Forthcoming in Robin Jeshion (ed.), Essays on Singular Reference (Oxford: Oxford University Press).

DEMONSTRATIVE REFERENCE, THE RELATIONAL VIEW OF

EXPERIENCE AND THE PROXIMALITY PRINCIPLE

BY JOHN CAMPBELL

This paper is about perceptual demonstratives, terms like ‘this’ and ‘that’ used to refer to

a currently perceived object. How is it that you know what such a term refers to? The

natural answer is that your perceptual awareness of the object is a relation to it that

provides you with knowledge of what the term stands for.

In a recent essay, Tyler Burge has argued that this relational view of experience

conflicts with what he calls the ‘Proximality Principle’, that he thinks is a consequence of

results in vision science. I will argue that this is precipitate. Bringing classical vision

science to bear on the characterization of perceptual consciousness is not straightforward.

In this paper my aim is not to motivate the relational view of experience from the

ground up. Rather, I want to look at whether it can withstand Burge’s attack. In §1 and

§2, I will rehearse some of the main ideas of the view. In §3, I will look at Burge’s

Proximality Principle itself. §4 and §5 diagnose the problems with this Principle.

1. Causing and Justifying the Pattern of Use of a Demonstrative

2

Knowledge of the reference of a demonstrative causes and justifies the pattern of use that

you make of the term. I begin by explaining this claim. One way to begin on the

explanation is to consider a parallel between perceptual demonstratives and the classical

propositional constants.

In the case of a classical propositional constant, such as ‘or’, we can distinguish

between the truth-table for the sign, and the rules of inference that govern it. So for

example, the classical elimination-rule for ‘or’ is:

[A] X [B] Y

C C

A or B, X, Y therefore C

And the introduction rules are:

A B

A or B A or B

These rules define what I will call the ‘pattern of use’ for ‘or’. They tell you how to

confirm an ‘or’ statement, and they tell you the implications of an ‘or’ statement. These

rules give a concise summary of the inputs to an ‘or’ judgement, and the outputs from an

3

‘or’ judgement. We shall see that we can give a parallel characterization of the pattern of

use of a perceptual demonstrative.

A classical propositional constant, like ‘or’ makes a difference to the truth-

condition of any statement containing it. A characterization of the ‘semantic value’ of

the constant is a description of the contribution that the constant makes to the

determination of a statement containing it as true or false. A classical propositional

constant has a semantic value by being associated with a truth table. In the case of ‘or’,

the classical truth table would of course be:

A B A or B

T T T

T F T

F T T

F F F

So we can make a distinction between the pattern of use of a propositional constant,

which is described by the rules of inference for it, and the semantic value of the constant,

which is supplied by its being associated with a truth table. I said that we will be able to

describe the pattern of use of a demonstrative. We will also be able to contrast that

pattern of use with the semantic value of the demonstrative, which is supplied by its

being associated with an object as its referent.

On a classical view, the pattern of use for a propositional constant is justified by

appealing to the truth-table for that constant. By appealing to the truth-table, you can see

4

that the use of the introduction-rule for ‘or’ will indeed preserve truth from premises to

conclusion. And indeed, you can see that this is the least demanding introduction rule

that does guarantee the truth of the premises. Similarly, by appealing to the truth-table,

you can see that the elimination-rule for ‘or’ preserves truth from premises to conclusion.

And you can see that it is the strongest elimination-rule to do so. If we suppose a speaker

who knows the truth table for a propositional constant, we can say that this speaker

knows what justifies the pattern of the use of the constant; this speaker knows the truth

table that justifies the associated rules of inference.

This familiar case provides a model for our knowledge of the reference of a

demonstrative. Just as we can characterize the introduction and elimination rules for a

logical constant, so we can characterize the inputs and outputs of a perceptual-

demonstrative judgment. Just as the constant has a semantic value by being associated

with a truth table, the demonstrative has a semantic value by being associated with an

object as its reference. Just as knowledge of the truth table justifies your use of the

standard introduction and elimination rules for ‘or’, so too knowledge of the reference of

a perceptual demonstrative justifies your making the pattern of use of it that you do.

Indeed, in one respect the classical model works better in the case of perceptual

demonstratives than it does in the case of propositional constants. It is not obvious what

it comes to, that the ordinary speaker has knowledge of the truth table for a propositional

constant; what is it to have knowledge of a truth table? Can we make anything of the

idea that the ordinary speaker is somehow confronted with the truth table that justifies

that pattern of use of the constant? In the case of a perceptual demonstrative, however,

there is no analogous puzzle. What is it to be confronted with the referent of a perceptual

5

demonstrative? Consider an ordinary perceptual demonstrative, like ‘that mountain’, or

‘that tree’, or ‘that woman’. Well, there the thing is – the mountain or tree or person or

whatever it is - staring you in the face. The idea that in this case you know what the term

stands for seems peculiarly inescapable.

