Defending the Content
View of Perceptual
Experience
Defending the Content
View of Perceptual
Experience
By
Diego Zucca
Defending the Content View of Perceptual Experience
By Diego Zucca
This book first published 2015
Cambridge Scholars Publishing
Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
Copyright © 2015 by Diego Zucca
All rights for this book reserved. No part of this book may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording or otherwise, without
the prior permission of the copyright owner.
ISBN (10): 1-4438-8275-5
ISBN (13): 978-1-4438-8275-0
TABLE OF CONTENTS
Preface ......................................................................................................... x
Acknowledgments ..................................................................................... xii
Introduction ................................................................................................. 1
Chapter One: The Semantics of Seeing and Related “Experiential”
Predicates
Introduction ................................................................................................11
Part I .......................................................................................................... 13
A Methodological Remark
Part II ......................................................................................................... 15
Seeing Something
II.1 Basic Conditions
II.2 Some Objections
II.3 Transparency
Part III........................................................................................................ 26
Seeing That P, Seeing That a Is F
III.1 Propositionality
III.2 Factivity
III.3 Opacity
III.4 Literal/Metaphorical
III.5 Object-Seeing and Fact-Seeing
III.6 Conceptuality
III.7 Definition
Part IV ....................................................................................................... 36
Seeing-As, Seeing Something As Something
IV.1 Implicativity and Normative Evaluability
IV.2 Recognition
IV.3 “Thick” Categories and Sensible Profiles
IV.4 Definition
vi
Table of Contents
IV.5 What Is a Sensible Profile? The SCM-Properties
IV.6 The Features of Seeing-As
Part V ......................................................................................................... 46
Looking, Seeming, Appearing
V.1 Looks/Seems/Appears: Analogies and Differences
V.2 A Principle Governing “Looks” Ascriptions
V.3 Three Uses of “Looks”
V.4 Is There a Phenomenological “Looks”?
V.5 The Phenomenological “Looks” Is Not Independent
V.6 “Looks F” Depends on “Is F”
V.7 To Sum Up
Concluding Remarks
Chapter Two: Some Basic Features of Perceptual Experience
Introduction ............................................................................................... 67
Part I .......................................................................................................... 68
The Belief Theory: Philosophical Arguments For and Against
I.1 The Belief Theory
I.2 Virtues of the Belief Theory
I.3 The Argument From Illusion, the Belief Theory and the
Adverbialist View
I.4 The Problems of the Belief Theory. Objections and Possible Replies
I.5 The Belief Theory and the Phenomenological Adequacy Constraint
I.6 Beliefs, Inferences, Concepts: Rationality Constraint and Generality
Constraint
I.7 Further Difficulties of the Belief Theory
Part II ......................................................................................................... 88
Against the Belief Theory: Arguments from Experimental Evidence
II.1 Inattentional- and Change- Blindness: There Is Seeing Without
Noticing
II.2 The Sperling Experiment and What It Tells With Respect to the
Belief Theory
II.3 The Case of Visual Associative Agnosia and the Belief Theory
II.4 The Case of Optic Ataxia and the Belief Theory
II.5 The Case of Blindsight and the Belief Theory
Concluding Remarks
Defending the Content View of Perceptual Experience
vii
Chapter Three: The Content View
Introduction ............................................................................................. 105
Part I ........................................................................................................ 106
The Core Idea of the Content View (CV)
I.1 Introducing the Content View
I.2 The Content View and the Belief Theory
I.3 Phenomenal Character and Representational Content
I.4 Transparency and Richness of Details
I.5 The Scenario Content Introduced
Part II ........................................................................................................117
Some Prima Facie Virtues of the Content View
II.1 Distinctive Features of States with Intentional Content
II.2 Perceptual Experience and Accuracy
II.3 The Content View and Ordinary Semantics of “Seeing”
and “Looking”
Concluding Remarks
Chapter Four: The Content View Articulated
Introduction ............................................................................................. 129
Part I ........................................................................................................ 130
Layers and Components of Perceptual Content
I.1 Beyond the Scenario Content
I.2 Proto-propositional Content and Seeing-As
I.3 Scenario Content and Object-Seeing
I.4 Three Layers of Content
I.5 The Limits of Dretske’s Theory of Seeing
Part II ....................................................................................................... 149
Objects and Properties: How They Feature in Perceptual Content
II.1 Back from Recognition to Discrimination
II.2 Object-Seeing through Property Discrimination
II.3 Which Basic Semantic Ingredients Shape the Content of Visual
Perception?
II.4 Object-Dependency and Singularity
II.5 Demonstrative Contents and Semantic Gap
viii
Table of Contents
Chapter Five: Phenomenal Character and Kinds of Perceptual
Content
Part I ........................................................................................................ 174
Phenomenal Character and Representational Content
I.1 Reconsidering Looks within the Content View
I.2 The Case of Perceptual Constancy
I.3 How Are We Aware of the Mode of Our Perceptual Experiences?
I.4 From Content Externalism to Phenomenal Externalism?
Part II ....................................................................................................... 215
Fregean vs. Russellian Content
II.1 To Recap Where We Are
II.2 Chalmers’s Third Way: the Double-Content View
II.3 Weaknesses of the Double-Content View
II.4 Perceptual Content and Character Are Wide, External, Russellian
Concluding Remarks
Chapter Six: Seeing-As and the Range of Represented Properties
Part I ........................................................................................................ 232
Seeing-As between “Thin” Properties and “Thick” Properties
I.1 Sensible Profiles and the Properties They Are Made Out Of
I.2 Liberals and Conservatives
I.3 A “Thin” and a “Thick” System? The Dual Content View
Part II ....................................................................................................... 244
The Phenomenal Contrast Method
II.1 Introducing the Phenomenal Contrast Method
II.2 Cyrillic Words and Pine Trees
II.3 Does Visual Agnosia Support the Liberal View?
II.4 Do Ambiguous Figures Support the Liberal View?
Part III...................................................................................................... 266
The Dual Content View: Beyond the Phenomenal Contrast Method
III.1 The Dual Content View Principle
III.2 Preparatory vs. Constitutive Cognitive Processes
III.3 The (Blurry) Superior Bound of Perceptual Content
III.4 Perceptual Content Is Moderately Rich and Outstrips Phenomenal
Content
Concluding Remarks
Defending the Content View of Perceptual Experience
ix
Chapter Seven: Bringing the Disjunctivist Challenge into the
Intentionalist View
Introduction ............................................................................................. 297
Part I ........................................................................................................ 298
Disjunctivism Introduced
I.1 What Is Disjunctivism?
I.2 The Reasons for Disjunctivism: The Detachment Problem
Part II ....................................................................................................... 308
The Good, the Bad and the Neutral: A Moderately Disjunctive
Intentionalism
II.1 The Priority of the Successful
II.2 Function and Content
II.3 Where Do We Put Illusions?
II.4 A Problem: The “Function” of Hallucinations
II.5 The Semantic Gap of Hallucinatory Contents
II.6 Beyond the Detachment Problem
Concluding Remarks
Conclusions ............................................................................................. 357
Bibliography ............................................................................................ 359
Index of Authors ...................................................................................... 378
PREFACE
This book is a defense of the content view (CV) on perceptual experience.
The CV is the idea that our perceptual experiences represent the world as
being a certain way, so they have representational content. Three main
issues are addressed in this work.
First, I show that the CV fits very well with the logical behavior of
ordinary ascriptions of seeing-episodes and related experiential episodes,
as well as with our pre-theoretical intuitions about what perceiving and
experiencing ultimately are: this preliminary analysis speaks to the prima
facie plausibility of such a view.
Second, I put forward a detailed account of perceptual episodes in
semantic terms, by articulating a specific version of the content view. I
provide arguments for the following theses: perceptual content is twolayered so it involves an iconic level and a discrete or proto-propositional
level (which roughly maps the seeing-as ascriptions in ordinary practices).
Perceptual content is singular and object-dependent or de re, so it includes
environmental objects as its semantic constituents. The phenomenal
character of perceptual experience is co-determined by the represented
properties together with the mode (e.g., visual mode), but not by the
perceived objects: I label such a view as impure representationalism.
Perceptual content is “Russellian”: it consists of worldly objects, properties
and relations. Both perceptual content and phenomenal character are “wide”
or determined by environmental factors, thus there is no Fregean narrow
perceptual content. In addition, there are two layers of properties that can
be represented in perception: a “thin” layer—for example, for visual
perception: spatial, chromatic, morphological properties—and a thicker
layer, which may depend on perceptual learning and includes properties
other than the “thin” ones but is nevertheless not as “thick” as natural
kind-properties.
Third, I show that such a version of the CV can cope with the typical
objections put forward by the advocates of (anti-intentionalist versions of)
disjunctivism. I myself put forward a moderately disjunctivist version of
the CV, according to which perceptual relations (illusory or veridical)
must be told apart from hallucinations, and as mental states of a different
kind. Such disjunctivism is “moderate” insofar as it allows genuine
Defending the Content View of Perceptual Experience
xi
relational perceptual experiences and hallucinations to share a positive
phenomenal character, contrary to what radical-disjunctivism-cum-naïverealism believes.
Ultimately, this will show that the CV vindicates our pre-theoretical
intuitions and does justice to our ordinary ascriptive practices. I will
articulate a detailed and argued version of the CV, and show that such a
version is not vulnerable to the standard objections recently placed on it by
the disjunctive branch. This study can therefore be considered a global,
multifaceted argument for the CV
ACKNOWLEDGMENTS
I wish to thank first and foremost Theodor Scaltsass and Matthew Nudds
for their philosophical and methodological advice and constructive
criticism. I am grateful to Christopher Peacocke for the kind and helpful
support I received during an eight-month sabbatical spent at the Columbia
University of New York. While there I also had the good fortune to
encounter Achille Varzi, who made my visit particularly fruitful and
interesting. I owe a special debt to Luigi Ruggiu, who has followed and
encouraged my work during the years.
I have also benefited greatly from the passionate discussions, reading
groups, philosophical dinners and criticisms with Andrew McKinlay,
Jonas Christensen, Andreas Paraskevaides, Roberto Loss, and Matteo
Giannasi.
This work is dedicated to my love who made it possible, Carolina
Ozan.
INTRODUCTION
This book is about the content view (CV) on perceptual experience—in
particular, on visual experience—and about the idea that perceptual
experiences have representational content. Its global aim is to argue for a
certain version of the CV, one that can meet the desiderata of a
satisfactory theory of perceptual experience and be defended from the
main criticisms against it, especially from the disjunctivists. The discourse
is articulated in several arguments, discussions and specific proposals that
develop into three major lines of inquiry.
First, I will show that the CV fits well with ordinary ascriptions of
visual episodes and visual experiences. Ordinary ways of talking about
seeing and experiencing embed deep and pre-theoretical intuitions about
what the ascribed episodes and states are, or at least seem to be. Capturing
the commonsensical intuitions about matter and making sense of the
ordinary ways of talking about it are relevant virtues for a philosophical
theory, even if our pre-theoretical intuitions are wrong and our ordinary
ways of talking are confused. From a methodological point of view,
showing that the CV respects and vindicates these pre-theoretical
intuitions and ways of talking is not an arbitrary celebration of vulgarity.
Having such virtues does not amount to it being true, of course; still, a
theory that possesses such virtues is, ceteris paribus, to be preferred over a
theory that lacks them. Therefore, showing that the CV has these virtues
counts as a prima facie argument in favor of its plausibility.
Second, I will argue for a certain detailed version of the CV by
discussing the main issues raised within the debate on it and taking a stand
on each of them so as to produce a systematic picture involving arguments
and commitments. These issues concern the many types and layers of
perceptual content, the semantic structure and the way objects and
properties feature in perceptual content, the relation between phenomenal
character and representational content, the externalism/internalism debate
on perceptual content and character, the issue about whether perceptual
content is Fregean or Russellian and the issue about whether the properties
represented in perception are just “thin” (e.g., colors, shapes and spatial
properties for the visual mode) or also “thick.”
Third, I will try to defend the CV—after having spelled out a specific
and detailed version of it—from the principal criticisms of the recently
2
Introduction
revived naïve realism, a new form of radical disjunctivism, revolving
around the relation between successful and deceptive perceptions. I will
show that some of these criticisms, instead of defeating the theory, can be
embedded into it. Indeed, the CV I articulate is a form of moderate
disjunctivism. However, I will argue that some other criticisms, according
to which the CV should just be abandoned, can be addressed and coped
with. As a result, my specific version of the CV is vindicated and shown to
be the most promising, avoiding the problems ascribed to it and the
dismissal of it as such by its opponents.
To develop these three lines of inquiry organically, I will proceed as
follows.
Chapter one is the introduction. I systematically analyze the logical
and semantically relevant features of ordinary ascriptions and selfascriptions of visual episodes and experiences. First, I propose a list of
necessary and sufficient conditions for a subject to be said to be seeing
something, then I argue that such conditions are highly plausible. The
logical behavior of seeing-ascriptions is analyzed, like “S sees O”, “S sees
an F.” Second, I move on to consider how seeing-that ascriptions behave
and what seeing-that involves in cases like “S sees that P”, “S sees that a
is F.” I individuate certain features of ascriptive contexts (opacity, conceptinvolvement, propositionally, factivity) and argue that seeing-that is a
fully-fledged propositional attitude that amounts to coming to know by
visual means. Then I consider seeing-as ascriptions and their behavior,
expressed by a three-place relation as “S sees a as an F.” Seeing-as is
intermediate between object-seeing and seeing-that; it presupposes the
first and is presupposed by the second. I show that ascribing a seeing-as
episode amounts to ascribing a positive, recursive, vision-based
recognitional disposition, and I discuss the relation between such
ascriptions and the evaluability of such episodes as mistaken or accurate.
After evaluating object-seeing (seeing-X), propositional seeing (seeing
that P), and recognitional seeing (seeing a as an F), I move on to consider
the ascription of experiential predicates like “looking,” “seeming” and
“appearing” in their different uses, roles and applications in ordinary
sentences. In discussing look-ascriptions specifically—helped by the
relative literature—I make explicit the relationship between the different
senses in which “looking” may be ascribed an epistemic sense (it looks to
S as if a is F), a comparative sense (A looks like B) or a phenomenological
sense (“the penny looks elliptical to me from here”). I question the
independence of the phenomenological use of “looks.” Then I argue, by
re-articulating a point held by Sellars, that the understanding of “looks F”
conceptually and logically depends on the understanding of “is F”: I
Defending the Content View of Perceptual Experience
3
explore the ways of this dependence.
Such a taxonomic survey is a way of getting into perception with a
clearer grip on certain fundamental distinctions, both conceptual and
terminological. As Austin suggested, ordinary language is not the last
word; yet, it is the first.
In Chapter Two, I introduce the belief theory of perception and point
out some of its basic virtues. I start with introducing the belief theory itself
and locating it in the classical debate on perception between sense-data
theorists, direct realists and adverbialists. By showing its advantages in
treating perception in representational terms (beliefs are representations), I
consider its difficulties—the philosophical ones, on the one hand, and the
problems it encounters before experimental evidence is obtained on the
other. The philosophical problems are related to its phenomenological
inaptness and to the different behavior of perceiving something to be F
from believing that something is F (concept-involving, entailing inferential
sensitivity and demanding constraints on rationality). The experimental
evidence I have focused on (inattentional- and change blindness, the
Sperling experiment, visual associative agnosia, optic ataxia, blind sight)
suggests that there is seeing without noticing, seeing without believing,
and also belief-acquiring through perception without perceptual
experience, so perceptual experience cannot be reduced simply to beliefacquiring.
In Chapter Three, I go on to introduce the CV as a view that can embed
the virtues of the belief theory—as a semantic, representational account of
perception—without suffering from the philosophical and experimental
weaknesses focused on above. The first crucial move is that of introducing
the notion of non-conceptual content and substituting it for the doxastic
account involved in belief theory. I argue that if perceptions are considered
as non-conceptual representations, the CV can avoid all the difficulties
encountered by the belief theory. Non-conceptual content is
phenomenologically apt, does justice to the difference between something
looking F to S and S’s believing that something is F (this being conceptinvolving, entailing inference-sensitivity and rational capacities on the part
of S), and has no special problems with the experimental data (a nonconceptual representation is pre-doxastic, can occur without its content
being believed, can outstrip conscious attention, and so on). Then, I take
into consideration the relation between phenomenal character and
representational content to suggest that fineness of grain and the lack of
structure of non-conceptual content can do justice to perceptual
phenomenology, which is profuse and rich in details in a way a doxastic
state cannot be. I introduce Peacocke’s notion of scenario content as a
4
Introduction
very promising way of semantically characterizing perceptual content that
does justice to the distinctiveness of perceptual phenomenology.
Afterwards, I isolate and briefly discuss some general reasons for
favoring the CV, namely, some of its fundamental explanatory virtues with
respect to certain apparent features of perceptual experiences: aspect,
absence, accuracy, aboutness. Since the CV is in a position to account for
such apparent properties—representations typically exhibit such
features—it is a highly promising view worth taking very seriously. This is
not a trivial point because, surprisingly enough, the CV is very seldom
argued for as such. Rather, it is presupposed, and one or the other version
of it is defended or attacked.
Finally, I consider some interesting analogies with the CV and the
ordinary ways of ascribing seeing-episodes. The difference between “seeing
something” and “seeing that” maps the difference between perceptuallynon-conceptually representing and coming to believe by visual means that
things are a certain way. I will also argue that looking-ascriptions are
consistent with the representational conception of perceptual experience,
and that the CV vindicates our pre-theoretical intuition that our perceptual
experiences can be veridical, partially illusory or totally illusory. No nonrepresentational account of that intuitive matching/mismatching relationship
is available.
Chapters four to six are the pars construens core of this book, where I
examine the matter and articulate a certain version of the CV (based on the
options made available in the current debate, of course).
In Chapter Four, I first argue for a two-layered view of visual content
(Part I). Beyond the scenario content, which is specified as ways of filling
out the space around the perceiver and has spatial-chromaticmorphological properties, another semantic layer is introduced, the protopropositional content. With Peacocke, I argue that the scenario content
cannot capture all there is in perceptual representation; in particular,
certain acts of property-recognition can be present or absent without
impacting scenario content. I show that introducing a perceptual protopropositional content between the scenario content and the doxastic
content of perceptual beliefs maps the pre-theoretical necessity, testified in
ordinary ascriptive practices, to distinguish object-seeing from seeing-as,
and both of them from seeing-that. Indeed, seeing-as ascriptions basically
ascribe visual episodes with proto-propositional content. I criticize
Dretske’s theory of seeing, which distinguishes simple seeing from
epistemic seeing and fatally overlooks the intermediate level of seeing-as, or
recognitional seeing. Without that level, the semantic and epistemological
transition from object-seeing to visually-based propositional knowledge
Defending the Content View of Perceptual Experience
5
remains an unaccountable mystery.
Second, I argue for the object-dependency and singularity of visual
content (Part II) and against the generality thesis held by Searle, McGinn
and others. In my view, a visual experience is individuated by a subject, a
Content (uppercase) composed by a perceived object and by a set of
represented properties (the content, lowercase), a perceptual mode. I show
that the generality thesis is false so the singularity thesis must be true.
Visual perception involves particulars in its Content, so visual Contents
are de re, demonstrative Contents. Then I profile the big puzzle that the
singularity thesis opens with respect to hallucinatory contents since
hallucinations do not have worldly particulars as constituents of their
putative contents. I have labeled it the semantic gap problem, but I deal
with it systematically only in the last chapter (Chapter Seven, Part II.5).
In Chapter Five, I first defend a form of impure representationalism
about phenomenal character (Part I). In this view, the phenomenal
character of a conscious perception is made out of represented properties
but represented under a mode (e.g. the visual mode). So in my view there
is a dependence-without-reduction rather than an identity between the
phenomenal and the intentional, between character and content (lowercase).
In particular, the object does not determine the phenomenal character (that
is why a hallucination can share its character with a veridical perception);
it is instead determined by the content (lowercase) plus the mode.
To argue for such an impure representationalist account of phenomenal
character, I start by considering the phenomenon of perceptual constancy.
This phenomenon seems to show that there can be a change in “look”
without a change in represented properties; for example, if you tilt a coin
on its side, it will “look” elliptical but it will keep appearing round. I reply
that perceptual constancy is not an argument against representationalism—
because the orientation of the coin is represented in vision, something does
change in the represented properties. There is no phenomenal change
without representational change. Nonetheless, I suggest that the
phenomenology involved in perceptual constancy does show perceptual
experiences to be egocentric perspectival representations of the world. For
some, the fact that egocentric contents are represented in perception
explains the perspectival phenomenology of visual experience, but
egocentric contents (representations of the very relations between the
world and the perceiver) can only partially account for the egocentric
character of visual experience. I argue that, in order to exhaustively
account for the egocentric character of visual experience, we need to
appeal also to the mode. The world is represented under a mode, and this
enables the experience to represent egocentric contents. Visual
6
Introduction
representation is perspectival in a way that goes beyond representing our
perspective on the world. Rather, both the world and our contingent
perspective on it are perspectivally represented. Thanks to the mode,
perceptions represent the world “from here.”
Following this line of thinking, I consider the inverted spectrum
hypothesis and the Inverted Earth thought experiment as potential
objections to representationalism. I show that the inverted spectrum
scenario, upon closer inspection of how our color-experience holistically
involves interwoven relations between color-properties (brightness, hue,
saturation), is less conceivable than it appears prima facie. Since each
color has a place in a virtual three-dimensional space with brightness,
saturation and hue as coordinates, inverting two colors would ruin all the
other representable relations between colors. I also analyze Block’s
Inverted Earth thought experiment and argue that it does not show that
representationalism is false unless you already take it as a given: in other
words, it is an interesting argument to make our intuitions explicit but it is
circular in the end. I accept that the conjunction of representationalism
with respect to the phenomenal character and externalism with respect to
the perceptual content entails phenomenal externalism. Since I hold both
representationalism (though impure) and content externalism to be true, I
must accept phenomenal externalism, even though it is counter-intuitive
(content externalism also sounded outrageous in the past). So I commit to
phenomenal externalism.
In Part II, I examine a very important issue for the CV so as to
complete my global picture of the semantic characterization of perceptual
episodes: the issue of whether perceptual content is Fregean or Russellian
in nature. I discuss Chalmers’s double-content view, the proposal that
perceptual experience has two kinds of content, one Russellian and the
other Fregean. Chalmers aims to save phenomenal internalism and content
externalism by distinguishing a Fregean narrow content, on which
phenomenal character supervenes, and a wide Russellian content. The
Fregean content would be specified, for example, as the property that
normally causes the phenomenal property F; the Russellian content would
be what normally causes that phenomenal property in the subject’s
environment. I provide many arguments against that proposal: the first
being that “normal causation” entering perceptual content is implausible.
Perceptual contents are not that sophisticated. Moreover, it does not seem
at all likely that visual phenomenology includes the representation of
properties as “being normally caused by” something. Perceptual content
should be ascribed in a way that respects perceptual phenomenology.
Second, Chalmers assumes that a phenomenal property can be picked out
Defending the Content View of Perceptual Experience
7
independently on any worldly represented property, but that possibility is
far from uncontroversial. In addition, I show that such a possibility would
entail a separation (a totally contingent relation) between phenomenal
character and representational content, therefore this view inherits the
same problems typical of qualia-realism. Such a separation does not do
justice to the transparency of visual phenomenology: our perceptual
experiences seem to attribute to the surrounding world those properties we
are aware of in perceiving.
In addition, references to “normality” and appropriateness of causation
are highly problematic. Any normality clause on causation implicitly
refers to an environment, but then the Fregean content is not narrow
anymore. I show there is no normality that is not environment-indexed, so
there is no narrow normality. If there were a Fregean content of
perception, it would be wide, so we would be better to get rid of it and
hold on to a wide external Russellian content. Chalmers’s third way is
flawed.
I conclude that perceptual content is Russellian and wide and (impure)
representationalism about phenomenal character is true, so phenomenal
character is wide and phenomenal externalism is true.
In Chapter Six, I take a stand on the issue that properties can be
represented in perception (I mainly consider the case of visual perception
as a paradigm, as I do in the whole book).
In the first part, I introduce the basic terms of the debate between
“liberals” and “conservatives” on perceptually representable properties.
In the second, I critically discuss the so-called phenomenal contrast
method, a method of comparative introspection that is supposed to lead—
according to some of its advocates, especially Susanne Siegel—to a liberal
view, namely, to the idea that visual perception can well represent other
“thicker” properties than colors, shapes, distances, shapes and the like. I
argue that such a method is flawed because the phenomenal difference
between the two contrasted experiences can also be explained within a
“conservative” framework. In my experience, non-visually representing a
“thick” property as a consequence of a visual episode can have a top-down
effect on the “thin” properties, without the very “thick” property itself
(like [being a lemon] or [being a pine tree] when seeing a lemon or a pine
tree) being visually represented. Therefore, the phenomenal difference
between an experience E1 (seeing a lemon as a lemon) and E2 (seeing a
lemon without seeing it as a lemon) can be explained without accepting
that [being a lemon] is a visually represented property in E1.
In the third part of Chapter Six, I argue for a moderately liberal view
on properties represented in perception on another basis than the (flawed)
8
Introduction
phenomenal contrast method: thanks to perceptual learning, a perceiver
can expand the range of perceptually representable properties beyond the
range of “thin” properties (for example, in visual experience, beyond the
spatial-chromatic-morphological properties). Nonetheless, such “thick”
contents can outstrip visual phenomenology so they are not detectable by
means of the phenomenal contrast method. I provide a criterion—based on
whether the way of learning to represent a given property is purely
perceptual or inferential—for determining whether a certain property is
perceptually represented or not.
In Chapter Seven, I take at face value the objections to the CV
typically made by those disjunctivists who advocate naïve realism.
First (Part I), I take into consideration the core idea of disjunctivism
and the principles it rejects. Then I present the reasons disjunctivists
provide for being against the CV: I hold that these reasons
(phenomenological, epistemological, semantically, metaphysical) are all
amenable to what I call the detachment problem. It seems that, on the CV,
a veridical perceptual experience must be conceived of as separate from
the world, characterized and type-individuated independently of its being a
genuine relation to the world. Indeed, if perceptual experiences are
individuated by their semantic properties or contents, and the content they
possess is independent of being exemplified or not, then veridical
experiences and hallucinations should be states of the same kind, and not
even the first can be thought of as an essentially world-involving state.
Here are the basic facets of the detachment problem. Perceptual
phenomenology is presentational (phenomenological facet), perceptual
knowledge entails that veridical experiences make available to us more
than what hallucinations make available to us, on pain of skeptic
consequences (epistemological facet); perceptual beliefs and judgments
can be de re and anchored to the world only if perceptual experience is a
direct presentation of worldly particulars (semantic facet): if veridical
experiences are genuine manifestations of the world, they cannot be
mental states of the same fundamental kind as hallucinations
(metaphysical facet).
Second (Part II), I argue for a moderately disjunctive version of the
CV, a version that should embed the demand of cognitive contact raised by
the disjunctivists, so avoiding the detachment problem in all its facets. I
argue that there is a conceptual, explanatory and metaphysical
asymmetrical dependence between the bad case and the good case:
disjunctivists are right in taking the good case as basic and in
characterizing the bad case in terms of it.
I argue that, from a naturalistic point of view, mental states are to be
Defending the Content View of Perceptual Experience
9
type-individuated according to their natural functions. I suppose that a
teleo-semantic version of the CV is true. Rather than arguing for its truth
(it would take another book), I show that the CV could meet the demands
and the worries raised by disjunctivists, especially with respect to the
good/bad asymmetrical dependence. Moreover, it would provide a
naturalistic explanation of that asymmetry. A wired-in teleo-function is
acquired through evolutionary selection thanks to its success and is thus
defined by reference to its successful exercises, so its failed exercises are
essentially a failure of the function they are exercises of. However, if
perceptual states are teleo-functional states, a veridical experience and a
deceptive experience will share their function of representing the
environment a certain way, even if one is a successful exercise and the
other is not. So a teleo-functional type-individuation of mental states rules
out radical disjunctivism insofar as it predicts that veridical and nonveridical perceptions have relevant properties in common.
Nonetheless, I argue that we should opt for a disjunctive treatment
having hallucinations and perceptual experiences (veridical or illusory) as
disjuncts rather than contrasting veridical and deceptive experiences.
Veridical perceptions and illusions are genuine relations to the world, they
are world-involving states with de re, object-dependent Contents, whilst
hallucinations are not relational states but states that introspectively seem
to be what they are not—namely, relational states. Disjunctivists are right
in thinking that subjective indiscriminability is not sufficient for sameness
in kind—it is not sufficient for sameness in Content either. They are right
in thinking that, for two mental states, having the same proximate causes
is not sufficient to be of the same mental kind. Indeed, hallucinations are
objectless states even if they could have the same proximate causes as
perceptions (veridical or illusory), whereas perceptual experiences
(accurate or not) are essentially relational states involving a worldly object
as a target.
Next (Part III), I deal with two related apparent problems for the CV,
which I shall call an item awareness problem and a semantic gap problem.
The first addresses the question of what we are aware of when
hallucinating; the second addresses the question of how hallucinations can
be inaccurate states, as they intuitively seem to be, if they lack an object
the represented properties could match or mismatch. Concerning the first
problem, I rule out the Meinongian proposal according to which
hallucinations have non-existent particulars as genuine objects. I argue
against the extravagant the idea that hallucinatory objects are genuine
particulars but have the bizarre property of not existing; if it were so, then
hallucinations would be a priori true. Indeed, the hallucinated pink rat is
10
Introduction
pink even though it lacks existence. In this way not only is the inaccuracy
of hallucination not vindicated, it even becomes impossible. Then I
consider a more promising option—when hallucinating, we are aware of
structured complexes of instantiated properties (property view) even if we
wrongly seem to be confronted with particulars. After raising some
perplexities about this proposal, I consider a more radical alternative to it;
i.e., the idea that while hallucinating we are not aware of anything, neither
of particulars nor of properties (no-item view). For property view, the
conscious character of our state depends on the mode and represented
properties, but a conscious state’s representing certain properties does not
entail that state’s involving the awareness of these properties. I do not
adjudicate between the property view and the no-item view; rather I point
at the virtues and weaknesses of both, and then I conclude that one of them
must be true. I also point out that the item awareness problem is not a
problem peculiar to the CV. It is shared by other views on perception so it
cannot be held against the CV. Anyway, the second issue of the semantic
gap of hallucinatory contents is independent of whether we prefer the
property view and the no-item view. In neither case would we be aware of
these particulars so no worldly object can work as a truth-maker or as an
accuracy-maker for hallucinatory states. My dealing with the problem
consists of dropping the intuition of inaccuracy of hallucinations and
explaining its origin and its apparent force on us. Hallucinations are not
inaccurate states; rather, they are states that seem to be worldly particular
and seem to have accuracy-conditions but are neither accurate nor
inaccurate. Intuition, on the contrary, depends on the hallucinations having
immediate cognitive effects. These are inaccurate so we tend to project
their inaccuracy (of beliefs or belief-like states) onto the hallucinations
that normally produce them.
Finally, I go back to the original detachment problem to show that the
CV can avoid it. The presentational phenomenology, the justificatory
power of the veridical perceptions, the possibility of having demonstrative
thoughts about the surrounding world, the relational metaphysics of
veridical perceptions can all be vindicated by my version of the CV.
Let us begin our journey into the CV!
CHAPTER ONE
THE SEMANTICS OF SEEING AND RELATED
“EXPERIENTIAL” PREDICATES
Introduction
This chapter is divided into five parts.
Part I is introductory and presents the general aim of the chapter,
which is that of producing a taxonomic survey of the ordinary ascriptions
of episodes of “seeing” as well as of episodes of seeing-experiences like
“seeming,” “appearing” and “looking.” The methodological sense of such
a survey on ordinary language is also made clear. Although my central
concern is the perceptual phenomenon itself rather than the typical ways it
is ordinarily characterized in everyday language, an analysis of the logical
behavior of ordinary ascriptions of perceptual experiences seems to be a
privileged starting point to make our basic intuitions concerning the
phenomenon explicit. Maybe a substantial theory of seeing and perceptual
experiencing will correct or even eliminate the intuitions underlying those
ascriptive uses, but there is no other way to start shaping a positive theory
than articulating its putative objects as they are manifest to us in ordinary
experience and language. To recall an Austinian saying, although ordinary
language is not the last word, still it has to be the first.1 Addressing the
preliminary question “what do we ascribe when we ascribe an episode of
seeing (or ψ-ing)?” can at least shed light on the much more relevant
question: “what does seeing (or ψ-ing) consist of?”
In Part II, a list is drawn up of necessary and sufficient conditions for
an ordinary seeing-ascription to be true in non-abnormal contexts. Second,
the logical behavior of basic seeing-X-ascriptions is analyzed, where the
verb is used as an objectual attitude without clauses according to the
1 See Austin 1961.
12
Chapter One
simple two-place scheme: “S sees (an) O” like “Fido sees a tree”; “Diego
sees a table” and the like.
Part III concerns the ascriptions of seeing-that cases, which behave
like propositional attitudes according to the scheme “S sees that P”;
“Diego sees that the table is brown” and the like. I argue that such
ascriptions are factive and logically opaque just as propositional attitudes
are, but they are not only ascriptions of perceptual episodes. Rather, they
are ascriptions of certain empirical propositional knowledge acquired as
consequences of perceptual episodes.
Part IV takes into account seeing-as ascriptions; i.e., ascriptions of a
sui generis three-place relation—S sees O as (an) F—whose logical
behavior is irreducible to either the objectual seeing-O or the propositional
seeing-that-P. Such ascriptions are neither factive nor logically
transparent, and presuppose the ascription of a certain “cognitive stand” by
the perceiver, like recognition or a categorization of some sort. I will argue
that this is the only context of ascription where the perceptual mistake can
come into play. Neither seeing-O episodes as such—unless they are
constituents of seeing-as episodes—nor seeing-that episodes can be
incorrect, false, mistaken and the like. Either you see an O or you don’t.
You just cannot falsely see an O (given ex hypothesi that we are not
talking about seeing the O as something). Likewise, you cannot falsely see
that P, because this is a factive ascription, just like “knowing.” The
reciprocal relations between three distinctive ascriptions (seeing O, seeing
O as an F, seeing that O is F) will be carefully articulated.
Part V will change the focus from the seeing-predicate to some basic
experiential predicates such as “seeming,” “appearing” and “looking.”
They will be considered as they behave in paradigmatic ascriptive
constructions like “seeming-that,” “looking-like,” “looking-as-if,”
“looking-as-though,” “appearing-that” and the like. Such verbs do not just
ascribe perceptions (as “sees,” “hears,” “smells” do); they ascribe
conscious perceptual experiences. As in the previous cases, I will critically
discuss the relative literature on the matter.
At the end of the chapter, I will provide a summary of the results of
each part and summarize the results in concise points.
PART I
A METHODOLOGICAL REMARK
What do we ordinarily ascribe when we ascribe or self-ascribe an episode
of seeing something, a case of seeing that something is such and such or
an episode of seeing something as something? What do we ascribe to S or
to us when we say that it seems to S (or to us) that such and such is the
case, when we say that an O looks F to me, that this O appears F to her or
that this looks like that and so on?
Some find it plausible to categorize visual perceptions and visual
experiences under natural kinds. Aren’t those phenomena distinctive
byproducts of the biological evolution of certain animal species? So it may
well be that the rough and intuitive individuation-criteria, applied by those
who ordinarily ascribe such mental states, actually pick out a cluster of
different phenomena whose ordinary grouping does not genuinely “track”
the objective division into natural kinds. Maybe the superficial properties
exhibited by the ordinary referents of “seeing”-episodes ascriptions are not
shared by other genuine cases of seeing. Maybe very different natural
kinds happen to be the referents of people’s ascription of seeing. Were this
the case, only scientists of vision (for example) would know the real
reference and the genuine extension of the term “seeing.” Ordinary people
would just be able to vaguely fix the reference through attaching the
meaning of the term to a cluster of manifest, superficial and non-essential
properties.1 Just as speakers can successfully refer to water without
knowing at all the nature of water (be it H2O), so too can they successfully
master and apply terms like “seeing,” “visually experiencing” and the like
without knowing the nature of the phenomena they ascribe. Even so, that
view would not per se entail the uselessness of a systematic consideration
of the ordinary uses as well as the related shared intuitions underlying
these uses. Generally speaking, any explanation must have the
individuation of an explanandum as its inevitable starting point. In order to
1 That is, at least, the Kripke-Putnam theory about the reference of natural kind
terms. See Putnam 1975.
14
Chapter One Part I
meaningfully ask, for example, “what is seeing?” the very question must
make sense before the answer (the explanans) can be obtained, before
coming to know what seeing is. What are we asking if not about the
manifest phenomenon we can intuitively pick out as folk speakers in the
first place? Even if discovery of the nature of X can give feedback on the
starting characterization and reveal it as flawed (confusing, naïve,
illegitimate, to be abandoned), nonetheless its status as a starting point of
the inquiry would be a precondition of the final cognitive success.
Therefore, a reconstructive taxonomy of the basic ways of ascribing visual
perceptions and experiences, a survey on the related vocabulary, is
methodologically useful at least in order to make explicit our unreflected
intuitions on the matter. Although the ways certain paradigmatic
expressions behave in ordinary language should not be considered as
normative to establish the way things are,2 still a reflective analysis of
those ways could successfully orient and prepare the substantial inquiry as
its preliminary rough material.
2 This was the “quietist” way some Oxford linguistic philosophers seemed to
consider their language analyses. For example, see Malcolm 1942, Moore 1962.
PART II
SEEING SOMETHING
II.1 Basic Conditions
We consider cases of seeing as perceptual episodes occurring to a subject
in an environment. “Seeing” is a determinate of the determinable
“perceiving” (as “hearing,” “tasting” and so on). Are there necessary and
sufficient conditions for truly ascribing to S an episode of seeing
something? Which contexts and circumstances are ordinarily and
implicitly taken to entitle a speaker to say that she or someone else is
seeing something? First of all, seeing-something is a certain sort of real
dyadic relation involving a perceiver and an environment as relata. Here is
a list of trivial conditions for seeing-X:
S sees X if
a) S is a perceiver with a visual apparatus
b) X is there in the S’s surrounding environment
c) Through the very episode, S discriminates X in some way from the
environment
d) X causes the very episode of S seeing-X
e) Such a discrimination must involve a presentation with a
phenomenological salience; it must give rise to a “looking” or a
“seeming.”
You cannot see X if you are blind or do not possess a perceptual-visual
apparatus. You cannot see X if X is not there. You cannot see X if you do
not discriminate it in any way from its surrounding environment. You
cannot see X if X does not provide any causal contribution to your seeing
it, and you cannot see X unless X looks some way to you. On the other
hand, if you have a working visual apparatus, and X’s impact on it causes
your discrimination of X in such a way that X looks some way to you, all
this is intuitively sufficient for you to see X. In short, seeing (X) is a
certain episode consisting of a discrimination-relation between a perceiver
16
Chapter One Part II
and an environmental object, where the object is causally responsible for
being discriminated through appropriately impacting on the subject’s
visual apparatus in such a way that the object looks to S a certain way.
Let us recap the basic meaning underlying the points listed above (a-e):
Point (a) captures the trivial reference to eyes implicitly involved in the
very mastery of the folk seeing-concept;
Point (b) captures the so-called implicativity of seeing: “S sees O”
presupposes that O is there to be seen (differently from “S wants
O”, for example);
Point (c) depends on seeing being a success verb,1 as other perceptual
verbs are. Perceiving something is certainly a kind of cognitive
achievement, the occurrent exercise of a dispositional capacity to
achieve a certain positive state;
Point (d) is meant to capture what has been notoriously emphasized by
the causal theories of perception:2 perceptions are episodes
appropriately caused by the perceived environment itself.
Perception can provide a form of contact with the world insofar as
it consists of a certain sort of world-to-subject causal impact;
Point (e) involves that object-seeing has some minimal phenomenological
constraints to the effect that in understanding “S is seeing-X”
uttered in non-abnormal contexts, a speaker is entitled to take it that
there is a way X looks to S.
II.2 Some Objections
Now I will consider some possible objections to the above conditions
for S to be seeing something and I will briefly reply to them.
Challenging the a-condition are the well-known experiments of
prosthetic vision, which realize cases of “vision-through-touch” (Bach-yRita 1972), reported and discussed by Dennett3 among others. A device
involving a small low-resolution video camera was mounted on eye-glass
frames so that the signal from the camera—an array of black-and-white
pixels—spread over the back or the abdomen of the subject in a grid of
vibrating tinglers. Surprisingly, subjects were able to interpret the patterns
of these tingles on their skin after a few hours of training: they recognized
1 Ryle 1949, Austin 1962, Armstrong 1968.
2 See Grice 1961, Strawson 1974, Lewis 1980.
3 See Dennett 1991, 337-344. On tactile-vision substitution systems (TVSS), see
also Back-y-Rita/Kercel 2003.
Seeing Something
17
a face, identified objects and so on. Were they seeing those objects despite
no sight being involved? Let us assume that it was a case of vision. First,
we should consider that part of the prosthetic device—the camera—may
be taken as an artificial visual apparatus. After all, it is causally sensitive
to light-waves and carries a signal consisting of a certain distribution of
gradients of light-energy, just like biological retinas and animals’ eyes.
Therefore, the “no eyes-no vision” principle embodied in condition a is
respected. Second, the capacity of seeing-with-touch comes with training
that necessarily involves the exercise of canonical vision in order to match
certain tactile information with a certain visible scene. Such a capacity is
therefore parasitic on proper vision and can be ascribed only to subjects
endowed with a working visual apparatus. Third, it is no surprise that such
an artificial integration of our natural biological capacities could constitute
a borderline case (both of vision and of touch), but the existence of
borderline cases does not undermine canonical demarcations. Finally, and
most importantly, at this stage of our inquiry we are talking about the
ordinary concept of seeing and its folk application in normal contexts. We
do not learn to master the concept of seeing and ascribe cases of seeingsomething by being shown abnormal contexts like prosthetic tactuo-vision.
If a speaker does not know that seeing something presupposes using one’s
eyes, we would not ascribe to that speaker the mastery of the concept of
seeing or of the respective word “seeing.”4 There may be dark samples of
H2O and maybe that unusual circumstance is known to a scientist who
claims to know the nature of water—the real reference of “water”;
nonetheless, the ordinary concept of water involves transparency as a
superficial reference-fixing property.
Condition b could raise perplexities insofar as some ascriptions of
seeing-X cases do not seem to be captured by b. For instance, where X is
known to be not there, like “Mary sees phantoms” or “Even if he’s in front
of his wife, Mister P. keeps seeing a hat.” To address this objection, we
should consider that apparently simple seeing-X ascriptions can be elliptic
ascriptions of more complex cases, like cases of seeing-as or seeing-that,
which we will carefully treat below. Mister P does not see any hat at all.
He sees his wife as a hat; he actually mistakes his wife for a hat.
Therefore, the above case is a case of seeing-as not a basic case of seeing
something simpliciter. Likewise, Mary cannot see a phantom unless there
is a phantom there to be seen. Rather, Mary sees something that looks to
4 I am making the plausible assumption that ascribing the mastery of a certain
word in ordinary-language contexts is sufficient for the mastery of the relative
concept. A subject’s using, understanding and correctly applying the word “water”
is sufficient evidence for crediting the subject with the concept of water.
18
Chapter One Part II
her as if it were a phantom, or she hallucinates a phantom and falsely takes
her subjectively seeing-like experience as an episode of seeing. In both
cases, Mary just believes she sees phantoms, but she doesn’t. Harman
distinguishes “seeing” (implicative) and “seeing*” (intentional):5 whilst
seeing is implicative and presupposes the existence of the seen object,
seeing* can have non-existents as complements, like “Jack sees a
unicorn.” As we have noticed, though, Harman stipulates seeing* as not
the simple “seeing” we are talking about here. The very fact that he needs
to introduce a special stipulation (*) entails that he is not talking about the
ordinary application of seeing-something ascriptions. Furthermore, we
should not be misled by the fact that sentences like these sound fine and in
order. The superficial grammatical form of “seeing-X” can hide the
contraction of more complex ascriptions. By nominalizing the
complement, I can treat any form of seeing as a case of object-seeing: “S
sees the train’s stopping at the station at 8 o’clock”; “S sees the difference
between a phantom on his right and a unicorn on her left” and so on. But
now we are treating object-seeing in a more specific sense, where “object”
is not meant in such an abstract way or in a superficially grammatical
sense.6 As Heil remarks,7 when an episode of seeing-X is ascribed, the Xcomplement can be meant to express either the object or the content of the
ascribed perceptual episode. Up to now we have been considering the
object rather than the content so we are interested in the direct complement
of seeing on the objectual interpretation. For example, one can ask: “Can
you see the boat there in the distance?” meaning “can you recognize a boat
in that which you are seeing?” This would be a case of seeing-as, not just
a case of seeing-X. Likewise, we are not concerned with cases where “X”
is a propositional clause, be it nominalized or not. “I saw the cat running
away from a dog” is not just an example of object-seeing, at least under
the most natural interpretation of it, because it is a case of seeing a fact
having one or more objects as its constituents (see below). Some think that
perceptions have certain objects in virtue of having certain contents. Be
this the case or not, perceptual object and perceptual content should not be
5 Harman 1990, 36ff. On the intentionality of seeing, see Anscombe 1965, Travis
2011.
6 Dretske (2000a, 117) refers to this as the difference between “concrete” and
“abstract” objects of seeing. Abstract objects of seeing are grammatical objects as
abstract noun phrases (seeing the bus arriving, seeing the difference, the number,
the answer), interrogative nominal clauses and so forth.
7 See Heil 1991, p. 9.
Seeing Something
19
confused,8 not even when analyzing ordinary ascriptions. As we will see,
the typical basic scheme for perceptual content-ascriptions is: S perceives
[O’s being F], namely, an object having a property.9 But now our focus is
object-seeing, not the content of seeing. So far so good.
Condition c—call it the discrimination condition—may sound very
ambiguous at first. What does (visually) discriminating consist of? An
objection to c, for example, could be that one can well see X without
noticing it, without consciously discriminating it from the environment.
That could happen because of tiredness, lack of attention, rapid
disappearance of the stimulus,10 bad conditions of visibility and so on. “He
saw it but he did not discriminate it” is a quite common and intelligible
sentence. As a first response to that, we should point out that seeing an F
does not involve discriminating it as an F, recognizing it, noticing it,
taking it that there is an F there, and the like. Nonetheless, seeing-X is
meaningfully ascribed if the subject is taken to discriminate X from the
environment in some way. For example, in order to be seeing a boat in the
distance, it is not necessary to see it as a boat, or as anything else, but for
the subject the object must be a potential object of individuation and
characterization (even if extremely vague). It must be possible for
subjects, if they focus their thematic attention on such seen objects, to ask:
“What is that?”11 Such a condition—discriminating O in some way from
the surrounding environment—does not even require a belief or explicit
“taking it” that there is something there. What is required is that the
allegedly seen object must make a potential difference for the subject be
that difference doxastic or just behavioral. If the presence of O does not
8 As we will see, many philosophers think that perceptions have objects but not
content; i.e., that the very notion of content is misleading as applied to perceptual
phenomena. See Chapter Six.
9 Fish 2009, p. 52.
10 This is the case raised by the Sperling experiments (see Sperling, 1960 and
below, chapter II, Part1.2.2), where subjects are exposed for a short time to a
visual stimulus they cannot notice and report. Nonetheless, thanks to an acoustic
marker, they can access the visual information after the very stimulus has
disappeared, so that they must have seen that stimulus.
11 See Dretske 2006, Siegel 2006b.
20
Chapter One Part II
make any difference—not even a potential one12—for the perceivers, they
just do not see it.13
Condition e could be resisted by pointing out that X causing itself to be
seen by S is an empirical rather than a conceptual condition of S’s seeingX. In other words, the ordinary concept of seeing an O does not involve
O’s causation of O’s vision, even though it is true that S sees O if O causes
the very episode. A straightforward test to make explicit the cluster of
intuitions underlying the ordinary mastery of a concept involves
constructing a counterfactual situation in which one of those allegedly
primitive intuitions are not satisfied. Imagine someone who understands
the expression “S sees an O” and can successfully apply it to normal
contexts. Could he really ignore that the episode involves a causal impact
between the object and the perceiver? Of course he doesn’t have to know
which causal interaction is going on—we do not need to be scientists of
vision to master the ordinary concept of seeing and to successfully ascribe
episodes of object-seeing—but at least he needs to intuitively type the
episode as an interaction between the subject and its environment. No
interaction without causation so no vision without causation.14
Condition e can be seen as a way to specify the nature of discrimination
introduced in condition c.
As poor as this discrimination can be, it must be conscious and involve
a phenomenological difference. It does not matter how S sees O or how O
looks to S; it does matter that O must look some way to S, be that way as
vague and indeterminate as a dark point in the distance, a rough shape,
something having a certain color, something moving and so on. Seeing an
12 I can be seeing something that is completely uninteresting for my behavior and
my belief-updating, so there is nothing actually changing in my behavior and in my
system of beliefs. But if I see O, then it must be true at least that if O were
behaviorally, pragmatically or doxastically relevant for me, it would make a
difference for me now.
13 If seeing an object is to make a difference, even a slight one, for the subject,
then such an episode must enable certain discriminations, which are potentially
manifest in behavior or in reasoning as difference-makers in the subject’s treating
the surrounding environment. Otherwise, the subject has not seen the object.
14 According to an ancient theory of vision—for example, Democritus’ view—the
eyes are the active causal part of the visual phenomenon; they “launch” a certain
quantity of rapid atoms toward the object, acting as a sort of “illuminating factor”
of it—namely, making it visible. This empirically false and rudimentary theory
does not undermine the causal factor as embedded into the concept of seeing. On
the contrary, the theory presupposes the idea of a causal interaction between the
perceiver and the perceived environment, even though it is wrong in specifying
type, mode and direction of the causal interaction.
Seeing Something
21
object involves having phenomenological consciousness of at least one of
its properties, even a perspectival, relational property like its distant
location. The way O looks to S in S’s seeing O could well be wrong. O
can look F to S even though O is not F. Still, S would be seeing O insofar
as, among other conditions, O looks some way to S (no matter how wrong
that look is). What matters is that the phenomenology associated with O’s
looking F to S is sufficient for S to consciously discriminate O from the
surrounding environment. Suppose you see a green landscape where an
unmoving chameleon is perfectly disguised through a perfect mimicry.
Suppose you even focus your attention on that point of the objective scene,
which globally looks to you just as a uniform greenish vegetable tract. Do
you see the chameleon? Intuitively you don’t because your visual
phenomenology does not differentiate it from the immediate surroundings.
There is no phenomenologically salient discrimination from the scene,
even if the object is there. You are in visual contact with the surface of the
chameleon, you focus on the portion of the scene corresponding to its skin,
you receive information about its color, your well-working visual
apparatus is in appropriate causal contact with it and so on.15 Such a
thought experiment shows by subtraction, so to speak, the necessity of the
phenomenal discrimination condition (e) for object-seeing. If visual
discrimination of O from its immediate surroundings is missing, then S is
not seeing O according to the ordinary concept and ascription-conditions
of seeing something. Although minimal, object-seeing has
phenomenological constraints.
A natural objection to condition e could appeal to some counterexamples
—namely, to cases of object-seeing that lack phenomenology and a
fortiori differentiation-phenomenology. For example, so-called blindsight
is vision lacking visual phenomenology. Subjects with blindsight report a
lack of any visual experience despite having certain abilities to guess what
is where and detect visually perceivable properties. So, subjects with
blindsight form conscious empirical (true) beliefs grounded on
15 Siegel 2006b constructs a thought-experiment on the same lines: Franco loves
doing stunts in the sky, dresses in red and uses invisible fibers to stay suspended in
the air. But today Franco has dressed in a blue uniform that perfectly matches the
actual color of the sky so you cannot individuate him. You just see a homogeneous
blue scene (which includes Franco’s surface despite you not knowing that). Siegel
argues that you do not see Franco because you are not in a position to form any de
re mental state about Franco. Seeing O should put you in a position to have de re
mental states about O, given other conditions that we do not need to specify. See
also Dretske 1969, 18-33.
22
Chapter One Part II
unconscious episodes of vision.16 Well, if such patients do see objects,
then seeing cannot have phenomenological constraints. Another case
against e could be inattentional blindness.17 According to an interpretation
of this phenomenon, you do see objects in a scene within your visual field
but you fail to notice them.18 Therefore, there are also non-pathological
cases of object-seeing that do not involve conscious and phenomenologically
salient discrimination of O from its immediate surroundings.
Another counterexample by Siegel19 is that of zombies. My zombietwin is a cognitive system functionally and physically identical to mine but
lacking any conscious phenomenology at all.20 Such zombies would
behave like me, form my empirical beliefs, perform the same actions,
make the same verbal reports and so on so it would be arbitrary to deny to
zombies any object-seeing, and we could consistently refuse a notion of
object-seeing involving it. This would therefore count as an objection for
phenomenological constraints on object-seeing.
We could address the three objections together by reminding ourselves
what we have been doing up to now: we are making explicit the intuitive
conditions for ordinary “object-seeing” ascriptions. Again, appealing to
borderline cases of object-seeing in such a methodological context does
not do; it is premature and misleading. Making explicit the intuitions
underlying the ordinary concept of seeing is not absolutely normative for
what we should take object-seeing to be. On the contrary, experimental
evidence can well be incompatible with ordinary intuitions and lead us to
amend, change or even abandon the ordinary concept. Blindsight is a
pathology so it is an abnormal behavior; it even puzzled the scientists who
discovered it in the first place. In addition, one could legitimately doubt
that it is opportune to describe blindsight as an example of genuine cases
of vision. As regards inattentional blindness, the very fact that there is a
vivid debate on its interpretation—is it not seeing but only potentially
seeing? Is it unconscious seeing? Is it conscious seeing without
noticing?21—shows that it is a non-ordinary case of object-seeing, so it is a
borderline case. Spelling out the conditions for ordinary ascription of X
16 See Weiskranz 1997, Clark 2001, Hilbert 1994.
17 See chapter II, part 1.2.1.
18 See Mack and Rock 1998, Dretske 2000a, Block 2007.
19 Siegel 2006b.
20 See Chalmers 1996.
21 As Siegel 2006b points out, the question whether there can be phenomenology
outside attention does not undermine per se the differentiation-phenomenology
constraint. If there can be phenomenology outside attention, then cases of
inattentional blindness would count as cases of object-seeing.
Seeing Something
23
well tolerates borderline cases not perfectly meeting those conditions, or
even lacking one of them. As for zombies, it is ex hypothesi a (doubtful)
thought experiment that depicts a possible world quite different from the
one in which our ordinary concepts are socially established, individually
formed, acquired and applied in everyday contexts. In a world inhabited
by zombies, the ordinary perceptual concepts would probably be different
and presuppose a different cluster of basic intuitions for their application.
Another objection accepts the discrimination condition c but denies
that it must involve phenomenological constraints along the following
lines: the discrimination of O from its surroundings could be behavioral,
consisting of sensory-motor dispositions, of potential influence on beliefformation and intention-formation or otherwise without necessarily
involving a phenomenological difference. Again, apart from pathological
and other borderline cases, such an objection would entail that S can see
O—according to the intuitive meaning of object-seeing—without O
visually looking to S in a way sufficient for S to discriminate O from its
immediate surroundings. But if you see O, O must look some way to you.
So a certain visual phenomenology is constitutive of intuitive objectseeing, such that a certain causal contact of S’s visual apparatuses with an
object O counts as an episode of S seeing O insofar as O looks some way
to S, where such a “looking” relation consists of a certain distinctive visual
phenomenology. To sum up again our list of object-seeing conditions:
S sees O if
a) S is endowed with a working visual apparatus
b) O is there in S’s surrounding environment
c) S discriminates O in some way from the immediate surroundings
d) O causes the very episode of S seeing X
e) O’s being discriminated by S is enabled by S’s visual phenomenology to the
effect that O looks some way to S.
II.3 Transparency
An important property that characterizes the logical behavior of
seeing-X ascriptions is transparency:22
(S sees X) and (X=Y) → S sees Y
22 Transparency of object-seeing is rightly emphasized by Dretske 1969, 1990
24
Chapter One Part II
Object-seeing is extensional; its truth-conditions are indifferent to the
descriptions through which the seen object is referred to. The seen object
is a particular the subject is causally related to. That particular, no matter
how you describe or characterize it in the ascription, is the seen object
when a-e conditions are satisfied. Transparency and extensionality of
object-seeing depend on it. The truth-conditions of “S sees O” do not
depend on the particular way O is cognitively appreciated or categorized
by S, nor do they depend on the particular way O looks to S. The
description used by the ascriber does not matter for the truth-conditions of
the ascription either. The fact that O is cognized by S in a certain way or
another, does not matter for whether S sees this particular O. S either sees
it or does not. In this sense, as Dretske points out, object-seeing is “noncognitive.”23
It is worth noticing that conditions d and e—discrimination and
phenomenological constraints—are not incompatible with the
transparency and extensionality of object-seeing ascriptions. If an object
can be seen by S only if it visually looks some way to S and is
discriminated from the surroundings by virtue of the way it looks, such a
condition does not turn object-seeing ascriptions into intensional contexts.
It is irrelevant for the truth-conditions of the ascription which way the
object looks to S; what matters is that the way it looks determines the
subject’s visual phenomenology to the point she can individuate O and
differentiate it from its immediate surroundings. So seeing O is compatible
with mistaken acts of categorization—mistaking one’s wife for a hat is
still seeing one’s wife—and more generally with false and illusory
perceptions. Actually, as we will see later, illusions and other deceptive
perceptions are possible only insofar as they are successful acts of objectseeing. What I believe I am seeing, what I take it to be, does not have any
relevance for what I am actually seeing. Seeing-O as such is neither
correct nor incorrect, neither true nor false, neither right nor wrong. Either
it is the case, or it is not.
What determines what S is seeing is the causal context connecting S
and its surroundings: the particular object causing the very visual episode
is the object S is seeing, even though only the phenomenological
conditions—among the other conditions spelled out above—make this
causal contact into a proper case of seeing. To sum up, in order for S to
see O, O must look some way to S, where “way” is salient in visual
phenomenology. But it is not the way O looks that determines what S is
seeing; rather, it is the fact that S’s visual apparatus is in certain
23 See Dretske 1969, 1981, 1990, 1995.
Seeing Something
25
appropriate causal contact with O, even though such a causal contact
counts as a seeing-episode only if it gives rise to a phenomenological
difference involving at least a way O looks to S.
In this part, I have spelled out the conditions for object-seeing
ascriptions in ordinary contexts then I addressed some possible objections
to there being necessary conditions. I have argued that object-seeing is
non-cognitive and causally-contextually determined, but has some
minimal phenomenological constraints basically consisting of putting S in
a position to visually discriminate O from its immediate surroundings. I
have also noticed that the logical behavior of seeing-X ascriptions is
characterized by transparency: substitution of O’s description with coreferential expressions does not affect the truth-conditions of “S sees O.”
Seeing-O is an extensional context.
PART III
SEEING THAT P, SEEING THAT A IS F
III.1 Propositionality
We ordinarily claim to see objects like tables, chairs, people, trees and so
on but object-seeing in this sense (for O = a physical particular) is not the
only kind of “seeing-X” ascriptions in ordinary discourse. We see events,
like the sunrise or the fall of a glass on the ground. We see properties, like
the yellow of the car and the rectangular shape of the sofa. We see
relations, like the difference between the scarlet of your bicycle and the
scarlet of my pullover. We also take ourselves and others to see facts or
states of affairs, like the fact that just now in my garden a cat is running
after a mouse. So the grammatical object of “seeing-X” ascriptions does
not always consist of a “mundane” object like an apple, a chair, a table and
the like. This last kind of object-seeing is the one we have been analyzing
at length up to this point. Now we will turn to seeing-that ascriptions.
Later on we will consider the relations between object-seeing in the basic
sense and other less basic “object”-ascriptions (like property-seeing,
relations-seeing, events-seeing), as well as the relations between basic
object-seeing and seeing-that ascriptions. Let us consider the following
scheme of ascription:
S sees that: [a is F]
Such ascriptions are ascriptions of a propositional attitude. Indeed, the
that-clause is filled by a proposition, just as it happens for thinking-that,
hoping-that, knowing-that and the like.1 Without entering into the logical
debate about the nature of propositions, let us assume that a proposition is
an abstract entity consisting of a certain structured semantic content in
such a way that an entity—a thought, a sentence, an utterance—expressing
a certain proposition is semantically evaluable as true or false in virtue of
expressing that proposition, where expressing a certain proposition amounts
to possessing a certain (propositional) content. Whilst propositions are
1 See Searle 1983.
Seeing That P, Seeing That a Is F
27
either truth-conditions or bearers of truth-conditions for the entities
expressing them, propositional contents are truth-conditions. The sentence
“the vase of flowers in front of me is red” expresses a proposition that
makes the sentence true if the vase of flowers in front of me is red, and
false if this is not the case. Therefore, a proposition is or bears2 a content
consisting of a possible fact or state of affairs. Thoughts or sentences (or
other content-bearing entities) possessing certain propositional content are
true or false according to whether such content is satisfied by a
corresponding state of affairs or not.
III.2 Factivity
What does it take to be true that S is seeing that P, for example, equals (a
is F)? What logical behavior characterizes ordinary seeing-that
ascriptions? The first relevant property of seeing-that ascriptions is their
factivity:
Factivity: S sees that P → P
It is a very strong intuition that if S sees that P, P is the case. If P were
not the case, nobody could ever see that P. So seeing-that ascriptions
behave like knowing-that ascriptions, at least from the point of view of
their factivity.3 Just as you cannot know P unless P is the case, you cannot
see that P unless P is the case. Just as you cannot know something false,4
you cannot see something false either. This is why ascribing a seeing-that2 For example, Stalnaker 1984 takes it that propositions are contents, rather than
being content-bearers. Anyway, such an abstract difference is not relevant at all for
our current concerns.
3 Some philosophers coin “seeing”-like expressions as non-factive corresponding
counterparts for seeing-that. Millikan (2000) introduces the technical notion “to
visage,” which is meant to stand to “to see” as believing stands to knowing, such
that S can falsely visage that P. Likewise, Johnston (2004) proposes to accept a
non-factive reading of seeing-that and defines it as “visually entertaining a
content.” Byrne (2011) also characterizes visual perception as a sui generis attitude
he calls “exing” (non-factively experiencing that P). As interesting as those
proposals could appear from a theoretical point of view, they just confirm what we
are arguing for here: the ordinary logical behavior of seeing-that ascriptions does
involve factivity.
4 Of course, you can know that a certain proposition is false. But the proposition
that a certain proposition is false can be known by you just insofar as it is a true
proposition. Although some deny it—for example, Hazlett 2010—factivity of
knowledge is mainly accepted by epistemologists.
28
Chapter One Part III
P episode to S involves endorsing the truth of P5 besides making a
statement about S or S’s occurrent mental episode.6
Let us note, for the sake of clarity, that ascribing to S an episode of
seeing that P obviously is not to say that S sees a proposition. Rather, it is
to claim that S sees that fact or occurring state of affairs, which makes the
proposition P true. We will soon explain what “seeing a fact” can mean.
III.3 Opacity
The second interesting property of seeing-that, shared by propositional
attitudes in general, is referential opacity:
Referential opacity – From (S sees that [a is F]) and (a=b), you cannot infer (S sees
that [b is F])
Referential opacity characterizes intentional contexts, where you
cannot substitute salva veritate co-referential expressions within the
proposition towards which S’s attitude is ascribed to be. Opacity is the
opposite of transparency, so this is a first important difference between
seeing and seeing-that. Just as you cannot infer that S believes that the
Morning Star is such and such from the circumstance that S believes the
Evening Star is such and such even if the Morning Star is the Evening
Star, likewise you cannot infer that S sees that b is F from the fact that S
sees that a is F, even if a = b. On the contrary, if you see a, and a = b, you
see b. So the conditions for ascriptions of seeing-that to be true are not the
same as the conditions for ascriptions of seeing-O to be true. As we will
see, the a-e conditions spelled out above are not enough for seeing-that.
Provided it is factive and opaque, what does seeing-that P amount to?
According to Dretske,7 seeing-that does not consist of a purely visual
episode; rather it consists of an episode of coming to know by visual
means. If S sees that P, S comes to know that P is the case through the
contribution of one or more visual episodes. The epistemic result of S
seeing-that P is that S knows that P. In order for such coming-to-knowthat- P to be a case of seeing-that-P, some perceptual-visual episode must
be at least among the ways in which S has gained that knowledge. Given
5 See Sellars 1953, 223: “To characterize S’s experience as a seeing is, in a
suitably broad sense, to apply the semantical concept of truth to that experience.”
6 That is what Armstrong calls “success-grammar” of the determinable perceivethat and its “determinates.” See Armstrong 1968, 212.
7 Cfr. Dretske 1969, 1990, 2000. Also Williamson 2000 and Crane (talk in
Manchester, 2008) agree.
Seeing That P, Seeing That a Is F
29
that Dretske’s proposal seems very plausible—it seems to be intuitive and
it well explains both opacity and factivity—we can ask what the strictly
visual conditions are for an episode of coming-to-know-that-P (where P is
piece of empirical knowledge) to count as an episode of seeing-that-P.
III.4 Literal/Metaphorical
To address this question, it is opportune to first make clear that we are
exclusively talking about literal cases of seeing-that. Even though the
literal/metaphorical distinction may be blurry and involve borderline
cases, we should leave aside the seeing-that ascriptions that have nothing
to do with the visual way of gaining the respective knowledge, cases
where “seeing” just generically means “becoming aware,” “coming to
know (through whatever means),” “realizing” and the like. For example,
cases like “S sees that the government is in trouble” do not concern us,
provided that S could see that even if completely blind.8 So, even though
seeing-that does not always mean “coming to know by visual means” in
ordinary discourse, we are exclusively interested in the subset of those
ascriptions for which the visual involvement is taken by the ascriber to be
essential.9 Just in those cases, “seeing” is not ascribed metaphorically but
8 Of course, it can be the case that S sees that the government is in trouble by
visual means. For example, S can read it in a newspaper. Cases like this will be
analyzed below: they consist of coming-to-know a certain fact through directly
coming-to-know other facts by visual means; for example, facts about what is
written in the newspaper. Suppose the news you read is false: you acquire a false
belief but you do not visually misperceive: the mistake does not rest on the seen
facts but on other matters (like trusting the journalist). Knowing that S by visual
means is forming a perceptual belief that P, which is justified on the basis of how
things look to one. A belief is perceptual when it is the output of one’s perceptual
system, without being the result of inferential combination of the output with other
collateral information. On the difference between perceptual and non-perceptual
beliefs, see Lyons 2005.
9 In order to realize that seeing-that is not just a perceptual episode but an
epistemic result of it, consider ascriptions of other non-visual perceptual episodes:
we do not ordinarily say that S touches that P, S smells that P, S tastes that P, just
because we cannot perceive facts, even though we can well come to know facts in
virtue of perceiving those objects, properties and relations that are their
constituents. Instead, hearing-that ascriptions are also quite in order because they
ascribe episodes of coming-to-know by auditory means. Indeed, a paradigmatic
case of hearing-that is testimonial knowledge, namely, understanding a spoken
sentence and become aware of its content as true. You hear sounds, maybe words,
30
Chapter One Part III
literally, namely, it entails the reference to a perceptual activity involving
the eyes, not just a generic way of obtaining knowledge.10
With this in mind, we can distinguish at least three kinds of seeing-that
episodes, according to the nature of visual involvement, in the process of
knowledge-acquisition the seeing-that is a result of:
1) Literally and directly seen facts: S sees that P when P is made true by a fact
whose constituents are seen by S. I see that the cat is on the sofa. I see the
cat, I see the sofa, I visually appreciate the on-ness spatial relation.
2) Facts I come to know as holding on the basis of literally seeing other facts.
As soon as I see that the mailbox is empty—through seeing the
constituents of the fact: the mailbox, its being empty—I come to know that
the postman hasn’t arrived yet. I do not see this last fact, rather I come to
know it as holding by actually seeing another fact (where seeing that last
fact does involve seeing its constituents).11 So, case 2) involves case 1). I
cannot indirectly see that Q unless I literally see that P.
3) Facts I “see” only metaphorically. I come to know that Q but I do not
literally see constituents of Q, or constituents of the facts that are the
proximate sources, or epistemic basis, for coming-to-know that Q.
In both cases 1 and 2, my seeing-that P rests on being in certain visual
contact with the constituents of a fact (objects, properties, relations). Only
the first counts as literally and properly seeing-that, whilst the second case
but you do not literally hear the propositional content, rather you understand it in
virtue of hearing sounds and words.
10 Also object-seeing holds the same literal/metaphorical distinction. Our analysis
of ordinary object-seeing ascriptions in Part 2 does not concern ascriptions like “S
sees the Nothingness”; “S sees the Uselessness of Existence” and the like. Those
cases do not concern philosophy of perception because do not have to do with
perception. Any episode of seeing that could be truly ascribed to a blind subject
does not concern our inquiry.
11 It is in this second sense that sometimes we claim to see past facts, facts whose
holding in the environment are no longer current: I see that someone came in with
muddy boots last night. As Armstrong remarks (see Armstrong 1968, 211) in such
cases we can only speak of seeing-that P, never of seeing the objects and
properties that constitute the fact that makes P true. We wouldn’t appropriately
claim that we saw the muddy boots or the person, for example. Rather, what we
see is the boot-marks on the ground, so the only literally and directly seen fact is
that there are boot-marks on the ground. A negative test for a case being of the
kind 1 or 2 is that of asking: do I also see the constituents of the fact I claim to see?
If it is not the case, then I am self-ascribing a case of kind 2—namely, coming to
know a fact in virtue of seeing another fact (= in virtue of coming to directly know
that fact just by visual means). Kind 2 can even concern the past, as in the case
above.
Seeing That P, Seeing That a Is F
31
is a derivative ascription where the visual means are only indirect. They
concern knowledge of another fact, Q, on which the knowledge of the P
involved in seeing-that-P is epistemically based. This involves an
inference or another kind of cognitive transition from one proposition to
another. While cases of type 3 do not concern us at all, cases of type 2 do,
although only insofar as they involve instantiations of cases of type 1. So
our focus now is on type 1.
What does it take then to see that P—or that a is F—in the sense of
case1?
Dretske calls seeing-that “epistemic seeing”12 to differentiate it from
“simple seeing” or object-seeing. Not only is epistemic seeing a propositional
state—an attitude toward a proposition—but being in such a state involves
conceptualizing the constituents of such a proposition. So seeing-that the
vase is on the table amounts to coming-to-know-that [the vase is on the
table] and this entails that S possesses and exercises the concepts [vase]
and [table]. No cognitive system not possessing the concepts a and F could
ever see-that [a is F]. So seeing-that is propositional, cognitive,
conceptual, and its ascription generates intensional and opaque contexts,
whilst seeing-X ascriptions are not necessarily cognitive or conceptual.
Instead, they generate extensional and transparent contexts.
Now, under this interpretation of the ordinary ascription of seeing-that
P, such an ascribed episode is not just a perceptual episode, although it
involves at least a perceptual episode as (one of) the means to obtain the
perceptual knowledge that seeing-that P entails. At this point of our
inquiry, we can ask: which relation holds between such a visual episode
and the propositional-conceptual state of perceptual knowledge resulting
from it?
III.5 Object-Seeing and Fact-Seeing
To begin with, we could notice that, even though seeing something is
not a sufficient condition to see that something is something, seeing
something seems to be a necessary condition for an episode of seeing-that
to occur. So
S sees that [a is F] → S sees a
Let us consider what I call the object-property scheme [a is F] for P.
Of course, a proposition can also have relations as its constituents. In
12 Dretske 1969, 1990, 2000.
32
Chapter One Part III
addition, it is quite normal to come to know by visual means that
something has a certain relation with something else; for example, that the
table is bigger than the chair. I can gain that elemental piece of perceptual
knowledge at once just by looking at the table and the chair. Nonetheless,
the object-property scheme is a more primitive way to start. Actually,
coming to know that a and b entertain a certain relation entails coming to
know that a has a certain property and b has a certain property in the first
place. So, it seems legitimate to begin with the more simple cases. Two
objects can entertain a relation only by virtue of having certain properties,
so that perceiving a relation entails perceiving properties of the relata.
As Jackson points out,13 seeing-objects is even conceptually more
primitive than seeing-facts. An argument for this is that seeing an object a
can be immediate—not mediated by seeing some other object b in virtue of
whose vision S can be taken to see a. On the contrary, Jackson successfully
argues, S seeing a fact (i.e., seeing that [a is F]) must be mediated by S
seeing objects and properties, the constituents of the allegedly seen fact.
However, I prefer to leave aside the “mediated/immediate” distinction
because it can capture different notions and so its meaning is potentially
ambiguous (just like the direct/indirect dichotomy).14 In any case, if seeing
that [a is F] entails seeing a, the a-e conditions spelled out in Part II for
seeing-X are necessary though not sufficient conditions for seeing-that [a
is F]. For any seeing-that ascription, there is an entitlement to ascribe a
respective episode of seeing-O, but such an entitlement does not hold the
other way round. We can see an object a by virtue of it looking F to us
without seeing that a is F.
Until now, we have not directly considered the role of seeing
properties besides the cases of seeing objects and seeing facts. But we
have pointed out that when S sees O, there must be a way O looks to S.
Actually this “way” is a property, the one through which the seen object is
given to S in her visual perception. So we basically see particular objects,
but only insofar as they are given in perception in some way. They look
some way to us, where “way” is a property.
As we have already said, no matter which way the particular object is
given in perception, in order for the particular to be seen, some property of
it must be given to S, which determines the way O looks to S. This could
be the color, the contour, the shape, the property of moving in a certain
way or direction, the size, or anything else. So object-seeing requires a
perceptual relation to a property, which enables the subject to discriminate
13 See Jackson 1977, Chapter I.
14 For a criticism of the direct/indirect distinction, see Austin 1962, Martin 2008,
Part 4. Also Travis 2004 (p.66) argues that such a distinction is occasion-sensitive.
Seeing That P, Seeing That a Is F
33
the object and be visually-phenomenologically conscious of it (“under” the
property, so to speak). In other terms, we see objects (also) in virtue of
becoming visually conscious of some of their properties.
III.6 Conceptuality
But if S seeing an object entails this object visually looking some way
to S, that condition is far from being sufficient for seeing that the object is
that way (the way it looks). To repeat the point, in order for S to see that a
is F, a looking F to S is a necessary but not sufficient condition. What else
is needed?
First, S must possess and exercise the concepts involved in the
proposition that expresses the seen fact. If S is to see that a is F, S must
conceptualize a and F. Without possessing the concepts of [dog], [tree],
[running-toward], S cannot see that the dog is running toward the tree,
even if S could well see the dog, the tree, as well as the movement of the
running dog.15 Seeing-that amounts to coming-to-know. Knowledge is true
belief and beliefs involve concepts, so seeing-that involves concepts.
Second, S must cognitively grasp the structure of the fact. Seeing-that
is a propositional attitude and propositions are structured entities, such that
attitudes toward those structured contents are structured states in the same
way. Suppose I am seeing a red cube under a blue triangle such that I am
visually conscious not only of the [red] and [blue] properties but also of
the [being-under] spatial relation. Moreover, suppose I also possess the
respective concepts and deploy them when perceiving the scene. If seeing
the cube and the triangle and being visually conscious both of the [red][blue] properties and of the [being-under] relation were a sufficient
condition for seeing that the red cube is under the blue triangle—given
also the subject’s relevant concept-possession—then such a condition
would always entail such a seeing-that. But this is patently false: such a
condition holds also when I see a blue triangle under a red cube, when I
see a red triangle under a blue cube, when I see a blue cube under a red
triangle, and so on. Given the same visual awareness of the same objects,
properties and relations, what makes the difference is exactly the (grasp of
the) structure of the seen fact. Given a certain scene including a certain
15 Of course there can well be de re attitudes, like demonstrative thoughts and
beliefs. I can see that dog, and of/about that particular I see, I can see that it is
running toward a tree. In any case, even if I can see-that a is F, where a is to be
interpreted extensionally and demonstratively—I see that that is F, no matter how
you ascribe or fix the reference of that particular, still I must at least possess the
concept of F in order to see that that is F.
34
Chapter One Part III
structured fact—certain objects having certain properties and standing in
certain relations—the visual awareness of its constituents is necessary but
not sufficient for seeing this fact; for example, for seeing that aRb (bRa is
another fact despite being made out of the same constituents).
If S possesses the relevant concepts and a looks F to her, then S is in a
position to see that [a is F]. It usually happens, but it is not necessary.
Indeed, usually possessing the relevant concepts and seeing O, which
visually looks F, causes her to come to believe that a is F, which is her
coming to know that a is F if a is F and the perceptual relation to this
circumstance is appropriate.16 As we will see in the second part of this
work, perception is not to be identified with either belief or beliefacquiring, even if perceptions usually lead to doxastic states. Provided
seeing-that is acquiring a certain knowledge and knowledge involves
belief, it is clear that neither seeing O nor O’s looking F to S entails as
such S’s seeing-that O is F.
Let us sum up the points we have established so far: seeing-that is the
ascription of a propositional attitude. It is the ascription of a factive state,
so that seeing that P entails P. Such ascriptions generate referentially
opaque or intensional contexts. Both factivity and opacity are well
explained by interpreting the ordinary meaning of seeing that as
amounting to coming to know by visual means. So seeing-that ascriptions
not only ascribe a perceptual episode but—when used not
metaphorically—also the acquiring of a piece of empirical knowledge on
the basis of a perceptual episode. When the acquiring is not inferential,
seeing-that (coming to know facts by visual means) entails seeing the
objects, properties and relations that are the constituents of those known
facts as well as possessing the relevant concepts and having a cognitive
grasp of the structure of the perceptually known fact. If having
propositional attitudes is a privilege of rational and conceptual beings,17
then only rational and conceptual beings can see that something is the
case.
16 There can well be “veridical illusions.” For example, suppose S’s perceptual
system is not working properly—O looks F to S and it happens that O is actually F.
If S comes to believe that O is F on the basis of a wrongly-caused perception, she
would not come to know that O is F thereby because the visual means grounding
her perceptual belief would be not appropriate to justify her belief and she would
not be able to turn it into knowledge. Such cases are a sub-class of Gettier cases
concerning perceptual knowledge. On veridical illusions, see Chapter Three, Part
2.4.
17 See Davidson 1982.
Seeing That P, Seeing That a Is F
35
III.7 Definition
We could embody all the relevant acquisitions in a working definition:
S sees that P if, thanks to becoming visually aware of those objects, properties
and relations that are the constituents of the environmental fact that makes P
true, by conceptualizing those constituents and by grasping the structure of that
very fact, S comes to know that P is the case.
PART IV
SEEING-AS, SEEING SOMETHING
AS SOMETHING
IV.1 Implicativity and Normative Evaluability
Besides objectual seeing (seeing O) and propositional seeing (seeing-that
P), there is another important kind of seeing-ascription to be carefully
considered, namely, the seeing-as ascription.1 The basic scheme for this
kind of ascription is:
S sees a as (an) F
Unlike seeing-that, seeing-as is not factive. The implication [S sees a
as (an) F] → [a is F] is false indeed. S can see a particular as F even if that
particular is not F. I can see a dark moving object on the street as a mouse
even if it is actually a piece of paper moved by the wind in the dark.
Like seeing-O, seeing-as is implicative. In order to see a particular as a
mouse or as a piece of paper, I must see it in the first place.2 The
implicativity of seeing-as depends on a respective object-seeing episode
being involved in each seeing-as episode as its constituent:
1 Surprisingly enough, Dretske does not consider that kind of seeing at all. As we
will see later, this omission has fatal consequences not only for his semantics of
“see” but also for his substantive philosophical theory of perception. As far as I
know, all he has to say about it is in a small note (see Dretske 1990, 133, footnote
1) where he recognizes seeing-as as a hybrid way of seeing between object-seeing
and fact-seeing.
2 Again, we should keep in mind that we are considering only literal ascriptions of
seeing-as, those essentially involving an ascription of a visual perception. Just as S
can see taxation as state robbery, S can see unicorns as nice animals. We are not
interested on such non-perceptual uses of seeing-as so we will not talk about them.
When literally ascribed, seeing-as is implicative so S cannot see a as an F unless
she sees a.
Seeing-As, Seeing Something As Something
37
[S sees a as (an) F] → [S sees a]
Maybe what I see as an F is not an F, still I must be successful in
seeing it to wrongly see it as an F. This entails that, as spelled out in Part
II, the a-e conditions for seeing-O are at least necessary conditions for
seeing-as as well. In order to see my wife as a hat,3 I must see my wife in
the first place, so the conditions for seeing my wife as a hat include those
for seeing my wife simpliciter.
Therefore, in order to see a as F, I must see a on the one hand, but I do
not need to see that a is F on the other. Indeed, seeing-that a is F is
incompatible with a not being F (factivity), whilst seeing a as F is well
compatible with a not being F insofar as it is not factive.
Therefore, a very important property of these kinds of ascriptions is
that they can be correct or incorrect; in other words, there are correctnessconditions associated with them. Seeing-O as such cannot be correct or
incorrect. Either you see O or you don’t. Likewise, seeing-that P cannot be
correct or incorrect either. Either you see that P or you don’t. You cannot
incorrectly see that P, nor can you correctly see that P for the same reason.
Seeing-that is an epistemic achievement, it is factive indeed; seeing-O is
also an achievement insofar as it is implicative and therefore presupposes
a real seen object. Seeing-O and seeing-that-P are essentially successexpressions as we have already noted.
Only seeing-as, then, exhibits the peculiar capacity of being
normatively evaluable—it can be right or wrong. With seeing-as
ascriptions, the possibility of error concerning seeing-episodes definitely
comes into play. Indeed, until now, seeing-as episodes are the only seeingepisodes that can be mistaken. Let us analyze the ordinary contexts in
which seeing-as is truly ascribed to a perceiving subject.
What do we ascribe when we claim that S sees a as (an) F? Is it a
cognitive and epistemic achievement based on a perceptual episode (like
seeing-that), or is it rather a thoroughly perceptual episode (like seeingO)? To address these questions, let us start from the paradigmatic case of
seeing-as made famous be Wittgenstein,4 that of the duck-rabbit
ambiguous figure. Ambiguous figures are a very specific subset of seeingas episodes so this is just a way to start our analysis. We are interested in
ambiguous figures only insofar as they are a well-known case of seeingas. In the case mentioned, a figure is presented—a group of signs and
lines—which can be “interpreted” as depicting a rabbit (be it F) or as
3 See Sacks 1985.
4 See Wittgenstein 1953, Part II, Section xi; Wittgenstein 1991, p.1-29. On this,
see Bozzi 1998.
38
Chapter One Part IV
depicting a duck (be it G). So the same seen O can alternatively be seen as
an F or as a G. As is well known, the Gestalt-shift from one visual
interpretation to another is discrete and can be prompted by a voluntary
act. Whenever I want to, provided that I am able to visually “see” both the
figures in the same drawings, I can suddenly “shift” from one figure to the
other. So seeing-as ascriptions seem to be ascriptions of acts of
recognition. If S sees a as an F, S recognizes an (example of) F in the seen
object a. Seeing-as is recognizing-as.
IV.2 Recognition
But there is a caveat. Although useful, the example of ambiguous figures
can also be misleading for our concerns for the following reasons: firstly,
it is a case of cognitively sophisticated visual interpretation of stylized and
two-dimensional drawings, which presupposes a fine-grained competence
the idea of abstract figurative representation and the capacity of
suspending judgment about what the seen object really is. Perceivers
sensitive to visual flip-flopping obviously know that what they are actually
seeing is neither a rabbit nor a duck—and possess many other high-level
cognitive skills. Seeing-as can be far more primitive than that. For
example, it can be meaningfully ascribed to an animal. Secondly, and most
importantly, considering what you “can see” in a stylized drawing could
fatally conceal the fundamental property of seeing-as, namely, the
circumstance that it can be mistaken. Whatever you see in a drawing, is
something you recognize insofar as you are playing a sort of game in
which you tell “what you see.” It is not the ordinary situation where
someone would tell you: “You are wrong; there isn’t anything like what
you claim to see!” Recognition is factive. If you recognize an F, what you
have so recognized is an F. For example, you cannot visually recognize a
real rabbit unless what you are seeing is a rabbit. If this is not the case, at
least you can believe, think, take it that you have recognized a as a rabbit
(or a rabbit in a), but you are wrong. You have not recognized anything.
So seeing-as can consist of an episode of recognition when it is correct
and successful, but it does not have to consist of that. It can happen to be a
failure as well.
Therefore, in successful cases, seeing-as is perceptual recognition, the
exercise of a positive recognitional disposition or capacity. But on closer
inspection, also in unsuccessful cases, seeing-as is still the (failed)
exercise of a positive recognitional capacity. This means that by ascribing
to S an episode of seeing a as an F, you credit S with the general positive
capacity of recognizing examples of F even in case you are consciously
Seeing-As, Seeing Something As Something
39
ascribing a mistaken occurrent exercise of that general recognitional
capacity. I can wrongly see a bat as a bird only if I have a general capacity
to see a bird as a bird, even though my actual exercise of such a
recognitional capacity is a failed occurrence of my positive cognitive
disposition.
So ascribing to a subject an episode of seeing-as, be it ascribed as
being a right or wrong token, entails ascribing a positive recognitional
capacity that can be recursively exercised, and whose actual occurrence is
the essentially repeatable token of that type. If S sees a as an F, then F is
something S has cognitive familiarity with; it is something S can generally
recognize. Seeing-as entails general sensitivity to a type. What kind of
recognitional capacity is involved in seeing-as? Seeing-as can be taken as
a three-place relation whose constituents are a subject, an object and a
property satisfying the as-complement, so we can put the question as
follows: which types of properties can stand in the F-place within such a
three-place relation [S sees a as (an) F?] in literal cases of seeing-as?
To begin with, seeing-as involves a cognitive stand toward the
perceived object. “Seeing a as an F” means taking something as an
(example) of F. If I see a as an F, I see an a, and I take what I see as an F. I
can be wrong in seeing-as because seeing-as is a cognitive “taking,” an
exposure to error.
Second, according to the kind of property that stands in the F-place,
seeing-as can consist of different kinds of cognitive acts. Let us carefully
make the point more explicit.
We have already ruled out all ordinary uses of the scheme [S sees a as
an F] where a is not the kind of thing that can be literally seen. Seeing an
apple, a table or a rabbit as an F concerns us, whilst seeing the
government, life or justice as F does not insofar as “seeing” here has a
metaphorical, non-perceptual meaning. So what about F? Provided that
seeing-as involves a taking-as directed at the (literally) seen object, which
properties are the right complement for the episode to be of interest to a
theory of perception? In other words, which arguments for F are such that
seeing-as is a perceptual episode though with cognitive significance? For
sure, seeing a rabbit as the meaning of my life, seeing a tree as God’s gift
and the like are not perceptual episodes of seeing-as, even if they remotely
involve respective cases of object-seeing because the argument for a in the
ascriptive scheme is filled by typically perceivable particulars.
40
Chapter One Part IV
IV.3 “Thick” Categories and Sensible Profiles
Often the place of F is filled by sortal terms, such as “rabbit” and “duck”
in the example above. Let us take the example of seeing an object as a real
rabbit. Now, is [being a rabbit] a property that can be perceptually
detected? [Rabbit] is a natural kind. What determines whether an animal
belongs to a given natural kind are certain genetic-evolutionary features
whose discovery can be the object of a long-lasting scientific inquiry. So
such a natural-kind property does not seem to be the property a subject can
perceptually see-as. Seeing-as amounts to taking a certain cognitive stand
toward the seen object, and this stand is specified just by the F-property
the object is taken to be. But when we ordinarily ascribe to subjects their
seeing something as a rabbit, we ascribe the exercise of the ordinary
[rabbit] concept or category—namely, sensitivity to a stereotypical
intuitive class of particulars whose exercise is activated by the recognition
of a certain sensible profile in the seen object. When seeing-as episodes
are perceptual, the “as-an-F taking” act must depend on certain visible
traits that are typically and evidently associated with examples of Fs. To be
a perceptual episode, seeing-as must be recognition of something as an F
“just by sight,” without inferences grounded in collateral knowledge or
other cognitive transitions involving guessing, reasoning, hypothesizing
and the like. For example, I see something as a rabbit only if: 1) I possess
the concept [rabbit] or at least the category {rabbit}, on which a positive
recognitional capacity of examples of rabbits is grounded; 2) I am able to
exercise this recognitional capacity towards a so as to see a as an F just by
looking and visually appreciating a certain sensible profile that
immediately prompts in my previously established perceptual scheme the
activation of a category or a concept [F].
I say “a category or a concept” because seeing-as need not be
conceptual, even if it is essentially cognitive and involves the subsumption
of a particular seen object under a general type. It is true that when an
adult human being is ascribed an episode of seeing something as a rabbit,
she is ascribed the possession and exercise of the (ordinary and
stereotypical) [rabbit]-concept thereby. Nonetheless, seeing-as-F
ascriptions as such not only do not necessarily entail S’s possession of the
[F]-concept, they do not entail any concept-possession at all. We can
legitimately ascribe to a dog an episode of seeing his master as his leader
of the pack without ascribing it either the possession of the [leader of the
pack]-concept, or any concept-possession at all. We can claim that a lion
sees a running gazelle as prey without committing ourselves to ascribing
to the lion concept-possession (possession of the [prey]-concept or any
Seeing-As, Seeing Something As Something
41
other concept). However, we could not ascribe to the dog an episode of
seeing his master as the leader of the pack, nor could we ascribe to the lion
an episode of seeing a gazelle as prey unless the dog and the lion possess
the respective recognitional categories in their “cognitive spaces,” so to
speak.
For sure the general and recursive recognitional disposition concerning
Fs is a rougher and more inarticulate dispositional state than the
possession of concept F. Possessing the concept of [leader of the pack]
involves being able to make many inferences, being able to understand
propositions containing the concept, and so on.5However, in order for a
dog to have the “category” of {leader of the pack}, it needs to manifest a
recursive recognitional capacity on the basis of which its behavior is
generally interpretable. The dog’s {leader of the pack} category has a
vaguely similar content to our [leader of the pack] concept, and such a
similarity is sufficient to ascribe to a cognitive system a certain seeing-asF state using a concept in the ascription without committing oneself to
attributing possession of that concept to the subject. To sum up this last
fundamental point: seeing-as is cognitive but it is not necessarily
conceptual nor is it epistemic. It presupposes a general recognitional
capacity we have provisionally called “category.” Category-possession is
much less demanding than concept-possession and can be ascribed just on
a behavioral basis. It does not entail inferential and rational capacities, for
example.6
5 See Chapter Two, Part I.6. Unlike concept-possession, category-possession does
not require the Generality Constraint to hold: in order for S to possess the
categories F and G, F does not have to grasp what it is to see a as G and b as F just
because S sees a as an F and b as a G. Likewise, seeing-as or recognitional
categorization does not require what I call the Rationality Constraint to hold (see
next chapter): in order to see a as (an) F, you do not need to be able to draw a set
of relevant inferences. You do not need to have inferential abilities at all insofar as
perceptual seeing-as is proto-propositional rather than a fully-fledged propositional
attitude, like seeing-that.
6 So “seeing-as” denotes the occurrent episodic exercise of a recognitional predoxastic capacity/disposition or pace. Tye (2000, 215) writes: “Object or shape
recognition in vision […] is a matter of seeing that such and such a type of object
is present. Seeing that something is the case, in turn, is a matter of forming an
appropriate belief or judgment on the basis of visual experiences or sensations […]
there are two components in visual recognition, a belief component and a looking
component.” There are indeed a recognitional, cognitive component (thick
categorization) and a looking-, strictly visual component (sensible profile), but the
recognitional component is not necessarily a belief nor does it involve a seeingthat episode. If it were the case, it would sound inappropriate to ascribe a seeing-as
42
Chapter One Part IV
Let us now consider an example of seeing-as made by Crane.7 An
infant and a scientist are both looking at a cathode ray tube. Even though
both of them see it, the difference between them is that the scientist sees it
as a cathode ray tube because he possesses the concept of cathode ray tube
and exercises it while perceiving, whilst the infant cannot see it as a
cathode ray tube just because she lacks the respective concept.
Well, it is true that seeing something as a cathode ray tube entails
possessing the [cathode ray tube] concept, which in turn entails having a
certain net of beliefs and a certain inferential capacity concerning
reasoning and propositional knowledge about cathode ray tubes, their
functions and so on. In the same vein, we cannot see an object as a toaster
without possessing the [toaster]-concept. However, such a condition does
not depend on seeing-as as such; it depends on the specific properties
under which the seeing-as subsumes the seen objects. The properties of
[being a cathode ray tube] and [being a toaster] cannot be grasped by a
non-conceptual being—nor can they be grasped by a conceptual being
lacking those respective concepts, unless grasping them is the very
acquisition of the respective concepts—but this does not mean that any
general recognitional capacity or categorical object-typing needs to be
conceptual thereby. On the contrary, there is a non-conceptual seeing-as,
even though it cannot concern the visual recognition of those properties
only conceptual beings can be cognitively sensitive to. By seeing-as, I just
recognize something—known or familiar to me—in an object by virtue of
visually appreciating its sensible profile. Not every recognition involves a
conceptualization. Some recognitions do (i.e., of a cathode ray tube, of a
toaster), some don’t.
IV.4 Definition
Let us try to provide a definition of ordinary seeing-as.
S sees a as an F if: S sees an F in a thanks to the exercise of a general
recognitional capacity for Fs prompted by the visual appreciation of a’s
objective sensible profile.
From all we have said it is clear that an important feature of seeing-as
ascriptions is their referential opacity. Such opacity depends on the fact
episode to animals as we do not normally credit them with beliefs, concepts and
propositional attitudes. Unless we do accept a very undemanding notion of belief,
then, Tye seems, to me, plainly wrong in assimilating seeing-as to seeing-that.
7 Crane 1992, 3ff.
Seeing-As, Seeing Something As Something
43
that they are ascriptions of a certain cognitive stand of the subject toward
an object, of a certain taking-as. Even if taking-as is non-conceptual but
“categorical” or recognitional, nonetheless in “S sees a as (one) F” you
cannot salva veritate substitute the “F” with co-referential expressions. Of
course, you can substitute a with a co-referential expression b, because
seeing-a is transparent and extensional. What is not transparent and
extensional is not the specification of what you see (a); rather it is the
specification of the way you cognize or categorize what you see (of the
“F” you take a as being). So seeing-as ascriptions are referentially opaque,
even if they entail transparent seeing-O ascriptions as their constituents (as
was the case in seeing-that ascriptions as well).
IV.5 What Is a Sensible Profile? The SCM-Properties
A last remark is worth examining. Even if the arguments for F in [S sees a
as (an) F] are usually sortal terms—prey, a rabbit, a cathode ray tube, a
duck, an apple, a table and so on—actually this is not always necessarily
the case. The argument for F may well be the property of [being a red
thing], [being of a certain shape] and so on. Despite the fact that an
expression like “seeing something as red” sounds less natural than “seeing
something as an apple,” the logic of ascription is the same. S sees an
object as having a certain color that S can recursively recognize, a color to
which S is perceptually and cognitively sensitive.8 In cases of seeing
something as red, as square, as having a certain size and so on, there is not
just a categorization prompted by the appreciation of a sensible profile;
there is also recognition of properties directly belonging to the strictly
visually appreciated sensible profile. So there are seeing-as episodes that
concern purely perceptual properties even if they are cognized or
recognized within the episode. Accordingly, there are two possible kinds
of mistake involved in seeing-as: a mistake consisting of prompting the
wrong category in front of a particular exhibiting a sensible profile that is
correctly appreciated by the subject, and a mistake consisting of failing to
8 The very same happens with the locution “O looks F” or “O looks as an F”: the
argument for F can be an adjective, like “red,” or a sortal term like “a tomato,” but
both refer to properties, the property of being red and the property of being a
tomato. See Price 2011, 141. In the same vein, we can say that S sees O as F or S
sees O as an F, according to whether F denotes a property to be expressed with an
adjective, like [red], or a sortal property to be expressed by a name, like [a tomato].
Anyway, both “seeing something as a red thing” and “seeing something as a
tomato” are ascriptions having the same structure although the properties referred
to are quite different.
44
Chapter One Part IV
appreciate properties the very sensible profile is made of, so to speak. An
example of the first mistake could be where S correctly sees a shape, a
color and a size (a case of seeing the sensible profile as it is) but she takes
the particular exhibiting the profile to be an F when it is not an F. An
example of the second mistake could be where you see a particular as
having a shape, and/or a color, and/or a size that the particular does not
objectively exhibit. The latter is a purely visual mistake, so to speak. The
first is a cognitively thicker mistake.
What do we mean by “the recognized properties which directly belong
to the sensible profile”? We mean a set of properties the visual system is
sensitive to in the first place. In what follows, I will call them spatialchromatic-morphological properties (SCM)—namely, those properties the
subject must be visually aware of in order to be able to take that visually
appreciated sensible profile as being something known, as being an
(example of) F.9 Without seeing an object as having certain colors, certain
shapes and sizes and certain locations and spatial relations with the
surroundings, a subject would not be in a position to take any seen object
to be anything, because she would not be in a position to see anything in
the first place. I call an instantiated cluster of those basic SCM-properties
an objective “sensible profile.”10 A particular O is seen by S by virtue of S
being aware of some properties of O’s sensible profile, by virtue of O
looking some way to S, as expressed above.
IV.6 The Features of Seeing-As
Let us finally sum up the long but necessary digression about seeing-as
ascriptions:
1) [Seeing a as (an) F] is implicative (a must be there to be seen) but it
is not factive (a may not be an F). It has correctness-conditions insofar as
it may be wrong/incorrect/mistaken.
2) It does not entail seeing-that a is (an) F even though it puts the
subject in a position to see-that a is an F, provided the subject possesses
the concept of F and the relevant propositional capacities. However, seeing
a as an F does not necessarily involve either the [F]-concept-possession or
conceptual abilities.
9 In the same vein, McGinn 1982 (p.42) calls those properties the “manifest
properties” of an object.
10 I take the expression from Johnston 2004, but I do not mean to commit myself
to his way of characterizing it.
Seeing-As, Seeing Something As Something
45
3) Nonetheless, seeing-as is cognitive and involves a taking-as stand
toward the seen object, consisting of the exercise of a positive
recognitional “categorical” capacity. Such ascriptions are referentially
opaque insofar as they ascribe a way an object is cognized by the subject
and from the point of view of the subject.
