Journal of Philosophy, Inc.

Does Perception Have a Nonconceptual Content?

Author(s): Christopher Peacocke
Reviewed work(s):
Source: The Journal of Philosophy, Vol. 98, No. 5 (May, 2001), pp. 239-264
Published by: Journal of Philosophy, Inc.
Stable URL: http://www.jstor.org/stable/2678383 .
Accessed: 04/09/2012 18:24

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Journal of Philosophy, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal
of Philosophy.

http://www.jstor.org
 DOES PERCEPTION H4AVE A NONCONCEPTUAL CONTENT? 239

DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT?*

T n he question posed in my title is one that has been vigorously
debated in philosophy for almost twenty years now. In one
form or another, the idea that perceptual experience has a
content that is nonconceptual is found in the writings of, among
others, Jose Bermuidez, Tim Crane, Fred Dretske, Gareth Evans,
Susan Hurley, myself, and Michael Tye.' The idea has been strongly
opposed byJohn McDowell, Sonia Sedivy, and, most recently, by Bill
Brewer.2 The question has generated so much discussion because the
possibility-or otherwise-of nonconceptual content is inseparable
from some fundamental issues: about the individuation of conceptual
content; about the nature of concept possession; about the nature of
rationality; about the relation between animal and human percep-
tion; and even about our conception of objectivity-which all turn, in
part, on the possibility of nonconceptual content in perception. That
my question is inseparable from these wider issues I shall try to make

* The present paper was delivered at the Certosa di Pontignano, Siena in May
1999, under the auspices of the Universita degli Studi di Siena. I am grateful to
participants at the meeting for discussion, and to later audiences at Bonn, Essen,
Oxford and New York University. I have retained the largely informal style of a
conference talk. Special thanks to Bill Brewer and Stephen Schiffer for valuable
comments. Some of the work on the present text was carried out while I held a
Leverhulme Research Professorship; once again I thank the Leverhulme Trust for
this invaluable support.
1 Bermiidez, The Paradox of Self-Consciousness(Cambridge: MIT, 1998), especially
chapter 3; Crane, "The Nonconceptual Content of Experience," in Crane, ed., The
Contentsof Experience:Essays on Perception(New York: Cambridge, 1992), pp. 136-57;
Dretske, Naturalizing the Mind (Cambridge: MIT, 1995), especially chapter 1, pp.
19ff., and earlier in Knowledgeand the Flow of Information (Cambridge: MIT, 1981);
Evans, The Varieties of Reference(New York: Oxford, 1982), see index entry for
"Conceptual and Non-conceptual Content"; Hurley, Consciousnessin Action (Cam-
bridge: Harvard, 1998), especially chapter 4; my "Analogue Content," Proceedingsof
theAristotelianSociety,Supplementary Volume LX (1986): 1-17, and A Study of Concepts
(Cambridge: MIT, 1992), chapter 3; Tye, Ten Problemsof Consciousness(Cambridge:
MIT, 1995), especially pp. 137ff. A rather different but related use of the notion of
nonconceptual content is found in the work of Adrian Cussins: see his paper "The
Connectionist Construction of Concepts," in Margaret Boden, ed., The Philosophyof
ArtificialIntelligence(New York: Oxford, 1990), pp. 368-440.
2 McDowell, Mind and World (Cambridge: Harvard, 1994), especially Lecture III
and Afterword, Part II; Sedivy, "Must Conceptually Informed Perceptual Experience
Involve Non-Conceptual Content?" Canadian Journal of Philosophy, xxvi (1996):
413-31; and Brewer, Perception and Reason (New York: Oxford, 1999), especially
chapter 5.

0022-362X/01/9805/239- 64 ?C)2001 The Journal of Philosophy, Inc.

240 THE JOURNAL OF PHILOSOPHY

plausible as we proceed. My specific aim here is to argue further for

the thesis that perception does indeed have nonconceptual content.3
I. THE CLAIMS OF NONCONCEPTUAL CONTENT
The nonconceptual content of perception that is the topic of this
paper is meant to be one species of representational content. It is
content that is evaluable as correct or as incorrect. So we are not
concerned here with whether there are nonrepresentational con-
scious properties of perceptions, what are sometimes called sensa-
tional properties. The question I am now addressing concerns solely
the class of representational contents of perception. It is the issue of
whether some of the representational content of perception is non-
conceptual.
The discussions of recent years have focused on the fine-grained
representational content of experience. When you look at the new
Guggenheim Museum in Bilbao, or see a new abstract sculpture, or
the face of a person, you see each of these objects as having a quite
specific shape and size. Similarly, you see them as having quite
specific shades of colors, surface textures, and contours. Equally,
when you hear a musical tone, there is a sense in which you perceive
its pitch. You may not recognize the tone-you may not have absolute
pitch-but you can discriminate that pitch from many, many others
if you are asked to compare it with another. This discrimination is
based on the way the tone sounds to you. All parties to these discus-
sions have acknowledged the fine-grained character of this represen-
tational content. What has been at issue is not its existence, but its
character.
We shall not do justice to the fine-grained phenomenology of
experience if we restrict ourselves to those contents which can be
built up by referring to the properties and relations that the per-
ceived objects are represented by the experience as possessing. We
must, in describing the fine-grained phenomenology, make use of the
notion of the way in which some property or relation is given in the
experience. The same shape can be perceived in two different ways,
and the same holds for the shape properties, if we regard them as
within the representational content of experience. Ernst Mach's ex-
ample of one and the same shape that can be perceived either as a

3 I shall confine myself in this paper to the fundamental philosophical motiva-

tions for acknowledging nonconceptual content in perception, and the philosoph-
ical challenges such acknowledgment must address. There are many further
important issues about the nature and role of nonconceptual content which arise in
the light of psychological experiments on perception, and on action: discussion of
them will have to be left to another occasion.
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 241

square or as a regular diamond is a familiar example. These are not

different shapes. The shape of an object need not alter when it
moves; and an object can be perceived either as a square, or as a
diamond, in either of the standard orientations relative to the per-
ceiver.
The need to use the notion of the way in which something is
perceived is by no means special to the perception of spatial proper-
ties and relations. Suppose middle C and the F-sharpjust above it are
played simultaneously on a piano. The interval thus sounded may be
heard either as an augmented fourth, or as a diminished fifth. In the
former case, the higher tone of the interval is heard as the seventh of
the presumed tonic scale.4 In the latter case, it is heard as the fourth
of that scale. In neither case need these music-theoretic descriptions
be ones known to the perceiver, as part of his personal-level concep-
tual repertoire. We use them in a plausible theory about the nature of
the two ways in which the interval can be perceived. It is the way itself,
and not the materials of music theory, which enter the description of
the subject's experience. Even one and the same time interval may be
perceived in two different ways. It is perceived in one way, when it is
perceived as the familiar time interval between your switching on
your computer and its emitting the start-up sound. It may be per-
ceived in another way when the same interval is exactly that of the
silence in a complete, but empty measure (bar) of a Haydn string
quartet, in which case it may be heard as composed of two beats.
So, in characterizing the fine-grained content of experience, we
need the notion of the experience representing things or events or
places or times, given in a certain way, as having certain properties or
as standing in certain relations, also as given in a certain way. Hence-
forth I use the phrase the content of experienceto cover not only which
objects, properties, and relations are perceived, but also the ways in
which they are perceived. The ways I have mentioned all contribute
to the representational content of the experience. That is, when
something is perceived in one of these ways, the claim that the object
really is the way it is experienced as being is one which has a
correctness condition.
Here, now, are five at least moderately intuitive claims about non-
conceptual content which one might expect a believer in noncon-
ceptual content to defend. Some of these claims are, prima facie,
incompatible with the position that the representational content of

