Representationalism and
Anti-Representationalism
About Perceptual Experience
by Keith A. Wilson
Thesis submitted in partial fulfilment of the
requirements for the degree of Doctor of Philosophy
University of Warwick
Department of Philosophy
May 2013
Contents
1. Introduction
1.1. Background
1
1.2. Perceptual Experience and Representation
3
1.2.1. Perceptual experience ........................................................................3
1.2.2. Phenomenal character .......................................................................5
1.2.3. Representational content ...................................................................5
1.2.4. Representationalism and relationalism ..............................................9
1.3. Overview
10
2. The Silence of the Senses
2.1. Introduction
14
2.2. Perceptual Representation
15
2.2.1. Representing such-and-such as so ....................................................17
2.2.2. Face value .......................................................................................18
2.2.3. Represention-to, not representation-by............................................21
2.2.4. Recognisability ................................................................................23
2.2.5. Looks-indexing................................................................................26
2.2.6. P-representation ..............................................................................27
2.3. The Case Against Representationalism
29
2.3.1. The argument from misleading appearances ....................................29
2.3.2. The argument from looks ................................................................32
2.3.3. The argument from recognisability ..................................................37
2.3.4. The argument from unmediated awareness .....................................41
2.4. Conclusion
45
3. The Sound of Silence
3.1. Introduction
47
3.2. An Argument from Non-Comparative Looks
47
3.2.1. ‘Zeeing’ and the problem of illusion ................................................48
3.2.2. Non-comparative looks ...................................................................51
3.2.3. Phenomenal content ........................................................................53
3.2.4. Responses ........................................................................................56
3.3. An Argument from Visual Appearances
57
3.3.1. Siegel’s Content View ......................................................................58
3.3.2. The Argument from Appearing .......................................................60
3.3.3. Property-types and property-instances .............................................62
3.3.4. Responses ........................................................................................66
3.4. An Argument from Perceptual Seeming
70
3.4.1. Schellenberg’s Master Argument .....................................................70
3.5. Conclusion
74
Representationalism & Anti-Representationalism About Perceptual Experience
iii
4. Looks
4.1. Introduction
75
4.2. The Varieties of Looks
76
4.2.1. Evidential and non-evidential looks .................................................77
4.2.2. Comparative and non-comparative looks ........................................78
4.2.3. Senses, uses and the regress argument .............................................80
4.2.4. Phenomenal and non-phenomenal looks .........................................82
4.2.5. Non-comparative phenomenal looks ...............................................85
4.2.6. Phenomenal content ........................................................................90
4.3. Parsimony about Appearances
92
4.3.1. A deflationary analysis ....................................................................93
4.3.2. What’s in a look? ............................................................................96
4.3.3. The problem of conflicting appearances ..........................................98
4.3.4. Consequences of Parsimony...........................................................101
4.4. Conclusion
105
5. Recognisability and Phenomenal Character
5.1. Introduction
106
5.2. Recognisability
106
5.2.1. The case for Recognisability ..........................................................107
5.2.2. Availability to the subject ..............................................................109
5.2.3. Recognisability and phenomenal character ....................................111
5.2.4. Two questions about perception ....................................................114
5.2.5. Consequences of Recognisability ...................................................117
5.3. Phenomenal Character and Awareness
120
5.3.1. The phenomenological objection ...................................................120
5.3.2. The argument from unmediated awareness ...................................124
5.4. Conclusion
126
6. Disambiguation Strategies
6.1. Introduction
127
6.2. Content and Experience
128
6.2.1. Transmission and convergence models ..........................................128
6.2.2. Explanatory strategies ...................................................................129
6.3. Externalist Strategies
130
6.3.1. Teleosemantics and biosemantics...................................................131
6.3.2. Externalism about self-knowledge .................................................133
6.3.3. Phenomenal externalism ................................................................138
6.3.4. Asymmetric dependency ................................................................140
6.4. The Capacities Approach
140
6.4.1. Demonstrative content ..................................................................141
6.4.2. Background knowledge .................................................................143
Representationalism & Anti-Representationalism About Perceptual Experience
iv
6.4.3. Discriminatory capacities...............................................................145
6.4.4. Intuitional content .........................................................................148
6.5. Conclusion
149
7. Conclusion
7.1. Summary
151
7.2. Implications for Representationalism
153
7.2.1. Perception and thought .................................................................154
7.2.2. Constraints upon representationalism ...........................................155
7.2.3. Explanatory roles of representation ...............................................156
7.2.4. Responding to Travis.....................................................................158
7.3. Directions for Future Research
159
7.3.1. Perceptual discriminatory capacities ..............................................159
7.3.2. The possibility of hybrid views ......................................................160
7.3.3. The role of phenomenal character..................................................161
7.3.4. Experience and belief .....................................................................161
Bibliography
163
Representationalism & Anti-Representationalism About Perceptual Experience
v
Acknowledgements
Many thanks to my supervisors, Bill Brewer and Matthew Soteriou, for their
invaluable help and advice throughout the writing of this thesis, to my
examiners, Tim Crane and Naomi Eilan for their generous comments and
suggestions, and to the Arts and Humanities Research Council for funding my
research. Thanks also to Avery Archer, Thomas Baldwin, Alex Byrne, Quassim
Cassam, Wesley Chai, Tony Cheng, Sarah Cooper, Joe Cunningham, Nadja El
Kassar, Juan Camilo Espejo-Serna, Michael Fenton, Craig French, Johan Gersel,
Thomas Hodgson, Ivan Ivanov, Mark Kalderon, Alex Kelly, Hasen Khudairi,
Peter Lamarque, Hemdat Lerman, Guy Longworth, Brent Madison, Mike
Martin, Anna Marmodoro, Richard Moore, John Morrison, Alva Noë,
Christopher Peacocke, Simon Prosser, Louise Richardson, Johannes Roessler,
Kranti Saran, Miguel Angel Sebastian, Susanna Schellenberg, Liam Shields,
Matthias Somers, Karen Somers-Hall, Tom Stoneham, Charles Travis, Michael
Tye and Barnaby Walker for their helpful suggestions, discussions and support
along the way. (Any errors and inaccuracies of course remain my own.)
Most of all, I dedicate this thesis to my wife for her endless love and patience,
and to my parents, who always encouraged me to think.
Declaration
This thesis is my own work and has not been submitted, in whole or in part, for
a degree at another university.
Representationalism & Anti-Representationalism About Perceptual Experience
vi
Abstract
Many philosophers have held that perceptual experience is fundamentally a
matter of perceivers being in particular representational states. Such states are
said to have representational content, i.e. accuracy or veridicality conditions,
capturing the way that things, according to that experience, appear to be. In this
thesis I argue that the case against representationalism — the view that
perceptual experience is fundamentally and irreducibly representational — that is
set out in Charles Travis’s ‘The Silence of the Senses’ (2004) constitutes a
powerful, but much misunderstood and neglected argument against this
prevailing philosophical orthodoxy.
In chapter 2, I present an interpretation of Travis’s arguments that poses a
dilemma for the representationalist concerning the indeterminacy and
availability of perceptual content. Chapters 3 and 4 evaluate a variety of
arguments in favour of such content based upon the nature of appearances, or
‘looks’, including those by Byrne (2009), Siegel (2010) and Schellenberg
(2011b), each of which I find to be problematic. Finally, chapters 5 and 6
examine the relationship between representational content and phenomenal
character, i.e. what perceptual experience is subjectively like, outlining some
potential responses to Travis’s anti-representationalism. These include the
external individuation of content and self-knowledge, and the operation of
perceptual discriminatory capacities (the latter of which does not necessarily
favour a representationalist account of experience).
I conclude that Travis’s arguments establish substantive constraints upon the
nature and role of perceptual content. Moreover, I argue that the debate centres
less upon the existence of such content than its explanatory role, particularly in
relation to phenomenal character and the contents of other mental states: belief,
intention, thought, knowledge, and so on. This in turn highlights the need for
representationalists to better clarify the role of the contents their theories posit,
and why such theories constitute a better explanation of the relevant
phenomena than the corresponding non-representational view.
1. Introduction
Representationalism & anti-representationalism
1.1. Background
The notion that perceptual experience is representational or has representational
content has become something close to orthodoxy in recent philosophy of
perception. Drawing upon the notions of belief and thought, philosophers have
sought to characterise perceptual experience in terms of its ‘content’, or how, in
experience, a perceiving subject ‘represents’ the world as being. An alternative
philosophical tradition, however, sees perceptual experience as being essentially
presentational. That is, to perceive something consists in being immediately and
directly acquainted with some mind-independent object and its properties. To
the extent that this presentational view is incompatible with the more formal
notion of representation — and it is not clear that they are incompatible — some
of its advocates are concerned to argue that there is no need, or room, for the
notion of representation in a philosophical account of perceptual experience. It
is, to borrow an image from Wittgenstein (1953: §271), a wheel that may be
turned, but that does not form part of the mechanism.1 Indeed, some doubt the
very coherence of the representationalist account.
This divergence of views may be traced back at least as far as David Hume,
who famously argued that the objects of perception cannot be the familiar
mind-independent objects of the physical world, but are rather ideas or ‘sense
impressions’ in the mind (Hume 1999: XII.i, 201). Whilst few modern-day
philosophers would endorse Hume’s ‘argument from illusion’ for the conclusion
that the objects of perception are mental and not physical in nature,2 the above
views can be seen as offering different responses to the challenge it presents. The
presentational view rejects Hume’s argument from illusion as unsound,3 instead
1
A comment that is echoed by Travis (2004: 86).
2
Robinson (1994) being a notable exception.
3
Cf. Reid 1997: VI.xxii, 186; Snowdon 1990.
Chapter 1, Introduction
2
characterising perception as having an ‘act–object’ structure in which the subject
is directly confronted by, or acquainted with, external objects and/or their
properties. This yields a relational conception of experience according to which
perception cannot be explained in terms of anything that omits either of its
relata (e.g. representation). The representational view, on the other hand,
accepts Hume’s contention that perception cannot be accounted for in terms of
the presentation of external objects alone. Unlike Hume’s ‘ideas’, however,
representational contents are not themselves perceived, but rather are a way of
capturing the connection between perceptual subjects and objects in a
metaphysically neutral way — something that many advocates of the
presentational view consider to be either misguided or impossible. The precise
reasons for thinking that subjects perceive the world by representing it (albeit
more or less distinctly) to be some particular way, however, are less than clear,
and have become a matter of considerable dispute within the literature.
In this thesis, I aim to contribute to this debate by articulating and
evaluating a series of challenges to the view that perceptual experiences have
specific representational contents, and consequently whether representation is a
necessary condition for perception. In particular, I examine a number of
arguments arising from Charles Travis’s ‘The Silence of the Senses’ (Travis
2004) that purport to show that there is no role for the notion of representation
in a philosophical account of perceptual experience. If sound, these arguments
present a direct challenge to the representationalist strand of Hume’s legacy,
which has often — somewhat misleadingly in my view — been placed in
opposition to a so-called relational view of experience. According to the
relational view (of which there are many variants), perceptual experience is not
most fundamentally a matter of representing the world to be some way, but of
standing in a particular relation — the perceptual relation (Crane 2006)— to the
mind-independent objects of experience. It is therefore fundamentally
presentational in character. Supporters of such a view include Charles Travis
(ibid; forthcoming), John Campbell (2002), M. G. F. Martin (2002b) and Bill
Brewer (2011), all of whom deny that the notion of representation plays a
fundamental role, or in some cases any role at all, in a philosophical account of
perceptual experience.
In what follows, I examine the effectiveness of Travis’s arguments in ruling
out various forms and accounts of representational content in order to clarify
both (i) the nature of the challenge to the representational view, and (ii) the
extent to which this supports a non-representational view of perceptual
experience. I conclude that the arguments set out in chapter 2 below present a
serious and important challenge to the prevailing representationalist orthodoxy
— one that has yet to be adequately addressed by its advocates. Moreover, these
arguments offer a means of framing the debate in terms of the theoretical
commitments and explanatory role of representational content — in particular
its role in explaining the subjective phenomenological character of experience —
in a way that helps to clarify the various issues at stake. Thus, instead of the
potentially misleading dichotomy between ‘representational’ and
Chapter 1, Introduction
3
‘presentational’, or relational, views of experience we find that there are (at
least) two distinct roles that representational contents are thought to play, one
of which turns out to be highly problematic. Thus, not only do Travis’s
arguments rule out many otherwise seemingly plausible accounts of
representational content (chapters 3 and 4), but they also present a challenge to
the representational theory’s ability to account for how we are able to recognise,
or come to know, the representational contents of experience (chapters 5 and 6).
Indeed, the issue of recognisability, as I will call it, turns out to lie at the heart of
the debate.
1.2. Perceptual Experience and Representation
Before providing an overview of the various arguments and chapters of my
thesis below, it will be useful to first establish some terminology concerning the
concepts to which it relates. Foremost amongst these are the notions of
perceptual experience, phenomenal character and representational content.
1.2.1. Perceptual experience
The term ‘perception’, at least as it applies to the present debate, refers to
sensory perception: touch, taste, smell, sight, hearing, and so on. It is therefore
distinct from what might be called, in a more general sense, mere perception, as
this term is sometimes used to emphasise the subjectivity of human belief or
experience. Nevertheless, there is a distinctively subjective aspect to sensory
perception in terms of the subjective character of the experiences that individual
perceptual subjects enjoy, or what I will call their phenomenal character (1.2.2).
In the sense in which I use the term below, an experience is a perceptual
episode in one or more sense modalities that occurs to some individual subject
and is of some determinate duration.4 Such token (i.e. particular) occurrences of
experiences might more properly be called ‘experiencings’, since the term
‘experience’ is potentially ambiguous between a general experience-type (a
universal) and its particular instantiations. Nevertheless, I do not propose to
take a stand on this issue here. Similarly, in referring to (perceptual) experience
simpliciter, I do not mean to exclude the possibility of other types of
experiences, such as thinkings, feelings, imaginings, and so on. Since this thesis
is primarily concerned with the philosophy of perception, however, I will use the
term ‘experience’ to refer to perceptual experience throughout, except where
otherwise noted.
Somewhat unfortunately, philosophers have tended to take visual perceptual
experience to be paradigmatic of perceptual experience in general. This is
perhaps due to its rich and complex structure, which is more obviously
amenable to introspection than, say, touch or smell. Alternatively, it may be
because out of all the external senses, vision most obviously seems to connect us
4
I will remain neutral on the question of whether experiences can be literally instantaneous,
or must be of non-zero duration, though philosophical considerations concerning our
experience of the time and change point to the latter (Gallagher 1998).
Chapter 1, Introduction
4
with external objects in a — at least prima facie — peculiarly immediate and
direct way. Ironically, these very features may be what sets vision apart from the
other senses as something of a special case, at least in terms of its complexity,
though each of these points can to a greater or lesser extent also be said to apply
to other sensory modalities. Indeed, many advocates of representational content,
whether explicitly or implicitly, take their accounts to generalise more or less
unproblematically across the senses. For present purposes, however, I will
follow philosophical convention in taking visual experiences to be at least
illustrative of perceptual experience in general, if not paradigmatically so. As
such, none of the main arguments I give below are intended to be peculiar to
vision. In cases where particular senses — smell, touch, proprioception, etc. —
may be thought to pose particular difficulties or problems, I will set these to one
side to concentrate on the more general point. It is a substantive question,
however, to what extent both representational and relational theories of
perception apply to each of the sense modalities, as well as to the multi-sensory
nature of perceptual experience as a whole. These, however, are not problems I
consider in detail here.
There exists a further issue concerning whether ‘perception’ and its
cognates, as well as ‘seeing’, ‘hearing’ and so on, should be taken as success
terms. Some philosophers, as well as many non-philosophers, use perceptual
terms to denote only genuine (i.e. non-deviant or illusory) perceptual episodes.
On this usage, a subject’s ‘seeing’ a tree, for example, entails the existence of the
tree that is seen. Or, to put the point another way, if there were no such tree,
then they could not have seen it. On another usage of ‘seeing’, however, one
might be said to ‘see’ a dagger even though one is exposed to no such object —
when experiencing a hallucination, for example. This use of ‘seeing’ is neutral
about the existence of its objects, instead answering to the disjunctive criterion
that either the perceptual episode is a genuine perceptual seeing (i.e. that there is
such an object) or it is subjectively indistinguishable from such an experience
(e.g. it is a hallucination or imagining). What goes for ‘seeing’ also goes for
(mere) ‘perceiving’ such that many philosophers take sensory imaginings or
hallucinations to be cases of perceptual experience, as opposed to some other
kind of experience that is merely subjectively indistinguishable from it. On this
terminological point, I take ‘perception’ and ‘experience’ to be committed to the
existence of the objects or properties to which they refer; i.e. they are success
terms. In deference to common usage, however, I will extend the term
‘perceptual experience’ to include hallucinations and other perception-like
episodes, and not only genuine (i.e. appropriately caused veridical) perceivings.
This should not be taken to beg important philosophical questions concerning
whether hallucinations and genuine perception are of the same psychological
kind — a point that disjunctivists deny.5 However, for convenience and
consistency with the literature it will be useful to group all such experiences
under a single heading.
5
Hinton (1973), Snowdon (1990; 2005), McDowell (1982; 1986), Martin (2002; 2004;
2006).
Chapter 1, Introduction
5
1.2.2. Phenomenal character
It is intuitively evident that perceptual experiences have a certain subjective
aspect that constitutes (or can constitute) part of the subject’s conscious state. I
will call this its phenomenal character. The phenomena in question are perfectly
familiar ones and reflect the sense in which, for example, the experience of a red
tomato differs from that of a green one, or the feel of a rough surface differs
from the feel of a smooth surface, and so on. This may variously be described as
‘subjective character’, ‘phenomenology’, ‘what it’s like’, ‘raw feel’, and so on,
and is no doubt due to some combination of the qualities of the relevant
experience, its objects, and the particular sensations that they evoke. The precise
definition and explanatory role of phenomenal character, however, are
contentious — indeed, this turns out to be one of the central issues in the present
debate (chapters 5, 6). Some theorists reject the idea that perceptual experiences
have any additional qualitative aspect over and above their content, whether
this be representational or object-involving. Nevertheless, it is important to
mark the role that phenomenal character plays both in conscious experience and
in philosophical accounts of perception more generally.
For present purposes, I take the term ‘phenomenal character’ to be neutral
both between the various metaphysical accounts of it (e.g. Naïve Realist or
representational) as well as between what is subjectively distinguishable to the
subject having the experience and what M. G. F. Martin (2002a: 186) has called
its ‘phenomenal nature’. Some philosophers take it as definitional that two
experiences which are subjectively indistinguishable from one another — e.g. a
genuine perception and a qualitatively matching hallucination or illusion — have
the same (i.e. type-identical) phenomenal character. Others, notably Martin and
other Naïve Realists, argue that an experience’s phenomenal character extends
beyond what is discriminable to the subject so that two qualitatively matching
experiences may possess differing phenomenal characters, the precise nature of
which may not be apparent to the subject. For the avoidance of confusion, I
take phenomenal character to include at least what is discriminable to the
subject in terms of the subjectively distinguishable phenomenology of perceptual
experience, but that it may extend to aspects of experience that are not readily
available to introspection.6 Thus, according to this usage, an experience’s
phenomenal character may, though need not, outstrip what is subjectively
distinguishable to the subject of the relevant experience.
1.2.3. Representational content
Since I will be considering whether experiences have representational content, it
is important to be clear about what does and does not constitute
‘representation’ in this context. Paradigm cases of representation include: man6
The term ‘introspection’ is another somewhat unfortunate term of art, suggesting as it does
a perception-like faculty for gaining self-knowledge concerning the contents of one’s
experience. For present purposes I take this term to be synonymous with first-personal
reflection upon experience without prejudicing the nature of, or processes involved in, selfreflection (cf. 6.3.2).
Chapter 1, Introduction
6
made artefacts such as photographs, which represent the visible appearances of
objects; maps, which represent spatial and geographical features; and language,
which represents the content of thoughts and discourse. Representations are
standardly thought to admit of a content–vehicle distinction such that wherever
we find representation, we find both representational vehicles — photos, maps,
utterances, etc. — and representational contents, i.e. what is represented by those
vehicles. It is therefore of the essence of representation that what does the
representing is logically distinct from what is being represented — in this case the
objects and properties of perception. 7
Applying the notion of representation to the realm of mental phenomena is
notoriously problematic, but typically involves the attribution of content to a
series of representational vehicles that includes the brain and central nervous
system.8 It is a substantive philosophical question whether the vehicles of mental
representation and content are limited to the brain and body or whether they
extend to features of the subject’s physical or linguistic environment. This
question has particular implications in the philosophy of perception, this being
the faculty (if any) by which subjects become aware of, or are connected to,
their surrounding environment, thereby creating the possibility of objective
thought. Indeed, the ability of our perceptual faculties to ground thought and
knowledge is one of the issues that drives much of the present debate. According
to various forms of semantic externalism, the content of mental representations
not only supervene upon the subject’s bodily state, but also upon aspects of their
environment (Kripke 1972; Putnam 1975; Burge 1979, 1986). Similarly, various
forms of vehicle externalism attribute a representation-bearing role to objects
and events beyond the subject’s physical body (Hurley 1998: ch. 8). Again, I will
not take a stance on these issues here except insofar as they impact upon the
status and role of representational content in perception (chapters 5 and 6). My
thesis, however, concerns the contents of perceptual experiences rather than
their vehicles as I take the precise nature of the latter to be a primarily empirical
question. The nature of perceptual content, however, is not necessarily
something that can be settled solely by scientific investigation since it also
concerns the function(s) and explanatory role(s) of such content in accounting
for a subject’s mental life, thought and behaviour. There is also, as I discuss in
chapters 2, 6 and elsewhere, room for the idea that experiences may have
multiple contents, each fulfilling a different explanatory role (cf. Chalmers
2006; Crane forthcoming). This approach is not, however, immune to the
criticisms set out below (2.2.2).
Central to the notion of representation and the content–vehicle distinction is
the idea of one thing varying in accordance with another, or covariance.
Contrary to Berkeley, as Reid (1977: V.xiii, 74) pointed out, there is no need for
mental representations to literally resemble their contents — indeed, the idea
seems absurd. My belief that I am seeing a red rose is not itself red, nor is it a
7
A notable exception is Sedivy (2004), who has questioned whether this distinction is
applicable to mental representation.
8
Unless otherwise noted, by ‘content’ I mean representational content.
Chapter 1, Introduction
7
rose, and so on. The relevant covariance must therefore take place across
different qualities, such as correspondence between the perceptible attributes of
some external object and the firing patterns of neuronal assemblies in the optic
nerve and brain, for example. Such covariance is not, however, sufficient for
representation in the sense that is relevant here. In order for there to be
representation, there must also be the possibility of misrepresentation such that
the relevant content can either be accurate or inaccurate. For example, if I think
that some actual violets are red when they are in fact blue, then my thought
does not accurately represent the world, and is therefore non-veridical, or false.
Conversely, when the conditions specified by my mental content are satisfied,
then my thought can be said to be an accurate representation, and so is
veridical, or true. The notion that content entails a series of accuracy or
veridicality conditions is central to the notion of representation outlined here,
though the precise nature of these conditions is open to debate. Any notion of
content that rejects this entailment is therefore not a form of representation in
the sense that I mean here, but rather relates to a more generic use of the term to
mean what a perception (thought, belief, etc.) is of — or what I shall call its
objects.9
Complications arise when we consider that even representational contents
which are satisfied, and so veridical, can be the result of deviant perceptual
experiences — the direct stimulation of the brain via electrodes, or perceptual
hallucinations, for example. In such cases, even though the content of the
experience may be accurate, the mental state that bears that content is not
related to its objects in the appropriate way, yielding a subjectively matching
hallucination.10 Thus, whilst genuine perception (i.e. not including perceptual
illusions and hallucination) is by definition veridical, not all veridical (in the
present sense) perceptual experiences are genuine perceptions. According to the
representational view, then, a perceptual state must both (a) possess
representational content whose accuracy conditions are satisfied, and (b) be
appropriately related — whether causally or otherwise — to the objects that
make the content veridical in order to constitute genuine perception.11
Opponents of the view deny that any such characterisation can be given. For
present purposes, however, I will differentiate between veridicality, meaning
accuracy or satisfaction of content, and types of experience — perception,
hallucination, illusion, and so on.
The above conditions on representational states — the content–vehicle
distinction, the ability to misrepresent, veridicality — rule out a large number of
ostensible cases as being representational in the appropriate sense. These
include, though are not limited to:
9
Cf. Martin forthcoming a: ch. 2.
10
11
Sometimes called a ‘veridical hallucination’; e.g. Johnston 2006: 273.
Siegel (2010: 36) calls these conditions ‘weak veridicality’ and ‘strong veridicality’,
respectively, where the latter entails the former. However, this terminology is potentially
misleading as it places what are arguably two quite different kinds of conditions under the
same banner.
Chapter 1, Introduction
8
(i)
Causal traces, such as tree rings, which are sometimes said to ‘represent’
the age of the tree. Here, ‘representation’ is synonymous with mere
causal correlation since, in the normal case, the number of rings
correspond to the age of the tree in years. However, as Travis (2004: 58–
59) points out, this cannot be considered representation in the stronger
sense since we would not say that number of rings misrepresented (i.e.
represented falsely) the tree’s age where the number of rings did not so
correspond. Similarly, the mere activity of neurons in the brain or
nervous system — for example in the retina or visual cortex — in response
to external stimuli is insufficient for representation, though under
appropriate circumstances such activity may constitute vehicles of
representations. Such stimulus–response patterns are merely causal traces
of external events and interactions that, whilst they may carry
information, do not amount to representation unless part of a more
complex intensional system.12 Rather, such traces indicate or are
factively correlated with their immediate or distal causes.
(ii)
Stand-ins, markers or proxies such as a piece in a board game or a pin in
a map. In the final case, such a marker may be said to ‘represent’ a
location of some real-world object or geographical feature that
corresponds to its position on the map. However, strictly speaking, it is
the location of the marker within the representational system of the map
that represents a geographical location, and not the marker itself. Such a
marker does not represent a real world object unless it depicts it in some
literal sense — as a picture of church depicts a church, for example.
Similarly, a piece in a board game does not represent the player in the
sense that is relevant here, but rather serves as a proxy whose location,
rather than the marker or piece itself, represents that player’s location in
the game. Again, such ‘stand-ins’ may form part of, or draw upon the
representational properties of, a more complex representational system,
but they do not themselves represent in the sense of ‘representation’ that
is relevant here.
Whilst it is commonplace to talk about each of the above examples as forms of
‘representation’, they are not the kinds of representation that theorists take to
constitute mental representation. This is perhaps better captured by the
philosophical notion of intentionality or aboutness, i.e. intentional directedness
towards some object or property which that state or content is about. Indeed,
since all other kinds of representation, whether directly or indirectly, owe their
representational status to some form of mental representation, the notion of
intentionality may be thought to have a certain kind of priority or privileged
status within the debate about representational content.
12
In neuroscience and other scientific literature, ‘representation’ is often used to denote just
this sort of information carrying. This is, however, a weaker sense of representation than the
one intended here.
Chapter 1, Introduction
9
In the following chapter I establish the more precise notion of prepresentation — intended to suggest ‘perceptual representation’ — with which I
will be concerned throughout the rest of this thesis. It is an open question at this
point whether perceptual experience, as I have defined it, involves prepresentation, and if so, what the conditions for p-representation are. Indeed,
whether there is such a thing as p-representation, and what form its contents
might take, are two of the main questions with which I engage.
1.2.4. Representationalism and relationalism
As suggested above, much of the recent debate about representational content in
experience has been concerned with the distinction between views that accord
representational content a fundamental, and so irreducible, role in a
philosophical account of perceptual states, and those that deny that it has any
such role. I will label these positions representationalism and antirepresentationalism, respectively.
The central claim of representationalism can be summarised as follows:
REP
The notions of representation and representational content make
a fundamental, and so ineliminable, contribution to explaining
the nature and role of perceptual psychological states in thought
and behaviour.
Anti-representationalism is simply the denial of this claim. Call this NREP. Note
that representationalism is not to be confused with the position commonly
known as intentionalism, which is a thesis about the representational basis of
phenomenal character (Byrne 2001). Intentionalism presupposes
representationalism in the sense that I use that term here, but it is not identical
to it since one could coherently hold representationalism to be true and yet
reject intentionalism (chapter 5). Unfortunately, some philosophers have used
the term ‘representationalism’ as a synonym for intentionalism, but I propose to
keep them separate.
A further position claims that the perceptual relation between subject and
object plays a similarly irreducible role in the explanation of perceptual
psychological states. I will label this position relationalism. Relationalists claim
that:
REL
The subject’s relation to the mind-independent objects (and/or
their properties) of sensory perception makes a fundamental, and
so ineliminable, contribution to explaining the nature and role of
perceptual psychological states in thought and behaviour.
Note that relationalism, as I have formulated it, is a claim about genuine (i.e.
non-deviant) perception. REL is therefore silent on the status of non-veridical
perceptual experiences, such as hallucinations and illusions. Anti-relationalism is
simply the denial of REL. Call this NREL.
Chapter 1, Introduction
10
Now, it is a substantive question which combinations of these theses are
compatible with one another; that is, whether it is possible, or indeed coherent,
to be both a representationalist and a relationalist — or indeed an antirepresentationalist and an anti-relationalist (the ‘pro’ and ‘anti’ versions of each
view are obviously incompatible with one another). Those who take these
positions to be compatible can be said to adopt some form of hybrid or
compatibilist view concerning the nature of perceptual states, as discussed in
chapters 5 and 6. The conjunction of REP and REL — call this COM — therefore
entails that:
COM
Both representational content and the subject’s relation to the
objects or properties of perception play a fundamental, and so
ineliminable, role in explaining the nature and role of perceptual
psychological states in thought and behaviour.
There are further substantive philosophical questions as to whether REP, REL
and COM are plausible or coherent positions, and what the motivations for
holding them might be. These chiefly fall under the headings of
(i)
Psychological: explaining the nature or types of perceptual states for the
purpose of giving a comprehensive account of thought and behaviour;
(ii)
Epistemological: explaining the justificatory role of such states in the
ability of conscious subjects to gain knowledge and awareness of their
environment;
(iii)
Phenomenological: accounting for the distinctive nature and
contribution that the phenomenal character of perceptual experiences
make to the mental life of perceptual subjects.
(iv)
Metaphysical: accounting for the nature of the relation between
perceivers, experiences, objects and properties in the world.
These explanatory projects are not mutually exclusive, nor do they neatly map
onto the taxonomy of views I sketched out above. Indeed, a large part of this
thesis will be devoted to clarifying and untangling the roles of the above claims
and positions in satisfying these various desiderata for a philosophically
satisfying account of sensory perception, as is summarised below.
1.3. Overview
Chapter 2 presents an interpretation of Charles Travis’s influential but, in my
view, often misunderstood paper, ‘The Silence of the Senses’ (2004;
forthcoming). In it, Travis aims to show via a series of arguments that the
notions of representation and representational content have no place in a
philosophical account of perceptual experience. Foremost among these
arguments is the ‘argument from looks’ in which Travis sets up a dilemma for
Chapter 1, Introduction
11
the representationalist about perceptual experience. In order for perceptual
representations, or what I call p-representations, to perform the role their
proponents assign them, they would, according to Travis, have to pick out some
way that the world must be in order for them to be veridical; i.e. their accuracy
conditions. However, p-representations are incapable of doing this in a way that
is consistent with our being able to recognise, i.e. grasp or gain reflective access
to, those representational contents. Since such recognisability is, according to
Travis, a necessary condition for representation, there can be no such
representations, and so representationalism is false.
I examine these ideas at greater length in chapters 3 and 4 which consider a
number of oft-cited objections to Travis’s arguments, along with the role that
perceptual appearances — ‘looks’ in the case of vision — play, respectively.
Chapter 3 addresses objections to Travis made by Alex Byrne (2009), Susannah
Siegel (2011) and Susannah Schellenberg (2011b). These centre upon Travis’s
notion of ‘looks-indexing’, according to which the representational contents of
perceptual states are recognised by the subject in virtue or because of how
things, in those experiences, look (mutatis mutandis for non-visual modalities).
Byrne concurs with Travis that perceptual states cannot be looks-indexed, but
nevertheless goes on to propose a kind of content based on the phenomenal
character of experience that he alleges is capable of meeting Travis’s
recognisability condition. I argue that Travis’s argument from looks is equally
applicable to such ‘non-comparative looks’, which do not — as Byrne, Siegel and
Schellenberg seem to think — constitute a satisfactory response to it. Indeed, the
argument simply reiterates at the level of non-comparative looks with the same
problematic outcome for representationalism as for comparative looks. Further
arguments from Siegel and Schellenberg similarly fail to establish REP due to
equivocation or because they beg the question against the opposing antirepresentationalist view.
Chapter 4 takes up the question of whether it is possible to ground any
satisfactory account of p-representation on the semantics of ‘looks’ or the nature
of appearances. Since phenomenal character, at least insofar as this is accessible
to introspection, is (pace Byrne) equivocal between different representational
contents, it is incapable of fixing univocal, mind-independent accuracy
conditions for p-representation. The only accuracy conditions that are fixed in
this way are, as argued in chapter 3, the mind-dependent conditions for the
subject’s having that very experience.13 This does not, however, constitute
grounds for thinking that such contents generalise to all suitably individuated
experience-types in the way that representationalism requires, as my argument
against Siegel shows. Furthermore, the availability of alternative comparative
analyses of the semantics of ‘looks’, such as Martin’s ‘parsimonious’ view
(Martin 2010) and Brewer’s ‘relevant similarities’ (Brewer 2011), means that
any argument for representational content on the basis of looks is at best neutral
in the debate between representationalism and anti-representationalism about
experience. Concerns about the contents of non-veridical perceptual states turn
13
Cf. Glüer (2009).
Chapter 1, Introduction
12
out to be similarly equivocal, though do highlight an important methodological
difference in the construction of a theory of perception — namely, whether to
consider perception from a primarily epistemological (i.e. knowledge-based) or
primarily phenomenological (i.e. subjective character-based) perspective. Such
methodological differences, however, are not easily resolved on the basis of the
distinctions or concerns outlined here, and to a large extent are as much
determined by the outcome of the present debate as they inform it.
In chapter 5, having rejected the notion of looks-indexing, as well as the
possibility of grounding p-representational content in appearances to the
subject, I examine some further responses to Travis’s anti-representationalism.
These divide into two categories: (i) those that reject and (ii) those that attempt
to satisfy Travis’s recognisability requirement. I argue that the first of these
routes is successful in delivering a kind of representational content, but one that
cannot — at least not without substantive further theoretical commitments and
exposition — explain the phenomenal character of perceptual experience. This
amounts to a rejection of the intentionalist claim that the phenomenal character
of experience supervenes upon representational content and so favours a nonrepresentational account of perceptual phenomenology. The second route aims
to satisfy the recognisability requirement, but faces a further question with
regard to the contribution of phenomenal character to perceptual awareness —
the so-called ‘phenomenological objection’. I argue that this objection is not well
motivated and so without substantial further argument does not necessarily
represent a genuine problem for representationalism.
In chapter 6, I address the question of whether any views that explain the
individuation and recognisability of p-representational content via a single
common element are able to meet the challenge posed by Travis. (Whilst it is
difficult to rule out the possibility of independent explanations for these two
aspects of perceptual experience, I argue that there are substantial theoretical
barriers to doing so.) Teleosemantic and other externalist accounts of prepresentation are particularly problematic in this regard due to the externally
individuated nature of the content that they posit. One solution to this is to
adopt an externalist account of self-knowledge. This satisfies Travis’s
recognisability requirement, but leads to a form of externalism about
phenomenal character, or phenomenal externalism, that many philosophers find
implausible or problematic. A more promising line of response lies with the
conceptualism of McDowell (1998), Brewer (1999) and Schellenberg (2011a;
2011b). By accounting for both the individuation of content and its subjective
availability through a single common element — namely, so-called recognitional
or discriminatory capacities — such accounts are potentially able to meet the
challenge that Travis presents. However, this raises a series of further questions
concerning the relation between the operation of such capacities and the
resulting content. Depending on how these questions are answered, a
comparable account may also be given by the anti-representationalist, making
the role of such content less than fundamental. This may explain why such
views have, to a large extent, been abandoned by their original proponents in
Chapter 1, Introduction
13
favour of moderate or strong versions of relationalism. Nevertheless, if these
problems can be overcome, such views offer the potential of unifying
representationalism and relationalism in a form of compatibilism that accounts
for perceptual phenomenology in terms of both representation and the
perceptual relation.
I conclude that Travis’s arguments present representationalists with a
challenge. They must either (a) abandon the recognisability condition, thus
severing the link between representation and perceptual phenomenology, or else
(b) explain how representational content is recognisable to the subject,
supposedly on the basis of perceptual phenomenology, despite being subjectively
indistinguishable from other such contents with differing accuracy conditions.
This in turn highlights the need for advocates of representationalism to be more
explicit about the explanatory role of the representational content of experience,
which is often employed for multiple, sometimes conflicting, purposes.
Relationalist accounts, on the other hand, do not aim to explain perceptual
phenomenology in terms of representational content, and so do not face this
dilemma. Furthermore, I argue that there is no reason why such accounts should
be taken to be incompatible with the existence of representational content,
provided that such content is not taken to explain perceptual phenomenology.
This in turn opens up the possibility of hybrid or ‘compatibilist’ views upon
which both representational content and the perceptual relation have distinct or
combined explanatory roles. In the meantime, Travis’s objections pose a
powerful and important challenge to representationalist views, as well as
offering an important way of sharpening understanding of the debate whilst
ruling out a number of otherwise apparently plausible views of perceptual
content.
2. The Silence of the Senses
The Case Against Representationalism
2.1. Introduction
Kant famously stated that ‘[t]houghts without content are empty, intuitions
without concepts are blind’ (Kant 1998: A51/B75). In ‘The Silence of the
Senses’, Charles Travis (2004; forthcoming)1 argues that perceptual experience
is not only ‘blind’, but ‘silent’, in that it does not deliver representational
content to the subject. Consequently, according to Travis, perception ‘is, in a
crucial way, not an intentional phenomenon’ (ibid: 93). This chapter sets out the
principal arguments that Travis presents for this view along with the conditions
upon perceptual representation that he takes to underpin it. My aim here is to
provide a clear expression of these arguments along with their relations to the
various conditions upon representation that Travis describes. This in turn
enables the construction a taxonomy of various forms of representationalism
according to which of Travis’s premises and conditions they endorse. Responses
and counterarguments, both simple and subtle, to these arguments are
considered in subsequent chapters.
I begin by outlining a notion of representation that I will call prepresentation, short for perceptual representation (2.2), which characterises the
form of representation to which Travis takes many advocates of
representationalism to subscribe. This is captured by a number of conditions
that an adequate notion of perceptual representation must conform to in order
to meet various demands that representationalists place upon it. The second half
of the chapter presents a series of arguments that can be found in Travis (2004)
for the conclusion that perception does not constitutively involve prepresentation (2.3). This helps identify a number of possible responses that the
1
Where possible, citations below refer to the original 2004 version of Travis’s paper. I have,
however, used the amended terminology and other clarifications of the revised version
throughout, which at the time of writing had yet to be published.
Chapter 2, The Silence of the Senses
15
representationalist might make, according to which of Travis’s premises or
conditions upon p-representation they wish to reject. These responses are then
taken up in greater detail throughout chapters 3 to 6, which evaluate the
effectiveness and implications of these arguments.
2.2. Perceptual Representation
The primary target of Travis’s (2004) paper is the view that perceptual
experience is essentially a representational phenomenon. This is the position
characterised in chapter 1 as representationalism, or REP for short (1.2.4). Since
Travis denies that perception is essentially representational, his position
constitutes a form of anti-representationalism, or NREP (ibid.).2 Travis contrasts
such representationalism with the view that perception simply presents the
world to us (as opposed to representing it) thereby making concrete external
objects and properties available for thought, belief, judgement, etc. On Travis’s
preferred view, those aspects of experience which, according to
representationalists, favour the existence of representational content are instead
situated within the processes of judgement and/or interpretation that take place
on the basis of experience, rather than forming part of experience itself. For
example, Travis considers perceptual illusions to be cases in which the subject is
misled by appearances rather than cases of misrepresentation since, according to
him, perceptual experience does not involve any representation (2.3.1).
Travis contrasts his view with those of Davies (1992), McDowell (1994),
McGinn (1982; 1991), Peacocke (1992), Searle (1983), Tye (1995), and others
who take perceptual experience to possess determinate representational content
in both veridical and non-veridical cases.3 The analysis of both genuine
perception and other perceptual experiences such as hallucination and illusion in
terms of a single underlying ‘common factor’, i.e. their representational content,
is one of the central features of the accounts of perception to which Travis is
opposed. I will refer to this as the principle of intentionality, since it is
equivalent to the claim that p-representations can occur in the absence of those
objects that they are ‘about’; i.e. that they have intentional content. Not all
representationalists would agree with this claim, at least its unqualified form.
However, since providing a uniform account of veridical and non-veridical
perceptual experiences is one of the chief motivations for endorsing
representationalism, most if not all representationalists would subscribe to the
following closely related principle:
2
Whether REP is genuinely incompatible with the kind of presentational view that Travis
favours is something I examine further in chapters 5 and 6. For the time being, I simply focus
on explicating and evaluating his arguments against representationalism.
3
I use the term ‘determinate’ in contrast to ‘determinable’, and not to ‘vague’. Accuracy
conditions that involve some degree of vagueness — about the precise number of items that
are present, for example — can, on this usage, still be perfectly determinate.
Chapter 2, The Silence of the Senses
16
Common Content: subjectively indistinguishable veridical and nonveridical perceptual experiences share a common representational
content.
This principle is weaker than a straightforward identity claim between the
representational content of ‘good’ (i.e. veridical perception) and ‘bad’ (i.e. nonveridical) cases. Indeed, such an identity is denied by those forms of
representationalism whose contents incorporate an externally individuated or
demonstrative aspect (6.3, 6.4.1). All of the above principles, however, are
denied by Travis, who regards non-veridical perceptual experience as being
explained by the phenomenon of misleading as opposed to the existence of any
such common representational element (Travis 2004: 64).
Travis characterises the target forms of representationalism as endorsing
four key constraints upon the notion of representation that features in accounts
of perceptual experience. These constraints, set out in detail below (2.2.1–4), are
largely interrelated and collectively define the notion of representation to which
Travis is opposed (2.2.6). To avoid prejudicing the issue of whether there exists
any other form of representation that does not satisfy Travis’s constraints (see
5.2.4), but which meets at least some of the representationalists’ requirements, I
will call such representation p-representation. This term is intended to be
suggestive of ‘perceptual representation’ and is a placeholder for the kind of
representation that advocates of representationalism, pace Travis, hold to exist.
According to Travis, only p-representation can support his opponents’ claims
concerning the role of representation in perceptual experience (2.3). On his
view, such a notion is unmotivated and the constraints upon it inconsistent,
such that no form of representation could ever satisfy them. Consequently, there
can be no such thing as p-representation. Travis's aim, therefore, is to
demonstrate that representationalism, at least in its most common form, is
incoherent, and so that perceptual experience cannot be p-representational.
It is important to realise that Travis’s arguments do not rule out the
possibility that there is some sense in which representational contents — in the
form of phenomenal content, for example (3.2.3, 4.2.6) — may be associated
with individual perceptual experiences; for the purposes of neuroscience or
behavioural psychology, for example. Rather, the claim is that such content
does not — indeed, according to Travis, it cannot — capture or exhaust the
nature of perceptual experience. In particular, as will become clear in later
chapters, these arguments present specific problems in accounting for the
phenomenal character of experiences in representational terms — a position
known as intentionalism. Furthermore, if such content fails to satisfy the
conditions below, it would not, in Travis’s view, correspond to any familiar
notion of representation, and consequently be unable to sustain the claims that
representationalists make for their theories, such as the ability to explain the
contents of perceptual beliefs, judgements or illusions. Such alternative uses of
the term ‘representation’ are therefore besides the point, which is to exhaustively
Chapter 2, The Silence of the Senses
17
characterise perceptual experience in terms of its representational content —
something that Travis and other anti-representationalists hold to be impossible.
2.2.1. Representing such-and-such as so
The first condition by which Travis characterises his target notion of
representation is that ‘[t]he representing in question is representing such-andsuch as so’ (Travis 2004: 58). Thus, p-representation represents the world as
being some way or other; i.e. it has determinate representational content (1.2.3).
Travis contrasts this form of representation with that of ‘an effect or trace of
something’, such as the way that tree rings may be said to ‘represent’ the
amount of growth or climatic conditions in a particular year, or some kind of
‘stand-in, or substitute, for what is represented’ (ibid.). The former qualification
rules out much of the kind of ‘representation’ that, for example, neuroscientists
find in the brain and nervous system, which are mere causal traces of perceptual
stimuli. Such traces cannot, by definition, occur in the absence of what they are
causal traces of, and so do not accord with the principle of intentionality. Nor
can such traces misrepresent. The latter qualification is intended to rule out the
sense in which a piece in a game ‘represent[s] an infantry division in a game of
strategy’ or ‘[a] squiggle on a map may represent the Lot’ (ibid.). Such
representations provide a stand-in or proxy for another object by means of some
established rule or convention (1.2.3). P-representations, on the other hand,
possess content in virtue of their structure or logical form in a way that is not
merely conventional, but revealed through their role in rational thought and
behaviour, either by introspection or by reflection upon the nature and
character of experience (see 2.2.4).
Central to the idea of representing such-and-such as so is the notion of
correctness or accuracy conditions. Such conditions specify the situation, or
situations, under which the relevant content is accurate or veridical (1.2.3). In
the case of a given perceptual experience, this occurs when the content of that
experience corresponds to the way that things, independently of how the subject
perceives them as being, actually are. P-representational content, then, must be
assessable for truth or accuracy. 4 That is to say, for any given representational
content, there is a definite and principled answer as to whether it is accurate or
not, depending upon its correspondence with the subject’s immediate perceptual
environment. Thus, such states purport to represent the way that the subject’s
environment objectively is, rather than how things subjectively seem or appear.
(A state that represented only subjective features of the subject’s own mental
state would not qualify as objective in the relevant sense — see 3.2.3.)
Consequently, on the representationalist’s view, the phenomena of seeming or
appearing are a matter of the subject’s representing the world to be some
particular way, rather than anything that features explicitly in the content of
such representations. Appearing, we might say, is all in the attitude (cf. Travis
2004: 60).
4 ‘Truth’ normally indicates a bivalent notion of correspondence, whereas ‘accuracy’ may
admit of degrees. For present purposes, it does not matter which of these terms we use.
Chapter 2, The Silence of the Senses
18
We can summarise this first condition upon the notion of p-representation
as follows:
Objectivity: p-representation represents things as being objectively so in
a way that is capable of being assessed for truth or accuracy.
Note that Travis’s use of the phrase ‘representing such-and-such as
so’ (Travis 2004: 58) might be taken to suggest that p-representational content
must be propositional, and so by implication conceptual. However, since two of
his stated targets — namely, Peacocke (1992) and Tye (1995) — both endorse
notions of non-conceptual content, we can assume that Travis intends for his
arguments to apply to both conceptual and non-conceptual content. Henceforth,
I will use the term ‘propositional’ to cover any content that is assessable for
truth, and ‘content’ as being assessable for accuracy, regardless of whether it is
conceptual or otherwise. Thus, Travis may be taken to argue not only against
the existence of conceptual p-representational content, but against any form of
content that possesses determinate accuracy or correctness conditions.
2.2.2. Face value
Travis’s second condition upon p-representation is intended to capture the
representationalist intuition that every perceptual experience has a univocal, i.e.
single and determinate, face-value content. The condition is motivated by the
intuitive notion that things can be just as they appear, and has two distinct
aspects. The first relates to how the way that experience represents the world as
being — its representational content — can be ‘read off’, as it were, the
experience itself. This corresponds to Travis’s Recognisability condition below
(2.2.4). It is not uncommon, however, to talk of taking experience ‘at face
value’, meaning that one simply accepts the diktats of experience without
doubting that things are in fact the way that they appear. The prevalence of
such expressions might be taken to suggest that there is some particular way (as
opposed to many different ways) that any given experience represents the world
as being. It is this aspect of the ‘face value’ claim that is captured by Travis’s
second condition, which may be summarised as follows:
Face-value: p-representations have a determinate and univocal ‘face
value, at which [they] can be taken or declined’.5
This condition forms the basis for what is sometimes called Travis’s
‘indeterminacy objection’ (Schellenberg 2011b: 7), and is a key premise in his
argument from looks (2.3.2).
Two aspects of the above condition are necessary for Travis’s purpose. First,
and most importantly, it entails that p-representations have precisely one
content, as opposed to many or none. Thus, each perceptual experience
5
Travis (2004: 63).
Chapter 2, The Silence of the Senses
19
represents the world to be some particular way, and not a disjunction of
possible alternatives, each of which shares the same perceptible appearance. For
example, if there appears to be a lemon in front of me then, according to the
representationalist, the representational content of my experience includes that
there is a lemon, and not the open-ended disjunction there is either a lemon, a
wax imitation lemon, a lemon-shaped bar of soap, or …, even if each of these
alternatives would appear, from the same location and angle, perceptually
indistinguishable from an actual lemon. An account that meets Face-value may
therefore be contrasted with the view that perceptual experiences have
disjunctive content in which a potentially infinite range of possible ways that the
world might be are ‘represented’, but without any one of these possibilities
being singled out as the content of experience.6 Travis claims that such a view
would make representation in perception incoherent, [therefore] not
intelligibly representation at all. Too many things would thereby be
represented as so at once. There are just too many things things look like.7
The second aspect of Face-value is that p-representations have the kind of
content that can either be accepted or rejected. Thus, when enjoying a given
perceptual experience, a subject is not thereby committed to believing or judging
that something in particular is the case. Instead, experience is supposed to
provide the grounds upon which such beliefs or judgements may be formed or
withheld according to the context and dispositions of the perceiver. This further
consequence of Face-value is addressed below (2.2.3).