There are two points here. One is that perception of the reference of a

demonstrative confronts one with that which does in fact justify the particular pattern of

use that one makes of the term. Second, this is not just a coincidence. Typically, your

knowledge of reference is what causes you to use the demonstrative in the way you do. I

sum these points up by saying that knowledge of reference causes and justifies the pattern

of use of a term.

One reason that the classical propositional constants are easy to think about for us

as theorists is that for them it is easy to give a concise characterization of the pattern of

use that has to be justified. It is not so straightforward to describe the pattern of use of a

perceptual demonstrative. How should we go about it? We have a clue from vision

science. Much experimental work on vision goes like this. The subject is asked to make

a verbal report, or to make some sign equivalent to a verbal report, involving a perceptual

demonstrative. ‘Push the button when you see that square turn red’, for example. So a

push on the button is equivalent to the judgment, ‘that square is red’. Theorizing about

the results of these experiments is typically an attempt to find the information-processing

inputs to such a demonstrative judgment. So we can use the results of this approach to

describe the basis on which we make use of a demonstrative, what the perceptual inputs

are to the use of the demonstrative. This will provide us with a parallel to the

introduction rule for a logical constant.

6

In current vision science, the dominant approach to this question is work in

Treisman’s tradition, which supposes that the basis for a demonstrative verbal report of a

perceived object, ‘that is red’, will include the content, ‘Redness at place p’, in which the

feature redness is located on a feature map. So we could state the introduction rule for

the demonstrative in some such terms as this:

FEATURE MAP: Redness at place p

JUDGMENT: That is red

This gives us the introduction rule for the demonstrative. We can take a similar approach

to the output rules for a visual demonstrative. What is the output of a visual

demonstrative judgment, such as ‘That is red’? Actions are outputs from such judgments.

For example, if you want to pick up something red, then once you have formed the

visual-demonstrative judgment, ‘that is red’, you are able to do it. So the judgment, ‘that

is red’ is able to identify the object for the benefit of the motor system; and the evidence

suggests that once again, location is critical (Jeannerod 1997). So as an output rule for a

visual demonstrative, we might consider:

JUDGMENT: That is red

MOTOR SYSTEM: To reach something red, move towards p

The question now is, how should we characterize knowledge of the reference of a

demonstrative? What provides you with the knowledge that causes and justifies this

7

pattern of use for the demonstrative? It is here that we find the role of perceptual

experience. It is by experiencing the thing that you are confronted with the referent

whose association with the demonstrative constitutes the semantic value of the term.

And, evidently, it is not merely having the thing somewhere in your visual field that

provides this knowledge of reference; the thing might be somewhere in your visual field

even though you had never noticed it. It is, rather, conscious visual attention to your

object that provides you with your knowledge of the reference of the term, that causes

and justifies the pattern of use that you make of it. These points summarize a much fuller

discussion in Campbell 2002. I rehearse them here to orient the following discussion.

My main points now are:

(a) this conception of the role of experience in providing knowledge of the reference

of a demonstrative term requires us to think of perceptual experience in relational

terms, and

(b) Burge’s ‘Proximality Principle’ is not supported by experimental results in vision

science, and does not threaten the natural picture of our knowledge of the

reference of a visual demonstrative.

2. The Role of Experience of the Object in Justifying Pattern of Use

8

What makes it possible to regard the pattern of use of a demonstrative as given a

semantic anchoring by your knowledge of what the term stands for? So long as we think

of experience as a kind of glow that accompanies representations, or as a kind of structure

to the representations we have, the idea of appealing to experience of objects to justify

pattern of use seems hopeless. On this kind of picture, you have the network of

representations in your mind or brain, and you make the use of them that you do, but the

fact that you have conscious experience adds nothing that could possibly add up to a

semantic foundation for the pattern of use of a demonstrative.

What changes the situation is the possibility of appealing to a relational

conception of experience. On the relational conception, experience is a relation between

the perceiver and a distal element of the environment, the object referred to. On the

relational conception, the objects and properties about which we think demonstratively

can constitute the qualitative characters of our experiences.

