THE ECOLOGICAL
APPROACH TO VISUAL
PERCEPTION
Classic Edition
James J. Gibson
TABLE OF CONTENTS
Preface xi
Introduction xiii
Introduction to the Classic Edition xvii
PART I
The Environment to be Perceived 1
1 The Animal and the Environment 3
2 Medium, Substances, Surfaces 12
3 The Meaningful Environment 28
PART II
The Information for Visual Perception 39
4 The Relationship Between Stimulation and
Stimulus Information 41
5 The Ambient Optic Array 58
6 Events and the Information for Perceiving Events 85
x Contents
7 The Optical Information for Self-Perception 104
8 The Theory of Affordances 119
PART III
Visual Perception 137
9 Experimental Evidence for Direct Perception:
Persisting Layout 139
10 Experiments on the Perception of Motion in the World
and Movement of the Self 162
11 The Discovery of the Occluding Edge and Its
Implications for Perception 180
12 Looking with the Head and Eyes 193
13 Locomotion and Manipulation 213
14 The Theory of Information Pickup and Its Consequences 227
PART IV
Depiction 253
15 Pictures and Visual Awareness 255
16 Motion Pictures and Visual Awareness 279
Conclusion 290
Appendix 1: The Principal Terms Used in Ecological Optics 294
Appendix 2: The Concept of Invariants in Ecological Optics 297
Bibliography 299
Index 305
PART I
The Environment to
be Perceived
In this book, environment will refer to the surroundings of those organisms that
perceive and behave, that is to say, animals. The environment of plants, organisms
that lack sense organs and muscles, is not relevant in the study of perception
and behavior. We shall treat the vegetation of the world as animals do, as if it
were lumped together with the inorganic minerals of the world, with the physical,
chemical, and geological environment. Plants in general are not animate;
they do not move about, they do not behave, they lack a nervous system, and
they do not have sensations. In these respects they are like the objects of physics,
chemistry, and geology.
The world can be described at different levels, and one can choose which
level to begin with. Biology begins with the division between the nonliving
and the living. But psychology begins with the division between the inanimate
and the animate, and this is where we choose to begin. The animals themselves
can be divided in different ways. Zoology classiﬁes them by heredity and
anatomy, by phylum, class, order, genus, and species, but psychology can classify
them by their way of life, as predatory or preyed upon, terrestrial or aquatic,
crawling or walking, ﬂying or nonﬂying, and arboreal or ground-living. We
are more interested in ways of life than in heredity.
The environment consists of the surroundings of animals. Let us observe that
in one sense the surroundings of a single animal are the same as the surroundings
of all animals but that in another sense the surroundings of a single animal
are different from those of any other animal. These two senses of the term can
be troublesome and may cause confusion. The apparent contradiction can be
resolved, but let us defer the problem until later. (The solution lies in the fact
that animals are mobile.) For the present it is enough to note that the surroundings
of any animal include other animals as well as the plants and the nonliving
1
THE ANIMAL AND THE
ENVIRONMENT
4 The Ecological Approach to Visual Perception
things. The former are just as much parts of its environment as the inanimate
parts. For any animal needs to distinguish not only the substances and objects
of its material environment but also the other animals and the differences
between them. It cannot afford to confuse prey with predator, own-species
with another species, or male with female.
The Mutuality of Animal and Environment
The fact is worth remembering because it is often neglected that the words
animal and environment make an inseparable pair. Each term implies the other. No
animal could exist without an environment surrounding it. Equally, although
not so obvious, an environment implies an animal (or at least an organism) to be
surrounded. This means that the surface of the earth, millions of years ago before
life developed on it, was not an environment, properly speaking. The earth was
a physical reality, a part of the universe, and the subject matter of geology. It was
a potential environment, prerequisite to the evolution of life on this planet. We
might agree to call it a world, but it was not an environment.
The mutuality of animal and environment is not implied by physics and the
physical sciences. The basic concepts of space, time, matter, and energy do not
lead naturally to the organism-environment concept or to the concept of a
species and its habitat. Instead, they seem to lead to the idea of an animal as an
extremely complex object of the physical world. The animal is thought of as a
highly organized part of the physical world but still a part and still an object.
This way of thinking neglects the fact that the animal-object is surrounded in
a special way, that an environment is ambient for a living object in a different
way from the way that a set of objects is ambient for a physical object. The term
physical environment is, therefore, apt to get us mixed up, and it will usually be
avoided in this book.
Every animal is, in some degree at least, a perceiver and a behaver. It is
sentient and animate, to use old-fashioned terms. It is a perceiver of the environment
and a behaver in the environment. But this is not to say that it perceives
the world of physics and behaves in the space and time of physics.
The Difference Between the Animal Environment and the
Physical World
The world of physics encompasses everything from atoms through terrestrial
objects to galaxies. These things exist at different levels of size that go to almost
unimaginable extremes. The physical world of atoms and their ultimate particles
is measured at the level of millionths of a millimeter and less. The astronomical
world of stars and galaxies is measured at the level of light-years and
more. Neither of these extremes is an environment. The size-level at which the
environment exists is the intermediate one that is measured in millimeters and
The Animal and the Environment 5
meters. The ordinary familiar things of the earth are of this size—actually a
narrow band of sizes relative to the far extremes. The sizes of animals, similarly,
are limited to the intermediate terrestrial scale. The size of the smallest animal
is an appreciable fraction of a millimeter, and that of the largest is only a few
meters.
The masses of animals, likewise, are measured within the range of milligrams
to kilograms, not at the extremes of the scale, and for good physiological
reasons. A cell must have a minimum of substances in order to permit biochemical
reactions; living animals cannot exceed a maximum mass of cells if they are
all to be nourished and if they are to be mobile. In short, the sizes and masses
of things in the environment are comparable with those of the animals.
Units of the Environment
Physical reality has structure at all levels of metric size from atoms to galaxies.
Within the intermediate band of terrestrial sizes, the environment of animals
and men is itself structured at various levels of size. At the level of kilometers,
the earth is shaped by mountains and hills. At the level of meters, it is formed
by boulders and cliffs and canyons, and also by trees. It is still more ﬁnely structured
at the level of millimeters by pebbles and crystals and particles of soil, and
also by leaves and grass blades and plant cells. All these things are structural
units of the terrestrial environment, what we loosely call the forms or shapes of
our familiar world.
Now, with respect to these units, an essential point of theory must be
emphasized. The smaller units are embedded in the larger units by what I will
call nesting. For example, canyons are nested within mountains; trees are nested
within canyons; leaves are nested within trees; and cells are nested within
leaves. There are forms within forms both up and down the scale of size. Units
are nested within larger units. Things are components of other things. They
would constitute a hierarchy except that this hierarchy is not categorical but
full of transitions and overlaps. Hence, for the terrestrial environment, there is
no special proper unit in terms of which it can be analyzed once and for all.
There are no atomic units of the world considered as an environment. Instead,
there are subordinate and superordinate units. The unit you choose for
describing the environment depends on the level of the environment you
choose to describe.
The size-levels of the world emphasized by modern physics, the atomic and
the cosmic, are inappropriate for the psychologist. We are concerned here with
things at the ecological level, with the habitat of animals and men, because we
all behave with respect to things we can look at and feel, or smell and taste, and
events we can listen to. The sense organs of animals, the perceptual systems
(Gibson, 1966b), are not capable of detecting atoms or galaxies. Within their
limits, however, these perceptual systems are still capable of detecting a certain
6 The Ecological Approach to Visual Perception
range of things and events. One can see a mountain if it is far enough away and
a grain of sand if it is close enough. That fact is sufﬁciently wonderful in itself
to deserve study, and it is one of the facts that this book will try to explain.
The explanation of how we human observers, at least some of us, can
visualize an atom or a galaxy even if we cannot see one will not be attempted at
this stage of the inquiry. It is not so much a problem of perception as it is of
thinking, and there will be more about this later. We must ﬁrst consider how
we can perceive the environment—how we apprehend the same things that our
human ancestors did before they learned about atoms and galaxies. We are
concerned with direct perception, not so much with the indirect perception got
by using microscopes and telescopes or by photographs and pictures, and still
less with the kind of apprehension got by speech and writing. These higherorder
modes of apprehension will only be considered in Part IV of this book, at
the end.
Units of the Ground Surface
The literal basis of the terrestrial environment is the ground, the underlying
surface of support that tends to be on the average ﬂat—that is to say, a plane—and
also level, or perpendicular to gravity. And the ground itself is structured at
various levels of metric size, these units being nested within one another. The
fact to be noted now, since it is important for the theory of perspective in Part II,
is that these units tend to be repeated over the whole surface of the earth. Grains
of sand tend to be of the same size everywhere, and so do pebbles and rocks.
Blades of grass are all more or less similar to one another, and so are clumps of
grass and bushes. These natural units are not, of course, perfectly uniform like
the man-made tiles of a pavement. Nevertheless, even if their repetition is not
metrically regular, it is stochastically regular, that is to say, regular in a probabilistic
way. In short, the component units of the ground do not get smaller as one
goes north, for instance. They tend to be evenly spaced; and if they are scattered,
they tend to be evenly scattered.
The Time Scale of the Environment: Events
Another difference between the environment to be described and the world of
physics is in the temporal scale of the process and events we choose to consider.
The duration of processes at the level of the universe may be measured in
millions of years, and the duration of processes at the level of the atom may be
measured in millionths of a second. But the duration of processes in the environment
is measured only in years and seconds. The various life spans of the
animals themselves fall within this range. The changes that are perceived, those
on which acts of behavior depend, are neither extremely slow nor extremely
rapid. Human observers cannot perceive the erosion of a mountain, but they
The Animal and the Environment 7
can detect the fall of a rock. They can notice the displacement of a chair in a
room but not the shift of an electron in an atom.
