unreliable when there is a certain kind of discrepancy

between the fictional story told and the text itself. The problem
is to say exactly what that discrepancy is. The mark of textual
unreliability is not that the story cannot be deduced from the
text; no fictional stories of any interest or complexity can be
straightforwardly deduced from their texts alone, for texts need
to be taken in conjunction with relevant background informa-
tion. And even in a reliable text, what is literally said and what
is supposed to be part of the story can be in conflict. That is
how it is when the text contains irony, metaphor or other non-
literal devices. And a text containing irony is not necessarily an
unreliable text.
The standard characterization of unreliability appeals to the
idea of a narrator. The Turn of the Screw is an unreliable narrative,
261
Interpretation
or so we may suppose for the sake of the argument. How so?
Simplifying a little, we can say that the story is told to us by a
narrator, the governess. She tells of events of which she has
knowledge and in which she was deeply involved. But in fiction,
as in real life, people may tell what is not true, either through
error or through the desire to deceive. And so it is with the
governess, who is mistaken - radically mistaken - about the
events she recounts. But since the governess is our source of
information and we have no independent access to the facts of
the case, how do we know that she is mistaken? The standard
account says that narrative unreliability is the product of a dis-
crepancy between two perspectives. One of these perspectives
is that of the implied author, a figure who may in a sense be
fictional or imagined, because his or her mental economy does
not necessarily correspond to that of the actual author, but who
is not to be thought of as occupying a position within the work
itself. Rather he is conceptualized as the agent responsible for
the story qua fiction.' The implied author's perspective is always
external to the story.
The other perspective is that of the narrator. Narrators, ac-
cording to the standard view, come in two kinds: internal and
external. An internal or (in the favoured lingo) intradiegetic nar-
rator belongs within the world of the story, telling of what he
or she knows or believes to be true, or trying to mislead us about
what has happened (as in an unreliable narrative). An external
or extradiegetic narrator is outside that world telling us a fictional
story. Watson is the internal narrator of the Sherlock Holmes
stories, while the narrator of Tom ]ones is external. 2
' So the implied author, as I use that notion, is always "extradiegetic" in Ge-
nette's sense (Narrative Discourse).
2
Here I shall ignore a tendency which some writers on narrative have of using
the terms "narrator" and "author'' to refer to things which could not literally
be narrators or authors because they are not persons. Thus, Christian Metz
speaks of the viewer's perception that the film images are organized by a
''grand image-maker'', which Metz then goes on to describe as "the film itself
as linguistic object" (Film Language: A Semiotics of the Cinema, pp. 2o-21). See
also remarks on Seymour Chatman in note 14 to Chapter 8, this volume.
In my view, the category of external narrators is redundant;

there is no need to distinguish between the external narrator and
the implied author. Advocates of the external narrator point to
works in which the narrative voice represents an outlook which
the reader seems to be intended to reject; it is natural to say that
there is in such cases a tension between narrator and implied
author and consequently that we must conceptualize the work
by invoking two distinct personalities. But cases of this kind can
be handled without appeal to the external narrator by assuming
that the implied author is speaking ironically; it is the implied
author who speaks, but what is said is intended by that same
speaker to be ridiculed or rejected. I shall simply assume that
we can dispense with the idea of an external narrator in favour
of the implied author. This will have the advantage of greatly
simplifying the discussion, and will not prejudice the case in my
favour; most of the arguments about unreliable narrators con-
cern narrators agreed to be internal.
From now on, I shall use the term "narrator'' exclusively to
refer to the possessor of this internal perspective. The governess
in The Turn of the Screw is an internal narrator in this sense. She
is a narrator who stands in a particularly intimate relation to
the events she relates, for she was a participant in them, and
much of what she tells us concerns herself, even if she is deluded
about her role in it all. Not all narrators are like that; some of
them tell of events of which they were passive witnesses, or
about which they have come to know via the testimony of oth-
ers. But as long as the narrator is narrating about that of which
she purports to have knowledge or at least belief, I count her
internaP For it must then be part of the story that the narrator
3
A distinction is sometimes made between intradiegetic narrators who are
characters and narrators who are not (cf. Genette's distinction between homo-
diegetic and heterodiegetic narrators in his Narrative Discourse, pp. 244-245).
But this is a distinction without a difference. The narrator is someone who,
within the scope of the story itself, knows of certain events and recounts them
to us (reliably or unreliably). His knowing requires that he belongs to the
world of the fiction, and so he is a character in it, if an unnamed one. Whether
he knows because he took part in the action or because he was told about it
Interpretation
belongs within the world of the story, just as Julius Caesar and
I belong to the same world (the actual world), even though what
I tell you about him I know only through hearsay.
The reader trying to figure out what happens in this story
now has two voices to attend to: the (implied) author's and the
narrator's. The authorial voice may not be explicitly signalled,
but readers are aware of its presence if they know that what
they are reading is fiction, because fictions have to be thought
of as works produced with certain intentions. Sensitive readers
may be able to spot a tension between these two voices, the
authorial and the narratorial. The narrator insists that there are
ghosts. But why has the author made this a story in which no
one else, apparently, sees or has knowledge of them? These and
like clues suggest that, perhaps, the author's intentions concern-
ing the story conflict with the narrator's certainties. But in such
a conflict the author always wins. Author and narrator do not
speak with the same authority. The narrator, within the world
of the story, is a mere reporter of events, and a fallible one like
all such reporters. The author, standing outside the story, is not
conveying to the audience a set of possibly erroneous beliefs
about the story world; the author makes things be true in that
story world (makes things "be fictional", as I say) simply by
decree or stipulation. If we as readers find it reasonable to infer
that the author intended that there not be any ghosts in the
story, then we must conclude that there simply aren't any. So
narrators are unreliable when their claims contradict the inferred
intentions of the author. This view of narrative unreliability, or
something broadly like it, is widely accepted. It is expressed in
a summarizing remark of Wayne Booth's: "I have called a nar-
rator reliable when he speaks for or acts in accordance with the
norms of the work (which is to say, the implied author's norms),
unreliable when he does not." 4 That, roughly, is how it is in
later (or because of the operation of magic, if that's how things are in the
story) is of no moment to his being a character. There is a distinction here,
but it is best understood as that based on epistemic proximity to the action,
and it admits of degrees. It is of no special relevance to our present concerns.
• Booth, Rhetoric of Fiction, pp. 158-159, emphasis in the original. One mislead-
ing feature of this remark is the implication that narrative unreliability is
Ford's The Good Soldier, in Camus's The Fall, in Ishiguro's The

Remains of the Day, and in many other literary narratives we
commonly describe as unreliable. 5
But what of those narratives which are intuitively unreliable,
but where the unreliability is not, or at least not obviously, at-
tributable to a narrator? In the literary case you might insist that
there is always a narrator to blame for the unreliability, however
unobvious her presence may be.6 This seems rather ad hoc; it is
worth asking whether there is some more elegant solution to the
difficulty. And with film, the idea of a narrator such as would
be postulated in order to save the Boothian definition of unre-
liability strains the bounds of coherence. It will take a moment
to see why.
9.2 THE ASYMMETRY BETWEEN

LITERATURE AND FILM
I want to introduce a distinction between kinds of narrators. The

distinction is one between controlling and embedded narrators.
Narrators as I have defined them are characters within the world
of the fiction who are to be thought of as telling us facts, lies or
deluded ravings? Narrators tell by making utterances, and we
can speak of the text of that utterance. Now that text- the text
of which it is fictional that it is uttered by the narrator - may
coincide with the text we are reading when we read the work.
In that case, we imagine the narrator to be controlling; it is fie-
always and exclusively a matter of value (Booth's "norms"), which is certainly

not the case, as many of Booth's examples attest. Booth's definition is taken
over, more or less, by Chatman, Story and Discourse, p. 233, and is repeated
in Gerald Prince, A Dictionary of Narratology, p. 101.
5
The film version of The Remains of the Day is not an unreliable narrative.
6
"Some important narrators are only implicit in the text, that is, their 'presence'
must be inferred and constructed by the spectator'' (Branigan, Narrative Com-
prehension and Film, p. 75).
7
This, of course, is a simplification, since there are stories within stories where
a narrator internal at one level is external at another; in these stories it is
fictional that the narrator is telling us that it is fictional that ... etc. Inclusion
of such cases into our present taxonomy would further complicate an already
complex structure, and ignoring them will not affect the argument.
Interpretation
tional that he or she is the source of the text before us. We know,
of course, that the text is fictional and we probably know the
identity of its real author, but we may think that it is part of the
fiction itself that the narrator is the source of this text, and accept
the fiction's implicit or explicit invitation to imagine exactly that.
In this sense, Watson is a controlling narrator of the Holmes
stories, and to some extent an unreliable one in the sense of
Booth. 8 But some narrators are embedded rather than controlling.
Where there is an embedded narrator, the text we read tells us
of someone's telling, but that teller is not, fictionally, responsible
for the text we read. Rather, it is fictional that the text we read
reports that person's telling. 9 Embedded narrators are common
in film: think of all those conflicting accounts in Rashomon, the
bits of Charles Kane's life told by various folk in Citizen Kane
and Walter Neff's disillusioned narrative in Double Indemnity.
Now this distinction between controlling and embedded nar-
rators is very important for understanding narration in film.
There are embedded narrators in film: I've just mentioned some
examples of them. But there is something awkward - indeed,
something close to incoherence - about the idea of a controlling
narrator in film. With literature it is often natural to imagine
that what one is reading is a true account of certain events wit-
nessed or otherwise known about by someone, who then went
to the trouble of setting it all down for us in writing; some of
John Buchan's adventure stories, we are to imagine, are the
product of a careful editor who has heard from the parties con-
cerned and has created a judicious account on paper from their
8
F. K. Stanzel supposes that all "first-person" narrators (in my terms: internal
narrators) are necessarily unreliable, because of the limitations on their knowl-
edge (A Theory of Narrative, p. 89). But failure to be omniscient is one thing
and failure to be reliable another. Perhaps the thought here is that a non-
omniscient narrator could not be certain of the truth of any of his beliefs. But
it is an error of the Cartesian tradition to suppose that lack of certainty trans-
lates into unreliability.
9
The term "embedded narrator'' is sometimes used to refer to any character-
narrator. This strikes me as· misleading usage; a character-narrator who is
controlling in my sense is not necessarily embedded in the story. He tells the
story, but he does not tell of his own telling. See, e.g., Wallace Martin, Recent
Theories of Narrative, p. 135.
266
reports. It is that report we imagine ourselves to be reading, and