There could be different patterns of use associated with two demonstratives used

to refer to the same object. For example, if you see one person right in front of you, and

a person reflected in a mirror, you might perfectly well associate different patterns of use

with the two demonstratives ‘that man (right in front of me)’ and ‘that man (seen in the

mirror)’. You might be able to find out a lot about each individual, even though you are

not quite sure whether they are one and the same. In this kind of case, we would say that

you are visually experiencing these people (this person) in different ways, and those

different ways of perceiving the person are causing and justifying the different patterns of

use you associate with the two demonstratives. (For more on the notion of a ‘way of

experiencing’ here, see Campbell 2008).

9

In §3, I will set out Burge’s Proximality Principle, and the problems he thinks it

presents for a relational view of experience. Before doing so, though, I want to set out

two background differences between us. In his essay, Tyler Burge says:

I accept that the way in which one is given the object (a way that marks the

psychological state) plays a role in justifying, or warranting, the pattern of use

made of the demonstrative. This much seems trivially true.

(Burge 2005, p. 54)

Let me explain just where I part company with Burge on this. The question whether, in

general, the pattern of use of a term can be given any kind of justification, and if so, what

kind of justification, is usually taken to be quite difficult. Let me give some examples of

people who have found it difficult.

The problem Wittgenstein is grappling with in his discussion of rule-following

arises from the fact that he does not believe there is any justification at all to be given for

the pattern of use that we make of a term. In particular, there are no notions of truth or

reference in terms of which we can justify pattern of use; the ideas that the use of a

propositional constant is justified by appeal to the truth table, or that the use of a singular

term is justified by appeal to the reference of the term, are rejected. That means that he

has a considerable problem in explaining how there can be such a thing as going right or

wrong in one’s use of a term; and that is what he is grappling with in his discussion of

following a rule. Quine famously rejected the idea that the pattern of use of a term can be

given any semantic justification; the structure of the waves that ripple across the field of

10

force is not to be justified by appeal to the notions of truth and reference, but if justified

at all, to be justified by the demands of science. Dummett, followed by Brandom, argued

that there is a justification to be given for the pattern of use that we make of a term, but

that the justification is not to be given in terms of truth and reference.

My own view is that we can appeal to the relational view of experience, and in

particular, to the notion of being given an object in a particular way in experience, to

provide a semantic foundation for the use of a demonstrative. (I set out the issues over

the possibility of justifying a pattern of use in Campbell 2007.) So my own view is that

these philosophers – Wittgenstein and Quine and Dummett and Brandom– were

mistaken. But the questions they raised, about the possibility of justifying patterns of use

in semantic terms, do not seem to me trivial. This is where I part company from Burge.

The second divergence emerges, I think, when one notices that Burge shifts, in a

few lines, from talking about the ‘justification’ of a pattern of use, to talking about the

‘justification’ of a belief. Thus, he quotes me as saying that the object has to be there in

order for there to be a ‘justification’ of the pattern of use. To this he objects as follows:

An individual can be ‘justified’ – epistemically entitled to his belief – whether or

not an object is there.

(Burge 2005, p. 55)

Burge’s view thus appears to be that the justification of the pattern of use of a term is the

same thing as the justification of a particular belief involving the term. Here I differ from

him. Justification of the pattern of use of a term is one thing, and justification of a belief

11

involving the term is another. The two types of justification are indeed related, which is

perhaps why Burge identifies them; but they are not identical, and the relation between

them is not straightforward.

To illustrate, my point is that justification of the inference rules for a logical

constant is one thing, and justification of particular beliefs involving the logical constant

is another. Consider the position of someone who takes the Classical View of the

propositional constants, and says that the classical rules of inference are justified by

appeal to the classical truth tables. This is not the same thing as saying that the classical

truth table is somehow an element in the justification of any particular belief involving

the logical constant. Consider the case of classical disjunction. A glance at the

introduction rule shows that what it takes to justify a particular belief involving

disjunction is one or the other of the disjuncts. The truth table plays a quite different role

here; it is not an element in the immediate justification of the belief. Consider the case of

a perceptual demonstrative. What I am arguing is that an experiential relation to the

object justifies the pattern of use that one makes of the demonstrative. It is a different

claim to say that the experiential relation to the object is an element in the immediate

justification of a belief involving the demonstrative. In fact, on the view I am advocating,

the experiential relation to the object will not be an element in the immediate justification

of a belief involving the demonstrative. Rather, the immediate justification of a belief

involving the demonstrative will be provided by an information-processing content at the

level of the feature-map, such as: ‘Redness at place p’. Burge argues at some length that

the immediate warrant for a perceptual-demonstrative belief must be provided by a state

whose existence is independent of the existence of the external object referred to. In fact

12

that is an implication of the view he is trying to criticize, for the information-processing

state, ‘Redness at place p’, could perfectly well exist whether or not the external object

exists.