In endorsing the Face-value condition, McDowell (1994: 26) states:
That things are thus and so is the content of the experience, and it can also
be the content of a judgment. It becomes the content of a judgment if the
subject decides to take the experience at face value.
Thus, when a perceptual subject has an experience with the content there is a
greenish lemon in front of me, and where they have no reason to suspect the
contrary, they will typically judge, and thereby come to believe, that there is
indeed a greenish lemon in front of them. Conversely, in cases where one has a
positive reason to doubt that things are as they appear — if one knew or had
some reason to believe that the scene were illuminated by green light, for
example — one might conclude that, contrary to the experience's ‘face value’
content, the lemon is not green, but some other colour, and only appears green
due to the unusual lighting conditions. Alternatively, one may choose to
withhold judgement altogether. Furthermore, McDowell takes the contents of
6 Such views should not be confused with the position known as disjunctivism, which claims
that perception and non-veridical perceptual experience are of fundamentally different
psychological kinds.
7
Ibid: 87. The alleged ‘incoherence’ is presumably based on the fact that the various
disjuncts would be inconsistent and the set of possible disjuncts too large to constitute any
recognisable form of p-representation.
Chapter 2, The Silence of the Senses
20
the experience and the corresponding judgement to be identical (though both
are presumably more complex than in this simple example). In this way,
individual experiences provide defeasible warrant for believing that things are
‘just as they appear’, with their contents, or elements of it, figuring in the
contents of perceptual judgements and beliefs. This further commitment,
however, is not mandated by Face-value.
No doubt the principle that the content of any perceptual judgement or
belief is identical to, or straightforwardly derivable from, the content of the
experience is appealing and would be endorsed by representationalists of
various stripes. Such a principle offers a simple and straightforward explanation
of how perceptual judgements and beliefs are formed: one simply comes to
believe what one sees, hears, tastes, and so on. Indeed, this is one of the benefits
that advocates of representationalism suppose to flow from it: that we can judge
or believe things to be just as they appear. Perceptual judgement, for such
representationalists, is simply a matter of a subject’s endorsing the relevant part,
or parts, of the content of one’s perceptual experience. This may be contrasted
with anti-representationalist accounts according to which the process of belief
and judgement formation play a more substantive role, expanding to assume
one of the functions that representationalists take to be fulfilled by experience.
A commitment to such a principle, however, is unnecessary for Travis’s purpose
or for p-representation, which turns on the presence of univocal face value
content as opposed to the role that such content plays in judgement or belief.
We must therefore be careful to avoid to avoid conflating Face-value with the
stronger, theoretically distinct claim that the contents of experience are identical
to those that figure in perceptual judgements and belief.
Accounts of perceptual content that do not satisfy Face-value include the
‘content pluralism’ of Chalmers (2006) and Crane (forthcoming), both of which
entail that experience has not one, but many such contents, or ‘content
relations’, each of which has a distinct explanatory role. Whilst such views are
not obviously targets of Travis (2004), 8 it is nevertheless possible to run the
same arguments against them with respect to each individual content that such
pluralists claim perceptual experience to have. In this case, the arguments will
relate to the individual contents that occur within the context of some particular
explanatory role, rather than the plurality of contents as a whole. This differs
from the disjunctive form of content discussed above that takes experience to
have only a single content with a disjunctive logical structure. In the latter case,
Travis’s arguments will apply to the disjunction as a whole, as opposed to each
of the individual disjuncts as in the case of content pluralism. Thus, Travis’s
arguments may be taken to apply to any representational content that meets the
conditions necessary for p-representation, even where no one content may be
singled out as ‘the’ content of experience.
8
Cf. ibid: 82.
Chapter 2, The Silence of the Senses
21
2.2.3. Represention-to, not representation-by
The third of Travis’s conditions for p-representation distinguishes between two
different forms, or roles, of representation, which Travis calls
‘autorepresentation’ and ‘allorepresentation’ (Travis 2004: 61), respectively.
Autorepresentation corresponds to ‘representing to … oneself’ (ibid.) or taking
something to be so. It is therefore a form of representation-by a conscious
subject. Examples of autorepresentation include belief, judgement and any other
propositional attitude in which a subject accepts something to be true, or indeed
false. Consequently, autorepresentation does not admit of a neutral attitude
towards a content since to autorepresent p is already to accept (or reject) that p.
Such representation is, as Travis puts it, ‘all in the attitude’ (ibid: 60).
Allorepresentation, on the other hand, occurs when whatever is being
represented can be either accepted or rejected; it is representation-to the subject.
Travis goes on to identify two subspecies of allorepresentation, which he calls
‘committed’ and ‘uncommitted’ allorepresentation. In committed
allorepresentation, the content of the representation has some kind of authority
which, all other things being equal, recommends it to the subject, as in the case
of testimony from a normally reliable source. Uncommitted allorepresentation,
on the other hand, has no such authority, there being nothing in the means of
representation to determine or suggest that things are one way rather than
another — English sentences being a case in point.9 Travis argues that only
committed representations can have a face value in the relevant sense, since
uncommitted representation involves no commitment to any particular state of
affairs being the case, and thus is incapable of providing a reason for belief or
judgement (ibid. 62).
The term ‘representation’ may be used to denote both auto- and
allorepresentation, and is therefore ambiguous. The location of a marker on a
map, for example, may (allo)represent one’s own location, whereas one can
(auto)represent something to be the case by deciding that it is so, the difference
being in whether the thinker is the ‘consumer’ or ‘producer’ of the
representation.10 The key point, however, is that since Face-value provides for
the rejection as well as acceptance of contents (cf. 2.2.2), the relevant form of prepresentation cannot be autorepresentation. In contrast to judgement (at least
on some accounts), we cannot choose or decide what we perceive in experience,
but rather its representational content is (on the representationalist view) in
some sense ‘given’ to us. We simply find ourselves, to use McDowell’s phrase,
saddled with content.
We can capture this condition upon p-representation as follows:
9
Travis does not deny that such sentences can have truth or accuracy conditions, merely that
the means of representation (i.e. English) does not itself recommend nor guarantee its
representational content to be true (ibid. 61).
10
Cf. Millikan 1989.
Chapter 2, The Silence of the Senses
22
Givenness: p-representation consists in content being ‘given’ to the
subject (representation-to), not in subjects taking things to be the
case (representation-by).
P-representations are thus defeasible in the sense that, in circumstances where an
ideally rational subject has reason to believe their experience to be deceptive —
when viewing a familiar perceptual illusion, for example — they can judge that
things are not the way that their experience represents them to be. Similarly, if
the subject has reason to believe that their perceptual experience may be
misleading or faulty — when undergoing a blind clinical trial for a
hallucinogenic drug, for example — they may choose to withhold judgement as
to whether their experiences accurately reflect the way things are. Conversely,
where the subject is aware of no such reasons, they will typically judge that
things are the way that they appear.11 This reflects the committed nature of prepresentation, which possesses a form of authority that, under normal
circumstances at least, recommends its content to the subject.
Givenness comes into play in Travis’s argument from recognisability (2.3.3)
and entails that experiences are not perceptual ‘takings’. As such, whatever fixes
the content of experience cannot solely depend upon how that experience is
interpreted, or what it is taken it to indicate. This would in any case arguably be
inconsistent with the role of perception as the basis for judgement or belief,
which places the order of explanation precisely the other way round. If the
content of experience were itself to depend upon such ‘takings’, this would give
rise to a circularity since perceptual takings, in the form of perceptual belief or
judgement, themselves depend upon experience. On accounts of experience in
which perceptual belief and experience are separate and distinct mental states,
the former is generally held to be explanatorily dependent upon the latter,
precluding the possibility of the reverse dependence. This does not rule out the
possibility that experiences may be identified with dispositions to believe, or
‘potential beliefs’ as suggested by Armstrong (1968: 242), provided that such
dispositions are not themselves beliefs; i.e. they are not perceptual ‘takings’. In
either case, however, what a given experience represents is supposed to be
intrinsic to that very experience, with its content being simply ‘given’ to the
subject.
Note that the notion of givenness employed here need not fall into the socalled ‘Myth of the Given’ (McDowell 2008) since there is no suggestion that
experiential contents are independent of the subject’s cognitive capacities (the
thing that is supposed to make McDowell’s ‘Given’ mythical). Rather, the idea
is that, according to representationalism, the faculty of perception is what gives
us representational contents that we are subsequently able to judge as true or
false, guide our actions, and so on. In McDowell’s terms, p-representational
content is ‘given’ with a small ‘g’, not a capital ‘G’. Similarly, we should not
conclude that in order for the contents of experience to be given to the subject in
the relevant sense, there must be some entity over and above experience itself
11
This presumption may of course turn out be incorrect, thus granting scepticism a foothold.
Chapter 2, The Silence of the Senses
23
that does the giving. Just as an arrow on a map can ‘give’ us a spatial location
relative to the features it depicts, perceptual experience itself can give us
information about how things are in our immediate surroundings. The sense in
which p-representational content is ‘given’ is one that holds between an
experience and a subject, rather than between two subjects in the manner of, for
example, verbal testimony. The relevant representing is done by experience itself
rather than, for example, any further sub-personal agent or
‘representer’ (though neither does this rule out the existence of such agents).
2.2.4. Recognisability
The fourth and final condition that Travis places upon p-representation is that
the contents of experience must be recognisable to the perceptual subject. This is
perhaps the most important and contentious of Travis’s conditions, and consists
of two main claims. First, perceptual experience must be recognisable as a form
of representation since ‘you cannot represent things to people as so in a way
they simply cannot recognize as doing that’ (Travis 2004: 63). This is not to say
that we are always, or even normally, aware of being represented to in
perception, or that perceptual subjects are necessarily capable of accurately
describing the nature and content of their experiences — something that is
clearly not the case. Rather, it must at least be possible in principle — perhaps
through philosophical reasoning and reflection — to recognise the
representational nature of experience as such.
Second, and more importantly for present purposes, we must be capable of
recognising the content of our experiences. That is to say, we must be able to
know what it is that is being represented to us, or equivalently, what it would
take for any given perceptual experience to be veridical. Again, this does not
mean that we are in practice able to determine whether the relevant state of
affairs obtains, or to elucidate the relevant accuracy conditions in a formal
manner. Indeed, these tasks may require considerable conceptual and/or
empirical investigation. Rather, the suggestion is that we must be able to grasp
or recognise how the world would need to be in order for perceptual experience
to accurately represent it. When perceptually encountering an object that looks
like a lemon, for example, it should be apparent to the subject — that is to say, it
is cognitively available to them — that their experience is as of a lemon, or that
the experience should lead them to form beliefs concerning lemons and not, say,
potatoes or bars of soap. If the object turned out to be a potato or a ringer for a
lemon, such as an identical looking lemon-shaped bar of soap, then this
discrepancy would be something that is at least potentially discernible to the
subject, perhaps by occasioning surprise or disbelief upon discovering the
unexpected object.
We can characterise this condition upon p-representation as follows:
Recognisability: perceptual subjects must be capable of recognising the
representational content of any given p-representation solely in
virtue of having that very experience.
Chapter 2, The Silence of the Senses
24
By this we mean that the subject must be able to recognise both what their
perceptual experience represents to be the case, and how the world would have
to be in order for their experience to be veridical; i.e. having some intuitive or
implicit grasp of their experiences’ accuracy or truth conditions. This leaves it
open both whether they know that the relevant conditions obtain (something
that cannot be required for perceptual experience alone), and in what such
knowing consists.
The condition is motivated by epistemological considerations concerning
Face-value and Givenness since if experience presents the world to the subject as
being some particular way (Objectivity), then it must be possible to recognise
what way that is on the basis of that experience (5.2.1). This marks out prepresentation as occurring at the personal, rather than sub-personal, level,
making experience on the representationalist model analogous to testimony.
Experiences, we might say, testify or report to the subject how things are, with
the contents of these reports corresponding to individual p-representational
contents. Of course it is precisely this analogy with testimony that antirepresentationalists are concerned to reject, since they deny that experiences
have any such content.
For the time being, I will leave the sense of ‘recognition’ with which Travis
is operating as largely intuitive, though this will require further clarification in
due course (5.2). The relevant sense must allow for the possibility of subjects
being capable of coming to know both the contents of their experiences and that
their experience is representational. However, it need not be defined in terms of
explicit knowledge. Rather, some form of tacit understanding or grasp of the
conditions required for an experience’s accuracy will suffice. These may be
cashed out in terms of practical or cognitive abilities, i.e. knowledge-how rather
than knowledge-that. In particular, such recognisability may (though need not)
manifest itself in the phenomenology of experience, as reflected by its
phenomenal character; e.g. in an object’s looking red or as located at a
particular point in egocentric space.
A notable aspect of the above condition is that the content of perceptual
experience must be recognisable ‘in virtue of the experience itself’. This is
intended to rule out the possibility that such content is recognisable on the basis
of prior or subsequent knowledge, expectations or behaviour, since if this were
the case then it would be unclear in what sense this shows experience to be
representational as opposed to that other thing. Similarly, it cannot be the case
that one recognises the content of experience solely as a result of beliefs and
judgements formed on the basis of it. This would serve to demonstrate that
belief or judgement were representational, but not experience, since it would
give no reason to attribute the representational content to the latter, other than
perhaps as an explanation of how the former get their content, which is
precisely the kind of error that anti-representationalists accuse their opponents
of making. Furthermore, since such forms of autorepresentation entail that the
subject already accepts their contents as true, Givenness would fail to obtain.
Perceptual belief and judgement alone therefore appear to be incapable of
Chapter 2, The Silence of the Senses
25
making the representational contents of perception (if there are any)
recognisable in the relevant sense. Rather, Travis argues, there must be some
special feature or aspect of perceptual experience in virtue of which its
representational content is apparent, as discussed below (2.2.5).
The precise formulation of Recognisability, along with the precise sense of
‘recognition’ in this context, is central to Travis’s argument from looks (2.3.2),
much of which rests upon it. Indeed, if a representationalist were to reject this
condition outright, then this argument (though not some of Travis’s subsequent
arguments) would fail to go through.12 However, such a rejection is not without
its costs, since the basic principle is closely connected not only with Face-value
and Givenness, but with the connection between representational content and
perceptual phenomenology (5.2.3). Indeed, it is a consequence of some forms of
intentionalism that the elements of p-representation that figure in phenomenal
character are in some sense ‘recognisable’ to the subject. Thus, to give up on
Recognisability would entail a denial of some forms of intentionalism.13
Furthermore, without Recognisability it is difficult to see what work the notion
of representation is supposed to be doing, except as a placeholder or term of art.
Such ‘non-recognisable’ forms of p-representation would not constitute a
familiar form of representation as such. If the notion of representation that
representationalists employ is supposed to be the familiar one that we apply to
maps, photographs, language, etc., it would present a serious issue if the content
of perception turned out to be opaque to the experiencing subject. As Travis
(2004: 86) puts it,
[T]hat we are represented to in experience is meant to be a familiar
phenomenon; something we can tell is happening. It is not just events
occurring in visual processing mechanisms of which we are all ignorant. It
should not come to us as a complete surprise someday, to be sprung on us
by future neurophysiologists, that we are thus represented to (uselessly, of
course, since we were all ignorant of it).
Travis’s claim, then, is that p-representational content must be grounded in
a subject’s ability to recognise what is represented to them in experience, which
in turn grounds any subsequent judgements or beliefs made on the basis of it.
Furthermore, there must be some principled way of determining in any
particular case precisely what the content of a given experience is. If not, then
we leave open the possibility that it is the contents of judgements or beliefs that
fix the content of experience, rather than, as representationalists typically claim,
the other way around.
12
I consider such a rejection in connection with Burge’s (2010) account of representational
content in chapter 5.
13
Intentionalists who hold a supervenience rather than identity thesis of phenomenal
character can endorse the Face-value condition whilst rejecting Recognisability since not
every difference in representational content need be reflected in perceptual phenomenology.
Chapter 2, The Silence of the Senses
26
2.2.5. Looks-indexing
Having argued in favour of Recognisability, Travis goes on to ask what feature,
or features, of experience could make p-representational contents recognisable
in the relevant sense. In the case of visual perception, it seems plausible that this
should be the way that things look. On this view, p-representational content is
recognisable to the subject — or ‘indexed’ to use Travis’s expression — on the
basis of how, in the relevant visual experience, things look to the subject.14
Travis notes that the relevant sense of ‘looks’ should not be conflated with the
contents of p-representations themselves, stating (ibid. 63):
I take it that it would be cheating if, say, ‘looks like things are thus and so’
turned out just to mean ‘things are represented to the perceiver as being thus
and so’. Looks in that sense might be representational content; but they
could not be that by which an experience is recognizable as having the
representational content that it does.
Looks, therefore, are supposed to be what enables us to recognise the contents
of experience, thereby making those contents available to the subject. We can
summarise this condition as follows:
Looks-indexing: the content of any given p-representation must be
recognisable to the subject solely in virtue of how, in the relevant
experience, things perceptually seem to that subject.15
It is important to realise that Looks-indexing is a provisional and temporary
assumption not to be confused with Recognisability itself (2.2.4). Rather, Travis
makes this assumption on behalf of the representationalist as a means of
explaining how Recognisability may be met, but it is ultimately dropped in the
final section of Travis’s paper. This presents the representationalist with a
number of options. First, they may wish to endorse both of these conditions and
so specify a notion of ‘looks’ or appearances that satisfies them without falling
prey to the arguments that Travis presses (2.3.3). This is the route that Siegel
(2010), Schellenberg (2011a; 2011b) and others take, and which Travis claims is
impossible to maintain, as I discuss in chapters 3 and 4. Alternatively, the
representationalist may reject Looks-indexing, but retain or reformulate
Recognisability, in which case the onus is upon them to specify an alternative
way of satisfying the latter. This appears to be Byrne’s (2009) preferred option
(chapter 3). Finally, they may reject Recognisability outright, yielding a form of
representational content that is not available to the subject on the basis of
reflection alone, as per Burge (2010). Further variations are possible by
14
15
Similar considerations presumably apply to each of the other sensory modalities.
In general, where Travis uses the term ‘looks’ with specific reference to visual experience, I
will prefer the more generic ‘seems’ or ‘appears’, which I take it also apply to each of the
other sensory modalities.
Chapter 2, The Silence of the Senses
27
weakening or strengthening the notion of ‘recognition’ that is employed by these
conditions accordingly (5.2.2).
A corollary to Travis’s point about looks is that what enables us to
recognise the content of perceptual experience need not be the same as what
fixes or determines that content. Whilst the first requirement is cognitive or
epistemic, the second is semantic, and so the two may come apart. Whilst Travis
does not always clearly distinguish between these different aspects of perceptual
experience, Recognisability is clearly concerned with the former. This leaves
room for forms of representationalism in which what fixes the content of
experience is distinct from whatever makes that content recognisable — one may
be internally individuated and the other externally individuated, for example.
The coherence and plausibility of such views will be addressed in chapter 6
where I argue that only those forms of representationalism that explain both
content and recognisability in terms of a single common factor are likely to be
capable of avoiding Travis’s arguments from looks and recognisability. These
arguments may therefore be seen as providing constraints upon the forms of
representationalism that are capable of satisfying some form of Recognisability
condition.
2.2.6. P-representation
Excluding Looks-indexing, the above conditions jointly characterise the form of
representation that Travis takes representationalists to think is involved in
perceptual experience, namely p-representation. These conditions are not
intended to be exhaustive or definitional, but rather identify key theoretical
commitments that are either explicitly or implicitly endorsed by many, though
not all, representationalists. This in turns helps to differentiate p-representation
from other forms of representation, such as those I reject in 1.2.3. Moreover,
each of these conditions is individually necessary for p-representation insofar as
it underpins the explanatory role that such representations play in
representationalist accounts of perception. Without Objectivity, for example,
perceptual experience could not form the basis for knowledge of objective
aspects of the external world. Without Face-value it would be impossible to
accept or reject the content of perceptual experience. Without Givenness, we
would be incapable of judging or believing on the basis of experience. And
without Recognisability, we would be unaware of being represented to at all, or
be incapable of grasping on the basis of first-personal reflection or perceptual
phenomenology what it is that our experiences represent.
To summarise, then, p-representation is a form of representation in which
things are represented to the subject as being some particular way (Objectivity
and Face-value). Moreover, it does not commit the subject to believing or
judging that things are that way, since such contents may be either accepted or
rejected (Givenness). Where such a judgement is made, its content may,
according to some versions of representationalism, be identical to or else
straightforwardly derivable from the content of the corresponding perceptual
experience, though this is not required for Travis’s purposes. The
Chapter 2, The Silence of the Senses
28
representational content of experience, however, must be discernible — e.g.
knowable — on the basis of first-personal reflection upon the experience itself or
perceptual phenomenology (Recognisability), though this does not require that
subjects are necessarily able to articulate or ascertain the satisfaction of its
accuracy conditions in practice. In the case of visual perception, and mutatis
mutandis for each of the other sensory modalities, this plausibly occurs on the
basis of how, in the relevant experience, things look (Looks-indexing), though
again this condition is not essential to Travis’s argument — indeed, he later gives
reasons for rejecting it.
Travis aims to show that the above notion of p-representation is incoherent,
and so unable to fulfil the explanatory role that representationalists assign to it.
Consequently, given that some variant of this form of representation is defended
by many leading representationalists, perception cannot on those accounts be
representational. This argumentative strategy admits of several possible
responses. First, representationalists may choose to deny the validity, or one or
more of the premises, of Travis’s arguments, as detailed in the following section.
Second, they may deny that one or more of the above conditions are necessary
for p-representation. Travis’s responds to the latter by stating that such
representationalists are not his target (Travis 2004: 82, 86), observing that the
first two conditions are taken directly from the work of prominent
representationalists, including McDowell, Peacocke, Davies, Tye and McGinn
(ibid. 86). However, it is arguable that such a denial would result in a notion of
representation that is too weak to fulfil the representationalist’s purpose in
providing an adequate explanation of various perceptual phenomena, such as
illusions, or the links between perceptual experience and judgement or belief.
The notion of p-representation is therefore closely connected with the
explanatory purposes of representationalism and its alleged benefits — purposes
upon which Travis’s arguments place strict limits.
The arguments in ‘The Silence of the Senses’ do not, however, rule out the
possibility that perceptual experience might involve some other form of
representation than p-representation, such as the causally covariant,
information-theoretic notion identified in the previous chapter. I return to this
question in chapter 5 when I consider Burge’s rejection of Recognisability. For
the time being, however, I will assume that, if successful, Travis’s arguments
only secure the weaker conclusion that perception does not constitutively
involve p-representation, rather than ruling out any form of representation
whatsoever. Arguably, such ‘representations’ do not warrant use of the term, at
least in its familiar sense. Terminological quibbles aside, however, I will argue
that Travis’s arguments constitute a powerful case against many otherwise
seemingly plausible accounts of perceptual representation, and so present a
serious and important challenge to representationalism about experience in
general.
Chapter 2, The Silence of the Senses
29
2.3. The Case Against Representationalism
For present purposes, I will concentrate upon four main arguments that may be
found in Travis’s ‘The Silence of the Senses’ (Travis 2004). Each is directed
against a particular motivation towards or kind of representationalism.
Together, the arguments provide a case against representationalism. Probably
the best known and widely discussed (though not always understood) argument
is Travis’s argument from looks (2.3.2), which aims to establish that it is
impossible to ground p-representational content in visual appearances or how
things look. This argument assumes Looks-indexing, and so is vulnerable to
forms of representationalism that reject either this or the Recognisability
condition outright. Such forms of representationalism are instead addressed by
what I will call the argument from recognisability (2.3.3). Additional arguments
are intended to show that the mere existence of perceptual illusions does not
present any particular advantage to representationalism over antirepresentationalism (2.3.1), and that representational content is incapable of
providing the perceiver with unmediated awareness of an object (2.3.4),
respectively.
2.3.1. The argument from misleading appearances
The first and simplest of Travis’s arguments concerns the significance of
perceptual illusion in the debate about perceptual representation. Historically,
Hume’s argument from illusion and others like it have been powerful weapons
in the representationalist’s arsenal (cf. Smith 2002). The traditional argument
from illusion purports to show that the objects of perception cannot be the
external physical objects or events that we ordinarily take experience to involve,
but must be mental entities such as sense impressions or ‘ideas’. Although the
resulting ‘theory of ideas’ has long since fallen into disrepute (cf. Reid 2002:
II.ix, 136; Snowdon 1990), vestiges of it can arguably be found in the
representationalist theories of perception that took its place. However, few
representationalists take representations, as opposed to external objects, events
and their properties, to be the objects of perceptual experience. Rather,
representing is supposed to be the means by which external physical objects and
events are perceived. Thus p-representation does not function as an epistemic or
perceptual intermediary between the subject and the external world. Rather,
representational content is intended to capture what is common to both
perceptions and their phenomenally matching counterparts, such as
hallucinations and illusions.
Travis’s argument from misleading appearances does not engage directly
with Hume’s traditional argument. Instead, it seeks to counter the view that the
phenomenon of perceptual illusion — i.e. experiences in which a subject
‘misperceives’ the state of the environment — provides any advantage to the
representationalist view. Byrne (2009), for example, considers the existence of
such illusions to be a positive motivation for endorsing representationalism, as
considered in the next chapter (3.2.1) — a suggestion which Travis rejects.
Typically, such representationalist arguments depend upon some version of
Chapter 2, The Silence of the Senses
30
what M. G. F. Martin calls the common kind assumption, which he states as
follows: ‘whatever fundamental kind of mental event occurs when one is
veridically perceiving some scene can occur whether or not one is
perceiving’ (Martin 2004: 40). This assumption, however, along with the
Common Content condition that flows from it (2.2), need not be accepted by
the anti-representationalist on the basis that perceptual illusion may be
considered parasitic upon veridical perception, and not assimilable to it.
Consequently, perceptual illusion and hallucination may require different forms
of explanation to veridical perception.16 Anti-representationalists, who reject the
common kind assumption, therefore offer alternative accounts of illusion that
may be evaluated upon their own merits.
This argumentative strategy is captured by Travis’s argument from
misleading appearances as follows:
MA1! Representationalism can explain the existence of perceptual
illusions in terms of misrepresentation.
MA2! Anti-representationalism can explain the existence of perceptual
illusions in terms of misleading or misinterpretation.
MA3! The mere existence of perceptual illusions provides does not
constitute a reason to suppose veridical perceptions to be
representational. (From MA1 and MA2)
MA1 grants that the representationalist is able to explain perceptual illusions in
terms of inaccurate, i.e. non-veridical, representations. This explanation,
however, is placed on a level with the anti-representationalist account in MA2.
This generates the conclusion in MA3 that, since that both accounts are able to
offer an explanation of perceptual illusion, the mere existence of illusory
experience does not favour either view. In essence, the availability of an
alternative anti-representationalist account of illusion is held to neutralise the
representationalist’s alleged advantage in this regard.
Of course, it is open to either side to argue that their explanation is better
than their opponent’s on grounds of consistency, parsimony, etc. This is
precisely Byrne’s (2009) strategy, as discussed in the following chapter.17 For the
time being, however, it will suffice to note that there are explanations of illusion
available to anti-representationalists that are perfectly consistent with their
explanation of veridical perceptual experience, but that do not depend upon the
notion of representation. Such explanations generally account for misleading
perceptions in terms of misinterpretation or misjudgements that individual
perceptual experiences indicate something that they do not. That is to say that
under normal circumstances an appearance of some particular type T which
16
Although relationalist explanations of hallucination typically differ from that of veridical
perception, the former may share the same explanatory structure as the latter, as described in
3.2.1.
17
For the converse claim, see Brewer 2006: 90–91.
Chapter 2, The Silence of the Senses
31
reliably indicates that things are some particular way P may easily be mistaken
for a similar appearance T′ that does not indicate P. Thus, by mistaking T′ for
T, one is inclined to judge P when in fact not-P. For example, a navy-coloured
jacket might appear black under fluorescent lighting, but blue in natural light. It
is therefore easy to misjudge the former appearance as indicating that the jacket
is black when in fact it is blue. Many such perceptual illusions may be explained
in terms of their subjective similarities to experiences of objects or properties
that, by custom, habit or inclination, we take them to resemble.
The main difference between the above explanation and the
representationalist’s relates to where one places the error. Representationalists
typically take perceptual illusions to be errors of perception that furnish the
subject with false or inaccurate representations — or rather misrepresentations.
The resulting experiences are therefore in some sense erroneous. Antirepresentationalists, on the other hand, typically regard perception as a success
term, and so there can be no errors of perception as such.18 Instead, illusions are
cases in which we are misled by appearances, or fail to interpret them correctly,
and thus the error is not one of experience per se, but of discrimination or
judgement. In effect, whatever significance the representationalist places upon
the notion of misrepresentation, the anti-representationalist attaches to
whatever cognitive faculty ‘downstream’ of experience — judgement,
discrimination, etc. — is responsible for misleading without any need to appeal
to representation. The force of the representationalist’s objection from illusion
can thus be seen to rest upon their commitment to the common kind assumption
and the representational nature of experience, thus rendering their argument
inconclusive. The anti-representationalist, on the other hand, can appeal to
subjective similarities between objects and their ‘ringers’ (relative to human
discriminatory capacities) as the reason for which we can be misled by
experience without claiming that experience is thereby erroneous. Naturally,
there may be differences between the sorts of similarities that individual
perceivers find misleading, since their perceptual sensitivities may differ, but
such errors are not errors of perception in the strictest sense.
Travis’s argument from misleading appearances addresses the phenomenon
of perceptual illusions, but does little to explain the nature of hallucinations.
These cases, in which no perceptual object — or at least none of the ordinary
kind — is present, can be explained by representationalists in a similar manner
to illusions, i.e. as cases of misrepresentation. It is unclear, however, that antirepresentationalists need commit themselves to a substantive account of the
nature of perceptual hallucinations other than that these constitute a form of
perceptual experience that is easily mistaken for genuine perception. That is to
say, hallucinations are subjectively indistinguishable from veridical perceptual
experiences, but do not share the same categorial grounds. This position is
defended by Martin (2004) as part of his ‘phenomenal disjunctivism’ (Byrne &
Logue 2008: 68) as well as some representationalists, such as Williamson (2000)
and McDowell (1994). The existence of hallucinations thus presents a further
18
Cf. Brewer 2007: 169.
Chapter 2, The Silence of the Senses
32
opportunity for representationalists to press their case, with the argument
proceeding in a manner analogous to the question concerning illusion. This no
doubt presents a serious challenge to anti-representationalism. However, if the
disjunctivist is right, this challenge need not undermine their account of the
central cases of perception, which are taken to have explanatory and conceptual
priority. Such arguments reveal a difference in the methodologies of
representationalists, who aim for the most general account of perceptual
experience possible, including hallucinations and illusions, and antirepresentationalists, who arguably offer a more parsimonious account of
veridical perception upon which their accounts of illusion and hallucination are
parasitic.
2.3.2. The argument from looks
The best known, but perhaps most misunderstood, of Travis’s arguments
purports to show that (i) the notion of p-representation plays no fundamental
role in a philosophical account of perception, and (ii) the Face-value and Looksindexing constraints that many representationalists place upon p-representation
are inconsistent, and so that there can be no such form of representation. Byrne
(2009: 14–15), for example, sets out Travis’s argument in terms of providing an
account of illusion, thereby portraying it as aiming to secure the weaker
conclusion that representationalism lacks support, whilst Siegel (2010: 60–65)
takes it to be primarily about the semantics of ‘looks’. Travis, however, clearly
takes himself to be arguing for the much stronger conclusion that ‘perception is
not [p-]representational’ (Travis 2004: 57), and so ‘is, in a crucial way, not an
intentional phenomenon’ (ibid. 93). The truth, I will argue (5.2.4), lies
somewhere in between these two extremes.
Travis’s argument from looks proceeds on the assumption that what makes
the representational content of perceptual experiences recognisable to a subject
must, by hypothesis, be the way that things in that experience appear, i.e.
Looks-indexing. In the case of visual perception, this corresponds to how things
look to the subject, raising the question of whether any notion of looks is
capable of indexing representational content. As such, this argument does not
rule out forms of representationalism that reject Face-value, Recognisability or
Looks-indexing. 19 Rather, it aims to establish that p-representation cashed out
in terms of mere appearances, i.e. looks, cannot perform the function that many
representationalists assign to it, and so cannot constitute an adequate
explanation of experience.
19
Rejections of the latter principle are dealt with by the argument from recognisability
(2.3.3).
Chapter 2, The Silence of the Senses
33
The argument may be summarised as follows:
L1!
The content of any given p-representation must be recognisable
to the subject solely in virtue of how, in the relevant experience,
things perceptually seem to that subject. (Looks-indexing)
L2
If visual experiences were p-representational, then their contents
would be indexed according to the way that things look [to the
subject]. (Corollary of L1)
L3!
Visual (i.e. comparative) looks are unfit to index representational
content since they are indeterminate between multiple contents,
which would contravene Face-value.
L4!
Thinkable (i.e. epistemic) looks are unfit to index
representational content since they cannot be determined solely
on the basis of what is perceptually available to the subject in
experience, which would contravene Recognisability.
L5!
No further notion of looks is capable of satisfying both Facevalue and Recognisability.
L6!
The contents of visual experiences cannot be indexed according
to the way that things look [to the subject]. (From L3 to L5)
L7!
Visual experiences are not p-representational. (From L2 and
L6)20
The transition from L1 to L2 is justified by inference to the best explanation
as the most plausible account of how visual experiences are able to satisfy
Recognisability. Travis’s notion of ‘indexing’ relates to whatever feature of
perceptual experience is supposed to make its representational content
recognisable to the subject. Thus, the content of any perceptual experience that
is looks-indexed is recognisable in virtue of how, in it, things visually appear to
the subject (2.2.5). L3 and L4 are concerned with the various notions of looks
that might perform such perceptual indexing. Here, Travis identifies two distinct
forms of looks, claiming that neither is capable of performing the required
function. From this, along with considerations about the conflicting demands
upon the notion of looks (L5), Travis draws the conclusion that visual
experience (L6) — and by generalisation all forms of perception — cannot be prepresentational (L7).
The argument from looks presents the representationalist with an apparent
dilemma. They must either (a) elucidate a notion of ‘looks’ that is capable of
indexing the relevant representational content, thereby rejecting one of L3, L4
or L5, or else (b) reject Looks-indexing, Face-value or Recognisability. Rejecting
20
Although the argument is presented in terms of visual perception, it is presumably
intended to generalise to cover other perceptual modalities.
Chapter 2, The Silence of the Senses
34
Looks-indexing places the onus upon the representationalist to find some other
way of satisfying Recognisability, or else justify dropping this apparently
plausible condition upon p-representation. (Travis has a separate argument
against such views that I discuss in 2.3.3.) Rejecting Face-value, on the other
hand, loses one of the supposed benefits of representationalism, namely that
perception represents the world as being some particular way. To understand
why the dilemma is pressing, however, it is necessary to gain a clear
understanding of the notions of visual and thinkable looks that Travis
considers,21 along with the reasons why he considers it impossible that any
further notion of looks could meet the representationalist’s requirements.
The first notion of looks that Travis considers is what are more commonly
called comparative looks (cf. Chisholm 1957: 45). Such looks involve some kind
of explicit or implicit comparison with the visual appearance of another object,
or objects, and are characterised in terms of what something ‘looks like’ in order
to appear the way it does (Travis 2004: 69–70). Thus, they are visual looks
(Travis forthcoming: §3). For convenience, I will abbreviate this as looksv. For
example, if the object before me looksv like a lemon, then it has the
characteristic look that lemons have — call this looking lemonish. Furthermore,
since many things share that very same way of looking, anything that looks
lemonish also looks like a wax imitation lemon, or (to the untrained eye under
appropriate circumstances) like a yellowish lime, the front surface of a
hollowed-out lemon, a lemon-shaped bar of soap, and so on. In fact, there are
innumerable ways that something can look like a lemon. Thus, for all that
things lookv a particular way, there are any number of ways that the world
might actually be, all of which share the same visual appearance.
Importantly, the corresponding resemblance relationships are symmetrical; if
something looksv like a wax lemon then a wax lemon also looksv like it.
Consequently, claims Travis, there can be nothing about an object’s lookingv
like p that identifies the content of that experience as representing p; for
example, that there is a lemon before me. Rather, the very same perceptual
experience might equally be said to represent any (or all) of the numerous ways
in which it can lookv to me just like there is a lemon before me — that there is a
wax lemon, for example. As per L3, comparative or visual looks fail to meet the
Face-value condition (2.2.2), since they ‘do not decide any particular
representational content for any given experience to have’ (Travis 2004: 69).
Visual looks, in this sense, are equivocal.
The second notion of looks that Travis considers is what are generally called
epistemic looks (cf. Chisholm 1957: 44). Whereas visual looks relate to
resemblances between appearances, epistemic looks are ‘very much a matter of
what can be gathered from, or what is suggested by, the facts at hand, or those
21
Travis (2004) uses the terms ‘looks like’ and ‘looks as if’ in place of visible and thinkable
looks, but this is apt to generate the impression that the argument concerns the semantic
properties of certain linguistic forms rather than two roles that such appearances can play. I
therefore adopt the terminology of Travis (forthcoming) in order to remain neutral about the
nature of looks and their semantics, which are considered in greater depth in chapter 4.
Chapter 2, The Silence of the Senses
35
visibly (audibly, etc.) on hand’ (Travis 2004: 76). They are what might be called
thinkable looks (Travis forthcoming), or lookst for short. Thinkable looks refer
to a specific way that things are in order to look the way that they do. Thus,
they are ideally suited to indexing p-representational contents like ‘there is a
lemon before me’. Indeed, thinkable looks just are those contents that
perceptual experiences incline a perceiver to believe or judge under the relevant
circumstances.22 Whereas one might say, for example, that a particular painting
looksv like a Vermeer despite our knowing that it was in fact painted by van
Meegeren (an artist with an indistinguishable visual style), we would not
normally say that it lookst as if it is a Vermeer since we already know that it is
not (ibid. 75).23
It is doubtful, however, that the contents that are indexed by or identical to
thinkable looks are recognisable to the subject solely on the basis of information
that is perceptually available to them in experience. For example, if Amy sees
what she takes to be a lemon in front of her, then there is a sense in which the
relevant object lookst to her as if it is a lemon (which it is). But for all that
things look that way, she might equally have taken the same object to be a
lemon-shaped bar of soap had she had a subjectively indistinguishable
experience of it in a different context — upon walking into a chemist’s shop, for
example. This is not due to any difference in the information that is perceptually
available to the subject, but rather a matter of what she takes to be the case on
the basis of that information. Plausibly, what is different about these two
situations is not a matter of how things look in any perceptual sense, but of
what the subject is inclined to infer from their total evidence base under the
circumstances. Such inferences are a matter of judgement and not of experience,
which conveys only how things appear. It is therefore unclear how one could
differentiate solely in virtue of perceptual experience, as Recognisability
requires, which of various otherwise indistinguishable possibilities are prepresented. Moreover, thinkable looks are a form of autorepresentation which,
Travis argues, cannot be ‘given’ to the subject since its content is already
accepted as true (2.2.3). Thus thinkable looks are in danger of collapsing into
what Travis (2004: 67) calls mere ‘indicating’. That is, as suggesting to a subject
that, under the circumstances, p may be the case. But this cannot be what makes
representational content recognisable for similar reasons. Thinkable looks
cannot index p-representational content on the basis of how things perceptually
seem to the subject since the information that distinguishes them is not
22
Travis (2004) associates thinkable looks with the English locution ‘looks as if’ in the
indicative mood, as distinct from ‘looks like’, which is typically (though not always)
comparative.
23
The issue is complicated by the fact that several different locutions of ‘looks‘ — including,
on occasions ‘looks like‘ — can be used to signify either comparative or epistemic looks.
Travis (2004) claims that ‘looks as though … were …’ in the subjunctive mood only ever
expresses the former notion, whereas ‘looks as if … is …’ in the indicative mood only
expresses the latter. His argument, however, does not turn on this point, and the linguistic
claim is dropped in Travis (forthcoming).
Chapter 2, The Silence of the Senses
36
perceptually available to them, either in experience or otherwise, and thus
Recognisability fails. Thinkable looks are not perceptual looks.
It is important to emphasise that Travis’s argument from looks is not a
merely semantic dispute about the meaning of ‘looks’, ‘appearance’, ‘seems’ and
their cognates. As such, the point of the argument is not that there is no widely
accepted meaning of ‘looks’ that individuates perceptual content. Rather, the
intended point is that there can be no such notion of looks on the basis that the
very idea of a univocal, objective and wholly perceptual ‘look’ is itself
incoherent. Either (a) looks are visual, in which case they fail to pick out any
specific way that the world must be in order for things to look that way, i.e.
they are equivocal, or (b) looks ‘index’, or are identical to, thinkable contents
about the subject’s perceptual environment, in which case they are not wholly
perceptual, since nothing that is perceptually available to (as opposed to
knowable by) the subject could pick out such univocal content.
Having established the two horns of the dilemma that Travis presses, the
question naturally arises as to whether some other notion of looks is able to
index representational content in a way that both has face value and is
perceptually available to the subject. As Travis (2004: 79–82) makes clear in his
discussion of McDowell’s notion of ‘ostensible seeing’, however, his answer to
this question is emphatically negative. According to McDowell’s disjunctive
notion of looks — call this looksd — every possible state of affairs matching the
relevant perceptible appearance is said to be represented. This either robs prepresentation of its univocality, since experience does not represent things to be
a determinate way, or, if the disjunction itself is the face value, it removes the
possibility that such content can misrepresent, thus undermining standard
representationalist explanation of perceptual illusion (2.3.1).
In essence, the need to satisfy Recognisability pushes towards visual looks,
which contravene Face-value, whilst attempts to satisfy Face-value push towards
thinkable looks, which contravene Recognisability. The requirements to
establish both univocal representational content (Face-value) and to make that
content recognisable solely in virtue of how things appear to the subject (Looksindexing) thus work against each other in a way that is, according to Travis,
fundamentally irreconcilable. Thus there cannot be any intermediate notion of
looks or middle ground between these two alternatives, and so the idea of looksindexed p-representation is incoherent.24 Nevertheless, many of Travis’s critics
(e.g. Byrne 2009 and Schellenberg 2011b) take there to be a further ‘noncomparative’ notion of looks that shares the relevant features of visible and
24
One way around this would be to abandon the notion that p-representational concepts
concern the subject’s perceptual environment in favour of their being facts about the
experiences of subjects (cf. Glüer 2009). But this robs representationalism of much of its
explanatory power, since the resulting propositions are no longer capable of grounding
objective knowledge or belief about external objects.
Chapter 2, The Silence of the Senses
37
thinkable looks in order that both Face-value and Looks-indexing may both be
met.25 Such rejections of L5 are examined in greater detail in chapters 3 and 4.
An alternative response to Travis’s argument from looks is to reject Looksindexing in favour of the weaker Recognisability condition, or else to reject
Recognisability outright. The former option places the onus upon the
representationalist to explain how perceivers are able to recognise the contents
of their perceptual experiences in a way that does not fall foul of Travis’s
argument from recognisability (2.3.3). It is, however, difficult to see how this
can be done other than through Looks-indexing, at least in the case of visual
perception. This response may be combined with a rejection of one or more of
Givenness, Common Content or Face-value, each of which is independently
plausible for reasons given above. The second option of rejecting or
reformulating Recognisability is more radical, though once again raises the
question of in what sense the resulting states would involve a form of
representation as opposed to mere causal covariance. Nevertheless, if prepresentational content is thought to be externally individuated, then one might
not expect it to be recognisable — at least not in the standard sense — to the
subject, depending upon how this condition is formulated. This possibility is
examined in greater detail in chapters 5 and 6.
2.3.3. The argument from recognisability
The third of Travis’s main arguments against representationalism appears in the
final section of his paper.26 This argument, which I will refer to as the argument
from recognisability, is directed against those representationalists who reject
Looks-indexing (L1 of the argument from looks) in favour of the weaker
Recognisability condition. As noted above, Travis does not take his case against
representationalism to depend upon Looks-indexing and, whilst the appropriate
formulation of Recognisability is a matter of some delicacy, it seems at least
superficially plausible that the content of p-representation might in some sense
be recognisable directly from experience, as opposed to how things look
(appear, seem, etc.) in it. Perhaps, as Travis puts it, ‘we can just tell how things
are thus represented to us; there is no saying precisely how we can tell’ (Travis
2004: 84). In other words, even if there were no rule-based algorithm describing
how we are able to recognise the content of perceptual experiences — on the
basis of looks, for example — we nevertheless possess some innate or acquired
capacity to do so. Travis raises two questions for such views: (i) precisely which
contents do such experiences have, where the explanation for this must give
some principled reason for thinking that one answer to this question can be
correct here over another, and (ii) what is the relation between our being
represented to in this way and our seeing (hearing, tasting, etc.) the things we
25
Byrne (2009) claims to reject Looks-indexing, but it is unclear that he takes it to play the
role that is described here (3.2.1).
26
Mislabelled section 4 (actually section 5) in Travis (2004: 82–93), or section 6 in Travis
(forthcoming).
Chapter 2, The Silence of the Senses
38
do, since if there is no such relation, then representing would be nothing but ‘a
(very annoying) wheel idling’ (ibid. 86)?
In relation to the first of these questions, Travis presents the following
argument:
R1!
Perceptual subjects must be capable of recognising the
representational content of any given p-representation solely in
virtue of having that very experience. (Recognisability)
R2
The notion of representation that is involved in perceptual
experience should be one that is familiar to us.
R3!
The contents of p-representations cannot be recognisable by their
(external) objects since, according to the principle of
intentionality, there may be no such objects.
R4!
The contents of p-representations cannot be recognisable by what
the subject takes to be the case, since for perceptual takings to be
part of experience would contravene Givenness.
R5!
The contents of p-representations cannot be recognisable by what
they indicate, or are taken to do so, since indicating and
expecting are not wholly perceptual, which would contravene
Recognisability.
R6!
P-representational content cannot be recognisable by what
experience is like (i.e. its phenomenal character), since this is
indeterminate between multiple contents, which would
contravene Face-value.
R7!
No other feature of p-representation is capable of satisfying all of
the above constraints.
R8!
We cannot recognise the representational nature and content of
p-representation solely in virtue of having that very experience.
(From R7)
R9!
Perceptual experiences are not p-representational. (From R1 and
R8)
The structure of the above argument closely mirrors that of the argument
from looks (2.3.2), with Looks-indexing (L1) replaced by Recognisability (R1),
and R5 and R6 corresponding to L4 and L3, respectively. Additional premises
(R3 and R4) deal with the recognition of p-representational content via its
external (i.e. intentional) objects and the notion of indication discussed above
(2.3.1, 1.2.3). In addition to drawing upon all four necessary conditions for prepresentation (2.3.1–4), the argument also assumes what I have called the
principle of intentionality (2.2) — a consequence of the representationalist’s
Chapter 2, The Silence of the Senses
39
explanation of illusion in terms of misrepresentation (2.3.1). Rejecting this
would therefore undermine both the relevant explanation of illusion and R3,
yielding an externalist form of content disjunctivism in which the external
objects of experience individuate its content. However, this still presents the
difficulty of explaining precisely how such content may be recognised on the
basis of its objects, as discussed in chapter 6.
Travis motivates R2 on the basis that ‘[i]t should not come as a complete
surprise [to us] that we are thus represented to’ (Travis 2004: 86). This ties in
with the justification of Recognisability as a consequence of Face-value and
Givenness, since it is the subject of experience, as opposed to some sub-personal
mechanism, that is supposed to be presented with representational content. The
relevant notion of representing is therefore supposed to be something that is
already clear, or else graspable on the basis of reflection, perhaps of a
philosophical nature, upon ordinary experience. This restricts the range of
possible answers to the question of what feature, or features, of experience
make its contents recognisable to the subject — or, alternatively, to what makes
it the case that there is a correct answer to what the content of a given
experience is — to a relatively small number of candidates. These are then
considered and rejected by Travis in R3 through R6 as follows:
(R3)
If the objects of perceptual experience (i.e. concrete external objects and
their properties) were what made experiential content recognisable, then
it would only be possible to satisfy Recognisability if these were proper
parts of experience. However, the representationalist explanation of
illusion in terms of misrepresentation means that one can have a typeidentical experience in the absence of the relevant objects. Thus they
cannot be what explains the corresponding content’s recognisability,
since it would still remain to be explained what makes the illusory
experience’s content recognisable, and where such an explanation would
render the original explanation superfluous.27
(R4)
The feature of experience that makes p-representational content
recognisable cannot be that the subject takes that content to be true
since, according to Givenness, p-representation does not involve taking
anything to be the case. Rather, such content is ‘given’ to the subject; i.e.
p-representation is allorepresentation, not autorepresentation (2.2.3). If
this were not the case, it would reverse the order of explanation between
perceptual content and judgement or belief, since subjects are supposed
to make such judgements on the basis of perceptual experience, and not
the other way round, generating a potential circularity in the justification
of perceptual belief. Consequently, recognisability cannot depend on the
subject’s taking or judging things to be the case.
27
This criticism parallels the representationalist’s argument from illusion. It differs in its
dialectical force, however, due to the representationalist’s endorsement of the principle of
intentionality, whereas the anti-representationalist is committed to no such principle.
Chapter 2, The Silence of the Senses
40
(R5)
If p-representational content cannot be recognised on the basis of what
the subject takes to be the case (R4), then perhaps the relevant feature is
a matter of indicating. This would make p-representation akin to
thinkable looks (2.3.2) which indicate what the subject would be
justified in believing under the circumstances, but can still be accepted or
rejected as per Givenness. The problem with this suggestion is that, like
thinkable looks, such indications cannot be recognised solely on the
basis of what is given in experience, as Recognisability demands, since
they are not wholly perceptual, but rely on the subject’s additional
background knowledge or the context of the experience. Therefore the
relevant feature of p-representation cannot be indication.