its imagined author counts as a controlling narrator. But what
are we to imagine that would be analogous to this in the filmic
case - that the person in the know has gone to the trouble of
recreating it all for us on camera, spending millions of dollars,
employing famous actors and a vast army of technicians? That
seems implausible, especially in cases where the narrator, if
there is one, would most naturally be thought of as living in the
precinematic age.
Yet writers on film have, unfortunately, spoken of narrators
in film as controlling, or at least as if they have partial control.
Take the example of Hitchcock's Stage Fright, a filmic narrative
which tells, in part, of a character named Johnny, who relates
how he discovered the body but has been mistakenly identified
as the murderer. And the film tells of Johnny's telling, not by
reproducing for us his verbal account, but by translating that
verbal account into images which constitute a flashback. We in-
itially assume the flashback is correct. But at the end we learn
that it is not, that Johnny was really the murderer, that his story
of having found the body was a lie. The images we saw at the
beginning, of Johnny discovering the body, tum out to be un-
reliable. In one sense Johnny is the source of this unreliability:
it is his (fictional) act of false telling which causes us to be mis-
led. But is Johnny also to be regarded as the source of the un-
reliable images we see? Seymour Chatman says yes: Johnny, he
says, "is 'responsible' for the lying images and sounds that we
10
see and hear". That cannot be right. Johnny, like the other char-
acters, exists within the story, and it is not part of that story that
he produced and edited cinematic images in order to convince
his fictional fellows (and us?) of his innocence- a transparently
self-defeating enterprise. There are films in which some of the
images we see onscreen are the product of a character, and it is
fictional, of that character, that what we see onscreen is pro-
duced by him or her - Peeping Tom is perhaps the most famous
'" Chatman, Coming to Terms, p. 132. Chatman also says, of Alain Resnais's
Providence, that "Langham's voice-over, we eventually surmise, is somehow
constructing the images filling the screen'' (p. 133). Compare this with the
discussion in his Story and Discourse, p. 237.
Interpretation
example of this. So there is nothing incoherent in the idea of an
embedded narrator who tells, partly or wholly, by means of
cinematic images of which he or she is the source. But that is
manifestly not the situation in Stage Fright, and an analysis like
Chatman's, which puts Stage Fright and Peeping Tom in the same
category, must be wrong.
If controlling narrators in film are ruled out, how can we have
unreliability in film? One way would be by ascribing unrelia-
bility to an embedded narrator: 11 we have a filmic narrative in
which it is fictional that one of the characters speaks falsely, and
that narrative may tell us what is falsely spoken in a number of
ways- by voice-over, or, more commonly, by transforming the
character's words into images without at the same time making
it clear that what is spoken is false. In these cases we needn't
suppose it fictional that what we see onscreen proceeds from
the character. The filmic source of the images lies outside the
diegesis. That, it seems, is how it must be for any unreliable
filmic narrative that aspires to coherence: unreliability, where it
occurs, must be the product of an embedded narrator, because
that's the only narrator which comfortably fits the constraints of
filmic presentation.
It certainly is that way with many filmic narratives which
aspire to a kind of unreliability, as with Stage Fright. The ques-
tion is, is all filmic unreliability of this kind? George Wilson
claims to have found a case which is not: Fritz Lang's You Only
Live Once. The film tells, apparently, of the problems of a young
12
man who has been in trouble with the law and who is subse-
quently and falsely accused of a murder-robbery. Wilson argues
that this natural interpretation of the film is, on closer exami-
nation, not supported by, and is in fact at certain crucial points
undermined by, the film's narration and its studiedly selective
presentation of events. On Wilson's view, a right interpretation
of the film would have us withhold judgement as to Eddie's
guilt or innocence. But that is certainly not what most viewers
" I agree with Christian Metz that "the explicit enunciators in the film are
always embedded" (Metz, "The Impersonal Enunciation", p. 768).
12
Wilson, Narration in Light.
268
of the film have done; generations of critics and lay viewers have
accepted the view that Eddie is an innocent victim. But there
does not seem to be any embedded narrator in the film to whom
unreliability might be described. And, given the film's dramatic
structure, it is not easy to come up with a covert embedded nar-
rator of any plausibility.' 3
I am not entirely convinced by Wilson's interpretation, but it
does mesh well with parts of the film which are otherwise hard
to understand. And if Wilson's interpretation does not apply in
all detail to Lang's film, it is not difficult to imagine a film,
different from Lang's in minor ways, to which it does. So we
may conclude that it is possible for there to be unreliable narra-
tion in film where no embedded narrator is present - and when
it comes to definitions, possible counterexamples are as telling
as actual ones. In that case we really will have to look for an-
other definition of narrative unreliability if we are to accom-
modate You Only Live Once and like cases. Anyway, I shall
assume, for the sake of the argument, that Wilson is right about
the film.
So You Only Live Once is an unreliable narrative without an
unreliable narrator.' 4 What, then, is the source of its unreliability,
if not the tension between author and embedded narrator? Un-
reliability can have its source, I claim, in a certain kind of com-
plex intention on the part of the implied author. Let me explain.
An agent can do something with an intention of the following
complex kind: she creates or presents something which she in-
tends to be taken as evidence of her intentions, and she intends
that a superficial grasp of that evidence will suggest that her
intention was <1>, whereas a better, more reflective grasp of the
' Perhaps you think it analytic that narrative must have a narrator (as does
3
Sarah Kozloff: "Because narrative films are narrative, someone must be nar-
rating"; Invisible Storytellers, p. 115, quoted approvingly in Chatman, Coming
to Terms, p. 133). But then you simply object to my terminology, and I could
avoid the objection by using another term to refer to the vehicle of narrator-
less story-telling. Consider it done.
'• As Cary Groner pointed out, An Occurrence at Owl Creek (1961) seems to be
a case of unreliability without a narrator. But here the unreliability is tem-
porary; that the film's images are largely the imaginings of a character is
made evident at the end.
Interpretation
evidence will suggest that her intention was 'I'. An example:
Freida compliments Fred on his sophisticated sense of humour.
Her flattery seems out of line with what we know of Freida's
acerbic personality. But now we see that it was all ironic and
intended to be recognized, ultimately, as more of the abuse she
usually heaps upon Fred. Freida's performance was unreliable
in my sense, and there may be people, Fred among them, who
didn't get to the second stage, because that took just a little more
calculating than some of us can be relied on to make.
That seems to be what is going on in You Only Live Once; we
take the images and sounds that make up the film as intended
one way. But if we are scrupulous in our examination of those
images, we find peculiarities, incongruities and apparently un-
motivated elements that start to fall into place when we see that
it can be interpreted in another way. Their falling into place
consists in their being seen as intended to suggest that second,
less obvious interpretation. Narratives which are the product of
this kind of two-tier system of intentions constitute a distinctive
and especially challenging class, and I do not think that they are
very well understood. I hope to change that somewhat in the
rest of this chapter.
9·3 KINDS OF UNRELIABILITY

Defining unreliable narrative in terms of complex intentions at-
tributable to an implied author allows us to count a narrative
as unreliable when there is no narrator we can identify as the
source of unreliability. What are the relations between this kind
of unreliability and the cases of narratorial unreliability covered
by Booth's definition? Some of the cases that are unreliable on
Booth's definition would not be unreliable on mine. In Lardner's
"Haircut," for example, we have an internal narrator whose out-
look (his "norms", as Booth puts it) is different from, and un-
dermined by, that of the implied author. But this is not a case
where we should attribute a complex intention to the implied
author. The disparity of outlooks is too obvious in this case for
us to conclude that the implied author has intentions which can
be grasped only on reflection. Of course, it is possible to imagine
270
someone who is misled by this or by any narrative; someone
might think that the speaker in "Haircut" is saying things
which, within the framework of the story, represent a correct
assessment of the situation. But if the test of the implied author's
complexity of intention is the mere possibility that someone will
be fooled in this way, we shall end up saying that all narratives
are unreliable. We need a criterion of unreliability that is more
discriminating. We should say that a narrative is unreliable in
my sense only if it has a tendency to mislead an attentive, in-
telligent reader with some experience of the relevant genre -
here the short story. I do not think any such reader will be long
misled by Lardner's tale. The warranted conclusion is surely
that the implied author intends us to see, straight off, the moral
shortsightedness of the narrator. "Haircut" is unreliable in
Booth's sense, but not in mine.
That is not to say that the extensions of Booth's definitions
and mine are disjoint, for it is possible for a work to satisfy both
of them. Such a work would be one where the implied author
intends the plot to be understood one way on a superficial read-
ing, another way on a more attentive reading, and where there
is a lying or deluded narrator. In fact, it is common for an un-
reliable narrator to be the mechanism whereby unreliability in
my sense is achieved. At first we take, and are intended to iake,
the narrator at her word and so interpret the story one way; on
reflection we see her, as we are intended to, as unreliable, and
revise our reading.' 5 So there is overlap between our definitions,
but they are not the same; mine was introduced, after all, to
cover cases that Booth's does not cover- and so the definitions
characterize different concepts.
But I shall not endorse the comfortably ecumenical position
that these definitions are merely different but equal. I believe
'
5
In that case we have a "seductive" unreliable narrator. The classic case is
The Tum of the Screw. Where Booth's definition applies and mine does not,
we have an unseductive one. See James Phelan, "Narrative Discourse, Lit-
erary Character and Ideology", in idem, Reading People, Reading Plots, p. 137.
Sometimes cases of unobviously unreliable narratives are described as "am-
biguous" (e.g., by Shlomith Rimmon-Kenan, Narrative Fiction, p. 103), but I
wish to use this term for another purpose. See Section 9-4-
Interpretation
that my characterization of unreliability in terms of a complex
intention - call it "complex unreliability" - is of greater theo-
retical and critical interest than the familiar Boothian character-
ization in terms of a disparity of outlook between the narrator
and the author- call it "Booth unreliability". There are of course
interesting cases of Booth unreliability, but they tend also to be
cases of complex unreliability, that is, cases in which the nar-
rator's unreliability is to some degree intended as unobvious.
We are past the point where a narrator's unreliability is intrin-
sically interesting, because we no longer presume that narrators
will be reliable. Without that presumption, narratorial unrelia-
bility is, of itself, no more significant than the mendacity of a
dramatic speaker - no more significant, that is, from the point
of view of a theory of narrative. But unreliability that is to some
degree hidden is of theoretical interest because its operation de-
pends on delicately balanced inferential strategies that the
reader must undertake. To get a good sense of the nature of
these inferences will require the introduction of another narra-
tive concept, that of an ambiguous narrative.
9·4 AMBIGUOUS AND UNRELIABLE