The point is rather that the internal information-processing state could not have its

status as part of the pattern of use of a demonstrative term unless the external object

exists. Suppose you try to argue that you could justify the pattern of use of a

demonstrative term without appealing to the existence of the external object referred to.

Here I think you generate the same kind of mystification that would be produced by

arguing that the rules for inference of a classical propositional constant could be justified,

even if there were no truth-table that could be produced for that constant which would

justify those rules of inference. If the term is a classical propositional constant, then to

assign it a semantic value is to assign it a truth table. Similarly, if a term is a visual

demonstrative, then to assign it a semantic value is to assign it an object as reference.

And it is only once the semantic value of the term has been fixed that there is any

particular pattern of use for it that can be justified by appeal to its semantic value.

On the relational view of experience, the object itself is a constituent of the visual

experience. In perceiving the object, you do indeed experience it in a particular way; and

the way in which that object is given plays a role in determining that one rather than

another pattern of use for the term is correct. That is the point of the example I gave

above, of the two different demonstratives referring to one and the same person.

It would be a mistake to suppose that we can somehow miss out the object

altogether, leaving behind only the ‘way’ of thinking of that object, to justify a particular

pattern of use. You might suppose, for example, that we could appeal to a conception of

13

perceptual experience on which the experience is exhausted by its representational

content. That content will include, you may say, demonstratives. And can’t the

demonstratives that are elements in the content of experience themselves be

anaphorically linked to the demonstratives one uses in propositional judgment? If so,

isn’t that all we require for a semantic justification of the pattern of use of the

demonstrative?

The demonstrative contents that allegedly figure in the alleged representational

content of experience will themselves have some pattern of use. That pattern of use

might be thought to have some interesting relation to the pattern of use made of the

demonstratives in propositional judgment. The relation might be identity, or it might be

something more indirect. But the pattern of use of these demonstrative elements cannot

itself constitute a semantic foundation for the pattern of use of the demonstrative in

propositional judgment. The pattern of use is one thing, and the reference of the term is

another. To justify a pattern of use by appeal to another pattern of use is not the same

thing as justifying a pattern of use by appeal to the reference of the term. To appeal to

the reference of the term, you have to be considering the object referred to. That requires

that there be such an object, and that you be appropriately related to it.

3. The Proximality Principle

Why do we need a relational view of experience? Only a relational view can

acknowledge the role of experience in making it possible for us to think about the objects

14

around us. It is, on the face of it, my current visual experience of the tree in front of me

that makes it possible for me to think about it. But if you think of experience as, for

example, a kind of sensation that the object produces in you, that makes it very difficult

to understand why consciousness – the mere occurrence of a sensation – can have any

role at all to play in your thought about the thing. Or if you think of experience as itself

already a representational state, then the experiential state presupposes, and cannot

explain, the ability to grasp those representations.

So far I have filled this out by remarking that knowledge of the reference of a

term causes and justifies the pattern of use of the term; that in the case of a perceptual

demonstrative, it is your experience of the object that causes and justifies the pattern of

use that you make of the term; and that only experience, relationally conceived, can cause

and justify the pattern of use that you make of a demonstrative.

Perceptual-demonstrative reference to objects depends on our perceiving them.

And this demands that we view perceptual experience as a relation between the perceiver

and the things perceived. This relational view of experience says that the content of

visual experience is constituted by the objects and properties in the scene perceived. This

relational view is an analysis of ordinary perception. It provides no analysis of the

separate phenomenon of hallucination.

Let us look, now, at Burge’s objection to this relational view of experience. This

relational view of experience says that the content of visual experience is not determined

merely by proximal stimulation of the visual system. In his essay Burge says that vision

science demands that we recognize only characterizations of perception that satisfy his

‘Proximality Principle’:

15

Holding constant the antecedent psychological set of the perceiver, a given type

of proximal stimulation (over the whole body), together with associated internal

afferent and efferent input into the perceptual system, will produce a given type of

perceptual state, assuming that there is no malfunctioning in the system and no

interference with the system.

(Burge 2005, p. 22)

Experience relationally conceived demands that the objects be there for you to see.

Holding constant the proximal stimulation is not enough for the objects to be there. So

the relational conception of experience violates Proximality, since Proximality says that

sameness of proximal stimulation (given the mentioned background conditions) is

sufficient to guarantee sameness of perceptual state. Therefore, Burge says, the relational

conception of experience has to be abandoned. For, Burge says, vision science depends

on Proximality.