(R6)
Finally, p-representational content is not recognisable on the basis of
‘what experience is like’, i.e. its discernible phenomenal character,
because perceptual phenomenology alone fails to identify any univocal
way that the world might objectively be. This parallels Travis’s argument
against visual looks (2.3.2) as the locus for univocal content in L3 of the
argument from looks. If what experience ‘is like’ were to represent every
state of affairs matching the phenomenology of that experience, then
‘[t]oo many things would be represented as so at once’, making prepresentation ‘incoherent, thus not intelligibly representation’ (Travis
2004: 87); i.e. it would have massively disjunctive contents. Since
phenomenal character contravenes Face-value, it cannot be what fixes
representational content, nor can it be what makes such content
recognisable.28
Having ruled out the obvious candidates for the recognition of prepresentational contents solely on the basis of experience, and in the absence of
any other plausible options, Travis concludes that p-representation simply has
no such feature (R7). Consequently, subjects cannot recognise prepresentational contents solely in virtue of having the relevant perceptual
experiences, and so those experiences cannot be p-representational (R9).
Much as Travis’s argument from looks aims to characterise p-representation
as an intermediate notion between visual and thinkable looks, the argument
from recognisability aims to characterise it as intermediate between taking
something to be the case (i.e. autorepresentation) and indicating. In both cases
the problem, according to Travis, is that there is and can be no such
intermediary. Just as McDowell’s attempt to combine the two notions of looks
into a single notion of ‘ostensible seeing’ collapses (Travis 2004: 81), the idea
that experience represents something to us on the basis of what experience
makes perceptually available is either incoherent or superfluous. It is incoherent
because, on the one hand, the need for univocal face value pushes us towards
28
The type of content that perceptual phenomenology does gives rise to, namely phenomenal
content, cannot be considered univocal since it does not identify any particular way for
things to objectively be. Rather, it individuates how things are with the subject, irrespective
of the external world. For further argument on this point, see 3.2.3.
Chapter 2, The Silence of the Senses
41
forms of autorepresentation that contravene Givenness. On the other hand, the
notion of indicating is essentially context-dependent and so cannot be solely a
matter of what is perceptually available to the subject in experience.
Furthermore, if p-representation were ‘a mere re-rehearsing of what experience
has otherwise made plain’ (ibid. 88), then there would be no further role for it
to play, and so such representation would be otiose (see 2.3.4). In order to play
the relevant explanatory role, p-representation must be what conveys the
contents of experience to the subject, which, by Travis’s lights, it cannot do. The
present argument may therefore be regarded as a stronger and more general
form of the argument from looks, whose structure it entails.
The above argument leaves it open to the representationalist to avoid its
conclusion by denying one of Travis’s conditions upon p-representation, such as
Recognisability, or by proposing some other method or notion of recognition
that is capable of respecting these conditions. To deny Givenness would result in
a doxastic account of perceptual content, with all the problems that this
entails.29 Denying the principle of intentionality undermines the
representationalist’s account of illusion, which relies upon the notion that
perception is an intentional phenomenon and so the relevant representational
elements can occur in the absence of the objects or properties that it represents
(2.3.1). A similar problem occurs if the representationalist denies that the
relevant representing should be something that is familiar to us (R2). In both
cases, it is no longer clear what work the role of representation is supposed to
do in the explanation of perceptual experience, except perhaps as a placeholder
for some technical notion quite unlike the ordinary use of this term. Face-value
is, as we have seen, a prerequisite for many representationalist accounts of
experience, including some that allow for disjunctive content (2.2.2). None of
these responses is therefore without its problems.
Finally, the formulation and necessity of Recognisability itself may be called
into question. For example, if what is meant by recognise the content of
experience is to be able to know what is represented therein, then it is possible
to give an externalist account of knowledge that satisfies this condition without
contravening R2 (6.3.2). Such a move towards externalism, however, may end
up entailing much the sort of relation between subject and object that antirepresentationalists argue is necessary for perceptual experience, and so
constitutes a kind of hybrid position. Any view that denies Recognisability
outright entails rejecting the link between representation and phenomenal
character altogether, and thus rejects intentionalism. These issues are taken up
in detail in chapters 5 and 6.
2.3.4. The argument from unmediated awareness
Travis’s fourth and final argument against representationalism addresses the
second of the above questions concerning the relation between the role of prepresentation and perceptual states like seeing, hearing, tasting, and so on. The
29
Cf. Pappas 1976.
Chapter 2, The Silence of the Senses
42
argument falls into two distinct parts. The first part aims to show that if
experience involved p-representation, then this representation must be the sole
source of perceptual information available to the subject. That is, in the case of
veridical perception, representation should not be an unnecessary ‘wheel
idling’ (Travis 2004: 86). The second part of the argument concludes that any
awareness of an object in virtue of p-representation would at best be mediate
awareness. Such awareness is comparable to testimony, and is constituted by the
representation of content plus the satisfaction of ‘certain quite substantial
contingencies’ (ibid. 89), such as believing the content to be true, the obtaining
of the relevant correctness conditions, the reliability of the source, and so on.
However, many representationalists think that perception delivers unmediated
awareness. Seeing the pig constitutes a form of awareness of the pig that being
told there is a pig does not — at least not until the various contingent conditions
are satisfied, and perhaps not even then. What remains mysterious is how a state
(i.e. p-representation) one can be in without the presence of its intentional
objects — in hallucinations, for example — can play an explanatory role in the
awareness of those objects in the ‘good’ case. Rather, whatever is doing the
explanatory work must be something over and above p-representation, thereby
rendering its content if not unnecessary then at least lacking in explanatory
value.
The first part of this argument can be summarised as follows:
U1!
P-representation is either (a) a separate source of perceptual
information to seeing 30 or (b) constitutes one’s perceptual
awareness.
U2
If p-representation were a separate source of perceptual
awareness to seeing, then it would be unnecessary for the
awareness of objects simpliciter.
U3
We have no phenomenological evidence for the existence of
multiple sources of perceptual information in visual experience.
U4!
P-representation and seeing are not separate sources of
perceptual information.
Travis’s rejects the ‘two-source’ model of perceptual awareness on the basis that
if p-representation and seeing were separate sources of information, the latter of
which was sufficient for perceptual awareness, then p-representation would be
effectively redundant (U2). (The relevant notion of ‘source’ here is proximal, not
distal, and so is more akin to an informational conduit than a causal origin.)
Therefore p-representation and seeing are not separate sources of perceptual
information (U3), and so p-representation must be constitutive of the perceptual
awareness of objects, as per U1(b).
30
The same applies to hearing, tasting, and so on for each available sense modality.
Chapter 2, The Silence of the Senses
43
Having argued for a ‘one-source’ model of perceptual awareness, Travis
aims to establish that p-representation can only ever furnish us with ‘mediated’,
as opposed to ‘unmediated’, awareness of objects. He argues this on the basis of
the following principle or test (Travis 2004: 89):
As a rule, awareness of something else plus satisfaction of surrounding
conditions cannot add up to that awareness of [the presence of o] which we
have in seeing [o] (to be there) — in my terms, unmediated awareness. Or,
more exactly, if X is something there might be even without Y, then
awareness of X (and whatever accompanies it per se in a particular case)
cannot qualify as unmediated awareness of Y — the sort one might have in
seeing Y.
In familiar forms of representation, such as testimony, it may be represented to
the subject that X. However, since such representations can occur even though
not-X — e.g. in the case of false testimony — such representation cannot
constitute unmediated awareness due to the presence of a representational
intermediary, in this case a report. According to Travis, what goes for testimony
goes for perception since, as we have seen, the representationalist also allows
that representational content may be present without its corresponding
intentional objects. Thus, awareness of a particular content cannot constitute
unmediated awareness of an object since the same state can occur in the absence
of that object.
We can formalise the second part of the argument as follows:
U5!
Subjectively indistinguishable veridical and non-veridical
perceptual experiences share a common representational content.
(Common Content)
U6
In non-veridical experience, p-representation cannot yield
unmediated awareness of intentional objects, since there may be
no such objects.
U7
The same representational content cannot explain the
unmediated awareness of objects in some cases, but not in others.
U8!
P-representational content cannot yield unmediated awareness of
an object in veridical perception. (From U5, U6 and U7)
This yields a kind of reverse argument from illusion that combines
representationalist’s account of illusion (U5) with a ‘base case’ (U6) and
‘generalising’ or ‘spreading step’ (U7) to force the conclusion that prepresentational content cannot yield the unmediated awareness of objects (U8)
even in the ‘good’ case (cf. Snowdon op. cit.). More schematically, if a given
representational content does not constitute unmediated awareness in nonveridical experience, then it cannot constitute unmediated awareness in veridical
Chapter 2, The Silence of the Senses
44
perception either.31 Since many representationalists hold that such
representations, or their constituent elements, can also occur in the absence of
that which they represent — indeed, this is often supposed to be one of the
strengths of representationalism — then the argument from unmediated
awareness, if successful, would render perceptual awareness mediate in a similar
way to testimony.
It is important to note that the term ‘mediated awareness’ does not imply
that one is aware of the intermediary; i.e. it is not indirect or secondary
epistemic perception in Dretske’s sense (1969: 79). Rather, the claim is that this
is not the kind of awareness that we take ourselves to have in perceptual
experience. Perceptual awareness cannot, according to the present view, be
factorized into p-representation plus the satisfaction of some additional
surrounding conditions such as taking something to be the case, factivity,
reliability, entitlement, and so on (Travis 2004: 89). This claim has important
epistemological consequences since if to see x is to have unmediated awareness
of x or to have one’s cognitive responses shaped by x’s presence, then ‘no
substantial entitlement is needed … to take there to be [x]’ (Travis 2004: 90).
Unmediated awareness of objects is therefore uniquely suited to yielding
knowledge of the world in a way that mediate, that is to say representational,
awareness is not — at least not without the satisfaction of a host of as-yetunspecified, and quite possibly unspecifiable, further conditions.
The argument from unmediated awareness is completed by the following
premises:
U9!
Visual experience (seeing) is a source of information that yields
unmediated awareness of objects.
U10!
Visual experience is not p-representation. (From U4, U8 and U9)
In summary, and generalising to cover other sense modalities, if p-representation
were our sole source of perceptual information but cannot deliver unmediated
awareness of objects, then perceptual experience cannot be p-representational.
As with the historical argument from illusion (2.3.1), Travis’s argument
from unmediated awareness may be challenged at several key points. Why, for
example, should the representationalist accept that the same p-representational
content cannot yield unmediated awareness in some cases and not in others
(U7)? On the face of it, this principle seems intuitive, but if p-representation
were ‘narrow’ in the sense of being internally individuated, whilst the mental
state of awareness were ‘wide’ or externally individuated, then the same narrow
content could be present in both cases without ruling out that it could also
feature in the explanation of unmediated awareness in the veridical, but not the
non-veridical case. Similarly at the level of content, some but not all
representational elements may be common between ‘good’ and ‘bad’ cases,
31
Note that the relevant content may be only part of the full p-representation, thus leaving
room for the kind of response to the argument discussed in 5.3.2.
Chapter 2, The Silence of the Senses
45
generating a form of content disjunctivism or indexicality. This is perfectly
consistent with Common Content and would explain the immediate awareness
of objects in terms of those representational elements or contents that are
present in the ‘good’ case, but not the ‘bad’.
Such responses, however, suggest there must be more to unmediated
awareness than perceptually experiencing an appropriately caused prepresentation of an object. This ‘something more’ is arguably precisely the
relation between subject and object that anti-representationalists claim is
fundamental to perceptual experience, and which appears to be omitted from
the representationalist model. This in turn suggests that representationalism
gives at best a partial characterisation of perceptual experience with much of the
explanatory work being done by the relational element. This suggestion is taken
up further in chapters 6 and 7, which explore the possibility that
representationalism and relationalism may in some sense be two sides of the
same coin.
2.4. Conclusion
Travis’s case against representationalism consists of a series of arguments, each
of which targets the view that perceptual experience constitutively involves a
particular form of representation — namely p-representation. Such
representations meet the conditions of Content, Face-value, Givenness and
Recognisability set out above, each of which is individually necessary for prepresentation. In responding to Travis, the representationalist must therefore
either challenge the soundness of these arguments or else provide an alternative
notion of representation that rejects one or more of these conditions upon prepresentation whilst maintaining the distinctive thesis that perceptual
experience may be characterised in essentially representational terms, i.e. REP. If
Travis is right, then perceptual experience cannot simply be a matter of a
perceiver’s being in a particular representational state, but rather constitutively
depends upon the subject’s relation to the objects they perceive, i.e. REL. On this
view, the notion of p-representation omits the very aspect of perception, namely
‘openness to the world’, that explains its ability to make an objective world
available to us in conscious thought and experience.
The following chapters examine a range of responses to Travis’s antirepresentationalism. These range from positing a third, ‘non-comparative’
notion of looks (chapters 3 and 4) to rejecting Recognisability (chapter 5) or
positing an alternative mechanism for its satisfaction (chapter 6). Even if Travis
arguments were successful, however, this would not show that there is no
notion of representation that is applicable to perception. Rather, Travis’s claim
is that such a notion would be (a) unrecognisable to us as representation; i.e. it
would be a form of ‘quasi-representation’ that shares some of the properties of
p-representation but, for example, does not show up in conscious perceptual
experience; or (b) unable to support many of the claims that representationalists
typically make for their theories — as a means of explaining phenomenal
Chapter 2, The Silence of the Senses
46
character (i.e. intentionalism) or what is common to both experience and
judgement, for example.
The challenge for representationalists, then, is to elucidate the
representational ‘language’ of the senses without thereby committing to the
existence of the kind of perceptual relation that many anti-representationalists,
typically relationalists, take to be essential to perceptual experience. If it is the
relationality, and not the representational content of experience, that accords
perception its special role within the philosophy and epistemology of mind, then
representation simply drops out of the picture since the senses would, as Travis
claims, be ‘silent’.
3. The Sound of Silence
Three arguments from perceptual appearances
3.1. Introduction
In this chapter I examine three arguments in favour of perceptual experience
having p-representational content. Although each is presented in connection
with a particular philosopher in the literature — namely, Alex Byrne (2009),
Susanna Siegel (2010) and Susanna Schellenberg (2011b) — they all exemplify a
common approach or assumption concerning experience. That is, they take the
nature or phenomenology of perceptual appearances to be decisive in
establishing the existence of representational content. Furthermore, despite
ostensibly responding to Travis (2004), each of these arguments fails to engage
with central aspects of the case against representationalism by underestimating
its force and generality. Consequently, these arguments fail to address the
central issues that Travis raises, rendering them unsuccessful as responses to
anti-representationalism or in establishing the existence of p-representational
content. Whilst this does not rule out the possibility that experience may indeed
have representational content, or that certain forms of relationalism about
experience may be compatible with this view — as Siegel and Schellenberg, for
example, claim — to establish this, a better way of adjudicating between the
relevant positions will need to be found.
3.2. An Argument from Non-Comparative Looks
The first argument in favour of representationalism I wish to consider involves a
third notion of looks — namely, non-comparative or phenomenal looks — which,
unlike Travis’s visible and thinkable looks, is supposedly capable of grounding
representational content. A recent exponent of this approach is Alex Byrne
(2009), who argues in favour of non-comparative looks (3.2.2). Byrne’s
principal criticism of Travis’s anti-representationalism is that it fails to provide
an adequate account of perceptual illusions, such as the Müller-Lyer figure (ibid.
445). Representationalism, on the other hand, is tailor-made to provide such an
Chapter 3, The Sound of Silence
48
explanation in terms of the notion of misrepresentation, as previously discussed
in chapter 2. However, Byrne’s argument on this point begs the question by
characterising anti-representationalism as involving a ‘factive propositional
attitude’ (ibid. 437) when the existence of any such attitude in experience is
precisely what anti-representationalists like Travis deny (3.2.1).
Furthermore, I argue that Byrne fails to elucidate a coherent noncomparative notion of looks, arguably committing him to a form of phenomenal
content that is unable to support the standard representationalist account of
illusion since it entails that any experience that has the relevant content is
necessarily veridical simply in virtue of being experienced (3.2.3). If so, Byrne’s
own account of experience suffers from precisely the difficulty that he levels at
Travis since it is incapable of explaining how the relevant perceptual content is
able to misrepresent in the case of perceptual illusion. Given that, according to
Byrne, this was supposed to be one of the main benefits of representationalism
about perceptual experience, and the reason for which he rejects Travis’s antirepresentationalism, this renders his response to Travis seriously flawed, or at
least incomplete. Thus I argue that the attempt to ground representational
content in ‘non-comparative’ looks fails to establish representationalism (3.2.4).
3.2.1. ‘Zeeing’ and the problem of illusion
Byrne begins his argument by spelling out the thesis to which Travis is opposed,
which he refers to as the Content View.1 After rejecting an initial formulation of
this thesis on the basis that it involves a ‘special philosophical sense’ (ibid. 433)
of perceptual experience that Byrne rejects, he settles on the following
characterisation of the view (ibid. 437):
CV′
Perception constitutively involves a propositional attitude rather like
the non-factive attitude of believing.
According to this thesis — a form of representationalism — whenever one
perceives, one is in a contentful representational state that does not entail the
truth of its contents.2 Thus ‘perceiving’ is akin to ‘believing’ or ‘hoping’, and
unlike ‘knowing’, which is generally taken to be factive. From this, we can infer
that Byrne endorses Travis’s Objectivity condition, since propositional attitudes
‘represent things as being so’, and are therefore assessable for truth or accuracy
(2.2.1).
In order for there to be some particular way that things are represented as
being, however, this content must have a single face-value content, and so CV′
may also be taken to entail Face-value. Whether Byrne endorses Givenness is
less clear, since his analogy with belief might be taken to suggest that the
relevant propositional attitude entails a level of commitment compatible with
1
2
Not to be confused with Siegel’s use of this term (3.3.1).
Byrne’s use of the term ‘propositional attitude’ is intended to be neutral between conceptual
and non-conceptual content, and so CV′ should not be taken to entail a purely conceptual
view of perceptual content (ibid.).
Chapter 3, The Sound of Silence
49
perception being autorepresentation, or representation-by. However, as Travis
argues, perception cannot be autorepresentation since only committed
allorepresentation — i.e. representation-to — can have a face value (2.2.3). I will
therefore assume for the sake of argument that Byrne also endorses Givenness,
though nothing in what follows turns on this. Finally, I argue below that Byrne
endorses a form of Looks-indexing, which in turn entails Recognisability, albeit
under a different notion of looks than the ones Travis considers (3.2.3). As such,
the content of a perceptual state must be accessible to the subject solely in virtue
of their being in that very state. From this, we can conclude that the content of
the propositional attitude to which Byrne refers is a form of p-representation,
and thus a potential target of Travis’s argument from looks (2.3.2).
In order to determine whether CV′ is true, it will be necessary to compare it
with the corresponding anti-representationalist position from some suitably
neutral standpoint. Byrne’s attempt to do this centres upon his analysis of
perceptual illusion in terms of the propositional attitudes of ‘exing’ and
‘zeeing’ (Byrne 2009: 437).
Byrne’s positive argument for representationalism takes the form of an
inference to the best explanation. The idea is to show that there are theoretical
considerations that count in favour of CV′ — a claim which Byrne takes Travis
to deny — thereby providing a counter-argument to Travis’s argument from
misleading appearances (2.3.1). He begins by characterising antirepresentationalism in terms of a factive attitude — ‘zeeing’ — which is
somewhat akin to knowing (ibid.). According to this view, perceptual illusion
may be explained in terms of misinterpreting one’s ‘zeeing that p’ as indicating
q, and thus arriving at a false conclusion on the basis of one’s perceptual
experience. Since zeeing is factive, zeeing that p logically entails p, and so any
error that occurs must be one of interpretation or judgement, rather than of
perceptual experience per se. This is compatible with Brewer’s (1999: 169)
maxim that ‘[e]rror, strictly speaking … is never an essential feature of
experience itself’ on the basis that one can only perceive, or ‘zee’, what is
actually there. Byrne characterises representationalism, on the other hand, in
terms of CV′, which claims that perception constitutively involves a non-factive
attitude — ‘exing’, intended to suggest ‘experiencing‘ — somewhat akin to
believing (Byrne 2009: 437). On this view, perceptual illusions occur just in case
a subject ‘exes that p’ when in fact not-p; i.e. it is the content of the perceptual
state that is in error rather than the subject’s interpretation of it. Byrne argues
that this explanation of illusion is simpler than the anti-representationalist’s and
so should be preferred upon grounds of parsimony.
Byrne’s formulation of the dispute is misguided in several respects. Firstly,
anti-representationalists not only deny that perception constitutively involves a
non-factive propositional attitude, but that it involves any propositional attitude
at all. Indeed, this is precisely the conclusion that Travis’s arguments aim to
establish. Since the factive attitude of zeeing is as much a contentful, and
therefore representational, state as exing, these arguments would count as much
against the former as it as they do against Byrne’s representationalism. The
Chapter 3, The Sound of Silence
50
characterisation of anti-representationalism as constitutively involving a factive
propositional attitude is therefore both beside the mark and question-begging.
The point here is not whether the attitude involved in perception is factive or
non-factive, as Byrne appears to suggest, but whether it involves any
propositional attitude at all. Consequently, an argument against zeeing is not an
argument against anti-representationalism, because anti-representationalism
involves no such attitude.
As we saw in chapter 2, anti-representationalists typically explain perceptual
illusions in terms of misleading or misinterpretation of perceptual experience
(2.3.1). It is therefore instructive to compare the explanatory structure of
genuine and illusory perceptual experiences on the resulting account. In the case
of perception, a subject is presented with an object o which she judges to be F.
Since o is in fact F, the experience is veridical and may subsequently form the
basis for a judgement to this effect. In the illusory case, on the other hand, the
subject is similarly presented with an object o that she judges to be F. However,
in this case o is not F, and so the judgement is incorrect. Again, this is not a
matter of ‘zeeing’ o to be one way and then judging it to be another, as Byrne
describes it, but of mistaking the look of o under the relevant circumstances as
indicating that it is F when it is not. In neither case does any representation —
accurate or otherwise — form a part of the experience. Instead, representation
only comes into play at the level of judgement and belief. Put this way, the antirepresentationalist account of illusion can be seen to be no more complex than
their account of perception — or, for that matter, than Byrne’s own
representational account. The only significant difference lies in where one places
the ‘error’. For Byrne, such errors lie in the subject’s representational state,
which fails to accurately reflect how things are in the world. For the antirepresentationalist it lies with whatever faculty is responsible for interpreting
perceptual experience, i.e. belief- or judgement-formation. Byrne’s assertion that
the representationalist account of illusion is simpler and more parsimonious is
therefore false.
A further point worth noting is that even if Byrne’s characterisation were
correct, it is unclear why this should constitute a problem for the antirepresentationalist. Whilst parsimony is undoubtedly a virtue of philosophical
or other theories, it is not the final arbiter of truth. Given that many antirepresentationalists take illusory or hallucinatory experiences to be parasitic
upon perception, rather than the central cases with which a theory of perception
should be concerned, it would be considered no defect if their theory were to
handle the former in a different and potentially more complex way than the
latter (although, as shown above, this is not necessarily the case).3 Indeed, for
anti-representationalists who take ‘perception’ to be a success verb, illusions and
hallucinations are not, strictly speaking forms of perception at all, but rather
experiences that can easily be mistaken for perception. Provided that a theory of
perception is compatible with the existence of illusions and hallucinations — a
3
Disjunctivists, for example, characterise illusions and/or hallucinations in terms of their
subjective indistinguishability from veridical perceptual experiences (Martin 2004: 72).
Chapter 3, The Sound of Silence
51
fact that all parties accept — it is not mandatory that such cases be explained in
precisely the same manner as genuine perceptions (though they may be). It is
therefore difficult to see why Byrne takes this to be a decisive factor in the
debate, especially if anti-representationalism were, as some of its proponents
suggest (e.g. Brewer 2006; 2008), able to offer a more plausible account of
veridical perception.
If Byrne’s aim is to show that the existence of perceptual illusion offers more
support to CV′ than it does to anti-representationalism then it fails because,
without begging the question against the latter, he has not demonstrated that
this is the case. As Travis points out, this motivation for supporting
representationalism is neutralised by the availability of an alternative
explanation for such phenomena in the form of the misinterpretation of
perceptual appearances (2.3.1). Admittedly, this approach, which dates back to
Berkeley and Reid, is not without its problems, but these do not appear to be
Byrne’s target. Furthermore, even if the anti-representationalist explanation of
illusion were more complex than that offered by representationalism, then this
cannot be taken to settle the matter. Thus, whilst the availability of a simple
account of the nature of illusions might be taken as a motivation for endorsing
representationalism, it cannot be the decisive factor in the present debate.
Rather, it is one issue among many that must be taken into consideration when
evaluating a theory of perception.
3.2.2. Non-comparative looks
The remainder of Byrne’s response to Travis focuses upon the various notions of
looks that Travis employs in connection with the ‘indexing’ or recognition of
perceptual content. In particular, Byrne rejects premise L5 of Travis’s argument
from looks, claiming that there is a third notion of looks that is capable of
meeting both Face-value and Looks-indexing, thereby defusing Travis’s
dilemma. Indeed, he claims that the need for such a notion is apparent from
Travis’s (2004: 69–70) own definition of visible looks (‘looking like’), which is
as follows:
[S]omething looks thus-and-so, or like such-and-such, where it looks the
way such-and-such, or things which are (were) thus-and-so, does (would,
might) look.
Unpacking the somewhat convoluted grammar of this passage, it is apparent
that ‘looks’ appears in both the explanandum (‘looks thus-and-so, or like suchand-such’) and explanans (‘looks the way … does’). The latter form of looks
must therefore be distinct from the former or else Travis’s definition would be
circular. Byrne takes this to show that, even by Travis’s lights, a third notion of
Chapter 3, The Sound of Silence
52
looks is required — a notion which, following Chisholm (1957), he calls noncomparative looks.4
Byrne describes non-comparative looks, or ‘looksnc’ (Byrne 2009: 443), as a
kind of ‘Gestalt’ (ibid.) for which the subject possesses some innate or learned
recognitional capacity (cf. Chisholm 1957; Jackson 1977). Thus red things
looknc reddish, cows looknc cowish, old things looknc oldish, and so on, where in
each case the explanans picks out a non-comparative look that is typically
possessed by things of the relevant kind. Furthermore, Byrne contends that
looksnc are explanatorily prior to both looksv and lookst, thereby avoiding the
alleged circularity in Travis’s account.
Breckenridge (2007) makes a similar proposal, characterising looks as
properties of perceptual events (‘lookings’) that correspond to the manner in
which things, according to them, look. However, the resulting ‘adverbial’
account (ibid. 4) suffers from the same problem as Travis’s visible looks, since
such ‘ways of looking’ would be incapable of indexing content that is any more
finely grained than those very looks. For example, many things lookv like a
lemon, only some of which have the property of being a lemon. Either each of
these either counts as the same way of looking, or it counts as a different way of
looking. If each counts as the same way of looking — looking lemonish, as it
were — then this way of looking cannot index the proposition that something is
a lemon, since, despite having this look, the object might equally well be said to
exhibit the appearance of a lemon-shaped bar of soap or innumerable other
things. Thus Face-value cannot be met. Conversely, if each of these possibilities
is to count as a different way of looking — the look of a lemon, the look of a
lemon-shaped bar of soap, and so on — then such looks are epistemic, and so
not discernible on the basis of perceptual information alone since they are
perceptually indistinguishable. Thus Looks-indexing cannot be not met. We are
back with Travis’s original dilemma. Consequently, if looksnc are supposed to
solve this problem, then they cannot be adverbial ways of looking.
Unfortunately, whilst Byrne’s overall strategy is clear, he fails to provide an
adequate account of just what he takes looksnc to be, and how they are
supposed to individuate the sort of content that representationalists posit —
namely, p-representation. In the absence of such an account, ‘non-comparative
looks’ becomes a mere placeholder for some notion of looks that both has
univocal face-value and is recognisable, as per McDowell’s ‘ostensible
seeing’ (2.3.2). However, such a stipulation does not solve the problem that
Travis raises, but merely restates it since it remains unclear how the alleged
looks could be something that is perceptually given in experience.
Byrne goes on to reject the notion that content is ‘looks-indexed’,
concluding (Byrne 2009: 444):
4
In fact, Travis’s notion of visible looks can be explained in terms of, for example, visually
relevant similarities, and so does not require an additional sense of ‘looks’ (cf. Brewer 2011:
§5.3; Martin 2010; 4.2.3).
Chapter 3, The Sound of Silence
53
The upshot is that Travis is in one way right. Perceptual content, if there is
such a thing, is not ‘looks-indexed’, at least as that notion has been
explained here. But Travis is wrong to conclude that our ordinary talk
provides no support for [representationalism].
Of course, Travis argues for no such conclusion. Rather, as described in chapter
2, he argues that the very notion of p-representation is incoherent. Furthermore,
it is unclear that Byrne really does oppose Looks-indexing as Travis formulates
it. Rather, he takes this term to mean that there is some sense of the word
‘looks’ that is ‘best understood as implicitly reporting the content of the exing
attitude’ (Byrne 2009: 439). Whilst it is true that there may be no such sense,
this does not undermine Looks-indexing, which relates to the recognisability of
perceptual content and not its expression in everyday language.5 Indeed, Byrne
goes on to defend looksnc as constitutive of p-representational content (ibid
443). Despite the above denial, then, I take it that Byrne does endorse a form of
Looks-indexing, at least for the particular notion of looks he describes — i.e.
looksnc — even if it has no common expression in everyday language.6
3.2.3. Phenomenal content
Byrne claims that, given its role in explaining illusion, we should ‘expect
perceptual content to be relatively thin’ (Byrne 2009: 449; 450). By ‘thin’,
Byrne means that p-representational content would not contain cognitively
sophisticated concepts like lemons, pigs or Vermeers, but instead relates to more
‘primitive’ sensory concepts, such as yellow or square. This contrasts with
theorists like Siegel (2010), who argue that experience has ‘rich’ content that
involves precisely such natural-kind properties. On Byrne’s view, the
propositional content arising from non-comparative looks would, like Jackson’s
‘phenomenal looks’ (Jackson 1977: 33), ‘concern properties like shape, motion,
colour, shading, orientation and the like, not properties like being tired,
belonging to Sarah or being a lemon’ (Byrne 2009: 449). The difficulty for this
view, however, arises when we try to spell out precisely which properties are
represented in perceptual experience.
Representationalists typically claim that the contents of perceptual
experience concern, amongst other things, the appearance-independent
properties of objects, such as being yellow, being round, and so on. On this
view, that objects look yellow or look round, etc. to a subject is supposed to be
explained in terms of the subject’s representing them to be yellow, round, etc.
That is, an object o appearing to be F to subject S is explained in terms of S’s
representing that o is F, in accordance with Travis’s Objectivity condition
(2.2.1). Seeming or appearing is therefore a function of the propositional
attitude under which the relevant content is entertained — ‘exing’ in Byrne’s
terminology (3.2.1) — as opposed to something that features explicitly in the
content of experience itself.
5
Byrne himself seems happy to identify looksnc as a sense of ‘looks’ (ibid.).
6
A more detailed analysis of the semantics of ‘looks’ is given in chapter 4.
Chapter 3, The Sound of Silence
54
What Byrne and others appear to overlook, however, is that the argument
from looks applies equally to so-called perceptual primitives like being round
and being yellow as it does to cognitively sophisticated ones like being a lemon
or being a Vermeer. This leaves such views open to the following form of
objection: it is not S’s propositional attitude towards the representational
content of experience which makes it the case that o appears F — being yellow,
for example — to S, but rather the properties that are represented in that
content; i.e. they are appearance properties. In order for the representationalist
explanation of illusion to go through, every case in which o appears F to S must
involve the same content regardless of whether or not o actually is F. Only in
this way can the relevant content be said to misrepresent. However, according
to the present objection, if non-comparative looks index anything, it is how
things appear to a subject and not how things are in the world. That is to say,
the content that S’s perceptual experience makes recognisable is not that o is F,
but rather that o appears F. Thus, instead of appearances being a matter of the
subject’s representing that p under a particular propositional attitude, they
become embedded into the contents of the relevant p-representations, yielding
what I will call phenomenal content.
Why should we think that non-comparative looks index properties of the
subject’s experience, rather than those of the external objects that give rise to
them? After all, it is part of the representationalist’s view that things appear to
be, for example, yellow by representing them to be yellow, not by representing
them to look yellow. The problem stems from the nature of looks-indexing and
the need for the relevant content to be recognisable to the subject. If what is
supposed to make the relevant content recognisable is the way that things
looknc, then this must be (as one might expect) common to both epistemically
‘good’ and ‘bad’ cases. A subject who undergoes a perceptual illusion is just as
aware of experiencing a yellow non-comparative ‘look’ as one who sees a ripe
lemon under normal lighting — they are, according to Byrne, experiencing the
same phenomenal look. The point that the objection presses is: what is it about
such looks, which are common to both good and bad cases, that makes the
content o is yellow recognisable, as opposed to o looks yellow? Since both cases
involve the same phenomenal look, then why should we think of the look as
favouring one case — e.g. (the object’s being) yellow — over the other, i.e. that it
(merely) looks yellow)?
The above point is precisely the one that Travis presses against visual, or
comparative, looks. Given all that Byrne has said about non-comparative looks,
it is as yet unclear how they could index propositions concerning the way the
world objectively is since there is no determinate state of affairs that must
obtain in order for things to look the way that they do. Something that looks
yellow, for example, could just as well indicate a white object illuminated by
yellow light as it does something being yellow. That is, there is no specific way
that a given object must be in order for it to look the way that it does, and thus
nothing objective to represent, except perhaps a disposition to cause a certain
type of experience. Moreover, the presence of any given non-comparative look
Chapter 3, The Sound of Silence
55
co-varies not with o’s being F, but with o’s looking F. Thus, if looksnc index
some content, i.e. make it recognisable to the subject, then it should concern
whatever property of experience corresponds to this look, and not the properties
of the external objects that cause it. What is required here, and what Byrne’s
account fails to provide, is an explanation of how the appearance-independent
properties of objects such as being yellow come to figure in the contents of
perceptual experience at all. The representation of such objective properties
cannot simply be taken for granted by the representationalist since this is
precisely what is at issue in Travis’s argument from looks.
At first blush, representing that objects look thus-and-so rather than being
thus-and-so, may seem unproblematic, at least for Byrne who denies that
perceptual content need play any strong epistemic role (Byrne 2009: 450).
However, if non-comparative looks, and therefore p-representations, relate only
to the subjective appearances of objects (how they look) rather than their mindindependent properties (how they are), then it becomes impossible for the
resulting content to misrepresent. An object o that looks yellow to S, for
example, cannot fail to have the property of looking yellow to S under those
particular circumstances, quite independently of whether it is yellow or not.
This difficulty arises from the fact that being yellow and looking yellow can be
regarded as two distinct properties, one of which relates to the objects of
experience and the other to experience itself. The fact that the English locutions
that describe these properties both happen to include the word ‘yellow’ is a
contingent feature of surface grammar that does not itself ground any deep
philosophical connection between the two. Consequently, it is easy to overlook
the fact that, for all that Byrne has said, there is nothing to connect looking
yellow with being yellow — they are, to all intents and purposes, quite separate
and distinct properties.
Byrne effectively concedes this when he states that ‘perceptual content, if
there is such a thing, goes with the way things look when they looknc F, which
need not include Fness’ (Byrne 2009: 443). Just as yellow objects do not always
look yellow, non-yellow objects can sometimes look yellow — when viewed
under yellow light, for example. Thus something can look yellow without being
yellow, and vice versa.7 Moreover, since being yellow is a monadic property of
objects, whereas looking yellow to S is a relational property whose relata
include both subject and object, the two properties cannot be numerically
identical. Rather, they must be connected in some other manner. The objection
to Byrne’s account is that it fails to specify what this connection is, thus
rendering it at the very least incomplete, and at worst incoherent since in any
experience in which things seem to S to be F, it is true that they seem this way to
S, and so perceptual experiences effectively become their own truthmakers.
7
Lest the metaphysics of colour be thought to present a particular problem in this regard,
compare the case of circular objects, which can sometimes exhibit a ‘non-comparative’
elliptical look when viewed at an angle. This is not to say that circular objects look to be
elliptical when seen at an angle. Rather, they look like ellipses do when seen face-on such
that they might, under certain circumstances, be mistaken for an ellipse on the basis of this
shared visible look.
Chapter 3, The Sound of Silence
56
The form of objection sketched above does not deny that representational
contents may be individuated by something other than looks. Rather, it shows
that p-representational contents that are indexed — that is to say, made
recognisable — by looksnc alone cannot privilege the representation of objective
properties, i.e. being F, over phenomenal properties, i.e. (merely) looking F.
Consequently, the resulting p-representations cannot misrepresent; they
constitute phenomenal content. This counts against the view set out in Byrne
(2009) to the extent that it supports Looks-indexing, linguistic objections
notwithstanding. It is particularly pressing given that the ability to account for
illusion is the reason Byrne gives for favouring representationalism over Travis’s
anti-representationism (3.2.1). If, on the other hand, Byrne rejects Looksindexing, the problem is circumvented, but this leaves him without an account
of p-representation since the notion of non-comparative looks would then be
explanatorily inert. The objection therefore counts against any theory that takes
p-representational content to be grounded in non-comparative looks. This
includes both Siegel (2010) and Schellenberg (2011b), who explicitly cite Byrne
(2009) as a proponent of this form of content, and whose arguments I examine
in further detail below.
3.2.4. Responses
Having established a potential objection to the non-comparative looks based
view, I will now consider some responses that a proponent of such a view might
give to the above argument, along with the reasons why they are unsatisfactory.
The most plausible line of defence is perhaps that the semantics of prepresentational content is quite independent of an account of how such content
is recognised by the subject. The former determines how content is fixed or
individuated, whilst the latter explains how it is accessible to the subject. For
example, one might adopt a counterfactual (Dretske 1994), informational
(Fodor 1987) or teleosemantic (Millikan 1993) account according to which the
contents of perceptual experience refers to the mind-independent properties of
objects and not their looks. On this view, what makes it the case that a subject
represents, for example, the presence of a lemon, as opposed to something that
merely looks like a lemon, is explained in terms of something external to
experience itself, such as the evolutionarily selected ability to represent lemons
as a source of nutrition.
The problem with this approach, however, as Travis points out, is that it
fails to explain how such content is recognisable to the subject. According to the
Recognisability condition, the representational content of experience — i.e. what
would need to be the case in order for that content to be veridical — is supposed
to be recognisable to the subject in virtue of having that very experience (2.2.4).
This is just part of what it means to be represented to, i.e. Givenness. However,
if the accuracy conditions of p-representations depend upon evolutionary
contingencies, counterfactual possibilities or informational links with the
environment, it is difficult (though not impossible — see 5.3) to see how this
condition can be satisfied. More generally, if what makes such contents
Chapter 3, The Sound of Silence
57
recognisable differs from what fixes that content, it remains to be explained
how the former is able to track the latter. Travis aims to drive a wedge between
these two requirements by arguing that whatever makes content recognisable
(e.g. non-comparative looks) will dictate the kind of content that one can
thereby recognise (i.e. phenomenal content) instead of the kinds of contents that
representationalists typically claim experiences to have. I will return to this
point in chapters 5 and 6.
An alternative response that Byrne might (and indeed does) give is that for
something to look yellow is just for it to ‘look the way that yellow things are’. 8
This might be taken to mean one of two things, the first of which is trivially true
and the second of which is question-begging. If ‘the way that yellow things are’
is taken to mean the property that, as a matter of necessity, all yellow things
have in common — namely, their yellowness — then the point is trivial, since it is
hardly explanatory to say that ‘to look yellow just is to look yellow’.
Alternatively, if Byrne intends for his maxim to establish some kind of
connection between looking yellow and being yellow, as his argument requires,
then being yellow must itself be a way of looking yellow. To the extent that this
is true — something that is far from clear since things that are yellow need not
look it — then this is presumably just one of the ways in which things can look
yellow, since many non-yellow things also look yellow under appropriate
circumstances: a white piece of chalk in yellow light, for example. In order for
the relevant experiences to represent being yellow rather than being white in
yellow light, there must be something that privileges the former content over the
latter. This cannot simply be the fact that one is true and the other false in any
given experience, since the content of experience being satisfied is supposed
determine whether or not that experience is veridical, and not the other way
round. Thus Byrne’s response either falls under the previous point concerning
the recognisability of perceptual content, or else assumes the very thing it is
supposed to explain — namely that to look yellow is to represent something as
being yellow. I therefore conclude that the response is unsuccessful and that
non-comparative looks, or looksnc, like visible looks, are equivocal and so unfit
to index representational content.
3.3. An Argument from Visual Appearances
The second argument for representationalism I will consider is Susanna Siegel’s
‘Argument from Appearing’ (Siegel 2010: 44). Like Byrne, Siegel takes Travis to
be making a ‘semantic objection’ (ibid. 59) against the claim that looks report
the content of visual experiences. However, instead of focusing upon the kind of
‘non-comparative’ looks that Byrne describes, Siegel presents an argument for
the view that all experiences have representational content — a position which
she, like Byrne (3.3.1), calls the Content View, or CV for short. According to
Siegel, however, even proponents of relationalism about experience are
8 Alex Byrne in conversation at the SNNP Emotion and Perception Workshop, University of
Glasgow, October 2010.
Chapter 3, The Sound of Silence
58
committed to CV. If correct, this would mean that representationalism and
relationalism are not competing theses, as is often assumed, but compatible. By
arguing that perceptual experiences have accuracy conditions regardless of
which account of experience one adopts, Siegel aims to establish CV as common
ground between representational and relational accounts of experience, thereby
altering the terms of the debate.
In this section I assesses whether Siegel’s Argument from Appearing is
successful in establishing her claim that proponents of relationalism are
committed to CV. I argue that her account begs the question against
relationalism at more than one point — particularly in the way she characterises
the phenomenology of visual experience as involving the presentation of
properties, where this is conceived of in terms of property-types and not
property-instances (3.3.2). Consequently, the Argument from Appearing fails to
provide any reason for thinking that relationalism entails CV, though neither
does this rule out their compatibility. Thus, whilst there may be arguments that
can adjudicate between or demonstrate the compatibility of relationalism and
CV, this is not one of them. Possible rejoinders to these objections are discussed
in 3.3.3.
3.3.1. Siegel’s Content View
In her book The Contents of Visual Experience, Susanna Siegel aims to alter the
terms of the present debate by arguing that what she calls the Content View
(CV) applies equally to both representational and relational accounts of
experience, i.e. REP and REL (1.2.4). According to Siegel (2010: 28), for an
experience to have content is just for it to be capable of accuracy or inaccuracy
according to a set of accuracy conditions that are conveyed to the subject of that
experience. The accuracy conditions for a visual experience as of a red circular
object, such as one might experience upon seeing a tomato, for example, would
include (amongst other things) there being a circular red object — e.g. a tomato
— at the relevant egocentrically defined spatial location. If the object were not
red, or were in a different location from where it was perceived to be, then this
content would be inaccurate, and therefore non-veridical (cf. 1.2.3).
Nevertheless, the same experience may correctly predicate other aspects of the
object — for example its shape. Thus CV can account for the sense in which an
experience can be ‘more’ or ‘less’ accurate according to how much or how little
of its content is veridical (ibid. 32).
According to Siegel, content may be conveyed to a subject in one of three
ways (ibid. 51). First, it may become, or be systematically related to, the content
of an explicit belief — ‘there is a red tomato’, for example — that it is natural for
the subject to form as a result of having had the relevant perceptual experience.
Second, experiential content may guide conscious bodily action, such as
reaching out to grasp an object in the appropriate manner. Third, the content of
experience may be consciously available to the subject through the faculty of
introspection. In each case, however, the content of the experience is in some
sense ‘conveyed’ to the subject in thought or action, and not a merely formal or
Chapter 3, The Sound of Silence
59
theoretical feature of experience as described from a third-personal point of
view. This is in keeping with Travis’s Recognisability condition, though it is
unclear why first two cases should be taken as the evidence of representational
content of experience, rather than of belief or action, since such evidence would
also be compatible with anti-representationalist accounts of experience.
Nevertheless, the issue for the CV theorist will be to explain both how and why
such conveying is possible.9
Contents, for Siegel, ‘are a kind of accuracy condition’ (ibid. 30). Thus, she
argues, if it can be shown that perceptual experiences have the right kind of
accuracy conditions, i.e. those that are conveyed to the subject in experience,
then it will have been shown that experiences have contents. This is exactly
what the Argument from Appearing aims to achieve (3.3.2). This argument is
itself a refinement of a simpler argument, called the ‘Argument from
Accuracy’ (Siegel 2010: 34), which is as follows:
C1!
All experiences are accurate or inaccurate.
C2!
If all experiences are accurate or inaccurate, then all experiences
have accuracy conditions.
C3!
All experiences have accuracy conditions.
The Argument from Accuracy is unsatisfactory because, as Siegel herself
admits, it fails to secure the right kind of accuracy conditions for experience.
For any given experience e, there exists a trivial accuracy condition, namely that
e is accurate if and only if e is accurate. Such accuracy conditions correspond to
the application of the predicate ‘x is veridical’ and are not conveyed to the
subject via experience, which was one of the requirements for establishing the
existence of content, and therefore CV. The problem is that the application of a
predicate — in this case accuracy — does not itself entail the existence of content.
If it did, then the application of any predicate, such as being blue or being tall
would also entail the existence of content (cf. ibid. 30). Such trivial accuracy
conditions, however, are unable to differentiate between the contents of one
experience and another, since all experiences would have precisely the same
veridicality condition — namely, that the experience is accurate (blue, tall, etc.) if
and only if it is accurate (blue, tall, …) — and so the same ‘content’.
Consequently, we cannot conclude from the mere existence of accuracy
conditions that experiences have contents, since C1 might be true solely in virtue
of such conditions and yet CV be false since the relevant conditions are not
conveyed to the subject via experience.
The Argument from Accuracy fails to establish the desired conclusion that
all experiences have content (i.e. CV) as it is unable to yield the kind of content
— namely, representational content — that CV requires. It therefore needs to be
9
Not all representationalists accept that perceptual representations need be conveyed to the
subject. Burge (2010), for example, rejects this in favour of non-conscious representational
content (5.2.4).
Chapter 3, The Sound of Silence
60
refined not only to rule out such trivial accuracy conditions, but to explain both
why experiences are assessable for accuracy in the first place, and how these
conditions are conveyed to the subject (ibid. 44). Siegel’s Argument from
Appearing attempts to remedy these flaws.
3.3.2. The Argument from Appearing
Siegel’s argument begins with a supposedly uncontentious premise concerning
the phenomenology of visual experience. This is followed by a series of steps
which aim to derive the existence of accuracy conditions that are conveyed to
the subject, and so content. The idea is to show that advocates of REL, who
accept that we perceptually experience the properties of objects, are thereby
committed to experience having accuracy conditions that are conveyed by that
experience, and so content that is discernible to the subject. To do this, the
argument must derive CV from suitably neutral assumptions concerning the
nature of perceptual experience that are acceptable to proponents of REP and
REL alike. If successful, this would establish CV as common ground between the
representational and relational accounts, thus redefining the terms of the debate
concerning the nature of perceptual experience. I will argue that the argument
fails to do this.
Though Siegel’s articulation of the argument is somewhat complex, its
general structure is reasonably straightforward, proceeding from the presence of
visual properties in experience to the existence of representational content and
the Content View. Although I will mainly be concerned here with only the first
three premises, the entire argument is reproduced below for completeness (Siegel
2010: 45; minor formatting changes only).
A1!
All visual perceptual experiences present clusters of properties as
being instantiated.
A2!
If an experience e presents a cluster of properties F as being
instantiated, then:
A2b
A3!
Necessarily: things are the way e presents them only if propertycluster F is instantiated.
If necessarily: things are the way e presents them only if propertycluster F is instantiated, then:
A3b
e has a set of accuracy conditions C, conveyed to the subject of
e, such that: C is satisfied in a world only if there is something
that has F in that world.
A4!
If e has a set of accuracy conditions C, conveyed to the subject of e,
such that e is accurate only if C, then:
A4b!
A5!
e has a set of accuracy conditions C*, conveyed to the subject
of e, such that e is accurate iff C*.
All visual perceptual experiences have contents. (CV)
Chapter 3, The Sound of Silence
61
The first thing to note about this argument is the formulation of its initial
premise. Based on phenomenological reflection upon how visual experience
seems to us, experiences are said to ‘present clusters of properties as being
instantiated’. What precisely does this claim amount to, and what grounds does
Siegel offer for accepting it? Perceptual experience, it can be agreed, presents
objects. In most, if not all, cases those objects are perceived as possessing
particular properties: colour, shape, spatial location, and so forth. Moreover, it
is plausible that the function of experience is to make external objects and their
properties available to the subject in a way that enables them to form
appropriate beliefs, judgements, intentions, carry out actions, and so on. As
Siegel (ibid. 46) points out, objects are never presented in experience as bare
particulars, shorn of all their properties. This certainly constitutes grounds for
thinking that both objects and their properties — themselves, at least on some
views of what it is to be a property, aspects of objects — are presented in
experience. When we see a red tomato, we see both the tomato (an object) and
its redness (a property). Indeed, one might hold that it is only because we see the
tomato’s shape, location, and so on that we can be said to see the tomato, even
though seeing each of these individual properties may be neither necessary nor
sufficient for seeing tomatoes in general. Does this demonstrate that perceptual
experience presents properties ‘as being instantiated’? It is not clear that it does.