NARRATIVES
An ambiguous narrative is one which does not enable us to

answer all the questions which arise concerning the story. In a
sense, every narration is ambiguous, since no narration can pos-
sibly provide complete information about the characters and
events it describes. What did each character have for breakfast
that morning? At what exact moment was each character born?
You might say that these are not meaningful questions within
the context of the fiction, as they would be in the real world,
and to suppose that they are is to endorse the fallacy of taking
fictional characters for real people concerning whom every ques-
tion, interesting or not, has an answer. But such questions are
not meaningless. Most fictional narratives present stories in
which the characters are assumed, except where there is evi-
dence to the contrary, to have the sorts of characteristics that
real people have. No one doubts, I think, that Marlowe in The
272
Big Sleep is a man who, like others, was born at a particular

time; there is no suggestion, for instance, that he is immortal. So
there must, according to the story, be a time of his birth. It is
just that the narrative does not elect any particular time as that
time. And if it is part of the story that Marlowe was born at a
certain time, it can hardly be senseless to ask when.
But it would be a fallacy to assume that there must be an
answer to that question. That would be to assume that fictional
characters are real people. The question of Marlowe's birth date
is not meaningless, but it is insignificant, and The Big Sleep is
not in any interesting sense an ambiguous narrative just because
it fails to resolve that question. It is only when a question is
significant that the narration's failure to answer it gives grounds
for saying that the narration is ambiguous. When is a question
significant? One answer is this: when members of the audience
are (normally) inclined to ask it concerning that narrative. The
friend of this theory can note with satisfaction that the question
about the time of Marlowe's birth is not a question that viewers
are inclined to ask. But this proposal will not do. Many people
are inclined to wonder what will happen to Rhett and Scarlett
at the end of Gone with the Wind (as the recent and long-awaited
sequel indicates), but this would not be grounds for saying that
the narrative (either the book or the film) is ambiguous in the
sense I am interested in here. Questions about the continuation
or noncontinuation of relationships are ones we are almost al-
ways inclined to ask- at least they arise fleetingly in our minds
- at the end of the work. This proposal is going to make too
many narratives ambiguous.
A proposal with a similar defect has it that the narrative is
ambiguous if it leads us to expect an answer to a question when
in fact it does not provide an answer - though the proposal
would at least explain the intuition that Gone with the Wind is
not ambiguous, since that narrative does not lead us to expect
that an answer will be given to the question, "What happens to
them after the narrative breaks off?" But this proposal is neither
necessary nor sufficient for ambiguity. The narrative might
make it clear from the start that a certain question is not going
to be answered (in the case of a film that might require a voice-
273
Interpretation
over to be convincing). So we are not led to expect an answer,

and indeed the question may not be answered by the narrative.
Yet the question might be one such that, in not answering it, the
narrative takes on the character of ambiguity. So satisfaction of
the proposed criterion is not necessary for ambiguity. And a
question may arise to which we expect an answer, but where
we put down failure to provide an answer to incompetence in
the construction of the narrative. In that case we would be doing
the work a favour it does not deserve by calling it ambiguous.
The case of narrative incompetence is the clue to solving our
problem. When a question arises which is not answered in the
narrative, but where we ascribe the nonanswer to incompetence,
we do not think that a question has been deliberately raised by
the implied author, and deliberately left unanswered; we think,
exactly, that either the raising or the failure to answer was due
to some oversight or other failure of execution. So I propose the
following as the criterion for determining when a narration is
ambiguous: when it raises a question in the reader's/viewer's
mind which it fails to answer, and where the raising and the
nonanswering seem to have been intentional. This proposal
gives the result that The Big Sleep is not ambiguous, even though
it fails to determine the answer to the question of Marlowe's
age: the narrative gives no indication that the maker intended
that question to be raised. It does the same for Gone with the
Wind; while viewers and readers may wonder about the future
of Rhett and Scarlett, and the makers may have expected that
they would wonder about it, and while all this may be common
knowledge between audience and maker, the question does not
seem to be intentionally raised and intentionally left unan-
swered by the narrative.'6
6
' The film narratives used by Bordwell and Thompson to illustrate ambiguity
(Day of Wrath and Last Year at Marienbad) would count as ambiguous on my
definition. Bordwell and Thompson associate ambiguity closely with causal-
ity (Film Art, p. 250). Their idea seems to be that the work is ambiguous to
the extent that the causes or effects of narrative elements are unclear. Since
most narrative events, like most events in real life, have many distinct partial
causes and many distinct effects, we shall need to distinguish the significant
274
It is easy at the level of theory to see the differences between

unreliable and ambiguous narratives. But it is not always easy
to say which kind a particular work belongs to. Is Rashomon an
ambiguous or an unreliable narration? If there figure within it
embedded narrators who are unreliable, we may grant that it is
unreliable in the narratorial sense of Booth. It would be ambig-
uous if it left it an open question which of the conflicting ac-
counts is true. But is it unreliable in the sense that I have
defined, that it is possible to detect in its making the influence
of a complex intention of the kind I have described in connection
with You Only Live Once? That would be so if we thought of
Rashomon this way: as intended, first, to suggest to us that the
problem is to decide which account is true, and second, to sug-
gest on deeper reflection the relativity of truth and, in conse-
quence, the falsity of our first question's presupposition that
there is a right answer. For some of us, the easy relativism of
the last option is too banal to be an attractive candidate for in-
terpretation - but this may be just an indication that there is
sometimes no neutral perspective from which to choose between
ambiguity and unreliability, a situation we sometimes experi-
ence with other interpretive choices.' 7
But while ambiguity and unreliability are distinct interpretive
options, they are compatible, not merely in the sense that there
is sometimes no principled choice between them, but in the
stronger sense that a single interpretation of the work may re-
quire the application of both. We might, for instance, take Rash-
oman as complex unreliable in that at first glance the options are
between the explicit accounts of the various narrators, while on
reflection we see that there is another option - the relativistic
one - and that the story is ambiguous between those collected
at the first round and this one. (I would count that as only mar-
ginally less banal than straightforwardly opting for relativism,
but it might still be the best thing we can come up with.) On
that view, Rashomon is both complex unreliable and ambiguous.
from the nonsignificant causes and effects. My proposal can be read as doing
that.
7
' See my "Interpretation and Objectivity".
275
Interpretation
Note that You Only Live Once is, on Wilson's account, both
ambiguous and unreliable. At first the question "Is Eddie guilty
of murder?" seems to be answered by the narrative. But we see,
on closer inspection, that it is left open by it. That also seems to
me the best case that can be made on behalf of a psychological
interpretation of The Turn of the Screw: reflection doesn't show
that there are no ghosts; at best it shows that another hypothesis
does about as well as the supernatural one when it comes to
explaining the text.' 8 In that case might there be some internal
connection between unreliability and ambiguity?
Complex unreliability does not necessitate ambiguity. But
complex unreliability is an easier effect to achieve when it in-
volves ambiguity than when it does not. The task of the maker
of a complex-unreliable narration is difficult. It is to set clues at
two levels: at level1, where the clues are more obvious but only
superficially persuasive; and at level 2, where they are less ob-
vious but more weighty when reflected upon. But the degree of
difficulty of the task varies from case to case, and one deter-
minant of it is what we might call the epistemic distance between
the two levels: increase the distance and you increase the diffi-
culty. By "distance" I mean the disparity between what you
want to convey at a first impression and what you want the
audience to grasp on further reflection. The greater the distance
in this sense, the greater the subtlety and complexity of the rea-
soning the audience will have to go through to cover the gap,
and the less likely it is that they will succeed. Trying to raise
the probability of success in such a case by reinforcing the clues
at level 2 may simply undermine the whole project by making
the inference to level 2 more obvious and natural than that to
level1.
8
' Jack Clayton's The Innocents, a film version of The Tum of the Screw, is inter-
esting in this regard. One difficulty the film makers had to contend with was
that a significant proportion of the film's audience would bring with them
their knowledge of the unreliability in James's story, which would make it
impossible for the film to achieve the same effect; the audience would be
primed for the discovery of the higher-level clues from the start. As I un-
derstand the film, their solution, intelligently enough, was to forgo any at-
tempt at complex unreliability and to settle for ambiguity.
In this sense of "distance" there will generally be a greater

distance between the two levels if at one level we are given, say,
a yes answer to a question, and at the other a no answer, than
if at one level we are given an answer (yes or no) and at the
other we are told that no answer is forthcoming. In the first case
you first have to persuade the audience to conclude that the
answer to a certain question is, say, yes, and then to revise their
opinion and conclude that the answer is no. In the second case
you first have to persuade the audience to conclude that the
answer to the question is yes (or no), and then simply to with-
draw that opinion, remaining thereafter agnostic on the ques-
tion, thinking of the work now as ambiguous. So the second is
an easier thing to do than the first, and we would expect to find
the second more commonly than the first in cases of complex
unreliability. Indeed, that seems to be how things are with no-
table cases of complex unreliability like You Only Live Once and
The Turn of the Screw.
When an unreliable narrative is one that seems, superficially,
to close a certain issue but reveals on reflection that the question
is left open, as in You Only Live Once, let us say that we have a
"transition to openness". Consider a transition in the opposite
direction- a "transition to closure"- where the narrative seems,
superficially, to leave a certain issue open but is seen on reflec-
tion to answer the question one way or another. Would a tran-
sition to closure be easier or more difficult to effect than a
transition to openness? Taking into account only what I have
called epistemic distance suggests that it would be neither more
nor less difficult, since distance is symmetrical; the distance be-
tween two things is independent of the order in which they are
taken. However, there are grounds for saying that it would,
other things being equal, be more difficult to effect a transition
to closure than to openness. The task, in creating a complex-
unreliable narration, is to suggest one hypothesis by means of
more obvious but ultimately less convincing evidence, and to
suggest another by means of less obvious but more convincing
evidence. The difficulty is to ensure that the more convincing
evidence will in fact be less obvious, without having it disappear
entirely from view. That difficulty will be greater the stronger
277
Interpretation
is the hypothesis for which it is evidence. After all, it takes more
evidence to get us to believe a strong conclusion than a weak
one. It's not hard to convince me that either Oswald shot Ken-
nedy or someone else did; it's much harder to convince me that
someone else did. So, other things being equal, you will have to
provide stronger evidence to support a definite conclusion than
to support a mere ambiguity. But then the transition to closure
requires stronger evidence at the level of the less obvious, and
that kind of transition is going to be more difficult to effect than
one to openness. So we would expect to find transitions to open-
ness more frequently in literature and other narrative forms than
transitions to closure. And that, I believe, is exactly what we do
find.' 9
Perhaps there is a lesson here. The old structuralist project
was to find patterns and regularities- even to some degree nom-
ological ones - within the apparent diversity of literature. This
is not generally thought to have been a success. But perhaps the
project was not entirely doomed. Perhaps there is something like
an "order of narrative" to be discovered by appeal not to struc-
ture but to reason. If tales, cinematic or literary, are generated
by someone for someone, a major determinant in the evolution
of tales, and in the patterns of in variance detectable through that
evolution, will be what we might call the constraints of reciprocal
reason. Tellers have to decide what they can expect hearers to
pick up by way of clues to their intentions, and hearers have to
decide what tellers will think hearers can pick up, and so on.
As story-telling evolves, as new narratives piggyback on the
communicative breakthroughs of their predecessors and as new
and more complex narrative intentions become common knowl-
edge in the community of tellers and hearers, these inferences
will become more complex, with the most inventive tellers al-
ways operating on the thin boundary between surprising the
audience and having the audience completely miss the point.
'
9
George Wilson argues that there is what I am calling a "transition to closure"
in Ford's Searchers: that the narrative appears to be ambiguous on the ques-
tion of Ethan's motives, but that it can, on reflection, be seen to disclose a
motive. See Wilson, Narration in Light, pp. 45-50.
But we shouldn't expect a simple linear development of increas-
ingly complicated narratives, for engineering constraints will
quickly exhaust the capacity of any given narrative trope. Con-
sider narratives which display what I have called complex un-
reliability. Tellers cannot move forward, if that's the word, by
steadily increasing the epistemic distance between the obvious
and the unobvious interpretation; as I have pointed out, episte-
mic distance and unobviousness pull in different directions, and
beyond a certain point increases in epistemic distance become
self-defeating. At that point tellers have to look to a different
trope or play against the rising expectations of hearers by going
into reverse - for example, presenting what hearers will expect
to be a complex-unreliable narrative which turns out, surpris-
ingly, to be quite straightforward. With constraints pulling
against one another in this way we can expect that narrative
will not be a wholly amorphous, infinitely stretchable bag of
tricks, but something with identifiable boundaries and perhaps
a centre, and with tellers and hearers chasing each other away
from and back towards that centre in various directions. The
structure here will not be set by the laws of genre, of literary
archetypes or by the structure of language, but by the rules for
assessing the reasonableness of inferences about people's inten-
tions.
I have suggested that there is a kind of narrative unreliability