4On the musical case, see the fine contribution of Mark DeBellis, Music and
Conceptualization(New York: Cambridge, 1995), especially chapter 2.
242 THE JOURNAL OF PHILOSOPHY

experience is exclusively conceptual. Before we start discussing them,

here is a brief, wholly dogmatic, statement of each of the claims.
(1) The nonconceptual content of perceptual experience contrib-
utes to making available to a thinker various perceptually based
concepts. Only a thinker who has a perceptual experience with a
certain kind of nonconceptual representational content can employ
such perceptual-demonstrative concepts as that shape, that texture,that
interval of time. The nonconceptual content of experience also helps
to make available such general, nondemonstrative recognitional con-
cepts as the concept regular-diamond shaped and the concept red. It also
makes available more specific, but still general, recognitional con-
cepts such as the concept Cambridgeblue and the concept yellow ochre.
Perceptual states with nonconceptual content make these general
concepts available to a thinker by providing the canonical, noninfer-
ential basis for the application of these concepts to things given in
experience. The nonconceptual contents that make available these var-
ious perceptually based conceptual contents cannot, however, be iden-
tified with any of the conceptual contents that they make available.
These claims are all to be construed as constitutive. What it is to
have the perceptually based concepts is to be elucidated, philosoph-
ically, in terms of the relations of those concepts to the nonconcep-
tual content of experience.
(2) Experiences with finer-grained nonconceptual contents can
also provide an empirical basis for the acquisition, and so enter the
causal explanation of the learning, of such general concepts as
regular-diamondshaped.
(3) An experience with a certain nonconceptual content can make
rational a judgment of a conceptual content suitably related to the
nonconceptual content that the experience represents as correct.
(4) Some of the nonconceptual content of our experience can be
identical with the representational content of the experience of
creatures that either possess no concepts, or possess only a set of
concepts far more rudimentary than our own.
(5) The nonconceptual content of experience can enter the ex-
planation of features of intentional action. There is an important
discussion in psychology of whether this content, or its precursors,
enter the explanation of action only via memory, rather than control
concurrent action on-line.5 But all agree that it has an explanatory
role in some cases.

5 For an illuminating overview, see A. David Milner and Melvyn A. Goodale, The
Visual Brain in Action (New York: Oxford, 1995).
 DOES PERCEPTION H4AVE A NONCONCEPTUAL CONTENT? 243

The formulation of these five claims should make clear some of the
wider ramifications of the otherwise apparently rather local issue of
whether perception has a nonconceptual representational content.
We need to be very clear what we mean by 'conceptual'. I shall be
taking it that conceptual content is content of a kind that can be the
content of judgment and belief. Concepts are constituents of those
intentional contents which can be the complete, truth-evaluable,
contents ofjudgment and belief. Conceptual content and concepts I
take to have identities conforming to, indeed answerable to, Gottlob
Frege's criterion of identity for senses. Complete contents p and q are
distinct if and only if it is possible for someone for whom the question
arises rationally to judge that p withoutjudging that q, and even while
judging not-q. So the content 'This country is Italy' is distinct from the
content 'This country is this country'; the content 'The floor-plan is
square' is distinct from 'The floor-plan is a regular diamond'; the
content 'Your lost pen is there' (pointing under a pile of papers) is
distinct from 'Your lost pen is where you last used it'. Concepts C and
D are distinct if and only if there is some completing content E such
that the complete content E (C) is distinct from the complete content
(D); or, in other words, if and only if there is some content E(C)
such that someone for whom the question arises can rationallyjudge
E (C) without judging E (D). So the pairs

this country Italy

square regulardiamond
there(pointing appropriately) whereyou last used thepen

are all pairs of distinct concepts by this criterion. As these examples

show, concepts as intended here may be demonstrative or indexical.
On this approach, any further connections between concepts and,
for instance, language have to be earned by further argument. If
someone holds that a concept user must have a language in which he
can express at least some of his concepts, that is a substantive,
nondefinitional thesis that needs to be established. The same applies
to the thesis that concepts, so characterized, conform to Evans's
"generality constraint" (op. cit.).
It should, however, be uncontroversial that any content that can be
expressed in language by the use of an indicative sentence, including
sentences containing indexicals and demonstratives, will be a con-
ceptual content. That follows in the presence of the less controversial
premise that any utterance of an indicative sentence, in a given
context, expresses a content that is of a kind that could also be the
content of a belief or judgment.
244 THE JOURNAL OF PHILOSOPHY

The Fregean criterion as I have formulated it applies only to those

contents which can be the content ofjudgment or belief. It would be
a fair observation that the Fregean criterion as I have formulated it is
an instance of a more general criterion, under which we can rightly
say that there are distinct ways W1 and W2 in which a shape, or an
interval, is given in perception simply because it is not obvious (even
at the level of perception) that something given in way W1 is the same
shape, or interval, as something given in way W2. So there certainly is
some intuitive notion of "mode of presentation" that applies uncon-
troversially at the level of perceptual content. That more general
notion of mode of presentation is not, however, the level at which the
question here is posed. My question is whether the ways involved in
perceptual content are the same as the conceptual contents that
feature in the content of judgment and belief. The stipulations and
terminology introduced so far do not settle this substantive question.
So much for minimal stage-setting. I turn now to argue in support
of at least the claims (1) - (4), and, if it is agreed that they are true,
to consider why they are true.6
II. DISTINCT FROM CONCEPTUAL CONTENT, BUT MAKING IT AVAILABLE
We can label as conceptualists those who claim that the representa-
tional content of experience is always conceptual. Conceptualists are
committed to the falsity of the first claim (1), with its consequence
that there is a level of representational content that cannot be eluci-
dated in terms of conceptual content. Conceptualists have been
tempted to say that the phenomena for whose description noncon-
ceptual content has been invoked are better described by invoking
conceptual contents containing perceptual-demonstrative concepts,
or certain recognitional concepts. I take these two kinds of concepts
in turn, and ask whether they can really meet the conceptualist's
needs.
(a) Perceptualdemonstratives.A perceptual-demonstrative concept
that shape is a concept made available by an experience as of a
particular shape. This perceptual-demonstrative concept will, in the
context of its use in thought, refer to a shape, and the shape may be
finely individuated. Similarly for the perceptual-demonstrative con-
cepts that shade, that sound, that taste.Can the conceptualist say that it
is these concepts which enter the representational content of expe-