Firstly, it is not at all obvious what A1 is supposed to mean. Siegel notes
that this premise admits of both de re and de dicto readings. According to the de
re reading, which is the intended one, ‘for each visual perceptual experience,
there are some properties (colour, relative location, etc.) such that the
experience presents those properties as instantiated’ (ibid. 47). On the de dicto
reading, ‘there need be no specific properties such that [each experience]
presents them as instantiated’ (ibid.). However, this fails to explain what it is for
a property to be ‘presented as instantiated’. Given that the argument is supposed
to be neutral between REP and REL, it cannot be that for a property to be
presented in this way is for experience to have a predicational structure, since it
is a feature of relationalism that experience need have no such structure. More
importantly, given the vanishingly small distance between having a predicational
structure and having representational content, this would render A1
objectionable on grounds of circularity. Since experience having representational
content is supposed to be the conclusion of Siegel’s argument and not one of its
premises, we must therefore assume that A1 cannot mean (though it may entail)
that experiences have predicational structure.
A second possibility is that ‘as being instantiated’ is supposed to indicate
that perceptual experience is in some sense committal. That is, that the
experience of a red circular patch is not neutral concerning the way the world is,
in the manner of a supposition or imagining, but that it carries some kind of
prima facie epistemic force or authority. Of course, perceptual experience is not
committal in the same manner as judgement or belief, since we can doubt that
things are the way that they appear — when we experience a familiar visual
Chapter 3, The Sound of Silence
62
illusion, such as the Müller-Lyer, for example.10 That is to say, p-representation
is not autorepresentation (2.2.3). Rather, the intended distinction is supposed to
be something like the difference between merely entertaining some association
between an object and a property and the sense in which perceptual experience
might be thought of as reporting the way that things are, making it analogous to
a kind of testimony. This analogy with testimony, however, is problematic, since
if perceptual experience is a kind of report then it trivially has representational
content — namely, what is reported — and so the argument would again assume
that experience was representational from the outset. Indeed, whether
perceptual experience is in the relevant sense analogous to testimony is precisely
what is at issue between proponents of REP and REL, since the suggestion that
experience has face-value content is exactly what many relationalists, including
Travis, deny.
There is, however, one sense of ‘presentation’ in which relationalists can
agree that experience presents properties. This is the sense in which instances of
properties such as redness, roundness, egocentric location, and so on, are
presented in (veridical) perception. This is not the same as the claim that these
properties are represented in experience, since relationalists also hold that
experience cannot be exhaustively characterised in these terms. Nevertheless, if
objects and their properties are relata or constituents of perceptual experiences,
as REL claims, then it makes sense to say that they are presented to the subject in
experience. If ‘presented as being instantiated’ just means for instances of those
properties to be presented in experience, then it is unobjectionable, since this
claim is neutral between REP and REL. If, on the other hand, to be presented in
this way is to be predicated of an object, or to have propositional structure, then
A1 must be rejected as begging the question against REL. I will therefore assume
the following reading of A1, this being the one that is compatible with both
relational and representational models of experience, upon which I will proceed
to evaluate the remainder of the argument accordingly:
A1"!
All visual perception involves the presentation of clusters of
properties in experience.
3.3.3. Property-types and property-instances
The subsequent premises of Siegel’s Argument from Appearing are supposed to
constitute independently plausible claims about the notions of instantiation,
accuracy conditions and content. A2 makes the transition from the presentation
of properties in experience to a notion of accuracy that is based upon whether
those properties are instantiated. Roughly: things are the way that a visual
perceptual experience presents them as being only if the property-clusters that it
presents are actually instantiated. A3 moves from this to accuracy conditions
that are conveyed to the subject, claiming that if things are the way that a visual
10
For relationalists like Travis, this locution is misleading since they hold that there is no
such way that perceptual experience portrays the world as being (2.3.2).
Chapter 3, The Sound of Silence
63
perceptual experience presents them as being only if the property-clusters that it
presents are instantiated — i.e. A2b — then that experience has accuracy
conditions that are conveyed to the subject. Finally, A4 makes the transition
from accuracy conditions to content. Assuming Siegel’s definition of content as
a kind of accuracy condition that is conveyed to the subject in experience, as is
required for the validity of the argument, this step is relatively trivial.
To evaluate each of the above claims, we must consider the ways in which
properties may be presented in experience, which will differ between
representational and relational views. According to the representationalist,
properties are presented in experience by being represented to the subject. Since
such representations can be either true or false, there is a clear sense in which
the relevant properties are either instantiated or not, according to whether they
are possessed by the object or objects in question. Thus the representation of a
property and the instantiation of that property are separate and distinct entities.
The former constitutes part of the representational content of the subject’s
mental state, and is thus mental. The latter (or absence thereof) is some aspect
of the external world to which that representation is answerable, and is thus
non-mental.11 Under REL, however, the relevant property-instances (or those
aspects of an external object which correspond to the instantiation of its
properties) are directly present in experience. On this picture, there is no
difference between the property as presented in experience and its instantiation
in the world — they are one and the same thing. This makes A2 trivially true on
the relational view for all veridical perceptions, since such experiences
necessarily involve the presentation of property-instances and not of propertyrepresentations or types as is the case for the representational view.
The problem for Siegel’s argument is that it trades upon an equivocation
between these two senses of ‘presenting properties’. An advocate of REL will
accept the reading of A1′ according to which visual perceptual experiences
present property-instances since on their view objects and their properties are
themselves constituents of experience. The sense upon which the presentation of
properties would deliver accuracy conditions that are conveyed to the subject,
however, is the representationalist reading. To assume such a reading from the
outset would beg the question in favour of CV, since this assumes that
experience has a propositional structure, and so need not be endorsed by the
proponent of REL.
Furthermore, on the relational reading of A1′, only instantiated properties,
i.e. property-instances, are presented in the relationalist’s sense of ‘presentation’.
Thus A2 comes out as necessarily and trivially true for all veridical visual
perceptions, since all of the properties that are (according to REL) presented in
experience are instantiated, since only property-instances can be presented in
this way. On the relationalist reading of A2, then, A2b may be taken to mean:
11
On a nominalist account of properties, for a property to be instantiated just is for the
relevant objects to fall within the extension of the predicate that is represented. Siegel (2010:
58) also claims that a similar modification to the theory may be made to accommodate trope
theory, though I dispute this modification below (3.3.4).
Chapter 3, The Sound of Silence
A2b′
64
Necessarily: things are the way e presents them only if propertycluster F consists of property-instances.
which, in the case of veridical experience, is always true. However, on the
corresponding representationalist reading, A2b means:
A2b″ Necessarily: things are the way e presents them only if the
property-types in property-cluster F are instantiated by the
relevant object(s).
But being a property-instance in A2b′ and being instantiated by an object in
A2b″ are two quite different concepts; the former signifies that a particular falls
under a general metaphysical kind — i.e. the category of instantiated properties
— whilst the latter indicates the satisfaction of a predicate by an object. Indeed,
it is only due to a play on words that A2 can be taken to encompass both
readings. The term ‘instantiated’ as it appears in A2 is therefore also equivocal.
This presents a problem for Siegel’s argument because A3 only obtains if we
assume the representational reading of A1 and A2, i.e. A2b″, which take
‘properties’ to mean property-types. But on a non-question begging reading of
the argument, the relationalist will take ‘properties’ to mean property-instances,
in which case A2 is trivially true and A3 is false, since such accuracy conditions
are not conveyed to the subject by experience (see below). This renders the
Argument from Appearing invalid either due to (a) circularity, if the questionbegging reading of A1 is assumed, or (b) equivocation over the meaning of
‘presenting properties’.
Equivocation aside, however, there is a more fundamental problem with the
above reading of Siegel’s argument. If the relevant property-instances are only
‘instantiated’ in the metaphysical and not the representational sense, then the
argument fails to deliver the kind of accuracy conditions that CV requires. Recall
that contents, for Siegel, are a kind of accuracy condition that is conveyed to the
subject. But whether perceived instantiations of properties are in fact propertyinstances (as opposed to being subjectively indistinguishable from presentations
of property-instances—see 3.3.4) is not conveyed to the subject in experience.
For this to be the case, perceptual experiences would, according to the
relationalist interpretation of Siegel’s argument, have to pronounce upon
whether they were veridical or not — something that both sides agree they do
not.12 Consequently, the only kind of accuracy conditions that the above
reading of the argument entails are of the trivial kind, the likes of which Siegel
rejects in her Argument from Accuracy (3.3.1). Without a representational
understanding of ‘presenting properties’ and ‘instantiation’, the Argument from
Appearing fails to deliver or explain the link between the phenomenology of
12
Even those theorists who take perceptual contents to be presented as if they are veridical
would not endorse this stronger claim since non-veridical experiences are subjectively
indistinguishable from the corresponding veridical experiences.
Chapter 3, The Sound of Silence
65
perceptual experience and its content, and so fails to show why an advocate of
REL should endorse CV.
It is important to understand precisely why the presentation of propertyinstances, as opposed to property-types, cannot deliver the required accuracy
conditions. If, as relationalists claim, all that experience delivers is a collection
of objects and/or property-instances without thereby representing those entities,
then it would be impossible from a first-personal perspective to discern what the
accuracy conditions for that experience were. This is not to deny that such an
experience has accuracy conditions. But, as Siegel’s Argument from Accuracy
shows, the mere existence of accuracy conditions is not sufficient for content. In
this case, the resulting accuracy conditions will be of the form: experience e is
accurate if and only if e presents those particular (i.e. numerically identical)
objects and/or properties. Such conditions, which correspond to the experience’s
identity conditions, are not conveyed to the subject in experience, and thus do
not constitute content in Siegel’s terms. Without a specification of the types that
each individual property-instance must fall under, there is no way of deriving
accuracy conditions of the form ‘o is red’, as would be required to establish CV.
Thus the only accuracy conditions that the relationalist need admit that
experiences have are its identity conditions, which are insufficient to establish
the existence of content as they are not conveyed to the subject in experience.
Thus A3 is false.13
Furthermore, the relational reading of A1′ is only true for veridical
perception since, according to REL, non-veridical experiences do not present
property-instances. Rather, such experiences are subjectively indistinguishable
from experiences in which properties-instances are presented. Indeed, for
disjunctivists such as Martin (2004) this may be the only positive
characterisation we can give of them. Presumably, however, Siegel does not
intend for ‘visual perceptual experiences’ to include only veridical perception,
but also non-veridical experiences, such as hallucinations and illusions.14
Unfortunately, when we take these into account, the proposed reading of A1
breaks down since under REL , non-veridical experiences cannot be
straightforwardly characterised in terms of the presentation of the properties of
objects on the grounds that there may be no such objects. This is not necessarily
a problem for REL, which supports other ways of accounting for illusions (e.g.
Martin op. cit; Brewer 2008), but makes it impossible to evaluate Siegel’s
argument according to A1′. We must therefore either reject this reading of A1,
or reject A2 as false under the relational account since it only holds for veridical
perception.
13 One might object that for every property-instance that is presented in experience, there
must be some corresponding property-type of which it is an instance. I consider this
objection below (3.3.4).
14
Indeed, she explicitly states in her defence of A1 that ‘[t]hese considerations about the kind
of visual phenomenology involved in seeing ordinary objects … apply equally to cases of
merely seeming to see objects’ (Siegel 2010: 48, my italics).
Chapter 3, The Sound of Silence
66
The problem for the Argument from Appearing at this point may be stated
in the form of a dilemma. If we confine the argument to veridical perceptual
experience (this being the set of experiences for which A1′ holds), then the
argument only succeeds in showing that perceptual experiences have the kind of
trivial accuracy conditions that indicate whether they are veridical or not. This
is precisely the reason for which Siegel rejected the Argument from Accuracy,
since such accuracy conditions fail to explain how or why experiences should
convey the kind of propositional contents to the subject that CV requires (3.3.1).
If, on the other hand, we do not confine the argument to veridical perceptual
experience, as appears to be Siegel’s intention, then it is unclear how A1 can be
understood in a way that is neutral between Siegel and her opponents without
begging the question against REL. This renders the argument circular, and
therefore ineffectual. As it stands, then, the Argument from Appearing is
unconvincing because it fails to connect with the very view it is designed to
engage — namely, REL. Nor, it seems, is this dilemma easily resolvable without
substantial reformulation of the argument.
3.3.4. Responses
Having established the structure of, and some problems with, Siegel’s Argument
from Appearing, I now wish to consider a number of possible rejoinders on
behalf of the Content View. I will argue that none of these is satisfactory, and
consequently that Siegel’s argument should be rejected.
The first kind of response concerns the nature of CV itself. As formulated
above, CV is a claim about the presence of content in visual experience.
However, or so the response goes, this need not be understood as
representational content. Rather, the mere presentation of property-instances
itself constitutes a form of content, and so Siegel’s argument goes through even
on a relationalist understanding of perceptual experience. Whilst it is certainly
the case that philosophers have used the term ‘content’ in widely differing ways
— McDowell (2008) being a case in point — Siegel quite clearly ties the notion
of content to the existence of accuracy conditions that are conveyed to the
subject. Since such accuracy conditions are tantamount to representational
content as it is here conceived, the issue is not merely terminological. On the
above reading, the Argument from Appearing fails to demonstrate the existence
of such accuracy conditions, and so fails to establish CV, regardless of the notion
of content that Siegel takes herself to be employing. Furthermore, Siegel’s
characterisation of CV meets each of the conditions — including Face-value —
that Travis sets out for p-representation (2.2), and so is a legitimate target for
Travis’s arguments against representationalism. Plausibly, Siegel’s content is just
representational content, and the initial response is misplaced.
A second worry is that Siegel’s argument is based upon the phenomenology
of vision and so metaphysical concerns about the nature of experience are beside
the point. Whilst I do not wish to claim that representationalists and
relationalists experience differing visual phenomenology, there remains the issue
of how we are to understand the claims that Siegel is making, particularly in A1
Chapter 3, The Sound of Silence
67
of her Argument from Appearing. That theorists from both camps can sign up
to the same form of words is not sufficient to make the argument valid if, as I
have argued, it relies upon an equivocation over the meaning of ‘presenting
properties’, which is ambiguous between presenting property-instances and
representing property-types. Thus my objection is not that these two views of
perceptual experience deliver different verdicts about ‘what vision is like’,
phenomenologically speaking. Rather, my claim is that, to the extent that
Siegel’s phenomenological claim can be considered neutral between these two
views, it fails to establish what she thinks it does. Moreover, if ‘properties’ is
understood in the way that is required for the argument to go through, then its
initial premise can be rejected by the relationalist as question-begging.
References to the metaphysics of experience are merely intended to make this
fact conspicuous, rather than biasing the argument one way or the other.15
A more substantive response on behalf of Siegel at this point would be to
reformulate A1 using a disjunctive sense of ‘presenting properties’ that
accommodates both veridical and non-veridical experiences — call this
presents*. For something to present* F as being instantiated is just for it to
present F as being instantiated or to be subjectively indistinguishable from an
experience that presents F as being instantiated. Making the appropriate
substitutions, this results in the following reading of A1:
A1*!
All visual perceptual experiences present clusters of properties as
being instantiated or as subjectively indistinguishable from
experiences that present clusters of properties as being
instantiated.
Whilst this premise is true even under REL, according to which ‘present[ing]
clusters of properties as being instantiated’ would be interpreted to mean
presenting a series of property-instances, this generates problems later on in the
argument. If we take the consequent of A2 and antecedent of A3, for example,
to mean:
A2b*! Necessarily: things are the way e presents* them only if propertycluster F is instantiated.
then according to the disjunctive reading of presents*, A2b* will be true for
both veridical and non-veridical experiences, as expected. It does not follow
from this, however, that the resulting accuracy conditions will be conveyed to
the subject. This would only be the case if some set of property-types were
presented to the subject in experience, as occurs on the representationalist view.
The proponent of REL, on the other hand, need not be committed to this, which
15
A related objection is that my objection to Siegel draws upon a particular understanding of
the metaphysics of properties. However, this is arguably more of a problem for Siegel, who
employs the notion of ‘property-clusters’ without adequately specifying what she means by
this, leaving room for a level of ambiguity upon which, I claim, her argument depends.
Chapter 3, The Sound of Silence
68
goes beyond the claim that there are some sets of property-instances that, if
presented, would be subjectively indistinguishable from the corresponding nonveridical experience. This is not the same as saying that any specific propertytypes are presented in non-veridical experiences, since there could be an
innumerable number of experiences of properties-instances that were
subjectively indistinguishable from that experience, none of which would qualify
as the content of the experience. This is just another form of the dilemma that
Travis presses. In any case, the argument still conflates the representational and
relational senses of ‘presentation’, making A3 false on the disjunctive reading of
A1 given above.
A fourth response on behalf of the representationalist is that even if we
accept that property-instances and not property-types are presented in
experience, as per the relationalist account, for every property-instance
presented there must be some corresponding property-type of which it is an
instance that is thereby also present in experience. That is, one cannot have a
property-instance that is not an instance of some determinate type, and it is
these types, not their instances, that give rise to the experience’s accuracy
conditions. Siegel hints at this when she states that ‘[i]t is hard to see how …
accuracy conditions could fail to be conveyed to the subject in whatever way the
properties they derive from are’ (Siegel 2010: 53) as well as in her trope
formulation of the argument (below). This response is appealing because it
acknowledges the relationalist account of experience whilst allowing Siegel to
argue that such experience nevertheless has accuracy conditions and therefore
content.
The objection, however, now becomes one of which property-types are
represented in experience. According to Travis’s argument from looks (2.3.2),
there is simply no good answer to this question since there are a multitude of
properties that match any given visual appearance. The constraints upon what
would constitute an answer are the familiar ones of Recognisability, Givenness
and perceptual availability to the subject, thereby reinstating Travis’s dilemma.
Unless relevant property-type is discernible to the subject, the corresponding
accuracy condition cannot be said to be conveyed to the subject, and thus fails
to establish the existence of content in Siegel’s terms (3.3.1). Even if it is
discernible to the subject, then this cannot be solely on the basis of what is
perceptually available to them, but rather must draw upon what they are
inclined to believe or judge. This reintroduces circularity worries about the
explanatory role of experience with respect to judgement and belief, since prepresentation is representation-to and not representation-by (2.2.3). This
appears to be the route that Siegel herself favours (ibid.), but does not constitute
a reason to suppose that experience, as opposed to belief, judgement or action,
has content. It is therefore unclear that the original objection can be
circumvented in this way.
The final counter-argument that I wish to consider is that my objection to
Siegel commits the relationalist to the existence of tropes. Relationalism about
perception, as I have described it, involves the presentation in experience of
property-instances where such instances are either concrete individuals (i.e.
Chapter 3, The Sound of Silence
69
tropes) or aspects of external objects. Thus, what is presented when visually
experiencing a red tomato is not the abstract universal redness, but rather the
redness of that tomato, where this is distinct from the redness of other objects,
including other tomatoes that are of precisely the same colour. The existence of
tropes is contentious and so not necessarily something the relationalist would
want to commit herself to, and certainly not on the basis of visual
phenomenology alone. Thus, so the counter-argument goes, the objection stands
or falls with the existence of tropes. Moreover, Siegel claims that her argument
‘could easily be reformulated to accommodate this position’ (Siegel 2010: 48)
simply by substituting ‘clusters of properties’ for ‘cluster of F-tropes’, where F
indicates the relevant trope types.
Taking the latter point first, this way of stating the argument in terms of
trope types begs the question discussed above of precisely which types are
presented in experience. The point about invoking tropes from the relationalist
perspective is that they show how particulars need not be presented as falling
under any particular type. To assume that they are presented in this way
without further argument is not a response to this objection. Indeed, if we take
‘a cluster of F-tropes’ in the reformulated A3 to mean a cluster of those very
tropes — i.e. a token- rather than type-identical experience — we arrive back at
the original objection since the resulting accuracy conditions will be none other
than the identity conditions for that very experience.16 Such identity conditions
are not conveyed to the subject in experience, and so the argument fails to
establish CV.
On the former more substantive point, the trope-like formulation of the
relationalist objection is merely intended to make explicit the commitment to a
representationalist conception of properties and property-types that is required
to make Siegel’s argument valid. Unfortunately for Siegel, this also undermines
any possible dialectical force that it might have against anti-representationalism,
which is its intended target. Another way of making the same point would be to
claim that only objects, including those aspects of them that instantiate
properties, are presented in experience. Such a formulation would be neutral
with respect to the existence of tropes, but presses the same objection that
objects need not be presented under a particular property-type in order for them
to feature in perceptual experience. The presence of visual properties such as
redness, roundness, being over here, and so on would then be explained in terms
of our perceptual sensitivity to certain types of stimuli, but without the relevant
types figuring at the level of perceptual experience. That such types may show
up in the content of judgements and beliefs is not, on the relationalist picture,
16
Cf. ‘A trope version of [A3] would look like this:
If necessarily: things are the way E presents them only if a cluster of F-tropes is instantiated,
then:
E has a set of accuracy conditions C, conveyed to the subject of E, such that:
C is satisfied in a world only if there is something that has a cluster of F-tropes in
that world.’ (Siegel ibid. 58, fn. 29).
Chapter 3, The Sound of Silence
70
reason to attribute contents to experiences proper. Thus the relationalist need
not commit herself to trope theory in order to make the relevant objection.
3.4. An Argument from Perceptual Seeming
The final argument in favour of representationalism that I wish to consider is
Susanna Schellenberg’s ‘Master Argument’ (Schellenberg 2011b). Like Siegel’s
Argument from Appearing (3.3), the Master Argument aims to derive the
conclusion that perceptual experience has representational content that is
assessable for truth or accuracy, this time from a supposedly neutral notion of
perceptual awareness or ‘seeming’. As with Siegel’s argument, however, the
relevant notions turn out to be far from neutral, instead being biased in favour
of representationalism, thus rendering the argument circular and so
unpersuasive.
3.4.1. Schellenberg’s Master Argument
The first part of Schellenberg’s argument, given below, is intended to establish
the existence of content. This is followed by a second line of argument for the
view that such content necessarily has accuracy conditions.17 Assuming that by
‘content’ Schellenberg means representational content, the latter part of the
argument is relatively uncontroversial since all parties can agree that perception
has ‘content’ in the sense of the objects and/or properties that are in fact
perceived. I will therefore restrict most of the following discussion to the part of
the argument given in Schellenberg (2011b: 6).
M1
If a subject is perceptually related to the world (and not suffering
from blindsight etc.), then she is aware of the world.
M2
If a subject is aware of the world, then the world seems a certain
way to her.
M3
If the world seems a certain way to her, then she has an
experience with content C, where C corresponds to the way the
world seems to her.
M4
If a subject is perceptually related to the world (and not suffering
from blindsight etc.), then she has an experience with content C,
where C corresponds to the way the world seems to her.
Taking each of the above premises in turn, M1 is intended to be something
upon which all perceptual theorists can agree. Bracketing concerns with the
formulation of this premise in terms of the subject’s relation to and awareness of
‘the world’, as opposed to the objects and properties in it, this premise may
therefore be endorsed by Schellenberg’s opponents. Schellenberg goes on to
17
This contrasts with Siegel’s Argument from Appearing (3.3.2), which takes the order of
precedence to be the other way around.
Chapter 3, The Sound of Silence
71
claim that ‘[o]n the face of it, the fact that a subject is aware of the world entails
that the world seems a certain way to her’ and that ‘there is no obvious reason
why [pure] relationists should not accept [M2]’ (ibid. 8). However, again
bracketing similar concerns about the precise formulation of these claims, it is
far from clear that this is the case. The problem lies with an ambiguity in the
word ‘seems’.
There are two relevant readings of ‘seems’ in M2, neither of which appears
capable of delivering the intended result. The first corresponds to the relatively
uncontentious claim:
M2′
If a subject is aware of the world, then her experience has a
certain phenomenal character.
That is to say that for every perceptual experience, there is something it is like to
have that experience (1.2.2) — the particular reddish and roundish character of
the experience that one typically has upon seeing a ripe tomato, for example.
This reading does secure a notion of representational content (C), but it is a
form of phenomenal content that corresponds to the phenomenal character of
perceptual experience — call this Cp — and not to the non-appearance properties
of perceptual objects — call this Co.18 As I argued in relation to non-comparative
looks (3.2.2), this notion of content does not secure any specific way that the
world would need to be in order for an experience to be veridical. Rather, it
delivers a ‘thin’ notion of phenomenal content that is indeterminate between
innumerable ways that the world might be.
This objection constitutes what Schellenberg, following Travis, terms the
indeterminacy objection (ibid. 18). According to the objection, the accuracy
conditions for Cp relate to properties of the subject’s experience and not the
objective properties of the objects perceived. If the objection holds, then premise
5 in the second part of Schellenberg’s argument is no longer coherent
(Schellenberg 2011b: 6):
M5
The world is either the way it seems to [the subject] or it is
different from the way it seems to her.
since the accuracy conditions for Cp do not relate to the world in any objective
sense, but rather to the subject’s experience of it. Consequently, these conditions
cannot fail to obtain even in non-veridical cases (3.2.2). Thus, on the reading of
‘seems’ given in M2′, Schellenberg’s Master Argument only succeeds in
establishing that phenomenal content has accuracy conditions, and not that
experience has any objective representational content.
No doubt many representationalists would reject the above objection on the
basis that the truthmakers for phenomenal content are not supposed to be
18
According to some theories of phenomenal character, such as Naïve Realism, these
amount to the same thing. But this is a substantive claim that cannot simply be assumed
from the outset.
Chapter 3, The Sound of Silence
72
experiences, but objective states of the world. Thus, even though such content
may be individuated by the subjective phenomenal character of perceptual
experience, it is still answerable to how things objectively are in the world for
the satisfaction of its truth or accuracy conditions. However, this again raises
the problem of precisely which contents experiences have. If they are supposed
to have objective content, then the representationalist must show why they have
those particular objective contents as opposed to any other which shares an
indistinguishable perceptible appearance; i.e. Travis’s indeterminacy objection.
If they are supposed to be satisfied by any perceptible objects that yield the same
subjective phenomenal character then it is no longer clear how the relevant
accuracy conditions can be specified independently of the subject’s experience. If
they cannot, then the accuracy conditions for any given perceptual experience e
are just that the world is such as to gives rise to an experience whose
phenomenal character is subjectively indistinguishable from e. However, this
cannot be the right way of specifying such content given that any experience
could, under appropriate circumstances, give rise to such an experience — for
example, if the subject had taken a hallucinogenic drug that predisposes her to
hallucinate that P on the basis of seeing objects for which not-P. In order to
circumvent this kind of objection, some kind of normative constraint must be
introduced, such as whether such objects routinely give rise to P-experiences (cf.
Siegel 2010: 57). The problem remains, however, as to how the phenomenal
properties of experiences — reddishness, roundishness, and so on — can be met
by some particular state of the world without simply assuming the point at
issue; namely, that the relevant experiences have objective accuracy conditions.
Although more could be said about the above issue, it is clear that the
subjective phenomenal reading of ‘seems’ is not what Schellenberg has in mind.
Rather, the relevant seemings are supposed to specify some objective state of the
world (Schellenberg 2011b: 10). This suggests that M2 should instead be
interpreted as follows:
M2″
If a subject is aware of the world, then there is some determinate
and univocal way that the world seems to her to be.
The trouble with this interpretation of M2 is that it is unacceptable to any antirepresentationalist who, like Travis, denies that perceptual experiences have a
univocal face value — that is to say, a way that the world, in the relevant sense
seems to be (2.2.2). On this reading of ‘seems’, M2″ assumes precisely this
principle from the outset. Thus, Schellenberg’s argument that experience has
representational content assumes that her opponent’s position is false. This
seriously undermines her argument against any anti-representationalist who
denies Face-value, for whom this reading of M2 will be unacceptable. Given
that Travis (2004) is such a representationalist, this renders Schellenberg’s
Master Argument ineffectual against his position.19
19
This is not, of course, to say that Schellenberg’s conclusion is false. Rather, the Master
Argument is dialectically ineffective against Travis’s anti-representationalism.
Chapter 3, The Sound of Silence
73
Schellenberg goes on to discuss this ‘seems-content link’ (Schellenberg
2011b: 10) that is presupposed by M3 at some length. She concludes that the
link holds in at least the case of non-comparative looks, and possibly also in the
case of visual and thinkable looks if one rejects Travis’s argument from looks.20
However, as we saw above, the link does not hold in the case of noncomparative looks, which turn out to be as indeterminate between various ways
that the world might be as Travis’s visual looks. Furthermore, Schellenberg’s
grounds for rejecting Travis’s argument, which she takes from Byrne (2009), are
problematic since Byrne appears to reject the very ‘seems-content’ link (i.e.
Looks-indexing, 2.2.5) to which Schellenberg appeals on the basis of noncomparative looks. This undermines Schellenberg’s claim that her Master
Argument succeeds in establishing a form of objective perceptual representation,
since it is precisely this link that secures the necessary reading of ‘seems’ in M2
without which the move to M3 is blocked. However, even if, as I argued above,
Byrne does endorse a version of Looks-indexing, this would only secure the
existence of ‘thin’ phenomenal content, which is insufficient for the
representationalist’s purpose. Thus, while the Master Argument may, according
to the phenomenal reading of ‘seems’, be successful in establishing the notion of
phenomenal content, it is unsuccessful in establishing the existence of objective
representational content in perceptual experience.
Furthermore, it is unclear from the above argument exactly what the
argumentative force of M1 through M3 is supposed to be. There does not
appear to be any logical reason to accept M2 on the basis of M1, or M3 on the
basis of M2. After all, it is hardly part of our pre-theoretical intuitions that
perceptual awareness entails the world’s seeming to be some particular way, or
that such seeming entails representational content. Indeed, these are precisely
the points that are at issue in Travis’s argument from looks. Rather, these
premises appear to be intended as a kind of ‘intuition pump’ that facilitates a
slide from ‘awareness’ in M1 to ‘content’ in M3. From there, it is possible to
derive the conclusion for which Schellenberg is aiming in M4 and beyond.
Alternatively, perhaps Schellenberg intends for her argument to allude or
‘point’, in a more metaphorical sense, to some notion of content entailed by the
notion of perceptual awareness. However, far from being the innocent moves
that Schellenberg appears to suggest, M2 and M3 are highly contentious and, in
the case of M2, potentially question-begging from the anti-representationalist’s
perspective. 21 It is therefore possible for the anti-representationalist to simply
reject M2 on the basis that the required ‘way of seeming‘ — for example,
epistemic seeming — is unavailable to the subject on the basis of perceptual
experience alone, thus preventing Schellenberg’s Master Argument from getting
off the ground.
20
Indeed, the relevant readings of ‘seems’ are precisely those that are dealt with in Travis’s
argument from looks (2.3.2).
21
Schellenberg herself assumes that M3 will be the point at which anti-representationalists
seek to resist the argument, whereas M2 is clearly more objectionable by Travis’s lights.
Chapter 3, The Sound of Silence
74
3.5. Conclusion
I have argued that, in the face of Travis’s arguments against representationalism,
none of the above approaches is successful in establishing that perceptual
experience has representational content. Byrne’s ‘non-comparative’ looks only
establish the existence of phenomenal content, i.e. content whose accuracy
conditions vary with experience itself, rather than with the objects and
properties towards which it is directed. Such content fails to meet Travis’s
Objectivity condition and cannot be used to ground claims to objective
knowledge or belief as it is necessarily veridical. Siegel’s Argument from
Appearing suffers from a similar flaw as her Argument from Accuracy, since it
fails to distinguish between the notion of accuracy conditions as a predicate and
accuracy conditions as content (cf. Martin forthcoming b). Furthermore, it
conflates a familiar fact about looks or appearances — namely that they may be
paradigmatic of certain types of objects and their properties — with a
contentious philosophical claim about the structure of perceptual experience,
namely that it presents properties ‘as being instantiated’, where this is taken to
presuppose a distinction between property-types and their instances.
Consequently, the argument fails to have any dialectical force, being instead a
‘dramatic expression of conviction’22 that merely asserts what
representationalists already believe — i.e. that perceptual experience has
representational content — without showing why anyone who does not already
hold this view need be committed to it. Finally, Schellenberg’s Master Argument
relies upon an equally question-begging notion of perceptual seeming, which
presupposes exactly the contents whose existence it was supposed to
demonstrate. As such, each of these arguments must be rejected as a means by
which representationalism may be established from a neutral starting point.
More generally, each of the above arguments can be seen as an instance of a
tempting line of thought: namely, that the existence of accuracy or veridicality
conditions, and therefore of representational content, can be derived from the
mere existence of perceptual appearances. Unfortunately for the
representationalist, as I argue in the following chapter, mere appearances — or
‘looks’ — are neutral between representational and relational views of
experience, and so cannot be taken to settle the debate one way or the other.
Whilst none of this rules out the possibility that experience may indeed have
representational content, demonstrating that is does so turns out to be a difficult
philosophical problem that is not easily resolved by an appeal to appearances.
For all that the arguments presented above show, Travis may yet be right that
the senses are indeed ‘silent’, as he seeks to demonstrate.23
22
23
This phrase was suggested to me by Mark Kalderon (2011).
Parts of this chapter have benefited from invaluable feedback and discussion on
presentations given at the Universities of Oxford, Warwick, Columbia and Essex, including
comments from Johannes Roessler and Anna Marmodoro.
4. Looks
A deflationary analysis
4.1. Introduction
In the previous chapter I suggested that Byrne’s (2009) account of perceptual
experience may commit him to a form of content, i.e. phenomenal content, that
is incapable of meeting his own objection to Travis — namely, the need to
provide a satisfactory account of perceptual illusion. In doing so, I identified
two possible strategies by which this problem could be overcome whilst
providing a satisfactory response to Travis’s argument from looks (2.3.2): (i) by
identifying a genuinely non-comparative notion of looks that supports both
Face-value and Looks-indexing, and (ii) by defending an externalist account of
how subjects are able to gain first-personal knowledge of the contents of their
experiences. The first of these strategies is examined below whilst the second is
discussed in chapter 6. Moreover, I argue that considerations concerning the
nature and semantics of ‘looks’ are insufficient to establish the existence of prepresentation. Thus, representationalists require an alternative motivation for
their position.
According to Looks-indexing, p-representational contents must be
recognisable to the subject in virtue of how, according to the relevant
experiences, things appear — or in the case of visual perception, how things look
(2.2.4). In order to refute Travis’s argument from looks without abandoning the
notion of p-representation, the representationalist must therefore either (a)
elucidate an alternative notion of looks that satisfies Face-value, but is not
reducible to either visible or thinkable looks, neither of which is capable of
indexing representational content on the basis of what is perceptually available
to the subject (2.3.2), or (b) reject Looks-indexing (or the weaker
Recognisability condition) outright, in which case Travis’s other arguments will
apply — specifically, the argument from recognisability (2.3.3) and the argument
from unmediated awareness (2.3.4), responses to which are examined in
chapters 5 and 6.
Chapter 4, Looks
76
The first part of this chapter examines various notions and analyses of looks
in order to establish whether there is a notion of ‘non-comparative’ or
phenomenal looks that is capable of making representational contents
recognisable (4.2). The second part of the chapter examines the prospects for an
alternative ‘deflationary’ analysis of our everyday looks talk in purely
comparative terms (4.3). By neutralising the challenge from looks-based
accounts of visual experience, such an analysis would render the notion of noncomparative looks effectively redundant, closing off a possible line of response
to Travis by the representationalist about perceptual experience. This in turn
forces the representationalist to find an alternative way of justifying their
position (4.4).
4.2. The Varieties of Looks
Much of the terminology surrounding the various notions of looks originates
with Chisholm (1957), who first proposed the distinction between epistemic,
comparative and non-comparative looks. These roughly, though not always
precisely, correspond to what I have been calling thinkable, visible and
phenomenal looks, respectively. I now wish to present a more nuanced account
of these distinctions that differentiates between the evidential and non-evidential
senses of ‘looks’ (4.2.1), comparative and non-comparative analyses of
‘looks’ (4.2.2), and the relation between looks and the phenomenal character of
experience (4.2.4). I then consider whether the phenomenal use of ‘looks’ may
itself be analysed in comparative terms, thus rendering it unsuitable for indexing
the representational content of perception (4.2.6). Finally I conclude that, even
if Jackson (1977) and Brogaard’s (forthcoming) arguments are successful in
establishing a distinctive phenomenal usage of ‘looks’, this use is nevertheless
insufficient for the representationalist’s purpose, thus making room for the
essentially comparative or ‘deflationary’ analysis presented in 4.3.
Before I proceed to examine these various notions, it is important to
distinguish between, on the one hand, the semantic analysis of ‘looks’, ‘seems’,
‘appears’ and their cognates and, on the other, the metaphysics of appearances
and phenomenal character. In this chapter I concentrate mainly upon the
former. In 4.3, however, I will argue that the availability of alternative analyses
of ‘looks’ effectively blocks the use of such semantic analyses to draw
conclusions about the metaphysics of experience concerning the existence and
nature of representational content. Given the weight that some of the recent
literature places upon the semantics of ‘looks’, and the misreading of Travis
(2004) as making a narrow point about the meaning of ‘looks’ rather than
about experience more generally, there is good reason to be interested in the
semantic question in its own right. However, even if this were to suggest the
existence of the kind of ‘non-comparative’ looks that Byrne and others describe,
I argue below that it would be of little consequence to the argument over
representationalism and, as such, the debate should be conducted upon other
more fertile grounds.
Chapter 4, Looks
77
4.2.1. Evidential and non-evidential looks
The distinction between epistemic and comparative looks originates with
Chisholm in connection with the evidential use of appearance expressions
(Chisholm 1957: 44). However, as Chisholm also points out (ibid. 47), these
two varieties of looks are closely related in that many sentences may be used to
express either or both senses. For example,
(1)
The ball looks red to S
can mean either that the subject, S, has ‘adequate evidence’ (ibid. 44) for the
proposition that the ball is red (the epistemic use, or ‘thinkable looks’) or that,
in respect of its colour, the ball visually appears to S the way that red things do
(the comparative use, or ‘visible looks’). Indeed, that the ball in some respect
visually resembles red things may itself constitute S’s evidence for the ball’s
being red. In this case, the comparative reading of (1) is also being used
evidentially, demonstrating that these two uses of ‘looks’ need not be mutually
exclusive.
‘Looks’ can also be used in non-perceptual contexts, as in ‘It looks as if the
government will be re-elected’ (Jackson 1977: 31). Such uses are clearly
evidential, but need not be limited to just visual evidence as in the case of
perceptual looks.1 The existence of the non-perceptual use, however, suggests a
useful test for multiple senses of ‘looks’. As Brogaard (forthcoming: 18) points
out, words may exhibit two distinct forms of ambiguity: (i) lexical ambiguity, as
in the case of ‘bank’, which can mean either the edge of a river or a financial
institution; and (ii) polysemy, as with ‘fine’ which has multiple related
meanings, as in ‘a fine restaurant’ or ‘finely shaped features’ (ibid.). The former
reflects a mere accident of syntax where two distinct concepts just happen to
share the same spelling, though may be tokened by different words in other
languages. The latter is ‘systematic’ in that the various meanings are
semantically, and not just syntactically, related (ibid.). By mixing different uses
of ‘looks’ within a single sentence we can identify whether this term is
systematically ambiguous and what its various senses are.2
For example, if one were reading a newspaper under green light, then the
sentence
(2)
It looks as if the government will be re-elected and the paper is
green
1
I take it that looks are a kind of appearance that, excluding non-perceptual uses, are visual
in nature, just as tastes, smells, sounds, and so on, are forms of appearances to the other
senses. In what follows, I will treat ‘looks’ and ‘visual appearances’ as synonymous.
2
Lexical ambiguity can be ruled out because there are many languages in which each of the
uses of the English ‘looks’ may also be expressed using a single term (ibid. 19).
Chapter 4, Looks
78
will have an infelicitous reading on which ‘it looks as if’ has wide scope and so a
single instance of ‘looks’ is used to convey both perceptual and non-perceptual
meanings. Similarly,
(3)
It looks like the train is late and over there
has an erroneous reading on which ‘over there’ is used to demonstratively
identify the relevant object (the train). (Its legitimate reading is one on which
both conjuncts employ the evidential use of ‘looks’, as when reading the
platform number and expected time of arrival from a departures board, for
example.) The infelicity of such sentences suggests that there are (at least) two
senses of ‘looks’, which I will refer to as its evidential and non-evidential senses,
respectively.
The evidential sense of ‘looks’ relates to what one’s total evidence base
indicates to be the case, and so corresponds with Travis’s notion of ‘thinkable
looks’ (2.3.2). For example, ‘seeing Luc and Pia’s flat strewn with broken
crockery’ might, under certain circumstances, be taken to indicate that they have
had a tiff (Travis 2004: 67). The evidential sense can be further subdivided into
perceptual and non-perceptual uses, the former of which includes visual,
auditory, tactile, and other sensory evidence. When ‘look’ is used in this sense,
(1) means that, according to the visual or other perceptual evidence available to
S, the ball looks to be red, giving epistemic support to the proposition
(4)
The ball is red
This may be contrasted with the non-evidential sense of ‘looks’, according to
which (1) is intended to be neutral as to whether the ball is red or not. On this
reading, (1) merely states how the ball appears to S without any commitment as
to how it actually is. This is what we might refer to as a mere look. Indeed, one
could hold that the ball both non-evidentially looks red but that it is purple
without contradiction. The non-evidential sense thus provides no epistemic
support for (4) and is what a subject might retreat to when challenged about, or
given reason to doubt the veridicality of, their experience. Upon discovering that
the ball is not in fact red, for example, one might continue to assert that it looks
red, meaning that it could easily be mistaken for such.
4.2.2. Comparative and non-comparative looks
Common locutions of the evidential sense of ‘looks’ include ‘looks as if’ — for
example, ‘it looks as if it will rain’ — and some cases of ‘looks like’, such as ‘it
looks like [it will] rain’. More commonly, however, ‘looks like’ introduces the
comparative use of looks. The comparative use also has two forms: explicitly
comparative and implicitly comparative. The former invites an explicit
comparison between the appearances of two or more objects, as in
(5)
Charles looks like the prime minister
Chapter 4, Looks
79
Here, the visual appearance of one object (Charles) is being compared with the
appearance of another (the prime minister). Moreover, (5) has both comparative
evidential and comparative non-evidential readings, which may be paraphrased
as follows:
(5′)
The [visual] evidence indicates that Charles is the prime minister
(5″)
The look of Charles is similar to the look of the prime minister
The latter of which may be analysed as
(5*)
∃x∃y (look(x) ∧ (has(Charles, x) ∧ (look(y) ∧ (has(The Prime
Minister, y) ∧ SIM(x, y)))))
Note that (5*) need not predicate the same look of both objects. Instead, it
posits a visually relevant similarity between the two (4.3). However, in cases
where two objects share precisely the same look, we can simplify this to:
(5**) ∃x∃y∃s (look(s) ∧ (has(x, s) ∧ has(y, s))) 3
The above analyses bring out two important features of comparative looks.
First, they do not specify how a thing has to be in order to be the way it looks,
merely that it shares some relevant feature, or features, with another object or
look. Thus, the value of s in (5**) is not fixed by the semantic properties of (5),
but rather relies upon various contextual factors, including the subject’s prior
knowledge or perceptual abilities, to establish what the relevant similarity is. It
is therefore a substantive question just what the referents or bearers of
comparative looks are. Second, comparative looks are symmetrical in that the
referents of (5′) or (5″) can be reversed without altering the statements’ truth
values. This has important implications for the debate about looks-indexing,
since if x comparatively looks like y then y also comparatively looks like x.
Consequently, as Travis argues, comparative looks alone cannot make a specific
univocal face-value content recognisable to the subject, since it is ambiguous
whether it is x or y that is being represented. To that extent, a look that x is
equally a look that y. An essentially comparative analysis of ‘looks’ is therefore
unacceptable to any representationalist who accepts Looks-indexing since it is
incompatible with the relevant content having a univocal face value, as per
premise L3 of Travis’s argument from looks (2.3.2).
An implicitly comparative looks report, on the other hand, is one that does
not exhibit the surface structure of (5) — i.e. ‘φ looks like ψ’ — but which is best
understood in terms a comparison with some other object or look. For example,
3
Cf. Brogaard 2010: 6. For simplicity, I assume that ‘the prime minister’ is a singular
referring term. Such an assumption is not required for a comparative analysis of ‘looks’ and
may be replaced with a suitable account of definite descriptions; e.g. Russell (1905).
Chapter 4, Looks
(6)
80
Johan looks Scandinavian
might plausibly be given an analysis of the form:
(6*)
∃s (look(s) ∧ (has(Johan, s) ∧ has(φ, s)))
where φ is some real or imagined individual of characteristically Scandinavian
appearance.4 As with explicitly comparative looks reports, implicitly
comparative reports can have both evidential and non-evidential senses. More
importantly for present purposes, as with explicitly comparative looks,
implicitly comparative looks are symmetrical and so similarly incapable of
indexing univocal representational content. One might object to this on the
basis that φ has a specific privileged role, and so in the above example is
capable of indexing the univocal content being Scandinavian. However, if φ is
simply an exemplar of the relevant predicate then, assuming that (6*) is true, φ
equally looks like Johan, comparatively speaking, as Johan looks like φ. That
we don’t typically speak in this way is arguably more a matter of convention
given our linguistic interests, than anything of deep metaphysical significance.
4.2.3. Senses, uses and the regress argument
The above considerations raise the question of whether everyday non-evidential
comparative looks reports that are not explicitly comparative are implicitly
comparative. If this were the case, then such looks would be unable to index the
kind of univocal representational contents that representationalism requires. In
order to resist this charge, the representationalist must provide a noncomparative analysis of ‘looks’ that cannot be expressed in terms of an
equivalent comparative looks report since, if such an equivalence were to hold,
then ‘looks’ cannot index univocal content and Travis’s argument goes through.
Conversely, if looks can be given an irreducibly non-comparative analysis, then
premise L3 of Travis’s argument from looks is false since it is possible to meet
Looks-indexing without contravening Face-value. I consider the case for and
against such an equivalence throughout the rest of this chapter.
The fact that both comparative and non-comparative looks can be evidential
or non-evidential suggests that either (a) the comparative and non-comparative
uses of ‘looks’ are not distinct senses, or (b) there are (at least) four senses of
‘looks’ corresponding to each of the possible combinations of evidential or nonevidential versus comparative or non-comparative looks. To resolve this
question, we can carry out a similar test for the existence of comparative and
non-comparative senses of ‘looks’ as for the evidential and non-evidential
senses. For example,
(7)
4
The bread looks burnt and like a brick
For an account of ‘looks’ that rejects the comparative analysis of (6), see Byrne (2009).
Chapter 4, Looks
81
seems perfectly felicitous, as does its evidential analogue:
(8)
The president looks as if he will be re-elected and govern
differently to his predecessor
As Brogaard (forthcoming: 21) points out, the acceptability of sentences like (7)
may also be explained by providing a comparative analysis of both conjuncts.
That is, there may be a reading of this sentence such that its first conjunct,
which is not explicitly comparative, is analysed as:
(7*)
∃s (look(s) ∧ (has(The Bread, s) ∧ SIM(Burnt, s)))
This would remove any potential ambiguity by making both looks comparative.
A similar analysis, however, cannot be given for (8), since ‘being re-elected’ is
clearly non-comparative. This suggests that the comparative element of ‘looks’ is
a feature of its analysis or use, rather than a distinct sense in its own right. If
this is correct, then the presence of an explicit or implicit comparative element
should be considered an attribute of particular looks reports, rather than part of
the meaning of ‘looks’ itself (cf. ibid.).
One argument that is sometimes given for the primacy of the noncomparative notions of looks over the comparative notion (e.g. by Byrne 2009:
441 and Brogaard forthcoming: 22) is that the existence of the former is
presupposed by the latter. Call this the regress argument. According to this
argument, comparative analyses like (7*) presuppose that there is some way that
things look — s in the above example — that grounds the truth or falsity of
comparative looks reports. On pain of circularity, such looks cannot themselves
be comparative, and must instead be some other, more primitive form of looks,
namely non-comparative looks.
The regress argument, however, is too quick. There is nothing to say that s
in the above analysis must exclusively be a look, or if it is, that such looks must
be analysed non-comparatively. All that is required to avoid the alleged regress
is that s must not itself be the same comparative look, or some precursor of it,
that is being analysed. This leaves considerable scope for such analyses to
quantify over, for example, the external objects of perception or their properties
(4.3). Furthermore, any look that grounds the truth of a comparative looks
report may itself be comparative in nature. For example, if phenomenal looks
turned out to be analysable in terms of visually relevant similarities between
objects, then those similarities would ground the truth of the corresponding
comparative looks reports. Since the notion of similarity is itself comparative,
then looks of this kind would be irreducibly comparative as they cannot be
expressed in terms of any more primitive non-comparative notion. Thus Byrne
and Brogaard’s regress argument against the primacy of comparative looks fails
to go through.
Chapter 4, Looks
82
4.2.4. Phenomenal and non-phenomenal looks
Having identified the existence of evidential, non-evidential, comparative and
non-comparative looks reports, and the various combinations thereof, there
remains the further question of which, if any, of these is used to give
phenomenological reports. That is, are there uses of sentences like (1) — ‘The
ball looks red to S’ — that report the phenomenal character of S’s perceptual
experience, or are such sentences essentially descriptive of objects in the world?
The prospects of a looks-indexed account of perceptual content will turn upon
the answer to this question. Moreover, if any use of ‘looks’ is capable of
indexing univocal perceptual content, then it will be the non-comparative use
since, as we have seen, comparative looks are symmetric and so equivocal. A
potential candidate for such a non-comparative use is Jackson’s
‘phenomenological’ or ‘phenomenal’ use, which he describes as follows (Jackson
1977: 33):
The phenomenal use is characterized by being explicitly tied to terms for
color, shape, and/or distance: ‘It looks blue to me’, ‘It looks triangular’, ‘The
tree looks closer than the house’, ‘The top line looks longer than the bottom
line’, ‘There looks to be a red square in the middle of the white wall’, and so
on. That is, instead of terms like ‘cow’, ‘house’, ‘happy’, we have, in the
phenomenal use, terms like ‘red’, ‘square’, and ‘longer than’.