4) Finally, when it is used non-metaphorically but literally as referring
to episodes of recognition-by-vision, seeing-as presupposes the visual
awareness of a sensible profile—made out of spatial-chromatic-morphological
properties—as a strictly perceptual basis for the thicker categorization (in
case F is not itself a SCM-property, like being of a certain color, shape,
size, location and the like).
So far so good for seeing-as ordinary ascriptions.
PART V
LOOKING, SEEMING, APPEARING
V.1 Looks/Seems/Appears: Analogies and Differences
Ascriptions of visual perceptions often involve predicates—like
“seeming,” “looking” and “appearing”—belonging to the “experiential”
vocabulary, which characterize the ascribed visual episode from the point
of view of the subject, so to speak.1 Let us start our analysis by using
everyday examples. Consider some common ordinary uses of “looks”:
1) a looks F (to S)
2) a looks like (an) F
3) a looks (to S) as if it is F
4) it looks (to S) that [a is F], it looks as though [a is F], it looks as if [a is F]
5) There looks to be an a which is F
The looking-ascriptions are but only one relevant type among the
ascriptions involving what Chisholm calls the “language of appearing.”2
Indeed, analog statements could be constructed with “appears” and
“seems”—when used in visual contexts—in the place of “looks.”3
Although, as Austin suggests,4 there are relevant and philosophically
interesting differences between appears-, seems-, and looks- ascriptions in
ordinary discourse. Firstly, suitable contexts of ascriptions for “looks” do
normally involve vision. At least, they involve episodes of seeing, so that
1 As we will see, “looks F” often characterizes something as having an objective
look. When I say that expressions with “looks” have to do with the point of view
of the experiencing subject, I mean that even objective looks are potential ways of
being experienced by one (by everyone, by a “normal” perceiver) as looking in
such-and-such way. Without any subject of experience at all, even objective looks
would just be purely potential looks.
2 See Chisholm 1957, Chapter 4.
3 Austin 1962, 33-43.
4 Ibidem.
Looking, Seeming, Appearing
47
when they are used with reference to non-visual episodes, they are
metaphorical applications stretching the proper sense of “looking” in the
same way as “seeing” itself can be stretched to cover non-visual cases (see
above). On the contrary, “appears” and “seems” may or may not carry an
implicit reference to vision or literal seeing. An a may well nonmetaphorically appear or seem F without a being seen or, a being a
typically visible entity without its appearing F being the consequence of a
visual episode (for example, of the appreciation of a visually given
sensible profile). According to Austin, while “looks” involves a visual
appearance, “appears” tendentially involves a reference to special
circumstances and “seems” tendentially involves a reference to evidence.
Let us consider this taxonomy by considering an example: a) “He looks
guilty”; b) “He appears guilty”; c) “He seems guilty.” Approximately, they
respectively mean: a) his visual appearance is typical of guilty men (be he
guilty or not); b) Given such unusual circumstances, one may take him as
being guilty (even if he may be not such); c) There is prima facie evidence
that he is guilty.
Therefore, a man can seem guilty without looking guilty. Independently
of his visual appearance, there may be evidence that he is guilty.
Conversely, a man can look guilty without seeming guilty. Despite
exhibiting a visual appearance that is typically associated with guilt, there
is non-visual evidence that he is not guilty. Therefore, a man can appear
guilty without either looking or seeming such. For example, he can appear
to be guilty given such and such extraordinary coincidences, but,
fortunately for him, there is independent evidence that these are just
coincidences, and so on. Of course, these three uses are not unrelated: the
a’s distinctive visual appearance (look) may well be used as evidence
(seem) supporting the proposition that a is F. Likewise, a distinctive look
exhibited by a may be what you appeal to when you state that a appears F
in special circumstances. But there is no immediate implication from
looking to seeming, or from looking to appearing, or the other way round.
You could understandably say that the moon looks no bigger than a
sixpence even if you would never say that the moon seems no bigger than
a sixpence.5 “Seeming” has to do with non-conclusive evidence, and is
5 “Seems”-, “looks”- and “appears”- ascriptions are occasion-sensitive, so to
speak. Depending on the context, even their relation changes. For example, as
Austin suggests, football players seen from the highest seats of the stadium can be
said to “look like ants” without this implying at all that they “seem ants” thereby.
That an animal looks like a pig in the distance could eventually be used as prima
facie evidence that it is a pig, so that one could say that it also seems a pig. But if a
48
Chapter One Part V
therefore appropriate for cases when it may or may not be the case. If you
know that a is not F, you do not generally say that a seems F, nor do you
say so if you know that a is F. A particular feature of “seems,” having to
do with its reference to evidence, makes its logical behavior quite different
from both “appears” and “looks.” You can say that judging from its
appearance, something is such and such, you can say that judging from its
look, something is such and such, but you cannot say that judging from its
seeming, something is such and such. This depends on “lookings” and
appearances being facts on which a judgment can be based, whilst
“seeming” is already a prima facie judgment or an inclination to judge. In
other words, “seems” refers to a certain body of evidence, “looks” and
“appears” refer to facts that may eventually count as non-conclusive
evidence for a proposition that a “seeming” consists of. Given all this, and
also that “appears,” when used as involving vision, can be substituted by
“looks” plus a reference to some special circumstances,6 it is clear that the
constructions involving “look” are of the greatest interest for our concerns.
Indeed, in their most natural use, they seem essentially to have to do with
visual perception, differently from “appears” and “seems.” Consequently,
let us focus on those.
V.2 A Principle Governing “Looks”-Ascriptions
In the Austinian example provided above, the predicate standing as a
complement of “look” does not actually denote a typically visible
property: [guiltiness], whatever it is, is not the kind of property we would
take as observational, as something the eyes could be directly sensitive to
in the first place.7 Nonetheless, seeing someone and appreciating a certain
cloud is said to look like a pig instead, there would not be any temptation
whatsoever to say that the cloud seems a pig thereby.
6 Of course, not every occurrence of “appears” implicitly or explicitly entails the
presence of special circumstances. Ordinary discourse is occasion-sensitive,
context-dependent, variable and heterogeneous. Austin himself does not mean to
argue that a reference to special circumstances is necessary for “appears” to be
appropriate. Rather, he tries to reconstruct the use of such an expression according
to what a normal speaker usually and generally does with that expression.
7 I argued above (see Part 2) that we can see an object only in virtue of being
visually aware of some of its properties. In this sense, if S sees O, O must look
somehow to S. Now, nothing can be literally seen only by virtue of looking guilty.
Rather, it can be seen by virtue of looking blue, so-and-so shaped, moving and the
like. For the time being, this intuitive test should be enough to demarcate
observational properties from non-observational ones.
Looking, Seeming, Appearing
49
cluster of visual SCM-properties typically associated with that “thick”
property ([being guilty]) can make it possible that one looks guilty,
appears guilty according to its look, seems guilty to one if one is to judge
him just on the basis of his actual look. So we can express this as a general
principle:
In order for a to look F, when F is not an SCM-property, a must look G in the
first place, where G is a complex of SCM-properties that individuate a certain
visual profile.
Let us forget guiltiness then, whose recognition just by looking, by the
way, presupposes high-level skills at recognizing facial patterns, at finegrained psychological interpretations of expressions matched with verbal
and non-verbal behavior and so on. But even if we take another non-SCM
complement of “looks” involved in the ascriptive scheme “a looks F”—
like “a pig,” “a table,” “a tree”—still the principle just mentioned holds.
Something can look F (say a pig) only if it exhibits a sensible profile made
of a certain cluster of SCM-visual properties.8 For example, those
distinctive features typically characterizing pigs, like a pinkish color, a
certain shape and size, certain movements and the like. So what I have just
noted also entails that, for a to look F to S, S must possess the
recognitional capacity for Fs, namely, S must be able to be prompted to a
certain recognition by being able to associate a known/familiar type of
things with the as-familiar typical, distinctive sensible profile exhibited by
things of that type. So, we should distinguish expressions like “looks
angry, looks (like) a pig, looks European” and the like, from expressions
like “looks blue, looks round, looks to be moving toward me, looks big”
and the like. When “looks” is used literally, the first type of ascription to
one’s experience always and necessarily entails the second type. Roughly
put, there is a purely visual-perceptual “look” on the one hand, and a
cognitively loaded recognition based on the appreciation of a purely
8 For now, I keep totally neutral about whether properties like [table] or [pig] can
be represented in perception. According to the Content View on perception, O’s
looking F to S can, in some cases at least, consist of S’s visual experience
representing the property F as instantiated by O. According to our principle above,
if the Content View is true and if properties like [table] and [pig] could be contents
of perceptual experiences, then it would be true that, in order for an experience to
represent something as a pig or as a table, the very same experience must represent
the object as instantiating such-and-such SCM properties, like [pink], [of that size]
and the like in the first place.
50
Chapter One Part V
visual-perceptual look on the other.9 There is no thick recognition just by
looking without sensitivity to strictly visual properties.
V.3 Three Uses of “Looks”
Not only are there many possible constructions involving “looks,” there
also seem to be different and irreducible uses of this expression. A
classical taxonomy of these uses has been provided by Chisholm.10
Chisholm distinguishes three uses, which he respectively calls epistemic,
comparative and non-comparative. Jackson has successively articulated a
threefold taxonomy as well,11 though Jackson’s version is slightly different
as we will soon see.
Let us introduce and consider such a proposal for a critical assessment
in the first place.
The epistemic use is paradigmatically (but not only) given in the
propositional construction “it looks as if P”; for example, “it looks as if the
dog is running after a cat” or “it looks as if it will rain soon.” Such
sentences can be indexed to a subject (“It looks to me/to S that P”). Their
use is called epistemic insofar as they allegedly mean there is a body of
visually acquired evidence that supports the proposition P.12 Hence “it
looks (to me) as if P” roughly captures the situation referred to by Austin
when a certain visual look is considered potential evidence for a
proposition, resulting in a “seeming.” But while “seeming” is already
committal, it is a guarded assent; the epistemic use of “looking” is not
necessarily committal. That there is a body of visually acquired evidence
that supports the proposition P can well be compatible with the knowledge
that non-P, while such a circumstance is ruled out for “seeming”-claims. If
9 Here “based” does not entail inferential processes from premises to conclusions
of course, but only that given a capacity of a certain kind, a more basic capacity of
another type must be already in play.
10 See Chisholm 1957, Chapter 4.
11 See Jackson 1977.
12 Such epistemic uses of “looks,” individuated by Chisholm and Jackson,
correspond to the Austinian analysis of “seems,” where the evidence supporting
the proposition must be of a visual nature. For the epistemic sense of “looks,” see
also Alston 1999 and Lyons 2005. Lyons distinguishes a “pure” and “experiential”
sense of what he calls epistemic-doxastic “looks.” According to the pure use, X
looks F to S if S is disposed to believe that X is F; according to the experiential
use, X looks F to S if the way X looks to S disposes S to believe that X is F. We
are interested in the experiential use here. I have called the pure use, at least when
not essentially connected with perceptual means, “metaphorical” (just as seeing
when not involving the visual apparatus), and I leave it to one side.
Looking, Seeming, Appearing
51
I know or believe that it is not going to rain, I can say that it looks as if it
is going to rain,13 but I would not say that it seems that it is going to rain.
Maybe I have heard the official forecast so I know it will be sunny, but at
the same time it looks as if it is going to rain soon because the sky is
darkish and the air is wet. So “it looks as if” is a noncommittal claim that
there is visually acquired prima facie evidence supporting a proposition,
independently of whether P is known to be true or false for other reasons
(non-visual collateral knowledge and the like).
However, there is a non-propositional construction not considered by
Jackson, consisting of the scheme “a looks (to me) as if it is F” (let’s say:
F = triangular). As Maund points out,14 in such ascriptions (or selfascriptions) the object a must be a seen object that plays a causal role in
the ascribed experience of a looking F to one. Unlike “it looks as if [a is
F],” “a looks as if it is F,” for example, is not compatible with S not seeing
a at all. On the contrary, it may (even visually) look to me as if a is F
because I see that b is G and b being G supports the proposition that a is
F.15 So, in sentences like the latter, two distinct elements are involved: a
seen object a, and a distinctive way it looks, which can eventually provide
a reason for a certain belief about a (so to ground a “seeming” concerning
a). The non-propositional construction—let us call it de re-looking
construction—therefore involves reference to a distinctive visual
appearance of the seen object,16 while the propositional construction does
not.17
13 Yet it would not be something you say naturally. This may be why Travis
(2004, 76) does not agree: “It cannot look as if X on this notion where it is
perfectly plain that X is not so.” I find that if you say “It looks as if P even if notP,” you are saying that all the evidence immediately available supports P, but you
have other sources of knowledge that defeat such evidence. Maybe it sounds a bit
unnatural but it seems to be perfectly understandable.
14 I take the remark by Maund (2003, 137ff).
15 For example, it seems to me I recognize the postman coming toward my house
so I say: “It looks as if the mailbox is still empty.” I see the postman and his
coming toward my house, so that seen fact supports the proposition that the
mailbox is still empty (given that I have already taken home the post yesterday).
Seeing b (postman) and seeing that b is G (him coming to my house) supports the
proposition that a is F (the mailbox is empty). No causal role is directly played by
the mailbox in the “looking”; I don’t even need to be seeing it. Differently, if I say
that “the mailbox looks as if it is empty,” presumably I am seeing it, maybe
without opening it so that there can be room for doubt (otherwise I would just say
that the mailbox is empty).
16 As Sellars points out, in ascribing “X looks red to S,” I am endorsing the
existence of X, while in ascribing “It looks to S as if there is a red object over
52
Chapter One Part V
The second use of “looks” is called comparative by both Chisholm and
Jackson. It typically (but not exclusively) comes in look-like constructions,
like “X looks like (an) F,” “X looks like Y” and so on. For example, “Jack
looks like his brother”; “that zebra looks like a horse from here”; “Tom
looks like a poodle”; “the football players, seen from the high terraces of
the stadium, look like ants, or like black spots” and the like. This use
explicitly establishes a comparison between two things with respect to the
way each of them respectively looks (where only one thing needs to be
actually seen). They look the same way, namely, they exhibit a similar
visual appearance. The comparison may be indexed to a subject, to more
or less special circumstances, to a time, to a point of view and so on. For
example, to me, now, with this light and from my actual point of view,
that zebra over there looks like a horse. Indexing does not imply that
comparative looking-claims are intrinsically subjective. On the contrary, A
can look like B, or like an F, without further qualifications. Two things
can share a shape, a color, a whole sensible appearance just because that is
the way they are; they look similar appearing as they are and seen in
standard circumstances by anybody. Moreover, when a looking-claim is
made that is indexed to a subject and to her actual point of view, there is
not necessarily a reference to the individual subjectivity involved in the
particular episode. I can well say that X looks like Y to me, from here, in
these circumstances, meaning that such and such conditions make X and Y
look the same (or similar) to whoever looks at them in these conditions I
now find myself in.18 A football player seen from the highest seats of the
stadium, objectively—at least intersubjectively—looks like an ant, like a
black small rolling ball and so on. Ways of looking, and the relational
property of something’s looking like something else, are publicly
assessable features of objects.19 The comparative use, though, must be
there,” I am not even endorsing the existence of an object seen by S and looking to
her a certain way. See Sellars 1953, § 21. For example, the first can be a successful
perception or an illusion but not a hallucination; the second could be also a
hallucination without any object being there to be seen. Snowdon’s disjunctivist
treatment of “looking”- ascriptions also follows these lines: if (it looks as if there is
an F), either (there is something that looks to S to be F) or (it is to S as if there is
something that looks to S to be F). See Snowdon 1981, 185.
17 So, if it is true that “if I see something it looks somehow to me” (Shoemaker
1975, 299), as we have argued above, it must also be true the other way round—
namely, that if something looks somehow to me, I must be seeing it.
18 By “these conditions,” I mean the types of conditions of which my actual
conditions are a token.
19 As Austin brilliantly points out, “the way things look is, in general, just as
much a fact about the world, just as open to public confirmation or challenge, as
Looking, Seeming, Appearing
53
distinguished from the epistemic use because it does not entail any
evidence supporting a proposition, any “seeming” to S as if things are a
certain way. X looking like Y or like an F to S does not entail at all that it
looks to S as if X is Y or an F. This drawn shape looks like a horse to S,
but in no way does it look as if it is a horse. Appreciating the similarity of
the way X looks to the way F looks is by no means the same as being
inclined to believe that X is an F, nor is it the same as taking X’s look as
eventually supporting the proposition that X is F. So comparative looking
has essentially nothing to do with evidence (even though, obviously, a
similarity between the looks exhibited by two objects may be used as
evidence for propositions concerning the respective things). A relevant
feature of the comparative use is that, when the ascriptive scheme involves
an object and a sortal term as in “a looks like an F to S,” ascribing a
comparative looking-like experience to a subject entails ascribing her the
general capacity to recognize Fs just by looking; i.e., to have both
familiarity with F’s sensible profile and the recognitional capacity
concerning examples of F as such. If a can look like a table to me, then
generally I must be able to recognize tables just by looking, I must be
familiar with tables, I must know what they are (where “knowing” is to be
meant loosely, as not necessarily involving the possession of the
respective concept but only the recognitional category. See the seeing-as
part above).
Both the epistemic and the comparative use of “look” have to do with
visual appearances as experienced by a subject. The epistemic use
involves a certain body of evidence being visually acquired so it implicitly
ascribes a visual experience to a subject. The comparative use, by
comparing ways of looking, presupposes the subject’s visual sensitivity to
each of both ways of looking, which are compared. For example, with the
way a looks as grasped as being the same as the way Fs look.20 In order to
the way things are. I am not disclosing a fact about myself but about petrol when I
say that petrol looks like water” (Austin 1962, 64). To use Searle’s expression, the
reciprocal property of looking the same (characterizing two things) is not
“ontologically subjective” (Searle 1992). If it presupposes ontologically subjective
properties characterizing the experience of a subject who appreciates such a
looking-like relation, this has to be argued on independent bases. Looking-like
relations are objective relations appreciated by subjects through their experiences.
20 On pain of infinite regress, “X looks like (an) F” must presuppose a noncomparative grasp of the general way Fs look. If this way were specified by F
looking like Gs, an implicit reference to the way Gs look would be made thereby.
Now, if G looking such-and-such were also comparatively specified, then a
foothold for this last comparison should be pointed, and so on. We need to stop at
54
Chapter One Part V
visually appreciate a looking-like relation, I must be able to visually
appreciate the involved relata in the first place.21 Indeed, looking like an F
is looking the way an F looks, so the comparative use is explained using
“looks F.” If the comparative use were not grounded on another noncomparative sense of “looks,” this explanation would be patently circular
and involve a regress.22
V.4 Is There a Phenomenological “Looks”?
It seems that both these uses presuppose a third, more fundamental use of
“looks” that is neither the epistemic one nor the comparative one but must
be implicitly involved in them. Chisholm negatively calls it “noncomparative” while Jackson positively labels it “phenomenal.” Such a use
refers to the way or to a way something visually looks, which can ground a
looking-like claim (a looks like an F because both a and Fs look that way)
as well as an epistemic use (a’ s looking this way may be evidence that a
is this or that way). How are we to specify the way something visually
looks without appealing to what it looks like or to what it seems prima
facie to be? By describing a sensible profile as an arranged and unified
complex of SCM-properties. For example, if a looks like a pig, it must
look the way pigs typically look. So how do pigs typically look?
Excluding a circular comparative appeal to some other type of things pigs
could look like, we are to individuate some typical “piggish” visible
features, a complex pattern of SCM-properties exhibited by pigs in
standard conditions: colors, shape, movements and patterns of movements,
size, shape-details, and so on. Such a complex visual appearance would
constitute a complex look decomposable into elements like: it looks [pink],
some point and refer to the way things of a certain kind generally look,
independently of them looking like something else.
21 By pointing a looking-like relation, either we can compare the typical way Fs
look with the typical way Gs look, or we can compare the particular way that F
looks now in the actual circumstance with the typical way Gs look, or we can
compare the typical way Fs look in certain circumstances with the typical way Gs
look in other circumstances.
22 The remark is made by Byrne 2011, 72. Travis 2004 complains that looking-like
claims are normally silent on what way both the relata just look. “Pia looks like
her sister”: well, what way, how, do Pia and her sister look? He takes that as an
argument against the idea that such lookings cannot determine the content; but this
is not our concern here. To take in Travis’s suggestion about looking-like, we need
to individuate another non-comparative use and sense of “look” to make sense of
the comparative use. Travis does not recognize any other non-comparative use
apart from the one Chisholm and Jackson call the epistemic use.
Looking, Seeming, Appearing
55
it looks [round (in a certain manner)], it looks [with this and that shape]
and so on.23 Recognizing a pig just by sight presupposes being able to be
visually aware of these look-properties as well as of their unified complex
as a global sensible profile, and appreciate them as potential presentations
of pigs.
Such a “phenomenological” use of looks24 therefore would concern
strictly visual properties and sensible profiles. According to Jackson,25 the
phenomenal use—this is what he calls it—is limited to properties like
“red,” “triangular,” “moving,” and the like, so that “it looks blue” can be
phenomenal but “it looks old” cannot.26 Moreover, he argues that “looks”
has such a special meaning when (and only when) it is followed by what I
have been calling SMC-predicates, namely, terms for colors, shapes, sizes,
distances and the like. Why is there a need for introducing the third use? Is
its application really limited to SCM-properties? What Jackson calls
“phenomenal use” is the one involved in cases like the following:
1) There is a sense in which a tilted coin looks elliptical to S even if it seems
round to S.
2) There is a sense in which a red surface at night looks brown even if it seems
red to S.
23 The verbal description of F’s general sensible profile may not be as fine-grained
as needed for individuation of that sensible profile. This is not the issue at stake
here, though. I am talking about distinctive visual profiles whose indication could
eventually need to be partially demonstrative. For example, “the round shape
typical of pigs” would be circular as a part of the description of a general way pigs
look, but it could make sense as a demonstrative way to pick out a certain peculiar
round shape in order to be understood by one who is familiar with pigs and their
visual recognition. What matters here are visual properties we are sensitive to, not
our verbal description of them.
24 While Jackson calls it “phenomenal,” I prefer the looser characterization of it as
“phenomenological” because Jackson’s positive account of that use takes it as
involving mental objects that strongly resemble the more classical and notorious
sense data. But just recognizing the existence of a phenomenological use in the
ordinary discourse is still neutral on the existence of intrinsically subjective
phenomenal properties having a suspicious ontology like sense data. See below.
25 See Jackson 1977, 77ff.
26 Also Tye (2000, 54) agrees with Jackson that phenomenal looks talk involves
locutions of the form “X looks F to S” where “F” must express a sensory property.
Other philosophers, who hold a) that high-level properties like [dog] or [lemon]
can be represented in visual perception, and b) that the phenomenology of
experience supervenes on represented properties, do not limit the phenomenal
look-ascriptions to strictly visual properties.
56
Chapter One Part V
3) There is a sense in which the blade of grass before S’s eyes looks bigger to
S than a huge tree in the distance (which is the same sense in which the
Moon looks no bigger than a sixpence).
4) There is a sense in which a straight stick in water looks bent, even if it
seems straight and not bent.
Now, these cases do not hint at all at a body of visually acquired
evidence for a proposition so the epistemic use is ruled out as a candidate
for interpreting the sentences above. The coin’s looking elliptical in that
context is by no means even prima facie evidence that it is elliptical (at
least it may be evidence that it is round). The surface looking brown in that
case is not evidence for it being brown (at least it is evidence for it being
red). The blade of grass looking bigger than the tree in that circumstance is
not evidence for the proposition [the blade is bigger than the tree]; the bent
appearance of the stick partially immersed in water is not evidence for its
being bent.27 Nor are these essentially comparative cases. If they were, in
any case they would presuppose implicit reference to a look to be picked
out non-comparatively (maybe demonstratively, or otherwise) on pain of
circularity. So, they need to belong to a third irreducible use or sense of
“looking.”28
Perhaps some would not agree that this is a legitimate ordinary use of
“looks”; indeed, some hold that it is actually just a piece of philosophical
jargon. But do we really need any philosophical background at all to
immediately understand 1-4, to understand, for example, that the moon
looks no bigger than a sixpence in a certain sense? I do not think we do.
On the contrary, we talk this way quite naturally. Provided we do, it has to
be made explicit what we do when we do so. According to Jackson, in
doing so we implicitly quantify over mental objects having intrinsically
subjective phenomenal properties, namely, mental items that are
analogous to the classical sense data.29 But I think it is quite possible to
27 Jackson’s proposal of a “phenomenal use” limited to SCM-properties does not
mean that any look-ascription followed by SCM-properties must be phenomenal.
An SCM-look ascription may well be epistemic. An object may be said to look red
because there is prima facie evidence, in certain circumstances of (non-optimal)
observation, that it is red, so it looks to be red. It is important to remark that an
object can keep its “purely visual” epistemic look (that size, that shape, that color)
and change its “non purely visual” epistemic look: a DVD can look intriguing and
then unappealing, because of a change of interest in the subject, without changing
its visual looks. The example is by Price 2011.
28 A similar distinction, between “intentional look,” public and referential, and
“qualitative look,” private and subjective, is to be found in Block 1990.
29 See Broad 1925, Price 1932, Ayer 1940.
Looking, Seeming, Appearing
57
hold that there is a “phenomenological” use of “looks” in ordinary
discourse about experiences and experienced things without accepting that
there is an even implicit commitment to sense-data-like entities as a result.
I call that third use phenomenological because it sounds more neutral than
“phenomenal.” Indeed, its use has to do with the ways a thing can appear
to a perceiver, but this in no way entails that such ways of appearing are
necessarily “mental” or intrinsically subjective.
V.5 Phenomenological “Looks” Is Not Independent
I argue that when we say a tilted coin surface looks elliptical or that the
moon looks no bigger than a sixpence, we are implicitly making the very
sophisticated counterfactual operation of comparing the actual experienced
situation with another possible perceptual situation in which we would be
looking at an elliptical object (or at a sixpence-sized object). Namely, we
compare two objective ways of appearing, one of which is actual, the other
of which is only possible. So, comparing two ways of appearing entails the
capacity of non-comparatively appreciating the way each relatum looks,
but it does not entail at all that the “appearance-property” standing as a
complement of “looks” is intrinsically mental or subjective. These are
sorts of complicated comparative uses that presuppose a non-comparative,
“phenomenological,” but not a phenomenal, use of “looks” in the sense
meant by Jackson. Now I will try to make my point explicit.
Take the sentence “the penny looks elliptical” said in front of a tilted
coin that my experience does not present as being elliptical at all but rather
as being a round object seen from a certain perspective. Now, in saying
that it looks elliptical, I am indirectly referring to the way elliptical things
look. I am referring to the distinctive appearance exhibited by things that
are elliptical, seen in a given circumstance I take as paradigmatic for
fixing a typical way of looking as a mark constituting the visual memory
of that property. Likewise, if I suddenly visualize my brother, I “put before
my mind” his face seen from before, even if I would recognize him in
many other perceptual circumstances. What I am saying is even if the
penny looks the way round things look to one when tilted, it also looks in
a similar way to the way elliptical things look to one when not tilted.
Likewise, the moon looks no bigger than a sixpence because it looks—in a
certain, quite abstract respect—the way a sixpence looks up close. So a red
thing looks brown insofar as it looks the way a brown thing looks in
daylight, which is the paradigmatic circumstance that fixes the visual
memory of “the” way a red thing looks when it looks as it is. Being able to
abstract away the phenomenological similarities between objective looks
58
Chapter One Part V
from the respective perceptual circumstances is a very sophisticated skill.
Indeed, the tilted coin looks to be round because of our visual appreciation
of perspectival properties, the tree looks to be bigger than the blade
because binocularity and focus make us able to appreciate the distance and
the object looks to be red because our visual system is able to keep the
color looking (to be) constant despite illumination-changes but in a really
sophisticated sense (although an ordinary and not just an exclusively
philosophical sense), the tree may be said to look smaller than the blade of
grass, the red object may be said to look brown and so on. When an object
moves away from us, despite the size-constancy (it looks to be a constant
size), still something changes in our visual experience. Indeed it is thanks
to that change that we can perceive it is moving away from us.30 In order
to express the change, we can lean on a counterfactual situation, which
would be in some way similar to our actual one: we can imagine the way a
stationary object would appear when becoming smaller and smaller before
our eyes. This is what “looks” means in such a special “phenomenological”
sense. If we called looking1 the “looking” involved in the epistemic
constructions (= looking to be = visually seeming) and looking2 the
phenomenological one, then we could say that:
X looks2 F if it exhibits the same visual appearance an F would exhibit in a
possible experience of an F seen in paradigmatic circumstances in which it
would look to be an F; i.e., look1 (an) F.
As clumsy as this characterization may sound, I think that it captures
what matters. Such phenomenological use of “looks” is distinct but
nonetheless parasitic on the use of “looks” involved in epistemic
constructions where looking F means looking to be F. The tilted coin does
look2 elliptical without looking to be elliptical because it looks as an
elliptical coin would look to be in certain paradigmatic circumstances
other than the actual ones. What I am rejecting, then, is not the existence
of a phenomenological use of “look” but the fact that it must concern the
ontologically subjective properties (Jackson) referred to in a sort of
parallel phenomenal language. Rather, it concerns publicly assessable
properties that characterize worldly objects, though it implicitly refers, for
30 Peacocke 1983 (p.12-14) discusses the case of seeing two trees of the same size
on the same street. The one that is farther away will look smaller in some sense
even though it does not look to be smaller. On the contrary, in optimal conditions it
looks to be the same size as the nearer one. In short, there is a phenomenal contrast
between the trees, which does not seem to be due to a contrast in which properties
are experienced as characterizing the trees.
Looking, Seeming, Appearing
59
a sophisticated comparison, to a non-actual, counterfactual perceptual
situation in which other circumstances hold and different objects would be
seen.
V.6 “Looks F” Depends On “Is F”
According to this interpretation of the “phenomenological” use of “look,”
“looks F” (looks2) in this special sense is parasitic on “looking to be F”
(looks1).
Now, I argue—on the lines of a famous thesis held by Sellars31—that,
on its own, “looking to be F” is logically dependent on “is F” so that
“being F” is prior to “looking F” even for F = SCM-properties (like
[green], [square], [round] and the like). If this is the case, not only is
looking-F (looks2) parasitic on looking-to-be-F (looks1) but looking-to-beF on its own is also parasitic on being-F in a strong sense.
Given the sentence “a looks F to S,” with respect to its superficial
grammar we have a triadic relation among [looking] involving an object, a
subject and a property as its arguments. Sense-data theorists analyze such
a relation by appealing to sense data.32 For example, when the (round)
penny looks elliptical to John, provided that nothing elliptical is out there,
the [elliptical]-property would be instantiated by a mental sense datum,
which would be the “real” argument for F. Others would take the
[looking]-relation as primitive and therefore not analyzable.33 Without
entering into this substantial debate, I will take advantage of some
illuminating remarks by Sellars to show that we do not need to introduce
any intrinsically subjective entity to either explain or analyze the
ordinarily ascribed looking-relation “X looks F to S.”34 The core idea is
31 See Sellars 1953, §10-23.
32 Other sense-data theorists, like Broad 1925, introduce sense data not to analyze
the relation above but to explain it in cases when, as in the example above, there is
nothing elliptical. Also Jackson 1977 and Robinson 1994 accept sense-data-like
mental objects as the only possible explanation of perceptual illusions. For the time
being, we do not need to enter into this substantial debate here because now we are
concerned with the ordinary discourse.
33 The so-called theory of appearing, originally proposed by Chisholm 1950 and
then developed by Alston. See Alston 1988, 1999, 2005.
34 Of course, there may be many other substantive reasons for introducing sense
data that are independent from trying to make sense of the ordinary discourse
insofar as it involves looking-expressions and the like. Again, I am not interested
in arguing sense-data theories as such; rather I am pointing out that we can make
perfect sense of ordinary looking-sentences without taking that such sentences
(rightly or wrongly) implicitly quantify over sense data or analog mental entities.
60
Chapter One Part V
taking reports like “a looks F to me” as saying less than “I see that a is F”
but as implicitly referring to this positive case, namely, as characterizing
one’s visual experience as being just like one in which one sees that a is
F—or a as being F just as it is—but at the same time without endorsing
that content [a is F]. If “a looks F” means “a is F according to my visual
experience,” the ascription of something looking red to one is the
ascription of an experience individuated by its being as though one were
seeing that something is red, just as when seeing something as it is.35 So,
all that looking red involves is the ability to recognize red things just by
looking, plus the possibility of talking of one’s experience as withholding
its authority about how things are according to it. As Sellars says,
x looks red to S has the sense of ‘S has an experience which involves in a
unique way the idea that that x is red and involves it in such a way that if
this idea were true—and if S knew that the circumstances were normal—
the experience would be correctly characterized as a seeing that x is red.’36
The logical priority of “is red” over “looks red” entails that the
perceptual standard conditions for recognizing something as red—which
must be known by one who masters the [red] concept—are conditions in
which things look as they are (= red), conditions that give content to the
very concept of being red. So, instead of analyzing [red] as “what looks
red in standard conditions,” the very standard conditions for something
being recognized by one as being red also fix what it means to just look
red to one without being such. If a just looks red to one, one’s experience
is as if one were experiencing a red thing, without there being such an
experience (no red thing is experienced indeed).
Sellars illustrates that priority of “is F” over “looks F” through a
fictional story,37 which I will now re-tell and sum up a bit freely. Suppose
John, the owner of a necktie shop, has only experienced colors in standard
conditions in daylight and has learned the respective concepts that way, so
that when he says “This is green, that is red,” people nod and approve.
Electric light is invented, but John resists the innovation for a while and
goes without it. As soon as he installs a light bulb in his shop, the troubles
starts. When he says “That necktie is green,” a customer disagrees by
saying “No, it is blue, come out here and see.” John replies “Now it is
blue, inside it was green though,” and so on. First, he would realize
35 Sellars’ account of “looking” is actually the first appearance of a disjunctive
theory of experience, whose paternity is generally ascribed to Hinton 1967.
36 Sellars 1953, §22.
37 See Sellars 1953, §14-17.
Looking, Seeming, Appearing
61
something like this: “Well, it is blue, in fact—I rule out that electric light
suddenly changes the color of my neckties but still it is as though I were
seeing a green necktie when I looked at it here in my shop.” After some
time he would learn to make another kind of report, like “It looks green in
here, but it is blue.”
He has learned that things can look different from the way they are
because of variations in perceptual conditions, so the experience of an F in
special conditions can be just like the experience of a G in normal
conditions. He also learns to characterize the visual experience as of
seeing that a is G—when knowing that a is F—as “looking-G.” In Sellars’
terms, he can now ascribe to his experience a certain propositional claim
(e.g., that is G) without endorsing it. Of course, by learning that a green
thing can change its look according to the circumstances, he is still
learning the concept of [being green], because before that awareness, his
very concept [of green] was quite rudimentary. At some point, the fact that
looking at the blue necktie in the shop is similar to seeing a green necktie
will become so obvious that John would say that it “Looks green” only in
a very sophisticated sense, while he could more plausibly say that it looks
blue in the first place. It actually looks the way blue things look in here; it
begins to look as it is as soon as its variations are taken as the normal
behavior of blue things when displaced.
In any case, “something merely looking green to S” is defined, as an
experience, by reference of “S seeing that something is green” through the
subtraction of the endorsement conveyed by the latter. Now, John can
become familiar with the general fact that blue things look this or that way
in this or that circumstance, where those “ways” are fixed by paradigmatic
experiences of other properties in normal conditions. For example, a
yellow thing seen in the dark looks a bit like a green thing seen in perfect
illumination under the sun. Learning that that is the way yellow things
look in the dark is learning both the [yellow]-concept and a use of “looks”
that is no longer bound to certain specific circumstances. So, even though
something can look some way (F) while it does not look to be that way, it
can look F (green, elliptical) because it looks the way F-things look to be
in paradigmatic circumstances. On the one hand, it can look to be a way
(yellow, round) because the experience of it is an experience as of seeing
that it is that way (where “as of” is neutral about whether it is such an
experience or not).
To sum up this last point, the concept and the understanding of
“looking F” depend on the concept and the understanding of “being F”
even for F = SCM-properties, not the other way round. When “looks F” is
used in a phenomenological sense (as not entailing that something is F, at
62
Chapter One Part V
least according to one’s visual experience), it means that something’s visual
appearance in one respect is just like the one F-things have when seen in
some paradigmatic conditions. In turn, the way F-things look in normal
conditions is the way they are, so that “looking F” means “looking for one
respect as F-things look when they look as they are” (phenomenological
sense), “looking to be F (epistemic sense),” or “looking the way F-things
look” (comparative sense). The phenomenological sense is also interpretable
as a sort of sophisticated species of the comparative use, and both the
comparative and the phenomenological uses presuppose the epistemic one.
I can grasp what it is for something to look like an F or to look2-F (like
“looks elliptical” said about something that looks to be round) only if I can
grasp what it is for something to be F, and at the same time I am able to
visually recognize something as being F. On the other hand, the epistemic
use presupposes the ability to recognize Fs (say, elliptical things or red
things) just by looking, an ability involving a certain visual
phenomenology. That visual phenomenology is actually what is involved
in looking-talk as such, provided that “it looks F to me” in its more
fundamental sense means “my visual experience in some relevant respects
is (phenomenologically) as if I were experiencing an F.” So, the three uses
are deeply and reciprocally interwoven but they are not to be confused
with each other. The phenomenological sense of “look,” importantly, does
not entail a commitment to intrinsically subjective properties (like sense
data or “phenomenal colors” or the like); it only entails the ability to
recognize certain properties just by looking at the objects having them,
which in turn entails that experiences do have a visual phenomenology,
making it possible that things look some way to the perceiver.
V.7 To Sum Up
“Looking”-, “appearing”- and “seeming”- constructions have been
analyzed with respect to their similarities and differences. If “appearing”
ascriptions usually hint at special circumstances and “seeming” ascriptions
have mostly to do with non-conclusive evidence, neither necessarily has to
do with visual episodes, as is the case of “looking”-ascriptions. For that
reason, “looking”-constructions are what I have specifically focused on.
At least three distinctive, irreducible uses of “looks” can be individuated:
an epistemic use having do to with visual acquired evidence supporting a
proposition (like “seeming”), a comparative use having to do with
similarities in ways different (kinds of) things distinctively appear in their
sensible profiles, and a more controversial phenomenological use, which is
involved in the other two and interwoven with them. The last use appears
Looking, Seeming, Appearing
63
to be irreducible, especially when its complement are properties, say, F or
G—things are not taken to objectively possess even though they may be
said to look F or G (this round coin looks elliptical, that yellow shirt looks
green now). Nevertheless, this use does not involve anything more than
those publicly assessable properties that may objectively characterize seen
things. Indeed, looking-F in such a peculiar sense means looking the way
F-things look to be in other paradigmatic circumstances. In addition, I
have argued that “looks F” in that last more natural sense (a looks to be F)
is on its own conceptually dependent on “is F.”
Another very important feature of “looks F” is that when F is not a
spatial-chromatic-morphological property (as [square], [bigger-than],
[red], [round], and the like), in order for a to visually look F, a must
exhibit a certain distinctive sensible profile made of SCM-properties in the
first place, properties the subject must be generally sensitive to.
Concluding Remarks
In this first chapter, I have taken into consideration the ordinary basic
vocabulary for perceptual ascriptions concerning vision. Namely, I have
analyzed the most relevant terms and constructions involved in ascriptions
and self-ascriptions of episodes of visual perception. Seeing-ascriptions, in
their different constructions, on the one hand, and the related experiential
terms like “looking,” “seeming,” “appearing,” on the other, have been
analyzed with respect to their logical behavior at certain length, with
particular attention to their reciprocal relations.
In Part I, I briefly argued for the methodological necessity and
legitimacy of starting a substantive philosophical inquiry on perception
from the analysis of ordinary language and pre-theoretical discourse.
Making our intuitions explicit about what seeing, for example, is
implicitly taken to be according to the way it is ascribed in everyday
discourse is a way of individuating the explanandum in the first place—
what we are interested in when we ask, for example, “what is seeing?” If
the very question is to make any sense to us before we come to know the
answer, then we must have partial, rough and general access to the
inquired object. Even though making such pre-theoretical access explicit
does not answer any substantive question (like the one above), still it is a
way of starting the inquiry by bringing to awareness what we already take
the inquired object to be insofar as we show when talking to and
understanding each other. In talking about someone seeing something, we
show what we take seeing to be, maybe mostly in a non-thematic form.
We could even be systematically wrong, but only a positive theory could
64
Chapter One Part V
establish it with substantive arguments. Nonetheless, by waiting for a
theory to stand such a “manifest image” on its head, it seems
philosophically wise to carefully consider that manifest image in the first
place. Even if it is not normative in a strong sense, nevertheless the
capacity of saving and explaining our basic intuitions on the matter would
be at least an advantage for a philosophical theory of visual perception.
In Part II, I took into consideration ordinary contexts of object-seeing
ascriptions. Firstly, I spelled out a list of necessary and sufficient
conditions for seeing-something to be truly ascribed, such that seeing-X
has been defined as: a certain episode consisting of a discriminationrelation between a perceiver and an environmental object, where the
object is causally responsible for its being discriminated through
appropriately impacting on the subject’s visual apparatus in such a way
that the object looks to S a certain way. Second, I argued that seeing-X is
implicative, that it is a success verb, that seeing-X ascriptions generate
transparent and extensional contexts. I also argued that, even if what you
see is a matter of what you are causally connected to in your surroundings,
object-seeing has some minimal phenomenological constraints to the
effect that there must be a way O visually looks to you if you are to see O.
In Part III, I examined seeing-that ascriptions and their logical
behavior. As a result, I concluded that seeing-that is a fully-fledged
propositional attitude that gives rise to opaque contexts (where substitution
fails). I also argued that seeing-that is conceptual (i.e., you need to possess
the concepts of F if you are to see that something is F), and factive (if you
see that P, then P). All these features of seeing-that are best explained by
taking the meaning of ordinary instances of seeing-that as equivalent to
“coming to know by visual means.” As a consequence, I have provided a
definition according to which S sees that P if: “thanks to becoming visually
aware of those objects, properties and relations which are the constituent of
the environmental fact which makes P true, by conceptualizing those
constituents and by grasping the structure of that very fact, S comes to
know that P is the case.” This definition is meant to capture only those
cases of true ascription where “seeing-that” is neither metaphorical—
knowing or coming to know that P by whatever means—nor referred to in
inferential knowledge only remotely dependent on a visual episode.
In Part IV, I moved on to examine seeing-as ascriptions. I showed that
seeing-as is peculiar and irreducible to seeing-O or seeing-that, even if it
entails the first and it is entailed by the second. I also argued that seeing-as
ascriptions give rise to non-transparent contexts, and that this logical
feature is best explained by seeing-as being cognitive, categorical and
recognitional. It entails a cognitive stand the perceiver takes toward a seen
Looking, Seeming, Appearing
65
object, although such a stand or “taking” need not be either conceptual or
epistemic-propositional. Seeing-as is implicative, like object-seeing, and it
is not factive, unlike seeing-that. Being both cognitive and non-factive,
seeing-as can be right or wrong, namely, its ascription makes room for a
normative evaluation involving the possibility of mistake: seeing-as can be
mistaken. When “seeing an a as F” is not metaphorically ascribed (as
equivalent to considering-as), it necessarily involves the visual
appreciation of a sensible profile, namely, of a unified complex of spatialchromatic-morphological properties (I called them SCM-properties)
typically exhibited by Fs, where S possesses the general capacity of
recognizing Fs just by looking. To concentrate all these acquisitions, I
have defined seeing-as as follows: “Seeing a as an F is seeing an F in a
thanks to the exercise of a general recognitional capacity prompted by the
visual appreciation of a’s objective sensible profile.”
In Part V, I left seeing-ascriptions behind to focus on experiential
predicates like “seeming,” “appearing” and “looking” insofar as they are
ordinarily used to ascribe or self-ascribe visual experiences. If seeming
normally hints at non-conclusive evidence (“It seems that P”; “X seems
F”; “It seems to S as if P” and the like) and appearing normally involves
an implicit reference to special circumstances, neither of them necessarily
entails a reference to seeing or vision. On the contrary, looking when nonmetaphorically used does involve reference to the visual presentation of a
seen object. For this reason I focused my analysis on looking. First, I
distinguished propositional constructions (“It looks as if P”, “it looks as
though P”) from a de re construction (a looks F, a looks as if it is F) that
involve reference to a seen object looking some way, so I focused on the
latter.
An important distinction I have made is that between SCM-properties
like [square], [yellow], [big], and non-SCM-properties like [old],
[European], [pig] and other “thick” properties. To this end I established
the following principle: when F is not an SCM-property, a must look G in
the first place, where G is a complex of SCM-properties that individuates
a certain visual profile.
Subsequently, by critically relying on the relevant literature, I have
distinguished an epistemic, comparative and phenomenological use of
looks. When used epistemically, as in “it looks as if a is F,” “a looks as if
it is F” means that there is a body of visually acquired knowledge that
non-conclusively supports the proposition [a is F]. So the epistemic use is
an analog to “seems” when used with reference to vision, but nonetheless
it is not as committal as “seems” is. Something can look F to me even if I
know it is not F, but something cannot seem F to me if I know it is not F.
66
Chapter One Part V
The comparative use covers cases like “a looks like b,” “a looks like an F”
and the like. It roughly means that a’s sensible profile is similar to b’s (or
to F’s) sensible profile. If that cow looks like a horse, its visual profile is
similar to the typical visual profile horses exhibit. On pain of an infinite
regress, the comparative use presupposes a non-comparative one (how do
horses look? If they just look like Gs, then how do Gs look? And so on).
The third use, also presupposed by the epistemic use (how to analyze a
“visually acquired body of knowledge”) is what I have called the
phenomenological use. When a looks F to S, a has a certain visual profile,
the appreciation of which by S is a condition for a visual episode counting
as evidence (epistemic) to S and for S to appreciate the similarity between
two visual profiles (comparative). In particular, the phenomenological use
explains how it is possible that, even when F is an SCM-property, a looks
F without looking like an F or without that look being evidence for
claiming a is F. The phenomenological use is in play, for example, when
we say that a tilted coin looks elliptical even if we know just from that
very visual episode that it is round and looks to be round (epistemic), even
if it does not even look like elliptical things (comparative). However, I
have argued that that use does not entail at all a commitment to the
existence of intrinsically subjective properties, least of all to the existence
of mental objects having them. On the contrary, following Sellars among
others, I have argued that “looks F” logically and conceptually depends on
“is F” in such a way that, when we claim that something looks F without
looking to be F—according to the phenomenological use—we are
implicitly positing a counterfactual perceptual situation, different from the
actually experienced one in which things would look to be F. Calling the
phenomenological look “look2”, I have stated that view in the following
principle:
X looks2 F if it exhibits the same visual appearance an F would exhibit in
a possible experience of an F seen in paradigmatic circumstances, in
which it would look to be an F.
Something can look F without being known to be F thereby because it
can look, in a certain respect, the same way something else would look in
looking as it is in a paradigmatic presentation.
So far so good for the preliminary analysis of the basic visual and
experiential terms. Before using these acquisitions to enter into the matter
of perception theory and articulate a positive view (Chapters Three and
Four), I will present and critically discuss the most classical version of the
belief theory in the next chapter.
CHAPTER TWO
SOME BASIC FEATURES OF PERCEPTUAL
EXPERIENCE
Introduction
This chapter is divided into two parts.
Part I presents the classic belief theory (BT) of perception and
discusses at length some of its virtues and the many problems it faces. Part
II raises some problems with BT: I show that BT is incompatible with
experimental data about vision and visual experience so it must be either
deeply revised or abandoned.
PART I
THE BELIEF THEORY:
PHILOSOPHICAL ARGUMENTS
FOR AND AGAINST
I.1. The Belief Theory (BT)
We cognitively relate to our surroundings through our perceptual
experiences in such a way that we come to know how our environment is
arranged, which properties are exemplified in it, which objects are present,
what relations such objects and properties entertain with each other and
which events occur that involve those objects, properties and relations. It
is by means of our perceptual experiences that we keep in constant touch
with the physical world surrounding us: we then act upon and judge based
on our perceptions. Indeed, the exercise of our perceptual capacities often
and immediately gives rise to true beliefs about the world. For this reason,
a well-known classical view on perception identifies perception with belief
or, more precisely, with episodes of belief-acquiring. Even though
nowadays the so-called belief theory1 is not a mainstream view, presenting
it critically seems a good way to enter into the matter by both contrasting
and demarcating perception from belief, and evaluating the nature of their
intimate connection. The core of the belief theory of perception, in
Armstrong’s words, is:
Veridical perception is the acquiring of true beliefs; sensory illusion is the
acquiring of false beliefs.2
The reason Armstrong talks about belief-acquiring instead of belief
simpliciter is that a belief is a dispositional state that can produce
occurrent events such as judgments. Whilst judging-that-P is a mental
1 See Armstrong 1968, Pitcher 1970, Heil 1983.
2 Armstrong 1968, 208.
The Belief Theory: Philosophical Arguments For and Against
69
event, believing-that-P is not. Even though, on the one hand, acquiring the
respective belief may be an event, on the other, that belief typically
produces mental events with the same content P, namely, correspondent
judgments that P. Once I have acquired a belief and until I envisage good
reasons to revise or reject it, I can ascribe that belief as a permanent
disposition on which to make judgments and inferences and take actions
consistent with it. Judgments, actions and inferential reasoning can be
occurrent manifestations, or expressions, of related doxastic dispositions.
Not only do such states produce events but also may well be produced by
events. Indeed, certain empirical beliefs are produced by the occurrence of
perceptions.
So, if perceptual experiences are by their very nature events or
episodes,3 they cannot be identical with states or dispositions, as beliefs
are, rather they are to be identified with, the episodes of acquiring certain
empirical beliefs. Of course, there are acquisitions of empirical beliefs by
means of senses, not by whatever means. A visual experience will be an
eye-dependent belief-acquiring, for example.
I.2 Virtues of the Belief Theory
Prima facie, such a view has some important advantages. First, it clearly
explains the intentionality of perception by appealing to the intentionality
of beliefs. As Armstrong points out:
The intentionality of perception reduces to the intentionality of the belief
acquired. (1968, 210)
This is a belief that P has the propositional content P. A belief-state is a
state with content, so the acquiring of such a state with content P can be
taken as an episode having the same intentional content characterizing the
so-acquired state. In that view, intentionality is grounded in content,
provided representing something as being some way is, for a state, a way
3 The first person who stressed that perceptual experiences are events rather than
states was Ryle (1949). Indeed, in his view, perceptual verbs are achievementverbs, namely, success-verbs denoting the occurrence of a cognitive
accomplishment. Even better would be to take perceptions as episodes, because an
episode can last a while through time as a perception, like staring at a relatively
stable scene for a while, even though the event/episode distinction is blurred and
vague. Compare this with the state/event-or-episode distinction, which is much
more important and clear-cut.
70
Chapter Two Part I
to be about something.4 If beliefs are intentional states insofar as they are
representations, so will be their acquisition as perceptions.
Second, such a view elegantly explains how it is that having a certain
perceptual experience so naturally involves forming a certain perceptual
belief about the world. In BT, the very experience consists of the
acquisition of the respective belief so no special transition is needed.
Third, BT explains the intuition that we can be right or wrong in
perceiving the world, where being wrong amounts to being in a false state,
and being right amounts to being in a true state. Perceptual experience can
be veridical or falsidical insofar as it can be, respectively, the acquiring of
a true or a false belief. Illusory and delusive perceptions are sensory
acquisitions of false beliefs. Veridical perceptions are (appropriate)
acquisitions of true beliefs, so are episodes of production of empirical
knowledge about the surroundings.
I.3 The Argument from Illusion: the Belief Theory,
the Sense Data Theory and the Adverbialist View
From a historical point of view, BT presented some very important virtues
that avoided the embarrassing consequences of the classical sense datum
theory, which was the mainstream view of perception during the first half
of the past century.5 According to the sense datum theory, what we are
directly conscious of in perception are mental items, intrinsically
subjective entities having the properties they look to have, or being as they
appear so infallibly known by the perceiver through the very act of
experiencing them. Material objects that populate the external world
would otherwise be accessed only indirectly by the perceivers, who would
experience subjective sense data somehow resembling a correspondent
worldly object. Leaving aside the details, among the many problems of
4 The notion of intentionality was first introduced by Brentano (1874) as that
property, characterizing mental states, of being “directed upon” or “about”
something: “In presentation, something is presented, in judgment something is
affirmed or denied, in love loved, in hate hated, in desire desired, and so on”
(Brentano 1874, 88). Having an object for a mental state is the basic way of being
an intentional state. Whilst having content entails having intentionality, it is not
obvious that having intentionality entails having content: object-intentionality, for
example, desiring that object, may be a property of states that do not exhibit
content-intentionality. More on that later.
5 See Moore 1910, Russell 1912, 1922, Broad 1925, Price 1932, Ayer 1940. More
recently, updated versions of the sense data theories have been held by Jackson
1977 and Robinson 1994.
The Belief Theory: Philosophical Arguments For and Against
71
BT, the most relevant is the extravagant ontology attributed to sense data.
Non-physical, subject-dependent entities—so their position is a priori
incompatible with a physicalist view of mind and nature—can be
indeterminate insofar as experiences can be such6 and even contradictory
insofar as experience can be such.7 Another quite interesting objection
made by Armstrong is that such mental objects should be so peculiar that
the relation of “being identical in a respect” cannot be transitive—in that A
can be exactly identical with B in respect to property F, and B can be
exactly identical with C in respect to F without A being exactly identical
with C in respect to F thereby. Suppose I cannot distinguish A’s color
from B’s color or B’s from C’s but I can distinguish A’s from C’s color. If
sense data’s ex hypothesi are as they appear and are incorrigibly known,
subjective indistinguishability entails identity. However, it can
paradoxically happen that A is identical with B in a respect (e.g., color), B
is identical with C in that very respect, but A is different from C!
Something must be wrong with that.8
Within the belief view of perception, such difficulties do not seem to
arise. First, there is nothing surprising in the indeterminacy of beliefcontents since I can well perceptually acquire the belief that the hen has a
large number of speckles without having to believe that it has a certain Nnumber of them. Second, if I look at A and B and acquire the belief they
are the same color, and then I look at B and C and acquire the belief they
are the same color, I acquire two beliefs that turn out to be incompatible or
false only once I have looked at A and C at the same time. That last belief
about their difference I acquired makes me revise (at least one of) the two
6 For example, I can visually experience a speckled hen without being visually
aware of a determinate number of speckles the hen allegedly presents in my
perception. If the “mental hen” is a sense datum, which by definition is as it
appears, it must be an indeterminate object having an indeterminate number of
speckles. On that, see Chisholm 1942. Sanford 1981 argues that sense data, if they
exist, should have contradictory properties.
7 For a discussion of such problems, see Armstrong himself, 1968, and Pautz
2007. For example, the waterfall illusion is such that the waterfall looks to be both
moving and not moving at the same time. On that case and its relevance for the
theory of perception, see Crane 1988. Another example of contradictory perceptual
contents is given by impossible figures (Escher’s or Reutersvärd’s drawings, to
give two well-known examples). On that, see Crane 1992.
8 Russell accepted this consequence and held that “identity in a respect” is not a
transitive relation. It is so intuitively obvious that if A and B share the same
property and B and C share the very same property that property must be shared
also by A and C, that denying it inevitably looks an ad hoc and unsatisfactory
solution.
72
Chapter Two Part I
previously acquired beliefs. So, “identity in a respect” is a transitive
relation but my experience can lead to me wrongly acquiring the belief
that certain things are identical in one respect whilst another perception
can correct the previous acquisitions just insofar as it is another, more
authoritative, act of acquiring.
In addition, BT purports to remove the very ground on which the sense
datum theory has been proposed as the only possible “solution”, namely,
the so-called argument from illusion and the analog argument from
hallucination.9 Both arguments have often been provided as inescapable
cases against direct realism, which is, in short, the ordinary and intuitive
idea that in perception we are only directly aware of the perceived world.
The first (AI), in a plausible and summarized reconstruction, runs as
follows: in perceptual illusion, something appears F without being F
(intuitive definition of illusion); whenever something perceptually appears
F to S, given that the property F is somehow presented, S is aware of
something that has the property F (sometimes called “phenomenal
principle”);10 but ex hypothesi the perceived object is not F so S must be
aware of something else (Leibniz’s law). We therefore need to posit a
mental sense datum, which is it is F that S is aware of (inference to the
best explanation); but if illusions involve awareness of sense data and are
potentially indistinguishable from veridical experiences, then we are
always aware of sense data in experience (generalization).11
The second argument (AH) is analog to the first but slightly different.
For any veridical experience, there could be a subjectively indistinguishable
hallucination (intuitive assumption).12 In hallucination, one is aware of
9 For a more detailed and critical analysis of those well-known arguments, see
Austin 1962, Robinson 1994, Dancy 1995, Smith 2002.
10 The phenomenal principle has been variously formulated by sense data
theorists. Robinson (1994, 31) renders it as follows: “If there sensibly appears to a
subject to be something which possess a particular sensible quality, then there is
something of which the subject is aware which possess that sensible quality.”
11 The argument for illusion is presented by Moore 1910, Broad 1925, Price 1932,
Ayer 1940.
12 An illusion is an experience where some object in the world is perceived but the
experience presents it as being a way it is not. A hallucination is an experience
where something is “presented” that does not exist at all in the world. For example,
seeing a green lemon as being yellow is an illusion; experiencing seeing a yellow
lemon when there is no lemon (neither green nor yellow) to be seen is a
hallucination. This intuitive distinction is more philosophical than scientific. For
example, the “Hermann-Grid” illusion—looking at a grid of white lines on a black
background—you “see” grey spots at each intersection between the lines but there
The Belief Theory: Philosophical Arguments For and Against
73
something; for example, of an object experienced as being such and such
(phenomenological evidence). The object you are aware of does not exist
ex hypothesi so it must be a sense datum (inference to the best
explanation). But if hallucinations are subjectively indistinguishable from
veridical experience, then in experience we are never aware of worldly
objects but we are always aware of sense data (generalization). So, given
that these arguments seem to make the option of direct realism
impossible—in illusion and in hallucination, by definition, you cannot be
aware (just) of the external world—sense data need to be posited as the
only direct objects of perceptual awareness.
I have briefly introduced these very controversial arguments only to
stress a purported virtue of BT without meaning at all that BT is the only
alternative in town to the sense datum theory as a “solution” for these
arguments or that these arguments are as conclusive and knockdown
against direct realism as they are historically purported to be. In any case,
if perceptual experience is belief-acquiring, what you are aware of in
illusions and hallucinations are the contents of your acquired false beliefs.
For example, you can well be aware that you are experiencing something
blue even if you are perceiving something yellow (illusion) or you are not
perceiving anything (hallucination). The content of your beliefs is nothing
non-physical; it is just the way you come to believe the surrounding world
is arranged.
Another traditional attempt to tackle the illusion argument without
accepting the dubious ontology of sense data comes from the adverbialist
view. Adverbialists (Chisholm 1966, Ducasse 1942)13 reject the very actobject analysis of perceptual experience, so they claim that when you
visually hallucinate a red tomato, you are not aware of a sense datum; that
is, of a non-physical particular having the visible features of a red tomato.
Instead of being aware of a red and circular mental item—as it could seem
at first sight by trusting the superficial grammar of experienceascriptions—[red] and [circular] are ways of S’s experiencing rather than
objects of S’s experience, so S visually experiences “circularly” and
“redly” rather than experiencing something circular and red in the same
way someone can be said to dance a merry dance just to mean that he
dances merrily. When you experience (as of) a red circular object, in fact,
[red] and [circular] are ways of your experience being modified.
are no grey spots at all. In the standard jargon philosophy of perception, that is a
hallucination.
13 The adverbialist view was defended by Sellars (1967). For a more recent
defense of the adverbialist view, see Tye 1984. For persuasive criticisms of the
view, see Lycan 1987 and Butchvarov 1980.
74
Chapter Two Part I
So, there is no need to hypostatize alleged mental objects of awareness
to justify the possibility of hallucinations and illusions. Without entering
into further details here, I want to point out the two main weaknesses this
presumed solution has been conveniently charged with. First, there is a big
problem with phenomenological adequacy.14 When we introspect our
experiences, we seem to be aware of objects and properties presented to us
but we do not seem to be aware of ourselves being modified or affected in
certain ways. On the contrary, in that view, the ways of S being affected
by a given experience and the ways in which things appear to be to S in
virtue of having that experience are neatly separated even if the second is
an effect of the first. As Martin points out, “[...] we cannot separate our
knowledge of what it is like to be in a state from knowledge of the subject
matter presented to one in being in such a state of mind […] to know what
such experience is like is in part to know how things are presented to one
as being” (2008 II, 33).15 Experience appears to have a “subject matter”
we can attend to, and the adverbialist view is inadequate insofar as it does
not do justice to that immediate phenomenology of experiencing.16 An
14 On the phenomenological inadequacy of adverbialism, see Maund 1994, 200ff.
and Martin 2008.
15 This is not the so-called transparency of experience we will deal with later on.
Transparency or “diaphanousness” (see Moore 1922, 25ff.) is that property of
experience in virtue of which, if we try to describe the intrinsic properties of our
experience, we end up describing the way the experience presents the objective
world as being. But whilst transparency is denied by sense data theories, that less
committal phenomenological observation is compatible with the sense data theory
(though not with adverbialism). Also, sense data are “topics” for the experience
presented as being certain ways. Those topics are taken to be mental though, but
the Act-Object- or “Topic”-principle, which captures our immediate
phenomenology, is respected both by sense data theories and BT but not by
adverbialism. Differently, transparency, which also seems to capture our
phenomenology, is not respected by either sense data theories or adverbialism but
it is respected by BT. For BT, such topics concerned in and by experience are
contents rather than objects (even if contents can well be object-dependent, as we
will see later).
16 Martin 2008 (II), 30ff., also points out that another relevant feature of our
perceptual phenomenology is incompatible with adverbialism: in introspecting our
experiences, we not only come to know the objects and properties we are
purportedly attending to, we may also come to discover how we can experience
those objects, properties and relations; for example, by exploiting chromatic
contrasts, shadows, illumination condition, differences in focus and the like.
Experiencing involves a conscious point of view of something so that introspection
makes explicit also those very elements of experiencing that make possible that
The Belief Theory: Philosophical Arguments For and Against
75
adverbialist could reply that phenomenology is tricky, and that even if the
phenomenology of experience turned out to be incompatible with the
account, so much the worse for the phenomenology. Phenomenology
would turn out to be systematically misleading.
But there is another worry that is not only dependent on
phenomenological constraints. Suppose you see a red cube over a green
circle. The way things appear to you having that experience exhibits a
semantically articulated content. You see a red cube over a green circle
rather than seeing a red circle over a green cube, or a green cube over a red
circle, or a red cube under a green circle, and so on. You do not just see
properties and objects and relations, you experience a certain relation
between certain objects having certain properties—each having their own
properties—where that relation is not symmetrical: the cube is over the
circle and not the other way round; the red thing is over the green thing
and not the other way around, for example. Now, if the purported
properties given in experience E are just ways of E being modified, how is
it that experiencing [redly] plus [greenly] plus [cubically] plus [circularly]
plus [overly] makes me come to be in a cognitive state with such
semantically structured content? In short, a juxtaposed sequence of
adverbs cannot render the articulateness of the information we acquire
through experience.17
An adverbialist could reply that seeing a red cube over a green circle is
a certain complex way of being modified that typically causes a belief that
there is a red cube over a green circle, instead of other beliefs even
involving the same (concepts of) objects, properties and relations. The
reply would only shed light on the inadequacy of the theory itself. Either
the relation between experiences and what they make us take to be the
case is just inscrutable and mysterious or the elements that at least seem to
be given in experience are not reducible to ways of being modified, no
matter how complex these ways are.
With respect to that, BT seems to involve a more promising treatment
of the argument from illusion/hallucination than the adverbialist proposal,
with the advantage that it avoids the commitment to ontologically suspect
entities like sense data. A belief that [there is a red cube over a green
circle] can bear content as structured as propositional content could be. In
point of view on the surrounding world. For further criticism of the adverbialist
view, see Butchvarov 1980 and Lycan 1987.
17 That articulateness of perceptual information is not to be conflated with the
propositional structuredness of perceptual beliefs we will see in the next chapter.
For the time being, it will suffice to point out that perceptual information has such
an articulation that the adverbialist view cannot capture and explain.
76
Chapter Two Part I
addition, the committal nature of belief explains how it is that in
experiencing (as well as in hallucinating) a red cube over a green circle,
we take things to be a certain way insofar as they appear to be that way in
the experience, not only “because of” or “due” to the experience. To
conceive that such experience can be a hallucination/illusion, there is no
need to posit a mental non-physical red-and-cubic item or take the
experience as a complex cluster of ways of being affected, which is never
witnessed in phenomenology. Rather, it is sufficient to take the experience
as the acquiring of a belief whose content—that there is a red cube over a
green circle—may not be exemplified at all (hallucination: there is no cube
and no circle) or partially exemplified (illusion: e.g., there is a cube and a
triangle but they are not respectively red and green, they are not one over
the other, etc.)
To sum up, BT seems to explain some fundamental features of
perceptual experience on the one hand, and it was historically appealing as
an alternative to the sense datum theory as it sidestepped that theory’s
main embarrassing difficulties on the other.
I.4 The Problems of the Belief Theory.
Objections and Possible Replies
Nonetheless, there are other types of evidence and arguments that seem to
jeopardize the identification of perceptual states with states of beliefacquiring. Let us now consider some of these problems.
First, beliefs are propositional attitudes toward structured contents in
such a way that having a certain belief involves the possession and
exercise of those concepts the beliefs are, so to speak, made out of. I
cannot believe that Gs are Fs without possessing the concepts {G} and
{F}.18 So, if perceptions were acquisitions of beliefs and beliefs involve
concept-possession, then only conceptual beings would have perceptual
states. This is still not a knockdown objection because the idea that
perceptual experiences are conceptual states is open to debate, as we will
see later, but it puts a strong constraint on the acceptability of the view to
the effect that BT is true only if perceptual conceptualism is true. In
addition, if BT is true, then either non-linguistic animals do not have
perceptual states and perceptual experiences or we should ascribe to nonlinguistic animals beliefs and concept-possession. In which case, we
should accept a very undemanding idea of what it takes to both possess a
18 To avoid confusion, from now on I will adopt the following notation in the rest
of this work: [F] = the property of being F; {F} = the concept F.
The Belief Theory: Philosophical Arguments For and Against
77
concept and be a believer.19 The first option seems unnecessary and
arbitrary; the second option entails a trivialization of the view because it
identifies perceptual states20 with belief-acquisitions at the cost of
adopting a very poor notion of belief, without explaining what it really is
and how it relates to propositional attitudes like our fully structured
perceptually acquired empirical beliefs.
Second, it seems that providing a case of perception without belief, or
even a case of perception without any acquiring of belief, would be
sufficient to defeat the view: if perceptions were episodes of beliefacquiring, neither perceptions without respective beliefs nor perceptions
without respective acquisitions of belief would be possible. Both the first
and the second case seem quite common however. In the first case, we
often have a perceptual experience as if things were a certain way without
believing they are such. This typically happens with perceptual illusions,
like the Müller-Lyer and Ponzo illusions where two equal lines seem to be
different in length (Figures 1 and 2).
Figure 1 Müller-Lyer Illusion
(from: http://mathworld.wolfram.com/Mueller-LyerIllusion.html)
Figure 2 Ponzo Illusion
(from: : http://mathworld.wolfram.com/PonzosIllusion.html
19 In fact, that entailment is accepted by Armstrong himself who writes: “If
perception is the acquiring of belief then clearly it must involve the possession of
concepts. For to believe that A is B entails possessing the concept A and B. But
since perception can occur in the total absence of the ability to speak, we are
committed to the view that there can be concepts that involve no linguistic ability”
(1968, 210).
20 To respect the standard way of talking, I keep talking of perceptual states, but it
is clear that perceptions are to be considered events or episodes rather than states.
78
Chapter Two Part I
The interesting and well-known phenomenon is that even once we are
told that the lines are equal, the experience of them continues to be as if
they were unequal. That feature of perceptual experience, often called
cognitive impenetrability and allegedly due to the informational encapsulation
of perceptual modules,21 makes perceptual experience insensitive to
collateral knowledge, for example, to the fact that we judge and believe
that the lines are equal because we are told it is so. So, we have a
perception that does not give rise to a corresponding belief whereas the
persistence of the illusion—despite our belief on the matter—entails that
we believe and judge things to be other than the way they are experienced
in perception. Such perceptual illusions appear to make a strong case
against BT.
The second case is also very common. Suppose you know your desk is
black because you see it every working day. You visually experience it but
there is no event of acquiring the belief that the desk is black because you
cannot acquire what you already possess. If there can be a perceptual
experience without any doxastic change or increase, then perception is not
belief-acquiring. It must be something else. According to the first case,
perception cannot be a belief or necessarily involves belief. According to
the second case, ad abundantiam, even if perceptual states could be belief
states, they could not be acquisitions of belief states.
Advocates of BT have replied to such challenges, but their replies are
not fully satisfactory. Firstly, the case of perception without belief is faced
by treating perception as an inclination to believe. For example, we
perceive the lines as looking unequal but we know on independent
grounds that they are equal. There is an inclination to believe they are
unequal but this is successfully contrasted by a more authoritative nonperceptual previous knowledge so the inclination is “held in check by a
stronger belief” (Armstrong 1968, 213) or it is a “suppressed inclination”
(Pitcher 1970, 78).22 However, this reply involves a substantive
modification of the original view to the effect that now perception has
become an inclination to acquire a belief and not the acquiring itself. It is
quite one thing to say that perceptions are belief-acquiring inclinations and
another that perceptions are belief-acquisitions full stop.
21 See Fodor 1983, Pylyshyn 2006, Crane 1992, Palmer 1999.
22 Pitcher (1970) distinguishes three cases: a) Full belief, when an experience just
coincides with acquiring a belief; b) Half-belief, when we suspend the judgment
although we have an inclination to believe certain content; c) Anti-belief, when we
resist the very inclination by virtue of possessing more authoritative collateral
knowledge.
The Belief Theory: Philosophical Arguments For and Against
79
Things get even worse if we consider that sometimes perception does
not even involve any inclination to acquire a belief. For example, one’s
own image in a mirror looks like the bodily presence of a doppelgänger,
but in no way do we find ourselves inclined to believe it is the case. So
things can perceptually appear to be a way we do not believe they are;
things can even perceptually appear a way we are not even prima facie
inclined to believe. Moreover, things (as in the second case above) can
perceptually appear to be a way we already believe they are, so there is
neither acquisition nor inclination to acquire at all.23 Armstrong and
Pitcher counter-reply by an appeal to counterfactual analysis on the
following lines: granted that I do not believe that the lines are unequal, or
that there is a doppelgänger behind the glass, still we can say that had I
not known that I was facing an optical illusion, I would have been at least
inclined to believe that the lines would be unequal; had I not known how
mirrors work in our world and other facts of that kind, I would have been
inclined to believe that I was face to face with a doppelgänger. In the same
vein goes the treatment of perceptions-without-acquisitions. Had I not
already believed that my desk was black, that experience would have been
the (inclination to an) acquiring of the belief that my desk is black.
Armstrong calls such cases “the acquiring of a potential belief” (1968,
215). So the view gets more and more modified as well as less and less
clear as soon as it embeds the objections in itself, so to speak. Perception
is not belief nor is it the acquiring of belief but it is a suppressible
inclination to acquire a belief; then it is characterized as a suppressible
inclination to acquire a potential belief.24 What is stated then if not that
perceptual experience is intimately connected with perceptual beliefs, and
in normal conditions gives rise to beliefs in systems capable of believing?
In addition, why are such episodes inclinations to believe one or another
content? What does it take for a perceptual experience to predispose me to
acquire the belief that P rather than the belief that Q? There is a difference
between me being told that my desk is black, and me seeing that my desk
is black. Now, if a theory of perception is meant to capture what
perception essentially is, and perceiving my desk is only one of the infinite
23 Once the modification from “belief-acquiring” to “inclination to believe” is
accepted that objection could be accommodated: if you already believe that your
desk is black, still your actual perception contributes to reinforce your belief state
as being an inclination to believe what you already believe. Even if you cannot
acquire a belief you already have, still you can be inclined to believe—maybe
uselessly—a content you already believe.
24 As Smith (2001) rightly remarks, a potential belief is no more a belief than a
potential chair is a chair.
80
Chapter Two Part I
ways I can come to believe that it is black, the peculiarity of that special
way of coming to believe that a is F perceptually, rather than otherwise,
should be embedded in the general characterization provided by the
theory. In other words, even granted that perception is a special way of
prima facie coming to believe certain empirical propositions about our
surroundings, what is it that makes that ‘way’ special? What is it that
makes such episodes what they are beyond just a way of being inclined to
believe that something is the case? Perceptual means—namely,
involvement of sense-organs—is too poor an additional characterization to
enucleate the distinctive way our perceptual experiences enrich our
doxastic lives.25
I.5 The Belief Theory and the Phenomenological
Adequacy Constraint
So, BT does not seem to meet what we may call the constraint of
phenomenological adequacy, something that a good theory of perception
should meet. As we will see more in detail, perceptual experience exhibits
a conscious phenomenology, and that phenomenology is relevant both for
individuating the very perceptual episode exhibiting it and explaining how
it is that that experience gives rise to one belief over another. In particular,
perceptual experience exhibits a phenomenology of bodily presence of the
environment, as Husserl puts it,26 which is neither explained nor captured
by the reductive analysis provided by belief theorists.27 Just taking
25 A variant of BT is the judgment theory articulated by Craig 1976, according to
which perceptions are to be equated to judgments rather than to beliefs. Since the
notion of judgment must be defined in terms of belief—can you judge that P
without believing or coming-to-believe that P?—the criticism of BT seems to me
to be as challenging as that more recent variant of the view.
26 See Husserl 1910 and 2006. On the phenomenology of bodily presence, which
has called for an explanation in perceptual theory, see Smith 2001, 2002. Moved
by ontological worries, Armstrong unsatisfactorily reduces perceptual
phenomenology to a doxastic element: “The content of our perceptions, which so
many philosophers want to turn into a non-physical object, is simply the content of
the beliefs involved (1968, 215).”
27 It is worth pointing out that by “belief theory” we mean here the reductive
analysis that just equates perceptual experience with belief-acquiring, as in the
Armstrong-Pitcher view. A non-reductive analysis would take belief-induciveness
as an essential feature of perceptual experience, without denying the irreducible
and essential presence of a sensory aspect. For example, in Reid’s classical theory,
perception is constituted by “sensation plus conception”; a sensory episode and a
consequent belief. Armstrong’s main worry is avoiding a two-factor view (raw feel
The Belief Theory: Philosophical Arguments For and Against
81
perceptual experience as an inclination to come to believe certain
propositions and externally adding that such inclination is exerted on the
subject by perceptual means leaves unexplained the intimate relation
between the peculiar phenomenology determined by these “means” and
the (potential) beliefs that such conscious episodes give rise to. Likewise,
even though the phenomenal character of experience does not need to be
hypostatized into qualia or non-physical sense data, nonetheless it needs to
be accounted for in some way. The apparent bodily presence exhibited by
the phenomenology of illusions and hallucinations is too recalcitrant to be
reduced to incoming awareness of belief-contents. Something essential is
missing in this picture.
Thus, BT, at least as it has been classically articulated, to my
knowledge does not seem phenomenologically adequate. Even if it were
true, it is deeply incomplete, failing to capture and explain an essential
element of its object.
Another aspect of this lack of phenomenological adequacy is the
reductive way of treating perceptual introspective awareness. According to
Armstrong, given that perceptual experience is the inclination to acquire a
potential belief—where that potentiality is counterfactually specified—
“introspective awareness of perception would be the awareness of the
acquiring of such potential beliefs (1968, 218).” But perceptual
introspection, whatever it is, involves access to the phenomenology of
one’s experience, which is irreducible to one’s awareness of acquiring a
set of potential beliefs. In particular, the theory entails that perceptual
introspection, by means of which we have access to the content of our
perceptions, would involve the grasp of a counterfactual of the kind
discussed above. For example, “had I had different beliefs and knowledge
than I have now, that state I am just having would have inclined me to
acquire such and such beliefs”: can this really be what perceptual
introspection amounts to? Whatever introspection is, it seems strongly
counter-intuitive that it is made out of the grasp of such abstract and
logically sophisticated counterfactual contents.28 Furthermore, if
introspection is the self-conscious access to perceptual phenomenology,
and beliefs as such lack phenomenal character,29 why should the mere
+ judgment) because he is trying to provide a physicalist alternative to the sense
datum theory, which hypostatized the qualitative element of experience as a
domain of non-physical entities.
28 On the constraint of phenomenological adequacy, see Smith 2001, Crane 2011.
29 Although there is an ongoing debate about that—see Horgan and Tienson 2002,
Pitt 2004, Kriegel 2009—the mainstream view is still that beliefs do not have a
proprietary phenomenal character. In any case, it is the distinctively perceptual
82
Chapter Two Part I
acquiring of an empirical belief exhibit a distinctive phenomenology
instead? If there is no distinctive phenomenology in believing that a is F,
why should there be one in being inclined to acquire the potential belief
that a is F? By reducing perceptual experiences to belief-acquiring, belief
theorists do not do justice to the essential role of phenomenology in both
individuating perceptual experiences and explaining their cognitive role.
Even if an ad hoc theory of introspection could fill that gap, the more ad
hoc collateral views are added, the weaker and less plausible the global
picture becomes.
I.6 Beliefs, Inferences, Concepts: Rationality Constraint
and Generality Constraint
In any case, even if we accept as a substantive and clear view the thesis
that perceptual experience is the inclination to believe or the suppressible
inclination to acquire a belief by perceptual means, even if we leave aside
the doubts of phenomenological inadequacy hinted at above, there are
further problems concerning the radically different behavior of perceptual
experiences from perceptual beliefs in the economy of our cognitive lives.
Let us explore these problems more closely now.
Beliefs are mental states involving concept-possession and inferential
abilities; their ascription is subject to the normative constraint of
rationality typical of fully-fledged propositional attitudes.30 Beliefs,
concept-possession and inferential abilities are intimately connected to the
effect that possessing a concept entails the ability to compose it with other
concepts so as to form one or more beliefs. Having a belief on its own
entails the subject’s ability to compose it with other beliefs or judgments
with the aim of drawing inferences and producing rule-governed reasoning.
If this is the case, attribution of rationality—meant as sensitivity to
reasons plus aiming at truth—is a precondition for crediting a subject with
belief states and concept-possession. So, insofar as beliefs necessarily
involve concepts, any argument against perceptual conceptualism turns out
to be an argument against BT as well. I will reconsider conceptualism
later; now I will only anticipate some main criticisms of that view.
phenomenal character (for example, visual) that must be done justice to within a
theory of perceptual experience, not just a generic phenomenal character of any
sort. The sensuous character of perception cannot be disregarded.
30 On that, see Davidson 1973, 1982, Evans 1982, Crane 1992.
The Belief Theory: Philosophical Arguments For and Against
83
According to many philosophers,31 a relevant condition for conceptpossession is what Evans introduced as the generality constraint (1982).32
Roughly speaking, according to GC, if S possesses the concept {F} and is
able to think [a is G], {F} can be meaningfully applied to a and {G} can
be meaningfully applied to b, then S must be able to think [a is F] as well
as [b is F] (provided that she knows b). I cannot be credited with
possession of the concept {love} unless, if I think that John loves Mary, I
am also able to think that Mary loves John, or that Diego loves Mary,
provided that I know Diego.33 Of course I must be able to entertain these
thoughts even if I believe they are false: “meaningfully applied” does not
mean “truly applied” indeed. GC entails that a concept is never possessed
and exercised in isolation but always as a constituent of actual or potential
propositional structures as thoughts, beliefs or judgments. Likewise, GC
entails that in entertaining a single thought—or in possessing/acquiring a
single belief—we manifest capacities that can be recursively exerted in a
potentially infinite number of different cases.34 Intentional propositional
states “come not as single spies but in whole battalions” as Crane vividly
remarks (1992, 11), so if you have a belief, you must have many other
beliefs rationally interacting with each other as well as with other types of
intentional states (desires, intentions, hopes, fears and the like). In order to
have the concept of {F}, I must have many beliefs in which that concept
figures;35 likewise, in order to have a belief, I must have many other
beliefs interacting under constraints of logical and semantic consistency
and being revisable in the light of other incoming beliefs. Namely, I must
have articulated inferential abilities.
Conversely, in order to play their logical and semantic role in
reasoning, beliefs must be internally structured into discrete constituents—
the concepts—whose composition produces and explains the inferential
properties of the beliefs in which such concepts occur. As Evans (1981,
132) puts it, “behind the idea of a system of beliefs lies that of a system of
31 Cfr. Evans 1982, Peacocke 1992, Brandom 1994, Crane 1992. For a radically
different view, see Fodor 1998.
32 See Evans 1982, 100 ff.
33 {John}, {Mary} and {Diego} as constituents of potential thoughts are singular
concepts, namely, concepts that apply only to an individual.
34 In fact, GC is a way of stating the so-called compositionality of thought as
involving both its productivity and its systematicity. See Fodor 1975.
35 Possessing a concept C amounts to knowing the conditions of its satisfactions,
and knowing when C is satisfied involves knowing that certain beliefs involving C
would be true in certain circumstances, false in others.
84
Chapter Two Part I
concepts whose structure determines the inferential properties of the
beliefs.”
These features of beliefs and concepts are often characterized as
normativity and holism of the intentional. If I ascribe to you the belief that
the table is red, I must ascribe to you the belief that the table is colored,
the belief that the table has a surface, the belief that the table is not
transparent, the belief that if just one other thing in the room was red, there
would be just two red things in the room and so on. In ascribing to you a
belief that a is F, I need to ascribe all the beliefs sufficient for you to
possess the concepts contained in the ascribed belief (according to GC);36 I
also need to ascribe to you the belief that many propositions immediately
incompatible with that belief are not true, and so on. By “normativity” is
meant that if you believe P, if you are rational you ought to believe other
propositions, so in ascribing to you a belief, I am treating you as a rational
being with a whole belief-system aiming for consistency and therefore
sensitive to reasons and logical consequences.37 By “holism” is meant that
intentional contents of your propositional attitudes are reciprocally codetermined in such a way that you cannot have a single attitude toward
content without having many other attitudes toward that content as well as
many other propositional contents.
Now, if having a belief-state involves all of this and the BT of
perception is true, perceptual states would likewise involve everything just
discussed and exhibit similar behavior in our cognitive lives. But this does
not seem to be the case. Consider the reason-sensitivity of belief-forming,
on the basis of which our beliefs are open to constant revision through the
income of new evidence and the increase of our knowledge. Revisability
in the light of evidence is an essential feature of belief but, as we have
seen, persistence of illusions, though disbelieved, shows that perceptual
experience does not behave just like belief. A perception can very well
36 Which beliefs are sufficient for a subject to be ascribed possession of a certain
content is a substantive problem.
37 You do not need to consciously think all the content you “must believe” if you
are rational and believe that P. A belief, as a dispositional state that may or may
not produce conscious judgments and thoughts, can well be unconscious. For
example, I believed that the first minister has two legs even before making that
judgment now. It was true that if one had asked me, I would have answered in a
certain way. If I heard that the first minister lost a leg, I would have thought that he
has just one leg, and so on. Even if I do not know it explicitly, I believe many
contents that it is rational for me to believe given that I believe certain propositions
and I am in a position to envisage those contents as necessarily entailed by these
propositions. Beliefs need not be conscious.
The Belief Theory: Philosophical Arguments For and Against
85
make you revise a previous belief but it cannot be revised in light of a
belief that is incompatible with it. If perceptual experience could be
reduced to belief-acquiring, it would be a state sensitive to reasons and to
the availability of new evidence so that the inclination to believe in the
experience would just disappear as soon as we came to believe otherwise.
But it does not disappear—nor does it lose its “phenomenology of
immediacy”—therefore perceptual experience does not meet the essential
constraints of belief-ascription like revisability, sensitivity to reasons and
evidence, logical constraints of consistency and so on.
I.7 Further Difficulties of the Belief Theory
Another case against this identification has been made by Crane (1988).
Some experiences have contradictory contents, as in the case of the
waterfall illusion where the waterfall looks to be both moving and not
moving. Now, contradictory beliefs at the same time do not occur to the
same subject because our doxastic life is constrained by rationality and the
requirements of consistency in such a way that the contradictions nested in
our whole system of beliefs will be removed as soon as they are made
explicit and conscious as such—namely, as logically incompatible. A
subject can very well doubt and oscillate between P and not-P for a while
of course, but she cannot explicitly believe P and not-P at the same time.
So, if experiences with contradictory contents are possible in a way
contradictory beliefs are not, perceptual experiences are not beliefs.
Indeed, perceptions lack inferential structure and do not entertain logicaldeductive relations as beliefs do. As such, they do not exhibit the essential
features of belief such as normativity and holistic dependence. As Crane
remarks, there is nothing else you ought to perceive just because you
perceive something as being such and such, but there are many
propositions you ought to believe if you are to believe that something is
such and such (1992, 18). Perceptual experiences are evidentially relevant
inputs of new information and contents to our reasoning systems and can
make us revise our beliefs, but not the other way round. That asymmetry,
of great importance from an epistemological point of view, is incompatible
with BT and, all the more so, cannot be explained within it.
In addition, there is a relevant difference between the way certain
information is acquired in perception and the way information is
doxastically acquired that remains unexplained within the framework of
BT. Consider the visual experience of a cup of coffee on the table. In
having the experience, I acquire the belief that [the cup on the table before
me has coffee in it]. I could acquire the belief that [the cup on the table
86
Chapter Two Part I
before me has coffee in it]—suppose I am looking elsewhere and you tell
me—without having to acquire beliefs with detailed contents concerning
the shape, size and color of the coffee, the table and the cup, their
orientation, their distance from me and so on. However, I cannot visually
experience the presence of coffee in the cup without experiencing the
particular shape of the cup, its particular shade of color, its orientation,
how much coffee is in the cup, how dark it is and many other details of the
seen scene. That difference speaks in favor of an asymmetry between the
perceptual way and the doxastic way we acquire information about the
world. A belief contains a discrete piece of information the believer
commits to without having to commit to more specific information
thereby. Of course, if I come to believe that the cup has coffee in it, I need
to come to believe (though maybe implicitly) that the coffee must have a
color, the cup must have a shape and so on. But first, this is something I
must infer from the content of my belief. Second, these inferences, unlike
the visual experience, may well leave indeterminate what color the coffee
is, what shape the cup is, what size the table is and so on. In Dretske’s
terms,38 this is a distinction between the analog and digital coding of
information involved respectively by belief and perception, so that any
belief I can come to acquire through having a perceptual experience is a
different way of “digitalizing” the rich, profuse and detailed content of my
visual perception. I will focus on this important distinction later.39 Here it
is enough to point out that, in the same way that a “picture is worth a
thousand words,” we could well say that an experience is worth thousands
of beliefs. In having a visual experience, I am in a position to form at once
many specific beliefs about a certain range of objects, properties and
relations without having to infer one from another. For example, seeing an
object involves seeing its actual shape, color, orientation, illumination,
distance, size, spatial relations and so on in a way in which acquiring a
belief about that object does not involve acquiring other beliefs about its
actual properties. A belief theorist could reply as follows: 1) Nothing
prevents perception from being the simultaneous acquiring of a very large
number of beliefs.40 A perception is an episode of belief-acquiring, not the
acquiring of a single belief; and 2) my acquiring certain specific beliefs
about actual shapes, sizes and colors (SCM-properties) in visually
38 See Dretske 1981, 135-153, Dretske 1988, 1990, 1995, 2000.
39 See this chapter, Part II; Chapter Three, Part III.
40 Still, in order for BT to be plausible and consistent, that number of acquired
beliefs must be finite. The information acquired cannot be other than determined
and limited with respect to the means, the ways of the information being coded, the
time of causal exposure to the environmental stimuli and so forth.
The Belief Theory: Philosophical Arguments For and Against
87
experiencing a scene does not depend on that experience being reducible
to belief-acquiring; rather, it depends on the fact that the visual way of
acquiring beliefs entails that certain specific beliefs need to be acquired
together. In short, perception is a special case of belief-acquiring where
the subject must acquire many beliefs at once, and these beliefs need to
contain certain specific information about certain kinds of properties,
depending on the perceptual means involved. For example, acquiring by
touch the belief that something is edged will also involve the acquiring of
specific beliefs about the particular texture of the touched object, whereas
seeing the same edged object need not involve belief-acquisitions about
the texture; rather, it will involve belief-acquisitions about the color.
Hearing something moving on your left (say, someone is calling you) will
not involve belief-acquisitions about the color of the moving object but it
will involve belief-acquisitions about the pitch and the tone of the sound it
is producing, whilst seeing the very same moving object will keep “beliefneutral” about what sound it is producing but it will need to involve beliefacquisitions concerning its color, size and shape.
Nevertheless, this reply leaves unexplained why visually coming to
believe that X is moving should entail visually coming to believe that X is
so-and-so shaped, so-and-so colored, at this or that distance and so on. In
fact, the concept {red} does not entail a certain way of being shaped, a
certain way of being distant and so on so while beliefs acquired through a
given experience have specific contents that do not entertain inferential
relations and do not entail each other, still these beliefs need to be acquired
together. The way contents are perceptually acquired determines these
contents being in certain relations that are not those inferential relations in
which belief-contents normally stand to each other. It must have to do with
the nature of perceptual information itself. Belief-contents are insensitive
to their informational origins in a way perceptual contents are not.41 In
perceptual experience, what information I come to acquire is inextricably
connected with how I come to acquire it.42 That does not hold for the
semantic content of belief.
41 On that, see Jacob 1999, 82-85.
42 On that, see Evans 1982, specially the Chapters 6-8.
PART II
AGAINST THE BELIEF THEORY:
ARGUMENTS FROM EXPERIMENTAL EVIDENCE
II.1 Inattentional- and Change Blindness: There Is Seeing
Without Noticing
Another case against BT is constituted by the available experimental
evidence that visual perception does not always involve perceptual belief.
In what follows, I will discuss a bit of such evidence by dividing the
material into two parts. First, I will present and consider the well-known
phenomenon of seeing-with-noticing involved in normal vision as well as
in experimental cases of inattentional blindness and change blindness.
Then I will introduce the Sperling experiment and discuss some
interpretations of it. Second, I will consider some evidence coming from
the clinical study of pathological cases such as some forms of visual
associative agnosia, optic ataxia and blindsight. Finally, I will argue that
all this evidence taken together constitutes another big, insurmountable
problem for the BT of perception.
Inattentional blindness1 is the failure to visually notice a fully visible
but unexpected object because attention was engaged in another task,
event, object or local part of the scene. A quite bizarre example of this is
the “invisible gorilla” experiment (Simon & Chabris 1999).2 Upon
watching a video of a basketball match, subjects are asked to keep track of
the passes or are given some other special task involving attentional focus
on details. At some point, a woman dressed as a gorilla and holding an
umbrella enters the playground and comfortably plays with the other
players. The majority of subjects do not notice or recognize the gorilla,
1 See Mack & Rock 1998, Simon & Chabris 1999.
2 Here is a link to the video of the experiment:
http://www.youtube.com/watch?v=vJG698U2Mvo
Against the Belief Theory: Arguments from Experimental Evidence
89
even though it is so evident to them once they are told about its presence
that they find it hard to believe they could have failed to notice such a
macroscopic object perfectly falling into their visual field.3
Another analog, interesting and well-known phenomenon is change
blindness.4 In the most classical experiment, a photograph is briefly
presented to the subject, followed by a blank, then sometimes an identical
photograph but other times a similar but not identical photograph. The
difference can be an object that has changed in color or shape or has
disappeared. The exposure can go on for fifty times or more, and still
subjects are unable to see the difference between the two pictures, even
though they sequentially and carefully “explored” both of the pictures.5
Change blindness (CB) in fact is an analog but different phenomenon from
inattentional blindness (IB) insofar as it involves memory and the
comparison of a past percept with a present one, whilst IA is a failure to
notice a present object or visible feature within a single scene we are
presently facing. An example considered by Dretske (1994, 2006) consists
of the presentation of two pictures of the same group of people in the same
positions and circumstances, apart from the fact that in the second picture
one of the eight people is missing. Another example by Dretske (2000,
125) asks us to glance at two figures A and B, which look to be perfectly
identical (See Figure 3).
However, figure B has a well-visible black spot that is absent from
figure A. On the one hand, we do not notice that spot otherwise we would
easily be in a position to detect and report how A and B differ. On the
other, we glance attentively at all details of both figures so we have a
strong intuition that we must have seen both that black fully visible spot in
figure B and the correspondent fully visible spotless white area in figure
A. How is it then that we see things we did not even realize we were
seeing?
3 According to Pani 2000, in cases of IA “you see the stimulus but you do not keep
conscious of it in a belief-way.” Therefore, IA would be a case of unconscious
seeing. According to another interpretation, more liberal on what “being conscious
amounts to,” IA could be a case of consciously seeing a feature without noticing it.
4 On change blindness, see the debate between Dretske 1994, Dennett 1994. See
also Simons 2000, Rensink 2002.
5 See Block 2007. The cause of change blindness is well known. In experiments,
things are arranged so that changes are introduced into the scene during the
subject’s saccades. Saccades are ballistic eye movements that can happen several
times a second. During the saccade, the visual system takes in little or no
information about the scene.
Chapter Two Part II
90
A
B
Figure 3
(a modified version of an example by Dretske 2000, p. 125)
Both IB and CB seem to show that there are things we see without
noticing so that seeing should not be identified with visually noticing. We
see more than we notice, in short. Why is that relevant with respect to
assessing the plausibility of BT? It matters a lot because acquiring a belief
entails exerting the concepts the belief is composed of and committing to
its content or endorsing it, and this cognitive act can only be a conscious
event that involves noticing what you come to believe and the
object/objects about which you acquire your belief. If there are things we
see but we do not notice, there are things that are contents of our
perceptions but fail to be contents of any allegedly acquired beliefs. This
eventuality constitutes a case against BT. Without noticing and
recognizing the seen objects as some kind of thing or as having this or that
property, there can be no acquiring of beliefs about the objects seen. So
seeing cannot be coming to believe.6
Intuitively, we are inclined to say that the subjects do see the invisible
gorilla, that the subjects do see the eighth person missing in the second
photograph, that the subjects do see the point making the difference
between the figures, but they do not notice them so they do not acquire
conscious beliefs about those seen objects. If acquiring a belief involves
conceptualizing the constituents of its content, and conceptualizing
presuppose noticing and recognizing as its condition, it seems
straightforward that if within your perceptual experience there are things
you see but do not notice, perception cannot be just equated to beliefacquiring.
6 In fact, it is possible to visually notice something without recognizing it as a
certain kind of thing. I can notice something that suddenly begins to move within
my visual field without being able to recognize it. In any case, what matters to us is
that recognizing entails noticing, and believing entails recognizing. Even if I can
notice O without recognizing O, and maybe I can recognize O without acquiring
beliefs about O, acquiring a belief about O entails recognizing O, which on its own
entails noticing O. So, seeing without noticing entails seeing without beliefacquiring.
Against the Belief Theory: Arguments from Experimental Evidence
91
In Dretske’s efficacious terms (Dretske 2000), the cases above show
that we can have visual object-awareness without having fact-awareness
of circumstances involving the object of which we are aware. For
example, we are visually aware of the black point making the difference
between the two figures but we are not aware that they differ; we are
visually aware of the gorilla moving around but we are not aware that
there is a gorilla moving around; we are visually aware of the eighth
person in the first picture as we are aware of the correspondent portion of
space no longer occupied by him in the second picture but we are not
aware that a person we saw in the first picture is missing in the second
picture.7
Actually acquiring a belief involves what Dretske calls factawareness—to believe is always to believe that something is the case
insofar as believing is a propositional attitude. So, object-awareness without
fact-awareness means perception without belief; it means becoming
perceptually aware of an object without becoming fact-aware of anything
about it and it means perceiving something without acquiring any related
belief about it.
Now, a counter-strategy of the belief theorists would be denying that
seeing without noticing is possible. You wrongly believe you saw
something without noticing it but, in fact, you do not see anything until
you notice it. So, even if you think, “I must have seen it, it was there
before me and it is so big and fully visible after all!” you think wrong—
you just didn’t see the gorilla, or the black spot, or the eighth person and
the like.8
Another more subtle reply would restrict the validity of BT to
perceptual experience rather than concern perception as such. Perceptual
experience is conscious perception. Maybe you do see the unnoticed
objects above but you do not experience them, but perceptual experiences
are belief-acquisitions and therefore they must involve noticing and
recognizing the seen objects. Seeing is not sufficient for visually
experiencing so it is not a problem if seeing is not sufficient for beliefacquiring. The modified theory is that visually experiencing, not just
seeing, is to be equated with belief-acquiring.
7 Imagine you see twenty-seven children on the playground for a brief time.
Probably you will see all the twenty-seven children so you will become conscious
of all those perceptual objects, but you will fail to become conscious of the fact
they are twenty-seven. On the difference between object-awareness and factawareness, see Dretske 1981, 146 ff., Dretske 1993, 1999.
8 This is the move made by Dennett 1991, 1994, and Nöe 2004.
92
Chapter Two Part II
Let us consider one of the examples above and call Mario the eighth
man missing in the second picture. We can ask: 1) Did S see Mario?; 2)
Did S visually experience Mario?
If S saw Mario but did not consciously experience him, then perception
cannot be belief-acquiring but the possibility that perceptual experience is
belief-acquiring is left open. If S visually experienced Mario, then not
even perceptual experience can be belief-acquiring.9 If the subject reports
that she cannot see the difference between picture A and picture B, for
sure she cannot have noticed Mario (nor did she notice his absence of
course)10 and acquired beliefs about him.
Now, if S did not see Mario—as subjects have to believe if they
believe that seeing = noticing = believing—as a consequence, S did not
see any of the people in the picture. In fact, it is counterfactually true that
had S been put before two pictures whose difference was another person
missing rather than Mario (say Gino), S would not have noticed the
difference either, so S would not have seen Gino either. But the theory’s
entailing that S did not see any one of the people in the picture seems to be
an eloquent reductio ad absurdum of the theory itself. Mario was seen.11
Generally speaking, the attentional focus by foveal vision12 involves a
(blurry) difference within the scene falling under our visual field at a given
9 An advocate of BT could reply that even if S did not consciously experience
Mario, S could have acquired an unconscious belief concerning the presence of
Mario. That seems a desperate move though. Granted that unconscious beliefs are
quite possible (see above), it is much less obvious that there can be unconscious
events of belief-acquiring. In addition, if a belief-acquiring involves a
conceptualization, we should be able to think of an exercise of conceptualization
happening at a certain time as a punctual event but without any consciousness. In
any case, a belief must be essentially apt to become conscious and interact with
other beliefs in producing inferences. As we have seen, the doxastic realm is
normative; beliefs are states we are responsible for. How can a subject be
responsible for and rationally responsive to beliefs he does not even know he has
acquired?
