Back to CQ Homepage









Institute of Experimental Psychology,

Oxford, December 1955

Dear Professor Gibson,

At the close of our conversation after your class on

Wednesday evening, you indicated your dissent from my

statement that we can only describe our experience by

reference to the physical state of affairs which is normally

the case when we have an experience of that kind. May I be

permitted to anticipate what I suggest from your published

work would be your objections to my thesis and put my case

before you in writing, as I usually find I can express

myself more effectively in this way.

You agreed with me when I said that `red' is a property

of lights and objects, but I fancy that your interpretation

of my statement was different from my own. I meant that

`redness' is a real physical property of lights and objects

in the physical environment [JJG's gloss reads: "It is that

by virtue of the differential reflectivity of the surfaces,

i.e. red is specific to a surface"]. You, I fancy, would

want to say that it is a property not of physical objects

and phenomena, but of phenomenal objects in the visual world

[JJG's gloss reads: "No"]. I think the visual world, except

as a name for those aspects of the physical world which we

can discriminate visually [JJG's gloss reads: "It's simply a

name for perception"], is a myth. It is part of the

homunculus theory which you deplore [JJG's gloss reads: "It

is that, too"]. It is the television screen in the brain

which the homunculus looks at before deciding what to do or


A crucial case is the red after-image. You, I suspect,

would want to say that in this case we see a phenomenal

surface which is literally red. [JJG's gloss reads: "No. We

see a colour patch in the visual field, not a surface"] I

would disagree, there is only a physical surface which looks

red, i.e., we are having the sort of experience we have when

we look at a patch of red light projected onto a surface

[JJG's gloss reads: "How about closed eyes?"]

The view I suspect you of holding has some awkward

consequences. Firstly where is this surface with its red

patch. It is clearly not in the physical environment, but

equally clearly it is not in the brain. There is nothing in

the brain which could get red when the after image starts

and cease being red when it fades [JJG's gloss reads: "Why

not?"]. Brain processes just aren't the sort of things to

which colour concepts apply. If therefore there really is

something red, that something must exist outside the

physical world known to science. [JJG's gloss reads: "an

ancient paradox"] Perhaps you don't find this conclusion

embarrassing. Many people don't; but you say in your book

(p.1) that the view we get when we look at things "is in

every detail a nervous process". [JJG's gloss reads:

"Redness is an abstraction"]

Another difficulty is this; if "it looks as if there is

a red patch on the card though in fact there isn't" means

"the card in my visual world, unlike the corresponding card

in the physical world, has a red patch on it", does "This

house looks at least 200 years old, although in fact it

isn't," mean "The house in my visual world, unlike the

corresponding house in the physical world, is at least 200

years old"? If not, why not? If so, then what on earth

does it mean to say that I have got a 200 year old house in

my private visual world?

In ordinary English which is after all the language

spoken and understood by the naive subject in a

psychological experiment, to say "It looks as if there were

a red patch of light on the card", is, I should have

thought, the only obvious and natural way of describing what

happens when the subject has what in the technical language

of psychology is known as `a red after image', and this, I

would suggest, means simply that the subject is having the

sort of visual experience he usually has when confronted

with a red patch of light projected onto a surface like the

one in question. Similarly a house which looks 200 years

old is one that gives him the sort of visual experience that

200 year old houses usually give.

Of course to say that there is nothing red at all in

the after-image situation, appears extremely paradoxical at

first sight. One feels like objecting "but dammit! You can

see the redness right there in front of you." The trouble

here, I think, is that, having learnt to identify redness by

means of the special type of experience we get when we look

at red things in daylight illumination, we tend to think of

`red' as the name for the experience by which we identify

redness; whereas in fact redness is a property of physical

objects, not a property of experiences or of `phenomenal'


Colour, of course, is a specially difficult case. The

physicists tell us that redness does not correspond to any

simple characteristic of the physical stimulus, and that our

classification of things in terms of red, yellow, green,

blue, etc., is based on the peculiarities of our visual

receptor system and does not correspond to any distinctive

features of the physical stimulus. This, however, does not

alter the fact that the word `red' as used in ordinary

English is the name of a property of physical objects and

light sources. It is if you like the property of giving

rise in normal daylight illumination to a characteristic

type of visual experience; but the word `red' refers to the

property of objects whereby they excite that experience, not

to the experience itself. It is poppies, hot pokers and

pillar boxes that are red and they don't stop being red when

the light goes out, though they may cease to look red.

Incidentally the history of colour as the evil demon of

perception theory is interesting. It goes back at least to

Locke who classified it along with smells and tastes as a

`secondary quality', a quality aroused by the object in the

mind of the beholder as distinct from primary qualities like

shape and size which belonged to the object itself. As you

point out in your book, it was the only quality which could

be attributed to Wundt's visual sensations.

There is, however, another and more fundamental mistake

underlying the belief both in sensations and a phenomenal

world of objects. I have described it in a paper which is

coming out in the February number of the British Journal of

Psychology as

"the mistaken assumption that because our ability

to describe things in our environment depends on

our consciousness of them, our descriptions of

things are primarily descriptions of our conscious

experience and only secondarily, indirectly and

inferentially descriptions of the objects and

events in our environments. It is assumed that

because we recognize things in our environment by

their look, sound, smell, taste and feel, we begin

by describing their phenomenal properties, i.e.

the properties of the looks, sounds, smells,

tastes and feels which they produce in us, and

infer their real properties from their phenomenal

properties. In fact, the reverse is the case. We

begin by learning to recognize the real properties

of things in our environment. We learn to

recognize them, of course, by their look, sound,

smell, taste and feel; but this does not mean that

we have to learn to describe the look, sound,

smell, taste and feel of things before we can

describe the things themselves. Indeed, it is only

after we have learnt to describe the things in our

environment that we can learn to describe our

consciousness of them. We describe our conscious

experience not in terms of the mythological

`phenomenal properties' which are supposed to

inhere in the mythological `objects' in the

mythological `phenomenal field', but by reference

to the actual physical properties of the concrete

physical objects, events and processes which

normally, though not perhaps in the present

instance, give rise to the sort of conscious

experience which we are trying to describe."

(Place 1956, p.49)

I would submit that the view I am putting forward is

the logical conclusion to which your own view leads. You

reject the homunculus in the skull story and Helmholtz's

doctrine of unconscious inference and yet both these

doctrines follow logically from the doctrine of an interior

visual world which you appear to accept. You say in your

book (p.8) that "the stimulus variable within the retinal

image to which a property of visual space corresponds need

only be a correlate of that property, not a copy of it."

You also indicated in class that you reject the Gestalt

theory of isomorphism in so far as it implies that in order

for perception to be veridical an accurate reproduction of

the stimulus environment must be constructed in the cortex

from the input arriving at the cortex from the retina. Why

not take the final step and recognise that the various

elements of visual experience are correlates, but not

copies, of the corresponding properties of the stimulus

environment? Unless you take this further step, so it seems

to me, stages F and G in the diagram on p.18 of your book

will never lose their nebulous character, since you will

never find anything in the physical world which has the

properties you ascribe to them.

Yours sincerely

Ullin T. Place


Institute of Experimental Psychology,

Oxford, 10th January 1956

Dear Professor Gibson,

As I understand it from our conversation last month

your view of the `visual world' may be formulated as


The concept of the `visual world' and the descriptions

which are given of it are a shorthand device for describing

the responses, verbal or otherwise, which an organism [JJG's

gloss reads: "a `naive' organism"] makes to a stimulation

[JJG's gloss reads: "to `ordinal' stimulation (i.e.

focussable light) not to retinal images"] impinging on its

visual receptor system. On this descriptive system these

responses are characterised in terms of the state of affairs

in the physical environment which would have to be the case

[JJG's gloss reads: "and which ordinarly are the case"] for

the response in question to be `appropriate' or `adapted' to

the actual situation confronting the organism.

This is undoubtedly a logically defensible use of the

term, but I very much doubt whether, if applied

consistently, it would allow you to say all the things you

say about the `visual world' in your book. In particular I

find it difficult to see what can be meant by the term

`visual field' if you adopt this definition of `visual

world' [JJG's gloss reads: "I mean visual sense-data"].

The concept of the `visual field', I take it, is

introduced in order to explain what it is that we are

describing when we adopt a sophisticated attitude and

describe what we see in fixed monocular regard as if framed

by the outlines of the nose and eyebrows and the rather

indeterminate limits of visual discrimination in other

directions. [JJG's gloss reads: "Yes, very roughly."]