If this is correct, then sentences like
(9)
It looks red and very old
should have an infelicitous reading since only ‘red’ and not ‘very old’ has an
obviously phenomenal use.5 However, this does not appear to be the case. This
apparent discrepancy can be explained either in terms of the corresponding
comparative reading
(9′)
It looks the way that red things look and the way that very old
things look
‘obscuring’ the erroneous reading (Byrne 2009: 442), or by the phenomenal
reading itself being analysable in comparative terms — a possibility that Jackson
rejects for reasons I examine in the following section. For present purposes,
however, it will suffice to clarify precisely what Jackson’s phenomenal use is
intended to capture, and why one might be tempted to posit it as a means of
indexing representational content.
Perhaps the strongest argument for the existence of distinctly phenomenal
looks comes from the following scenario described by Jackson (1977: 36). A
‘super-achromatic-color-sorter’ (Martin 2010: 192) — call her Mona — is
5
Cf. Thau 2002: 230. Byrne (op. cit.), following Jackson, claims that even predicates like
‘old’ have a suitable non-comparative reading.
Chapter 4, Looks
83
entirely colour blind and sees only monochromatic degrees of luminosity
without any polychromatic colour. Consequently, Mona has none of the visual
phenomenology that we normally associate with seeing colours, or else she has
very different visual phenomenology corresponding to her monochromatic
experiences. Mona’s ability to discriminate luminance levels, however, is
sufficiently acute that she can sort objects into categories by their colour — red,
green, and so on — despite being unable to ‘see’ these colour in the conventional
sense. Red objects, for example, appear to Mona as a particular shade of grey,
or as having some unique surface reflectance property that differs from green
and other coloured objects. Thus, by stipulation, all of Mona’s talk and
behaviour concerning colour precisely matches those of a normally sighted
subject. Furthermore, since red-looking objects look to Mona the way that red
things normally look to her — that is to say, as a particular shade of grey — then
they can be said to comparatively ‘look red’ to her. There is a sense, however, in
which we want to say that nothing looks red to Mona, since she lacks the
relevant phenomenology. This sense, argues Jackson, corresponds to the
phenomenal use of ‘looks’.
Despite its intuitive appeal, one might be tempted to doubt the coherence of
the SORTER scenario. The term ‘red’ covers a wide variety of perceptible shades
across an even wider range of environmental conditions, and it is implausible
that an achromatic perceiver like Mona could differentiate unfamiliar red
objects from similarly varying shades of other hues based upon their luminance
and reflectance properties alone. Moreover, even if there were some uniquely
identifying property of red appearances other than their subjective colour, it
seems that such a perceiver would be tracking a different property of objects to
the one picked out by normally sighted perceivers by the predicate is red, i.e.
their colour. This is borne out by the counterfactual conditionals that hold for
Mona, which include monochromatic-illusory cases in which objects merely
appear to her to be red in virtue of possessing the relevant luminosity or
reflectance profile, but which would not appear red to a normally colour-sighted
perceiver.
Even if we grant the coherence of SORTER, however, there is a further
question concerning the connection between an object’s looking some particular
way and the phenomenal character of that experience. Like many other views,
representationalism typically holds that the properties of the objects of
experience are distinct from the properties of experience itself (cf. 3.3.3). Whilst
an object of experience can, for example, be square in the ordinary sense, an
experience of that object can only be described as ‘square’ insofar as it
represents something as being square. Thus, the experience is not itself square,
but rather exhibits the phenomenal correlate of squareness — call this square*.
Now, the connection between being square and square* is either contingent or
necessary. If it is contingent, then there must be some further explanation as to
why square* represents being square as opposed to some other property, such as
being red. But even if this connection were to hold necessarily, then there is still
a further question as to why an experience with a certain phenomenal character,
Chapter 4, Looks
84
i.e. square*, should attribute the property of being square to some external
object, given that, for all that has been said so far, the two are quite separate
and distinct properties.
Naïve Realists are able to avoid this objection by asserting the identity of
these two properties since, for them, to experience something to be square is
simply to be presented with its squareness. Sense-datum theorists make a similar
manoeuvre but at the level of mental objects. The question for the
representationalist, then, is to explain beyond mere causal covariance why a
particular quality of experience should be taken to represent a given objective
property.
To illustrate by way of an example, if (1) were analysed as
(1′)
The ball phenomenally looks red* to S
then would be unclear how a property of experience — namely, red* — can be
attributed to a physical object (the ball) since only experiences and not objects
can be red*. At this point, the representationalist has a number of options. First,
(1′) may be read as saying that S has an experience of phenomenal redness (i.e.
red*) that is, or appears to be, caused by the ball. Here, the connection between
red and red* is contingent since, in a different perceiver, the same object might
have caused an experience with a different phenomenal character, such as blue*
or pain. This would make the correlation between red and red* a mere accident
of grammar, since the only connection between the two properties is a
contingent causal one. Consequently, on this view the claim that
(10)
Things which are red look red
would, on the comparative use of ‘looks’, not be a necessary truth, but an
‘empirical generalization’ (Chisholm 1957: 51). Such a claim would be
equivalent to:
(10′)
Things which are red phenomenally look red* in k
where k is the relevant set of conditions for seeing red things.
Another option would be to say that S represents the ball as looking red by
having an experience of red*. That is, experiences with a red* phenomenal
character are intrinsically representational. The question of whether there are
any intrinsically representational phenomenal characters lies outside the scope
of this chapter. However, the suggestion seems to parallel at the level of
properties, rather than objects, the representationalist notion that one
experiences objects by representing them to be some particular way, and so is in
keeping with the general tenor of the representationalist’s account.
Finally, red* might be understood as a kind of narrow content that
supervenes upon the properties of experience, but not its external objects. In this
case, phenomenal looks provide a way of capturing the aspect of perceptual
Chapter 4, Looks
85
experience according to which things appear to be some particular way to the
perceiver without entailing that they actually are that way — surely the essence
of the notion of ‘looks’.
Although Jackson’s SORTER scenario does not provide us with any positive
characterisation of, for example, red*, his intention, along with that of
Chisholm (ibid.), seems to be to posit a use of ‘looks’ that is applicable only to
experiences that have a certain phenomenal character. Since Mona’s experience
lacks this character, nothing phenomenally looks red to her, despite red things
having the relevant comparative look. It remains to be shown, however, that
there is any connection between the proposed phenomenal use and our everyday
looks talk. I argue below (4.3) that such talk can be explained equally well, if
not better, without any reference to phenomenal looks. If this is correct, then
the phenomenal use of ‘looks’, in so far as such a use exists, may be a distinctly
philosophical one to which we only need appeal in order to describe various
apparently logical possibilities, such as SORTER or inverted spectrum scenarios.
Thus, even if there is a distinctly phenomenal use of ‘looks’, this in itself does
not necessarily support the truth of representationalism (4.2.6).6
A further account of ‘non-comparative’ looks offered by Byrne (2009: 444)
seeks to extend Jackson’s notion of phenomenal looks to include more complex
looks, such as looking old, cow-ish or Scandinavian. However, it does this by
combining multiple phenomenal looks involving colour, shape and distance
terms into more complex visual Gestalten which are recognisable on the basis of
the relevant phenomenally basic looks. This renders Byrne’s phenomenally nonbasic non-comparative looks a form of linguistic shorthand for more complex
combinations of phenomenally basic looks. As such, they are only as capable of
indexing representational perceptual contents as the phenomenally basic looks
that form their building blocks. I will therefore not give an extended treatment
of them here since any argument against the latter will count equally against the
former.
4.2.5. Non-comparative phenomenal looks
Having established the threefold distinction between evidential, comparative
and phenomenal looks, I will now evaluate the arguments offered by Jackson
and Brogaard in favour of the existence of non-comparative phenomenal looks.
Both Byrne and Schellenberg, amongst others, claim that a non-comparative use
of ‘looks’ may be used to ground the content of perceptual experience. Jackson
(1977) presents several arguments that appear to support this, claiming that the
phenomenal use is irreducible to either comparative or epistemic looks.
6
It is worth noting that the proposed equivalence between phenomenal and comparative
looks in this scenario is only supposed to hold between looks to a subject. That is, there is
nothing to say that looking red to Mona necessarily means the same as looking red to Mary
(Jackson 1982) or any other perceiver. This in turn suggests that the flaw in the comparative
analysis of phenomenal looks, if there is one, lies not in its comparative nature, but in its
subject specificity. This suggestion is taken up by the alternative analyses of ‘looks’ described
in 4.3, which primarily deal with objective looks, of which subject-specific looks turn out to
be a special case.
Chapter 4, Looks
86
However, as I argue below, these arguments are unconvincing, and the resulting
non-comparative looks too thin to support the kind of representational contents
that Byrne and others propose, as Byrne (2009: 444) himself admits.
Furthermore, the availability of alternative comparative analyses of phenomenal
looks removes the need to appeal to this notion in accounting for most, if not
all, of our everyday looks talk, thereby neutralising the force of Byrne and
Schellenberg’s arguments (4.3). Note that my intention in this section is not to
offer arguments against Jackson’s own sense-datum account of perceptual
experience, which he takes to be supported by the existence of phenomenal
looks. Rather, I aim to refute the use of Jackson and Brogaard’s arguments in
favour of the existence of phenomenal looks as a means of supporting
representationalism.
Jackson aims to establish that statements of the form:
(11)
x phenomenally looks F to S
are not systematically equivalent to statements of the form:
(12)
x comparatively looks the way things that are F normally look to
S in k
where k is a specification of normal, ideal, current or some other kind of
circumstances (Jackson 1977: 34). That is to say, Jackson rejects any
equivalence between the phenomenal and implicitly comparative (or indeed
evidential) uses of ‘looks’.7 Jackson’s objections to this equivalence are twofold.
First, he presents a number of apparent counterexamples which purport to show
that (11) may be true when (12) is false, or vice versa. Second, he argues that
there is no relevant set of circumstances, k, such that the alleged equivalence
always holds. I will address each of these objections in turn.
Jackson’s first objection to the equivalence between (11) and (12) is that
something can look F to S even when nothing is F. Such cases, he argues, render
(12) false despite (11) being true, since nothing can normally look the way to S
that things that are F do if there are no such things. For all we know, argues
Jackson, this may be the case in our own world with respect to certain narrowly
defined colour properties, which may never be instantiated despite the
appearance of being so, as in the case of Hume’s ‘missing shade of blue’ (ibid.
35).
Jackson anticipates the obvious response to this objection, which is to
replace (12) with the subjunctive conditional:
(13)
7
x comparatively looks the way things that are F would normally
look to S in k
Jackson also rejects accounting for non-comparative looks in term of the sensory states that
they typically evoke (ibid.). However, since this argument is of a similar form, I will not deal
with it separately here.
Chapter 4, Looks
87
Thus, even if no actual things are F, things can look the way that Fs would
(under k to S) were such object to exist. According to Jackson (1977: 35), the
problem with such an analysis is that
[T]o say that [x] looks red to me is to say something about how things
actually are, it is not to say anything about how things would be if the
world were different.
However, this response is based upon a misconception. How things would
appear to S were she, per impossibile, to experience some red object is a fact
about the actual world since it concerns the constitution and perceptual
capacities of S. That this may be explained in terms of a modal notion seems
beside the point.
To illustrate this, consider Jackson’s DEMON scenario. This involves ‘a
Cartesian evil demon who hates red things but tolerates non-red things looking
red on odd occasions’ (ibid.). The demon’s dislike of red things is such that he
prevents any such things from coming into existence. Furthermore, if even a
single red thing were to come into existence, the demon would destroy the entire
world along with everything in it. Consequently, nothing in this world can look
the way that red things normally do, as per (13), since if any red thing were to
come into existence then the entire world would cease to exist.
Call the demon-world w and the closest possible world in which there are
red things w′. Contrary to Jackson’s claim, the existence of an evil demon at w
does not make (13) false; only the absence of red things in w′ could do that. But
since w′ is the closest possible world in which there are red things, then (13)
cannot be false unless no possible world contains red things. Given that our own
world does contain red things, then (13) cannot fail to be true whenever (11)
obtains. Thus Jackson’s counterexample to the proposed equivalence fails.
The above response to Jackson might be considered uncharitable on the
basis that his objection to (13) is less about semantics than the very idea that
how things look in this world can be answerable to merely possible states of
affairs rather than what is actually the case. After all, in the demon world, w,
some things ‘look red’ to S despite there being nothing in that world that is red,
and so nothing that such looks may be compared with. However, this objection
is question-begging since it presupposes that we antecedently know what
phenomenal looks are — or at least what they are not — within the context of an
analysis of phenomenal looks. This suggestion cannot be given any weight
unless some concrete reason to reject it is provided. Unfortunately, beyond the
autobiographical statement that Jackson himself finds the view to be
implausible, he says little to back up this claim, and so we must reject this
objection for the time being.
A similar issue is raised by another scenario described by Jackson (1977:
35). This involves a subject — call him Frank — who is unusually sensitive to red
light such that he is dazzled by anything that is red. On occasion, however,
some non-red things — for example, white walls at sunset — look red to Frank,
although nothing which is red looks this way due to his dazzlement. To a large
Chapter 4, Looks
88
extent, this scenario relies for its apparent plausibility upon the value of k in the
above formulation. For example, though red things may dazzle Frank where k is
‘full sunlight’, the same objects would presumably look red to him under
suitably low conditions of illumination. If not, and no red objects look red to
Frank, as Jackson seems to suggest, then it is debatable to what extent anything
can be said to look red to Frank, since he is presumably incapable of perceiving
the relevant colour. At best, any looks that Frank might perceive under such
circumstances would be akin to a kind of optical illusion in which certain
conditions cause a phenomenally red experience in Frank. However, Jackson
gives us no reason to suppose that this experience would in any way correspond
to other perceivers’ experiences of red, i.e. red* (4.2.4), since it is entirely
unconnected with the perception of any red object. To simply assume such a
correspondence from the outset would be question-begging, since the existence
of such phenomenal looks is precisely the point at issue, rendering the argument
circular. Conversely, the comparative analysis of (11), i.e. (12), quite plausibly
suggests that the phenomenal character of Frank’s experience would not be
red*, but whatever property is shared by all the objects that look red to him —
call this F*. Since any object that looks F* to Frank also looks the way that
objects with the relevant property (whatever that might be) normally look in k,
the proposed equivalence between (11) and (13) continues to hold without
exception, and so Jackson’s scenario fails constitute a convincing
counterexample to the comparative analysis.
The problem with the above scenarios is that each attempts to divorce the
phenomenon of looking red from things being red. This may be plausible for
philosophers like Jackson who take there to be a distinctly phenomenal sense of
‘looks’ such that one can have an experience of phenomenal redness (red*) quite
independently of one’s capacity for seeing red things. However, this assumption
is of dubious coherence, and potentially objectionable to any theorist who
rejects the separation of phenomenal and comparative looks — or indeed the
separation between the perceived properties of objects and the phenomenal
character of experience. Thus the very plausibility of the alleged counterexamples that Jackson describes depends on the assumption that phenomenal
looks (e.g. looking red*) are distinct from comparative looks (looking like a red
thing) — the very claim they are being used to demonstrate — and so must be
rejected on grounds of circularity.
Having established that none of the above scenarios are wholly convincing,
this leads us to Jackson’s second objection concerning the specification of the
conditions k required by (12) or (13), which he puts as follows (Jackson 1977:
36, with appropriate modifications to the numbering):
In everyday chat about color, we take reasonably bright daylight as normal
circumstances, but this quite obviously will not do in (12). (11) does not
mean anything like
(14) x looks the way things that are F normally look to S in daylight.
Chapter 4, Looks
89
Jackson proceeds to consider several possible candidates for such ‘normal
circumstances’ and finds each of them wanting. His discussion, however,
conflates two different aspects of (12): (i) the specification of the relevant
comparison conditions, k, and (ii) the meaning of the term ‘normally’. Jackson
appears to take the value of k to be ‘normal circumstances’, which in most cases
(according to Jackson) means in full daylight. However, in Chisholm’s original
analysis of comparative looks, there is no suggestion that k must remain fixed
across all experiences, or that it should represent normal conditions. Rather, as
Chisholm (1957: 45–46) is at pains to point out, k is indexical and may vary
according to context. This is hardly surprising given the context-dependence of
other implicitly comparative expressions; e.g. ‘that animal is small [for an
elephant]’. Jackson’s assumption that k may generally be taken to be
‘reasonably bright daylight’ in everyday discourse, even if true, is therefore
irrelevant to the evaluation of (13) in the scenarios he describes. Indeed, the
circumstances under which (11) is true need not bear any obvious relation to the
corresponding value of k in (13).
Take the case of Frank. If this scenario is coherent, and I have argued that
there is reason to doubt this, an object’s looking red to Frank can be analysed
not in terms of (14), but as
(15)
x looks the way things that are F [would] normally look to Frank
in conditions of low illumination
where F is ‘red’. As I suggested above, if (15) is never true, then it would be
questionable to what extent Frank is capable of seeing red at all. Conversely, if
(15) is true under some set of conditions k, then (11) will be equivalent to (15)
irrespective of the conditions under which non-red things look red to Frank.
Thus k specifies the conditions under which Frank sees red things, not things
that merely appear to be red.
Allowing k to vary according to context removes much of the force of
Jackson’s objection, but raises the further question of whether there is any
principled, non-arbitrary way of determining the relevant conditions. After all, if
we are free to choose whichever value of k makes (12) or (13) true in each
individual case, then the above response begins to appear dangerously ad hoc. It
is plausible, however, that for many subjects most of the time, (11) does mean
the same as (13) when k specifies present conditions; i.e. the conditions under
which x is seen. The divergence of Jackson’s and other scenarios from this norm
is explicable purely in terms of the unusual nature of the subject or situation
concerned, which thereby cause a different set of conditions to apply. Thus, for
all that Jackson has said, there appears to be no reason to suppose that
phenomenal looks reports cannot be analysed in comparative terms, calling into
question the need for non-comparative phenomenal looks.
Chapter 4, Looks
90
4.2.6. Phenomenal content
I have argued that the evidential, comparative and phenomenal uses of looks,
far from being mutually exclusive, represent orthogonal aspects of the analysis
of looks reports (cf. Brogaard forthcoming: 21). As such, Jackson, Brogaard and
Byrne’s arguments for the primacy and irreducibility of phenomenal looks may
be seen as attempts to establish the existence of non-evidential non-comparative
analyses of phenomenal looks. However, each involves a number of
questionable assumptions, most notably Jackson (1977), who assumes that
relevant conditions for perceiving appearances must remain fixed across all uses
of ‘looks’ as opposed to varying with context — a natural consequence of any
comparative analysis of looks. Indeed, the only one of the above arguments that
offers any support to the proposed non-comparative analysis is Jackson’s
SORTER scenario, which is itself of dubious coherence, raising serious issues
concerning how phenomenal properties relate to the objects of experience. As
noted above, this not a problem for Jackson himself, who is a sense datum
theorist, but it presents a serious obstacle to those representationalists who wish
to adopt Jackson’s account of looks whilst rejecting his conclusion, i.e. the sense
datum theory. Even if we grant the coherence of non-comparative phenomenal
looks reports, however, it remains to be seen whether such a use is capable of
indexing objective representational content. If it is not, as I argue is the case,
then Jackson’s argument offers no support to the representationalist.
In order to index perceptual content, in Travis’s terms, a perceptual
appearance must make exactly one such content recognisable to the subject, as
per Looks-indexing (2.2.5). This is the case because it is a necessary condition
for autorepresentation (i.e. representing-to) that the content of any given
experience be recognisable to the subject (2.2.4). Furthermore, in order to
support a representationalist account of illusion, such contents must also be
capable of misrepresentation — that is, of being false or non-veridical — this
being one of the supposed benefits of representationalism (2.3.1). Now, if the
proposed non-comparative phenomenal use of ‘looks’ relates to the presence of
a given phenomenal character in experience — for example, red* — as opposed
to a specific property of an object — i.e. being red — then neither of these
requirements can be met. To see why this is the case, consider the following
proposition as it applies to Mona, the super-achromatic colour sorter (4.2.4):
(16)
The ball phenomenally looks red* to Mona
As it stands, (16) is false since, ex hypothesi, nothing looks phenomenally red to
Mona as her experience lacks the relevant phenomenal character. This is true
despite the corresponding comparative reading
(17)
The ball comparatively looks the way that red things [would] do
to Mona under relevant circumstances
Chapter 4, Looks
91
being true, and regardless of whether the ball in question is red or not.
Conversely, for (16) to be true, Mona would need to experience red*, in which
case the content would be veridical. Furthermore, (16) is true of ordinary
perceivers in all cases where the ball appears red, regardless of whether it is red
or not. Thus, the only content that (16) can make recognisable to the subject on
the basis of phenomenal looks-indexing is not that the ball is red, but that it
(merely) looks red; i.e. that the subject’s experience has a particular phenomenal
character, just as Jackson suggests. There are two problems with this view from
the representationalist’s perspective.
First, the proposed content attributes a property to the subject’s experience,
not to the objects of experience, or in the case of (16), the ball. If the
representational contents of experience were of this kind, then they would fail to
predicate anything of external objects since it is only the phenomenal character
of the experience that would be represented. This is an even weaker notion of
representation than Byrne’s ‘relatively thin’ content, since it does not yield the
content that the ball is red, but rather that the ball (merely) looks red. On this
view, what makes (16) true would be something like:
(18)
Mona’s experiencing the ball involves phenomenal redness [red*]
Alternatively, if — by the application of an externalist theory of perceptual
content, for example — (16) is supposed to represent the ball’s being red, then it
remains to be explained how (16) could make that content recognisable given
that phenomenal looks are equivocal between the ball’s being red and its
(merely) looking red, not to mention the multitude of ways in which something
can look to be red; e.g. by shining a red light on it.
Note that simply insisting, perhaps on the basis of an externalist theory of
content, that Mona’s experience represents being red and not her experience
involving red* does not constitute an adequate response to the objection. The
present argument concerns the content that Mona’s experience makes
recognisable to her, rather than what content her experience can be described as
possessing from some independent third-personal point of view.8 If the
recognisability of experiential content is based upon phenomenal looks,
themselves features of experience, then it remains entirely opaque how this
could make content that is about the objects of experience available to the
subject. At best, this approach yields phenomenal content (3.2.3) whilst failing
to say anything about the objective world outside of experience.
Second, and equally problematic for the representationalist, is that the
resulting content is unable to support the standard account of perceptual
illusion. Given that a subject’s experience involves phenomenal redness (i.e.
red*) if and only if something looks phenomenally red to them, then (16) can
only falsely represent in cases where the ball is represented as phenomenally
looking red, i.e. (16), but where Mona’s experience lacks the relevant
8
Appeals to externalist accounts of self-knowledge, e.g. Burge (1979), on the other hand,
insofar as these bear upon Looks-indexing, are relevant, and are dealt with in chapter 6.
Chapter 4, Looks
92
phenomenal character, i.e. (18), is false. This form of illusion, if indeed it is
coherent, is not perceptual but cognitive, since it describes a case in which the
subject is mistaken about the nature of her own experience.9 Conversely, if the
possibility of such illusions is rejected, as I believe it should be, then the
resulting content, i.e. (16), can never be said to misrepresent, since the subject
only ever represents something as phenomenally looking red* when they are
experiencing a red phenomenal character, and so the content is necessarily
veridical. Thus the content of (16) becomes its own truthmaker. Note that even
if the possibility of misrepresentation is allowed for, it is not of a kind that
would enable the representationalist to explain the existence of perceptual
illusions, since at no point is any object represented as being red; merely that it
looks so.
Putting both of these points together, and in the absence of any convincing
counter-argument, we can conclude that non-comparative phenomenal looks, as
described, are unfit to index the representational content of perceptual
experience. That is not to say that we cannot describe the phenomenal character
of experience in representational terms. Rather, in providing a response to
Travis’s argument from looks, such representational content (a) can only make
the phenomenal character of experience recognisable and not the properties of
external objects, and (b) fails to provide an adequate explanation of perceptual
illusion. Indeed, this is precisely the notion of phenomenal content that was
rejected in the previous chapter in connection with Byrne’s account of noncomparative looks (3.2.3), and which Byrne (2009: 443) himself admits to be
unsuitable for indexing the content of perceptual experiences. Thus, even if
SORTER is taken to justify the existence of some form of non-comparative
phenomenal looks — and it is far from clear that it should be — such looks
cannot index the kind of content that many representationalists take perceptual
experiences to have. This is not a problem for Jackson (1977: ch. 3), who argues
that phenomenal looks justify the existence of sense data rather than
representational content. It is, however, a problem for Byrne, Schellenberg and
any other representationalist who wishes to ground their notion of prepresentational content upon it.
4.3. Parsimony about Appearances
Having established the difficulties with both Jackson’s phenomenal looks and
Byrne’s ‘non-comparative’ looks for indexing representational content, I will
now to turn to an alternative analysis of ‘looks’. According to this essentially
comparative account, looks statements are grounded in primitive visually
relevant similarities between objects, and not in additional appearance or
phenomenal properties. It is therefore a deflationary analysis as it holds
comparative, and not phenomenal, looks to be explanatorily more basic.
Furthermore, this account treats both objective and subject-relative looks — i.e.
9
This is not the situation of Mona, to whom objects do not even phenomenally look red
(compare: a subject representing the ball as being red when it is not).
Chapter 4, Looks
93
‘looks F’ and ‘looks F to S’, respectively — as points on a single continuum
(4.3.4). It is therefore not only more parsimonious (4.3.1), but also more
powerful than Byrne and Jackson’s non-comparative accounts (4.3.2). Finally, I
consider a potentially objectionable feature of the parsimonious account, along
with how this might be resolved (4.3.3).
Since this deflationary analysis is essentially comparative, it cannot support
the indexing of representational content by looks in the way that
representationism requires. Instead, it remains strictly neutral on the question of
whether experience has representational content. The availability of such an
account, however, suggests that representationalists should not seek to motivate
their view via the semantics or metaphysics of ‘looks’ since, in the absence of a
stronger argument than those considered above, there is no reason to suppose
that a non-comparative account of looks must apply. Rather, one could equally
adopt a comparative analysis of ‘looks’, thereby, in conjunction with Travis’s
argument from looks, neutralising the force of the representationalist’s
argument.
4.3.1. A deflationary analysis
One of the most thorough and detailed treatments of comparative looks in the
recent literature comes from M. G. F. Martin, who claims that many of the
cases of what Jackson takes to be paradigmatically ‘phenomenal talk’ are
implicitly comparative (Martin 2010: 162). The central claim of his account is
captured by the following principle, which Martin (ibid. 197) labels Parsimony:
Parsimony: looks statements are made true just by properties of objects
that we need to appeal to in order to explain the truth of
sentences that are not explicitly looks sentences.
According to Parsimony, there is no need to posit an additional phenomenal use
of ‘looks’ or appearance properties to explain the meaning of looks statements.
Instead, such statements are made true or false by an object’s ordinary
appearance-independent properties, such as their shape or colour. In effect, the
proponent of Parsimony claims that the role that was played by phenomenal
properties above (4.2.4) can equally well be played by those appearanceindependent properties that give rise to the relevant looks; i.e. the intrinsic
properties of shape, size, colour, and so on, that we ordinarily take physical
objects to have independently of how they look. Consequently, there is no need
to appeal to additional appearance properties or non-comparative looks in order
to analyse many — or indeed most — everyday looks statements.
Although Martin’s account of the semantics of ‘looks’ is characteristically
subtle and complex, his central thesis is relatively straightforward and may be
divided into two lemmas. First, instead of an object’s look (or looks) being some
property distinct from its intrinsic properties, it is simply the product of its
visually relevant similarities to other objects. For example, the ball’s looking red
is just a matter of one object (the ball) bearing a certain visually relevant
Chapter 4, Looks
94
similarity — i.e. its apparent colour — to the characteristic appearance of red
things, regardless of whether it is red or not. On this view, looks are merely a
way of quantifying over similarities between the multitude of ways that objects
can appear to us.
The second lemma is that what grounds the explanation of these visually
relevant similarities is the mind-independent properties of the objects themselves
— shape, colour, and so on — along with our perceptual sensitivities to them.
What makes a property visually relevant is the way that it perceptually affects us
through the sensory modality of vision, with the same applying mutatis
mutandis for each of the other senses. Thus it is us and our sensory faculties that
determine the similarities that we find visually salient, rather than anything
peculiar to the objects of perception and their properties. Nevertheless, the
objective properties of those objects are what ground such similarity relations.
Consequently, according to Parsimony, our sensory faculties are capable of
perceiving intrinsic properties of objects, including their shape, colour, and so
on, along with similarities between these properties, without the need to posit
additional appearance-dependent properties.
According to Parsimony, then, sentences like (1) — ‘The ball looks red [to
S]’ — means that a certain object (the ball) visually strikes S in a way that is
characteristic of red things relative to a certain set of contextually defined
conditions. Things ‘look red’ because they strike us in the way that is
characteristic of objects that are red; i.e. it is the way that red things (typically,
under circumstances that are conducive to perception) look. Whilst this requires
that there be various ways in which things can visually strike us, i.e.
phenomenal characters, these are not looks per se. Rather, it is the similarities
between objects that ground the truth of looks statements. Since the account is
based upon similarities, this yields the following comparative analysis of
sentences like (1):
(19)
∃s (has(x, s) ∧ (look(s) ∧ φ(s)))
where x is some physical object (e.g. the ball) and φ is a function that identifies
the relevant ‘way of looking’, such as looking red (Martin 2010: 172). This also
leaves open what kind of entities s and φ pick out, each of which I address in
turn below.
Taking the function first, as we have already established, φ cannot itself be
‘red’ since (1) does not mean that the ball is red, or that the relevant look is red
(looks do not themselves have colours; they are of colours), merely that it
appears that way. According to Jackson’s phenomenal account of looks, φ
would take the value red*, or more generally ψ*, where this term identifies the
phenomenal counterpart of ψ, the precise method of establishing which is
unclear. Instead of specifying ways of looking in terms of some property,
however — phenomenal or otherwise — Martin takes φ to be a similarity
function, or SIM for short. This function takes a predicate (‘is red’) and a look
(s), and obtains if and only if the latter is relevantly similar to those objects that
Chapter 4, Looks
95
satisfy the predicate. For example, if s is a reddish look, then SIM(‘is red’, s) will
be true. An additional getting-the-characteristic function, or C for short, is then
required to map what Martin calls the ‘complement adjectival phrase’ (ibid.) of
the looks predicate — in this case ‘red‘ — to the property of objects that is
characteristic of that look under circumstances k. Thus, C(red, look, k) would
return the property that is characteristic of things that look red under the
relevant circumstances — namely being red.10
Putting all the above elements together yields the following analysis:
(20)
∃s (has(x, s) ∧ (look(s) ∧ SIM(C(ψ, look, k), s)))
Thus (1) may be analysed as x (the ball) having a look that is relevantly similar
to the look of things that are red (this being the property characteristic of things
that have this look) under circumstances k, where ψ is the adjectival-phrase
complement of ‘looks’ in (1), i.e. ‘red’. The apparent complexity of the analysis
masks the underlying simplicity of the sense in which things that look red look
like (a comparative notion) the red things, thus corresponding to Travis’s notion
of a visible look (2.3.2).
Before moving on to examine the consequences of Parsimony, it is
important to note that (20) is intended as a logical, and not psychological,
analysis of the semantics of ‘looks’. There is no suggestion that the process
outlined above parallels some sequence of psychological events that we go
through when evaluating looks statements, though in some cases it may do.
Rather, it is intended to explain the underlying semantic structure of such
statements in a way that reveals their essentially comparative nature, as distinct
from the phenomenal analysis offered by Jackson and others (4.2).
What Martin’s deflationary analysis shows is that both similarity
judgements, in the form of SIM, and the properties that are characteristic of a
given look, C, are central to the meaning of looks statements. Thus, it is not that
there is any intrinsic connection between an object’s looking red and being red,
but rather that being red is (under certain circumstances) a visually relevant
characteristic of things that look red. This need not render judgements like (1)
inferential or epistemically indirect since subjects may become expert at
recognising such properties without the need to perform any such inference.
However, such linguistic competence only becomes possible once one knows
what red things ‘look like’ in the sense that one associates the appearanceindependent property of being red with the phenomenal character that one
experiences upon seeing red objects.11
10
Note that in many everyday cases the correspondence need not be trivial. A reddish look
may, under certain lighting conditions, be characteristic of white objects, or someone who
‘looks pregnant’ may have the property of being a particular shape (Martin 2010: 173).
11
As before, if the metaphysics of colour are thought to be problematic in this regard,
consider the analogous case of shape.
Chapter 4, Looks
96
4.3.2. What’s in a look?
In certain respects, Martin’s deflationary analysis of ‘looks’ is not dissimilar to
Byrne’s ‘non-comparative’ looks (4.2.5). Both make reference to distinct
properties, or sets of properties, that relate to a given look. However, whereas
Byrne claims that such a set of phenomenal characters just is the look, according
to Parsimony, looks are constituted by similarities between properties, and not
those properties themselves. Thus, according to the parsimonious account, such
looks are implicitly comparative. This does not yet explain, however, what kind
of entities s and the results of the C function in the above analysis are. Indeed, it
is possible that these are precisely the kind of phenomenal looks that Jackson
and Byrne posit. On such a view, the visually salient similarities that SIM picks
out would be between the bundle of phenomenal properties that the subject is
experiencing and the bundle of phenomenal properties returned by C that is
characteristic of the look in question; e.g. looking red. On this interpretation,
Martin’s analysis would entail the existence of non-comparative looks. (I argue
below that this view is incorrect, and so there is no such entailment.) In the case
of complex looks, the main point of difference between Martin and Byrne is that
to ‘look red’ is not to non-comparatively look red*, but to look relevantly
similar to the characteristic look of red things, where a ‘look’ is something like a
Kantian sensory manifold. This difference may seem somewhat obscure, but
there is a further point that is central to Martin’s account.
In contrast to Jackson, Martin argues that the sentence
(21)
The spread looks splendid12
is ambiguous between the following readings:
(21′)
There is a splendid look that the meal has
(21″) The meal has a look that is relevantly similar to the look of
splendid things
Or, to put this more formally (Martin 2010: 185):
(21*) ∃s (has(the meal, s) ∧ (look(s) ∧ splendid(s)))
(21**) ∃s (has(the meal, s) ∧ (look(s) ∧ SIM(C(splendid, look, k), s)))
According to (21′), it is the look of the meal that is splendid. In (21″), the meal
itself has a look that is relevantly similar to that of meals that are splendid.
Conversely, however, Martin argues that comparable sentences involving shape
or colour terms contain no such ambiguity. That is, (13) does not admit of the
readings
12
Martin uses the term ‘spread’ to mean a large and impressive meal.
Chapter 4, Looks
(13′)
97
There is a red look that the ball has
(13″) The ball has a look that is relevantly similar to the look of red
things
Instead, according to Parsimony, the two readings effectively collapse into (13″)
since to have a red look just is to have a look which is relevantly similar to the
look of red things. Unlike ‘splendid’, which conceivably applies to both looks
and objects, a look is not the sort of thing that can itself be red; in (13′), for
example. Rather, it is the look of something red.13
Parsimony entails that quantification over the looks of objects does not
require those objects to possess any particular properties other than looking that
way; i.e. there need not be any red thing that is seen in order for something to
look red. This runs counter to Jackson’s account, which posits the existence of
precisely such entities — namely, sense data. But if looks are not phenomenal
properties of experiences or objects, then what are they? Martin’s answer is at
once simple and surprising: they are the ordinary appearance-independent
properties of objects.
The second aspect of Parsimony, then, is that it identifies the phenomenal
character of looks with the appearance-independent properties of objects: being
red, being square, and so on. On this view, the properties of which one is aware
in perceptual experience are not the phenomenal correlates of an object’s
appearance-independent properties, but those very properties themselves. When
one sees the red ball, for example, one experiences or is aware of the redness of
that very object, and not merely some phenomenal aspect of an experience
caused by the object. Thus, instead of a duality between the world of ‘external’
objects and their appearance-independent properties and an ‘internal’ world of
experience and its phenomenal properties, Parsimony entails — at least in the
case of colour, shape and distance perception — that we experience the
appearance-independent properties of external objects themselves. Furthermore,
in the case of objects that merely look red, it explains why there need be no red
thing present. In such cases, the red look of an object is explained by its
qualitative similarity, as determined by the similarity function SIM, to things
that are red. This may, though need not, cause one to mistakenly judge on the
basis of this look that something is red when it is not. Thus in the case of mere
looks and perceptual illusions, nothing in the subject’s experience need
instantiate the properties that are perceived. Rather, one is confronted with an
appearance that is relevantly similar to the characteristic look of some other
object, causing one’s thoughts and actions to be guided accordingly.
13
More accurately, it has a look that is relevantly similar to the look of red things, since the
inference from:
∃x (experiencing(S, x) ∧ looksred(x))
to
∃x (experiencing(S, x) ∧ red(x))
is fallacious.
Chapter 4, Looks
98
In summary, then, Jackson identifies a class of non-comparative or
‘phenomenal’ looks that he takes to underpin the meaning of comparative looks
statements. This, along with the wider range of non-comparative looks —
including Byrne’s Gestalten, which arise from combining multiple phenomenal
looks — may indeed be used to ground a limited notion of representational
content. However, the resulting content lacks the objective import that prepresentation requires. Martin, on the other hand, argues that although noncomparative uses of ‘looks’ may exist, they are not themselves looks. Rather,
looks arise from salient similarities between the experiences of objects that
possess the relevant property, and in some cases those that don’t.
Moreover, one can become visually aware of an object’s properties through
direct perceptual experience of them such that the experienced properties of the
object are themselves constituents of that experience. This removes the need to
posit additional entities or properties that correspond to an object’s looks,
instead locating looks in the similarities between those properties of objects in
the world that subjects find visually salient. As such, to experience an F-look
need not be to experience F-ness per se, but rather receives the disjunctive
analysis of either experiencing F-ness or some other property, the experience of
which bears visually relevant similarities to it. Thus, a stick that is halfimmersed in water is not itself bent, nor does it have a ‘bent look’ in the sense of
possessing an appearance property of ‘bentness’. It merely looks, in certain
respects and from particular angles, like bent sticks do. Similarly, a white wall
illuminated by red light does not present the subject with the look of a red wall
per se, but rather with a look that one might, under certain circumstances,
mistake for a red wall due to the similarity between the two experiences, and so
on.14 Such an explanation of perceptual illusion is consistent with Travis’s
argument from misleading appearances (2.3.1).
4.3.3. The problem of conflicting appearances
Perhaps the most serious problem for the parsimonious account is one that
Martin (2010: 219–22) himself considers. It is not unusual in philosophical
discussions of perception to appeal to cases in which the appearance of an
object is altered whilst its intrinsic properties remain unchanged. Hume
famously used just such a case to argue that the direct objects of perception
cannot be external physical objects, which ‘suffer no alteration’ despite changes
in their appearances, but must instead be ideas or impressions in the mind
(Hume 1999: XII.i, 201). This was later refuted by Reid (1997: VI.xxii, 186),
whose response has since been taken up by Snowdon (1990) and others.
However, this problem of conflicting or incompatible appearances also presents
a problem for Parsimony. If judgements about the sensible properties of objects
in the epistemically ‘good’ case are just judgements about their intrinsic
properties, then how are we to explain cases in which an object’s sensible
appearances change whilst its intrinsic properties do not? A paradigm example
14
See 4.3.4 for further qualification on this point.
Chapter 4, Looks
99
of such a case is that of a stick that when partially immersed in water appears
bent. That is to say, in certain respects from the relevant angle, it looks
relevantly like a stick that is bent would do under normal circumstances, i.e.
when it is not immersed in water (Martin op. cit.).
There are two obvious responses to this problem in connection with
Parsimony, both of which Martin rejects. The first is to claim that an object’s
appearance depends not only upon its intrinsic properties, but also its relational
properties. The appearance of the stick, for example, varies when it is partially
immersed in water due to its acquiring the relational property of being
immersed in water. This would be entirely in keeping with the principle of
Parsimony, which claims only that how an object looks is explained by
reference to its appearance-independent properties, and not what kinds of
properties these are. We can therefore identify two variants of the principle
according to which types of properties are admissible in explaining looks
statements: (i) Relational Parsimony, according to which looks statements are
made true just by those properties of objects and their relations to objects that
we need to appeal to in order to explain the truth of sentences that are not
explicitly looks sentences, and (ii) Non-Relational Parsimony: looks statements
are made true just by the non-relational properties of objects that we need to
appeal to in order to explain the truth of sentences that are not explicitly looks
sentences.
The problem with Relational Parsimony from Martin’s (2010: 220) point of
view is that:
[T]he ways of looking that the stick has just are among its basic visible
properties, most saliently its length and shape, and potentially its surface
colour. These simply do not change when it is placed half in the water
Martin does not give any particular argument for this assumption, though he
clearly aims to give the simplest possible account of our perceptual awareness of
objects and their properties. Admitting that relational properties have an effect
upon appearances would, perhaps, weaken the sense in which we can be said to
be directly aware of the non-relational properties of objects, since various
combinations of relational and non-relational properties can combine to
produce the same or similar appearances. Such combinations, however, admit of
a disjunctive analysis in a similar manner to disjunctive accounts of perception
and perceptual illusion. On this view, the ‘bent’ appearance of the stick would
be explained in terms of the distorting effects of the water in which it is
immersed, the refractive properties of which combine with the shape of the stick
to yield the resulting look. Thus, upon seeing a stick that is immersed in water,
we veridically see the properties of the both stick and water combined, rather
than just the shape of the stick. Presumably, however, even in the normal case,
minute distortions from the air or other intervening medium would mean that
we rarely, if ever, saw the non-appearance properties of objects simpliciter. The
resulting account is therefore unattractive to Naïve Realists, like Martin, who
Chapter 4, Looks
100
wish to claim that we have direct, and therefore unmediated, awareness of
external objects in the normal case.
An alternative response would be to claim that looks statements are
‘principally reports of our psychological states’ (Martin 2010: 220). This view
explains how objects can change in appearance without changing their nonrelational properties in terms of the subject having entering a different
psychological state. However, this would seem to imply that objects do not have
any particular look when they are not being experienced, which is contrary to
the way we use ‘looks’. This drawback could perhaps be avoided by a move
similar to the one that Jackson makes in the case of phenomenal looks, with
each look corresponding to what subjects would experience were they to be
confronted with that object (4.2.5). However, this again involves identifying
looks with the psychological responses of subjects, thereby weakening the sense
in which perceptual experience can be said to deliver direct awareness or
knowledge of external properties, and so is similarly inhospitable to Naïve
Realism.
Martin’s own response to this challenge is to claim that the appearances of
objects do not, despite evidence to the contrary, change. Indeed, he observes
that ‘[i]f we endorse Parsimony, then we must reject the claim that the stick has
changed in appearance’ (Martin 2010: 220). Instead, according to Martin, what
changes is the appropriateness of asserting that the object has certain
appearances according to the ‘perspective of comparison’ that one adopts (ibid.
221). On this view, the stick has both the property of ‘looking straight’ and
‘looking bent’, but one is only in a position to assert the former upon seeing the
stick out of water, and the latter when it is partially immersed. The intrinsic
properties of the stick that ground both appearances, however, remain fixed
regardless of whether it is immersed or not. This position is highly
counterintuitive since we can, and frequently do, say that the appearances of
objects change. However, it follows from Parsimony that if an object’s
appearance-independent properties do not alter, its appearance — or
appearances — must remain similarly fixed. In a sense, then, Parsimony entails
that objects possess all of the objective appearances that they can possibly have,
irrespective of which of these is currently available to observers at any given
time. What makes it the case that things appear one way rather than another is
instead the perspective or psychological state of the observer.
This unpalatable consequence of Parsimony, however, may be avoided since
the principle does not require that there be a simple one-to-one correspondence
between perceptual appearances and the appearance-independent properties of
objects. Instead, appearances may be realised by many different physical
properties with the similarities between them being explained by the
constitution and perceptual sensitivities of the subject. The ‘bent’ appearance of
a stick in water, for example, need not be explained in terms of a single physical
property that explains both the appearance of a stick in water and one that is
genuinely bent. That we identify both appearances as forms of ‘bentness’ is not
explained by sameness of truthmakers, but by their subjective — or rather
Chapter 4, Looks
101
intersubjective — similarity to paradigm cases of bent objects. Provided that such
similarities are grounded in the relevant appearance-independent properties,
which may include the perceptual constitution of the subject, Parsimony is
maintained. In most paradigm cases of perception, however, such properties
typically will be just the shape, colour, distance properties, and on so, of
external objects. However, in cases of illusion or subjective similarity (e.g. the
submerged stick), this need not be the case.
A modified version of Martin’s view may therefore be constructed in which
changes of appearance are explained by the tendency of the human perceptual
system to group appearances or looks together, based around certain paradigm
cases. What an object looks like — a comparative notion — is explained in terms
of its visual similarities to other objects, which in turn instantiate those
properties of which they are paradigmatic. Precisely which objects constitute the
relevant paradigms will be a partly empirical matter upon which the Parsimony
theorist need not take a view. Indeed, it will not matter whether the relevant
cases differ between times and subjects provided that the same sets of
similarities are maintained. In that respect, a genuinely bent stick is as good an
exemplar of bentness as, for example, a banana.15 Thus, it is no objection to
Parsimony that the resulting looks are equivocal, since this is an inevitable
consequence of any comparative analysis of ‘looks’. Rather, it suggests that
what determines an object’s appearance is its similarities to other objects. This
in turn may be explained by reference to both the object’s and the perceiver’s
non-appearance properties, as Parsimony suggests. That we can describe the
presence of a particular appearance in terms of a potentially complex
disjunction of physical properties is neither here nor there. Indeed, the problem
with doing so will be precisely the representationalist’s difficulty in attempting
to index p-representational content by visible looks (2.3.2). Provided that we
identify looks with sets of similarities rather than looks-independent properties,
however, this issue need not arise.
4.3.4. Consequences of Parsimony
Thus far, we have only considered what might be termed objective looks; that is,
the looks that objects have to any suitably perceptual observers under certain
narrowly defined conditions. But Jackson and Byrne’s non-comparative
accounts of looks are relative to individual subjects; i.e. subjective looks. How
are ‘looks to S’ to be accounted for in the parsimonious account? In fact,
Parsimony offers an elegant way of unifying these two varieties of looks since
subject-relativity is already built into the parsimonious analysis in the form of
the similarity function (4.3.1). Since the experiences that various subjects find
relevantly similar may differ from one individual to the next, subject-relative
looks are simply what constitutes a visually relevant similarity for a particular
subject. That is, sentences like (1) — ‘The ball looks red to S’ — exhibit an
identical logical structure to (13), as formalised by (20), but employ a subject-
15
Cf. Wittgenstein on the grammar of sensation terms (1953: §293).
Chapter 4, Looks
102
specific, as opposed to global, similarity function. We can formalise this as
follows:
(22)
∃y (has(x, y) ∧ (look(y) ∧ SIMS(C(φ, look, k), y)))
where the subscripted similarity function SIMS captures those similarities that
are visually relevant to subject S.
An objective look, on the other hand, involves a set of similarities that all
sufficiently acute perceivers have in common. Thus, for something to objectively
look red is for the subject’s perceptual experience of an object to exhibit the
same similarities to their experiences of red things as other subjects’ experiences
do to their experiences of red things. This does not entail that one subject’s
experience of an object or property need be subjectively similar to another’s
experience of the same object or property — something that would in any case
be impossible to determine. Rather, the intersubjectivity applies to what the
subjects’ similarity functions have in common; i.e. the kinds of experiences that
the subjects find to be visually similar. The mapping from subjective to objective
looks thus reflects the similarities that all normally perceiving subjects share. To
that extent, the global similarity function SIM may be thought of as a subset (or
in some cases, a superset) of SIMS. It is therefore a feature of Parsimony that it
takes both subjective and objective looks to share the same logical structure.
Another important aspect of Parsimony is the role it accords to a subject’s
prior knowledge. In fact, two distinctive kinds of knowledge are involved: (i)
knowledge of the characteristics of each given type of look, i.e. getting-thecharacteristics, or C (4.3.1); and (ii) knowing which experiences exhibit relevant
similarities to those characteristics, i.e. SIMS. Such ‘knowledge’ need not be
explicit, but may correspond to an innate capacity — for example, to perceive
certain experiences as being inherently similar — or be acquired through
experience. Children below the age of five, for example, find it difficult to
identify the colours of unfamiliar objects despite being familiar with the relevant
colour terms in other contexts — they have no difficulty in naming a random
selection of colours, for example (Bornstein 1985). This suggests that such
associations, or the ability to reflect upon the sensory experiences that support
them, may be acquired.
Parsimony accords a role to this type of knowledge not only in acquiring the
ability to name colours according to their appearance, but in recognising
relevant similarities between different perceptual experiences. Once learned, this
knowledge becomes implicit such that we can identify a red-looking object as
‘red’ without any conscious thought or inference even though this term covers a
wide range of actual and mere appearances — for example, under different
lighting conditions. Conversely, a perceiver’s conception of green and blue may
differ from other perceivers with similar levels of acuity not only because they
differ in their phenomenal character (they may not) but because one finds
certain pairs of colours to be more similar than others on the basis of one’s prior
experience and training. Some subjects, for example (including the present
Chapter 4, Looks
103
author), may find certain shades of bluish-green to be more subjectively similar
to paradigm examples of blue than do those who class them as ‘green’. This
need not be merely a matter of using the terms ‘blue’ and ‘green’ in slightly
different ways. To such a perceiver, these colours may really more closely
resemble blue than green. Thus, the implicit knowledge of looks plays a central
role in the parsimonious analysis (cf. 6.4.2).