not covered by the standard account. I have given a general
characterization of it, and suggested how it stands in logical
relation to the more familiar kind of unreliability. I have also
described what I call ambiguous narrative and suggested that
there is a close connection between this and the kind of unreli-
able narrative I defined - a connection forged by the difficulty
readers and viewers face in making the inferences to the implied
author's intentions that are necessary if unreliability is to be de-
tected. I conclude that unreliability in narrative makes no sense
without appeal to the concept of an implied author, but that the
concept of a narrator is required by only one kind of unrelia-
bility. The implied author, we may say, is an absolute presup-
279
Interpretation
position of unreliability, the narrator a merely conditional one.
That is why there can be narratives which are genuinely unre-
liable - rather than merely ambiguous or misleading - even
when there is no narrator in whom we can locate the source of
the unreliability.
280
In conclusion
A fiction film is a visual and a pictorial work. It is visual because

our mode of access to it is visual; it is pictorial because its mode
of representation is pictorial. Its material is moving pictures -
pictures which really move rather than simply create the illusion
of movement. Nor is film typically productive of any cognitive
illusion to the effect that what it represents is real; our standard
mode of engagement with the film is via imagination rather than
belief. Imagination is, however, parasitic on belief, for it consists
of running our belief (and desire) system off-line, disconnected
from standard inputs and outputs. But while the pictures of film
are not productive of illusions, they are typically realistic pic-
tures: pictures which are like, in significant ways, the things they
represent. And it is partly in virtue of their likeness to these
things that we are able to recognize the depictive content of
these pictures. For this reason film is not a linguistic medium,
nor is it in any interesting sense like a linguistic medium.
There is something distinctive about our imaginings in re-
sponse to film. It is not, as many theorists have claimed, that
we imagine ourselves to be witnesses of the action, placed where
the camera is. Rather, it is that our imaginings have a distinc-
tively visual structure. Nor do we imagine that the action rep-
resented is occurring in the present as we watch; film is
preeminently an art of time, but it does not represent fictional
things as co-occurrent with our watching. Despite the fact that
film is an essentially nonlinguistic medium, it is possible to de-
281
In conclusion
velop a general theory of interpretation which accounts for both
literature and film. According to this theory, interpretation is
explanation - explanation by reference to intentional causes.
With a novel or a film, the interpreter's task is to formulate
plausible hypotheses about the story-telling intentions produc-
tive of the work. Still, literature and film offer rather different
kinds of narrative possibilities and, in particular, the role for
unreliable narrators is more restricted in film than in literature.
Film shows the need for a category of unreliable but narratorless
narratives.
Named propositions
(Starred propositions are accepted.)
Illusionism: That film typically causes the viewer to undergo an illu-

sory experience.
There are two versions of this claim:
Cognitive Illusionism: That film creates an illusion of the reality of the

fiction it presents.
Perceptual Illusionism: That film typically creates an illusion of move-

ment.
Presentation Thesis: That photographs and cinematic images do not

merely represent the things they are of, but rather present those things
to us.
As a special case of this:
Transparency: That when we see a photograph or cinematic image of

X, we see X.
Perceptual Realism*: That pictures in general, and cinematic images in

particular, are in significant respects like the things they represent.
Imagined Observer Hypothesis: That the standard mode of imaginative

engagement with film involves the viewer imagining that she is watch-
ing the events of the story from the position of the camera.
Named propositions
Claim of Presentness: That fiction films typically represent the fictional

events as occurring in the viewer's present.
Implied Author Intentionalism*: That narrative interpretation requires
us to discover the narrative intentions of the work's implied author.
Real Author Intentionalism: That narrative interpretation requires us
to discover the narrative intentions of the work's real author.
Bibliography
Affron, Charles, Cinema and Sentiment, University of Chicago Press,

1982.
Akins, Kathleen A., ''What Is It Like to Be Boring and Myopic?" in B.
Dahlbom (ed.), Dennett and His Critics: Demystifying Mind, Cam-
bridge, Mass.: Basil Blackwell, 1993.
Altman, Rick, Sound Theory, Sound Practice, London: Routledge & Ke-
gan Paul, 1992.
Anstis, Stuart, "Motion Perception in the Frontal Plane: Sensory As-
pects", inK. Boff, L. Kaufman and J. Thomas (eds.), Handbook of
Perception and Human Performance, vol. 1, New York: Wiley, 1986.
Amheim, Rudolf, Film as Art, London: Faber & Faber, 1983.
Aumont, J., ''The Point of View", trans. Arthur Denner, Quarterly Re-
view of Film and Video, 11 (1989): 1-22.
Aumont, J., A. Bergala, M. Marie and M. Vemet, Aesthetics of Film,
trans. R. Neupert, Austin: University of Texas Press, 1992.
Bach-y-Rita, P., et al., ''Vision Substitution by Tactile Image Projection",
Nature, 221 (1969): 963-964.
Bal, Mieke, "Description as Narration", in On Story-Telling, Sonoma,
Calif.: Polebridge, 1991.
Balazs, Bela, Theory of the Film: Character and Growth of a New Art, New
York: Amo, 1972.
Baron-Cohen, S., "The Autistic Child's Theory of Mind: A Case of Spe-
cific Developmental Delay", Journal of Child Psychology and Psychi-
atry and Allied Disorders, 30 (1989): 285-297.
Baron-Cohen, S., A. Leslie and U. Frith, "Does the Autistic Child Have
a 'Theory of Mind'?" Cognition, 21 (1985): 37-46.
Baudry, Jean-Louis, "Ideological Effects of the Basic Cinematic Appa-
Bibliography
ratus", in Bill Nichols (ed.), Movies and Methods, vol. 2, Berkeley
and Los Angeles: University of California Press, 1985.
Bazin, Andre, What Is Cinema? vol. 1, trans. Hugh Gray, Berkeley and
Los Angeles: University of California Press, 1971.
"The Myth of Total Cinema", in Mast and Cohen (eds.), Film Theory
and Criticism, 3d ed., New York: Oxford University Press, 1985.
Jean Renoir, trans. W. W. Halsey II and W. H. Simon, New York: Da
Capo, 1992.
Biederman, Irving, ''Visual Object Recognition", in Alvin Goldman
(ed.), Readings in Philosophy and Cognitive Science, Cambridge,
Mass.: MIT Press, 1993.
Boghossian, Paul A., and J. David Velleman, "Colour as a Secondary
Quality'', Mind, 98 (1989): So-103.
Booth, Wayne, The Rhetoric of Fiction, 2d ed., University of Chicago
Press, 1983.
Bordwell, David, Narration in the Fiction Film, Madison: University of
Wisconsin Press, 1985.
Making Meaning: Inference and Rhetoric in the Interpretation of Cinema,
Cambridge, Mass.: Harvard University Press, 1989.
Bordwell, David, and K. Thompson, Film Art: An Introduction, New
York: Random House, 1986.
Branigan, Edward, ''The Point-of-View Shot", in Bill Nichols (ed.),
Movies and Methods, vol. 2, Berkeley and Los Angeles: University
of California Press, 1985.
Narrative Comprehension and Film, New York: Routledge, 1992.
Brown, E. L., and K. Deffenbacher, Perception and the Senses, New York:
Oxford University Press, 1979.
Browne, Nick, The Rhetoric of Filmic Narration, Ann Arbor, Mich.: UMI
Research Press, 1976.
Brysan, Norman, et al. (eds.), Visual Theory: Painting and Interpretation,
Cambridge: Polity, 1991.
Burch, Noel, Theory of Film Practice, trans. Helen Lane, Princeton, N.J.:
Princeton University Press, 1981.
Carroll, Noel, ''The Power of Movies", Daedalus, 114, no. 4 (1985): 79-
103.
Mystifying Movies: Fads and Fallacies in Contemporary Film Theory, New
York: Columbia University Press, 1988.
Philosophical Problems of Classical Film Theory, Princeton, N.J.: Prince-
ton University Press, 1988.
The Philosophy of Horror: Paradoxes of the Heart, New York: Routledge
& Kegan Paul, 1990.
286
Bibliography
"Review of Gregory Currie, The Nature of Fiction", Mind, 103 (1994):