The same thing holds for frequencies as for durations. The very slow cycles
of the world are imperceptible, and so are the very rapid cycles. But at the
level of a mechanical clock, each motion of the pendulum can be seen and each
click of the escapement can be heard. The rate of change, the transition, is
within the limits of perceptibility.
In this book, emphasis will be placed on events, cycles, and changes at the
terrestrial level of the physical world. The changes we shall study are those that
occur in the environment. I shall talk about changes, events, and sequences of
FIGURE 1.1 The structure of the terrestrial earth as seen from above.
In this aerial photograph only the large-scale features of the terrain are shown.
(Photo by Grant Heilman)
8 The Ecological Approach to Visual Perception
events but not about time as such. The ﬂow of abstract empty time, however useful
this concept may be to the physicist, has no reality for an animal. We perceive not
time but processes, changes, sequences, or so I shall assume. The human awareness
of clock-time, socialized time, is another matter.
Just as physical reality has structure at all levels of metric size, so it has structure
at all levels of metric duration. Terrestrial processes occur at the intermediate
level of duration. They are the natural units of sequential structure. And
once more it is important to realize that smaller units are nested within larger
units. There are events within events, as there are forms within forms, up to the
yearly shift of the path of the sun across the sky and down to the breaking of a
twig. And hence there are no elementary units of temporal structure. You can
describe the events of the environment at various levels.
The acts of animals themselves, like the events of the environment they
perceive, can be described at various levels, as subordinate and superordinate
acts. And the duration of animal acts is comparable to the duration of environmental
events. There are no elementary atomic responses.
The natural units of the terrestrial environment and the natural units of
terrestrial events should not be confused with the metrical units of space and
time. The latter are arbitrary and conventional. The former are unitary in one
sense of the term, and the latter are unitary in a quite different sense. A single
whole is not the same as a standard of measurement.
Permanence and Change of the Layout
Space and time will not often be referred to in this book, but a great deal will
be said about permanence and change. Consider the shape of the terrestrial
environment, or what may be called its layout. It will be assumed that the layout
of the environment is both permanent in some respects and changing in some
other respects. A living room, for example, is relatively permanent with respect
to the layout of ﬂoor, walls, and ceiling, but every now and then the arrangement
of the furniture in the room is changed. The shape of a growing child is
relatively permanent for some features and changing for others. An observer
can recognize the same room on different occasions while perceiving the
change of arrangement, or the same child at different ages while noticing her
growth. The permanence underlies the change.
Permanence is relative, of course; that is, it depends on whether you mean
persistenceoveraday,ayear,oramillennium.Almostnothingisforeverpermanent;
nothing is either immutable or mutable. So it is better to speak of persistence under
change. The “permanent objects” of the world, which are of so much concern to
psychologists and philosophers, are actually only objects that persist for a very long
time.
The abstract notion of invariance and variance in mathematics is related
to what is meant by persistence and change in the environment. There are
The Animal and the Environment 9
variants and invariants in any transformation, constants and variables. Some
properties are conserved and others not conserved. The same words are not
used by all writers (for example, Piaget, 1969), but there is a common core of
meaning in all such pairs of terms. The point to be noted is that for persistence
and change, for invariant and variant, each term of the pair is reciprocal to the
other.
Persistence in the Environment
The persistence of the geometrical layout of the environment depends in part
on the kind of substance composing it and its rigidity or resistance to deformation.
A solid substance is not readily changed in shape. A semisolid substance is
more easily changed in shape. A liquid substance takes on whatever may be the
shape of its solid container. The upper surface of a liquid substance tends to the
ideal shape of a plane perpendicular to gravity, but this is easily disturbed, as
when waves form. When we speak of the permanent layout of the environment,
therefore, we refer mainly to the solid substances. The liquids of the
world, the streams and oceans, are shaped by the solids, and as for the gaseous
matter of the world, the air, it is not shaped at all. I will argue that the air is
actually a medium for terrestrial animals.
When a solid substance with a constant shape melts, as a block of ice melts,
we say that the object has ceased to exist. This way of speaking is ecological, not
physical, for there is physical conservation of matter and mass despite the change
from solid to liquid. The same would be true if a shaped object disintegrated,
changing from solid to granular. The object does not persist, but the matter
does. Ecology calls this a nonpersistence, a destruction of the object, whereas
physics calls it a mere change of state. Both assertions are correct, but the former
is more relevant to the behavior of animals and children. Physics has sometimes
been taken to imply that when a liquid mass has evaporated and the substance
has been wholly dispersed in the air, or when an object has been consumed by
ﬁre, nothing has really gone out of existence. But this is an error. Even if terrestrial
matter cannot be annihilated, a resistant light-reﬂecting surface can, and
this is what counts for perception.
Going out of existence, cessation or destruction, is a kind of environmental
event and one that is extremely important to perceive. When something is
burned up, or dissolved, or shattered, it disappears. But it disappears in special
ways that have recently been investigated at Cornell (Gibson, 1968a). It does
not disappear in the way that a thing does when it becomes hidden or goes
around a corner. Instead, the form of the object may be optically dispersed or
dissipated, in the manner of smoke. The visual basis of this kind of perception
will be further considered in Part II on ecological optics.
The environment normally manifests some things that persist and some that
do not, some features that are invariant and some that are variant. A wholly
10 The Ecological Approach to Visual Perception
invariant environment, unchanging in all parts and motionless, would be
completely rigid and obviously would no longer be an environment. In fact,
there would be neither animals nor plants. At the other extreme, an environment
that was changing in all parts and was wholly variant, consisting only of
swirling clouds of matter, would also not be an environment. In both extreme
cases there would be space, time, matter, and energy, but there would be no
habitat.
The fact of an environment that is mainly rigid but partly nonrigid,
mainly motionless but partly movable, a world that is both changeless in many
respects and changeable in others but is neither dead at one extreme nor chaotic
at the other, is of great importance for our inquiry. This fact will become
evident later when we talk about the geometry of the environment and its
transformations.
ON PERSISTENCE AND CHANGE
Our failure to understand the concurrence of persistence and change at the
ecological level is probably connected with an old idea—the atomic theory of
persistence and change, which asserts that what persists in the world are
atoms and what changes in the world are the positions of atoms, or their
arrangement. This is still an inﬂuential assumption in modern physics and
chemistry, although it goes back to Democritus and the Greek thinkers who
followed him. There will be more about the atomistic assumption in
Chapter 6 on events and how they are perceived.
Motion in the Environment
The motions of things in the environment are of a different order from the
motions of bodies in space. The fundamental laws of motion hold for celestial
mechanics, but events on earth do not have the elegant simplicity of the motions
of planets. Events on earth begin and end abruptly instead of being continuous.
Pure velocity and acceleration, either linear or angular, are rarely observable
except in machines. And there are very few ideal elastic bodies except for
billiard balls. The terrestrial world is mostly made of surfaces, not of bodies in
space. And these surfaces often ﬂow or undergo stretching, squeezing, bending,
and breaking in ways of enormous mechanical complexity.
So different, in fact, are environmental motions from those studied by Isaac
Newton that it is best to think of them as changes of structure rather than
changes of position of elementary bodies, changes of form rather than of point
locations, or changes in the layout rather than motions in the usual meaning of
the term.
The Animal and the Environment 11
Summary
The environment of animals and men is what they perceive. The environment
is not the same as the physical world, if one means by that the world described
by physics.
The observer and his environment are complementary. So are the set of
observers and their common environment.
The components and events of the environment fall into natural units.
These units are nested. They should not be confused with the metric units of
space and time.
The environment persists in some respects and changes in other respects.
The most radical change is going out of existence or coming into existence.
4
THE RELATIONSHIP BETWEEN
STIMULATION AND STIMULUS
INFORMATION
Having described the environment, I shall now describe the information available
to observers for perceiving the environment. Only then will we be
prepared to consider how they perceive, what the activity of perception consists
of, and how they can control behavior in the environment.
For visual perception, the information is obviously in light. But the term
light means different things in different sciences, and we shall have to sort out
the different meanings to avoid confusion. Most of us are confused, including
the scientists themselves. The science of light is called optics. But the science of
vision is also called optics, and the textbooks are not at all clear about the difference.
Let us try to distinguish light as physical energy, light as a stimulus for
vision, and light as information for perception.
What I call ecological optics is concerned with the available information for
perception and differs from physical optics, from geometrical optics, and also
from physiological optics. Ecological optics cuts across the boundaries of these
existing disciplines, borrowing from all but going beyond them.
Ecological optics rests on several distinctions that are not basic in physical
optics: the distinction between luminous bodies and nonluminous bodies; the
difference between light as radiation and light as illumination; and the difference
between radiant light, propagating outward from a source, and ambient
light, coming to a point in a medium where an eye might be stationed. Since
these differences are fundamental, they should be stated at the beginning. Why
they are so important will become clear.
42 The Ecological Approach to Visual Perception
The Distinction Between Luminous and Illuminated Bodies
Some material bodies emit light, and others do not. Light comes from sources
such as the sun in the sky and from other sources close at hand such as ﬁres or
lamps on the earth. They “give” light, as we say, whereas ordinary objects do
not. Nonluminous objects only reﬂect some part of the light that falls on them
from a source. And yet we can see the nonluminous bodies along with the
luminous ones. In fact, most of the things that need to be seen are nonluminous;
they are only seen “by the light of” the source. The question is, how are they
seen? For they do not stimulate the eye with light in the same way that luminous
bodies do. The intermediate case of luminescent bodies is exceptional.