Martin (2010) primarily presents a semantic analysis of looks statements.
However, one might equally be interested in the psychological analysis of looks.
What makes something look the way it does, and in what sense can this be said
to be similar or identical to the looks of other things? As noted above, Martin
does not attempt to provide such an analysis, but his account is suggestive of the
basic outline for one. The look of an object is determined by the similarities that
a subject is able to draw between experiences — in particular, similarities
between this and one’s prior experience of paradigm examples of particular
colours, shapes, and other visible properties. Subjects are able to move from
looks to classes of experiences; for example, in being able to group together
objects that look the same, or similar. Conversely, in moving from an experience
to a look, one is able to recognise the property that is paradigmatic for that
look — being red, being square, and so on, as encoded by the function C.
Consequently, a subject’s grasp of looks runs both from a given looksexpression to a class of experiences, and from a given experience to the
corresponding looks-expression. Both of these abilities are therefore necessary
for the understanding and communication of looks statements. Whether these
constitute two forms of knowledge or two different uses of the same knowledge
is an empirical question to be determined by the presence or absence of the
corresponding abilities in actual subjects.
The final aspect of the deflationary account of looks that I wish to consider
is what, if anything, it tells us about the various uses of ‘looks’. Whilst Martin
talks about ‘comparative’ and ‘non-comparative’ looks in the sense used by
Chisholm, Jackson and Byrne (4.2.2), this terminology is potentially misleading.
What Parsimony shows is arguably not that there are several different uses of
‘looks’, but that there are (at least) two different uses of looks statements.16 A
sentence like (1) can be used (a) comparatively, as in the ball looks the way that
red things do; and (b) phenomenally, as in the ball has a red* look. This does
not reflect any difference in the meaning of ‘looks’, but rather in the intention
that the speaker wishes to communicate. In the first case, the speaker is drawing
an implicit comparison between her experience of the ball, which seems to be
red (even though it may not be), and the characteristic appearance of red things.
The second case draws attention to the phenomenology of the subject’s
experience, which has a red* phenomenal character. According to Parsimony,
both cases employ the same sense of looks, which is implicitly comparative.
The same is true of the evidential and epistemic uses of ‘looks’. When (1) is
offered as evidence of the proposition that the ball is red, the intended sense of
‘looks’ is also implicitly comparative. The only difference from (i) is that the
16
This distinction was pointed out to me by Mark Eli Kalderon.
Chapter 4, Looks
104
speaker intends her audience to take this as evidence for, or justification of, the
latter proposition. Similarly, sentences like
(23)
It looks like rain
may be taken to mean it looks as if it is going to rain or those clouds look like
rainclouds do. However, even in the former case, it is precisely because the
clouds look like rainclouds that one concludes it is going to rain, and not
because the term ‘looks’ takes on a different meaning. Rather, the difference
between these two readings is pragmatic rather than semantic. This in turn helps
to explain why both Chisholm (1957: 47) and Jackson (1977: 33) find epistemic
and comparative looks statements to be so closely connected since they are
effectively two different uses of the same proposition.
Parsimony therefore suggests that we should abandon the idea of distinct
comparative, non-comparative and epistemic uses of ‘looks’. Instead, it offers a
single logical analysis of ‘looks’ in the form of (20) and its subject-relative
variant (22), and two, or perhaps three, uses of looks statements. ‘Looks’, on
this account, is implicitly comparative in both its epistemic and phenomenal
uses. These two uses only come apart, if at all, in certain very specific cases,
such as when talking about the phenomenal character of one’s experience. In
most cases, however, including all of Jackson and Byrne’s examples (4.2), it is
the comparative use that is being employed. As a general rule, we have little
application for the phenomenal usage in everyday conversation, and even when
we do use it — when doing philosophy, for example — it is unsuitable for
indexing p-representation for the reasons given above.
Furthermore, on the assumption that the deflationary analysis is correct,
which seems at least plausible, most uses of ‘looks’ are either explicitly or
implicitly comparative and so subject to Travis’s argument from looks (2.3.2).
This conclusion is unfavourable for the representationalist insofar as it suggests
that experiential contents cannot be looks-indexed, since comparative looks are
equivocal, thereby contravening the Face-value condition. The
representationalist is therefore faced with the choice of rejecting either Looksindexing, Face-value or Parsimony. However, the mere plausibility of the
deflationary analysis of looks is itself a problem for representationalism, since in
the absence of an effective counter-argument, the mere possibility that looks
may be comparative undermines the argument for representational content.
Indeed, given the problems with Jackson and Byrne’s alternative accounts of
phenomenal looks and the resulting notion of phenomenal content, one could
argue that the burden of proof lies very much upon the representationalist to
give a substantive account of looks that does not suffer from the problem of
phenomenal content, or else — assuming that Face-value forms part of the view
that they wish to defend — to reject Looks-indexing. Either way, the prospects
for looks-indexed p-representational content do not look good.
Chapter 4, Looks
105
4.4. Conclusion
In this chapter, I examined three different notions of looks: comparative,
epistemic and phenomenal, or ‘non-comparative’, looks. I argued that Jackson’s
case for the primacy of the latter is unconvincing given its reliance on the
unwarranted and implausible assumption that there must be some particular set
of conditions according to which all looks statements are judged. Moreover, the
phenomenal or non-comparative use of looks is only capable of indexing
phenomenal content, which is veridical even in cases of perceptual illusion or
misleading — i.e. ‘mere’ looks — and so incapable of misrepresenting. Such uses
are arguably equally well explained by a comparative analysis of looks
statements based upon the intersubjective similarities between experiences of
different objects and properties. Combined with the principle of Parsimony,
which states that the properties that make looks statements true or false are just
the appearance-independent properties of objects, this deflationary account
delivers a unified analysis of looks encompassing both objective and subjective
appearances. Furthermore, the deflationary account renders the notions of
comparative and epistemic looks redundant, replacing them with comparative
and evidential uses of looks statements, respectively.
The availability of an alternative, more powerful deflationary analysis of
‘looks’ renders Travis’s provisional assumption of Looks-indexing untenable. If
looks are essentially comparative, as Parsimony suggests, then they are
equivocal and so incapable of indexing p-representational content, as per
Travis’s argument from looks (2.3.2). The only sense of ‘looks’ that is not
comparative, therefore, is Jackson’s phenomenal sense, but this is similarly unfit
to index p-representation. Therefore the content of experience cannot be looks
indexed and Looks-indexing is false. In the absence of any further argument
against the plausibility of the deflationary analysis, this neutralises the
representationalist claim that the existence of representational content may be
derived from an analysis of appearances. Any appeal to the nature of
appearances, or the semantics of ‘looks’, may be defused by a counter-appeal to
Parsimony. Whilst this does not constitute a knock-down argument against
representationalism, it does shift the grounds of the debate away from the
analysis of looks-statements and towards the issue of recognisability, or
availability of content to the subject. This is perhaps unsurprising since the
semantics of ‘looks’ or appearances is a peripheral, though related, issue to the
metaphysics of experience. If one cannot ground the representational content of
experience in appearances, however, then Byrne, Siegel and Schellenberg’s
arguments are beside the point. The following chapter therefore considers the
nature of recognisability, along with some possible responses to it, in greater
detail.
5. Recognisability and
Phenomenal Character
Objective representation & perceptual awareness
5.1. Introduction
Having rejected various appeals to visual phenomenology, appearances and the
semantics of ‘looks’ as responses to Travis in chapters 3 and 4, the following
two chapters examine the relation between representationalism, Travis’s
Recognisability condition and the phenomenal character of experience. I
approach this by way of two distinct, but related questions. First, should
representationalists endorse Recognisability, and if not, what are the
consequences of rejecting it (5.2)? Second, how might some form of this
condition be met by those representationalists who do endorse it (ch. 6)?
I begin by considering the motivation for Recognisability in relation to Facevalue (5.2.1) and phenomenal character (5.2.3), as well as what form the
condition should take (5.2.2). I then examine Tyler Burge’s (2010) account of
‘objective representation’ and its relation to Travis (2004), arguing that they
each relate to quite different explanatory projects (5.2.4). The second part of the
chapter identifies a potential challenge, or ‘phenomenological objection’ (5.3.1),
to representationalism concerning the role of phenomenal character in
explaining our awareness of mind-independent objects that features in Travis’s
argument from unmediated awareness (5.3.2). I argue that the challenge is
question-begging and so does not, without substantial additional argument,
constitute a genuine problem for representationalism.
5.2. Recognisability
I introduced Recognisability in chapter 2 as a means of capturing Travis’s
requirement that the contents of individual experiences be cognitively available
or otherwise accessible to the subject (2.2.4). In this section, I examine the
Chapter 5, Recognisability and Phenomenal Character
107
arguments for and against this condition in greater detail (5.2.1), how it should
be interpreted (5.2.2) and the relation between Recognisability and phenomenal
character in order to establish whether the representationalist need be
committed to it (5.2.3). This in turn clarifies two different questions that may be
asked about representational content in perceptual experience (5.2.4), along
with the consequences of accepting or rejecting this condition upon prepresentation (5.2.5).
5.2.1. The case for Recognisability
As described in previous chapters, Travis’s arguments from looks and
recognisability rely heavily upon the notion that the contents of experience are
not just attributable to perceptual experiences from a third-personal theoretical
point of view, but accessible to or ‘recognisable’ by the subject. Without this
condition, which forms an explicit premise in the argument from recognisability
(2.3.3) and is also referenced in the argument from looks (2.3.2), Travis’s
arguments would fail to go through. It is therefore essential to his case against
representationalism that the condition obtains. But precisely what does
Recognisability entail, and why should the representationalist endorse it? Here
is how Travis (forthcoming: §4) motivates this condition:
If we are going to be represented to in experience, then the relevant
representing must be something we can appreciate for the representing that
it is. If, in a perceptual experience, things are represented to us as being thus
and so, then we must be able to appreciate the experience as representing as
so what it thus does; to appreciate what it is that is so according to it. This
need not mean that we can characterise such representational content
accurately, or formulate it explicitly. But we should be able to recognise,
where needed, of particular ways things may or may not be, whether that is
what the experience represented to us as so — whether that is what one
would take be so in taking the experience at face value — whether, for
example, the experience is one according to which a certain stick is bent, or
rather one according to which that stick is straight. The core idea is: you
cannot represent things to people as so in a way they simply cannot
recognise as doing that.
The above passage highlights the connection between Recognisability and
other conditions that Travis places upon p-representation, in particular Facevalue (2.2.2) and Givenness (2.2.3). Jointly, these two conditions entail that
experiences have a determinate and univocal (i.e. single) ‘face value’ at which
they may be taken or declined that is represented to the subject. Naturally, not
all representations are recognisable in this way. Brain states, for example, may
have representational content that need not be manifest at the level of conscious
thought. Instead, these may form part of subconscious, sub-personal
mechanisms, the nature of which is known only to neuroscientists if at all.
Crucially, however, according to Givenness, p-representation is a form of
Chapter 5, Recognisability and Phenomenal Character
108
representation to the subject and so operates at a personal, rather than subpersonal, level.
Givenness alone, however, does not entail recognisability. According to
Tyler Burge’s (2010) view of perceptual representation, perception is ‘a type of
objective sensory representation by the individual’ (ibid. 368; original italics)
that is ‘constitutively a representational competence’ (ibid. 379) and ‘where
representation begins’ (ibid. 366). Perception, according to Burge, has objective,
subject-level contents. These are not, however, recognisable to the individual,
but rather play a distinctive role in scientific theorising. They are not
representations-to, but representations-by the individual; autorepresentation
rather than allorepresentation in Travis’s terminology (2.2.3).
What, then, appears to motivate Recognisability is the notion that
experiences have a specific ‘face value’ at which they are capable of being taken
or declined. It is constitutive of being able to take something at face value that
what is taken or declined, i.e. its representational content, must be consciously
available or accessible to the subject. When Travis (op. cit.) states that ‘you
cannot represent things to people as so in a way they simply cannot recognise as
doing that’, his thought appears to be that in order to facilitate the taking or
declining of contents in perceptual judgement, it must be apparent to the subject
what those contents are. If this were not the case, then it would be difficult to
make sense of the idea that subjects are able to take experiences ‘at face value’,
or to recognise the distinction between how things appear to them and how they
believe them to be at all, since only the latter contents would be available to
conscious reflection.1 On such a view there would be no determinate way that
things appear to the subject to be, and so nothing capable of being judged as so.
Rather, perceptual judgements and beliefs would simply spring fully formed, as
it were, into consciousness — something that is clearly not the case.2 Thus, for
Travis, the recognisability of content is a necessary condition for representing to
a subject, though not necessarily for representations in general.
Having established at least one motivation for endorsing Recognisability (I
give another in 5.2.3 below), we can capture this condition as follows:
Recognisability: perceptual subjects must be capable of recognising the
representational content of any given p-representation solely in
virtue of having that very experience.
This formulation, however, leaves at least two important questions unanswered:
(i) what does such ‘recognition’, or availability to a subject, consist in?, and
1
2
A related objection to representationalism is given in 5.3.1.
The situation here might seem no worse than for the anti-representationalist, who denies
the existence of such experiential contents outright. However, since representationalism
specifically appeals to representational contents as being constitutive of perceptual
experience, the problem constitutes an objection to the coherence of the representationalist
view.
Chapter 5, Recognisability and Phenomenal Character
109
(ii) in virtue of what are such experiences recognisable? I address each of these
questions below.
5.2.2. Availability to the subject
The qualification at the end of the above formulation of Recognisability is
intended to capture the need for the p-representational contents to be
recognisable as contents of experience, and not, for example, as the contents of
beliefs or actions. This is important because whilst it is widely accepted that
perceptual beliefs, judgements and actions have intentional content, this cannot
in itself be taken to prove the existence of p-representation. Provided that an
alternative explanation of the relevant phenomenon — namely, perceptual
experience — is available, then representation is just one of many several
possible explanations, each of which is compatible with the existence of the
contents of belief and action. Moreover, if the notion of p-representation is itself
incoherent, as Travis argues, then there can be no such representations. It is
therefore necessary for p-representational contents to be attributable directly to
experience, and not simply to beliefs or judgements, which are traditionally
thought of as being ‘downstream’ of experience. Consequently, cognitive access
to the intentional content of perceptual belief or action is not sufficient to meet
Recognisability. Similarly, a subject who is connected to a brain-scanner, the
output of which is interpreted by a machine or team of scientists that informs
the subject of the ‘contents’ of their experience would fail to meet the relevant
condition. Rather, it must be apparent to the subject — perhaps on the basis of
reflection or via perceptual phenomenology — what the content of their
experience is. I will talk of such content as being available to or accessible by the
subject.
One way in which p-representational contents may be available to a subject
is through introspection, or reflective awareness, of their own mental states.
That is, in reflecting upon their experience, the subject may be able to tell that it
represents, for example, that there is a lime in the fruit basket. For this to count
as evidence of the content of experience, rather than of belief, such content
should exhibit a degree of independence from the subject’s beliefs. If they do not
believe there is a lime in the fruit basket, for example, since they bought only
lemons, then experience must still recognisably represent there to be a lime —
perhaps due to some trick of the light — in spite of their inclination to believe
otherwise. Similarly, the lines of the Müller-Lyer illusion continue to look
different lengths even to subjects who know them to be otherwise — because
they have measured them, for example — quite independently of the subject’s
beliefs. Finally, there must be some specific way that experience represents the
world as being in order for it to qualify as p-representation. Contents that are
equivocal about what is represented, or that merely reflect what the subject is
inclined to judge under those circumstances, do not constitute evidence of prepresentation.
Chapter 5, Recognisability and Phenomenal Character
110
The above considerations suggest a stronger interpretation of
Recognisability based upon the subject’s ability to gain knowledge of the
content of their experience as follows. Call this Knowability:
Knowability: perceptual subjects must be capable of coming to know the
representational content of any given p-representation solely in
virtue of having that very experience.
As with Recognisability, Knowability does not require that subjects routinely
know or are capable of formulating the accuracy conditions for each individual
experience. Rather, it suffices that they are in principle capable of knowing how
any given experience represents the world as being. This renders the resulting
contents cognitively available or accessible to the subject as part of their overall
mental economy, and so capable of playing a role in motivating action or belief.
Nevertheless, Knowability might be thought to be too strong a condition upon
p-representation for two reasons. First, such knowledge may be tacit or implicit,
rather than explicitly held by the subject. Second, this may require the subject to
employ other capacities, such as conceptual competencies or draw upon other
background knowledge, and so not be ‘solely in virtue of having that very
experience’.
In relation to the first point, Travis’s account of Recognisability does not
require subjects to know the content of their experiences in any explicit sense.
Rather they should be ‘able to appreciate the experience as representing [what
it] does’ and to ‘recognize, where needed, of particular ways things may or may
not be, whether that is what the experience represented to [them] as so’ (Travis
2004: 62). This suggests a weaker interpretation of Recognisability according to
which the subject need only possess a tacit or implicit grasp of what it would
take for their experience to be veridical. That is, they would recognise upon
further inspection, for example, whether the experience was veridical or not
based upon their interactions and expectations of the world.3 Such an
understanding may be reflected in thought and action by guiding subjects’
actions and judgements according to what is represented. We can therefore
formulate a weaker condition than Knowability as follows:
Recognisability′: perceptual subjects must be capable of grasping what it
would take for the content of any given p-representation to be
accurate or veridical solely in virtue of having that very
experience.
Note that even on this weaker formulation, it must still be possible for a
suitably self-aware and conceptually sophisticated subject to consciously
recognise that experience is representational as such. If this were not the case,
then the relevant representations would be opaque to reflective awareness
3
This counterfactual claim is intended to apply both to subjects who are actually capable of
such investigation and those who are not.
Chapter 5, Recognisability and Phenomenal Character
111
(though not, perhaps, to empirical or scientific investigation; cf. 5.2.4). Not all
perceptual subjects who satisfy Recognisability′, however, need possess such
reflective self-awareness.
In relation to the second of the above points, Looks-indexing characterised
the recognisability of perceptual content in terms of how, according to the
relevant experience, things perceptually seem — or look in the case of vision — to
the subject. In this case, the relevant contents were recognisable solely in virtue
of experience, since to have that experience was just to be aware of things
looking some particular way. Having rejected Looks-indexing, however, we
must now consider how else a subject might understand or grasp what is
represented to them. Plausibly, this may involve the operation of various
cognitive capacities, including conceptual capacities or other forms of
background knowledge — a knowledge of appearances, for example (6.4.2).
Whether such recognition is solely in virtue of experience will depend upon the
nature of the particular capacities or knowledge involved, and whether this is
taken to be built into experience or external to it.
In the case of background knowledge, we can rule out the use of inference
or deductive reasoning, since p-representation is intended to capture how
experience represents the world as being, and not what can be concluded or
deduced on the basis of it. However, to avoid ruling out the non-inferential
application of the relevant capacities, we can further weaken Recognisability′ as
follows:
Recognisability″: perceptual subjects must be capable of grasping what it
would take for the content of any given p-representation to be
accurate or veridical solely in virtue of having that very
experience plus the non-inferential operation of perceptual
capacities and/or tacit knowledge of appearances.
This formulation of Recognisability retains the spirit of Travis’s requirement
that the relevant recognition be perceptual, and not simply a matter of reasoning
or deduction, whilst not unnecessarily limiting the forms that such recognition
might take.
5.2.3. Recognisability and phenomenal character
In addition to motivating Recognisability via its connections with Face-value
and Givenness (5.2.1) — both conditions that many representationalists would
accept — there are good reasons to think that p-representation must also be
recognisable in the following sense. To perceptually experience something
entails a modification of the subject’s conscious state such that there is
something that it is like for them to have that experience. 4 This ‘what it’s like’ness of experience is identified as its phenomenal character (1.2.2). Given that
representationalists explain the nature of experience in terms of its
4
I discuss unconscious perceptual experiences below.
Chapter 5, Recognisability and Phenomenal Character
112
representational content, and that every such experience has some specific
phenomenal character, it follows that such content must play a role in
determining phenomenal character. Several variants of this view are possible.
Chief amongst these are weak and strong intentionalism:5
Weak intentionalism: the phenomenal character of a perceptual
experience supervenes upon the (p-)representational content of
that experience.
Strong intentionalism: the phenomenal character of a perceptual
experience is identical to the (p-)representational content of that
experience.
In the case of strong intentionalism, every phenomenal character that an
experience can have corresponds to some specific representational content since
the two are, on this view, identical. Since phenomenal character captures what
experience is like for the subject, and assuming that the subject is or can become
consciously aware of this phenomenal character, it follows that such
representational content will in some sense also be recognisable to the subject.
This claim needs to be stated with some care since on a representationalist
view of perception, the object of the subject’s conscious awareness is not the
phenomenal character of experience per se, but worldly objects and their
properties. Nevertheless, if we take phenomenal character to be a modification
of the subject’s conscious state, there is a sense in which they are, or can
become, aware of what their experience ‘is like’, i.e. its phenomenal character.
Such awareness of phenomenal character is derivative of the subject’s experience
of the objects of perception, and is what is subjectively indistinguishable from
the case of a phenomenally matching hallucination or illusion. Since, according
to strong intentionalism, every phenomenal character is identical to some
representational content, it follows that this content is also recognisable to the
subject on the basis of their experience in virtue of that experience’s phenomenal
character.
The position is less straightforward for weak intentionalism, which posits a
supervenience rather than identity relationship with representational content.
On this view, different contents may correspond to the same or subjectively
indistinguishable phenomenal characters (though not vice versa). Consequently,
representational content is underdetermined by phenomenal character. This
means that contents need not be recognisable on the basis of phenomenal
character alone, giving the weak intentionalist grounds to reject Recognisability
on the basis of phenomenal character. (The above argument from Face-value
and Givenness, however, still applies.) Nevertheless, according to many weak
intentionalist accounts, there is some subset or core of representational content
that determines phenomenal character. Where this is the case, it may in some
5
These positions are also sometimes referred to as weak and strong representationalism. For
the avoidance of confusion, however, I will adhere to the present terminology.
Chapter 5, Recognisability and Phenomenal Character
113
(though not necessarily all) cases be possible to recognise that subset of content,
though not the content of experience in general, on the basis of the phenomenal
character to which it gives rise. That is to say, the subject will be capable of
grasping what it would take for their experience to be veridical on the basis of
its phenomenal character alone. On such views, a version of Recognisability will
also hold.6
It is notable that neither form of intentionalism requires that the content of
experience be exhausted by its phenomenal character. As such, there may be
elements of p-representation that do not contribute to phenomenal character
and so cannot be recognised in virtue of it. If so, however, this will place
constraints upon the explanatory role that those elements of conscious
perceptual experience can play. If certain aspects of a representational content
make no contribution to what experience is like for the subject, for example,
then it is difficult to see how they can form part of that experience’s face value. 7
Whilst there is no reason that such content could not play some other
explanatory role — in scientific theorising, perhaps (5.2.4) — it is unclear what it
could possibly say about perceptual experience as such. Nevertheless, an
interpretation of Recognisability that ruled out the existence of such content
elements would be beg the question against those forms of intentionalism that
include them. We can therefore place a further gloss upon Recognisability″ as
follows:
Recognisability‴: perceptual subjects must be capable of grasping what
it would take for that content which gives rise to the phenomenal
character of any given p-representation to be accurate or
veridical solely in virtue of having that very experience plus the
non-inferential operation of their perceptual capacities and/or
tacit knowledge of appearances.
It is this version of Recognisability that I will assume in further discussion and
that I take representationalists to be committed to (modulo underdetermination
of content by phenomenal character in weak intentionalism, as described
above).
In discussing perceptual experience, I have so far focused upon conscious
experience. That is, those experiences that have a distinctive phenomenal
character of which the subject is, or may become, consciously aware. It remains
an open question whether the same kinds (i.e. type-identical) experiences can be
had by subjects who are not consciously aware, or not actively attending to,
whatever it is they are experiencing. Block (2007), for example, defends an
account of perceptual experience which distinguishes between what he calls
‘phenomenal’ and ‘access consciousness’, the first of which may be present in
6
7
For a representational view that denies Recognisability, see 5.2.4.
A possible exception to this might be externally individuated or demonstrative elements,
though these arguably do make a contribution to perceptual phenomenology, at least in a
general sense, and so form part of the resulting experience’s phenomenal character.
Chapter 5, Recognisability and Phenomenal Character
114
non-conscious perceptual experience and the latter of which corresponds to
conscious access to the contents of experience. These come apart in cases such
as blindsight or deaf-hearing, which are claimed to exhibit the second but not
the first kind of perceptual consciousness.
Nevertheless, the nature of conscious perceptual experience has — rightly or
wrongly — assumed a certain priority in the discussion of perception due to the
latter’s relation to thought and consciousness more generally. A subject whose
perceptual experiences have a specific phenomenal character, for example, may
be able to form demonstrative thoughts about the objects of their experience
(5.3.1). This seems not to be true of subjects, such as blindsight or deaf-hearing
patients, who lack the relevant phenomenal character in at least part of their
perceptual field, or whose experience exhibits phenomenal but not access
consciousness, similarly preventing phenomenal character from playing an
epistemic role (cf. ibid.). The notion that phenomenal character plays a role in
demonstrative thought does not of course rule out the existence of nonconscious perception. However, if experience is to be explained in
representational terms, then it should be clear what role it plays in conscious
awareness. Plausibly, this occurs through the phenomenal character of
experience which, according to intentionalism, is a function of its
representational content. Thus if intentionalism, i.e. the view that phenomenal
character supervenes upon or is identical to representational content, cannot
account for the nature and role of conscious perceptual experience, then it
should be rejected.
5.2.4. Two questions about perception
Having clarified the interpretation of Recognisability, along with some
motivations for endorsing it, I will now consider the consequences of rejecting
this condition upon p-representation. One theory that does this is Tyler Burge’s
account of perceptual representation, or what he calls ‘objective’ or ‘empirical
representation’ (Burge 2010: 3). I will argue that in rejecting Recognisability,
Burge’s account fails to account for the nature of perceptual experience, and so
does not constitute a form of representationalism as such. This in turn helps to
clarify the claims made by Travis (2004), showing that these two views are not
necessarily incompatible, but rather are responding to two different questions
about the nature of perception and experience, respectively.
According to Burge (2010), ‘[p]erception is a type of objective sensory
representation by the individual’ (ibid. 368; original emphasis) and
‘constitutively a representational competence’ (ibid. 379). The contents that
perceptual states have, however, is not in virtue of their phenomenal or
conscious character, since they may have none, but rather a function of certain
distal facts about the evolutionary history and representational capacities of the
organism. Combined with the proximal stimuli of the organism’s sensory
surfaces, this yields representational contents that play a non-trivial role in the
scientific explanation of the relevant individual’s psychology and behaviour. On
this view, perceptual states have representational content irrespective of whether
Chapter 5, Recognisability and Phenomenal Character
115
those contents are consciously available to the subject via introspection. Rather,
such content has a purely formal or descriptive role in explaining the
functioning of the organism from a scientific perspective. Burge therefore rejects
Recognisability as a necessary condition for perceptual representation, claiming
in relation to vision, for example, that ‘[a]t the level of conscious access,
individuals are oblivious to what they perceive’ (ibid. 375).
In contrast to Burge (2010), whose account aims to explain the emergence
of representation and objectivity in general, Travis (2004) claims that
‘perception is not representational’ (ibid. 57), that ‘perceiving what we do has
no representational content’ (ibid. 92) and so is ‘not an intentional
phenomenon’ (ibid. 93). It is tempting to read Burge (2010) as a refutation of
Travis (2004) on the basis that the latter makes precisely the mistake that Burge
warns against; namely, attributing a capacity for representing certain general
conditions for perceptual representation — the availability of content,
phenomenal character, conscious awareness, and so on — to the subject.
According to Burge, such attributions ‘constitute hyper-intellectualization of
constitutive requirements on perception’ (Burge 2010: 13) which can equally
occur in the absence of such personal-level rational capacities. Rather, objective
representations are attributable to the subject — in virtue of initiating action
(ibid. 373), for example — but are not representations to the subject in the sense
described above (5.2.1). It is unclear, however, that Burge’s and Travis’s
accounts actually conflict. Rather, we can understand each as addressing quite
separate and independent questions about the nature of perception.
The question to which Burge is responding concerns whether perceptual
states have representational contents, and if so, what those contents are. The
answer to this question will determine whether such content features in an
account of what it is for a creature to perceive external objects. Such content
need not be consciously or otherwise accessible to the subject in order to play an
explanatory role in the relevant scientific theories. To say that perceptual states
may be assigned representational content from some third-person theoretical
point of view, however, is not to say that perceptual experiences are to be
explained solely in terms of their content — the position I have been referring to
as ‘representationalism’. What a subject experiences when they are in a given
perceptual state is intimately related to their perceptual phenomenology, which
need not be explained in purely representational terms. Indeed, Burge himself
endorses a non-representational account of perceptual sensations, denying that
intentionalism constitutes an adequate account of phenomenal character (ibid.
14, fn. 2). Thus, the question of whether perceptual states possess or can be
assigned representational content need not entail — or at least not
straightforwardly so — a representationalist explanation of perceptual
experience, where this is taken to involve conscious phenomenal character.
In contrast to Burge, Travis (2004) is concerned to give an account of the
constitutive nature of perceptual experience; that is, of the presentation of an
external, mind-independent world in consciousness. This brings considerations
about phenomenal character to the fore, since this is — at least in a large number
Chapter 5, Recognisability and Phenomenal Character
116
of central cases — an aspect of the conscious experiences of some conscious or
self-aware subject. This is not to say that consciousness or self-awareness is a
necessary condition for perception — a point that Burge (sometimes erroneously)
makes against what he calls ‘Individual Representationalism’ in contrast to his
own ‘anti-individualist’ account of perception (Burge 2010: 12–22). Rather, it
underscores the need to explain the role of phenomenal character in the
subject’s consciousness or perceptual awareness of the world. If perceptual
experience is to be explained in terms of representational content, then there
must presumably be some relation between that content and the phenomenal
character of experience — hence Travis’s Recognisability condition (5.2.3). Not
all representationalists endorse Recognisability. But where they don’t, the
burden is upon them to explain precisely how the contents of perceptual
experience give rise to its phenomenal character, and how this in turn enables us
to be consciously aware of the external world, objects, properties, and so on.
The above two questions about the existence of representational content and
the nature of perceptual experience therefore come apart, and may — to some
extent, at least — be considered independently of one another. On the one hand,
Burge argues in favour of representational content on the strength of its role in
scientific explanation, but without explaining how this relates to the
phenomenology of conscious perception. Travis, on the other hand, argues that
perceptual experience cannot be explained in terms of representational content,
since the relevant content cannot explain what is recognisable to us (2.3.3) or
how perceptual awareness is possible (2.3.4). So far, there is no reason to think
that these two claims are incompatible. It would, for example, be perfectly
possible to given a Burgean account of representational content whilst accepting
that perceptual experience, in terms of its conscious phenomenal character,
cannot be explained in this way — indeed, Burge himself takes this view.
Intentionalist accounts of perception attempt to answer the second question
concerning the nature of phenomenal character in terms of the first, i.e. the
existence of representational content. Anti-representationalists, on the other
hand, who endorse REL (1.2.4), deny that any answer to the question
concerning phenomenal character that draws upon the representational content
of experience can be given. Strictly speaking, however, they are neutral on the
first question. Rather, their principal claim concerns the role that representation
can play in an account of perceptual experience, rather than ruling out its
existence altogether.
Distinguishing these two aspects of the debate — the existence of
representational content and its role in accounting for conscious experience —
enables us to more precisely pinpoint and evaluate the claims made by Travis
(2004) and Burge (2010). Travis’s concern with perception is directed towards
the experiential dimension of the debate. His arguments aim to show, based
upon what is available to the subject through the phenomenal character of
perception, that conscious perceptual experience cannot be explained in purely
representational terms. This does not, however, rule out those views that do not
appeal to representation to explain perceptual phenomenology, such as Burge’s
Chapter 5, Recognisability and Phenomenal Character
117
objective representation or Naïve Realist accounts of experience. Conversely, by
abandoning Travis’s Recognisability condition, Burge (2010) is able to avoid the
arguments from looks and recognisability since, for him, representational
content has an entirely different explanatory role relating to scientific
explanation and not the phenomenal character of experience.8 However, it does
so at the cost of failing to make any substantive claims about the phenomenal
character of experience, which remains unexplained. This is not to say that a
Burge-style account of phenomenal character could not be developed. Such an
account would, however, require substantive additions to Burge’s theory which
would then be subject to Travis’s critique of representational content, as
discussed in the following chapter (6.3).
Similarly, when Travis claims that ‘[p]erception is not the stuff of which
things might be represented to us as so’ (Travis 2004: 93), he must be
interpreted as making a claim about conscious experience, since nothing in the
arguments he gives in this paper rules out the possibility that we might assign,
for scientific or other reasons, representational content to perceptual states. 9
Provided there is some way — teleosemantic, anti-individualism, asymmetric
dependence, etc. — of disambiguating the contents of these states from all the
other possible contents that might be attributed to that state, then such content
cannot be objectionable on the basis of the arguments Travis provides.10 Rather,
his claim is that such content cannot yield an adequate characterisation of
perceptual experience, where this is taken to include conscious phenomenology.
Understood in this way, ‘The Silence of the Senses’ does not establish that there
is no role for representational content in the philosophy of perception — indeed,
it is silent on precisely this point. Instead, Travis (2004) denies that we should
think of perceptual experience, i.e. the conscious phenomenal character of
perception, in representational terms. No form of representation, according to
Travis, is capable of doing the required explanatory work.
5.2.5. Consequences of Recognisability
Having argued that Burge (2010) does not engage with the issues that concern
Travis (2004) concerning the nature of conscious experience, there remains the
question of whether any account that rejects Recognisability could do so. I
8
Whether it can in fact play this role is debatable since science can only inform us about
causal relations between perceptual ‘inputs’ and experimental ‘outputs’. To the extent that
scientific explanation posits the existence of externally individuated representational
contents, these are a purely descriptive tool or convenience for simplifying the relevant
explanations, and do not indicate an ontological commitment to the existence of
representations plural, as Burge (ibid. 379, fn. 15) himself admits. Furthermore, the role of
the sciences in grounding the nature of such content raises issues concerning the priority of
scientific over philosophical explanation, especially given that the alleged consensus over the
existence of the relevant contents is much weaker than Burge claims, if indeed there is such a
consensus. I will not, however, pursue these issues further here.
9
Travis (forthcoming: §1) hints at this possibility when he says that ‘[t]he representing this
essay is thus not about is, I suggest, enough to serve the purposes of serious psychology’.
10
See 6.3 for further discussion.
Chapter 5, Recognisability and Phenomenal Character
118
argued above that if we understand conscious perceptual experience as being at
least partly constituted by its phenomenal character, then according to a
representationalist account of experience, some form of intentionalism must also
obtain (5.2.3). This follows because if all there is to experience is its
representational content, then the phenomenal character that partly constitutes
that experience must also be representational. The alternative — a
representational account of experience that did not account for phenomenal
character — would not say anything about the conscious aspect of such
experience or what perception is like for the subject. Furthermore, it is highly
plausible that phenomenal character plays a positive epistemic role in perceptual
awareness — for example in enabling the subject to entertain demonstrative
thoughts about an object, or to know which object it is they are perceiving
(5.3.1). Moreover, since phenomenal character includes at least those aspects of
the subject’s conscious state that are discernible via introspection, these aspects
of experience are accessible to the subject in the sense of being available to
consciousness.
According to strong intentionalism, an experience’s phenomenal character is
identical with its representational content. It follows that if the phenomenal
character of experience is accessible to the subject, then so is its representational
content since they are one and the same thing. This in turn entails a form of
Recognisability since the content of experience may be grasped by the subject
simply in virtue of that experience’s phenomenal character (5.2.3). Where the
subject’s access to phenomenal character is incomplete — for reasons of ‘antiluminosity’ (Williamson 2000: ch. 4), for example — or where the experience’s
content is not exhausted by its phenomenal character, then only the content that
corresponds to the accessible phenomenal character will be recognisable.
Nevertheless, Recognisability in the sense discussed above will hold. Strong
intentionalism is therefore incompatible with the denial of Recognisability.
Rejecting Recognisability amounts to the claim that the representational
content of experience need not be graspable on the basis of that experience —
which for the intentionalist includes its phenomenal character — alone. This
might be the case under weak intentionalism where content is underdetermined
by phenomenal character due to different contents yielding the same, or
subjectively indistinguishable, characters. In such cases, it may not be possible
for the subject to grasp which of the relevant contents is represented in virtue of
the resulting phenomenal character, and so Recognisability can fail. This might
happen for a number of reasons. For example, two subjectively indistinguishable
experiences might be thought to confer differing warrants for justification or
belief according to the epistemic situation of the perceiver — a position known
as content disjunctivism. Alternatively, the contents of experience may vary
according to aspects of the perceiver’s environment, evolutionary history or
other externally individuated factors (6.3.1). In each instance, subjectively
indistinguishable experiences may plausibly lead to different beliefs, judgements
or epistemic warrant in a way that is indistinguishable to the experiencing
subject.
Chapter 5, Recognisability and Phenomenal Character
119
The weak intentionalist may therefore wish to reject Recognisability on the
basis that representational content is not indexed by phenomenal character.
Whilst this might be true of weak intentionalism in general, however, the above
examples need not count against the version of Recognisability according to
which only those content elements that give rise to phenomenal character need
be available to the subject (5.2.3). Indeed, it would be surprising if subjectively
distinguishable differences in phenomenal character were not closely correlated
with corresponding differences in representational content, at least at a coursegrained level. Such qualitative differences may enable a limited form of
Recognisability, i.e. Recognisability‴, to be preserved despite the global
underdetermination of content by phenomenal character. This will be the case
for any form of intentionalism in which phenomenal character unambiguously
maps onto elements of representational content, at least in a wide variety of
(though not necessarily all) cases. Such cases are not ruled out by weak
intentionalism per se, but will depend upon the details of individual
representationalist accounts.
If the weak intentionalist were to reject Recognisability outright, however,
then it would be possible to avoid Travis’s arguments from looks and
recognisability. The cost of doing so would be to deny that the corresponding
differences in content form part of what is given to the subject in experience.
This raises the question of what explanatory role such ostensible contents can
play in perceptual experience, since they cannot be manifest to the subject
through the phenomenology of experience. This in turn puts pressure on
whether, given their cognitive inaccessibility, such contents should be considered
a part of experience at all. Perhaps such elements, which may include references
to external particulars, might facilitate the perceptual ‘tracking’ of externally
individuated factors or demonstrative reference of the kind described above.
However, such a role must be confined to a level below that of phenomenal
consciousness, since such differences will be indiscernible to the subject via
perceptual phenomenology.
The weak intentionalist is therefore faced with a choice. They can either
reject Recognisability, in which case it remains to be shown how it is possible
for the relevant representational content to perform a substantive cognitive or
epistemic role at the level of the subject. Moreover, if such content does not
form part of the phenomenal character of experience, then it will be
questionable to what extent it should be regarded as the content of experience at
all, rather than of judgement, belief, or some other mental state. Alternatively,
the intentionalist may endorse Recognisability, in which case an account of
precisely how such content is recognisable is required. Again, the resulting
contents must be demonstrably attributable to experience rather than any
consequent beliefs, judgements or actions, since any evidence on the basis of the
latter will be equally compatible with an anti-representationalist account of
experience, and so not decisive in the present debate. For the weak
intentionalist, this will involve explaining how it is possible to disambiguate the
multiple possible contents of experience that correspond to any given
Chapter 5, Recognisability and Phenomenal Character
120
phenomenal character given that the latter is held to underdetermine the former.
This topic is addressed in the following chapter.
5.3. Phenomenal Character and Awareness
5.3.1. The phenomenological objection
One of the objections that is sometimes levelled against representationalism
concerns the relationship between the phenomenal character of experience and
the perceptual awareness of objects. Specifically, Schellenberg (2011b: 18)
identifies the ‘phenomenological objection’ as the charge that representationalist
views of perception ‘which explain phenomenology in virtue of relations to
anything other than the mind-independent objects, properties, and events that
perceiving subjects are aware of sever phenomenology from what we are aware
of’ (ibid.). A version of this objection can be found in Travis’s argument from
unmediated awareness (5.3.2), which claims that any mental state that can
occur independently of the objects and properties of which it is a perception —
in a hallucination or illusion, for example — will be incapable of explaining how
that state can constitute an awareness of those objects and properties. This
objection requires stating with care, since if the claim is supposed to be that we
cannot gain knowledge of qualitative features of the world towards which we
do not stand in a relation of direct acquaintance, then it risks begging the
question against representationalism.
In describing the phenomenological objection, Schellenberg (ibid.) cites
Campbell (2002), Martin (2002b) and Brewer (2007) as claiming that
‘representational views misconstrue the phenomenological basis of perceptual
experience insofar as they detach the phenomenology of experience from
relations to qualitative features of the world’ (Schellenberg 2011b: 3). Of these,
the clearest exposition of a problem of this kind is to be found in Campbell
(2002: 120–26), who claims that the explanatory role of experience can only be
accounted for in relational terms as follows (ibid. 122–23):
On the Relational View, experience of objects is a more primitive state than
thought about objects, which nonetheless reaches all the way to the objects
themselves. In particular, experience of an object is what explains your
ability to grasp a demonstrative term referring to that object.
For Campbell, the grasp of demonstrative terms is what enables us to gain
knowledge of particular objects, and in particular which objects it is we are
perceiving (cf. Evans 1982: 65). The principal objection, however, does not
concern the acquisition of perceptual knowledge, which Campbell (op. cit. 125)
admits may be explained in both representational and relational terms provided
that the relevant representations are of particulars and not only general features.
Rather, Campbell claims that only a relational understanding of experience ‘can
explain how it is that we can have the conception of objects as mindindependent’ at all (ibid. 121).
Chapter 5, Recognisability and Phenomenal Character
121
Campbell’s objection, then, is that a purely representational account of
experience that does not constitutively involve the objects of experience cannot
explain how we can conceive of objects as being independent of our perceiving
them. If, as Campbell puts it, the contents of experience ‘do not reach all the
way to the objects themselves’, then any resulting conception would only be of
our own subjective experience, which can occur without any such objects, rather
than of an objective world as such.
This problem does not arise for the relationalist, according to whom
perceptual experience is a matter of standing in a relation of direct acquaintance
with perceptual objects and their properties, since acquaintance is not an
intentional relation and so cannot obtain in the absence of the relevant objects.
Moreover, according to certain versions of relationalism — Naïve Realism, for
example — the phenomenal character of experience is not merely individuated,
but metaphysically constituted by the properties of whatever external objects are
being perceived. On this view, the phenomenal character of experience is
directly explicable in terms of its objects, rather than something that is also
present in the non-veridical case, as in the case of representational content.11
Consequently, our conception of objects as mind-independent may be explained
in terms of the subjective experience of such objects in perception through the
phenomenal character of experience.
Travis’s argument from unmediated awareness contains a related objection:
without experience being grounded in the presence of worldly objects that are
brought into view by perception, experience cannot explain our awareness of
external objects and the world. Underlying both of the above claims is an
assumption that thought and representation must at some point ‘bottom out’ or
be grounded in some kind of unmediated or primitive connection with the
world. Plausibly, this connection occurs in perception through the phenomenal
character of experience, which reflects how the world is revealed to us in
consciousness. Call this the grounding assumption.
Grounding assumption: perceptual awareness of external mindindependent objects must be grounded in the phenomenal
character of experience.
According to this assumption, an account of experience should explain both
how phenomenal character enables us to grasp demonstrative references to
objects, and how this makes an awareness of an external mind-independent
world possible. That representationalism cannot do so is what I will refer to as
the phenomenological objection.
A number of responses to this objection are available. First, the
representationalist may simply reject the grounding assumption, as per Burge
(2010). According to Burge, the objectivity of perceptual representation comes
11
As discussed in 3.2.1, the phenomenal character of non-veridical experience may be
explained as being in some sense parasitic upon, or secondary to, that of veridical
experience.
Chapter 5, Recognisability and Phenomenal Character
122
from its teleological or functional nature, and not from the availability of
content to the conscious subject. 12 However, as we saw above (5.2.4), this fails
to explain the connection between phenomenal character and demonstrative
thought, since it is not account of perceptual experience as such. Nor does it
explain how we are able to form a conception of a mind-independent world,
since the relevant representations need not be conscious. Furthermore, the
possibility of blindsight, in which a perceiver lacks the usual visual
phenomenology and yet is able to respond at better-than-chance levels to
perceptual stimuli strongly suggests that there is such a connection, since
without conscious access to the relevant phenomenal character, the blindsighter
appears unable to form demonstrative thoughts about objects to which he is
otherwise able to respond. Rejecting the grounding assumption may not,
therefore, be the representationalist’s best option.
A second response appeals to the relation between the subject and object on
the representationalist account. According to intentionalism, phenomenal
character is either identical to or supervenes upon representational content, and
not relations to the qualitative features of objects or property-instances (5.2.3).
Moreover, most representationalists hold that a perceiving subject does stand in
a relation to the objects and properties they perceive — namely, a causal
relation. Moreover, this relation may be considered perceptual in virtue of its
being constitutive of seeing, hearing, and so on. On such a view, the state or
process of being perceptually aware of some external object or property may be
factored into (at least) two distinct components: (i) a representational state, the
truth or accuracy conditions of which, in the case of veridical perception, are
met by the perceiver’s physical environment, but not in the case of
hallucinations and illusions, and (ii) the satisfaction of some external, typically
causal, relation between the intentional objects of that representational state and
the state itself. Thus, according to this kind of representationalism, perceptual
experience consists of no more than being in an appropriately caused
representational state that there is (so far) no reason to suppose cannot occur in
the absence of the relevant external objects and properties. Indeed, this is
typically how perceptual illusions and hallucinations are to be explained on a
representationalist view (2.3.1).
If the phenomenological objection is that representationalism detaches
phenomenal character from the objects of experience, however, then it is
difficult to see how an appeal to causation is supposed to help here. According
to intentionalism, the phenomenal character of experience is explained by its
representational content, not the causal relation of experience to its objects.13 If
the awareness of objects were to be explained in terms of a causal or other
relation, then it is no longer clear what explanatory work is being done by
12
‘The developmental, phylogenetic, psychological, and constitutive sources of objectivity in
perception lie below the level of individual representation, control, awareness, or
responsibility’ (ibid. 547).
13
Building the causal element into the content, as in Searle (1983), does not seem to help
here.
Chapter 5, Recognisability and Phenomenal Character
123
representational content, and in particular the content that gives rise to the
experience’s phenomenal character. To the extent that the original objection
was to explain how phenomenal character enables a grasp of demonstrative
reference, this explanation does not seem to address the problem. Consider, for
example, two distinct experiences, only one of which is veridical and which
possess the same phenomenal character in virtue of their contents, but where
only the veridical experience constitutes perceptual awareness. Since the only
thing that differs between these experience is the presence of the causal relation,
then (so the argument goes) it must be this and not their phenomenal character
that explains the resulting awareness. Thus the representationalist fails to
explain the role of phenomenal character in perceptual awareness since the
causal relation is doing all of the explanatory work.
Travis makes an objection of this form when he claims that ‘awareness of
something else plus satisfaction of surrounding conditions cannot add up to that
awareness of [an object] which we have in seeing one (to be there)’ (Travis
2004: 89). From this, he derives the following principle (ibid.):
[I]f X is something there might be even without Y, then awareness of X (and
whatever accompanies it per se in a particular case) cannot qualify as
unmediated awareness of Y — the sort one might have in seeing Y.
This principle, however, is contentious for a number of reasons. First, it relies
on two different kinds of awareness: the awareness of external objects and the
awareness of representational content. According to the representationalist, we
are aware of external objects by, or in virtue of, representing them to be some
particular way. We are only aware of representational contents, however, in an
indirect or derivative sense — for example by reflecting upon perceptual
phenomenology. Thus, for the representationalist, we are not aware of
representational content in the same way that we are aware of external objects,
that is to say perceptually aware. Second, the objection is too quick. Like any
‘common factor’ account, intentionalism entails is that representational content
is not sufficient for perceptual awareness. This does not mean that it is not
necessary. If representational content plus some causal or other relation were
held to be jointly sufficient for such awareness, then phenomenal character, or
the representational content that underlies it, will form part of the explanation
of perceptual awareness. This meets the objection that representationalism
detaches phenomenal character from awareness. Thus the phenomenological
objection can be met.
In order to constitute an objection against representationalism, the antirepresentationalist would need to claim that phenomenal character alone should
explain perceptual awareness and the subject’s grasp of demonstratives. But
such a strong principle does not seem to be motivated and would be potentially
question-begging. It is not clear, therefore, that the phenomenological objection
as stated above constitutes a genuine constraint upon representationalism. This
is not to say that Travis’s principle is without merit, but to uncover the nature
of his argument we will need to make it more precise.
Chapter 5, Recognisability and Phenomenal Character
124
5.3.2. The argument from unmediated awareness
A formulation of the phenomenological objection that does not appeal to the
causal relation or other contentious metaphysical principles may be found in
Travis’s argument from unmediated awareness (2.3.4), reproduced below for
convenience.
U1!
P-representation is either (a) a separate source of perceptual
information in seeing or (b) constitutes one’s perceptual
awareness.
U2
If p-representation were a separate source of perceptual
awareness to seeing, then it would be unnecessary for the
awareness of objects simpliciter.
U3
We have no phenomenological evidence for the existence of
multiple sources of perceptual information in visual experience.
U4!
P-representation and seeing are not separate sources of
perceptual information.
U5!
Subjectively indistinguishable veridical and non-veridical
perceptual experiences share a common representational content.
(Common Content)
U6
In non-veridical experience, p-representation cannot yield
unmediated awareness of intentional objects, since there may be
no such objects.
U7
The same representational content cannot explain the
unmediated awareness of objects in some cases, but not in others.