Such descriptions, I should have thought, are responses

appropriate to the environment which would be the case if

the light waves impinging on the eye were reflected from a

surface at a uniform though indeterminate distance from the

eye. [JJG's gloss reads: "This is not an environment, in any

sensible meaning of that term."] If so they are, on the

above definition, descriptions of the individual's visual

world at the time when this peculiar attitude is adopted.

[JJG's gloss reads: "?".] In other words the visual field

is a special type of visual field, not a separate category

which can be contrasted with the visual world. On this

interpretation you would not be able to speak as you do as

if the visual world were somehow a compound of a number of

visual fields. [JJG's gloss reads: "Speaking thus was a

mistake I now regret. The visual field is simply an

artifact of human picture-perception."]

I would submit that your use of the term `visual world'

in your book is not consistent with the above definition,

that you can only say what you want to and do say about it

by treating it as if it were a kind of internal television

screen [JJG's gloss reads: "this must be avoided, we

agree."], and that your introduction of the terms `visual

world' and `visual field' in your theoretical analysis of

visual perception is both unnecessary and misleading.

Rather than trying to defend the view that your use

of these terms is not consistent with your definition by

quoting chapter and verse, let me try to show how the main

argument of your book could be introduced without using

these terms at all. What follows is an introduction to the

psychology of visual perception from a purely behaviorist

point of view, which to my mind would provide an ideal

starting point from which to develop the main argument of

your book:


The psychology of perception may be defined as the study of

those features of its environment which an organism can

discriminate by virtue of the stimulation impinging on its

receptor organs [JJG's gloss reads: "Good. But one must

define stimulation."], the stimulus conditions under which

such discriminations are possible, and the stimulus

conditions under which discrimination breaks down.


To discriminate a featre of the environment is to make a

response which is specific to the feature in question in the

sense that it is made when and only when the distinctive

pattern of stimulation characteristic of the feature in

question impinges on the appropriate receptor organ. [JJG's

gloss reads: "Good".]]

In studying the perceptual capacities of human

subjects, the ability to apply a descriptive epithet

appropriate [JJG's gloss reads: "specific"] to the

environmental feature in question is usually taken as

evidence of discrimination. Animals do not possess this

complex repertoire of verbal habits which enable the human

adult to make an immediate discrimination of some kind when

presented with almost any stimulus. It is therefore

necessary in most cases to train the animal to perform some

simple movement or some other type of response in such a way

that the response is elicited when and only when the

environmental feature in question is presented to it. If

the animal can learn to restrict its responses in this way,

we conclude that it can discriminate the feature in

question. [JJG's gloss reads: "Correct".]

It should be noted that, in the case of animals, a

failure to learn does not necessarily imply a failure to

discriminate, since we know that learning can often be

impeded by other factors. Consequently it is often

extremely difficult to demonstrate conclusively that an

animal cannot discriminate a particular feature of its

environment. [JJG's gloss reads: "Correct".]

In the case of human subjects failure to discriminate

is shown either by a failure to respond or, more frequently,

by the use of the `wrong' descriptive epithet, i.e. one that

is inappropriate [JJG's gloss reads: "unspecific"] to the

actual situation presented to him. A misdescription of this

kind indicates a failure to discriminate between the

stimulus situation presented to the subject and the type of

stimulus situation where the epithet used would be

appropriate. (N.B. It would be in talking about these

`misdescriptions', as I have called them, that you would use

the concept of the `visual world' as defined above. A case

where the subject says, "It looks round, two feet away,

etc., although I know it can't be", would count as a

misdescription in this sense.) [JJG's gloss reads: "This

sounds like nonsense to me."]

Misdescriptions and failures to respond can, of course,

merely be an indication of an inadequate vocabulary,

mistaken use of words or a failure to understand the

instructions; but with proper instructions and subjects of

reasonable intelligence and education, such factors can

usually be neglected.


The features [JJG's gloss reads: "You mean substances not

features don't you? Substances have many discriminable

features"] of the environment which can be discriminated by

eye are basically of three types:

1. light sources,

2. gases liquids and solids which transmit some but not

all the light impinging on them and therefore

`transparent' [JJG's gloss reads: "You can't be glib

about this problem. It's difficult"],

3. gases, liquids and solids which absorb and/or reflect

the light impinging on them.

Any light source, any concentration of gaseous, liquid

or solid substance which transmits, reflects or absorbs

light may be said to constitute a `visual object' [JJG's

gloss reads: "No. A gas which transmits light without

refraction or reflections is invisible."]. The principal

features of a visual object which a human subject can detect

by eye are the brightness of a light source, the degree of

transparency-opacity of a transparent object, the albedo

(reflecting capacity) of an opaque object, the level of

illumination in different parts of the environment, crude

differences in the wave length of light generated by a light

source (the colour of light sources), crude differences in

the differential reflecting and transmitting capacity of

opaque and transparent objects (the colour of objects as

distinct from the colour of lights), the size and shape of

objects, the texture of surfaces, the distance and position

of objects relative to one another and to the observer, the

movement of objects and their velocity and direction of a

movement. [JJG's gloss reads: "Your list constitutes a set

of problems, not definitions."]

Before we can understand the conditions under which a

particular environment feature is discriminated, we need to

understand the stimulus conditions under which we are able

to discriminate differences in these general features of

visual objects, such as brightness, colour, size, shape,

texture, position and movement, since it is differences of

this kind which provide the basis for our discrimination of

all those myriads of features of our environment which we

can discriminate by eye. The psychology of visual

perception is, therefore, primarily concerned with the

conditions under which these basic discriminations are made.

(N.B. This section may be described as a brief survey of the

`visual world' in the sense of those features of the

physical environment which can be discriminated by eye.

Your description of the `visual world' in your book is, I

suggest, primarily a description of the visual world in this

sense, rather than in the sense defined earlier on.)


Differences in the albedo, colour, size, shape, texture,

position and movement of transparent and opaque objects can

only be detected by eye when the objects in question are

adequately illuminated, that is when light of a certain

intensity is reflected or transmitted by them onto the eye.

Nevertheless, although there are levels of illumination

below which such features are no longer discriminated,

within the range of illumination conditions where

discrimination of detail is possible changes in illumination

do not materially affect the discriminations which a subject

makes in spite of considerable changes both in the pattern

and in the absolute intensity of the light reaching the eye

from the environmental feature in question [JJG's gloss

reads: "and why do you think this is?"]. Similarly

discrimination of such features as the size, shape, relative

distance and velocity of objects is maintained in spite of

vast changes in the pattern of stimulation which is

projected from the object onto the retina of the eye which

result from changes in its distance from and orientation

with respect to the observer [JJG's gloss reads: "True. But

how can this be so?"]. Such changes in the conditions of

illumination , distance and orientation with respect to the

observer are, of course, discriminated as such; but the

marked changes in the pattern of stimulation which they

imply do not normally affect the organism's capacity to

discriminate those features of the environment which are

independent of the observer and the conditions of

illumination [JJG's gloss reads: "a paradox"].

Under normal circumstances an organism will only

respond to those aspects [JJG's gloss reads: "?"] of the

stimulus impinging on the eye which correspond to identities

[JJG's gloss reads: "?"] and differences [JJG's gloss reads:

"?"] in the physical constitution of its environment, and

will ignore those which merely reflect the particular

conditions under which observation takes place [JJG's gloss

reads: "?"]. Thus a penny will be discriminated as a round

object in spite of the considerable changes in the pattern

of stimulation projected from it onto the retina which occur

with alterations in its orientation with respect to the eye,

and its similarity, when viewed at a slant, to an elliptical

object viewed full face by virtue of a similarity in the

cross-section of the light rays passing to the eye will be

ignored. [JJG's gloss reads: "???"]

How far this tendency is a product of the innate

constitution of the organism and how far it is a product of

learning is open to dispute [JJG's gloss reads: "In short,

this theory has no proper explanation of constancy"]; but

its biological utility is obvious.


Although the discriminations which it normally makes are of

features of the environment which are independent of the

special conditions under which they are observed, it is

possible for the human organism, at least, to learn to

discriminate differences in the pattern of stimulation

impinging on the eye corresponding to differences in the

conditions of observation [JJG's gloss reads: "`conditions

of observation' - vague"] rather than to differences in the

actual constitution of the environment. [JJG's gloss reads:

"The question is, what proximal stimuli do correspond to the

actual constitution of the environment?"]

The ability to make discriminations of this kind only

becomes important in the normal course of events when the

problem arises of making an accurate representation of a

tri-dimensional scene on a two-dimensional surface. In any

course in drawing or painting one is taught to ignore the

discriminations of the shape size and colour of objects that

we `habitually' make. One learns to pay attention to the

shape as projected instead of `thinking in terms of' the

tri-dimensional shape or its conventional representation,

the elevation. One learns to `see' distance in terms of

perspective, the `smallness' of large objects far away, the

`whiteness' of the highlights on a shiny black object.