The trouble with this argument is that the ordinary notion of seeing also violates

Proximality. So, by the lights of this argument, the ordinary distinction between seeing

and hallucinating should also be abandoned. You might have thought that seeing and

hallucinating were states of quite different perceptual types. But vision science has put

paid to that. There is no distinction between seeing and hallucinating. And Burge does

seem to draw this conclusion:

16

One of the most basic things we know from the science of vision is that the same

perceptual type can be a perception, a misperception, or a perceptual illusion that

fails to be a perception of anything in the environment.

(Burge 2005, p. 24)

Even if you are an opponent of the Relational View of Experience, you might hope to

find an objection to the view that does not wind up denying the distinction between

seeing and hallucinating. The simplest way to temper Burge’s argument so that it does

not wind up denying that distinction is something like this. You might try to hold on to

something like Proximality, while acknowledging that there is a significant distinction

between seeing and hallucinating. You might say:

(a) There is a ‘difference of perceptual type’ between seeing and hallucinating, and it

has to do with a difference at the level of conscious experience. In seeing the

things around you, the content of your consciousness is constituted by the

ordinary objects and their properties. In cases of hallucination, you are in a

conscious state. But even if the proximal stimulation is the same in the

hallucination as in a case of ordinary seeing, the content of the conscious state is

different. For you are not conscious of the ordinary objects and their properties.

(b) Nonetheless, there is a ‘sameness of perceptual type’ between seeing and

hallucinating, but this is a sameness of ‘Proximal’ type. There is a sameness of

proximal type in that the same states of the visual system in the brain are

17

involved. You might think of this sameness as a sameness of biological type, or

as a sameness at the level of information-processing content, using the notion of

information-processing content standard in vision science.

So we could hold on to the distinction between seeing and hallucinating by saying they

are states of different conscious types. We could still acknowledge that there is a sense in

which there can be similarities of perceptual type between seeing and hallucinating at the

levels of brain biology and information-processing. The trouble is that now Burge’s case

against the Relational View of Experience has collapsed. The relational view is an

attempt to characterize the conscious content of ordinary vision. Vision science accepts

that there is a difference between seeing and hallucinating. It is entirely consistent with

vision science to locate that difference, as the Relational View does, in the difference

between experiencing objects and properties and having experiences of a quite different,

hallucinatory type. We can acknowledge that there may be underlying proximal

similarities between seeing and hallucinating, at the level of brain biology and visual

information-processing. But vision science does not after all demand that we obliterate

the distinction between seeing and hallucinating.

Here is another way in which to think of the issue. Russell thought our

knowledge of truths has to be distinguished from our knowledge of things. In particular,

the kind of knowledge of things that he calls acquaintaince does not consist in knowledge

of truths at all:

18

Knowledge of things, when it is of the kind we call knowledge by acquaintance,

is essentially simpler than any knowledge of truths, and logically independent of

knowledge of truths, though it would be rash to assume that human beings ever, in

fact, have acquaintance with things without at the same time knowing some truth

about them.

(Russell 1912, p. 46.)

And on Russell’s view, our acquaintance with objects provides us with knowledge of the

references of demonstrative terms, such as ‘this’ and ‘that’:

it is scarcely conceivable that we can make a judgement or entertain a supposition

without knowing what it is that we are judging or supposing about. We must

attach some meaning to the words we use, if we are to speak significantly and not

utter mere noise; and the meaning we attach to our words must be something with

which we are acquainted.

(Russell 1912, p. 46.)

This is a prototype of the relational view of experience: according to Russell, there is a

kind of acquaintance with things that is not a matter of grasping some truth about them,

and the content of the acquaintance is constituted by the object with which one is

acquainted.

Suppose we liberalize Russell’s official view, to allow that acquaintance can be a

relation that holds between the perceiver and an ordinary distal object seen. Much of

19

vision science does not appeal to any such conception of acquaintance as a relation of

consciousness between oneself and the object. Vision science classically appeals to a

notion of ‘representation’ that is recognizably propositional. Can we say therefore that

Russell’s view has been refuted by vision science? Only if classical vision science makes

an authoritative claim to comprehensive description of all the phenomena of vision. It

should go without saying that classical vision science makes no such claim, and has no

experimental support for such a claim. In particular, the phenomena of visual awareness

have proven quite difficult for classical vision science to describe and explain. How to

characterize visual awareness, and its relation to the phenomena described by classical

vision science, are open questions. Any responsible use of something like Russell’s

notion of acquaintance with seen objects must, of course, try to characterize its relation to

phenomena described by classical vision science, such as Treisman’s information-

processing characterization of visual attention. But I see no difficulty of principle in

giving such an account. So an appeal to Proximality to refute the notion of non-

propositional acquaintance with objects carries little weight.