10 There is still the possibility that she noticed Mario and suddenly forgot but we
are talking of normal subjects not showing special impairment of the memoryskills. In any case, at least in IA (inattentional blindness), the scene is present so
the memory has nothing to do with the phenomenon.
11 Suppose in the second picture Mario is substituted with Franco, who looks
completely different. In that case, both Mario and Franco would be seen but not
their difference. In Dretske’s view, a difference is a fact about two relata, the fact
that they differ indeed such that you cannot see a fact without noticing or believing
it. See Dretske 1993.
12 Foveal vision is that involving the fovea. Having the highest concentration of
cones, the fovea is the part that grants the maximal visual acuity.
Against the Belief Theory: Arguments from Experimental Evidence
93
time and between what is focused and a thematic object of visual attention
on the one hand, and what is “in view” but only as a potential topic of
attention, as being “virtually present” (Nöe 2004), on the other. Whilst I
am looking at my desk, the door of my office is on the periphery of my
visual field. Even if I am not visually focusing on the door, still it is now
counterfactually true that if the door moved I would notice its movement.
It seems to me quite natural to infer from that fact that I must be seeing the
door now but only certain kinds of change in the scene would make me
notice it.
Of course, in normal visual perception we notice and recognize (and
believe) a lot of things so it could be maintained that perception, globally
considered, essentially involves episodes of noticing and taking things as
being some or another way. In fact, you cannot overlook everything.13 At
any given time, your perceptual experience has some “topics” that attract
your attention and involve conscious recognition, but what is at stake here
is whether perception can be identified and reduced to those episodes of
noticing-recognizing-believing not just whether perception typically
involves such episodes.
It is perhaps worth considering that noticing is not identical to
believing. Noticing an object and some of its properties is a condition for
putting us in a position to acquire a belief about that object; for example,
the belief that the object has such and such properties. So if I acquire a
belief through perceptually experiencing a certain environmental
condition, I must have noticed the object I acquire the belief about, but not
the other way round. Specially, if believing has a normative, holistic and
rational dimension as I have argued for above, this is not the same for
noticing. Noticing a tree and some of its properties (e.g., it being distant)
does not involve acquiring a belief about the tree. Animals not possessing
concepts and propositional attitudes can well notice certain objects and
even overlook certain others. As I will argue later, there are ways of
perceptually noticing and even of recognizing kinds of objects that do not
involve acquiring a belief. But what matters here for challenging BT does
not concern the identity/difference between noticing and believing; for the
time being, it is enough to accept that in order to perceptually come to
believe that a seen object is a certain way, that object has to be noticed in
the first place. Believing entails noticing, independent of whether noticing
entails believing or not. So, seeing without noticing entails seeing without
believing.
13 On that, see Smith 2001, Husserl, Ideen I, 84.
94
Chapter Two Part II
Those who hold that there is no seeing without noticing/believing, like
Dennett and Nöe,14 appeal to the so-called “refrigerator light illusion.” As
soon as you open the fridge, you find the light on so you could be led to
believe that the light is always on. In fact, the light being on depends on
you opening the fridge to look into it. Likewise, in this vivid analogy, our
visual phenomenology presenting us with a whole scene continuously in
view, where we seem to see more than the details and portions we are
specially attending to, would be an illusion insofar as what we are seeing
at a time t depends on our shift of attention to a newly noticed detail. As
the light is off unless you open the fridge, you only potentially see a detail
or an object unless you focus on it and notice it. In short, the refrigerator
light illusion involves mistaking a potential event for an actual event. You
take yourself to see something just because you are aware that you could
notice it, but in fact you could see it only as soon as you noticed it.15 What
you take to be seeing without noticing is just potentially seeing, which
amounts to nothing more than potentially noticing. The seeing = noticing
equation is thus vindicated.
II.2 The Sperling Experiment and What It Tells
with Respect to the Belief Theory
Despite such a subtle reply having some force, the so-called Sperling
effect shows that we must see more than we notice.16 Sperling17 showed
14 See Dennett 1991, 1994, 2002; Nöe 2004, O’Regan and Nöe 2001. Actually,
Dennett and Nöe do not identify perception with belief-acquiring à la Armstrong;
they rather take perceptions to be dispositions to believe.
15 Dennett 1991, O’Regan and Nöe 2001 and Nöe 2001 argue in many ways for
the unreliability of our introspective phenomenology by appealing to phenomena
like the blind spots in our visual field. It is well known that a relevant portion of
the surface at the center of our retina is covered by nervous fiber such that there are
no photoreceptors on it. But our experience does not introspectively present itself
with black holes in the visual field. In addition, the visual exploration of the
environment is not as continuous and uniform as it introspectively seems to us;
rather it involves jumpy and discontinuous saccadic movements. All that is of the
greatest interest but, as Dretske rightly remarks (2006, 164), what matters for
whether in change blindness cases we see more than we notice is not whether we
see everything we introspectively seem to see, it is rather whether we see more
than we notice. In other words, that we are wrong in claiming to see the whole
scene falling under our visual field in detail does not mean that we don’t see
something even if we do not notice it. See also Nöe 2006.
16 I have introduced a certain argumentative use of the appeal to the refrigerator
light illusion concerning the reducibility of seeing to noticing. Another use
Against the Belief Theory: Arguments from Experimental Evidence
95
his subjects arrays of alphanumeric characters; for example, three rows of
four characters each, for 50 milliseconds, followed by a blank field.
Subjects reported that they could see almost all characters. However, when
asked to say what letters they had seen, they could remember
approximately only four letters, fewer than the half of the presented letters.
Until that point, the situation is analogous to the examples above
concerning inattentional blindness and change blindness, namely, we are
puzzled as to whether the subjects did really see the “forgotten” letters or
if they properly saw only the ones they noticed and were able to truly
report. On the one hand, S must have seen them—they were within her
visual field and there was a phenomenal “feeling” of having in view a
continuous portion of space populated of many letter-shapes. On the other
hand, how is it possible that S saw all the letters if she can identify just
four? Sperling found a genial way to settle the question experimentally.
He played a tone just after the array was replaced by the blank. Subjects
were asked to report the letters composing the top row when the tone of
the sound was high, to report the letters in the middle row when the tone
was in the middle range and to report the letters in the low row when the
tone was low. In these cases, subjects were able to report all the letters of
the “auditorily intimated” row (but not those in the other rows, of course).
Now, in this case, saying that you must have seen the letters becomes
more than a strong intuition; actually, it is a necessity. Indeed, the
acoustically induced decision to attend to one or another row is made after
the visual stimulus has already been removed. So, provided that you
cannot see objects that are not there, you must have seen them before they
disappeared. The Sperling effect entails the existence of an “iconic
memory” (Neisser 1967), a very short-term store of visual information that
is accessible only when the subject is prompted in certain ways to attend to
that information. Of course, as soon as the acoustic marker prompts the
subject to attend to a row, the accessibility to the other rows irremediably
vanishes forever.
There is a controversial and fascinating debate, which I will not enter
into now, concerning whether the perception of the unattended letters is
conscious or not.18 Whilst the experiment neatly shows that we can see
concerns phenomenology, so it says that a perceptual state has phenomenology
only as soon as you attend to it. These uses are intimately connected but still
different. For example, maybe there can be a phenomenology not involving any act
of noticing or recognizing, much less involving an act of believing.
17 See Sperling 1960, Baars 1988.
18 See Block 2007. Block takes the experiment as showing that “phenomenology
outstrips cognitive accessibility.” But others take it to show that vision outstrips
96
Chapter Two Part II
things without noticing them, it is less obvious that it also shows we can
consciously see or experience something without noticing it. In any case,
the already seen rows can be accessed, though not all at once, after they
have disappeared, so there is no room for arguing that you begin to see
something only as soon as you notice it, as some tried to argue for with
respect to change blindness and inattentional blindness. How could I begin
to see something that is not there anymore? It would be a sort of magic
performance.19
II.3 The Case of Visual Associative Agnosia and the Belief
Theory
Additional examples of experimental evidence are of the greatest interest
for evaluating BT, like certain clinical reports about patients with dramatic
cognitive and perceptual dysfunctions. In particular, I want to briefly take
into consideration the case of (a kind of) visual associative agnosia first,
then the case of optic ataxia, and finally the more familiar case of
blindsight.
Visual agnosia20 involves impairments in perceptual capacities that are
not due to elementary sensory malfunction. Associative visual agnosia is
one kind of that pathology—typically caused by specific kinds of brain
damage—involving an inability to recognize familiar objects despite being
normally able to perceive and describe forms and other visible features.
So, object-recognition21 is impaired despite low-level visual perception
working perfectly. The patient is able to recognize the objects through
phenomenology. According to them, the episode of vision acquires a
phenomenology as soon as the subject attends to the objects. See also Dretske
2006.
19 There is an auditory analog of the Sperling effect consisting of an “echoic
buffer” (not to be confused with short-term memory). Our system “iconically”
stores auditory information for a short while, which can be accessed after the
stimulus has ceased only if we focus our attention on the just-stored material. To
report an example by Fodor 2007, it may happen that I hear the clock bells ringing,
say, for four times, when I am not attending to them at all. Suddenly, I shift my
attention to that perceptual intake because I need to know what time it is. There is
a short interval within which I am still able to “count” the times of ringing.
20 On visual agnosia, see Farah 2004, Milner & Goodale 1995, Goodale & Milner
2004.
21 There are different kinds of visual associative agnosia depending on the kind of
objects the subject cannot recognize. For example, patients with prosopagnosia
cannot recognize faces and patients with alexia cannot recognize words.
Against the Belief Theory: Arguments from Experimental Evidence
97
other sensory ways (e.g., touch). Her visual system is not impaired and nor
are her intellectual, descriptive or inferential skills. She just cannot
“extract” from the percept the thick information concerning the kind of
thing she is perceiving. A scientist who had worked a lot on visual
associative agnosia described the pathology as “a normal percept is
stripped of its meaning” (Teuber 1968, 293). Despite the form perception
being good22—at least in pure visual associative agnosia—the objectrecognition is blocked so the perceptual input is taken in but does not put
the subject in a position to categorize the object in many relevant ways;
therefore, she cannot acquire relevant beliefs about the object. As we have
already noticed, there may be recognitional acts in perception that are
more basic than fully-fledged beliefs so that recognizing does not entail
believing. In other words, there may be something like a non-conceptual
or “proto-conceptual” recognition but, conversely, fully-fledged perceptual
beliefs do involve conceptualization of the percept. If perceptual
experience is possible without object-recognition, a fortiori it is possible
without belief. Provided visual associative agnosia shows dissociation
between perception and belief, it speaks against identity as a result.
A possible reply to the argument based on visual associative agnosia is
that such pathology only shows that there can be a perceptual experience
without the certain kinds of belief-acquiring that occur in normal subjects
rather than that perception is possible without any belief-acquiring. In fact,
these subjects do report that the seen object is red, wide, has such and such
shape and so on. For example, a subject cannot recognize a comb as a
comb (even if she possesses the concept of {comb} and can be prompted
to exert it perceptually, for example, through touch) and wonders if it is a
pipe, and another subject cannot see a bottle opener as a bottle opener and
asks whether it could be a big key,23 but they could wonder whether that
was a key or this was a pipe insofar as they acquired true beliefs about the
seen objects in the first place, concerning shape-properties, formproperties, color-properties and the like. What these cases prove is that
certain pathologies can reduce the beliefs perceptually acquirable but they
do not prove that perception is not belief-acquiring. For example, as we
will see later, some philosophers take the content of perception to be
confined to the representation of “thin” spatial-chromatic-morphological
22 Form perception is impaired in visual apperceptive agnosia, which is often but
not always combined with visual associative agnosia. The most interesting case
with respect to the belief/perception relation is the pure visual associative agnosia,
namely, object-recognition impairment without form impairment.
23 Such examples are provided by experiments in Rubens & Benson 1971.
98
Chapter Two Part II
properties (SCM).24 So if we combine BT with the view that the only
properties represented in perception are SCM, “thick” recognitional
properties involved in representing a comb as a comb or a key as a key
would not be properties represented in perception at all, so such cognitive
failures at recognizing things would not be strictly perceptual deficits. In
that case, such visual experiences would be acquisitions of beliefs
concerning SCM-properties, as visual experiences had by normal subjects.
So, in perception, we conceptualize objects as certain kinds of things and
therefore acquire beliefs about them, but in cases like visual associative
agnosia, subjects fail to conceptualize objects as certain kinds of things (a
comb, a watch, a known person, a word) but they do conceptualize them
as red, bulgy, thin, distant and so on as normal perceivers do.
This reply is sound, but it weakens the original theory by making it
compatible only with a particular view on the range of properties
representable in perception. Furthermore, it does not explain why these
impaired subjects are able to prompt the exercise of the right thick
recognitional concepts just by touching the very same objects they so
“poorly” see. If touch is (immediately and non-inferentially) “recognitionally
rich” whilst vision of the same objects is “recognitionally poor,” there
must be an impairment in vision as well as in the system of cognitive
categorization of the percept. Maybe a further reply could be that the
impairment is neither in the visual perception nor in the system of
categorization-recognition-conceptualization but rather in the mechanism
of integration between the information strictly delivered by visual
experience and certain non-perceptual concepts, whilst the mechanism of
integration of tactile information with conceptualization is working well.
But the very fact of introducing so many positive auxiliary hypotheses
seems to make this global line of defense more and more ad hoc.
II.4 The Case of Optic Ataxia and the Belief Theory
There is other clinical evidence concerning optic ataxia25 that is clearly
immune to the above scenarios. Optic ataxia is another kind of impairment
that presents exactly an inverted situation of visual apperceptive agnosia
(which is, unlike associative agnosia, an impairment in recognizing and
reporting information about forms and spatial properties). Patients with
24 For example, McGinn 1982, Clark 2000, Byrne 2011. Theorists who hold that
“thick” properties (like {comb} or {key}) are also represented in perception are
Siegel 2006, 2011, Bayne 2011 and Prinz 2006.
25 On optic ataxia, see Clark 2001, Milner and Goodale 1995.
Against the Belief Theory: Arguments from Experimental Evidence
99
visual apperceptive agnosia are normally good at manipulating, reaching
and acting upon objects; in short, they have perfect visuo-motor control.
On the contrary, visually ataxic patients are normally good at recognizing
objects and verbally reporting what the objects are, where they are and
how they are spatially- and shape-arranged, but they show dramatic
impairments at performing elementary tasks like reaching for an object,
putting an object into a hole and the like. Both impairments, considered
together, provide evidence for the existence of two different relatively
independent mechanisms implemented respectively in the so-called
“dorsal stream” and “visual stream.”26 Indeed, visual agnosia is the result
of lesions on the ventral stream, whilst in optic ataxia the lesions are on
dorsal stream. Optic ataxics behave as if they could not use the spatial
information inherent in any visual scene despite their perfect recognitional
and descriptive competence.
For example, while a subject with visual apperceptive agnosia—
involving impairment of form perception—cannot see the orientation of a
displayed slit but can nonetheless post a letter through the slit, a subject
with optic ataxia can recognize and truly report the orientation of the slit
but is unable to post the letter through it despite the fact that he does not
have any bodily impairment. The first is bad at using perceptual
information for recognition, recall and reasoning but is good at using that
information for fine-grained behavior and action upon the perceived scene;
the second is bad at this second task but is good at the first.
Both impairments provide evidence for the double coding of visual
information also found in unimpaired subjects. For example, consider the
Tichener circles illusion, a circle surrounded by an annulus of small circles
appears bigger than an equally big circle surrounded by an annulus of big
circles.
Figure 4 The Tichener Circles Illusion
(from: http://mathworld.wolfram.com/TitchenerIllusion.html)
26 The ventral stream leads from primary visual cortex, through V4, to temporal
areas; the dorsal stream leads from the primary visual cortex, through MT, to the
parietal cortex. See Milner and Goodale 1995, Jennerod and Jacob 2006.
100
Chapter Two Part II
This is what the experience makes you think or believe (provided you
are not told you are the victim of an illusion). The size of the circle is
misrepresented to the effect that two equal sizes appear to be different in
virtue of surrounding “distractors,” just as with the lines in the MüllerLyer illusion. Now, you can enlarge the central disc of the second figure to
make it as big as to (falsely) appear equal to the first central disc. This is
what Algioti et al., ingeniously, did with their colleagues.27 They used
poker chips as the central discs and asked the subjects to immediately pick
up the target disc on the left if the two discs appeared equal in size and to
pick up the one on the right if they appeared different in size. Surprisingly
enough, despite them being under an illusion (both phenomenologically
and doxastically), they made the right choices, acting according to the
actual sizes of the discs. The experiment showed that even in unimpaired
perceivers, an intake of reliable perceptual information can be shown in
fine-grained behavior without becoming explicitly conscious as to produce
beliefs. On the contrary, that intake of perceptual information can happen
despite wrong perceptual beliefs being acquired about the same scene.
Both optic ataxia and the Tichener circles illusion show ad
abundantiam that perception cannot be belief-acquiring insofar as some
level of perceptual information is used in action-guiding, even if it is
absent from any content consciously considered or believed or in
contradiction with the content consciously considered. So, in these cases,
not even the ad hoc replies above will do. Here we have correct and useful
visual perceptions of objects as having certain SCM-properties without
any recognition-conceptualization of them as having these properties, not
just without recognition-conceptualization of other thicker properties
belonging to the same objects. Here the only explanation is that the
behaviorally displayed information that the discs are unequal is not
conscious; it can co-exist indeed with other conscious (wrong) information
that they are equal. So again, it could still be the case that perception is not
belief-acquiring but that conscious perception, or perceptual experience,
is. Indeed, the content of such action-guiding perceptions not only is not
the propositional content of empirical beliefs, it is not even the content
exhibited by our perceptual phenomenology (of course those contents for
belief theorists are one and the same). So such a piece of evidence is a
good argument against the equation Perception = Belief-Acquiring but it
does not work against the more restricted equation, Conscious
“Experiential” Perception = Belief-Acquiring. So far so good.
Nonetheless, the evidence that there is perception without belief at least
27 See Algioti et al. 1995.
Against the Belief Theory: Arguments from Experimental Evidence
101
should constrain the belief theorist to explain why there cannot also be
perceptual conscious experiences without belief. Moreover, if those
perceptions are intentional—as a matter of fact, they are about the circles,
which are perceived, though unconsciously, as unequal in size—the main
reason to introduce BT is flawed. Such a theory purports to explain
intentionality of perceptual experience by reducing it to intentionality of
perceptual beliefs. But if there is perceptual intentionality without
perceptual beliefs, then the fact of acquiring a belief with a certain content
cannot be the only way to explain perceptual intentionality. In short, an
appeal to the intentionality of beliefs as the best explanation of
intentionality of perception is “screened off” by the existence of perceptual
episodes—though non-conscious—that are intentional without involving
beliefs.
II.5 The Case of Blindsight and the Belief Theory
On the other hand, the idea that only conscious perception is beliefacquiring meets an obstacle in the experimental evidence concerning the
existence of perceptual belief-acquiring without any related conscious
experience. The so-called blindsight is such a phenomenon. In most cases
of blindsight, damage to the primary visual cortex (V1) results in the tragic
absence of any visual phenomenology to the effect that the subjects claim
to be completely blind.28 Nonetheless, if prompted to a forced response or
invited to guess about a surrounding scene, they show a surprising ability
to answer correctly at levels significantly much above chance. Hilbert
reports the case of M.S., suffering of cerebral achromatopsia (color
blindness due to brain damage).29 He denies seeing color, he cannot sort
objects by color, he cannot name the colors of objects shown to him and
he mostly behaves as if he were completely color-blind but in many
experiments he is able to recognize shapes and figures whose boundaries
are specified exclusively by color. He acquires true empirical beliefs30 by
28 On blindsight, see Cowey 2010.
29 The source of Hilbert (1994, 447) is Mollon et al. 1980. On achromatopsia, see
Beauvois & Salliant 1985.
30 I say that he acquires true beliefs rather than knowledge because it is far from
clear whether such beliefs, although they are gained through a relatively reliable
process, are or are not justified. In any case, even according to a reliabilist view of
knowledge, treating these true acts of guessing as knowledge would entail a very
undemanding sort of reliabilism. The subject does not know whether she knows,
does not even believe she knows and does not have any idea of whether and why
102
Chapter Two Part II
purely visual means but without any related visual experience. Vision
operates directly to give him beliefs about what is seen without any
mediation of visual experience.31
Now, considering these well-known experimental data (inattentionaland change blindness, Sperling effect, visual associative and apperceptive
agnosia, optic ataxia and blindsight), it seems we cannot escape the
conclusion that visually experiencing is neither sufficient nor necessary for
coming to believe by visual means. So not even just conscious experience
can be equated with believe-acquiring. As a result, both the more
ambitious version (seeing is believing) and the more restricted version
(visually experiencing is believing) of BT must be rejected.
Concluding Remarks
In Part I of this chapter, I presented and critically discussed the belief
theory of perception. First, I pointed out some virtues of that view—i.e.,
explaining intentionality, semantic evaluability, and perceptual beliefforming without apparent transitions—which made the view historically
attractive in facing the argument from illusion better than direct realism
and adverbialism but without committing to the ontological extravagances
involved in sense data theory.
Second, I produced many arguments to show that BT is to be rejected
despite those theoretical advantages over other views. These arguments
were grouped into two kinds—the more philosophical arguments, and
arguments based on experimental evidence (Part II).
With respect to the first kind, I considered the normative, rational and
holistic constraints for belief-ascription as well as the generality constraint
holding for concept-possession. Were perception identical to belief, its
content would satisfy those constraints, but this is not the case.
Furthermore, I argued that persistence of illusion cannot be accommodated
that process of coming-to-believe is reliable and truth-conducive. In any case, this
is not our concern here.
31 Another interesting case is the so-called unilateral neglect. Neglect patients
often behave as if half of their world no longer exists. Due to brain damage, they
are totally oblivious of objects and features of half of the room, they may eat only
from a half of the plate, they may shave only one side of their face and so on.
However, they indirectly show sensitivity to objects and features falling under the
neglected half when prompted. So this is another example of perception without
perceptual awareness, just like blindsight. On unilateral neglect, see McGlincheyBerroth 1997, Driver and Villeumier 2001, and Dretske 2006 for a more
philosophical interpretation of data.
Against the Belief Theory: Arguments from Experimental Evidence
103
within BT since any attempt to retool the theory by the notions of
“suppressed inclination” or “potential belief” is doomed to fail. In
addition, contradictory and impossible contents may feature in perceptual
experience in a way they cannot feature in belief.
With respect to the second kind of arguments, in Part II I considered
different types of cases clearly showing that perceptual belief or beliefacquiring are neither necessary nor sufficient for perception. Change
blindness and inattentional blindness show that we can see objects and
properties without noticing them, without recognizing them and without
acquiring beliefs about them. On the other hand, any attempt to deny that
we really see what we do not notice-recognize-form beliefs about is
dashed by other evidence, such as the Sperling effect. The Sperling
experiments indubitably show that we must see at least something we do
not either notice or believe we see. Since we can be prompted to notice
objects we are not in visual contact with anymore, we must have seen
them before.
I also considered visual associative agnosia (VAA), a pathology
involving the dramatic absence of object-recognition despite the fact that
one’s vision is not impaired. VAA shows that there can be a visual
perception of objects without a respective belief-acquisition of what these
objects are, so that divorce between (impaired) recognition or beliefacquiring on the one hand, and (unimpaired) visual-perceptual capacities
on the other, advocates a distinction between the two.
Moreover, optic ataxia also shows that there can be perception of
spatial properties, which suffices for fine-grained action-guiding, without
any associated doxastic awareness of those perceived contents. This
happens also in normal subjects, as the Tichener illusion and other similar
cases show. People behave as if they “knew” that two circles are the same
size even though the very same circles consciously “look” to be of
different sizes; there can be perception of spatial properties not only
without a respective belief-acquisition but also associated with the
perceptual acquiring of a belief that contradicts the unawarely perceived
content. The belief theorist could reply that at least conscious perceptual
experience is belief-acquiring even if there might be an absence of beliefacquisition in perceptions without experience. I have argued also that that
line of defense is challenged by the well-known case of blindsight, where
there is perceptual belief-acquiring without any perceptual conscious
experience. So, if there is perception without belief-acquiring as well as
perceptual belief-acquiring without conscious experience, neither
perceptions nor perceptual experiences can be identical with beliefacquisitions. If conscious perception were identical with perceptual belief-
104
Chapter Two Part II
acquiring, we would expect that a perceptual belief-acquisition without
conscious perception would be impossible. Blindsight, however, shows
that it is.
In the next chapter, I will introduce the content view as the only
possible evolution of BT that is not vulnerable to these objections.
CHAPTER THREE
THE CONTENT VIEW
Introduction
In this chapter, I introduce and start discussing the content view. In Part I, I
introduce the content view as a promising way of saving the virtues of the
BT without facing its main problems. In Part II, I argue that the content
view best captures our intuitions and does the greatest justice to the
ordinary way of ascribing perceptual experiences, which I analyzed at
length in Chapter I.
Part II introduces the main issues within the content view and so
prepares the ground for the more substantive discussion of the following
chapter. Finally, I put these provisional conclusions together in order to
sum up and state the points just described.
PART I
THE CORE IDEA OF THE CONTENT VIEW (CV)
I.1 Introducing the Content View
In the previous chapter, I pointed out some important virtues of BT. To
recap them shortly, BT explains the intentionality of perceptual
experiences, it explains their semantic evaluability—i.e., perceptual states
can be taken as correct or mistaken, as veridical or falsidical—and it
explains why experiencing in this way naturally involves taking things to
be a certain way. Moreover, it was a better alternative to direct realism
than sense data theory on the one hand (in facing the arguments from
illusion and hallucination without committing to a suspicious and
redundant ontology), and a more plausible approach than adverbialism (in
doing justice to the articulated quality of perceptual information as well as
to the “purported objectivity” of perceptual experiences witnessed in their
phenomenology, made out of worldly objects and properties rather than of
“ways”) on the other.
I also considered many reasons why the view should be rejected, and I
argued that these definitely outweigh the advantages. Among these
reasons, I considered that belief involves concept-possession, its ascription
is normative and it is subject to constraints of consistency and rationality,
unlike perception. In addition, belief is committal in a way perceptual
experience is not (I can disbelieve my experience), illusory experiences
persist despite being disbelieved (as beliefs do not), experiences can
present contradictory and impossible elements and so on. Furthermore, I
have recalled some global empirical evidence showing that believing is
neither necessary nor sufficient for perceiving and/or perceptually
experiencing. Another weakness of the view I have stressed so far is its
phenomenological inadequacy. Experiencing does not introspectively
seem to consist just of us acquiring an inclination to believe certain
propositions, especially if that inclination is counterfactually specified.
There must be something more in perceiving than being potentially
committed to abstract propositional content.
The Core Idea of the Content View (CV)
107
A more promising view would be one that embeds the advantages of
BT without its fatal problems. That is what the content view (CV) aims to
do.1 According to CV, perception has intentional content and it is normally
belief-inducing but it is not belief or belief-acquiring. As some of the bestknown advocates of CV put it:
A perceptual experience represents the world as being a certain way. (Evans
1982, 226; Peacocke 1992, 66)
Experiences may be correct or incorrect […] In short, experiences have
representational or semantic properties; they have content. (Davies 1992, 22)
[...] experiences have contents, where contents are a kind of condition under
which experiences are accurate, similar in many ways to the truth-conditions
of beliefs. (Siegel 2010, 4)
In perceptual experience, things are represented as being a certain way.
If that is the case—if CV is correct—then the intentionality of perceptual
experience can be explained by experiences having content. Likewise,
their semantic evaluability as correct or incorrect, veridical or falsidical,
can be explained by them having correctness-conditions determined by
their contents. Finally, the acquisition of beliefs, which perceptual
experiences involve in normal conditions, can be explained by the subject
doxastically committing to the content of her perceptual experience so that
she takes things to be the way her experience represents them as being.
I.2 Content View and Belief Theory
So the main fundamental virtues of BT can prima facie be embedded in
CV. Can the problems of the BT pointed out above also be accommodated
within CV? Let us see.
Actually, BT is a specific form that CV can take. Indeed, if experiences
are belief-acquisitions, experiences are contentful states. They do represent
the world as being a certain way; representing things as being such and so
is just what a belief does. However, CV as such is not committed to taking
perceptions to be belief-acquisitions or perceptual content as being the
same in nature as the content of belief. First, a CV-theorist can distinguish
perception from perceptual belief and attribute content to both. Second,
given that believed contents are committal,2 conceptual and propositional,
1 Brewer 2008. See also Siegel 2011.
2 Believing that P is committing oneself to the truth of P.
108
Chapter Three Part I
as well as holistically ascribed and reason-sensitive, as seen above, a CVtheorist could attribute to perceptions different kind of content from that
which characterizes perceptual beliefs. In such a version of CV, not only
do perceptions have content on their own without being belief-acquisitions
but their content is also of a different kind from that of beliefs. How
should such content be characterized so that CV preserves the virtues of
BT without running up against its difficulties? Let us proceed in steps and
cursorily retrace the main difficulties to assess how the alleged perceptual
content must not be characterized at least.
First, belief states are fully-fledged propositional attitudes involving
possession and exercise of the concepts constituting the believed contents.
Belief-ascription is subject to the rationality constraint and conceptpossession ascription to the generality constraint, but perceptually
experiencing something as being so does not need to meet these
demanding constraints. Hence, if perceptions are to be contentful episodes,
their contents must be free from both rationality and generality
constraints. Indeed, no such content can be a belief or be identical in kind
to the content of a belief.
Second, perceptually experiencing a as F cannot be as committal as
believing that a is F. It must be at least possible that I take what I perceive
“at face value” and disbelieve the content of my perceptual experience, as
happens, for example, in illusions known as such. So, if perceptual
experiences are committal in any way, they cannot commit us to their
contents the way beliefs do. Indeed, beliefs are commitments to the truth
of certain propositions, to their contents being exemplified.3
Third, perceptual content cannot be as indifferent to its informational
origins as belief-content is; rather the special way of accessing that content
must matter. Perceptual content needs to be phenomenologically adequate
so it needs somehow to embed in itself the “vehicle” through which it is
acquired.4 It cannot be just the abstract proposition that constitutes
believed contents, even if it is in virtue of having certain content that a
perception induces in the subject acquiring certain empirical beliefs rather
than others. That calls for an epistemologically and semantically
satisfactory account of the transition from perceptual experiences to
perceptual beliefs, but I will deal with that later on.
In addition, perceptual content needs to be compatible with the
experimental evidence I have recalled to make a case against BT. If
3 This is clearly shown by Moore’s Paradox: “I believe that P but not-P” is a
paradox just because believing P is committing to the truth of P, so that selfascribing the belief that P cannot be consistent with judging non-P.
4 On the content/vehicle distinction and phenomenology, see Crane 2008, 23.
The Core Idea of the Content View (CV)
109
perceptions are contentful states, the kind of content they bear must allow
that we are in such contentful states, we don’t notice (a part of) the
contents of these very states (inattentional- and change blindness, Sperling
experiments); we don’t recognize the perceived objects as being of a given
kind of thing (visual associative agnosia); or we are not even perceptually
conscious of them (optic ataxia, Tichener illusion, blindsight).
Maybe only a view involving multi-layered content could accommodate
all that evidence. Anyway, what is certain is that perceptual experiences
cannot be beliefs and their content cannot be of a kind that is subject to
rationality and generality constraints. Moreover, if GC is a good criterion
for concept-possession, and if the arguments above are sound, then
perceptual content needs be non-conceptual. In other words, some
arguments I have introduced above against BT are good arguments against
conceptualism as well.5 Perceptual conceptualism6 is the version of CV
that holds that perceptual experiences have conceptual content. Even
though experiences are not beliefs, in acquiring perceptual beliefs on their
grounds, a subject endorses the very same kind of content that
characterizes the experiences so having a perceptual experience with
certain content involves a display of the very same conceptual capacities
operative in forming the respective perceptual belief. This idea is
incompatible with the evidence just recalled, since seeing does not
necessarily involve noticing, recognizing, conceptualizing, believing. In
seeing, for example, the very same “conceptual capacities that are
operative in believing” are not necessarily found, even though seeing often
involves attentive noticing and recognitional acts as well as acts of
conceptualization and belief-formation in those cognitive systems able to
conceptualize and believe at all, but more on that later.
I.3 Phenomenal Character and Representational Content
I have charged BT with the accusation of phenomenological inadequacy.
Before asking whether and how some version of CV could do any better in
that respect, I want to briefly introduce the notion of phenomenal
character, a central notion for the debate about the nature of perception.
Whilst perceptual experience is taken as having a distinctive phenomenal
5 As we have remarked, beliefs involve concepts and concept possession involves
having beliefs containing the possessed concept as well as being able to make
certain inferences. A concept determines and explains the inferential properties of
the beliefs in which it occurs as a constituent.
6 See McDowell 1994, Brewer 1999.
110
Chapter Three Part I
character, CV is a “two-component view”7 insofar as it takes perceptual
experiences to exhibit two fundamental features: phenomenal character
and representational content. While the representational content is the way
experience represents the world as being—or the way the surrounding
world seems to be to the subject according to her perceptual experience—
the phenomenal character is the “what it is like” or subjective aspect of
experience.8 For example, the something it is like to see (a certain shade
of) red is subjectively different from what it is like to see (a certain shade
of) green. A given conscious mental state has phenomenal properties if
there is something it is like to be in it. Perceptual conscious states are such
that there is something it is like to be in them so I will say that they are
phenomenal states insofar they have phenomenal properties.9 Also,
particular elements of a given perceptual experience can be said to have
phenomenal properties. For example, if there is something it is like to see
something bulgy, red and round, there is also something it is like to see
something red, where such a phenomenal property is involved in the first
complex phenomenal state. Of course, a phenomenal state can very well
be multimodal. For example, there is something it is like to see certain
objects and properties and hearing certain sounds at a given time. A global
phenomenal state can also be not just perceptual but involving other
emotional or “feeling-like” conscious properties of any sort. With Siegel
(2010, 20ff.), we can call overall experience a certain global conscious
state with its many facets, at a time t, like looking at the clouds, hearing
the wind, feeling the cold air on one’s face, touching and guiding the
handlebars and feeling happy when cycling into the forest. Taking an
overall experience, we can zoom in, for example, on its visual component,
namely, on the visual phenomenal state involved in the overall
experience.10 We can further zoom in on an aspect of that global visual
experience and consider what it is like to see a certain property (e.g., a
certain color of a certain cloud). That single phenomenal property could be
shared by other visual experiences as well as by other very different
7 See Maund 1994, Siegel 2010, Chalmers 2008.
8 See Nagel 1974, who famously introduced the “what it is like” characterization.
9 As Bayne writes, “phenomenal states are states which it is ‘something it is like’
to instantiate” (2011, 17). In Tye’s words, “consider your visual experience as you
stare at a bright turquoise color patch in a paint store. There is something it is like
for you subjectively to undergo that experience. What it is like to undergo the
experience is very different from what it is like for you to experience a dull brown
color patch. This difference is a difference in what is often called “phenomenal
character” (Tye 2007).
10 On that, see also Horgan and Tienson 2002.
The Core Idea of the Content View (CV)
111
overall experiences.
So a mental state has the determinate phenomenal character according
to what it is like for the subject to be in it. The phenomenal properties of
the state constitute “what it is like” to be in that state. Often such
properties are called qualia but this word has many meanings and uses
rather than a uniform and shared sense within this debate11 so I prefer to
neutrally talk of the subjective, qualitative character of perceptual
experiences, their phenomenal character indeed. I am conscious that such a
characterization is not very clear but more than providing a definition of
the “what it is like” aspect of experience on the whole, an appeal is made
simply to our intuition and introspective capacities. In fact, there is no way
of ostensibly pointing out such a subjective aspect in a publicly accessible
manner so faute de mieux each one needs to do so by itself.
In any case, according to CV, perceptual experiences exhibit two
fundamental features whose relation will be carefully examined later on:
representational content and phenomenal character. A certain state S is a
representational state if it has representational properties (= a content). The
very same state S is also a phenomenal state insofar as it has phenomenal
properties (= its phenomenal character). So the phenomenal character of a
visual experience consists of the properties of your conscious experience
you can introspectively attend to whilst the representational content of a
visual experience consists of properties your experience represents things
as having.12 When I visually experience a red square, in normal conditions
something I see is represented as both red and square in my experience so
these are the representational properties providing my experience with its
intentional content. The phenomenal properties of that very experience
make up the subjective visual phenomenology typically associated with
experiencing something as red and square so they constitute the
phenomenal character. So far so good.
11 Tye individuates at least four uses of the word: 1) Qualia as those
introspectively accessible qualities making up the phenomenal character of your
experience; 2) Qualia as those intrinsic and non-representational features of nonphysical sense data responsible for the phenomenal character; 3) Qualia as those
intrinsic, non-representational properties of experiences solely responsible for their
phenomenal character, where that characterization is ontologically neutral about
whether there are non-physical mental objects or not (Peacocke 1983, Nagel 1974);
and 4) Qualia as intrinsic properties of experiences that are also ineffable, nonphysical and given incorrigibly to their subject (Dennett 1991). We are concerned
only with the first use, the most theory-neutral, so that no-one would ever deny that
each conscious experience has qualia; i.e., has a distinctive phenomenal character.
On the different notions of qualia in the current debate, see also Martin 2008 (II).
12 On that, see Dretske 1995, Tye 2000.
112
Chapter Three Part I
I.4 Transparency and Richness of Details
What about the phenomenological adequacy of CV then? What perceptual
content would be phenomenologically adequate? It should be one that
would do justice to the phenomenal character of experience; that would
“respect” or be compatible with the way perceptual experience presents
itself insofar as it is introspectively accessible by us. Of course, it is not a
priori ruled out that the spontaneous phenomenology of experience is
systematically misleading but a theory consistent with the way experience
introspectively seems to oneself to be featured, other things being equal,
would be a much better theory than one that violated that phenomenological
adequacy constraint. Therefore, even if what I am calling phenomenological
adequacy is not an absolute theoretical necessity, still it is a relevant
desideratum for a theory concerning our experiences and their manifest
features, therefore concerning us as subjects of experience.
Now I want to introduce two fundamental features of visual
phenomenology that are introspectively assessable: transparency and
richness of details.
Transparency is that property of perceptual experience in virtue of
which when you introspectively attend to the properties of your own
experience, you end up by attending to the properties your experience
attributes to the objects you are perceiving.13 In Harman’s clear words:
Look at a tree and try to turn your attention to intrinsic features of your visual
experience. I predict you will find that the only features there to turn your
attention to will be features of the presented tree, including relational features
of the tree ‘from here.’ (Harman 1990, 667)
As soon as I search for the intrinsic properties of my experience, I only
find the world and its properties or—to put it more exactly—the way the
perceived world seems to be according to my experience. But that is
nothing more than the representational content of my experience.
Transparency has been appealed to, to argue for different and
incompatible views.14 I want to argue that CV is at least compatible with
transparency. The phenomenology of our perceptual experiences is such
that they present themselves as having an “objective purport,” as
13 Transparency was first introduced by Moore: when we try to introspect the
sensation of blue, all we can see is the blue: the other element is as if it were
diaphanous (1922, 25). On transparency, see Tye 2000, Siewert 2003, Stoljar 2004,
Martin 2008.
14 See Martin 2002.
The Core Idea of the Content View (CV)
113
attributing features and properties to the real perceived world we are
facing. That is why CV is, in that respect, phenomenologically more
adequate than adverbialism or sense data theory. On the other hand, it does
prima facie better than direct realism because it can better accommodate
the argument from illusion.15 If our experience is falsidical, it cannot
involve the real presence of (all) the properties we find in introspecting the
experience itself, but it can well falsidically attribute to things properties
they do not in fact have. So such a phenomenology of objective purport
and “direction to the world” is vindicated within CV, but also the
eventuality that our experiences are illusory and may “mismatch” reality is
no longer a puzzle.16
Often transparency is used by the representationalist advocates of CV17
to argue for representationalism, namely, the view that phenomenal
properties are identical to representational properties. But other versions of
CV do not hold that identity18 so I will leave that debate to the side for the
time being and take transparency to be a just fit with the general
framework of CV.
As I have said, BT is not phenomenologically adequate even if taking
perceptions as belief-acquiring did explain the objective purport
experiences seem to have. Take visual experience, for example. What BT
disregards is the relevance of the special way of acquiring information so
phenomenologically salient in visually experiencing, especially insofar as
that special way characterizes visual phenomenology in a way that could
not be explained if perceptual experiences were nothing more than beliefacquisitions—namely, the richness of details or fine-grainedness of the
visually conveyed information.
Assigning content to a visual experience must involve a kind of
content consistent with the fine-grained and profuse way of (purportedly)
“taking in concrete reality” (Crane 2008, 24) that vision introspectively
seems to be. The semantic characterization of perceptual experiences
needs to fit their phenomenological features if it is to explain their
relevance in our first-person access to the content itself. If content consists
15 Advocates of a disjunctivist version of direct realism (see Martin 2006, 2008,
Fish 2009) claim they can accommodate the problem of illusion and hallucination
without giving up direct realism. I will take this proposal into consideration later.
See Chapter Seven.
16 I know things are not so easy because the presentational phenomenology of
hallucinations is still to be explained, but I will discuss the problem later on (see
Chapter Six, II.5). For the time being, these introductory remarks are sufficient.
17 For example, Dretske 1995, Tye 2000.
18 For example, Peacocke 2007, Block 1998.
114
Chapter Three Part I
of correctness-conditions and we access (come to know) such correctnessconditions of our experiences in virtue of them having a distinctive
phenomenal character, then that phenomenal character must be somehow
“witnessed” in content. As we saw, perceptual content is not separable
from its informational origins, unlike the content of beliefs.19 Perceptual
content and the perceptual vehicle are originally entangled.
The difference between a picture and a sentence may be a good
metaphor to render the difference between perceptual content as it is
witnessed in visual phenomenology on the one hand and the content of
perceptual beliefs acquirable on the basis of accessing that content on the
other.20 Perceptual content presents itself as analog, unit-free,21 continuous
and profuse. Perceptual beliefs are ways of “extracting” pieces of discrete
and “digitalized” content from a more basic experiential content, which is
recalcitrant—phenomenologically in the first place—to be identified with
the potential extractions it can give rise to. Extracting pieces of
information in a “digital” form—as Dretske 1981 characterizes the
doxastic form—involves a gain in classification (conceptualizing is
classifying) but an impoverishment in terms of informational richness, a
gain in quality against a loss in quantity of information. Consider an image
as representing a scene. It is neither true nor false in itself; it can represent
a real scene in a more or less accurate way with respect to different visible
properties (colors, sizes, orientation, fidelity in morphological details and
so on). In visual experience, the attended scene presents itself as a topic
for further exploration where infinite details are potentially available for
access. They are within our reach, so to speak.22
Therefore, if we take the picture/sentence analogy seriously, we can
make the first important point about a phenomenologically salient
difference, as a kind of content, between our experiences and our
perceptual beliefs. Perceptual experiences can be more or less accurate
whilst beliefs are either true or false.
19 As Crane (2008, 23) puts the very same point, “It is central to the
phenomenology of experience that what is conveyed to the subject includes its
specific vehicle.” What is conveyed to the subject in experience is the content.
20 Besides Dretske 1981 and Crane 1992, 2011, also Fodor 2007 adopts that very
analogy by taking perceptions as iconic (picture-like) representations on the one
side and perceptual beliefs as discursive (sentence-like) representations on the
other.
21 For example, you do not visually experience a distance in feet or in meters. On
perceptual content being analogous and unit-free, see Peacocke 1986.
22 As Nöe 2006, 422, remarks, “Phenomenologically, the world is given to
perception as available.”
The Core Idea of the Content View (CV)
115
I.5 The Scenario Content Introduced
If representational content consists of correctness-conditions, the correctnessconditions for beliefs are their truth-conditions whilst the correctnessconditions for perceptions are their accuracy-conditions in the first place.
Accuracy is a gradual notion that can do justice to the profuse and
continuous nature of perceptual content as it is conveyed in our concrete
experiences.
The well-known notion of scenario content (SC) articulated by
Peacocke is a way of semantically characterizing perceptual experiences
that does justice to their fine-grained phenomenology. SC is
specified by the ways of filling out the space around the subject which are
consistent with the representational content’s being correct. (Peacocke 1992,
105)
If the surrounding world instantiates the spatial type under which those
“ways” fall, then the content is exemplified and the perception is correct.
That set of ways is to be determined egocentrically by first fixing an origin
(for example, the chest of the perceiver) and axes (for example, the
directions back/front, up/down, left/right) with respect to the origin or
center; then determining for each point within that centered space—
identified by its direction and distance from the fixed origin—“whether
there is a surface there, and if so what texture, hue, saturation, brightness
and temperature it has at that point, together with its degree of solidity”
(ibidem, 106). Once that SC is ascribed to a perceptual experience, the
experience can be semantically evaluated as soon as it is cast into a
positioned scenario, namely, each point of the scenario content is matched
with a correspondent point in the real surrounding environment where the
subject is and its “origins” are located according to real directions, places
and times. A positioned scenario is easily assessable for correctness and
accuracy. For example, a piece of the content it carries could be specified
as follows: “At that point (= at a certain distance and direction from the
origin) of the positioned scenario there is a red solid surface with a certain
brightness, saturation, hue and orientation.” Now, if the real space around
the perceiver is characterized that way at that point, and that point of the
spatial type is instantiated, then the perception is accurate. If not, it is
inaccurate. That content is a spatial type used to semantically characterize
the experience; it is not found somewhere “in the head.” Nonetheless, that
style of content-ascription is phenomenologically apt because it embeds
reference to typically visual properties, as represented from the “point of
view” of the perceiver, with a profuseness and concrete richness of details
116
Chapter Three Part I
typical of conscious visual experiences that without that perceptual content
is reduced to beliefs or abstract concepts. Indeed, the subject does not need
to possess the concepts used to specify the content ascribed to her
experience. For example, the concepts {red}, {saturation}, {distance} and
the like. A fortiori, the subject does not even need to acquire a belief
having these concepts as its constituent in order to have such an
experience assessable for accuracy and semantically evaluable throughout.
Now, SC is not the only layer of content with which experience is
characterizable but it is a basic layer that already shows how a version of
CV can be phenomenologically adequate, saves the virtues of BT and
avoids the inescapable problems hinted at above at the same time.
As we will see later, there are other layers of perceptual content that
are more semantically articulated than SC insofar as they involve explicit
reference to “discrete” objects as having types of properties, as involving
“thicker” recognitional acts and so on. But for now I am mostly interested
in showing how CV can meet the main general desiderata a satisfactory
theory of perception should meet, independently of how other more
specific issues can be detailed and filled.
PART II
SOME PRIMA FACIE VIRTUES
OF THE CONTENT VIEW
II.1 Distinctive Features of States with Intentional Content
In fact, CV is too rarely defended as such against its alternatives.1 Rather,
it is articulated in specific ways, and some sub-options within CV are
argued for against others. For these reasons it seems opportune to
preliminarily argue for its plausibility, especially insofar as an alternative
view—the disjunctivism-cum-naïve-realism (DJ-NR)—is affirming itself
and challenging the very basic assumption of CV, namely, the idea that
perceptual experiences are contentful states.2 According to DJ-NR,
perceptual experience is relational in a way experience could not be if CV
were true. In that view, experiences, when veridical, are relations having
the world itself as a constituent rather than being representations of it;
when falsidical, they are events of another kind that do not share anything
with veridical experiences apart from their subjective indistinguishability
from them (each from its subjectively matching counterpart). So veridical
cases are genuine relations to the surrounding world, which are then
“directly manifested” rather than represented through them, and do not
share anything essential or important with illusions and hallucinations, not
even their alleged content. That is the well-known version of DJ-NR
1 For example, Evans 1982 and Peacocke 1992—the champions of CV—never
defend the truth of CV as such. A relevant exception is Siegel 2010.
2 See Martin 2002, 2004, 2006. Actually, a naïve realist does not need to deny that
experience has content—it will suffice for her to hold that the notion of content is
not so relevant in individuating and characterizing experiences. On the contrary,
according to CV the content of a perceptual experience is an essential feature
making the state what it is. Siegel 2010 argues that for that reason naïve realism is
compatible with CV; however, embedded in NR, the notion of content becomes
redundant. It ceases to be a key-notion to say the least.
118
Chapter Three Part II
(Martin, Fish, Brewer) anyway.3 Other versions of DJ-NR are available
but I will deal with them later on (Chapter Seven).
Of course, in order to defeat that global challenge, it will be necessary
to develop a detailed version of CV that is able to tackle the criticisms
presented by the advocates of the recently revived naïve realism. But since
CV is perhaps slowly ceasing to be the mainstream view, some
preliminary arguments at least for its prima facie plausibility need to be
introduced.
The first reason for taking CV as promising has already been recalled.
Perceptions have intentionality, namely, they are states directed-upon
worldly objects. Representing is a way of being about a represented target
in that it represents the target as being a certain way.
Intentionality as a property of representations has been characterized as
having three main features. First, the power to misrepresent, which would
occur when perceptions are incorrect or mistaken (their correctnessconditions are not instantiated).4 Second, the aboutness itself, insofar as an
intentional state refers to some object or condition, is about something
indeed. This is also a feature that essentially characterizes perceptual
experiences—they are about objects or circumstances in the environment.
Third, the aspectual shape. In being about an object O, a representation of
O always represents it (as being) a certain way, as F or G. Again, this is
also a typical feature of perceptual experience; for example, when I see an
object and it looks some way to me (red, square, big), that would be the
way my experience represents the object as being if experience was an
intentional-representational state. As Dretske 1995, 31, puts it:
“Experiences are about objects, but one cannot experience an object
without experiencing it under some aspect.”5 So perceptual experience has
intentionality, and these three fundamental features of intentionality would
be well explained by appealing to the notion of intentional content.
Crane likewise identifies (2010, 86) three main ideas that make it
necessary to introduce the notion of content with respect to perceptual
experience: aspect, absence, accuracy. By “aspect,” he means what we
have just called the aspectual shape, so that seeing something involves
something looking a certain way or its being visually given under some
aspect. “Absence” refers to the possibility of an intentional state being
3 See Martin 2004, 2006, 2008; Fish 2009; Brewer 2008.
4 See Chisholm 1957.
5 If I perceive an object, it looks some way to me. An object can look some way F
to me if it is experienced by me as F. But experiencing an object as being a certain
way F seems to require representing it as being that way, especially if we consider
that S can experience O as F even if O is not F. See also Tye 2011, 185ff.
Some Prima Facie Virtues of the Content View
119
about an object or a state of affairs that does not actually exist.6 For
example, I can desire something that is not there so my state has an
intentional object that does not exist. I can believe that a state of affairs
exists but that possible state of affairs is not ‘satisfied’ by reality, and so
on. Absence is intimately connected with the power to misrepresent
introduced above. If a state represents that something is the case and that
fact does not hold, the absence of the represented fact is what makes the
state a misrepresentation of the way things are. Visual experiences can
have absent objects, for example, in hallucinations or illusions.7 Last but
not least, accuracy is a salient feature of experiences. So, aboutness,
(potential) absence, aspect and accuracy are salient features of both
perceptual experiences and contentful, representational states. A
straightforward way of accounting for that striking parallelism is taking
perceptual experiences to be representational states. Nothing other than a
representation seems to have in itself these distinctive features of
(possible) absence, (possible) accuracy and (necessary) aspectuality.
II.2 Perceptual Experience and Accuracy
That visual experiences are assessable for accuracy means that they can
more or less match or mismatch reality, so their accuracy is “measured” by
the world.8 How can that happen if not by experiences being about reality
and representing it in a certain way?9 If that argument does not prove the
truth of CV, at least it seems to be a natural conclusion and a very
plausible inference to the best explanation.
A possible reply could be that many other processes or activities are
more or less accurate without being representations or contentful states, so
being assessable for accuracy does not entail any possession of
representational properties. For example, an occurrent example of digestion
can be more or less accurate with respect to a standard of “digestive
efficiency” determined by the natural function, evolutionarily fixed, of the
dedicated mechanism, but no-one would take a case of digestion to be
thereby a representation. But that objection is resistible because the
6 See Brentano 1874, Husserl 1913/70, Searle 1983.
7 Hallucinations lack a worldly object at all; illusions can represent a real
perceived scene in such a way that it appears that a certain object is part of the
scene but actually that object is something else. So a partial detail presented in an
illusory experience can also be “about” non-existent objects, not only a
hallucination.
8 In Searle’s words (1983), perceptions have a world-to-mind direction of fit.
9 An argument along these lines can be found in Siegel 2010, 30ff.
120
Chapter Three Part II
accuracy involved in perceptual experience is not just a kind of generic
bio-functional well-working; rather, it involves a positive matching
between the experience and the environment. In other words, there is a set
of circumstances, or at least a cluster of environmental properties, the
experience can match (veridical experience), partially match (illusion) or
totally mismatch (hallucination), all of which involve the idea of a
correspondence or a mapping-relation between two domains that is more
than an appropriate reaction to some external condition. Accuracy
concerns such a correspondence whose accessibility is a matter of how
much of a domain is instantiated in the other domain.
Therefore, as Siegel 2010 rightly remarks, the best explanation for the
pre-theoretical distinction between partially veridical, falsidical and
completely veridical experiences is that experiences have accuracyconditions in virtue of which they can be more or less accurate, where the
conditions of such accuracy are indeed the contents. Accuracy of
perceptual experience has to do with experiences having the power of
being veridical, partially veridical or falsidical, and these ordinary
characterizations employ intentional and clearly semantic notions. In
addition, the accuracy-conditions of perceptual experiences are conveyed
to the conscious subject and have a phenomenological salience to the
effect that the properties presented in experience make it accurate only if
they are instantiated in the environment. Other functions like digestion can
be normatively evaluated as correct/incorrect and even as more or less
accurate, but they do not involve the very same properties that need to be
instantiated in reality in order for the very episode to be accurate. That is
the difference between a response to an environmental condition, which is
also normatively evaluable as more or less accurate, and a representational
response to an environmental condition whose accuracy is a sui generis
accuracy consisting of a distinctive kind of correspondence where a certain
spatial type may or may not be instantiated by a worldly token.
A set of accuracy-conditions conveyed to the subject through her
experience’s presenting properties, which are instantiated when the
experience itself is accurate, is a content. So perceptual experiences have
accuracy-conditions conveyed to the experiencer in such a way that they
cannot be other than contents. All that is enough to render CV a very
promising working hypothesis at least.
I will close the third chapter by arguing that many fundamental
properties of the ordinary ascriptions of episodes of vision and visual
experiences—analyzed at length in the first chapter of this work—fit very
well with CV, so that CV would mostly “save” our pre-theoretical
intuitions about seeing and visually experiencing insofar as these intuitions
Some Prima Facie Virtues of the Content View
121
are implicitly embodied in the ordinary perceptual vocabulary we adopt in
everyday discourse. Not only does CV seem to be phenomenologically apt
then, it also seems to be apt at vindicating the manifest image of us as
subjects of experience capable of being in sensory contact with our
environment.
Normally, when S sees an object O, O looks some way to S, say F. But
O looking F to S does not entail that S comes to believe that O is F. Seeing
is not believing or coming to believe. Nonetheless, in normal conditions
and in absence of collateral knowledge, when a seen object O looks F to
us, we do tend to come to believe that O is F. When O is F, the episode of
seeing occurs in normal conditions (without abnormal causal deviations,
etc.) and our visual apparatus is working well, we justifiably come to
believe the true fact that O is F, so we come to know that O is F. In other
words, we see that O is F. It may also happen that we come to believe by
visual means that O is F but O is not F; for example, the lines are not
unequal as they appear to be according to the illusory experience we
falsidically trust or endorse, so we come to acquire a false belief by visual
means. In cases of both knowledge and false belief, we acquire a
perceptual belief through an episode of seeing something through a visual
perception, but this visual perception is not identical to that which
acquiring it often and normally gives rise to. We could say that BT
wrongly equates seeing with seeing-that (or visually coming to believethat).
Furthermore, not only can there be seeing without seeing-that, there
can also be seeing without seeing-as as well as seeing-as without seeingthat.10 In fact, S can see O without recognizing O as a kind of thing (as an
F), even if it is true that seeing an object normally prompts certain
cognitively “thick” recognitions, namely, certain acts of seeing-as. So
seeing is not necessarily either doxastic or recognitional but it does
normally bring about perceptual beliefs in systems having beliefs and
10 The relation of entailment is: seeing-that (O is F) → seeing (O) as (F) → seeing
(O), but it does not hold the other way round. No visual beliefs without visual
recognition, no visual recognition without vision, but nothing prevents that there
are episodes of vision without visual recognition as well as episodes of recognition
without belief-acquiring and related acts of conceptualization. The absence of
entailment defeats both BT and the conceptualist CV. Let us remember, for the
sake of clarity, that being a conceptualist does not entail being a belief theorist. For
example, McDowell 1994, 1998 holds that perceptual content is conceptual but
perceptions are not belief-acquiring. Rather, by acquiring a perceptual belief based
on a perception, the subject just endorses the very same content of her perceptual
experience so that those contents need to be the same nature (i.e., conceptual).
122
Chapter Three Part II
conceptual abilities, and recognitional acts in systems having beliefs and
conceptual abilities as well as in less sophisticated cognitive systems not
endowed with conceptual capacities. In other words, not every recognition
is a conceptualization, so recognitional seeing is more basic than visual
belief-acquiring, and object-seeing is more basic than seeing-as.
According to CV, seeing-episodes are representational episodes.
According to a certain non-conceptualist version of CV, such episodes are
more basic acquisitions of information than beliefs, and they do not need
concept-possession. Perceptual content can be—and it normally is—an
object for further extractions so it can be material, so to say, for “takings,”
recognitions, and finally for beliefs.
We saw in Chapter One that object-seeing ascriptions are extensional
and transparent ascriptions of an episode of discrimination involving the
visual apparatus and caused by the discriminated environmental object
itself. So the seen object is causally, contextually determined. Now,
cognitive ascriptions are not transparent but cognitively loaded; objectseeing as such is non-cognitive, at least in the broad sense that whether
you see an object or not does not depend on whether you cognize it in a
specific way. It is enough that you discriminate it some way, no matter
what that is.
As we saw in Chapter Two, within a perceptual experience, there may
well be object-seeing without noticing but any perceptual experience
involves some noticed objects at least that look some way to the subject.
In any case, experimental evidence shows that object-seeing does not
necessarily involve either noticing or recognition so the extensionality of
its ordinary ascription is empirically vindicated. What we see does not
depend on what we believe we are seeing, nor does it depend on what we
recognize in what we see—it does not even depend on what we notice.
Seeing-that ascriptions are ascriptions of fully-fledged propositional
attitudes instead so they involve respective ascriptions of conceptpossession and beliefs. They are referentially opaque and factive to the
effect that seeing that O is F is a case of coming to know by visual means
that O is F. Anyway, seeing-that ascriptions cover the successful cases of
the more general process of perceptual belief-acquiring, which may well
be a cognitive failure, namely, the acquisition of a false belief. Again,
everything fits very well with CV, according to which perceptual
experiences have a content of their own that may or may not give rise to
perceptual beliefs that conceptualize and endorse some aspects of that
content by casting it into a propositional structure appropriate for it to be
believed.
So, “seeing facts” is no more than seeing objects as having certain
Some Prima Facie Virtues of the Content View
123
properties and relations and—truly and justifiably—coming to believe that
a seen object has this or that property or entertains such and such relations
with other seen objects. Seeing objects, properties and relations puts us in
a position to come to know facts concerning those objects, properties and
relations, but the episodes of one kind are not identical with episodes of
the other kind.
We also discussed seeing-as ascriptions in Chapter One. Seeing-as
when successful is recognitional seeing. It is neither transparent nor
factive and is normatively evaluable as right or wrong. It exhibits
correctness-conditions measured by the way the world is arranged. Even
when unsuccessful, seeing-as presupposes the possession of a positive
recognitional capacity concerning a certain property so it is recursive and
involves a categorization of the seen object. That is not conceptualization
nor does it involve belief-acquiring; rather it is the subsumption of a token
under a type and does involve a “taking,” a certain cognitive stand we
could call a point of view on the perceived object. Such a cognitive stand
is a belief-like state but not a belief since it does not necessarily draw on
inferential abilities and conceptual capacities. Actually seeing O as (an) F
is the cognitive outcome of a perceptual episode, which is a condition for a
perceptually acquired belief that O is F but does not entail that belief. It
amounts to seeing an F in O thanks to the exercise of a recognitional
capacity prompted by the visual appreciation of O’s sensible profile.
Now, according to CV, seeing O as (an) F is representing O as being an
F thanks to the representation of O’s sensible profile in the first place. A
sensible profile is made out of SCM-properties, so representing an object
as having a certain sensible profile is representing it as having a complex
of visible properties like colors, size, shape, spatial properties, orientation,
solidity, texture and the like. In perceptual seeing-as, any thicker
categorization (e.g., seeing O as a [pig]) entails the appreciation of that
complex of SCM-properties typically exhibited by examples of Fs, so it is
the recognition of a visual type in the first place.
Couched in terms of CV, in order to represent O as an F when F is not
a visual property, S’s visual experience needs to represent O as having
certain visual properties or exhibiting a distinctive sensible profile,
recursively recognizable just by sight.
II.3 Content View and the Ordinary Semantics
of “Seeing” and “Looking”
In Chapter One, I discussed some relevant “experiential” vocabulary,
namely, the logical behavior and the semantics of “looks”-/“seems”-
124
Chapter Three Part II
/“appears”- ordinary ascriptions. I argued for the same idea I have just
stated here: In order for O to look F when F is not an SCM-property, O
must look G in the first place, where G is a complex of SCM-properties
that individuates a certain visual profile. Provided that in CV, O looking F
to S means S’s visual experience is representing O to S as being F, if we
are to visually represent properties other than SCM, we would be able to
do it by virtue of visually representing SCM-properties in the first place.
Therefore, we can say that SCM-properties are the basic contents of visual
experience. Actually, the scenario content introduced above is made out of
those “thin” properties: colors, spatial relations, distances, shapes and so
on.
Therefore, there seems to be both “thick” properties and “thin”
properties involved in perceptual content. Their reciprocal relationship has
to be articulated; it is to be inquired whether “thick” represented
properties—and which ones, eventually—may be considered as properties
represented in perception instead of being represented by the subject just
in virtue of the subject’s having a perception in which only “thin”
properties are represented. I will leave this important issue aside though.11
In any case, insofar as seeing O as an F is (pre-conceptually and predoxastically) taking something as a token of the type F. It amounts to
representing something as being an F so provided perceptual experiences
involve seeing-as episodes, their nature is well consistent with CV.
In addition, CV also could explain the intimate entanglement between
the different uses/meanings of “looks.” If perceptual experiences represent
the world as being a certain way, it is natural that what we have called an
“epistemic look” and a “phenomenological look” are inextricably blended
with each other. The phenomenal character of experience “maps” its
representational content, as transparency shows, so that as soon as we
attend to our own visual experience, in attempting to find visual “looks,”
we end up finding ways perceived things look to be according to our visual
experience. There are phenomenal properties in experience but they cannot
be picked out independently of the way things are represented to be when
the properties are instantiated. A visual experience presenting O as looking
F in fact is prima facie evidence that supports the proposition that O is F,
because F is the way O looks to be in experience.
Now I can go back in terms of CV to an argument I put forward with
respect to the different uses of “looks.” Any way of describing one’s own
visual experience by phenomenal “lookings”—for example, “looks F”—in
11 See Chapter Six, Part I, where I explain why I am more sympathetic with the
liberal view.
Some Prima Facie Virtues of the Content View
125
non-representational terms must finally rest in a hypothetical situation
where O looks to be F so O is represented as being F. For example, when
we say the penny “looks” elliptical even if we know it is circular and it
only looks to be elliptical, still we are saying that another experience
presenting the penny from a “frontal” point of view would represent the
penny as being elliptical. When we say that the big tree in the distance
looks smaller than the small tree nearby, we are saying that in another
experience sharing with the actual one some phenomenal properties
(perhaps not the “focal” ones), the distant tree would be represented as
smaller than the other, namely, such a hypothetical experience would
incline S to take it that that tree is smaller.
In short, when visual properties are somehow instantiated but not
represented as being had by perceived things, in fact these properties are
still “potential representational properties” of other experiences we need to
counterfactually refer to when we adopt that sophisticated phenomenal
introspective stand toward our ongoing experience.
Within the framework provided by CV, the related dependence of
“looks F” on “is F” I argued for in Chapter I is also explained. Something
can “look F” to S even if it is not believed by S to be F only if S is able to
recognize Fs just by looking, so only if S is able to correctly apply [—is F]
to seen objects. When O is claimed to look F even if it is known not to be
such, what is claimed is “my experience is, at least with some relevant
respect, as if I was experiencing an F.” So, visually experiencing
something that is F as being F works as a “standard experience” for
ascribing cases of “just looking so” and for characterizing the experience
itself with the phenomenal property [F]. Successful cases of seeing and
seeing-as are paradigmatic in ascription of visual experiences so that the
ascriptions of an experience with a certain phenomenology and a certain
content is asymmetrically dependent on ascriptions of veridical and
successful experiences.
That asymmetry between ascriptions of veridical experiences and
noncommittal “veridicality-neutral” ascriptions of experiences is explained
within CV by considering that the accuracy-conditions of an experience
cannot be specified other than normatively; i.e., other than referring to the
circumstance in which such conditions would be fully satisfied (and the
related contents would be instantiated). Just as belief-contents are
specified by the circumstances that would be obtained were the belief true,
so perceptual contents are individuated by the way the world would be
were the perception accurate. Meaning is not truth but its individuation is
truth-dependent.
126
Chapter Three Part II
Accordingly, the phenomenology of an experience—with its
introspective looking-properties—is specified by implicit reference to the
case in which an experience with exactly the same phenomenology was an
accurate presentation of that purportedly experienced scene.
Concluding Remarks
In the first part of this chapter, I outlined the core idea of CV. I argued that
a non-conceptualist version of CV can embed the virtues of BT without
having to face its insurmountable difficulties. CV can be compatible with
the experimental evidence speaking against the identity between
perception and belief-acquiring, and can be immune to the main
philosophical arguments showing the structural difference between beliefcontents and perceptual content. In particular, CV does not exhibit the
phenomenological inadequacy that afflicts BT.
I briefly introduced the fundamental notions of phenomenal character
and representational content, the two basic elements of perceptual
experience according to CV. Perceptual experiences present themselves as
having certain phenomenological features so a good theory should at least
try to be compatible with these manifest features. Two important features
of experience that are graspable by introspection are transparency and
richness of details. Both features fit quite well with the non-conceptualist
version of CV that I will articulate in detail later in this work.
In considering visual content, I briefly introduced the idea of an SC,
originally proposed by Peacocke, and I argued that SC is a semantical
characterization of perceptions that is phenomenologically apt, at least as a
first basic layer of perceptual content. More generally, CV is in a position
to make room for both transparency and richness of details so CV does not
turn perceptual introspection into a systematically misleading activity. In
addition, a style of ascription such as Peacocke’s SC allows us to assign
“iconic” and gradual accuracy-conditions to perceptual episodes rather
than the “yes-or-no” truth-conditions belief-contents consist of, so it
allows us to posit an unstructured content that can very well be both nonconceptual and profuse in nature.12
In Part II, I put together some putative virtues of CV, which are so
seldom argued for.
First, I advanced some substantial reasons for taking CV as a
promising view. Aspect, (possible) absence, aboutness and accuracy are
12 As we will see, perceptual content can well be structured in a certain sense to be
clarified even if it does not exhibit the semantic structure of a proposition.
Some Prima Facie Virtues of the Content View
127
features of perceptual experiences that, taken together, strongly support the
idea that perceptual experiences have representational content.
In particular, experiences are assessable for accuracy and such
accuracy-conditions are conveyed to the subject in the experience in such
a way that experience can match or mismatch reality. But only a contentful
mental episode can have correctness-conditions involving the instantiation
in reality of properties presented to the subject in the very episode.
In addition, I argued that CV “saves” our pre-theoretical intuitions
concerning the possibility that our perceptual experiences are veridical,
partially veridical or totally illusory. That idea may be given a sense only
if we take perceptual experience to be able to partially match, totally
match or totally mismatch the surrounding world, where the matchingrelation is throughout representational.
Finally, I argued that CV is well compatible with the pre-theoretical
intuitions conveyed by the ordinary ascriptions of “seeing”-episodes and
experiential predicates like “looking”/“seeming”/“appearing” analyzed at
length in Chapter One. So, CV is consistent with the manifest image of
perception and experience as is implicitly shown and witnessed in
everyday discourse.
The difference between seeing-O and seeing-that maps the difference
between perceptually representing and coming to believe by visual means.
Seeing-that is a (true) doxastic representation based on a visual nonconceptual and non-doxastic representation. For CV, O looking F to S
amounts to S’s experience representing O as being F. Seeing-that is not
necessary for seeing-O, even if quite often seeing an object O as having a
property F makes us acquire the belief that O is F.
Seeing-as or recognitional seeing is a middle-way representation that is
not a belief or a fully-fledged propositional attitude but involves a
cognitive stand toward a perceived object consisting of representing the
object as falling under a type. When F is not an SCM-property, seeing O as
an F entails that the related experience is a perceptual representation of a
complex of SCM-properties composing a unitary and distinctive sensible
profile. Seeing-as is necessary for seeing-that but not the other way round.
There is seeing without recognition, there is recognition without belief.
I also argued that CV explains why the phenomenal use and the
epistemic use of “looks” are so inextricably meshed. The phenomenal
character of our perceptual experience maps its representational content,
so introspecting an experience in search of “phenomenal looks” turns out
to be just finding those properties that perceived things look to have, those
128
Chapter Three Part II
ways these things look to be.13
In the next chapter, I will articulate in more detail a specific version of
CV with the aim of showing later how CV can well face the challenges by
the recent disjunctivist revival of naïve realism.
13 But for a partial revision of that acquisition, see Chapter Four.
CHAPTER FOUR
THE CONTENT VIEW,
ARTICULATED AND DEFENDED
Introduction
This chapter is in two parts. In this chapter, I will articulate and defend a
certain version of the content view that can address the problems typically
raised by its opponents.
PART I
LAYERS AND COMPONENTS
OF PERCEPTUAL CONTENT
I.1 Beyond the Scenario Content
In the previous chapter, I introduced the notion of non-conceptual scenario
content as a phenomenologically apt characterization of perceptual
experience in semantic terms. Nonetheless, SC cannot capture other
distinctive elements of perceptual content as well as the correlative
phenomenological richness of perceptual experiences. Indeed, the very
same SC or “way of filling out the space” around S may characterize two
of S’s experiences that differ in terms of represented properties as well as
in terms of phenomenology. This just means that the position of SC is
necessary but not sufficient insofar as SC does not exhaust the semantic
and phenomenal richness of perceptual experience. Two examples by
Peacocke 1992 make this clear. Consider a square rotated 45 degrees in
such a way that its four angles are located up, down, left and right with
respect to your point of view (see Figure 1). As Palmer 1983, 292 notes,
that figure is often immediately seen as an upright regular diamond rather
than as a tilted square, but it can also be seen the second way through a
sort of voluntary perceptual “switch.”
Figure 1 Square/Regular Diamond
The square/regular diamond was first cited by Mach 1897, who considered
it a case of an ambiguous figure.1 You can see ambiguous figures
1 Also Palmer 1983 and Macpherson 2006 take it as an example of ambiguous
figure.
Layers and Components of Perceptual Content
131
alternatively in two different ways due to a sudden change known as a
Gestalt switch, which happens due to the saccadic nature of our visual
perception. In any case, the square/diamond figure can be seen two ways
such that these ways are not taken by the perceiver as two rotational
variants of the same shape but rather as two instances of different shapes, a
(regular) diamond or a (tilted) square. So, if the same SC is compatible
with different perceptual contents, SC cannot exhaust the whole intentional
content of perception.
According to Peacocke, whether we see the figure as a square or as a
diamond depends on which symmetries we respectively perceive. When
the figure is perceived as a regular diamond, the perceived symmetry is
about the bisectors of its angles. When the figure is perceived as a tilted
square, the perceived symmetry is instead about the bisectors of its sides.2
Different symmetries are immediately considered/disregarded so the visual
experience changes accordingly in both content and phenomenal character.
Another example is related to spatial grouping. Consider a twodimensional array made of nine ordered points forming a sort of square
(see Figure 2).
Figure 2 Rows or Columns?
Now, you can alternatively see the array as a set of three rows or as a set of
three columns. Provided the array in view is the same in both cases so the
SC is fixed, such a shift also involves the presence of an additional
perceptual content beyond SC.3 Peacocke called this additional content
proto-propositional (PPC). Proto-propositions are made out of individuals,
2 For a detailed and critical discussion by Peacocke, see Macpherson 2006.
3 Seeing certain figures as exhibiting certain symmetries or seeing a set of points
as part of a column or row makes a difference in the nonconceptual perceptual
content as well as in the phenomenology of the experience. Indeed, those
differences do not rest on the exercise of concepts like [symmetrical to] or
[column]. Noticing a certain symmetry does not require the conceptualization of
something as symmetrical, grouping points into rows (horizontally) does not
require the conceptualization of something as a row, so no concept-possession is
needed.
132
Chapter Four Part I
properties and relations,4 and a perceptual experience with a certain PPC
somehow “attributes” some property or relation to individuals having that
PPC. For example, in a certain visual experience, a perceived individual
may be represented as [square], [equidistant from], [parallel to] and the
like. So that layer of perceptual content, even if it is non-conceptual and
not fully structured (it is not propositional), involves a segmentation of the
global scene into individuals and the properties they are represented as
having, and/or relations they are represented as entertaining. So the
examples above are examples of experiences with the same SC but
different PPC. PPC is essential to recognition, or seeing-as. Consider
seeing two objects as square despite their difference in terms of size,
orientation, color and the like. In this case, two very different SCs may
share a same (piece of) PPC: [square]. So two different ways of filling out
the space around S may be associated with the same way of perceptually
representing the perceived objects constituting the scene; likewise, the
same way of filling out the space around S may be associated with two
different ways of representing the perceived scene: [square]/[diamond]. In
short, you can have one SC with different PPCs, or one PPC with different
SCs. Therefore, SC and PPC aren’t independent but neatly distinguished.
Perceptual experience has a PPC by virtue of having a SC, but having a
certain SC does not determine alone which PPC the experience has (nor is
it the case the other way round).
Not only does PPC contribute to determining the representational content
of an experience, it also contributes to determining the phenomenology of the
experience. Seeing something as a set of rows and seeing something as a
set of columns are two phenomenally different visual experiences despite
their sameness of SC, for example. As we will see, there is an intimate
relation between phenomenal character and representational content so it
is no surprise that a change of one involves a change of the other.
On the one side, visual phenomenology is profuse, continuous and
fine-grained; on the other, it presents the subject with objects experienced
as having certain properties so visual phenomenology also exhibits a
“discrete” aspect. Visual experience is not just as if detailed global scenes
were presented to the subject; it is also as if objects contained in the global
scene were presented to the subject as being certain ways and as
4 Such contents, made out of individuals, properties and relations, are usually
labeled as “Russellian.” Indeed, a Russellian proposition is a proposition made out
of individuals, properties and/or relations (instead of “Fregean” constituents like
concepts, senses or other). On that, see Chalmers 2006.
Layers and Components of Perceptual Content
133
entertaining certain relations with each other.5
I.2 Proto-Propositional Content and Seeing-As
Now I want to posit a parallel between Peacocke’s notion of PPC and the
notion of seeing-as I analyzed in Chapter One (Part IV). I want to argue
that such a layer of perceptual content captures and explains what is
involved in the ordinary ascriptions of seeing-as episodes. Likewise I will
argue that the necessity of introducing PPC between SC and the
propositional-conceptual content of perceptual judgments and beliefs
explains and parallels the existence of seeing-as ascriptions as an
intermediate and irreducible level between seeing and seeing-that.
Firstly, PPC is assessable for truth. In perceiving something as a
square, as symmetrical to something, as equidistant from something, you
may be right or wrong; in a true or false state according to how the world
is arranged.
Secondly, PPC needs to be introduced as an intermediate kind of
content between the SC and the conceptual content of our perceptual
beliefs. In particular, if explaining the acquisition of observational
concepts like [square] or [symmetrical] is not to be circular, such
conceptual mastery must rest on paradigmatic experiences having a
corresponding PPC.6 I cannot acquire mastery of the concept [square]
without having some preceding experiences with the very content
[square], which cannot feature in SC for the reasons hinted at above.7 So
PPC also grounds and makes conceivable the transition from experiences
to perceptual judgments and beliefs, just like seeing-as somehow mediates
the transition from simple object-seeing to epistemic seeing-that.
Thirdly, PPC is of the greatest importance for recognition, cognitive
maps and spatial reasoning because recognizing the same individual,
5 As Chalmers 2006, 110 remarks, “Phenomenology of vision seems to present a
world that is carved into objects at its joints. One does not simply perceive a
distribution of mass and color: one perceives objects on top of other objects, each
of which may be articulated into objectual parts.”
6 “[...] if we are to have a non-circular and individuating account of mastery of this
perceptual concept [straight], that mastery must be related to some feature of
experience which does not have to be explained in terms which presuppose
possession of the concept.” (Peacocke 1992, 121). On the argument from
circularity, see Crane 1992, Speaks 2005, Bermudez 1995, 2007.
7 I can accurately experience an instantiated way of filling out the space around
me, which involves the presence of a square, without perceiving that region of
perceived space as a square.
134
Chapter Four Part I
place, relation or property over time and across different contexts of
presentation is fundamental for memory and explains the intelligent
behavior of animals not credited with concepts, beliefs and inferential
abilities. It would be impossible to retain in one’s memory many global
SCs and compare them to some aspects, or parts of them, in order to plan
actions or produce integrated representations of the environment, but it is
quite possible to identify and re-identify two individuals over time on the
basis of perception of some of their concurrent properties and recognition
of them on different occasions. Therefore, a segmentation of the scene into
discrete elements like individuals, properties and relations is a
precondition for perceptual recognition, memory and spatial reasoning.
Thus, PPC is typically recognitional and must be posited not only to
explain our acquisition of observational concepts but also to explain the
intelligent behavior of non-conceptual beings.
These are just the basic distinctive features of seeing-as episodes.
Seeing-as episodes do not involve concept-possession or conceptualization
so they are non-conceptual, just like states with PPC.
Seeing-as is normatively evaluable as right or wrong and it is
essentially recognitional. Indeed, it is the positive exercise of a recursive,
positive recognitional capacity, which entails the general sensitivity to a
certain type. Just as a proto-proposition attributes a property to an
individual and so brings the particular under a type—the experience
represents an O as F; for example, one line as symmetrical with another—
a seeing-as episode takes an object as being of a certain type, so it is
categorical, or category-involving. A possessed category is a disposition
that can be recursively prompted, or activated, by the vision of a certain
scene but it is not necessarily entailed by the vision of the scene as such. I
can see two symmetrical lines without seeing them as symmetrical. So,
given an SC, there can be different PPCs. If PPC is equated with the
content of seeing-as episodes, given a seen scene, there can be different
ways of categorizing its more basic content—say its SC—according to
what is noticed, what is recognized, according to how the seen scene
immediately and pre-doxastically “strikes” the subject.
Seeing-as episodes, which typically have PPC, involve the cognitive
act of bringing a perceived particular under a certain pre-conceptual
generality. Seeing-as contents are therefore general but still nonconceptual and exhibit a certain “bringing-under” structure, but still they
are not propositional. Thus, it is plausible to think that proto-propositional
content is recognitional insofar as it is the typical content characterizing
seeing-as episodes. Seeing-as episodes have PPC contents.
There is but one important asymmetry between seeing-as as I have
Layers and Components of Perceptual Content
135
characterized it and proto-propositional perceptual states as Peacocke
characterizes them. Peacocke invokes perceptual PCC only with respect to
SCM-properties and relations,8 like [curved], [distant from], [parallel],
[bigger than], [diamond-shaped], whilst in an episode of seeing an O as an
F, the F-category may well be “thicker” than that, like, for example,
[duck], [rabbit], [lemon], [prey].9 In my discussion of seeing-as, I also
argued (1.4.3) that in order for S to see an O as F when F is not an SCMproperty, S must visually appreciate O’s sensible profile in the first place.
Namely, S must see O as having a certain typical and unified cluster of
SCM-properties. So there is a “thin” seeing-as and a “thick” seeing-as that
presupposes the first, and perceptual states with Peacockean PPC can only
be equated to “thin” seeing-as episodes.
Yet the issue about the nature of that layer of PPC is one thing, the
issue about the kind or range of properties that could feature in such PPC,
is another. Peacocke holds as a substantive additional thesis that only
SCM-properties can feature in visual content,10 but his position of PPC as
8 By SCM-properties, I mean Spatial-Chromatic-Morphological properties; i.e.,
strictly visual properties.
9 On the issue of which properties can be represented in perception, there is a
“conservative” view (McGinn 1982, Tye 1995, 2000, Dretske 1995, Price 2011)
according to which only low-level or thin properties can be perceptually
represented (those I have called SCM), and a “liberal view” (Siegel 2006, 2010,
Prinz 2006, Bayne 2011) according to which also thick or high-level properties
like [chair], [dog], [table] can be perceptually represented. I am sympathetic with
the liberal view: genuinely perceptual seeing-as episodes can involve thick
categorical contents provided that the thick property—[table], [chair], [my
mother], [animal], [something I have already encountered], [something edible] and
the like—is categorized thanks to the immediate appreciation of its typical sensible
profile (a type-complex of SCM-properties), and the association between that
sensible profile and the thick category that is established by perceptual learning,
without any inference or reasoning involved: once a new perceptual scheme is
stabilized and becomes an immediate recognitional disposition “just by sight,” the
recognized thick property is to be considered a genuine part of the perceptual
content. So, SCM-representations are systemic and the relative mechanisms are
wired-in whereas perceptual representations of thick contents are typically
dependent on individual learning and on the perceiver’s history: they are cognitive
exploitations of low-level representations based on the appreciation of a visible
profile as a type associated with a thick property.
10 If one holds that only SCM-properties can feature in PPC, then “seeing-as F”
ascriptions where F is not an SCM-property – for example, seeing O as a rabbit, as
a face, as prey, as food, as the same I met before – would not be literal ascriptions
of seeing-as episodes but, rather, metaphorical ascriptions amounting to ascriptions
of “taking-as” cases, which have nothing essential to do with perception.
136
Chapter Four Part I
a necessary layer of perceptual content is independent of this additional
thesis. For the time being, let us say that PPC is necessary to do justice to
both phenomenology and representational powers of perceptual experience,
regardless of whether one thinks that in perceptual experience only
properties like [red] and [square] can be represented or that properties like
[being a rabbit] or [being a lemon] can also be represented. In short, be it
thin or thick, perceptual content must involve PPC, a kind of content that
fits very well with ordinary ascriptions of “seeing-as” episodes.
I.3 Scenario Content and Object-Seeing
In Chapter One (Part II), I spelled out a set of conditions for a subject to be
truly ascribed an episode of seeing something. Among those conditions
was the subject’s phenomenologically salient discrimination of the seen
object from the surrounding environment. Let’s call it the discrimination
condition (DC). DC requires that the seen object must look some way to
the subject so that it can be discriminated from its surroundings by virtue
of the way, or ways, it looks. DC is a very basic requirement. For example,
you can well see a square by virtue of the square looking some way to
you, but in order to see the square, you do not need to see it as a square.11
You just need to discriminate it somehow from the environment; for
example, appreciating certain contrasts and boundaries by being visually
sensitive to their contours or by some other way.12
Now it is arguable that DC can be satisfied just by reference to an
experience having a certain SC. Indeed, an experience with a certain SC
presents the subject with a certain volume of surrounding space—dense
and with discriminable features. For example, SC involves the
representation of a certain surface at a certain point of the objective space,
with a certain color, orientation, brightness, saturation and so on. The
subject can well be credited with seeing an object by virtue of seeing its
surface by locating and discriminating it from the surrounding space.13 In
11 An “object” you can see need not be a chair or a table. If a neutron stream were
visible, such that it made a visible path, then you could see a neutron stream even
if you had no idea that you were seeing a neutron stream.
12 Suppose somebody throws a square object and its trajectory crosses your visual
field. You see the speedily moving object without being able to appreciate its
squareness. Nonetheless, you have seen the square object by virtue of having
discriminated it from its surroundings (otherwise, you would not have visually
appreciated its movement).
13 At the level of SC, objects are not represented. Rather, an experience with a
certain SC is sufficient for you to see an object by virtue of discriminating certain
Layers and Components of Perceptual Content
137
order to see an object, there is no need to individuate it in a specific way,
to perceive it as being a thing of a certain kind, to take or recognize it as
something. This has been duly emphasized before by pointing out that
object-seeing is transparent and extensional (Chapter One, Part II.3). So,
SC is a basic layer of content that appears to satisfy the conditions for
object-seeing. In order to see something, you do not need to explicitly
represent individuals as having certain properties and relations. It suffices
to represent some properties that characterize the seen object in such a way
that representing those properties makes the object available for
discrimination.
Nonetheless, there is something to be noted here. In order for S to see
something (O), S must have a visual apparatus, O must be there to be seen
and O must (appropriately) cause the very episode of seeing O (Chapter
One, Part II.1). Now, a certain experience could have a certain SC without
any object being there to cause the experience itself, maybe even without
that S actually being equipped with a well-working visual apparatus.
Visual hallucinations, indeed, can be semantically characterized by
ascribing them a certain SC, but they are not cases of seeing insofar as
seeing involves a real relation to a seen object whilst visually hallucinating
something does not involve any real perceptual relation. In addition, it is
possible to imagine the visual experience—at least a conscious mental
state subjectively identical to a visual experience14—had by a brain in a
vat or (more realistically) induced in a subject without a working visual
apparatus. That experience would have a certain SC but it would certainly
not be a case of seeing. So, given that all causal, existential and relational
conditions for seeing something are satisfied, SC is apt to capture the other
condition for object-seeing (DC). If you see an object also by virtue of
your visual experience having a certain content, that content is an SC, even
features of it. SC is still silent on objects as having properties. According to Clark
2000, our visual experience attributes colors to locations rather than to objects. I do
not agree in general, but that view can well be applied to visual SC, namely, to the
basic semantic layer of visual experience. As we will see, there are other layers of
perceptual content involving objects, properties and relations.
14 Whether we can call “visual experience” a certain kind of hallucination depends
on the way we interpret the adjective “visual.” If an experience is visual only if it
involves an act of vision, a hallucination cannot be a visual experience (it may
occur without any genuine act of vision). In that case, certain hallucinations
introspectively seem visual experiences without being such. In case an experience
is visual if it involves the phenomenology typically involved in experiences of
visual perception, then a hallucination can well be a visual experience insofar as it
exhibits a certain phenomenology indiscriminable from that associated with an act
of visual perception.
138
Chapter Four Part I
if for a given visual experience to have a certain SC it is not sufficient for
that experience to be a case of seeing. So far so good.
I have just proposed a double parallelism relating object-seeing to SC
on the one hand, and seeing-as to PPC on the other. Then, provided that
there cannot be (literal) seeing-as without object-seeing, an important
question is whether there can be object-seeing without seeing-as. Even if
their conditions are different, it may well be that one cannot occur without
the other. Given our parallelism, the above question is identical to the
question of whether or not there can be visual experience with SC but
without PPC, provided there cannot be visual experiences with PPC
without SC.15 I do not want to take a definitive stand on that so I only
argue, according to intuitive plausibility, that “normal” perceptual
experience16 is always imbued with acts of recognition, of noticing, and
that it naturally involves taking the seen objects as being such and so.
Maybe it is possible for some (abnormal) experience to have just an SC,
but it is plausible to think that SC is a basic layer that acquires cognitive
significance for knowledge, reasoning and action only insofar as it allows
acts of recognition and categorization, insofar as it exhibits PPC.
Especially if we consider the essential belief-inducing role of perceptual
experience, it appears quite reasonable that PPC is a fundamental feature
of perceptual content, given that its presence enables the transition from
perception to belief. So, given an experience with SC, it seems that there
must be PPC. Given a case of object-seeing, it seems there must be an act
of seeing-as, but it is clear that in order for an episode of object-seeing to
occur, there is no need for a particular act of seeing-as to occur.17
Likewise, given an experience with a certain SC, there is no determinate
PPC the experience must have in order to be possible. Yet it seems that
there must be a PPC.18 As we said before, you cannot overlook everything,
15 Peacocke 1992 answers no to that question. For him, SC is not autonomous.
16 In Chapter Two, Parts II.1-II.3, I discussed experimental evidence that there is
seeing without noticing. But the point here is not whether there can be something
we see without noticing it; the point is, rather, whether there can be a perceptual
experience in which we neither notice nor recognize anything at all.
17 As Dretske remarks, “[…] in order to qualify as a perceptual state (seeing s), a
structure must be coupled to a cognitive mechanism capable of exploiting the
information held in sensory representation.” (Dretske 1981, 258, Chapter 6,
footnote 29). See also Dretske 2000. But no particular way of exploiting that
perceptual incoming information is required for the information to come in through
a perceptual channel.
18 Also Peacocke (1992, 124) denies autonomy to SC with respect to PPC, even if
he does not really argue for it beyond the remark that that thesis is reasonable: “I
doubt that we could ever justify the attribution of genuinely spatial content to an
Layers and Components of Perceptual Content
139
so even if in a given experience you can see many things without noticing
or recognizing them, you cannot fail to recognize something.
I.4 Three Layers of Content
In Chapter One I analyzed seeing-that ascriptions as well as object-seeing
and seeing-as. I argued for the view that these ascriptions attribute to the
subject an episode of coming to know a fact by visual means. So seeingthat is visually acquired propositional knowledge—it is a propositional
state involving a perceptual episode or state rather than just being a
perceptual episode or state. In particular, seeing-that is the factive,
successful ascription of a more general case of coming to believe by visual
means, namely of perceptual judgment or belief. I have also argued that if
you are to (literally) see that a is F, you must see a as an F; likewise, if you
are to see a as an F, you must see a in the first place. So there is a
transition involving three respective layers of content, from the SC that
characterizes object-seeing (seeing a) to PPC, which characterizes seeingas (seeing a as F) to the conceptual and propositional content that
characterizes the resulting perceptual belief or judgment that [a is F]. Such
threefold representational transition may be usefully represented by a
scheme (see below) that displays our understanding about the special
nature of perceptual content (Chapter Two):
organism’s state, of a kind going beyond sensitivity to higher-order properties of
stimulation patterns, unless the subject were on occasion to employ states with
these contents in identifying places over time,” where such an identification is an
example of experience with PPC. Anyway, it is one thing to say that there are no
organisms having experiences with SC that never have experiences with PPC, and
another to argue that, in organisms having experiences with both SC and PPC, an
experience can occur on a given occasion, which does exhibit SC without
exhibiting any PPC. Anyway, I do not think that question is among the
fundamental ones.
Chapter Four Part I
140
[Seeing O]
→ C1 (Scenario Content)
[Seeing O as an F] → C2 (Proto-(propositional content)
[Seeing-that: O is F] → C3 (propositional content of empirical judgments)
C1
SCENARIO
CONTENT
C2
PROTOPROPOSITIONAL
CONTENT
{recognition}
{knowledge}
Kind
of
Content
Type
of
Correctness
Nature
of
Content
iconic
analog
accurate/
inaccurate
dense
homogeneous
points in the perceptual
space
experience
true/false
discrete
heterogeneous
individuals
properties
relations
true/false
discrete
abstract
discrete
halfstructured
C3
fully
PROPOSITIONAL
structured
CONTENT
Constituent
Entities
{discrimination}
Kind of
Mental
State
perceptual
experience
empirical
recombinable
judgment/
concepts
belief
Nature
of the
State
nonconceptual
nonconceptual
(categorical)
conceptual
These intimately related levels of content leave open the possibility for
different kinds of mistakes to occur in perceptual judgment or belief. I can
have an experience with an inaccurate SC such that it constrains the
possible PPC and gives rise to a false belief. Or I can have an experience
whose SC is accurate but I fail to rightly recognize certain properties at a
PPC level, so I come to acquire a false perceptual belief despite the basic
SC of my experience being accurate. Or I can have an experience with an
accurate SC and a veridical PPC so that I see Fs as Fs and Gs as Gs but,
nonetheless, I form the false belief that a is not F and b is not G; for
example, on the basis of fallacious collateral knowledge that makes me
distrust my experience, or by virtue of some other cognitive defaillance of
the “central” belief-forming or inferential system. When everything goes
well—which is mostly the case—perceptual beliefs are brought about by
their correctness-conditions insofar as perceptual experiences are brought
about by their correctness-conditions, and the content of perceptual
experience is successfully conceptualized and given a propositional
structure accordingly. In any case, it is very important to keep in mind that
some mistakes in perceptual judging are not “strictly perceptual” mistakes
on the one hand, and that if perceptual mistakes occur, they can be of very
different kinds (inaccurate SC, inaccurate PPC or both) on the other.
Layers and Components of Perceptual Content
141
I.5 The Limits of Dretske’s Theory of Seeing
In the next subpart, I want to critically evaluate Dretske’s famous theory of
seeing. I will argue that this view presents some very important
weaknesses—it rests on a false dichotomy between two kinds of
seeing19—so that it must be either rejected or integrated.
I.5.1 Two Ways of Seeing
The core of Dretske’s well-known theory of seeing (Dretske 1969, 1981,
1988, 1995) has remained the same through the years despite some
specific advancements and theoretical enrichment.20 To sum it up briefly,
Dretske distinguishes between “simple” or “non-epistemic seeing” (SS)
and “epistemic seeing” (ES); the first is cognitively neutral and is equated
to object-seeing while the second is cognitively loaded and equated to
fact-seeing. SS is of things like tables, chairs and the like, whilst ES
involves knowing facts about seen things. In other words, ES amounts to
seeing-that. These different states respectively involve different kinds of
awareness—object-awareness and fact-awareness. In seeing a, I become
visually aware of a; in seeing that a is F, I become aware of the fact that a
is F. Seeing a in itself is concept-free while seeing that a is F is a conceptcharged mental state and hence involves a belief so it is propositional
(perceptually acquired) knowledge.
If S is aware that x is F, then S has the concept F and uses (applies) it in his
awareness of x […]. Perceptual awareness of facts is a mental state or attitude
that involves the possession and use of concepts, the sort of cognitive or
intellectual capacity involved in thought and belief. (Dretske 2000, 134)
While sensory perception (object-seeing) is the pick-up and delivery of
information, cognitive perception (fact-seeing) is its utilization for
identification, classification, recognition and so on. Seeing objects
19 As I will show, it is not the dichotomy itself that is false but its supposed
completeness.
20 In Dretske 1969, the theory is meant to address epistemological worries as well
as capture some theoretically relevant features of ordinary language. In Dretske
1981, the theory is embedded in a more general theory of information so it is a
theory of different ways for certain states of carrying information and for cognitive
systems of picking up environmental information. In Dretske 1995, the
information-based theory is further enriched with a bio-functional and teleosemantic component.
142
Chapter Four Part I
amounts to taking in information about them21 in an analog form. A piece
of information is carried by a state in analog form when it is not the most
specific information carried by that state; rather, it is carried with other
information it is nested into. For example, the experience of a cup may
carry the information that the cup has coffee in it in an analog way. Indeed
it is not the most specific information carried to the subject by the
experience: you also necessarily experience how big the cup is, where it is,
what color it is and so on. On the contrary, the most specific carried
information of that belief is the fact that [the cup has coffee in it]. The
same information carried by a perceptual experience in an analog way
could be carried, for example, by a sentence in a digital way when you
come to know that the cup has coffee in it because you are told about it. So
the cognitive use of perceptual experience is an extraction or
“digitalization” of information carried by experience in an analog way, it is
an analog-to-digital conversion, but perceptual experiencing is taking
information, not extracting/converting it.
Perception is a process by means of which information is delivered within a
richer matrix of information (hence in analog form) to the cognitive center for
their selective use […] cognitive activity is the conceptual mobilization of
incoming information and this conceptual treatment is fundamentally a matter
of ignoring differences, of going from the concrete to the abstract, of passing
from the particular to the general, (Dretske 1981, 142)
but
perception in itself is cognitively neutral. (153)
Later on, Dretske distinguishes conventional and natural representations
—defined as states having indicator functions—on the one hand, and
natural representations into sensory and conceptual representations on the
other (1995, 19ff.). Sensory representations (like experiences, sensations
and feelings) have systemic, innate and philogenetically fixed indicator
functions, while conceptual representations (like thoughts, judgments and
beliefs) have acquired and ontogenetically determined indicator functions.
Experiences are to be identified with states whose representational properties
are systemic. Thoughts (conceptual states in general), on the other hand, are
states whose representational properties are acquired. (Dretske 1995, 15)
21 “E is a visual experience of x in S if E carries information about x, the
information is extracted from light by photoreceptors in S’s eyes, and this
information is directly available for control of S’s action.” (Dretske 2006, 152)
Layers and Components of Perceptual Content
143
This picture, though enriched with new bio-functional aspects, is
consistent with the previous theory of seeing referred to above (1969,
1981). It still encompasses a clear dichotomy between sensation and
cognition where the cognitive element involved in perceptual knowledge
is neatly distinguished from the sensory component and associated with
conceptual capacities involving belief and thought.
I.5.2 Two Ways Aren’t Enough
Now I want to argue that something essential must be missing in that view.
Dretske starts from a linguistic-ascriptive distinction in the first place then
tries to show that this ascriptive distinction tracks a real distinction
between two different objective phenomena. Then, unfortunately, he
fatally conflates two different criteria of distinction—one strictly
epistemological, the other having to do with levels of cognition in a more
liberal sense. Let us start with the more general criticism though.
What makes the view incomplete is the total absence of an
intermediate level between object-seeing and fact-seeing, namely of what
we have called seeing-as and semantically characterized with possession
of PPC. Actually, the position of that hybrid mode of seeing appears to be
inevitable—and Dretske’s view appears to be thereby insufficient—by
considering the following intuitively evident facts:
1) Not every “digitalization” is a conceptualization. To extract
information from an experience so as to convert its analog and profuse
content into more discrete contents, as happens when perceiving
something as F, I do not need to possess the concept F; actually, I do not
need to have conceptual abilities at all. Non-conceptual animals can
perceive objects as being so-and-so, for example as moving or as being
distant, without deploying the concepts [moving] or [distant].22 If objectseeing is meant to be cognitively neutral, i.e., if the only cognitive use of
perceptual information were conceptualization, then how could a nonconceptual animal classify, identify and recognize objects over time and
place? There must be a “general” perceptual content that is still preconceptual.23
22 On “seeing x as an F” as being “conceptually undemanding,” see also Johnston
2006.