[JJG's gloss reads: "i.e. to see the visual field"]

In the past the ability to make discriminations of this

specialised kind in which objects are differentiated

according to the pattern of light and shade and the size and

shape of the images projected from them onto the retina has

been studied by psychologists far more extensively than is

warranted by the actual importance of such discriminations

in the individual's adaptation to his environment. There

have been two main reasons for this interest, one valid

reason and one whose validity is extremely doubtful.

The valid reason for studying these `artist-type'

discriminations, as we may call them, is that when made by a

skilled observer they seem to correspond very closely with

the pattern of stimulation impinging on the retina [JJG's

gloss reads: "seem to, but do not. This was the mistake of

my book."]. Direct observations of the retinal image by

existing methods tell us comparatively little, and although

the pattern of stimulation can be inferred from our

knowledge of the optics of the light stimulus, this is an

extremely cumbersome method of determining the pattern of

stimulation at any given moment. Consequently, although

they are not strictly speaking observations of the retinal

image itself, these artist-type discriminations do provide a

very convenient method of assessing the main characteristics

of the retinal image at any given moment [JJG's gloss reads:

"No."]. Since it is the pattern of stimulation on the retina

or, more precisely, the combination of the patterns of

stimulation on the two retinas to which the organism is

responding when it makes its visual discriminations, this

information can be extremely useful in helping us to

understand how the more normal and biologically more useful

discriminations are made.

(N.B. The concept of the `artist-type visual

discrimination' corresponds to your concept of the `visual

field'. In this section I have tried to bring out the

contrast between the `visual field' in this sense and the

`visual world' in the sense of the features of the

environment we normally discriminate, and the sense in which

knowledge about the `visual field' helps us to understand

the `visual world'.)



The second and less valid reason for studying what I have

called the `artist-type' of visual discrimination was the

assumption that these discriminations are somehow more basic

and fundamental than the discriminations of shape, size,

colour and distance that we ordinarily make. It was thought

that we begin by noting the shapes, sizes and colours which

the artist notices and then learn to infer from these

`apparent' shapes, sizes and colours the `real' shapes,

sizes and colours of the objects themselves. As we do not

normally make any conscious inferences of this kind, it was

assumed that some sort of `unconscious inference' must

occur. [JJG's gloss reads: "Good"]

In view of the difficulty which most children experience in

learning to make these discriminations when required to do

so by their preceptors in the visual arts, it is perhaps

difficult to understnad why so improbable a hypothesis

should ever have been seriously entertained. It must be

remembered, however, that when psychologists first started

to think about these problems a great deal of work had

already been done by physicists, on the nature of the light

stimulus, and by anatomists and physiologists, on the

mechanism of the eye, and that this information provided the

jumping off point for psychological research in this field.

The physicists had shown that light is a physical

energy which varies only in intensity and wave length; while

the anatomists and physiologists had shown that the eye is

like a camera with a photosensitive surface (the retina)

onto which a two-dimensional pattern of light is reflected

in the environment.

It seemed to follow from this that the basic

discriminations made by the visual apparatus and hence by

the organism must be of differences in the intensity and

wave length of the light projected from different points in

the visual field (not in your sense), and the distribution

of these differences over the surface of the retina. In

fact the basic discriminations, in the sense of the ones the

organism normally and naturally makes, are between features

of the environment corresponding to certain rather complex

relationships within the pattern of stimulation on the

retina [JJG's gloss reads: "of the focussable light entering

the eyes"].

There is, of course, a sense in which the

discrimination of intensity wave length and their bi-

dimensional distribution is more basic [JJG's gloss reads:

"I now doubt this"] than the discriminations we ordinarly

make of the shape, size and colour of objects in that

differences in intensity, wave length and bi-dimensional

distribution must be `detected' by the visual apparatus

before the organism can respond, as it normally does, to the

more complex relationships between them which correspond to

similarities and differences in the stimulus environment;

but the word `before' here should be taken to refer only to

priority in the sequence of events between the impingement

of light energy on the retina and the resulting response.

It should not be taken to imply priority in an ontogenetic

or phylogenetic sense. Nor should differentiation or

`detection' by the eye be confused with discrimination by

the organism.

(N.B. In this section I have tried to state what I take to

be the essence of your theoretical position `vis a vis'

earlier work in the field. The historical story is, of

course, grossly oversimplified. At this point the stage is

set for a discussion of the stimulus conditions which

provide the basis for our normal discriminations of distance

shape and size in a three dimensional environment along the

lines of your book.) [JJG's gloss reads: "You seem to agree

with everything except the "field vs world" distinction.

That admittedly needs clarification. But it is necessary, I

am sure, for the explanation of constancy."]

I would very much like to hear your comments on this

effort, if you can manage to wade through it.

Yours sincerely

U.T. Place


Institute of Experimental Psychology,

Oxford, 27/1/56

Dear Prof. Gibson,

Thank you for your pencilled comments on my last

screed. It would appear from these, that apart from the

question of how `experience' comes into the picture, a

problem which I deliberately left out in my letter, there is

no substantial disagreement between us.

The only remarks which you make which seem to call for

comment concern the problem of constancy. You seem to think

that there is something that I have failed to explain and

that your `visual world' and `visual field' concepts are

required to do this.

In this connection I should like to begin by saying

that I do not regard my previous letter as presenting a

`theory' in any proper sense of the world. A theory is

either a set of presumptive scientific laws designed to

coordinate an existing body of empirical facts and to

predict new ones or a hypothesis as to the particular

application of known or presumed laws and principles in a

given case. I have not tried to formulate any laws or

principles, nor have I put forward any hypotheses. I have

not tried to give an explanation, in the scientific sense,

of anything. I have merely tried to present the problems in

the field of visual perception as clearly and simply as

possible, and in such a way as to avoid generating pseudo-

problems, like the traditional problem of constancy, which

arise when the real problems are badly stated.

The problem of constancy as traditionally discussed is

a pseudo-problem because it is based on the mistaken

assumption that the only sort of discriminations which can

be made on the basis of the stimulation impinging on the

retina are discriminations of the `artist', `picture' or

`visual field' type.

As I see it there are only two genuine empirical

problems tangled up with this ghastly conceptual muddle.

The first is the problem with which you deal so admirably in

your book of defining the `proximal stimuli' to which the

organism is responding when it discriminates the albedo,

size, shape, distance and velocity of objects under varying

conditions of stimulation. The only other psychological

question, as distinct from the physiological questions about

the underlying neurological processes, is why the organism

discriminates these rather complex aspects of the stimulus

more readily than it discriminates, in so far as it ever

does, the relative size, shape and intensity of different

parts of the stimulus itself.

Unfortunately at this point in my last letter I made

some rather misleading remarks as to the type of answer to

be given to this latter question. I pointed out the obvious

fact that `constancy' discriminations are biologically

useful, whereas the `retinal image' discriminations of the

artist are not; but I also raised the further problem of

whether our tendency to make the former type of

discrimination more readily than the latter is due to an

innate predisposition or to learning.

I must emphasise that in talking about learning here I

was not talking about `learning to perceive' in the sense

used by some writers who suppose that we start with a

`visual field' type of discrimination and have to learn

somehow to modify our visual experience in the direction of

constancy or have to learn to make the correct unconscious

inference. This is pure mythology.

Nor was I intending to overlook the obvious fact that

all the human discriminations we investigate are learned in

the sense that we have to learn to use words correctly. In

many cases, however, this verbal learning is clearly

superimposed on earlier discriminations involving other

types of response, such as the basic locomotor and

manipulatory adjustments. Most of these discriminations, no

doubt, are also learned; but there are some responses such

as reaching out for a near object which seem to imply some

degree of distance and size discrimination which may well be

innate. There is certainly nothing intrinsically improbable

in such an assumption, since there are many innate reactions

of animals which involve constancy discriminations. Many

organisms for example have innate escape reactions triggered

off by moving objects larger than themselves, but not so far

as one can gather by objects projecting a retinal image of

any particular size.

It may be of course that in the case of the human

organism the question of which stimuli shall trigger off

which responses is entirely decided by learning. Even so it

is still possible that the innate constitution of the

organism might be such that `gradients of stimulus

generalisation', to use Hull's phrase, develop more readily

along dimensions of the stimulus, such as those which you

have described, than they do along simple dimensions of more

or less uniform patches of retinal stimulation.