6
The discussion of (5) would involve more delving into the psychological liter-
ature than is feasible here. It is also significant that the case for the existence of
nonconceptual content in perception can be made without reliance on any partic-
ular thesis relating it to action.
 DOES PERCEPTION H4AVE A NONCONCEPTUAL CONTENT? 245

rience, and thereby fully capture the content that I have been claim-
ing is nonconceptual?
An initial problem with this version of the perceptual-demonstrative
route is the presence of the general concept in the perceptual
demonstrative, as the concept shapefeatures in that shape.No one can
enjoy perceptual states, or indeed any other conscious mental states,
with the conceptual content that shapein their representational con-
tent unless he possesses the general concept shape.It is, though, quite
implausible that one must have that general concept in order to
perceive objects as having various specific shapes. Nor is it at all clear
that two perceivers must have some more or less general predicative
concept in order for them both to see some object as having the same
specific shape. One perceiver may think of a presented shape as that
rectangle, the other perceiver may think of it as that straight-sidedfigure.
This difference in their thoughts need not prevent them from seeing
it in exactly the same way.
It is in any case not always required that a good, successfully
referring perceptual demonstrative contain some general concept.
For those who think it must, a useful intermediate case that forces the
softening of such a hard position is provided by the demonstrative
there.One difference between thereand that is that the former must
refer to a location, rather than anything present at that location. But
the demonstrative theredoes not do this by having the general con-
cept of place as one of its constituents. In fact, it is not at all obvious
that a thinker, in order to have the rather unsophisticated capacity to
think of a perceived location as there,must also have the general
concept place or location in his conceptual repertoire. Perhaps he
must have the resources for introducing it, on the basis of concepts
he already has: but that is different from already deploying it.
If this intermediate case is granted, it could hardly be denied that
we could conceive of a family of demonstratives, modeled on there:
maybe that-Cfor the color apparently at a location; that-T for the
surface texture there; and so forth. But I do not think this would be
a stable stopping point as a limit on the possibility of perceptual
demonstratives unsupplemented with general concepts, nor as a
suitable limit for the conceptualist position. This is so, first, because
we do not actually have such devices in our conceptual repertoire.
Our experience nevertheless still has the determinate, fine-grained
character concerning textures, shades, and the rest all the same. The
fact that we could introduce such conceptual devices just serves to
emphasize that what makes such devices available, namely, the rich
representational content of experience, exists in advance of the
246 THE JOURNAL OF PHILOSOPHY

conceptual apparatus it makes possible. That makes it implausible

that it can be identified with anything in that conceptual apparatus.
Second, there seem to be cases in which a wholly unsupplemented
perceptual demonstrative 'that' still secures reference in a suitable
perceptual context. You may think, when viewing a vase in a museum,
'That is beautiful', and be referring in thought to the color, not to the
shape or the surface texture. You may think, watching a Balinese
dancer, 'That's beautiful', and be referring to the movement or the
gesture rather than anything else. Perhaps sophisticated thinkers can
serve up general concepts if asked 'What kind of thing or event are
you thinking of?'. But it would be a stretch to insist that some general
concept must have been entering the singular demonstrative compo-
nent of their perception or their thought all along. It seems, for
instance, that someone could be introduced to the general concept
timbre,applicable to sounds, by his first having an experience leading
him to judge 'That's beautiful', referring specifically to the timbre of
(say) a clarinet. It may be that our listener only later applies the
concept timbreto the instance he had already perceived and thought
about. ('That sound' could be too unspecific to capture what he
experienced as beautiful.)
The natural treatment of such cases is that the type of the percep-
tual demonstrative involved is individuated by a demonstrative ele-
ment (which would expressed in an utterance by 'that')-an element
to which I shall return-together with some particular way that enters
the representational content of the perceptual experience that makes
the whole perceptual demonstrative available. The way in question is
the way in which the shade, or movement, or shape, or whatever is
referred to by the demonstrative, is perceived.
For any given way in which a particular quality may be perceived,
there is a specific kind such that that way is intrinsically a way for
something of that kind to be perceived. The way in which a shape is
presented-for example, as a regular diamond-is intrinsically a way
for a shape to be perceived, and not anything else. The same applies
to the way in which a texture, or a musical interval, is perceived. It
does not make any sense to suggest that a shape might be perceived
in the way a musical interval is perceived. In synaesthesia, the subject
does not literally perceive a vowel sound in the way in which a color
can be perceived. On reflection, it should not be a surprise that for
each way, there is a specific kind such that the way is intrinsically a way
for something of that kind to be perceived. Since the way contributes
to a correctness condition, this general connection between ways and
kinds is equivalent to this proposition: that for an object to have the
property a given way presents it as having, the object must meet a
 DOES PERCEPTION H4AVE A NONCONCEPTUAL CONTENT? 247

certain condition, and this condition is applicable only within objects

of a given kind.
This general connection between ways and kinds is important
because it helps to contribute to an explanation of why a general
concept is not alwaysneeded to supplement the 'that' in a perceptual
demonstrative to secure manifest reference to one kind of quality of
the perceived object or event, rather than another. The kind of
property referred to-a shape property rather than a sound property,
say-is fixed by the perceptual way that contributes to the individu-
ation of the perceptual demonstrative in question. Provided that the
demonstrative 'that' is linked to a given way, there will be a general
kind such that the reference of the demonstrative must be of that
general kind.
Ways of being perceived are inherently general. For instance, the
shapes of two different objects may be perceived in exactly the same
way, either on one occasion or on different occasions. By contrast,
perceptual-demonstrative reference to shapes themselves (to con-
tinue with that example) proceeds fundamentally via the shape of
particular presented individuals. When a thinker thinks 'That would
make a good logo', referring to a shape, he is characteristically
thinking of the shape as given in the perception of some particular
object in his experience; and the truth conditions of his thought
concern the perceived shape of that particular object. Maybe one can
refer perceptually demonstratively to the shape common to several
objects, each of whose shape one perceives, and perceives to be
the same for all of them-thus 'That shape, instantiated equally in all
of those objects, would make a good logo'. But this still involves a
tie to particulars, several rather than just one. This particularity,
which seems distinctive of, and essential to, perceptual demonstra-
tives, is one reason that one cannot simply identify the perceptual-
demonstrative sense with the relevant way of being perceived. The
way, which is general, lacks the tie to particulars which is character-
istic of the perceptual demonstrative.
The link to a particular perceived individual that has the property
demonstratively thought about is a characteristic of the demonstra-
tive element in the perceptual-demonstrative concept. It would not
be right to regard a perceptual-demonstrative concept that shape as
thought on a particular occasion as built up from a way in which a
shape may be perceived plus some entirely generic constituent. The
link to a particular, given in perception or experience, must be
captured if we are to characterize fully the perceptual-demonstrative
concept used on this occasion.
248 THE JOURNAL OF PHILOSOPHY