23 Dretske 2003, 81, footnote 3, explicitly distinguishes “experience of M from
recognizing (i.e., conceptually representing) something as M. The same
assumption is made by Fodor 2007, for whom “representing as” amounts to
conceptualizing.
144
Chapter Four Part I
2) To make the same point in other terms, it is not only conceptual
animals that learn from experience. Perceptual learning presupposes a
cognitively loaded experience, namely, the possibility for perceptual
experiences to acquire new representational functions through repeated
perceptual encounters with environmental objects and conditions. If
perceptual learning is possible for non-conceptual beings, the assimilation
of the sensation/cognition distinction to the systemic/acquired distinction
and the conflation of both of them into the perceptual/conceptual
distinction must be flawed, unless Dretske buys into a very undemanding
view on concepts such that any form of acquired recognitional,
classificatory and identificational capacity involves concepts.24 In that
case, any learning animal would be a conceptual cognitive system, but the
concept of “concept” entailed by such a view fails drastically to capture
the cognitive richness of our conceptual abilities insofar as they are
connected with inferential abilities as well as rationality (see II, 1-1.6).
Anyway, Dretske does not take that extremely reductive approach on
concepts so the inconsistency remains.
3) Another way of saying the same thing is pointing out that it is not
only beliefs that are acquirable representational states pace Dretske. No
conceptual or inferential ability is involved in representing things as being
a certain way through a perceptual experience in an ontogenetically
acquirable way. A dog’s experience of a doorbell ringing, if repeated, may
well come to represent to the dog the presence of someone behind the
door, for example, but no-one—or very few people—would credit a dog
with the concepts [presence], [people], [being behind of], [door] and the
like, or with a propositional belief that there is someone at the door. A
representation with certain content such that the subject having it need not
possess the concepts canonically used to specify that very same content is
just a non-conceptual content by definition.25 Thus, there must be nonconceptual but acquired and cognitively relevant content; there must be
24 That there must be acquired but non-conceptual representational states (nonconceptual seeing-as states) does not entail that every seeing-as state must be
ontogenetically acquired. For example, there seems to be evidence that in many
mammals there is an innate (i.e., not acquired) recognitional ability, or perceptual
category, for [animal movement] or [biological movement] as distinct from
movement of inanimate, non-living objects. See Johansson 1973, Blake 1993,
Vallortigara 2000, 46ff. Likewise, it seems that the auditory perception-recognition
of a [voice-of-a-conspecific] is also an innate ability: a non-acquired seeing-as. Of
course, there are also views—like Fodor’s, 1998—in which concepts themselves
may well be innate.
25 Cussins 1990 introduces non-conceptual content exactly like that.
Layers and Components of Perceptual Content
145
seeing-as in other terms. If we believe in the Kantian motto that intuitions
without concepts are blind but concepts without intuitions are empty, we
still need to posit an intermediate level of perceptual representation that is
neither completely blind nor completely empty, which makes the
application of concepts to “intuitions” possible.26
I.5.3 Too Many Distinctions at Once
Strictly speaking, cognition has to do essentially with truth, just as
knowledge has to do with truth. A cognitive achievement is a successful
way of being in the right relation with a certain domain in one’s
surrounding environment. For Dretske, cognitive perception (see Dretske
1990, 133ff.) amounts to knowledge; for example, cognitively perceiving a
cat is visually acquiring the knowledge that the seen cat is a cat. On the
contrary, simple seeing is not cognitive just because, as such, it does not
have to do with truth (and knowledge): either you see the cat or you don’t,
but to see it, you do not have to come to know that it is a cat or something
else. You just see it in the first place. So you have a seen object (SS) and
knowledge about it (ES); i.e., that it is a cat or some other [F]; and these
are different episodes even if an ES-episode presupposes an SS-episode
and not vice versa.
There are but two ways that something—a state, a process, an
episode—can be characterized as cognitive. One way depends on the
state/process/episode being an epistemic achievement, so on its being a
success such that the state/episode/process is true or at least results in a
true state. In that strict sense, untrue states are not cognitive states by
definition and nor are those states that are neither true nor false (like
object-seeing). In another sense, a state/process/episode is cognitive
because it has a certain role in the system’s use of information that
normally allows the system to behave appropriately or to produce reliable
representations, maps or “views” on the surrounding environment. In that
second sense, a state may well be cognitive and untrue at the same time.
Indeed, a cognitive failure is no less cognitive than a cognitive
achievement. A state produced by a mechanism that has the function of
producing knowledge or veridical representations of the environment is
cognitive even if it happens to be a “bad” or unsuccessful token of the type
26 Another important difference between perceptual seeing-as and full-fledged
deployment of concepts in perception is that the very same deployed concepts can
be activated by the subject in many other non-perceptual circumstances, while
basic seeing-as episodes may well be passive, domain-specific and contextdependent.
146
Chapter Four Part I
of states produced by that very mechanism. Dretske does not distinguish
these two senses.
In addition, a certain state can constitute a cognitive achievement
without being evaluable as true/false or as veridical/falsidical thereby. For
example, it may be argued that seeing an object is a cognitive achievement
despite the fact that seeing an object is not something that can be
characterized as such as true or false. Indeed, you might fail to see an
object that is within your visual reach so seeing it is a successful
perceptual performance on your part.
With all this in mind, let us consider again Dretske’s distinction
between two ways of seeing. SS is non-cognitive insofar as it does not
require knowledge of the thing seen whereas ES is cognitive insofar as it is
just the episode of acquiring knowledge that the seen thing is such and
such. But that distinction cannot be confused with another distinction:
between states that can be true-veridical (or false-falsidical thereby) on the
one hand, and states that cannot be true-veridical (nor false-falsidical
thereby). For example, a belief is a certain kind of mental state that can be
true or false and, more significantly, can be knowledge or not.27 You can
see X and come to believe by visual means that X is a cat, or that x is
square. The episode of belief-acquiring may or may not be an episode of
knowledge-acquiring; for example, surely it is not when the belief is false
because X is not a cat, or X is not blue as it seems to be to you in
experience. So for Dretske that episode should be non-cognitive because it
does not require knowledge of the thing seen! Such conflation of two
senses of “cognitive” fatally undermines Dretske’s global theory of seeing.
For example, incredibly, the only time he has anything to say about seeingas in the whole corpus of his writings is in a footnote—and he just leaves
it aside as something unimportant.
[Seeing-as is] a hybrid form of perception, a way of seeing that goes beyond
sensory perception […] but falling short of full cognitive perception
(knowledge not being required). One sees a stick as a snake. The stick
obviously does not have to be a snake for one to see it as a snake. Hence, this
cannot be cognitive perception. (Dretske 1990, 133, footnote)
So Dretske only recognizes two levels of seeing—either “seeing X,
that is an F” or “seeing that X is an F.” In the first case (SS), F is just one
of the possible descriptions the ascriber can use to pick out the object seen
27 If the belief is true, it can be knowledge (given justification and some other
ingredient); if the belief is false, it cannot be knowledge since knowledge is
factive.
Layers and Components of Perceptual Content
147
by S. In the second case (ES), F is a concept possessed and exercised by S
within the very propositional state that amounts to visually acquired
knowledge. So, in the second case, the exercise of F is ascribed; in the first
case, it is used by the ascriber to ascribe something else.
With his purportedly exhaustive cognitive/non-cognitive distinction,
where “cognitive” is implausibly equated to both “epistemic” and
“conceptual,” Dretske cannot see any independent way of seeing besides
his insufficient dichotomy of seeing an object and seeing that an object is
such and so. For this reason, he assimilates cases of seeing “where the cat
is” or “how big the tree is” with cases of seeing that the cat is there and
seeing that the tree is that big, so with cases of conceptual and
propositional epistemic states of knowledge.28 On the contrary, a dog can
well see where the cat is; indeed it cleverly runs in a certain direction to
catch it; even though it cannot see that the cat is down there, provided that
seeing-that-P involves a conceptualization of the constituents of the fact
making P true, and the endorsement of the fully structured propositional
content of P. There is pre-propositional cognition, there is pre-conceptual
recognition, there is seeing-as between simple object-seeing and seeingthat, in other words. Dretske fails to give the right mediating role to
seeing-as because its ascription is a non-factive context, and he does not
consider that, just as successful belief may be knowledge, so successful
seeing-as may be genuine recognition; “animal knowledge,” if you like.29
28 Dretske holds that recognizing something as a triangle is seeing that X is a
triangle (Dretske 1990, 131), where seeing-that involves conceptualization of the
[triangle]-property. But this is too demanding a condition for the recognition I am
arguing for: any recognition would be conceptual by definition, which is highly
implausible and empirically puzzling, given the overwhelming evidence for
recognitional abilities of non-conceptual animals.
29 See Prinz 2006, 436ff. As Prinz notes, it sounds odds to say that S sees a fork as
a fork because “seeing-as” is more often used—as for example in Wittgenstein
1953—for cases involving a special interpretive act, like that involved in seeing a
cloud as a warthog, or a drawing as a duck. But seeing-as is more basic than that—
it is any act of visual recognition of a type T whose correspondent ability is stored
in memory: it is a recognition triggered by the presence of a token of T when
successful, but it may well be triggered by an object that is not a token of T when
unsuccessful. Seeing-as may be wrong, unlike seeing-that and like coming-tobelieve-that-P-by-visual-means. Since recognition is factive, visual recognition
stands to seeing-that just like seeing-as stands to coming-to-believe-that-by-visualmeans from an epistemological point of view, but visual recognition stands to
seeing-as just like seeing-that stands to coming-to-believe-by- visual-means from
the point of view of cognitive articulateness of the state, so to say. Of course the
unsuccessful cases are conceptually parasitic on the successful cases insofar as
148
Chapter Four Part I
Seeing-as stands to genuine recognition as visually-coming-to-believe
stands to seeing-that, so to conceptual-propositional knowledge (based on
visual perception). Seeing-as involves recognitional abilities whereas
seeing-that involves fully-fledged concepts able to be molded into
beliefs.30
Seeing-as is a necessary level to explain the epistemic value of
perceptual experience, to its very possibility of producing perceptual
knowledge. How could you know that a is F by visual means if such visual
means are conceived as being just episodes of object-seeing? If seeing a as
an F is not somehow included in the visual means by which S comes to
know that a is F, that transition from simple perception to perceptual
knowledge is a mystery from an epistemological, semantical, and even a
phenomenological point of view, but more on that later.
seeing is a success-verb in the first place. You pick out a wrong case of seeing-as
by reference to a positive capacity; you pick out a falsidical case of coming-tobelieve-by-visual-means by reference to what it would be for such a mental state to
be a case of seeing-that.
30 Also Tye 2000, 215 has a doxastic model of object-recognition: object
recognition is taken to be “a matter of seeing that such and such type of object is
present […] a matter of forming an appropriate belief or judgment on the basis of
visual experiences […] there are two components in visual recognition, a belief
component and a looking component.” The same view is that of Lyons 2005, 242:
“to recognize an object, to categorize it, to identify it, is at least typically to judge
it to be a member of a certain category.” Also for Fodor 2007, seeing a as F is
conceptualizing it as an F, so forming the respective belief that it is an F. I am
arguing that the two-component view is a consequence of neglecting the seeing-as
level as a pre-doxastic one, a level that stands between these two alleged
components (look and judgment). Recognitional acts are not necessarily
conceptualizations or judgments that something is the case, otherwise nonlinguistic animals could not possess recognitional abilities at all nor could they
ever learn.
PART II
OBJECTS AND PROPERTIES:
HOW THEY FEATURE IN PERCEPTUAL
CONTENT
II.1 Back from Recognition to Discrimination
In Chapter One (Parts I, 4.3, 4.4, 4.5), I argued that seeing-as is
recognitional and presupposes both possession and activation of a positive
category or ability. That recognitional scheme is stored in memory such
that the recognition of an a as (an) F rests on a “matching” between the
stored category and the perceived object a. Now, that capacity can concern
different kinds of properties. I have argued that when F is not an SCMproperty (a property constituting a’s visible profile), seeing a as F
presupposes the appreciation of a’s sensible profile and the association of
it with the property F. In other words, strictly visual properties must be
appreciated in the first place in order to (literally) see a as F when F is not
itself a property that shapes a’s visible profile. So the basic capacity to
consider is that appreciation, namely the ability to see an a as having a
certain visible profile, or as being F, G and H where F, G and H are SCMproperties, like [red], [square], [big], [distant-from], [symmetrical-with]
and so on. For example, in order to see a, say, as [prey], I must see a as
[so-and-so shaped], [so-and-so colored], [so-and-so big], [so-and-so
moving] and the like. Now what does it take to see, say, something as [red]
provided that seeing-as concerning SCM-properties is more basic than
seeing-as concerning non-SCM-properties? In order to see something as
red or as square, you must possess the [red] or [square] category, namely,
you must be able to (non-conceptually) recognize red or square objects as
tokens of a same type, say [red] or [square].
Let us call seeing-as concerning SCM-properties “basic seeing-as.” Basic
seeing-as is still recognitional and therefore involves memory so it is a
matching between a seen object and an already stored perceptual ability.
All that presupposes an even more basic visual capacity that we need to
150
Chapter Four Part II
consider: discriminating SCM-properties is a more basic operation than
recognizing SCM-properties. It is well known that discriminatory powers
outstrip memory. For example, I can recognize a shade of red as the same
shade of red of an object I saw on another occasion, but the shades of red I
can presently discriminate are more than the shades of red I can retain in
memory and use later for recognitional and comparative purposes. So,
discrimination is more fine-grained than recognition. In short, on the one
hand, recognition entails discrimination; on the other, discrimination
outstrips recognition.1
If perceptual discriminatory powers can be associated with objectseeing and accounted for by appeal to SCs (see above), perceptual
recognitional powers can be associated with seeing-as episodes—episodes
involving memory and accounted for by appeal to PPCs.
How could one or more episodes of object-seeing enable respective
episodes of seeing-as? To answer, we need to step back to object-seeing
and see how reiterated object-seeing can prepare the constitution of a
perceptual scheme, which comes to be stored in one’s memory as an
acquired recursively available recognitional ability. In order for this to be
possible, there must be a way of visually representing an object as F
already at the level of object-seeing. How does object-seeing involve the
representation of properties as possessed by seen objects? What is the
semantic structure of the content possessed by states of object-seeing?
II.2 Object-Seeing Through Property Discrimination
Visual discrimination of an object through a property involves prerecognitional representation of a certain property. Seeing an object O
through discriminating its contours—say, thanks to chromatic contrasts
against the surroundings—is not yet seeing the object as having certain
properties F, G, Z. Seeing-as is recognitional indeed. On the contrary, you
can see an object thanks to being visually sensitive to a property of it, be it
F, without explicitly recognizing the property as the property F you have a
recognitional ability for; i.e., without exerting the recognitional act of
matching the perceived property to a category already stored in memory.2
So far I have been using “seeing” as an implicative verb, always
having particular objects as its possible direct complements. Instead of
talking of seeing properties, I prefer to talk of seeing objects through
visually discriminating some of their visible properties. Whenever we are
1 See Tye 1995, Raffman 1995, Martin 1992.
2 See Prinz 2006.
Objects and Properties: How They Feature in Perceptual Content
151
perceptually conscious of a property, we are seeing an object that
possesses that property.3 This is why “seeing properties” is nothing more
than seeing particular objects that have those properties.4 So, seeing the
instantiation of a property like [yellow] amounts to seeing a yellow object
thanks to visually discriminating its yellowness.
However, it may also happen that we see an object through
representing it as having a property that in fact it does not have; yet this
circumstance can put us in a position to see the object. This is the case of
visual illusion, which involves a seen object although falsidically
represented. If I see a red apple through visually representing it as green, I
am undergoing an illusory experience; still I am having a perceptual
experience of the apple by representing it as having a certain property,
albeit one it does not have. I am in visual causal contact with the apple, my
visual system represents that object as green, that visual representation is
caused by the object and it is indeed about it, but the property the object is
represented as having is absent.
So, object-seeing always happens through property discrimination but
it may also happen through representation of the object as having a
property it does not have. Visual illusion is still a successful case of objectseeing though a unsuccessful case of property-representation. It is an
intentional relation grounded in a real relation between the subject and the
perceived environment.5 It is just in being a real relation that it can be an
intentional (falsidical) relation about the real perceived relatum. In fact,
visual illusions are perceptions so whenever a state of visual illusion is
instantiated, a case of object-seeing is ipso facto instantiated. Unlike visual
hallucinations—which are not perceptions despite their perception-like
phenomenology—perceptual illusions are always about environmental
objects the subject is genuinely related to. But the objectless content of
3 Here I use the expression “perceptually conscious of” in a factive or implicative
way so if you are perceptually conscious of a property, it is instantiated by an
object perceived by you. The problem of what we are conscious of in illusion
(when the property is not instantiated) or in hallucination (when not even an object
is perceived) is another huge issue I will face later. I am just arguing that when you
perceive a property, so when you are appropriately related to an instance of that
property, you are perceiving an object thereby that has the property.
4 See Chalmers 2006. It is an oversimplification: we see the sky, shadows, a
rainbow, and other objects that are clearly not particulars. In addition, our
experience is never just of an object as having a property. That is a very scholastic
way to describe experience by abstracting away certain features from the concrete
flux and organized totality of experiencing a continuous, complex scene over time.
5 As McGinn 1982, 50 notes, “You do not cease to see a thing just because your
experience credits it with properties it does not objectively have.”
152
Chapter Four Part II
hallucinations will need to be dealt with in another chapter (see Chapter
Seven, Part II 4-5). So far, it is enough to remark that 1) object-seeing
involves property-representation, 2) property-representation need not be
seeing-as or recognitional seeing, and 3) a case of wrong propertyrepresentation may well put the subject in the position of seeing an object.
II.3 Which Basic Semantic Ingredients Shape the Content
of Visual Perception?
Perceptual experiences are related to environmental objects given to the
subject “under” certain properties through a certain perceptual apparatus
(vision, touch, hearing etc.) that also determines the way they are
experienced. Call it a mode. The very same property can be given through
experience in a visual or tactile way (e.g., [square]), in a visual or auditory
way (e.g., [overhead]), in an olfactory or in a taste-way (e.g., [sour]), and
so on. The mode is not just a matter of which apparatus is causally
involved but also which distinctive phenomenological effects that
involvement has. So seeing something and visually representing it as
square is not like touching the same thing and realizing it is square despite
the sameness of the perceived object and represented property.6
Therefore, a certain perceptual experience is a mental episode
individuated by the following elements:
1. a subject who/which undergoes the perceptual experience;
2. an object in the environment the subject is causally and perceptually
related to;
3. a property (or properties) the object is represented as having in the
experience;
4. a mode or a way both the object and its represented properties are given in
the experience; and
5. a time in which the episode takes place.
6 This is why I call it mode rather than modality. While perceptual modalities are
individuated by mechanical facts, by the apparatus involved, the mode as I mean it
has to do with the phenomenological dimension related to a specific modality, not
just with the modality. I agree with Lyons 2005, 241 in that “what distinguishes
one kind of perceptual system from another is the kind of information they process,
rather than any phenomenal experiences they produce.” Nonetheless, the conscious
character associated with each perceptual modality is different and exhibits a
proper phenomenological salience. That such salience is not the criterion for
distinguishing perceptual systems in cognitive sciences is another matter.
Objects and Properties: How They Feature in Perceptual Content
153
So, a certain perceptual experience (PE) could be individuated as
follows:
In PE, at t the object O is represented as being F by the subject S in the visual
mode.
We have an object-dependent content made out of the perceived object
on the one hand, and the properties the object is represented as having on
the other. Apart from the experiencer and the time at which the very
experience takes place, there is another element that individuates a
perception, which is to be distinguished from the content: it is the mode. It
is often said that what is perceived in perceptual experience is always
given in a certain way so a distinction is made between what is perceived
and how it is perceived. The what/how distinction is very important and
useful provided the two possible interpretations of it are not conflated or
confused with each other. By “how O is given through experience,” one
could mean: a) how the object is represented as being = the properties it is
represented as having in the experience; or b) how both the object and the
represented properties are experienced = in which mode they are
perceptually represented (visually, auditorily, and so on). I will call the
first content and the second mode. So, the Content (upper case) of a
perceptual experience is constituted by its object and its content (lower
case):
content + object = Content
The perceived object, together with the properties it is represented as
having, is then the Content of a given perceptual experience. The mode
does not directly enter to constitute the Content, even if it contributes to
determining it. In fact, the perceptual mode constrains the range of
properties that can or cannot be represented in a given perception. For
example, the visual mode determines that the experience will represent the
object as having certain colors but not as having certain smells.7
The Content of a perceptual experience (PE from now on) is that in
virtue of which the PE is semantically evaluable as correct/incorrect,
accurate/inaccurate, veridical/falsidical. The Content constitutes the world7 Of course, our perceptual activity is essentially multimodal and integrated on the
top of involving a continuous income of information of a whole complex “scene”
over time, so this representation of “a” perception as indexed to a specific mode
and involving “an” object and “a” property is a philosophical abstraction. It is
important to keep this in mind.
154
Chapter Four Part II
to-mind conditions of satisfaction of the PE;8 namely, the way the world
should be in order for the PE to be accurate-correct-veridical. The
perceived object both causes the PE and constitutes its target, whereas the
properties the targeted object is represented as having, the content (lower
case) of the PE, are what is “measured” by the properties the target
actually has. Thus, the perceptual content exhibits an analogous structure
with that of a standard propositional function, which can acquire a certain
truth-value according to the argument with which it is “saturated.”9 For
example, {------- is wise}, can get the value (true) if it is applied to the
individual Socrates, or the value (false) if it is applied to someone else
who is not wise. So the argument for the Content of a PE is furnished by
the world, namely, by the perceived object that causes the PE itself, and
the value (accurate/inaccurate, veridical/falsidical) is given by the eventual
matching between the represented properties and the properties actually
possessed by the worldly target. If the seen object is F, a PE that represents
it as F is accurate; a PE that does not is inaccurate. So, worldly objects are
accuracy-makers for perceptual contents insofar as they possess or lack
those very properties the PE represents them as having. Perceptual
Contents are de re Contents.
I introduce the notion of accuracy-maker instead of talking about truthmakers because that notion is compatible with the basic content of
perceptual experience being neither conceptual nor propositional, as I
believe and will argue is the case. Only sentences and sentence-like mental
states like thoughts and beliefs can properly be true and be made true by
truth-makers accordingly. But if true and false are yes/no notions,10 on the
contrary, accuracy may well come in degrees. Not only can a perceptual
experience be accurate or inaccurate full stop, it can be semantically
evaluated as more or less accurate or inaccurate. Therefore, it makes sense
to evaluate how accurate a perception is in a way it does not seemingly
8 To use the language of Searle 1983, 1992, perceptual experience has a world-tomind direction of fit, unlike other types of intentional states. For example, a desire
or an intention has a mind-to-world direction of fit.
9 See Frege 1891/1980. Russell 1910, 28, defines a propositional function as
“something which contains a variable x, and expresses a proposition as soon as a
value is assigned to x. That is to say, it differs from a proposition solely by the fact
that it is ambiguous: it contains a variable of which the value is unassigned.” Only
when a value is assigned to the variable can the function become a genuine, truthevaluable sentence. In the same vein, only by considering the real object that is
perceived does a perceptual representation become an accuracy-available semantic
episode.
10 At least that holds in standard logic. Fuzzy logic and other non-standard logics
admit degrees of truth, but I am not concerned with them here.
Objects and Properties: How They Feature in Perceptual Content
155
make sense to evaluate how true a proposition is.
So we have a worldly object O with its real properties, a certain set of
represented properties putatively belonging to the perceived object, and a
matching-relation between the set of real properties of O and the set of
properties represented in the PE as had by O. The matching-relation can
make the PE accurate (PE content and object match), inaccurate (PE
content and object mismatch), or partially accurate (PE content and object
match with respect to certain properties but they mismatch with respect to
others), according to the way the worldly relatum is.
We can stipulate that a PE is veridical when it is fully accurate,
namely, when all the relevant properties the perceived object is
represented as having are actually had by it. Likewise, a PE will be
falsidical when it is inaccurate in many relevant respects, namely, in case
many or all the relevant properties the object is represented as having by
the PE are not actually had.
What matters, though, is that accuracy comes in degrees, even though
it can make sense and be theoretically useful to semantically characterize a
PE as veridical or falsidical. Likewise, instead of considering global
veridicality, we can introduce the notion of veridical-with-respect-to-aproperty so the more accurate a PE is, the bigger the number of properties
with respect to which it is veridical. In any case, the Content of a PE
consists of its accuracy-conditions, namely, the objective conditions under
which the PE would be accurate. Those conditions—the Content—involve
the perceived object as a constituent, and the properties represented as had
by the object—the content—as another constituent. Only together can they
provide an accuracy-value for the PE.
II.4 Object-Dependency and Singularity
II.4.1 Introducing Object-Dependency
I have been arguing that perceptual Content is object-dependent11
11 I have called Content of a PE the semantic structure involving both the object
and the properties it is represented as having in PE. I have called content the set of
these properties. So by definition Content is object-dependent. To avoid confusion,
I want to argue that what is standardly called perceptual content in the debate is
object-dependent, and it is identical with what I call Content. What I call content
(the properties O is represented as having) is not object-dependent just because it is
only an ingredient of the perceptual content standardly meant. The reason why I
call such set of properties “content”—at the price of raising some confusion in the
reader—will become clear later.
156
Chapter Four Part II
insofar as it involves the perceptual object, a worldly particular, as one of
its semantic constituents. This picture also fits with the transparent
behavior and implicativity of object-seeing ascriptions, which depends on
object-seeing being a real relation. If perceptual Content contains worldly
particulars, then it is partially determined by external factors. Holding
object-dependency involves committing to Content externalism. Indeed,
there are external extra-mental factors that co-determine perceptual
Content.
It is difficult to deny that in perception we become conscious of
worldly particulars, but is it necessary to hold that those worldly
particulars also constitute the perceptual Content itself instead of being
just among the items that make the contents accurate/veridical?
In what follows, I will present some arguments for the necessity of
including the particular worldly perceived entities in the Content of PE.
II.4.2 The Generality Thesis
As has been made explicit since Chapter One, it is among the conditions
of object-seeing that the seen object must (appropriately) cause the very
episode of seeing it; that object-dependency of perception does not only
rest on our ordinary concept of perceiving. It appears to be necessary, from
an intentional point of view at least, to make logical room for the very
ideas of both non-veridical perception and “veridical” hallucination. A
non-veridical perception is an illusion, for example; a case where you see
an object but your PE credits it with properties the object does not have.
The content of your perception is non-veridical but it is such just because
it is non-veridical of the seen object, so the object makes the PE true or
false insofar as it is seen.
Hallucinations are experiences where it seems to you that you perceive
an object with certain properties but there is no perceived object at all, so a
fortiori those properties are not instantiated by anything you perceive.
“Veridical” hallucinations are cases where you hallucinate a scene that
happens to match the real scene before you but the latter is not
(appropriately) causing your experience so the experience is not an
experience of it. Here is a well-known example by Grice (1962):12 you see
a clock on the shelf then a neuroscientist stimulates your visual cortex in
such a way that you do not notice anything changing in your experience.
He removes the clock. As Grice points out, there is a strong intuition that
you did not see the clock, even when it really was before your eyes, as
12 On veridical hallucinations, see also Lewis 1980.
Objects and Properties: How They Feature in Perceptual Content
157
soon as you began to be artificially stimulated just because your
experience was then caused by the visual cortex stimulation rather than by
the clock itself. “Veridical” hallucinations are cases where the content of
your experience perfectly matches the surrounding reality, but that
matching is intuitively not enough for the experience to be a veridical
perception. The properties the experience represents are actually
instantiated but they are not instantiated by a seen object insofar as no
object is seen, even though in the hallucination it seems that there is an
object located there having this and that property. That is not a veridical
perceptual experience but a “veridical” hallucination because the causal
factor, the environmental relation to an object, is missing. Accounting for
the difference of their contents is a problem for those who hold that
perceptual content does not involve particulars; and that worldly
particulars with their properties are what (may) satisfy the content rather
than being part of the content itself. How can the perceptual content of a
PE be sensitive to its causes if none of its causes are part of it? However,
how could hallucinations be “veridical” if they are completely detached
from the environment they purport to present to the subject?
According to the generality thesis (GT), when a subject perceives the
world, the content of her perception “is not to be specified by using any
terms that refer to the object of experience” (McGinn 1982, 51):13
Generality thesis = perceptual content is always general and does not
contain particulars.
If GT is true, perceptual content must embed an existential component.
For example, a certain PE could be ascribed the following content:
There is an object that is F at location L.14
Thus, GT entails that perceptual content is existentially quantified.15
13 Others who hold the GT are Lewis 1980, Searle 1983, McLaughlin 1989,
Davies 1992. In discussing the GT, I will mainly follow the extraordinarily clear
and insightful Soteriou 2001. I find that his criticism very convincing.
14 Location L should be specified egocentrically, like “at such a distance and
direction from me,” and so on.
15 McGinn 1982 explains the GT. Davies 1992 explicitly holds that if perceptual
content is general, it must be existentially quantified. Such a transition seems to be
inevitable. Either perceived objects are demonstratively presented and content is
not general, or those objects need to be non-demonstratively present through the
PE representing certain properties of them, where such a representation cannot be
158
Chapter Four Part II
However, that sort of content could well by satisfied by the clock on the
shelf in Grice’s thought experiment. Provided content consists of the
accuracy and veridicality conditions of the PE, if there is something that is
a clock located at a certain distance and direction from the cortexstimulated subject, then her PE is veridical throughout. But if we are to
save the intuition that it is a hallucination after all, we need to do justice to
its “strange” hallucinatory veridicality. Specifically, it is implausible that
the veridicality (or accuracy) of a perceptual experience could be
evaluated independently of whether any object is perceived and of which
object it is that it is perceived by the subject.16
II.4.3 Searle’s Account of Perceptual Content
Assuming that the content is general and then existentially quantified, in
order to distinguish the accuracy-value of “veridical” hallucinations from
those of veridical perceptions, we need to further enrich the general
content with a causal element. That is, notoriously, Searle’s move (Searle
1983, 1991).17 In Searle’s view, perceptual content must contain a
reference to a causal element as well as a (self)-reference to the very
perceptual experience the content is content of:
If I have a visual experience so that it visually seems to me that an F is G:
visual experience: [that the F is G and the fact that F is G is causing this visual
experience]18
So, we have a general, existential, causal and self-referential content of
perceptual experiences.19 There is an object O, which is F, and the object’s
neutral about these properties being instantiated in the environment. But a property
is instantiated when there is some object that possesses it.
16 On this point, see Soteriou 2001, Sainsbury 2006. In addition, the GT is also
phenomenologically implausible. In perceptual experience, it does not
introspectively seem to us as if we were entertaining existential and general
content, like there being an O that is such and so. The felt reality of perception
involves us purportedly referring to particulars in a demonstrative way. We
introspectively seem to experience this and that as such and so.
17 See also Chalmers 2004, 2006.
18 That account is exposed in Searle 1992, 288ff. See the critical discussion of that
view by Soteriou 2001.
19 I will leave aside the idea that perceptual content is introduced as a that-clause.
Searle holds that perceptual content is propositional but that idea is relatively
independent of its account of perceptual content as general and involving causation
on the one side and self-reference to the experience itself on the other.
Objects and Properties: How They Feature in Perceptual Content
159
being F is causing that very experience according to which there is an
object O that is F. If the world satisfies all that, then the PE is true.
Evidently that veridicality condition is not satisfied by a “veridical”
hallucination insofar as the clock’s being on the shelf out there is not
causing the experience as if there was a clock on the shelf out there.
Rather, the brain-manipulation is causing that experience. So “veridical”
hallucinations are not veridical after all20 if the content of perceptual
experience is articulated that way.
If GT is true, and the veridicality of a PE has to do with the question of
whether an object is perceived as well as which object is perceived, then
Searle’s move appears inevitable: we need to embed into existential
content a causal and a self-referential component.
One could still reject the intuition that veridicality of PE must depend
on whether an object is perceived, and if so, which object. However, such
intuition can also be positively argued for rather than just stating that it is
too strong an intuition to be dropped. Soteriou 2001 provides a powerful
argument—based on cases of veridical misperception—for the view that
visual experiences cannot be given accuracy-conditions independently of
how we settle the questions of whether an object is perceived and which. I
freely restate it.
According to the content view, if S misperceives a part of his
surrounding environment, then S’s experience represents that part as being
different from the way it actually is. Now suppose you are wearing
displacing glasses such that you see a round red object but it looks to you
to be located to the left of where it actually is. You misperceive the object
because your PE represents it as being in a wrong location. At this point, if
we put an identical round red object at the real location where your
experience wrongly represents the seen object to be, we obtain a veridical
misperception: a seen object is causally responsible for the experience.
Your experience as if there is an object with such and such properties at a
certain location is now the case so that general content is satisfied,21 thus
20 “Veridical” hallucinations are not veridical as perceptual experiences. That
does not prevent certain correspondent beliefs from being true. If I form the belief
that there is a clock of the shelf at a certain distance from me in virtue of having a
certain hallucination, I form a true belief, even though it is an unjustified belief of
course.
21 If the new object is also perceived on the left of the location it actually
occupies, we can put a third round object in the real location where the second
round red object is represented to be by S’s experience. We will end up by
constructing a perfect veridical misperception as soon as the new object is
displaced by the glasses, out of S’s visual field. Suppose the surroundings covered
160
Chapter Four Part II
we have a “veridical” misperception. But to really call it veridical entails
treating it as fully accurate, therefore we should implausibly reject the
following principle: if some part of the subject’s environment is different
from the way it is represented to be, the experience cannot be fully
accurate so it must be partially non-veridical. How is it possible that parts
of the perceived environment are different from the way they are
represented in the PE, and the PE still be veridical? It is not; it cannot be.22
So we have a case in which there is misperception, and so non-veridical
perception, even if the general content is satisfied. This means that the
veridicality conditions of a PE cannot be settled independent of which
objects are perceived and a fortiori of whether any object is perceived.
Tye 2011 provides a similar example. Suppose that unknown to you
there is a mirror in front of you placed at a forty-five-degree angle; behind
it there is a yellow cube. To the right of the mirror and reflected in it there
is a white cube, which, due to unusual lighting conditions, looks yellow.
Now, in your PE, it looks to you as if there is a yellow cube at a certain
location, and that is the case, since there really is a yellow cube at that
location, therefore the existential-general content is fully satisfied. If that
is the content of PE, then that PE should count as fully accurate despite the
evident fact that you are perceiving a white cube as being yellow, and as
being in a location where it is not. So, if the question of veridicality could
be settled independently of which object is being perceived, we should
treat as accurate experiences that clearly represent certain objects as being
by S’s visual field contains three round red objects O1, O2, O3 respectively at
locations L1, L2, L3 and S’s experience represents three objects as located at L1, L2,
L3, as is the case. However, O2 is located where O1 is represented to be, O3 is
located where O2 is represented to be, O3 is not represented because it is out of S’s
visual field.
22 Soteriou’s argument could be challenged as follows: In “veridical
misperception” there is actually no part of the environment that is represented as
different from the way it is. Take an objective portion of the surrounding
environment and describe how it is according to your PE: you will correctly
describe what is there. At L there really is a round red object, for example.
Soteriou would reply that your PE represents that object you perceive, as being at
L, not the object that actually is at L. So he is re-stating that perceptual
representation is demonstrative, namely, that it has particular contents. I do not
want to say that the argument is fully circular or formally wrong. Rather, I take it
as a good argument, but it is based on another intuition, namely, that if you
perceive O and in virtue of that perception your experience represents that there is
an object at L, then your experience represents that perceived object as being at L
not just that there is an object at L. It is still a (legitimate) appeal to your intuitions.
Objects and Properties: How They Feature in Perceptual Content
161
other than they are.23 We need to justify the intuition that I misperceive
that cube, and that I misperceive that round red object, so the experience is
a misperception insofar as that cube is not as it seems to me, that that red
round object is not where it seems to be to me, even though another cube
is the way PE represents that cube to be, and another red object is where
PE represents that object to be.
As the above examples neatly show, strictly speaking there are neither
veridical perceptual hallucinations nor veridical misperceptions—given
that veridicality entails accuracy—even if the world satisfies a certain
general content associated with them, like “there is an O that is F at L.”24
Something is missing from that sort of content: the very perceived
particulars seem to be the best candidates for filling that semantic gap,
thus the GT seems to be flawed.
Alternatively, Searle’s causal plus self-referential conditions provide
the resources to make these cases of misperception non-veridical because
the experience that there is an object O at location L is not caused by the
object that is at location L. So it is true that there is an object that is F and
G at location L (the general-existential content is satisfied), but it is not
true that there is an object F and G at location L and that object is causing
this very experience of there being an object F and G at location L. Given
the Soteriou/Tye argument summed up above for the view that veridicality
23 As in the previous example by Soteriou, that case is not a knockdown
argument—rather it is a way of making our intuitions more evident and
compelling. In fact, if you hold that perceptual content is general and existential,
you may hold that the above PE is accurate. The intuition that it is inaccurate
depends on the intuition that your PE represents this particular as being such, not
only that there is a particular that is such. But this is exactly the thesis to be argued
for; therefore it cannot be used as a premise to defeat the opposing view. Still, the
intuition is so strong that to abandon it entails abandoning in toto our ordinary
notion of perception as well as our relative spontaneous ascriptions of
accuracy/inaccuracy. That can constitute an argument to the best explanation, in
my view.
24 One could defend the GT by arguing that a hallucination that perfectly matches
the surrounding environment is “veridical” after all, at least in some sense, and if
perceptual content is to be evaluated by reference to a perceived object, then
hallucinations would be not evaluable as “veridical” or as “falsidical.” As I will
argue for later (see Chapter Six, Part II.5), hallucinations are in fact neither
veridical nor falsidical, and the intuition that a “perfect” hallucination is veridical,
at least in some sense, depends on the veridicality of the beliefs immediately
produced by hallucinatory states as their natural cognitive effects. A belief that
“there is an O that is T at location L” is well veridical and truth-evaluable, but it is
not the case for the hallucination the belief is brought about by.
162
Chapter Four Part II
cannot be settled independently of which object is perceived and on
whether an object is perceived, then we need either to drop GC itself—so
to admit particulars into the content—or accept the causal and selfreferential characterization of experiential content provided by Searle.
There is no other way.
II.4.4 The Implausibility of Searle’s Account
Searle’s view is implausible in many respects. Many opponents, including
Soteriou himself, have emphasized some of its weaknesses.25 Let us
consider the most obvious ones.
First, there is an issue of phenomenological adequacy.26 If perceptual
content has somehow to reflect perceptual phenomenology,27 then it is
difficult to accept that the immediate phenomenology of perceptual
experience involves our awareness of a causal relation between the
apparent object and the experience itself. From that point of view, the selfreferential component is even harder to accept than the causal one.28 Can
we really make sense of a child or a non-conceptual animal having an
experience whose content involves a reference to its very experience being
caused by an object’s being such and so? Such articulated content seems to
be adequately ascribed only to cognitive systems endowed with
introspective powers and capable of structured propositional attitudes.
A possible defense could insist that just because perceptual content is
non-conceptual, the subject does not need to possess the concepts of
[causation], [one’s own experience] and the like in order to have a mental
state with these objective contents.29 Nonetheless, even if perceptual
content is non-conceptual, there is a threshold of semantic complexity that
cannot be crossed in the ascription of intentional mental states to nonconceptual beings. In order for a mental state to have certain contents, at
least the subject credited with that mental state must be a conceptual
being. You cannot entertain the content [quark] or [neuron] or [toaster] if
you don’t possess conceptual abilities at all. Likewise, it is implausible
that the self-referential content [myself being caused by the object’s being
the way I represent it to be] can characterize the perceptual experience of
25 See Soteriou 2001, Armstrong 1991, Burge 1991, McDowell 1991, Millar
1991.
26 On that, see also Chalmers 2006.
27 On that, see Part I of the next chapter.
28 For example, scholars like Siegel 2006, 2010 and Butterfill 2011 hold that in
visual experience we do represent causal relations.
29 This line of defense is taken by Chalmers 2002.
Objects and Properties: How They Feature in Perceptual Content
163
an animal or a child. In fact, from a developmental point of view, our
experiences have contents before we as children begin to entertain the very
content of [experience],30 be it conceptually or non-conceptually.
In addition, there is also the issue of deviant or non-standard causal
chains to consider.31 An object being a certain way may well cause my
experience of that very object being that way, but it may cause it in an
inappropriate way. Either these cases of non-standard causation satisfy the
content and make PE veridical and accurate, or the normative clause of
appropriateness should also be included within the content to the effect
that a PE has content like: this F is G, and the fact that F is G is causing
this very experience that this F is G in an appropriate and non-deviant
way. Is this not too much?32 As is well known, it is very problematic to
spell out what this “appropriateness” amounts to.33 Such a highly
theoretical notion, which is so puzzling even for theoretical thought itself,
should be experiential content! So the phenomenological adequacy
constraint is violated on the one hand, and the view is developmentally
implausible on the other. Indeed, the content of perceptual experiences of
children and non-conceptual animals should contain a sort of sketchy
theory of perceptual experience, including clauses of appropriateness in
causation!
Secondly, content-ascription should be constrained not only by
phenomenological elements but also by considerations about explanatory
relevance with respect to behavior and observable cognitive abilities.34
Ascribed contents should not be over-sophisticated with respect to the
discriminatory abilities shown in behavior, especially by non-linguistic
animals,35 if such contents are posited just to account for these very
30 On children’s development of the power to ascribe and self-ascribe experiences
and perceptions, see Nudds 2011.
31 See Grice 1961, Lewis 1980.
32 On this point, see Tye 2011.
33 As is well known, that is perhaps the problem for causal theories of content. If
the content of a mental state is taken to be determined by what causes it in “normal
conditions” or in “appropriate causal chains” and the like, the problem is just that
such notions of “normality” or of “appropriate causation” are normative notions
that go beyond the merely causal characterization against the reductive intentions
of the causal semantics theorists.
34 This point has been duly emphasized by Soteriou 2001.
35 Obviously, in evaluating cognitive abilities of non-linguistic animals, we only
have their observable behavior.
164
Chapter Four Part II
abilities.36 Thus, not only are causation and self-reference not testified to
in our visual phenomenology, they are not even necessary to explain the
discriminatory abilities of perceivers.
So the phenomenological adequacy constraint and behavioral
discriminatory constraint may be thought of as not absolutely normative if
considered distributively. For example, it may well be the case that certain
visual contents are not testified to in our visual phenomenology37 but they
are to be postulated in order to explain certain discriminatory abilities we
have. Conversely, our visual phenomenology may make our experience
seem richer and the contents more detailed than it really is. So, nothing
prevents visual content from being richer or poorer than what
phenomenology and introspection “tell” us. Nonetheless, we assess that
gap just by evaluating certain discriminatory abilities or certain
discriminatory inabilities. But if a certain content does not explain either
distinctive features of phenomenology or any discriminatory ability, then
ascribing that content is explanatorily redundant. What I am arguing for is
that the two constraints for content-ascription introduced above are weakly
normative if each of them is considered independently but they become
strongly normative if taken together to the effect that if a certain contentascription violates one of them, there must be a good reason, and that
reason must be grounded in the necessity of respecting the other constraint
(for example, to explain certain discriminatory abilities or inabilities it
may be necessary to “discredit” certain elements of our visual
phenomenology). But an ascription that violates both constraints must be
mistaken or arbitrary at the very least.
Thus, no independent reason other than the necessity of saving GT can
support Searle’s account of perceptual content but the implausibility of
that account is theoretically much more pressing than the opportunity of
saving GT.38 If perceptual content is purely general, we need to buy
36 As Soteriou points out, “What discriminatory abilities are left unexplained if
one does not include the causal component in the content of visual experience?”
(2001, 183)
37 For example, I have considered the cases of blindsight, inattentional blindness
and the Sperling effect. See Chapter Two, Part II.
38 As Tye 2011 notes, in any case the GT cannot commit to a purely general
content; rather it is inevitable to refer to some demonstrative elements in
specifying that general, existential content. For example, there must be a reference
to a particular time, like now, as well as to a particular subject, me. For example,
according to my PE, there is a round red cube at a certain distance and direction
from me, now, not yesterday or with respect to another point in space. So a
particular subject, a particular time and a particular egocentrically specified place
need to influence the accuracy-conditions of the PE as well as its content. Even if
Objects and Properties: How They Feature in Perceptual Content
165
Searle’s implausible account. Therefore, perceptual content is not purely
general but contains particulars.39
II.5 Demonstrative Contents and the Semantic Gap
Given this argumentative vindication of the particularity of visualperceptual content, we can go back to our Content/content distinction
introduced above. The object-dependent Content of PE is made out of the
object plus a content (= properties the object is represented as having). A
de re or demonstrative Content takes the following form:
of/about O: PE represents properties F, G, H
At least when there is a perceived object, perceptual Content is objectdependent. If the object is an accuracy-maker, it is such only insofar as it
has certain properties. It can be an accuracy-maker by being part of the
evaluable content because, strictly speaking, its properties are the entities
whose eventual matching with the experience need to be evaluated. So the
object is what the experience is an experience of, what the represented
properties belong to, if the experience is accurate. To be more precise, the
object does not make the experience accurate/inaccurate; rather it makes
the experience accuracy-evaluable by being part of its Content, whereas
the accuracy depends on the matching-relation between the content of the
PE and the properties of the perceived object.40
Illusions are made falsidical not by the object but by the properties
belonging to the object the perception is a perception of. The seeingrelation fixes the aboutness of the representational episode involved in that
seeing, where such aboutness is not itself represented, not even in terms of
causal factors pace Searle. Seeing-episodes are contentful episodes in
which certain properties are represented as possessed by the very seen
content were general, it should be impurely such. If there were no reference to
perceived particulars, still there would be reference to other particulars.
39 For further arguments in favor of the particularity of visual perception, see
Sainsbury 2006. Chalmers 2004 and Siegel 2010 have a mixed view according to
which PEs have both singular and non-singular contents.
40 The perceived object need not be a thing like an apple or a table. There can be
objects like a rainbow, the sky, or a foggy area in the air, or whatever. Normally
perceptual experiences are relations of an objective global scene populated with
many objects having properties and entertaining relations with each other.
Therefore, when I talk of “the perceived object,” it is just an oversimplification to
make the exposition and the reasoning clearer.
166
Chapter Four Part II
object, even if it is only by representing some properties of it that a seeingepisode can take place. So, seeing-episodes have an object by virtue of
having a content, but they have a certain Content by virtue of having a
certain object as part of that Content itself. Change the seen object and the
Content changes. That is the object-dependency of PE. Relationality is
essential to the Content of a PE insofar as a PE is essentially related to the
world.
There must be an object not only for a PE to be accurate but also for a
PE to be inaccurate. Misperceptions are inaccurate representations of the
respectively perceived object in question.
What are hallucinations inaccurate representations of? That delicate
question calls for a different treatment of illusory Contents on the one
hand, and hallucinatory “contents” on the other. Illusions are perceptions;
hallucinations are not. Perceptual illusions and veridical perceptions have
the same semantic structure, an object-dependent Content, made accurate
or inaccurate by certain properties of the very perceived object.
To recall our analysis of look-ascription in Chapter One, perceptual
Contents are fittingly expressed by ascriptions like “O looks F to S”
instead of by ascriptions like “it looks to S as if O is F.” The latter, indeed,
is neutral on the existence of O as well as on S’s PE being a perceptual
relation to O.41
Particularity of visual experience also fits with visual phenomenology.
In perceptual experience it is not just as if there was an X that is such and
so, rather we seem to be demonstratively related to particulars, to this and
that, so that this and that visually look such and so. Perceptual awareness
is the basic way of being acquainted with worldly objects. If perceptual
content were general and existential, our perceptual access to the world
would be description-like, not an acquaintance.
However, the issue of hallucinatory contents hinted at above specifically
arises by considering that hallucinations have a “demonstrative”
phenomenology as well, despite the absence of an object and, therefore,
despite the absence of Content (object + content). If perceptual Content is
object-dependent, the ascription of hallucinatory content, if any, is a
dilemma; since hallucinations are objectless states by definition, they are
non-relational states. In short, they are not perceptions.
In addition, there is a strong intuition that hallucinations are inaccurate,
but that does not fit with the real properties of perceived objects being the
accuracy-makers of perceptual experiences. By definition there is nothing
whose properties can make a hallucinatory content accurate or not, just
41 See Chapter One, Part V.3.
Objects and Properties: How They Feature in Perceptual Content
167
because there is nothing a hallucination is a perception of. If in order to be
inaccurate an experience must be accuracy-evaluable in the first place, it
must involve an object in its Content; we conclude then, against strong
intuition to the contrary, that hallucinatory experiences are neither accurate
nor inaccurate but rather they lack semantic value.
Object-dependency of perceptual content, together with the absence of
an object that characterizes a hallucination, leads to what I will call the
puzzle of the semantic gap, expressed by the following scheme:
Perceptual Experience (PE):42
object + content = Content accuracy-value
a) Veridical PE: content exemplified by the object→satisfied
veridical
b) Illusory PE: content not exemplified by the object→not satisfied
falsidical
Hallucination (H):
------- + content = ? falsidical?
The semantic gap is a problem that any view on perceptual content
faces, which takes particular perceived objects as essential ingredients for
the truth- or accuracy-evaluable Content of PE. If the object is essential to
perceptual content, hallucinations should be Content-less mental episodes
insofar as they are objectless mental episodes. This is highly counterintuitive, though, since in hallucinations the world seems to us to be
certain ways—at least it seems to seem so—and hallucinations may be
subjectively indiscriminable from genuinely perceptual experiences. So
how can they have same phenomenology but no Content? How could they
be Content-less in the first place? If these problems were not satisfactorily
addressed, the particularity thesis (PT) would be flawed.
Actually, GT straightforwardly avoids the problem of the semantic
gap: if perceptual Content is existential and general, hallucinations are not
42 Even if it is unusual in the literature, I will use “perceptual experience” only for
illusions and veridical perceptions, not for hallucinations. The reason is that
perception is a relation to the world; hallucinations are not genuine relations, so a
perceptual experience is really perceptual only if it involves a perception. A visual
hallucination is an experience that is indiscriminable from a perceptual experience.
That indiscriminability, though, is not sufficient for making it a perceptual episode.
A hallucination is an experience as if a certain perceptual experience was taking
place. That experience is perceptual, so to say, only from the very (misleading)
point of view of the experiencer.
168
Chapter Four Part II
semantically gapped at all; instead they are just like veridical and illusory
perceptions. Hallucinating a red cube, veridically perceiving a red cube or
perceiving a green cube that falsidically looks red would be states with
exactly the same existential content or that there is something that is red,
which is a cube, which is located at L. But, as we have seen, in order to
distinguish “veridical” hallucinations from both illusions and veridical
perceptions, the view must implausibly embed a causal and self-referential
element into the content.
So we are left with a dilemma: the GT (in its existential version) is
implausible, and the PT only seems much more plausible—for the many
reasons hinted at above—until it is asked to account for the possibility of
hallucinatory Contents.
A reason for GT, sometimes stated and more often presupposed, is the
intuition that if two experiences are phenomenally identical, they must
have the same Content. For example, McGinn argues that perceptual
content must be general “on pain of denying that distinct objects can seem
precisely the same” (1982, 39). Take two experiences—PE1 and PE2—
with respective objects O1 and O2, where O1 and O2 are visually identical.
Now, if perceptual content is object-dependent, then PE1 and PE2 have
different Contents, since PE1 is of O1 whilst PE2 is of O2, even though they
are phenomenologically identical because O1 and O2 are visually
indistinguishable. The properties O1 is represented as having in PE1 and
the properties O2 is represented as having in PE2 are the same properties;
the experiences share their phenomenology but their Content is different
since it involves different environmental objects whose properties are
respectively represented.
The example of the twin objects O1 and O2 can be used with the case of
a hallucination of an object identical to them. Again, the hallucination has
ex hypothesi the same phenomenology of PE1 and PE2: in having it, it
seems to S that there is an object O3, and O3 “seems the same” as O1 and
O 2.
However, this internalist intuition in favor of GT is not necessary.
Externalist theories of Content just deny that a causal, external factor
contributes to the determination of the Content in a way that may be not
reflected in phenomenology. After all, it is O1 that is represented in PE1,
not O2. After all, a hallucination “of” O3 is not a perception; still it
introspectively seems exactly the same as a PE.
Two different but connected principles seem to be implicitly making
people find GT so compelling. One could be called the Cartesian principle
(CP): according to which the very nature and type of a certain conscious
experience must be fully available to the subject by introspection so that:
Objects and Properties: How They Feature in Perceptual Content
169
Cartesian principle: If two mental states are subjectively indiscriminable, they
have the same nature and the same type.
The CP entails that a veridical experience (VE) and a hallucination (H)
subjectively indiscriminable from VE must be the same kind and type of
mental state.43 So, if we combine that principle with the intentionalist idea
that the Content of a mental state individuates it and is essential to it, then
we are compelled to conclude that a VP and an indiscriminable H must
share their Content.
According to another principle, which could be called the accessibility
principle (AP), the Content of a certain experience must be fully available
by introspection so that:
Accessibility principle: If two experiences are subjectively indiscriminable,
they must have the same Content.
According to the CP, if a veridical perception indiscriminable from a
given hallucination is a contentful state, then the hallucination is a
contentful state thereby. It is ruled out by that principle that a state could
subjectively seem to be contentful without being such.
According to the AP, not only is being indiscriminable from a
contentful state sufficient for being contentful but also being
indiscriminable from a state that has a certain Content is sufficient for the
state having exactly that Content.44 That rules out object-dependency,
because not only can a hallucination be indiscriminable from a perception
without having objects at all but also a perception of a qualitatively
identical but numerically different object than the object of a given
perception.
So, hallucinations must be contentful, experiences of identical objects
must be identical in Content, therefore perceptual Content cannot involve
43 For a strong criticism of what I am calling the Cartesian Principle, see Martin
2002, 2004, 2006. Martin argues against a similar principle to support a
disjunctive, naive realist view of perceptual experience. As I will argue later on
(see Chapter Six), not only may such a principle be denied by an intentionalist but
it must be denied by a coherent intentionalist.
44 It could be, though, that AP “screens off” CP in being sufficient to rule out that
two subjectively indiscriminable states are of the same nature and type, as well as
having the same Content. But without CP, AP could well allow that two
indiscriminable states have the same Content without being of the same kind and
type. For example, one could be a perceptual experience with content C; the other
could be another type of contentful state (say a desire, an intention, a belief) with
content C.
170
Chapter Four Part II
particulars. GT must be true if CP and AP are true.
Actually both principles are false, as I will argue at length later on.45 I
will also argue that hallucinations are contentful states in a certain special
sense but not because of the truth of the CP. Indeed, the CP is false and
hallucinations are mental states of a different kind from perceptions, be
they veridical or falsidical. The GT must be rejected for the reasons
already provided above but it is important to make clear that the implicit
principles that seemingly make it so compelling are false principles.
Before saying anything more about hallucinatory contents, I want to
stress a relevant consequence of the PT with respect to the relation
between intentionality and phenomenology, between representational
content and phenomenal character. Intimate as the relation is between
perceptual Content and phenomenal character, that relation cannot be one
of identity. At least the Content can change despite constancy of
phenomenal character; for example, you can remove O1 and substitute it
with O2 or with nothing if you rightly stimulate the subject’s brain so as to
bring about a matching hallucination. Content changes, phenomenal
character does not: for sure this is plainly incompatible with the identity
theory, often called strong representationalism.46
Therefore perceptual Content, as involving the perceived object, is
wide insofar as it depends on the world the subject is causally connected
to, namely, on which object she is perceiving as well as whether she is
perceiving anything at all. Whether what I have called content—the
45 See Chapter Seven, Part II.5.
46 Byrne 2001 distinguishes weak representationalism (the supervenience-thesis)
from strong representationalism (the identity thesis) about phenomenal character.
Another relevant distinction is that between intermodal and intramodal
representationalism: for intermodal representationalists (Dretske 1995, Tye 2000),
the represented properties fully determine the phenomenal character, so seeing
something as [overhead] and hearing something as [overhead] is phenomenally
different only because in each case other properties are represented besides the
[overhead] property (for example, colors in the first case; pitch, volume and tone in
the second case). Another distinction is between restricted and unrestricted
representationalism: according to the first, any conscious experience has a content
that determines its character (Dretske 1995, Tye 2000, Bain 2003); according to
the second, only certain phenomenal states (like perceptual experiences) have
content, and that determines the respective phenomenal character of the state. Here
I am arguing for a weak and intramodal repressentationalism. I find the
unrestricted representationalism implausible but since I am concerned with
perceptual experience, that issue will be not relevant: what matters is that
perceptual experiences are contentful states and their contents determine their
conscious character.
Objects and Properties: How They Feature in Perceptual Content
171
properties the perceived object is represented as having—is wide or
narrow is another huge but independent problem.47 Even if it was narrow,
that could not prevent the Content it is a part of from being wide.
In the next part, I will deal with the issue of the relation between
phenomenal character and representational Content in a general way. In
order to finally face the problem of hallucinatory contents—the semantic
gap problem—we have to patiently wait until Chapter Seven.
We are left with the difficult problem of the semantic gap. As we will
see in the next chapter, phenomenal character is determined not by the
object but by the properties the PE represents the object as having, so
Content is object-dependent but phenomenal character is not. Now, the
properties the PE represent the object as having are “wide” in another
sense from that in which the Content is wide as being object-involving:
they depend on external factors in a historical way, which is still to be
explored. In any case, given that the object does not determine the
phenomenal character, we still need to account for the content of
hallucinatory states, namely, for the semantic evaluability of objectless
mental states that are phenomenally identical to perceptual states whose
evaluability essentially involves perceived objects. Before finally facing
this problem, in the next chapter I will complete my positive articulation
of the content view on perceptual experience.
47 To report Chalmers’s clear and simple definition, “A property is narrow when
necessarily, for any individual who has that property, an intrinsic duplicate of that
individual has that property (regardless of environment). A property is wide when
it is possible for an individual to have the property while an intrinsic duplicate
lacks that property” (Chalmers 2004, 108). So, if the instantiation of a mental
property, like having a certain content or having a certain Content, depends on
external factors, that property is wide. If it is independent from external factors,
then it is narrow.
CHAPTER FIVE
PHENOMENAL CHARACTER AND KINDS
OF PERCEPTUAL CONTENT
PART I
PHENOMENAL CHARACTER
AND REPRESENTATIONAL CONTENT
I.1 Reconsidering Looks Within the Content View
The content view can vindicate a relevant sense of ordinary ascriptions of
“O looks F” by equating them to ascriptions or self-ascriptions of
experiences that represent a seen object as being F. That is the sense I have
called epistemic, following Chisholm and Jackson among others.1
Something O looks F when it looks to be F according to one’s experience
of it, so when there is prima facie visual evidence that it is F. Such lookattributions are objective. O exhibits a certain objective look, a way of
appearing, which may well justify the proposition that O has a certain
property, the very property it looks to have.2 When O looks F in that sense
of looking, F is the objective property the experience attributes to the seen
object. The content view suitably accommodates the idea that something
can look F without being F: representational contents may not be satisfied
by the world.
So, when O looks F to S, S has a visual experience of a perceived O
that represents O as F, where O is the object, F is the content and the
couple [object + content] together composes the Content.
Now the Content of a PE consists of its veridicality conditions. So if
the seen object has the property the PE represents it as having, then the PE
is veridical or accurate, which means that the represented property is the
very same property actually had by the seen object if the PE is accurate.
1 Actually, it would be more opportune to call it doxastic because calling it
epistemic may give the impression that it has to essentially do with knowledge,
whereas it has to do with prima facie evidence for beliefs or belief-like states
instead. Moreover, it could be confused with Dretske’s “epistemic seeing” with
which it has nothing in common.
2 If other collateral knowledge does not defeat that evidence, other things being
equal, and so forth.
Phenomenal Character and Representational Content
175
Properties or relations like F could be [red], [square], [moving in S’s
direction], [distant from S], [bigger-than], and the like: they are intentional
properties, or representational properties shaping the Content of PE,
together with the perceived object/objects.
A PE is a conscious episode with a phenomenal character, a way it is
like to be in it. Let us call a property, among those that constitute the
phenomenal character of a PE, a phenomenal property.3 So, a PE has
intentional properties and phenomenal properties. Intentional properties
are those by virtue of which the PE is accuracy-evaluable so it is a
contentful state; phenomenal properties are those by virtue of which the
PE is a conscious episode so it has a certain subjective character.
What is the relation between the intentional and phenomenal properties
of a PE? How are we to pick them out in a given PE? How could we ever
assess whether and how they are related, whether they are identical or
different, whether they overlap or not?
A way to investigate the matter is to consider the other ordinary lookascriptions analyzed in Chapter One in order to determine whether and
how the content view can also vindicate those other uses,4 or whether, at
least, it may be compatible with them. It may well be that the nonepistemic uses—other than (prima facie) “looking to be”—somehow map
or track other non-intentional features of PEs.
The comparative use attributes to two things a common or similar
distinctive appearance, as in “A looks like B.” Attributing such a
commonality of appearances by no means represents a sameness of the
compared things. A can look like B without evidence that they are
objectively the same in some respect, apart from that partial sameness in
their appearances themselves.5 I have already argued that comparative
look-ascriptions presuppose another non-comparative use of “looks,”
otherwise the comparative use would be circular: A looks like B, but how
does B look? Like A, of course, but how does A look? Like B? If B looks
like C, well, how does C look? And so on and so forth. There must be a
way A and B look on the basis of which the very comparison is made, and
3See Chalmers 2006, Siegel 2010.
4Of course, the issue of the relation between phenomenal character and
representational content is an issue internal to CV. Indeed, only for those who hold
the CV does PE have representational content at all.
5If something a looks like something else (an F), they must have something in
common, but not necessarily the property F used to specify what a looks like. For
example, a looks like an F so it must have some property in common with F (at
least a certain distinctive appearance), but that property need not be the property F
itself.
176
Chapter Five Part I
that “way” cannot be attributed by an epistemic look-ascription because if
it were, then “A looks like B” would amount to “there is (prima facie)
evidence that A is F and there is (prima facie) evidence that B is F as
well.” But that would amount to “there is (prima facie) evidence that A
and B are the same, namely, items of the same kind or type F.” But we
ruled out the idea that comparative looks involve such objective
commitments, as then your looking like a poodle—in certain
circumstances—should entail prima facie evidence that you are a poodle!
The non-epistemic and non-comparative use of “looks” implicitly
involved in both these uses6 is what I have called the phenomenological
look. The existence of a phenomenological look explains why something
can look red without looking to be red; something can look elliptical
without looking to be elliptical and so forth. This look does not seem to
directly concern intentional properties since intentional properties shape
perceptual Content and perceptual Content consists of the PE’s veridicality
conditions—O’s being F would make veridical and accurate the PE
according to which O looks F, but in that case “looks F” should amount to
“looks to be F” given that the PE is true if O is F.
I.2 The Case of Perceptual Constancy
I.2.1 Perceptual Constancy and the Ways of Looks
The well-known phenomenon of perceptual constancy is a good account
of such non-intentional use of “looks.” Standard cases of perceptual
constancy, for example, are provided by Peacocke 1983, 2007 as evidence
for the existence of “primed” or “sensational” properties, namely,
properties that subjectively characterize the perceptual experience without
being represented in the experience as had by the perceived objects
themselves, as instantiated by the perceived world: the ways things look
that are not the ways things look to be.
Here is an example, which is not a case of perceptual constancy but it
still involves it. You see two trees, the bigger one is far away from you and
6 As I made clear in Chapter One, the epistemic look also involves the
phenomenological look. The visually acquired evidence that a is F, in “a looks F,”
is visually acquired indeed: such a visual acquisition of evidence about how the
world is arranged must consist of a conscious appreciation of certain visible ways
things appear. That does not prevent the phenomenological use from being
dependent on the other uses as well (especially on the epistemic or intentional
one). The uses are deeply interwoven, without any of them being causally or
explanatory prior to the others.
Phenomenal Character and Representational Content
177
the smaller one is very near to you. Still, there is a seemingly legitimate
sense in which their sizes look the same from your point of view without
looking to be the same.7 Actually, it is the same visual phenomenon
involved in size-constancy, only comparatively considered. By virtue of
size-constancy, when you see an object moving toward you, even if it
comes to occupy a bigger and bigger portion of your visual field as it gets
nearer, and despite the phenomenological change, your PE represents it as
being constant in size. But even if it looks to be the same in size, it seems
to look different as well; at least it looks different as given in visual
phenomenology. Something changes in the way the object appears.
Imagine the object is round but while moving toward you it slowly
rotates in such a way that at certain moments its oblique orientation
changes how its actual size occupies your visual field so that it is
sometimes an ellipsis.8 Despite the phenomenal changes due to the
respective changes in orientation when the object is rotating, your PE
keeps representing the object as being constant in shape, namely, as round.
Still, it is hard to deny that at certain times the object “looks” elliptical as
well, even though your PE represents it as being round and makes you
believe that it is round.
Imagine also that parts of this rotating object are perfectly illuminated
by the sun while on other parts some shadow is cast. There is a sense in
which your visual experience of the object is not chromatically
homogeneous and, indeed, you notice a sharp difference between the fully
illuminated parts and the shadowed parts. Suppose the object is yellow.
Your PE represents the object as being yellow but the shadowed parts of it
look “greenish” in some sense, despite there being no doubt that they look
to be yellow in the first place. Even if the distribution of shadowed and
non-shadowed parts changes during the rotation, that phenomenal change
does not amount to your PE representing parts of the object as suddenly
changing color. This is color-constancy.
Why should size-constancy, shape-constancy and color-constancy
make a case against the idea that the only properties of perceptual
experiences are intentional properties? Because your PE represents things
as being the same through time or space, but something else, something
phenomenally relevant, is changing.9 Conversely, your PE may represent
7 Peacocke (1983, 12) calls it the “problem of Additional Characterization.” See
also Byrne 2001, 221ff.
8 On sensational properties as irreducible to represented properties, see Peacocke
2007.
9 It is important to remark that the sameness of size, shape and color despite other
phenomenal changes is not just judgmental, it is phenomenologically given in PE.
178
Chapter Five Part I
things as changing despite a sort of phenomenal sameness of their visual
look. Suppose an object becomes smaller and smaller in such a way that
its moving toward you is phenomenally “compensated” for by its
becoming smaller, so it “looks” constant to you, in a certain sense of
“look” at least. Suppose you experience a wall whose surface
phenomenally looks uniform despite the fact that parts of it are clearly
shadowed. Suppose a round object really gets elliptical but at the same
time changes its orientation in a way that the portion of your visual field
covered by it is constant, such that it “looks” to you constant—same
represented properties with phenomenal change; different represented
properties without phenomenal change. So there must be phenomenal
properties beyond the representational ones, and which are irreducible to
them, in PE. This is the inference, in short. If A and B do not
systematically co-vary, they cannot be either identical or related by a
relation of supervenience. Given A, it is not given B thereby, and vice
versa. After all, if both “look”-properties were representational, our PE
would have contradictory contents. The same object would be represented
by the same PE as being round and elliptical at the same time, as being big
and small at the same time, as being yellow and green at the same time.
But that is impossible, not so much because PEs with contradictory
contents are a priori impossible10 but because we experience that
phenomenon of sameness-despite-phenomenal-difference and differencedespite-phenomenal-sameness as something quite natural, as what exactly
is to be expected in PE, as something far from requiring contradictory
I do not just judge that the tilted coin is round “despite appearances”; rather the
coin visually appears to be round (despite it presenting, in another sense, also an
“elliptical appearance”). If it did not violate the phenomenology, treating the
sameness as a result of a judgment could have been a way out. But it is not an
available way, given the perceptual phenomenology of constancy. Indeed, your
judgment of sameness does not contradict your experience at all, as may happen in
cases of perceptual illusions known as such; rather it endorses the experiential
content itself. On the contrary, it is only by reflection that I become conscious of
the “elliptical appearance” of the tilted round object, of the “different appearance”
of the shadowed and unshadowed parts of the surface, of the different “sizeappearance” of the same object seen from near and from far away. Instead,
perceptual constancy is easily achieved by infants much before they show any
introspective abilities. On perceptual constancy in infants—even in newborns!—
see Slater 1998.
10 On the contrary, there are experiences with contradictory and/or impossible
contents. For example, Escher figures, the waterfall illusion, the Penrose triangle,
and so on. See Crane 1988, Ernst 1996. For experiences of impossible colors, see
Churchland 2005.
Phenomenal Character and Representational Content
179
beliefs endorsed by us!
In addition, if the same objective color can look many ways (e.g.,
green, yellowish, orange), then “looks F” cannot just individuate one
objective color represented as had by an object in PE. The same goes for
sizes and shapes. Conversely, if a certain way of looking (be it F) can
represent to the subject different objective colors (or shapes or sizes), F
cannot be a representational property such that the PE representing it is
true if the perceived object is F full stop.
The relation between the phenomenal level and representational level
of PE must be more complicated so it is in need of a positive account. Let
us ask: given that there are phenomenological looks that are not directly
and straightforwardly amenable to intentional looks, what is the relation
between them? What are phenomenal properties in the first place?
It would be natural to think that the phenomenological look is
irreducible to the epistemic look the same way as the phenomenal
properties of PEs are irreducible to their intentional-representational
properties. Still, there are some ways one could reply to that idea.
I.2.2 A Possible Reply Against Representationalism
A radical counter-move is to plainly deny that there are ways things look
without them looking to be these ways thereby. Therefore, if O really
looks F in S’s PE, then O looks to be F such that if O were F, S’s PE would
be accurate and veridical. There are only intentional properties in PE’s
phenomenology; there are only intentional looks that can be correctly
attributed to the world as experienced.
For example, when you see the two trees in the case above, the far one
looks to be much bigger than the near one, whereas there is no clear and
independent sense in which the two trees look the same size. They just do
not look the same in size because distance is visually represented, so the
distant tree appears the way distant trees of a certain size are supposed to
appear to an observer. The same applies to the closer tree: it appears as it
should appear when seen from a point.
When you see a tilted round object, it just looks to be circular; there is
no sense at all in which it looks elliptical. Orientation in space is visually
represented so the way the object looks is the way a tilted round object is
supposed to look to an observer so positioned. Nothing looks elliptical.11
11 One could reply that the retinal image produced by the tilted round object is in
fact an ellipsis. But that is a bad reply: why should the way something looks be the
same shape as the retinal image? The retinal image is just a certain gradient of light
distribution; it is a step of the total information processing involved in vision,
180
Chapter Five Part I
When you see a yellow object whose surface is partially shadowed, the
surface looks chromatically uniform; it looks to be yellow not differently
colored in parts. Illumination conditions, as well as shadows, are visually
represented so the way the surface looks is the way a yellow partially
illuminated surface is supposed to look to an observer. Nothing looks
greenish at all.12
In short, it is not true that there are phenomenal changes without
representational changes in perceptual constancy. What phenomenally
changes depends on representational changes; for example, on variations
in represented distance, orientation, illumination and so forth. Phenomenal
differences despite constancy are differences in what is represented by PE
rather that differences in how PE represents it. So, there is no
phenomenological look that is not amenable to an intentional look:
phenomenal properties are completely exhausted by representational
properties. That is the representationalist version of CV. In this version of
CV, not only are perceptual experiences representational episodes but also
their representational properties are the only properties we are aware of in
PE. The phenomenal character of a PE is completely determined by its
representational content. Thus, nothing looks F in PE unless it looks to be
F according to the PE.
I.2.3 A Counter-Reply
I do not want to argue against that reply; rather I want to point out its
insufficiency. Indeed, that view overlooks a genuine phenomenological
feature of visual experience. It is true that the trees do not properly look
the same size, that the surface does not properly look differently colored in
its shadowed and unshadowed parts, that the round object does not
nothing more. Of course, the retinal image has to do with the final conscious vision
of the object, so also with its distinctive looks in experience. But there is no need
for a shape-similarity between looks and causal-physical basis of the visual
episode. Otherwise, why don’t we argue that the way O looks must share a shape
with the excitation of the optical nerve at a certain point? On the inappropriateness
of calling “image” the retinal information, see Nöe 2004, Chapter 2.2.
12 Besides the example of the shadowed surface, color-constancy is responsible
for our experiences of colors being relatively independent from the time of the day.
For example, both in daylight and in the dark, red surfaces appear to be red despite
changes of illumination, just because the illumination conditions are visually
represented. Also, in this case one could reply that in daylight as well as in the
dark, red surfaces just look to be red, so any look is intentional. So it is a mistake
to take it that red things look brown when scarcely illuminated. Instead, they are
scarcely illuminated red surfaces and they look to be such.
Phenomenal Character and Representational Content
181
properly look elliptical. Although that way of ascribing looks is
linguistically inappropriate, still we understand what it means that in
experience the trees “look” the same size, the surface “looks” differently
colored, the object “looks” elliptical. The immediacy in understanding, or
at least vaguely grasping, that strange way of speaking must be somehow
rooted in phenomenology and calls for an explanation. Just glossing over
it is unsatisfactory.
So, what do we grasp in the idea that the trees look the same size, the
surface looks differently colored, the object looks elliptical? As I have
already argued in Chapter One (V.5), we grasp a certain partial similarity
between our actual experience and another eventual experience of
something different. For example, if I were seeing a small but close tree, in
some respects my experience would present it in a similar way to the way
my actual PE presents that big but distant tree. Of course, distance is
visually represented; indeed that similarity holds only in some respects
and not in others. For example, the comparison abstracts from focus,
perspectival relations with other objects and any other features of my PE
that are cues for appreciating distance-properties.13 Grasping such a
hypothetical similarity rests on a highly sophisticated counterfactual
operation but it is grounded on our introspective powers of comparing
actual and possible (or past) experiences. Likewise, illumination
conditions and shadows are visually represented but this does not prevent
a subject from appreciating that the actual experience of a portion of
shadowed yellow is similar in some respects to a hypothetical experience
of an unshadowed green.14 The comparison abstracts from illumination
conditions, shadows and so forth but it is grounded in actual experience.
There is nothing illegitimate in appreciating that this yellow surface with
distributed shadows on it looks the way an unshadowed surface colored
13 We could distinguish a holistic phenomenology of PE, and an atomistic
phenomenology. The overall visual phenomenology of a PE is such that distance is
represented together with size, color is represented together with illumination,
shapes are represented together with orientation, and so forth. Besides that
spontaneous, holistic phenomenology, an introspective focus on details and
particular features of our experienced scene may allow us to isolate certain
“apparent properties” in abstraction from their relations to the global visual
experience. So we can take a sort of atomistic attitude toward our own experience,
and even compare two phenomenological atoms, say, belonging to different PEs.
That sophisticated operation underlies our mastery of phenomenal looksascriptions.
14 I want to point out that it is an objective fact about yellow surfaces and green
surfaces that a yellow surface in the dark is similar to a green surface under the
sun. Nothing intrinsically private is involved in that circumstance.
182
Chapter Five Part I
with green and yellow parts would look to be.
In the same vein, a round tilted object is present in PE in a partially
similar way to the way an elliptical non-tilted object would be presented
by another hypothetical PE. Of course, orientation is visually represented
so in order to make the comparison, the subject needs to disregard those
visual properties that make her appreciate the orientation and take the
object as round instead of elliptical. Still, under certain conditions, an
elliptical, frontally oriented object would be presented in a PE in a similar
way.
What does all this amount to? Is the meaningfulness of comparative
talk about lookings and visual experiences evidence against representationalism?
Different representational properties may be phenomenally similar, and
vice versa, but picking out a phenomenal property F is possible only by
reference to a paradigmatic experience in which something is represented
as F. So, there seems to be a way something looks without looking to be
that very way but only insofar as it looks the way something else would
look to be in another PE.
That asymmetry is of the greatest importance, but what matters here is
the possibility of comparing two different representational properties as
being “experientially” similar despite their objective difference. If yellow
things in bright light, in some respects, look the way green things in dark
light look, then two things can look to be different despite looking similar
in that very respect.
In Chapter One, I argued that the comparative use presupposes a
phenomenological use of “looks” otherwise the comparative use would
involve a regressus. Given that A looks like B, how does B look? So there
must be the phenomenological look of A and of B in order for them to be
compared.
Now we have just seen that the attempt to reduce the phenomenological
look to the intentional look (“looks F” to “looks to be F”) is blocked by the
possibility of comparison between PEs that represent different properties. I
think the following moral could be drawn from all this: if you reduce the
comparative look to the phenomenological look, then you need to admit
the phenomenological look is distinct from the intentional look. If you
reduce the phenomenological look to the intentional look, then you need to
admit the comparative look is distinct from the intentional look. In any
case, the intentional look alone does not cover our ascriptive uses, our
intuitions and our phenomenology. As a consequence, the phenomenological
look is dependent on but irreducible to the intentional look, and that
irreducibility despite the dependency is what also makes the comparative
look irreducible to the intentional. The phenomenon of the elliptic look of
Phenomenal Character and Representational Content
183
the tilted penny (i.e., the phenomenological variations of “apparent shape”
despite perceptual constancy of roundness representation) testifies to the
irreducibility of the phenomenological look to the intentional look. Their
dependency has been argued for before (Chapter One, Part V.5):
something phenomenologically looks elliptical if experiencing it is, in
some specific respect, as if one were seeing an elliptical object from a
privileged (frontal) perspective in paradigmatically optimal circumstances.
So even assuming that we are aware only of representational properties
in PE, our introspective access to our PEs allows us to compare and couple
certain ways these representational properties are given to us in different
PEs. This is true even if we pick out those “ways” as ways a certain object
is represented to be in another paradigmatic experience. These “ways of
being given” are not reducible to the ways the world is represented to be in
certain PEs, even if they are picked out only by reference to the content of
another PE. We can meaningfully claim that a looks F (without looking to
be F) in the actual PE because we know what it is for something (else) to
look to be F; in other words, we know what it is like to experience an F.
This view gives explanatory and phenomenological priority to
representational properties: what is fundamentally given in PE are objects
and the ways they are represented as being. But our introspective powers,
conjoined with both a sophisticated counterfactual capacity and the
memory/imagination of a non-actual experience, allow us to compare two
experiences with different representational properties. So, no phenomenal
property could be picked out if it were not a “potential” representational
property.15
This view is well compatible with the transparency of PE. In
introspectively attending to your own experience of a penny, you will find
the properties that characterize the penny according to your PE of it. Yet,
the [roundness] represented in your PE can be taken by you as being
represented in such a way that it has some distinctive appearances in
common with the way an elliptical, frontally oriented penny would have if
you experienced it. The property [elliptical] is not directly represented in
your PE but the property [round] being given can remind you of the way
“it would be like” to see an ellipsis oriented differently.16
15 That fits well with the conceptual and explanatory dependence of “looks” from
“is” I argued for—with Sellars—in Chapter One (5.6).
16 That is why these apparent properties, like the elliptic look of an object that
looks to be round, are no less public properties than the properly intentional
properties of the PE, like [round]. The “elliptical” way of appearing of a tilted
round object is something intersubjectively assessable; it is an objective feature of
the perceived entity. There is nothing intrinsically subjective in these apparent
184
Chapter Five Part I
I.2.4 Perceptual Experiences Are Egocentric Representations
The complex properties of being [round + at-that-orientation], [big +
at-that-distance], [red + in-that-illumination-condition] are representational
properties throughout. They belong to the Content of PE. Skipping over
the color-case for the moment, let us note that something with a certain
size appears differently according to its distance from me; something with
a certain shape appears differently according to its orientation with respect
to me. And that, even if we accept the argument that the respective
phenomenological variations are representational variations (change in
distance or in orientation), these represented spatial properties must still be
represented egocentrically, as ways the world is with respect to me, rather
than allocentrically, as ways the world is full stop.17
If the phenomenological change of the cases above is accounted for in
terms of representational change, then the very same perceiver must be
somehow involved in the Content itself. If perceptual constancy does not
speak in favor of the awareness of non-representational properties in PE,
and so any phenomenal variation is in fact a variation in Content, then any
phenomenal feature of the PE that is clearly dependent on my actual point
of view must be accounted for by including myself, the perceiver, in the
Content.
The availability of the world for us to act upon it is the fundamental
character of our perceptual encounters with the world, so it is quite natural
that perceptual representations are egocentric and subject-indexed.
Something is represented as being at a certain place, distance, orientation
with respect to me and my body. When something is distant, it is distant
from here in the first place, not from a neutral point among the points in
the represented surrounding space. The location of my body is not one
among other places that my PE represents; rather it is where all the other
places are spatially represented. Likewise, a coin is tilted not just from my
point of view.
Although this to-me component of perceptual phenomenology is
something we are aware of,18 it is not a non-representational property of
our experience of the world; it is rather a property of the way the world is
represented in the experience. So, a representationalist view could
properties so they are not qualia in the strongest sense; rather, they are waysthings-seems-to-be-from-that-point-of-view, where anyone else could occupy that
point of view.
17 On PE as having an egocentrically-indexed content, see Burge 2005, 68ff,
Peacocke 1992, 70ff, Evans 1982, Chapter 7. See also Dretske 2003, 78ff.
18 On the to-me component of perceptual representations, see also Crane 2011.
Phenomenal Character and Representational Content
185
accommodate that egocentric aspect by including it within the
representata we are aware of in PE without having to accept that it is a
property of the representation itself, namely, of the state.
As Dretske notes (2003, 78), nothing prevents a perceptual experience
from having “egocentric” properties among its contents. For example, the
trees can be represented as being one smaller and nearer than the other, on
one hand, and as occupying the same portion of the subject’s visual field—
what we have called “looking the same size”—on the other. The ways the
trees “appear” to the subjects are represented properties, which means the
representation has egocentric as well as allocentric contents. Relational
properties are then represented in PE—properties that characterize the
objective spatial relation between the perceived objects and the perceiver.
Our point of view on the world is something we access in experiencing and
being aware of the egocentrically specified properties of the perceived
world; it is not something we access by being “directly” aware of the
experience itself or of its alleged phenomenal, non-representational
properties. For example, the perspectival properties of the visual
appearances are ways in which PE represents objective features with
respect to our actual location, so “from here” they are representations of
objective relations between us and the perceived world.
Thus, neither the phenomenon of perceptual constancy nor the
egocentric character of PEs seems to be enough to make an argument
against representationalism. If we are able to compare ways of appearing
of different properties, it is because our PEs have Content involving
information about the relation between us and the world besides the
information about the world. So, a tilted coin, if seen-from-here, looks a
bit like a coin non-tilted but seen-from-there.
Movement-perception is another good example of spatial representations
being subject-indexed and therefore relational. The very idea of an
objective space rests on the immediate grasp of the difference between
certain apparent movements due to the movement of the perceiver, and
other apparent movements due to the objective movements of perceived
things.19 In addition, the appreciation of objective movement also depends
19 See Evans 1982, Chapter VI. In the very phenomenology of PE, objects are
given as subject-independent on the one side, and the experiential relation to the
perceived world is given as perspectivally connected on the other (see Siegel 2010,
Chapter 7, on this). If I change my perspective on the object, I do not take the
object to be moving thereby (subject-independence). Instead, if I change my
position, my phenomenology will change as an effect of that, so in changing my
position, I expect certain distinctive phenomenological changes (perspectival
186
Chapter Five Part I
on the movement of the eyes. If you track a constantly moving object by
gradually turning your head, your retinal image of the object is constant
but you see it moving because the brain receives coordinated information
of the movement of your eyes as well as your head. So, visual
representation of movement involves a complex representation of the
relative spatial relations between the perceiver’s body, the perceiver’s eyes
and the moving object.
To draw an overall conclusion: perceptual constancy itself does not
speak in favor of the existence of phenomenal non-representational
properties of PEs we are allegedly aware of but it shows that in
representing the world as being a certain way through our PEs, we can
have at once introspective access to a) certain egocentric elements, which
are themselves represented rather than being properties of the
representational state itself; and b) certain relational properties connecting
us and the world, which are also represented in the experience rather than
being properties of the experience.
We have the capacity to compare these “egocentric ways” across
experiences with different representational properties. We are points of
view on the world, and our being such is introspectively accessed by us
through visually experiencing the world, not through seeing our own
experience of the world as some sort of internal eye.20 There is no need to
be aware of non-representational qualitative properties of the experience
itself to become aware of the world as perceived from a (certain) point of
view.
I.3 How Are We Aware of the Mode of Our Perceptual
Experiences?
The phenomenal character of PE cannot be identical with its Content
because a PE’s Contents are object-involving whereas the phenomenal
character of a given PE can be shared by the PE of another visually
indistinguishable object and also by an objectless hallucination.21 But a
representationalist could still argue that the phenomenal character of a PE
is identical with its content rather than its Content: it may be that the
connectedness). Subject-independence and perspectival connectedness are two
sides of a same coin.
20 To get how it is possible to go beyond the perception model of introspection, I
benefited from the reading of Conor McHugh’s PhD dissertation, Self-knowledge
in Consciousness, Edinburgh 2008, available online.
21 Disjunctivists about phenomenal character deny it (Martin 2006, Fish 2010). I
will argue later why they are wrong (see Chapter seven).
Phenomenal Character and Representational Content
187
properties the PE represents the (putative) object as being—which
constitute the content and so compose the Content of a PE together with
the object the PE is of—are identical with the PE’s phenomenal properties,
namely, the PE’s phenomenal character. In other words, it may be the case
that even though perceptual Content is object-dependent in a way
phenomenal character is not so they cannot be just identical, still the
phenomenal character is identical to the content, namely, to the properties
the PE represents the (putative) object as having. For example, I see an
apple and my PE represents it as round and green. The phenomenal
character cannot be identical to the Content because another identical
apple or even a hallucinated apple could be the “object” of PEs with the
same phenomenal character. Nonetheless, the phenomenal character could
be identical to the [round]/[green] represented properties.
In considering this hypothesis (representationalism about properties),
we notice that the represented properties are always egocentrically
represented from a point of view and in relation to the perceiver. Properties
like [distant-from-here], [so-oriented-with-respect-to-my-position] and the
like are represented relations involving both the perceiver and her
surroundings as relata. In experience, we are perceptually aware of the
world from our point of view.
If the point of view is “made out” of the properties things are
represented as having, then property-representationalism is true. Now it
seems that any representation of properties—as objectively related in
certain ways to the perceiver—cannot exhaust the egocentric
representation of the objective properties themselves alone. Let us explore
more closely the apparent gap between represented egocentric features and
elements on one hand and the egocentric way of representing features and
elements on the other.
In visual perception, any worldly object or feature is represented
within an egocentric frame of reference, within an egocentric space.22 The
perceiver is not just represented as an object among other objects: that
would be an allocentric representation, which includes myself within its
targets. Likewise, whenever I hear a sound coming from a certain
direction, I am immediately disposed to do various things so my spatial
representations are strongly egocentric; they are not just objective
representations of many things among which there is also myself or my
body occupying a certain position in the objective space. Rather, all
perceived things are perceived in a way that presents them from my point
22 See Evans 1982, especially Chapter 7: “There is only an egocentric space
because there is only a behavioral space.”
188
Chapter Five Part I
of view, for me, to me, for me to act upon, for my further exploration. The
primitive connection of perception to (possible) action deeply influences
the way the information is taken in, and this is reflected in perceptual
phenomenology as well. There is “up” and “down”, “left” and “right”, “in
front” and “behind” just by reference to myself as an agent in a way that is
completely different from the way my PE may represent an object as
behind another object, under or above another object, to the left or to the
right of another object. In these cases, both objects would be represented
from my point of view and in spatial relation to myself anyway, but not the
other way round. So there is an asymmetry between the represented
surroundings and myself that we cannot account for by appeal to the mere
fact that I am also represented as occupying a place just like the other
objects in my PE. Indeed, my body need not even be visually represented
in order for my visual PE to egocentrically represent all the surrounding
objects and their properties—that is, to represent them with respect to
myself. On the contrary, my own body is usually not represented in visual
PE whilst anything could be perspectivally represented. The very fact that
PE’s phenomenology exhibits certain ways of making apparent objective
properties besides the properties themselves—for example, the “elliptical”
mode of presentation of the tilted round object—is due to the fact that PEs
are egocentric ways of representing the world, not just representations of
the world that include the subject among their many contents. That “tome” dimension cannot be reduced to the fact that my experience also has
my relative position among its contents.
To explain this point in other terms, it seems that the egocentric aspect
of PE’s phenomenal character cannot be exhausted by the egocentric
contents of PEs, which anyway explain many aspects of the first-person
phenomenology of PEs. In PE, relations are represented not only between
perceived objects and their represented properties but also between
objects, their properties and myself, the perceiver. Nonetheless, this
important fact is not sufficient to explain the egocentric character of
perceptual experience. Where does that special first-person character come
from if not from the content? Should we concede that in PE we are aware
of qualia or property of the experience in the first place?
There is a third way between reducing phenomenal character to
represented properties and accepting the existence of qualia-like entities as
direct objects of perceptual awareness.
In Part IV.3 of Chapter Four, I introduced the basic ingredients for
individuating a certain PE: a subject, an object, a set of properties and a
mode. I ruled out that the phenomenal character of PE depends on the
object, so on the Content as composed by the object and the content. Then
Phenomenal Character and Representational Content
189
I argued that the egocentric component of PE cannot be exhausted by
content alone either. In other words, the egocentric content of PE alone
cannot account for the egocentric conscious character of PE. There is one
element left, which may be the key to explaining the gap between
representational content and phenomenal character. That element is the
mode.
The perceptual experience of an object, with a certain content, is a
different mental state from a belief about the very same object and with
the same content. The two intentional states differ in their modes—they
are different kinds of attitudes—toward the same object and content as
well as towards the same Content. Now, by “mode” I do not just mean the
kind of attitude like perceiving or believing or desiring or hoping or
fearing23 but also the determinate perceptual modality of the determinable
“perceptual” attitude.24 For example, vision is a different mode from
hearing or tasting so a perceptual experience in the visual mode can share
an object and a content (so a Content) with a perceptual experience in the
auditory mode, or with a perceptual experience in the tactile mode. Seeing
something moving is different from hearing something moving, seeing a
square is different from touching a square and so on. Is the mode of a PE
relevant for its phenomenal character?
As Chalmers (2004) and Crane (2007) put it, we can distinguish a pure
and an impure representationalism. According to pure representationalism,
either the content of a state fully determines its phenomenal character
(weak pure representationalism) or it is identical to it (strong pure
representationalism). According to impure representationalism, the content
of a state together with its mode fully determines its phenomenal character
(weak impure representationalism) or it is identical with it (strong impure
representationalism).25 I will argue for weak impure representationalism,
23 What I call mode here, with Crane 2007, is called quality by Husserl 1901 and
manner by Chalmers 2004, 2006.
24 By “attitude” I do not mean propositional attitude but semantic attitude in
general. An attitude toward content does not have to be a propositional attitude
insofar as there are non-propositional contents. If perceptual content is not
propositional, as I have argued above, then the “perceptual attitude” is not a
propositional attitude although it is a certain attitude toward content.
25 A parallel distinction similar to the pure/impure one is that between intermodal
and intramodal representationalism, proposed by Byrne 2001. According to the
first view, phenomenal character is fully determined by representational content
given a modality (see Lycan 1996, Harman 1996). For example, two visual
perceptions with the same content must have the same phenomenal character.
According to the second view, sameness of content fully determines sameness of
phenomenal character independently of modalities (seeing, hearing, tasting and so
190
Chapter Five Part I
namely, for the view that the mode co-determines PE’s phenomenal
character together with the content.26 I will then argue that the egocentric
character of PE depends on its distinctive mode as well as its egocentric
and relational contents. Phenomenal character is not fixed by content
alone.
If the mode is sufficient for explaining the phenomenal difference
between PEs with the same content, then there is no need to appeal to
phenomenal non-representational properties, namely, to qualia-like
properties of the experience we would be conscious of in the first place
when we enjoy a PE.27
It strikes me as compelling evidence that visual awareness of a
property “feels” different from auditory or tactile awareness of the same
property. Seeing a square feels different from touching it, hearing something
moving toward me feels different from seeing something moving toward
me. This fascination with primitive intuition is not shared by everyone,
however. Pure representationalists (Dretske, Tye, Lycan, Bain) usually
account for this phenomenal difference in terms of the difference of the
represented properties.28 For example, when you see a square, your PE
on). Another distinction is that between unrestricted and restricted
representationalism. In the first view (Dretske 1995, Tye 1995, Bain 2003), any
conscious state is a representation so it has content, including orgasms, tickles,
moods, depression, physical pain and so on. In the second view, only certain kinds
of conscious state have content and intentionality, like perceptions, beliefs and
thoughts (McGinn 1988). I think that unrestricted representationalism is
implausible but I will not try to argue against it. Instead, my concerns are
perceptual content and perceptual experience, not consciousness or intentionality
in general. Here I am not interested in whether depression has an object or content
of any sort.
26 Also, shifts in attention can change the phenomenal character of a PE but
attention may be taken to modify phenomenal character via determining which
among the properties falling under your visual field are represented and which are
not. In addition, each perceptual mode allows and enables certain distinctive shifts
of attention. In any case, differences in content as given under the mode account
for phenomenal differences due to attention. On attention as co-determining
phenomenal character, see Fish 2010, 58ff., Price 2011.
27 Qualia realists are Block 1990, 2003, Shoemaker 2006, Thau 2002.
28 The argument is put forward by Dretske 1995, 2003; Tye 2006; Bain 2003;
Byrne 2001. Byrne provides an argument for pure intentionalism, which,
unfortunately, begs the questions: 1) If S has two consecutive PEs that differ in
phenomenal character, will she notice the difference? 2) If she notices the
difference, will the way things seems to her change? 3) The way things seem to S
in PE is the content of the PE, so does difference in phenomenal character entail
difference in content? The argument does not work because it presupposes that a
Phenomenal Character and Representational Content
191
also represents its color; when you touch a square, your PE also represents
its texture or at least in a much more fine-grained manner than a visual PE
would. So the represented properties vary. It is just a contingent fact that
vision does not represent pitch and timbre; that touch does not represent
color, that hearing does not represent texture and so on. Still, the mode has
only an indirect role in determining phenomenal character insofar as it
determines which properties are represented, so it determines representational
content. It remains true that phenomenal character is fully fixed by
representational content alone.
That reply, although apparently straightforward, seems to deeply
violate our phenomenological intuitions. Of course, we will never have an
auditory PE of a moving object without a related perception of timbre,
pitch and other auditory properties. We will never have a visual PE of a
moving object without any perception of some chromatic contrast or of
some other morphological properties of the moving object and so on. But
the distinctive phenomenal character of a visual experience of a square
does not introspectively seem exclusively due to other concomitantly seen
properties like colors. Likewise, touching a square does not seem to feel
different from seeing it just because you also perceive texture and fail to
perceive colors. There is a distinctive phenomenology of seeing certain
properties that cannot be reduced to the experience of that property being
or not being accompanied by simultaneous experiences of other properties.
Moreover, in having a visual experience of an F, you are in a position
to immediately self-ascribe not only a mental state with that content (an F)
but also a mental state with that mode (a visual perception). Just by seeing
something, you become immediately aware that you are seeing it instead
of hearing or touching it; a primitive fact that can be hardly explained by
appealing to the presence or absence of certain other represented
properties. In being immediately and non-inferentially conscious of
enjoying a visual experience, I do not “check” the range of properties my
PE is representing, I just feel myself seeing and enjoy the distinctively
visual appearance of the seen things.29 My awareness is awareness under a
change in phenomenal character is identical with a change in the way things seem
to be to the subject. But that assumption just amounts to the assumption of pure
intentionalism itself! For a similar criticism, see Chalmers 2004, Crane 2007.
29 Noë 2004 points out that being visually conscious of a property involves
appreciation of certain sensory-motor dependencies of visual sensations from
certain possible bodily movements, and such dependencies will be different when
you are conscious of the same property through touch. Noë’s so-called enactive
theory is not in principle incompatible with CV; rather, it is a view on how
perceptual content is determined. I will not be concerned with enactivism but that
192
Chapter Five Part I
certain mode, even though, of course, my overall experience is normally
an integrated and multimodal flux. Consider seeing and hearing an object
moving toward you. If you focus your attention just on the property of
movement toward you, you may well still introspectively tell apart the
visual and the auditory phenomenology of that very same property. This is
not to deny that there is distinctive phenomenology of the complex, bimodal experience of seeing-with-hearing that movement; rather, this is to
claim that we can introspectively attend to the basic “mono-modal”
elements of that complex phenomenology in a separate way.30
In addition, we cannot imagine what it would be like to hear colors or
see pitches and timbres so it is distinctive of the visual phenomenal
character that there are certain exclusive visual properties like colors, not
the other way around. In hearing an object moving from the left, passing
overhead and going down to the right, and in seeing the same moving
object, we have two PEs that share their phenomenal character only
partially. They share it insofar as the phenomenal character is determined
also by the content. They do not share it fully for two reasons: first, the
mode is different and the shared represented properties are given under a
different mode, under a different phenomenal character; second, there are
other properties represented in one PE that are not represented in the other
PE. So the difference in represented properties determines a difference in
the overall phenomenal character of the two PEs but that fact alone does
not account for the difference in the phenomenal character of perceptual
representation of the common represented properties (like [object moving
from my left, overhead, down to my right]). For example, the perspectival
representation involved in vision is something that does not characterize in
the same way the phenomenal character of touch. You can very well
represent something as round by touch but it will not “look” elliptical at
all, not even in the sophisticated sense in which I have claimed it is
legitimate to talk with respect to visual appearances.31 So the same
represented property can have a different “look” of that very represented
remark about the dependence of qualitative character of perception from certain
expectations connected to bodily movements well explains the influence of the
mode on the phenomenal character.
30 Imagine you close your eyes when hearing that the object—say a small toy
plane—is moving toward you. If you are attending just to the object and its
movement, the phenomenal difference before and after closing your eyes does not
seem to be reducible to the fact that you do not see colors or other visual
properties. There is a further element that makes a phenomenal difference: you are
not visually representing that property.
31 See Chapter One, V.5.
Phenomenal Character and Representational Content
193
property. Phenomenal looks depend on modes as well as the content,
which is why they are irreducible to intentional looks.
So the phenomenal character of PE depends on both the content and
the mode. In particular, the mode determines the relevant egocentric
character of the PE, the “to-me”-component of the character.