In other words, all I am asking when I ask how far

constancy is learned and how far it is an innate tendency,

is how far the difficulty most of us experience in learning

to discriminate objects according to their retinal

projections can be accounted for by the incubus of past

learning which has been exclusively of the constancy type

and which conflicts with the super-imposed picture

perception habits. That is certainly part of the

explanation; but the possibility of an innate predisposition

towards constancy-type discriminations cannot be ruled out,

and I see no reason why there should be any need to prejudge

this issue which can be safely left to future research

without seriously affecting any important theoretical issue.

If you think there are any other empirical issues in

the constancy problem, I should be very interested to know

what you think they are and how the `visual world' and

`visual field' concepts would help.

Yours sincerely

U.T. Place



James J. Gibson

Cornell University

Theories of perceiving have generally taken it for

granted that the process is not directly dependent on the

stimulation of receptors but only indirectly dependent.

[UTP's gloss reads: "Perceiving is an achievement not a

process."] Perception is supposed to be based on sensation.

The latter depends on stimulation, to be sure, but the

former involves some additional process the nature of which

must be discovered. The theory to be outlined, however,

assumes that perception [UTP's gloss reads: "i.e.

discrimination?"] can be explained as a direct function of

stimulation when the stimulus is analysed with sufficient

care. The proposal is to dispense entirely with the concept

of sensation, using the term only to refer to a rare and

sophisticated kind of experience instead of an elementary

one, and relegating it to the status of a psychological


This approach is not, then, a theory of perception in

the ordinary meaning of that world. Existing theories of

perception are concerned with how sensations or sense-data

get converted into percepts. The formulas have been

various: interpretation of the data by the mind,

supplementation of the data by memory, or organization of

the data by the brain. In the present theory all such

formulas are superfluous since no process of conversion is

assumed. Not only the qualities of things but also their

very thingness itself is taken to be discoverable in

stiumulation, [UTP's gloss reads: "a confusing way of

putting it."] and the assumption extends to places and

events in space and time. They are all supposed to be, in

the rejected terminology, "sensed". This basic assumption

can be tested by psychophysical experiments. [UTP's gloss

reads: "?"] The theory incorporates, then, certain aspects

of both perceiving and sensing in the older meanings of the

terms; it is concerned with the facts of perception on the

one hand the methodology of sensation on the other but

rejects the theoretical distinction between them.

Classical psychophysics has accumulated the most solid

body of scientific facts in the whole of psychology by

studying the dependence of so-called sensory experience on

stimulus variables. The commonest psychophysical method is

to show that differences in experience (as evidenced by

judgements) correspond to differences in stimulation. In

order to do so, the experimenter must have previously built

a device to isolate and control a variable of stimulus-

energy in which he is interested. The essence of the

experiment is that a judgement of "greater" is always given

when the experimenter sufficiently varies the energy in one

way and a judgement of "less" when he varies it the other

way. The output is specific to the input, within limits

which are investigated. The success of the psychophysical

procedure has been so great that the specialists in it like

to claim that they are scaling or even measuring experience

(20). The correspondences obtained, whether or not

mathematical functions can be written to express them, have

yielded a mass of evidence for physiological theory on the

ways in which the neural process is specific to the

stimulus, and consequently sensory physiology nowadays goes

hand in hand with traditional psychophysics.

The contemporary workers in this tradition tend to

cultivate their own garden of so-called sensory problems and

shy away from the perceptual problems. If they are

physiologists they are apt to say the perception is "mental"

and not their business. If they are psychologists they are

apt to feel distrust for difficult speculations about

gestalten. But surely this shyness is a mistake. The

possessors of a good method in science ought to try it on

any problem where it can be used. The psychophysical

experiment can be extended, for example, to the

investigation of visual space-perception if devices can be

built for the control and systematic variation for the

stimulus conditions which arouse it (13). Whether these are

to be called cues for depth or stimuli for depth is not so

important as is the trying out of the experiment. Presumably

no one doubts that a psychophysics of perception would be

desirable if it could be achieved.

Over and above the argument for extended psychophysical

experiments, however, there is the possiblity of a general

psychophysical theory of experience. The central idea of

such a theory would be this: for every aspect or property of

the phenomenal world, [UTP's gloss reads: "The phenomenal

world is a myth and so are its properties. If you say `For

every aspect or feature of the environment which we can

discriminate..' O.K."] however subtle, there exists some

variable of the energy flux at receptors, however complex,

with which the aspect or property is in psychophysical

correspondence. This says that there are not only stimuli

for colours and sounds and tastes, but also for objects and

places and events, and for meanings and values and abstract

things. [UTP's gloss reads: "???"] It implies that when no

stimulus for a quality or kind of experience is known, one

should be sought. It is a bold hypothesis, not to say a

grandiose one. Nevertheless it is worth examining for its

presuppostions and implications.

In the first place, it assumes that the phenomenal

world can be resolved into phenomenal dimensions, aspects,

or properties. It coes not admit that exepreince is

composed of irreducible units, forms, or figure-ground

impressions. [UTP's gloss reads: "We can say nothing about

what experience is composed of except perhaps that it is

composed of patterns of neural activity."] It asserts that

awareness is fundamentally discriminative and that the

difference between any one experience and any other is a

transitional difference, not an absolute one. These

dimensions are not taken to be those of the "sensory solids"

of analytic introspection, however, for these depended on a

doctrine. They are the dimensions of everyday experience

and since this is enormously variable, they are vastly

greater in number. [UTP's gloss reads: "I do not see that we

can speak of `dimensions of experience'. All you need here

surely is the notion of parameters of response'".]

In the second place, it assumes that there exist in the

flux of energy surrounding us variables of stimulation of

high order as well as those of low order. The latter are

mostly the ones which psychophysical experimenters have

learned to isolate and control but this is no reason for

doubting that the former exist, nor that they might be

brought under experimental control with the exercise of

ingenuity. Light, heat, sound, and force vary not only in

such ways as frequency intensity, direction, and the like

but also with respect to serial order and succession. The

energy flow at the surface of an organism can be measured as

the amount at a fixed point, and sensory physiologists have

done so, but the differences, rations, rates, gradients, and

derivatives of this flow are the variables in which the

psychologist is interested. They are the stimuli for the

activity of the whole organism as distinguisehd from that of

a cell. These higher order variables are the ones that

correspond to the physical properties of objects and events

in the environment, or at least to the properties that are

important for an animal's behavior. They explain why it is

that stimuli at the sense organs can specify objects beyond

the sense organs.

Many or most of these high order variables in the

surrounding sea of energy are not responded to by a given

individual or by the individual at a given moment. An

animal may not react because its species does not possess

the necessary receptor mechanism, or because it has not

learned to discriminate the variable in question, or because

it is responding instead to some other variation of evergy.

The way to find out whether an animal is equipped to react

specifically to an energy charge, or can be trained to react

specifically, or can be "set" to react specifically is to

perform a discrimination experiment. In the case of man,

this can be a psychophysical experiment. [UTP's gloss

reads: "I should have thought that the precise relationship

between human and animal psychophysical experiments requires

more detailed discussion."]

The psychophysical theory of perceptual experience does

not assume, it should be noted, an innate repertory of

sensory responses to a fixed list of energy variables. It

asserts that the energy variables around us are far richer

and more complex than we have imagined and the the ability

to discriminate these variables increases first with the

maturation of the nervous system and later with the progress

of learning. They are capable of becoming stimuli insofar

as an individual develops the ability to make differential

reactions to them, that is, to perceive them. The

phenomenal world [UTP's gloss reads: "I don't like this ?"]

of one individual, consequently, may be quite different from

that of another individual, and it may be different for the

same individual on different occasions. But the physical

environment is the same for all, and the differences in

perceptual reactions are differences in what the individual

is responding TO. They are not differences in the process

of constructing or organizing a world out of meaningless

sensations. The process of perception is one of

discriminating and selecting and abstracting a world, not

one of building it.

The psychophysical theory of perception, as was

suggested at the outset, proposes to dispense with the

concept of sensation or, more exactly, to dispense with the

distinction between sensation and perception. This proposal

involves difficulties of terminology for which the reader

should make allowances. Terms like "sensing" and "the

senses" cannot be abandoned without substitutes, which are

not easy to invent. What should be dispensed with, in the

last analysis, is the persistent notion that sensory

impressions [UTP's gloss reads: "All impressions are surely

impressions of something. What are sensory impressions

impressions of?"] are prerequisite to, or the basis of,

other impressions. Sensations are the symptoms not the

cause of perception. They are the products of introspection

and, not only this, the products of a doctrinal kind of

introspection that expects to find them at the core of every

perception. Introspection is a very useful guide to the

study of why things look, sound, and feel as they do, and is

the usual preliminary to a good psychophysical experiment,

but a cultivated naivete about what the world does look,

sound, and feel like is a better guide than the fixed

expectations of a premature structuralism. The danger of

structuralism is that a list of modes, attributes, and

qualities gets written down in accordance with supposedly

known variables of stimulation. This results in a tendency

thereafter to find just these elements in experience and not

to look for new variables of stimulation in the environment.