A terrestrial surface that gives light is usually, although not always, distinguishable
from one that does not; it is visibly luminous, as distinct from being
visibly illuminated. In physical optics, the case of reﬂected light is reduced to
the re-emission of light by the atoms of the reﬂecting surface. But in ecological
optics, the difference between a luminous and an illuminated surface is crucial.
Where a reﬂecting surface in physical optics is treated as if it were a dense set of
tiny luminous bodies, in ecological optics a reﬂecting surface is treated as if it
were a true surface having a texture. There will be more of this later.
The Distinction Between Radiation and Illumination
Radiant energy as studied in physics is propagated through empty space at
enormous velocity. Such energy can be treated either as particles or as waves
(and this is a great puzzle, even to physicists), but it travels in straight lines, or
rays. The paths of photons are straight lines, and the perpendiculars to the wave
fronts are straight lines. Moreover, light comes from atoms and returns to
atoms. They give off and take in energy in quantal units. Matter and energy
interact. There are elegant laws of this radiation, both at the size-level of atoms
and on the grand scale of the universe. But at the ecological level of substances,
surfaces, and the medium, we need be concerned only with some of these laws,
chieﬂy scattering, reﬂection, and absorption.
WHY ECOLOGICAL OPTICS?
The term ecological optics ﬁrst appeared in print in an article with that title
in Vision Research (Gibson, 1961). It seemed to me that the study of light,
over the centuries, had not produced a coherent discipline. The science
of radiant energy in physics, the science of optical instruments, and the
science of the eye were quite different. The textbooks and journals of
optics gave the impression of monolithic authority, but there were deep
contradictions between the assumptions of the various branches of optics.
The Relationship Between Stimulation and Stimulus Information 43
When I discovered that even an occasional physicist recognized these
cracks in the foundations of the optical establishment (Ronchi, 1957),
I ventured to suggest that optics at the level appropriate for perception
should have a new name.
In daylight, part of the radiant light of the sun reaches the earth in parallel rays,
but another part is scattered by being transmitted through an atmosphere that
is never perfectly transparent. This light is even more thoroughly scattered
when it strikes the textured ground, by what can be called scatter reﬂection. (This
is not to be confused with mirror reﬂection, which is governed by the simple law
of equal angles of the incident ray and the reﬂected ray. Mirror reﬂection
seldom happens, for there are no mirrors on the ground, and even water
surfaces, which could act as mirrors, are usually rippled.) The scatter-reﬂected
light is in turn reﬂected back from the sky. Each new reﬂection further disperses
FIGURE 4.1 The steady state of reverberating light in an illuminated medium under
the sky.
Although at any point in the air the illumination comes from all directions, the
prevailing illumination is from the left in this diagram because the direct radiation
from the sun comes from the left.
44 The Ecological Approach to Visual Perception
the incident rays. The light thus ﬁnds its way into shelters that are not open to
the sun, or even to the sky. In semienclosed spaces the light continues to bounce
back and forth at 186,000 miles per second. It ﬁnds its way through chinks and
crevices and into caverns, until the energy is ﬁnally absorbed. This light can
hardly be thought of as radiation now; it is illumination.
Illumination is a fact of higher order than radiation. In physical optics,
experimenters try to avoid what they call stray light in the dark room. But in
ecological optics, this light that has gone astray is just what interests us. The
opticist works with rays of light, rays that diverge in all directions from their
source and never converge to a point unless they are focused by a lens. But an
organism has to work with light that converges from all directions and, moreover,
has different intensities in different directions.
Many-times reﬂected light in a medium has a number of consequences that,
although important for vision, have not been recognized by students of optics.
Chief among them is the fact of ambient light, that is, light that surrounds a
point, any point, in the space where an observer could be stationed.
The Distinction Between Radiant Light and Ambient Light
Radiation becomes illumination by reverberating between the earth and the sky
and between surfaces that face one another. But that term, referring as it does
to sound, does not do justice to the unimaginable quickness of the ﬂux or to the
uncountable multiplicity of the reﬂections back and forth or to their unlimited
scattering. If the illumination is conceived as a manifold of rays, one can
imagine every point on every surface of any environment as radiating rays
outward from that point, as physicists do. Every such radiating pencil is
completely “dense.” One could think of the rays as completely ﬁlling the air
and think of each point in the air as a point of intersection of rays coming from
FIGURE 4.2 Radiant light from a point source and ambient light to a point in the
medium.
A creature with eyes is shown at the point in the air, but it need not be occupied.
The Relationship Between Stimulation and Stimulus Information 45
all directions. It would follow that light is ambient at every point. Light would
come to every point; it would surround every point; it would be environing at
every point. This is one way of conceiving ambient light.
Such an omnidirectional ﬂux of light could not exist in empty space but
only in an environment of reﬂecting surfaces. In any ordinary terrestrial space,
the illumination reaches an equilibrium, that is, it achieves what is called a
steady state. The input of energy from the sun is just balanced by the absorption
of energy at the surfaces. With any change in the source, a new steady state is
immediately reached, as when the sun goes down or is hidden by a cloud. No
matter how abrupt the rise or fall of intensity of the light coming from a lamp,
the rise or fall of illumination in the room is just as abrupt. The system is said
to be open rather than closed inasmuch as addition of energy to the airspace and
subtraction of energy from it are going on all the time, but the structure of the
reverberation remains the same and does not change. What could this structure
be? It is possible to conceive a nested set of solid angles at each point in the
medium, as distinguished from a dense set of intersecting lines. The set of solid
angles would be the same whatever the intensity of illumination might be
(there will be more about this later). They are angles of intercept, based on the
environment. The ﬂow of energy is relevant to the stimulation of a retina, but
the set of solid angles considered as projections is more relevant to stimulus
information.
Consider the differences between radiant light and ambient light that have
so far been stated or implied. Radiant light causes illumination; ambient light
is the result of illumination. Radiant light diverges from an energy source;
ambient light converges to a point of observation. Radiant light must consist of
an inﬁnitely dense set of rays; ambient light can be thought of as a set of solid
angles having a common apex. Radiant light from a point source is not different
in different directions; ambient light at a point is different in different directions.
Radiant light has no structure; ambient light has structure. Radiant light
is propagated; ambient light is not, it is simply there. Radiant light comes from
atoms and returns to atoms; ambient light depends upon an environment of
surfaces. Radiant light is energy; ambient light can be information.
The Structuring of Ambient Light
Only insofar as ambient light has structure does it specify the environment. I
mean by this that the light at the point of observation has to be different in
different directions (or there have to be differences in different directions) in order
for it to contain any information. The differences are principally differences of
intensity. The term that will be used to describe ambient light with structure is
an ambient optic array. This implies an arrangement of some sort, that is, a pattern,
a texture, or a conﬁguration. The array has to have parts. The ambient light
cannot be homogeneous or blank. (See the illustrations in Chapter 5.)
46 The Ecological Approach to Visual Perception
What would be the limiting case of ambient light without structure? It would
arise if the air were ﬁlled with such a dense fog that the light could not reverberate
between surfaces but only between the droplets or particles in the
medium. The air would then be translucent but not transparent. Multiple
reﬂection would occur only between closely packed microsurfaces, yielding a
sort of microillumination of things too small to see. At any point of observation
there would be radiation, but without differences in different directions,
without transitions or gradations of intensity, there would be no structure and
no array. Similarly, homogeneous ambient light would occur inside a translucent
shell of some strongly diffusing substance that was illuminated from
outside. The shell would transmit light but not structure.
In the case of unstructured ambient light, an environment is not speciﬁed
and no information about an environment is available. Since the light is undifferentiated,
it cannot be discriminated, and there is no information in any
meaning of that term. The ambient light in this respect is no different from
ambient darkness. An environment could exist behind the fog or the darkness,
or nothing could exist; either alternative is possible. In the case of ambient light
that is unstructured in one part and structured in an adjacent part, such as the
blue sky above the horizon and the textured region below it, the former speciﬁes
a void and the latter a surface. Similarly, the homogeneous area between
clouds speciﬁes emptiness, and the heterogeneous areas specify clouds.
The structuring of ambient light by surfaces, especially by their pigmentation
and their layout, will be described in the next chapter. Chieﬂy, it is the opaque
surfaces of the world that reﬂect light, but we must also consider the luminous
surfaces that emit light and the semitransparent surfaces that transmit light. As
far as the evidence goes, we will describe how the light speciﬁes these surfaces,
their composition, texture, color, and layout, their gross properties, not their
atomic properties. And this specifying of them is useful information about them.
Stimulation and Stimulus Information
In order to stimulate a photoreceptor, that is, to excite it and make it “ﬁre,”
light energy must be absorbed by it, and this energy must exceed a certain characteristic
amount known as the threshold of the receptor. Energy must be transduced,
as the physiologist likes to put it, from one form to another. The rule is
supposed to hold for each of a whole bank of photoreceptors, such as is found
in the retina. Hence, if an eye were to be stationed at some point where there
is ambient light, part of the light would enter the pupil, be absorbed, and act as
stimulation. If no eye or any other body that absorbs light is stationed at that
point, the ﬂying photons in the air (or the wave fronts) would simply pass
through the point without interfering with one another. Only potential stimulation
exists at such a point. Actual stimulation depends on the presence of
photoreceptors.
The Relationship Between Stimulation and Stimulus Information 47
Consider an observer with an eye at a point in a fog-ﬁlled medium. The
receptors in the retina would be stimulated, and there would consequently
be impulses in the ﬁbers of the optic nerve. But the light entering the pupil
of the eye would not be different in different directions; it would be unfocusable,
and no image could be formed on the retina. There could be no retinal
image because the light on the retina would be just as homogeneous as the
ambient light outside the eye. The possessor of the eye could not ﬁx it on
anything, and the eye would drift aimlessly. He could not look from one
item to another, for no items would be present. If he turned the eye, the
experience would be just what it was before. If he moved the eye forward
in space, nothing in the ﬁeld of view would change. Nothing he could
do would make any difference in what he could experience, with this single
exception: if he closed the eye, an experience that he might call brightness
would give way to one he might call darkness. He could distinguish between
stimulation of his photoreceptors and nonstimulation of them. But as far as
perceiving goes, his eye would be just as blind when light entered it as it would
be when light did not.