That the phenomenal character is determined by the mode does not
entail that we can introspectively pick out certain properties of our
experience independently of the content of our PE. When we attend to our
visual experience, what we find are properties things seem to have but we
are also aware that they are properties things visually seem to us to have.32
That awareness is essentially intertwined with the awareness that they are
properties things seem to have “to me” or from my point of view with
respect to me. The egocentric aspect of PE is essentially determined by the
mode. It is by having a mode that PE is not just a representation of the way
things are but a point of view on the way things are.33 It is to the subject
that things seem a certain way in a PE.34 However, we are never aware of a
mode as such independently of what is given under that mode, of the
properties the PE represent things as having. So we cannot ever be
“directly” conscious of a perceptual mode; rather we become conscious of
the represented properties as given under that mode; for example, of
32 Lycan 2003, a pure representationalist, admits that the modes or “guises” can be
“phenomenal properties of some sort, perhaps higher-order properties” of the
intentional properties of PE. But, he remarks, these guises cannot be supposed to
come apart from the intentional qualia they present. Fair enough, the property of
visually seeming F cannot be told apart from the property of seeming F in the first
place; still, visually seeming F is phenomenally different from auditorily seeming
F, for example. I cannot get how Lycan can maintain a pure representationalist
view if, for him, phenomenal character may depend on “guises” in which the
intentional properties are presented. The core principle of pure representationalism
is the supervenient claim “same content → same phenomenal character.” But that
does not work if “same content + different mode → different phenomenal
character.” If something that is not the content (the mode) makes a phenomenal
difference, then pure representationalism must be false full stop.
33 On that constitutive point-of-view-ness of PE, Crane rightly insists in Crane
2007, 19ff.
34 As Crane 2007, 22 remarks, impure representationalism is more plausible than
pure representationalism because “impure representationalism accommodates the
way in which the “seeming” itself can enter into the phenomenal character of the
experience.” Within the framework of impure representationalism, the
irreducibility of the phenomenal look to an intentional/evidential look can be
accommodated without the need to commit to qualia (Block, Shoemaker,
Peacocke) or reduce the phenomenology of visually representing certain properties
of the very properties represented (Tye, Dretske, Stalnaker).
194
Chapter Five Part I
[square] as visually given. That is why the contribution of the mode in
determining phenomenal character does not entail that we are aware of the
intrinsic properties of our PEs independently of our PEs having the
contents they have. A mode is always a mode for intentional properties to
be given in experience. This is why the view articulated here is still a form
of representationalism though “impure” or non-reductive.
The idea that phenomenal character is co-determined by content and
mode may also account for certain features of perceptual phenomenology
that seem to be non-intentional at first. A quite usual example is that of
blurry vision.35 As soon as you take off your glasses, your visual
experience becomes blurry without attributing blurriness to anything in the
world. Some may argue voilà! a property we are aware of in visual
experience that does not belong to its representational content, namely, to
the ways PE represents the perceived world as being. Since blurriness is
not attributed to anything perceived, it must be a phenomenal property of
the experience itself: a quale.
Now this property could be accounted for as a property of the mode,
namely, blurry is the mode through which your PE represents the world.
The vision is blurry not the world, according to the PE. You do not
perceive your own vision but the visual mode immediately contributes to
your “blurry” phenomenology in such a way that you come to take the
seen world not as blurry but rather as “blurrily” seen by you; so only an
impure representationalism can handle cases like blurriness.
Since your conscious PE involves an awareness of the mode besides
the conscious entertainment of the content,36 the experienced blurriness
rests on the visual mode rather than on special qualia as primary objects of
awareness. Blurriness is not a represented property, it is a feature that
characterizes the mode of representing, namely, the visual vehicle of
perceptual information, and that feature of the mode is reflected in
35 See Bach 1997, Boghossian and Velleman 1989, Baldwin 1990, Dretske 2003,
Crane 2006.
36 It is important to emphasize that consciously entertaining content and being
aware of the mode under which that content is given are two inseparable and
intertwined elements of a conscious PE. You are not aware of the mode as a quale,
as an entity; rather consciously entertaining content under a certain mode puts you
in a position to immediately and non-inferentially know that it is under that mode
that you are entertaining that content. You would never introspectively pick out the
mode if that mode were not a mode for a determinate content to be perceptually
given to you in experience. The irreducibility of the mode-consciousness to
content-consciousness does not deny that perceptual consciousness and perceptual
phenomenology are essentially intentional.
Phenomenal Character and Representational Content
195
phenomenology.37 So such a phenomenon does not speak against
representationalism as such; at the very least, it speaks against pure
representationalism according to which the represented properties exhaust
the phenomenal character of PE.
So, impure representationalism is phenomenologically adequate and
compatible with the transparency of PE. In attending to your PE, you
always discover the properties things are represented as having, but in
experiencing, you are conscious that represented properties are under a
mode that contributes to shaping the distinctive phenomenology of your
PE.
This view is different from pure representationalism about phenomenal
character on the one hand, and qualia-realism on the other. By qualiarealism, I mean the view that in PE we are aware of intrinsic properties of
our own experiences, something like a “mental paint” (Block 2003) or
nameless “appearance-properties” (Shoemaker 2006, Thau 2002).38 Pure
representationalism does not do justice to certain fundamental
phenomenological features of PE; qualia-realism does not do justice to
transparency of PE either. PEs seem to attribute intrinsic properties to
things, exactly those properties we are thematically aware of as soon as we
introspectively attend to our own PEs. Moreover, qualia-realism is
unsatisfactory in explaining the intimate, seemingly non-arbitrary
connection between the phenomenology of PE and its intentionality.39
Impure, mode-based representationalism accounts for the egocentric
character of PE by appealing to the mode, so it does justice to the point of
view-ness, or the “to-me component” of experiencing. On the other hand,
it also accounts for transparency and the essentially intentional nature of the
phenomenal character itself. Indeed, the properties we find in introspection
37 Besides the appeal to the mode, a further important way of rejecting blurry
vision as evidence against representationalism is pointing out that PE may
represent properties of the channels and states of the perceiver as well. For
example, the property [blurry] could be represented by the PE as a bad state of
acuity of the visual mode in the actual circumstance. That consideration is
compatible with the appeal to the mode, not alternative to it. For representationalist
treatments of the case of blurry vision, see Dretske 2000, Lycan 2003, Tye 2002.
38 Under that label (qualia-realism), I mean to subsume any view that explains
perceptual looks in terms of the production of a certain quale the subject becomes
aware of, where that quale can be picked out independently of the content of the
PE and/or on the potential judgments the subject could make on the basis of that
PE. See Pettit 2003.
39 Also Fregean views on phenomenal content (Chalmers 2004, 2006, Thompson
2003), as we will see soon, have problems positing an internal connection between
phenomenal character and the properties it represents.
196
Chapter Five Part I
are the properties things are represented as having, although from our
point of view and under a distinctive mode.40
I.4 From Content Externalism to Phenomenal
Externalism?
I.4.1 If Content Is Not in the Head, Where Are Qualia?
We have been arguing that the phenomenal character of PE supervenes on
perceptual content plus the distinctive perceptual mode under which that
content is given in PE. Perceptual Content is wide insofar as it is objectdependent, it is a de re Content. What about perceptual content; i.e., the
properties the object is represented as having? Is it narrow or wide?
Representationalism about phenomenal character is generally committed
to content externalism.41 The represented properties—the content—may
depend on factors external to the psychophysical constitution of the
individual perceiver so that two microphysical twins can in principle differ
in the representational contents of their PEs: such contents are called
“wide.”
We will deal later with the issue of how mental states like PEs may
come to acquire and possess the content they possess. In any case, we may
already legitimately assume that the content of our PEs must somehow
depend on the evolution of our perceptual systems in a given environment,
on the survivor-adaptational value associated with our ability to detect
certain environmental properties.42 Thus, the reasons why a certain (type
40 Tye 1995, 134-7 argues that pure representationalism is the simplest
explanation of why a change in phenomenal character always involves a respective
change in content: because they are one and the same. Impure representationalism
seems the simplest explanation of why phenomenal character and content are so
intimately coordinated and why each sensory modality has an additional
phenomenological salience that does not only depend on which property is
represented but also on how the property is represented.
41 You can be representationalist and content-internalist about phenomenal
properties. You need to hold that your PE has only phenomenal content, namely, a
content that just supervenes on the phenomenal character of the individual
perceiver. It is hard, though, to individuate a set of veridicality conditions of your
PEs that would not change by changing deep causal facts about the perceiverenvironment relation, the properties that systematically cause this or that
experience and so on. On phenomenal content, see Kriegel 2002, Horgan and
Tienson 2002, Chalmers 2004, 2006.
42 This is the most plausible naturalistic story about how we are endowed with
perceptual systems.
Phenomenal Character and Representational Content
197
of) perceptual state represents the properties it represents are outside our
phenomenological accessibility. Such reasons go back to causal and
evolutionary relations between certain types of states and certain
environmental properties. Nonetheless, we can immediately access the
content our PEs have by virtue of their having the phenomenal character
they have. So, the phenomenal character somehow “tracks” the represented
properties and makes them accessible to us. That basic fact may be
explained by considering that the conscious character of PE is also an
outcome of evolution. Our capability to consciously entertain content must
have grown out of our capability to possess contentful mental states so, as
an unsurprising consequence, conscious perceptual experience tracks and
“traces” perceptual contents, making us conscious of them, or at least a
part of them.43 Conscious information may be used in many ways, far
beyond the context of its acquisition. The content of a certain PE depends
on the function of the type the state is a token of, and representational
functions, as with any other proper functions, are acquired by evolution.44
So, not only is the Content object-involving but the content (represented
properties) is environment-dependent in the broader historical sense.
Now, if content also depends on external factors, there may well be
elements or layers of content that are not present in phenomenology. So,
phenomenology makes content subjectively available on the one hand but
may not exhaust its informational richness on the other.
Inasmuch as phenomenal character makes content available to us, a
representationalist view of phenomenal character, even an “impure” one,
faces the “problem” of phenomenal externalism. If phenomenal character
is made out of intentional properties, then, since intentional properties are
wide, phenomenal character is also wide thereby. Phenomenal externalism
seems prima facie a “problem” because the way it feels, the what-it-is-like
dimension of my conscious experiences, seems to be a purely subjective
matter, an intrinsic, private feature hardly dependent on external factors. It
intuitively seems that I could be hallucinating everything, I could be a
brain in a vat, I could even have had a totally different evolutionary
history, yet the conscious character of my actual experiences would be
identical provided I am in that psychological, internal state.
Let us sum up. If content is wide, then it does not just supervene on
individual subjects but also depends on the environment: two identical
twins located in different environments could have PEs with different
contents. Here I have assumed that content is wide, a view I will try to
43 There are also unconscious and unnoticed contents in our perception, as we
have seen. See Chapter Two, Part I.2.
44 See the next chapter for further details of that view.
198
Chapter Five Part I
argue for later. Now I will focus on a further question: even if content
were wide, what about the phenomenal character, which is determined by
content plus mode? Would it be wide or narrow? Does content externalism
entail phenomenal character-externalism? Or can an impure representationalist
block that entailment? The answer is no; there is no way of blocking the
entailment. To give up phenomenal externalism, you must give up
representationalism about phenomenal character, content externalism or
both.
If perceptual Content consists of veridicality conditions, the properties
that compose it together with the object are physical features of the
environment normally detected by our perceptual systems. If an S’s PE is
veridical and has [O’s being F] as its Content, then F (the content) is the
very physical property instantiated in S’s environment by the perceived
object O. Now, the phenomenal character of a PE is made out of
intentional properties as egocentrically given under a mode, and for an
impure representationalist, the intentional properties determine it.
Let us call qualia—in a very liberal and generic use of this
controversial term—the properties that make up the phenomenal character
of a PE. Now:
6. content is wide (content externalism)
7. qualia either (a) are the content or (b) supervene on it
(representationalism)45
8. qualia are wide (phenomenal externalism).
So assuming content externalism and representationalism, phenomenal
externalism must follow. Maybe another assumption would be needed in
case one opts for 2b (supervenience rather that identity), the assumption
that if a mental property supervenes on another wide mental property, it
45 Case (a) is called strong representationalism; case (b) is called weak
representationalism (see Tye 2012). A case of the same content with different
phenomenal characters would refute both (a) and (b); a case of different content
with the same phenomenal character would refute only (a). If phenomenal
properties are identical with intentional properties, then the implication must be biconditional. If phenomenal properties only supervene on intentional properties, it
may still be the case that the same phenomenal property could be determined by
different intentional properties. As noted above, even if an object-dependent
Content determines phenomenal character, that very phenomenal character may be
had by a PE with different Content; for example, a PE of an another
indistinguishable but numerically different object, or even an objectless
hallucination with the same represented properties.
Phenomenal Character and Representational Content
199
too must be a wide mental property. In any case, that assumption is really
difficult to question. If it is true that given content C the phenomenal
character P is given thereby, then C’s depending on external factors entails
P’s depending on the very same external factors.46
To repeat the above point, impure representationalism is also
committed to the thesis that the content (co)determines the phenomenal
character of PE. Indeed, if the content C given under the mode M
determines phenomenal character, it suffices that content C is sensitive to
external factors for phenomenal character being sensitive to external
factors accordingly. In other words, even if the mode were narrow—shared
by any intrinsic duplicates across different environments—the pair [mode
+ content] would be wide anyway, provided that a constituent of the pair,
the content, is wide. The very notion of “wide” as employed by
externalists implies that the property is being determined also (even if not
only) by external factors. As a consequence, both pure and impure
representationalism seem to be committed to phenomenal externalism, at
least those versions of them committed to content externalism. But
phenomenal externalism is strongly counter-intuitive, prima facie at least,
and this is often considered a positive reason for rejecting those views that
entail it, as our view (representationalism about phenomenal character plus
content externalism) does. But we have seen above that perceptual
introspection reveals intentional properties in the first place, and that we
access perceptual content by virtue of our perceptual experiences having
the phenomenal character they have. So, the phenomenal character must
systematically track the content—the represented properties—and the best
explanation of that tracking is that phenomenal properties are just the
properties things are represented as having (properties represented by the
subject’s PE under a certain mode).
We need to show that phenomenal externalism is less implausible a
view than it prima facie appears. So, there is no need at all to reject any
46 For entailment from 1 and 2 to 3, see Byrne/Tye 2006. If you want to reject
qualia externalism, you must reject either content externalism or
representationalism, or both. To see the incompatibility, take Mario and Franco
with phenomenally different experiences. We can construct two intrinsic duplicates
of them in two different environments that make identical the intentional content of
their experiences. So, if content is external and qualia are internal,
representationalism is false. If representationalism is true and qualia are internal,
content is not external. As Ellis 2012, 2 straightforwardly put it: “If experience’s
phenomenal character is exhausted by its representational content, and if
representational content is externally individuated, then phenomenal character
itself is externally individuated.”
200
Chapter Five Part I
view that necessarily entails it.
I.4.2 Inverted Earth and Inverted Spectrum
Another powerful argument usually advanced against representationalism
is the inverted spectrum (IS) hypothesis,47 which purports to depict a
scenario of qualia inversion without illusion. Two subjects’ PEs (in this
world or in another possible world) could have some inverted phenomenal
properties, such that the experience caused by instances of color A in S1
are phenomenally the same as the experiences caused by instances of color
B in S2, and vice versa. S1 and S2 would perfectly understand each other—
they would be apt to detect instances of A and B in their common
environment, their perceptual phenomenology would successfully (though
differently) track different properties, their beliefs about colors would be
justified (“that is blue”), and finally the intentional contents of their PEs of
real A-color would be the same despite the phenomenal property of the
experience of A undergone by S1 being the same as the phenomenal
property of an experience of B undergone by S2, and vice versa. As Blocks
puts it, the IS hypothesis opens a scenario where “things we agree are red
look to you the way things we agree are green look to me” (1996, 511).
Here “look” is meant phenomenally rather than epistemically or
intentionally, of course.
So, if IS is theoretically possible or at least consistently conceivable,
then phenomenal properties could be associated with different intentional
properties in such a way that the first cannot simply supervene on the
second much less be identical with the second. So representationalism, be
it weak or strong, is false.
Many philosophers reject the hypothesis as inconsistent or impossible.48
Anyway, without entering into the details of a difficult and multifaceted
debate, I want to consider a famous intrapersonal version of the IS
hypothesis that does not have the many problems the classic interpersonal
version has. The intrapersonal IS is at the core of a well-known thought
experiment proposed by Block, the so-called Inverted Earth experiment
(IE from now on).
As is well known, the Twin Earth thought experiment was first
introduced by Putnam (1975) to argue in favor of meaning externalism. On
Twin Earth, everything is as it is on Earth, apart from the fact that a liquid
XYZ, superficially indistinguishable from water but with a different
47 See Shoemaker 1982, Palmer 1999.
48 For example, Dennett 1991.
Phenomenal Character and Representational Content
201
molecular structure, occupies the place of water, and in Twin-Earthian
language, people call XYZ “water.” Putnam brilliantly showed that
Mario’s and Twin-Mario’s utterances about “water” would have different
meanings in Earth and in Twin Earth, so “meanings ain’t in the head”;
rather, they are sensitive to causal, external factors. Therefore, at least
certain mental contents are “wide” and not supervenient on individual
psychologies and neuroanatomies: they do not need to keep constant
across intrinsic duplicates. Later on, wideness was extended to other kinds
of contents, like object-dependent, de re thoughts and beliefs (McDowell
1984, Burge 1979, 1991), memory-contents (Burge 1991, Davidson 1982)
and more generally to the contents of propositional attitudes (Stich 1980,
Fodor 1980).
Block uses a similar thought experiment to argue not only for the
wideness of perceptual content—which he accepts anyway—but also for
the narrowness of qualitative character as well as for the consequent
irreducibility of (narrow) phenomenal character to (wide) content. In short,
IE (=Inverted Earth) is directed against representationalism in the first
place (1990, 1996).
In IE, things have complementary colors to the colors of their
counterparts on Earth. The sky is yellow, grass is red, ripe tomatoes are
green and so on. The population of IE visually detects these properties,
experiences the respective intentional contents and forms perceptual
beliefs that the grass is red, the sky is yellow and the like. However, IE
people have an inverted language so they call the color of their sky “blue”
and the color of their grass “green,” just like us.
One night, while you are sleeping, you are kidnapped by a team of
scientists and transported to IE. Before you wake up, they put inverting
lenses that shift your visual experience from a color to its complementary
color behind your retinas. Now, when you see the yellow sky in IE, your
experience is like your normal experience of the sky you’re acquainted
with on Earth, so you think and say that the sky is blue just like the people
on IE say. You do not notice any difference from the day before.49 If
experience is unknowingly wrong—after all, the sky is yellow but you see
it as blue because of your inverting lenses, and you do not realize you are
wrong because of a linguistic coincidence (IE people call “blue” the
yellow that you mistake for blue)—in the long run, the intentional content
of your PEs and your utterances about colors would attune with those of
IE people. By keeping in long and repeated causal contact with the
49 The scientists also pigment your skin so that it looks to you as it did before
despite the presence of the inverted lenses.
202
Chapter Five Part I
environmental properties of IE, your mental states would finally shift their
original intentional content. With the strong intuition that your qualia
would not change accordingly, the IE thought experiment shows that the
phenomenal character of PE cannot be either solely identical with or
supervenient on intentional content. The first can keep constant throughout
the change of the second. The IE experiment holds a basic advantage over
the usual IS hypothesis: it is a science-fiction case, of course, but nobody
seems to doubt its theoretical possibility. On the contrary, the hypothesis
of IS as such, the idea that my experience of yellow objects “feels like”
your experience of blue objects and vice versa—but the phenomenal
difference does not make any intentional difference, neither perceptual nor
linguistic—has been criticized as inconsistent and empirically or
metaphysically impossible.50 Once your perceptual contents are attuned
with your new environment and community in IE, you may remember
what the sky was like years ago, before your (unknown) transportation to
IE, so your memory would remind you of an experience just like the
experience you have now when you look at the yellow sky. That reference
to your memories perfectly shows that the “blue” phenomenal character is
introspectively constant—we assume your memory is well-working—but
that very same type of experience had the [blue] property in the past and
has the [yellow] property now as its intentional content. Therefore,
representationalism about phenomenal character must be false. This is the
core of the argument.
The argument is a way of making conceivable the idea that our
distinctive experience of property A (e.g., [green]) may have been
typically caused by the property B (e.g., [yellow]) so a phenomenal
property A could have represented the objective property B by tracking it
and being related to it with a systematic causal relation. This possibility
undermines representationalism because it is incompatible with
phenomenal properties being the very properties that PE attributes to
perceived objects as intrinsic properties of them. For a representationalist,
all phenomenally identical experiences must attribute the same objective
50 Block had already proposed an intrapersonal version of IS (1990). Inverted
lenses are put on Mario’s eyes. He adapts and comes to use the language of the
community even though he remembers the past (the sky looks now as lemons used
to look before). Eventually, he has an amnesia about the past. Since it is
implausible that the qualia are re-inverted just because of the amnesia, he still has
inverted qualia. This version, as often is the case in the IS hypothesis, is an
example of qualia inversion despite intentional constancy. On the contrary, IE is
an example of intentional change despite phenomenal constancy. The result is the
same: that of telling apart phenomenal properties and intentional properties.
Phenomenal Character and Representational Content
203
properties to the perceived objects. If this were the case, the possibility
that another phenomenal property could have represented the same
objective property would be ruled out a priori but these scenarios (IS and
IE) seem to consistently allow for that possibility.
I.4.3 A Representationalist Reply
The IE thought experiment may be used to argue against functionalism or
against reductive “pure” representationalism about phenomenal character.
Representationalism is often associated with functionalism so qualia are
taken to be functionalizable qua intentional, or intentional qua
functionalizable.
I am concerned with representationalism and content externalism here,
not functionalism. The three are related but different issues. My critical
target will rather be the idea that:
given content externalism, if IE is possible, then representationalism must be
false.
There are some basic assumptions in the depiction of the IE scenario
provided by Block:
1. The intentional properties of my visual experience will change through
long-term exposition to the IE-environment.
2. My PEs will become veridical despite my having inverting lenses on my
retinas.
3. The phenomenal character of my PE will not change even through longterm exposition to the IE-environment.
I want to argue that these assumptions are far from being obvious; rather,
they are especially compelling for people already anti-representationalist
and/or internalist about qualia. Take away one of the assumptions above
and the IE scenario will not be relevant anymore. So, if we have reasons
for rejecting 1 or 2 or 3, or if we just do not have compelling reasons to
accept all of them, then we have no reason to endorse the conclusion that
qualia cannot be intentional properties.
Let us begin with the first assumption above.
You come to IE with a perceptual apparatus, a brain and a global
cognitive system, which are designed for you—as a member of your
species—to successfully cope with your environment. Your basic
cognitive functions, at least those that are phylogenetically fixed rather
than acquired through individual learning, are relational functions that can
204
Chapter Five Part I
be picked out only by reference to the environment they are supposed to
relate you to; that is the environment in which one you live in insofar as it
is common to the one your ancestors lived in.
Now it is controversial whether, and to what extent, visual perception
is sensitive to learning, other background knowledge and cognitively
higher information (memory, expectations, reasoning and so on). There
seems to be evidence that at least color-perception, as it depends on the socalled early visual system, is cognitively impenetrable and informationally
encapsulated in Fodor’s sense.51 Thus, color-perception is a wired-in
proper function of a relatively independent module and its contents are not
influenced by other acquired knowledge or individual learning. Now, if
that is the case, the type-state your visual system occupies when you see a
blue object in normal conditions is supposed to represent [blue], no matter
if a token of that state is caused by something yellow, anything else in
another Earth you are knowingly or unknowingly transported to or comesto-be normally or systematically caused by another property in an another
environment to the one in which your visual apparatus had been selected
because of what it did there. So, if color-perception is a wired-in proper
representational function, and wired-in proper representational functions
are species-specific and acquired by selection within a given environment,
your mental state normally caused by [blue] on Earth will continue to
represent [blue] in IE, even after long exposition to complementary
conditions and independently of linguistic factors and social mutual
understanding. Even if your belief-contents, your memories and other
intentional states changed, your visual contents, like color-representation,
would not. So there is no visual (at least chromatic) “content-attunement”
in IE despite the fact that you and IE people understand each other and
communicate. You will be wrong, at least perceptually wrong, about the
color of the IE sky.
Your perceptual states will ever indicate the presence of [blue] when
you see instances of yellow, even though you call that instance “blue” and
that instance of yellow wrongly looks to be blue to you due to the lenses.
There is illusion, then, no matter if you or others do not notice it. If you
took off your lenses, then you would veridically see the IE sky as it is—
yellow. So the intentional content of your PEs does not change the day
after the transportation or twenty years later.
Thus, the argument aims to show that in IE, in the long run, there will
be an intentional shift of your color-PEs without a qualitative shift. But it
51 See Fodor 1983, Pylyshyn 2006. For a reply, see also Tye 1998, Lycan 2003,
27.
Phenomenal Character and Representational Content
205
is far from obvious that there would be an intentional shift in the first
place. There is an empirical basis to think that this would not be the case.
According to the second assumption, which is connected to the first,
inverting lenses do not prevent your PE of the sky in IE from being
veridical, at least in the long run. You look at the yellow sky and it looks
blue to you. Your internal state is that which is supposed to occur in the
presence of blue, but there is nothing blue before you. You mistake yellow
for blue: the lenses deceive you rather than save you from error. That your
systematic color-deception saves you from social isolation in IE is another
matter. Even in the long run, your mutual understanding with IE people
rests on a coincidence; it is a misunderstanding. Your mental state
represents what it is supposed to represent in normal conditions, but the
presence of inverting lenses in your organism makes your conditions
abnormal. The lenses make you see things other than they are. This
remains true wherever you are transported independently of the time of
your permanence in the destination world.52
The key point is that “normal conditions”—on which intentional
content depends—are relative to the environment where the perceptual
system has been selected, not to the actual environment. Making
conditions “normal” in another environment through the use of inverting
lenses and waiting a long time does not help to change the original
content. The content of your lenses-influenced PEs is wrong and will ever
be such. Take them off after ten years and you will see things as they are.
Of course, my PEs in IE make me individuate objects, contours,
movements and every visible feature whose detection depends on colordetection. My behavior would be successful. Moreover, I communicate
perfectly with IE people. These conjoined facts lead us to think that
everything is all right with my PEs but this is a false intuition: my PEs will
be veridical in many other respects besides allowing me to speak IE52 Another empirical possibility is that the effect on the lenses would be
compensated in a while so that the sky would begin to appear yellow (as it is), the
grass red and so on. On that case, the subject would suspect that either he is
dreaming, or he is not on earth, or he is hallucinating, or something like that.
Kohler (1964) experimented with adaptation to strange color-distorting spectacles
with surprising results. He gave subjects goggles with vertically bisected lenses, in
which the left part was blue and the right part yellow. After fewer than 50 days, the
subjects recovered their previous color-detection skills, their ability to detect and
categorize colors. As soon as the goggles were removed, they had a complete
distortion again, but opposite to the first. Why could such adaptation to
complementary shifting lenses not happen also on IE? This empirical issue could
indeed undermine the very presuppositions at work/in the works in constructing
the IE scenario.
206
Chapter Five Part I
language but they will not be veridical with respect to colorrepresentation.
Again, it is far from obvious that there is intentional change at all. On
the contrary, it is plausible to suppose that no repeated causal contact with
IE will magically make inverting lenses into a virtuous means for turning
my deceiving PEs into veridical PEs.
By critically discussing assumptions 1 and 2, I have shown that there is
a plausible view on the content of color-perception that does not entail the
supposed intentional change over time. But some could refuse the “teleofunctional” account of perceptual content and take these replies to be ad
hoc. In another theory that grounds content in actual causal co-variation or
in anything else, for example, these replies would not be available. Still,
they are a possible way out: maybe it could also be argued that such a
teleo-functional view on visual content appears plausible because it
“saves” representationalism, which is a plausible option for many other
independent reasons, from objections based on IE thought experiments.
Alternatively, it suffices to question the third assumption for blocking
the anti-representationalist conclusion from IE scenarios. Even if
assumptions 1 and 2 were true or inevitable, if assumption 3 was not
inevitable the argument would be undermined anyway. Assumption 3 is
that there would not be any qualitative change in your experiences before
and (a long time) after your trip. Indeed, even if there were intentional
changes thanks to time and lenses (assumptions 1 and 2), it can be argued
that there would not be any qualitative change. In fact, if there were both
intentional and qualitative change, then representationalism is not refuted
but reinforced by the experiment.
What supports assumption 3? First, an internalist intuition about
phenomenal character. Your internal state when you see yellow on IE is
the same as what you normally had on Earth before blue-instances, so your
related experiences must be qualitatively the same. That intuition is
undeniably strong and hard to get rid of but do not forget that it is already
hidden in assumption 3, so it cannot be taken as a conclusion of the
argument. If you think—or do not a priori rule out—that phenomenal
character can be partially determined from external matters, you may well
not accept that the phenomenal properties of your PEs of the sky are
constant across transportation, especially if you have accepted that there is
intentional change.
For a representationalist, qualia are the properties experience represents
things as having (given under a mode for an impure representationalist).
They are the properties that would be actually had by perceived objects if
the PE were veridical. Now, if the content of your PE changes after ten
Phenomenal Character and Representational Content
207
years on IE so that your PEs of the sky really come to represent it as
yellow (as it is), then the qualia must have changed accordingly.
Otherwise, the qualia of your PE could not be the properties perceived
things are represented as having. Indeed, your PE before the trip could not
represent a different property from the PEs long after your trip if the
before-PE and the after-PE had the same qualia. If this seems to be a
question-begging reply, it is because it is such. Now, the experiment does
not show that qualia do not change, it rather postulates it and therefore it is
no less question-begging, but the burden of showing that representationalism
is false should be borne by the argument against it. Certainly a
representationalist cannot be asked to concede it in advance.
Even if we concede that the intentional properties change (assumptions
1 and 2), all the experiment shows is that intentional change across Earths
entails phenomenal externalism, namely, that content externalism entails
phenomenal externalism. It does not show that qualia are not intentional
properties (unless one assumes that qualia must be internalistically
determined).53
Are there any substantive arguments, besides the internalist intuition,
to support the view that qualia would not change over time? Block argues
for unchanging qualia across different Earths by appealing to what your
visual memory would be after transportation (1990; 1996).
For example, if you look at the yellow sky on your birthday on IE, you
may remember how looking at the sky used to feel a long time ago. Given
that your memory is working well, your experiences would introspectively
seem the same to you so their phenomenal character must be the same
despite their difference in content ([blue], [yellow]).
However, such an appeal to memory and comparative introspective
judgments presupposes that phenomenal memory is reliable across Earthto-IE transportation. But if phenomenal character is externally determined,
we should conclude instead that such a memory is deceiving. So the
absolute reliability of phenomenal memory across transportation is
intuitively irresistible only on the basis of an internalist intuition about
phenomenal character! Tye (1998) questions the strength of that intuition
53 Note that the qualia-externalism involved in historically based teleological
views of color-content is an externalism compatible with invariability of qualia
under transportation. Once an individual is born, it is given its evolutionary history
as a member of its species and you can displace it wherever you want, for however
much time you want, a very basic layer of its perceptual contents (including colors)
will remain unchanged. What matters for color-contents are not the actual causal
connections but the causal connection of the species with the environment where
the species evolved.
208
Chapter Five Part I
by first pointing out that classical Putnamian externalism about belief- and
thought-contents involves the prima facie counter-intuitive abandonment
of the first-person authority of memory. If you move to Twin Earth for a
long time, your utterance will change in content: “Water is drinkable”
would come to mean “XYZ is drinkable.” For example, suppose you think
that you did drink much more water on your tenth birthday than you are
drinking now on your twentieth. Unknowingly you are thinking that you
drank much more water then than XYZ now so you would be deceived
about your own past in thinking that you drank the same stuff then and
now. Therefore, despite intuitions, the memory of our own past may be
tricky for a content-externalist. Why could what happens for propositional
memory not happen for experiential memory? As Tye argues, memorycontents must actually be wide and not necessarily shared by intrinsic
duplicates: a native inhabitant of Twin Earth remembers her past drinking
binge of XYZ while you on Earth, before interplanetary transportation,
remember your own drinking binge of water. This is evident without
disregarding magic trips. Both of you utter, “I had a big drinking binge of
water”; you are microphysical duplicates yet your thought and utterance is
about a water-drinking binge, her thought and utterance is about her XYZ
drinking binge. So, external factors matter for memory-contents;54 what is
remembered is external to you. Change the past and your memory
becomes inaccurate, other things being equal.
Now we could suppose that phenomenal memory works a similar way
across transportation and long-term permanence. I remember that the sky
used to look blue to me just like now. However—a phenomenal externalist
could say—it is a mistake because now the sky looks yellow. That color is
yellow indeed; I just call it “blue” because I am an IE-language speaker. I
don’t notice any difference but there is one.
Block replies, not without reason, that such an appeal to memoryexternalism neglects a very strong intuition on privileged access and
authority of introspection on the one hand, and “assumes that as far as
memory goes, phenomenal character is representational content” (1996,
45) on the other. That is true but it also holds the other way around: the
very intuition that qualia would not change under transportation and longterm permanence, as memory allegedly shows, just assumes that
phenomenal character is not representational content as long as the latter is
externally determined. Likewise, it assumes that phenomenal memory is
internally determined.
Therefore, it seems that the IE experiment does not show that
54 On externalism about memory-contents, see Davidson 1982, Burge 1998.
Phenomenal Character and Representational Content
209
representationalism is false unless one already has anti-representationalist
intuitions!55 Perhaps it helps to make these intuitions explicit.
In any case, the IE experiment seems to show that if you are a contentexternalist and a representationalist, you must be externalist about
phenomenal character or about phenomenal memory.
It may still be said that phenomenal character cannot shift without the
shift being unnoticed by the subject. But the very experiment supposes a
long-term and gradual change of content. Even if we accept that an
unnoticed sudden shift of phenomenal character is inconceivable, that does
not prevent the phenomenal character from gradually shifting without
being noticed in such a way that this gradual shift perturbs the reliability
of phenomenal memory. In addition, if the phenomenal character changed
gradually, that change would be no more mysterious than the gradual
change of content from [blue] to [yellow]. As Tye 1998 notes, the
privileged access that characterizes perceptual consciousness holds with
respect to the present states, to what it is like to have that experience now,
does not hold also with respect to the past states, to what it used to be like
to have that remotely past experience. If you have independent reasons for
representationalism, you have room for rejecting the absolute authority of
phenomenal memory, in which case you can accept that there is an
intentional shift after transportation to IE.
A reason we resist this strategy may be the naïve idea of phenomenal
memory as a sort of inner picture you store and look at again when you
remember the experience. That “photography-model” of phenomenal
memory, as Tye opportunely calls it, may depend, I would suggest, on a
perception-model of perceptual introspection. If introspecting one’s own
experience amounts to seeing it with an inner eye or something like that,
then remembering your past experience amounts to somehow recalling to
the mind that picture-like inner image and “seeing” it again. I do not want
to enter into the intricate discussion of which model of introspection best
fits the empirical data on visual memory and other philosophical
constraints; nonetheless, I do want to point out that if the inner-eye model
is a problematic model of introspection, the photograph-model of
phenomenal memory is even less compatible with evidence about the way
55 Note that also belief-internalism, before the Putnam/Burge arguments, was a
very compelling intuition. Its strength has been gradually dismissed by the force of
the arguments against it but the removal of a natural, strong intuition is not sudden,
it needs to be metabolized. Phenomenal internalism seems to be even more
recalcitrant to negative arguments yet that is not a reason for its truth. In general,
the first-person authority on our own mental states seems absolute to us but it is
false, and accepting that falsity may be not easy at all at first.
210
Chapter Five Part I
visual memory works. If we consider that phenomenal memory is a
reconstruction, which starts from poor cues and instructions and is filled
through using actual beliefs and knowledge, then it becomes less
implausible that our access to our past experiences could mislead us. All
that considered, a representationalist could take it that if intentional
content changes after a long time on IE, then my phenomenal image of the
past on Earth now inaccurately represents the earthly sky as yellow, even
in the case where my phenomenal memory inaccurately made me take the
experiences as phenomenally indistinguishable.
To sum up this cursory discussion of the IE experiment: for a
representationalist (who is a content-externalist) there is more than one
way to resist the argument based on the IE experiment. Representationalists
may reject assumption 1 and hold that there is no intentional change over
time because color-contents are teleo-functionally fixed so they are
insensitive to trips, lenses, learning, and belief- and language-change.
They may also question assumption 2 on the same basis: in being teleofunctionally fixed, color-contents are also insensitive to the presence of
inverting lenses: the internal state I am in as I look at the sky on IE is a
state I am supposed to have before blue not yellow things, period.
Consequently, there is no attunement of perceptual content, no learning or
socializing that can ever compensate the deceiving effect of inverting
lenses. They may question assumption 3 and reject the idea that
phenomenal character does not change over time after transportation. Note
that it suffices that if one of the assumptions is false—especially 1 and 2—
then the experiment ceases to threaten representationalism at all. Either
intentional content does not change, in which case phenomenal character
not changing is what a representationalist would predict to be the case, or
phenomenal character does change with intentional content, so it is again
something a representationalist would predict. However, the
representationalist needs to choose a unique line of defense: neither 1 nor
3 seem absolutely compelling but at least one of them must be true if
representationalism is true. Indeed, rejecting both would entail that
intentional content does not change whereas phenomenal character does.
Paradoxically, that would be an anti-representationalist conclusion!
Anyway, what matters here is that the IE experiment does not entail at
all the negation of representationalism because it is based on hidden
assumptions a representationalist would have the right to deny. Actually,
presupposing that phenomenal character keeps constant means committing
to phenomenal internalism, and so to anti-representationalism if you are
content-externalist. But presupposing it amounts to begging the question.
Likewise, rejecting that the phenomenal character remains the same just
Phenomenal Character and Representational Content
211
because it is intentional throughout begs the question in favor of
representationalism. That means that the IE experiment as such is neutral
about the issue of representationalism since it becomes a case against
representationalism only by surreptitiously adding anti-representationalist
assumptions to it. In that case, the argument is question-begging.
A positive lesson we can draw from the IE experiment is that content
externalism is in conflict with phenomenal internalism, even if it is not in
conflict with representationalism. But no argument based on IE consistently
speaks against phenomenal externalism, as we have seen.56
Until this point, I have argued that even if the IE scenario were
possible, there would be interpretations of it that might be perfectly
compatible with representationalism: either there is both qualitative and
intentional change or there is neither a qualitative nor an intentional shift
en route to IE.
I want to point out that there are reasons for rejecting the possibility of
the IE scenario, reasons that may involve a parallel rejection of the more
classical IS scenarios.
IS depicts a situation where there is a qualia inversion without illusion
and IE an intentional inversion without qualitative inversion despite the
fact that everything else is identical: communication, language, beliefformation, recognitional abilities, behavior and so forth.
If on IE there are sunrises, sunsets and nights, we need to suppose that
the color of the sky gradually changes according to the Twin-Sun
illumination in such a way that the apparent color of everything else
changes. Even if color-constancy also holds on IE (better: just because it
holds), variations of illumination will change the apparent color of IEarthly things so the reciprocal relations between the colors of things will
change accordingly. But there are reasons to doubt that every behavior and
belief directly or indirectly involving colors would be identical despite
such color-inversion. The many possible relations between colors across
changes of illumination are likely to be not perfectly preserved in a world
where the color of everything is complementary of the color it has on
Earth. For example, the perception of the color of an object is influenced
56 Often thought-experiments against phenomenal externalism just beg the
question and assume phenomenal internalism within the argument. For example,
Rey 1988 holds that a brain identical to yours in a vat would have the same
phenomenal character your experience has now so phenomenal character does not
depend on external matters. But it is only because Rey supposes that phenomenal
character is insensitive to external matters (= narrow) that he finds the brain-in-avat case compelling. See Lycan 2003, Tye 2010.
212
Chapter Five Part I
by the colors of the background and of the objects contiguous to it.57 Now,
we could expect that the influence of a yellow background on a red object
is slightly different from the influence of a blue background on a green
object, especially if we consider that influence diachronically, over time
and across gradual illumination-change. Change of illumination will
change every relation between every color. The extraordinarily complex
and “holistic” color-representation of an IE native’s PE is not likely to be
preserved in toto if we change each color with its respective complement.
We can expect that some judgments made by natives about colors, like “In
the early morning, with those lighting conditions, the color-shade of that
surface has a hue, a brightness and a saturation that has this and that
relation with the hue, brightness and saturation of that other color-shade of
that other surface,” will be different from your judgment. If even one of
these indefinitely many relations is not totally preserved, then, even with
inverting lenses, there will be a phenomenal difference between the
natives on IE and you after you are transported. So, even with perfect
inverting lenses, the IE surroundings will not appear to you exactly as it
appears to a native. Sooner or later, due to some strange judgment or
behavior toward colors, a native would suggest you have your eyes tested
by an IE-ophthalmologist. Thus, the very possibility of the thought
experiment is undermined despite its prima facie conceivability.
This objection is similar to more immediate objections that could be
placed against the very possibility of an IS. Invert one quale with another
and you will have a change somewhere in some representation within the
global system of color-relations. For example, colors can be placed on a
scale from the darkest to the brightest, from the maximal to the minimal
hue, from the maximal to the minimal saturation. Each shade of color you
can discriminate will occupy a specific point on each of these three scales
so the discrimination-relations between each shade of discriminable colors
is relational and incredibly complex throughout.58 Invert a quale of a
darker color with the quale of a brighter color if you do not want people
and inverted people to judge differently about which color or shade is
57 Many color-illusions are built and based upon the well-known phenomenon of
influence-by-contiguity: for example, the checkerboard illusion among others (see
Adelson 1995). It is remarkable, for our concerns, that these illusions generally
work only with certain contiguous colors so a local spectrum-inverted subject is
likely to be representationally different from a normal subject with respect to their
sensitivity to certain color-illusions.
58 See Churchland and Churchland 1997, Matthen 2005. In inverting a color with
another, you will end up disrupting many other relations of similarity or
dissimilarity between perceived colors.
Phenomenal Character and Representational Content
213
darker as you will upset the representation of “darker-than” and “brighter
than.” But by changing the “darker-than” relation, you will end up
changing the saturation-relation, the hue-relation and other represented
properties thereby and you will not have the desired scenario of qualia
inversion without illusion, other representational facts being equal. Each
discriminable color has a point on a global space where that point is
individuated by its relations to any other point so you cannot arbitrarily
invert two points without changing many relations between all other
points. Conceivability of IS is considered sufficient to prove the falsity of
representationalism but even if that were conceded—I have argued above
that there is no compelling reason to do so—it can be argued that such a
conceivability becomes more and more problematic as soon as we detail
our hypothesis and take its consequences seriously. In addition, according
to the empirical data available about color-discrimination, which confirm
its relational and “holistic” dimension, the possibility of IS seems to be
definitely undermined.59
In this Part I, I have broadly considered the relation between the
phenomenal character and representational content of PE. I argued for an
impure representationalist version of the content view. I considered the
case of perceptual constancy as a possible basis for an antirepresentationalist argument, and I have given a negative answer:
phenomenal change involved in constant representations of sizes, shapes
and colors can be accounted for in terms of represented variations (of
distance, orientation, illumination). Nonetheless, a genuine look at the
phenomenology involved in visual constancy shows that PEs are
egocentric representations. I have argued that the egocentric character or
PE can be only partially explained by its egocentric, subject-involving
contents. To do justice of the point of view-ness or “to-me” component of
PE’s phenomenology, we need to take into account the mode under which
59 According to the so-called “opponent process theory” of vision, three types of
color-sensitive cones are to be found on the retina, each type sensitive to different
visible spectrum wavelengths. One type of cell registers the value of incoming
light on a continuum from red to green, another type on a continuum from yellow
to blue, another on a continuum from black to white. Each cell computes the ratio
between the inputs from some combination of cone types (that combination varies
through cell types). Each cell has an activation level that constitutes a triplet of
output that conjointly produce a certain value in a three-dimensional space of
possible activations. In that global model for the neural implementation of visual
discrimination, you cannot just change one color with another without disrupting
the relevant properties of the whole discrimination-system. On the opponent
process theory originally proposed by Hurvich and Jameson 1957, see Churchland
2005, 2007; Pautz 2006; Byrne/Tye 2006.
214
Chapter Five Part I
perceptual contents are given. That mode co-determines phenomenal
character: impure representationalism, which is a tertium between qualiarealism and reductive representationalism, can embed the advantages of
these alternatives without inheriting their problems, especially insofar as it
respects the perspectival-aspectual and the intentional-representational
components of perceptual phenomenology and accounts for their intimate
relation.
In succession, I have taken into consideration the anti-representationalist
arguments based on IS and IE scenarios. I have tried to show that such
arguments do not seriously threaten representationalism since they rest on
hidden assumptions that would and should not be shared by a
representationalist. However, reflection on the Inverted Earth shows that,
given representationalism about phenomenal character, content externalism
entails phenomenal externalism. Again, that is an objection to
representationalism only if phenomenal internalism is assumed. Phenomenal
internalism must be resisted though, despite its intuitive force on us.
PART II
FREGEAN VS RUSSELLIAN CONTENT
II.1 To Recap Where We Are
Perceptual experiences are states with phenomenal character on the one
hand, and states that represent the world on the other. In the version of the
content view I have been arguing for, the phenomenal character of a PE
(the way it is like to be in that PE) cannot be picked out independently of
the intentional content (the way the PE represents the world to be), which
co-determines the phenomenal character together with the mode. Indeed,
phenomenal properties are intentional represented properties given to the
subject under a mode. The conscious character and the intentional content
of PE content cannot be told apart,1 even if the first is not sufficient for
fully determining the second.
I have argued that Content is constituted by the perceived object and a
set of properties (the content) the PE attributes to that object. Perceptual
Content is a condition of satisfaction in the world for the PE: it is the way
the world would be if the PE were accurate-veridical. If the perceived
object has the properties the PE attributes to it, then the PE is accurate and
veridical.
I have also argued that the phenomenal properties of a PE are
determined by the ways the PE represents the object to be (i.e., by the
content) on the one hand, and by the mode under which those properties
are given to the subject on the other, but they are not determined by the
perceived object as such.2 However, if not only the Content is wide but
1 That does not prevent there being conscious states without content, or contentful
states without consciousness. I think both of these cases hold.
2 So perceptual experiences are individuated by their respectively perceived
object, their Contents are also individuated by their perceived object, but their
phenomenal characters are not (so, nor are they individuated by their objectdependent Content); rather, their phenomenal character is individuated together by
216
Chapter Five Part II
also the content (represented properties), then the phenomenal character,
as determined by the content, must be wide as well. That the Content is
wide in that view is evident from the fact that it is singular and objectinvolving: remove the object, or change it with another identical object,
without changing the internal constitution of the subject, and the Content
will change. That is just what “wide” means.
The properties the PE represents the object as having are wide in
another, historical, sense. Whilst the object is actual, what determines
which properties a PE represents “its” object as having is the function of
the type of state the actual PE is a token of, where such a function is
acquired evolutionarily by wired-in mechanisms of our perceptual
apparatuses.3 Now, if the representational power of our internal states
depends on past causal connections between members of our species and
certain environmental properties, the represented properties must be wide,
namely, they cannot supervene on the intrinsic constitution of the actual
perceiver. Rather they diachronically supervene on the environment in
which the selection of these representational functions in the perceiver’s
species occurred. Assuming that is a plausible story, I have argued that, if
phenomenal character is inseparable from represented properties,4
phenomenal character must also be wide, or externally determined.
Our spontaneous and inevitable resistance to phenomenal externalism
shows how rooted phenomenal internalism is in our intuitions; how
psychologically compelling it is, so to speak.
For that reason, other options have been proposed to save phenomenal
internalism and narrowness. One is the combination of content-internalism
with representationalism (McGinn 1982, Siewert 1998, Horgan and
Tienson 2002) where if two PEs with the same phenomenal character
the content and by the mode, where the content is the set of properties the object is
represented as having.
3 As I will argue later, only the basic perceptual contents of our PEs are
phylogenetically fixed (for example, color, size, shape, position, distance and
movement for vision). There is another layer of perceptual content, corresponding
to seeing-as episodes, which may well be determined by individual learning. Our
base of species-specific representations can be enriched by ontogenetically
determined contents like recognition, identification, categorization and the like.
There is perceptual learning indeed, which is made possible by previous wired-in
representational mechanisms.
4 I stress the word “represented” because the standard term “representational” may
result in ambiguity: a property of a mental state can be representational without
being identical with the property the state represents something as being (= the
represented property). For example, even a quale may be taken to be
representational insofar as attributed to it is the property of representing things.
Fregean vs. Russellian Content
217
share their content, content needs to be narrow if phenomenal character is
narrow. Instead, I argue that content externalism is the most plausible
option so rejecting it is too high a price for saving an even stronger
intuition in favor of phenomenal internalism. It is better to save intuitions
but good arguments cannot be ignored.5
Another option is anti-representationalism about phenomenal character,
combining content externalism with phenomenal internalism. This is the
option of qualia-realists (Block 1990, 1996, Shoemaker 1994, 2001, 2006,
Peacocke 1983, 2007). I have argued against this option because, among
other reasons against it, it entails an implausible separation between
phenomenal character and representational content, which does not do
justice to perceptual phenomenology and, moreover, cannot account for
the circumstance that our PEs are assessable for accuracy just by virtue of
having a certain phenomenal character.6 If the relation between
5 Horgan and Tienson 2002 appeal to our phenomenological intuitions to show
that conscious phenomenology is throughout intentional. Their examples and their
detailed first-person description of our experience as intrinsically representational
are undoubtedly efficacious but then they seem to argue the same way (appeal to
first-person intuitions) to show that phenomenological intentionality is narrow. But
that phenomenology is narrow cannot be shown by appeal to phenomenology
itself: phenomenology, just because it is so “near” to us and directly given to us, is
silent about whether an intrinsic duplicate of ourselves would have an identical
phenomenology or not. Phenomenal character is silent about its eventual essential
relation with external and causal matters so it is silent also about the absence of
such an essential relation. It rather seems that such an appeal to intuitions conceals
a philosophical prejudge.
6 To be more precise, phenomenal character constrains which contents perceptual
experiences have so in virtue of that fact we may consciously assess our own PEs
for accuracy. As we have seen before (see Chapter Two), there can be perceptual
contents that are not testified at all in phenomenology: there is evidence from the
Sperling experiment, blindsight, the distinction between the dorsal and ventral
stream dramatically shown by cases of visual associative agnosia and optic ataxia
and so forth. Often we are not phenomenally conscious of the perceptual cues our
systems use to detect certain properties; for example, to recognize a face, to
classify things and the like. The chicken-sexers case – people who are able to
visually tell males and females apart but do not have access how their visual
experiences enable them to do that – is only a curious example of a very common
phenomenon. Not all perceptual contents are conscious but we can access the
conscious ones by virtue of the phenomenal character of our PEs. More generally,
CV as such does not entail that we can introspectively access “the” content of our
experiences. There may be features of that content that are introspectively
accessible and features that are not. In any case, there is phenomenal content of
PE, the content that is subjectively accessible in virtue of the PE’s having a certain
218
Chapter Five Part II
phenomenal character and content is contingent and extrinsic, what does
that make the access to the second through the first except arbitrary and
mysterious? This is also an epistemological worry besides phenomenological
issues.
Now I want to take into consideration a third option, which accepts the
representationalist constraint on one hand but posits two different kinds of
contents for PEs on the other: a Fregean content, which is supposed to be
narrow and intimately connected to phenomenal character, and a
Russellian content, which is supposed to be wide so only contingently
connected with the (internally determined) phenomenal character. That is
Chalmers’s third option (2004, 2006), defended also by Thompson,
Kriegel, Rey and Levine among others.7 I will focus on Chalmers’ version
of the view, which seems to me the most well-worked and detailed one.
II.2 Chalmers’s Third Way: The Double-Content View
Frege explained the informativeness of identity-statements by appeal to
the idea that the same referent can be given in different ways or according
to different senses (1892). The Morning Star and the Evening Star are two
modes of presentation of the same referent, Venus. Their different
cognitive roles accounting for the fact that the Morning Star is the Evening
Star can be the object of discovery so it is far from being a trivial and
uninteresting identity. More generally, we can see beliefs made out of
concepts, each of them having certain extensions. For example, “The
Morning Star is bright” is composed of the concepts “the Morning Star”
and “bright,” having respectively Venus and the property [bright] as their
extensions. That merely extensional content of a belief is what is usually
called its Russellian content, a content made out of objects, properties and
relations.8 Besides a Russellian content or extension, each concept
composing the belief-content is associated with a certain condition on
worldly entities for being the extension of the concept itself. For example,
Venus is apt to be the extension of the concept the “Morning Star” insofar
as it is the first star that appears in the sky in the early morning, so Venus
satisfies the condition of that concept. A concept picks out a referent in the
conscious character. On the notion of phenomenal content, see Kriegel 2002,
Horgan and Tienson 2002, Chalmers 2004, McGinn 1988.
7 See Thompson 2003, 2008, Kriegel 2002, Rey 1998, Levine 2003. See also
Horgan and Tienson 2002.
8 That is the content of propositions according to Russell, which is why that
content is called Russellian. See Russell 1919. Anyway, I am not concerned with
historical matters here so I use the common label.
Fregean vs. Russellian Content
219
world (object, property or relation) insofar as that referent satisfies that
condition, which can be said to be the sense or mode of presentation the
referent is given according to that concept. So concepts have both sense
and reference; they are specific ways certain referents are given or
presented.9
Now, if we shift from belief and linguistic contents to perceptual
contents, we may apply the same distinction between referents/extensions
of a PE and their modes of presentation: call the first Russellian content
and the second Fregean content. So, a PE may be thought of as having a
Russellian content, made out of objects, properties and relations, and a
Fregean content, made out of the modes of presentation of those objects,
properties, and relations (MP from now on). Of course, perceptual contents
are not made out of concepts as beliefs are provided perceptual content is
non-conceptual (see Chapter Three). Nonetheless, the notion of the MP of
an object, property or relation may play the same role as a conceptual way
of an object/property/relation being given in belief, thought and other
language-like propositional attitudes in PE, namely, a non-conceptual MP
or a way for an extension to be given to the subject.
If objects, properties and relations are given in perception under
respective MPs, then the phenomenal character could be taken as
determined by them rather than by the very represented properties and
relations. In other words, phenomenal character could be determined by
Fregean content instead of Russellian content. If we add to that the idea
that MPs are narrow, we will obtain a double-content view according to
which PEs have a narrow Fregean content, supervenient on phenomenology
and only internally determined, and a wide Russellian content not
supervenient on phenomenology and also externally determined.
It is important to point out that MPs are not to be confused with what I
have been calling the mode. The mode is a manner for a content to be
given to the subject—for example, visually; it is not part of the content
itself. On the contrary, MPs are (Fregean) contents. They are accuracyconditions.
What kinds of conditions of satisfaction on Russellian extensions are
Fregean contents? What makes them accurate if not the Russellian
properties the object is represented as having? The answer given by the
advocates of the double-content view, like Chalmers, is: the most natural
condition a property must satisfy in order to be the property attributed by
the PE is that of being the normal and appropriate cause of experiences of
that phenomenal type the PE in question is a token of:
9 See Peacocke 1983, Chalmers 2004, 2006.
220
Chapter Five Part II
(PE): Russellian content: O is: [F] (= Russellian, physical property, like [red])
(PE): Fregean content: O is: [the property that normally causes
experiences]
A more articulated Fregean content for a PE representing O as being F
would be:
The object causing this experience has the property that usually causes
experiences of phenomenal redness. (Chalmers 2008, 110)
So the MP of the Russellian property [red] is another content than the
Russellian one, not less objective, consisting of a causal condition of a
type of experience. That Fregean content, made out of MPs, has
straightforward conditions of satisfaction so it is semantically evaluable
vis-a-vis to the world. If the seen object has the property that normally
causes my phenomenally-red experiences, the Fregean content of my PE is
satisfied; otherwise it is not. Given an environment in which that
phenomenally-red type-experience would be normally caused by green
rather than red objects, the very same MP could be a way for another
Russellian property ([green]) to be given. The same Fregean content could
be associated with different Russellian contents according to environmental
extrinsic matters of causal co-variations.
The double-content view easily accommodates the possibility of the
IS: two identical phenomenal states, say identical to your usual redexperience, could represent their objects as having respectively a [red] and
a [green] color, given that each of these phenomenal states is usually
caused respectively by red and green objects. My red-experience could
have been caused by the [green] property. Namely, that very MP could
have been the MP of another Russellian property than the property it is a
MP of. The very same MP may pick out different properties in different
environments as well as in different creatures.
Therefore, representationalism is true, but the phenomenal character of
PE supervenes on Fregean content rather than on Russellian content.
Phenomenal character supervenes on contents involving a causal
component, a self-referential component concerning the experience itself,
and especially involving reference to experiential types, like
“phenomenally-red” or “phenomenally F” in general. In this way, it can be
true that a given phenomenal character unequivocally determines a narrow
content, made out of MPs, so phenomenal internalism and
representationalism are vindicated without even denying the possibility of
IS scenarios or IE-like scenarios.
Before critically discussing the double-content view, I want to briefly
Fregean vs. Russellian Content
221
clarify a point in order to avoid possible confusion. The Fregean content
exemplified by Chalmers in the above quotation includes reference both to
the object that causes the very experience the content is content of on the
one hand, and to the property that usually causes that very experience on
the other. However, I have already argued that perceptual Content is
object-dependent so the real seen object, not a mode of presentation of it,
enters into the Content. The double-content view in itself is neutral about
whether perceptual Content is object-dependent, existential or anything
else.10 So, given that I have already argued for the object-dependency of
perceptual Content, I will consider the double-content view as a view
concerning the represented properties, what I have called the content
(lower case). A version of the double-content view that embeds objectdependency would have the following de re characterization:
Fregean de re content: of the seen object O—PE represents [the property that
normally causes this phenomenally-red type-experience]
Russellian de re content: of the seen object O—PE represents [red] (= physical,
Russellian property)
So far so good. Now, I will argue that the double-content view is untenable
and must be abandoned.
II.3 The Weaknesses of the Double Content View
To begin with, Chalmers specifies Fregean content by reference to a
causal element and a self-referential element concerning the very
experience, the same way as Searle (1983, 1991) specifies perceptual
content. I have already argued (Chapter Four, Part II.4.4) that such a view
is implausible so I will only cursorily recall those critical remarks here.
First, there is an issue of phenomenological inadequacy: no causal element
seems to be reflected in visual phenomenology; moreover, it does not
seem the case that a visual experience presents the subject with features of
the visual experience itself. In addition, such contents are too sophisticated
to plausibly characterize the experience of a child or a non-linguistic
animal, even if we consider such content is non-conceptual so the
perceiver does not have to deploy the respective concepts. Our experiences
have content before we begin to entertain contents like “my own
experience” or “my experience being caused by P,” be that conceptually or
10 Chalmers himself holds a multiple-contents view according to which a PE can
have a singular content, which hallucinations lack, an existential content, which is
shared by perceptions and hallucinations, and so on.
222
Chapter Five Part II
non-conceptually. Furthermore, the causal element included in the content
must be enriched with a normative clause or appropriateness to avoid
deviant or abnormal causal chains satisfying the content and making the
perception veridical. Chalmers explicitly inserts that clause into the
Fregean content: “the property which normally and appropriately causes
phenomenally-red experiences” (2004, 106). That such a problematic
notion as “appropriate causation” should enter experiential, nonconceptual contents seems highly implausible, and not only from a
phenomenological point of view. As I noted above by discussing Searle’s
analog view, there are no discriminatory abilities in perceivers one could
not account for without positing such sophisticated content, so positing it
is explanatorily redundant if not simply illegitimate.
Other difficulties are germane to Chalmers’s specification of the
Fregean content. Here are some.
First, besides the causal, the normative and the self-referential element,
there is reference to a phenomenal property. That is why Chalmers takes
his narrow representationalism to be non-reductive insofar as it does not
aim at getting rid of phenomenal notions by reducing them to nonphenomenal notions. Now, the Fregean content of my PE of a red object is
accounted for by appeal to “phenomenally-red” experiences. That means
that I am supposed to be able to pick out such a phenomenal property
independently of the property my PE seems to attribute to red things,
independently of what the double-content advocate calls the Russellian
red, the red as represented “out there.” I argued in Chapter One that the
“phenomenal” looks are conceptually and phenomenologically parasitic on
intentional looks: on the one hand, I can master and understand the notion
of “looking F” as long as I master and understand the notion of “looking to
be F”; on the other hand, I can master and understand the notion of
“looking to be F” as long as I can master and understand the notion of
being F (see Chapter One, Part V, 4-6). It is only by detecting, sorting and
recognizing red things that I could learn to detect, sort and recognize
phenomenally-red looks, so “red experiences.” Moreover, there is no
uniform phenomenal experience caused by red things that can be picked
out without considering what worldly things our visual experiences
prompt us to take as being red. Perceptual constancy teaches us that “red
experiences” may be significantly different from each other so they fall
into the same type only insofar as they are all experiences that make us
take things as objectively red.11 Even if Chalmers presents his view as
11 It will not do to restrict the phenomenally red experience to the experience
caused by red things in standard or paradigmatic conditions. Indeed, there is no
way to single out standard conditions for color-vision. See Hardin 1987, 1993.
Fregean vs. Russellian Content
223
(Fregean) representationalism, actually the representational aspect of the
PE’s phenomenal character is not so intrinsic to the latter to prevent the
independent individuation of phenomenal properties. The Fregean content
is an abstract condition on a property, that of appropriately and normally
causing a certain type of phenomenal property in my PE. If perceptual
phenomenology is intrinsically representational in that way, then that view
is not clearly distinguishable from a form of qualia-realism: qualia are
intrinsic properties of the experience we are aware of in experiencing.12 In
fact, that criterion of ascription of Fregean content needs the position of
phenomenal properties as qualia: if the Fregean content is specified as
“whatever objective property appropriately causes that property of my
experience,” then the phenomenal red is a property of the experience in the
first place, which has only an extrinsic connection—via normal
causation—with the represented property.
Paradoxically enough, that view is supposed to save certain basic
phenomenological intuitions (especially the phenomenal internalism
intuition), but it radically neglects perceptual phenomenology. Visual
experience is transparent: the properties we attend in introspection present
themselves as being the properties of perceived things not the properties of
the experiences normally caused by certain properties of perceived things.
Visual experience attributes to worldly objects those very properties that
are present in visual phenomenology, not other properties that purportedly
cause them in normal conditions and in the appropriate way.
Chalmers considers this issue of phenomenological adequacy13 and
replies that in fact we “see through” the MPs in such a way that we are
perceptually aware of both the MPs and the properties given under them
so transparency is vindicated. But that reply is unsatisfactory: indeed, what
he calls the MP of a property is only contingently such. It is a quale
normally caused by that property, which could well be the MP of another
property if it were normally caused by it. MPs are not phenomenologically
connected, so to speak, to the properties they present. Actually, in that
12 Pettit 2003, 225ff., presents the qualia theory as follows: a color looks red if it
produces a red quale in the perceiver such that: a) the way the object looks is
independent of the perceptual abilities typically enabled by the object having such
a look; b) the way the object looks is manifest to the perceiver; c) the object’s
looking so enables perceptual abilities like sifting the object from other objects of
different colors, and sorting the object in a same category of same-color objects
and so on. So, the quale is a property of the experience you are aware of. It enables
certain judgments, recognitions and comparisons but it can be picked out
independently of the way you take the world to be. See also Lewis 1995.
13 See Chalmers 2004, 2006.
224
Chapter Five Part II
model, perceptual introspection consists of attending to the qualia
included in the Fregean contents or MPs; it does not consist of attending to
Fregean contents or MPs. In introspection, you do not attend to whatever
property is causing your phenomenally-red experience; you rather attend
to your phenomenal red, to your quale. That is incompatible with
transparency pace Chalmers. According to transparency, in introspection
you attend to properties of seen things.
Another big problem that can be envisaged concerns the interpretation
of the “normal causation” and “appropriate causation” clauses included in
the Fregean content of PE. What fixes normality and appropriateness in
the first place? If normality were just a statistical notion, then the Fregean
veridicality conditions of my PE would be satisfied even if I were
systematically misperceiving colors and other visual properties. Suppose
both green and red things “normally” cause in me phenomenally-red
experiences. So, a PE of a green thing would be veridical insofar as
[green] normally causes phenomenally-red experiences in me. My
perceptual system could even be so disrupted that my phenomenal colortypes do not track or map any types of properties in the world. Suppose
my phenomenal red is “normally” caused by a certain shade of red, a
certain shade of blue and a certain shade of yellow whilst my phenomenal
green is “normally” caused by another slightly different shade of red,
another shade of blue and another shade of yellow. So, in that case the first
three shades would make veridical my PE instantiating phenomenal red;
the other three shades of the same objective color would make veridical
my PE instantiating phenomenal green. In both cases, the perceived
property is that which normally causes—among others14—a certain type of
experience. Fregean contents would make such fuzzy PEs into veridical
PEs.
If “normally” is not a statistical notion but has a stronger normative
dimension, such an additional normativity must be explained within the
view. If by “normal” is meant the same as appropriate, again the
appropriateness clause must be accounted for. It seems hard to define the
appropriateness of the causation of a phenomenally-red experience without
reference to red things as appropriate causes. But, of course, that move is
not allowed by the double-content view advocate: physical red is
Russellian indeed, but by definition the phenomenally-red experience
could have had the same Fregean content even if its Russellian content had
14 At least, the normal cause is the disjunctive set of properties, say [red45] or
[blue57] or [green69].
Fregean vs. Russellian Content
225
been the extension or property [blue].15
In addition, there seems to be a problem in that view similar to the socalled disjunction problem, which affects causal theories of mental
content.16 In accounting for content in terms of normal causation,
informational-causal theories have trouble making room for mistakes or
misrepresentations. A misrepresentation is a circumstance where a
representational state is tokened but its content is not exemplified. Now, if
the content is whatever causes the type of the tokened mental state, any
token of the state will always be true, its content coming to include the
actual property causing it.17 If I have a phenomenally-red experience
before a blue object, then what normally causes my phenomenally-red
experience will become the disjunctive property ([red] or [blue]), so the
content will be satisfied and the experience veridical. It will not do to say
that the instance of blue—in our environment and for our perceptual
systems—is not the “normal” cause of the phenomenally-red experiences,
but it is an “abnormal” cause. Many visual illusions are perfectly
“normal”; indeed, this is why they are intersubjectively shared. Not every
circumstance our PE misrepresents is an abnormal one.18 Perceptual
mistakes are quite common when we are in perfect shape too.
Besides, if by “normal and appropriate” causation of phenomenally-red
experiences is meant normal causation in me, then it is to be decided how
far in the past—or perhaps in the future—the “normality” goes. For
example, if I were transported to IE, would my Fregean visual contents
change? In the long run, as Block notes, the normal causes of my
phenomenally-red experiences would become the green things, so the
15 Chalmers 2004, 114, considers the possibility of specifying the Fregean content
in terms of community rather than individually on these lines: “the property that
normally causes phenomenally red experiences in my community.” That
possibility has been rejected by Chalmers himself, who points out that in that case
a spectrally inverted individual in her community would have systematic illusions.
16 See Fodor 1990, Dretske 1986, Jacob 1999.
17 Suppose you mistake a cow for a horse, so the concept [horse] is tokened in
your mind. Now, if the content of your mental state is whatever property causes it,
or whatever property that type of mental state counterfactually depends on, then its
content should be [cow OR horse] rather than [cow], but in that way the
representation would be correct.
18 At this point, one could appeal to optimal conditions or other similar normative
notions. The problem is that is seems hopeless to specify optimality of conditions
for phenomenally red experiences without referring to red things as the optimal
causes. One should be able to tell why a circumstance where a phenomenally red
experience is caused by a blue thing is not optimal without any reference to the
[red] property as the good referent of that type of PE.
226
Chapter Five Part II
property [green]. Now, if that is the case, then the Fregean content
changes across transportation but that is contrary to the original hypothesis
that Fregean content is narrow, different from Russellian content, which is
wide. So in that case, Fregean content would not be able to vindicate
phenomenal internalism either. If we want Fregean content to be narrow,
perhaps we need to bind the “normality” to my more remote past. Maybe
we should specify the Fregean content like this: “The property that
normally causes phenomenally-red experiences in me provided that I stay
in my original environment or that no inverting lenses or other artificial
modifiers are put on my eyes.” As is self-evident, these clauses would in
fact dash the initial hope of keeping the Fregean content narrow, namely
exclusively supervenient on the subject’s internal constitution, which was
the only reason for introducing Fregean content as well as the Russellian
content. In short, since the normality clause must be more than statistical
but more strongly normative, and both “normality” and “appropriateness”
have a normativity depending on them being indexed to an environment,
then a change of environment, for example from Earth to Inverted Earth,
would change what “normally and appropriately caused” refers to, thus
changing also the Fregean content. Not only does the normal cause change
but also what it is for a certain type-experience to be normally caused; i.e.,
not only the Russellian content but also the Fregean.
Another natural move would be that of changing the reference to me
with a reference to my species, like: “The property that normally causes
phenomenally-red experiences in members of my species with wellworking visual apparatuses,” or something like that. First, that would
prevent the possibility of an inverted spectrum without illusion for
members of the same species, contrary to the desideratum of
accommodating IS scenarios. Second, and most importantly, that would
prevent again the Fregean content from being narrow since the PE of an
intrinsic duplicate belonging to another species (or a Swampman
belonging to no species at all) would have different or no Fregean content
at all.
II.4 Perceptual Content and Character Are Wide,
External, Russellian
To sum up our critical point on the double-content view: the view is
untenable so it cannot be taken as a virtuous “third way” mediating
between the extreme views of qualia-realism and externalist
representationalism. The attempt to save phenomenal internalism by
introducing a narrow content on which it should supervene—a content like
Fregean vs. Russellian Content
227
[the worldly property that normally causes that type of phenomenal
property]—is doomed to fail. I have shown it to be so by means of three
main arguments.
First, the causal and self-referential components cannot plausibly enter
into perceptual content because they are phenomenologically inapt and
there are no discriminatory abilities only they could account for. Even
more unpalatable is the inclusion of an appropriateness clause into the
content. Rather, the PE’s being caused by an object makes the PE to be
about it, the PE’s being appropriately caused by the properties had by the
perceived object, results in a successful representation of these properties.
The contextual and causal conditions, which make a PE about an object
and represent its properties due to the appropriateness of that causation,
are objective conditions obtained in the perceiver-environment relation;
they make the representation possible but they are not represented as such
by the PE. A perceptual state successfully representing its causes by being
appropriately caused by them does not have to represent itself being
caused by these causes, nor does it have to represent that its own causation
is appropriate. It suffices that the latter is such in fact. Perceptual contents
are satisfaction conditions, not representations of all it would be
contextually required for these conditions to be satisfied.
Second, the posited Fregean content includes the phenomenal property
[F] as one that can be picked out independently of the reference to any
worldly represented property F. Indeed, [F] is only contingently an MP
of F; it could be as much an MP of G or of whatever else could normally
cause it. Such a priority of the phenomenal over the intentional seems to
turn priorities upside down insofar as it presupposes the original
givenness of a quale of [F], introspectively accessible as an intrinsic
property of our PE. That is hardly compatible with the view being a
representationalist view, though of a non-reductive sort; rather it makes
it analog to qualia-realism. Indeed, it inherits its problems; for example,
that of the separation between phenomenal character and represented
properties (phenomenal/intentional pairs taken as contingently related and
reciprocally invertible), and that of the inability to do justice of
transparency of perceptual phenomenology (which attributes to things the
properties we are aware of not their normal causes).
Third, the reference to “normality” and “appropriateness” of causation
in the content is highly problematic for many reasons. First, many illusions
and misperceptions are “normally” caused, the apparatus is working well
and the conditions are not abnormal so there must be accurate and
veridical misperceptions, at least according to their Fregean content.
Second, the notion of “normality” is vague with respect to temporal and
228
Chapter Five Part II
modal extension. If you are transported to IE and stay there for a while—
or perhaps it suffices that it is just simply possible—not only do the
properties that normally cause your phenomenally-red experiences change
(and so the Russellian content) but also what-it-is-to-be-normally-caused
for that type of experience (and so the Fregean content). Any move to
introduce into the content clauses that index the content to one’s own
environment would only emphasize that Fregean content is not narrow,
contrary to the desideratum of saving narrowness of phenomenology
through the introduction of Fregean content. There is no “normality” that
is not environment-indexed so there is no narrow normality. Therefore,
there cannot be any narrow content including normality. If there were
Fregean content, it would be wide, but then we would be better to use
Occam’s razor and be content with the Russellian one.
If “normality” is just a statistical and not an environment-involving
notion, like “most of the time caused in me,” then a perceiver who
systematically misrepresents, or has experiences sorted into phenomenal
types that do not map at all color-properties in her environment, would
have accurate experiences with satisfied Fregean contents. The purely
statistical interpretation of “normal” would also open a disjunction
problem for Fregean contents. You have a phenomenally-red experience
caused by an example of worldly blue so the property that mostly causes
your phenomenally-red experience becomes now the disjunctive property
([red] or [blue]); therefore, any mistake is turned into a veridical
perception by contributing to change what “normal causation” is for that
phenomenal type. Finally, any other maneuver to get a more-thanstatistical normativity for the “normally caused” clause, like interpreting
the condition as “normally caused in members of my community” or
“normally caused in members of my species,” would patently re-introduce
wideness into the content and so prevent IS scenarios within communities
or species.
It seems that to make room for misrepresentation—even the Fregean
one—we need to refer to a certain environment we are causally connected
to not just to “whatever environment I could be in,” so it seems that
perceptual content cannot be anything other than wide.
The maxim I draw from all this is that perceptual content is Russellian,
external and wide, and so it is thereby phenomenal in character as long as
representationalism is the best account of the phenomenal character of PE.
The double-content view fails in trying to save phenomenal internalism by
introducing a narrow content of PE. Such content would not be narrow
either so any reason for introducing it becomes redundant. The overall
point is that phenomenal internalism cannot be saved.
Fregean vs. Russellian Content
229
We are left with our wide, Russellian, impure, non-reductive
representationalism. It is non-reductive not because there is a content that
cannot be specified without qualia-like phenomenal notions but because
the represented, Russellian properties our PEs attribute to perceived things
are represented by the PE under a mode; for example, the visual mode or
the auditory mode, and modes are phenomenologically salient. The mode
is not an MP if by MP we mean another sort of content besides the
Russellian properties attributed to things: there are no other represented
properties in PE besides them. The mode is a way for that content to be
given to the subject, and it makes a phenomenal difference without being
part of the content. The most relevant difference between the mode and the
Fregean MPs is that MPs are supposed to be picked out independently of
the properties they present. That is why there can be inversion between
properties and the MPs of them such that a phenomenally green MP could
come-to-present the property [red] and vice versa. On the contrary, the
mode is a way for certain Russellian properties to be given to the subject,
but the phenomenal character as co-determined by that mode still cannot
be in any way separated by those Russellian properties that co-determine
it.
To a certain extent, the mode does the explanatory job the MPs are
supposed to do: it accounts for the aspectuality of perceptual
phenomenology as well as for its egocentric dimension. In PE, the world is
given in such a way that the perceiver is a point of view. That point of
view-ness of experiencing is intimately connected with the perceptual
mode of a given PE.19
Concluding Remarks
In this chapter, I have examined two issues. First (Part I), I proposed a
form of impure representationalism on phenomenal character of visual
experience. Second, I critically considered the double-content view—
proposed by Chalmers and others—according to which perceptual
experiences have two kinds of content: Fregean and Russellian.
According to my impure representationalism, the phenomenal
character of visual perception is made out of intentional properties
represented under a mode: the mode co-determines the phenomenal
19 I am focusing on visual experience here a bit abstractly but our concrete
perceptual experience, considered as a continuous, unitary flux involving
integration of many modalities, exhibits a complex phenomenal character with
many levels of aspectuality and “perspectivity” due to the composite interplay of
the different modes.
230
Chapter Five Part II
character so my view is opposed as much to reductive versions of
representationalism, like identity theories as to anti-representationalist
views (for which phenomenal properties are just others than the ways
perceived things look to be). I considered the phenomenon of perceptual
constancy as a potential challenge to representationalism. I argued that the
phenomenology involved in perceptual constancy is not a real challenge,
at least for impure representationalism: indeed, in visual experience,
certain environmental properties are represented egocentrically and such a
perspectival dimension of visual phenomenology is made possible by the
mode under which such objective properties are represented.
Successively, I considered the IE and IS hypotheses as arguments
against representationalism, and I have shown that no valid arguments
against representationalism can be construed out of such hypotheses unless
the falsity of representationalism is already circularly presupposed. In
addition, I provided some arguments against the prima facie conceivability
of IS scenarios.
In Part II, I argued that perceptual content is thoroughly Russellian so
perceptual experiences have no Fregean contents. The notion of perceptual
Fregean content proposed by Chalmers—something like [the property that
is (normally) causing this phenomenal property of my experience]—is
highly problematic. First, it involves the notion of causation, even that of
normal causation, which is implausible as a constituent of a perceptual,
non-conceptual content. Second, such content is phenomenologically inapt
insofar as it is incompatible with transparency: when we attend to our own
experiences, we seem to be confronted with the properties things look to
have not with the normal causes of these properties. Third, Chalmers
assumes that one can pick out a phenomenal property independently of the
way things seem to objectively be arranged according to one’s experience,
but that possibility is far from obvious. From these and other arguments I
have concluded that the double-content view is untenable, and that visual
content is wide, Russellian and external, and as such must be the
phenomenal character as well. The narrowness of phenomenal character,
despite a compelling intuition in favor of it, cannot be saved, not even by a
double-content theory. Phenomenal externalism must be true because
content externalism and (impure) representationalism are true.
CHAPTER SIX
SEEING-AS AND THE RANGE
OF REPRESENTED PROPERTIES
PART I
SEEING-AS BETWEEN “THIN”
AND “THICK” PROPERTIES
I.1 Sensible Profiles and the Properties They Are Made
Out Of
In Chapter One, Part IV, I introduced seeing-as ascriptions. When not
metaphorically ascribed (like “S sees Obama as a good president”),
seeing-as episodes are ascribed as vision-based recognitional acts, as
occurrent exercises of a general, recursive recognitional ability toward a
certain type. Being able to recognize examples of F as Fs amounts to
possessing the {F}-category: seeing-as may well be non-conceptual
insofar as categories are not concepts.1 Although they are general and
entertain a subsumption-relation with their examples, as concepts do, their
possession does not require the generality constraint (GC) to hold.2 In
order for S to possess the categories F and G, S does not have to grasp
what it is to see a as a G and b as an F just because S sees a as an F and b
as a G. In short, context-free recombinability and inferential roles in an
open-ended range of possible propositional attitudes are hallmarks of
concept-possession but not of category-possession, least of all of
possession of perceptual categories.3 Recognitional abilities are not
concepts, seeing-as episodes are not seeing-that episodes, acts of visual
recognition are not perceptual beliefs or judgments. Nonetheless,
recognitional seeing involves a cognitive stand toward a seen object: when
1 See Prinz 2002 on that difference.
2 See Chapter Two, Part I.6.
3 As Prinz 2002 argues, another difference between conceptualization and
categorization is that concepts are representations that must be able to be actively
tokened by the subject, whereas it is allowed that categorical representations may
be tokened only passively.
Seeing-As between “Thin” Properties and “Thick” Properties
233
S sees O as (an) F, S takes O to be an F.4 As a “taking,” seeing-as is a
belief-like state with cognitive import, even if it is not a fully-fledged
belief. As a belief-like state, it is truth-evaluable. Because of its cognitive
import, it is essentially an exposure to error.
I have also argued that seeing something O as (an) F can take two
forms according to the value taken by F: F can be a SCM-property,5 like a
certain size, color, shape, motion and position or a thicker, non-SCMproperty like [prey], [duck], [table] and the like. Provided that seeing a as
F entails seeing a, and seeing a entails discriminating some of a’s visible
properties (SCM), when F itself is a SCM-property not only is that
property visually discriminated but it is also recognized as such. Seeing
something O as [red] presupposes seeing O through discriminating its
color-property, but it involves more than that. It also requires that the
recognitional ability concerning the [red] property is activated for a
categorization: O is seen as an example of the general type [red]; it is
brought or subsumed under a general type. Now, the sensitivity to SCMproperties is the basis for recognitional seeing, concerning the very same
SCM-properties (which can be recognized-categorized only if
discriminated)6 and the thicker properties like [prey], [duck] and the like. I
have also stated a principle concerning such relation between “thin” and
“thick” properties with respect to seeing-as ascriptions: when F is not an
SCM-property, seeing O as (an) F, as an occurrent exercise of a
recognitional capacity for Fs, must be prompted by the visual appreciation
of O’s sensible profile. I have characterized a sensible profile as a complex
and unified cluster of SCM-properties. Now we are in a position to
formulate that principle within the theoretical framework of the CV:
4 By “taking O to be an F” I do not mean “believing that O is an F” even though I
am referring to a belief-like state. A recursive behavioral attitude toward things
that are F, where such an attitude varies according to variation of contingent
desires and needs, counts as “taking O to be F” in the very basic sense I am using
that expression. Object-recognition is taking-as, but it is not necessarily believingthat.
5 By SCM-property, I mean “spatial-chromatic-morphological.” SCM-properties
are those properties the early visual system as such is basically sensitive to. The
early visual system can then deliver its outputs as inputs to other systems (to the
late visual system in the first place), which extract or represent higher-level
properties.
6 You can discriminate a property without being able to sort an instance of it as
being of the same type of other instances you have perceived before. That is why
seeing something as F is not presupposed by discriminating F. For discrimination,
object-seeing without seeing-as is sufficient. See Chapter Three, Part I.
234
Chapter Six Part I
In order for a PE to represent a perceived object as being F, when F is not an
SCM-property, that PE must represent certain SCM-properties of the perceived
object in the first place.
Perceptually recognizing something O as an F involves an appreciation
of O’s sensible profile as typically characterizing Fs. There is a set of
SCM-properties you must be visually aware of in order to successfully see
something as an F even when F itself is not an SCM-property.
A parallel principle may be held with respect to “looks F” ascriptions.
Indeed, if S sees O as an F, then O looks (an) F to S. So, in order for O to
look F to S when F is not an SCM-property, O must look G to S in the first
place, where G is a set of SCM-properties.7 In order for O to visually look
like a pig to S, O must look to S as having a certain shape, certain ways of
moving, a certain dimension, certain colors and so on: it must look to
exhibit a visible profile typically exhibited by pigs. In visual seeing-as-F
as well as in looking-F, either F is an SCM-property or the recognitional
category F is prompted by S’s PE by virtue of that PE representing a
complex of SCM-properties unified in a typical and recursively
identifiable sensible profile.
In Chapter Three, Part I, I argued that seeing-as episodes are
characterizable by Peacockean proto-propositional contents (PPC), whilst
object-seeing is to be associated with scenario content (SC). Insofar as an
episode of seeing-as entails an episode of object-seeing, ascription of a
perceptual PPC presupposes the ascription of a compatible SC. I also
argued that a given SC constrains which PPC the experience could have,
but that SC could be tokened without one or the other PPC (compatible
with it) being tokened. The same SC may make room for representations
of different properties, like certain symmetries instead of others, or rows
instead of columns, as in the square/diamond and rows/column cases
considered before.8 Not only is SC non-conceptual but PPC also: you do
not have to deploy the concept [symmetrical with] in order to be able to
visually detect symmetry; to see two lines as symmetrical. However, PPC
7 Here I am talking of (visually based) intentional or epistemic look: looking F as
looking to be F. I have argued that phenomenal properties are intentional
properties given under a mode so also in phenomenal “looks F”, either F is a
represented property or it is a property that could be represented in another
experience similar in some respect to the present experience. For example, the
tilted coin looks elliptical because it looks in some respects the way an elliptical
(not tilted) coin would look to be. In any case, here I am talking of those
intentional ways of looking present in phenomenal character, so of properties
things looks to have according to the phenomenal character of our experience.
8 See Chapter Three, Part I 1-4.
Seeing-As between “Thin” Properties and “Thick” Properties
235
is categorical insofar as you need to have the general recognitional
disposition for the [symmetrical with]-relation, in order to represent it in
your PE and see line a as symmetrical with line b. Now, certain properties
represented thanks to the experience having a PPC are “thin” visual
properties like [symmetrical with], [distant from]; but also properties like
[red] or [square] can be (not only discriminated but also) objects of
classificatory recognition. However, it prima facie seems that other
properties are also represented by the experience’s PPC that are not “thin”
visual properties but “thicker” ones, like [duck], [chair], [animal] and the
like. At least it is an open question whether PEs can do that or not.
In criticizing Dretske’s theory of seeing (Chapter Three, Part I.5), I
argued ad abundantiam that seeing-as or recognitional seeing is
irreducible to seeing-that or propositional seeing so there must be a level
of cognitive, “categorical” but pre-conceptual and proto-propositional
perception that epistemologically, semantically and phenomenologically
mediates between basic object-seeing (and relative SC) and acquisition of
fully-fledged perceptual knowledge. However, I have focused on the
semantic structure of “seeing O as (an) F” without trying to distinguish
when seeing-as is a perceptual recognition and when it is something more
than perceptual: if fact, not every non-conceptual cognitive activity must
be perceptual.
A very important issue for CV is that of establishing how “thick” or
rich perceptual content is, namely, establishing what the range of the
properties a PE can represent its object as having is. Are there perceptual
contents (represented properties), for example, visual contents, beyond the
SCM-properties typically computed by the (early) visual system? If there
are any, when do represented contents stop being perceptual? Another way
of couching the same question is: at what point of “thickness” of the value
of F does “seeing a as F” become not a literal but a metaphorical or
analogical ascription (for seeing = considering or coming-to-consider also
by visual means)? Can the property [animal] or [prey], for example, be
visually represented? And the property [duck]? How are we to decide
whether a property recognized in a seen object just by looking at it is
represented in visual perception? If the very fact of being recognized just
by looking was sufficient for a property to be represented in visual
perception, why not consider the property of [being my laptop] or [being
the newest model of Mac] a property represented in visual perception?
Indeed, now I have a learned non-inferential disposition to recognize
certain properties like being a certain model of a Mac, so that information
is outputted as an instantaneous visual recognition without inferential
236
Chapter Six Part I
process.9 So, provided that we need to distinguish between basic
discriminatory seeing and recognitional seeing, we are left with the
question of what object-recognition is visual and what is not visual, a
question that for CV becomes what properties are represented in visual
perception and what properties are represented thanks to a visual
perception plus some other cognitive, non-perceptual categorization or
inference.10
I.2 Liberals and Conservatives
There are two main views concerning the range of properties that can
feature in the content of visual perception. According to a “conservative”
view,11 perceptual content is “poor” or “thin” so only low-level properties
feature in it (for vision, only the properties I have been calling SCM);
according to a “liberal” view,12 perceptual content is “rich” or “thick” so
high-level properties, which are not SCM, also feature in it. Among the
advocates of the rich CV, there are more or less radical options, some
going as far as including natural kinds and causation into perceptual
content.13
9 It could be replied that the inferential process is implicit rather than absent. That
may be the case but then how are we to decide whether there is implicit inference
or no inference? One could take any perceptual recognition as the outcome of an
implicit inference. The problem is transferred to the lack of criteria for implicit
inference.
10 The very same problem also holds for the distinction between perceptual and
non-perceptual beliefs. I come to believe that the cat is on the mat by visual means,
so that seems to be a perceptual belief due to the way it is acquired. But I also
come to believe that my parents are on holiday by looking at their closed house:
that belief is perceptually acquired as well. There is collateral information, yes, but
also my recognition of the cat and the mat is partially due to my memory and other
not presently incoming information. Intuitively, as I have argued (Chapter One,
Part III.4), one episode is just the perceptual appreciation of a certain circumstance
(the cat’s being on the mat); the other is the appreciation of a circumstance (my
parents being away) on the basis of the perceptual appreciation of another
circumstance (the house being closed). What is difficult is to spell out a criterion
for distinguishing perceptual and non-perceptual beliefs that goes beyond
intuitions and neatly demarcates the two types of belief.
11 See McGinn 1982, Tye 1995, 2000, Dretske 1995, Lormand 1996, Price 2011.
12 See Strawson 1994, McDowell 1998, Carruthers 2000, Millar 2000, Peacocke
2003, Siegel 2006, Bayne 2011.
13 See Siegel 2010.
Seeing-As between “Thin” Properties and “Thick” Properties
237
The issue may concern cognitive science as well as philosophy of
perception. Marr’s famous work on vision, for example, addresses the
question of “what kind of information is vision really delivering” (Marr
1982, 35). If one thinks that vision and visual experience is representation,
then the question is about what kind of properties are visually
representable or represented in vision.
However, neurophysiology of vision and knowledge of the visual
system is not sufficient to get an answer: everything we know about the
physical systems that process information about “thin” visual properties on
the one hand and the physical correlates of thicker categorizations of the
percepts on the other leaves indeterminate where the visual job stops and
other ultra-visual cognitive operations begin. For example, a conservative
will identify visual information with the one delivered and processed by
the early visual system located in the areas V1 and V5, whereas a liberal
will take also the fusiform face area (FFA) and the inferotemporal cortex
(IT)—involved in the operations of the late visual system, like objectrecognition and classification—as neural correlates of visual experience.
Some even hold that the issue is more terminological than
substantial—concerning which cognitive processes, starting from vision,
we still call “visual”—but if we consider, for example, perceptual
experience with respect to its evidential role in acquiring perceptual
knowledge, then the question about what range of properties can be
perceptually experienced may have relevant consequences for the theory
of perceptual justification.14 In any case, even if the demarcation is
pragmatical and blurry, it would already be a substantial result: a
“continuist” view of the relation between perceptual and non-perceptual
representations could, for example, lead to a certain view in epistemology
of perception.
Usually, the liberal strategy starts with the representationalist
assumption that a variation in the phenomenal character of a PE entails a
variation in its representational content, so if, for example, seeing an
object as a pig makes a visual phenomenal difference from seeing—ceteris
paribus—the same object but not as a pig, then the represented property
[pig] must belong to the very visual content of my PE. I shall critically
discuss such a methodology, called the phenomenal contrast method,
(Siegel 2010) and take some cases and examples of its application at face
value. More generally, I will argue for the following points:
1) Perceptual content is thick, but not so thick as to include natural kinds
so the most plausible view is a moderately rich content view.
14 Siegel 2010, 7-11 and Bayne 2011, 18 argue that the debate is substantial.
238
Chapter Six Part I
2) The two-layered dimension of perceptual content is only partially
testified in phenomenal character. Not every property included in the
thick layer of perceptual content is phenomenally given (at least, not
with a distinctive perceptual, mode-dependent phenomenology, e.g.,
visual).
3) The phenomenal contrast method can be of help but it alone cannot
settle matters.
Dealing with this debate will offer me the occasion to delve deeper into
the relation between phenomenal character and the representational
content of perceptual experience.
I.3 A “Thin” and “Thick” System?
The Dual Content View
The so-called “low-level vision”15 enables the recovery of certain physical
properties like surface-orientation, depth at different points of the visual
field, motion, boundaries of objects, colors and the like. Such information
is processed by the early visual system, whose operations are bottom-up
processes starting from the incoming stimulus and using general principles
independent of collateral knowledge and higher-level information like
object-specific knowledge. Low-level vision processes are highly
specialized in subsystems operating in parallel;16 they are pre-attentive and
stimulus-driven. According to Marr’s theory, the outputs of these modules
are used to produce different stages of representation of the perceived
world. First, there are low-level subsystems that encode information about
small edges derived from discontinuities in incoming light on the retinas
then intermediate-level representations encode information about contours
and surfaces, constant colors, depth. That representation, which Marr calls
the “2 ½-D sketch”—where the properties of surfaces, such as depth, color
and orientation are explicitly represented—is further processed so to
obtain a 3D sketch, the “final product” of low-level visual processing.17 A
3D sketch is a description of the observer-independent three-dimensional
shape and geometry of an object, which prepares the subject for recognition
15 See Yuille/Ullman 1990, Pylyshyn 2006.
16 For example, starting from the retinal irradiation, one subsystem realizes the
edge-detection, another the color-detection and so on. See Yuille and Ullman 1990
for an overview.
17 Actually, the story is much more complex than that, and involves other
intermediate level stages of information processing although for our needs it is not
necessary to get into the details.
Seeing-As between “Thin” Properties and “Thick” Properties
239
of the object.
Object-recognition is the main function of “high-level vision” realized
by the late visual system. Such performance is not stimulus-driven and
bottom-up but top-down; it is not parallel but sequential; it is not
encapsulated but cognitively penetrable by attention, memory, expectancies
and collateral knowledge in general: object-recognition is “the activation
in memory of a representation of a stimulus-class—a chair, a giraffe, a
mushroom—from an image projected by an object to the retina”
(Biederman 1990, 42). The geometrical representation of the object
delivered by the low-level vision processes activates an entry-level
representation of that object in memory on the basis of certain shapeproperties shared by all objects of that class.18 There are many theories of
object-recognition (likewise there are many theories of low-level vision);19
I am not concerned with specific details and theoretical options but with
the general point that there are two co-operating visual systems, one of
which represents certain basic properties of the seen objects, the other that
enables the higher-level processes of recognition, classification,
identification and categorization of these seen and visually represented
objects. That second process essentially involves memory insofar as the
basic visual representation of the seen object activates, or is matched
against, a like representation previously stored in memory. An F impinges
on the subject’s sensory transducers (the retinas), this visual input is
processed so as to produce a basic representation (low-level vision), and
then such a representation is somehow matched against stored representations
of Fs so a recognitional episode occurs.
Now, it is far from clear at which point that multi-staged and complex
process becomes conscious and phenomenologically salient. For sure the
many parallel and sub-personal processes by the low-level modules, which
occur in parallel, are out of our conscious accessibility; rather, only their
byproduct enters our visual phenomenology. We seem to be visually
18 Besides the so-called entry-level classifications (giraffe, chair, mushroom and
the like), which are shape-indexed, there are also subordinate classifications (as a
certain sub-species of giraffe), which are also shape-indexed, and superordinate
classifications (as mammal, furniture, object), more abstract and not shapeindexed. See Biederman 1990, 48ff.
19 One of the most promising is the geon theory proposed by Biederman (a short
presentation is in Biederman 1990): a given view of an object can be represented
as an arrangement of simple primitive volumes (geons). Any object can be
specified by a few geons. The full family of geons has twenty-four members, the
recognition of which (in combination) allows us to recognize just-by-sight up to
thirty thousand types of objects.
240
Chapter Six Part I
confronted with objects having certain SCM-properties and belonging to
certain sorts (a table, a known face, a dog).
The issue we are facing now is whether “thick” properties are
represented in perceptual experience, and which, if any. Even if the
existence of two levels of vision and their respective dedicated systems is
not conclusive enough to answer these two questions, it is important to
keep such a double-level nature of visual perception in mind. There is a
more basic representational level of strictly visual properties, and there is
recognitional visual perception, where the first capacity is presupposed by
the second: only through representing certain strictly visual properties in
the first place can a subject be engaged in visual recognition. Perhaps the
difference between object-seeing (and respective SC) and seeing-as (and
respective PPC) may be at least roughly mapped by low- and high-vision,
so we would have two layers of semantical characterization of PE
mapping two operative levels of the visual machinery (at least in us). In
any case, at a phenomenological level, we can introspectively distinguish
certain objective appearances of shapes, colors and movements, from other
types the seen objects are taken to belong to by us just by looking at them.
The main difference between low-level vision and recognitional vision
is that the first depends on wired-in and phylogenetically fixed
mechanisms and modules whereas the second is an activity that involves
learning and individual storing of representations in memory.
Recognitional perception is memory-based so it is sensitive to individual
encounters. Recognitional categories or abilities are acquirable in a way
basic sensitivity to low-level visual properties is not. On the other hand, it
is only insofar as there are species-specific, wired-in mechanisms of
representation of SCM-properties that perceptual learning is possible at
all; in this way we acquire new recognitional abilities and make our
perceptual contents “thicker.” Perceptual identification, recognition and
categorization are performed based on basic discriminations of SCMproperties in the first place. In order to gain a perceptual recognitional
capacity, I must be sensitive to certain differences manifest in my PE, on
the basis of which I can exert higher-level sortings, identifications and
classifications. Nobody could ever learn ex nihilo. As Quine 1969, 306
puts it, “There could be no induction, no habit formation, no conditioning,
without prior dispositions on the subject’s part to treat one stimulation as
more nearly similar to a second than a third.” So, there cannot be any
learning without a previous innate capacity to measure perceptual
similarity, for example, between colors, shapes, ways of moving and so on.
To use my terminology, there cannot be perceptual learning without
sensitivity to SCM-properties and consequently similarities/dissimilarities
Seeing-As between “Thin” Properties and “Thick” Properties
241
between sensible profiles, which are made out of such properties.
Acquiring a recognitional ability could be equated to forming a new
perceptual scheme of immediate categorization such that you see O as an F
just by looking at it because your actual visual input is matched against a
stored representation of F. Perceptual recognition is normally the result of
an acquired recognitional ability20 such that S’s seeing an F now prompts
the automatic categorization of it as (an) F. So, seeing-as is an
ontogenetically acquired capacity depending on one’s individual learning
history. However, it is based on the more basic, philogenetically fixed
capacity of discriminating SCM-properties, namely, “strictly visual”
properties.
Episodes of recognitional seeing are cases where the incoming percept
takes on the semantic content of the stored percepts it is matched to, so it
is seen and represented as an example of (an) F. Through perceptual
learning, the organism groups discrete and unified complexes of SCMproperties (visible profiles) into types or sorts so a new perceptual
category, or recognitional scheme, is established. As soon as that
perceptual scheme is stabilized, a new semantic content is embedded into
the type of PE caused by examples of F when recognition is triggered by
the very incoming percept. Recognizing a seen percept as an F entails
treating it as such in behavior. That is why perceptual categories can be
ascribed also to learning by non-linguistic animals—in fact, they must be
ascribed to them—according to the fine-graininess of their behavioral
flexibility. Whether an object O can look like an F to S can be assessed by
checking if S behaves toward different examples of F in sufficiently
different contexts of presentation in a constantly appropriate way
(consistently with S’s actual desires and needs).
Intuitively, which properties are represented in visual perception
depends on the proper function of the visual apparatus. So, determining
the function of the visual system should put us in a position to determine
the range of properties that can be visually represented. However, that
attempt is not very promising—at least it is not sufficient—because the
function of the visual system is that of both detecting and representing
strictly visual properties (SCM), and enabling thicker recognitions of the
seen objects to enable behavioral success.21 We need a criterion for
20 I say “normally” because nothing prevents the existence of innate categories.
On the contrary, it seems that categories like [moving living being] or [voice of a
conspecific] are innate and not acquired. However, the most recognitional abilities
we have depend on perceptual learning, and so on individual encounters.