It is important to distinguish between ways of being perceived and

ways of having a certain generic quality. A way in which a shape
property may be perceived is to be sharply distinguished from a way
of being shaped. A way of being shaped has to do with shapes
themselves-it is a way of occupying space-and does not have to do
with the way in which shapes are perceived. On the natural, unforced
way of understanding the notion of a way of being shaped, the
property of being a square and the property of being a regular
diamond are actually the same way of being shaped. This means that
one could not individuate perceptual concepts, whether fine grained
or rough grained, in terms of which way of being shaped they are
concepts of. If one uses 'way' in this territory for 'way of being
shaped', or 'way of being colored', or more generally 'way of having
some generic kind of perceptible property or relation', one will need
additionally something else to capture the different ways in which
these ways of being shaped may be perceived.
Although a way contributes to a correctness condition, it is impor-
tant that which object is presented in a given way is not simply a
matter of the object's fitting that way. A distant aircraft in the sky may
be presented as being in a certain direction. It may not in fact be in
that direction, because the light rays are passing through refracting
bodies of differentially heated air. When an object is presented in a
given way in perception, which object is presented in that way is at
least partly a matter of causation, as H.P. Grice7 argued. It is a
consequence of this point that one cannot regard ways as individu-
ated by the condition for them to refer to some object or property. It
is, however, a highly plausible thesis about each Fregean sense (in-
cluding perceptual demonstratives) that it can be individuated by the
condition for it to refer to a given object or property.8 If this plausible
thesis is correct, then we have a further argument that ways are not
Fregean senses.
This distinctness of ways from senses is further confirmed by the
natural explication of the conditions for a perceptual-demonstrative
concept to refer to a given object. Those conditions concern the
causal relations of an object to a way in which the object is given in
perception. Such conditions take for granted the notion of a way, and
then draw upon it in elucidating the condition for the demonstrative
to refer to a given object.

7 "The Causal Theory of Perception," Proceedingsof the Aristotelian Society,Supple-

mentary Volume xxxv (1961): 121-52.
8 See Michael Dummett, The Interpretationof Frege'sPhilosophy (Cambridge: Har-
vard, 1981), chapter 3, pp. 45ff.
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 249

These remarks against treating fine-grained perceptual content as

explicable in terms of perceptual-demonstrative concepts should not
be taken to disparage the very idea of a fine-grained perceptual
demonstrative. There are perceptual-demonstrative ways of thinking
of presented shapes, colors, sounds, and the rest. Perception of an
object's shape may make available a way of thinking of that particular
shape, as 'that shape'. This is a way of thinking that also makes it
possible to wonder or to speculate whether some other object has that
very shape, the shape one perceives the presented object to have.
It is constitutive of perceptual demonstratives that they can be used
by the thinker only while the perceptual experience that makes them
available continues. McDowell suggests that such a dependence on
the occurrence of the experience "would cast doubt on its being
recognizable as a conceptual capacity at all" (op. cit., p. 57). This may
be one of the reasons he moves to endorse recognitional concepts
as the correct treatment of perceptual content. Of experience-
dependent demonstratives, he says that they look "like Wittgen-
stein's case of the person who says 'I know how tall I am', putting
his hand on top of his head to prove it" (op. cit., p. 57).9 But the
person in Ludwig Wittgenstein's example does not, by saying or
thinking 'this tall' come to know his height in meters or any
other unit, which is what knowing how tall one is requires on the
usual understanding. It is not even clear that such a person
thereby (I emphasize thereby) comes to know how tall he is, in the
sense that he could indicate which of the markings on a wall on the
opposite side of the room is roughly his height. By contrast, a
perceptual-demonstrative thought latches on to a magnitude, or
shade, or color, only if that magnitude, or shade, or color, is itself
given in the experience that makes the perceptual-demonstrative
concept available. There is no possibility of making perceptual-
demonstrative reference to a magnitude, or shade, or color, and
not knowing what magnitude, or shade, or color it is one is
thinking about. Such perceptual-demonstrative reference is the
most fundamental way of knowing what magnitude, or shade, or
color is in question.
We should also distinguish here the conditions for making demon-
strative reference in language from the conditions for employing a
perceptual-demonstrative concept. The person who utters 'that ta-
ble', pointing behind himself to a table he is not seeing, certainly

9 His reference is to Wittgenstein's PhilosophicalInvestigations, G.E.M. Anscombe,

trans. (Oxford: Blackwell, 1958, second edition), ? 279.
250 THE JOURNAL OF PHILOSOPHY

refers to a table; but he has no perceptual-demonstrative concept of

it. A similar point could be made for an utterance of 'that shade',
pointing to the wall behind oneself, without looking at it. There
are also important background conditions that must be met for
perceptual-demonstrative reference in thought to succeed. Strobo-
scopic lighting, undetected movement, and much else can under-
mine the holding of these background conditions. Perceptual-
demonstrative reference can fail even when it seems to the thinker to
succeed. (These background conditions will be all the more impor-
tant to one, like myself, who thinks that perceptual demonstratives
and observational concepts can be individuated in terms of the
conditions under which contents containing them can be not merely
rationallyjudged, but known.'0) But there are plenty of cases in which
the required background conditions are fulfilled, and there can be
perceptual-demonstrative reference to finely-individuated properties
and relations. My point has not at all been that such perceptual
demonstratives are nonexistent. Rather, my point is that they are
themselves individuated in part in terms of ways in which properties,
magnitudes, and relations are given in experience; and so cannot be
used to elucidate the nature of such ways of being experienced.
(b) Recognitional concepts.The other position tempting to some
conceptualists is that the fine-grained representational content of
experience is to be captured not by perceptual-demonstrative con-
cepts, but by fine-grained recognitional concepts. This is McDowell's
view in Mind and World,and it squares with his most explicit recent
discussion of the issues.11 McDowell writes in Mind and World:"Why
not say that one is... equipped to embrace shades of colour within
one's conceptual thinking with the very same determinateness with
which they are presented in one's visual experience, so that one's
concepts can capture colours no less sharply than one's experience
presents them?" (op. cit., p. 56). Soon after in the same book he makes
clear: "What is in play here is a recognitional capacity..." (op. cit., p.
57). The recognitional capacity associated with the recognitional
concept persists, if only briefly, after an experience which is of the

10 See my Being Known (New York: Oxford, 1999), chapter 2.

11 "Replies to Commentators," Philosophy and PhenomenologicalResearch,LVIII, 2
(June 1998): 403-31. In the part of these "Replies" which deals with me, McDowell
writes always of demonstratives, rather than recognitional concepts. But it is impor-
tant that for McDowell in his Reply, "'that way' in 'It looks that way' is predicative,
not substantival.... Expressions at this place in this construction express, rather than
refer to, ways things can look" (p. 418). On McDowell's theory, expressions in that
place express recognitional concepts. This is fully in accord with his position in
Mind and World-see p. 57.
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 251

kind recognized (op. cit., p. 57). McDowell seeks to capture part of the
fine-grained spatial content of experience with a conceptual content
of the following sort, which he would say is represented as correct by
the experience:
This is shaped R
Here R expresses, for McDowell, a recognitional concept of a way of
being shaped. Similarly for color, on McDowell's view we have the
conceptual content
This is colored S.
in the representational content of the experience, where S is a
recognitional concept of a fine-sliced shade.
If it was wrong to have the general concept shape in the demon-
strative attempts to capture the fine-grained content of experience, it
seems equally wrong to have the general relational concept shaped
in the representational content under this conceptualist
treatment. The same applies to the concept colored . I shall
not, however, trade on this point, and shall take it that McDowell
could use precisely the points I made earlier about the absence of a
need for supplementation by a general concept. He could move
simply to propose the conceptual contents 'This is R', 'This is S' as
the fine-grained representational content of experience. The recog-
nitional capacities underlying these recognitional concepts persist,
"possibly for quite a short period," according to McDowell in Mind
and World(op. cit., p. 172), where he holds that they can be employed
in the absence of perception of the finely-sliced properties to which
they refer.
As Diana Raffman'2 observes, it is well known that perceptually-
based discrimination of properties far outreaches memory and iden-
tification of those properties. Recognitional capacities are by their
very nature constrained by memory capacities. I would add to Raff-
man's point only two observations. First, even if memory were as
finely discriminating as perception, that still would not make it right
to regard recognitional concepts as entering the representational
content of experience. There cannot be recognition when the per-
ceptual property is encountered for the first time in a given way. A
first-time experience of a property nevertheless still has a specific,
fine-grained representational content. And second, in those counter-
factual circumstances in which memory were as discriminating as