This hypothetical case demonstrates the difference between the retina and
the eye, that is, the difference between receptors and a perceptual organ.
Receptors are stimulated, whereas an organ is activated. There can be stimulation
of a retina by light without any activation of the eye by stimulus information.
Actually, the eye is part of a dual organ, one of a pair of mobile eyes, and
they are set in a head that can turn, attached to a body that can move from place
to place. These organs make a hierarchy and constitute what I have called a
perceptual system (Gibson, 1966b, Ch. 3). Such a system is never simply stimulated
but instead can go into activity in the presence of stimulus information.
The characteristic activities of the visual system will be described in Chapter 12
of this book.
The distinction between stimulation for receptors and stimulus information
for the visual system is crucial for what is to follow. Receptors are passive,
elementary, anatomical components of an eye that, in turn, is only an organ of
the complete system (Gibson, 1966b, Ch. 2). The traditional conception of a
sense is almost wholly abandoned in this new approach. Stimulation by light
and corresponding sensations of brightness are traditionally supposed to be the
basis of visual perception. The inputs of the nerves are supposed to be the data
on which the perceptual processes in the brain operate. But I make a quite
different assumption, because the evidence suggests that stimuli as such contain
no information, that brightness sensations are not elements of perception, and
that inputs of the retina are not sensory elements on which the brain operates.
Visual perception can fail not only for lack of stimulation but also for lack of
stimulus information. In homogeneous ambient darkness, vision fails for lack of
stimulation. In homogeneous ambient light, vision fails for lack of information,
even with adequate stimulation and corresponding sensations.
48 The Ecological Approach to Visual Perception
Do we Ever See Light as Such?
The difference between stimulation and stimulus information can be shown in
another way, by considering two contradictory assertions: (1) nothing can be
seen, properly speaking, but light; and (2) light, properly speaking, can never
be seen. At least one of these assertions must be wrong.
Classical optics, comparing the eye to a camera, has taught that nothing can
possibly get into the eye but light in the form of rays or wave fronts. The only
alternative to this doctrine seemed to be the naive theory that little copies of
objects got into the eye. If all that can ever reach the retina is light in this form,
then it would follow that all we can ever see is this light. Sensations of light are
the fundamental basis of visual perception, the data, or what is given. This line
of reasoning has seemed unassailable up to the present. It leads to what I have
called the sensation-based theories of perception (Gibson, 1966b). We cannot
see surfaces or objects or the environment directly; we only see them indirectly.
All we ever see directly is what stimulates the eye, light. The verb to see, properly
used, means to have one or more sensations of light.
What about the opposite assertion that we never see light? It may at ﬁrst
sound unreasonable, or perhaps false, but let us examine the statement carefully.
Of all the possible things that can be seen, is light one of them?
A single point of light in an otherwise dark ﬁeld is not “light”; it speciﬁes
either a very distant source of light or a very small source, a luminous object. A
single instant or “ﬂash” of such a point speciﬁes a brief event at the source, that
is, the on and the off. A ﬁre with coals or ﬂames, a lamp with a wick or ﬁlament,
a sun or a moon—all these are quite speciﬁc objects and are so speciﬁed; no one
sees merely light. What about a luminous ﬁeld, such as the sky? To me it seems
that I see the sky, not the luminosity as such. What about a beam of light in the
air? But this is not seeing light, because the beam is only visible if there are illuminated
particles in the medium. The same is true of the shafts of sunlight seen
in clouds under certain conditions.
One can perceive a rainbow, to be sure, a spectrum, but even so that is not
the seeing of light. Halos, highlights on water, and scintillations of various
kinds are all manifestations of light, not light as such. The only way we see
illumination, I believe, is by way of that which is illuminated, the surface on
which the beam falls, the cloud, or the particles that are lighted. We do not see
the light that is in the air, or that ﬁlls the air. If all this is correct, it becomes
quite reasonable to assert that all we ever see is the environment or facts about
the environment, never photons or waves or radiant energy.
What about the sensation of being dazzled by looking at the sun, or the
sensation of glare that one gets from looking at glossy surfaces that reﬂect an
intense source? Are these not sensations of light as such, and do we not then see
pure physical energy? Even in this case, I would argue that the answer is no; we
are perceiving a state of the eye akin to pain, arising from excessive stimulation.
The Relationship Between Stimulation and Stimulus Information 49
We perceive a fact about the body as distinguished from a fact about the world,
the fact of overstimulation but not the light that caused it. And the experiencing
of facts about the body is not the basis of experiencing facts about the
world.
If light in the exact sense of the term is never seen as such, it follows that
seeing the environment cannot be based on seeing light as such. The stimulation
of the receptors in the retina cannot be seen, paradoxical as this may sound.
The supposed sensations resulting from this stimulation are not the data for
perception. Stimulation may be a necessary condition for seeing, but it is not
sufﬁcient. There has to be stimulus information available to the perceptual
system, not just stimulation of the receptors.
In ordinary speech we say that vision depends on light, and we do not need
to know physics to be able to say it with conﬁdence. All of us, including every
child, know what it is like to be “in the dark.” We cannot see anything, not
even our own bodies. Approaching dangers and collisions ahead cannot be
foreseen, and this is, with some reason, alarming. But what we mean when we
say that vision depends on light is that it depends on illumination and on sources
of illumination. We do not necessarily mean that we have to see light or have
sensations of light in order to see anything else.
Just as the stimulation of the receptors in the retina cannot be seen, so the
mechanical stimulation of the receptors in the skin cannot be felt, and the stimulation
of the hair cells in the inner ear cannot be heard. So also the chemical
stimulation of the receptors in the tongue cannot be tasted, and the stimulation
of the receptors in the nasal membrane cannot be smelled. We do not perceive
stimuli.
The Concept of the Stimulus as an Application of Energy
The explicit assumption that only the receptors of observers are stimulated and
that their sense organs are not stimulated but activated is in disagreement with
what most psychologists take for granted. They blithely use the verb stimulate
and the noun stimulus in various ways not consistent with one another. It is
convenient and easy to do so, but if the words are slippery and if we allow
ourselves to slide from one meaning to another unawares, we are confused
without knowing it. I once examined the writings of modern psychology and
found eight separate ways in which the use of the term stimulus was equivocal
(Gibson, 1960a).
The concept of the stimulus comes from physiology, where it ﬁrst meant
whatever application of energy ﬁres a nerve cell or touches off a receptor or
excites a reﬂex response. It was taken over by psychology, because it seemed
that a stimulus explained not only the arousal of a sensation but the arousal of a
response, including responses much more elaborate than reﬂexes. If all behavior
consisted of responses to stimuli, it looked as if a truly scientiﬁc psychology
50 The Ecological Approach to Visual Perception
could be founded. This was the stimulus-response formula. It was indeed
promising. Both stimuli and responses could be measured. But a great variety
of environmental facts had to be called stimuli because a variety of things can
be responded to. If anything in the world can be called a stimulus, the concept
has got out of hand and its original meaning has been lost. I suggest that we go
back to its meaning in physiology. In this book I shall use the term strictly. For
I now wish to make the clearest possible contrast between stimulus energy and
stimulus information.
Note that a stimulus, strictly speaking in the physiologist’s sense, is anything
that touches off a receptor or causes a response; it is the effective stimulus, and
whatever application of energy touches off the receptor is effective. The
photoreceptors in the eye are usually triggered by light but not necessarily; they
are also triggered by mechanical or electrical energy. The mechanoreceptors of
the skin and the chemoreceptors of the mouth and nose are more or less specialized
for mechanical and chemical energy respectively but not completely so;
they are just especially “sensitive” to those kinds of energy. A stimulus in this
strict meaning carries no information about its source in the world; that is, it
does not specify its source. Only stimulation that comes in a structured array
and that changes over time speciﬁes its external source.
Note also that a stimulus, strictly speaking, is temporary. There is nothing
lasting about it, as there is about a persisting object of the environment. A stimulus
must begin and end. If it persists, the response of the receptor tapers off and
ceases; the term for this is sensory adaptation. Hence, a permanent object cannot
possibly be speciﬁed by a stimulus. The stimulus information for an object
would have to reside in something persisting during an otherwise changing
ﬂow of stimulation. And note above all that an object cannot be a stimulus,
although current thinking carelessly takes for granted that it is one.
An application of stimulus energy exceeding the threshold can be said to
cause a response of the sensory mechanism, and the response is an effect. But the
presence of stimulus information cannot be said to cause perception. Perception
is not a response to a stimulus but an act of information pickup. Perception may
or may not occur in the presence of information. Perceptual awareness, unlike
sensory awareness, does not have any discoverable stimulus threshold. It depends
on the age of the perceiver, how well he has learned to perceive, and how
strongly he is motivated to perceive. If perceptions are based on sensations and
sensations have thresholds, then perceptions should have thresholds. But they
do not, and the reason for this, I believe, is that perceptions are not based on
sensations. There are magnitudes for applied stimuli above which sensations
occur and below which they do not. But there is no magnitude of information
above which perceiving occurs and below which it does not.
When stimulus energy is transformed into nervous impulses, they are said to
be transmitted to the brain. But stimulus information is not anything that could
possibly be sent up a nerve bundle and delivered to the brain, inasmuch as it has
The Relationship Between Stimulation and Stimulus Information 51
to be isolated and extracted from the ambient energy. Information as here
conceived is not transmitted or conveyed, does not consist of signals or messages,
and does not entail a sender and a receiver. This will be elaborated later.