4. Consciousness and the Proximality Principle

Let me set this point in broader context. A number of philosophers have argued that

scientific characterizations of perception do not address problems of perceptual

consciousness. Thomas Nagel argued that it is possible to have a comprehensive

scientific understanding of bats’ use of sonar without have any conception at all of the

20

perceptual experiences enjoyed by bats using sonar. Frank Jackson argued that a scientist

without color experience could in principle have comprehensive knowledge of the

processes involved in color vision without knowing anything about what color experience

is. Joseph Levine argued that a full scientific account of perception would leave it

unintelligible why one sensation – one color sensation, for example - rather than another

is correlated with one neural state rather than another. These arguments have been much

discussed and extremely popular. There has been something like a consensus that

scientific accounts of perception do indeed leave certain questions about perceptual

consciousness unaddressed.

The relational view of experience is intended to apply to the notion of perceptual

experience. One way to see the appeal of the idea is to remark that problems about the

qualitative nature of experience – problems of the sorts raised by Jackson and Levine –

apply most forcibly when we take it that the world experienced does not itself contain

qualitative properties such as colors; when we take it that out in the world there are only

the objects and properties described by basic physics. The puzzle then is how to locate

the qualitative character of, for example, color experience. But this view of the world we

observe is not compulsory. We could acknowledge that the world we observe can be

correctly described at many different levels, not just the level of basic physics. The

relational view depends on the idea that qualitative properties are in fact characteristics of

the world we observe; our experiences have the qualitative characters that they do in

virtue of the fact that they are relations to those aspects of the world. So looking for the

qualitative character of experience in the nature of a brain state is looking for it in the

wrong place; we have to be looking rather at the colors of the objects experienced.

21

Vision scientists are studying the brain. In effect, the methodology of cognitive

science is to look for a box-and-arrow diagram of the structure of the visual system in the

brain. The boxes and arrows are given their overall orientation by knowledge of the

adaptive value of the visual system for the organism in its usual ecological niche. But the

box-and-arrow diagram has also to be recognizable as a diagram of a particular set of

biological structures in the brain. In the end, what this means is that that the box-and-

arrow breakdown of how every sub-task is accomplished by the system has to be

ultimately involve processes that are achieved by ‘primitive processors’ such as variation

in the rate of firing of a neuron (cf Block 1995, Dennett 1981). The whole methodology

here is geared to providing an illuminating characterization of a brain system. Of course,

there us no presumption that any particular assembly of cells is fulfilling just one

function; a cell assembly may play many roles. The mission is to characterize those

roles, for each cell assembly.

Classical computational approaches to vision propose that the functional

architecture of the biological system can be characterized in terms of computations,

realized by a biology with something like the syntax of a computational language. The

methodology here was given its classic explicit formulation by Marr when he

distinguished his three levels of analysis:

1. Computation

2. Algorithm

3. Implementation

22

At the level of computation we specify the task being performed by a visual system. At

the level of the algorithm we specify the procedures that the system uses to perform that

task. And at the level of implementation we say something about how in detail particular

‘primitive processors’ in the brain can be viewed as executing those procedures.

Burge’s defense of his Proximality Principle consists of rehearsal of basic ideas

from classical computational approaches to vision, such as the idea of the inverse

problem (that vision has to recover characteristics of the distal stimulus from the

characteristics of the proximal stimulus), and some elements of the explanation of

apparent motion. The examples he provides all fall comfortably within the ambit of a

Marr-style three-level approach. Does theorizing within this tradition conform to

Proximality?

Specification of the task being performed by the system, at the computational

level, is usually driven by hypotheses about the ecological niche of the organism:

knowledge of what, from an evolutionary point of view, this system is for. This gives an

anti-individualist aspect to the methodology, which Burge acknowledges. We are

looking at the organism in its ecological niche. Two different organisms, occupying

different ecological niches, might be molecule-for-molecule identical yet their visual

systems could be computing the answers to different problems.

Whether the full Proximality Principle is demanded by work in this tradition is a

further question. You might think that something like Proximality is demanded because

the aim is to characterize the workings of a brain system, and once we control for

ecological niche the workings of the internal brain system are all that the vision scientist

in this tradition is looking at. Still, to acknowledge that the vision scientist is describing

23

the workings of a brain system is one thing, and to endorse Proximality is another. There

is a certain determinism implicit in Proximality that the vision scientist need not endorse.

Let us look again at the Principle:

Holding constant the antecedent psychological set of the perceiver, a given type

of proximal stimulation (over the whole body), together with associated internal

afferent and efferent input into the perceptual system, will produce a given type of

perceptual state, assuming that there is no malfunctioning in the system and no

interference with the system.