12 "On the Persistence of Phenomenology," in Thomas

Metzinger, ed., Conscious
Experience(Paderborn: Schoeningh, 1995), pp. 293-308, pp. 294ff.
252 THE JOURNAL OF PHILOSOPHY

perception, the experience would still have a specific content which

could exist even if the recognitional capacity did not. The memory
structures required for the recognitional concept and its correspond-
ing recognitional capacity, and their subpersonal underpinnings,
could all be absent, and yet the experience still have a fine-grained
representational content. The specific fine-grained content would
show up in matching tests that do not need to rely on memory, as
opposed to current perception. This modal point applies however
short the period for which the recognitional capacity is said to exist.
It seems to me, then, that neither perceptual demonstratives nor
recognitional concepts can capture the representational content of
perception. But however convincing it may be, candidate concept by
candidate concept, that each candidate concept cannot capture per-
ceptual content, it is reasonable to ask here for more theoretical
explanation. Is there some deeper, and more general, reason why
perceptual content cannot be captured by any kind of conceptual
content? I shall return to this, when we are in a better position to
characterize more fully the nature of the distinction between concep-
tual and nonconceptual content.
III. LEARNING OBSERVATIONAL CONCEPTS
Now let us take the phenomenon of learning a new observational
concept from instances of it that are perceived when one is taught, or
otherwise acquires, the concept. Consider the representational con-
tent of the subject's experience when she is learning the concept
pyramid, considered as an observational shape concept.
On the one hand, the representational content of this experience,
in the case of learning from a positive instance, must be sufficient for
someone rationally to apply the concept-must entitle her to apply
the concept-when experience is being taken at face value. If that
were not so, it would not be an observational concept after all. Yet on
the other hand, this representational content cannot include the
concept pyramid, for that would require the learner already to possess
the concept. The learner could have such experiences only if the
lesson were unnecessary.
The natural solution to this last quandary is to acknowledge that
there is such a thing as having an experience of something as being
pyramid shaped that does not involve already having the concept of
being pyramid shaped. What such an experience will have is a non-
conceptual content which, if correct, is sufficient for something's
falling under the observational concept pyramid.
In fact, that is to understate the case. Not merely is the holding of
the nonconceptual content represented as correct in the positive
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 253

cases involved in learning sufficient for the object in question to be a

pyramid. The experiences from which, as presenting positive in-
stances, one can rationally learn the concept, are those which have
nonconceptual representational contents of the sort mentioned in
the very condition for possessing the observational concept that is
being learned. There is a specific kind of nonconceptual content of
experience to which one's judgments 'That's a regular diamond'
must be rationally sensitive if one is to possess the observational
concept regulardiamond.There is a different specific kind of noncon-
ceptual content of experience to which one's judgment 'That's a
square' must be rationally sensitive if one is to possess the observa-
tional concept square. There is a specific kind of nonconceptual
content of experience to which one must be sensitive if one is to
acquire a recognitional concept for the musical interval of an aug-
mented fourth. It is a different kind from that to which sensitivity is
required in possessing a recognitional concept for a diminished fifth;
and so forth.'3
This rational sensitivity is not a matter of inference, even though
the occurrence of an experience with a certain nonconceptual con-
tent does make rational a perceptual judgment with a conceptualized
content. It is intuitively wrong to classify the case as one of inference.
It also could not be correct, if the perceptual content is nonconcep-
tual: inferential relations can hold only between (states with) concep-
tual contents. We can, as reflective thinkers, indeed introduce ways of
thinking about the nonconceptual ways in which things are given. But
this is built on a more primitive level of rational sensitivity to the
experience's nonconceptual content itself. Such reflective thought is
not necessary for having the capacity to make perceptual judgments
about one's environment.
IV. MAKING RATIONAL TRANSITIONS TO PERCEPTUAL JUDGMENTS
I have emphasized that the way in which some thing, or property, or
relation is given in the nonconceptual content of an experience is
something that contributes to what it is like to have that experience.
These ways which feature in nonconceptual content are then at the
conscious, personal level, and are not merely subpersonal. As features
of the subjective experience, their presence can entitle a thinker to
make a particular judgment, or to form a certain belief.

13
For some observational concepts, the kind may be amodal, and so instantiated
in experiences in more than one modality. This is plausible for straight and right-
angled, construed as observational concepts.
254 THE JOURNAL OF PHILOSOPHY

A thinker can be rational in making a transition from an experi-

ence with a certain nonconceptual representational content to a
judgment with a certain conceptual content, in particular in making
a transition to judging a content in which an observational concept is
predicated of presented objects or events. Such a transition is rational
when the thinker is entitled to take her experience at face value, and
when the observational concept is individuated in part at least as one
that the thinker must be willing to judge when experience has a
certain kind of nonconceptual representational content (and is being
taken at face value). Consider Mach's example of a closed right-
angled figure formed from four straight lines, and whose symmetry
about the bisectors of its opposite angles is perceived. Such an
experience can make rational the judgment That'sa regulardiamond.
It does not make rational the judgment That's a square.That latter
judgment is made rational rather by a different nonconceptual con-
tent, in which the perceived symmetry of the closed figure is a
symmetry about the bisectors of its sides. Which symmetries are
perceived in each case does not need to be something known to, or
even conceptualized by, the perceiver.
The nonconceptual content in question throughout this paper is
nonconceptual representational content, with a correctness condi-
tion. Consider any case in which an observational concept F is plau-
sibly individuated partly by its relations to a particular sort of
nonconceptual content. It is then a philosophical task to show that
the holding of the correctness condition for the nonconceptual
content in question ensures the holding of the correctness condition
for the conceptual content That'sF In our simple spatial cases, one
could show this by showing that the shape which something must be
when it is presented (veridically) in an experience with a certain
nonconceptual content is a shape which is sufficient for the observa-
tional concept is squareto apply to it.
This is an instance of a task that exists for any rational transition
that involves a concept or concepts at some point or other. The task
exists equally for the case of logical transitions. If a premise gives a
good reason for accepting a certain conclusion, because of the logical
form of the inference, there must be some explanation, in the
semantics of the logical constants involved in the transition, of why
this is a valid transition. That is why we need a semantics, and not
merely a proof theory, for a logic. (Even in the special case in which
the semantics is itself proof theoretic, the need for a semantical
theory is still being granted.) The ordinary thinker does not need to
know what the explanation of the validity of a particular transition is,
in either the logical or the perceptual case. But the explanation must
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 255