542-545·
Caughie, John (ed.), Theories of Authorship: A Reader, London: Routledge
& Kegan Paul, 1981.
Cavell, Stanley, The World Viewed: Reflections on the Ontology of Film,
''Ugly Duckling, Funny Butterfly: Bette Davis and Now, Voyager",
Critical Inquiry, 16, no. 1 (1990): 213-247.
Chatman, Seymour, Story and Discourse: Narrative Structure in Fiction
and Film, Ithaca, N.Y.: Cornell University Press, 1978.
Coming to Terms: The Rhetoric of Narrative in Fiction and Film, Ithaca,
N.Y.: Cornell University Press, 1990.
Churchland, Paul M., Scientific Realism and the Pl£lsticity of Mind, Cam-
bridge University Press, 1979.
"Perceptual Plasticity and Theoretical Neutrality: A Reply to Jerry
Fodor", Philosophy of Science, 55 (1988): 167-187.
Cohen, Ted, "What's Special About Photography?" Monist, 71 (1988):
292-305.
Comolli, Jean-Louis, "Machines of the Visible", in Stephen Heath and
Teresa de Lauretis (eds.), The Cinematic Apparatus, London: Mac-
millan, 1980.
Corballis, M. C., "On the Evolution of Language and Generativity",
Cognition, 44 (1992): 197-226.
Currie, Gregory, The Nature of Fiction, Cambridge University Press,
1990.
"Work and Text'', Mind, 100 (1991): 325-340.
"Interpreting Fiction", in R. Freadman and L. Reinhardt (eds.), Lit-
erary Theory and Philosophy, London: Macmillan, 1991.
"Text without Context: Some Errors of Stanley Fish", Philosophy and
Literature, 15 (1991): 212-228.
"Interpretation and Objectivity", Mind, 102 (1993): 413-428.
"Imagination and Simulation: Aesthetics Meets Cognitive Science",
in M. Davies and T. Stone (eds.), Mental Simulation, Oxford: Basil
Blackwell, 1994.
"Simulation-Theory, Theory-Theory and the Evidence from Autism",
in P. Carruthers and P. Smith (eds.), Theories of Theories of Mind,
Cambridge University Press, 1995.
Dagle, Joan, "Narrative Discourse in Film and Fiction: The Question of
the Present Tense", in Syndy M. Conger and Janice R. Welsch
(eds.), Narrative Strategies, Macomb: Western illinois University
Press, 1980.
Bibliography
Dahlbom, B. (ed.), Dennett and His Critics: Demystifying Mind, Cam-

bridge, Mass.: Basil Blackwell, 1993.
Davies, Martin, "Function in Perception", Australasian Journal of Philos-
ophy, 61 (1983): 409-426.
Davies, M., and T. Stone (eds.), Mental Simulation, Oxford: Basil Black-
well, 1994.
Davies, W. M., Experience and Content: Consequences of a Continuum The-
ory, Ph.D. dissertation, Flinders University, 1993.
Dawkins, Richard, The Blind Watchmaker, Harlow: Longman Scientific
and Technical, 1986.
Dayan, Daniel, "The Tutor-Code of Classical Cinema", in B. Nichols
(ed.), Movies and Methods, vol. 1, 1985.
Deleuze, Gilles, Masochism: An Interpretation of Coldness and Cruelty,
New York: Braziller, 1971.
Cinema 2: The Time-Image, trans. H. Tomlinson and R. Galeta, London:
Athlone, 1989.
Dennett, Daniel, Brainstorms: Philosophical Essays on Mind and Psychol-
ogy, Montgomery, Vt.: Bradford, 1978.
"Real Patterns", Journal of Philosophy, 88 (1991): 27-51.
Dick, Bernard F., Anatomy of Film, 2d ed., New York: StMartin's, 1990.
Durgnat, Raymond, review of Metz's Psychoanalysis and Cinema, Film
Quarterly, 36, no. 2 (1982-3): 58-64.
"Theory of Theory- and Bufiuel the Joker", Film Quarterly, 44 (1990):
32-44·
Dworkin, R., A Matter of Principle, Cambridge, Mass.: Harvard Univer-
sity Press, 1985.
Law's Empire, London: Fontana, 1986.
Eco, Umberto, A Theory of Semiotics, Bloomington: Indiana University
Press, 1979.
"On the Contribution of Film to Semiotics", in G. Mast and M. Cohen
(eds.), Film Theory and Criticism, 3d ed., New York: Oxford Uni-
versity Press, 1985.
Ellis, John, Visible Fictions: Cinema, Television, Video, London: Routledge
& Kegan Paul, 1982.
Fish, Stanley, Is There a Text in this Class? The Authority of Interpretive
Communities, Cambridge, Mass.: Harvard University Press, 1980.
Doing What Comes Naturally: Change, Rhetoric and the Practice of Theory
in Literary and Legal Studies, Durham, N.C.: Duke University Press,
1989.
Flitterman, Sally, "Woman, Desire, and the Look: Feminism and the
Enunciative Apparatus in Cinema", Cine-Tracts, 2 (1978), reprinted
288
Bibliography
in Caughie, John (ed.), Theories of Authorship: A Reader, London:

Routledge & Kegan Paul, 1981.
Fodor, Jerry, The Language of Thought, New York: Crowell, 1975.
"Observation Reconsidered", Philosophy of Science, 51 (1984): 23-43.
"Precis of The Modularity of Mind", Behavioral and Brain Sciences, 8
(1985): 1-42.
"Meaning and the World Order", in Psychosemantics: The Problem of
Meaning in the Philosophy of Mind, Cambridge, Mass.: MIT Press,
1987.
Friedberg, Anne, "Les Fldneurs du Mal(l): Cinema and the Postmodem
Condition", PMLA, 106, no. 3 (May 1991): 419-431.
Genette, Gerard. Narrative Discourse: An Essay in Method, trans. J. E.
Lewin, Ithaca, N.Y.: Cornell University Press, 198o.
Gilman, Daniel, "Pictures in Cognition", Erkenntnis, 41, no. 1 (July
1994): 87-102.
Glymore, Clark. "Freud's Androids", in Jerome Nue (ed.), The Cam-
bridge Companion to Freud, Cambridge University Press, 1991.
Goldman, Alvin, "Interpretation Psychologized", Mind and Language,
4, no. 3 (1989): 161-185.
"The Psychology of Folk Psychology'', Behavioral and Brain Sciences,
16, no.1 (1993): 15-28.
Goldman, Alvin (ed.), Readings in Philosophy and Cognitive Science, Cam-
bridge, Mass.: MIT Press, 1993.
Gombrich, E., "Image and Code: Scope and Limits of Conventionalism
in Pictorial Representation", in The Image and the Eye: Further Stud-
ies in the Psychology of Pictorial Representation, Oxford: Phaidon,
1982.
Goodman, Nelson, Languages of Art: An Approach to a Theory of Symbols,
Indianapolis: Hackett, 1976.
Gopnik, Allison, "What is It if It Isn't a Theory?" in Peter Carruthers
and Peter K. Smith (eds.), Theories of Theories of Mind, Cambridge
University Press, 1995.
Gordon, Robert M., "Folk Psychology as Simulation", Mind and Lan-
guage, 1 (1986): 158-171.
Grice, Paul, "Logic and Conversation", in Studies in the Way of Words,
Guarniero, G., "Experience of Tactile Vision", Perception, 3 (1974): 101-
104.
Hacking, Ian, "Do We See Through a Microscope?" in P.M. Church-
land and C. Hooker (eds.), Images of Science: Essays on Realism and
Empiricism, University of Chicago Press, 1985.
Bibliography
Halliwell, Leslie, The Filmgoer's Companion, 3d ed., New York: Hill &
Wang, 1970.
Hardin, C. L., "Color Subjectivism", in A. Goldman (ed.), Readings in
Philosophy and Cognitive Science.
Harman, Gilbert, "Eco Location", in M. Cohen and G. Mast (eds.), Film
Theory and Criticism, 2d ed., New York: Oxford University Press,
1979·
"Semiotics and the Cinema: Metz and Wollen", in M. Cohen and G.
Mast (eds.), Film Theory and Criticism, 2d ed., New York: Oxford
Harvey, Nigel, ''Wishful Thinking Impairs Belief-Desire Reasoning: A
Case of Decoupling Failure in Adults?" Cognition, 45 (1992): 141-
162.
Haugeland, John, "The Nature and Plausibility of Cognitivism", Be-
havioral and Brain Sciences, 2 (1978): 215-26o.
Heal, Jane, "Replication and Functionalism", in J. Butterfield (ed.), Lan-
guage, Mind, and Logic, Cambridge University Press, 1986.
Henderson, B., "Tense, Mood and Voice in Film (Notes after Genette)",
Film Quarterly, 36, no. 4 (1983): 4-17.
"Two Types of Film Theory'', in B. Nichols (ed.), Movies and Methods:
An Anthology, vol. 1, Berkeley and Los Angeles: University of Cal-
ifornia Press, 1985.
Hirsch, E. D., Validity in Interpretation, New Haven, Conn.: Yale Uni-
Holton, Richard, "Intentions, Response-Dependence and Immunity
from Error", in Peter Menzies (ed.), Response-Dependent Concepts:
Working Papers in Philosophy, no. 1, Canberra: Research School of
Social Sciences, Australian National University, 1991.
Humphreys, Glyn W., and Jane M. Riddoch (eds.), Visual Object Proc-
essing: A Cognitive Neuropsychological Approach, London: Erlbaum,
1987-
Irwin, Timothy, Meaning and Convention, M.A. thesis, Otago University,
1g88.
Jackson, Frank, Perception: A Representative Theory, Cambridge Univer-
sity Press, 1977.
"Metaphysics by Possible Cases", in Brian Garrett and Peter Menzies
(eds.), 1992 ANU Metaphysics Conference, Working Papers in Philos-
ophy, no. 2, Canberra: Research School of Social Sciences, Austra-
lian National University, 1992.
Jackson, Frank and Phillip Pettit, "Functionalism and Broad Content",
Mind, 97 (1g88): 381-400.
Bibliography
Jarvie, Ian, Philosophy of the Film, London: Routledge & Kegan Paul,
1987.
Jeannerod, M. "The Representing Brain: Neural Correlates of Motor
Intention and Imagery'', Behavioral and Brain Sciences, 17 (1994):
187-202.
Johnson, C. N., and PaulL. Harris, ''Magic: Special but Not Excluded",
British Journal of Developmental Psychology, 12 (1994): 35-51.
Johnston, Mark, "Dispositional Theories of Value", Aristotelian Society
Supplementary Volume, 63 (1989): 139-174.
Jones, R. K., and M. A. Hagen, "A Perspective on Cross-Cultural Pic-
ture Perception", in M. A. Hagen (ed.), The Perception of Pictures,
vol. 2, New York: Academic, 1980.
Juhl, P. D., Interpretation: An Essay in the Philosophy of Literary Criticism,
Princeton, N.J.: Princeton University Press, 198o.
Kanner, L., "Autistic Disturbances of Affective Contact", Nervous Child,
2 (1943): 217-250.
Khatchadourian, Haig, "Remarks on the 'Cinematic/Uncinematic Dis-
tinction in Film Art'", in Music, Film and Art, New York: Gordon
Breach, 1985.
"Space and Time in Film", British Journal of Aesthetics, 27 (1987): 169-
177·
Kozloff, Sarah, Invisible Storytellers: Voice-Over Narration in American
Fiction Film, Berkeley and Los Angeles: University of California
Press, 1988.
Lapsley, Robert, and Michael Westlake, Film Theory: An Introduction,
Manchester University Press, 1988.
le Poidevin, Robin, "Time and Truth in Fiction", British Journal of Aes-
thetics, 28 (1988): 248-258.
Leslie, Alan, "Pretence and Representation: The Origins of 'Theory of
Mind'", Psychological Review, 94, no. 4 (1987): 412-426.
"Pretending and Believing: Issues in the Theory of ToMM", Cogni-
tion, 50 (1994): 211-238.
Leslie, Alan., T. German and F. Happe, "Even a Theory-Theory Needs
Information Processing: ToMM, an Alternative Theory-Theory of
the Child's Theory of Mind", Behavioral and Brain Sciences, 16
(1993): 57·
Levinson, Jerrold, "Seeing, Imaginarily, at the Movies", Philosophical
Quarterly, 43, no. 170 (1993): 7o-78.
Levinson, Jerrold, and Philip Alperson, What Is a Temporal Art? Vol. 16
of Midwest Studies in Philosophy, French et al., ed. Notre Dame
Ind.: Notre Dame University Press, 1991.
Bibliography
Lewis, David K., Convention, Cambridge, Mass.: Harvard University