When a small packet of stimulus energy is absorbed by a receptor, what is
lost to the environment is gained by the living cells. The amount of energy may
be as low as a few quanta, but nevertheless energy is conserved. In contrast to
this fact, stimulus information is not lost from the environment when it is
gained by the observer. There is no such thing as conservation of information.
It is not limited in amount. The available information in ambient light, vibration,
contact, and chemical action is inexhaustible.
A stimulus, then, carries some of the meaning that the word had in Latin, a
goad stuck into the skin of an ox. It is a brief and discrete application of energy
to a sensitive surface. As such, it speciﬁes little beyond itself; it contains no
information. But a ﬂowing array of stimulation is a different matter entirely.
Ambient Energy as Available Stimulation
The environment of an observer was said to consist of substances, the medium,
and surfaces. Gravity, heat, light, sound, and volatile substances ﬁll the medium.
Chemical and mechanical contacts and vibrations impinge on the observer’s
body. The observer is immersed as it were in a sea of physical energy. It is a
ﬂowing sea, for it changes and undergoes cycles of change, especially of temperature
and illumination. The observer, being an organism, exchanges energy
with the environment by respiration, food consumption, and behavior. A very
small fraction of this ambient sea of energy constitutes stimulation and provides
information. The fraction is small, for only the ambient odor entering the nose
is effective for smelling, only the train of air vibrations impinging on the
eardrums is effective for hearing, and only the ambient light at the entrance
pupil of an eye is effective for vision. But this tiny portion of the sea of energy
is crucial for survival, because it contains information for things at a distance.
It should be obvious by now that this minute inﬂow of stimulus energy does
not consist of discrete inputs—that stimulation does not consist of stimuli. The
ﬂow is continuous. There are, of course, episodes in the ﬂow, but these are
nested within one another and cannot be cut up into elementary units.
Stimulation is not momentary.
Radiant energy of all wavelengths falls on an individual, that is, impinges
on the skin. The infrared radiation will give warmth, and the ultraviolet
will cause sunburn, but the narrow band of radiation in between, light, is the
only kind that will excite the photoreceptors in the eye after entering the pupil.
An eye, or at least a vertebrate chambered eye as distinguished from the faceted
eye of an insect, usually takes in something less than a hemisphere of the
ambient light, according to G. L. Walls (1942). A pair of eyes like those of a
rabbit, pointing in opposite directions, takes in nearly the whole of the ambient
52 The Ecological Approach to Visual Perception
light at the same time. Ambient light is structured, as we have seen. And the
purpose of a dual ocular system is to register this structure or, more exactly, the
invariants of its changing structure. Ambient light is usually very rich in what
we call pattern and change. The retinal images register both. And a retinal
image involves stimulation of its receptive surface but not, as often supposed, a
set or a sequence of stimuli.
The Orthodox Theory of the Retinal Image
The generally accepted theory of the eye does not acknowledge that it registers
the invariant structure of ambient light but asserts that it forms an image of an
object on the back of the eye. The object, of course, is in the outer world, and
the back of the eye is a photoreceptive surface attached to a nerve bundle. What
is the difference between these theories?
The theory of image formation in a dark chamber like the eye goes back
more than 350 years to Johannes Kepler. The germ of the theory as stated by
him was that everything visible radiates, more particularly that every point on
a body can emit rays in all directions. An opaque reﬂecting surface, to be sure,
receives radiation from a source and then re-emits it, but in effect it becomes a
collection of radiating point sources. If an eye is present, a small cone of diverging
rays enters the pupil from each point source and is caused by the lens to
converge to another point on the retina. The diverging and converging rays
make what is called a focused pencil of rays. The dense set of focus points on the
retina constitutes the retinal image. There is a one-to-one projective correspondence
between radiating points and focus points.
A focused pencil of rays consists of two parts, the diverging cone of radiant
light and the converging cone of rays refracted by the lens, one cone with its
vertex on the object and the other with its vertex in the image. This pencil is
then repeated for every point on the object. Thus, there is a limitless set of rays
in each pencil and a limitless set of pencils for each object. The history of optics
suggests that Kepler was mainly responsible for this extraordinary intellectual
invention. It involved difﬁcult ideas, but it was and still is the unchallenged
foundation of the theory of image formation. The notion of an object composed
of points has proved over the centuries to be sympathetic to physicists, because
most of them assume that an object really consists of its atoms. And later, in the
nineteenth century, the notion of a retinal image consisting of sharp points of
focused light did not seem strange to physiologists because they were familiar
with punctate stimuli, for example, on the skin.
This theory of point-to-point correspondence between an object and its
image lends itself to mathematical analysis. It can be abstracted to the concepts
of projective geometry and can be applied with great success to the design of
cameras and projectors, that is, to the making of pictures with light, photography.
The theory permits lenses to be made with smaller “aberrations,” that
The Relationship Between Stimulation and Stimulus Information 53
is, with ﬁner points in the point-to-point correspondence. It works beautifully,
in short, for the images that fall on screens or surfaces and that are intended to
be looked at. But this success makes it tempting to believe that the image on the
retina falls on a kind of screen and is itself something intended to be looked at,
that is, a picture. It leads to one of the most seductive fallacies in the history of
psychology—that the retinal image is something to be seen. I call this the “little
man in the brain” theory of the retinal image (Gibson, 1966b, p. 226), which
conceives the eye as a camera at the end of a nerve cable that transmits the
image to the brain. Then there has to be a little man, a homunculus, seated in
the brain who looks at this physiological image. The little man would have to
have an eye to see it with, of course, a little eye with a little retinal image
connected to a little brain, and so we have explained nothing by this theory. We
are in fact worse off than before, since we are confronted with the paradox of
an inﬁnite series of little men, each within the other and each looking at the
brain of the next bigger man.
If the retinal image is not transmitted to the brain as a whole, the only
alternative has seemed to be that it is transmitted to the brain element by element,
that is, by signals in the ﬁbers of the optic nerve. There would then be an
element-to-element correspondence between image and brain analogous to the
FIGURE 4.3 A focused pencil of rays connecting a radiating point on a surface with
a focus point in the retinal image.
The rays in the pencil are supposed to be inﬁnitely dense. Note that only the rays
that enter the pupil are effective for vision. (From The Perception of the Visual World
by James Jerome Gibson and used with the agreement of the reprint publisher,
Greenwood Press, Inc.)
54 The Ecological Approach to Visual Perception
point-to-point correspondence between object and image. This seems to avoid
the fallacy of the little man in the brain who looks at an image, but it entails all
the difﬁculties of what I have called the sensation-based theories of perception.
The correspondence between the spots of light on the retina and the spots of
sensation in the brain can only be a correspondence of intensity to brightness and
of wavelength to color. If so, the brain is faced with the tremendous task of
constructing a phenomenal environment out of spots differing in brightness and
color. If these are what is seen directly, what is given for perception, if these are
the data of sense, then the fact of perception is almost miraculous.
JAMES MILL ON VISUAL SENSATION, 1829
“When I lift my eyes from the paper on which I am writing, I see from my
window trees and meadows, and horses and oxen, and distant hills. I see
each of its proper size, of its proper form, and at its proper distance; and
these particulars appear as immediate informations of the eye as the colors
which I see by means of it. Yet philosophy has ascertained that we derive
nothing from the eye whatever but sensations of color . . . . How then, is it
that we receive accurate information by the eye of size and shape and
distance? By association merely” (Mill, Analysis of the Phenomena of the
Human Mind, 1829).
How is it indeed! Mill answered, by association. But others answered, by
innate ideas of space or by rational inference from the sensations or by interpretation
of the data. Still others have said, by spontaneous organization of sensory
inputs to the brain. The current fashionable answer is, by computerlike activities
of the brain on neural signals. We have empiricism, nativism, rationalism,
Gestalt theory, and now information-processing theory. Their adherents
would go on debating forever if we did not make a fresh start. Has philosophy
ascertained that “we derive nothing from the eye whatever but sensations
of color”? No. “Sensations of color” meant dabs or spots of color, as if
in a painting. Perception does not begin that way.
Even the more sophisticated theory that the retinal image is transmitted as
signals in the ﬁbers of the optic nerve has the lurking implication of a little man
in the brain. For these signals must be in code and therefore have to be decoded;
signals are messages, and messages have to be interpreted. In both theories the
eye sends, the nerve transmits, and a mind or spirit receives. Both theories carry
the implication of a mind that is separate from a body.
It is not necessary to assume that anything whatever is transmitted along the
optic nerve in the activity of perception. We need not believe that either an
inverted picture or a set of messages is delivered to the brain. We can think of
The Relationship Between Stimulation and Stimulus Information 55
vision as a perceptual system, the brain being simply part of the system. The eye
is also part of the system, since retinal inputs lead to ocular adjustments and
then to altered retinal inputs, and so on. The process is circular, not a one-way
transmission. The eye-head-brain-body system registers the invariants in the
structure of ambient light. The eye is not a camera that forms and delivers an
image, nor is the retina simply a keyboard that can be struck by ﬁngers of light.
A Demonstration that the Retinal Image is not Necessary for Vision
We are apt to forget that an eye is not necessarily a dark chamber, on the back
surface of which an inverted image is formed by a lens in the manner described
by Kepler. Although the eyes of vertebrates and mollusks are of this sort, the
eyes of arthropods are not. They have what is called a compound eye, with no
chamber, no lens, and no sensory surface but with a closely packed set of
receptive tubes called ommatidia. Each tube points in a different direction from
every other tube, and presumably the organ can thus register differences of
intensity in different directions. It is therefore part of a system that registers the
structure of ambient light.