The theory of sensation and the practice of analytic

introspection were necessary perhaps for the development of

our knowledge of sense-physiology to its present stage,

which is impressive. But they are nothing but a hindrance

to the development of what might be called perceptual

physiology, that is, the physiology of the afferent side of

the total behavior process. An adequate theory of behavior

depends on the latter kind of knowledge. They physiology of

the receptor-cell is beginning to be understood; what is

needed is information about the physiology of the receptor-

mosaic and the higher-order processes within it. We need a

molar instead of a molecular physiology to go with a molar




The psychophysical theory of perception has been applied, as

yet, only to vision (3). Its claims extend to the whole of

phenomenal experience and it tends toward a break down of

the classical sense-modalities, but our concern here will be

primarily with experiments about visual phenomena. To what

extent can it be shown that visual perception is a function

of optical stimulation?

The question actually is whether the visual world

(perception) is a function of stimulation instead of merely

the visual field (sensation). The visual world is an

experience [UTP's gloss reads: "No."] characterized by a

ground and other surfaces, receding, frontal, or slanting,

not by patches of colour as such. It includes edges, gaps

and depth-shapes, or objects, which possess constancy of

colour, size, and shape. It includes the point-of-view

itself as one pole of the experience, and impressions of a

changing point of view when the observer moves. It includes

such phenomena as the motions and rotations of visibly rigid

objects and the elastic or fluid deformations of non-rigid

objects like faces or the surface of the ocean. It includes

the perception of events, living things, and human artifacts

like pictures and symbols. Its properties and qualities

largely remain to be described and classified, and the above

list should only arouse the visual investigator to look for



The stimulus condition for determinate vision is light

reflected (or occasionally radiated) from substances through

clear air, that is, the focussable light from the objects of

the environment to any point in the environment where an

animal may place its eyes. The projective capacity of the

light to the point in question is analysed by talking about

rays, but the ray of light is not the elementary stimulus

for vision as sometimes has been imagined. This array of

rays converges to the point from all directions, although

animals like ourselves can pick up less than half of the

total sphere at any one moment. We explore the total array

by turning our eyes and head. We also sample the family of

all possible arrays by moving from one to another point of

view. As a matter of fact we are normally never limited to

the stimulation provided by the single array to a motionless

point; we have two eyes and we pick up two arrays at the

same time, converging to different points a short distance

apart. These facts might be expressed by saying that the

incoming optical ray-sheaf is dual and fluid, rather than

single and static (13, pp.1-2).

The higher order variables of optical stimulation on

which perception is said to depend should all be

discoverable in the patterning of the single ray-sheaf, the

duality of the patterning at the same moment, and the

transformation of its patterning over time. The duality

factor can be referred to as disparity and the

transformation as motion parallax. The momentary pattern of

the single ray-sheaf is projected as the momentary retinal

image and this consists of transitions of luminous energy

between areas of relative light and dark. We may speak of

this condition, for convenience, as optical texture.

Textured images are necessary for any impression more

definite than filmy or fog-like colour in a void. If one

wears diffusing plastic hemispheres over the eyes (half

ping-pong balls) the experience can be described as

"nothingness" in the sense that "thing quality" has

disappeared from the perception (9,14).

An important variable of optical texture is its

density, or the frequency of transitions. This may be

either the overall density or the density at a certain

region along one or another meridian or the array. There is

also the ratio of the horizontal to the vertical density in

a certain region, and perhaps other variables which may be

found significant. Finally, there is the gradient of

density over the array, for example an overall increase in

density upward. Such an increase is produced by a terrain

or ground surface in the optical array given to a

terrestrial animal.

It should be noted that under natural conditions the

gradient of texture density in each single projection is

accompanied by two coincident gradients, one of the degree

of successive incongruence of the elements in each

projection (the gradient of parallactic motion) and the

other the degree of simultaneous incongruence of the

elements of the dual projection (the gradient of disparity).



Granting that a surface in perception [UTP's gloss reads: "I

don't like this sort of talk."] is produced by retinal

images which are textured, the recession or increasing

distance of the surface should be produced by the triple

gradient defined above. There is psychophysical evidence

for this hypothesis (2,13). Moreover, the specific distance

of any point or object on the surface would be given by the

degree of density, disparity, and motility of that point in

the array relative to the gradient. The distance from "here

to there", in short, is probably produced by the difference

with respect to these variables between the image of the

nose in the field of view (the visual "here") and the image

of the object (which is "there"). It would seem that the

stimulus variables for the perception of distance [UTP's

gloss reads: "This is better."] in the visual world have

been literally under our nose all the time.

The length of a stretch of distance along the ground

from "there to there", at whatever distance the stretch may

lie, may be given by some such variable as the number of

transitions or texture elements in the stretch. This would

make possible objective judgements of such distance

stretches at different distances, and these are, in fact,

possible (19). The same reasoning should apply to stretches

of "width" along the ground at different distances as well

as to stretches of "depth", that is, to the frontal as well

as the longitudinal dimension of a receding surface. The

surface should possess a phenomenally constant scale, or in

other worlds, constancy of size aong both dimensions.

The property of the "optical slant" of a delimited

surface such as the face of an object, that is, its apparent

departure from the frontal plane or angle to the line of

sight, should be given by the degree of "one way

compression" of the texture (the ratio of vertical to

horizontal density) along with concomitant degrees of skew

of the texture as regards both disparity and motility (3,

p.173). This implies that constancy of shape of the elements

or structure of the texture is a necessary corollary of

slant-perception. The direction in which a surface slants

or recedes is given by the direction in which the density of

its texture increases and in which the supplementary

gradients run.


An edge in the visual world seems to be characterized by an

abrupt increase in distance as compared to the gradual

increase in distance of a receding surface. The quality and

amount of this depth at an edge should depend respectively

on a stepwise discontinuity in the gradients, and on the

degree of increasing density, disparity, and motility of the

texture at the step. Exploratory experiments suggest that

it does.

An edge should be distinguished from a margin, which is

given by a simple discontinuity of luminous intensity in the

array and which induces no clear depth impression. An edge,

especially if "closed", is what makes the surface of an

object appear in front of the surface of the ground or

background. This reminds one of the figure-ground

phenomenon. But an edge with depth may also surround a

hole, window, gap, or interspace, in which case the surface

inside the edge appears behind the surface outside the edge.

Such a phenomeon has what could only be called a

"frontground". The conclusion of gestalt theorists,

therefore, that a textureless closed contour necessarily

yields a figure-on-ground experience is incorrect, and the

inferences that the differences between figure and ground

are a "field phenomenon", and that a form constitutes a

prototype for visual perception are both equally dubious.

The gestalt theorists could not wholly rid themselves of the

old doctrine that two-dimensional form is primitive and

sensory while three-dimensional depth is secondary and

perceptual. If depth is given in stimulation, however, it

does not have to be derived from sensory organization. A

phenomenal object involves impressions of edges and sufaces,

not of tracings on paper (4).


[UTP's gloss reads: "This use of phenomenal is O.K."]

An observer can be made to see a substantial surface where

no substantial surface exists by manipulating the array of

light to his eyes, that is by artifically producing the

stimulus conditions necessary for the perception. An

optical device for this purpose has been constructed (13).

It is called a "pseudo-tunnel", since the perception it

induces is that of looking into a long cylindrical space or

room. It is a "place", not an "object". Physically it

consists of circular apertures or edges cut in sheets of

smooth material hung behind one another, of alternating low

and high reflectance, which are uniformly illuminated. The

number and frequency of the sheets in the series can be

varied. Optically it yeilds an array to each eye consisting

of either concentric or skewed circular margins, that is,

abrupt alternating transitions of intensity. Perceptually

it yields a solid and substantial tunnel when the

transitions are sufficiently frequent. As their frequency

decreases the perception becomes less solid or substantial.

The walls of the tunnel then lose the quality which Katz,

speaking of colour, called visual "hardness" or

"inpenetrability" (15). When the transitions are wholly

eliminated (by substituting sheets of the same reflectance)

the experience is that of fog filling the nearest aperture.