exist if the transition is to be valid. Only so is there an explanation of

why it is truth preserving, and thereby of how it can lead to new
knowledge.
My own view is that it is always a task of philosophy to explain why
a transition is a good one. This requirement, in my view, applies to
the conceptual-to-conceptual case of logic; to the nonconceptual (but
still representational)-to-conceptual case of perceptual judgment;
and to the arguably wholly nonrepresentational-to-conceptual transi-
tion from the occurrence of a pain in a subject to his conceptualized
self-ascription of pain.'4
The general position on nonconceptual content that I have out-
lined so far may seem intuitive, but it has been subjected to a
philosophical critique by McDowell. His most fundamental objection
is that experiences can have a rational bearing on thought only if
their content is conceptual: "The point of the claim that experience
involves conceptual capacities is that it enables us to credit experi-
ences with a rational bearing on empirical thinking" (op. cit., p. 52).
This claim needs an argument, and McDowell's argument for it in
Mind and Worldis that rational thinking involves the possibility of
scrutinizing the relations between experience and judgment. Noth-
ing is outside the limits of rational self-scrutiny (op. cit., p. 52), even
if, after scrutinizing the case in question, we decide to endorse it.
McDowell's view is that it is hard to see how this scrutiny would be
possible unless experience has only conceptual content (op. cit., pp.
52-53).
That last step seems to me a nonsequitur. I accept that we can
rationally scrutinize any instance of any putative reason-giving rela-
tion, including the relation between experience and judgment. Mc-
Dowell's concern is that on a theory like mine, the subject cannot
gain the required "comprehensive view of the two contents"-that of
the experience and that of the judgment (op. cit., p. 166). On my
treatment, however, a thinker can ask 'Is something's looking that
way a reason for judging that it's square?', for instance. On the
approach I advocate, 'that way', in this particular occurrence, refers
demonstratively to a way in which something can be perceived. The
reference itself is made bysomething conceptual: demonstrative con-
cepts can enter conceptual contents. There is no requirement that

14
I hope this and the preceding paragraphs clarify my intentions at around page
80 of A Study of Concepts,about which McDowell expresses some puzzlement: "it is
hard to see much in the way of a further issue about how the reason [supplied by
an experience with an allegedly nonconceptual content] can be a good one"
("Replies," p. 418).
256 THE JOURNAL OF PHILOSOPHY

the reference of the demonstrative be conceptualized. (I shall return

to such uses of 'that way' in a few paragraphs.) So thought can
scrutinize and evaluate the relations between nonconceptual and
conceptual contents, and obtain a comprehensive view of both. On
McDowell's theory, of course, the demonstrative 'that way' expresses
a sense (a recognitional concept), which does on his view enter the
representational content of experience. I have rejected that treat-
ment on other grounds-the required recognitional capacities do
not exist, and do not need to exist for the experience to have a
fine-grained content. But quite apart from that, the mere possibility
of the treatment I have offered shows that it is a nonsequitur to move
from an agreed premise of the required scrutinizability in conceptual
thought of the rational relations between experience and judgment
to the conclusion that experience itself must have only conceptual
representational content. As long as we can think about the noncon-
ceptual representational content, as we certainly can, we can have the
required scrutinizability.
McDowell also holds that there is a requirement of articulability of
reasons, including perceptual reasons, and that this too supports his
view that the representational content of experience must be wholly
conceptual. "[Peacocke] has to sever the tie between reasons for
which a subject thinks as she does and reason she can give for
thinking that way. Reasons that the subject can give, in so far as they
are articulable, must be within the space of concepts" (op. cit., p. 165).
I am not at all sure myself that there is such a requirement of
articulability. But if there is, it seems to me that the use of the
demonstrative 'that way', construed in the way it is on my approach,
would allow the friend of nonconceptual content to meet any such
demand. 'I believe it's square because it looks that way' is an articu-
lation of the subject's reasons. Once again, the conceptual character
of the conceptual constituent 'that way' must be sharply distinguished
from the nonconceptual character of its reference, a nonconceptual
way in which something is perceived.
I would invoke demonstrative ways of thinking of ways in address-
ing a closely related objection of Brewer's (op. cit., pp. 166-69). He
argues that a thinker must be capable of recognizing her reasons as
such, that is, as reasons. This demand does not, Brewer argues,
intrinsically raise difficulties of principle in the case of a mental state
that gives reasons, provided that the state has conceptual content. But
in the case of a mental state with an allegedly nonconceptual content,
6mere possession of the state involved in having such a 'reason' is
explicitly insufficient for the subject's understanding of the proposi-
tion whose association with this state grounds its putative status as his
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 257

reason for doing what he does.... The only alternative seems to be that
he must have some second-order knowledge of the relation between
mental states of the type in question and the truth of the belief...for
which he thereby, and only instrumentally, recognizes his having a
reason" (op. cit., pp. 167-68). Brewer says that the nonconceptualist is
"condemned to follow the discredited classical foundationalists and
coherentists" (op. cit., p. 166). In short, his position is that the
nonconceptualist can provide for the recognition of his own reasons
only by arguing as follows:
(i) This state is F
(ii) Anythingthat is F is a reason for believingthat p.
(iii) Hence I have reason to believe that p (op. cit.,p. 168).
Brewer adds that "what stands in place of 'F' will have, at best, to
be hideously complex" (op. cit., p. 168).
I agree with Brewer that, if the nonconceptualist could account for
the recognition of the reason that his experience gives him for
making a perceptual judgment only in this second-order, instrumen-
tal fashion, the position would be unacceptable. But I maintain that
the nonconceptualist's position does not have the consequence
Brewer claims. Here is a way that a thinker can come to recognize a
state with nonconceptual representational content as giving reasons,
without taking the second-order, instrumentalist route. The percep-
tual experience represents some presented object or event as having
some property, given in a certain way W So the experience has a
content (if we put it, conveniently but inessentially for the present
point, in object-dependent form):
(iv) x, given in ways, has propertyP, given in way W.
The way s in which the object is given will capture its apparent
distance and direction, egocentrically characterized, from the per-
ceiver.15Now suppose this way Wis, for example, one of the noncon-
ceptual ways in which a shape can be given, and is mentioned as
sufficient for being square in the possession condition for the con-
cept square.On my account, the subject is then entitled to move from
(iv) to the conceptualized content:
(v) That object [given in ways] is square.