Press, 1969.
"Veridical Hallucination and Prosthetic Vision", in Philosophical Papers,
vol. 2, New York: Oxford University Press, 1986. First published in
Australasian Journal of Philosophy, 58, no. 3 (1980): 239-249.
Postscript to "Truth in Fiction", Philosophical Papers, vol. 1, New
York: Oxford University Press, 1983.
"The Paradoxes of Time Travel", Philosophical Papers, vol. 2, New
"Counterfactual Dependence and Time's Arrow", Philosophical Pa-
pers, vol. 2, New York: Oxford University Press, 1986.
Lotman, Jurij, Semiotics of the Cinema, trans. Mark E. Suino, Michigan
Slavic Contributions, no. 5, Ann Arbor: University of Michigan
Press, 1976.
Lycan, W. G., "Form, Function, and Feel", Journal of Philosophy, 78
(1981): 24-50.
Lyons, John, Language and Linguistics: An Introduction, Cambridge Uni-
Mandler, J. M., Stories, Scripts and Scenes: Aspects of Schema Theory, Hills-
dale, N.J.: Erlbaum, 1984.
Margolis, Joseph (ed.), The Worlds of Art and the World, Amsterdam:
Rodopi, 1984.
Marr, David, Vision: A Computational Investigation into the Human Rep-
resentation and Processing of Visual Information, New York: Freeman,
1982.
Martin, Edwin, "On Seeing Walton's Great-Grandfather", Critical In-
quiry, 12 (1986): 796-800.
Martin, Wallace, Recent Theories of Narrative, Ithaca, N.Y.: Cornell Uni-
Mast, Gerald, Film/Cinema/Movie: A Theory of Experience, New York:
Harper & Row, 1977.
Mast, G., and M. Cohen (eds.), Film Theory and Criticism, 3d ed., New
Maynard, Patrick, "Drawing and Shooting: Causality in Depiction",
Journal of Aesthetics and Art Criticism, 44 (1985--6): 115-129.
McTaggart, J. E., The Nature of Existence, vol. 2, chapter 33, Cambridge
Mellor, D. H., "The Unreality of Tense", in R. Le Poidevin and M.
MacBeath (eds.), The Philosophy of Time, New York: Oxford Uni-
Mendilow, A. A., Time and the Novel, New York: Humanities, 1972.
Bibliography
Metz, Christian, Film Language: A Semiotics of the Cinema, trans. M. Tay-
lor, New York: Oxford University Press, 1974.
Language and Cinema, trans. Donna J. Umiker-sebeok, The Hague:
Mouton, 1974.
"Trucage and the Film", in W. J. T., Mitchell (ed.), The Language of
Images, University of Chicago Press, 198o.
The Imaginary Signifier: Psychoanalysis and the Cinema, trans. Celia Brit-
ton et al., Bloomington: Indiana University Press, 1982.
"Some Points in the Semiotics of Cinema", in G. Mast and M. Cohen
"On the Notion of Cinematographic Language", in B. Nichols (ed.),
Movies and Methods: An Anthology, vol. 1, Berkeley and Los An-
geles: University of California Press, 1985.
"The Impersonal Enunciation, or the Site of Film (in the Margin of
Recent Works on Enunciation in Cinema)", New Literary History,
22, no. 3 (Summer 1991): 747-772.
Millikan, Ruth Garrett, Language, Thought, and Other Biological Catego-
ries, Cambridge, Mass.: MIT Press, 1984.
Mitry, Jean, Esthetique et psychologie du cinema, I: les structures, Paris:
Editions Universitaires, 1963.
Moreno, Julio, "Subjective Cinema: The Problem of Film in the First
Person", Quarterly Review of Film, Radio and Television, 7 (1952-3):
341-372.
Mulvey, Laura, ''Visual Pleasure and Narrative Cinema" in G. Mast
and M. Cohen (eds.), Film Theory and Criticism, 3d ed., New York:
Oxford University Press, 1985.
Miinsterberg, Hugo, "The Means of the Photoplay'', in G. Mast and M.
Cohen (eds.), Film Theory and Criticism, 3d ed., New York: Oxford
Nehamas, Alexander, "The Postulated Author: Critical Monism as a
Regulative Ideal", Critical Inquiry, 8 (1981): 133-149.
Newman, E., and P. Hartline, "The Infrared 'Vision' of Snakes", Sci-
entific American, 246 (March 1992): 98--107.
Nichols, Bill (ed.), Movies and Methods: An Anthology, vols. 1 and 2,
Berkeley and Los Angeles: University of California Press, 1976,
1985.
Ogle, Patrick, "Technological and Aesthetic Influences on the Devel-
opment of Deep-Focus Cinematography in the United States", in
Bill Nichols (ed.), Movies and Methods, vol. 2, Berkeley and Los
Angeles: University of California Press, 1985.
293
Bibliography
Panofsky, Erwin, "Style and Medium in the Motion Pictures", in G.

Mast and M. Cohen (eds.), Film Theory and Criticism, 3d ed., New
Peacocke, Christopher, Sense and Content: Experience, Thought and Their
Relations, Oxford University Press, 1983.
Petitto, L. A., "On the Autonomy of Language and Gesture: Evidence
from the Acquisition of Pronouns in American Sign Language",
Cognition, 27 (1987): 1-52.
Pettersson, Anders, "On Walton's and Currie's Analysis of Literary
Fiction", Philosophy and Literature, 17 (1993): 84--97.
Pettit, Philip, "Realism and Response-Dependence", Mind, 100 (1991):
587-623.
Phelan, James, Reading People, Reading Plots: Character, Progression and
the Interpretation of Narrative, University of Chicago Press, 1989.
Piattelli-Palmarini, M., "Ever Since Language and Learning: After-
thoughts on the Piaget-Chomsky Debate", Cognition, 50 (1994):
315-346.
Pitcher, George, A Theory of Perception, Princeton, N.J.: Princeton Uni-
Premack, D., Gavagai! Or the Future History of the Animal Language Con-
troversy, Cambridge, Mass.: MIT Press, 1986.
''Minds with and without Language", in L. Weiskrantz (ed.), Thought
Without Language, Oxford University Press, 1988.
Premack, D., and G. Woodruff, "Does the Chimpanzee have a Theory
of Mind?" Behavioral and Brain Sciences, 1 (1978): 515-526.
Prince, G., A Dictionary of Narratology, Lincoln: University of Nebraska
Press, 1987.
Pylyshyn, Zenon, "Computation and Cognition: Issues in the Foun-
dations of Cognitive Science", Behavioral and Brain Sciences, 3
(198o): 111-132.
"Imagery and Artificial Intelligence", in Ned Block (ed.), Readings in
Philosophical Psychology, vol. 2, Cambridge, Mass.: Harvard Uni-
Ray, Robert, A Certain Tendency of the Hollywood Cinema, Princeton, N.J.:
Princeton University Press, 1985.
Reisz, Karel, and Gavin Millar, The Technique of Film Editing, London:
Focal, 1968.
Rimmon-Kenan, Shlomith, Narrative Fiction: Contemporary Poetics, Lon-
don: Methuen, 1983.
Robinson, Jenefer, "Style and Personality in the Literary Work", Phil-
osophical Review, 94 (1985): 227-247.
294
Bibliography
Roth, D., and A. Leslie, "The Recognition of Attitude Conveyed by
Utterance: A Study of Preschool and Autistic Children", British
Journal of Developmental Psychology, 9 (1991): 315-3.30·
Sartwell, Crispin, "Natural Generativity and Imitation", British Journal
of Aesthetics, 31 (1991): 58-67.
Savage-Rumbaugh, S., et al., "Spontaneous Symbol Acquisition and
Communicative Use by Pigmy Chimpanzees Pan paniscus'', Journal
of Experimental Psychology: General, 115 (1986): 211-235·
Schier, Flint, Deeper into Pictures: An Essay on Pictorial Representation,
Scruton, Roger, "Photography and Representation", in The Aesthetic
Understanding: Essays in the Philosophy of Art and Culture, London:
Methuen, 1983.
Sesonske, Alexander, "Time and Tense in Cinema", Journal of Aesthetics
and Art Criticism, 38 (1979'-'8<>): 419-426.
Silverman, Kaja, The Subject of Semiotics, New York: Oxford University
Press, 1983.
Skow, John, "Harold Brodkey's New Novel is Erotic- But Not to Ev-
erybody", Time, May 9, 1994.
Snyder, Joel, "Photography and Ontology'', in J. Margolis (ed.), The
Worlds of Art and the World, Amsterdam: Rodopi, 1984.
Sobchack, Vivian, The Address of the Eye: A Phenomenology of Film Ex-
perience, Princeton, N. J.: Princeton University Press, 1992.
Sorensen, Roy, "Self-Strengthening Empathy'', unpublished manu-
script.
Sparshott, Francis, "Basic Film Aesthetics", in G. Mast and M. Cohen
Stalnaker, Robert, Inquiry, Cambridge, Mass.: MIT Press, 1984.
Stampe, Dennis, "Toward a Causal Theory of Linguistic Representa-
tion", Midwest Studies in Philosophy, vol. 2, Morris: University of
Minnesota Press, 1977·
Stanzel, F. K., A Theory of Narrative, trans. Charlotte Goedsche, Cam-
bridge University Press, 1984.
Stephenson, R., and J. R. Debrix, The Cinema as Art, New York: Penguin,
1976.
Sternberg, Meir, Expositional Modes and Temporal Ordering in Fiction, Bal-
timore, Md.: Johns Hopkins University Press, 1978.
Stich, S., and S. Nichols, "Second Thoughts on Simulation", in M. Da-
vies and T. Stone (eds.), Mental Simulation, forthcoming.
Studlar, Gaylan, "Masochism and the Perverse Pleasures of the Cin-
295
Bibliography
ema", in Bill Nichols (ed.), Movies and Methods: An Anthology, vol.