In a chapter on the evolutionary development of visual systems (Gibson,
1966b, Ch. 9), I described the chambered eye and the compound eye as two
different ways of accepting an array of light coming from an environment
(pp. 163 ff.). The camera eye has a concave mosaic of photoreceptors, a retina.
The compound eye has a convex packet of photoreceptive light tubes. The
former accepts an inﬁnite number of pencils of light, each focused to a point
and combining to make a continuous image. The latter accepts a ﬁnite number
of samples of ambient light, without focusing them and without forming an
optical image. But if several thousand tubes are packed together, as in the eye
of a dragonﬂy, visual perception is quite good. There is nothing behind a
dragonﬂy’s eye that could possibly be seen by you, no image on a surface, no
picture. But nevertheless the dragonﬂy sees its environment.
Zoologists who study insect vision are so respectful of optics as taught in
physics textbooks that they are constrained to think of a sort of upright image as
being formed in the insect eye. But this notion is both vague and self-contradictory.
There is no screen on which an image could be formed. The concept
of an ambient optic array, even if not recognized in optics, is a better foundation
for the understanding of vision in general than the concept of the retinal
image. The registering of differences of intensity in different directions is
necessary for visual perception; the formation of a retinal image is not.
The Concept of Optical Information
The concept of information with which we are most familiar is derived from
our experiences of communicating with other people and being communicated
56 The Ecological Approach to Visual Perception
with, not from our experience of perceiving the environment directly. We tend
to think of information primarily as being sent and received, and we assume
that some intermediate kind of transmission has to occur, a “medium” of
communication or a “channel” along which the information is said to ﬂow.
Information in this sense consists of messages, signs, and signals. In early times
messages, which could be oral, written, or pictorial, had to be sent by runner or
by horseman. Then the semaphore system was invented, and then the electrical
telegraph, wireless telegraphy, the telephone, television, and so on at an accelerated
rate of development.
THE FALLACY OF THE IMAGE IN THE EYE
Ever since someone peeled off the back of the excised eye of a slaughtered
ox and, holding it up in front of a scene, observed a tiny, colored, inverted
image of the scene on the transparent retina, we have been tempted to draw
a false conclusion. We think of the image as something to be seen, a picture
on a screen. You can see it if you take out the ox’s eye, so why shouldn’t the
ox see it? The fallacy ought to be evident.
The question of how we can see the world as upright when the retinal
image is inverted arises because of this false conclusion. All the experiments
on this famous question have come to nothing. The reginal image is not
anything that can be seen. The famous experiment of G. M. Stratton (1897)
on reinverting the retinal image gave unintelligible results because it was
misconceived.
We also communicate with others by making a picture on a surface (clay
tablet, papyrus, paper, wall, canvas, or screen) and by making a sculpture, a
model, or a solid image. In the history of image-making, the chief technological
revolution was brought about by the invention of photography, that is, of
a photosensitive surface that could be placed at the back of a darkened chamber
with a lens in front. This kind of communication, which we call graphic or
plastic, does not consist of signs or signals and is not so obviously a message
from one person to another. It is not so obviously transmitted or conveyed.
Pictures and sculptures are apt to be displayed, and thus they contain information
and make it available for anyone who looks. They nevertheless are, like the
spoken and written words of language, man-made. They provide information
that, like the information conveyed by words, is mediated by the perception of
the ﬁrst observer. They do not permit ﬁrsthand experience—only experience
at second hand.
The ambient stimulus information available in the sea of energy around us is
quite different. The information for perception is not transmitted, does not
The Relationship Between Stimulation and Stimulus Information 57
consist of signals, and does not entail a sender and a receiver. The environment
does not communicate with the observers who inhabit it. Why should the world
speak to us? The concept of stimuli as signals to be interpreted implies some
such nonsense as a world-soul trying to get through to us. The world is speciﬁed
in the structure of the light that reaches us, but it is entirely up to us to perceive
it. The secrets of nature are not to be understood by the breaking of its code.
Optical information, the information that can be extracted from a ﬂowing
optic array, is a concept with which we are not at all familiar. Being intellectually
lazy, we try to understand perception in the same way we understand
communication, in terms of the familiar. There is a vast literature nowadays of
speculation about the media of communication. Much of it is undisciplined and
vague. The concept of information most of us have comes from that literature.
But this is not the concept that will be adopted in this book. For we cannot
explain perception in terms of communication; it is quite the other way
around. We cannot convey information about the world to others unless we
have perceived the world. And the available information for our perception is
radically different from the information we convey.
Summary
Ecological optics is concerned with many-times-reﬂected light in the medium,
that is, illumination. Physical optics is concerned with electromagnetic energy,
that is, radiation.
Ambient light coming to a point in the air is profoundly different from
radiant light leaving a point source. The ambient light has structure, whereas
the radiant light does not. Hence, ambient light makes available information
about reﬂecting surfaces, whereas radiant light can at most transmit information
about the atoms from which it comes.
If the ambient light were unstructured or undifferentiated, it would provide
no information about an environment, although it would stimulate the photoreceptors
of an eye. Thus, there is a clear distinction between stimulus information
and stimulation. We do not have sensations of light triggered by stimuli
under normal conditions. The doctrine of discrete stimuli does not apply to
ordinary vision.
The orthodox theory of the formation of an image on a screen, based on the
correspondence between radiating points and focus points, is rejected as the
basis for an explanation of ecological vision. This theory applies to the design
of optical instruments and cameras, but it is a seductive fallacy to conceive the
ocular system in this way. One of the worst results of the fallacy is the inference
that the retinal image is transmitted to the brain.
The information that can be extracted from ambient light is not the kind of
information that is transmitted over a channel. There is no sender outside the
head and no receiver inside the head.
5
THE AMBIENT OPTIC ARRAY
The central concept of ecological optics is the ambient optic array at a point of
observation. To be an array means to have an arrangement, and to be ambient at
a point means to surround a position in the environment that could be occupied
by an observer. The position may or may not be occupied; for the present, let
us treat it as if it were not.
What is implied more speciﬁcally by an arrangement? So far I have suggested
only that it has structure, which is not very explicit. The absence of structure is
easier to describe. This would be a homogeneous ﬁeld with no differences of
intensity in different parts. An array cannot be homogeneous; it must be heterogeneous.
That is, it cannot be undifferentiated, it must be differentiated; it
cannot be empty, it must be ﬁlled; it cannot be formless, it must be formed.
These contrasting terms are still unsatisfactory, however. It is difﬁcult to deﬁne
the notion of structure. In the effort to clarify it, a radical proposal will be made
having to do with invariant structure.
What is implied by ambient at a point? The answer to this question is not so
difﬁcult. To be ambient, an array must surround the point completely. It must
be environing. The ﬁeld must be closed, in the geometrical sense of that term,
the sense in which the surface of a sphere returns upon itself. More precisely,
the ﬁeld is unbounded. Note that the ﬁeld provided by a picture on a plane
surface does not satisfy this criterion. No picture can be ambient, and even a
picture said to be panoramic is never a completely closed sphere. Note also that
the temporary ﬁeld of view of an observer does not satisfy the criterion, for it
also has boundaries. This fact is obviously of the greatest importance, and we
shall return to it in Chapter 7 and again in Chapter 12.
Finally, what is implied by the term point in the phrase point of observation?
Instead of a geometrical point in abstract space, I mean a position in ecological
The Ambient Optic Array 59
space, in a medium instead of in a void. It is a place where an observer might
be and from which an act of observation could be made. Whereas abstract
space consists of points, ecological space consists of places—locations or
positions.
A sharp distinction will be made between the ambient array at an unoccupied
point of observation and the array at a point that is occupied by an observer,
human or other. When the position becomes occupied, something very interesting
happens to the ambient array: it contains information about the body of
the observer. This modiﬁcation of the array will be given due consideration
later.
The point of observation in ecological optics might seem to be the equivalent
of the station point in perspective geometry, the kind of perspective
used in the making of a representative painting. The station point is the point
of projection for the picture plane on which the scene is projected. But the
terms are not at all equivalent and should not be confused, as we shall see. A
station point has to be stationary. It cannot move relative to the world, and it
must not move relative to the picture plane. But a point of observation is never
stationary, except as a limiting case. Observers move about in the environment,
and observation is typically made from a moving position.
How is Ambient Light Structured? Preliminary Considerations
If we reject the assumption that the environment consists of atoms in space and
that, hence, the light coming to a point in space consists of rays from these
atoms, what do we accept? It is tempting to assume that the environment
consists of objects in space and that, hence, the ambient array consists of closedcontour
forms in an otherwise empty ﬁeld, or “ﬁgures on a ground.” For each
object in space, there would correspond a form in the optic array. But this
assumption is not close to being good enough and must also be rejected. A form
in the array could not correspond to each object in space, because some objects
are hidden behind others. And in any case, to put it radically, the environment
does not consist of objects. The environment consists of the earth and the sky
with objects on the earth and in the sky, of mountains and clouds, ﬁres and
sunsets, pebbles and stars. Not all of these are segregated objects, and some of
them are nested within one another, and some move, and some are animate.
But the environment is all these various things—places, surfaces, layouts,
motions, events, animals, people, and artifacts that structure the light at points
of observation. The array at a point does not consist of forms in a ﬁeld. The
ﬁgure-ground phenomenon does not apply to the world in general. The notion
of a closed contour, an outline, comes from the art of drawing an object, and
the phenomenon comes from the experiment of presenting an observer with a
drawing to ﬁnd out what she perceives. But this is not the only way, or even the
best way, to investigate perception.