15 No doubt the way Wcould also be said in some pre-theoreticsense to be part

of the way the object is given in perception. But it is important to distinguish
between those wayswhich help to determine whichobject is perceived from those
which do not. s is in the former class-though as emphasizedearlier, this determi-
nation is not a matter of fitting.
258 THE JOURNAL OF PHILOSOPHY

Similarly for any other way whose presence in an experience gives

good reason for applying an observational concept.
Now, what would be involved in a subject's reflecting on these
transitions, and in coming to appreciate his reasons as reasons? (iv)
and (v) do not seem to be hideously complex; complexity is not the
heart of the issue. But there are two interesting capacities a thinker
must have if she is to be capable of appreciating both the occurrence,
and the rationality, of her transition from (iv) to (v).
First, our reflective subject must have a special way of thinking of
the way W There seems to be a phenomenon here that is a partial
analogue, for the relations between nonconceptual contents and
demonstrative concepts of them, of the relations that hold between
concepts and canonical concepts of those concepts-"conceptual
redeployment," as I called it.16 When some property is given to a
thinker in perception in way W, then, if that thinker is sufficiently
reflective, there seems to be a conceptual way CQ0 of thinking of that
way W, where this conceptual way C,, seems to be made available to
the thinker precisely by her enjoying an experience whose noncon-
ceptual content includes that way W Presumably, this relation of
'being made available by' consists at least in part in a certain canon-
ical sensitivity of judgments involving the concept Cw to evidence
about the way W which the subject has precisely because she is
experiencing some property as given in way W
Second, in appreciating that the occurrence of an experience with
the representational content (iv) entitles her to judge the content
(v), our thinker will be displaying a sensitivity to a particular bound-
ary. This is the boundary between those states which are, and those
states which are not, mentioned in the possession condition for a
given concept such as square. Our thinker will have moved from the
ground level of satisfying a possession condition to the level of
thinking about it. One, perhaps the basic, way to make this transition
is to ask questions that are in the first instance not about concepts,
but about the world. Such a question might be 'Would something's
looking that way Wgive reason to think it is square?' One can answer
such questions by drawing on one's ordinary, ground-level abilities to
react rationally to one's perceptual states in coming to make judg-
ments about the world. Someone who satisfies a possession condition
can clamber her way up to reach some understanding of what that
possession condition is.

16 See my Being Known, sections 5.5-5.6.

 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 259

This description of the transition from perception to judgment,

and of the subject's appreciation thereof, is not second order. Nor is
it of the highly problematic kinds in the sorts of classical foundation-
alist and coherentist writings Brewer rightly criticizes. (iv) and (v) are
not themselves about mental states at all. It is true that (iv) becomes
pertinent because of its relation to the content of a mental state. But
that will equally hold on a conceptualist's account of the reason
relation and of appreciation of it as a relation of reason giving. The
content is conceptual, according to that theorist, and stands in vari-
ous reason-giving relations appreciable by the thinker. But that con-
tent is relevant only because it is the representational content, on the
conceptualist's view, of the perceptual experience in question. That it
is relevant only because of that relation does not make the concep-
tualist's view second order, or instrumentalist. Equally the corre-
sponding feature of (iv) and (v) cannot make the nonconceptualist's
account second order or instrumentalist either. It is the relation
between (iv) and (v) which is crucial in the account of the reason-
giving relation, and appreciation of it as such; and that does not
involve anything second order.
A full theoretical, philosophical account of why the transition from
(iv) to (v) is valid would of course involve some philosophical theory.
The theory would be part of a general theory of content. A full
theoretical account would equally be required in the case of logical
transitions between purely conceptual contents. The philosophical
theory does not need to be known to the ordinary thinker in either
of these rather different cases. In both cases, there is a willingness to
make transitions between certain contents, which is constitutive of
possession of the observational concept or the logical concept, re-
spectively. On my account, in the former case the content of the
entitling or reason-giving state is nonconceptual, and in the latter it
is conceptual. The ordinary thinker can appreciate his reasons as
reasons in either case. In either case, he may or may not go further
in reflecting on the nature of the good reasons involved.
A different question about this approach does not query the dis-
tinctions I have drawn, but raises at least a problem about how the
issues have been framed.'7 Surely, I would not object to a regimen-
tation of the content of a conceptual state in which, for instance, a
recognitional concept for a particular size of diamond shape were
captured by the very nonconceptual content mentioned in its posses-
sion condition? But if we can use nonconceptual content in capturing

17
Here I am indebted to Stephen Schiffer.
260 THE JOURNAL OF PHILOSOPHY

conceptual content, has the issue with the conceptualist been prop-
erlyjoined when we take it as the question of whether experience has
only conceptual content? I agree that we could employ such a means
of capturing a recognitional concept. But if we do use the apparatus
of nonconceptual content in characterizing conceptual states, we
must distinguish sharply between the kinds of relations in which a
state must stand in order to possess the original nonconceptual
contents, and those in which it must stand in order to possess those
conceptual contents which are, under this proposal, captured by
nonconceptual contents. The arguments of the next section will
further distinguish these two sets of relations.
V. KINDS OF CONTENT, ANIMAL PERCEPTION, AND OBJECTIVITY
Cats, dogs, and animals of many other species, as well as human
infants, perceive the world, even though their conceptual repertoire
is limited, and perhaps even nonexistent. These perceptions are
subserved by perceptual organs, and in the case of higher species,
subserved by brain structures similar in significant respects to
those which subserve mature human perception. By the "hard
line" on animal perception, I mean the thesis that none of the
conscious perceptual states with representational content enjoyed
by mature humans can be enjoyed by nonlinguistic animals with-
out concepts, or with only minimal conceptual capacities. By the
"soft line," I mean simply the denial of the hard line. So the soft
line says that some of the conscious perceptual states with repre-
sentational content enjoyed by mature humans can be enjoyed by
nonlinguistic animals without concepts, or with only minimal con-
ceptual capacities.
For what it is worth, pretheoretical intuition seems to find the hard
line too hard to accept. The hard line entails that the following
cannot be literally true: that the animal has a visual experience as of
a surface at a certain orientation, and at a certain distance and
direction from itself, in exactly the same sense in which an adult
human can have a visual experience with that as part of its content.
Abandoning all pretense at unbiased terminology, I shall call the
conclusion that that cannot be literally true the unintuitive conclusion.
The soft line, which prima facie at least is not committed to the
unintuitive conclusion, is naturally developed hand-in-hand with a
theory of nonconceptual representational content. According to the
soft line, it is nonconceptual representational content which can be
common to visual experiences which both you and a mere animal
may enjoy.
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 261

Those who take the hard line need not deny that animals have
some perceptual sensitivity to their environment, a sensitivity which
explains their actions. McDowell, who takes the hard line, insists that
animals have such a perceptual sensitivity: "It is a plain fact that we
share perception with mere animals" (op. cit., p. 114).18 Nor does
McDowell have in mind, for instance, only some kind of informa-
tional sensitivity that would make animals automatons of the sort
which Ren6 Descartes apparently believed them to be. In Mind and
World,McDowell speaks of "proto-subjectivity"(op. cit., p. 117). He
says that there is a legitimate kind of talk of what the features of the
environment are for an animal, which "expresses an analogue to the
notion of subjectivity, close enough to ensure that there is no Carte-
sian automatism in our picture" (op. cit., p. 116). All the same, it must
be only an analogue on McDowell's view. It is, on his account, not
literally true that the mere animal has a visual experience as of a
surface at a certain distance and direction in exactly the same sense
that mature concept-using humans do.
It is certainly a necessary condition of being reasonable in rejecting
the unintuitive conclusion that one address the arguments that in
perceptual experience, there is only conceptual, and not nonconcep-
tual, content. But if we can develop a theory of nonconceptual
representational content, and answer the conceptualist's objections
to it, as I have been trying to do, then much of the pressure to adopt
the unintuitive conclusion is substantially relieved.
Although I think, however, that the unintuitive conclusion is in-
deed to be rejected and the soft line is right, the truth in this area
seems to me much more interesting than those somewhat flat con-
clusions suggest. It seems to me that there is a good, Kantian point in
McDowell which should not get lost in the endorsement of the soft
line. There is something very plausible in McDowell's Kantian posi-
tion that "the objective world is present only to a self-conscious
subject, a subject who can ascribe experiences to herself.... It is.. .the
power of conceptual thinking that brings both the world and the
self into view. Creatures without conceptual capacities lack self-
consciousness and-this is part of the same package- experience of
objective reality" (op. cit., p. 114). Although of course a creature with
perceptions with nonconceptual representational contents has states
whose correctness conditions concern the objective world, it is a
question whether a creature without the specific conceptual appara-
tus McDowell mentions would be conceiving-would even have the