21 There are different systems for different levels of object-recognition and
categorization. The “entry level” categorization, at which we tend to spontaneously
242
Chapter Six Part I
establishing which vision-based recognition is genuinely visual and which
is not. On the one hand, cognitive sciences consider perceptual systems
those that “take the inputs from the world and not from the larger
organism” (Lyons 2005, 241); on the other hand, memory is storage of
information about the world in the “larger organism.” But—as I will argue
later in this chapter—top-down influences occurring in perceptual
recognition do not make perceptual recognition less perceptual just
because the mental representation is not purely stimulus-driven. Thus, the
following criterion:
the mental state M is a perceptual representation only if it is outputted by a
perceptual system22
is also defective or incomplete until it is made clear how we individuate
the boundaries of a perceptual system like the visual system. If only innate
modules and subsystems are considered part of it, then only representations
of SCM-properties are visual; if, instead, acquired recognitional abilities
resting on perceptual memory are also taken to become part of the global
visual system at a certain point of the organism’s learning history, then
perceptually learned categorizations will also be delivered by the visual
system considered as a function having a lower and higher representational
level. The visual system feeds its outputs into more central systems, like
the memory system, the reasoning system and so on. The problem is that
visual memory engenders feedback changes in the very visual sensitivity
of the organism toward certain previously encountered properties;23
therefore, by acquiring new vision-based recognitional dispositions, the
organism’s PEs can be taken to acquire new thick perceptual contents.
Thus, seeing-as as a prerogative of learning animals can be taken as a
way of expanding the thickness of the perceptual content possessed by the
organism’s PEs. Even if in a certain sense we do not learn to see because
identify visually presented objects, for example [table], [chair], [apple] is located
in the left hemisphere, while the “subordinate level” at which we classify an apple
as a Granny Smith or a computer as a Mac is in the right hemisphere. Both are
distinguished from a superordinate level at which we make generic categorizations
like [object], [fruit] and the like. See Marsolek 1999, Jolicoeur et al. 1984.
22 Something along these lines is proposed by Lyons 2005.
23 Also from a neural point of view, there is evidence of those pathways in the
input systems traveling forwards from our sensory receptors and backwards; for
example, from the inferotemporal cortex (associated with high-level vision) to
prestriate areas, associated with intermediate-level vision. See Lamme and
Roelfsema 2000, Kosslyn 1994. Certain recognitions may drive attention to better
explore and detect low-level properties, for example.
Seeing-As between “Thin” Properties and “Thick” Properties
243
sensitivity to SCM-properties is wired-in and species-specific, in another
sense we do learn to see because we can perceptually learn to see-as
through exploiting the visual memory of certain types of SCM-complexes,
of certain types of visual profiles.
I will call that two-layered conception of perceptual content the dual
content view.24 According to the dual content view, perceptual experiences
have a double semantic content, namely, there are two layers of properties
PEs represent their object as having: the SCM-properties and related
sensible profiles on the one hand, and other deeper, “thicker” properties
that are represented by virtue of being associated—through memory and
learning—with certain sets of SCM-properties on the other.
The dual content view is an answer to the question whether properties
other than SCM are or are not represented in visual perception. It is a
positive answer that corresponds to the view that seeing-as can be a
genuine perceptual activity. Recognitional seeing is a visual skill so the
respective recognized contents can be bona fide visual contents even if
they are not SCM.25
However, the dual content view as such is silent about which
properties, other than SCM, can be represented in visual experience.
Although the requirement for a property being representable in visual
experience is not being an SCM-property that does not mean that any
property whatsoever can be represented in a visual PE.
Now I shall test the phenomenal contrast method (Siegel 2006, 2010)
as a way of showing that PEs have thick contents by demonstrating that
the presence or absence of such thick contents make a phenomenal
difference. I will try to show that such a method, though useful, is not as
conclusive as its proponents take it to be. I will take some well-known
examples as a testing ground.
24 I freely take the expression from Prinz 2006 but, as will soon become apparent,
my view is different from Prinz’s.
25 That is to say that visual experience is not only an input to the process of
recognition where the input matches the memory representation but also an output
of such a process.
PART II
THE PHENOMENAL CONTRAST METHOD
II.1 Introducing the Phenomenal Contrast Method
Perceptual introspection is the act of attending to the felt character of the
conscious experience we have in perceiving. But introspection may not be
reliable. Furthermore, even if it were reliable, it may be harder than it
prima facie seems to decide what exactly introspection tells us about the
nature and features of our own experiences. Indeed, many incompatible
views on perceptual content and character appeal to introspection as
evidence for their plausibility.
However, introspection can reliably tell us if and when two PEs are
phenomenally alike or not: even if we remain skeptic about our
introspective access to what they differ in, still we can rely on
introspection for appreciating whether they phenomenally differ or not.
The phenomenal contrast method exploits this comparative use of
introspection: consider a target experience E1, an O recognized as having
the thick property F, and a contrasting experience E2, identical to E1
except O is not recognized as having the thick property F. In the
representationalist assumption that a phenomenal change entails a
representational change, if E1 and E2 phenomenally differ, then the
property F being represented in E1 but not in E2 must be the reason why. In
fact, the presence or absence of F is ex hypothesi the only difference1
between E1 and E2. For example, if experiencing a sequence of two
physical movements, one of which is caused by the other instead of just
being spatio-temporally contiguous, makes a visual phenomenal
difference, then the event being one of causation will be represented in one
PE and not in the other, and that will be why the PEs phenomenally differ.2
If seeing a sequence of Cyrillic words before learning Russian, and seeing
1 Contrast-arguments are proposed by Siegel 2006, 2010, Horgan and Tienson
2002, Bayne 2011, Pitt 2004, Siewert 1998.
2 On perceiving causation, see Siegel 2010, Chapter 5; Butterfill 2011.
The Phenomenal Contrast Method
245
it after learning Russian, are two PEs that phenomenally differ, then the
property [being a sequence of Russian words] being represented in the
second PE explains why it phenomenally differs from the first PE; if
seeing a pine tree before becoming able to recognize pine trees just by
looking is phenomenally different from seeing a pine tree after learning
how to visually spot them, then the property of being a pine tree
represented in the second PE must be the reason why it phenomenally
differs from the first PE. Let us more carefully consider some of these
examples in order to appreciate and evaluate the phenomenal contrast
method at work.
II.2 Cyrillic Words and Pine Trees
Let us start with Siegel’s favorite examples cited above.3 The first rests on
the different way it visually looks to read a Cyrillic text before and after
you have learned Russian.4 When you are still learning to read Russian,
you have to focus on each mark and word whilst after you have learned the
language, you will “see through” the shapes and the morphological
properties of the letters and just focus on the meaning of the text. Your
disposition to attend to the semantic properties of the text will overwhelm
your disposition to attend to its orthographic properties. That phenomenal
difference seems to be visual. Under the representationalist assumption, it
must be also representational.
The second example also involves a learned recognitional disposition:
you gradually learn to visually recognize pine trees so your visual
experience of pine trees is phenomenally different once you have become
able to spot them just by looking. Now your PE does not feel like that of a
generic tree but that of the tree you are facing. Pine trees are now visually
salient to you as such, which is why you spot them immediately. So, the
property of [being a pine tree] must be represented in your actual PE if
your previous PEs of pine trees before acquiring the recognitional
disposition were phenomenally different.
Many objections to these examples are possible. First of all, the
phenomenal change could depend on the presence of a “phenomenology of
recognition” in the case when the recognitional disposition is present: so
the difference may be in cognitive rather than visual phenomenology; for
3 See Siegel 2006, Siegel 2010, 100ff.
4 The same example is already in Peacocke 1992, Chapter 3, and in Searle 1983,
55ff. Block 1995 discusses the auditory analog: think of hearing meaningless
sounds and hearing the very same sounds but when you understand their meanings.
It seems that there is an auditory phenomenal difference between the two PEs.
246
Chapter Six Part II
example, as associated with taking the tree as a familiar object, as an
already encountered object, or as a tree of a known type or as a pine tree.
So, a feeling of familiarity,5 the belief that it is a familiar object or even
the belief that it is a pine tree could make a phenomenal but non-sensory
difference.6 Even leaving aside the controversial hypothesis of a cognitive
phenomenology associated with beliefs, there could be a cognitive
phenomenology of the belief-like state of “taking” the object as an F, so of
the pre-doxastic commitment to a certain thick content that is not visual
though accessed by visual means. Thus, the appeal to a distinctive
recognitional phenomenology could explain the phenomenal difference
without need of positing a difference in visual phenomenal character.7
Siegel (2010, 104-105) replies to that objection through a fictional
story. Suppose you are an expert pine-spotter but you are told that the
forest has been replaced with a hologram. So, if the phenomenal difference
depended on your commitment to the fact that you are facing a familiar
object (or a pine tree), your belief that you are facing a hologram would
make the PE lose its additional character and would be phenomenally alike
to PEs of pine trees before you became a pine-spotter. But this is
intuitively not the case: seeing the hologram would be like a spotter’s PE.
This reply is not very convincing. Once you have acquired the
recognitional disposition, any pine-tree-like sensible profile would
immediately trigger the exercise of that recognitional capacity. Collateral
testimonial information that you are not facing an F can change your
propositional beliefs and judgments but cannot have an immediate effect
on a stabilized recognitional disposition or its allegedly distinctive
phenomenology. The doxastic impenetrability—at least in the
immediate—of that recognitional “taking” does not speak against its being
a cognitive stand toward certain content.8
5 Likewise, also a feeling of novelty or unfamiliarity before an object never before
encountered can contribute to what it is like to have a certain experience. Think of
situations where you are trying to get what an object is, which kind of thing it
could be: that “guessing” exploration in search of some similarity to things you
already know is likely to have phenomenal effects.
6 On cognitive phenomenology, see Pitt 2004, Horgan and Tienson 2002.
7 On there being a phenomenology of categorization, see also Brewer 2008. A
phenomenological change, he argues, can be “change of conceptual classificatory
engagement” with the seen object. But the classificatory engagement need not be
conceptual in order to determine phenomenology; it can also be a non-conceptual
categorization.
8 Siegel 2010 says that hunches and intuitions are less resistant than beliefs,
whereas the recognitional phenomenology of a visual episode cannot be canceled
by a belief either: so, such a phenomenology cannot depend on a belief or on
The Phenomenal Contrast Method
247
In addition, as a pine-spotter I would recognize a hologram of a pine
tree as being such; I would not see it as a hologram of a generic tree. The
recognitional phenomenology would make the difference insofar as I
would take the hologram as a well-done hologram of a pine tree, like a
picture of my brother feels visually different to me from a picture of a
generic guy (even if I know it is a picture and not my brother) just because
it is a picture of my brother. In short, recognitional phenomenology can be
transferred to pictures, holograms and other well-worked representations.
Take a soap lemon: I know how real lemons look so even if I know that O
is a soap lemon, my recognitional ability toward lemons is activated by O.
O looks just like a real lemon and that is why it phenomenally differs from
the way it would have looked to me before I was able to recognize lemons.
Recognitional ability toward Fs entails recognitional ability toward Fs’
distinctive sensible profiles so that if an F-like familiar sensible profile is
exemplified, it does not cease to be familiar as soon as I come to know
that it is not an F: it is anyway a fake token of a familiar sensible profile
not of an unfamiliar one.
Another argument offered by Siegel also seems far from conclusive.
She presents the example (ibid. 108ff.) of being bombarded by captions on
billboards along the highway, of involuntarily hearing a cell-phone
conversation on the street and the like. If you recognize the letters and
meanings on the captions and the sentences in the spoken sounds, then
your PEs are phenomenally different from respective PEs of a sequence of
shapes or a stream of meaningless sounds even if you are passive and not
attending to these stimuli. She concludes that the taking in is purely
sensory so the respective phenomenology is sensory thereby. But an
income of certain information being passively received does not mean it is
sensory. Deploying a recognitional ability before a perceptual stimulus
could well be passive (“automatic”) and non-sensory at a time9 so the
hunches and intuitions. Bayne 2011, 26, also argues that the representation of
something, like a pipe, must be experiential content because it resists doxastic
penetration. Were it a belief, being told that X is not a pipe would be sufficient for
X to cease to look like a pipe. As I said, I think that is a bad argument: a
recognitional exercise passively triggered by a stimulus does not have to feature in
the visual phenomenal character of the PE in order not to be a structured belief. It
may be neither a belief nor content testified in the visual phenomenal character: it
could be a thick categorization acted by means of a visual “thin” representation.
9 As we have seen above, at least certain categories will be only passively exerted,
unlike concepts, whose possession entails the capacity of their active exercise and
use within inferences and judgments. That does not entail at all that the exercise of
a certain category must have a sensory phenomenal character.
248
Chapter Six Part II
argument does not do: the phenomenal change could well be representational
though non-visual (or sensory).10
Another objection is that E2 may represent a general sensible profile
that is invariant across differences in shapes of pine trees, what Siegel
herself—who considers that objection—calls a shape-Gestalt.11 So, it is
not a [pine tree] property but a typical shape-Gestalt, or what I have called
a visible profile made out of SCM-properties, that is represented in E2 but
not in E1. What makes a visual phenomenal difference is content; not
involving the [pine tree] property but a recognized type of visible profile
(shape-Gestalt) typically associated by me with pine trees. Insofar as the
Gestalt-shape is made out of SCM-properties, the phenomenal and
representational difference between E1 and E2 does not entail any thick
property to be represented in the PE.
Siegel replies in many ways, none of which I find conclusive. First, she
points out that “the more abstract the shape-Gestalt is, the less reason there
is to think that experience fails to represent it prior to one’s gaining a
recognitional disposition” (112). But if the Gestalt-shape is represented
before gaining the disposition, it is not apt to explain the phenomenal
contrast between E1 and E2. The counter-reply is easy to address. The
Gestalt-shape is abstract enough to subsume different examples of pine
trees (more or less leafed, more or less rich with branches and so on) but
not abstract enough to be shared by other kinds of trees or other objects, so
abstract enough to make a phenomenal difference after and before gaining
the disposition:12 the shared properties become relevant for you only once
10 Another objection could be that the feeling of familiarity is a nonrepresentational raw feel, which does not involve the representation of O as
familiar. Siegel opportunely replies that you can fail to get what is familiar in that
you feel familiar, but that would still be a less specific representation of something
as familiar. See Siegel 2010, 109ff.
11 We can distinguish between token-profiles and type-profiles. A token-profile is
a demonstrable, individual, unified complex of SCM-properties whilst a typeprofile is a more abstract and less determinate type of profile, which can be
compatible with a range of different token-profiles, just as pine trees have visible
invariant properties and visible individual differences. Recognizing a pine tree just
by looking does not mean visually representing the [pine tree] property; rather it
means visually representing an individual profile as a token of a more general
type-profile (the one typically associated with pine trees). In my language,
recognitional seeing would be seeing O as F, but F would only involve SCMproperties.
12 Another objection against the liberal view is by Byrne 2010, 80ff: suppose
lemons grown on Island A look like normal lemons, and lemons grown on Island B
look like cucumbers. S acquires a recognitional disposition both for the fruits of
The Phenomenal Contrast Method
249
you have gained the relative disposition whilst they were potentially
detectable but negligible before.
Second, Siegel considers the example of a face expressing doubt, a
property of Mario’s you learn to visually recognize. Provided that such
learning makes a phenomenal difference, it is implausible—she argues—
that before and after recognizing doubt in Mario’s expression, representation
of shapes or colors change. Therefore, the represented property making the
phenomenal difference cannot consist of colors, shapes or other SCMproperties.
But this argument can also be resisted. What could be making the
phenomenal difference could be that this example of sensible profile
typically associated with Mario’s doubtful face is just the example of a
sensible type I reliably use to spot Mario’s doubtfulness. So, [being the
example of a familiar expression] can be a property that, again, determines
cognitive phenomenology. An individual sensible profile is represented
before and after learning that a more general type of it tracks a thick
property but after learning, such a sensible profile is recognized as the
token of a type, so a distinctive recognitional phenomenology is added.
Siegel still counter-replies by supposing a counterfactual situation in
which Mario normally contorts his face that particular way when he is
bemused rather than when he is doubtful and I learn to spot it. In that case,
the property of being the example of a familiar expression would be
represented as well, but intuitively that counterfactual case would be
phenomenally different from the original case about Mario’s doubt. So, the
property of [being familiar], be it “felt” or represented, is not enough for
accounting for the supposed phenomenal difference.
However, that that phenomenal difference—between spotting doubt in
W1 and spotting bemusement in W2—would be visual is just an
assumption if not a question-begging conclusion. For example, it could
well be emotional provided that in spotting others’ emotions, I inevitably
have an emotional response of some sort. So, the visual component of the
alleged phenomenal difference could be accounted for by appeal to
familiarity or recognitional phenomenology on the one hand, and by
appeal to an emotional component associated with myself spotting Mario’s
specific emotions just by looking at his face on the other. My PE, then,
visually represents a certain visual profile as a token of a known and
familiar Gestalt-shape associated with a certain emotion, so that in nonIsland A and of Island B. If Siegel is right, if S sees an A-lemon and a B-lemon
side by side, S will visually represent both as lemons. But that view will absurdly
predict that S will come to believe that they are fruit of the same kind, and that
they will look visually more similar after S has learned to recognize them by sight.
250
Chapter Six Part II
visually representing that visually spotted emotion, I respond with a
certain emotion which influences the overall phenomenal character of my
global state. Gestalt-shape, familiarity and emotional response account for
the phenomenal difference between E1 and E2 even in the counterfactual
situation given that the emotional response to one’s bemusement is
different from the emotional response to one’s doubtfulness, without the
property [being in doubt] having to be visually represented in PE.13
Another important possibility that Siegel does not consider is that there
could be a phenomenal visual difference between E1 and E2 without there
being a difference in visual representation of the thick property even
recognized by visual means in E2. The non-sensory representation of
something O seen as an F can affect the ways O are visually represented,
without F being itself visually represented in the very PE. High-level
representations may determine the visual character indirectly, namely, by
determining a change in the representation of low-level properties
(SCM).14
For example, suppose that recognizing something as a pine tree is not a
visual content even though such recognition can be realized by visual
means. Therefore, even if there was a visual difference between E1 and E2,
taking the percept as an example of a type may be a non-visual cognitive
act but it may nonetheless orient, for example, my visual attention
differently, and attention is a factor that undoubtedly determines the visual
phenomenal character.15 Recognition of the pine tree can give relevance to
certain details in my PE—it can change the way of exploring the percept,
it can trigger different mental images and different expectancies—all of
which has to do with the phenomenal character. For example, as a pine
tree spotter, I am more likely to spontaneously focus my attention on what
differentiates pine trees from other trees; that is, on pine tree markers.
13 The reference to an emotional component in the overall phenomenal character
can also partially account for the Cyrillic case. Once I understand Russian, I get the
meaning of what I read. According to what I am reading, I will have additional
emotional feelings, mental images, expectations and so forth: all that can change
the overall phenomenal character of my visual perception without visual
representation of thick properties.
14 Hansen et al. 2006, for example, show that recognizing something as an F can
change the perception of its color. They presented subjects with pictures of fruit
that could be manipulated so to make the fruit appear any color. When asked to
manipulate the image of a banana so as to make it appear grey, the subjects made it
bluish, as if they were compensating for the knowledge bananas are yellow.
Recognitions create new visual expectancies and integrations.
15 On attention and phenomenal character with respect to visual perception, see
Treisman 1998.
The Phenomenal Contrast Method
251
Likewise, as a Cyrillic reader, I am more likely to concentrate my visual
attention on those details that differentiate one letter from another and one
word from the other rather than exploring a sequence of shapes without
spotting special markers. Our spontaneous search for markers of familiar
types of objects or for markers of known individuals is why we
notoriously tend to misperceive perceived objects and people for familiar
objects and people. The identification of certain features in a perceived
object prompts a mistaken recognitional act insofar as a representation of
those features is stored in perceptual memory as a marker of a certain type
(F) or as a marker of a certain individual (Mario, Fido, my bicycle):
recognition can be either classificatory or identificational.
Gaining the recognitional disposition will provide my visual sensitivity
with a new contrast class: for example, not only can I spot something as a
tree and differentiate it from other objects, I can also spot a kind of that
kind thanks to more specific visible features that track the presence of
certain types of tree. Now, trees will no longer appear as generic visible
profiles; some of them will appear to instantiate the more specific sensible
profile to which I have become sensitive.
In spotting pine trees, I will use certain visible details as markers of a
certain kind; therefore, the immediate encounter with these visual details
will trigger certain responses. Change in attention, the relevance given to
certain details, in the way of exploring, in expectancies relative to the
unseen parts of the recognized object, in the contrast-classes: all these
visual differences can be consequences of representing something as an F
other than visually (for example, categorically or conceptually)16—a visual
consequence of a non-visual representation.
Hence, something is wrong in the very phenomenal contrast method:
showing that two experiences differ in visual phenomenal character if they
16 A similar point is persuasively made by Tye 2000, 61ff. With respect to hearing
a known language, he claims that “there are differences in emotional and imagistic
responses, feelings of familiarity that weren’t present before, differences in efforts
of concentration involved as one listens to the speaker […] there are also
phenomenal differences connected to the phonological processing. Before one
understands French, the phonological structure one hears in the French utterance is
fragmentary.” The idea is that there is a top-down causal influence on sensory
phenomenal character by the high-level categorization or recognition, without
which the recognized properties are sensorily represented. They just make a
sensory difference through a top-down causal influence. If I am told “Find the red
object or you will die!” I would explore the scene in search of a red object so that
the aim will condition the visual way of exploring a scene, my focal attention and
so on. That causal influence does not entail that the very content [red saves my
life] should be visually represented as such.
252
Chapter Six Part II
are identical in everything but only one of them results in recognition of a
thick property does not amount to showing that that very thick property is
visually represented. The recognition of the thick property could have
backward effects on the visual character of the PE without having to be
represented in the PE; that is to say, without itself belonging to the visual
phenomenal content.
In sum, Siegel’s examples and related arguments are not conclusive
because there are plausible alternative accounts of the phenomenal
contrast she starts from. The phenomenal difference—be it sensory or nonsensory—between merely seeing an F and visually recognizing it as an F
can be explained without having to commit to the idea that F itself must be
visually represented.
I am not claiming that the phenomenal character of visual experience is
exhausted by the representation of low-level properties (SCM), as the
conservative view holds. Rather, I am claiming that Siegel’s arguments
based on the phenomenal contrast method do not show that the
conservative view is false. If liberalism is to be vindicated, it must be
vindicated in some other way.
II.3 Does Visual Agnosia Support the Liberal View?
According to the representationalist assumption, there is such an intimate
connection between representational content and the phenomenal character
of PE that a phenomenal change entails a representational change. That is
why the phenomenal contrast method aims at showing that thick properties
are sensorily represented on the basis of the fact that their being
represented makes a difference in sensory phenomenal character. Even if
there are layers and parts of the perceptual content that lack a respective
conscious character (see Chapter 2, Part I.2), the properties shaping the
conscious character vary with variations in content. This is how we can
construct the notion of phenomenal content, the content a PE has by virtue
of having a certain phenomenal character. Such a notion successfully
captures the representationalist assumption about phenomenal character; it
allows us to focus on the perceptual content that is phenomenologically
testified, so to speak, and set aside the eventual other non-conscious
components of perceptual content. So, Siegel’s thesis is that the
phenomenal content of PE contains thick properties. The phenomenal
contrast method aims at showing, case by case, whether a given thick
property (causation, natural kinds, other non-natural kinds) belongs to the
phenomenal content or not.
Now I want to take into consideration a contrast argument in favor of
The Phenomenal Contrast Method
253
liberalism—originally presented by Van Gulick (1994) and then developed
by Bayne (2011)—based on visual associative agnosia (VAA). As we have
seen (Chapter Two, Part I.2.3), VAA is a deficit in object-recognition
whereas the perceptual skills are unimpaired,17 so it is a case of successful
seeing without seeing-as, or of visual representation without visual
categorization. VAA patients are perfectly able to draw the seen objects, to
pair two identical objects, even to accurately describe their SCM-details
but they are unable to recognize the even familiar objects they see (for
example, a pipe, a comb, a key, a can opener) unless they are named. If
you misname the object, the patient will make a visual misidentification of
it according to the naming-driven categorization.
Bayne argues that it is extremely plausible that the phenomenal
character changes in subjects affected by VAA but there is evidence that
the low-level properties are normally represented by their PEs—visual
modules are unimpaired—so the supposed phenomenal change must
concern high-level properties, the ones the subject becomes unable to
recognize (pipe, comb, key, can opener). Therefore, those thick properties
are most likely included in the phenomenal content of unimpaired
subjects’ PEs.
This argument is intuitively powerful only prima facie because in fact
it is question-begging: only if you already share the intuition that the
phenomenal character of visual experience includes representation of highlevel properties will you find it extremely intuitive that your visual
character changes when you are affected by VAA. The unquestionable
intuition is rather that the overall phenomenal character changes, not its
strictly visual component. What it is like for you to see a gun is probably
different from what it is like to see a hairdryer or a drill even though they
may look alike in certain circumstances. There are also non-visual
components of the overall phenomenal character associated with the
recognition of a property by visual means. At least this is an alternative
possibility one could reasonably appeal to.
For example, the way you categorize what you see triggers certain
possible responses connected to what you could do with the seen object,
what way you could behave toward it, which affordances it would allow
you, whether it may be dangerous or useful and so forth. After recognition,
you will also imagine a different internal constitution and a different
degree of solidity, a different weight and material, different ways of
manipulating the object. All these automatic responses are likely to make a
17 Actually, these are rare cases of pure VAA. Mostly, VAA is accompanied by
impairments of low-level visual skills. On VAA, see Ruben and Benson 1971.
254
Chapter Six Part II
phenomenal difference though not necessarily a visual phenomenal one.
If there is a visual phenomenal difference, it can be accounted for by
appeal to the top-down effects of recognition (or lack of recognition) on
the representation of the low-level properties. You will probably explore a
hairdryer, a gun and a drill differently, with attention to different parts and
details. In seeing a part of the object, you will imagistically “construct” a
slightly different mental model of the object and have a different image of
the unseen parts of the object. Your visual attention will be differently
distributed and exerted in different patterns. As in the case of the pine tree,
visual phenomenal difference does not necessarily entail visual
representation of the properties causally responsible for that difference.
The case of VAA is a case in which the absence of recognition of familiar
objects is likely to determine the absence of respective specific ways of
visually exploring the object, the absence of a certain distribution of focal
attention, of certain specific ways of imagining the unseen parts and so
forth.
Bayne takes the example of normal recognition like seeing O as a dog
(2011, 24):
Suppose you are looking at a dog in the distance. The light is poor, you have
difficulty in identifying the seen object. Suddenly, the recognition dawns […]
must there be low-level differences between these two percepts? I do not see
why.
Bayne can take the VAA case as an argument for liberalism only because
he assumes that: 1) VVA affects the visual phenomenal character, and 2)
change is not only in low-level properties. But if there are other options to
rule out, either that the phenomenal change is visual or even the visual
change could be a top-down change in low-level properties, then the
argument that results is very weak if not question-begging. The fact that
the visual system is unimpaired and well-working in a VAA patient does
not prevent it being used in a different way than if they were able to
categorize what they see. Different patterns of attention, different ways of
exploring the object and different ways of visually imagining unseen
details and parts are plausibly present even if the dedicated systems are
technically unimpaired.
The example of recognizing a dog in fact merely restates the
assumption that the change in phenomenal character due to recognition is
not a change in low-level representations. Perhaps Bayne uses this
example to make us aware of the intuitive implausibility of the view that a
sudden recognitional event could have an instantaneous effect on the
visual representation of SCM-properties. But that circumstance is far from
The Phenomenal Contrast Method
255
being implausible: the intuition may be misleading.
First of all, the recognition reactivates a mental image of a stored
category18 that may well change immediate projections about unseen parts
and expectations of possible movements. It can condition the further
exploration as guided by a specifically distributed focal attention and so
on. Perception is active exploration guided by specific goals and tasks.
Any representational updating of a percept inevitably conditions the
further exploration of it, hence a categorization can well involve a
manifold cascade of visual differences in representation of SCM, or lowlevel, properties.
Once the recognition consciously occurs, the search for the right stored
perceptual category to which the seen percept could be successfully
matched has already taken place unknown to you, hence it is not a mystery
that the visual phenomenal character changes suddenly as a top-down
modulation from recognition to change in low-level represented properties.19
In short, there is plausibly a non-sensory phenomenal change associated
with recognition, and there is a visual change as a top-down consequence
of the thick recognition. A conjoined appeal to these two sorts of change
puts us in a position to explain away the phenomenal change concerning
both the VAA and the normal cases compared in contrast arguments, such
as dog-recognition, Cyrillic reading, pine-tree spotting and the like. VAA
does not offer any further substantive support to the liberal view on
phenomenal content because it can also be accounted for from a
conservative point of view.
18 See Kosslyn 1994. According to Kosslyn, the back-projections from the inferior
temporal cortex to the prestriate area of the brain—therefore, from high-level
vision areas to low-level vision areas—are to be explained as the use of imagery to
improve the perceptual representation: for example, when the stimulus is poor
(occlusion, bad illumination, insufficient foveation and so forth), a guess about
object-identity guides the completion of the seen object, so it fills in the details on
the basis of the recognitional category that has been activated. If Kosslyn is right,
visual imagination and visual perception are deeply interwoven so top-down
effects are what is to be expected.
19 Recognizing a dog will take a few dozen milliseconds then a backward
projection can fill in the details in terms of represented SCM-properties. Such a
backward effect can be phenomenologically salient, but you just become conscious
of a difference without becoming conscious of that phenomenal difference being
due to such backward effects.
Chapter Six Part II
256
II.4 Do Ambiguous Figures Support the Liberal View?
A salient application of the phenomenal contrast method concerns the
aspect-switching of ambiguous figures. I have already considered the
example of the square/regular diamond, which is considered an ambiguous
figure.20 The properties [square] and [regular diamond] that can be
alternatively represented by or attributed to the seen figure could be
considered low-level properties, although quite sophisticated ones. In fact,
they are types of shapes so they belong to the SCM-class of properties.
Instead, more interesting, at least as a contrast-case for phenomenal
content liberalism, is the aspect-switching in ambiguous figures where the
properties alternatively represented are not SCM-properties but thick
properties like [duck]/[rabbit], [old girl]/[young woman], [faces]/[vase], as
in the example below I will be now considering.
DUCK/RABBIT*
OLD WOMAN/YOUNG GIRL*
VASE/FACES*
20 See Chapter Three, Part I.1-2.
* The three figures are taken from: http://mathworld.wolfram.com
The Phenomenal Contrast Method
257
According to Searle, the representation of the switching properties
must be part of the content of visual experience insofar as the figure
visually looks different according to whether it is seen as an F or as a G.
Searle’s example is that of a stylized figure (below) that could be taken as
“a table with a line with two large balloons underneath, as the numeral
1001 with a line over the top, as a bridge with two pipelines crossing
underneath, as the eyes of a man wearing a hat with strings hanging down
each side, and so on” (Searle 1983, 54).
Likewise, looking at the duck/rabbit ambiguous figure will be a
phenomenally different experience according to whether you see a duck or
a rabbit in it. The switch is instantaneous and can be voluntarily produced
so we seem to have a case in which the phenomenal contrast is
straightforwardly “felt” rather than supposed, counterfactually imagined,
or intuitively ascribed to others’ experiences (VVA). Another advantage of
such cases is that the change is not gradual as in the case of learning to
read a language or recognize pine trees. In these cases, the phenomenal
contrast is more projected or hypothesized than lived as such because of
the gradual nature of the acquisition of the recognitional disposition, so
that at least long-term memory has to be involved21 in order to do a firstperson, subjective comparison.
The very same figure, which keeps in view and does not change with
respect to its low-level properties, looks phenomenally different according
to whether you represent it under a high-level property or another, as a
duck or a rabbit (or as a young girl or an old woman). Hence, according to
the phenomenal contrast method, the high-level property that makes the
phenomenal difference must be represented in the visual experience. The
phenomenal shift is best explained by a visual representational shift. That
is the argument, in short.
Now I shall discuss these examples and argue that aspect-switch can
also be well explained without committing to liberalism on phenomenal
content.22 Again, liberalism—the dual content view, as I have called it—
21 As we have seen with respect to Inverted Earth (Chapter Three, Part III.4.2),
phenomenal memory can be unreliable.
22 In what follows, I will freely draw and develop some of the arguments from the
illuminating paper of Price 2011 on this issue.
258
Chapter Six Part II
needs to be argued for on some other basis because ambiguous figures do
not speak in favor of it either. Let us begin.
First of all, I have already pointed out that the role of attention is
fundamental in shaping the visual phenomenal character. So, if a different
categorization of the object had an influence on the distribution of the
focal attention as well as on the way of exploring the parts of the object,
then the phenomenal shift could be accounted for by appeal to attention;
so, without involving the visual representation of the thick property, the
object is recognized as being (duck or rabbit, old woman or young girl,
vase or faces). It seems very plausible that the recognition of a known
shape engenders shifts in attention and respective ways of exploring it. As
Price persuasively points out, that difference can be even introspectively
verified.
When seeing the duck/rabbit as a rabbit, I tend to look at the picture from left
to right, and when seeing it as a duck, I tend to look at it from right to left.
Also, when seeing it as a rabbit, I attend to the rabbit’s mouth and eye together;
when I see it as a duck, I attend to the duck’s eye and beak together.23
Such a difference in the sequence and direction of exploration, in
election of points of focal attention, as well as in grouping of the attended
parts can differ from one subject to another. Perhaps there are differences
in patterns of attention that are invariant across subjects and other
differences that are idiosyncratic to individuals: that is an empirical matter.
What matters instead is that a difference in patterns of attention, when
seeing the figure as a duck or a rabbit, is phenomenologically evident; it
suffices to “look at oneself” during the two successive acts of visual
exploration to realize it. Therefore, the phenomenal shift may be
accounted for by appeal to a shift in patterns of visual attention, which
concerns low-level instead of high-level properties. Hence, conservatism
is not threatened by aspect-switch.
It can be replied that such an explanation only partially accounts for
the phenomenal shift: in fact, I can force myself to fix my attention on the
same parts and voluntarily start the recognitional shift even though with a
much greater effort.24 So, assuming that the visual attention can be kept
23 Price 2011, 144. A similar point is made by Tye 1995, 140: “Where a figure has
an ambiguous decomposition into spatial parts, concepts can influence which
decomposition occurs.” Also a non-conceptual categorization, I would add, can
have top-down effects of the spatial decomposition, therefore on the visual
phenomenal character, without being part of it.
24 A similar point is made by Macpherson 2007.
The Phenomenal Contrast Method
259
perfectly constant, the recognitional shift still engenders a phenomenal
shift. Therefore, the shift of attention and the changes in the sequence of
exploration cannot be the whole story.
Therefore, another element needs to be appealed to. As Price himself
argues, that element could plausibly be visual imagination. Any seen
object presents itself in perception as a whole three-dimensional entity
perceived from a certain point of view: even if only a part of it is in view,
strictly speaking, the experience takes the unseen back of the object into
account.25 Now, when recognition takes place, not only is the experience
not neutral about there being an unseen part of the perceived object but
also about how that unseen part is arranged. Moreover, ambiguous figures
are often poor in detail; this is probably one reason why they are
ambiguous. Your visual imagination can fill in the missing details in two
individually consistent though reciprocally incompatible ways. Then,
according to the way you see the object—as a duck or a rabbit, as a young
girl or an old woman—you will visually imagine different details and your
imagination will automatically “project” certain features onto the unseen
part of the object. Take the old woman/young girl figure, and carefully
attend to your own phenomenal shift as soon as you “switch” from seeing
it as an old woman or as a young girl. While seeing the picture as an old
woman, you expect that its unseen part is half of an old woman’s face. You
will, for example, “project” the back part of her neck just behind the hair
you are seeing. While seeing the picture as a young girl, you will imagine
the unseen part to be half of a nice young face so all the three-dimensional
arrangements of what is behind the seen surfaces would be projected in a
totally different way. Thus, it is clear that the recognitional shift has a
cascade of effects in the way you instantaneously imagine the unseen parts
on the one hand, and more generally in the ways you fill in other details,
also in poorly determined areas of the part of the object in view. In
addition, it can be supposed that your virtual expectancies about possible
moves the object could make also make a phenomenal contribution. To
imagine the old woman turning her head, or to imagine the young girl
turning her head, is to imagine two visually different presentations and so
two very different arrangements of low-level properties.
At this point, a liberal may reply that the account of the phenomenal
difference just offered is in fact an implicit commitment to liberalism. If
visual perception and visual imagination were inseparably interwoven, as,
25 Husserl’s notion of “adumbration” or “shading off” is just meant to capture that
presence/absence of unseen portions of the objects. See Husserl, Ideen I, §84. As
Nöe 2004, 2006 remarks, the unseen parts of objects are “virtually present” in
perceptual experience.
260
Chapter Six Part II
for example, Kosslyn and others hold,26 the contents of visual imagination
that integrate the representation of the seen percept should be included in
the very accuracy-conditions of the PE. To make an analogy, visual
imagination in our cases could be taken as a sort of detailed visual
completion: visual completion is the phenomenon for which, for example,
your PE represents certain occluded (then unseen) parts of a seen object.27
Now, if the visually added elements are part of the perceptual content, then
it would be true that recognizing the figure as an old woman or a young
girl would make a phenomenal visual difference.
That reply can be addressed by remembering a point I have already
argued for: making a visual phenomenal difference, for a thick
categorization of O as F, does not entail that the very property F directly
enters into the visual phenomenal character. That is the Achilles’ heel of
the phenomenal contrast method: seeing O as F can make a visual
phenomenal difference as an indirect effect on the low-level represented
properties. In our case—granted that at least a part of such visually
“imagined” contents belong to the genuine visual contents—what visual
imagination adds is not the properties of being an old woman or a young
girl but rather, for example, the low-level properties as constituents of the
sensible profiles respectively associated with an old woman or a young
girl. So, seeing the figure as an old woman involves representing the seen
figure as the seen part of a certain three-dimensional sensible profile, and
that involves the visual phenomenal difference at the level of different
unseen low-level properties. So, seeing the figure as an old woman or as a
young girl will make a big phenomenal difference, but that phenomenal
difference may be only causally, indirectly dependent on the thick
properties being non-sensorily represented.
The appeal to visual imagination is even more important in accounting
for phenomenal shift in very stylized ambiguous figures, as in the example
used by Searle. The poorer in detail the figure is, the more relevant the role
of visual imagination becomes in filling in the lack of details.28 The more
abstract the figure is, the easier it is for us to isolate a “neutral” figure, a
pure shape, so to speak, and so to freely imagine what it could plausibly
26 See, for example, Kosslyn 1994.
27 On visual completion, see Kanisza 1980.
28 There are more stylized versions of the duck/rabbit figure, for example, for
which the completion of details is mostly accomplished by visual imagination.
This is why, in these versions, it is also easier to isolate the drawing and consider it
independently of what it may represent, namely, as a set of abstract signs or lines.
The Phenomenal Contrast Method
261
be.29 In that case, the phenomenal shift is less “constraining” and more
malleable by our free will because the active imagination outstrips passive
recognition due to the absence of details constraining the interpretation
into fewer but more specified interpretations.
Finally, I want to point to another way of accounting for the
phenomenal shift in terms compatible with a conservative view. Visual
experience presents us with three-dimensional objects located in space and
oriented in certain ways with respect to us. The way an object looks is
determined by the way its silhouette is oriented and placed in egocentric
space. For example, how the object looks, what part is behind what other
part, is relevant from the point of view of an observer.
Now, look at the duck/rabbit figure above: while seeing it as a duck,
the inferior and the superior part of its beak appears to be on the same
ideal depth level they are not one behind the other; while seeing the object
as a rabbit, the very same parts do not appear to be aligned with respect to
their distance from me anymore. On the contrary, when interpreted as the
ears of the rabbit, one of them is nearer to the observer, the other is a bit
further away, and they are no more one over the other as when they were
represented as parts of a beak; rather they are horizontally aligned. That
means that all the perspectival dimensions of the visual representation
change, so with such a change in the representation of low-level
properties, a phenomenal visual change is expected. Therefore, an
assumption of the argument from aspect-switch to liberalism is simply not
true. It is true that there is phenomenal change but it is not true that this
phenomenal change occurs by the low-level properties being constant.
What is constant is the “patch” on the page, but what is relevant is not the
two-dimensional patch but rather the low-level properties that are
represented once that “patch” is interpreted as one three-dimensional
figure or as another. The renowned psychologist Gregory (1997, 195) says
29 Siegel 2010, 20-21, makes the example of a cloud that can be “seen as” a
crocodile or a hotdog. That is a case in which the free play of visual imagination is
the most relevant factor. What is visually represented in the first place is the cloud,
so there is a thin content made out of SCM-properties and a thick content
consisting of the recognition of the object as a cloud. In alternatively representing
the visually imagined properties [hotdog] and [crocodile], a change in visual
phenomenal character could be explained by: a) change in attentional relevance,
shifting from the spatial properties the cloud could share with a huge hotdog to the
spatial properties the cloud could share with a crocodile, or b) change in visual
imagination of different ways of filling in the details both of the seen object and of
the “supposedly” unseen parts of the respectively recognized kinds (hotdog,
crocodile.)
262
Chapter Six Part II
that when we look at ambiguous figures, the “phenomenal phenomenon”
changes but the “stimulus input” remains constant. Such descriptions can
be misleading if given a hasty philosophical interpretation. That both the
distal stimulus (the drawing on the page) and the proximal stimulus (the
retinal images) are constant does not mean at all that the represented lowlevel visual properties are constant thereby.30
The same point can be applied to the young girl/old woman figure.
When I look at the figure and see it as a young woman, her neck appears
to be nearer to me than her shoulder, and the neck appears “attached” in a
continuous line to the shoulder. They are parts of the very same smooth
surface. The necklace rests between the neck and the shoulder, which are
“continuous” parts of the smooth bodily surface. The seen part of the girl’s
face is more distant from me than her neck and her nose is more distant
from me than her neck and jaw. As soon as I see the figure as an old
woman, however, all these spatial relations significantly change. Now the
“supposed” nose of the young girl becomes a sort of protuberance of the
supposed nose of the old woman so it is not behind it anymore (“it”, the
nose of the old woman, corresponds to the face of the young girl); rather, it
is aligned with it on the same ideal level of depth, so to speak. The
supposed chin and mouth of the old woman are before the woman’s
shoulder with respect to me (nearer to me), whilst before it was attached to
the girl’s shoulder (it was the young girl’s neck). Additionally, the old
woman’s nose is oriented toward my side whilst before it was oriented
toward the other direction (it was actually the young girl’s face). I could
continue with such a parallel spatial description but these remarks are
enough to get to the fundamental point: it is not true that the low-level
properties represented by the two cases of recognition are the same.
Therefore, it is a fortiori not true that the only way to account for the
phenomenal difference is the visual representation of the very thick
properties F and G the object is alternatively seen as being. The contrast
argument is that there is a visual phenomenal difference but it cannot come
from the low-level representation, which is identical; it must come from
the high-level properties being represented in the respective visual
experiences. Even if the argument was conclusive—I argued above that
this is not the case—a premise is clearly false so the conclusion is not
30 In fact, the retinal images are not the same. Some instantaneous images can be
shared but if I am right in supposing that patterns of attention change, the sequence
of the retinal images in the two cases changes accordingly; insofar as the saccadic
movements will change, the points of foveation will change and so forth. The
proximal stimulus over time will be different after and before the shift, and that
difference will mirror the difference in representation of low-level properties.
The Phenomenal Contrast Method
263
ensured. The phenomenal difference may well come from the huge
differences in representation of low-level properties, especially the
representation of the spatial arrangements of the “seen” silhouettes.
This explanation is also available for more abstract ambiguous figures
like the [vase]/[faces]. As soon as you see the two faces one before the
other, the black background will appear to be behind them. If you shift to
the vase interpretation, then the black surface will be before the white
surface, which becomes the background. Moreover, the visually imagined
unseen part will change a lot in the two cases: the back of the vase will be
round in a certain way; the back of the two faces would be face-like, with
specific volume and features. These are differences in the sensible profiles,
in particular the ways the space is occupied by the represented objects, so
they are differences in low-level properties.
Furthermore, a counterfactual consideration can show that the property
[vase] being represented in the first case and the property [faces facing
each other] represented in the other does not have much to do with the
visual phenomenal shift. That shift would occur even if the vase was not
represented as a vase but as an unfamiliar object with exactly that threedimensional shape (typical of certain vases), and the faces were
represented as two pieces of plastic with that shape (typical of human
faces). Such a counterfactual example shows that what matters for the
visual phenomenal shift is not the visual representation of certain thick
properties ([vase], [faces]) but the visual representation of certain
complexes of thin properties like being so-and-so arranged threedimensional shapes. That the representation of such three-dimensional
shapes is easily activated by the subject non-visually representing the
properties [vase] and [facing faces] is another matter, which is only
causally and indirectly relevant in determining strictly visual phenomenal
differences.
Actually, the [vase]/[faces] figure involves an analog aspect-switch to
the one involved in the Necker cube:
(from: http://mathworld.wolfram.com/NeckerCube.html
264
Chapter Six Part II
You can see the cube in two ways: as soon as you change the way of
seeing it, the back becomes the front of the cube, the front becomes the
back and the sides are inverted with respect to the property of being
occluded or being occluding (though transparent). Another global
difference in visual representation is that the cube appears as seen from
below in the one case or from above in the other: the cube appears to
change its orientation in other words.
All these variations in represented properties are variations in lowlevel properties (position of the sides, orientation of the cube, being behind
or before and so on). The phenomenal change is evident, and it does not
introspectively seem to be qualitatively different from the [vase]/[faces]
change or the other changes exemplified above. But in the Necker cube
example, the cube is seen as a cube in both cases so there is no change in
what you see the object as being. You keep seeing it as a cube, before and
after the shift. That the phenomenal shift in this case seems introspectively
similar, for example, to the phenomenal shift in the [vase]/[face] case
speaks ad abundantiam in favor of the fact that the thick properties [vase]
and [faces] do not directly enter into the visual phenomenal change. In the
Necker cube example, not only is there no shift in thick properties (there is
no recognitional shift at all), but the only recognized property is a thin one
([cube]). Therefore, the visual aspect-switch does not entail shifts in the
representation of thick properties because the relevant property recognized
in the object is thin ([cube]) or a shift in recognized properties at all
because the cube keeps being seen as a cube across the shift.
In subpart II.4, I considered the phenomenal contrast method as a
theoretical tool for establishing whether high-level properties are
represented in perception or not. I have argued that such a method is not
conclusive and is insufficient for establishing the thickness of perceptual
content. Well, nor is it sufficient for establishing the thinness of perceptual
content.
First (II.4.2), I considered Siegel’s examples of learning to read a
language and acquiring the ability to recognize pine trees by sight. I have
shown that the phenomenal difference determined by such acquisitions of
a recognitional disposition can be accounted for in conservative terms. On
the one hand, the phenomenal difference could be not strictly visual but
resting on recognitional phenomenology or/and on a feeling of familiarity.
On the other hand, even a visual phenomenal difference would not show
that a recognized thick property such as [pine tree] must be visually
represented. Instead, exerting a non-sensory representation as a
consequence of a visual episode could indirectly determine differences in
sensorily represented properties through orienting the patterns of attention,
The Phenomenal Contrast Method
265
the way of exploring the object, the way of imagining the unseen parts of
the seen object and so on.
Subsequently (II.4.3), I addressed whether the case of visual
associative agnosia supports a liberal view of phenomenal content, as
some philosophers think. My answer was negative. Again, I have shown
that even if the inability to categorize seen objects despite one’s visual
system being unimpaired can have consequences on visual phenomenal
character, this does not show that the properties corresponding to these
categories ([comb], [key], [table], [dog]) are visually represented in
normal subjects. The recognitional failure could make a difference in nonsensory phenomenal character on the one hand and in sensory phenomenal
character—for example, by indirectly influencing attentional factors and
ways of exploring—on the other.
Finally (II.4.4), I considered the phenomenon of ambiguous figures. I
have variously argued that the phenomenon of aspect-switch does not
support liberalism in phenomenal content either. Any phenomenal visual
shift involved in aspect-switch can be explained away by appeal to
changes in representation of low-level properties as a consequence of the
recognitional change; again, without the fact that the high-level recognized
properties have to enter into the visual content and character. Besides
changes in patterns of visual attention and visual imagination, the
recognitional shift involves a global shift in the three-dimensional
representation of the silhouettes in space, of their orientation and of their
position with respect to the observer: therefore, a shift at the level of
represented SCM-properties.
In sum, all of the analysis and arguments considered in this subpart,
taken together, constitute a case against the validity of the phenomenal
contrast method as a way to support liberalism. Such a skeptical
conclusion about this method does not entail that liberalism in perceptual
content has been shown to be false; it only entails that it is not proved or
supported by the usual contrast arguments.
Even if the phenomenal content of PE did not include high-level
properties, it may well be that the perceptual content does, in such a way
that perceptual content is not exhausted by phenomenal content. In the
next subpart, I will argue for the dual content view—the view that
perceptual content has a thin and a thick layer—on another basis than
introspection and the phenomenal contrast method.
PART III
THE DUAL CONTENT VIEW:
BEYOND THE PHENOMENAL CONTRAST
METHOD
III.1 The Dual Content View Principle
When you see an object, you discriminate some of its visible features,
some of those properties I have been calling SCM-properties. When you
see an object as F, you satisfy the conditions required for seeing an object
so you discriminate some visible features of that object. F can be an SMCproperty or a thick property. If F is a thick property, when you see an
object as F you must also see it as G, where G is a set of SCM-properties.
Explained in other terms, recognitional seeing, when the recognized type
is not itself a thin property, like [red], [round], [moving], [overhead],
presupposes recognition of some thin properties in the first place. In order
to see something as F, you must be sensitive to a typical visible profile
typically associated with Fs. If you recognize a human just by looking at
him/her, it is because there is a way humans typically look, where “look”
here means “exhibit certain SCM-features”: those your visual system is
sensitive to foremost.
On the one hand, there are properties the visual system is “technically”
sensitive to: colors, shapes, sizes, position, movement, distances and the
like; on the other, there are properties that the basic sensitivity of your
visual system puts you in a position to represent thanks to the association
of types of unified visible profiles with certain thicker properties, like
[human being], [table], [dog], as well as properties like [being Mario] (=
that already encountered individual), [that place] (= the place already
encountered). 1 The latter are properties attributed to objects when we exert
1 Even if it is obvious, I want to point out that by “being Mario” I mean being a
certain already encountered individual, not “being that individual who is named
The Dual Content View: Beyond the Phenomenal Contrast Method
267
toward them a classificatory or an identificatory visual capacity.
If seeing an object as an F amounts to coming-to-recognize the object
as an F by visual means, seeing-that an object is F amounts to coming to
know by visual means that a certain object is F. That propositional state is
the factive species of the corresponding genus: coming to believe by visual
means. If you come-to-recognize an object O as an F by visual means, and
in addition you are a conceptual being capable of propositional attitudes
and you possess the relevant concepts for O and F, then you will come to
believe by visual means that O is F. Provided that your recognition of O as
F is successful, so O actually is an F, then you will come to know that O is
(an) F.
Now there is a relevant asymmetry between coming to know by visual
means that O is red and coming to know by visual means that O is a table.
Indeed, in order to come to know by visual means that O is a table, you
must come to know by visual means that O has a certain shape, size, color
and so on. In coming to know that O is a table by visual means, the very
“visual means” involve your sensitivity to certain SCM-properties but not
vice versa; in order to come to know by visual means that O is rectangular,
has a certain complex shape, a certain size, a color and so on, you do not
need to come to know by visual means that O is a table. The very same
asymmetry comes between seeing O as, say, rectangular and seeing O as a
table: in order to recognize a table in O, you need to recognize O as having
a certain visible profile involving properties like [being rectangular], but
not vice versa. Indeed, you can see it as rectangular without seeing it as a
table.
Therefore, when F is a thick property, then its recognition “by visual
means” consists of its association with a recognized visible profile, call it
G. Thick recognitions are visual only insofar as they pass through thin
visual recognitions. Any visual episode represents thin properties in the
first place.
The dual content view is the view that visual perceptions possess two
layers of content, a thin, or low-level, one, and a thick, or high-level, one,
where the high-level content is represented thanks to representations of the
low-level content but not vice versa. According to the dual content view,
seeing O as an F can be a genuinely and literally perceptual episode, even
when F is not an SCM-property. So, thick properties can be represented in
Mario.” That second property (being named a certain way) is not among those
properties that can be recognized by visual means. No visible features track the
way someone is named but some visible features can track the property of being a
certain individual as long as these features are possessed only by that individual, at
least, within the usual environment of the perceiver.
268
Chapter Six Part III
perception, and liberalism of perceptual content is true. In what follows, I
will argue for the dual content view of visual perception.
I have argued that the phenomenal contrast method is unable to prove
the truth of the dual content view: introspection of phenomenal character
cannot tell whether thick properties are represented in visual experience. It
can tell that two experiences phenomenally differ but it could be unreliable
in telling how. Anyway, an assessed phenomenal visual difference between
experiences E1 and E2—one with and the other without a certain thick
recognition—does not tell us that the recognized property itself is
represented in visual experience. How are we to decide then whether a
given recognitional content is possessed by the very perceptual state or
whether it is non-perceptual content represented by virtue of another
perceptual content being represented (and eventually influencing that
perceptual content itself)? Is that a really substantive question or just a
terminological matter concerning what we stipulate as “perceptual”? I
think it is a substantive question and it can be answered by finding a
positive criterion.
Our visual apparatus is basically and innately sensitive to certain thin
properties. Indeed, the basic visual representations of SCM are modular,
cognitively impenetrable, so they are fixed. The relevant mechanisms and
subsystems are phylogenetically set up to be set off by certain properties
(color, movement, shapes and so on). The way they are is the way they
have evolved, a specific way for each different species whose members are
equipped with them.
But our perceptual systems—as well as those of all other beings able to
learn—are also set up to be altered by our individual experiences. Such
phylogenetically fixed mechanisms are supposed to enable the
identification, classification and recognition of objects and properties
already encountered in our environment so that we can use past
experiences to better represent the present environment, to be prepared for
future encounters, to bring about desired states of affairs.
Perceptual learning consists of acquiring such identificational and
classificatory dispositions through exploiting our basic innate sensitivity to
thin properties. Therefore, visual recognitional episodes also have
ontogenetically determined contents, which depend on individual learning.
I can visually recognize my mother only because I have encountered her
before. I can recognize dogs and tables now because I have encountered
many tokens of them in my life. Recognition involves memory and
therefore it “goes beyond” the present stimulus and exploits information
stored in the past. It does not make the episode less perceptual but rather
enables it to have thick content. Perceptual learning is a way of expanding
The Dual Content View: Beyond the Phenomenal Contrast Method
269
the range of perceivable properties and the very contents of our PEs.
Such a view can be expressed by the following principle, which I will
call the dual content view principle:
DCVP: When there is a unified set of visually accessible properties (SCM) that
reliably track other deeper properties (F), and that tracking itself has been
embedded into a recognitional category through perceptual learning, then a
visual perception PE of (SCM) has the thick content F besides the thin content
(SCM). F comes to be part of the set of accuracy-conditions for that PE; that is,
of the PE’s semantic content.
DCVP needs to be integrated with a clause in order to stop the
extension of perceptually representable properties at some point. We do
not want the property of [being on holiday] to be represented in a PE
consisting of my seeing my parents’ house with closed windows, even if
that property can be truly represented as a property of my parents by visual
means or by virtue of my visual PE. Actually, I come to know by visual
means that my parents are on holiday, where the “visual means” include
my seeing their house with closed windows. I can also recognize the last
president of the United States as such by visual means so I see him as the
last president of the United States, but we do not want our view to allow
the property [being the last president of the United State] to be included in
the content of a visual perception.
Thus, there are now two theoretical tasks to be fulfilled. First, it has to
be better shown why at least certain thick recognitional contents
associated with visual episodes—recognizing an F by visual means—must
be considered bona fide perceptual contents. If the phenomenal contrast
method is insufficient, why not hold that all perceptual content is thin?
Which other method can be of help?
Second, a criterion must be found by which we can demarcate
perceptual thick contents and non-perceptual thick contents; for example,
[being a dog] or [being a table] on the one hand, and [being the last
president of the United States] or [being on holiday] on the other. We
cannot just rely on intuition to ascribe a certain content to a visual episode
and another to a belief or a recognitional state that would only be
(co)caused by a visual episode.
In the following subpart (III.2), I shall undertake the first task; then in
III.3, I shall turn to the second.
270
Chapter Six Part III
III.2 Preparatory vs. Constitutive Cognitive Processes
Perceptual learning involves the acquisition of new perceptual schemes,2
such that an encounter with a percept—once the perceptual scheme is
acquired and stabilized—immediately and non-inferentially triggers the
recognition of it as a certain type of thing. Perceptually acquiring a
recognitional ability presupposes the ability to perceive certain SCMproperties as composing a unified sensible profile, which can be identified
on different occasions of presentation: any perceptual categorization that
occurs by visual means occurs by exploiting visible cues.
Perceptual learning can also concern abstract SCM-properties, like
[triangle]. For example, a rat can be trained to jump at panels with
varieties of triangles painted on them.3 After the training, we could say that
the rat sees the figures as triangles insofar as it perceptually groups them
into the same type and waits for the reward as soon as it jumps to tokens
of that type. A monkey can be trained to choose the larger of two
rectangles so that it can come-to-recognize the [larger-than]-relation
independently of the absolute sizes of the relata. After perceptual training
and learning, the monkey immediately categorizes the [larger-than]
relation and acts accordingly to obtain the desired reward.4 Once an animal
becomes able to visually recognize these properties ([larger-than],
[triangle]), its respective PEs are enriched with a new PPC. Such
acquisitions of PPC by the type of experiences caused by certain objects
are possible because certain properties are represented at the level of SC,
from which the new contents are somehow “extracted” through learning.
New categorizations can expand the PPC of a given PE with a certain SC.
The latter, however, is fixed and depends on innate mechanisms. That is
not to say that there are no innate categorizations and no seeing-as
acquired; the point is rather that perceptual categorizations can be acquired
and often are, whereas basic SCM-representations corresponding to SC
draw on wired-in and philogenetically fixed capacities.
So, visually thin properties can be represented at the level of both SC
(colors, position in space, distance) and PPC ([larger-than], [symmetricalwith], [triangle] and so on). Two PEs may share their SCs but differ in
their thin PPCs. Now, the dual content view is not just the view that PEs
have an SC and a PPC. A conservative view of perceptual content can also
hold that PE has these two layers of content. Peacocke, who was the first
2 See Neisser 1967, Piaget 1937.
3 I took the example from Sellars 1981.
4 Such experiments by Kluver are cited in Gibson 1969, 283ff., and also by
Dretske 2000a, 131ff.
The Dual Content View: Beyond the Phenomenal Contrast Method
271
to elaborate the distinction between SC and PPC, is a conservative
himself; that is, he holds that only SCM-properties can be represented in
visual perception.5 Properties and relations like [symmetrical with],
[equidistant from], [square], [regular diamond] and the like are still SCMproperties that may or may not be “extracted” by the SC. When extracted,
they enter into the PPC of the PE; for example, a square is not just visually
represented, it is also represented as a square—it is seen or recognized as a
square. The dual content view is not just the idea that PEs have an SC and
a PPC but it is the liberal idea that the PPC of PE includes not only thin
properties like [square] and [symmetrical with] but also thick properties
like [dog], [table] and [prey]. For a conservative, in contrast, both SC and
PPC are made out of thin properties.
Thus, it is important not to confuse the thesis that perceptual content is
two-layered (SC and PPC) with the thesis that perceptual content is dual
(thin and thick).6 I have already argued for perceptual content being twolayered (Chapter Three, Part I), and now I am arguing for the liberal view;
that is, for the dual content view. The dual content view entails that when I
see something as an F, and F is not an SCM-property, it may well belong
to the very perceptual content of that seeing-as-episode, a content a certain
type of PE can acquire through perceptual learning.
A natural objection to the dual content view is that the visual content is
the one outputted by the visual modules, and only then are such visual
representations fed into other cognitive systems like memory, reasoning
and so forth. Therefore, any thick representation that presupposes a
cognitive mediation beyond the stimulus-driven information outputted by
the visual modules is not genuine visual content but non-sensory content
represented by virtue of another perceptual content being represented, but
not belonging to the latter.
To tackle that objection, I want first to spell out a distinction concerning
the notion of “cognitive mediation.” It is true that perceptual recognition
5 See Peacocke 1992.
6 I want to avoid another possible confusion in the reader: the dual content view
has nothing to do with what I have called the double content view (Chapter Five,
Part II), according to which PE has both a Russellian and a Fregean content. I have
argued against the double content view whereas I am arguing for the dual content
view: that means that, in my opinion, the properties represented in PE are
Russellian properties (contrary to the double content view), and these Russellian
properties included in perceptual content are both low-level, or thin, and highlevel, or thick (dual content view). Perceptual content is “two-layered” (scenario
content and proto-propositional content) and it is “dual” (thin and thick) but it is
not “double” (Fregean and Russellian).
272
Chapter Six Part III
involves the contribution of memory, and therefore of stored representations
the incoming stimulus is matched with, so there is a “cognitive
integration” of the incoming stimulus that fills the gap between the poor
proximal stimulus and the rich represented content. That is why visual
recognitions can fail despite the fact that the SCM-properties are
successfully represented at the level of SC, when you see O as an F but it
is not an F even if the object has just the colors, shape, size and position
your PE represents it as having.
Nonetheless, we need to distinguish between two kinds of cognitive
integration that could be involved in a perceptual episode: a preparatory
and a constitutive integration. To freely use the working-characterization
by Ben-Zeev, “preparatory” means “making ready beforehand for
something following” whilst “constitutive” means “making something
what it is” (1988).7 A preparatory process precedes and is separate from
the cognitive product it is preparatory of whereas a constitutive process is
intrinsic to the state it gives rise to; it is part of what makes the produced
state or episode what it is. Maybe the boundaries between preparatory
processes and constitutive processes are blurry and vague but that does not
undermine the theoretical usefulness of the distinction.
Let us test that difference by means of some examples. A paradigmatic
case of a cognitive process being preparatory with respect to a resulting
cognitive state is the case of an inferential process with respect to a belief
representing its conclusion. Sometimes I come to know that P by visual
means, but the visual experience involved, together with other collateral
information, is only preparatory with respect to the resulting cognitive
product consisting, for example, of a certain belief. I see the newspaper, I
know how to read the English language, I have reasons for trusting the
journalist, so I come to know that the government is in trouble (be it P). I
come to know that P by visual means but the visual experience is
preparatory not constitutive of the mental state I come to be in when
believing-that-P. That is why the government is not part of the content of
my visual perception, even though it is among the means by which I come
to know about the government. Such a belief is the result of a visual
experience with a proprietary content (say, certain written marks are on the
page and so on) that I need to semantically interpret on the one hand, and
of other beliefs about the reliability of the source, about the UKparliament and constitution and the like on the other. So I infer that the
government is in trouble starting from certain premises, most of them
7 Ben Zeev 1988 adopts this distinction to criticize the idea that “non-pure
perceptions”—that is, contaminated with memory, expectations and the like—are
indirect.
The Dual Content View: Beyond the Phenomenal Contrast Method
273
implicit because they are obvious, plus a genuinely perceptual belief about
what I am seeing and reading in the newspaper in front of me. Likewise,
when I come to know that my parents are on holiday by seeing their house
with closed windows, that final state has my perceptual experience and
other beliefs as constituents of a preparatory cognitive mediation not as its
constitutive processes. There is perceptually acquired information but it is
inferentially combined with previous background beliefs and used to form
a belief, which is another further, different state.
One could also say that the case of perceptual recognition involves the
influence of memory and collateral, previous knowledge (or information,
at least), and therefore the recognition is not to be characterized as
perceptual. But once a perceptual scheme is stabilized after learning, the
cognitive process of matching the percept with a representation “stored” in
memory has become constitutive of the perceptual episode itself; it is not
merely preparatory as it was before learning and perhaps during the
learning phase. When I recognize a table just by looking at it, I do not
have to first represent a certain object then consult my memory as if it was
a static and separate database and finally match the content of my visual
experience with previously stored content. Rather, through previous
learning, a new perceptual scheme has been embedded into my very
perceptual system—a system including both innate mechanisms and
acquired mechanisms—so now I visually categorize a table immediately
and un-mediately; that is, without need of genuine cognitive mediation.8
That is why my very perception of a table has the respective thick
classification as a constitutive content. Likewise, once I know Mario and I
am able to recognize him just by looking at him, the respective visual
identification has the identity of Mario as its constitutive content even
though the actual visual stimulus is “enriched” by the identificational
content by means of memory, so by a virtual connection to past
experiences. Such a connection is not inferential or real cognitive
mediation of any other sort.
Still, it could be argued that it is a process anyway. Again, the objection
would rest on the ambiguity of the very notion of process. Any cognitive
achievement of any sort, including the many operations of the early visual
8 There is a lot of evidence for high plasticity of the visual stream: the encounter
with a novel object can make a certain group of cells become responsive to that
object in the future. That evidence is overwhelming for face recognition, but not
only for that (see Gross and Sergent 1992). In that way, a perceptual system
acquires new stable functions by learning. Learning can be considered a sort of
individual evolution, which exploits the basic phylogenetic functions established at
the level of specie-evolution.
274
Chapter Six Part III
system, are processes; i.e., causal chains made out of respective rings, but
a given process should count as cognitive mediation when the starting
point of that process is preparatory for, separated from and independent of
the final point. Mediation should not count as cognitive when the
processes that bring the resulting state are constitutive of that very state.
Not every ring of the causal chain in a cognitive process is also a cognitive
step of it, so to speak.
I have spoken of a representation stored in memory and matched with
the percept on a given occasion but such a façon de parler must not be
taken as if perceptual memory were a database that needs to be consulted
after the environmental information is taken in. Perceptual learning is not
the literal storage of information in a box; rather, it is a structural change
of the immediate response before a certain type of stimulus, a change
embedded in the perceptual system at a certain point of a subject’s learning
history.9 After such a structural change, the presence of the stimulus just
triggers a different perceptual response with different content. Now the
object is seen as an F without any cognitive mediation or extrinsic
preparatory process: rather, the PE now possesses a new constitutive
content.10 As Ben-Zeev notes, there is no sequential transition of separated
and juxtaposed cognitive acts in perceptual recognition but one unitary
one.
Perceiving and the exercise of a cognitive schema […] take place together
within the same experience. There is no perceptual raw material waiting to be
organized by perceptual schemas; the very first perceptual awareness is
organized. Perceptual recognition is not a process behind the ongoing state of
9 As Prinz 2006, 436 remarks when talking of the matching involved in perceptual
recognition “Matching does not necessarily involve comparing two token
representations […] matching can also occur when an input triggers a stored
representation.”
10 Granit 1982 (quoted by Ben-Zeev 1988, 320) did experiments on training dogs
and reports interesting information about the neurophysiological difference before
and after training them to handle specific problems: “Before the dog has learned
the problem facing him, the electrical activity accompanying the effort is spread
over large portions of the brain. In the fully trained animal, the active site has
shrunk to a small focus responding to the challenge. The learned behavior is then
fully automated and the voluntary effort is restricted to a trigger function.” That is
to say, the acquisition of new perceptual schemes of categorization make
immediate and effortless what was object of a positive effort, so the response is
now automatically triggered, it is embedded into the very perceptual episode, it is
no more a further cognitive activity. A preparatory “interpretation” of the
represented has become a constitutive way of representing it in perception.
The Dual Content View: Beyond the Phenomenal Contrast Method
275
perceiving. Perceiving is a type of direct understanding which does not take
place over time; it is happening at any point in time. (1988, 322)
Insofar as there is perceptual categorization, there is thick perceptual
content; that is, a high-level content constitutive of the very perception.
The perceptual episode is not merely preparatory of the representation of
that content at this point, it involves it. We can visually perceive O to be F
even if F is a high-level property; we do not just take O to be F thanks to
visual perception. Our sensitivity to SCM-properties is put to work
through perceptual learning so we acquire visual sensitivity to new types
by associating these types with certain types of visual profiles. This is the
spirit of the dual content view principle proposed above. As soon as the
thick property belongs to the perceptual content through the subject’s
learning a perceptual scheme, that property enters into the accuracyconditions of the PE. For example, failing to recognize a table as a table in
optimal conditions would be a perceptual mistake not just a cognitive
mistake of another kind connected to a purely thin accurate perception.11
Rather, my perception of the table may be accurate with respect to the thin
properties it represents the object as having, but inaccurate with respect to
the thick property ([being a table]) it fails to represent the object as having.
Our perceptual mechanisms are phylogenetically set up to be set off by
certain properties (SCM-properties for vision), but perceptual learning—
whose principles are also wired-in—can shape these very mechanisms to
implement more specific representational functions. For example, a certain
recognitional mechanism is set up to be set off by dogs. I can perceive a
thing as a dog as long as the representation of a certain visual profile unmediately—not just in a preparatory way—activates the representation of
that type [dog].
Before or during the learning phase—that is, the process of embedding
and stabilizing a perceptual schema—it would not have been a
misperception for my PE to represent the thin properties without the thick
one. After learning, however, it would be a misperception as my PE’s
accuracy-conditions have changed: as a consequence, what represented
content is perceptual and what is non-perceptual depends on the learning
history of the individual as perceptual schemas are actually embedded into
the individual’s cognitive architecture.
11 That involves visual associative agnosia being a perceptual impairment not just
a cognitive one. If VAA makes the patient lose the previously acquired ability to
categorize ordinary objects, they lose a perceptual ability because the acquired
recognitional abilities were embedded into their system of perceptual
responsiveness.
276
Chapter Six Part III
Likewise, which beliefs are perceptual and which are not does not
merely depend on the kind of content the belief possesses or on the way
the belief is acquired (e.g., by visual means); it also depends on the
learning history of the believing subject. Such a history determines what
one is able to visually recognize without cognitive preparatory mediations;
consequently, it determines which correspondent belief, formed by visual
means, is perceptual or not. If I needed to make inferences or reasoning to
exert a certain categorization, I could, un-mediately and just by looking.
Perceptual content is dual; it is both thin and thick: which properties enter
into the thin content depends on the species-specific equipment so it is a
phylogenetic and evolutionary matter. The question of what properties can
enter into the thick content is open and depends on individual learning
history so it is an ontogenetic matter.12
III.3 The (Blurry) Superior Boundary of Perceptual
Content
The DCVP says that those visually recognized thick properties, which are
tracked by recognized thin visible profiles, are genuine visual contents as
long as that very tracking relation has been structurally embedded into a
recognitional schema through perceptual learning.
Now I can learn to recognize many types of things just by looking: I
can see that you are saying hello to me just by looking at your moving
hand, I can see that the postman has not arrived yet just by looking at the
empty mailbox. Not only can I recognize Mario just by looking, because
he just looks like Mario, I can also come to believe just by looking that
someone must have stolen my car because it looks like someone must
have stolen my car, given that the garage is empty and the door open. I
could learn to recognize just by looking that something is a mahogany
chest of the XIX century made in the south of France, that someone is the
last president of the United States, that someone is my great-uncle, that
this bird is an extremely rare variety of African goldfinch and so forth.
12 Sober 1984, 147ff., distinguishes a developmental explanation from a
selectional explanation of behavior. A selectional explanation addresses the
question of why organisms of a certain species behave a certain way;
developmental explanations (appropriate only for learning systems) address the
question of why the individual organism behaves a certain way. According to the
general spirit of that distinction, given that mental contents are ascribed to explain
behavior, ascriptions of thin perceptual contents are part of selectional
explanations and ascriptions of thick perceptual contents are mostly part of
developmental explanations.
The Dual Content View: Beyond the Phenomenal Contrast Method
277
Does it mean that, if the DCVP is true, all these sophisticated contents
must be able to be perceptually represented? In fact, once the perceptual
recognition of something as [a certain rare variety of African goldfinch]
has been learned and the respective cognitive response has been made unmediated, it seems that such a sophisticated property should be ascribed as
perceptual content. The learning process has brought about a structured
association between a given fine-grained type of visible profile and a finegrained property so, according to DCVP, it has enriched that type of PE
with such content. Is there something wrong with DCVP then?
To begin with, let us distinguish a proposition P we come to know by
visual means from an object O we come-to-recognize as an F by visual
means. Let us start with visually acquired beliefs.
First, there are different ways a proposition can come to be known by
visual means (see Chapter One, Part III.4-5). I literally and directly see
that P when I come to know P through being visually aware of the
constituents of the fact that makes proposition P true. If I literally and
directly see that the cat is on the mat, I must become visually aware of the
cat, of the mat and of the [on-ness] relation in the first place. I come to
know that fact just by looking because I see and categorize the
constituents (the cat as a cat, the mat as a mat and so on) and conceptualize
and combine them, so to speak, into a structured propositional attitude.
When I come to know that the postman has not arrived yet just by
looking at the empty mailbox, when I come to know that someone must
have stolen my car just by seeing my empty garage and the door open,
when I come to know that my parents are on holiday just by looking at
their house with closed windows, actually I am not visually aware of the
constituents of these facts: I do not see either the postman or his arriving
to my home, I do not see either my car or anyone stealing it, I do not see
my parents or their wonderful holiday destination. Rather, I only see the
empty mailbox, the empty garage, the house with closed windows. In
these cases then, there is no doubt whether such unseen constituents of the
known facts are contents of a perceptual episode. They are not. Visual
contents are objects, properties and relations I become visually aware of
just by looking, not the constituents of facts I infer from what I see. It does
not matter that certain inferential transitions are quick and obvious so they
do not involve my conscious, explicit focus on each of their steps. They
are inferences because the information I have available is about something
I am in perceptual contact with, whilst the information I come to “extract”
is about something I am not in perceptual contact with so there is a
cognitive, inferential mediation. This is why I would not be perceptually
wrong in case the postman has arrived but surprisingly he did not have any
278
Chapter Six Part III
letters for me; my brother took my car for a sudden emergency and left the
door open; or my parents are at home with an unusual headache so they
have chosen to keep their windows closed. I would not be wrong in having
certain perceptual beliefs, rather in having drawn certain inferences from
certain true perceptual beliefs. A perceptual belief is based on the
perception of the constituents of the believed fact.13 Therefore, even if
certain cognitive transitions from the perceptual knowledge of a fact to
beliefs about other facts becomes natural and effortless through the
formation of inferential habits, this does not entail that the latter believed
contents are perceptual contents thereby. The DCVP does not entail it
because the postman, the car thief and my parents are not visually
recognized thanks to their typical visible profile; they are not visually
recognized at all so in our examples they simply cannot be perceptual
contents I am aware of.
I have just distinguished genuine perceptual beliefs from other
remotely perception-based beliefs. A perceptual belief rests on the
perception of the constituents of the fact that makes the believed
proposition true. In order to be combined into a proposition suitable to be
believed, such constituents are conceptualized. I have argued at length that
a more basic operation—perceptual recognition—is presupposed to the
conceptualization involved in perceptual belief. So, the content of a
perceptual belief is conceptual but the content of the perceptual episode
the perceptual belief is based upon is not. The non-conceptual content of
the perceptual episode is two-layered: it involves an SC and a PPC that is
categorical-recognitional. Thus, the constituents of the perceptual belief
are perceptually categorized before being conceptualized. So, in seeingthat a is F, a is to be seen as F in the first place.
Now, when seeing-as is genuinely perceptual, F belongs to perceptual
content; that content on which the (conceptual) content of the belief that a
is F is based. Seeing-as is perceptual when a certain property is visually
recognized in a seen object thanks to the recognition of a typical visible
profile that F-tokens exhibit, and such an association of the typical profile
with the property has become a stabilized perceptual scheme. That is
DCVP indeed.
13 A definition compatible with the idea, expressed above, that which belief is
perceptual depends on the learning history of the perceiver. A perceptual belief
that P is based on the visual awareness of the constituents of the fact that makes P
true. Such constituents are objects, properties and relations. Which properties can
be perceptually represented depends on the learning history of the perceiver so
which beliefs are perceptual depends on the learning history of the perceiver too: a
perceiver can expand the range of properties it can perceptually represent.
The Dual Content View: Beyond the Phenomenal Contrast Method
279
The problem we are left with now is that DCVP seems to entail that
contents like [being a mahogany chest of the XIX century made in the
south of France], [being the last president of the United States], [being my
great-uncle], [being an extremely rare variety of African goldfinch] can be
perceptual. In these cases, the distinction proposed above between
“literally” seen facts and facts inferred from seen facts is not sufficient to
rule these contents out of bona fide perceptual contents. In fact, in these
latter cases, I do see the object I come to take as an F by visual means; I
am visually aware of the F, for example, the mahogany chest of the XIX
century made in the south of France, of the last president of United States,
of my great-uncle, of the bird belonging to an extremely rare variety of
African goldfinch. So, if I develop a visual recognitional ability for the
respective properties, why should they not be perceptual contents, and so
properties represented in my visual perception?
Some philosophers, like Prinz, go just as far as holding that such thick
sophisticated properties can very well be perceptual contents.14 Now I
shall briefly consider that view—which could be called a boundless
liberalism on perceptual content—and argue against it. We must go liberal
but only moderately so.
According to Prinz’s radical view, we can perceive anything we can
conceive because the content of a perceptual episode may well depend on
the meaning we transiently assign to it: each time we use visual
appearances to confirm that an abstract property is instantiated, our
perceptual representation of that appearance gets the supplementary
meaning it is used as an indicator of. For example, I want to know which
of my students is a philosophy major so I ask them to raise a hand in case
they are. Seeing a raised hand in that context amounts to perceiving the
property of [being a philosophy major] because such a visual state has
been given that transient meaning by me.15
Without entering into the details of the eccentric Prinz’s view, I want to
make a general point. Using visual appearances to confirm whether an
abstract property is instantiated is a way of using our perceptual
representations to go beyond perceived contents. What makes the
perception of a raised hand the representation of [being a philosophy
major] is not just the visual perception of the raised hand nor the
perceptually learned and stabilized recognitional scheme; it is the
14 Prinz 2006 argues that we can perceive anything we can conceive, and therefore
any abstract property whatsoever can enter into perceptual content: not only a finegrained property like [being my great-uncle] but also a property like [injustice],
which no one would be prepared to take as perceivable.
15 The example is provided by Prinz himself, 2006, 446.
280
Chapter Six Part III
contextual, contingent use of visual information that presupposes a large
number of sophisticated beliefs as well as the evidential relations between
them and the mastery of certain social conventions, linguistic and nonlinguistic symbolic language and so on. This is a case in which the visual
perception is only preparatory with respect to the obtained belief that there
is a philosophy major since the recognition of a philosophy major in the
student who raised her hand is not constitutive of the visual perception.
Suppose the student has not heard my question properly, or she is just
pretending to be a philosophy major: there is no reason to think that my
perceptual state is mistaken but if it is not, then the representation of the
student as being a philosophy major is not part of the accuracy-conditions
of my PE. The accuracy-conditions of a PE turn it into a misperception
when they are not satisfied. What would be wrong would be the belief
assigning that transient meaning to my perception rather than the
perception itself: the latter did not have that meaning so it was not
unreliable as such. The thick content was beyond the content of the PE; it
was not content constitutive of the PE.
Cases where you make certain preparatory inferences that make you
conclude that if something is the case (a is F), then the event E happens or
the property G is instantiated and E or G are visually available are not
cases where you perceive a as an F; they are just cases where you perceive
a as a G (or the event E consisting of a’s being G), and use that perceptual
representation as evidence for another proposition: [a is F]. I can have
reasons to think that if the candidate shows up at a certain congress, she
will lose the elections. I see the candidate at the congress so I use that
visual appearance—together with a lot of other inferentially obtained
information—to confirm that the abstract property of [not going to win the
next elections] belongs to her, so I take a to be F by means of seeing a as
being G. Perceiving the candidate as being at the congress does not entail
perceiving the candidate as being about to lose the election just because a
set of bridging inferences and connecting beliefs has made me think that a
property being instantiated reliably indicates the other property being
instantiated.
The inferential bridge can also become a sort of habit but this does not
prevent the newly bestowed thick content from being connected to the
visually perceived property only inferentially and, therefore, doxastically.
The “automation,” so to speak, of certain connecting beliefs does not
entail that the visually obtained knowledge does not also rest on such
beliefs and their evidential relations mediating from the perceptual
experience to the obtained “thick” knowledge.
The Dual Content View: Beyond the Phenomenal Contrast Method
281
Let us take an example of what Dretske calls “displaced perception.”16
A displaced perception is exactly a situation where I come to know a fact I
do not perceive by virtue of coming-to-know another fact whose
constituents I am visually aware of, so by virtue of “directly” coming to
know another fact by visual means. For example, I come to know by
visual means that I weigh sixty kilos by looking at the pointer of my scale.
I do not perceive my weighting sixty kilograms; rather I perceive the
display of my scale and its pointer. The fact that the pointer indicates a
certain value, a fact which is perceptually known by me, enables me to
come to know another fact by exploiting connecting beliefs about the
relation between the weight-pointer and the weight of things put on the
scale. Among such connecting beliefs—which enable me to construct an
inferential bridge from a perceptually known fact to another more abstract
one—is the belief that the pointer is calibrated in certain ways so to
reliably co-vary with the weight of things put on it, that for each kilo of
the load, the pointer will move in a certain direction and so on. Thanks to
many connecting beliefs, I am able to use a visual perception to know
truths about things I may not be seeing (I see my body but it would be not
necessary; I certainly do not see its property of weighing sixty kilos
anyway). Once I become confident with weights, displays and pointers,
there is no more need to engage any explicit inferential process from what
I see to what I come to believe on that basis: I just take a look at the
pointer and realize that I have gained two kilos. Now, I can realize what
my weight is just by looking but my visual perception, again, is only
preparatory to my belief that I weigh sixty kilos. The content is not
constitutive of my visual perception, it is a belief resulting from my visual
perception combined with a lot of other connecting beliefs for as quickly
as the transition from the perception to this resulting belief can be. This is
why, if the scales are broken, there is error in the resulting belief but there
is not visual misperception as a result. The weight of my body is not a
content of my visual PE; otherwise broken scales would ipso facto make
my PE a visual misperception.
Therefore, DCVP does not commit us to accepting that the content of
such displaced perceptions is a perceptual content. The limiting clause I
am trying to make explicit through the examples above is actually already
present in the very notion of “perceptual learning” contained in DCVP.
DCVP says that thick contents stably associated with typical visible
profiles by perceptual learning become contents constitutive of the
respective perceptions of such visible profiles. An inferential, sophisticated
16 See Dretske 1995, Chapter 2.
282
Chapter Six Part III
preparatory process is not really an example of perceptual learning. On the
contrary, recognitional categories acquired by perceptual learning do not
depend on evidential relations with other collateral beliefs; they do not
depend for their exercise on previous inferential transitions. Inferential
reasoning can very well be combined with perceptual learning. For
example, I learn that a certain pattern of movement, the raising of hands, is
a request for speaking in certain contexts. Now, this is a conventional code
that requires, in order to be understood, many beliefs, inferences and
sophisticated concepts. Now I get, just by looking, that someone is asking
to join the discussion, but what I have perceptually learned is that this
typical way of raising hands is associated with something. That this
“something” is a conventional way of asking to speak is the content of a
connecting belief that works as an epistemic bridge between what I
perceptually recognize and the final belief that someone would like to
speak. To repeat myself for the sake of clarity, the fact that I just by
looking take a certain gesture to be a certain request does not prevent the
connecting beliefs from being essential to that recognition. So, the
recognition is not purely perceptual insofar as it rests on the many
evidential relations the final recognition entertains with other connecting
beliefs, besides the evidential relation to the visually perceived objects and
properties. DCVP is safe, but better with a limiting clause specified into it
in order to rule out properties that are too abstract—whose representation
is essentially concept-involving and inference-involving—as candidates
for being part of perceptual content
Dual content view principle (integrated):
When there is a unified set of visually accessible properties (SCM) that
reliably track other thicker properties (F), and the tracking itself has been
embedded into a recognitional category through perceptual learning without
preparatory inferences and connecting beliefs, then a visual perception PE of
(SCM) has the thick content F besides the thin content (SCM). F comes to be
part of the set of accuracy-conditions for the PE; that is, of the PE’s semantic
content.
The introduction of such a clause puts us in a position to treat the other
examples of implausible perceptual contents—fine-grained properties
visually detectable thanks to a typical visible profile associated with their
instantiation. Let us assume that a thick property like [table] can be
represented in perception. Now, on which non-arbitrary basis would it be
not the case, for example, that we perceptually represent the property of
[being a French cherry table in Imperium style inlaid with mother-of-pearl
decorations]? I become an expert antiquarian and refine my recognitional
The Dual Content View: Beyond the Phenomenal Contrast Method
283
dispositions to such a point that I am able to recognize that complex and
sophisticated property by sight. As an expert naturalist, I can become so
good at birdwatching that I can immediately spot the property of being an
extremely rare variety of African goldfinch. These are parallel cases to that
of the pine-tree spotter used by Siegel, in fact (Part I).
In order to address these cases, let us ask ourselves how we have come
to acquire such a fine-grained recognitional disposition. For sure our
visual sensitivity to certain features as type-markers is of the greatest
importance so there is perceptual learning involved. But how have we
associated these visible markers with what they mark now? For example,
how have I learned to associate certain SCM-features of the bird to the
property of [being an extremely rare variety of African goldfinch]? I must
have had abstract ornithological knowledge, I must already have concepts
of natural kinds, beliefs about how certain kinds are spread and distributed
in different areas of the planet, I need to know that there are African and
non-African goldfinches, I need to know that African goldfinches are
goldfinches, that goldfinches are birds and so forth. There is a huge
number of inferences I must be sensitive to if I am to recognize such a
complex property at all (perceptually or not). So, the perceptual learning
component in such a recognitional ability is only a component amongst
other collateral knowledge, including inferences and high-level scientific
concepts. I have perceptually learned to recognize a certain vague, rough
and indeterminate class of things, the class of those things that share
certain visible markers or a specific Gestalt-shape. But that such a Gestaltshape tracks the property of [being an extremely rare variety of African
goldfinch] is not something that can be embedded into the perceptual
scheme as such. There is not just perceptual learning involved in such a
recognitional category, there is much more than that, therefore such thick
content is not represented in perception; rather, it is contextually
represented also, thanks to a perceptual episode. A test for certain content
being perceptually representable, provided that perceptual content is nonconceptual (see Chapter Two, Part II), is the possibility of entertaining the
respective content without having the concepts canonically used to
characterize the content. Now, entertaining the content [an extremely rare
variety of African goldfinch] is not possible without possessing the
respective concepts; for example, the concepts of [being African], [being
rare] and so on. A non-conceptual being could well be perceptually
sensitive to a certain variety of African goldfinches. Suppose that such a
variety is poisonous and has certain visible markers, so a certain species
develops a sensitivity to these markers, avoids these birds and maximizes
survival. In this case, the perceptual sensitivity to that variety of African
284
Chapter Six Part III
goldfinches would not amount to the ability to perceptually represent the
property of [being an extremely rare variety of African goldfinch = F] but
to another rough and non-conceptual property that happens to track the
sophisticated property F.
Let us consider the example by Crane (1992), already cited, of a
scientist who sees something as a cathode ray tube. What are the
conditions for seeing O as an F for F = [being a cathode ray tube]?
A child who happens to look at the cathode ray tube will see it as
having certain SCM-properties, a certain visual appearance, certain parts
with certain colors, shapes, solidity, sizes; she will not see it as having the
function it has been designed for because she probably does not possess
the concepts of (ray), (cathode) or others involved in understanding what a
cathode ray tube is for. In fact, a whole piece of scientific theory is needed
to master these concepts. As I have argued (Chapter One, Part IV.3;
Chapter Two, Part I.1.6), to possess a concept means to grasp its
inferential role at least in a subset of relevant inferences it can enter into
by being a constituent of propositions these inferences are made of. So,
concept-possession in general involves inferential abilities in general;
furthermore, possession of a certain concept involves sensitivity to certain
respective inferential patterns. A child can well see a cathode ray tube and
even recognize it as something, an F or a G—of course, not as a cathode
ray tube—even if she has no conceptual abilities at all yet. It suffices for
her to have recognitional abilities, namely, the capacity to exploit her
basic, wired-in sensitivity to visual properties (SCM) to recognize visual
profiles and type them some way; that is, the capacity of perceptual
learning. A recognitional category that can enter into the content of a
perceptual episode must have been acquired through perceptual learning; it
must be embedded into what stably shapes the responsiveness of the
perceptual system. The case of recognizing something as a cathode ray
tube does not fall into perceptual categorization because it presupposes
inferential and doxastic learning over and above perceptual learning. Here
a certain visual experience with a proprietary representational content is
conducive to thicker representations by virtue of the evidential relations
between the visual appearances of the seen object and many other beliefs
and inferences.
So, first, categorizing something as a cathode ray tube entails
conceptualizing something as a cathode ray tube. This indicates that such a
categorization is not merely perceptual because perceptual content— be it
thin or thick—is non-conceptual. Second, the visual experience of the
cathode ray tube is preparatory with respect to the recognition of it as a
cathode ray tube because the PE contributes to this thick recognition
The Dual Content View: Beyond the Phenomenal Contrast Method
285
together with a complex set of beliefs that constitute a complex inferential
bridge from the visual experience to the recognition of the seen object as a
cathode ray tube.
The cases of the cathode ray tube, the rare species of African goldfinch
and Siegel’s pine tree are to be distinguished from cases like Prinz’s one of
recognizing a philosophy major by visual means (by seeing the raised
hand). The first are cases where the property F, as sophisticated as it is,
possesses a typical visible profile that contains marker-features reliably
indicating the property F being instantiated: pine trees, as well as African
goldfinches and cathode ray tubes (as far as I know) do possess a
distinctive visible appearance that reliably tracks the respective thick
property. So, visual sensitivity to that property can be learned in a stable
and recursive way. On the contrary, there is no visible property that can
reliably track the property of being a philosophy major (besides the
contingent particular context of the example): this case is to be
distinguished from the first cases because the visual perception of the
raised hand is used to get information that is not reliably connected to
raising hands as such, but perhaps to that event together with a previous
question having a certain semantic content, with beliefs about certain
conventions and so on and so forth.
Let us sum up: the property of being a philosophy major is not
perceivable at all, even if one can come to know that someone is a
philosophy major by giving a transient meaning to a certain visual episode
in a particular context. This meaning, indeed, is not constitutive of the
visual episode as such; rather, it rests on collateral evidential relations
between the visual episode in a context and other beliefs relative to that
context. However, properties like being a rare variety of African goldfinch,
being a pine tree and similar kinds can be reliably connected to certain
distinctive visible profiles. Whilst a philosophy major does not look like
anything at all, there is a distinctive way African goldfinches and pine
trees look and so on. Nonetheless, I am arguing that neither of these
properties can ever become bona fide perceptual contents.17 The tracking
relation connecting the thick property and the type of distinctive visual
17 Take the property of being my great-uncle. Although I recognize my greatuncle just by looking at him, what the respective PE represents cannot be the very
property of [being my great-uncle]: there is just no distinctive look that tracks
great-uncleness. At most, my PE represents the property of being a certain
individual I am already familiar with, an individual I descriptively pick out as [my
great-uncle] because I know—by non-visual means—that he is my great-uncle.
The same holds for the last president of United States, and any individual object I
can recognize by visual means on different occasions of presentation.
286
Chapter Six Part III
appearance is not embedded into the perceptual system just by perceptual
learning. That is why representing these contents involves their
conceptualization. Exerting a concept entails possession of many beliefs
containing the concept and sensitivity to inferences in which the
conceptualized content has a distinctive role. A perceptual content does not
depend on possession of independent beliefs. At best, the content of
certain perceptual beliefs depends on certain perceptual contents, not the
other way around.
If there is a superior theoretical boundary for perceptually representable
contents, it is embedded in DCVP: any content acquirable by perceptual
learning without contribution of inferences and collateral beliefs can become
perceptual content once so acquired. Nonetheless, this characterization is
quite abstract and formal. To work out a more operative criterion for
establishing whether C is a perceptual content belonging to a PE or not
given a certain content C, empirical considerations are needed about how
our perceptual system works as well as the relation between our perceptual
modules, the mechanisms of perceptual memory and imagery and the
higher cognitive centers.
The constraints on perceptual content posed by DCVP are quite general
so they leave room for further empirical determinations. It may well turn
out that the superior boundary of perceptually representable contents is
blurry so that certain ambiguous cases are a matter of a more or less
pragmatic, if not arbitrary, stipulation. Nonetheless, the debate is a
substantial one: perceptions are not perceptual beliefs; perceptual beliefs
are connected to certain perceptual episodes in specific ways. Other
empirical beliefs acquired via perceptual beliefs are connected with
perceptual episodes (and their proprietary contents) only indirectly, by
being inferentially connected to certain perceptual beliefs. In any case, the
learning history of a cognitive system determines the perceptually
representable contents for the system itself.
III.4 Perceptual Content is Moderately Rich and Outstrips
Phenomenal Content
I have argued for the dual content view; that is, for the view that
perceptual content involves thin or low-level properties (SCM-properties
in the case of vision) on the one hand, and thick or high-level properties on
the other. The phenomenal contrast method does not show that perceptual
content is rich, contrary to what Siegel holds. But perceptual content is
rich. What makes its richness highly plausible is the existence of
perceptual learning and the concrete way it works. By expanding the
The Dual Content View: Beyond the Phenomenal Contrast Method
287
perceptually representable properties by embedding new recognitional
schemes into our perceptual system, it endows the perceptual system with
new perceptual sensitivities so that certain contents become constitutive of
certain perceptual episodes rather than being contents whose
representation is merely prepared by these perceptual episodes. Therefore,
certain thick recognitions are perceptual throughout.
Nonetheless, although perceptual content is rich, I want to argue that it
is not as rich as Siegel holds. In particular, I shall argue that perceptual
content does not include natural kinds. I rather argue for a moderately rich
view against the more radical views proposed by Siegel and Prinz among
others.18 A natural kind, like [tree] or [pine tree], [bird] or [goldfinch], is a
kind to which a perceptually encountered object may or may not belong.
As I have already noted (Chapter One, Part IV.3), what determines
whether an object belongs to a natural kind are certain geneticevolutionary features and other “deep structure” features whose discovery
may be an object of a very long-lasting scientific inquiry. The immediately
apparent properties of the object are usually explained as epiphenomenal
effects of more structural properties, which are assessed, determined and
described by the branch of science investigating the natural kind in
question. For an expert on Fs, something can look like an F without being
an F (where F is a natural kind) or vice versa. The famous Twin Earth case
by Putnam19 is just the case of a planet where a liquid XYZ exhibits
exactly the same apparent properties as our water despite its different
internal molecular structure. Now, this possibility entails that there is no
distinctive appearance-property that can put a subject in a position to
appreciate whether a sample of XYZ belongs to the [water] natural kind or
to the [XYZ] natural kind. This is also the reason why the genuine waterconcept is not an observational concept. Its acquisition is not just
perception-based; rather, it can be acquired and possessed only if the
subject possesses a complex inferential knowledge about molecular
structures, atomic bonds and so on.
Now, if anything that could become thick perceptual content is a
property that can become perceptually recognizable thanks to perceptual
learning, then natural kinds cannot ever be perceptual contents because it
cannot be by perceptual learning that I learn that an object belongs to a
18 Prinz does not explicitly argue that PE can represent natural kinds but he argues
that we can perceptually represent anything we can conceive and conceptually
represent. Therefore, his view entails that we can perceptually represent natural
kinds insofar as we can form a concept of a certain natural kind and mentally
represent it as such.
19 See Putnam 1975.
288
Chapter Six Part III
certain natural kind. Indeed, a natural kind is individuated by a deep
structure, the knowledge of which must be conceptual, inferential and
scientifically acquired. At most, I can learn to visually recognize samples
or members of that natural kind but inasmuch as my sensitivity to a certain
Gestalt-shape can track the presence of a member of a certain natural kind,
that very tracking relation cannot be embedded in a perceptual scheme
without the inferential bridge of connecting beliefs (see above). So, if one
sees a sample of XYZ and mistakes it for water, the mistake is not
perceptual, therefore the liquid’s being water or being XYZ is not part of
the accuracy-condition of the respective PE.
But then, when I claim that a subject can see X as a tree or a dog, what
is it that I am claiming if not that the property [tree] or [dog] are
perceptually represented? Aren’t they standard examples of natural kinds?
I have been arguing that perceptual content is non-conceptual. Its thick
component is categorical and categories are not concepts; still, categories
have some relevant analogies with concepts. For example, categorizing an
object is bringing it under a generality as is conceptualizing an object.
Seeing something as an F means taking it as an example of F; likewise,
conceptualizing something as an F means taking it as an example of F. But
exerting the concept F is much more demanding than activating the
category F. First of all, concept-possession is subject to the generality
constraint (GC) so exerting a concept means being able to draw a lot of
inferences involving that concept, having a lot of beliefs having that
concept as a semantic constituent and so on.
The content of a mental state is conceptual when the subject, in order
to be in that state, must possess the concepts used to canonically
characterize that very content.20 But such a requirement does not apply to
categorical content. For example, I can categorize something as an F
without possessing the corresponding concept F. To possess a
recognitional category is not to possess the respective concept; it is a much
less demanding ability. For example, possession of a recognitional
category may well be a merely passive ability—prompted only by and in
the presence of a concerned object—whereas possession of a concept
entails the ability of spontaneously tokening it in reasoning, planning,
voluntary recalling, imagination or whatever else.
With all this in mind, let us come back to the content-ascriptions S sees
O as a tree, S sees O as a dog and the like. Take [F] as the natural-kind
property of being a tree. If S’s seeing O as a tree is the ascription of a
20 Which is, in short, the definition of non-conceptual content proposed by
Cussins 1990. It then became a sort of shared definition in the relative debate.
The Dual Content View: Beyond the Phenomenal Contrast Method
289
perceptual episode with proprietary and constitutive perceptual content,
then the “tree” property involved cannot be [F], namely, it cannot be the
natural kind F. It must be something else. In fact, to represent something as
an [F], the subject would need to possess the respective concept, many
other related concepts, fully-fledged inferential skills, and some sophisticated
scientific knowledge about botanical taxonomies, genus/species relations
and so forth. All that, though, is not needed to be able to perceptually
represent something as a tree. This is something that a dog and other nonlinguistic mammals can do despite lacking scientific knowledge as well as
inferential and conceptual abilities. The content of this recognitional
seeing must then be a much more indeterminate and rough content that we
canonically characterize by means of certain concepts without the
perceiver possessing and exerting such concepts.