2, Berkeley and Los Angeles: University of California Press, 1985.
Todorov, Tzvetan, The Poetics of Prose, trans. Richard Howard, Ithaca,
N.Y.: Cornell University Press, 1977.
Turim, Maureen, Flashbacks in Film: Memory and History, New York:
Routledge, 1989.
Vermazen, Bruce, "Expression as Expression", Pacific Philosophical
Quarterly, 67 (1986): 196-224.
Vygotsky, L. S., "Play and Its Role in the Mental Development of the
Child", Soviet Psychology, 5 (1967): 6-18.
Walton, Kendall, "Transparent Pictures: On the Nature of Photographic
Realism", Critical Inquiry, 11 (1984): 246-277.
"Looking Again Through Photographs: A Response to Edwin Mar-
tin", Critical Inquiry, 12 (1986): 8ot-8o8.
Mimesis as Make-Believe: On the Foundations of the Representational Arts,
Warburton, Nigel, "Seeing Through 'Seeing Through Photographs'",
Ratio, new series, 1 (1988): 64-74.
Weiss, Paul, Cinematics, Carbondale: Southern Dlinois University Press,
1975·
Wellman, H., and K. Bartsch, "Young Children's Reasoning about Be-
liefs", Cognition, 30 (1988): 239-277.
Whyte, Jamie, ''The Normal Rewards of Success", Analysis, 51 (1991):
65-'73·
Wilensky, R., "Story Grammars Revisited", Journal of Pragmatics, 6
(1982): 423-432.
Williams, Bernard, "Imagination and the Self", in Problems of the Self,
Wilson, George M., Narration in Light: Studies in Cinematic Points of View,
Baltimore, Md.: Johns Hopkins University Press, 1986.
Wimmer, Hans, and Josef Perner, "Beliefs about Beliefs: Representation
and Constraining Function of Wrong Beliefs in Young Children's
Understanding of Deception", Cognition, 13 (1983): 103-128.
Wollheim, Richard, "Imagination and Identification", in On Art and the
Mind: Essays and Lectures, London: Lane, 1973.
Art and Its Objects, 2d ed., Cambridge University Press, 1980.
Painting as an Art, London: Thames & Hudson, 1987.
Woolley, Jacqueline D., and Katrina E. Phelps, "Young Children's Prac-
tical Reasoning about Imagination", British Journal of Developmental
Psychology, 12 (1994): 53--67.
Index
Affron, C., 165n Birds, The (Hitchcock), 176-7

Akins, K. A., 59n Black Widow (Rafelson), 212n
Alperson, P ., 92n, 101n Blow Up (Antonioni), 52
ambiguous narrative, 272-274; and Boghossian, P. A., 3m, 39n
unreliable narrative, 275-278 Booth, W., 245n; on narrative
anachrony: filmic and ellipsis, 220; and unreliability, 264-266, 270o-272,
its direction, 213-218; and tense, 275
199-200, 201-2o6;and untensed Bordwell, D., xilin, xxii, 105n, 107,
temporal relations, 207-2o8; weak, 190n, 204Jli on ambiguity, 27~; on
211-212 implied author, 24/248
Anna Karenina (Tolstoy), 152 Branigan, E., 2on, 1o6n, 16~, 265n; on
Anstis, 5., 45n point of view, 193-194
Amheim, R., 19-20, 5~ Brennan, A., xxi
Attack of the Fifty Foot Woman (Guest), Brown, E. L., 46n
103 Browne, N., 26n, 175-176n
Bryson, N ., xvii
Bach-y-Rita, P ., 59n Buchan, J., 266
Bal,M., 9~ Burch, N., 216
Balazs, B., 23, 166, 171, 178, 200
Baron-Cohen, 5., 143n, 15~ Camus, A., 265
Bartsch, K., 145n Carroll, N., xiiin, xxii, 27, Son, 16on,
Battleship Potemkin (Eisenstein), 104 16m, 1~
Baudry, J.-L., 30n Castle, W., 167, 168n
Bazin, A., on transparency and realism, Cavell, 5., 1n, 48, 50n
xxiii, 19-21, 5o-51, 79, 165-166, 187, Chatman, 5. 1411, 212n, 222, 244n, 245n,
191-192 26o, 26211, 265n, 267-268, 269n
behaviorism, 25, 235-236 Chomsky, N., xxiii, 115n
Bergman, 1., 23, 48 Churchland, P. M., 29n, 6411
Berkeley, G., 113 cinematic image: the reality of, 32-34;
Biederman, 1., 123n, 136n as representation, 2, 9-12 (see also
Big Sleep, The (Hawks), 272-274 pictures); as tensed and anachrony,
Bigelow, J., xxi 198-2o6
297
Index
cinematic movement: as apparent but film, 1-16; and controlling narrator,
nonillusory, 42-46; the reality of, 34- 266-268; documentary I fiction, 14-16;
42 engendering cognitive illusion, 22-23,
Citizen Kane (Welles), 14, 181, 266 27; engendering perceptual illusion,
Cohen, T., 58n 29-30, 35, 39; and imagining
colour: and cinematic image, 33; and someone's visual experience, 193-
film, 6-7; the reality of, 31-32, 39- 196; and impersonal imagining, 179-
40 18o; and personal imagining, 165-
Comolli, J.-L., 26n, u6n 179;realism, 19-22, 1o6-1o8;and
Constable, J., 52 seeing the world, 48, 69; as spatial
Cooper, N., xxi art, 103-106, 218-219; as temporal
Corballis, M. C., 123n art, 9(), 99-103, 218-219; unlike
Couvalis, G., xxi, 2m language, 114, 117-124, 128-131, 134-
Crossfire (Dmytryk), 205 135; see also photography
Culloden (Weir), 173 Fish, 5., 126n
Flitterman, 5., 2~
Dagle, J., 200n Fodor, J. A., 29n, 8m, 85-86, 8~
Davies, M., xxi-xxii, 53n Ford, F. M., 265, 2j'Sn
Davies, W. M., 29n, 59n, 6~ Freud, 5., xxiv
Dawkins, R., 59 Friedberg, A., xviiin
Dayan, D., xixn, 194
Debrix, J. R., 2oon Galileo, G., 111n
Deffenbacher, K., ~n Genette, G.,~' 199n, 20~, 212n,
Deleuze, G., 141n, 201n; on film and 262n, 263n
physics, xviii-xix Gilman, D., 46n, 86n
Dennett, D., 37-38, 8~ Glymore, C., xvn
Deren, M., 21~ Godfather, The, Part 2 (Coppola), 216n
Descartes, R., 83 Goldman, A. 1., 143n, 153n
Dick, B. F., 216n Gombrich, E., xvi
Dickens, C., 229 Gone with the Wind (Fleming), 14, 273-
Dombey and Son (Dickens), 228 274
Don't Look Now (Roeg), 216 Good Soldier, The (Ford), 265
Dostoyevsky, F., 244-245 Goodman, N., xvi, 12n
Double Indemnity (Wilder), 216n, 266 Gopnik, A., 143n
Durgnat, R., xviin Gordon, R. M., 143n, 154n
Dworkin, R., 225, 249n Gospel According to St. Matthew, The
(Passolini), 99
Eco, u., 11~, 131-133 Grant, C., 23, 38, 40'-41, 48, 189
Eisenstein, 5., 104 Great Expectations (Lean), 181
ellipsis, 212, 215, 220; in literture, 219- Gregory, R., xvi
222 Grice, P., 12~, 215n, 253n
Ellis, J., 22n Groner, C., 59n
Fall, The (Camus), 265 Hacking, 1., 59n