The solid phenomenal tunnel appearing when the overall

texture of the array is dense consists of a surface which

recedes from the periphery of the field of view to the

center; it slants inward. It appears thus because the

density of the array increases from the periphery to the

center and because (if both eyes are open and the head is

mobile) there is a decreasing crossed disparity and

decreasing crossed displacement of the texture from the

periphery to the center. When an observer peers through the

holes in the sheets the tree gradients are mutually

congruent. The perspective of the light and dark rings is

in geometrical agreement with the perspective of binocular

parallax and that of head-movement parallax. In these

circumstances the depth and distance of this synthetic

visual space is compelling. Two thirds of the observers

report that the walls of the tunnel seem parallel even when

asked if they seem to converge, that is, the far diameter

appears as large as the near diameter (13, p.8).



It is generally recognized that the problem of the size

constancy of objects in perception is somehow related to the

problem of space perception, but the question is precisely

how? This question might be clarified by examining the

physical distinction between things and space. We are

concerned with things of visible magnitude, and we also

concerned with spaces, not with space as such. Things (like

books, students, trees, and cows) are seen in spaces (like

rooms, corridors, gardens, and fields). Spaces are

determined by their surfaces, and so are things. A thing is

physically bounded by its surface but a space is a surface

[UTP's gloss reads: "No."], or at least it always has a

floor or ground and things either rest on it or come to rest

on it. In general, a space is an unbounded surface. The

surface of a thing may be detached from the surface of the

surrounding space, but some things are more separable from

the ground than others; a cow in a field is detachable from

it but a hillock in the field is less so. A thing usually

intercepts a small angle of the optic array; a space is what

fills the whole array. The biggest space we are capable of

seeing is the surface of the terrain (we can only imagine

the spherical surface of the planet) in which case half of

the optical array represents the sky. The sky,

paradoxically, presents scarcely any stimulation for space

perception although it is what we are tempted to call space,

for reasons discoverable in the history of scientific

thought. [UTP's gloss reads: "Space, surely, is the gap

between surfaces into and through which one is free to move.

The distinction you want is between the surfaces which

define a space and the surfaces which define `thing' or

`material object'."]

Once this distinction is clear it is possible to

realize that the difference between a thing and a space is

not absolute. A thing is normally a part of a space, and a

space is a collection of more or less adjacent things. A

surface itself is a collction of adjacent elements of

texture, which look like things on close inspection. A

thing therefore is located in a space relative to other

adjacent structures. Consequently a phenomenal thing is

located in a phenomenal space relative to adjacent

structures in the optical array and relative to the

gradients in it.

The investigators of the size constancy of objects have

not paid sufficient attention to the spaces in which they

set their standard and variable objects. Instead, they have

been concerned with the "cues for depth" of the object

itself, assuming one must have sensations before one can

possibly have a perception. Size constancy is a matter of

discriminating the surface which lies between the observer

and the object, the edges which separate the object from the

surface, and the gradients and steps of stimulation which

determine them. The question is why, when we see an object

in a space, does it not look smaller when observed from a

greater distance in proportion as its retinal image gets

smaller? The answer is that when the object is seen as part

of its space the space of which it is a part does not look

denser or more compact or smaller when that part is observed

from a greater distance. It certainly does not look smaller

in proportion as the retinal image gets denser. The density

is the stimulus for increasing distance, or an aspect of it.

This is true so long as the observer attends to the space of

"world" and is not reminded that he is supposed to have

sensations or to see the perspective of the scene. The

perspectives of double imagery and of parallactic motion are

also usually not seen as such (but then we have not been

trained to attend to them as we have to ordinary

perspective). With fixed monocular stimulation the size of

an object is given by the size of its image relative to the

size of the elements of texture or structure at that part of

the gradient - a ratio rather than a simple magnitiude, that

is, a higher-order variable. Size is perceived relative to

the size-scale of the place where the object is seen. [UTP's

gloss reads: "I like this sort of talk."] When phenomenal

objects are seen of constant size, it is because the

phenomenal space in question is seen as possessing constant

dimensions. If it is not, the objects are not. [UTP's gloss

reads: "But not this."]

An advantage of this theoretical approach is that it

can explain the constancy of shape of objects with the same

hypotheses. It can also account for constancy of velocity

in the same way. The phenomenal rigidity of the world

during locomotion (despite the expansion of the visual field

ahead follows from the same line of reasoning (12). There

is even some promise that constancy of brightness and color

will prove to be consistent with a surface theory of space

perception. The arguments will not be extended here.

Some investigators of space perception are convinced

that phenomenal space does not have constant dimensions at

its extremities from the observer. It can certainly be

pointed out correctly that the spaces one sees are not like

the space of an "empty box without sides" with three

dimensions extending to infinity - the abstract space of

Euclidean-Cartesian geometry. It is therefore easy to make

the assumption that phenomenal space is non-Euclidean as

this term is understood by geometers. This assumption

sounds very interesting, the more so since Luneberg adopted

and elaborated it in the effort to establish a non-Euclidean

metric for binocular visual space (16). The essence of the

position seems to be that perceived space is finite while

physical space is infinite. It is based (as the writer

understands it) on two supposed facts about space

perception, said to be grounded on experimental results.

The first is that perceived distance reaches a limit

asymptotically as physical distance increases without limit.

The second is that perceived size approaches zero as

physical distance increases without limit. It is true, of

course, that the visual field has a horizon. For an

unlimited textured surface (a plane, for the sake of

simplicity) the horizon is the point in the gradient of the

optical array where density becomes infinite - the

"vanishing point" of classical perspective. It is also true

that the optic apparatus is limited by an acuity threshold

below which any increase of density of the array becomes

indiscriminable. But is it a fact that the visual world has

a horizon? [UTP's gloss reads: "I don't know what this

question means."] Does the naive uncritical experience of

trying to make out something in the distance involve seeing

the horizon? The writer doubts it, and the reader should

look or himself in order to decide the question.

It seems to me that perceived size does not vanish and

perceived distance does not reach a maximum as one looks

into the distances. Instead, they both tend to become

indefinite or indeterminable, which is a different matter.

By analogy, when a letter of an acuity chart becomes too

small to read, it is because the separation of its parts has

become imperceptible, not perceptibly zero. It can be

argued, then, that the limits of phenomenal size and

distance on which the Luneberg theory is based are not facts

at all. Phenomenal size depends on a variable of ratio in

the retinal image, not a variable of "retinal size", and

phenomenal distance depends on a density variable, not on

"cues" or sensations.


Every perception or judgement in the so-called third

dimension involves, as one pole, the impression of the point

of view or "here". A report that the variable object is

more distant than the standard is convertible into the

report that "I am more distant from the variable object than

I am from the standard". The theory of sensory cues

neglects this fact. The theory of stimulus gradients,

however, find it implicit in the gradients themselves. The

bottom of the dual and fluid projection for a space like the

terrain (or the periphery of the projection for a space like

a room or tunnel) is unique with respect to density,

disparity, and parallactic motion. The nose is a prominent

feature of the bottom of the optical array; it projects as

the largest structure in the field of view (the minimum of

density), as the maximum of crossed retinal disparity, and

as the maximum of crossed parallactic motion. According to

the present theory it should arouse the maximum possible

impression of nearness or, what is the same thing, the zero

of distance away (3, p.228). Not only a visual space but

also the lcoation of oneself in that space is determined by

optical stimulation.

A point of view has not only position in a space but

also direction in a space. This also is given in

stimulation. A surface at the point of fixation has a

quality which has been termed optical slant. It is

definable as the apparent inclination of the surface to the

plane perpenticular to the line of sight (8). But it is

equally definable as the apparent angle at which I am

looking at the surface. It is given by the slope, or rate

of change, of density of texture at the center of the

incoming projection, or by the degree of one-way compression

of the texture at that point (the density variable being

normally correlated with the other two variables). The

direction of the gaze-line in space is given by optical

stimulation, whether or not it is also given postural

sitmulation. In short, we are sensitive to the particular

sample of the total optical array which our eyes at the

moemt are picking up. Since we do sample the total

spherical array by exploratory eye movements and head

movements - we look around - we are able to survey the whole

of a new environment in a few seconds, and to orient

ourselves in it.

The perception of the whole of an environment,

including where the observer is and where things are, is not

achieved in a single eye-fixation. It depends on

stimulation over time and seemingly on stimulation which

changes with time. Nevertheless what the observer is

responding to are the properties of stimulation which are

invariant over time. He ordinarily pays no attention to the

properties which change with time when he is attending to

space. These invariants, or high-order variables, are the

basis for our perception of the world as boundless,

permanent, concurrent, and as "public" rather than "private"

(3, p.160-162).