U8!
P-representational content cannot yield unmediated awareness of
an object in veridical perception. (From U5, U6 and U7)
U9!
Visual experience (seeing) is a source of information that yields
unmediated awareness of objects.
U10!
Visual experience is not p-representation. (From U4, U8 and U9)
Premises U1 through U4 argue that, according to representationalism, visual
representational states must be constitutive of perceptual awareness, and not
merely co-present with it. This is followed in U5 by the representationalist
principle of Common Content (2.2), which is used to argue that representational
content cannot yield unmediated awareness of an object, even in veridical
perception (U6 through U8). Finally, since we do think that visual perception
yields unmediated awareness of objects (U9) — that is, of having one’s cognitive
responses shaped by the presence of perceptual objects (Travis 2004: 90) — then
visual experience cannot be p-representational (U10).
Chapter 5, Recognisability and Phenomenal Character
125
There are a number points at which the representationalist may resist this
argument. First, as noted above, representationalism is not necessarily
committed to the view that representational content alone must account for
perceptual awareness. Rather, causal or other conditions, such as the ability to
perceptually track or re-identify objects over extended periods of time (cf. Evans
1982: 175), may also need to be satisfied. Nevertheless, since none of these
conditions itself constitutes an information source in the way that
representational content is envisaged as doing, it does not seem that such
considerations provide sufficient grounds for rejecting U4. 14 Moreover, if
experience were both a source of information about external objects and wholly
representational, there would be no room for any additional or supplementary
sources of information such as might be required for perceptual awareness.
A more promising line of attack can perhaps be mounted against U7, the
claim that the same representational content cannot explain the unmediated
awareness of an object in some cases and not others. By arguing that what holds
for non-veridical experience must also hold for perception, this move is
analogous to the ‘generalising step’ in the traditional argument from illusion — a
move that is often taken to be unsound (e.g. Snowdon 1990; Reid 1997). It is
U7 that enables the inference from U6 to U8 — the claim that p-representation
cannot yield unmediated awareness — from which it is but a short step to the
conclusion in U10 that seeing, and by extension other perceptual states, do not
involve p-representation. As with the traditional argument from illusion, this
argument may be rejected as unsound on the basis that just because
representational states do not yield awareness in one case — for example,
hallucination — does not mean that they cannot (in part, at least) constitute it in
any cases, such as veridical perception. Indeed, it is difficult to see this premise
as anything other than begging the question against the representational view.
Even given the rejection of U7, however, there remains an explanatory
burden upon the representationalist to show how representational content can,
in certain cases given the satisfaction of certain further conditions, constitute
perceptual awareness. The challenge for the representationalist, then, is to
provide an account that explains perceptual awareness at least partly on the
basis of representational contents that are not in themselves sufficient for such
awareness. Or, to put it another way, how the perceptual awareness of objects
depends upon the representational content of perception despite veridical and
non-veridical experiences sharing a single common factor — Martin’s ‘common
kind’ assumption (2.3.1). Unlike Naïve Realism, representationalism does not
deny the common kind assumption. Indeed, the existence of a common factor is,
according to intentionalism, precisely what explains the phenomenal character
of experience.
14
Travis’s use of the term ‘source’ here is somewhat misleading, since it is clearly perceptual
objects and properties that constitute the ultimate source of perceptual information.
According to the target views, however, representational content is part of the means by
which such information is perceptually available to the subject, and so constitutive of the
way that objects and properties are presented to the subject in experience.
Chapter 5, Recognisability and Phenomenal Character
126
5.4. Conclusion
In this chapter I have considered whether and how the representationalist ought
to meet the Recognisability condition proposed by Travis. This highlighted the
distinction between the question of whether perceptual states, as considered
from a third-personal scientific point of view, have representational content, and
whether the phenomenal character of perceptual experience may be explained in
such terms. Consequently, views such as Burge (2010) that seek to establish the
representational nature of perceptual states quite independently of any concern
with conscious phenomenal character do not engage with the point at issue in
Travis (2004), which relates to the explanatory role of representational content
in experience and our conscious awareness of mind-independent objects. As
such, Burge (2010) should not be interpreted as providing any kind of response
to Travis. Rather, he offers a largely orthogonal account of how representation
might figure in the philosophy of perception, though not necessarily in
conscious perceptual experience. The debate, therefore, is not so much between
representationalism on the one hand and relationalism on the other, but rather
between intentionalist theories of phenomenal character and their rivals, such as
Naïve Realism.
Furthermore, representationalists that accept intentionalism — the view that
phenomenal character is identical to or supervenes upon representational
content — face a number of problems in explaining how Recognisability can be
met. I argued that this condition should be endorsed by any theorist who
accepts strong intentionalism, and will also apply to some versions of weak
intentionalism, at least for that content upon which phenomenal character is
held to supervene. If it does not, then this will severely restrict the explanatory
role of the relevant content. I also rejected the ‘phenomenological objection’ that
representationalism detaches phenomenal character from perceptual awareness,
as claimed in Travis’s argument from unmediated awareness (2.3.4). This is
undoubtedly the weakest of Travis’s arguments, and arguably begs the question
against representationalism. To the extent that this objection represents a
genuine constraint upon representational accounts of experience, it does not
appear to present any particular problem for the representationalist, though it
does raise questions as to the explanatory role of phenomenal character in
perceptual experience.
6. Disambiguation Strategies
Externalism about self-knowledge &
discriminatory capacities
6.1. Introduction
In chapter 3, I argued that some straightforward, superficially plausible attempts
to derive the representational content of experiences from the way that things
appear or look fail to establish the existence of objective face-value content.
Further considerations concerning the semantics of appearance epxressions — in
particular the availability of an alternative comparative or deflationary analysis
of ‘looks’ (chapter 4) — led to the rejection of Travis’s provisional Looksindexing condition on perceptual representation. This was replaced by the
weaker Recognisability condition that the content of any conscious perceptual
experience should be in some sense ‘recognisable’, or consciously available to
the subject, solely in virtue of having that experience (5.2). I argued that
Recognisability should taken as a condition upon perceptual representation, or
p-representation, by those representationalists who take experience to have facevalue content that is given to the subject or is otherwise cognitively accessible
via the phenomenal character of conscious perceptual experience. Furthermore,
rejecting this condition would result in severe constraints upon the explanatory
role of the resulting content and its relation to phenomenal character, i.e.
intentionalism.
In this chapter, I consider various ways in which a representational account
of experience and phenomenal character might meet Recognisability. To
overcome Travis’s objections, the representationalist must propose some
plausible means by which the multiple possible contents of experience are
‘disambiguated’ to yield a univocal face-value content (6.2.1). I argue that the
most plausible way in which this may be achieved is through an account of
perception in which both the individuation and recognisability of prepresentational content share a single explanatory source (6.2.2). I then
examine a range of candidates for such a mechanism including externalism
Chapter 6, Disambiguation Strategies
128
about self-knowledge (6.3.2), demonstrative content (6.4.1), and recognitional,
conceptual or other discriminatory capacities (6.4.3). Of these, I conclude that
the latter capacities approach offers the most promising response to Travis, but
also raises a series of questions about the nature and role of such capacities in
individuating representational content, which do not exclusively favour
representationalism over relationalism.
6.2. Content and Experience
6.2.1. Transmission and convergence models
In the previous chapter, I distinguished between the following two questions
about perception and representational content:
(i)
Do perceptual states have representational content, and if so, how is this
content individuated? (Individuation)
(ii)
Is representational content recognisable to the subject in experience, and
if so, how is this possible? (Recognisability)
Accounts of representational content that provide a positive explanation of
Individuation, but reject Recognisability, e.g. Burge (2010), are representational,
but not representationalist as I have been using this term. Those that give a
positive explanation of both Individuation and Recognisability may be further
subdivided according to whether the individuation and recognisability of
perceptual content are explained by the same or different factors into what I will
call transmission and convergence models of experience. On the transmission
model, whatever fixes the content of perceptual experience also makes that
content recognisable with the relevant contents ‘transmitting’ through to the
subject’s grasp of that content. Examples of this approach include Byrne’s ‘noncomparative’ looks (3.2) (insofar as this may be taken to support Looksindexing), externalism about self-knowledge (6.3.1), as well as certain forms of
strong intentionalism; e.g. Glüer (2009). The convergence model, on the other
hand, explains the individuation of content and its recognisability via different
mechanisms that converge upon a single univocal face value, at least in the
majority of ordinary cases.1
The problem for advocates of the transmission model is to explain how, in
virtue of experience combined with the subject’s perceptual capacities and
background knowledge, p-representational content is recognisable to the subject
(5.2.3). Assuming that the relevant content has objective import — i.e. it is not
phenomenal content that represents only the properties of subjective experience
rather than of the objective world (3.2.3) — then what is recognised cannot be
explained on the basis of what experience is like alone, since perceptual
1
An account that gave separate explanations of Individuation and Recognisability that did
not converge upon a univocal face value would be a divergence model. I take it that this is
not an attractive position for the representationalist (see below).
Chapter 6, Disambiguation Strategies
129
phenomenology is equivocal, at least at the level of what is subjectively
distinguishable to the subject. Alternatively, if perceptual phenomenology is not
taken to be equivocal, then an account of how this is possible given the
existence of perceptual ‘ringers’, i.e. subjectively indistinguishable cases that are
non-veridical, and so equally candidates for the face-value contents of
experience. In either case, then, some additional factor that is present in
experience must feature in the explanation of how content is both individuated
and recognised. This additional factor aims to explain how both actual and
recognisable contents remain in step on the basis that both arise from the same
source. The question for the transmission model, then, is to specify what this
additional factor is and how it is able to perform its disambiguating role.
Advocates of the convergence model, on the other hand, must explain both
how p-representational content is individuated and how the resulting content is
recognisable to the subject on the basis of something other than what
individuates it. The problem here is to show how the content that is
recognisable from experience — typically relating to phenomenal character — can
correspond to, or track, the contents that those experiences actually have. If
these contents are divergent, the resulting account of perception will be an error
theory, since the representational content that the subject grasps will be different
to the content that experience actually has, and so not a form of prepresentation. Unless there is something about the way that content is
recognised by the subject that explains why, in the normal case, it corresponds
to the content of experience, this model will be unable to explain how such
content is recognisable to the subject.2 Whilst this possibility cannot be ruled
out in advance — due to the evolutionary pressures of natural selection, for
example — it seems implausible that the individuation and recognisability of
content can be explained by two separate mechanisms that just happen to
deliver precisely the same results in the majority of cases. In any case, this would
be a substantial commitment that supporters of a convergence approach would
need to defend. I therefore consider it unlikely that the convergence model is
capable of delivering an account of p-representation that satisfies
Recognisability.
6.2.2. Explanatory strategies
The primary objection that Travis (2004) sets up for those forms of
representationalism that endorse Face-value is one of indeterminacy; i.e. there
are too many possible contents of experience, each specifying a way that the
world would need to be in order to satisfy the relevant appearance, and all with
equal claim to being ‘the’ content of experience. As Tyler Burge puts this point
in relation to vision, ‘[a] key aspect of this problem is that effects of proximal
stimulation on the sensors of a perceptual system underdetermine both
2
Note that the problem here is not to guarantee that subjects cannot be mistaken about the
contents of experience, since this is something that representationalists can allow. Rather, it
is to explain how in a substantial number of ordinary cases that such recognisability is
possible.
Chapter 6, Disambiguation Strategies
130
representata of perception and the nature of the perceptual state that represents
such representata’ (Burge 2010: 344). A suitable response must therefore specify
both how determinate univocal content is to be assigned, and how it is
recognisable to the subject. That is, it must disambiguate between the multiple
possible contents of experience in such a way that explains how the resulting
content is capable of playing some substantive cognitive or epistemic role in the
perceiver’s mental life. Consequently, to determine precisely what the
representational content of experience is, some additional element or factor
must be appealed to.
Focusing on the transmission model (6.2.1), then, possible candidates for
explaining both the individuation and recognisability of contents via the same
mechanism include:3
(i)
anti-individualism about content (Burge 1979; 2010), self-knowledge
(Burge 1998; 1996) and phenomenal character (Tye 1995; 2002)
(ii)
distal causes, such as evolutionary function, that explain the function of
perceptual mechanisms in terms of the role they play for that organism
or genus (Millikan 1989; Burge 2010)
(iii)
asymmetric dependency upon the veridical case (Fodor 1987)
(iv)
demonstrative identification of the external objects and/or properties
being perceived (McDowell 1994; Brewer 1999)
(v)
the subject’s background knowledge of which appearances are correlated
with which objects and property types
(vi)
conceptual or discriminatory capacities invoked by individual perceptual
experiences (McDowell 1994; 2008)
Options (i) through (iii) fall under the heading of externalism about
representational content and self-knowledge (6.3) whilst options (iv) through
(vi) are related by their reliance upon perceptual discriminatory capacities (6.4).
Option (iv) exhibits characteristics of both groups, but is included in the latter
for convenience. In the following sections, I examine each of these candidate
views in turn.
6.3. Externalist Strategies
Externalist theories of content — for example, Burge (2010) — maintain that
representational content depends upon distal factors, such as the evolutionary
history of the organism’s genus or the representational capacities of its
perceptual systems (6.3.1). I will call this kind of view individuative externalism
(cf. Peacocke 2003). This may be contrasted with object-involving or Russellian
3
This list is not intended to be exhaustive, but is rather a representative sampling of
disambiguation strategies employed in the recent literature.
Chapter 6, Disambiguation Strategies
131
views that entail that perceptual objects or their properties are proper parts or
constituents of representational content. This is comparable to the views of
Naïve Realists, who take phenomenal character to be at least partly constituted
by external objects and their properties. I will call these kinds of view
constitutive externalism. When using the terms ‘externalism’ and ‘externalist’
without qualification, I will mean individuative externalism, or antiindividualism in Burge’s (1979) sense.
When considering the availability of externally individuated content to the
subject, however, it is not clear that such ‘external’ factors are accessible to
introspection, thus presenting a challenge to any externalist theory of experience
that wishes to uphold Recognisability. One solution to this would be to claim
that the subject’s awareness of their own mental states is also externally
individuated, thereby ensuring that the contents of experience ‘transmit’ through
to the resulting beliefs and judgements (6.3.2). This strategy aims to reconcile
the external individuation of perceptual content with the requirement that the
subject be capable of grasping such contents simply in virtue of having — or
rather being able to conceptualise — the relevant experience. This in turn
amounts to a rejection of premise R7 in Travis’s argument from recognisability
(2.3.3). Combined with intentionalism, however, this approach has the
consequence that the p-representational content upon which phenomenal
character supervenes or is identical to must also be externally individuated
(6.3.3). Finally, I consider an alternative externalist account that is due to Fodor
(1987) that I reject as an explanation of perceptual experience.
6.3.1. Teleosemantics and biosemantics
The first approach to disambiguating p-representational content that I will
consider is the teleosemantic approach, or TA for short. Like other externalist
views, this faces the difficulty of explaining how distal factors can influence a
subject’s recognition of experiential content. According to one version of this
view (Millikan 1989), the semantic content of a subject’s mental states depends
upon the role that those states play in suitably defined ‘normal’ conditions for
that organism. For example, a beaver’s splashing the water smartly with its tail
represents (for that animal) that there is danger, not because beavers routinely
splash their tails when there is danger (they also do this for all sorts of other
reasons), but because ‘only when it corresponds to danger does the instinctive
response to the splash on the part of the interpreter beavers, the consumers,
serve a purpose’ (ibid. 288). According to Millikan’s bio- or ‘consumer’
semantics, it is this purpose, which is typically linked to the biological and
evolutionary goals of the organism for survival, foraging, procreation, and so
on, that fixes the representational content of the resulting mental state, and not
the nature of the distal object or the organism itself. Whilst Millikan’s theory is
not specifically targeted at perceptual states, it may be naturally extended to
cover all forms of mental representation, including perception.
Chapter 6, Disambiguation Strategies
132
As discussed in chapter 5, Burge (2010) takes a similar view, emphasising
the evolutionary role or function of the representational states in question. For
example (ibid. 321):
[T]he specific range of attributes in the environment that an individual
perceptually represents … are constrained by causal interactions with the
environment, explained in ethology and zoology. The key interactions are
those that figured in molding [sic] the perceptual system shared by relevant
individuals.
Thus, according to Burge, it is to ethology and zoology that we must look to
settle the question of which properties are represented in perception.
Furthermore, ‘the relevant functional individual responses need not be by living
individuals’ (ibid.). Rather, the relevant facts are ones about the evolution of an
organism or genus that are common to all its individual members, and not facts
specific to the individual doing the representing. Both Millikan and Burge’s
views therefore constitute forms of ‘anti-individualism’ (Burge 1979), or
individuative externalism.
In each of the above cases, however, it is difficult to see how the distal
factors that fix representational content could enable a subject to grasp that
content, as Recognisability requires. The problem is particularly vivid in Burge’s
formulation, since the relevant facts may concern aspects of the evolutionary
history of an organism that are opaque to all but the most scientifically wellinformed observers — and even they can easily be wrong. However, even in
Millikan’s consumer semantics, there is no suggestion that an individual must
grasp, or be able to grasp, what their own mental states represent. Indeed, the
‘consumer’ of a perceptual representation may be a sub-personal cognitive
system, such as the system for bodily action, rather than the subject herself. 4
Consequently, such representational content need neither be conscious, nor
consciously available to the subject. Similarly, it need not be obvious to nor
recognisable by the subject what the normal function of a given representation
is. Rather, this will be a matter for empirical investigation and debate, and so
need not feature in the subject’s experience in the way that Recognisability
requires. By assigning a disambiguating role to distal causal and functional
factors, TA therefore makes it difficult to meet Recognisability, and so fails to
explain how the representational contents of experience are manifest in the
conscious experience of a perceiving subject.
As discussed in chapter 5, an advocate of TA might simply reject this
explanatory commitment, as per Burge (2010). This would fail to engage with
Travis since the resulting content no longer plays the same (or possibly any)
explanatory role in relation to the phenomenal character of conscious
experience. It is therefore not a representationalist view of experience as I define
this term.
4
Here Burge differs from Millikan, insisting that the biological functions that fix
representational content are ‘functions of the whole individual’ (Burge 2010: 320).
Chapter 6, Disambiguation Strategies
133
Alternatively, the representationalist might wish to deny that there is any
obstacle to a subject grasping the content of their own mental states on the basis
that they do not need to know the enabling conditions of a mental state in order
to know its content. Even if correct, however, this stipulation cannot explain
precisely how it is that a subject can grasp these contents, which is the main
point at issue. To do this, one might respond to the objection by appealing to
the kind of externalism about self-knowledge described below (6.3.2). On this
view, what individuates the representational contents of perceptual states also
individuates the subject’s reflective grasp or awareness of those contents,
thereby meeting the Recognisability constraint.
A third approach would be to argue that a subject’s grasp of the contents of
their experiences is at least partly conventional in that it is conditioned by their
past experience of the object and property types represented. This is a variation
of the background knowledge strategy (BKS) discussed below (6.4.2), which
claims that subjects are able to recognise the contents of their experience by
associating its phenomenal character, as determined by its content, with the
objects or properties represented therein. Such associations may themselves be
learned by experience or innate. Applying this strategy to TA, however, gives rise
to a convergence model of experience (6.2.1), since experiential contents would
be individuated and recognised by different means. This places constraints upon
the kinds of knowledge that may be brought to bear in recognising such
contents, along with the need to explain how BKS delivers contents that match
those delivered by TA, which are externally individuated by the subject’s
evolutionary history and environment.
6.3.2. Externalism about self-knowledge
Tyler Burge (1998) presents a strategy for explaining how a subject whose
mental states are externally individuated may nevertheless possess knowledge of
the contents of those states. His strategy can be summarised by the following
four claims:
(1)
Anti-individualism: first-order thoughts are at least partly
individuated by external physical objects and/or the shared
practices of a community of language users.
(2)
Privileged access: second-order thought has a reflexive, selfreferential character that is directly accessible to consciousness.
(3)
Enabling conditions: it is not necessary to know the enabling
conditions of a thought in order to know the contents of that
thought.
(4)
Redeployment: knowledge of one’s own thoughts ‘inherits’ its
empirical component from the concepts that those thoughts
employ.
Chapter 6, Disambiguation Strategies
134
The first of Burge’s claims is the thesis that mental content may be externally
individuated in various ways. He labels this anti-individualism in contrast to the
idea that a subject’s mental states constitutively depend only on the physical
state of the relevant individual, as well as to differentiate it from semantic
externalism in the philosophy of language. It is a form of individuative
externalism in the terminology introduced above (6.3.1). Whilst it lies outside
the scope of this chapter to discuss each of the above claims in full generality,
the representationalist may wish to employ a similar strategy to account for the
recognisability of p-representational content. Although this is not a move that
Burge himself makes, it is one that is available to advocates of externally
individuated content, such as McDowell (1994) and Tye (2007). The precise
method of individuation will not affect the present strategy, which is also
compatible with a teleo- or biosemantic approach (6.3.1).
Burge’s second claim posits a form of ‘second-order’ knowledge or
privileged access to one’s first-order mental states. This may be contrasted with
a perceptual model of self-knowledge according to which introspective thought
replicates the content of another mental state, thereby giving rise to a secondorder mental state with representational content distinct from the first-order
thought. This model of introspection is analogous to perception in that it
conceives of mental states as internal ‘objects’ that a thinker is able to inspect in
a way comparable to the sensory perception of external physical objects. Burge,
however, rejects the perceptual model, taking thoughts, beliefs, judgements, and
so on, about other thoughts to have a ‘reflexive, self-referential
character’ (Burge 1998: 659–60). On this view, the content of every secondorder thought about one’s own thinking contains a reference both to the subject
and to the first-order thought’s content. This enables the resulting belief state to
‘inherit’ the content of the first-order thought, rather than duplicating it, in such
a way that the two cannot come apart.
On this view, the content of the thought ‘I am thinking that there is some
water’ when I see water in front of me, for example, would have the following
structure:
(5)
I am thinking that there is some water
where the italicised proposition is not distinct from the first-order thought
thinking, but rather an instance of that thought itself.5 Thus the content of the
first-order thought (‘there is some water‘) is ‘inherited’ by the second-order
claim — or to borrow a metaphor from computer science, its content is passed
by reference rather than by value. This enables Burge to claim, contrary to
McKinsey (1991) et al., that we have a priori knowledge that we are thinking
about water without needing to know that there is H2O in our environment as
opposed to, for example, XYZ, as in Putnam’s (1975) ‘Twin Earth’ scenario.
5
This approach is similar to, though should be differentiated from, so-called higher-order
thought, or HOT, theories of consciousness (e.g. Rosenthal 1986), which employ a similar
self-referential structure to account for consciousness rather than self-knowledge.
Chapter 6, Disambiguation Strategies
135
Our water-thoughts about H2O are subjectively indistinguishable from
corresponding twin-water thoughts about XYZ (also called ‘water’ on Twin
Earth). However, the very fact that individuates our thoughts as being about
water — i.e. the presence of H2O in our physical environment — also
individuates our first-personal knowledge that it is water about which we are
thinking. Thus the first-personal awareness of the contents of thought can be
said to ‘track’ the contents of those very thoughts without requiring any
knowledge of the presuppositions or ‘enabling conditions’ (Burge 1988: 654) for
thinking about water, such as ‘water is H2O’ or ‘there is an external world’ —
Burge’s third claim.
According to Burge’s fourth and final maxim, the empirical component of
our second-order beliefs, etc., derives from the concepts that they employ rather
than any evidential connection between first- and second-order thoughts. Such
empirical content is therefore common to the subject’s first-order thought and
their second-order awareness of it, since both employ the same concepts. This is
what Peacocke (1996: 131) refers to as the Redeployment Claim and follows
from second-order thoughts inheriting their contents from first-order thoughts.
Thus we can know that we are thinking about water, where this term picks out
some externally individuated natural kind, in virtue of possessing the concept
water without having to know precisely which natural kind, i.e. H2O or XYZ,
our concept picks out.
Burge’s ESK strategy may be adapted by representationalists to explain how
content that is externally individuated is nevertheless available or recognisable
to the subject. In place of first-order beliefs we have perceptual experiences with
externally individuated content. In place of the second-order knowledge of
thoughts, we have the grasp, awareness or knowledge of perceptual content.
The resulting account would therefore claim that content recognised on the
basis of perceptual experience is ‘inherited’ directly from the experiential state.
Unlike on Burge’s account, however, this cannot be due to the relevant secondorder states having a reflexive, self-referential character since the corresponding
first-order states are experiences, not beliefs. Consequently, self-knowledge of
perceptual content cannot be self-verifying in the same way as self-knowledge of
thought since to think an experiential content is not thereby to perceptually
experience it. Nevertheless, the empirical content of such states could still derive
from the concepts that they employ. This approach is conducive to theorists like
McDowell (1994) and Brewer (1999; 2005) who argue that the representational
content of perceptual states must be conceptual in order to yield knowledge of
the world.6 However, even non-conceptual content theorists might apply a
similar strategy provided that their account posits some kind of constitutive
representational elements that are common to both p-representational contents
and our reflective grasp of them, and from which the relevant empirical
components may be derived.
6
Brewer has since abandoned conceptualism in favour of a relational view of perceptual
experience. McDowell (2008) still endorses conceptualism but no longer requires that the
contents of perceptual experience be propositional.
Chapter 6, Disambiguation Strategies
136
Adapting Burge’s maxims to the case of perceptual experience, then, yields
the following claims:
(1′)
Anti-individualism: perceptual experiences are at least partly
individuated by external physical objects and/or the evolutionary
history and representational capacities of the genus.
(2′)
Privileged access: perceptual belief has an indexical nature that is
directly accessible to consciousness.
(3′)
Enabling conditions: it is not necessary to know the enabling
conditions of a perceptual experience in order to grasp the
content of that experience.
(4′)
Redeployment: one’s grasp of the content of one’s own
perceptual experiences ‘inherits’ its empirical component from
the concepts or non-conceptual elements that those experiences
employ.
Here, the relevant enabling conditions in (3′) would be knowledge of the
relevant content types for perceptual experience, which under ESK will not be
required in order for the subject to grasp that they are in that state.
A first approximation of such a view might be modelled on the notion of
demonstrative reference. One way of thinking about ESK about experience
would be that the relevant second-order beliefs have demonstrative contents. On
this view, the content of a second-order thought about seeing a red tomato
would be of the form:
(6)
I am seeing that: there is a red tomato
where the italicised proposition is the content of a visual experience that is
demonstratively referenced by the second-order thought. This quotational model
of perceptual belief would, however, make the object of such beliefs a
proposition, which is — at least on some accounts of what it is to be a
proposition — an abstract object. This raises problems with indirect senses, and
does not seem a plausible account of experiential self-knowledge.
A more sophisticated view would be to take the content of the resulting
belief to incorporate demonstrative terms. For example:
(7)
That (there) is thus.7
Here, the sense of these demonstrative elements would be constrained by the
content of experience such that their reference is fixed by whatever objects and
properties the experience itself picks out — in the above case a red tomato. This
enables the representationalist to circumvent Travis’s argument from
7
Cf. Brewer 1999: 185; 6.4.1.
Chapter 6, Disambiguation Strategies
137
recognisability by giving an externalist account of self-knowledge according to
which p-representational contents and the first-personal awareness of them are
individuated by the same external factors; i.e. a transmission model of
experience. Thus it would be possible for a subject to know that they are, for
example, representing a tomato (round object, red, etc.) when they see one (or a
wax tomato, tomato-shaped ketchup bottle, etc.) on the basis that the content of
their second-order awareness ‘inherits’ the externally individuated content of the
first-order experience.
Note that this need not mean that subjects simply represent whatever objects
happen to be in front of them. Rather, depending upon the method of external
individuation employed, p-representational content may be determined by the
creature’s evolutionary history or representational capabilities. On this view
there is no need for the subject to be able to consciously disambiguate between
the multiple possible contents of experience. Rather, according to the above
version of ESK, their perceptual system does it for them. This then carries
through, or transmits, to any subsequent awareness, knowledge or belief about
the perceptual state itself, yielding precisely the same content. Thus, a secondorder belief about the perceptual experience in (7) would have the content:
(8)
I am seeing that: that (there) is thus.
Just as one’s water-thoughts track the presence of H2O as opposed to XYZ,
one’s tomato- (round-, red-, etc.) perceptions would track one’s capacity to
represent the presence of tomatoes (spheroids, red, etc.) in one’s perceptual
environment. Thus, Travis’s Face-value and Recognisability conditions may be
met by an ESK-based account of p-representation.
The application of ESK to the recognition of perceptual content has a
number of important consequences. First, it entails that a subject’s reflective
awareness or grasp of the face-value content of experience involves a secondorder thought with demonstrative or indexical content. This means that firstorder perceptual awareness and second-order reflective awareness of prepresentational content are, according to ESK, separate and distinct mental
states.8 Furthermore, ESK requires the resulting contents to be either
conceptually structured or concept-like. Although Recognisability does not
require that subjects are able to state the relevant content or accuracy conditions
for their experience (5.2.1), the possession of such concepts would presumably
be required in order for ESK to be capable of delivering the relevant empirical
content in accordance with Redeployment (4′). Such concept possession would
in turn require either (i) the subject to have had prior experience of the types
and entities that their experiences represent — shapes and colours, for example
— in order to acquire the relevant concept, or else (ii) that the grasp of such
concepts is innate. The precise conceptual apparatus required would depend
upon the relevant form of external individuation, but will involve some form of
recognitional concepts, i.e. concepts that are essentially tied to the recognition of
8
Note that the first-order state may, though need not, be conscious.
Chapter 6, Disambiguation Strategies
138
perceptual stimuli, with similar considerations applying in the case of nonconceptual content.
ESK, then, accounts for the way in which subjects are able to grasp the
contents of their perceptual experiences — for example, that is φ, where φ
corresponds to some perceptible property — on the basis of their recognitional
concepts, but without requiring them to be able to articulate the resulting
contents or differentiate them from cases that may be subjectively
indistinguishable; e.g. that is ψ. This is directly analogous to the external
individuation of thoughts, where there is some water (i.e. H2O) and there is
some twater (i.e. XYZ) may be indistinguishable from the thinker’s own
perspective. Despite not being in a position to know which of these thoughts he
is thinking, the subject can nevertheless know precisely what he is thinking due
to the second-order thought ‘inheriting’, or being constrained by, the first-order
content. In the case of perception, then, a subject would not need to know or
recognise whether their experience represents there to be a tomato, a bulbous
red spheroid or a wax imitation lemon in order to grasp what it would take for
that experience to be veridical. Rather, according to ESK, they are capable of
recognising that things would need to be like that in accordance with whatever
demonstrative concepts or concept-like elements are employed by their firstorder experience.
Whilst constituting a direct response to Travis, the form of recognisability
delivered by ESK therefore constitutes something of a limited victory for
representationalism. The sense in which subjects are able to grasp or know the
contents of their experience is tempered by their inability to discern which of a
range of properties is represented therein. Furthermore, if the resulting thoughts
are of the kind described by (7) and (8), it remains to be explained how such
demonstrative content plays a role in justifying or fixing the content of ordinary
beliefs such as ‘there is a tomato’ or ‘the tomato is red’. At the very least, the
relationship between the content of experience and the content of belief is not as
straightforward as it might first appear. Furthermore, since all perceptual beliefs
will take a similar form to (7), this content alone does not explain what makes
certain beliefs subjectively distinguishable and others not. If, on the other hand,
the relevant second-order thoughts or awareness are not of this kind, then it is
incumbent upon the representationalist to give an account of them. Thus, whilst
ESK offers a way of meeting Recognisability, this strategy is not without its
costs.
6.3.3. Phenomenal externalism
A further consequence of ESK is that if, as suggested above (5.2.4), we
understand representationalism as a thesis about the phenomenal character of
experience, this raises the possibility of the external individuation of
phenomenal character. The combination of intentionalism and ESK predicts that
not only p-representation, but also phenomenal character will be externally
individuated, since the latter is identical to, or supervenes upon, the former
(5.2.3). According to this position, which I will call phenomenal externalism,
Chapter 6, Disambiguation Strategies
139
what experience ‘is like’ for the subject will depend not only upon that
individual’s internal state, but the states of objects in the world. Thus the
experiences of two exact physical duplicates could have differing phenomenal
characters simply in virtue of the properties of the objects that they perceive
despite having identical patterns of activity throughout their brains and nervous
systems. Many philosophers find this consequence of phenomenal externalism
to be problematic and counterintuitive, though it is contentious whether any of
the arguments against it are decisive (see Lycan 2008, esp. §§4.4–4.5, for a
representative sample). Phenomenal character, it is tempting to think — at least
on the traditional conception of this term as picking out what is distinguishable
by the subject — must surely be internally individuated.9
There are two basic ways of responding to this issue. The first is to deny
that it constitutes a real problem for the representationalist. On this view, in
accepting ESK we must embrace an externalist notion of the mind and mental
states upon which the external individuation of phenomenal character is no
more problematic than, for example, the external individuation of thought and
knowledge. This is Michael Tye’s (1995; 2002) view, who argues that
phenomenal character is externally individuated in precisely this manner and
that any philosophical intuitions to the contrary are based on a mistaken
‘Cartesian conception’ of the mind. Alternatively, one could deny that those
aspects of representational content upon which phenomenal character supervene
are in fact externally individuated. This gives rise to the existence of ‘broad’ and
‘narrow’ contents according to which phenomenal character only supervenes
upon the internal aspects of representational content (cf. Schellenberg 2011a,
2011b; Tye 2007). This again raises questions concerning how Recognisability
can be met given the difference in individuation conditions between broad and
narrow contents, which effectively rule out the use of the ESK strategy.
It is also worth noting the parallel between phenomenal externalism and
Naïve Realism, according to which external objects and their properties are
considered to be partially constitutive of phenomenal character. These positions
differ in that the former is a thesis about what individuates rather than what
constitutes phenomenal character (in the metaphysical sense), as in the case of
Naïve Realism. Nevertheless, there is a degree of convergence and overlap
between the two theories. This in turn raises the question of whether we should
account for phenomenal character in purely internal terms. If not, this reduces
the grounds upon which the representationalist can object to Naïve Realism on
the basis that external objects determine phenomenal character, since their own
theory predicts the same, albeit with important metaphysical qualifications.
Indeed, if the relevant contents are demonstratively individuated, as they are for
Burge and Tye, then there is a sense in which the resulting representationalism is
just a form of relationalism (i.e. weak relationalism) in that the subject’s relation
to the external objects of perception plays a fundamental role in explaining the
9
This is a particular problem for the kind of demonstrative content posited by McDowell
and Brewer, which may vary according to the objects and properties the subject is presented
with (6.4.1).
Chapter 6, Disambiguation Strategies
140
nature and role of perceptual experiential. This yields a kind of compatibilism
that combines elements of both relationalism and representationalism within a
single hybrid view. Of course, such views will differ in how they account for
perceptual illusions and hallucinations. But as discussed in 2.3.1 and 3.2.1, this
need not constitute a knock-down objection to relationalism.
6.3.4. Asymmetric dependency
The final externalist approach that I wish to consider uses counterfactual
dependences to determine which of a range of possible, potentially disjunctive,
representational contents is present in any given case. Jerry Fodor’s (1987: 101–
10) solution to this ‘disjunction problem’, as he calls it, involves the
counterfactual dependence of, for example, a frog’s visually representing flies —
a potential food source — upon the presence of flies in its environment, as
opposed to, say, black dots. According to Fodor, the frog can be said to be
representing flies as opposed to the disjunction flies or black dots on the basis
that it would not represent black dots as a token of this type if there were no
flies in its environment. The converse claim — that it would not not represent
flies as a token of this type if there were no black dots in its environment — is,
however, false since flies are the frog’s food source and so the primary target of
its representational abilities. Consequently, the frog’s representation is
asymmetrically dependent upon the existence of flies, but not upon the existence
of black dots. Thus the frog may be said to represent the former, and not the
disjunction of flies and black dots.
Whilst this may present a plausible method of individuating prepresentational content from a third-person point of view — something upon
which many theorists would disagree (e.g. Burge 2010: 322; 307–8, fn. 27) —
the truth of the relevant counterfactuals are not obviously available to subjects
in experience. Indeed, given the contentious nature of modal facts in philosophy,
it is debatable whether they are even available to philosophers. It therefore
seems implausible to hold that in order to know what one’s perceptual
experience represents, one must have access to facts about which counterfactual
conditionals would hold in the absence of the thing that is ostensibly
represented; e.g. if there were no lemons then I would not represent a wax
lemon as a token of the same type as a real lemon. To suppose this — which
Fodor himself does not, since he does not endorse Recognisability — would
hugely over-intellectualise the notion of p-representation, the contents of which
would be inaccessible to all but the most philosophically sophisticated
perceivers. Thus, even if it constitutes a viable account of the representational
content of perceptual states, asymmetric dependency cannot form the basis for
the content of perceptual experiences whose content is recognisable to the
subject.
6.4. The Capacities Approach
The remainder of the candidate representationalist strategies for meeting
Recognisability I wish to consider highlight the role of perceptual capacities.
Chapter 6, Disambiguation Strategies
141
These range from recognitional and conceptual capacities (6.4.1) to the role of
background knowledge (6.4.2), other discriminatory capacities (6.4.3) and socalled ‘intuitional’ or non-propositional content (6.4.4). Each of these draws
upon some form of perceptual discriminatory capacity which forms the basis for
representational content that is recognisable to the subject. Whilst not
exclusively internalist in nature, I argue that such approaches constitute the
representationalist’s best opportunity to respond to Travis whilst avoiding
phenomenal externalism. Such approaches, however, do not necessarily favour
representationalism over relationalism, and so do not in themselves constitute
an argument for representationalism per se (6.5).
6.4.1. Demonstrative content
The notion of demonstrative content is introduced by McDowell (1994) and
developed by Brewer (1999) as a way of explaining how propositionally
structured perceptual contents can provide reasons for belief and action.
According to the latter account, such contents have the structure ‘that (there) is
thus’, where the first demonstrative refers to some externally individuated object
and the last to some determinate property-type that is predicated of that object
(cf. 6.3.2). Call this the demonstrative content strategy, or DCS for short.
According to McDowell (1994: 57), demonstrative concepts include potentially
short-lived capacities to pick out or ‘track’ the relevant object or property over
time, as well as more stable conceptual capacities as these are traditionally
envisaged. Demonstrative concepts may also be arbitrarily fine-grained so that
each colour shade or object-type corresponds to a different ‘concept’. Both
McDowell (ibid. Lecture I) and Brewer (1999: ch. 5) argue for a fully conceptual
account of representational content for reasons connected with the justification
of knowledge claims. This does not mean, however, that subject need possess
fully-fledged concepts for every object- or property-type that they represent, but
merely what I will refer to as the discriminatory capacity to identify them
throughout a given perceptual episode, with demonstrative ‘concepts’ being one
example of such capacities.
Since discriminatory capacities are general, i.e. they can apply to different
objects or property-instances, there is nothing to stop them featuring not only in
the representational contents of experience, but also in the subject’s firstpersonal beliefs or knowledge of that content. When seeing a red cube, for
example, the demonstrative ‘that’ might refer to the cube itself and the ‘thus’ its
colour — in this case a particular shade of red. Consequently, even a subject who
lacked the concept cube would arguably be able to recognise that her experience
was of that object, where the demonstrative picks out the relevant cubical
object. Such an experience would yield a different content, and so different
accuracy conditions, to the experience of an otherwise indistinguishable but
numerically distinct cube presented on a different occasion. Furthermore, this
difference is in some sense ‘recognisable’ to the subject in that on each occasion
it is manifest to them that they are representing that particular object, and not
merely some object that falls under a general kind; e.g. a cube. Thus a subject
Chapter 6, Disambiguation Strategies
142
who enjoys a visual experience with the content ‘that (there) is thus’ may, by
reflecting upon their experience, grasp that they are representing that spatially
bounded object thusly without requiring any means of individuating the relevant
object or property other than by demonstrative reference. In this way, the
‘particular’ (i.e. object-dependent) content of a visual experience, and not just its
‘general’ content, may be said to be recognisable to the subject solely in virtue of
the experience itself, as Recognisability requires, by means of the experience’s
phenomenal character, which is demonstrative in nature.
This form of DCS, then, would appear to be a candidate for a transmission
model of experience that is capable of meeting Recognisability and so refuting
Travis. Moreover, it is a view upon which different contents may be present in
subjectively indistinguishable (e.g. veridical and non-veridical) experiences, since
the identity of the object will affect the experience’s accuracy conditions, though
not its subjective character. This avoids Travis’s argument from unmediated
awareness (2.3.4) and similarly circumvents the argument from recognisability
(2.3.3), since the resulting contents are conveyed to the subject in experience.
The difficulty with this view, however, lies not in the structure of the
representational contents that it posits, but in the individuation of the relevant
objects and properties. In order to represent that ‘that (there) is thus’, it must be
possible for the perceptual system to single out both a particular object (‘that
there’) and a property-type (‘thus’) which is predicated of that object. A Travisstyle indeterminacy objection to the view would therefore be that it is not
recognisable from experience alone which object and property-type are thereby
individuated since there are numerous different and incompatible ways of
partitioning the world into objects and properties.10 The response that it is that
object or property — namely, the one demonstrated in experience — is unhelpful
as it fails to settle the question of how the relevant terms are individuated. In the
case described above, for example, it is the cube and its colour that are so
identified, rather than, say, the shape of its visible surfaces. The representational
contents of a cube-colour experience and a cube face-shape experience are
presumably different despite both falling under the same general description of
‘that (there) is thus’. Indeed, on a purely demonstrative view, every perception
will fall under this general schema, or some complex conjunction of terms of
this form. There is therefore still a case to be answered by the DCS theorist in
order to avoid falling prey to the indeterminacy objection, since the relevant
contents are as yet indeterminate between multiple sets of accuracy conditions,
depending upon the types of object and property that are demonstrated.
In practice, a suitable response to this problem must lie in the way that the
demonstrated objects and properties are perceptually individuated. To exercise a
visual discriminatory capacity for, say, a cube, the subject’s perceptual
apparatus must pick out and track the spatial extent and boundaries of that
object over time. Similarly, if it is the object’s colour that is being tracked as
opposed to its shape, then there must be some corresponding capacity that is
perceptually attuned to the particular colour of the object over time. Where
10
For an externalist solution to this objection, see 6.3.2.
Chapter 6, Disambiguation Strategies
143
multiple properties are being represented then many such capacities will be
active.
Moreover, there must be some definite answer to which discriminatory
capacities are active during any given perceptual episode. The answer to this
question will in turn fix the contents of p-representation, thereby resolving the
indeterminacy objection. To do this, the DCS theorist must appeal to further
facts about the individuation or activation of the relevant discriminatory or
conceptual capacities. Since the same (or indistinguishably similar) appearances
can be shared by multiple object types, the activation of a particular capacity
must be explained in terms of the perceptual abilities or conceptual repertoire of
the subject. This makes DCS a form of the discriminatory capacities approach
described below since the employment of such capacities on particular occasions
is what individuates p-representational content (6.4.3). Furthermore, the subject
must be capable of grasping which specific capacities are being employed in
order for the resulting content to be recognisable. The precise methods by which
this may be done are discussed in the following sections.
A variation upon the above response to Travis’s indeterminacy objection
would be to link phenomenal character to some other feature of experience,
such as the exercise of concepts or recognitional capacities, as per Schellenberg
(2011b). On this view, although the particular contents of experiences are
externally individuated, their phenomenal character is constituted by the
employment of concepts or conceptual capacities, rather than supervening
directly on its content. On this view, seeing a lemon under typical circumstances
would activate the concept is a lemon, thereby making it apparent to the subject
that they are representing a lemon, as opposed to, say, a lemon-shaped bar of
soap. This would be true even if the object were a lemon-shaped bar of soap
that looked suitably like an actual lemon, since in both cases the same concept
or recognitional capacity will be activated. It is unclear, however, that this is
more than a notational variant upon the standard intentionalist view in which
phenomenal character supervenes upon or is identical to the relevant contents.
Moreover, it is unclear why the exercise of such recognitional capacities should
result in the tokening of content. Indeed, it is debatable to what extent such
views need posit general contents at all since, according to them, phenomenal
character is a mere by-product of the operation of discriminatory capacities and
so not necessarily represented in the contents of experience as such.
6.4.2. Background knowledge
The background knowledge strategy, or BKS, is a variation upon the
discriminatory capacities approach that appeals to the subject’s explicit or tacit
knowledge of the correlation between their sensory impressions and object- or
property-types. For example, a subject who has visually experienced a number
of yellow-looking objects, the majority of which, upon inspection, in fact turned
out to be yellow rather than merely appearing to be, could plausibly come to
know what it is to see an object as yellow.11 If these objects become their
11
Cf. Jackson’s (1982) Mary.
Chapter 6, Disambiguation Strategies
144
paradigms — or ‘samples’, to use Wittgenstein’s (1953: §16) terminology — of
what it is to look yellow, then the subject can plausibly come to know that an
experience represents some novel object to be yellow by means of its similarity
to these paradigms. Such forms of background knowledge may be considered
‘external’ to perceptual experience in that the subject does not recognise an
experience to represent yellow in virtue of that experience alone. Rather, by
employing explicit (i.e. inferential) or tacit (non-inferential) knowledge of the
link between subjective appearances and the objective characteristics of objects,
a subject can come to recognise their experience as representing that an object is
yellow as opposed to merely looking yellow or some other way.
To take a simple example, the yellowish, roundish appearance of a lemon
may be taken to represent that there is something yellow, round, and so on in
front of one because one knows (associates, has previous experience of, …)
objects with this appearance as being yellow, round, lemons, etc. Such
knowledge, if it can be described as such, will consist in associations between
experiences of the relevant appearance types and the actual properties of objects
which, upon further investigation, turned out to instantiate the properties that
one comes to associate with the corresponding appearances; i.e. being yellow,
round, a lemon, etc.12 Thus a form of Recognisability may be derived on the
basis of the tacit knowledge of appearances.
By acquiring knowledge of the appearances of yellow objects, the subject
thereby acquires the capacity to recognise not only yellow objects, but the
representation of yellow in perceptual experience. On this view, what privileges
being yellow as the content of such representation — as opposed to, say, being
white and illuminated by yellow light — is its role in the relevant paradigm
cases. Typically, paradigm cases of yellow objects — a ripe lemon or banana, for
example — are in fact yellow and can be seen to be such under a wide variety of
environmental conditions.13 Thus whilst a white object, such as a white piece of
chalk illuminated by yellow light, shares the same yellowish appearance and so
‘looks yellow’, it is not paradigmatic of that appearance. Consequently, on the
present view, objects that look this way are represented in perceptual experience
as being yellow in virtue of their possessing a similar appearance to (i.e. they
look like) certain paradigm objects that are yellow. That is, under BKS, just what
it means to look yellow.
It is worth noting that something not unlike this kind of knowledge is
present in Martin’s comparative analysis of looks-statements in the form of the
similarity function (4.3.1). This takes a predicate (e.g. is yellow) and a look,
yielding truth if and only if the latter is relevantly similar to the look of those
objects that satisfy the predicate. Crucially, however, on Martin’s account this
12
Note that the application of such knowledge need not render access to representational
contents epistemically indirect since the relevant capacities may operate at a sub-personal
level rather than involving conscious inference, at least once the learning phase is complete.
13
Note, however, that at least one paradigm case of a yellow object — namely, the sun — is
not in fact yellow, but white; it merely appears yellow due to the effect of atmospheric
absorption (Wilk 2009).
Chapter 6, Disambiguation Strategies
145
need not be part of a psychological process that the subject goes through in
identifying particular looks. Rather, it is part of the semantic analysis of looksstatements, and so not necessarily even implicitly known by the subject. If is
therefore an important question for the BKS theorist whether subjects do in fact
possess such knowledge of appearances and whether it is acquired or innate. If
they do, then there seems to be no reason why this could not be applied at the
level of recognising the content of perceptual experience on the basis of
appearances.
According to BKS, then, a subject’s background knowledge of certain
paradigm cases of looks may be used to disambiguate p-representational
contents. However, in order for this to be more than a mere convention — i.e. a
way of reporting appearances, rather than constitutive of p-representation itself
— it must be explained how this knowledge comes to figure in the content of
perceptual experience in the first place. For this to be the case, such knowledge
must in some sense be exercised by the perception of objects, whether
consciously or otherwise. The exercise of tacit or explicit background
knowledge thus becomes equivalent to the possession of a conceptual,
recognitional or discriminatory capacity for the relevant object- or propertytype. When an object with the relevant appearance is perceived, this activates
the subject’s background knowledge, causing the corresponding type to be
tokened as part of that experience’s representational content. BKS is therefore
effectively a special case of the discriminatory capacity strategy described below.
6.4.3. Discriminatory capacities
A number of the above approaches — specifically DCS (6.4.1), BKS (6.4.2) and
certain forms of TA (6.3.1) — can be reduced to what I will call the capacities
approach, or CA for short. According to CA, what gives experience univocal
content is the exercise of conceptual, recognitional or discriminatory capacities,
each of which corresponds to a particular type of object or property that may be
represented. Since I take it that conceptual and recognitional capacities are
themselves forms of discriminatory capacity, I will prefer the latter term on the
grounds that it does not entail the existence of conceptual content (not all
representationalists are conceptualists), or the successful recognition of
perceptual stimuli (discriminatory capacities may misfire). However, variants of
the present approach may be constructed for each of these and other types of
capacity accordingly. Thus, if a given experience invokes the discriminatory
capacity for lemons (being yellow, ellipsoid, etc.), then the resulting content of
experience will represent a lemon (yellow, ellipsoid, etc.). This content is in turn
available to the subject in virtue of its employing a particular concept, or nonconceptual structure, that features in any second-order awareness of that
Chapter 6, Disambiguation Strategies
146
content — for example, ‘my experience is as of a lemon’, where ‘as of’ is not
taken to entail the presence of lemons.14
This becomes important when we consider a subjectively matching
hallucinatory experience of — to continue the above example — a lemon. In this
case, according to CA, the same discriminatory capacity will be operative, in this
case erroneously, causing the experience to have the same or subjectively
indistinguishable phenomenal character, which in turn supervenes upon, or is
correlated with, the discriminatory capacities that are employed. However, since
in this case there is no such object (colour, shape, etc.), the resulting
representation is inaccurate, yielding a false or non-veridical content. Thus CA
seeks to accommodate both (a) the sense in which veridical and non-veridical
contents share the same phenomenal character in virtue of the discriminatory
capacities that they employ, and (b) how the content of experience is
recognisable to the subject in virtue of the very same capacities that give rise to
its content. CA is therefore a transmission model of experience (6.2.1) that
adheres to the modified recognisability condition set out in the previous chapter
(5.2.3).