Feagin, 5., 46n Hagen, M. A., non
fiction: doing moral damage, 162-163; ''Haircut" (Lardner), 232-233, 27o-271
and imagination, 141, 147-148, 150, Haldane, J., xxi
152-155, 158-16o; and interpretation, Halliwell, L., 2o6n
242; visual, 169-170, 181-185 Hamlet (Shakespeare), 95
Index
Harman, G., 13211 knowledge of other minds, xv, 142-1~,
Harris, P. L., 163n 150, 158, 161
Hartline, P., 6411 Kozloff, 5., 269n
Harvey, N., 163n
Haugeland, J., 8411 Lacan, J., xiv, xvii, xix, 141
Heal, J., xxi, 143n LAdy from Shanghai, The (Welles), 128--
Henderson, B., xviin, 199n 129
Hills, D., xxi Lang. F., 131, 268
Hirsch, E. D., 244n Lapsley, R., 1nn
Hitchcock, A., 52, 153, 170, 174, 176, Lardner, R., 232-233, 270o-271
18<}, 195, 259· 267 lAst Year at Marienbad (Resnais), 101-
Holton, R., 43n 102, 21411
Hopkins, R., xxi lAwrence of Arabia (Lean), 5
Hudson, W., non le Poidevin, R., 20~
Humphreys, G. W., &)n Leslie, A, 142n, 143n, 15~. 161n
Hunt, 1., xxii Levinson, J., xxii, 3~, 9211, 101n, 17411
Lewis, D. K., xxii, 53n, 55n, 66n; on
imagination, 141-142; and belief and anachrony and time travel, 2o8--209;
desire, 151; perceptual and symbolic, on convention, 121; on flash-
184; personal and impersonal, 166; as stockman problem, 158--159; on time
simulation, 144-147, 149-150, 155- in fiction, 2n
161 Lippi, F., 95-96, 99
Incredible Shrinking Man, The (Arnold), Livingston, P., xxii
103 Lotman, J., 1nn, 2oon
Innocents, The (Clayton), 216n, 276n Lumiere Brothers, 24
interpretation: 226, 247-249; and Lycan, W. G., 84fl
appearance meaning. 255-258; Lyons, J., n~
indeterminacy in and truth, 25o-251;
and interpretive differences, 227-228; M (Lang), 131
literary, 237-2.42; narrative, 228, 23o- Malebranche, N., 62-63, 65
235; and real/implied author, 243- Malinas, G., xxii
246, 258--259; and utterance meaning. Mandler, J. M., 248n
255-258 Marr, D., 69n, 85n
Invasion of the Body Snatchers, The Martin, E., 55n, 5~
(Siegel), 186 Martin, W ., 266n
Irwin, T., 121n Mast, G., 3n, 411, 101n, 201n
lshiguro, K., 265 Maynard, P ., 5on, 76n
McTaggart, J. E., 2o6--207
Jackson, F., xxii, 36n, 18211, 240n Medlin, B., 4211
James, H., 234, 240 Mellor, D. H., xxii, 198n
Jarvie, 1., 23n, 58n, 200n Mendilow, A. A., 201n
Jeannerod, M., 163n Meshes of the Afternoon (Deren), 21411
Johnson, C. N., 163n Metz, C., xvi, xviin, xixn, 2411, 26n, 2~,
Jones, R. K., non no, n6n, n9n, 175n, 218n, 262n,
Juhl, P. D., 245n 268n
Millar, G., 190
Kanner, L., 15~ Millikan, R. G., 53n
Khatchadourian, H., 3411, 169 Mitry, J., 165-166, 168
Kitcher, P ., xxii Moreno, J., 16~
Index
Mulvey, L., xixn Prince, G., 265n
Miinsterberg, H., xxili, 34 Profane Friendship (Brodkey), 15~
psychoanalysis, xii, xiv, xv, xix, 27, 165,
Nabokov, V., 238 174-175
narrator, 262-263; controlling and Pudovkin, V., 190n
embedded, 265-266; reliable, 264; Pylyshyn, Z., 29, ~
unreliable, 237-238, 240, 264; and
unreliable narration, 26o-262, 265, Rafelson, R., 212n
268-270 Rashomon (Kurosawa), 266, 275
Nehamas, A., 245n Ray, R. B., 22n
Nerlich, G., xxii, 38n Rear Window (Hitchcock), 52
Newman, E., 6~ Reisz, K., 190
Nichols, S., 143n Remains of the Day, The (lshiguro),
North by Northwest (Hitchcock), 189-190 265
Renoir, J., 187
Occurrence at Owl Creek, An (Enrico), representation, 49; automorphic and
269n homomorphic, 97-98; linguistic, xvi-
Ogle, P., 1o8n xvii, 7...S, 113-137, 181-185, 189-190,
251-258; visual and pictorial, 7...S;
painting: as representational medium 1o-12
and art, 49, 51, 53-55, 61, 71, 77-'!8; Resnais, A., xviii, 198, 21~, 26~
and temporality, 95-96, 99 Rey, G., xxii
Panofsky, E., 23, 166, 168, 171, 178 Riddoch, J. M., 89n
Passolini, P. P ., 99 Rimmon-Kenan, S., 271n
Peacocke, C. A. B., 29n, 6~ Robbe-Grillet, A., xvili, 200
Peeping Tom (Powell), 261268 Robinson, J., 245n
Perkins, V., 164, 201n Rosen, G., xxii
Pemer, J., 142n 15~ Roth, D., 15~
Petitto, L. A., u6n Rules of the Game, The (Renoir), 187
Petterson, A., 161n
Pettit, P., xxii, 3m, 240n Saint, E. M., 189
Phelan, J., 27m Sartwell,C.,88n, 131n
Phelps, K. E., 163n Saussure, F., xxili
photography: as representational Savage-Rumbaugh, S., 123n
medium, 12-13, 51, 71, 77; and Schier, F., Son, 83n, 88, 131n
seeing the world, 48, 5o-51, 53-56, Scruton, R., 5on, 72--74t 75n, 13on
63-64t 66-68, 70 Searchers, The (Ford), 257
Piaget, J., 115n seeing, 53, 55, 57; egocentric and
Piattelli-Palmarini, M., 115n temporally extended information
pictures: and perceptual imagining, from, 66-68; and light, 59; and
184-185, 197; and representation, natural dependence, 61-62; paintings
xvii, So--81, 84...S7 and photographs, 53-62; and
Pigden, C., xxii uninterrupted transmission, 6o, 70
Pink Flamingos (Waters), 16~ semiotics, xvi-xvii, 116-117, 165
Pitcher, G., 182n Sesonske, A., 201n
Popper, K., xvi Sharp, R., xxii
Poussin, N., 95 Silverman, K., xvn
Premack, D., non, 123n simulation: and belief/desire, 149--151;
Price, A., xxii and fiction, 152-155;and
300
Index
imagination, 141148, 151-152; and intentions of implied author, 270o-
interpretation, 242-243; and other 272
minds, 144-147
Sirk, D., 259 Velleman, J. D., 31n, 39n, 111n
Skow, J., 15~ Verrnazen, B., 245n
Snyder, J., 50n Vertigo (Hitchcock), 170o-171
Sobchack, V., xviii visual perception, xv; as nonlinguistic,
Sober, E., xxii 136-137
Soli, I., xxii Vygotsky, L. S., 161
Sorensen, R., 146-147
Sparshott, F., 34- 20m Walton, K., xxii, 13n, 24n, 14~, 152n,
Spellbound (Hitchcock), 174-175, 179- 161n, 169n, 17~, 188n; on
tSo transparency, 20, 5o-51, 55n, 56n,
Squires, R., xxii 57n. 61-63, 75n, 173n
Stage Fright (Hitchcock), 195, 267-268 Wages of Fear, The (Clouzot), 1oon
Stagecoach (Ford), 175-176 War and Peace (Tolstoy), 246n
Stalnaker, R., 78n Warburton, N., 66n
Stampe, D., 78n Waters, J., 88, 167
Stanzel, F. K., 266n Weir, P., 173
Sternberg, M., 200n Weiss, P., 23n
Stich, S., 143n Welles, 0., 10
Stokes, J., xxii Wellman, H. M., 145n
structuralism: errors of, 114-116, 122n; Westlake, M., 111n
and narrative intentions, 278-279 Wetherall, P., xxii
Studlar, G., 141n White, M., 46n
Sturgeon, S., xxii Whyte, J., xxii, 149n
Wilder, B., 216n
Taylor, K., xxii Wilensky, R., 113n
They Shoot Horses, Don't They? Williams, B., 169, 178
(Pollack), 216n Williamson, T., xxii
Third Man, The (Reed), 1o-11 Wilson, G., xiiin; on cinematic images
Thompson, K., 27~ as unlike language, 128--130; on
Tingler, The (Castle), 168 imagining seeing, 165n, 166, 171-173,
Todorov, T., 115 178; on unreliability, 268--269, 276,
Tolstoy, L., 246n 278n
Tom Jones (Fielding), 262 Wimmer, H., 142n, 15~
Tooley, M., xxii Wittgenstein, L., 97, 123n, 136
Townsend, A., xxii Wollheim, R., 88, 90, 110n, 141, 169n
Turim, M. C., 22n, 198n, 2o6n Woodruff, G., 110n
Turn of the Screw, The (James), 232, 238, Woolley, J. D., 163n
240, 25o-251,261-262,276-277
Yalowitz, S., xxii
unreliable narrative, 261-262, 265; You Only Live Once (Lang), 268--270,
Booth, 268--270; in terms of complex 275-277
301

Currie, Gregory - Image and Mind PDF

Uploaded by

Copyright:

Available Formats

Currie, Gregory - Image and Mind PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Currie, Gregory - Image and Mind PDF

Uploaded by

Copyright:

Available Formats

This is a book about the nature of film: about the nature of

moving images, about the viewer's relation to film and about

.... ~····· CAMBRIDGE

© Cambridge University Press 1995

First published 1995

Library of Congress Cataloging-in-Publication Data

ISBN o-521-45356-9 Hardback

Transferred to digital printing 2004

Introduction: the essence of cinema 1

Part I Representation in film

Chapter 4 Languages of art and languages of film 113

5·4 Consequences 155

Part III Interpretation

9·3 Kinds of unreliability

This book aims to place film in relation to other things: to other

of style connects with the failure of doctrine: the lack of clarity

Crystalline narration will fracture the complementarity of a

I could go on. But extended polemic is a tedious thing; con-

Hunt, Frank Jackson, Philip Kitcher, David Lewis, Paisely Liv-

In writing this book I have tried to bring together certain aspects

Film is an art of recent invention. In fact it is just one hundred

Tired of the ponderous prescriptivism that dogged much film

Films, like samples of water, have something in common that

Representations come in a variety of kinds: words written and

I. 2 FILM AND THE VISIBLE

Films, like other kinds of pictures, are meant to be seen. The

• As Gerald Mast points out (ibid., p. 5).

1.3 DEGREES OF INESSENTIALNESS

colour is an obligatory determinable, while the precise choice of

1.4 THE VISIBLE AND THE PICTORIAL

It is one thing to say, as I have said, that the cinema is a visual

it as referring to Vienna, does not represent Vienna pictorially.

1.5 FICTIONAL PICTURES

There is a problem. Cinematic images certainly are represen-

have seems, intuitively, to be pictorial. Let us examine this idea

1.6 DOCUMENTARY AND FICTION

I have said that cinematic images, at least the photographically

It is easy to conflate the two representing functions of cine-

There are documentary films, just as there are literary docu-

So far, we have seen some of the things that are essential to

The myth of illusion

1.1 TRANSPARENCY, LIKENESS AND

Let us start by distinguishing three doctrines about cinema; all

nection with long-take, deep-focus style. But as I shall argue,

' See Kendall Walton, "Transparent Pictures". "Transparency" is used in other

sible our experience of that world. And many theorists have

1.2 COGNITIVE ILLUSIONS

capacity to make the viewer think that he or she is actually

132). In Chapter 4 I shall examine a way of taking the statements of Balazs,

The case is also not an example of the audience reacting to fic-

police. Nor is there in the experience of film watching any evi-

1. 3 FILM AND DREAMING

1.4 PERCEPTUAL ILLUSIONISM

provide some. But bear in mind a point already made: Percep-

1.5 THE REALITY OF CINEMATIC

I said in the Introduction that film is a pictorial medium.

the painting. The conditions for the presence of colours in the

1.6 ON THE MOTION OF IMAGES

So while I shall speak of our cinematic experience as represent-