How we perceive motion in an environment or space has always

been puzzling if, as seemed obvious, we have nothing to go

on except simple sensations. Actually, if successive series

are a fact of stimulation as well as adjacent series, the

phenomenal qualities of motion may find their explanation in

the resulting variables of high order, and there will be no

need to suppose that motion depends on inference or

intellectual processes. It will, howeve, be necessary to

discover how a man discriminates, or comes to discriminate,

the variables.

A great deal of confusion has resulted from the

seemingly simple assumption that the stimulus for motion is

the motion of a retinal image across the retina. This leads

into a theoretical morass when eye-movements have to be

taken into account. A really simpler assumption, because

more general, is that the stimulus is some transformation or

change of pattern in the optic array to the eye. The array

is not only fluid as a whole when the observer changes

position; it also includes changes of parts relative to the

whole when external objects move relative to their

surroundings. This array is what stimulates the eye (there

are two, of couse, each being a perspective transformation

of the other) and the retinal image is simply that part of

the sheaf which stimulates the retina in a given position of

the eye. The mathematical types of motions and

transformations in this array are in a very neat

correspondence with types of physical events, as the writer

has pointed out (5) and if these are chosen as variables of

stimulation a psychophysics of motion perception is


A whole class of changes in the retinal image, it

should be realized, are produced by responses of the

individual as distinguished from events of the environment.

At one extreme the maximizing of the definition of the image

and a considerabe alteration of the luminous intensity of

the image are produced respectively by accommodation

response and the pupillary response. A variety of

transpositions and rotations of the total image relative to

the retina are produced by the exploratory movements of the

eyes. At another extreme a set of transformation of the

total image are produced by the set of locomotor responses

of the individual (3, ch.7). They involve gradients of

velocity, and can be described by the term motion

perspective (12). All these modes of stimulation are

response-produced or, in a currently fashionable

terminology, they are "feedback" stimuli. Only the last

mentioned, however, has ever been supposed to have the

status of yielding a kind of visual sensation, presumably

because only this is easily introspectable. It has been

called a cue, a sign, or an indicator of space, and the

liklihood that it is a stimulus for the visual impression of

locomotion or change of the point of view has gone

unrecognized. Presumably all these kinds of stimulation,

whether or not conscious, do not have the function of

yielding sensations but of guiding or controlling the

responses which produce them.

The impression of locomotion through an environment is

unquestionably visual, over and above the contribution of

the classical proprioceptors. And now one can understand

better the phenomenal fact that the environment seems stable

and rigid during locomotion rather than moving and elastic.

If motion perspective is a stimulus for moving ahead in the

third dimension it cannot be a stimulus for an expanding

visual picture in two dimensions, or at least it cannot be

both at the same time. The phenomenal rigidity of surfaces

and objects in perception and their constancy of size and

shape as the observer moves in space is the obverse of the

fact that he senses that movement. Both aspects of the

total perception are given by the same stimulation, the one

by its invariant properties in time and the other by its

variation in time.

Phenomenal rigidity of a textured surface (along with

other properties like substantiality, continuity, and Katz's

surface-color) is preserved under all perspective

transformations of the proximal stimulus, but it seems to be

not perserved under non-perspective transformations of the

stimulus (7). These, in contrast, seem to yield the

perception of deformation or elastic movement, that is, of

non rigidity. The eye seems to be sensitive to different

specific kinds of transformation. The perception of elastic

movement is important since it is the kind of movement

characteristic of animate creatures or biological objects.



A psychophysics of the visual world, in order to be

complete, would have to include some account of the

perception of events such as cause-and-effect sequences,

some explanation of how we perceive other organisms

including the behavior and traits of persons, and above all

some reference to the apprehension of written symbols,

including pictures and images, which compose so large a part

of the human world as compared to that of other species.

The question is whether these more abstract features of the

visual world are specified by still higher high-order

variables of stimulation than any yet considered. If a

perceptual psychophysics is possible, how about a

psychophysics so global as to encompass meaning? This is

clearly not an issue to be settled in a few paragraphs but

perhaps the suggestion of an answer can be given.

If simple motions in the environment can be

mathematically specified in the optical stimulus, so can

complex motions. The events or happenings of the

environment are complex motions. They are projected as

transformations in the array, and if we knew the

transformations we should know the potential stimuli for the

perception of the events. They may prove to be not so

impossibly complex as might be supposed. Michotte has

already discovered regularities in the stimulus conditions

for the perception of causation by manipulating only simple

variables of motion, and he has shown that the resulting

impressions are immediate, which implies that events are

directly apprehended from the stimulation projected to the

eye (17). [UTP's gloss reads: "No. For the perception of

`pushing' perhaps. To say A causes B is to say `if A had

not occurred, B would not have occurred.' Whether or not

this is so in a given case can only be determined by

experiment. Causation as such gives rise to no

characteristic stimulation, though certain kinds of

causation, e.g. pushing, do. It is with them that Michotte

is concerned."]

The perception of persons has been said to be a matter

of apprehending "physiognomic" and "dynamic" qualities.

What these may be in the visual mode we do not know, but one

thing is certain: the visible responses of human individuals

from which we mostly judge their intentions, habits,

personality and character are all complex deformations of

the surfaces of their bodies, the facial surface being

especially important, and they are given to the observer as

complex deformations in the optical stimulus. The ability

to communicate by pantomime, even when it is mediated by a

motion picture, is proof of this. Little is known about the

significant variables of such deformations but there is

nothing in principle to prevent an experimental attack on

the problem. A valuable method will be the psychophysical

procedure of using artificial stimulation, with control of

the variables. All that is required is ingenuity in

choosing variables and constructing devices to control


The perception of human artifacts is perhaps the most

complex of all kinds of perception. Like the perception of

speech-sounds, it involves the psychology of communication

between persons, and the problem takes on aspects of this

higher order of stimulus-response relations. Words,

pictures, or symbols are far removed from the natural events

of the physical environment. They are said to be mere

tokens or substitutes for other events not "present to the

senses". Nevertheless it should not be forgotten that they

are concrete sources of optical stimulation (usually

tracings on a paper surface). And, what is more important,

they are specific to the natural objects or events said to

be "absent" (6). If they were not specific they would not

be symbols, i.e. would not signify anything.

The perception of artifacts, then, is genuinely

mediated or indirect perception, and the symbols are

genuinely indicators, clues, or cues to something else. The

perception is said to be based wholly on associative

learning. (It is interesting to realize that the classical

theory of sense data conceived them to be analogous to human

artifacts, as if the communicating and interpreting of a

message were the only way the process of sensation and

perception could be understood.) Nevertheless the fact of

specific correspondence between a set of different symbols

and a set of different percepts (or responses) makes this

kind of perception fundamentally like the kind studied in

the psychophysical experiment. [UTP's gloss reads: "This

sounds rather muddled."]

In what sense can it be said that there exist stimuli

for, say, the letters of the alphabet? There are stimuli in

the obvious sense that deposits of something like ink must

differentially reflect an array of light from a surface like

paper. But there are stimuli for letters in another sense -

that the variables of linear shape in the array specify

discriminable differences between each letter and every

other letter. After learning to discriminate them, there

come to be 28 responses to 28 specific patterns. And after

learning to "read", there may exist 28,000 specific

responses to 28,000 letter-combinations. The variables

which specify the higher order shapes of words are of very

high order indeed, but they certainly exist, and it is

probably to these that the reader specifically responds.

Investigators of the process have long since concluded that

a child does not learn to read by associating the separate

letters in various combinations. There exist not only

letter-forms but phrase-forms. And if so why do there not

also exist sentence-forms, and paragraph-forms, and even

book-forms? The process of learning to read can

legitimately be conceived as one of discriminating and

abstracting the variables of stimulation in a pyramiding

order. As we shall see, the process of learning to perceive

in general can be thought of in the same way.




Empiricism is the belief that knowledge of the world comes

through the sense organs, not from within, together with the

additional hypothesis made necessary by the seeming

unreliability of the senses that knowledge of the world

comes from past experience. The psychophysical theory of

perception makes no assumption about an influence of past

experience on perception. Does this omission make the

theory inconsistent with empiricism, and guilty of an

ativistic taint? Surely not. The emphasis on stimulation

makes it an even more radical empiricism than is customary

in theories of perception. If a postulate about the effect

of past experience on present experience is not necessary to

explain the latter it would be unparsimonious to make it.