(Burge 2005, p. 22)

There are a number of what we might call kludge terms here, jammed in with the hope of

making the thing work. These are ‘psychological set’, ‘malfunctioning’ and

‘interference’. The notion of ‘psychological set’ is usually used relative to a context in

which the subject is addressing some specific problem or other – ‘From where will the

target come?’, ‘Will the group accept me?’, and so on. And the experimenter usually

regards it as a responsibility to fill out exactly what they are including in ‘psychological

set’ in one context or another. So in the first case psychological set might include beliefs

about the likelihood of the target appearing from a particular direction, and in the second

case psychological set might include beliefs about one’s own gender and ethnicity and

about the group’s views on gender and ethnicity. Whether there is any well-defined

notion of ‘psychological set’ with the kind of generality Burge needs in stating

Proximality is moot. We need a notion of ‘psychological set’ that applies to any situation

24

whatever in which some kind of perceptual state might be produced. I don’t know that

we have any such notion, short of the ‘total psychology’ of the individual, whatever that

is. ‘Malfunctioning’ is another notion that is worth a second look. Burge needs a notion

of ‘functioning properly’ on which the visual system is functioning properly whether it

yields perceptions of one’s surroundings or mere hallucinations. Again, I don’t know

that we have a notion of proper functioning on which a visual system generating nothing

but hallucinations is ‘functioning properly’. And finally there is ‘interference’. I suppose

the idea is to acknowledge the possibility that something might come from outside the

brain system itself and affect its behavior. The thing is that even with all these kludges

jammed in, Proximality still might not work because the brain systems in question may

be inherently noisy and their workings may be probabilistic rather than deterministic. Put

in all the kludges, and put in the same proximal stimulation on two different occasions,

and you might get different results just because the brain biology is inherently noisy and

does not always work the same way, even though there are the statistical regularities that

the vision scientist characterizes and on which human survival depends.

Of course, you might reformulate Proximality probabilistically, and say

something like this: in an ordinary context, with nothing unusual going on, and all the

kludges jammed in, sameness of proximal stimulus in that context is fairly highly

correlated with sameness of conscious experience. The proponent of a relational view of

experience could agree with this. In the good case, sameness of proximal stimulus in that

context will be correlated with sameness of distal stimulus, after all.

These points are in a way irrelevant, though, because what is really driving

Burge’s argument is not Proximality itself but the more fundamental point that the vision

25

scientists are aiming to characterize the visual system in the brain, and what goes on in

the visual system in the brain is one thing and what goes on in the environment is

another. The point that is often made, though, is that this classical cognitive science

approach has no constitutive interest in problems of consciousness. A simple way to see

the point is to reflect that Marr’s approach could be and indeed has been fruitfully applied

to give computational accounts of the human immune system. You might argue that the

best way to view the immune system is as a computation system whose tasks include

learning about and remembering pathogens, and which develops a battery of responses to

them. The objective of a classical cognitive science approach to the immune system will

be to give specification of the computational tasks performed by particular subsystems,

explain what algorithms they use to perform them, and show how these algorithms are

biologically implemented. Yet the immune system has no particularly direct connection

with phenomena of consciousness. You might investigate the human visual system,

using this same methodology, without it ever occurring to you that the human visual

system has any particularly direct connection with phenomena of consciousness. That is

really the point that Nagel and Jackson and Levine were making: that the phenomena of

consciousness are not directly addressed by the scientific account of the visual system in

the brain.

The relational view of experience, as I said, is an attempt to characterize

phenomena of perceptual consciousness. The view is that the content of a perceptual

experience is constituted by the objects and properties in the scene perceived.

Consequently, there is nothing in common between the experiential content of a

perception when one ordinarily sees the scene before one, and the content of an

26

hallucination. It is, of course, consistent with this to acknowledge that there may be brain

states shared by ordinary perceiving subjects and subjects in the grip of hallucination.

Vision scientists might characterize those states in the very same information-processing

terms. But it does not follow from that there is no distinction to be drawn between those

states. There are important differences between ordinary perception and hallucination.

In a court of law, for example, life or death might turn on whether the witness really saw

something or merely had a hallucination. The relational view of experience is attempting

to characterize that difference between perception and hallucination in terms of the

contents of the subject’s conscious experience. You might indeed take a radical line here.

You might argue that the only distinctions we can legitimately draw between perceptual

states are distinctions that conform to Proximality, such as distinctions between visual

information-processing states in the brain. This would rule out the relational view of

experience. It would also imply that there is no difference between perception and

hallucination. Contemplating this strategy, I find myself with something of the emotions

one might feel on watching the Charge of the Light Brigade. C’est magnifique, mais ce

n’est pas la guerre.