Several features of CA are worth noting. First, in order to deliver univocal
face-value content, the operation of the relevant capacities must not be
underdetermined by the perceptual stimuli, since if the same stimuli were to
trigger multiple capacities whose representational contents conflict, then this
would yield multiple incompatible contents. This does not, however, prevent
multiple compatible capacities from being invoked by the same perceptual
stimuli. To take McDowell’s (2008: 3) example, seeing a red cardinal might
invoke discriminatory capacities for bird, animal and cardinal, as well as for red
and numerous other properties. If the same perceptual stimuli were to invoke
the recognitional capacities for cardinal, crimson rosella, imitation wax
cardinal, and so on, however, then the resulting content would be incoherent
and so necessarily false, since these properties are incompatible with one
another (a cardinal is not a crimson rosella, nor is it made of wax). This does
not mean that there can be no cases of incoherent or necessarily false contents.
Rather, such cases cannot be veridical or typical of experience in general
without rendering CA an error theory of perception. To avoid this, independent
justification that discriminatory capacities do not work in this way, or some
limit to their scope and complexity, must be given if the threat of indeterminacy
is to be avoided (see below).
The second notable feature of CA is that the particular (as opposed to
general) contents of experiences will necessarily outstrip what is recognisable on
the basis of discriminatory capacities alone. Such abilities are limited to what is
subjectively distinguishable in experience, and so cannot be used to support the
14
I take CA itself to be neutral between conceptual and non-conceptual approaches to
perceptual content. Given the role of the resulting elements in the recognition of content, one
might take the resulting view to be essentially conceptual in nature, which is McDowell
(1994; 2008) and Brewer’s (1999) position. For simplicity, in what follows I will therefore
speak only of conceptual content, though this should not be taken to rule out the possibility
of a parallel non-conceptual version of CA.
Chapter 6, Disambiguation Strategies
147
recognisability of particular aspects of experience. They may, however, generate
particular contents that include, for example, demonstrative references to
external objects and property-instances. Such references are not directly
discriminable from experience, but arguably do contribute to the phenomenal
character of experience in the sense that experience is presented as being an
experience of particulars, and not merely general types. CA theorists, such as
Schellenberg (2011b), would therefore do well to insist that the particularity of
experience is manifest in perceptual phenomenology, which is not distinct from
its particular content.
By combining CA with DCS (6.3.1) in this way, it is possible to explain how a
subject is able to grasp both the general content of their experience, based upon
the discriminatory capacities it employs, and its particular content, based on the
subject’s ability to identify and track objects in their perceptual environment. As
with DCS, the indexical character of particular content, e.g. ‘that cardinal (there)
is thusly red’, allows the subject’s grasp of demonstrative reference to ‘transmit’
to the relevant the self-knowledge claims; e.g. ‘I see a red cardinal’. Thus, by
employing a combination of discriminatory capacities and demonstrative
content, a potential response to Travis’s indeterminacy objection may be
constructed.
According to CA, then, the operation of perceptual discriminatory capacities
results in the tokening of concepts or corresponding non-conceptual structures
in the content of experience. This in turn raises further questions about the
nature of these capacities and contents. Why, for instance, should the operation
of discriminatory capacities yield representational content at all? And why
should we think of such capacities as being active in experience as opposed to
an anti-representational account of experience in which experiences give rise to
thoughts in which such capacities are active? In effect, CA assimilates perceptual
experience to a form of thought in which the exercise of discriminatory
capacities yields structured representational contents. However, whether
experience is assimilable in this way is part of what is in dispute. Whatever
account of discriminatory capacities is given by the representationalist at the
level of experience can be paralleled by the anti-representationalist in their
account of how experience gives rise to thought. Just as hallucinations and
illusions, in which representationalists posit the existence of false or nonveridical contents, can be explained by anti-representationalists in terms of the
misinterpretation or misjudgement of perceptual stimuli (2.3.1, 3.2.1), the role
of discriminatory capacities can equally be applied at the level of thought or
judgement. So even if CA constitutes a satisfactory response to Travis (2004), it
remains to be seen why this should be considered a better account of experience
than the corresponding anti-representationalist position.
In practice, one of the primary motivations for representationalism, and
specifically conceptualism, about experience comes from epistemological
concerns, such as the need to provide reasons for beliefs and judgements, or as
an explanation of hallucination and illusion. Given the Davidsonian dictum that
only beliefs, which are paradigmatically both representational and conceptual,
can constitute such reasons, this drives some theorists such as McDowell and
Chapter 6, Disambiguation Strategies
148
early Brewer to posit the existence of belief-like, i.e. conceptual, prepresentational contents. However, whether this motivation is sufficient to
establish the question of whether experience is representational or not is itself
highly contentious. Such considerations shift the debate away from the type of
representations involved in experience, or indeed whether it has any, to what the
explanatory role of such representations should be. Providing reasons or
justification for judgements, beliefs, and ultimately knowledge is just one such
role. Explaining the phenomenal character of experience, or what perceptions
and illusions have in common, is another. Accounting for the behaviour and
perceptual constancies in biological organisms is yet another. To conflate these
explanatory roles, or to assume that they are all played by a single notion or
type of representational content, runs together several different issues in a way
that is antithetical to philosophical clarity. It is therefore vital that
representationalists specify precisely what role, or roles, the contents they posit
are supposed to play, and why this constitutes a better explanation of the
relevant phenomena than the alternative anti-representationalist account.
For the reasons given above, it is insufficient for representationalists to
merely prove the existence of the relevant discriminatory capacities, since this
eventuality is compatible with both REP and REL. Rather, it is the role played by
the resulting representational content that must be justified. This leaves room
for accounts of perceptual experience in which discriminatory capacities are
active and may even be described in representational terms, but where the
resulting account of experience is primarily relational, or is neutral between
these two views. Again, the decisive factor in the debate will not be the fact that
a representational description of experience is possible, but rather the
explanatory role of that description in the explanation of thought and behaviour
more generally. Even supposing that the existence of the kind of discriminatory
capacities posited by CA were granted, this leaves a considerable amount of
justificatory work to be done in order to show that experience is fundamentally
representational. Moreover, if the operation of the relevant capacities
themselves turns out to be relationally individuated, or partly constituted by the
subject’s relation to the objects of perception, then the resulting view would
itself qualify as a form of relationalism since the operation of such capacities
would, on this view, form an essential part of perceptual experience.
6.4.4. Intuitional content
It is an interesting (though perhaps unsurprising in light of the above) historical
fact that two of the leading proponents of CA — namely McDowell and Brewer
— have since abandoned the view in favour of some form of relationalism.
Brewer (2011) now endorses a version of Campbell’s (2002) ‘Relational View’
and McDowell now rejects the idea that representational content is
propositional, though he still takes it to be conceptual — a notion he calls
‘intuitional content’ (McDowell 2008: 6).15 The precise reasons for this change
15
Burge (2010: 381) also holds a non-propositional view, but advocates a form of nonconceptual content.
Chapter 6, Disambiguation Strategies
149
are unclear, but in the case of McDowell it seems likely that it concerns the
difficulty of determining precisely which discriminatory capacities are active in
any given experience, along with the number of such capacities that would be
required to justify the range of true judgements that can result from experience
(McDowell 2008: 3–4). This change in the role of the discriminatory capacities
means that, for example, the concept cardinal need not feature in the content of
a subject’s experience in order for it to justify the judgement that there is a
cardinal. Rather, the recognitional capacities for red, bird plus certain shape and
colour concepts may be operative, with the act of judgement combining these
conceptual elements in the appropriate way to derive the content of the resulting
belief. On this revised view, the question of precisely which capacities are
operative in any given experience will presumably be an empirical one.
However, breaking the link between p-representational contents of experience
and the contents of beliefs or judgements brings McDowell a step closer to a
relational account of experience.
If the distinction between the capacities that are operative in experience and
those that are operative in judgement is to be a clear and principled one, then
further justification for this distinction must be given. Moreover, since
McDowell also admits the existence of discriminatory capacities at the level of
judgement, then why should all such capacities not work in this way, rather
than some being operative at the level of experience and others only in
judgement? If this were the case, however, then the resulting account would no
longer be representational, since this would leaves no role for representational
contents to play given that phenomenal character can already be explained in
terms of the employment of the relevant capacities.
Furthermore, it is unclear that McDowell’s ‘intuitional content’ even
constitutes a form of content in the philosophical sense since, being nonpropositional, it has no determinate accuracy conditions. Rather, McDowell’s
claim appears to be that it is a feature of experiences that they are fully
conceptualisable, or conceptually structured, rather than that they possess
representational contents per se. However, this is no grounds for disagreement
with the relationalist, who also holds that experiences are conceptualisable in
the formation of beliefs and judgement. It is therefore unclear that McDowell
(2008) should be taken to be offering a defence of representationalism as
opposed to simply introducing an alternative non-representational notion of
‘content’. Indeed, in the terms I have been using here, once one drops the claim
that p-representations are propositional, the resulting view more closely
resembles a form of relationalism than it does representationalism.
6.5. Conclusion
Of the views considered above, two possible strategies emerged as potential
responses to Travis. Each approach seeks to disambiguate the ‘face-value’
content of an experience from its subjectively indistinguishable alternatives in
order to avoid Travis’s indeterminacy objection. The teleosemantic approach
(TA) holds that the contents of experience are externally individuated by distal
Chapter 6, Disambiguation Strategies
150
causal relations, such as the evolutionary history or biological function of the
relevant organism’s genus. Since, by their nature, such factors are not directly
accessible to the perceiving subject, the ability to recognise the content of
experience must be explained by some further means, such as the external
individuation of knowledge (ESK) or the application of background knowledge
(BKS). That Travis’s arguments push the representationalist towards the external
individuation of content is interesting in itself, and has potential ramifications
for their account of phenomenal character; i.e. phenomenal externalism. The
demonstrative content strategy (DCS), on the other hand, as proposed by
McDowell (1994) and Brewer (1999), may be coupled with the view that the
content of perceptual experience is recognisable to the subject in virtue of the
operation of the same discriminatory capacities that are deployed in structuring
that content (CA). This raises important empirical and conceptual questions
concerning the nature and individuation of the discriminatory capacities it
proposes, which need not favour representationalism over antirepresentationalism.
In order to motivate their view, representationalists must not only show that
it is coherent and answers Travis’s challenge, but that it is superior to the
equivalent anti-representationalist view in a way that does not itself reduce to a
form of, for example, relationalism. Failure to do so gives representationalism
no explanatory advantage over the comparable relationalist view — a position
which in some cases it collapses into. McDowell’s (2008) notion of ‘intuitional
content’ appears to do precisely that, since the non-propositional ‘contents’ that
it posits are not assessable for truth or accuracy, and an equivalent explanation
can also be given at the level of judgement or belief. For these reasons, the
existence of representational content should not be seen as an end in its own
right, or as a silver bullet for solving problems in the philosophy of perception,
but rather as fulfilling some specific explanatory role, the nature of which needs
to be made clear from the outset. The existence of multiple such roles, including
epistemic justification and explaining phenomenal character, may require
multiple contents, the interrelations and connections between which need to be
made explicit, or else a detailed account of how such contents are individuated,
integrated and, where necessary, recognisable to the subject, given.
7. Conclusion
Beyond representationalism
7.1. Summary
In this thesis I have argued that the case against representationalism — the view
that perceptual experience is fundamentally and irreducibly representational —
set out in chapter 2 constitutes a powerful and much neglected argument against
the currently prevailing philosophical orthodoxy. Moreover, Travis’s arguments
establish substantive constraints upon the nature and role of perceptual content.
I also argued that the debate surrounding the content of experience centres not
so much upon the existence of representational content, but rather its
explanatory role, particularly in relation to phenomenal character and its
relation to other mental states — beliefs, thoughts, knowledge, intentions, and so
on.
In subsequent chapters I examined the implications of the above arguments
for various accounts of representational content including, though not restricted
to, looks- and appearance-based accounts (chapters 3 and 4), Burge’s ‘objective
representation’ (chapter 5), demonstrative, teleosemantic, counterfactual and
non-propositional accounts (chapter 6). In particular, the idea that perceptual
representation is ‘looks-indexed’ was rejected (pace Glüer 2009) on the basis
that such looks are either equivocal (in the case of comparative or visible looks)
or non-perceptual (epistemic or thinkable looks). Whilst the existence of a
phenomenal or ‘non-comparative’ notion of looks was not ruled out, this
proved to be unhelpful in establishing the existence of objective representational
content (4.2). Moreover, the availability of alternative comparative analyses of
phenomenal looks in the form of Martin’s (2010) ‘parsimonious’ account and
Brewer’s (2011) similarity-based account (4.3), neutralises this form of
argument for representationalism by reducing most such phenomenal uses to
their comparative equivalents, which are equivocal and so unsuitable for
indexing face-value content of the kind that many representationalists endorse.
Chapter 7, Conclusion
152
In chapter 5, I argued that the central issue at stake in the debate between
Travis and the representationalist is not the existence of representational
content, per se, but its role in explaining perceptual phenomenology. Thus,
views like Burge (2010), which claim that perceptual states have
representational contents with some other explanatory purpose — in scientific
explanation, for example — but deny that this explains the phenomenal
character of experience, i.e. intentionalism, are not in conflict with Travis
(2004). Rather, ‘The Silence of the Senses’ should be read as denying that
representation can fulfil a particular explanatory role in the mental life of
conscious subjects, not that representation has no role to play in the analysis of
perceptual states or abilities whatsoever.1 This criticism extends to the supposed
justificatory or reason-giving role of perceptual content, along with its relation
to belief, which can equally be explained in terms of conceptual or
discriminatory capacities operating at the level of judgement as at the level of
perception (chapter 6). Furthermore, I argued that the so-called
‘phenomenological objection’, or argument from unmediated awareness, that is
sometimes levelled at representational views is unmotivated and potentially
question-begging, despite raising substantive unanswered questions concerning
the role of phenomenal character in perceptual awareness (7.3.3).
In chapter 6, I considered how externally individuated representational
views, such as Millikan and Burge’s teleosemantics, might be extended to
account for the recognisability of phenomenal character through externalism
about self-knowledge (6.3.2). This view generalises to yield a form of
externalism about phenomenal character that many philosophers find
implausible, but which is able to escape the most serious of Travis’s arguments,
albeit with restrictions on the type of contents it can deliver.
Finally, I examined a number of alternative responses to Travis (2004),
including demonstrative, background knowledge- and capacity-based
approaches. Each of these involved the operation of perceptual discriminatory
capacities — whether these be conceptual, recognitional or otherwise — that
structure experience and its content. This raised a range of important and nontrivial questions about the precise nature of such capacities, their relation to
thought, concepts, content and phenomenal character. Whilst constituting the
outline of a possible response to Travis (2004) (7.2.4), such approaches fall
short of a providing conclusive argument for representationalism.2 Such an
argument would need to show why such representational content constitutes a
better explanation of perceptual phenomena than a comparable nonrepresentational view, such as Naïve Realism. As in the case of looks, any
argument based on the existence of discriminatory capacities that is available to
the representationalist at the level of experience is also potentially available to
the relationalist at the level of judgement, thereby neutralising its dialectical
1
This conflicts with an obvious reading of some of Travis’s stronger claims, most notably
that ‘perception is not representational’ (ibid. 57). Further arguments for this claim may be
found in some of Travis’s other work; e.g. Travis (2007) and Travis (unpublished).
2
See 7.1 below.
Chapter 7, Conclusion
153
force. Thus, even if a satisfactory response to Travis on this basis can be given
— something that it is far from clear (see 7.3.1) — this may not be decisive in the
wider debate concerning the fundamental nature of perceptual experience.
7.2. Implications for Representationalism
The case against representationalism presented above has a number of
important consequences for representationalist accounts of experience. First, it
rules out many otherwise apparently plausible accounts of representational
content, including the appearance-based accounts that I discuss in chapter 3.
Second, it refocuses the debate upon the existence of various constraints upon
the notion of representation that they employ (7.2.2) as well as the explanatory
roles that such representation is supposed to play (7.2.3). This in turn helps to
move the debate away from present, and in some cases artificial, divisions and
towards the more central issue of the distinction between perception and
thought (7.2.1) and the role of perceptual phenomenology (7.3.3).
Most importantly, however, Travis’s anti-representationalism sets up the
following challenge for representationalists about experience. Either:
(i)
they drop the requirement that p-representational content is apparent to
the subject in virtue of experience, including its phenomenal character,
i.e. Recognisability,
(ii)
they reject the claim that each experience has a single privileged ‘face
value’ content, i.e. Face-value, or
(iii)
they provide an alternative account of how the subject, in virtue of
perceptual experience, plus relevant perceptual capacities or background
knowledge, is able to grasp the representational content of their
experience.
As I argued above, none of these options is without its costs. The first either
leads to the rejection of intentionalism, or an acceptance that perceptual
representation is, as Travis claims, ‘silent’ in that it does not make
representational contents available to the subject, thus severely constraining its
explanatory role. Representational accounts of this sort are not forms of
representationalism per se, since they do not explain the nature of perceptual
experience as such (cf. 1.2.4). The second option rejects what many
representationalists take to be a basic datum about the phenomenology of
experience: namely, that things can be ‘just as they seem’ in the sense of having
accurate or veridical representational content. Rejecting this principle grants a
key premise of Travis’s argument — that looks are equivocal — which again
constrains the role of representational content in explaining perceptual
experience and the formation of judgements and belief. The third option
therefore seems like the most promising for any representationalist who wishes
to uphold some form of Recognisability. An account that aims to meet this
criteria, however, will need to do so in a way that does not itself reduce to a
Chapter 7, Conclusion
154
form of relationalism in order to qualify as a fundamentally representational
view. Alternatively, it may give rise to a ‘hybrid’ or compatibilist position in
which both representational and relational aspects of experience are taken to be
equally fundamental (7.3.2). I sketch a possible strategy for responding to this
challenge below (7.2.4).
7.2.1. Perception and thought
The debate between representationalism and anti-representationalism — or
between representationalism and relationalism, as it is sometimes, in my view
misleadingly, portrayed — is most fundamentally a question concerning the
division of labour between perception and thought. Assimilating the former to
the latter yields a representational picture of experience that is more obviously
conducive to explaining the reason-giving role of perception and the formation
of judgements and beliefs, which are representational. It does this, however, at
the cost of making perception into a kind of testimony of the senses, to be
accepted or rejected at face-value — a notion which Travis shows to be highly
problematic. Whether perceptual experience is, in the relevant sense, analogous
to testimony, or whether there are other ways in which it might satisfy this
function, are therefore central questions in the debate concerning the existence
(or otherwise) of representational content.
Moreover, Travis puts pressure on the assumption that the kind of content
that is needed to explain the phenomenal character of experience and the
representational contents that justify perceptual judgement or belief are one and
the same thing. This calls into question the role of phenomenal character in the
awareness and knowledge of the external world, and whether
representationalism can account for it.
Many representationalists whose accounts involves demonstrative or
otherwise externally individuated elements take themselves to have already
addressed these apparent shortcomings of the representational view. By
incorporating externally individuated elements into the contents of experience,
they aim to do justice both to the particularity, or fine-grained nature, of
experience and its role in enabling knowledge of the physical world. However,
as the arguments from looks and recognisability show, these issues are not
limited to the individuation of content, but also its availability to the subject —
an epistemic constraint.
This issue of ‘recognisability’, as I have called it, highlights the point at
which perceptual content becomes available to consciousness. This occurs most
obviously, though not necessarily exclusively, through the phenomenal character
of experience, or what experience ‘is like’. Only by being available to the subject
in some way can representational content play a substantive cognitive role in the
formation of thoughts and beliefs. Conversely, without explaining how such
availability to consciousness is possible, representationalism fails to meet its goal
of explaining the connection between thought and experience. Of course, the
representationalist can simply deny that experiential contents need be available
to consciousness, claiming instead that we simply find ourselves disposed to
Chapter 7, Conclusion
155
think, believe or do certain things as a result of our experiences. This approach,
however, fails to do justice to perceptual phenomenology which, at least in a
wide variety of cases, is available to conscious introspection, not to mention the
representationalist’s own notion that we can accept or reject perceptual
experience ‘at face value’ (2.2.2). To reject this principle therefore places severe
constraints upon the role that such content can play in our conscious mental
life.
Apart from drawing attention to the many and varied roles that
representational content may be thought to play (7.2.3), this highlights the need
for representationalists to explain precisely how experience makes it possible to
grasp or ‘recognise’ the content of experience. Indeed, it is upon precisely this
point that Travis’s arguments press. Any account that is unable to do this fails
to say anything about what it is to perceptually experience the world, regardless
of whether it posits the existence of representational content for other
explanatory purposes (cf. Burge 2010). Furthermore, the mere fact that beliefs,
judgements and knowledge possess such contents cannot be taken as a point in
favour of representationalism since the anti-representationalist can apply similar
explanations at the level of judgement as does the representationalist at the level
of experience. That is, whatever the representationalist takes to make
experiential contents available could instead be given as an explanation of what
gives perceptual judgements, beliefs and subsequent actions their contents. To
constitute an argument in favour of representationalism, then, as opposed to an
anti-representationalist view, such an explanation would need to be one that is
not available to the anti-representationalist ‘downstream’ of experience at the
level of perceptual interpretation or belief, or that otherwise rules out a nonrepresentational explanation of the relevant phenomenon.
7.2.2. Constraints upon representationalism
By challenging both the coherence of and need for objective representational
content, the arguments set out in chapter 2 present a series of constraints upon
representationalist theories of experience. Whilst this, in my view, falls short of
ruling out all representational accounts of perceptual experience, it also helps to
shift the debate onto potentially more fertile ground. As suggested above, the
alleged dichotomy between exclusively ‘representational’ views on the one hand
and exclusively ‘relational’ views on the other is a false one. Many so-called
representational views — for example, those that include demonstrative or
indexical elements — also contain a strongly relational aspect, whilst some
relational views may also be described in representational terms. Focusing solely
upon this axis of the debate only serves to obscure many other important
distinctions — and indeed similarities — between the candidate views. These
include the difference between individuative and constitutive externalism (i.e.
Fregean and Russellian content) and whether phenomenal character should be
explained in terms of purely internal or partly external factors, along with its
role in accounting for perceptual awareness (7.3.3).
Chapter 7, Conclusion
156
What constraints, then, do these arguments place upon representationalism?
To some extent these are set out in Travis’s Objectivity, Face-value, Givenness
and, crucially, Recognisability conditions (2.2), which provide a useful way of
categorising various forms of representationalism according to which of these
conditions that they endorse. But since not all representationalist accounts
endorse each of the conditions, what else does this tell us about
representationalism?
Travis’s arguments highlight the need to separate two important issues
about representational content. First, how is such content to be individuated —
phenomenologically, internally, externally, and so on. Second, whether and in
virtue of what it is it recognisable, or cognitively available, to the subject.
Accounts that aim to satisfy the latter criteria also need to explain why the
content that is recognisable from experience matches the content that is
individuated. Most plausibly this will be because both are determined in
precisely the same manner — the so-called transmission model described above
(6.2.1). Alternatively, the account may be an ‘error theory’ in which the
contents that are recognisable in experience are never, or only very rarely,
instantiated. Such an account, however, raises questions concerning the
desirability or advantages of such a theory over the available alternatives.
Finally, the relation between representational content, phenomenal character
and the role that each plays in conscious perceptual awareness must be
explained. If representational content is to explain phenomenal character, i.e. if
intentionalism is true, then at least some elements of that content may be
recognisable to the subject in virtue of this phenomenal character. Other content
elements may, however, be recognisable to the subject via other means.
Furthermore, if the same representational contents are to play a role in
explaining perceptual awareness and so ground the subject’s epistemic access to
perceived objects and properties, then this will also depend upon certain content
elements. Now, only where these two sets of content elements — i.e. those that
are recognisable to the subject in virtue of perceptual phenomenology and those
that explain perceptual awareness — are identical will the subject’s justification,
or reasons, for perceptual beliefs and judgements be cognitively available to
them. As the arguments in chapter 5 showed, the mere overlap or association
between these two sets of elements will be insufficient for such cognitive
availability. This impacts upon both the explanatory role of phenomenal
character in perceptual awareness as well as the kinds of explanation of
perceptual knowledge that can be given. Thus the issue of recognisability
highlights an important epistemological constraint upon the representational
content of experience.
7.2.3. Explanatory roles of representation
In ‘The Silence of the Senses’, Travis states that ‘[i]n no case I am aware of is
this view [representationalism] argued for’ (2004: 57). Whilst being something
of an overstatement, and certainly no longer the case, as demonstrated in
chapter 3, beyond a vague appeal to some form of argument from illusion,
Chapter 7, Conclusion
157
representationalists are often less than clear about the motivations for and
benefits of their view. Indeed, the notion of ‘representation’ has been seen as
something of a silver bullet in solving a wide variety of problems within the
philosophy of perception and action. Whether all of these can be solved by a
single form of content is, however, doubtful. Travis’s argumentative strategy
pits just two of these roles, or perhaps more, depending how finely they are
individuated, against one another to show that they cannot be made to coexist
without considerable difficulty. This highlights a general need for philosophers
of perception to be clear and explicit about precisely which explanatory role, or
roles, any given type of content is supposed to fulfil, as well as how it achieves
that goal and its relation to other content types.
By focusing exclusively upon the nature and existence of representational
content, it is easy to lose sight of the various overlapping roles it is often
supposed to play, many of which are frequently conflated or simply assumed
without substantive argument. Just a few of these — some of which have already
been mentioned above — are listed below, though the list is far from exhaustive:
(i)
Presenting objects and properties in the world in such a way as to make
them available to the subject in conscious thought and action
(ii)
Epistemic justification or giving reasons for beliefs and judgements
(iii)
Explaining the contents of beliefs, judgements and knowledge
(iv)
As the supervenience base for, or as constitutive of, phenomenal
character, or what perception ‘is like’
(v)
The nature of perceptual appearances, or how the world appears to the
perceptual subject to be
(vi)
Consciousness or awareness of objects and properties in the external
world
(vii)
Explaining the existence or nature of perceptual hallucinations and
illusions
(viii)
As a common factor between subjectively indistinguishable, but
epistemically distinct (i.e. veridical and non-veridical), experiences
(ix)
As part of the scientific explanation of an organism’s psychological
functioning and behaviour
(x)
As a way of individuating perceptual states
(xi)
Explaining the finely-grained nature of perceptual experience
That a single phenomenon — i.e. perceptual representation — might even be
considered to play each of the above roles may itself seem to be a point in its
favour. However, it is far from clear that any one form or notion of content is
capable of encompassing all of the above. If it did, the nature of, and responses
Chapter 7, Conclusion
158
to, Travis’s arguments from indeterminacy and awareness would surely be much
clearer.
Whilst it is tempting to posit different forms of representational content, or
different content elements, that fulfil various of the above roles, the resulting
contents will still face the above objections either individually (in the case of
multiple contents) or en masse (in the case of disjunctive or ‘multi-level’
contents). Moreover, in order to constitute an argument for representationalism,
each content would need to explain some aspect of perceptual experience that
could not equally well be explained by a comparable anti-representationalist
view. Simply assuming or insisting upon the existence of such contents from the
outset, particularly where this encompasses multiple explanatory roles, does not
constitute a justification for representationalism. If the content of experience is
to play a distinctive role in the philosophy of perception, then it needs to be
clear precisely what this role is and how it performs it, otherwise the supposed
explanation merely obfuscates the problem it is intended to solve. To this
extent, I concur with Burge that ‘the dispute would be better focused on
questionable psychological, representational, and epistemic roles that
[perceptual] representations have been given’ (Burge 2010: 379, fn. 15).
7.2.4. Responding to Travis
Having set out various roles of and constraints upon the content of experience,
how should representationalists respond to the challenges posed by Travis
(2004)? Clearly, if a representationalist account of experience is to be given,
then it must account for perceptual phenomenology. Furthermore, the resulting
content should play a constitutive, and not merely incidental, role in our
conscious awareness and epistemic access to the external world. Finally, the
representational content of experience must be available to the subject in
conscious thought and introspection, and not merely derivatively through the
contents of, for example, belief and action.
In chapter 6, I argued that the most plausible way of doing this would be via
a ‘transmission’ model of experience in which both the individuation and
recognisability of perceptual content are explained in terms of the same
mechanism or factor. Furthermore, I suggested that the most likely candidate for
such a mechanism would be some form of discriminatory capacities that are
responsible for structuring perceptual experience, and so its content, but that
also involve a demonstrative or indexical component that enables the subject to
grasp, as reflected in their perceptual phenomenology, particular and not only
general contents. Whether such capacities are conceptual or a precursor to
conceptualisation remains an open question. In either case, however, such
capacities will be grounded either in the subject’s implicit knowledge of
appearances, as gained through their past experience, or innate capabilities for
representing aspects of the world via the relevant sense modalities.
Whether perceptual experience is fundamentally representational, in the
sense required for representationalism, however, is uncertain. It is entirely
plausible that the discriminatory capacities that structure perceptual content, if
Chapter 7, Conclusion
159
indeed there are such capacities, would themselves be fundamentally relational
in character. That is, the operation of such capacities cannot be described in
isolation from the objects and properties to which they are perceptually
sensitive. Moreover, depending on the details of this account, it may make sense
to consider those objects and properties not only as individuating the contents
of experience, but as its constituents or relata. This would make the ‘contents’
of experience essentially relational as well as representational, and so a form of
compatibilism about experience. Provided that the phenomenal character of
experience can be described in terms of the operation of discriminatory
capacities that are fundamentally dependent upon the relevant objects or
properties that they track in the world, this need not divorce it from perceptual
awareness. However, any account of such awareness must in turn explain how
it is possible that a capacity that can ‘misfire‘ — in the case of perceptual
illusions, for example — can constitute awareness of an environmental
particular, rather than of some aspect of experience that can also occur in the
absence of its intended target. Alternatively, such capacities may operate at the
level of belief or judgement, rather than experience, in which case they could
also be used to support a non-representational account of experience.
Such a solution would represent only a partial victory for
representationalism, since the resulting view incorporates both representational
and relational elements. It is, therefore, an endorsement of CON and not only
REP or REL (1.2.4). Moreover, whether this constitutes a form of
representationalism at all will depend upon the explanatory work done by the
content that resulted from the operation of the relevant discriminatory
capacities, the precise mechanism for which remains unclear. Indeed, there are
reasons to be sceptical about the very distinction that philosophers draw
between perceptual experience and belief that motivates the need for such
content in the first place (7.3.4). An account of the above form, however, would
appear to be the representationalist’s best hope of giving an adequate response
to Travis. That such a response has yet to be given in any detail is a reflection of
the fact that, at the time of writing, there is no convincing argument in favour of
representationalism in its present forms.
7.3. Directions for Future Research
The above conclusions suggest the following areas for future research which
follow on directly from the issues raised in this thesis.
7.3.1. Perceptual discriminatory capacities
In chapter 6, I identified the role of what I called perceptual discriminatory
capacities as being central to a satisfactory account of perceptual experience
and/or belief. This is true not only for representationalism, but for relational
views as well, which require an explanation of the conceptualisation of
experience in belief, judgement and intentional action. Developing a more
detailed account of the nature and role of such capacities in experience is
therefore a priority for future research in this area that will in turn help inform
Chapter 7, Conclusion
160
the debate concerning the nature of perceptual experience. Such research will
involve answering the following questions concerning the individuation and
scope of such capacities, requiring a combination of philosophical and empirical
work to ascertain the precise role and mechanisms for such capacities along
with their relation to both representational content and perceptual
phenomenology:
(i)
How and why does the operation of discriminatory capacities instantiate
tokens of the relevant types in the content of experience?
(ii)
How are the relevant perceptual capacities to be individuated, and how
do they differ from the corresponding capacities involved in perceptual
judgement or belief?
(iii)
Precisely which discriminatory capacities are involved in any given
experience, and how is this to be determined?
(iv)
How and when do we acquire or develop the relevant capacities, and are
these learned or innate?
(v)
How does the operation of discriminatory capacities relate to the
conceptualisation, recognition and perceptual awareness of objects and
their properties?
(vi)
If the capacities that individuate experiential content differ from what
makes it recognisable, why should the latter contents coincide with the
former?
7.3.2. The possibility of hybrid views
It is surely no coincidence that a number of the positions described above (7.2.4)
are those more typically assumed by relationalists than by representationalists.
There is, however, no reason why these two views should be seen as opposing or
competing hypotheses, rather than two sides of the perceptual coin. Just as it is
possible to posit different representational contents to fulfil different
explanatory roles, it may be similarly advantageous to adopt, for example, a
representational account of non-conscious visual content and a relational
account of visual phenomenal character, or some other combination thereof.
Moreover, certain aspects of these positions are compatible such that it is
possible to combine relational elements within representational content, or to
even to describe fundamentally relational states as having propositionally
structured contents. Such ‘hybrid’ or compatibilist positions are not without
precedent, with Siegel (2010), Schellenberg (2011b) and McDowell (2008) all
advocating some form of hybrid view (cf. 6.4.4).
The possibility and potential advantages of such hybrid views is a
worthwhile and interesting development in recent philosophy of perception. By
letting go of the false dichotomy between representationalism and relationalism,
Chapter 7, Conclusion
161
it may be possible to move beyond the present debate and see representational
and relational elements not as two competing accounts of experience, but as
complementary ways of describing a single underlying perceptual reality.
Thought of in relational terms, representational content need not be seen as a
surrogate for, or epistemological barrier between, perceptual subjects and
objects, but as a descriptive tool for individuating and characterising the role of
certain perceptual states. Further work is needed, however, both on what forms
such hybrid accounts might take as well as the relation, or interrelation,
between the representational and relational elements of perception.
7.3.3. The role of phenomenal character
Much of the literature in the representationalist tradition is devoted to
explaining how representational content can account for the phenomenal
character of experience. However, comparatively little attention has been given
to the explanatory role of phenomenal character itself. If, as Campbell (2002)
and Johnston (2006) suggest, phenomenal character is not a mere by-product of
perceptual experience, but the means by which we are aware of the objective
world, this places further constraints upon the kinds of representational content
that can fulfil this role. In particular, representationalist accounts upon which
phenomenal character is epiphenomenal, as opposed to being constitutive of
perceptual awareness, may be ruled out. Furthermore, an explanation of how
the content that gives rise to phenomenal character contributes to perceptual
awareness will need to be given. The precise role of phenomenal character in an
overall account of perceptual experience is therefore a central question in the
philosophy of perception that, in my view, deserves much greater attention.
7.3.4. Experience and belief
Finally, as described above (7.2.1), the existence of representational content in
experience is often motivated by the division between perceptual experience on
the one hand, and perceptual belief, judgement and knowledge on the other.
This in turn assumes a division between experiential and belief states that,
whilst well established in the philosophical literature, is neither well defined nor
grounded in empirical research. In particular, the question of what makes a state
distinctively experiential, as opposed to doxastic or discursive, is not well
defined. Moreover, the individuation of particular experiences over time and
across multiple perceptual modalities is highly problematic. Whilst the multimodal nature of perception has received considerable attention in recent times,
the distinction between experience and belief that is so central to the present
debate has not. After all, it is difficult, if not impossible, to ascertain the
contents or metaphysical nature of experience when it is unclear how individual
experiences are to be defined or individuated. Most importantly, this issue holds
the potential to dissolve many of the puzzles surrounding the existence and
nature of perceptual content.
Some authors — notably Hinton (1973) and Byrne (2009: 431–32) — have
expressed scepticism that there is any such thing as perceptual experience in the
Chapter 7, Conclusion
162
philosophical sense of this term. Moreover, the notion of an ‘experience–belief
synapse’ (Byrne unpublished) at which experiential episodes become belief states
forms part of the motivation for the notion of perceptual content, which is
intuitively characterised as what experience and belief have in common, or from
which the content of belief may be derived. This in turn presupposes a model of
perception as a sequential ‘pipeline’ or ‘train of operations’ (Reid 1974: VI.xxi,
174) in which raw sensory input is processed by our perceptual faculties to yield
one form of content which is then conceptualised and transformed into belief
through the act of judgement. If, however, the relevant processes were not
sequential but, for example, mutually constraining such that perceptual
experience is the result of both top-down and bottom-up processes that
constrain or mutually reinforce one another to form a coherent and stable
percept, then the standard division between experience and belief may itself be
ill founded. On this view, instead of a causal relation between experience and
belief, experiences may themselves be constitutive of perceptual beliefs, with
experiential states playing a belief-like role in the psychology of the subject.
The existence of a constitutive relationship between perceptual experience
and perceptual belief would explain why many theorists — including Byrne
(2009: 450), Smith (2002), Armstrong (1968), Heil (1982) and Reid (1974), to
name but a few — have taken experience itself to be a doxastic state. However,
it would also negate the need to posit representational contents of experience
that can form the basis for the contents of belief in the first place, since
experience and perceptual belief would each be complementary aspects of the
same underlying mental state. Such an approach has the potential to unify
representationalism and relationalism in a single account of perceptual
experience centred around the kind of discriminatory capacities described above
(7.2.4). As with many other areas in the contemporary philosophy of mind,
however, a satisfactory resolution to this issue is only likely to be achieved
through a combination of philosophical and empirical investigation.
Bibliography
Armstrong, D. M. (1968). A Materialist Theory of Mind. London: Routledge and
Kegan Paul.
Block, Ned (2007). ‘Overflow, access and attention’. Behavioral and Brain Sciences,
30, pp. 530–42.
Bornstein, Marc H. (1985). ‘On the Development of Color Naming in Young
Children: Data and Theory’. Brain and Language, 26 (1), pp. 72–93.
Breckenridge, Wylie (2007). The Meaning of ‘Look’. Doctoral Thesis, New College,
University of Oxford.
Brewer, Bill. (1999). Perception and Reason. Oxford: Clarendon Press.
––––––––– (2005). ‘Do Sense Experiential States Have Conceptual Content?’. In
Contemporary Debates in Epistemology, M. Steup & E. Sosa (eds.).
Oxford: Blackwell, pp. 217–30.
––––––––– (2006). ‘Perception and Content’. European Journal of Philosophy, 14
(2), pp. 165–81.
––––––––– (2007). ‘Perception and Its Objects’. Philosophical Studies, 132 (1),
pp. 87–97.
––––––––– (2008). ‘How to Account for Illusion’. In Disjunctivism: Perception,
Action, Knowledge, ed. A. Haddock & F. Macpherson. Oxford: Oxford
University Press, pp. 168–80.
––––––––– (2011). Perception and Its Objects. Oxford: Oxford University Press.
Brogaard, Berit (forthcoming). ‘Do “Looks” Reports Reflect the Contents of
Perception?’. In The Language of Consciousness, B. Brogaard & D.
Chalmers (eds.), Oxford: Oxford University Press. Available from:
<http://sites.google.com/site/brogaardb/LooksReportsII.pdf> [Accessed
25/9/12].
Burge, Tyler (1979). ‘Individualism and the Mental’. In Midwest Studies in
Philosophy, IV, P. French, T. Uehling, and H. Wettstein (eds.).
Minneapolis: University of Minnesota Press, pp. 73–121.
––––––––– (1986). ‘Individualism and Psychology’. Philosophical Review, 95,
pp. 3–45.
–——–—– (1988). ‘Individualism and Self-Knowledge’. Proceedings of the
Aristotelian Society, 85 (11), pp. 649–63.
––––––––– (1996). ‘Our Entitlement to Self-Knowledge, I: Tyler Burge’. Proceedings
of the Aristotelian Society, New Series, 96 (1996), pp. 91–116.
––––––––– (2010) Origins of Objectivity. New York: Oxford University Press.
Byrne, Alex (2001). ‘Intentionalism Defended’. The Philosophical Review, 110 (2),
pp. 199–240.
Bibliography
164
––––––––– (2009). ‘Experience and Content’. Philosophical Quarterly, 59 (236),
pp. 429–51.
––––––––– (unpublished). ‘Perception and Belief’. Conceptual Contents: History and
Prospects Conference, University of Cambridge, 20 September 2011.
Byrne, Alex & Logue, Heather (2008). ‘Either/Or’. In Disjunctivism: Perception,
Action, Knowledge, A. Haddock & F. Macpherson (eds.). Oxford:
Oxford University Press, pp. 57–94.
Campbell, John (2002). Reference and Consciousness. Oxford: Oxford University
Press.
Chalmers, David J. (2006). ‘Perception and the Fall from Eden’. In T. S. Gendler &
J. Hawthorne (eds.), pp. 49–125.
Chisholm, Roderick M. (1957). Perceiving: A Philosophical Study. Ithaca, NY:
Cornell University Press.
Crane, Tim (2006). ‘Is There a Perceptual Relation?’. In T. S. Gendler &
J. Hawthorne (eds.), pp. 126–46.
––––––––– (forthcoming). ‘The Given’. In The Myth of the Mental?, J. Schear (ed.).
London: Routledge.
Davies, Martin (1992). ‘Perceptual Content and Local Supervenience’. Proceedings
of the Aristotelian Society, 92, pp. 21–45.
Dretske, Fred (1969). Seeing and Knowing. Chicago: University of Chicago Press.
Gallagher, Shaun (1998). The Inordinance of Time. Evanston:
Northwestern University Press.
––––––––– (1994). ‘If You Can't Make One, You Don't Know How It Works’. In
Midwest Studies in Philosophy, 19, P. French, T. Uehling & H.
Wettstein (eds.), pp. 468–82.
Evans, Gareth (1982). The Varieties of Reference, J. McDowell (ed.). Oxford:
Clarendon Press.
Fodor, Jerry A. (1987). Psychosemantics. Cambridge, MA: MIT Press.
Gendler, Tamar Szabó & Hawthorne, John (eds.) (2006). Perceptual Experience.
Oxford: Oxford University Press.
Glüer, Kathrin (2009). ‘In Defence of a Doxastic Account of Experience’. Mind and
Language, 24 (3), pp. 297–327.
Heil, John (1982). ‘Seeing is Believing’. Americal Philosophical Quarterly, 19 (3),
pp. 229–39.
Hinton, J. M. (1973). Experiences: An Inquiry into Some Ambiguities. Oxford:
Oxford University Press.
Hume, David (1999). An Enquiry Concerning Human Understanding, T. L.
Beauchamp (ed.). Oxford: Oxford University Press.
Hurley, Susan (1998). Consciousness in Action. Cambridge, MA: Harvard
University Press.
Bibliography
165
Jackson, Frank (1977). Perception: A Representative Theory. Cambridge:
Cambridge University Press.
–––––––––– (1982). ‘Epiphenomenal Qualia’. The Philosophical Quarterly, 32 (127),
pp. 127–36.
Johnston, Mark (2006). ‘Better than Mere Knowledge? The Function of Sensory
Awareness’. In T. S. Gendler & J. Hawthorne (eds.), pp. 260–90.
Kalderon, Mark Eli (2011), It is not an argument but a dramatic expression of
conviction… [Twitter] 31st July. Available from: <https://twitter.com/
PhilGeek/status/97634509058605056> [Accessed 20/5/13].
Kripke, Saul (1972). Naming and Necessity. Oxford: Blackwell.
Lycan, William (2008). ‘Representational Theories of Consciousness’. The Stanford
Encyclopedia of Philosophy (Fall 2008 Edition), E. N. Zalta (ed.),
<http://plato.stanford.edu/archives/fall2008/entries/consciousnessrepresentational/>.
Martin, M. G. F. (2002a). ‘Particular Thoughts & Singular Thought’. Royal
Institute of Philosophy Supplement, 51, pp. 173–214.
––––––––– (2002b). ‘The Transparency of Experience’. Mind and Language, 17,
pp. 376–425.
––––––––– (2004). ‘The Limits of Self-Awareness’. Philosophical Studies, 120,
pp. 37–89.
––––––––– (2006). ‘On Being Alienated’. In T. S. Gendler & J. Hawthorne (eds.),
pp. 354–410.
––––––––– (2010). ‘What’s in a Look?’. In Perceiving the World, B. Nanay (ed.).
Oxford: Oxford University Press, pp. 160–225.
––––––––– (forthcoming a). Uncovering Appearances.
––––––––– (forthcoming b). ‘Sensation’. Given at the Consciousness and SelfConsciousness Research Seminar, University of Warwick, 11 May 2011.
McDowell, John (1982). ‘Criteria, Defeasibility, and Knowledge’. Proceedings of
the British Academy, 68, pp. 455–79.
––––––––– (1986). ‘Singular Thought and the Extent of Inner Space’. In Subject,
Thought, and Context, P. Petit & J. McDowell (ed.). Oxford: Clarendon
Press.
––––––––– (1994). Mind and World. Cambridge, MA: Harvard University Press.
––––––––– (2008). ‘Avoiding the Myth of the Given’. In John McDowell:
Experience, Norm, and Nature, J. Lindgaard (ed.). Malden,
Massachusetts: Blackwell Publishing, pp. 1–14.
McGinn, Colin (1982). The Character of Mind, Oxford: Oxford University Press.
––––––––– (1991). The Problem of Consciousness, Oxford: Blackwell.
McKinsey, Michael (1991). ‘Anti-Individualism and Privileged Access’. Analysis, 51,
pp. 9–16.
Bibliography
166
Millikan, Ruth Garrett. (1989). ‘Biosemantics’. The Journal of Philosophy, 86 (6),
pp. 281–97.
Pappas, George S. (1976). ‘Seeinge and Seeingn’. Mind, 85 (338), pp. 171–88.
Peacocke, Christopher (1992). A Study of Concepts, Cambridge, Massachusetts:
MIT Press.
––––––––– (1996). ‘Our Entitlement to Self-Knowledge, II: Christopher Peacocke,
Entitlement, Self-Knowledge and Conceptual Redeployment’.
Proceedings of the Aristotelian Society, New Series, 96 (1996), pp. 117–
58.
Putnam, Hilary (1975). ‘The Meaning of Meaning’. Philosophical Papers, Vol. II:
Mind, Language, and Reality, Cambridge: Cambridge University Press.
Reid, Thomas (1997). An Inquiry into the Human Mind on the Principles of
Common Sense, D. R. Brookes (ed.). Edinburgh: Edinburgh University
Press.
––––––––– (2002). Essays on the Intellectual Powers of Man: A Critical Edition.
Edinburgh: Edinburgh University Press.
Robinson, Howard (1994). Perception. London: Routledge.
Rosenthal, David (1986). ‘Two concepts of consciousness’. Philosophical Studies,
49, pp. 329–59.
Russell, Bertrand (1905). ‘On Denoting’. Mind, pp. 479–93.
Schellenberg, Susanna (2011a). ‘Ontological Minimalism about Phenomenology’.
Philosophy and Phenomenological Research, 83 (1), pp. 1–40.
––––––––– (2011b). ‘Perceptual Content Defended’. Noûs, 45 (4), pp. 714–50.
Searle, John (1983). Intentionality, Cambridge: Cambridge University Press.
Sedivy, Sonia (2004). ‘Minds: Contents Without Vehicles’. Philosophical
Psychology, 17 (2), pp. 149–81.
Siegel, Susanna (2010). The Contents of Visual Experience. New York: Oxford
University Press.
Smith, A. D. (2001). ‘Perception and Belief’. Philosophy and Phenomenological
Research, 62 (2), pp. 283–309.
––––––––– (2002). The Problem of Perception. Harvard University Press.
Snowdon, Paul (1990). ‘The Objects of Perceptual Experience’. Proceedings of the
Aristotelian Society, Supplementary Volumes, 64 (1990), pp. 121–66
––––––––– (2005). ‘The Formulation of Disjunctivism: A Response to Fish’.
Proceedings of the Aristotelian Society, New Series, 105, pp. 129–41.
Thau, Michael (2002). Content and Consciousness. New York: Oxford University Press.
Travis, Charles (2004). ‘The Silence of the Senses’. Mind, 113, pp. 59–94.
––––––––– (2007). ‘Reason’s Reach’. European Journal of Philosophy, 15 (2),
pp. 225–48
Bibliography
167
––––––––– (forthcoming). ‘The Silence of the Senses’. Available from: <http://
mlag.up.pt/?article=the-silence-of-the-senses-new-version> [Accessed
25/9/12].
––––––––– (unpublished). ‘Unlocking the Outer World’. Given at Perception,
Consciousness and Reference Workshop, University of Warwick,
September 2009. Available from: <https://sites.google.com/site/
charlestraviswebsite/Home/current-and-recent-work/Outerworld.pdf?
attredirects=0> [Accessed 25/9/12].
Tye, Michael (1995). Ten Problems of Consciousness, Cambridge, Massachusetts:
MIT Press.
––––––––– (2002). Consciousness, Color and Content. Cambridge, MA: MIT Press.
––––––––– (2007). ‘Intentionalism and the Argument from No Common Content’.
Philosophical Perspectives, 21, Philosophy of Mind, pp. 589–613.
Wilk, Stephen R. (2009). ‘The Yellow Sun Paradox’. Optics & Photonics News, 20
(3), pp. 12–13.
Williamson, Timothy (2000). Knowledge and Its Limits. Oxford: Oxford University
Press.
Wittgenstein, Ludwig (1953). Philosophical Investigations. Malden, MA: Blackwell.