So, ascribing the state of seeing O as a tree to S does not ascribe to S
the representation of the natural-kind property [F], but of another property
[G] we descriptively pick out by referring to [F].
The correct determination of [G] (for example, “tree” for an animal
perceptually sensitive to trees) depends on the cognitive fine-graininess of
the perceiver in its (his/her) stand towards examples of trees. As I have
argued (Chapter One, Part IV), perceptual categorizing amounts to taking
a cognitive stand toward the object, so seeing-as ascriptions must capture
the point of view of the perceiver on its environment.21 This is why seeingas ascriptions are opaque contexts. Seeing O as a G or seeing O as a Z can
ascribe states with very different contents, even in case every G is Z and
every Z is G. Seeing-as is not extensionally ascribed because its ascription
must also capture the way the object is cognized by the subject. Even if an
animal develops a recognitional sensitivity to pine trees—perhaps even to
a particular species of pine trees—this is reliably correlated to and only to
pine trees; its respective perceptions will not have the natural kind [pine
tree] among their contents but some other thick content that must be
specified (in as far as it is possible) with reference to the behavior of the
animal toward examples of that property. Concerning our perceptions, the
most natural way of self-ascribing recognitional contents is that of using
the respective concepts by which we canonically pick out the property we
recognize. But the conceptualization of this recognitional act adds a lot of
information by connecting the perceptual content with inferential
knowledge and a sophisticated net of non-perceptual concepts. However,
we need to remember that categorical recognition, exerted by pure
21 A passive, immediate point of view, which can be changed by higher-order
beliefs. Ascribing these contents must account for both behavioral and
discriminatory abilities and their fine-graininess.
290
Chapter Six Part III
perceptual means, is much more coarse-grained than the concept we exert
as a consequence of the perceptual episode. For however much tougher it
may be to determine a rough category through concepts, it is certain that
perceptual contents cannot be as well-determined as to involve naturalkind properties, and more generally properties such that a fully-fledged
conceptual ability is needed to represent them. In fact, natural-kind
properties are among the properties that can be represented only by
possessing inferential knowledge and many inter-connected beliefs, so the
issue of perceptual representability of natural kinds is just a particular case
of the issue concerning the superior boundary of perceptual representability,
a necessary boundary explicitly embedded into the integrated version of
DCVP: anything that and nothing more than what can be recognized—
thanks to perceptual learning but without inferential integration coming
from collateral knowledge—can be perceptually represented. Natural-kind
properties simply fall out of that range of possible contents so they cannot
ever become perceptual contents.
A final point concerns the relation between Siegel’s argument for what
I have called the dual content view of perceptual content and my
arguments for the same view. I have tried to show that Siegel’s
phenomenal contrast method is flawed but the dual content view can be
argued for anyway on other bases; that is, through elaborating the idea of
perceptual learning and considering how perceptual recognition actually
works. It turns out that perceptual content is rich but only moderately so
because natural-kind properties and similar thick properties cannot be
perceptually represented. Perceptual learning alone does not go that far.
Besides the fact that phenomenal difference can be accounted for in
terms of thin contents, as I have shown in many ways above, there is
another reason why the phenomenal contrast method cannot do justice to
the thickness of perceptual content. By definition, this method can only
concern phenomenal content, the content that is purportedly testified in
conscious perceptual phenomenology, so to speak. But certain
recognitional contents may well be objectively possessed by perceptual
states without involving a proprietary, modality-specific phenomenal
change with respect to the case in which they are absent (the “contrastexperience”).
Even if we keep neutral about the existence of a recognitional
phenomenology, even if there were no phenomenology-based way to
establish that (some) PEs have thick contents besides their thin content,
our DCVP could be maintained anyway as it is a principle independent of
the assessability of phenomenal effects of the acquisition of a thick content
by a given type of perceptual episode. No reference to the conscious
The Dual Content View: Beyond the Phenomenal Contrast Method
291
character of PE is made in DCVP. A given mental content can well be a
visual content even though it does not directly enter into the visual
phenomenal character. In short, perceptual content can outstrip perceptual
phenomenal character. Besides the cases of seeing without noticing
already referred to (see Chapter Two, Part I.2.1-2) concerning normal
subjects and besides pathological cases like blindsight where there is
visual content represented without any visual phenomenology at all, there
are other situations where certain things start looking a certain way
without any visual phenomenal difference.22 An example is the so-called
“chicken-sexers.” Chicken-sexers are able to quickly spot, just by looking,
whether a chicken is male or female. It is an ability acquired by training,
which does not involve any introspectively noticeable phenomenal change
at all: indeed, chicken-sexers do not have any idea about which visual cues
their visual system uses to tell apart males from females. They just guess
without any observable basis.23 It is an acquired visual ability and the
acquisition of it does not modify the way things phenomenally look at all.
Chickens do not phenomenally look different from the way they used to
look before the subjects acquired such a vision-based recognitional
ability.24 No visually salient detail can be consciously individuated as the
difference-maker. Rather, a sort of instinct is developed, a visual
sensitivity without any visual conscious character associated with it.
According to DCVP, a property roughly corresponding to [being male]
and [being female]—perhaps even just [being of a sort] and [being of the
other sort] as subtypes of the [chicken]-type—is perceptually represented
because a perceptual sensitivity is developed that does not essentially
22 That is what Lyons 2005, for example, calls “non-experiential looking.” Prinz
2006 also commits himself to the view that the content of perceptual experience,
for example, the content of visual experience, can get thicker and thicker without
the visual phenomenal character changing accordingly. I agree with Prinz on that
point (as well as with Lyons on the same point), although I do not agree with his
main thesis that perceptual content can be expanded indefinitely, as far as to
contain any abstract property like [being unjust] or [being a philosophy major].
23 For Lyons 2005, there is a relevant sense of “looks” that is not experiential but
perceptual throughout: when I am prepared to take O as an F by seeing O but I
cannot point to a distinctive phenomenal look associated with O’s being F. The
case of chicken sexers perfectly responds to that characterization.
24 Perhaps some could argue that phenomenal character changes unknown to them
so they have no special authority about what phenomenal character their PEs have.
That seems to me a desperate move: postulating a phenomenal difference that is
not noticeable at all by the experiencing subject is a way of abusing the very
concept of phenomenal character. Phenomenal character is the “what it is like”
aspect of PE; it is conscious by definition.
292
Chapter Six Part III
involve any inferential preparatory operation. There is just perceptual
learning (such a sensitivity could be developed by a child before her
having the concepts of [being a male] and [being a female]). So, through
perceptual learning, a certain type of PE acquires new contents without
which such a semantic enrichment involves no phenomenal change, at
least no visual phenomenal change. Even if we keep neutral on the
existence of recognitional phenomenology—let us suppose that there is no
such thing—we should credit a PE with content that outstrips the PE’s
phenomenal character. Hence, there is no problem with taking it that a
given perception has certain visual contents without a correspondent visual
character. So, a part of perceptual content can lack phenomenal character.
Another example may be that of emotion detection. We can visually
recognize certain emotions in others’ faces but we are not able to tell
which visual detail made us attribute that emotion to that person in every
case. We just “see” it. For example, she just looks in pain; we don’t know
how a phenomenally accessible, fine-grained sensible profile tracks the
painfulness property.25
All this sheds light on why Siegel’s main question “which properties
are represented in perception?” is potentially ambiguous. If by “represented in
perception” is meant “testified in the phenomenal character of perceptual
experience,” it is one thing; if by “represented in perception” is meant
more broadly “represented by a perceptual experience,” it is another. If
part of PE content were not phenomenally salient, or at least not a part of
the phenomenal character associated with the specific modality of the PE
(e.g., visual phenomenal character), such a layer of content could not be
established through the phenomenal contrast method.
So, even if the phenomenal contrast method is conclusive, still there
could be bona fide perceptual contents whose presence in the E1 and absence
in E2 would not involve any introspectively appreciable phenomenal
contrast between E1 and E2.
To conclude Part III, I want to point out that the imperfect
correspondence between phenomenal character and representational
content—the second may outstrip the first—does not undermine in any
way our (impure) representationalist view on perceptual phenomenal
25 As Lyons efficaciously points out, “We should not assume that the low-level
representations in virtue of which high-level phenomenal content is fixed must
themselves be phenomenally conscious. We are often sensitive to the mental states
of others without being aware of the basis of our sensitivity.” Lyons is talking
about thick phenomenal content but it could be said that phenomenal content is
made out of low-level properties, and perceptual content can outstrip phenomenal
content just insofar as it can be thick (Lyons 2005, 31).
The Dual Content View: Beyond the Phenomenal Contrast Method
293
character.
Impure representationalism is the view that the phenomenal character
of a PE is determined by its content plus its mode. So, the properties the
PE represents the object as having and the mode under which the PE
represents these properties as instantiated by the perceived object
determine the phenomenal character of the PE. Now, this is perfectly fine
even if we allow perceptual content to outstrip phenomenal character.
Phenomenal character is made out of represented properties (under a
mode) but nothing prevents other properties being represented over and
above the properties that shape a visual PE’s visual phenomenal character.
I am saying that perceptual learning can enrich or “thicken” the content of
a visual perception without having to enrich or “thicken” in the same
proportion its proprietary (= visual) phenomenal character.26 In other
words, all phenomenal properties are represented properties, but not all
represented properties have to have a phenomenal counterpart.27 On
phenomenal character, representationalism is silent about the possibility of
a part of representational content being over and above the phenomenally
represented properties.
The visual mode, under which visual content is given to the subject in
a PE, may be only partially sensitive to thick visual contents. I do not want
to commit myself to a specific view on whether thick properties modify
the visual mode and enter into the visual phenomenal content, nor do I
want to commit myself to a view about which perceptually represented
thick properties modify the visual mode and which don’t. The main point
is rather that perceptual content is likely to be more extended than
phenomenal content: this can be held as a plausible idea, independent of
what we think about what part of the perceptual content is included in the
phenomenal content and what part is excluded. Perceptual learning may
only partially modify the phenomenal character of visual perception;
therefore our PEs could well acquire new contents without acquiring a
new corresponding conscious character.
26 Lyons 2005 notes that when we acquire a visual recognitional disposition, even
if there is a visual phenomenal change, that change is less relevant than the
representational change anyway: for example, if you recognize someone by sight,
the representational change from before is very relevant, whilst the visual
phenomenal change (assuming there is one) is not as relevant as the
representational one. That suffices for showing that not every property
perceptually represented is necessarily mapped in perceptual phenomenal
character.
27 In Chapter Three, Part III.4.1, I already held that conscious phenomenology
might not exhaust all the informational richness conveyed by a perceptual state.
294
Chapter Six Part III
Concluding Remarks
In the present chapter, I have treated the issue of what properties can be
represented in perception.
I first distinguished between “thin” (low-level) and “thick” (high-level)
properties and I argued that recognitional seeing is a perceptual episode
with a thick content that can be represented only if the perceptual episode
represents a visible profile made out of SCM-properties in the first place
(I.1). By distinguishing between a “liberal” view and a “conservative”
view of perceptual content (I.2), I presented the dual content view, which
is a certain version of liberalism (I.3). According to this view, perceptual
experiences have a dual content, a basic thin content and a thick content
that a certain type of PE can acquire by perceptual learning.
Second, I critically considered the phenomenal contrast method (II.1),
a method used by Siegel, among others, to prove the truth of the liberal
view. Such a method starts from a certain hypothesized phenomenal
contrast between two experiences E1 and E2—E1 having a thick content
and E2 identical to E1 apart from not having that thick content—and
concludes that the thick property’s being represented in E1 and not being
represented in E2 must be what accounts for the phenomenal difference
between E1 and E2. I have variously argued that such a method is not
conclusive because the phenomenal contrast can be accounted for even
from a conservative point of view; that is, by considering that a certain
thick, non-perceptual representation caused by the perceptual episode can
have top-down effects in representation of thin properties: by orienting
visual attention, by conditioning visual imagination, by changing the
temporal order of the explored parts of the seen object and so on. More
generally, a visual phenomenal contrast can well be caused by the
representation of a thick property without that very thick property being
represented in the very visual episode. I argued that neither visual agnosia
(II.3) nor ambiguous figures (II.4) supports the liberal view so I have
negatively shown that if the liberal view can be vindicated, it needs to be
vindicated on a basis other than the phenomenal contrast method, which
turned out to be flawed.
My positive arguments for the dual content view—and so, for
liberalism on perceptual content—rest on the consideration of perceptual
learning (III.1). I first expressed the DCVP, according to which, when a
subject is visually sensitive to a certain visible profile, that profile’s being
instantiated tracks the instantiation of another thick property and that very
tracking relation is embedded into a learned and stabilized perceptual
schema. The thick property in question is then considered part of the
The Dual Content View: Beyond the Phenomenal Contrast Method
295
perceptual content of the PE in which that perceptual recognitional schema
is successfully activated by a seen object.
I have considered some objections to DCVP; the first of which is that
if the principle holds, then we should allow any property occasionally
recognizable just by looking as possible perceptual content. I made a
distinction (III.2) between the preparatory and constitutive cognitive
processes associated with a PE: certain recognitional processes are made
constitutive of certain perceptual episodes through perceptual learning. In
other contexts, perceptual episodes are only preparatory to certain thick
recognitions together with other inferences, beliefs and collateral
knowledge. When we recognize an F just by looking, F is a bona fide
perceptual content only if it is constitutive of the perceptual episode; that
is, only if the respective perceptual scheme has been settled just through
perceptual learning and without supplementary inferences. Such a
difference between constitutive and preparatory processes allows us to
face the above objection: only certain properties recognizable just by
looking are constitutive contents of perceptual episodes.
In succession (III.3), I argued that there is a superior boundary of the
range of perceptually representable properties: properties the mental
representation of which involves the possession of a concept cannot be
perceptually represented. Indeed, perceptual content is non-conceptual; if
the mental representation of a property directly or indirectly involves
evidential relations between beliefs and inferential bridges, such a property
cannot be perceptually represented.
Finally (III.4), I argued that perceptual content is rich (this is the
liberalism of the dual content view) but only moderately so; that is, for
example, visual content includes properties other than colors, shapes, sizes
and movement (SCM-properties) but it cannot include natural-kind
properties, contrary to Siegel’s view. Representing a natural kind property
entails exerting certain concepts and inferences, possessing a certain
inferential knowledge and so forth: all these conditions are incompatible
with acquiring certain content just by perceptual learning; therefore,
natural kinds cannot be represented in perception.
Furthermore, I argued that the representational content of a PE can
outstrip its phenomenal character: our PE can be enriched with new thick
contents by perceptual learning but these semantic changes may also only
partially correspond to respective phenomenal visual changes.
Nonetheless, this view does not undermine the impure representationalism
argued for in Chapter Three, Part III: that the phenomenal properties are
just properties represented under a mode is true even though a part of the
perceptual content is not phenomenally salient. Things can literally look F
296
Chapter Six Part III
without their looking F being consciously testified in visual
phenomenology.
If perceptual content outstrips phenomenal character, it also outstrips
phenomenal content—the content a perceptual experience has by virtue of
having a certain phenomenal character—therefore the phenomenal
contrast method, even in cases where it was conclusive, would be
insufficient insofar as it concerns by definition only phenomenally salient
contents.
To sum up, perceptual content is dual, independently of whether the
thick level shapes the visual phenomenal character, and independently of
which, among the thick perceptual contents, can shape the visual
phenomenal character and which cannot.
CHAPTER SEVEN
BRINGING THE DISJUNCTIVIST CHALLENGE
INTO THE INTENTIONALIST VIEW
In this seventh chapter, I will present the disjunctivist view of perceptual
experience and the main reasons the advocates of this view adduce for
holding it.
Second, I will show that such reasons can be respected also within my
version of CV so there is no need to abandon it, and a moderately
disjunctivist version of CV itself is preferred to the anti-intentionalist
disjunctivisms like naïve-realist disjunctivism.
Third, I will argue that within a moderately disjunctivist version of
CV, the semantic gap problem—the problem of ascribing semantically
evaluable Content to objectless states like hallucinations—can be
successfully treated.
PART I
DISJUNCTIVISM, INTRODUCED
I.1 What Is Disjunctivism?
Disjunctivism about perceptual experience (DJ) comes in many forms.
The basic idea of DJ is that veridical experiences and deceptive
experiences (like hallucinations) should be considered as two different
kinds of mental states, namely, they should be given disjunctive treatment
along the following lines:
DJ: Either E is a manifestation of such and such worldly objects and properties
of the subject, or E is another kind of state that only seems to be a state of the
first kind.1
DJ rests on the rejection of a principle that—according to disjunctivists
like McDowell and Martin—often works as a hidden assumption: the
“common highest factor assumption” (McDowell 1982) or the “common
kind assumption” (Martin 2004).
CKA: Whatever kind of mental event M occurs when you veridically perceive,
that same kind of mental event M can occur when you hallucinate.2
1 As we will see, the “other kind” for hallucinatory states could be taken to be a
“mere appearance” (McDowell 1982) or a “spurious” kind consisting of anything
that is indiscriminable from a certain veridical perception (Martin 2004, 2006) or
something else. See McDowell 1982, 1994, 2008. Disjunctivism about PE was first
proposed by Hinton 1967, 1973. Other advocates of disjunctivism about PE are
Snowdon 1981, 2005, Putnam 1999, Martin 2002, 2004, 2006, 2008, Brewer 2008,
Travis 2004, 2006, Campbell 2002, Johnston 2004, 2006 and Fish 2009 among
others. Important contributions on the subject are collected in Haddock and
Macpherson 2008, and in Byrne and Logue 2008, Howthorne and Kovakovich
2006.
2 For a detailed discussion of CKA, see Martin 2004, 2006.
Disjunctivism Introduced
299
Both DJ and the related rejection of CKA entail the rejection of two prima
facie plausible principles: the first, already introduced in Chapter Three,
Part II.5, is the one I called the Cartesian principle.
CP: If two mental states are subjectively indiscriminable, they are mental
states of the same nature, kind and type.
Given that a certain hallucination could well be subjectively
indiscriminable from a certain veridical perception, then DJ and its related
rejection of CKA are ipso facto the rejection of CP, namely, of the idea
that the subjective indiscriminability of experiences E1 from E2 is
sufficient for their identity in type, nature and kind.
The second principle a disjunctivist must reject has to do with the
sufficiency of the proximate causes of a mental event for its individuation
as an event of a certain mental kind.3 According to the proximate cause
principle,
PCP: A mental event E1 is of the same kind as a mental event E2 if E1 is
brought about by the same kind of proximate cause as E2.4
Given that a certain hallucination could in principle be brought about
by a certain neural proximal stimulation, committing to DJ and the related
rejection of CKA is ipso facto committing to the rejection of the principle:
same kind of proximate cause → same kind of effect. Indeed, according to
DJ, a certain hallucination and a certain veridical perception could well
have the same kind of proximate cause without being mental events of the
same kind. DJ entails the rejection of PCP so there must be some noncausal conditions for two experiences to be of the same kind.
3 See Martin 2006, 356ff.
4 The rejection of CPC does not reject at all what Martin calls “experiential
naturalism” and that introduces as the principle that “our sense experiences are
themselves part of the natural causal order, subject to broadly physical and
psychological causes” (Martin 2006, 357), so our PEs are “subject just to broadly
physical causes […]” (Martin 2004, 273). Rejection of CPC only denies that the
identity in kind of proximate causes are sufficient for the identity in kind of two
mental events, so there must be identity in kind of distal causes, or some other noncausal condition must be satisfied, in order for two mental events to be of the same
kind. That is not to deny that experiences are subject, in general, to physical laws
and parts of the causal order, of course. That would be obscurantist at the very
least.
300
Chapter Seven Part I
To sum up, DJ’s rejection of CKA and CKA entails the rejection of both
CP and PCP.
I.2 The Reasons for Disjunctivism:
The Detachment Problem
With all this in mind, let us now turn to the positive reasons why,
according to its advocates, the disjunctive treatment of PE should be
adopted. Without any aim of suggesting a hierarchy or being exhaustive, I
will sequentially present four kinds of typical reasons for DJ:
phenomenological, epistemological, semantical and metaphysical.
Although such levels are deeply interwoven in the debate about DJ, it may
be useful to introduce them one by one in order to gain in clarity albeit at
the price of a bit of abstraction. Actually, all four kinds ultimately rest on
the idea that disjunctivism is the only way of solving what could be called
the detachment problem: any non-disjunctivist view on PE seems to drive
a fatal wedge between the perceptual experience and the world. This is the
key problem that underlies and shapes the debate in various forms and
dimensions.
I.2.1 Phenomenology
The phenomenology of perceptual experience exhibits the property of
transparency (see Chapter Two, Part II.4). In attending to your own
experience of an object, the only properties you will end up attending to
will be those your experience presents to you as the very properties of the
object you are perceiving; the properties that the object would have were
the experience true.
But there seems to be another property that PE introspectively seems to
have besides transparency, namely, the properties of actuality and
immediacy. In attending to your experience of an object, not only do you
attend to the “putative” properties your experience attributes to the object
(transparency), but your experience itself also stands as the very
presentation of such properties of the object; the so-presented properties
offer themselves, in experience, in their bodily presence “out there” as
actual and immediately given to you. So, besides a phenomenology of
transparency, in PE there is also a phenomenology of actuality and
immediacy; in experience, the world itself seems to be in view, manifested
to the subject. PE just seems to be, introspectively, such a similar
manifestation of the world with its objects, properties and relations. The
presentational phenomenology of PE is especially compatible with direct
Disjunctivism Introduced
301
or naïve realism, according to which PEs are direct relations of
presentation of the world to the subject.
As Martin 2002 points out, the classical sense data theory vindicates
actuality—sense data are there, present as mental objects genuinely related
to the subject5—but unfortunately it falls short of vindicating transparency.
According to such a theory, the properties you are aware of in your
experience are not properties things look to have; rather, they are
properties of sense data. Conversely, intentionalism vindicates
transparency but seems to fall short of vindicating actuality. In fact, if the
PE was falsidical, the content of PE would not be exemplified so it would
not be actual by definition. Only direct realism can vindicate both
transparency and actuality at once.6 According to direct realists, the
properties we are aware of in PE are properties things seem to have, and
they are also worldly properties in flesh and blood that the subject is
presented with in her experience. Therefore, direct realism appears to be
phenomenologically apt with respect to transparency and actuality on top
of capturing the commonsense view on perceptual experience.
However, the fact that illusions and hallucinations are possible is just
what intentionalism aims to justify through the notion of content. Indeed,
illusion is the “mother of intentionalism”: in illusion and hallucination,
actuality does not hold by definition since the property you experience as
had by an object is not actually had by the object (illusion), or the very
“object” is not actual because it does not exist (hallucination).7
Nonetheless, according to Martin, Fish and others, a disjunctive treatment
of deceptive experience along the following lines could save the virtues of
naïve realism without embedding its vices: when the PE is veridical, it is
the direct naïve-realist presentation of worldly items, and when it is not
5 Price 1932 asserts that when you have an experience as of a tomato, nothing
could be more certain that you are aware of something: something is there before
you to be experienced be you hallucinating or not. The actuality intuition is related
to the presentational phenomenology of PE. We can call that thesis, with Pautz
2006, item awareness.
6 I remark en passant that the other remaining classical option, adverbialism, does
not vindicate either transparency or actuality. By the adverbial view, experiences
are ways of a subject being modified: so in experience we are not aware either of
properties the experience represents things to have (transparency), or of worldly
and actual objects, properties and relations (actuality). So at least from a
phenomenological point of view, adverbialism is an error theory. I have already
suggested that the adverbial theory of perception is phenomenologically inapt and
leaves unexplained certain apparent properties of perceptual experience: see
Chapter Two, Part 1.3.
7 On the incompatibility between illusions and direct realism, see Smith 2002.
302
Chapter Seven Part I
veridical (illusion, hallucination), it is not a presentation of worldly
properties but something else, an episode of a different mental kind that is
subjectively indiscriminable from a certain genuine presentation of the
world.
Although not all disjunctivists are naïve-realists like Martin and Fish—
for example, McDowell does not reject CV—the anti-intentionalist use of
disjunctivism is very important in the debate. Insofar as such a version of
DJ posits itself as a deep criticism of CV, I will give special consideration
to it: I will call it disjunctivism-cum-naïve-realism (DJ-cum-NR).
In any case, according to DJ, veridical PEs are genuine manifestations
of the surrounding world to the subject, and their presentational
phenomenology would be best explained by taking such PEs as being
episodes of presentation, as being what they introspectively seem to be.
According to DJ-cum-NR, the very fact that our veridical PEs are genuine
manifestations of the surrounding world is incompatible with
intentionalism. A presentation of the world presupposes actuality in a way
that representation of the world does not: a represented world is not a
necessarily present world.
A conjunctive view of PE8 would posit a non-world-involving
common factor shared by veridical experiences and deceptive
experiences—for example, the content—plus some additional elements
that plausibly would be present in the veridical case but absent in the
deceptive case, whereas a disjunctive view starts from the “good” disjunct
considered an essentially world-involving manifestation, distinct from the
“bad” disjunct, considered a mental event of a different kind that only
(introspectively) seems to be a genuine manifestation of the surrounding
world.
On that account, according to the DJ-cum-NR version of disjunctivism,
deceptive cases do not speak against naïve realism because only the nonillusory/hallucinatory disjunct is given a naïve-realist account, whilst the
illusory or hallucinatory disjunct is told apart as something indiscriminable
from a certain veridical experience without it falling into the same mental
kind just by virtue of such an indiscriminability-property. Two things can
be subjectively indiscriminable without falling under the same mental
kind:9 this is another way of saying that the Cartesian principle is false.
To sum up, the presentational phenomenology of PE involves the
purported actuality of the manifested world, and it is best explained by this
actuality, at least in the veridical cases. Such a phenomenology is hardly
8 The opposition between disjunctive and conjunctive views is proposed by
Campbell 2002 and Johnston 2004.
9 On that, see Austin 1962, Martin 2006.
Disjunctivism Introduced
303
compatible with any conjunctive view of PE, which allegedly drives an
unbridgeable wedge between the subject and the world. If a veridical
experience is individuated and characterized independently of its being a
genuine manifestation of the world, the same experience could present the
subject with the world only “indirectly.” This is problematic from a
phenomenological point of view in the first place. Indeed, the issue of
phenomenological aptness is the first facet of the detachment problem.
I.2.2 Epistemology
Through our perceptual experiences, we come to acquire knowledge about
the surrounding world outside us. Therefore, perceptual experience has a
key epistemological role in “anchoring” our thoughts and empirical
judgments to the world,10 and providing our empirical thoughts with an
objective purport. Now, if perceptual experiences are to ground and justify
our empirical judgments about the surroundings, they cannot also make
available for us what illusion and hallucinations make available. If what is
given in perceptual experience were just an “appearance” shared by
illusions and hallucinations, then the justificatory power of a perceptual
experience would be as defective as the justificatory power of a
hallucination. In this case, the very external world would not make any
relevant contribution to our empirical knowledge,11 which is simply
absurd. The skeptical challenge would easily rise at this point and become
inescapable to the effect that our everyday perceptual judgments, even
produced in optimal conditions, could never count as genuine knowledge,
which is exactly what we do not want to be the case.
In other words, if we tell apart the “appearing” involved in perceptual
experience from the worldly items such appearing is a manifestation of in
the veridical cases—as do the conjunctive views that posit a neutral
“appearance” sharable by veridical perceptions and hallucinations—we
fatally introduce an irreparable gulf between our experience and the world.
That is, indeed, the epistemological facet of the detachment problem. On
the face of it, DJ aims to radically overcome any “interface model of
perception” (Putnam 1999, 42), or any idea of perceptual experience as a
“veil between us and the world” (McDowell 1982, 215), the very idea
originally responsible for the rise of the detachment problem. The latter
can be left behind only by conceiving two classes of epistemically distinct
10 This epistemological reason for disjunctivism is provided by McDowell 1982,
1994, 1995, 1998, 2008 as well as Travis 2004, 2006 and Campbell 2002.
11 The world itself would contribute to our knowledge of the world only in an
externally causal way, not in a constitutive way.
304
Chapter Seven Part I
experiences: those occurring when the way things are “makes itself
perceptually manifest” to a subject, and those occurring when it merely
seems to the subject that things are a certain way (where the experiences
of the second class misleadingly present themselves as belonging to the
first class). Thus, the only radical solution is the disjunctive treatment of
PEs.
I.2.3 Semantics
Russell used to distinguish descriptive knowledge from knowledge by
acquaintance: the second makes the first one possible, and involves direct
epistemic contact with an object without any inferential, predicational or
descriptive mediation.12 The typical way of getting acquainted with a
worldly object is perception. Perception offers the demonstrative,
referential basis for any predication, for any descriptive knowledge of the
surrounding environment. In other words, perceptual experience makes
available to us worldly particulars as “topics” for empirical judgments
about them and for propositional knowledge of them.
As Evans 1982 vigorously pointed out,13 perceptual discrimination is
the basic way our empirical thoughts can be anchored to the world,
namely, perceptual discrimination of certain objects provides original
referents for empirical judgments that “directly” involve such perceived
objects. In other words, perceptual experience basically makes
demonstrative thoughts about the world possible.14 In this way, perception
grants the very aboutness of empirical thinking.
According to Campbell, Johnston15 and Travis, only the “revelatory”
nature of perceptual experience could account for our capacity to make de
re judgments about the surroundings. The particulars being-directlypresented to the subject in experience, without any semantic intermediary
12 “We shall say that we have acquaintance of anything of which we are directly
aware, without the intermediary of any process of inference or any knowledge of
truths” (Russell 1912, 25). On the difference between acquaintance and descriptive
knowledge, see also Lewis 1983.
13 See Evans 1982, Chapters 4 and 6. See also McDowell 1984.
14 Campbell 2002 insists emphatically on these semantic reasons for
disjunctivism. See also Travis 2006. On singular thoughts and object-dependency,
the classical works are Evans 1982, Kaplan 1989, McDowell 1984, 1986. A
singular thought about a perceived object is made possible by the very existence of
the perceived object; i.e., that thought—with that content—could not be
entertained without the existence of the perceived object it is about.
15 See Johnston 2004, 2006.
Disjunctivism Introduced
305
like representations, enable us to judge directly about them not just
through mental representations of them. Johnston writes that “without
sensible consciousness we would not have any singular thought about
perceivable objects” (Johnston 2006, 265), our PEs “introduce particulars
as topics for thought” (Johnston 2004, 130) and “make external entities
available as objects of immediate demonstration” (Johnston 2006, 282).
Without the function of “bringing into view” realized by our PEs, not only
could the world not control our empirical judgments but our very scheme
of descriptive identification would collapse.
Empirical judgments are exposures to error just because they
presuppose that something worldly has been perceptually brought into
view in the first place, the worldly object the judgment is about that can
make the judgment true or false.
So, perceptual experience is not a representational phenomenon but a
way of directly “bringing the surroundings into view” (Travis 2004),
whereas empirical beliefs and judgments are representations about
portions of the world that perceptual experience has brought into view and
made available as a referent for true or false judgments. The role of PE is
not that of representing the world but rather, “the central task of seeing
[…] in a thinker’s life is to allow the world to bear, for that thinker, on
what he is to think (and do) according as it bears on what is so” (Travis
2006, §5).
Again, conceiving veridical PE as the direct presentation of particulars
as “topics” for empirical thoughts is a way of avoiding the detachment
problem that allegedly weighs on CV as well as on any other nondisjunctivist view. As is evident, the semantic issue is deeply interwoven
with the epistemological one. So, for the advocates of DJ, any semantic
and/or epistemic “interface” between the subject and the world would
confine the mind within itself and make impossible to us any knowledge
of the world on the one hand, and any reference to the world on the other.
At least the veridical experiences are to be thought genuine presentations
of worldly particulars in order to guarantee that mind-world cognitive
contact: this is the only way to cope with the detachment problem.
306
Chapter Seven Part I
I.2.4 Metaphysics
Perceptual experiences are mental states16 of a certain kind that have a
certain nature. As for any other entities we quantify, some criteria for the
identity and the individuation of a given perceptual experience must be
available, some type or kind of individuation must be possible.
Furthermore, among the many types and kinds under which a PE may fall,
there must be a “fundamental kind,”17 which the PE belongs to, that can
somehow capture the deep metaphysical nature of such an entity.
From an intentionalist point of view, PEs are essentially individuated
by their semantic properties, namely by their respective representational
content. A PE is what it is insofar as it possesses certain accuracyconditions, independently of whether such conditions are satisfied in the
world; i.e., independently of whether the content is exemplified or not. As
a consequence, the very nature of the mental state a certain PE consists of
must be indifferent to how the represented world is actually arranged.
Indeed, the semantic evaluability of a PE is possible just because the
individuation of such a state is independent of its semantic value
(true/false, accurate/inaccurate). This idea seems to entail that a certain
veridical PE could well have been an illusion or a hallucination—if things
had been different in the surrounding world—without being thereby of a
different kind or nature. In other words, if the fact that a certain PE
possesses certain content fundamentally individuates that PE, and such
content is possessed independently of whether the PE is veridical, illusory
or hallucinatory, the veridical perception being a genuine relation to the
world—say, a relation of manifestation of the world to the subject—must
be totally inessential to the nature of that PE. The PE is what it is
independently of the world and having a genuine relation with the world;
the metaphysical nature of the PE, even of veridical PEs, is essentially
world-independent.
16 I have already clarified (Chapter Two, Part 1.1) that PEs are episodes rather
than states. Here I just adopt the current use in order not to overload the discussion
with terminological subtleties.
17 Martin insists that a given PE must belong to a fundamental kind. The
fundamental kind to which an experience belongs is “its most specific kind; it tells
what essentially the event or episode is” (Martin 2006, 361). While an item may
have many natural properties, it must fall in just one fundamental kind. This is
what Wiggins 1980, 65, calls the “ultimate sortal” of a thing. I personally find this
notion highly suspicious if it is meant to metaphysically “carve the thing at its
joints” rather than being more modestly meant as the result of a pragmatic
taxonomy useful for certain descriptive or explanatory purposes.
Disjunctivism Introduced
307
Also from a metaphysical point of view, any conjunctive view for
which a veridical perceptual state can be factored into a world-independent
psychological state on the one hand, plus a certain arrangement of the
world on the other, is doomed to fail at accounting for the PE being a
cognitive contact to the world in which the world itself becomes manifest
and “disclosed” to the subject. In particular, if perceptual experience were
a representation, it could never be a presentation of the world itself
because a representation is what it is independent of its content being or
not being exemplified, while a presentation-manifestation of the world has
a relational nature that is essentially determined by the very world of
which the mental state (the PE) is a manifestation.18
As a consequence, only a disjunctive view is able to account for the
relationality of veridical perception by assigning to the latter a different
fundamental kind from that of illusions and/or hallucinations.19
Hallucinations must be thought of as having a different metaphysical
nature than veridical perception. Like the latter, they introspectively seem
to be genuine relations of manifestation but they are not, therefore
hallucinating involves a double mistake: about how the world is arranged
on the one hand and the very metaphysical nature of one’s own state on
the other.20
The metaphysics of perceptual states cannot be given conjunctive
treatment on pain of distancing the world from the subject; also with
respect to veridical cases, on pain of making the relationality of veridical
states metaphysically irrelevant; therefore on pain of closing the mind in
on itself and leaving the world outside. As is obvious, this is nothing more
than the metaphysical facet of the detachment problem.
18 On the opposition between an intentional view and a relational view on
perception, see Crane 2006.
19 As we will see, some disjunctivists oppose veridical perceptions to illusions and
hallucinations (e.g., Martin); other disjunctivists oppose veridical perception and
illusions to hallucinations (e.g., Snowdon, Langsam, Fish).
20 On hallucinatory states being inaccurate with respect to the world and to the
subject’s introspection of those states themselves, see Martin 2006.
PART II
THE GOOD, THE BAD AND THE NEUTRAL
II.1 The Priority of the Successful
Disjunctive views on perceptual experience take the veridical experience
as basic and analyze the deceptive cases (e.g., hallucinations) in terms of
the veridical experience. They make no room for any neutral experiential
states. Putatively neutral states are treated as either instances of the
successful disjunct or of the unsuccessful disjunct, where the disjunction is
usually interpreted exclusively. So, ascribing a “neutral state” can only
mean you do not know whether it is an instance of the one or of the other
disjunct rather than you have picked out a common ingredient of
successful and unsuccessful states. Since the unsuccessful state on its own
is analyzed in terms of the successful state, ascribing a “neutral” state
amounts to ascribing a state that at least seems to be a successful state,
where the seemingness is not completely endorsed by the one who
ascribes or self-ascribes the state.
Conjunctive views on perceptual experience take the neutral case as
basic, and analyze the successful case by factoring in the neutral case in
addition to other factors. The unsuccessful case will be characterized by
the lack of that further factor that determines success. The neutral case is
instantiated in both the successful and the unsuccessful case as a common
factor that makes up the successful case when conjoined with the further
ingredient.
Apart from radical disjunctivism, the disjunctive treatment of PE may
well involve a commonality of positive features between successful and
unsuccessful states (say, VP and H). What matters, for a view to be a form
of DJ, is that the common features are not seen as the most relevant
features in kind-individuating the two states as mental states, so that what
the successful and the unsuccessful states “fundamentally are” is common
to them. Given that a mental feature can be more or less relevant in
characterizing and individuating a mental state, it follows that there can be
more or less moderate versions of (non-radical) DJ. I will argue that there
The Good, the Bad and the Neutral
309
is not an aut-aut opposition between conjunctivist and disjunctivist views;
some features may be common to the successful and the unsuccessful case
and also mentally relevant to both, even if the unsuccessful case is
analyzed in terms of the successful case and so the successful case is
conceptually, explanatorily and metaphysically prior. I will argue that a
moderately disjunctivist intentionalism is such a view, on top of being the
most promising view in addressing the demand for cognitive contact
raised by disjunctivists; i.e., in putting us in a position to solve the
detachment problem.
In Chapter One, I argued that perceptual verbs and locutions like
“seeing” and “seeing-that” are success-verbs, so the ascription of them
encodes perceptual success. I also argued that “looking-F” conceptually
depends on “being F” so the idea of experiencing something as red
conceptually depends on the idea of successfully seeing something red.
Even at the level of ordinary ascriptions, the success-case is the basic one
by reference to which the deceptive cases are ascribed and understood.
CV can vindicate these ordinary intuitions about perceiving and
experiencing, as well as the core disjunctivist idea that the success-case
must be taken as the basic case by reference to which the neutral and the
unsuccessful cases are to be characterized. Nonetheless, the explanatory,
conceptual and metaphysical priority of the success-case does not rule out
the idea of a mentally relevant common factor between successful and
deceptive experiences between VP and H.
To show this, I will introduce the basic teleo-functional framework
first, a particularly promising version of CV. Without arguing for the truth
of this version, I will suppose it is true and argue that it is a good example
of how CV can embed the intuition of the good/bad asymmetry and also
meet the demand for cognitive contact that typically inspires the
disjunctivist proposals. Unlike the radical DJ-cum-NR, a moderately
disjunctivist CV does not entail any obscurantist consequence about the a
priori unaccountability of hallucinations: indeed, if the indiscriminability
between a veridical perception and its hallucinatory counterpart had an
explanation of any sort, it would appeal to properties common to the
hallucination and the veridical perception, but in this case it would no
longer be true that the H and VP have nothing relevant in common other
than their subjective indiscriminability, as radical disjunctivists like Martin
and Fish hold.
310
Chapter Seven Part II
II.2 Function and Content
II.2.1 Where Do the Semantic Properties
of Perceptions Come From?
No-one could reasonably deny that perceptual capacities have been
acquired by biological systems through evolution by virtue of the survivalvalue associated with them. A capacity realized by a mechanism because it
had survival-value is a teleo-function. If a function—a mechanofunction—is a causal role, a teleo-function is a causal role that has been
selected because it contributes to the survival and fitness of organisms
equipped with mechanisms that implement it. Biological functions are
teleo-functions, and so perceptual functions, being biological functions,
are teleo-functions too. According to the teleo-semantic accounts of
perceptual content (TS),1 the content of a perceptual state is determined by
the teleo-function of that state. The teleo-function of a certain type of
internal state is not just its actual causal role; rather, it is the causal role
that type of state has been selected to realize. The basic perceptual teleofunctions are selectionally acquired by organisms through the history of
the evolution of their species, and they are phylogenetically fixed and
wired-in. Then individual learning may expand and determine perceptual
deliverances by individually and ontogenetically “selecting” other more
fine-grained discriminatory capacities through the exploitation of the basic
species-specific capacities.
Let us consider an environmental property, like [red]. According to a
certain version of TS,2 the actual internal state you are in when you
successfully discriminate something red has the semantic content [is red]
because the state is a token of the type of state produced by a mechanism
selected because the states produced by it reliably co-varied with the
presence of something red, and having been selected amounts to
possessing the proper function of producing states that indicate the
presence of something red. Perceptual functions come from the evolutionbased “exploitation” of certain relations of co-variation between internal
states and environmental conditions.
Suppose this sketchy story is true and we have a plausible, naturalistic
account of why certain states have certain contents; indeed, such contents
are objectively determined by specific indicator functions historically
1 See Millikan 1984, 2000, Dretske 1988, 1995, Papineau 1990, Cummins 1994,
Neander 1995, Jacob 1999.
2 That is, for example, a view preferred by Dretske, Neander and Jacob. It is the
information-based teleo-semantics.
The Good, the Bad and the Neutral
311
established through evolution at the level of species. Thus, to schematize
things as briefly as possible, we have the following picture:
1) A token of a perceptual state derives its semantic properties from the type it
is a token of: that type of state has the function of indicating certain
environmental circumstances.
2) The type to which the token belongs derives its semantic properties from the
mechanism that has the function of producing states of that type.
3) The mechanism that produces that type of state derives its “content-giving”
function from evolutionary selection, which has maintained that mechanism as
a phylogenetic trait of the species because the states produced by it indicated
certain environmental conditions (the discrimination of which was somehow
relevant for acting in a survival-preserving way).
Hence, representational functions depend on biological functions, and
biological functions depend on the history of evolutionary selection. The
content represented by a certain token is the environmental condition in
the presence of which the mechanism producing the type has the
evolutionarily acquired function of producing a token of that type.
The teleo-functional proposal on perceptual content has the important
virtue of accounting for the possibility of misrepresentation, unlike nonteleological causal-informational semantics:3 the content is not fixed by
whatever may cause the state but rather by the condition the state has the
function of indicating. Normally the mechanism produces a token of its
type in presence of the “right” cause, but it can be improperly activated by
a different cause. Consequently, the content of an occurring state can very
well not be exemplified and so the state can be incorrect/inaccurate/false.4
If the content were fixed by whatever may cause the state, each token
would be a priori accurate/correct/true so misrepresentation would be
3 The classical informational semantics is that articulated by Dretske 1981 and
Fodor 1983. Whilst Dretske has realized the insufficiency of causal semantics and
has integrated the causal component into a teleo-semantic theory (Dretske 1986,
1988, 1995), Fodor is skeptic about the teleo-semantic approaches. See Fodor
1990.
4 Suppose a token of the type has the function of indicating [red] is actually caused
by an instance of [green]. It can be a misrepresentation because its content is not
fixed by whatever may cause it – in that case it would have the disjunction [red or
green] as (a part of its) content – but only by that sub-set of potential causes the
state has the function of indicating. That is known as the “disjunction problem”,
which may be overcome by a causal-informational semantics only insofar as a
teleo-functional component is added to the view. Causal roles alone cannot
generate the right normativity for the property of misrepresenting to be consistency
applicable. On this, see Jacob 1999, 78-140.
312
Chapter Seven Part II
impossible. In fact, perceptual states are normally caused by their
correctness-conditions, that is, their contents, but they don’t have to be.
That they are mostly caused by their correctness-condition is not a
mysterious pre-established harmony; rather, it is accounted for by the
selectional logic of function-acquisition. The very reason a mechanism
that produces that type of state has been acquired is that the occurrences of
this type of state are reliably correlated with the presence of a certain
property (which can be said to be “the content” of these states only once
the function has been established: then some tokens of the state can be
false). That means that the acquisition of a function by a mechanism
originally depends on a real relation between certain internal states and
those environmental properties that are now the contents of these states. A
function of φ-ing is acquired by a mechanism because the mechanism has
been successful in φ-ing, and φ-ing is relevant for survival and fitness.
That is why a function is defined by reference to its successful exercises
not its failed ones, even though failed exercises can well occur. But again,
they are failed exercises of that function just because the function is
defined by reference to its successful exercises.
To sum up this fundamental point: past successes in φ-ing are why a
mechanism that φ-es acquires the function of φ-ing. When φ-ing produces
certain states S in presence of certain conditions F, what is to be expected
is that tokens of S are normally caused by F; i.e., by their contents. The
reason a perceptual state has the content it has is the very same as a
mechanism having the function of producing states representing the
content selected in the first place.
With this in mind, let us first state a general principle about Good/Bad
asymmetry:
Principle of dependence (PD): There is a conceptual, explanatory and natural
asymmetrical dependence of the Bad cases on the Good cases.
A misrepresentation is such because it is a token of a type having the
function of indicating something else than the property that caused the
misrepresentation itself. A failed exercise of a perceptual function is what
it is because of the functional type it is a token of, so its characterization
and definition are conceptually, explanatorily and in primis naturally
(“metaphysically”) dependent on the function of its type, which is
determined by the cases in which the function is successfully exercised.
The occurrence of a state with content C, when C is not exemplified, is the
occurrence of a state whose type-individuation involves reference to a
circumstance in which C is exemplified, namely, to a Good token of the
same functional type. The content consists of the conditions under which
The Good, the Bad and the Neutral
313
the state would be correct so it is specified by a situation in which a state
of that type is caused by what it is “supposed” to be caused by. The Bad
case is analyzed by reference to the Good case, and not the other way
around. A Neutral case is not Good or Bad, but is ultimately to be
characterized only by reference to its Good counterpart.
Consider a hallucination of something being F. The content of that
experiential state cannot be specified other than by reference to how the
world would be were the state not a hallucination but a veridical
perception—a state caused by its content and representing it, and through
which a subject would become conscious of something perceived and of
the real way it is, of its being F.
Functions—be they successfully or unsuccessfully exercised—are
positive capacities and general dispositions. Perceptual functions are also
perceptual capacities. As such, their unsuccessful exercises are essentially
“privative” episodes, failures of the function of which they are an exercise.
The metaphysical relation between a function and its exercises is that
between a general positive capacity and the particular occurrences of that
capacity; some realize the very what-it-is of that capacity (e.g., φ-ing),
some do not, even though they are still exercises of it.
A functional disposition can be exercised recursively, and some tokens
of the functional type are failed tokens of their type—they do not do what
they are “supposed” to do as exercises of their function. Therefore, what
do a veridical and a correspondent deceptive experience have in common?
a) They are both occurrences of the same type, exercises of the very same
functional type.
b) One is a failed exercise of its functional type; the other is a successful
exercise.
What is common to them is absolutely relevant for their description,
their explanation, their very nature as mental episodes: they share a certain
representational function specified by their content. It is an essential
feature of a token-exercise of an objective bio-function, its being a tokenexercise of that function. It is not a contingent feature of it but a
maximally relevant one.
As we have seen, Martin introduces the idea of a “fundamental kind”
each of our perceptual experiences should fall into, and holds that Hs and
VPs cannot share a fundamental kind.5 I do not want to discuss subtle
issues concerning the (suspicious) essentialist metaphysics of “fundamental
5 The Aristotelian idea, according to which any item must fall in one and only one
fundamental kind, has been deeply articulated by Wiggins 1980.
314
Chapter Seven Part II
kinds” but a plausible and non-ideological way of kind-individuating
mental states like perceptual experiences, especially if we stay within a
broadly naturalistic framework, is that of appealing to the natural functions
of these states, something that is observer-independent,6 explains why
these states occur to us at all and why they have the features they have.
So, the taxonomy of perceptual states and their subsumption under
types and subtypes is an empirical affair dependent on facts about
evolution and natural history. A heart may have a morphology similar to
that of a closed fist, perhaps a color similar to that of a lung, it may make a
certain noise and so on, but such features are totally irrelevant for kindindividuating it. Rather, it is essentially defined by its bio-function, which
is that of pumping blood through the circulatory system. Any other
apparent feature of it has to be explained with respect to its fundamental
function: for biological organs, states or traits, their function just tells us
what they are; they are what they are insofar as they have a certain natural
function. This is also true for perceptual states.
A malfunctioning heart that does not pump blood is still a heart—only
in being a heart is it properly characterizable as malfunctioning. Likewise,
a successful and a correspondent deceptive experience have the same
function, and that also accounts for the fact that one is the deceptive
counterpart of the other. If perceptual states are functional states, as it is
very reasonable to hold, radical DJ must be false.
Instead, a moderately disjunctive treatment of the Good and the Bad—
one that does not rule out that the Good and the Bad share positive and
relevant properties—could be spelled out as follows:
Teleo-functional DJ: Either the state/process/event S is an actualization of
the function it has been produced to realize, or S is a failed exercise of that
function, i.e. it only seems an actualization of the function φ but it does not
perform φ.
Perceptual teleo-functional DJ: Either PE is an occurrent exercise that
successfully realizes the positive function/capacity of making one
conscious of certain environmental circumstances through representing
them accurately, or PE is an unsuccessful exercise of the same
function/capacity that only seems to successfully perform the capacity of
which it is a (failed) exercise.
6 Even if some deny that biological functions are objective and observerindependent (for example, Dennett 1978, 1991), among the philosophers of
biology and the biologists themselves functional realism is overwhelmingly the
mainstream view. On functions, see Allen/Bekoff/Lauder 1998, Buller 1999,
Ariew/Cummins/Perlman 2009.
The Good, the Bad and the Neutral
315
This exclusive disjunction embeds the idea that the Good is
definitionally, conceptually and explanatorily before the Bad; the idea that
the Bad is to be characterized just by reference to the Good. The Bad is a
failed example of what the Good is the paradigmatic and normative
example of, yet the Bad seems and “looks like” the Good and may be
indiscriminable from it.
In addition, the disjunctive treatment of PE in such terms also embeds
the idea that the Neutral is not an autonomously characterizable situation;
rather, its ascription rests on the subject’s condition of not being in a
position to tell whether she is in presence of the Good or the Bad disjunct.
There is no definitional, conceptual or explanatory independence of the
Neutral case so a fortiori there is no priority of the Neutral over the Good
and the Bad cases as in the conjunctivist views.
The asymmetric dependence of both the Bad and the Neutral on the
Good is compatible with the fact that the Good shares with the Bad
positive relevant features (functional facts above all),7 those by virtue of
which it is not a mystery that the Bad may seem an instance of the Good
until the point where it is subjectively indiscriminable. More on this
below.
II.2.2 Perceptual Capacities Realize Relational Functions
A perception is a certain kind of causal and cognitive relation to the
surrounding environment. Perceptual capacities are an example of those
functions that put the organism in relation to the environment by being
activated by a certain condition outside the perceiver. A perceptual state is
something that happens inside as a consequence of something that happens
outside, which the perceptual state itself, in being caused by it, somehow
indicates or represents.
In Chapter Two, I argued that visual perception is a real relation to the
environment, the seeing-relation (see also Chapter One), which involves a
representation of a perceived object as being such and so where these
properties are represented under a phenomenologically salient mode.
I have also argued that perceptual Content is object-dependent, so
perceptual experiences are de re states with a demonstrative Content. The
Content is made out of the object plus the content, where “content” (lower
7 As we will see, these functional facts, in the case of conscious perceptual
experiences, can be reflected in phenomenology. In fact, our very perceptual
phenomenology has the function of making us conscious (in certain distinctive
ways) of the representational states we are in, so of the environmental features they
represent when successful.
316
Chapter Seven Part II
case) means: the properties the PE represents the object as having (under a
mode). As a consequence, what makes a PE accurate or inaccurate is
whether—and to what extent—the properties the PE represents its object
as having are properties that the perceived object actually has. Being
essentially object-dependent, the PE’s Content and the kind it belongs to
are dependent on causal and external factors; they do not depend only on
subjective factors. So, for the version of the CV I have been articulating so
far, the Cartesian principle must be false.
⌐ CP: It is not true that if two mental states are subjectively indiscriminable;
they are mental states of the same nature, kind and type.
PEs are object-involving relational states so a state subjectively
indiscriminable from a certain PE obtained by the removal of the PE’s
distal object would not be a PE anymore. For the same reason, the
accessibility principle (see Chapter Three, Part II.5) must also be false.
⌐ AP: It is not true that if two experiences are subjectively indiscriminable
they must have the same Content.
Subjective indiscriminability does not even entail sameness in Content.
Change the object with another visually indiscriminable object and the
Content will change: the PE will represent another object as having certain
properties; the Content of the PE, being a de re content, would involve
another res. Remove the object completely and you will not have a PE
anymore, since the Content of a PE is essentially relational and objectinvolving.
But subjective indiscriminability—contra radical DJ—is grounded on
a positive phenomenal character that the indiscriminable experiences
share. Therefore, the phenomenal character cannot be determined by the
object; rather, it is determined by the properties the (putatively) perceived
object is represented as having and by the mode under which these
properties are represented (e.g., visual mode with its distinctive
phenomenology), according to the following scheme:
content + object
= Content
mode + content = phenomenal character
Therefore, the mode does not determine the Content,8 and the object does
not determine the phenomenal character.
8 It may be said to “determine” it only in an indirect sense insofar as, for example,
the visual mode determines that the represented properties cannot be sounds or
smells.
The Good, the Bad and the Neutral
317
Another principle that my version of CV must reject—so that, again, it
can embed another demand advanced by disjunctivists—is the proximate
cause principle (PCP):
⌐ PCP: It is not true that a mental event E1 is of the same kind as a mental
event E2 if E1 is brought about by the same kind of proximate cause as E2.
Again, the existence of a distal cause (an object) is essential for a certain
experience to be a perceptual experience, namely, a relational event. A
relational event and a non-relational event could have the same kind of
proximate causes but still they will not have to be mental states of the
same kind insofar as nothing will ever make the subjective appearance of a
perceptual relation to a distal object into a genuine perceptual relation if
the distal object or cause is removed. But sameness in kind of proximal
causes is compatible with absence of distal causes; therefore, that
sameness cannot be sufficient for making two mental states identical in
kind.
To sum up, subjective indiscriminability between two experiences is
not sufficient either for their sameness in kind, type or nature (CP is false)
or for their sameness in Content (AP is false), nor is sameness in kind of
proximate causes sufficient for sameness in kind, type and nature (PCP is
false). But indiscriminability is grounded in sameness of positive
phenomenal character, which depends on the mode and on the properties
the experiential state represents the object as being, where there is an
object at all, or the “putative” object where there is no object. Before
saying something about that puzzling introduction to the notion of a
“putative object,” I want to remark that this version of CV does represent a
form of moderate DJ but the relevant disjunction needs to be spelled out
not by contrasting veridical perceptions against hallucinations but by
contrasting genuine perceptual experiences (veridical and illusory) against
hallucinations, which are not perceptual experiences even though they
subjectively “look” to be such. Let’s see.
II.3 Where Do We Put Illusions?
Until now, we have considered the disjunctivist proposal as a view that
contrasts hallucinations (H) with veridical perceptions (VP). Tertium
datur, though: indeed, perceptual illusions, or inaccurate perceptions, are
neither hallucinations nor veridical perceptions. Let’s label them I, and
insert them between H and VP.
Chapter Seven Part II
318
Perceptual experiences are not intra-mental events; they are rather
subject-environment relations. Perceiving is being in causal and cognitive
contact with an external object: this is also entailed by the fact that the
perceptual object enters into the accuracy-conditions and so provides the
PE with a demonstrative semantic content. The object is the target of the
perceptual representation, and its having or not having the properties
represented by the PE of it is what makes the PE accurate or not. It is
through being of the object O that the PE is evaluable for accuracy with
respect to the represented property F. A PE is essentially about that
environmental object that is its distal cause. Therefore, such an aboutness
is causally and contextually determined, and makes the PE of its object
true of false.
Now, an illusory perceptual state is a perceptual experience that is still
successful in being a genuine perceptual relation but unsuccessful in
representing a property of the object it is a genuine relation to. For
example, an illusory visual experience is a case of “seeing-O” that
involves a certain discrimination-relation with the environmental object
that causes that very state (see Chapter One), but it is a case of
unsuccessful representation of the seen object as being in ways that
happen to be different from the ways this object actually is.
Now for Martin and others, the disjunctive treatment of PE should
contrast H and I with VP, namely, deceptive or Bad cases with Good
cases. Yet there is Bad and Bad: illusions and hallucinations are radically
different types of mental states—the first is a real relation with an
environmental object; the second is not a real relation at all but just a
subjective appearance of that. As a consequence, the right and relevant
disjunction should not be
a) PE: Either VP or (H, I)
but, rather
b) Either PE (VP, I) or H.
Whereas the first contrasts veridical perceptions with deceptive perceptions
independently of whether deceptive perceptions are hallucinations or
illusions, the second contrasts relational experiences with non-relational
experiences independently of whether the relational experiences (PE) are
deceptive or not. These contrasts are very different ones, and give rise to
very different sorts of disjunctivism.
According to b—for which I am arguing in favor of—hallucinations
are not perceptual experiences but illusions are. An illusion makes a
The Good, the Bad and the Neutral
319
wrong comment on a real perceptual target whilst a hallucination only
seems to make a comment on a real perceptual target but it is “targetless”;
it is only an appearance of a relation. In fact, it has no objectual target.
Consider a VP, an H and an I that are “correspondent” and subjectively
indiscriminable, so they share their phenomenal character. How are they
fundamentally different?
The VP is a real relation with an environmental object that it represents
as being the way it is, so it is successful both in making the subject
visually conscious of a seen object and in making the subject visually
conscious of its properties (by accurately representing these properties).
So VP is successful in a double sense. It is an example of the Good as a
real environmental-representational relation and as the particular
environmental relation it is, namely, as a representation that accurately
represents its environmental target: the content is exemplified.
The I is still a real relation with the environmental object but it
represents the object as having properties it does not have. So, it is
successful in making the subject visually conscious of an object and in
representing certain properties as of the object it is (successfully) about,
but it is an example of the Bad in being the particular representational
relation it is, namely, in having content that is not exemplified by the
perceived object.
The H is not a real relation with any environmental object so it is an
example of the Bad in failing at being a subject-environment relation and
in representing things as they are not. Not only is it just an apparent
relation to an environmental object but it represents the “putative” object
(it is an apparent relation with) as different from the way any real object in
the surroundings actually is.9
Accordingly, I and H, with respect to VP, involve different levels of
introspective mistakes. A VP introspectively seems to be a real relation
with a perceived object having the properties it appears to have in the PE.
Since the VP is a relation to a perceived object, and the object is as it
appears to be in the PE, a VP normally gives rise to two levels of
introspective success: the VP “looks” the way it is insofar as it looks like a
perceptual relation, and insofar as it looks like a successful perceptual
relation, namely the properties it attributes to the object are the right ones.
An illusion involves an introspective mistake because it subjectively
seems a VP, namely, a successful perceptual relation that attributes to the
9 Here I do not consider the case of veridical hallucinations. Normally, hallucinations
make us represent the surrounding world as different from the way it is, even if an
extreme coincidence could occur in which we hallucinate a world that happens to
be identical to the real world we are actually facing.
320
Chapter Seven Part II
perceived objects properties the object has. It is not what it seems because
the apparent properties of the object are not properties it actually has. Yet
an illusion involves introspective success because it subjectively looks like
a real relation to a perceived object, and from this point of view an illusion
looks just like what it is.
A hallucination involves two levels of introspective mistakes. On the
one hand, it subjectively looks like a perception (a genuine relation to an
environmental object) without being one; on the other hand, it subjectively
looks like a veridical perception, a perceptual relation that accurately
attributes certain properties to an object it is putatively related to, but it is
not as accurate as it seems.
An H, an I and a VP may well be subjectively indiscriminable but their
phenomenal character is radically deceiving in the case of H (an H
misleads us about the deep nature of the state we are in); it is moderately
deceiving in the case of I (an I misleads us about the accuracy of the state
but not about the deep nature of the state we are in), and it is not deceiving
at all in the case of VP. VPs are generally what they seem to be, perceptual
relations to an object that present to us the actual properties of that object.
We may then contrast I with VP or PE (I, VP) with H, but these
contrasts are very different ones.
1) Illusory/Veridical Disjunction:
Either PE is a subject-environment relation that successfully represents its
object, or it is a subject-environment relation unsuccessful in enabling the
subject to access (some of) the perceivable properties of the object, even if it
introspectively seems to be successful.
2) Perceptual Experience/Hallucination Disjunction:
Either E is a perceptual relation to the environment (be it accurate [=VP] or
inaccurate [=I]) or it is a state that seems to be a perceptual relation with the
environment but it is not [=H].
It is 2, not 1, that is the most significant disjunction. An accurate and
an inaccurate PE are the same type of relation—they have the same
function of representing an environmental target and they are successful in
doing it, although only VP does it accurately.10
10 That the right disjunction is the perception/hallucination one rather than the
veridical/deceptive one is embedded in the disjunctive treatment proposed by
Snowdon 1981, 202: “it looks to S as if there is an F: either [there is something
which looks to S to be F] or [it is to S as if there is something which looks to him
(S) to be F].” The first disjunct expresses a de re seeming whereas in the second
The Good, the Bad and the Neutral
321
Above all, we need to contrast relational states with non-relational
states. Only the first are genuine perceptual experiences, even if the
second kind may be indiscriminable from the first and have a misleading
relational introspective appearance. Disjunctivists are right in saying that
hallucinations have a different nature from veridical experiences but on the
one hand such a difference in nature does not prevent them from sharing a
positive phenomenal character with perceptual experiences pace radical
DJ; and on the other, the difference in deep nature as mental kinds is not
between veridical experiences and deceptive experiences in general but
between relational experiences (veridical and illusory perceptions) and
non-relational experiences (hallucinations). The point is relationality rather
than deceptiveness.
II.4 A Problem: The “Function” of Hallucinations
I have argued that mental states should be teleo-functionally typeindividuated so a functional identity is sufficient for sameness in type.
Since phenomenal character tracks intentional content (see Chapter Five)11
and content ultimately depends on teleo-functions, a sameness of functions
could account for the sameness of phenomenal character on which
indiscriminability is grounded.
On the other hand, I have proposed a disjunctive treatment of
perceptual experiences and hallucinations so the indiscriminability and the
commonality of phenomenal character between PE and H cannot be
sufficient for sameness in type. But a problem arises: given that a) typeindividuation should be based on teleo-functions and b) the phenomenal
character is also dependent on teleo-functions, wouldn't it be incoherent to
say that PEs and Hs belong to different types of mental states on the one
hand and that they may share their phenomenal character on the other? Do
PEs and Hs have the very same teleo-function then? If so, shouldn’t we
reject any disjunctive treatment of them?
The objection could be formulated as follows: granted that any
taxonomy and type-individuation of mental states should be based on
disjunct, the object falls under the scope of the seeming-operator. Note that the
first disjunct would be satisfied also by an illusory experience.
11 Remember that by content (lower case) I mean the properties the object is
represented as having in the PE. The object is the other component of the Content
(upper case), but the object does not determine phenomenal character. While the
Content is object-dependent, that is, de re content, the phenomenal character is not.
322
Chapter Seven Part II
teleo-functional factors,12 presumably also hallucinatory states have the
double function of
being a perceptual-representational relation with environmental objects, and
being an accurate and veridical relation with such environmental objects.
Both can be considered to entertain a determinable-determinate relation.
Although both of them are frustrated functions of H, I have argued so far
that being a successful or a failed example of a certain teleo-function does
not prevent a token being typed according to its function. Indeed, what
type-individuates a teleo-functional state is what it is supposed to do, not
what it actually does. Therefore, we have a problem in justifying why
hallucinations and perceptual relations should be told apart, even if they
share their teleo-functional type, since this seems inconsistent with the
teleo-functional account we have used to criticize radical and antiintentionalist forms of disjunctivism. Wasn’t sharing the most relevant
feature included in function-sharing?
The problem should be faced first by coming back to the very notion of
relational teleo-function.13 A relational function is doing something when
some external conditions occur. Perceptual functions are relational
functions par excellence because they just represent certain external
conditions when they occur. More precisely, given an object that causes
the perception and is its target, the perceptual episode represents it as
being such and so, say F and G.
A relational function is just a potential function before being applied to
the concrete context it is supposed to be set off by. Only the application of
the function to a certain external circumstance, context or domain of
reality can be considered a genuine exercise of it. Before being exerted, a
function is just an abstract disposition, a potential cognitive reaction to
certain circumstances. Now, it is completely different for a mechanism to
be activated without any pertinent circumstance at all, or to be activated by
the wrong circumstances. A wrong relation is not an absent relation.
Moreover, the environmental context provides the “adaptors” that
make the general function an adapted function, a function applied to a
concrete domain of reality. Let me give some examples.
12 At least, any taxonomy of perceptual states. One can doubt the efficacy of
naturalization based on a teleo-semantic account of thoughts, beliefs and
propositional attitudes in general but perceptual states are to be considered more
evidently, and less problematically, as produced by functions that are a byproduct
of evolution.
13 This notion has been articulated by Millikan 1984.
The Good, the Bad and the Neutral
323
Consider the mimetic function of a chameleon. His skin has the general
function of [becoming the color of the background]. Given a certain
background—say, a certain shade of green—the general function is
adapted to the concrete context and becomes the function [becoming that
shade of green]. Provided that the evaluation of accuracy for such a
function depends on the fine-graininess of the chromatic adaptability (that
depends on the very function of the chromatic adaptation, which is that of
not having been spotted by certain predators having a certain visual acuity
and normally being at certain distances and so on), the exercise of that
general function, once adapted to the concrete contexts, can be successful
or unsuccessful. Suppose the chameleon becomes yellowish and makes
itself more noticeable, then the token-exercise of that general function is
inaccurate because its accuracy is measured by the contextual domain of
reality. It is measured by the objective color of the background the
function is a function of “becoming-like.” The background in which the
animal is actually located is the occasional target of the mimetic function,
that which measures the success of that relational function. The colorchange is a real relation between the chameleon’s skin and the color of its
background. Given some other conditions—noise, danger, unknown
environment—the target of that function [becoming the same color of]
causes the activation of the function itself; it is an environmental input.
The skin has to become like that. Now, imagine that the mimetic
mechanism is activated without any relation to the surroundings, not as a
reaction to any environmental stimulus but just because of some internal
malfunction. The activation of a certain relational mechanism without the
presence of any relatum to which the mechanism is normally supposed to
react (suppose it is triggered during the night) would not be a genuine
exercise of the relational function. An exercise of a relational function, be
it Bad or Good, would be a genuine relation with an environmental target,
with a standard domain of application. Generally speaking, the animal is in
a “mistaken” or incorrect state because a mechanism is triggered for no
reason, perhaps with useless dissipation of energy and increase of risks.
But that probably would not be a wrong exercise of the function φ insofar
as it is a relational function that is exerted only when applied to a certain
target-domain. Rather, the state is a state that would be an exercise (failed
or successful) of φ if there were a causal context working as an “adaptor.”
Its context of application would also be the context of normative
evaluation for this exercise. Without causal context, without a proper
environmental cause/input/target, the state is not an inaccurate exercise of
the relational function; rather, it is the absence of the exercise of a function
despite the respective mechanism being triggered, which produces tokens
324
Chapter Seven Part II
of the type normally having the function of φ-ing. A failure at being a
relational function is something radically different from a genuine relation
that is a failure (as a relation). We should not confuse an unsuccessful
relation with failure at being a relation. The second is a mistake of a higher
order; it is not a mistake as a normatively evaluable relation or as a
reaction to a certain condition that is not as it should be. Here there is no
reaction to external conditions at all, neither proper nor improper. It is a
sort of meta-functional mistake, so to speak.
Consider the digesting function. The digestive system is a very
complex mechanism composed of many sub-mechanisms that cooperate
by implementing sub-functions that together enable digestion. It is the
function of metabolizing or processing incoming food. The digestive
apparatus does not have the function of being a relation to the food
ingested by the animal, though. Given a certain amount and type of food,
the digestive apparatus has the function of processing it. Given the target,
the function is that of reacting to the target in certain appropriate ways.
The digesting-relation involves the presence of food as its causal
triggering context; there is no digesting function of being a relation with
the food. Rather, to repeat the fundamental point, once there is incoming
food, a dedicated mechanism has the function of processing it. Imagine
that the digestive process is activated without reason, when no food is
coming in: broadly speaking, such a situation would be a “mistake” of the
mechanism of course, but by no means could this situation be compared to
one in which the incoming food is being processed in a wrong or
inaccurate way. In the latter case, the food is the real, normal relatum, the
domain of reality to which digestion applies. But an activation of the same
process without incoming food is not a digestive process at all; it is not a
genuine exercise of the digestive function, neither a Good nor a Bad one.
There is no relation with any food so no relational function is exercised;
rather, a mechanism, which should have been activated only to exercise a
relational function in presence of a certain relatum, is improperly
activated.
Now consider perceptual experience in the same vein. Perception is a
relational function such that, given an environmental object that causes the
perception, it has the function of representing it as being a certain way.
Perceptual functions do not include the function of being a causal relation.
Rather, given the causal relation with the relatum (an object), perceptual
states have the function of being a Good relatum, so to speak; that of
representing their target accurately. That is why hallucinations are not
perceptual experiences even if they have some functional analogies with
them (analogies plausibly responsible for the sameness in phenomenal
The Good, the Bad and the Neutral
325
character). No colors, no mimicry; no food, no digestion; no
environmental object, no perceptual experiences full stop.
Illusions and veridical perceptions possess a genuine “environmental
intentionality,” an aboutness determined by them being real perceptual
relations to worldly objects. Since their Content is demonstrative, their
object measures their veridicality so their having an object is the reason
they are semantically evaluable at all. It is about their target, that they
make certain “comments,” and it is the way their target is that makes these
comments veridical or falsidical.
Despite their subjective indiscriminability and commonality of
phenomenal character, perceptual experiences (illusory or veridical) and
hallucinations do not possess the same nature. They do share certain
functions though. Hallucinations are also states which “should” be
triggered by a perceived object so they “should” be perceptual relations,
but they aren’t; they are non-applied functions, non-demonstrative states.
They exhibit the appearance of perceptual relations because the original
function of those types of states should have been that of representing the
surrounding environment. This only potential commonality of basic
functions accounts for their commonality of conscious phenomenal
character, or at least this seems a very good working hypothesis.
Without entering into this delicate topic, I suggest it may be reasonable
to take our capacity of having perceptual states with a conscious character
(hence subjectively accessible and introspectable) as a byproduct of our
evolution. Even if it is far from clear why certain states have certain
phenomenal characters and other states have others, it is highly plausible
that the tracking relation between our phenomenal character and the
objective content of our perceptual states depends on the value of our
conscious access to our perceptual states for our lives, for our success in
acting, planning, reasoning and so on. Conscious experience tracks
representational functions “for us.”
In fact, in the case of hallucinations, the phenomenal character is still
made out of representational properties but nothing environmental is
represented as having these properties. That opens what I have called (in
Chapter Three, Part III.5) the problem of the semantic gap, which I will
address now.
326
Chapter Seven Part II
II.5 The Semantic Gap of Hallucinatory Contents
II.5.1 Introducing the Problem
Hallucinations exhibit a presentational phenomenology just like veridical
experiences (and illusions). Their conscious character of “felt reality”
needs to be given an explanation within an intentionalist view. If
hallucinating is sensorily entertaining a content that is not exemplified,
what is then exemplified or at least “present before the mind” when one is
hallucinating? What is the hallucinatory phenomenal character made of?
What do we attend to in introspecting an H?14
Besides this phenomenological issue, there is a related semantical issue
concerning the very content of hallucination. A very peculiar semantic
structure, different from that of perceptual experiences, needs to be
assigned to Hs. Indeed, according to the singularity thesis I have defended
(see Chapter Three, Part II.4), the Content of perceptual experiences
contains worldly particulars so it is de re demonstrative content. But
hallucinations are not perceptual experiences, even if they seem to be such
because of their tricky PE-like phenomenology. They have no object so
they must be taken to have no Content but rather just content. There are
properties represented but there is no worldly object these properties are
represented as belonging to. So, if PE’s Content is object-dependent and
H’s Content is not, what determines the accuracy-conditions of Hs? Of
what are hallucinatory comments veridical or falsidical, what do they
match or mismatch if not with an object?15
In fact, hallucinations intuitively seem to be mistaken, inaccurate
states. In order to be such, they must have a semantic content, but their
content cannot have the object-dependent semantic structure of a PE’s
Content. What makes Hs inaccurate or falsidical, if they are such at all,
provided that the accuracy of PEs can be semantically evaluated only
because their content is de re? If the singularity thesis is true for PE, we
14 See Smith 2002, 224-5: “We need to be able to account for the perceptual
attention that may be present in hallucination. A hallucinating subject may, for
example, be mentally focusing on another element in a hallucinated scene, and
then in another, describing in minute detail what he is aware of. In what sense is
this merely “mock”? […] The sensory features of the situation need to be
accounted for. How can this be done if such subjects are denied an object of
awareness?” On the topic of attention in hallucination, see also the remarks of
Johnston 2004.
15 The scheme proposed in Chapter Three, Part 2.5 makes the problem visible.
Hallucinations have no Content.
The Good, the Bad and the Neutral
327
have an asymmetry with hallucinatory contents that needs to be accounted
for. How can objectless sensory content be semantically evaluable? If
hallucinations have instead a general and not object-involving content,
why shouldn’t the generality thesis hold also for PEs? This is the semantic
gap problem, which I will address in what follows.
II.5.2 Perceptual Content and Perceptual Awareness
Intuitively an H that shared its conscious character with your current
experience would share with it also your current phenomenology of “felt
reality” (Smith 2002, Siegel 2008) or “bodily presence” (Husserl 1900).
Hs have a presentational phenomenology, just like VPs and Is. This
intuition can be made explicit and embedded into a principle we can call
the principle of item awareness, following Pautz 2007.
Item Awareness (IA): When S has a visual experience, there is something of
which S is aware.16
IA may lead to sense data theory if what S is aware of is identified
with a mental object.17 Given the possibility of illusions and
hallucinations, what S is always aware of when having a visual experience
cannot be mind-independent objects and properties, since in illusions some
properties you seem to see are not actually there and in hallucinations the
very object you seem to see as having certain properties is not there either
(so neither the putative object nor “its” putative properties are there). By
generalizing the case of H and I, sense data theorists conclude by
generalizing that in visually experiencing, we are always (at least directly)
16 Actually I consider “visual” a determinate of the determinable “perceptual” so a
hallucination cannot be a visual experience insofar as it is not a perceptual
experience. Here by visual it is meant “as of visual” or having the subjective
character normal bona fide visual experiences have. Visual experiences are seeingepisodes indeed.
17 IA is not identical to the principle Robinson calls phenomenal principle (PP):
“If there sensibly appears to a subject to be something which possess a particular
sensible quality then there is something of which the subject is aware which
possess that sensible quality” (1994, 31). IA just says that in having a sensible
appearance, you are aware of something; PP says that if you have an experience as
if something has a certain property, there must be an object that actually possesses
that property. PP naturally leads to sense data; IA is neutral about whether the
“something” you are aware of is an object having the property it seems to you
something has in experiencing or something else.
328
Chapter Seven Part II
aware of sense data, even in cases we are veridically perceiving (see
Chapter Two, Part1.1.3).
As already noted,18 CV has been proposed—first in the belief theory
version—as a way out from the unsatisfactory alternatives of sense data
theories and direct realism. If visual hallucinations also make us aware of
something (IA), then what visual experiences as such make us aware of
cannot be mind-independent items; since they cannot be sense data
either—given their extravagant ontology, their indeterminacy and other
problematic properties—what visual experiences make us aware of must
be intimately connected with the represented content, which, for the
representationalist version of CV, determines the phenomenal character of
the visual experience, namely, its being a state of conscious awareness.
A visual experience purports to present the subject with objects and
features. Such a purporting-to-present nature of visual experience is
analyzed by CV in terms of representing things as being a certain way. But
the content of a veridical experience is not its object. The object of your
VP is the perceived worldly object you become aware of, together with its
properties (those your VP veridically attributes to it). That object is part of
the object-dependent Content, so in consciously entertaining the Content
of a PE, we do become aware of its object as well as the certain properties
it possesses (those properties your PE veridically represents it as having =
the content of the PE).
According to certain versions of intentionalism, the items you are
aware of in H are particulars that do not exist, so-called Meinongian
objects.19 When visually hallucinating a pink rat, you do see a pink rat, a
bona fide particular you can demonstratively refer to, attend to, explore
and so on; it is just that it does not exist, which is why you are said to be
hallucinating it instead of perceiving it. So, there is a particular pink rat
you are aware of, though a non-existent one. The singularity thesis can be
straightforwardly maintained if we go Meinongian. Visual experience
always involves particulars, but sometimes—when we hallucinate—these
particulars do not exist. Existence is a property that only certain particulars
possess.
18 See this Chapter, I.2.1; Chapter Two, I.3.
19 According to Meinongianism (see Meinong 1904, Parsons 1980), there are
some objects that exist and some that don’t: intentional objects, imagined objects,
desired objects, fictional objects are of the second kind. A hallucination has a
Meinongian object; it is about a particular that does not exist. The Meinongian
ontology has been applied to solve the puzzle of hallucination first by Grossman
1974 (131ff.) and then by Smith 2002, Chapter 8.
The Good, the Bad and the Neutral
329
I do not want to discuss in too much detail the difficulties of this
extravagant solution. It is not clear to me whether such a version of CV is
really different from sense data theory. Is the non-existent rat minddependent or mind-independent? If it is mind-independent (so it is not a
sense datum), is it causing my very experience of it, as existent objects of
perceptual experiences are expected to do? Could a non-existent rat ever
cause anything? In addition, Meinongian particulars inherit all the puzzling
problems sense data carry along; for example their indeterminacy,20 their
bizarre ontological status21 and so on.22 Introducing an ad hoc ontology
does not look that promising.
Smith courageously advocates the Meinongian solution but he adds
that the awareness-relation of H with non-existent particulars should be
understood not ontologically or semantically but phenomenologically. The
restriction seems to be very reasonable but at the same time it apparently
makes the proposal a re-description of the problem rather than a solution:
the “phenomenology of singularity” exhibited by hallucinations is not
under discussion. No-one denies that in H it is phenomenologically as if
we become aware of particulars, but that is the problem itself, not the
solution. Given that we seem to be aware of something when
hallucinating, what are we aware of in fact? If the answer is that we are
aware of Meinongian particulars phenomenologically meant, are we
saying something more than hallucinations have a presentational
phenomenology in which it seems to the subject it is becoming aware of
seen particulars? Moreover, I am trying to address the semantic gap
problem, which is interwoven with the phenomenological issue but
concerns the gappy Contents of hallucinations and their apparent lack of
evaluability. If a PE is accurate/inaccurate only insofar as it has a worldly
object—that may or may not be the way the PE represents it to be—how
20 Smith 2002, 247 gets rid of the problem by pointing that Meinongian objects do
not exist so we should not worry about their indeterminacy. I find that answer
unsatisfactory. After all, positing non-existing objects as objects of awareness in
hallucination does not avoid the original puzzle: how can one be aware of
something that does not exist? For arguments in favour of Meinongianism, see also
McGinn 2004.
21 Where are they, for example? One could reply, “Well, they are nowhere, since
they do not exist, so where is the problem?” The problem is exactly that if they do
not exist, they cannot do the ontological and semantic job of being genuine relata
of a subject-world relation. They are neither “in” the mind (otherwise they would
be sense data) nor in the world (otherwise they would exist and have causal
powers).
22 A criticism of the Meinongian solution to the puzzle of the hallucinatory object
is to be found in Pautz 2007.
330
Chapter Seven Part II
can a Meinongian object make the hallucination accuracy-evaluable at all?
Ex hypothesi the Meinongian object is the way the H represents it to be
(the non-existent rat is pink). If that non-existent particular did the
semantic job the perceived object does in perceptual experiences (veridical
or illusory), then Hs would be de re states with the same semantic
structure as PEs, but unfortunately they would be always true. Indeed, a
PE is veridical when its object is the way the PE represents it as being. But
the Meinongian hallucinated rat is pink so the respective hallucination, as
with any other H, is accurate and veridical!
Such an outrageous result leads me to conclude that Meinongianism on
hallucinatory contents is either a truistic escamotage for hiding a problem
by re-stating it another way or it is a bad solution since it does not justify
the intuitive inaccuracy of hallucinatory contents.
Now, the problem of the semantic gap for H is how can an H have
accuracy-conditions at all if it is objectless, since at least no normal
worldly object is there to measure the accuracy of the H; i.e., since H lacks
an object whose actual properties could be matched or mismatched by the
“comments” the H makes? On what object-topic does the H make its
comments? On anything at all?
The item awareness issue for H is what are we aware of in H, if by
definition we are not aware of any worldly object? The H’s lack of an
object must be, somehow, consistently reconciled with the phenomenological
appearance that in H you are aware of “something” after all (item
awareness issue) as well as with the intuition that Hs have fully-fledged
accuracy-conditions that make them evaluable as inaccurate (semantic gap
problem). Both phenomenologically and semantically, “something” seems
to do the job the worldly object does in PE (veridical or not).
Of course, there is an intimate connection between the topic of item
awareness in hallucinations and the topic of hallucinatory contents. Before
addressing the semantic gap problem, let us then consider the item
awareness issue further because as we shall soon see it will be of help.
Having left behind the Meinongian proposal as patently inadequate,
another more promising idea is that when hallucinating we become aware
of properties if not of particular objects. That is a more attractive option.23
So the principle of item awareness holds, but in H we are aware just of
properties and never of particular objects. Let us call it the property view.
When veridically perceiving, we become aware of the perceived
objects as well as of their properties through representing them accurately.
23 See Dretske 1999, Johnston 2004, Tye 2005. The view is deeply analyzed and
discussed by Pautz 2007.
The Good, the Bad and the Neutral
331
The perceived object is the way our experience represents it to be, so
veridically perceiving it makes us aware of it and of the properties it
actually instantiates. VPs are episodes of object-awareness and propertyawareness.
When having an illusion, we become aware of a perceived object but
(some of) the properties that it looks to have according to our PE are not
instantiated by it. Now, are we aware of the properties the perceived
object looks to have, even if these properties are not instantiated by the
object?
When hallucinating, it is as if we are seeing a particular, and as if we
are becoming aware of certain properties of that putative particular.
Provided that we are not aware of any particular, are we aware at least of
the properties our H attributes to the putative object it purports to present
us with? In the illusory and hallucinatory cases, there are properties our
experience represents that are not instantiated by any perceived object. In
the illusory case, the experience represents the perceived object as having
such properties, so it is inaccurate and falsidical because of the way the
worldly object actually is. In the hallucinatory case, we also have the
semantic problem of justifying the inaccuracy. Illusions pose a
phenomenological problem (item awareness); hallucinations pose a
semantic problem (item awareness and semantic gap) as well.
Thus, even if we take it that in H we are aware of properties—
following the property view—the semantic gap problem would not
disappear as a result. However, this view about item awareness in H could
be a way toward a satisfactory treatment of it. Let us see if this is the case.
According to Dretske, Tye and Johnston,24 in hallucinations we are
aware of a “cluster of uninstantiated properties” (Dretske 1999, 102), a set
of “properties with no bearer” (Tye 2005, 169), “of uninstantiated sensible
profiles” (Johnston 2004, 135). The idea—which can be spelled out in
different ways25—is that in H we are not aware of objects but the
phenomenal character of “felt reality” depends on a complex of properties
our H represents as being instantiated, and of which we are aware. Such
properties are uninstantiated universals (Dretske 1999), which
misleadingly seem to us to be instantiated by something. So, in H we are
aware of no particulars at all but only of universals or uninstantiated
properties, while it just seems to us that we are aware of particulars. Our H
subjectively seems to be a PE. A cluster of uninstantiated properties we
are phenomenally aware of (e.g., [red], [round], [small], [smooth] and so
24 See also Foster 2000 and Forrest 2005.
25 See Pautz 2007 for a careful consideration of the many options on the market.
332
Chapter Seven Part II
on) can make us wrongly take our state as the factive awareness of a
particular apple, but there is no apple there: the hallucinated “apple” is not
a particular—not even a Meinongian particular—but a cluster of
uninstantiated properties. I add that such clusters must be thought of as
structured: visually hallucinating a red circle on a green square is not the
same as hallucinating a green circle on a red square, even if the two Hs
involve the same properties [red], [green], [circle], and [square]. The two
Hs makes us aware respectively of a different structured property complex,
even if these complexes are made out of the same (uninstantiated)
properties. The phenomenal character of an H is determined by the
properties the subject is aware of and by the specific way they are related
in the structured complex in which they are involved (and by the
distinctive mode; e.g., visual). Such structured complexes of universals are
all we are visually aware of in H. On the contrary, in PE we are aware also
of the particular objects that instantiate the properties, at least of those
objects our PEs represent as having certain properties. In illusions, we are
aware of genuine particulars and uninstantiated universals; in
hallucinations we are aware just of uninstantiated universals.
Such a view is consistent with the disjunctive treatment that contrasts
PE with H. Whilst in PEs we are genuinely related to worldly objects and
we become aware of these objects through perceiving them with their
properties, in Hs we only seem to be related to worldly particulars having
certain properties, therefore we seem to be aware of particulars but we are
aware only of structured complexes of uninstantiated properties. The
presentational phenomenology and the phenomenology of singularity
possessed by H are misleading features of H due to the fact that Hs seem
to be PEs.
I have argued that hallucinations have a positive phenomenal character
that they share with veridical and illusory perceptions, contrary to what
radical DJ holds. Since I have also argued for the singularity thesis,
hallucinations have an objectless, “gappy” content made out of intentional
properties, the properties that the hallucinatory experience, if it were a
perception, would represent its object as having. But Hs do not represent
any real object as having certain properties. Rather, in hallucinating, it
introspectively seems to us that we are aware of an object, and it also
seems that such a putative object is represented as having certain
properties. H’s phenomenal character is determined by these represented
properties; i.e., by the content. This content would be part of an objectdependent Content if the hallucination were a perception, if it were what it
subjectively seems to be.
The Good, the Bad and the Neutral
333
All this means that the phenomenal character does depend on
represented (structured complexes of) properties in a way that it is
independent of whether such properties are instantiated by particulars or
not. The phenomenal character of visual experience is then objectindependent, even if our experience is not neutral about whether there is
an object that has these properties. In visually experiencing, it does seem
to you that you are aware of particulars having these properties, but if you
are hallucinating it does wrongly seem so.
There is no use pretending that such an account of property-based
phenomenal character is not fully satisfactory. What does “being aware of
uninstantiated universals” mean? Is positing an awareness of uninstantiated
universals a phenomenologically apt account of phenomenal character of
hallucinations (as well as of perceptual experiences)? Can the “felt reality”
and the feeling of “bodily presence” be explained by the awareness of
universals? How can a mere awareness of universals exhibit a sensuous
character? Could I be aware of a universal [grayness] if my current PE
were a hallucination of my laptop? How can I be aware of something that
is not instantiated anywhere?26
However, this is at least a tentative explanation of the hallucinatory
phenomenal character and its indiscriminability from the phenomenal
character of veridical perceptions. Other views on perception are not in a
better position to explain this. As we have seen, DJ-cum-NR leaves such
indiscriminability as an insuperable mystery and has nothing substantive
to say about the positive conscious character of hallucinations; CV that is
committed to Meinongian objects is ontologically unsatisfactory and does
not account for the inaccuracy of hallucinations; CV that reduces
hallucinatory objects of awareness to beliefs—along the lines of the belief
theory—is even more phenomenologically implausible and so on.
Another radical solution is that of rejecting the very item awareness
principle (IA), namely, denying that whenever you have a visual
experience, you are aware of something. In ruling out particulars but
accepting uninstantiated properties as hallucinatory objects of awareness,
the property view tries to account for hallucinations in light of IA, which
does appear to be intuitively true. But if that intuition were unreliable, then
we would not need to commit to the idea that there must be some item we
are aware of in a hallucination. That is the radical view explicitly held by
26 We cannot say that the universal of [grey] is instantiated in my mind: at that
point, it would be natural to re-introduce a mental object that instantiates the
property, so rather than an uninstantiated universal it would be better thought of as
a property instantiated by a sense datum. We would be back to the sense data
theory with all its puzzles.
334
Chapter Seven Part II
Pautz 2007, and already endorsed by Evans 1982: in hallucination you are
not aware of any “item” at all, neither particulars nor properties.27 Where
does the prima facie plausibility of IA come from then?
According to Pautz, the intuition in favor of IA embeds the same
mistake into which sense data theorists were trapped: the need to
hypostatize the sensory content into “items” you are supposed to be in
genuine contact with. Instead of mental objects, the property view (as a
version of the CV that accepts IA) postulates that in H you are aware of
uninstantiated properties on the basis of the following reasoning: since
there are no worldly objects in H you are aware of (by definition), but you
must be aware of some “item” (IA), then you must be aware at least of the
properties your H represents a putative object as having. The reasoning is
the same as that which led to the position of sense data, even if now it
posits ontologically more plausible “items” (uninstantiated properties).
But sensorily entertaining content is not the same as being aware of
“items” that content is made up of. Having an H of a red and round object
amounts to sensorily entertaining a certain content involving the
properties [red] and [round], not to being aware of these properties.
Bearing a relation to content, for a given conscious state, does not amount
to bearing an awareness-relation to the properties that content is the
representation of. We are not aware of contents in the same way we are
aware of objects and properties. Objects and properties can be represented
by contentful states; they do not need to be objects of awareness for a
subject having those conscious contentful states.28 The relation of “being
represented” is not the relation of “being the object of awareness.” My
belief can represent unicorns as moving but I do not need to be aware
either of unicorns or of the property of movement in order to have the
belief. Rather, I need to entertain that content, so as to represent unicorns
and movement, but just as I am not aware of unicorns (they do not exist), I
am also not aware of their movement nor am I aware of the [movement]property even if I exert the respective concept.
27 See Evans 1982, 199ff: “[...] when a person hallucinates, so that it appears to
him he is confronting, say, a bus, then, whether or not he is taken in by the
appearances, there is literally nothing before his mind.” The appearance of having
something before the mind does not entail that there must be something before the
mind (not even something “mental”). Rather, “to hallucinate is precisely to be in a
condition in which it seems right to the hallucinating subject to say that he is
actually confronting something.”
28 Even though in case of perceptual contents the object is a constituent of the
Content itself.
The Good, the Bad and the Neutral
335
If we drop IA, then we can allow that in a hallucination we are not
aware of particulars or properties. Awareness is to be thought of as a
factive or implicative relation: if you are aware of X, then X is there as a
genuine object of your awareness. On the contrary, representation is not
factive or implicative. If your experience represents F, not only does it not
mean that F exists but nor does it mean that F is a genuine object of your
awareness.
So, hallucinated properties are not properties you are aware of; rather,
they are represented properties; properties your hallucination represents as
being instantiated and which then seem to you “items” you are aware of.
Just as the rat you hallucinate is not something you are aware of, its
pinkness is not something you are aware of, even if in hallucinating you
entertain a content involving the attribution of pinkness to something. The
phenomenal character of a VP makes you conscious of objects and
properties, even if they are not things you are sensorily conscious of; the
phenomenal character of an H fails to make you conscious of objects and
properties but it is not something you are sensorily conscious of. You do
not “see” your visual phenomenal character the way you see objects and
properties, or the way it phenomenally seems to you to see objects and
properties. On this basis, you self-ascribe certain content to your
experience but that self-ascription is done by looking at the world, or at
what seems to be the world to you according to your experience. In order
to self-ascribe an experience with certain content (made out of properties
represented as instantiated by a putative object), on the basis of being in a
certain conscious state, you do not need to see that conscious state itself;
i.e., to be sensuously aware of the very same conscious state of sensory
awareness.29
Instead of being aware of items in H, we consciously entertain content
that can be ascribed to our experience insofar as that experience has a
certain phenomenal character. Our being in a state with a certain conscious
character puts us in a position to self-ascribe a state with given content,
but we do not “see” the phenomenal character of our experience with an
inner eye. Since we do not come to grasp the content of our experience by
innerly perceiving the very experiences we have, there is no need to think
that in order to consciously entertain content, we must be aware of the
29 As Evans writes “There is no informational state which stands to the internal
state as that internal state stands to the state of the world” (Evans 1982, 227-8).
When it visually seems to you as if things are such and so, you self-ascribe a
conscious perception with content but you do not “see” either your own perception
or its content, rather you self-ascribe your experience on the basis of what it seems
to you to be there.
336
Chapter Seven Part II
properties that our experience represents as instantiated by something.30
Rather, we are aware of these properties only when the experience is
veridical so the properties are instantiated by a perceived object and the
experience makes us factively conscious of them.
An experience exhibits a distinctive phenomenal character: in having
the experience, you are in a state of awareness that has a phenomenal
character instead of being sensuously aware of your experience or of its
phenomenal character. You are aware of “items” (objects and properties)
only when they are exemplified; otherwise you enjoy a conscious state
with such phenomenal character that it puts you in a position to selfascribe an experience with certain content. Sensorily entertaining given
content, for an experience with a certain phenomenal character, does not
involve being aware of the properties and objects that would satisfy that
content were the experience veridical. Only when the content is satisfied
are you aware of such objects and properties. There is no need to be
“aware of uninstantiated properties” in having an H with a certain content;
you only need to be in a phenomenal state whose properties are the
properties your H represents as being instantiated. To repeat the point, the
properties your H represents as being instantiated do not need to be
properties you are aware of. If you become aware of the represented
properties of your hallucination, it is just because you self-ascribe an
experience with a certain content. But that self-awareness—the awareness
that you are having an experience with a certain phenomenal character and
with a certain content—is not a sensory awareness anymore; rather, it is a
conceptualization of your sensory state and involves belief that you are in
a state with such and such properties.31 The phenomenal character can well
be made out of intentional properties given under a mode—see Chapter
Three, Part Three—even if one is not aware of these intentional properties
and of the mode itself. Rather, these properties given under a mode are
what make your state a conscious state with a certain phenomenal
character, not something you are sensorily conscious of.
To conclude this survey of the available options: if we accept IA, the
“items” we are aware of in H can be either particulars or properties or
both. Particulars could be sense data or Meinongian objects, so within the
CV if H involves awareness of particulars, it must involve awareness of
Meinongian non-existent particulars. But Meinongianism on hallucinatory
content is implausible, not only because Meinongian objects are
30 That difference between “sensorily entertaining a content” and “being aware of
items” is articulated by Pautz 2007.
31 On the view on perceptual introspection, see Evans 1982, Chapter 7; Dretske
1995, Chapter 2; Tye 1995.
The Good, the Bad and the Neutral
337
ontologically problematic, just as classical sense data are, but also because
such a view would entail that hallucinations are a priori accurate, which is
absurd. The only plausible option that remains is that the items we are
aware of in H are properties. Since in H we do not see anything actually
having the properties we are aware of, these properties that H makes us
aware of must be structured complexes of uninstantiated universals. This
view is phenomenologically unsatisfactory and poses some problems. It is
not clear what it could be for a sensuous state to involve the awareness of,
say, the universal [red] and [round]. Nonetheless, no other view on
perceptual experience does any better so this perfectible account is better
than nothing.
Another possibility is that of denying the IA principle and holding that
in Hs, we are not aware of anything, neither particulars nor properties.
Rather, we consciously entertain content such that we are in a position to
self-ascribe an experience which, if it were veridical, would involve the
awareness of certain objects and properties. This view entails a rejection of
the perception-model of perceptual introspection. In introspecting your
perceptual experience, you do not become sensorily aware of the
phenomenal character of your own experience—rather, your experience is
a conscious state because it exhibits a phenomenal character, and you are
in that state—but you judgmentally self-ascribe an experience with a
certain character and content on the basis of how things seem to be to you
on the basis of that experience. That means that being in a sensory state
with a certain conscious character and content, and introspecting it, is not
being aware of that character and content as “items” you have a quasiperception of. You cannot perceive your own perceptual state, its character
and its content (where perceiving entails an awareness-relation); rather,
you can become judgmentally conscious of the fact that you are having a
visual experience with a certain conscious character and content (content
you entertain instead of being aware of it).
I find this account very convincing, although it entails a specific model
of perceptual introspection that could not be easily shared by many. In any
case, the only palatable alternative is that in H we are aware of
uninstantiated properties: the choice between the two alternatives also
depends on how we interpret the “awareness-relation.” If we allow that we
are aware of the contents of our experiences as well as of the properties
represented by the experiences as being instantiated by something, then we
can buy the property view, but we should keep in mind that we are
adopting a loose notion of awareness, which in any case is not the same
factive notion in veridical perceptions according to which we say we
become aware of worldly objects and properties. If we reject the IA
338
Chapter Seven Part II
principle, then we reserve the awareness of worldly objects and properties
so hallucinations do not make us aware of anything. Hallucinations seem
to be states of IA but they aren’t—they are deceptive about the world and
about themselves. That fits well with our moderately disjunctive version of
the CV. Either Hs seem to make us aware of particulars and instantiated
properties but they only make us aware of structured complexes of
uninstantiated universals (property view with IA), or Hs seem to make us
aware of particulars and properties instantiated by them but they do not
actually make us aware of anything at all (rejection of IA). Both ways are
plausible and present certain virtues; both ways leave us partially puzzled
and unsatisfied. I do not want to argue for either option. Our concern here
is more semantical than phenomenological.
Independently of which account one prefers, what remains to be
treated is the semantic gap problem. In both accounts, it is accepted that in
H we are not aware of worldly particulars so the content of H cannot be
object-dependent. How can H be accuracy-evaluable and inaccurate?
Here is a proposal compatible with our moderate disjunctivism:
hallucinations, strictly speaking, lack proprietary accuracy-conditions so
they are not inaccurate either; rather, they are neither accurate nor
inaccurate just because there is no worldly object they represent as having
the properties they represent, no object having properties that could be
matched with the hallucinatory content. Hallucinatory content is
essentially gappy—it is like a potential singular proposition where the box
for the singular term is empty. There is no truth- or accuracy-maker, in
other words. So there is not even inaccuracy. Hs make comments on no
particular worldly topic.
But on the other hand hallucinations have a cognitive role, just like
veridical experiences. They result in perceptual beliefs, and also in predoxastic stands towards the environment,32 which orient our behavior,
intentions, desires, planning, reasoning and so on. The cognitive effects of
a hallucination (due to the original function of this type of state)—their
32 Also, non-conceptual beings can hallucinate, and the apparent inaccuracy of
their state needs an account that does not appeal to beliefs. But there are cognitive
stands toward the environment that are pre-doxastic; for example, the acts of
“seeing-as” I investigated in Chapters One and Four. When hallucinating, animals
may have inaccurate stands toward the environment but their hallucinatory state is
not inaccurate in itself since its Content is gappy. Their mistake has to do with the
cognitive effects and consequences of the hallucinatory state: on the basis of their
H they come to represent the environment wrongly, but the hallucination in itself is
not a representation of any environmental object.