60 The Ecological Approach to Visual Perception
We obtain a better notion of the structure of ambient light when we think
of it as divided and subdivided into component parts. For the terrestrial environment,
the sky-earth contrast divides the unbounded spherical ﬁeld into two
hemispheres, the upper being brighter than the lower. Then both are further
subdivided, the lower much more elaborately than the upper and in quite a
different way. The components of the earth, as I suggested in Chapter 1, are
nested at different levels of size—for example, mountains, canyons, trees,
leaves, and cells. The components of the array from the earth also fall into a
hierarchy of subordinate levels of size, but the components of the array are quite
different, of course, from the components of the earth. The components of the
array are the visual angles from the mountains, canyons, trees, and leaves (actually,
what are called solid angles in geometry), and they are conventionally measured
in degrees, minutes, and seconds instead of kilometers, meters, and
millimeters. They are intercept angles, as we shall see. All these optical components
of the array, whatever their size, become vanishingly small at the margin
between earth and sky, the horizon; moreover, they change in size whenever
the point of observation moves. The substantial components of the earth, on
the other hand, do not change in size.
There are several advantages in conceiving the optic array in this way, as a
nested hierarchy of solid angles all having a common apex instead of as a set of
rays intersecting at a point. Every solid angle, no matter how small, has form in
the sense that its cross-section has a form, and a solid angle is quite unlike a ray
in this respect. Each solid angle is unique, whereas a ray is not unique and can
only be identiﬁed arbitrarily, by a pair of coordinates. Solid angles can ﬁll up a
sphere in the way that sectors ﬁll up a circle, but it must be remembered that
there are angles within angles, so that their sum does not add up to a sphere.
The surface of the sphere whose center is the common apex of all the solid
angles can be thought of as a kind of transparent ﬁlm or shell, but it should not
be thought of as a picture.
The structure of an optic array, so conceived, is without gaps. It does not
consist of points or spots that are discrete. It is completely ﬁlled. Every
component is found to consist of smaller components. Within the boundaries
of any form, however small, there are always other forms. This means that the
array is more like a hierarchy than like a matrix and that it should not
be analyzed into a set of spots of light, each with a locus and each with a
determinate intensity and frequency. In an ambient hierarchical structure,
loci are not deﬁned by pairs of coordinates, for the relation of location is not
given by degrees of azimuth and elevation (for example) but by the relation of
inclusion.
The difference between the relation of metric location and the relation of
inclusion can be illustrated by the following fact. The stars in the sky can be
located conveniently by degrees to the right of north and degrees up from
the horizon. But each star can also be located by its inclusion in one of the
FIGURE5.1Theambientopticarrayfromawrinkledearthoutdoorsunderthesky.
Inthisillustrationitisassumedthatilluminationhasreachedasteadystate.Theearthisshownaswrinkledorhumped,butnotascluttered.
Thedashedlinesinthisdrawingdepicttheenvelopesofvisualsolidangles,notraysoflight.Thenestingofthesesolidangleshasnotbeen
shown.Thecontrastsinthisdiagramarecausedbydifferentialilluminationofthehumpsoftheearth.Comparethiswiththephotograph
ofhillsandvalleysinFigure5.9.Thisisanopticarrayatasingleﬁxedpointofobservation.Itillustratesthemaininvariantsofnatural
perspective:theseparationofthetwohemispheresoftheambientarrayatthehorizon,andtheincreasingdensityoftheopticaltexture
towarditsmaximumatthehorizon.Theseareinvariantevenwhenthearrayﬂows,asitdoeswhenthepointofobservationmoves.
62 The Ecological Approach to Visual Perception
constellations and by the superordinate pattern of the whole sky. Similarly, the
optical structures that correspond to the leaves and trees and hills of the earth
are each included in the next larger structure. The texture of the earth, of
course, is dense compared to the constellations of discrete stars and thus even
less dependent than they are on a coordinate system. If this is so, the perception
of the direction of some particular item on the earth, its direction-from-here,
is not a problem in its own right. The perceiving of the environment does
not consist of perceptions of the differing directions of the items of the
environment.
The Laws of Natural Perspective: The Intercept Angle
The notion of a visual angle with its apex at the eye and its base at an object in
the world is very old. It goes back to Euclid who postulated what he called a
“visual cone” for each object in space. The term is not exact, for the object need
not be circular and the ﬁgure does not have to be a cone. Ptolemy spoke of the
“visual pyramid,” which implied that the object was rectangular. Actually, we
should refer to the face of an object, which can have any shape whatever, and to
a corresponding solid angle, having an envelope. A cross-section of this envelope
is what we call the outline of the object. We can now note that the solid angle
shrinks as the distance of the object from the apex increases, and it is laterally
squeezed as the face of the object is slanted or turned. These are the two main
laws of perspective for objects. Euclid and Ptolemy and their successors for
FIGURE 5.2 The ambient optic array from a room with a window.
This drawing shows a cluttered environment where some surfaces are projected at
the point of observation and the remainder are not, that is, where some are unhidden
and the others are hidden. The hidden surfaces are indicated by dotted lines. Only
the faces of the layout of surfaces are shown, not the facets of their surfaces, that is,
their textures.
The Ambient Optic Array 63
many centuries never doubted that objects were seen by means of these solid
angles, whether conical, pyramidal, or otherwise. They were the basis of
ancient optics. Nothing was then known of inverted retinal images, and the
comparison of the eye with a camera would not be made for a thousand years.
The ancients did not understand the eye, they were puzzled by light, they had
no conception of the modem doctrine that nothing gets into the eye but light,
but they were clear about visual angles.
The conception of the ambient optic array as a set of solid angles corresponding
to objects is thus a continuation of ancient and medieval optics.
Instead of only freestanding objects present to an eye, however, I postulate
an environment of illuminated surfaces. And instead of a group of solid
angles, I postulate a nested complex of them. The large solid angles in the
array come from the faces of this layout, from the facades of detached objects,
and from the interspaces or holes that we call background or sky (which Euclid
and Ptolemy seem never to have thought of). The small solid angles in the
array come from what might be called the facets of the layout as distinguished
from the faces, the textures of the surfaces as distinguished from their forms.
As already has been emphasized, however, the distinction between these
size-levels is arbitrary.
Natural perspective, as I conceive it, is the study of an ambient array of solid
angles that correspond to certain distinct geometrical parts of a terrestrial
environment, those that are separated by edges and corners. There are elegant
trigonometric relations between the angles and the environmental parts. There
are gradients of size and density of the angles along meridians of the lower half
of the array, the earth, with sizes vanishing and density becoming inﬁnite at the
horizon. These relations contain a great amount of information about the parts
of the earth. No one who understood them would think of questioning their
validity. It is a perfectly clear and straightforward discipline, although neglected
and undeveloped. But the environment does not wholly consist of sharply differentiated
geometrical parts or forms. Natural perspective does not apply to
shadows with penumbras and patches of light. It does not apply to sunlit surfaces
with varying degrees of illumination. It geometrizes the environment and thus
oversimpliﬁes it. The most serious limitation, however, is that natural
perspective omits motion from consideration. The ambient optic array is treated
as if its structure were frozen in time and as if the point of observation were
motionless.
Although I have called this discipline natural perspective, the ancients called
it perspectiva, the Latin word for what we now call optics. In modem times,
the term perspective has come to mean a technique—the technique of picturemaking.
A picture is a surface, whether it be painted by hand or processed
by photography, and perspective is the art of “representing” the geometrical
relationships of natural objects on that surface. When the Renaissance
painters discovered the procedures for perspective representation, they very
64 The Ecological Approach to Visual Perception
FIGURE 5.3 The same ambient array with the point of observation occupied by a
person.
When an observer is present at a point of observation, the visual system begins to
function.
properly called the method artiﬁcial perspective. They understood that this had
to be distinguished from the natural perspective that governed the ordinary
perception of the environment. Since that time we have become so pictureminded,
so dominated by pictorial thinking, that we have ceased to make
the distinction. But to confuse pictorial perspective with natural perspective
is to misconceive the problem of visual perception at the outset. The socalled
cues for depth in a picture are not at all the same as the information
for surface layout in a frozen ambient array, although pictorial thinking
about perception tempts us to assume that they are the same. Pictures are artiﬁcial
displays of information frozen in time, and this fact will be evident when
the special kind of visual perception that is mediated by such displays is treated
in detail in Part IV.
Natural perspective, as well as artiﬁcial perspective, is restricted in scope,
being concerned only with a frozen optical structure. This restriction will be
removed in what follows.
The Ambient Optic Array 65
Optical Structure with a Moving Point of Observation
A point of observation at rest is only the limiting case of a point of observation
in motion, the null case. Observation implies movement, that is, locomotion
with reference to the rigid environment, because all observers are animals and
all animals are mobile. Plants do not observe but animals do, and plants do not
move about but animals do. Hence, the structure of an optic array at a stationary
point of observation is only a special case of the structure of an optic array at a
moving point of observation. The point of observation normally proceeds
along a path of locomotion, and the “forms” of the array change as locomotion
proceeds. More particularly, every solid angle included within the array, large
or small, is enlarged or reduced or compressed or, in some cases, wiped out. It
is wiped out, of course, when its surface goes out of sight.
The optic array changes, of course, as the point of observation moves. But it
also does not change, not completely. Some features of the array do not persist
FIGURE 5.4 The change of the optic array brought about by a locomotor movement
of the observer.
The thin solid lines indicate the ambient optic array for the seated observer, and the
thin dashed lines the altered optic array after standing up and moving forward. The
difference between the two arrays is speciﬁc to the difference between the points
of observation, that is, to the path of locomotion. Note that the whole ambient
array is changed, including the portion behind the head. And note that what was
previously hidden becomes unhidden.