5. Qualitative Aspects of Experience

One element in the appeal of Burge’s Proximality Principle is that people often suppose

that since the conscious aspects of experience are ‘immediately given to the mind’, there

cannot be any mechanisms linking them to the mind. The line of thought is this:

27

On the relational view the qualitative aspects of experience are constituted by the

objects and properties around us. But vision science tells us that there are brain

mechanisms involved in perception of those objects and properties. The end

product of the operation of those brain mechanisms is to make the qualitative

aspects of experience immediately present to the mind. If we assume, as does the

Relational View, that the qualitative aspects of experience are not artifacts

generated as the end points of brain processing, but are rather the distal stimuli to

which the brain mechanisms link us, then we cannot acknowledge the immediacy

with which the qualitative aspects of experience are given to the mind. That

immediacy cannot be reconciled with the Relationalist’s idea that we are linked to

the qualitative aspects of experience by means of brain mechanisms.

This we might call ‘proximalism’ about the mind. We think of the mind crouching in the

center of the brain, receiving the input from brain processing which constitutes the

qualitative content that is ‘proximally’ given to the mind. Everything else, including the

distal objects and properties, will be given to the mind only ‘mediately’, in virtue of

consciousness having this proximally determined qualitative content.

Dismantling this line of thought is an extensive exercise. We have to look quite

critically at the sense, if any, in which the qualitative aspects of experience are

‘immediately’ given to the mind. And we have to articulate it fully enough to make it

apparent that there is no conflict between anything that might be right in the idea, and the

28

point that brain mechanisms are what make available the qualitative aspects of

experience.

I will not pursue the dismantling here. I will remark on why the dismantling is

worth the effort. Letting go of the idea that the conscious life is in this sense a ‘proximal’

phenomenon is what makes it possible for us to recognize the role of consciousness in

our cognitive lives. Once we think of consciousness as fundamentally a relation between

the subject and the surroundings, we get how it is that thought about the surroundings

could be made possible by consciousness of the surroundings. Articulating the role that

consciousness, relationally conceived, plays in an understanding of demonstratives is just

one piece in the resolution of this larger puzzle.

29

REFERENCES

Block, Ned. 1995. ‘The Mind as the Software of the Brain’. In D. Osherson, L.

Gleitman, S. Kosslyn, E. Smith and S. Sternberg (eds.), An Invitation to Cognitive

Science. Cambridge, Mass.: MIT Press.

Brandom, Robert. 1994. Making It Explicit: Reasoning, Representing, and Discursive

Commitment. Cambridge: Harvard University Press.

Burge, Tyler. 2005. ‘Disjunctivism and Perceptual Psychology’. Philosophical Topics

33, 1-75.

Campbell, John. 2002. Reference and Consciousness. Oxford: Oxford University

Press.

Campbell, John. 2007. ‘If Truth is Dethroned, What Role is Left For It?’. In Randall E.

Auxier and Lewis Edwin Hahn (eds.), Library of Living Philosophers: The

Philosophy of Michael Dummett. Chicago and LaSalle, Illinois: Open Court. 281-

300.

30

Campbell, John. 2008. ‘Consciousness and Reference’. In Brian McLaughlin and

Ansgar Beckermann (eds.), Oxford Handbook of Philosophy of Mind. Oxford:

Oxford University Press.

Dennett, Daniel. 1981. ‘Artificial Intellligence as Philosophy and as Psychology’. In his

Brainstorms. Brighton: Harvester Press.

Dummett, Michael. 1991. The Logical Basis of Metaphysics. Cambridge. Mass.:

Harvard University Press.

Jackson, Frank. 2002. ‘Epiphenomenal Qualia’. In David Chalmers (ed.), Philosophy of

Mind: Classical and Contemporary Readings. New York: Oxford University Press.

Levine, Joseph. 2002. ‘The Explanatory Gap’. In David Chalmers (ed.), Philosophy of

Mind: Classical and Contemporary Readings. New York: Oxford University Press.

Marr, David. 1982. Vision. San Francisco: W. H. Freeman.

Nagel, Thomas. 2002. ‘What is it Like to be a Bat?’. In David Chalmers (ed.),

Philosophy of Mind: Classical and Contemporary Readings. New York: Oxford

University Press.

Quine, W.V.O. 1960. Word and Object. Cambridge, Mass.: MIT Press.

31

Treisman, Anne. 1993. ‘The Perception of Features and Objects’. In A. Baddeley and L.

Weiskrantz (eds.) Attention: Selection, Awareness and Control. Oxford: Oxford

University Press.

Wittgenstein, Ludwig. 1953. Philosophical Investigations. New York: MacMillan.