Moreover the precise nature of such an effect, despite its

commonsense plausibility, is a troublesome problem which no

one has ever really solved. The problem of past experience,

however, is interesting if not relevant and it should be


The most radical solution to the past experience

problem is to assert that "past" and "present" are

subjective terms - that time is simply a protensive or

sequential attribute of experience and that likewise

sequential order is a dimension of stimulation. We are not

only surrounded by an array of energy; we are also immersed

in a flow of energy. Stimulation includes variables of

temporal change, and these may reach high orders of long

duration. We can apprehend events over longer temporal

spaces as we learn to discriminate these variables of higher

order. Consequently the only problem of so-called past

experience is how to develop a psychophysics of the

perception of "times", by analogy with the perception of

"spaces". There is simply a third sequential dimension in

the two-dimensional optical array which corresponds to the

fourth dimension of three-dimensional physics. In this

solution to the problem memory evaporates so far as

explanation is concerned, and "memories" are reduced to the

status of a psychological curiosity, like "sensations".

This is the most satisfactory theoretical position. The

only trouble with it is that traditional modes of thought

are violated, and that the position itself is undeveloped.

With greater orthodoxy and in more familiar terms the

question can be put in this form: what is the effect of past

stimulation, direct or indirect, on responses to present

stimulation? Since the words "past" and "present" have

subjective connotations, one can ask what is the effect of

previous stimulation on response to subsequent stimulation?

There can be no doubt that there is an effect of earlier

stimulation on the perception (or judgement or response)

induced by later stimulation. The effect is to alter it, as

compared with what the perception would have been in the

absence of the previous stimulation or in the presence of a

different previous stimulation. Just this is what a

properly controlled learning or transfer experiment

demonstrates. The usual answer to the question is that the

effect is indirect: a trace or memory of the previous

stimulation persists and somehow interacts with the direct

effect of the subsequent stimulation to alter the response.

This answer to the question has been termed the

"enrichment theory" of perceptual learning (10). There is

at least one serious objection to it, for it has an awkward

implication. It says that as memories accumulate with time

perception is in a progressively decreasing correspondence

with present stimulation. Perception becomes more

assumptive, more inferential, more imaginary and therefore

more subjective as learning progresses. It means that we

get less and less in touch with the environment with

practice in perceiving the environment. This seems to

contradict what we know about learning. There is a great

deal of evidence to show that the effect of practice on

perception is to make it more precise. Discrimination tends

to improve on repetition, as measured by decreases in the

variable errors and the constant errors of the

psychophysical experiment (1). It is therefore likely as

well as reasonable that we get more and more in touch with

the environment as a result of previous experience.

Presumably we do so by learning to respond specifically to

variables of the environment not previously responded to,

that is, variables of higher order. The effect of previous

stimulation on the perception of subsequent stimulation is

to make it more finely discriminative.

The latter position has been termed the

"differentiation theory" of perceptual learning (10, 11,

18). It admits an influence of the past on the present in

experience or behavior but it dispenses with the notion of

traces or memories as the explanation of this influence. It

also dispenses with the theory of association. It says that

an organism can learn in time to respond appropriately to

the permanent properties of its environment even when these

are subtle and intricate. It can respond to [UTP's gloss

reads: "To understand or appreciate something is to respond

to it not to its meaning or value."] the meanings, values,

or valences of things insofar as they are unique to the

objects, and inasmuch as each thing is discriminated from

other things. But it does not necessarily have to remember

or recall or associate in order thus to respond

appropriately, as the enrichment theory assumes. It may

simply discriminate, abstract, and identify the meanings

which are implicit in the subtle properties of stimulation.

The differentiation theory of perceptual development,

it should be noted, has to assume that the permanent

properties of the environment are implicit in the sea of

surrounding environmental energy. In the case of the

projective properties of light, which an animal samples with

every fixation of its eyes, this assumption has already been

discussed. The total environment is implicit in the

potential optical stimuli which an organism may get, or

learn to get, not in the momentary array at a given instant

of time. The difficulty with the past experience theory of

perceptual development arises when one tries to define what

is meant by present experience. Presumably by "present" not

theorist has ever meant an instantaneous cross section, but

a certain duration or span of temporal experience. What

span shall be chosen - a second, six seconds, an hour, or a

week? in the face of this difficulty the more radical

theoretical approach of sequential psychophysics, mentioned

earlier, seems desirable.

The organism learns to respond appropriately not only

to the "permanent properties" of the environment, but also

to its lawful changes, movements, events, and the directions

of events. Within the transformations of proximal

stimulation it distinguishes between those aspects of it

which are invariant over time and those not invariant over

time, as suggested earlier. The latter specify the changes

of things in the physical world and the former specify the

unchanging things. If the concept of stimulation is taken

broadly enough it can encompass not only perception but

cognition, knowledge [UTP's gloss reads: "No. This is on

the response side."], and thought [UTP's gloss reads:

"Perhaps - `pure stimulus act'"] as well.

To the extent that the energies at the surface of the

organism consist of a vast potential of stimuli which it may

react to in lesser or greater degree, the problems of

maturation, learning, attention, and habit formation, like

the problem of perception, can all be formulated as the

investigation of what the organism is specifically

responding to. Development is a matter of the increase in

the specificity of responses to stimuli, not a matter of the

connection of "old" responses to "new" stimuli, or of the

connection of "new" responses to "old" stimuli. Instead,

behavior becomes differentiated, with repetition. At the

level of neural physiology it is probably not that new

connections are formed between receptors and muscles but

that the nervous system itself becomes progressively more


This theory of the specific correspondence of the

qualitites of perception to variables of stimulation is

intended as an alternative to the theory that perception is

supplemented by memory. It does not deny the existence of

memory in some sense of that term; it only asserts that the

process of remembering, recalling, recollecting, or the

"reinstatement of past experience" is not a process which

contributes to the development of perception. [UTP's gloss

reads: "One must, however, remember the names of things and

what to do with them."] Remembering or reminiscing is a

symptom of learning, not the cause of it. No one will deny

that a man can remember, recall, recollect, etc. He often

does. But it need not be assumed that he has to mix

memories with sensations in order to have perceptions,

meanings, cognitions, or knowledge of the world in which he



1. Gibson, Eleanor, Improvement in perceptual judgments

as a function of controlled practice or training.

Psychol. Bull., 1953, 540, 401-431.

2. Gibson, J.J. The perception of visual surfaces. Amer. J.

Psychol., 63, 1950, 367-384.

3. Gibson, J.J. The perception of the visual world. Boston:

Houghton-Mifflin, 1951.

4. Gibson, J.J. What is a form? Psychol. Rev., 1951, 58,


5. Gibson, J.J. The visual perception of objective motion

and subjective movement. Psychol. Rev., 1954, 61, 304-


6. Gibson, J.J. Outline of a theory of picture perception.

Audio-Visual Aids Quart., 1954.

7. Gibson, J.J. Optical motions and transformations as

stimuli for visual perception. Film, Cornell Univ.

Photo-science service, Ithaca, N.Y., 1955.

8. Gibson, J.J. & Cornsweet, Janet, The perceived slant of

visual surfaces - optical and geographical. J. Exp.

Psychol., 1952, 44, 11-15.

9. Gibson, J.J. & Dibble, F.N., Exploratory experiments on

the stimulus conditions for the perception of a visual

surface. J.Exp.Psychol., 1952, 43, 414-419.

10. Gibson, J.J. & Gibson, Eleanor, Perceptual learning:

differentiation or enrichment? Psychol. Rev., 1955, 62,


11. Gibson, J.J. & Gibson, Eleanor, What is learned in

perceptual learning? A reply to Prof. Postman. Psychol.

Rev., 1955, 62, 447-450.

12. Gibson, J.J., Olum, P. & Rosenblatt, F. Parallax and

perspective during aircraft landings. Amer. J. Psychol.,

1955, 68, 372-385.

13. Gibson, J.J., Purdy, Jean, & Lawrence, Lois, A method of

controlling stimulation for the study of space

perception: the optical tunnel. J.Exp.Psychol., 50,

1955, 1-14.

14. Gibson, J.J. & Waddell, D. Homogeneous retinal

stimulation and and visual perception. Amer.J.Psychol.,

1952, 45, 263-270.

15. Katz, D. The world of colour. (Tr. by R.B. MacLeod and

C.W. Fox) London: Kegan Paul, Trench, Trubner & Co.,


16. Luneberg, R.K. Mathematical Analysis of binocular vision.

Princeton Univ. Press, 1947.

17. Michotte, A. La perception de la causalit. Studia

Psychologica, Louvain, 1954.

18. Postman, L. Association theory and perceptual learning.

Psychol.Rev., 1955, 62, 438-446.

19. Purdy, Jean & Gibson, Eleanor, A study of distance

judgement by the method of fractionation. J.Exp.Psychol.

in press.

20. Stevens, S.S. (Ed) Handbook of Experimental Psychology.

N.Y.: Wiley, 1951. Ch.I.