18 See also pp. 50, 64-65, and Lecture VI, section 4-7.
262 THE JOURNAL OF PHILOSOPHY

resources for conceiving-this world as objective. So the question

arises: Is there some way we can consistently acknowledge this point
while also maintaining the soft line? I shall be arguing that there
is-but to do so we need some deeper theory of the nature of the
distinction between conceptual and nonconceptual content.
I first draw a distinction within the possible grounds, or reason-
giving states, which lead, in a given subject's psychological economy,
to acceptance of a given kind of content. We can distinguish those
grounds being exhaustive,for a given content, and their being canon-
ical. Consider the minimal possible ways a creature capable of per-
ceptual states with nonconceptual content may come to accept of
three perceived objects a, b, and c that they are in a straight line, or
may come to reject that same content. The content may be accepted
on the basis of a perception with a nonconceptual representational
content to that effect. Such a creature may also store this information,
with or without memory images of the original perception. The
content may come to be rejected if, in moving or being moved
around, the subject does not have perceptions of the sort which, from
other angles, should be produced by a, b, and c if they are in a straight
line. Now suppose such perceptual and cognitive-map based ways of
coming to accept or reject exhaust the subject's ways of coming to
accept or reject such contents (or at least, exhaust those which are
distinctive of those contents, if this subject is capable of engaging in
more general modes of reasoning). Such exhaustiveness of the per-
ceptual-cum-cognitive-map based ways of coming to accept or reject
those contents contrasts sharply with the case of possession of the
concept of being in a straight line. For a thinker with the concept
collinear,any grounds for thinking that three things stand in the same
relation as is perceptibly instantiated when the thinker sees three
things to be collinear are equally grounds forjudging those first three
things to be collinear. There are no restrictions on what those kinds
of grounds may be. What those grounds may be is something to be
discovered. The grounds are in no way restricted to what can be
attained through perception and the use of a cognitive map. These
additional grounds need not be anything particularly sophisticated.
Even the child who moves some straight-edged physical object to
check whether three objects are in a straight line, when he cannot tell
this, for one reason or another, just by looking, is manifesting a
sensitivity to this open-endedness of grounds.
The presence in the thinker's thought of possible additional
grounds is entirely consistent with perception having some canonical
status in the individuation of the concept collinear,considered as an
observational concept. Some grounds can have a privileged status,
 DOES PERCEPTION HAVE A NONCONCEPTUAL CONTENT? 263

consistently with there being no restrictions in principle in a thinker's

thought on the range of possible grounds. The fact that certain
grounds are canonical does not mean that they are exhaustive. We
must distinguish the exhaustive from the (merely) canonical.
It is important to notice two points about the distinction between
the exhaustive and the canonical. First, when a certain body of
grounds is not exhaustive in principle in a thinker's thought, the
thinker is tacitly operating with the notion of the same property, or
same relation, in the world. It is the same relation that can be
detected, on the one hand, by perception with the use of cognitive
maps and, on the other, by the additional means-the ruler, or
whatever else the ingenious thinker may come up with. The thinker
for whom grounds involving perception-cum-cognitive maps are not
exhaustive for acceptance or rejection of a given kind of content must
be capable of kinds of reasoning that involve grasp of the notion of
sameness of property and sameness of relation. The same holds for
the case of sameness of objects, if we apply these points to the
distinction between genuine concepts of objects, and the ways objects
are given in the nonconceptual content of perception.
The second point to notice about this distinction is that it has to do
with the nature of objectivity. It is a consequence of the idea of things
being objectively thus-and-so, of its being objectively the case that
these objects stand in this relation, that their doing so is something
that can be known about in many different ways. It is not restricted in
principle to any one mode of access. We do not have fully objective
thought unless a thinker manifests some grasp of a conception of
objectivity with this character.
It is here that my discussion connects up somewhat more closely
with McDowell, despite our major differences. For he insists that only
with conceptual contents does one have a certain kind of objectivity-
and I have recently been asserting just that. The reasons are not, of
course, precisely the same. I have not, for instance, committed myself
to the view that there is properly objective thought, thought about
something conceived as objective, only if the subject actually has the
concept of perceptual experience. All the same, I am in agreement
with McDowell that the minimal requirements for thought about an
objective world that is in some way conceived as objective are not met
simply by the possession of perceptual states with nonconceptual
representational content.
A similar point holds also for a subject's conception of himself,
herself (or perhaps, in the case of some mere animals, itself). The
reasoning is essentially the same: it relies again on the distinction
between the exhaustive and the canonical. Suppose a thinker's only
264 THE JOURNAL OF PHILOSOPHY

way of establishing contents about himself is by the use of the follow-

ing:

(a) By perception-'I am in front of a lake', 'I am under a tree'.

(b) By memory- 'I was in front of a lake', 'I was under a tree'.
(c) By integrating the deliverances of perception into a cognitive map
and extracting the consequences thereof-'There's a wood behind
me', 'There's a lake an afternoon's walk from the wood in that
direction'.

That is, for this thinker perception, memory, and the operation of
his cognitive map in the ways illustrated above exhaust his ways of
establishing things about himself.19 This is not yet to have a full
conception of himself, as an object in the objective order of things. In
particular, it involves no conception of what it would be for some-
thing to be true of himself that is not establishable as true by the
above methods. Contents not so establishable include 'When I was
asleep, I rolled over four times', 'There are times I existed of which
I know nothing'. Full objective thought about oneself must, as in the
case of conceptual thought about properties and relations, involve
some conception of oneself as an object truths about which are not
exhausted by those which can be established in a certain restricted set
of ways. The subject who is oneself may exist in times, places, and
circumstances to which one's current means of establishing first-
person contents cannot reach.
I conjecture that this distinction, in point of grasp of objectivity,
between the minimal requirements for having states with nonconcep-
tual contents, and what is involved in conceptual content, is a deeper
reason why perceptual content cannot be explained in terms of
conceptual content. The most primitive aspects of representational
content in perception, which our subjective experience shares with
the mere animals, do not involve the grasp of objectivity required for
conceptual content. This is one of the reasons that trying to treat all
perceptual content as conceptual involves an overascription. We
should always distinguish between content that is objective, and
content which is not only objective, but which is also conceived of as
objective.
CHRISTOPHER PEACOCKE
New York University

19
On nonconceptual self-consciousness, see Hurley, chapter 4.