The Good, the Bad and the Neutral
339
causal and cognitive role in the economy of one’s mental life—are the
same as those of a certain veridical perception with the same content.
When I hallucinate a pink rat, there is nothing in the environment that I
mistake for a pink item and for a rat but I subjectively seem to be
presented with a particular, and my state subjectively seems to attribute to
“that” putative particular the [being a rat] and the [pink] properties. My H
is indiscriminable from a PE so I introspectively mistake my state for a
state of another nature. In particular, I mistake a non-relational, nonobject-dependent and non-accuracy-evaluable state for a relational,
accuracy-evaluable and object-dependent state. Because of that mistake, I
take a cognitive stand toward my surrounding environment—consisting of
beliefs and perhaps other more basic belief-like states—which is accuracyevaluable; it is inaccurate. These representations, which are typical
cognitive effects of the hallucinatory state, have a general content. For
example, “there is a pink rat there, which is moving in that direction.” Of
course I can believe a representation to be demonstratively referring to
“that” pink rat there, but my act is a failed act of demonstration grounded
on a false belief about the presence of a particular I am seeing. A failed act
of demonstration is not an act of demonstration; rather it is something I
take to be such but which fails at being such.
When having an H, we also form immediate (not inference-based)
beliefs that: a) we are having a VP; b) we are aware of particulars that are
presented as being such and so. Both beliefs are wrong—the one about the
nature of our state and the one about what there is in the world and how it
is arranged. But the hallucination is not inaccurate in itself; it is mistaken
for an accuracy-evaluable state because it is subjectively indiscriminable
from a state of that kind.
So, the intuition of the inaccuracy of hallucination is vindicated and
accounted for in two ways: 1) by arguing that it is a wrong intuition after
all, since Hs are not accuracy-evaluable despite seeming so; and 2) by
explaining why that intuition is so strong and apparently compelling. It is
such because a cascade of immediate and psychologically compelling
cognitive effects of H (beliefs and pre-doxastic cognitive stands) makes us
take the environment as being a certain way so we are in an inaccurate
state. The inaccuracy of the state is immediately dependent on the H, not
of the H itself.
To sum up the discussion above: in H, either we are aware of
uninstantiated properties or we are not aware of anything. Without having
to take a definitive stand toward these two options, what matters for the
semantic gap problem is that in any case hallucinatory states do not
involve awareness of particulars. Accordingly, Hs have gappy contents
340
Chapter Seven Part II
without objects, and that makes them not accuracy-available. The semantic
gap problem must be treated initially by recognizing that gappy nature of
hallucinatory contents, and then by making two more moves: first, we
must recognize that the intuition in favor of the inaccuracy of H is wrong,
and second, we need to account for the compelling appearance of the
intuition. The apparent inaccuracy of Hs depends on the cognitive effects
of them. They result in beliefs and other cognitive stands toward the
surrounding environment that are inaccurate. So, Hs are not inaccurate but
typically produce inaccurate states.
II.6 Beyond the Detachment Problem
II.6.1 The Content View and the Detachment Problem
In Part I.2 of this chapter, I introduced the detachment problem, and then I
argued that the different reasons adduced by most disjunctivists against the
CV—problems CV would entail and which allegedly could be solved only
by abandoning CV in favor of a non-CV version of DJ—are all amenable
to the detachment problem. I have grouped these reasons into
phenomenological, epistemological, semantical and metaphysical reasons.
Now that I have developed a moderately disjunctive version of the CV,
I will come back to these problems and argue that the above articulated
version of CV can well meet the demand for cognitive contact and so
avoid the various difficulties I have grouped under the idea of the
detachment problem. As a result, a moderately disjunctive version of CV
appears to be the most promising also in addressing those problems that
some disjunctivists wrongly take to be intrinsic to CV as such. The idea
that perceptual experience is representation does not prevent veridical
experiences from being episodes of genuine contact with the world; on the
contrary, the representational view on perceptual experience is the only
view that puts us in a position to satisfactorily explain how that contact
can be possible at all.
II.6.2 Phenomenology
The phenomenology of perceptual experience exhibits the properties of
transparency, actuality, immediacy and singularity. According to DJ,33 CV
33 In that context, with the label “DJ,” I will refer only to these forms of
disjunctivism (like radical disjunctivism cum naïve realism) that are incompatible
with CV, namely, the anti-intentionalist disjunctivism.
The Good, the Bad and the Neutral
341
cannot allow that perceptual phenomenology gets these properties right,
not even in the Good case. Indeed, if perceptual experience were
representational, it could not involve the genuine presentation of the
subject with the world, the direct contact to an actual world bodily present
and disclosed by the PE to the subject. If PEs were representations, they
could not be manifestative world-revealing episodes. Thanks to the
disjunctive treatment of the Good and the Bad, we should become direct
realists about the Good cases and abandon CV. In addition—so they
argue—not only is direct realism phenomenologically apt but it also
captures the commonsensical intuitions about what we seem to be doing
when we perceptually experience the world and have it simply “in view”
before us.
Now, transparency is compatible with an intentionalist version of CV
in which the phenomenal properties are properties represented as had by
something. I have argued for an impure intentionalism about phenomenal
character (see Chapter Five in which the phenomenal character is
determined by the represented properties plus the mode. In attending to
your PE, you will attend to the objects and features your PE purports to
present you with. CV analyzes that purporting-to-present relation in terms
of representing: the properties your PE represents the (putatively)
perceived object as having are the properties you attend to in introspecting
your PE.
Furthermore, attending to illusions and hallucinations is attending to
the purportedly presented objects and to the ways in which your
experience represents them as being. In the moderately disjunctive version
of CV I have been proposing, in H you only attend to represented
properties34 but you wrongly seem to attend to particulars that instantiate
these properties. So VP, I and H are transparent, but in introspecting VP,
you attend to the properties your experience veridically represents the
perceived object as having; while enjoying an illusion, you attend to the
properties your experience falsidically represents the object as having; and
in hallucination, you falsely seem to attend to properties instantiated by an
object you falsely seem to be in perceptual contact with.
It is true that perceptual phenomenology is presentational; it is also
true that in deceptive experience, our mistake is not only about the world
but also about our own experience. Illusions and hallucinations introspectively
34 If we buy the property view, according to which in H we are aware of
uninstantiated properties, then in H we attend to properties even if “attending to” is
used factively, as “being-aware of.” Otherwise, if we deny the item awareness
principle and at the same time interpret the “attending-to” relation as a factive one,
then in an H you seem to attend to properties but you do not attend to anything.
342
Chapter Seven Part II
seem to be episodes of presentation, but they aren’t.35 Veridical perceptions
seem to be episodes of presentation and they are such. The opposition
presentation/representation is a misleading and unnecessary opposition. In
fact, CV does not have to deny that in veridical perception we are
genuinely presented with the world. The representational episode the PE
consists of is just the way we manage to be presented with our surrounding
world.
As I have argued, the Good/Bad explanatory, conceptual and
metaphysical asymmetry is entailed at least by a teleo-semantic version of
CV (which I find particularly promising). The Bad case needs to be
characterized by reference to the Good case. A visual experience as of [a
being F] is an experience that looks just like an episode of seeing a being
F, so of being in “genuine contact” with the object a and the property F
instantiated by it. In a visual experience, successfully representing an
object as having a property simply amounts to being genuinely presented
with the object and its property. Representations are not veils between us
and the world; as a matter of fact, they are episodes that make it possible
for us to be in cognitive contact with the surrounding world.
Therefore, it is not true that CV must be an error theory with respect to
the phenomenology of immediacy, actuality and singularity. The apparent
immediacy or actuality of PE is not a systematic mistake; rather, it is a
mistake only when we have deceptive experiences, which do not do what
they seem to do. When I have a PE, it seems to me that I have the world in
view in its bodily presence manifested to me. That seeming is reliable
when my PE is veridical; it is also reliable when my PE involves an
illusion. Indeed, I am in real contact with environmental objects, although
in some respects it looks different from the way it is. When my PE
veridically represents the object O as being F, my PE manages to make me
aware of O as well as of F as instantiated by O. Veridical experiences
make me conscious of worldly objects and features. Their being
representational episodes does not prevent them from being factive
awareness-episodes. There is no reason for thinking that if perceptual
experience is a representational phenomenon, then it can make us only
indirectly aware of objects and properties. Representations are not mental
objects of awareness like sense data.
Disjunctivists characterize perceptual experiences as episodes of
“cognitive contact with the world” (McDowell 1982, 2008), as “states
which reveal the world we live in” (McDowell 1986), “fusions with the
35 To be more precise, illusions are episodes of presentation of an object, even if
the object is wrongly presented (at least in some respect).
The Good, the Bad and the Neutral
343
things themselves” (Sartwell 1995), ways of “having the surrounding
things in view” (Travis 2004). In experience, “facts make themselves
manifest” and “disclose themselves to the subject” (McDowell 1998,
2008) so our senses should be conceived as “windows” (Campbell 2002).
All these characterizations are interesting and emphatic metaphors, which
unfortunately are too seldom analyzed or explained by their proponents.36
Why could a representational episode not be a fusion with the world, a
way of revealing or manifesting the world to the subject? Why should our
sensory apparatuses not be like windows opened to the world itself if CV
is true?
Insofar as these proposals are conceived as incompatible with CV, it
seems that a misleading notion of representation is surreptitiously at work.
It is as if a CV-theorist had to hold that in PE we are directly conscious of
representations: if this were the case, we could be conscious of worldly
objects and features only indirectly.37 But as I have said, representations
are not like sense data; they are not direct objects of conscious awareness.
Rather, they are states that (can) make us conscious of worldly objects and
features. Far from being an interface between us and the world, perceptual
representations are states that enable us to have the world in view. Nothing
prevents a representational episode from being a presentational episode.
Sensorily entertaining exemplified content amounts to being conscious
of worldly objects and features that exemplify that content. Contents are
semantic properties of our conscious states; they are not the only objects
of awareness we are directly conscious of in PE. Instead, in veridical PEs
36 About the vagueness and metaphorical status of these typical characterizations,
Lowe (2000, 148ff.) and Burge (2005, 49) rightly complain.
37 For example, Campbell 2002, 188ff , proposes an opposition between the
“relational view” of PE and the representationalist view, and calls the first the
window-model of perceptual experience. Suppose you see a dagger through a pane
of glass: “To hold that the only way in which it can happen that you see a dagger
through a pane of glass is by having a representation of the glass appear on the
glass itself would plainly be a mistake.” It seems from that analogy that Campbell
thinks of representations as if they were sense-data-like entities, something like
pictures before the mind, such that we would see only the representations, the
effect of things on us rather than the things themselves. If that was
representationalism, then no doubt it should be abandoned. But CV is not to be
contrasted to the “relational view of experience” because it is compatible with the
idea that PEs are essentially relational, as we have seen. The metaphor of the pane
of glass is ill-conceived because the content of a conscious perceptual state is not
something you are aware of instead of the worldly objects and properties, in the
same way you would be aware of a representation of a dagger depicted on the glass
instead of the real dagger.
344
Chapter Seven Part II
we become conscious of the perceived world through being in such
contentful states as the PEs. When the content of a perceptual experience
is exemplified and the experience is in the right causal contact with its
object, the perceptual experience makes the subject “directly” conscious of
objects, aspects and features of the surrounding environment that is the
target or the experience and causes it.
The phenomenology of singularity is also accounted for by the
demonstrative structure of perceptual content, so by the singularity thesis:
perceptual content is de re; it is not general and existential so when we
attend to our PE, we attend to worldly objects “directly,” as objects we are
acquainted with and are available to us for comment, descriptions,
attributions or properties. Of this object here in view, my PE represents
this or that property. The perceptual object is there, made directly
available by the PE. So, perceptual experiences are genuine relations to the
world be they veridical or illusory. Only in H is the phenomenology of
singularity misleading because Hs introspectively seem to be de re states
without being such. This is the moderately disjunctive CV.
To sum up, it is not true that CV is not phenomenologically apt: the
jargon of “revelation,” “manifestation” and “disclosure” is interesting but
it does not explain much about what perceptual episodes are and how they
can be presentations at all. All these evocative ways of pointing to the
presentational nature of perceptual experience can be consistently
embedded into CV. Finally, CV can successfully vindicate the
phenomenology of transparency, immediacy and singularity that
distinctively characterizes our perceptual experiences.
II.6.3 Epistemology
Through perceptually experiencing, we acquire knowledge about the
surrounding world and form true and justified empirical beliefs. But if
what veridical perceptions make available to us were the same as what
illusions and hallucinations make available to us, not even veridical
experiences could justify our empirical beliefs and ground perceptual
knowledge.
This is why some disjunctivists argue that CV is epistemologically
inadequate; i.e., it falls short of accounting for perceptual knowledge.
Instead, they argue, in perception the world must be thought of as directly
available to us, not just as represented in an appearance-episode that is a
common factor sharable by certain illusions and hallucinations. If this
were the case, the justificatory power of our veridical experiences would
be as defective as that of correspondent illusions and hallucinations. In
The Good, the Bad and the Neutral
345
addition, the world itself would not make any direct contribution to our
empirical knowledge—apart from an exterior, causal contribution—which
is absurd. So CV necessarily opens the way to the skeptical challenge and
therefore it must be abandoned.
Now the polemical target of that criticism is a pure conjunctivist
version of CV, for which the fundamental nature of a perceptual
experience is that of being nothing more than a mere appearance—a
purely subjective state—which in certain cases is caused by the “right”
causes (VP), in other cases by the “wrong” causes (I) and even by no
external cause at all (H).
Even though the version of CV I have been defending does involve the
existence of a common factor (e.g., certain functional properties, a positive
phenomenal character), the common factor is not the only relevant element
for individuating and characterizing the states that share the factor itself.
Since the Cartesian principle is not true, it is not true that two
introspectively indiscriminable states have the same nature and are of
same mental kind. Since the accessibility principle is not true, it is not true
that two introspectively indiscriminable states have the same Content.
That is why hallucinations are not the same kind of states as perceptual
experiences, even if they share relevant properties. First, they do not make
available to the subject what perceptual experiences make available to the
subject. Indeed, a PE makes the subject conscious of worldly particulars,
and a veridical PE makes the subject conscious of worldly particulars and
their properties. So a PE makes the world itself available to the subject; an
H does not, even if it introspectively seems as to do so. Therefore, the idea
that the Good case cannot make available that which is made available by
the Bad case because otherwise the Good case would justify our belief no
more than the Bad case is not an objection to my version of CV. On the
contrary, the Good case is good exactly because it makes available to us
the surrounding world itself with its objects and features (by making us
conscious of them), whereas the Bad case does not make anything
available to us. The Bad case may be introspectively indiscriminable from
the Good case so I may not be in a position to know whether I am in a
Good state or in a Bad state. A skeptic may challenge this but no other
view will do any better in this sense: for example, a naïve realist à la
Martin must recognize that a hallucination can be subjectively
indiscriminable from a veridical perception. Even if the veridical case is a
direct manifestation of the world with a special justificatory power, I could
be in a second-order mistaken state in thinking that I am in a worlddisclosing state with an indefeasible justificatory power when instead I am
346
Chapter Seven Part II
in the Bad state.38 Thus, a moderately disjunctive version of CV entails a
difference in nature between perceptual experiences and hallucinations so
that a given experience is not just individuated by its being a certain “mere
appearance”: the nature of that “appearance” and the elements it factively
makes available for us are not just determined by its conscious character.
The subject could be deeply mistaken on the nature of her own experience
because things can be different from the way they seem, and perceptual
experiences are no exception.
The justificatory power of perceptual experiences essentially rests—
among other things—on their being genuine relations to the environment.
They can be such relations because their content is object-dependent, so
they are essentially episodes in which worldly particulars are made
available to the subject.
It is through perception that our empirical beliefs are originally
anchored to the world. Not only is such anchoring compatible with CV but
CV is also the best way to account for it. Only an object-dependent
content can guarantee the right “aboutness” for a belief to originally
concern the world, and the object-dependent content of perceptual beliefs
is a conceptualization and a judgmental endorsement of more basic,
object-dependent perceptual contents.
The epistemological facet of the detachment problem may constitute a
worry for certain particular versions of CV but a duly spelled-out version
of CV is actually the best way to cope with it.
II.6.4 Semantics
Our scheme of descriptive identification rests on the possibility of a nondescriptive acquaintance with the objects we identify, characterize and
know. Again, perceptual acquaintance with the world makes demonstrative
thoughts possible in the first place. Johnston, Campbell, Travis and others
hold that only the “revelatory” nature of perceptual experience could
account for our capacity to make de re judgments about perceived
particulars without any semantic intermediary (like representations)
between us and those particulars. Perceptions must be making particulars
available for immediate demonstration for de re judgments about them to
be possible at all.39 If perceptual experience did not bring such particulars
into view, not even the world could control our empirical judgments.
38 For this line of criticism, see Wright 2008.
39 According to Travis 2004, visual experience just “brings the world into view”
without involving any representation so that perceptual experiences are neither
accurate nor inaccurate; rather we are right or wrong in judging that what we see is
The Good, the Bad and the Neutral
347
I reply to these points by saying that all of this can and must be
accepted. None of these points are against CV as such: the singularity
thesis is precisely the ingredient of CV that makes the view semantically
apt in accounting for our direct reference to worldly particulars in
empirical judgments. The direct reference does rest on an immediate
availability of worldly particulars but this immediate availability is, again,
guaranteed by perceptual contents being de re contents. A PE makes
particulars immediately available for thought because its Content already
includes the object the PE represents as being such and so.
To recall the point made above, that a PE is a “direct presentation of
particulars” does not prevent PE from being a contentful state. It is exactly
by having a (exemplified) de re content that a PE manages to be the
conscious, direct presentation of a particular to the subject.
The singularity thesis also provides CV with the theoretical resources
to avoid the semantic facet of the detachment problem. Our perceptual
beliefs have de re particular-involving contents insofar as the perceptual
episodes on which they are grounded (and of which they are a cognitive
effect) have a particular-involving semantic Content.
such and so. He argues that looks as visible features cannot index a coherent
content because many different and incompatible things can share a look (a pig, the
half of a pig in view, a pig in the rear-view mirror, a wax pig, a hologram of a pig,
a hairless wild boar and on and on). But his argument does not work because first
he does not distinguish low-level contents, complex sensible profiles made out of
SCM-properties, from thicker contents like [being a pig], so from what the object
strikes you as being on the basis of your perception. SCM-looks are neither
inconsistent nor indeterminate: something can look pinkish, moving certain ways,
with this or that complex shape, at this or that distance and so on. The fact that
many different kinds of things can exhibit that objective sensible profile is another
matter: the sensible profile is represented determinately, and the visual
representation of it is accurate or not according to how the perceived object is in
fact. But also with respect to thick contents, Travis’s argument does not work: by
perceptual learning, I can acquire a visual recognitional disposition for a certain
type of object or for an individual even if the visible profile I represent could be
instantiated by other types of objects or other individuals. The fact that a robot or a
hologram could have the same visible profile as that of my mother does not entail
that when I recognize my mother just-by-looking, my experience does not
represent the object seen as being my mother but as being either my mother or a
hologram of my mother or a robot identical to my mother or whatever else one can
think of. Likewise, I can see x as a pig even if my visual experience could be
wrong in such a thick representation despite being accurate at the thin level of
SCM-properties: as I have argued, a very same scenario content is compatible with
different proto-propositional contents or seeing-as (see Chapter Four, Part 1).
348
Chapter Seven Part II
II.6.5 Metaphysics
According to the Martin-Fish disjunctivism-cum-naïve-realism, a veridical
experience is a state of a kind that could never be instantiated when
hallucinating. But CV entails that a state is fundamentally individuated by
its representational content, and that it can be shared by a veridical
perception and an indiscriminable hallucination. The content of a state
(namely something essential to its nature and kind) is individuated
independently of whether it is exemplified or not; i.e., whether the state is
accurate or inaccurate. So, for CV, it is argued, the fact that a veridical
experience is a genuine relation in which the subject is presented with the
surrounding world must be completely inessential to the nature of such a
state. Its being a cognitive contact is inessential to its nature so a VP is not
a world-involving state. CV must be a conjunctive view in which a
perceptual experience is individuated in isolation from the world. The
world is just externally added to the common factor—a merely
psychological, intra-mental element—to characterize veridical cases
without really conditioning their fundamental nature. So the mind is
separated from the world insofar as, even in the Good case, their relation is
merely causal rather than constitutive.
Therefore, CV, as a form of conjunctivism, is doomed to be trapped in
the detachment problem also with respect to the metaphysics of mental
states.
This line of criticism is grounded in good reasons but it should not be
addressed to CV as such, except for certain possible forms that could be
labeled as “purely conjunctive.”40
According to the moderately disjunctive version of CV, it is not true
that being a relation to mind-independent objects is inessential to the
nature of the perceptual experience as a mental state. The “mere
appearance” shared by veridical perceptions, illusions and hallucinations
does not individuate any common nature (DJ is right about that), although
it is a common property (radical DJ is wrong about that). The appearance
of a genuine relation to the world does not individuate anything else other
than a subjective-introspective commonality between something that is a
genuine relation and something that only seems to be a relation. H and PE
have a different “nature,” even if they may be subjectively mistaken one
40 For example, Crane’s version of CV (2006) is vulnerable to that objection
insofar as Crane holds that perceptual experiences are not relational states, and he
contrasts—differently from me—intentional with relational views on perceptual
experience.
The Good, the Bad and the Neutral
349
for the other. Indeed, hallucinations are not episodes that make us aware of
worldly objects and properties.
This means that my version of CV does not drive an unbridgeable
wedge between the mind and the world. The relationality of PE is essential
to it: that is why the contrast that really matters is the one between
hallucinations and perceptions (be they illusory or veridical).
I have argued that, if we are to attribute a “fundamental kind” to
mental states, such kind-individuation must track bio-functional facts. The
function of perceptual experience is a relational function: that of
accurately representing a perceived object. So a PE is a perceptual relation
with a determinate teleo-function. An H is a mental state of a type that
should have had the function of representing a perceived object accurately.
But a relational function without its proper environmental target or
relatum is not a function exerted inaccurately; rather it is the missed
exercise of a function, so an H can be only the appearance of a relational
function.
That is why, within the theoretical frame of CV, we need to commit to
a distinction in nature between hallucinations and perceptions; we need to
go moderately disjunctivist. Perceptions are essentially world-involving;
indeed their contents are de re content. The demand for cognitive contact
is met by CV, and the detachment problem is avoided by the moderately
disjunctive treatment of the relation between hallucinations and perceptual
experiences. They may have something in common, even something
relevant—phenomenal character and a “purported” teleo-function—but
that does not prevent hallucinations from being states of another kind,
namely, non-relational states. On the contrary, non-hallucinatory
perceptual states, be they illusions or veridical perceptions, are ecological
relations rather than merely psychological states. Perceptual states cannot
be detached from the world without losing their very nature.41
41 A criticism that could be moved to Martin and to other advocates of the
disjunction [hallucinations and illusions] vs. [veridical perceptions] is the
following: when having an illusion, we normally represent the object accurately in
some respects and inaccurately in other respects. That means that the very same
state (numerically the same) is a genuine “disclosure” of the world, on the one
hand, and a state of a totally different kind on the other. In fact, the state makes
certain properties manifest but falls short of making other properties manifest since
it is not a total illusion. Now, can we really conceive a state that is partially of a
manifestative kind and partially of that “spurious” kind which is subjectively
indistinguishable from the VP-kind? Since the perceptual mistake is normally
partial, radical DJ is at odds in accounting for a state that is partially veridical (for
certain properties) and partially falsidical (for other properties). Is the state an
instance of the Good or an instance of the Bad? Either answer is wrong, but the
350
Chapter Seven Part II
Concluding Remarks
In this chapter I have taken into consideration the disjunctivist (DJ)
objections to the content view about perceptual experience in order to
show that they are resistible. CV can cope with them.
In Part I, I first introduced the core idea of DJ, and individuated four
kinds of reasons typically provided in favor of it and often also provided
against CV. I suggested that such reasons (phenomenological,
epistemological, semantical, metaphysical) are all amenable to a basic
theoretical desideratum, that of avoiding what I have called the
detachment problem. The detachment problem is raised by any views on
perceptual experience that conceive of PEs as mental states that can be
type-individuated and characterized independently of their being or not
being a genuine relation to the world. Perceptual phenomenology is
presentational and apparently world-involving, a fact that would be best
explained by PEs really being genuine relations of presentation, at least in
the veridical cases. So, it is argued, presentations are not representations
since representations are not essentially world-involving: if veridical PEs
are also representations, they are what they are independently of being
genuine relations to the world, so CV entails that even veridical PEs are
“detached” from the world insofar as they do not essentially involve it.
Second, if a PE is a state that represents certain content that could be
veridical, illusory or hallucinatory, then a veridical perception would make
available to the subject what an illusion or a hallucination make available.
If so, a veridical PE could not provide reasons for perceptual knowledge
any more than an illusion or a hallucination. If we want PE to have an
epistemological role in justifying our perceptual knowledge, we must
believe that veridical PEs make available to us just certain represented
contents but they are relations to the world itself. That is the
epistemological facet of the detachment problem.
The semantic facet of the detachment problem goes like this: in order
to have demonstrative thoughts about the perceived world, perceptual
experience need not just represent the worldly particulars but rather make
them directly available for immediate demonstration to our perceptual
thoughts, beliefs and judgments. So, particulars need to be immediately
present, not represented. The metaphysics of mental states entailed in CV
need be conjunctivist metaphysics. Veridical experiences are mental states
individuated by their semantic properties, properties that can be shared by
Good and the Bad cannot be mixed since they radically differ in kind, so the view
cannot account for illusion.
The Good, the Bad and the Neutral
351
falsidical experiences. So, the relationality of veridical experience cannot
be fundamental to kind-individuate veridical PEs, and veridicality must be
taken as an external ingredient that does not determine the nature of the
mental state. On the contrary, DJ-cum-NR does consider veridical
experiences as relational states made out of the worldly elements that are
experienced and so “disclosed” to the subject, and deceptive experiences
will not be states of the same kind but they will be states subjectively
indiscriminable from states of this kind; i.e., from genuine manifestations.
In Part II, I argued that many critical points made by DJ can and must
be positively embedded into CV. Indeed, I have argued for a moderately
disjunctivist version of CV on the following lines. First, DJ is right in
taking the Good case (veridical perception) as basic and in characterizing
the Bad case by reference to the Good case. I have supposed that a teleosemantic version of CV is true (without positively arguing for that version)
in order to show that CV is consistent with an asymmetrical and
moderately disjunctivist treatment of the Good and the Bad. I then argued
that, within a naturalistic framework of the mind and mental properties,
the proper function of a mental state is what type-individuates it, on top of
determining its content when the state has the function of perceptually
representing the environment. Now, since the acquisition of a certain
representational function is a matter of evolutionary selection of a
respective mechanism, and the function was originally selected because it
was successful, a certain function so acquired is to be specified by
reference to its successful exercises. The Bad case is essentially a failed
exercise of the Good case, but not the other way around. Still, the Good
and the Bad have the same function so they share relevant properties: the
phenomenal character of PE is also plausibly related to the
representational function of that type of state, namely, to the semantic
function of the state.
Therefore, first of all, there is the conceptual, explanatory and natural
asymmetrical dependence of the Bad cases on the Good cases (principle of
dependence). Second, there are relevant common properties between the
Bad and the Good cases. They have the same function and the same
conscious character so radical disjunctivism is false as well as
incompatible with CV.
Nonetheless, I have suggested that the principles typically rejected by
disjunctivists are to be rejected also within a consistent version of CV.
First, it is false that if two states are subjectively indiscriminable, they
have the same nature, kind and type (Cartesian principle). Second, it is
false that if two experiences are subjectively indiscriminable, then they
have the same Content (accessibility principle). Third, it is false that
352
Chapter Seven Part II
mental event E1 is of the same kind as mental event E2 if E1 is brought
about by the same kind of proximate cause as E2 (proximate cause
principle). The moderately disjunctivist proposal is that of treating
perceptual experiences, illusory or veridical, as essentially relational
states, and to contrast them with hallucinatory states, which are only
subjectively indiscriminable from them but they are not relational states.
That perceptual experiences are relational states is entailed by the very
semantic structure of their Contents. Indeed, perceptual content is a de re
content, which includes a worldly, perceived object as a constituent.
Therefore, it is true that two states can have same subjective character, the
same kind of proximate causes, but a different “nature” (PEs are
essentially relations to the environment, Hs are not) and Contents (PEs
have de re contents, Hs do not because their contents are not objectdependent). In that view, the phenomenal character of an experience is
determined by the mode together with the properties the PE represents
something as having, but not by the object. Indeed, H is objectless but can
share the phenomenal character with a veridical experience because the
two states share the represented properties and the mode (e.g., visual). CV
must go (moderately) disjunctivist but the right disjunction is that between
perceptual experiences and hallucinations not that between veridical
experiences and deceptive experiences.
Additionally, I have pointed to a problem that seems to arise with
moderate disjunctivism and functional accounts of type-individuation of
mental states and the ascription of their contents. If mental states are typeindividuated, and plausibly hallucinations also are tokens of types of states
that have the function of accurately representing the environment a certain
way, why should hallucinations be contrasted with perceptual experiences
if an H and a PE share a basic teleo-function and functions are what typeindividuates mental states in the first place?
I have argued that such a problem can be coped with by appeal to a
closer reflection on the very notion of relational function. A function is
relational when it is supposed to be exercised toward a certain
environmental target. Perceptual functions are relational par excellence:
when a certain circumstance is given, a certain state is produced to
represent it (the circumstance) accurately. The object causing the
perception is not represented as it is, so the function is applied to an
environmental domain but unsuccessfully. But if there is no target at all,
no environmental domain to which the function is applied, then the
situation is better described as the failure to exert a relational function
rather than as an unsuccessful exercise of the function. Hallucinations are
states that should have been relational states but are not. Therefore, their
The Good, the Bad and the Neutral
353
proper function—that of representing their environmental target in an
accurate way—is not exercised at all, and a missing exercise of a relational
function due to the absence of a relatum is neither an accurate exercise nor
an inaccurate exercise; it is no exercise at all. Therefore, even if both H
and VP have the teleo-function of being a relation of accurate
representation of a certain environmental object, that function remains in
H only an abstract, potential, targetless function. Indeed, H is an
appearance of a representation of the world without really being about any
worldly object.
In fact, hallucinations raise big problems for CV. First, what are we
aware of when hallucinating if not of a worldly object? Second, if the
Content of PE is object-dependent (singularity thesis), how can objectless
states like hallucinations be semantically evaluable at all? And if they are
not semantically evaluable, where does the strong intuition that they are
inaccurate states come from?
The two questions are deeply interwoven; still they are not to be
conflated. I have addressed the first problem, which is not a problem
specifically for CV but for any view on perception. Given the intuition that
whenever we have a visual experience there is something we are aware of
(item awareness principle) and what we are aware of in H cannot be
mental objects like sense data —because of the well-known problems of
the sense data theory—two more alternatives are available. One is that in
H we are aware of Meinongian objects; i.e., of particulars having the
bizarre property of not existing. For example, when you hallucinate a pink
rat, you are aware of a genuine particular object, a rat; it is just that the rat
does not exist. I have argued that such an alternative is to be left aside
because such a suspicious ontology of non-existent particulars has the
same problems as the position of sense data (indeterminate objects,
impossible objects and so on). Above all, that is a hopeless solution for
treating H because in hallucinating a pink rat, the supposedly non-existent
particular we are aware of is pink, on top of being a rat. This means that
hallucinations are a priori true because their non-existent objects have the
properties the H represent them as having! We wanted to do justice to the
intuitive inaccuracy of hallucinations but our Meinongian treatment makes
all Hs true, so something must be wrong with this treatment.
The other alternative suggests that we are not aware of objects at all in
H; rather, we are aware of properties. Since no worldly object instantiates
these represented properties, we are aware of structured complexes of
uninstantiated properties, which are to be meant as universals. I have
pointed out that the property view is the best way to respect the item
awareness principle for Hs, but this view does not come without problems.
354
Chapter Seven Part II
It is not so easy, for example, to conceive of a sensuous and
phenomenologically salient awareness of universals. Could a phenomenal
character like the one your PE has just now be determined by your
awareness of universals alone? There is an issue of phenomenological
adequacy but there is not a more appealing proposal is actually in view.
Therefore, if in H we are aware of something——if the item awareness
principle holds—we must be aware of uninstantiated properties.
Another alternative is that of dropping the IA intuition and accepting
that in H we are not aware of anything at all—objects or properties
(neither instantiated nor uninstantiated). That is the Evans-Pautz proposal.
Sensorily entertaining content in a conscious experience, when the content
is made out of represented properties, does not amount to being aware of
these properties. Rather, being in a certain conscious state puts you in a
position to self-ascribe content but your PE’s having certain content—say,
representing [red] and [round]—entails that if your PE was veridical, you
would be factively aware of these instantiated properties. It does not entail
that you must be aware of these properties anyway but that they are
“uninstantiated.” The IA intuition comes from the wrong way to conceive
introspection and the relation between phenomenal character and content.
For a representationalist version of CV (as my view, see Chapter Five,
Part I), phenomenal properties are the properties your PE represents an
object as having (given under a mode, but that does not matter now) but a
perceptual state with a conscious phenomenal character and content is a
state you are in, it is not a state you must be aware of. In introspecting, you
may become aware of your own experience but that is not an inner
perception as if you “saw” your experience with its phenomenal character
and its related content. Phenomenal character is what makes your PE a
conscious visual state, not something you are visually conscious of. So,
visually entertaining content involving certain properties does not mean
being aware of these properties unless the PE is veridical, so it really
factively makes you aware of these properties and of the worldly object
that instantiates them. Call such a view on H the no-item awareness view.
I have not taken a definite stand toward the alternative between the
property view and the no-item awareness view. The choice between them
may depend also on a decision about how to stipulate the meaning of
“awareness-relation.” Anyway, this is a puzzling problem for any theory
of perception.
In any case, neither view solves the other problem I treated
successively, which is that of the semantic content of hallucinatory states.
I argued that the Content of PE is de re content, made out of the properties
the PE represents a perceived object as having, plus the worldly object the
The Good, the Bad and the Neutral
355
PE has both as its distal cause and its representational target. No worldly
target, no semantic evaluability.
But Hs are just states with gappy Content because properties are
represented, even under a mode (for example, visual) with its distinctive
phenomenology, but no object is there to make the content accurate or
inaccurate. So either we drop the singularity thesis or we take it that only
hallucinations have general and existential content, like [there is an object
which is pink and is a rat], so that Hs can be inaccurate. Since there are
good arguments for the singularity thesis (see Chapter Three, Part II), and
it is implausible that Hs and VPs, having the same potential function and
the same phenomenal character, have, the first, general and existential
content, and the second de re Content, I argued that hallucinations, despite
our intuitions, are not semantically evaluable so they are neither accurate
nor inaccurate in themselves. What is accuracy-evaluable, and inaccurate,
is not the hallucination as such but the beliefs and perhaps the more basic
belief-like states produced by hallucinations as immediate cognitive
effects of them. Hs subjectively seem demonstrative states so they seem to
be accuracy-evaluable states. Hs are not as accurate as they seem, but they
are not inaccurate either because they are not accuracy-evaluable. The
reason we have a strong intuition for their inaccuracy is that they have
false beliefs and falsidical cognitive stands toward the surrounding
environment as their natural and immediate effects.
After treating the problem of awareness in H and the problem of
hallucinatory contents, I came back to the detachment problem introduced
previously in order to show that the alleged problems that would weigh on
CV can be satisfactorily treated within the framework of the content view.
First, the phenomenology of singularity, actuality, immediacy and worldinvolving presentation does not speak against CV, nor must CV be an
error theory about perceptual phenomenology. I have tried to show that the
criticism rests on an unnecessary opposition between representation and
presentation. When a representational state like a PE is veridical, it does
make us aware of worldly objects and features so it enables the subject’s
capacity to be presented with the world. When the content of a PE is
exemplified and the PE is in the right causal contact with its target, then
the accurately represented world is a presented world. The singularity
thesis posits demonstrative contents so the PE’s content being originally
object-involving and world-involving fits very well with the
phenomenology of actuality, immediacy and singularity. So the version of
CV I have been articulating is phenomenologically apt.
Similar reflections can be made about the semantic and
epistemological facets of the detachment problem. It is not true that what a
356
Chapter Seven Part II
VP makes available to us is the same as what an H makes available to us
(a “mere appearance” or something like that). On the contrary, a VP makes
us aware of worldly objects and features, so it makes the surrounding
world available for us (through accurately representing it), whilst an H
does not make available for us any worldly objects and features. So it is
true that the justificatory power of a VP rests on it being a genuine relation
to the world rather than a “mere appearance” but that special power of the
Good case is not in conflict with CV. I have argued on similar lines that
the semantic worries about genuine references to the world are also easily
treated within CV. It is said that in order for our perceptual beliefs and
judgments to be about the perceived particulars, the role of perception
must be that of making worldly particulars immediately available for
reference and topics for empirical thought. But the singularity thesis just
guarantees that our perceptual beliefs and judgments are de re beliefs and
judgments. Indeed, our perceptual beliefs are de re beliefs because they
are grounded on de re perceptual contents in the first place, namely, on
particular-involving contents. There are no semantical concerns about the
detachment problem in my version of CV.
From a metaphysical point of view, I first argued that the only
plausible kind-individuation of mental states, at least within a naturalistic
framework, must rest on the (teleo-)functions of these states or, to put it
better, of the mechanism that produces the type of states a given state is a
token of. So, there is a “natural” divide between relational states
(representational functions exercised on an environmental target) and nonrelational states like hallucinations (representational states that lack a
target for being the genuine exercise of the function their type is supposed
to perform). Therefore, it is right that the kind of state I am having when
veridically perceiving an F is a type of state that could never be
instantiated when hallucinating an F. This is true—veridical states are
genuine relations so there is no metaphysical facet of the detachment
problem either. It is not true, in other words, that CV cannot give a special
metaphysical status to the Good case. Nothing, not even God, could make
a deceiving appearance of a perceptual relation (an H) into a genuine
perceptual relation (a PE). All that being said, veridical perceptions and
hallucinations have many relevant properties in common, for example,
their phenomenal character and certain teleo-functional features, so radical
disjunctivism—for which there is nothing relevant in common between the
veridical “manifestative” episodes and subjectively indiscriminable nonmanifestative hallucinations—is ultimately untenable.
CONCLUSIONS
The main points I have argued for in this book are summarized here:
• The content view on visual experience fits very well with ordinary
ascriptions of visual experiences and episodes. It also vindicates
our pre-theoretical intuitions about our experiences being accurate
or inaccurate, about things seeming/looking a certain way and so
forth.
• The content view itself, if duly articulated, embeds the virtues of
the classical belief theory, without inheriting its problems
(philosophical problems and incompatibility with experimental
data).
According to the global articulation of the content view I have
proposed:
• Perceptual content is two-layered, so it involves a “scenario” level
and a proto-propositional level (as originally argued by Peacocke).
• Perceptual content is singular and, de re, it is essentially an objectdependent content.
• Perceptual content is Russellian; it consists of environmental
properties and relations represented as possessed/entertained by
perceived objects so it is not double (Russellian and Fregean).
• The phenomenal character of our perceptual experiences is
determined by the represented properties and by the mode, but not
by the perceived object, which is represented as having the
properties (this is why it can be shared by veridical perceptions and
objectless states like hallucinations).
• This is an impure representationalism about phenomenal character:
it is impure because it entails that the conscious character is not
determined by the content alone but also by the mode. It is a form
of representationalism because the mode is a modality-dependent
way for a property to be represented, but that way is inseparable
from the respective property.
358
Conclusions
• Perceptual content is rich insofar as it may involve “thick”
properties beyond its “thin” properties (for visual contents: spatialchromatic-morphological properties). These thick properties can
come to enrich our perceptual contents through perceptual learning:
however, they may not be testified in the perceptual phenomenal
character.
• Perceptual content may well outstrip phenomenal content.
The content view as articulated is able to successfully cope with the
many facets of the detachment problem, the apparent wedge that the
notions of content and representation seem to drive between the mind and
the world: the disjunctivism-cum-naïve-realism is a bad and ultimately
untenable way to address the detachment problem. Not only is there no
need to abandon the content view in order to address it but a certain
version of the content view is the only way to address it properly. This
must be a moderately disjunctivist version of the content view, according
to which perceptual experiences (veridical or illusory) are to be told apart
from hallucinations: the former are genuine relations to the surrounding
world and may represent it accurately or inaccurately; the latter are not,
and—even if they may share a positive phenomenal character with
veridical perceptions—they have a gappy content that is, strictly speaking,
not even semantically evaluable. Nonetheless, hallucinations
systematically produce inaccurate beliefs about the surrounding world so
they are inaccurate states in a derivative sense.
Such points offer a plausible global picture that supports the content
view.
BIBLIOGRAPHY
Allen/Bekoff/Lauder 1998: Allen, C., Bekoff, M., Lauder, G. (ed. by),
Nature’s Purposes: Analyses of Function and Design in Biology,
Cambridge Mass., MIT Press, 1998
Algioti et al. 1995: Algioti, S., Goodale, M., Desouza, J.F.X., Size
Contrast Illusions Deceive the Eye but Not the Hand, «Current
Biology», 5, 1995, 679-685
Alston 1998: Alston, W., Perception and Conception, in K. R. Westphal
(ed. by), Pragmatism, Reason, and Norms, New York, Fordham
University Press, 1998, 59-87
—. 1999: Alston, W., Back to the Theory of Appearing, «Noûs»,
Supplement: Philosophical Perspectives, Epistemology, 1999, 181-203
—. 2005: Alston, W., Perception and Representation, «Philosophy and
Phenomenological Research», Vol. LXX, No. 2, 2005, 253-289
Anscombe 1965: Anscombe, G.E.M., The Intentionality of Sensation; A
Grammatical Feature, in R.J. Butler (ed. by), Analytical Philosophy,
Second Series, Oxford, Basil Blackwell, 1965, 55-75
Ariew/Cummins/Perlman 2009: Ariew, A., Cummins, R., Perlman, M. (ed.
by), Functions: New Essays in the Philosophy of Psychology and
Biology, Oxford, Oxford University Press, 200
Armstrong 1968: Armstrong, D.M., A Materialist Theory of Mind,
London, Routledge, 1968
—. 1991: Armstrong, D.M., Intentionality, Perception and Causality, in in
E. LePore and R. van Gulick (ed. by), John Searle and His Critics,
Oxford, Blackwell, 1991, 149-158
Austin 1961: Austin, J., A Plea for Excuses, in J. Austin, Collected Papers,
ed. by J. Urmson and G.J. Warnock, Oxford, OUP, 1961
—. 1962: Austin, J., Sense and Sensibilia, Oxford, Oxford Clarendon
Press, 1962
Ayer 1940: Ayer, A.J., Foundations of Empirical Knowledge, London, The
Macmillan Company, 1963 (ed. or. 1940)
Baars 1988: Baars, B., A Cognitive Theory of Consciousness, Cambridge,
Cambridge University Press, 1988
Bach 1997: Bach, K., Engineering the Mind. Review of Naturalizing the
Mind by Fred Dretske, «Philosophy and Phenomenological Research»,
360
Index of Authors
LVII (2), June 1997, 459-468
Back-y-Rita/Kercel 2003: Back-y-Rita, P., Kercel, S.W., Sensory
Substitution and Human-Machine Interface, «Trends in Cognitive
Neuroscience», 7 (12), 2003, 541-546
Bain 2003: Bain, D., Intentionalism and Pain, «Philosophical Quarterly»,
53, 213, 2003, 502-523
Baldwin 1990: Baldwin, T., G.E. Moore, London, Routledge, 1990
Bayne 2011: Bayne, T., Perception and the Reach of Phenomenal Content,
in Hawley/Macpherson 2011, 16-35
Beauvois & Salliant 1985: Beauvois, M.F., Salliant, B., Optic Aphasia for
Color and Color Agnosia, «Cognitive Neuropsychology» (2), 1985, 148
Ben Zeev 1988: Ben Zeev, A., Can Non-Pure Perception be Direct?, «The
Philosophical Quarterly», vol. 38, N. 152, 1988, 315-325
Bermudez 1994: Bermudez, J.L., Peacocke’s Argument Against the
Autonomy of Nonconceptual Content, «Mind and Language», vol. 9,
1994
—. 1995: Bermudez, J.L., Nonconceptual Content: From Perceptual
Experience to Subpersonal Computational States, in Gunther 2003,
183-215
—. 2007: Bermudez, J.L., Thinking Without Words, Oxford, Oxford
University Press, 2007
—. 2007: What is at Stake in the Debate About Nonconceptual Content?
«Stanford Encyclopedia of Philosophical Perspectives» 21, 1, 2007,
55-72
Biederman 1990: Biederman, I., Higher Level Vision, in
Osherson/Kosslyn/Hollerbach 1990, 41-71
Block 1990: Block, N., Inverted Earth (1990), reprinted in Block 2007b,
511-528
—. 1995: Block, N., On a Confusion about a Function of Consciousness,
«Behavioural and Brain Sciences», 18, 1995, 227-247
—. 2007: Block, Ned, Consciousness, Accessibility and the Mesh between
Psychology and Neuroscience, «Behavioral and Brain Sciences», 30,
2007, 481-199
—. 2007b: Block, N., Consciousness, Function and Representation:
Collected Papers, Cambridge Mass., MIT Press, 2007
Boghossian and Vellman: Boghossian, P., Vellman, J.D., Colors as a
Secondary Quality, «Mind», 98, January 1989, 81-103
Bozzi 1998: Bozzi, P., Vedere come. Commenti ai §§1-19 delle
Osservazioni sulla filosofia della psicologia di Wittgenstein, Milano,
Guerini & Associati, 1998
Defending the Content View of Perceptual Experience
361
Brandom 1994: Brandom, R., Making It Explicit: Reasoning, Representing
and Discursive Commitment, Cambridge Mass., Harvard University
Press, 1994
Brentano 1874: Brentano, F., Psychology from an Empirical Standpoint,
London, Routledge, 1995
Brewer 1999: Brewer, B., Perception and Belief, Oxford, Oxford
University Press, 1999
—. 2008: Brewer, B., How to Account for Illusion, in Haddock, A.,
Macpherson, F., Disjunctivism: Perception, Action, Knowledge,
Oxford, Oxford University Press, 2008
Broad 1925: Broad, C.D., The Mind and Its Place in Nature, London,
Routledge and Kegan Paul, 1925
Buller 1999: Buller, D. (ed. by), Function, Selection, and Design:
Philosophical Essays, Albany, NY, SUNY Press, 1999
Burge 1988: Burge, T., Individualism and the Mental, in Ludlow P., Martin
N. (ed. by), Externalism and Self-Knowledge, Stanford, CSLI
Publications, 1988, 21-84
—. 1991: Burge, T., Vision and Intentional Content, in E. LePore, R. Van
Gulik (ed. by), John Searle and his Critics, Oxford, Blackwell 1991,
195-213
—. 1998: Burge, T., Memory and Self-knowledge, in Ludlow, P., Martin,
N. (ed. by), Externalism and Self-knowledge, Stanford, CSLI
Publications, 351-370
—. 2005: Burge, T., Disjunctivism and Perceptual Psychology,
«Philosophical Topics», 33:1, 2005, 1-78
Butcharov 1980: Butcharov, P., Adverbial theories of consciousness. In
French, P.A., Uehling, T.E. & Wettstein, H (ed. by), Studies in
Epistemology. Midwest Studies in Philosophy, 5, Minneapolis:
University of Minnesota Press.
Butterfill 2011: Butterfill, S., Seeing Causings and Hearing Gestures, in
Hawley/Macpherson 2011, 36-59
Byrne 2001: Byrne, A., Intentionalism Defended, «The Philosophical
Review», 110, vol. 2, April 2001, 199-240
—. 2011: Byrne, A., Experience and Content, in Hawley/Macpherson
2011, 60-82
Byrne/Logue 2009: Byrne, A., Logue, H. (ed. by), Disjunctivism:
Contemporary Readings, Cambridge Mass., MIT Press, 2009
Byrne/Tye 2006: Byrne, A., Tye, M., Qualia Ain’t in the Head, «Nous», 40
(2), 2006, 241-255
Campbell 1994: Campbell, J., Past, Space and Self, Cambridge Mass.,
MIT Press, 1994
362
Index of Authors
—. 2002: Campbell, J., Reference and Consciousness, Oxford, Oxford
University Press, 2002
Carruthers 2000: Carruthers, P., Phenomenal Consciousness: A
Naturalistic Theory, Cambridge Mass., Cambridge University Press,
2000
Chalmers 1996: Chalmers, D., The Conscious Mind, Oxford, Oxford
University Press 1996
—. 2006: Chalmers, D., Perception and the Fall from Eden, in
Gendler/Howthorne 2006, 49-125
Cheesman/Merikle 1986: Cheesman, J., Merikle, P, Distinguishing
Conscious from Unconscious Perceptual Process, «Canadian Journal
of Psychology», 40, 1896, 343-367
Chisholm 1950: Chisholm, R., The Theory of Appearing, in Black, M. (ed.
by), Philosophical Analysis. A Collection of Essays, Ithaca, Cornell
University Press, 1950, 97-112
—. 1957: Chisholm, R., Perceiving. A Philosophical Study, NY, Cornell
University Press, 1957
—. 1966: Chisholm, R., Theory of Knowledge, Englewood Cliffs, NY,
Prentice Hall, 1966
Churchland 2005: Churchland, Paul, Chimerical Colors: Some
Phenomenological Predictions from Cognitive Neuroscience,
«Philosophical Psychology», 18, n. 5, 2005
Clark 2000: Clark, Austen, A Theory of Sentience, Oxford, Oxford
University Press, 2000
—. 2001: Clark, Andy, Visual Experience and Motor Action: Are the
Bonds Too Tight?, «Philosophical Review», 110:4, 2001, 495-519
Crane 1988: Crane, T., Concepts in Perception, «Analysis», vol. 48, n. 3,
June1988, 150-153
—. 1992: Crane, T., The Nonconceptual Content of Experience, in Crane,
T. (ed. by), The Contents of Experience: Essays on Perception,
Cambridge, Cambridge University Press, 1992, 1-21
—. 2003: Crane, T., The Waterfall Illusion, in Gunther, Y.H. (ed. by),
Essays on Nonconceptual Content, Cambridge Mass., MIT Press,
2003, 231-236
—. 2006: Crane, T., Is There a Perceptual Relation?, in
Gendler/Howthorne 2006, 126-146
—. 2009: Crane, T., The Given, in Schear, J.K. (ed. by), The Myth of the
Mental?, Cambridge Mass., Routledge, 2009 (available online)
—. 2011: Crane, T., Is Perception a Propositional Attitude?, in
Hawley/Macpherson 2011, 83-100
—. 2011b: Crane. T., Intentionalism (forthcoming), in McLaughlin, B.,
Defending the Content View of Perceptual Experience
363
Beckermann, A., Walter, S. (ed. by), Oxford Handbook of Philosophy
of Mind, Oxford, Oxford University Press, 474-493
Cowey 2010: Cowey, A., The Blindsight Saga, «Experimental Brain
Research», vol. 200, Issue 1, 3-24
Cummins 1975: Cummins, R., Functional Analysis, «Journal of
Philosophy», 72, 1975, 741-765
—. 1996: Cummins, R., Representations, Targets and Attitudes,
Cambridge Mass., MIT Press, 1996
—. 2002: Cummins, R., Neo-teleology, in Cummins, Ariew and Perlman
(ed. by), Functions: New Essays in The Philosophy of Biology and
Psychology, Oxford, Oxford University Press, 2002, 157-173
Cussins 1990: Cussins, A., The Connectionist Construction of Concepts, in
M. Boden (ed. by), The Philosophy of Artificial Intelligence, Oxford,
Oxford University Press, 1990, 368-440
Dancy 1995: Dancy, J., Arguments from Illusion, «Philosophical
Quarterly», 45, 1995, 421-438
—. 2004: Dancy, J. (ed. by), Perceptual Knowledge, Oxford, Oxford
University Press, 2004
Davidson 1973: Davidson, D., Radical Interpretation, «Dialectica», 27,
1973, 314–28
—. 1982: Davidson, D., Rational Animals, «Dialectica», 36, 1982, 317328
Davies 1991: Davies, M., Individualism and perceptual content, «Mind»
100, 1991, 461-484
—. 1992: Davies, M., Perceptual Content and Local Supervenience,
«Proceedings of the Aristotelian Society», 92, 1992, 21-45
—. 1997: Davies, M., Externalism and Experience, in Block, N.,
Flanagan, O, Guzeldere, C. (ed. by), The Nature of Consciousness.
Philosophical Debates, Cambridge Mass., MIT Press, 1997, 309-327
—. 2000: Davies, M., Interaction without Reduction: The Relationship
between Personal and Subpersonal Levels of Description, «Mind &
Society», vol. 1, 2000, 87-105
Dennett 1987: Dennett, D., The Intentional Stance, Cambridge Mass., MIT
Press 1987
—. 1991: Dennett, D., Consciousness Explained, London, Allen Lane
1991
—. 1991b: Dennett, D., Ways of Establishing Harmony, in McLaughlin
1991, 119-130
Dennett 2002: Dennett, D., Seeing Is Believing. Or Is It?, in Noë, A.,
Thompson, E. (ed. by), Vision and Mind: Selected Readings in
364
Index of Authors
Philosophy of Perception, Cambridge Mass., MIT Press, 481-496
Dretske 1969: Dretske, F., Seeing and Knowing, London, Routledge and
Kegan Paul, 1969
—. 1979: Dretske, F., Simple Seeing, in Dretske 2000, 97-112
—. 1981: Dretske, F., Knowledge and the Flow of Information, Oxford,
Basil Blackwell 1981
—. 1986: Dretske, F., Misrepresentation, in Bodgan, R.J. (ed. by), Belief:
Form, Content and Functions, Oxford, Oxford University Press, 1986,
294-313
—. 1988: Dretske, F., Explaining Behaviour. Reasons in a World of
Causes, Cambridge Mass., MIT Press, 1988
—. 1990: Dretske, F., Seeing, Believing and Knowing, in Oherson, D.N.,
Kosslyn, S.M., Hollerbach, J.M. (ed. by), Visual Cognition and Action,
vol. II, MIT Press, Cambridge Mass., 1990, 129-148
—. 1993: Dretske, F., Putting Information to Work, in Dretske 2000, 227241
—. 1995: Dretske, F., Naturalizing the Mind, Cambridge Mass., MIT Press
1995
—. 1999: Dretske, F., The Mind’s Awareness of Itself, in Dretske 2000,
158-177
—. 2000: Dretske, F., Perception, Knowledge and Belief. Selected Essays,
Cambridge, Cambridge University Press 2000
—. 2000a: Dretske, F., Conscious Experience, in Dretske 2000, 113-137
—. 2000b: Dretske, F., If You Can’t Make One, You Don’t Know How It
Works, in Dretske 2000, 208-226
—. 2003: Dretske, F., Experience as Representation, in «Philosophical
Issues», 13, 2003, 67-82
—. 2006: Dretske, F., Perception without Awareness, in Gendler/
Howthorne 2006, 147-180
Driver and Vuilleumier 2001: Driver, J, Vuilleumier, P., Perceptual
Awareness and Its Loss in Unilateral Neglect and Extinction, in S.
Deahene and L. Naccache (ed. by), Towards a Cognitive Neuroscience
of Consciousness: Basic Evidence and a Workspace Framework, 2001,
29-88
Ducasse 1942: Ducasse, C.J., Moore’s Refutation of Idealism, in P. Schlipp
(ed. by), The Philosophy of G.E. Moore, Chicago, Northwestern
University Press, 1942, 225-251
Ebner 2007: Ebner, M., Colour Constancy, West Sussex, Chichester,
England, Willey, 2007
Eriksen 1960: Eriksen, C.W., Discrimination and Learning Without
Awareness: A Methodological Study and Evaluation, «Psychological
Defending the Content View of Perceptual Experience
365
Review», 67, 1960, 279-300
Ernst 1996: Ernst, B., Optical Illusions, New York, Taschen, 1996
Evans 1982: Evans, G., The Varieties of Reference, Oxford, Oxford
University Press 1982
Farah 2004: Farah, M. J., Visual Agnosia, Cambridge Mass., MIT Press,
2004
Fish 2009: Fish, W., Perception, Hallucination, and Illusion, Oxford,
Oxford University Press, 2009
Fitch 2007: Fitch, G. and Nelson, M., "Singular Propositions", The
Stanford Encyclopedia of Philosophy (Fall 2014 Edition), Edward N.
Zalta (online ed.),
http://plato.stanford.edu/archives/fall2014/entries/propositionssingular/
Fodor 1983: Fodor, J., The Modularity of Mind, Cambridge Mass., MIT
Press, 1983
—. 1987: Fodor, J., Psychosemantics, Cambridge Mass., MIT Press, 1987
—. 1990: Fodor, J., A Theory of Content, in W. Lycan (ed. by), Mind and
Cognition, Oxford, Basil Blackwell 1990, 230-250
—. 1994: Fodor, J., The Elm and the Expert: Mentalese and Its Semantics,
Cambridge Mass., MIT Press, 1994
—. 1998: Fodor, J., Concepts. Where Cognitive Science Went Wrong,
Oxford, Oxford University Press, 1998
—. 2007: Fodor, J., The Revenge of the Given, in J. Cohen, B. McLaughlin
(ed. by), Contemporary Debates in Philosophy of Mind, Malden, MA,
Blackwell, 2007, 105-116
Foster 2000: Foster, J., The Nature of Perception, New York, Oxford
University Press, 2000
Frege 1891: Frege, G., Funktion und Begriff, translated as Function and
Concept by P. Geach in Translations from the Philosophical Writings
of Gottlob Frege, P. Geach and M. Black (eds. and trans.), Oxford,
Blackwell, third edition, 1980.
Gendler/Howthorne 2006: Gendler, T.S., Howthorne J. (ed. by),
Perceptual Experience, Oxford, Oxford University Press 2006
Gibson 1969: Gibson, E., Principles of Perceptual Learning and
Development, New York: Appleton-Century-Crofts, 1969
Gomez 2004: Gomez, J.C., Apes, Monkeys, Children and the Growth of
the Mind, Cambridge Mass., Harvard University Press, 2004
Goodale/Milner 2004: Goodale, M., Milner, A.D., Sight Unseen: An
Exploration of Conscious and Unconscious Vision, Oxford, Oxford
University Press, 2004
Grice 1957: Grice, P., “Meaning”, Studies in the Way of Words, Cambridge
366
Index of Authors
Mass., Harvard University Press, 1989
—. 1961: Grice, P., The Causal Theory of Perception, in Warnok G.J. (ed.
by), The Philosophy of Perception, Oxford, Oxford University Press
1977, 85-112
Grossman 1974: Grossman, R., Meinong’s Doctrine of the Aussersein of
the Pure Object, «Noüs», 8, 1974, 67-81
Gunther 2003: Gunther, Y.H. (ed. by), Essays on Nonconceptual Content,
Cambridge Mass., MIT Press, 2003
Haddock/Macpherson 2008: Haddock, A., Macpherson, F. (ed. by),
Disjunctivism, Oxford, OUP, 2008
Hansen et al. 2006: Hansen, T. et al, Memory Modulates Color
Appearance, «Nature Neuroscience», 9, 2006, 1367-8
Hardin 1987: Hardin, C.L., Qualia and Materialism. Closing the
Explanatory Gap, «Philosophy and Phenomenological Research», 48,
2, 1987, 291-298
—. 1993: Hardin, C.L., Color for Philosophers. Unweaving the Rainbow,
Indianapolis, Hackett, 1993
Harman 1990: Harman, G., The Intrinsic Quality of Experience,
«Philosophical Perspectives», 4, 1990, 31-52
Hawley/Macpherson 2011: Hawley, K., Macpherson, F., The Admissible
Contents of Experience, Oxford, Wiley-Blackwell, 2011
Heil 1983: Heil, J., Perception and Cognition, Los Angeles, University of
California Press, 1983
—. 1991: Heil, J., Perceptual Experience, in McLaughlin 1991 (ed. by), 116
Hilbert 1994: Hilbert, D., Is Seeing Believing?, «Proceedings of the
Biennial Meeting of the Philosophy of Science Association», Vol. One:
Contributed Papers, Chicago, University of Chicago Press, 1994, 446453
Hinton 1967: Hinton, M., Visual Experiences, «Mind», 76, 1967, 217-227
—. 1973: Hinton, M., Experiences, Oxford, Oxford University Press, 1973
Horgan/Tienson 2002: Horgan, T., Tienson, J., The Intentionality of
Phenomenology and the Phenomenology of Intentionality, in Chalmers
D. (ed. by), Philosophy of Mind: Classical and Contemporary
Readings, New York, Oxford University Press, 2002
Howthorne/Kovakovich 2006: Howthorne, J., Kovakovich, K., Disjunctivism,
supplement to «Proceedings of the Aristotelian Society», vol. 80, No.
1, 2006, 145-189
Husserl 1910/2006: Husserl, E., The basic problems of phenomenology
(from the lectures, Winter Semester, 1910–1911), Dordrecht, Springer,
2006.
Defending the Content View of Perceptual Experience
367
—. 1913/1970: Husserl, E., Logical Investigations, II vol., translated by
J.N. Findlay from the second German edition of Logische
Untersuchungen, London, The Humanities Press, 1970.
Hurvich and Jameson: Hurvich, L.M., Jameson, Dorothea, An OpponentProcess View of Color Vision, «Philosophical Review», 64, November
1957, 384-404
Jackson 1977: Jackson, F., Perception, Cambridge, Cambridge University
Press, 1977
Jacob 1999: Jacob, P., What Minds Can Do: Intentionality in a Non
Intentional World, Cambridge, Cambridge University Press, 1999
Jennerod & Jacob 2005: Jennerod, M., Jacob, P., Visual Cognition: A New
Look at the Two Visual Systems Model, «Neuropsychologia», 43, 301312
Johansson 1973: Johansson, G., Visual Perception of Biological Motion
and a Model for its Analysis, «Perception and Psychophysics», 14,
1973, 201-211
Johnston 2004: Johnston, M., The Obscure Object of Hallucination,
«Philosophical Studies», 120, 2004, 113-183
—. 2006: Johnston, M., Better than Mere Knowledge? The Function of
Sensory Awareness, in Gendler/Howthorne 2006, 260-290
Jolicoeur et al.: Jolicoeur, P., Gluck, M. A., Kosslyn, S. M., Pictures and
names: Making the connection, «Cognitive Psychology» 16, 1984,
243-275.
Kanisher 2001: Kanisher, N., Neural Events and Perceptual Awareness, in
Dehaene, S., Naccache, L. (ed. by), 2001, 89-113
Kanisza 1980: Kanisza, G., Grammatica del vedere, Bologna, Il Mulino,
1980
Kaplan 1989: Kaplan, D., Demonstratives, in Almog, J., Perry, J.,
Wettstein, H. (ed. by), Themes from Kaplan, New York, Oxford
University Press, 1989, 481-564
Kosslyn 1994: Kosslyn, S.M., Image and Brain: Resolution of the Imagery
Debate, Cambridge Mass., MIT Press, 1996
Kouider et al. 2007: Kouider, S., de Gardelle, V., Dupoux, E., Partial
Awareness and the Illusion of Phenomenal Consciousness, «Behavioral
and Brain Science», 30, 2007, 510-511
Kriegel 2002: Kriegel, U., Phenomenal Content, «Erkenntnis», 57, 2002,
175-198
—. 2009: Kriegel, U., Subjective Consciousness. A Self-representational
Theory, Oxford, OUP, 2009
Kripke 1982: Kripke, S., Wittgenstein on Rules and Private Language,
Oxford, Basil Blackwell, 1982
368
Index of Authors
Levine 2003: Levine, J., Experience and representation, in Q. Smith & A.
Jokic, ed. by), Consciousness: New Philosophical Essays, Oxford,
Oxford University Press.
Lewis 1980: Lewis, D., Veridical Hallucination and Prosthetic Vision, in
Dancy 2004, 79-91
—. 1983: Lewis, D., Individuation by Acquaintance and by Stipulation,
«Philosophical Review», 92, 1983, 3-32
—. 1995: Lewis, D., Should a Materialist Believe in Qualia?,
«Australausian Journal of Philosophy», 73 (1), 1995, 140-144
Lormand 1996: Lormand, E., Non-phenomenal Consciousness, «Nous»,
30, 1996, 242-261
Lycan 1987: Lycan, W., Consciousness, Cambridge Mass., MIT Press,
1987
—. 1996: Lycan, W., Consciousness and Experience, Cambridge Mass.,
Cambridge University Press, 1996
—. 2001: Lycan, W., The Case for Phenomenal Externalism,
«Philosophical Perspectives», 15 (supplement 15) 2001, 19-35
—. 2005: Lyons, J., Perceptual Belief and Non-experiential Look,
«Philosophical Perspectives», Vol. 19, Issue I, December 2005, 237256
Mack and Rock 1998: Mack, A, Rock, I., Inattentional Blindness,
Cambridge Mass., MIT Press, 1998.
Mach 1897: Mach, E., The Analysis of Sensations, Dover, New York 1959
Macpherson 2006: Macpherson, F., Ambiguous Figures and the Content of
Experience, «Noûs», 40:1, 2006, 82-117
—. 2011: Macpherson, F., Introduction: The Admissible Contents of
Experience, in Hawley/Macpherson 2011, 1-15
Malcolm 1942: Malcolm, N., Moore and Ordinary Language, in V. C.
Chappell, ed. by, Ordinary Language, Englewood Cliffs, Prentice-Hall,
1964, 5-23.
Marr 1982: Marr, D., Vision, New York, Freeman & Co., 1982
Marsolek 1999: Marsolek, C. J., Dissociable neural subsystems underlie
abstract and specific object recognition, «Psychological Science» 10,
1999, 111-118.
Martin 1992: Martin, M., Perception, Concepts, and Memory,
«Philosophical Review», vol. 101, 1992, 745-763
—. 1993: Martin, M., The Rational Role of Experience, «Proceedings of
The Aristotelian Society», 1993, 71-88
—. 2002: Martin, M., The Transparency of Experience, «Mind and
Language», Vol. 17, 4, 2002, 376-425
—. 2003: Martin, M., Sensible Appearances, in Baldwin, T. (ed. by), The
Defending the Content View of Perceptual Experience
369
Cambridge History of Philosophy: 1870-1945, Cambridge Mass.,
Cambridge University Press, 521-532
—. 2004: Martin, M., The Limits of Self-Awareness, «Philosophical
Studies», 120, 2004, 37-89
—. 2006: Martin, M., On Being Alienated, in Gendler/Howthorne 2006,
354-410
—. 2008: Martin, M., Uncovering Appearances (online manuscript):
http://www.homepages.ucl.ac.uk/~uctymfm/chapters.htm
Maund 2003: Maund, B., Perception, Chesham, Acumen 2003
McDowell 1982: McDowell, J., Criteria, Defeasibility and Knowledge, in
Dancy 2004, 209-219
—. 1984: McDowell, J., De Re Senses, «Philosophical Quarterly», xxxiv,
1984, 283-294
—. 1991: McDowell, J., Intentionality De Re, in LePore, E. and van
Gulick, R. (ed. by), John Searle and His Critics, Oxford, Blackwell,
1991, 215-25
—. 1994: McDowell, J., Mind and World, Cambridge Mass., Harvard
University Press, 1994
—. 1994b: McDowell, J., The Content of Perceptual Experience, «The
Philosophical Quarterly», Vol. 44, N. 175, Apr. 1994, 190-205
—. 1998: McDowell, J., Woodbridge Lectures: Sellars on Perceptual
Experience (Lecture I), «The Journal of Philosophy», vol. 95, n. 9,
Sept. 1998, 431-450
—. 2008: McDowell, J., The Disjunctive Conception of Experience as a
Material for a Transcendental Argument, in Macpherson, F., Haddock,
A.(ed. by), Disjunctivism: Perception, Action, Knowledge, Oxford,
Oxford University Press, 2008
McGinn 1982: McGinn, C., The Character of Mind, Oxford, Oxford
University Press, 1982
—. 1988: McGinn, C., Consciousness and Content, «Proceedings of the
British Academy», 74, 1988, 219-239
McGlinchey-Berroth 1997: McGlinchey-Berroth, R., Visual Information
Processing in Hemispatial Neglect, «Trends in Cognitive Sciences»,
vol. 1, N. 3, 1997, 91-97
McLaughlin 1989: McLaughlin, B., Why Perception is Not Singular
Reference, in Heil, J. (ed. by), Cause, Mind, and Reality, Dordrecht,
Kluwer, 1989, 112-128
—. 1991: McLaughlin, B. (ed. by), Dretske and Its Critics, Cambridge
Mass., Basil Blackwell, 1991
McLaughlin/Tye 1998: McLaughlin, B., Tye, M., Is Content-externalism
Compatible with Privileged Access?, «Philosophical Review», 107,
370
Index of Authors
1998, 349-380
McHugh 2008: McHugh, C., Self-Knowledge in Consciousness, PhD
Thesis – Edinburgh University (http://hdl.handle.net/1842/3488)
Meinong 1904: Über Gegenstandtheorie, Leipzig, Bart, 1904, translated
as The Theory of Objects in Chisholm, R.M. (ed. by), Realism and the
Background of Phenomenology, Glencoe, IL: Free Press, 1960; reprint:
Atascadero, CA: Ridgeview, 1981, 76–117.
Merleau-Ponty 1945: Merleau-Ponty, M., Phénoménologie de la
perception, Paris, Gallimard, 1945
Millar 1991: Millar, A., Concepts, Experience and Inference, «Mind», Vol.
100, n. 4, Oct 1991, 495-505
—. 2000: Millar, A., The Scope of Perceptual Knowledge, «Philosophy»,
Vol. 75, n. 291, Jan 2000, 73-88
—. 2007: Millar, A., What the Disjunctivist is Right About, «Philosophy
and Phenomenological Research», 74, 2007, 176-198
Milner/Goodale 1995: Milner, D., Goodale, M.A., The Visual Brain in
Action, Oxford, Oxford University Press, 1995
Millikan 1984: Millikan, R.G., Language, Thought and Other Biological
Categories, Cambridge Mass., MIT Press, 1984
—. 1989: Millikan, R., In Defence of Proper Functions, «Philosophy of
Science», vol. 56, n. 2, June 1989, 288-302
—. 1990: Millikan, R., Biosemantics, in Lycan W. (ed. by), Mind and
Cognition, Oxford, Basil Blackwell 1990, 221-229
—. 2000: Millikan, R.G., On Clear and Confused Ideas, Cambridge Mass.,
Cambridge University Press, 2000
—. 2009: Millikan, R.G., Biofunctions: Two Paradigms, in Ariew/
Cummins/Perlman 2009, 113-143
Milner/Goodale 1996: Milner, D., Goodale, M. (ed. by), The Visual Brain
in Action, Oxford, Oxford University Press, 1996
Mollon et al. 1980: Mollon, J.D, Newcombe, F., Polden, P.G., Ratcliff, G.,
On the Presence of Three Cones Mechanisms in Case of Total
Achromatopsia, in Verriest, G. (ed. by), Color Vision Deficiencies,
Bristol, Adam Higler, 130-135
Moore 1910: Moore, G.E., Some Main Problems in Philosophy, London,
Allen and Unvin, 1953
—. 1942: Moore, G.E., A Reply to My Critics, in Schlipp, P. (ed. by), The
Philosophy of G. E. Moore, Evanston, Tudor, III, 1942, 543-667
—. 1962: Moore, G.E., Philosophical Papers, New York, Collier Books,
1962
Nagel 1974: Nagel, Th., What is is Like to be a Bat?, «The Philosophical
Review», vol. 83, n. 4, Oct. 1974, 435-450
Defending the Content View of Perceptual Experience
371
Neander 1995: Neander, K., Misrepresenting and Malfunctioning,
«Philosophical Studies», 79, 1995, 109-141
—. 2006: Neander, K., Content for Cognitive Science, in
Macdonald/Papineau 2006, 195-207
Neisser 1967: Neisser, U., Cognitive Psychology. New York: AppletonCentury-Croft, 1967
Noë 2004: Noë, A., Action in Perception, Cambridge Mass., MIT Press
2004
Nudds 2009: Nudds, M., Recent Work in Perception: Naïve Realism and
Its Opponents, «Analysis», vol. 69, N. 2, 2009, 333-346
Osherson/Kosslyn/Hollerbach 1990: Oherson, D.N., Kosslyn, S.M.,
Hollerbach, J.M. (ed. by), Visual Cognition and Action, vol. II, MIT
Press, Cambridge Mass., 1990
O’Regan/Noë 2001: O’Regan, J.K, Noë, A., A Sensimotor Approach to
Vision and Visual Consciousness, «Behavioral and Brain Sciences»,
vol. 24, n. 5, 2001, 3-28
Palmer 1983: Palmer, S., The Psychology of Perceptual Organization: A
Transformational Approach, in J. Beck, B. Hope, A. Rosenfeld (ed.
by), New York, Academic Press, 1983
—. 1999: Palmer, S., Vision Science. Photons to Phenomenology,
Cambridge Mass., MIT Press, 1999
Papineau 1990: Papineau, D., Truth and Teleology, in Knowles, D. (ed.
by), Explanation and Its Limits, Cambridge, Cambridge University
Press, 1990
—. 1993: Papineau, D., Philosophical Naturalism, Oxford, Blackwell,
1993
Parsons 1980: Parsons, T., Nonexistent Objects, New Haven, Conn., Yale
U.P., 1980
Pautz 2006: Pautz, A., Can The Physicalist Explain the Colour Structure in
Terms of Colour Experience?, «Australian Journal of Philosophy», 84
(4), 2006, 535-554
—. 2007: Pautz, A., Intentionalism and Perceptual Presence,
«Philosophical Perspectives», 21, 493-541
—. 2011: Pautz, A., What are the Contents of Experience? In
Hawley/Macpherson 2011, 114-138
Peacocke 1983: Peacocke, C., Sense and Content. Experience, Thought
and their Relations, Oxford, Clarendon Press, 1983
—. 1986: Peacocke, C., Analog Content, «Proceeding of the Aristotelian
Society», 60, 1986, 1-17
—. 1991: Peacocke, C., Demonstrative Content: A Reply to John
McDowell, «Mind», Vol. 100, (1) Jan 1991, 123-133
372
Index of Authors
—. 1992: Peacocke, C., A Study of Concepts, Cambridge Mass., MIT
Press, 1992
—. 1994: Peacocke, C., Content, Computation and Externalism, «Mind
and Language», 9, 1994, 303-335
—. 1998: Peacocke, C., Nonconceptual content defended, «Philosophy and
Phenomenological Research», 58 (2), 1998 38,1-388
—. 2001: Peacocke, C., Phenomenology and Nonconceptual Content,
«Philosophy and Phenomenological Research», vol. 62, n. 3, May
2001, 609-615
—. 2003: Peacocke, C., Scenarios, Content and Perception, in Gunther,
Y.H. (ed. by), Essays on Nonconceptual Content, Cambridge Mass.,
MIT Press, 2003, 107-132
—. 2007: Peacocke, C., Sensational Properties: Theses to Accept and
Theses to Reject, «Revue Internationale de Philosophie», 62, 2008, 724
Perry 1979: Perry, J., The Problem of Essential Indexicals, «Noûs», 13, 1,
3-21
Pettit 2003: Pettit, P., Looks as Powers, «Philosophical Issues», 13 (1),
221-252
Piaget 1937: Piaget, J., La construction du réel chez l’enfant, Neuchâtel,
Delachaux et Niestlée, 1937
Pitcher 1970: Pitcher, G., A Theory of Perception, Princeton, Princeton
University Press, 1970
Pitt 2004: Pitt, D., The Phenomenology of Cognition, Or What Is It Like to
Think That P?, «Philosophy and Phenomenological Research», 69 (1),
1-36
Price 1932: Price, H.H., Perception, London, Methuen, 1932
—. 2011: Price, R., Aspect-Switching and Visual Phenomenal Character,
in Hawley/Macpherson 2011, 139-149
Prinz 2004: Prinz, J.J., Furnishing The Mind: Concepts and Their
Perceptual Basis, Cambridge Mass., MIT Press, 2004
—. 2006: Prinz, J.J., Beyond Appearances: The Content of Sensation and
Perception, in Gendler/Howthorne 2006, 434-460
Putnam 1975: Putnam, H., The Meaning of “Meaning”, in H. Putnam,
Philosophical Papers, vol. 2: Mind, Language and Reality, Cambridge
University Press, Cambridge, 1975
—. 1990: Putnam, H., James’s Theory of Perception, in Id., Realism with a
Human Face, Cambridge, Harvard University Press, 1990, 232-251
—. 1992: Putnam, H., Renewing Philosophy, Cambridge, Harvard
University Press, 1992
—. 1992b: Putnam, H., A Theory of Reference, in Putnam 1992, 35-59
Defending the Content View of Perceptual Experience
373
—. 1994: Putnam, H., Sense, Nonsense and the Senses: An Inquiry into the
Powers of Human Mind, The Dewey Lectures, «The Journal of
Philosophy», XCI, 1994, 445-517
—. 1999: Putnam, H., The Threefold Cord: Mind, Body and the World,
New York, Columbia University Press, New York, 1999
Pylyshyn 2006: Pylyshyn, Z.W., Seeing and Visualizing: It’s Not What You
Think (Life & Mind: Philosophical Issues in Biology & Psychology),
Cambridge Mass., MIT Press, 2006
Quine 1969: Quine, W.v.O., Natural Kinds, in Quine, W.v.O., Ontological
Relativity and Other Essays, New York, Columbia University Press,
1969
Raffman 1995: Raffman, D., On the Persistence of Phenomenology, in
Metzinger T. (ed. by), Conscious Experience, Lawrence, Kans., Allen,
293-308
Rensink 2000: Rensink, R.A., Visual search for change: a probe into the
nature of attentional processing, «Visual Cognition», 7, 2000, 345-376.
Rey 1998: Rey, G., Narrow Representationalist Account of Qualitative
Experience, in Tomberlin, J.E. (ed. by), Language, Mind, and Ontology
(«Philosophical Perspectives», Vol. 12). (Atascadero, CA: Ridgeview
Publishing.), 435-457
Robinson 1994: Robinson, H., Perception, London, Routledge 1994
Rosenthal 2005: Rosenthal, D., Consciousness and Mind, Oxford, Oxford
University Press, 2005
Rubens and Benson 1971: Rubens, A.B., Benson, D.F., Associative Visual
Agnosia, «Archives of Neurology», 24, 1971, 305-316
Russell 1910: Russell, B., Philosophical Essays, London: Longmans,
Green, 1910
—. 1912: Russell, B., The Problems of Philosophy, New York, Oxford
University Press, 1967
—. 1922: Russell, B., Our Knowledge of the External World, London,
Allen and Unwin, 1922
Ryle 1949: Ryle, G., The Concept of Mind, London, Hutchinson 1949
Sacks 1985: Sacks, O., The Man Who Mistook His Wife for a Hat, London,
Picador, 1985
Sainsbury 2006: Sainsbury, M., Reference Without Referents, Oxford,
Oxford University Press, 2006
Sanford 1981: Sanford, D.H, Illusions and Sense-Data, «Midwest Studies
in Philosophy», 6 (1),(1981), 371-385
Sartwell 1995: Sartwell, C., Radical Externalism Concerning Experience,
«Philosophical Studies», 78, 1995, 55-70
Searle 1983: Searle, J., Intentionality. An Essay in the Philosophy of Mind,
374
Index of Authors
Cambridge Mass., Cambridge University Press, 1983
—. 1992: Searle, J., The Rediscovery of Mind, Cambridge Mass., MIT
Press 1992
Sellars 1975: Sellars, W., Adverbial Theory of the Objects of Sensation,
«Metaphilosophy» 6, 1975, 144-160
—. 1953: Sellars, W., Empiricism and Philosophy of Mind, in DeVries, W.
A., Triplett T., Knowledge, Mind and The Given. Reading Wilfrid
Sellars’s “Empiricism and Philosophy of Mind”, Indianapolis, Hackett
Publishing Company, 2000
—. 1973: Sellars, W., Givenness and Explanatory Coherence, «The
Journal of Philosophy», vol. 70, n. 18, Oct. 1973, 612-624
—. 1981: Sellars, W., Mental Events, «Philosophical Studies», 81, 1981,
325–345
Sergent/Villemure 1989: Sergent, J., Villemure, J., Prosopagnosia in a
right emispherectomized patient, 12, 1989, 975-995
Shoemaker 1975: Shoemaker, S., Functionalism and Qualia,
«Philosophical Studies», 27, May 1975, 291-315
—. 1982: Shoemaker, S., The Inverted Spectrum, «The Journal of
Philosophy», 79, 1982, 357-381
—. 1994: Shoemaker, S., Self-knowledge and Inner Sense (lecture III):
The Phenomenal Character of Experience, «Philosophy and
Phenomenological Research», vol. 54 (2) June 1994, 291-314
—. 2006: Shoemaker, S., On the Ways Things Appear, in Gendler, T.S.,
Howthorne, J. (ed. by), Perceptual Experience, Oxford, Oxford
University Press 2006, 461-480
Siegel 2006: Siegel, S., Which Properties are Represented in Perception?,
in Gendler/Howthorne 2006, 481-503
—. 2006b: Siegel, S., How Does Phenomenology Constrain ObjectSeeing?, in «Australasian Journal of Philosophy», Vol. 84 (2), 2006,
429-441
—. 2006c: Siegel, S., Subject and Object in the Contents of Visual
Experience, in «Philosophical Review», Vol. 115, Vol. 3, 2006
—. 2009: Siegel, S., Do Visual Experiences Have Contents? available
online:
http://www.people.fas.harvard.edu/~ssiegel/papers/SPAWN_SIEGEL_
Sept1509.html
—. 2010: Siegel, S., The Contents of Experience, Oxford, OUP, 2011
—. 2011: Siegel, S., The Visual Experience of Causation, in Hawley, K,
Macpherson, F. (ed. by), The Admissible Contents of Experience,
Oxford, Wiley-Blackwell, 2011
Siewert 1998: Siewert, Ch., The Significance of Consciousness, Princeton,
Defending the Content View of Perceptual Experience
375
Princeton University Press, 1998
—. 2003: Siewert, Ch., Is Experience Transparent?, in «Philosophical
Studies», 117, 2003, 15-41
Simon & Chabris 1999: Simon, D.J., Chabris, C.F., Gorillas in our Midst:
Sustained Inattentional Blindness for Dynamic Events, «Perception»,
27, 1999, 1025-1040
Slater 1998: Slater, A., Visual Organization and Perceptual Constancies in
Early Infancy, in Walsh, V., Kulikowsky, J. (ed. by), Perceptual
Constancy: Why Things Look as They Do, Cambridge Mass.,
Cambridge University Press, 6-30
Smith 2001: Smith, A.D., Perception and Belief, «Philosophy and
Phenomenological Research», vol. 62, n. 2, Mar. 2001, 283-309
—. 2002: Smith, A.D., The Problem of Perception, Delhi, Motilal
Banarsiddas Publishers, 2005 (ed. or. 2002),
—. 2008: Smith, A.D., Disjunctivism and Discriminability, in
Haddock/Macpherson 2008, 181-204
Snowdon 1981: Snowdon, P., Perception, Vision and Causation, in Dancy
2004, 192-208
Sober 1984: Sober, E., The Nature of Selection, Cambridge Mass., MIT
Press, 1984
Soteriou 2000: Soteriou, M., The Particularity of Visual Perception,
«European Journal of Philosophy», 8 (2), 2000, 137-189
—. 2009: Soteriou, M., The Disjunctive Theory of Perception, in «Stanford
Encyclopedia of Philosophy», available online
Speacks 2005: Speaks, J., Is There a Problem about Nonconceptual
Content? «Philosophical Review», 114 (3), 2005, 359-398
Sperling 1960: Sperling, G., The Information Available in Brief Visual
Presentations, «Psychological Monographs», 74 (11, whole No. 498),
1960, 29-49
Stalnaker 1984: Stalnaker, R., Inquiry, Cambridge Mass., MIT Press, 1984
Stoljar 2004: Stoljar, D., The Argument from Diaphanousness, «Canadian
Journal of of Philosophy», Suppl. Vol. 30, 2004, 341-390
Strawson 1979: Strawson, P., Perception and Its Objects, in Dancy, J. (ed.
by), Perceptual Knowledge, Oxford, Oxford University Press, 2004
(ed. or. 1979)
—. 1994: Strawson, G., Mental Reality, Cambridge Mass., MIT Press,
1994
Sturgeon 1998: Sturgeon, S., Visual Experience, «Proceedings of
Aristotelian Society», 98, 1998, 179-200
—. 2008: Sturgeon, S., Disjunctivism about Perceptual Experience, in
Macpherson, F., Haddock, A., Disjunctivism: Perception, Action,
376
Index of Authors
Knowledge, Oxford, Oxford University Press, 112-143
Thau 2002: Thau, M., Consciousness and Cognition, Oxford, Oxford
University Press, 2002
Thompson 2003: Thompson, B., The Nature of Phenomenal Content, PhD
Dissertation, University of Arizona
—. 2008: Thompson, B., Representationalism and the Conceivability of
Inverted Spectra, «Synthese», 160, 2008, 203-213
Travis 2004: Travis, Ch., The Silence of Senses, «Mind», vol. 113, 449, Jan
2004, 57-94
—. 2011: Travis, Ch., Is Seeing Intentional?, in Ambroise, B., Lauger, S.
(ed. by), Philosophie du Langage Ordinaire, Hildesheim, Georg Holms
Treisman 1998: Treisman, A., Feature Binding, Attention and Object
Perception, «Philos Trans R Soc Lond B Biol Sci.», 29, August 1988,
1295–1306.
Tye 1984: Tye, M., The Adverbial Approach to Visual Experience, «The
Philosophical Review», vol. 93, n. 2, Apr 1984, 195-225
—. 1998: Tye, M., Inverted Earth, Swampman and Representationalism,
«Philosophical Perspectives», 12, 1998, 459-478
—. 2000: Tye, M., Consciousness, Colour and Content, Cambridge Mass.,
MIT Press, 2000
—. 2003: Tye, M., Qualia and Visual Content Revisited, in D. Chalmers
(ed. by), Philosophy of Mind: Classical and Contemporary Readings,
Oxford, OUP, 2002
—. 2006: Tye, M., Nonconceptual Content, Richness and Fineness of
Grain, in Gendler/Howthorne 2006, 504-531
—. 2006b: Tye, M., The Admissible Contents of Experience, in
Hawley/Macpherson 2011, 172-193
—. 2007: Qualia, in Stanford Encyclopedia of Philosophy, online at:
http://plato.stanford.edu/entries/qualia, 2007, Section 1.
Walker 1983: Walker, S., Animal Thought, London, Routledge and Kegan
Paul, 1983
Weiskrantz 1997: Weiskrantz, L. Consciousness Lost and Found: A
Neuropsychological Exploration, New York, Oxford University Press,
1997
Wiggins 1980: Wiggins, D., Sameness and Substance, Oxford, Blackwell,
1980
Williamson 1990: Williamson, T., Identity and Discrimination, Oxford,
Blackwell, 1990
—. 2000: Williamson, T., Knowledge and Its Limits, Oxford, Oxford
University Press, 2000
Wittgenstein 1953: Wittgenstein, L., Philosophical Investigations, Oxford,
Defending the Content View of Perceptual Experience
377
Blackwell 1953
—. 1980: Wittgenstein, L., vol. 1, G.E.M. Anscombe and G.H. von Wright
(eds.), G.E.M. Anscombe (trans.), vol. 2, G.H. von Wright and H.
Nyman (eds.), C.G. Luckhardt and M.A.E. Aue (trans.), Oxford:
Blackwell.
Wright 2008: Wright, C., Comment on John McDowell’s “The Disjunctive
Conception of Experience as a Material for a Transcendental
Argument”, in Haddock/Macpherson 2008, 390-404
Yablo 1992: Yablo, S., Mental Causation, «The Philosophical Review»,
vol. 101, n. 2, Apr. 1992, 245-280
Yuille/Hullman 1990: Yuille, A.L., Hullman, S, Computational Theories of
Low-Level Vision, in Osherson/Kosslyn/Hollerbach 1990, 5-40
INDEX OF AUTHORS
Allen, C., Bekoff, M., Lauder, R.E.,
314n
Algioti, S., 100, 100n
Alston, W., 50n, 59n
Anscombe, G.E.M., 18n
Ariew, A., Cummins, R., Perlman,
M., 314n
Armstrong, D.M., 16n, 28n, 29n,
68, 68n, 71n, 77n, 79n, 94n,
162n,
Austin, J., 11n, 18n, 32n, 46n, 47n,
48n, 50n, 52n, 53n, 72n
Ayer, A.J., 56n, 69n, 72n
Baars, B., 95n
Back-y-Rita, P., Kercel, S.V., 16n
Bain, D., 170n, 190, 190n
Baldwin, T., 194n
Bayne, T., 98n, 110n, 135n, 236n,
237n, 244n, 247n, 253, 253n,
254, 254n
Beauvois, M.F., Salliant, B., 101n
Ben Zeev, A., 272n
Bermudez, J.L., 133n
Biederman, I., 239, 239n
Block, N., 6, 22n, 56n, 89n, 95n,
113n, 189n, 193n, 202n, 245n
Boghossian, P., Velleman, D., 194n
Bozzi, P., 37n
Brandom, R., 83n
Brentano, F., 70n, 119n,
Brewer, B., 107n, 109n, 118, 118n,
246n, 298n
Broad, C.D., 56n, 59n, 70n, 72n
Buller, D., 314n
Burge, T., 162n, 184n, 201, 208n,
209n, 343n
Butterfill, S., 162n, 244n
Byrne, A., 54n, 98n, 170n, 177n,
189n, 199n, 213n, 248n, 298n
Campbell, J., 298n, 302n, 303n,
304n, 343
Carruthers, P., 236n
Chalmers, D., 22n, 110n, 132n,
133n, 151n, 158n, 162n, 165n,
171n, 175n, 189n, 191n, 195n,
196n, 218n, 219n, 221n, 223n,
224, 229, 230
Chisholm, R., 46, 46n, 50, 50n, 52,
54, 54n, 59n, 71n, 73, 118n, 174
Churchland, Paul, 178n, 212n, 213n
Clark, Austen, 98n, 137n
Clark, Andy, 22n
Crane, T., 28n, 42, 42n, 71n, 78n,
81n, 82n, 83, 83n, 85, 108n,
113, 114n, 133n, 118, 188,
178n, 184n, 189n, 190n, 193n,
194n, 284, 307n, 348n
Cowey, A., 101n
Cummins, R., 310n, 314n
Cussins, A., 144n, 288n
Dancy, J., 72n
Davidson, D., 34n, 82n, 201, 208n
Davies, M., 107, 157n
Dennett, D., 16, 16n, 89n, 91n, 94,
94n, 111n, 200n, 314n
Dretske, F., 19n, 20n, 22n. 23n, 24n,
28n, 31, 31n, 86, 86n, 89, 89n,
91,
113n, 114, 114n, 118, 118n, 141,
141n, 142, 142n, 143, 143n,
144, 145, 146, 147, 147n, 185,
190, 190n, 235, 281, 331, 331n
Defending the Content View of Perceptual Experience
Driver, J., 102n
Ducasse, C.J., 73
Ernst, B., 178n
Evans, G., 82, 82n, 83, 83n, 87n,
107, 117n, 184n, 185n, 187n,
304, 304n, 334, 334n, 335n,
336n, 354
Farah, M.J., 96n
Fish, W., 113n, 118, 186n, 190n,
298n, 301, 302, 309, 307n, 348
Fodor, J., 83n, 96n, 114n, 143n,
144n, 148n, 201, 204, 204n,
225n, 311n
Foster, J., 331n
Frege, G., 154n, 218
Gibson, E., 270n
Goodale, M., 96n, 98n, 99n
Grice, P., 16n, 156, 158, 163n
Grossman, R., 328n
Haddock, A., 298n
Hansen, T., 250n
Hardin, C.L., 222n
Harman, G., 18, 18n, 112, 189n
Heil, J., 18, 18n, 68n,
Hilbert, D., 22n, 101, 101n
Hinton, M., 60n, 298n
Horgan, T., 81, 110, 196, 216, 217n,
218n, 244n, 246n,
Howthorne, J., 298n
Husserl, E., 80, 80n, 93n, 119n,
189n, 259n, 327
Hurvich, L.M., 213n
Jameson, D., 213n
Jackson, F., 32, 32n, 50, 50n, 51, 52,
54, 54n, 55, 55n, 56, 56n, 59n,
70n,
Jacob, P., 87n, 99n, 225n, 310n,
311n
Jennerod, M., 99n
Johansson, G., 144n
Johnston, M., 27n, 44n, 143n, 298n,
302n, 304, 304n, 305, 305n,
326n, 330n, 331, 346,
Jolicoeur, P., 242n
Kanisza, G., 260n
Kaplan, D., 304n
379
Kosslyn, S.M., 242n, 255n, 260
Kovakovich, K., 299n
Kriegel, U., 81n, 196n, 218, 218n
Kripke, S., 13n
Levine, J., 218, 218n
Lewis, D., 16n, 156n, 157n, 163n,
223n, 304n
Logue, H., 298n
Lormand, E., 236n
Lycan, W., 73n, 75n, 189n, 193n,
190, 195n, 204n, 211n
Lyons, J., 148n, 152n, 242, 242n,
291n, 292n, 293n
Mach, E., 130
Mack, A., 22n, 88n
Macpherson, F., 130n, 131n, 258n,
298n
Malcolm, N., 14n
Marr, D., 237, 238
Marsolek, C.J., 242n
Martin, M., 74, 74n, 111n, 112n,
113n, 117n, 118, 118n, 149n,
169n, 186n, 298, 298n, 299n,
301, 302, 306n, 307n, 313, 318,
345, 348, 349, 349n, 309
Maund, B., 51, 51n, 74n, 110n
McDowell, J., 109n, 121n, 162n,
201, 236n, 298, 298n, 302, 303,
303n, 342, 343,
McGinn, C., 5, 44n, 98n, 135n,
151n, 157, 157n, 168, 216, 329n
McHugh, C., 186n
McGlinchey-Berroth, R., 102n
McLaughlin, B., 157n
Meinong, A., 328n, 329n, 330
Millar, A., 162n, 236n
Millikan, R.G., 27n, 310n, 322n
Milner, D., 96n, 98n, 99n
Mollon, J.D, Newcombe, F., Polden,
P.G., Ratcliff, G., 101n
Moore, G.E., 14n, 69n, 72n, 74n,
108n, 112n
Nagel, Th., 110n, 111n
Neander, K., 310n
Neisser, U., 95n
Noë, A., 91n, 94n
380
Index of Authors
Nudds, M., 163n
Kosslyn, 243n, 255n, 259n, 260
O’Regan, J.K, Noë, A., 94n
Palmer, S., 130n, 200n
Papineau, D., 310n
Parsons, T., 328n
Pautz, A., 71n, 213n, 301n, 327,
329n, 330n, 331n, 334, 336n,
354
Peacocke, Ch., 3, 4, 83n, 197, 111n,
113n, 114n, 115, 117n, 126, 130,
131n, 133, 133n, 135, 138n,
176, 177n, 184n, 193n, 217,
219n, 270, 326n, 245n, 271n,
357
Pettit, P., 195n, 223n
Pitcher, G., 68n, 78, 78n, 79, 80
Pitt, D., 81n, 244n, 246n
Price, H.H., 56n, 70n, 72n, 101n
Price, R., 43n, 56n, 135n, 190n,
236n, 257n, 258, 258n
Prinz, J., 98n, 135n, 147n, 150n,
232n, 243n, 274n, 279, 279n,
285, 287
Putnam, H., 199, 287n, 298n, 201
Pylyshyn, Z.W., 78n, 204n, 238n
Quine, W.v.O, 240
Raffman, D., 150n
Rensink, R.A., 89n
Rey, G., 211n, 218, 218n
Robinson, H., 59n, 70n, 72n
Rock, I., 22n, 88n
Rubens, A.B., Benson, D.F., 97n
Russell, B., 154n, 218n, 304n
Ryle, G., 16n, 69n
Sacks, O., 37n
Sainsbury, M., 158n, 165n
Sanford, D.H., 71n
Sartwell, C., 343
Searle, J., 5, 158, 159, 161, 162,
164, 245n
Sellars, W., 2, 59, 270n, 60n, 73n,
183n, 270n
Sergent, J., Villemure, J., 273n
Shoemaker, S., 190n, 193n, 195,
199n, 217
Siegel, S., 7, 22,19n, 21, 22, 98,
107, 107n, 110, 110n, 117n,
119n, 120, 135n, 162n, 165n,
175n, 185n, 236n, 237, 237n,
243, 244n, 245n, 246, 246n,
247, 248, 248n, 249, 249n, 250,
252, 261n, 264, 283, 285, 286,
287, 290, 292, 294, 295, 327
Siewert, Ch., 112n, 216, 244n
Simon, D.J., Chabris, C.F., 88, 88n
Slater, A., 178n
Smith, A.D., 72n, 79n, 80n, 81n,
93n, 301n, 326n, 327, 328n,
329, 329n
Snowdon, P., 52n, 298n, 307n, 320n
Sober, E., 276n
Soteriou, M., 157n, 158n, 159n,
161, 161n, 162
Sperling, G., 3, 19, 88, 94, 95, 95n,
96, 96n, 164, 217
Stalnaker, R., 27n, 193n
Stoljar, D., 112n
Strawson, P., 16n, 236n
Thau, M., 190n, 195
Thompson, B., 195n, 218, 218n
Tienson, J., 81n, 110n, 196n, 216,
217n, 218n, 244n, 246n
Travis, Ch., 298n, 303, 303n, 304,
305, 304n, 346, 346n, 347n
Treisman, A., 250n
Tye, M., 41n, 42n, 55n, 73n, 110n,
111n, 112n, 113n, 118n, 135n,
148n, 150n, 160, 161 163n,
164n, 170n, 190, 190n, 193n,
195n, 196n, 207, 211n, 213n,
236n, 251n, 258n, 330n, 336n
Vuilleumier, P., 102n
Wiggins, D., 306n, 313n
Williamson, T., 28n
Wittgenstein, L., 37, 37n, 147
Wright, C., 346n
Yuille, A.L., Hullman, S., 238n