66 The Ecological Approach to Visual Perception
and some do. The changes come from the locomotion, and the nonchanges
come from the rigid layout of the environmental surfaces. Hence, the nonchanges
specify the layout and count as information about it; the changes specify locomotion
and count as another kind of information, about the locomotion itself.
We have to distinguish between two kinds of structure in a normal ambient
array, and I shall call them the perspective structure and the invariant structure.
Perspective Structure and Invariant Structure
The term structure is vague, as we have seen. Let us suppose that a kind of essential
structure underlies the superﬁcial structure of an array when the point of
observation moves. This essential structure consists of what is invariant despite
the change. What is invariant does not emerge unequivocally except with a
ﬂux. The essentials become evident in the context of changing nonessentials.
Consider the paradox in the following piece of folk wisdom: “The more it
changes, the more it is the same thing.” Wherein is it true and wherein false? If
change means to become different but not to be converted into something else, the assertion
is true, and the saying emphasizes the fact that whatever is invariant is more
evident with change than it would be without change. If change means to become
different by being converted into something else, the assertion is self-contradictory, and
the paradox arises. But this is not what the word ordinarily means. And assuredly
it is not what change in the ambient array means. One arrangement does not
become a wholly different arrangement by a displacement of viewpoint. There is
no jump from one to another, only a variation of structure that serves to reveal
the nonvariation of structure. The pattern of the array does not ordinarily scintillate;
the forms of the array do not go from triangular to quadrangular, for example.
There are many invariants of structure, and some of them persist for long paths
of locomotion while some persist only for short paths. But what I am calling the
perspective structure changes with every displacement of the point of observation—
the shorter the displacement the smaller the change, and the longer the displacement
the greater the change. Assuming that the environment is never reduplicated
from place to place, the arrested perspective is unique at each stationary point of
observation, that is, for each point of observation there is one and only one
arrested perspective. On the other hand, invariants of structure are common to all
points of observation—some for all points in the whole terrestrial environment,
some only for points within the boundaries of certain locales, and some only for
points of observation within (say) a single room. But to repeat, the invariant
structure separates off best when the frozen perspective structure begins to ﬂow.
Consider, for example, the age-old question of how a rectangular surface
like a tabletop can be given to sight when presumably all that an eye can see is
a large number of forms that are trapezoids and only one form that is rectangular,
that one being seen only when the eye is positioned on a line perpendicular
to the center of the surface. The question has never been answered, but it
The Ambient Optic Array 67
can be reformulated to ask, What are the invariants underlying the transforming
perspectives in the array from the tabletop? What speciﬁes the shape
of this rigid surface as projected to a moving point of observation? Although
the changing angles and proportions of the set of trapezoidal projections are a
fact, the unchanging relations among the four angles and the invariant proportions
over the set are another fact, equally important, and they uniquely specify
the rectangular surface. There will be experimental evidence about optical
transformations as information in Chapter 9.
We tend to think of each member of the set of trapezoidal projections from
a rectangular object as being a form in space. A change is then a transition from
one form to another, a transformation. But this habit of thought is misleading.
Optical change is not a transition from one form to another but a reversible
process. The superﬁcial form becomes different, but the underlying form
remains the same. The structure changes in some respects and does not change
in others. More exactly, it is variant in some respects and invariant in others.
The geometrical habit of separating space from time and imagining sets
of frozen forms in space is very strong. One can think of each point of observation
in the medium as stationary and distinct. To each such point there
would correspond a unique optic array. The set of all points is the space of the
medium, and the corresponding set of all optic arrays is the whole of the available
information about layout. The set of all line segments in the space speciﬁes
all the possible displacements of points of observation in the medium, and the
corresponding set of transformation families gives the information that speciﬁes
all the possible paths. This is an elegant and abstract way of thinking, modeled
on projective geometry. But it does not allow for the complexities of optical
change and does not do justice to the fact that the optic array ﬂows in time instead
of going from one structure to another. What we need for the formulation of
ecological optics are not the traditional notions of space and time but the
concepts of variance and invariance considered as reciprocal to one another.
The notion of a set of stationary points of observation in the medium is appropriate
for the problem of a whole crowd of observers standing in different positions,
each of them perceiving the environment from his own point of view. But
even so, the fact that all observers can perceive the same environment depends
on the fact that each point of view can move to any other point of view.
REDUPLICATION
It is easy to make copies or duplicates of a picture but the world is never
exactly the same in one place as it is in another. Nor is one organism ever
exactly the same as another. One cubic yard of empty abstract space is
exactly the same as another, but that is a different matter.
68 The Ecological Approach to Visual Perception
The Signiﬁcance of Changing Perspective in the Ambient Array
When the moving point of observation is understood as the general case, the
stationary point of observation is more intelligible. It no longer is conceived as
a single geometrical point in space but as a pause in locomotion, as a temporarily
ﬁxed position relative to the environment. Accordingly, an arrested
perspective structure in the ambient array speciﬁes to an observer such a ﬁxed
position, that is, rest; and a ﬂowing perspective structure speciﬁes an unﬁxed
position, that is, locomotion. The optical information for distinguishing locomotion
from nonlocomotion is available, and this is extremely valuable for all
observers, human or animal. In physics the motion of an observer in space is
“relative,” inasmuch, as what we call motion with reference to one chosen
frame of reference may be nonmotion with reference to another frame of reference.
In ecology this does not hold, and the locomotion of an observer in the
environment is absolute. The environment is simply that with respect to which
either locomotion or a state of rest occurs, and the problem of relativity does
not arise.
Locomotion and rest go with ﬂowing and frozen perspective structure in the
ambient array; they are what the ﬂow and the nonﬂow mean. They contain
information about the potential observer, not information about the environment,
as the invariants do. But note that information about a world that surrounds
a point of observation implies information about the point of observation that is
surrounded by a world. Each kind of information implies the other. Later, in
discussing the occupied point of observation, I shall call the former exterospeciﬁc
information and the latter propriospeciﬁc information.
Not only does ﬂowing perspective structure specify locomotion, but the
particular instance of ﬂow speciﬁes the particular path of locomotion. That is,
the difference of perspective between the beginning and the end of the optical
change is speciﬁc to the difference of position between the beginning and the
end of the locomotor displacement. But more than that, the course of the optical
ﬂow is speciﬁc to the route the path of locomotion takes through the environment.
Between one place and another there are many different routes. The two
places are speciﬁed by their different arrested perspectives, but the different
routes between them are in correspondence with different optical sequences
between the two perspectives. There will be more of this later. It is enough
now to point out that the visual control of locomotion by an observer, purposive
locomotion such as homing, migrating, ﬁnding one’s way, getting from place
to place, and being oriented, depends on just the kind of sequential optical
information described.
It is important to realize that the ﬂowing perspective structure and the
underlying invariant structure are concurrent. They exist at the same time.
Although they specify different things, locomotion through a rigid world in
The Ambient Optic Array 69
the ﬁrst instance and the layout of that rigid world in the second instance, they
are like the two sides of a coin, for each implies the other. This hypothesis, that
optical change can seemingly specify two things at the same time, sounds very
strange, as if one cause were having two effects or as if one stimulus were
arousing two sensations. But there is nothing illogical about the idea of concurrent
speciﬁcation of two reciprocal things. Such an idea is much needed in
psychology.
The Change between Hidden and Unhidden Surfaces:
Covering Edges
We are now prepared to face a fact that has seemed deeply puzzling, a fact
that poses the greatest difﬁculty for all theories of visual perception based on
sensations. The layout of the environment includes unprojected (hidden)
surfaces at a point of observation as well as projected surfaces, but observers
perceive the layout, not just the projected surfaces. Things are seen in the
round and one thing is seen in front of another. How can this be? Information
must be available for the whole layout, not just for its facades, for the covered
surfaces as well as the covering surfaces. What is this information? Presumably
it becomes evident over time, with changes of the array. I will argue that
the information is implicit in the edges that separate the surfaces or, rather, in
the optical speciﬁcation of these edges. I am suggesting that if covering
edges are speciﬁed, both the covered and the covering surfaces are also
speciﬁed.
To suggest that an observer can see surfaces that are unseen is, of course, a
paradox. I do not mean that. I am not saying that one can see the unseen, and
I am suspicious of visionaries who claim that they can. A vast amount of mystiﬁcation
in the history of human thought has arisen from this paradox. The
suggestion is that one can perceive surfaces that are temporarily out of sight, and
what it is to be out of sight will be carefully deﬁned. The important fact is that
they come into sight and go out of sight as the observer moves, ﬁrst in one
direction and then in the opposite direction. If locomotion is reversible, as it is,
whatever goes out of sight as the observer travels comes into sight as the observer
returns and conversely. The generality of this principle has never been realized;
it applies to the shortest locomotions, in centimeters, as well as to the longest,
in kilometers. But it has not been elaborated. I will call it the principle of reversible
occlusion. The theory of the cues for depth perception includes one cue called
“movement parallax” and another called “superposition,” both related to the
above principle, but these terms are vague and do not even begin to explain
what needs to be explained. What we see is not depth as such but one thing
behind another. The new principle can be made explicit. I will attempt to do so,
at some length.
70 The Ecological Approach to Visual Perception
Projected and Unprojected Surfaces
There are many commonsense words that refer to the fact of covered and
uncovered things. Objects and surfaces are said to be hidden or unhidden,
screened or unscreened, concealed or revealed, undisclosed or disclosed. We
might borrow a technical word in astronomy, occultation, but it means primarily
the shutting off of the light from a celestial source, as in an eclipse. We need a
word for the cutting off of a visual solid angle, not of light rays. I have chosen
the word occlusion for it. An occluded surface is one that is out of sight or hidden
from view. An occluding edge is the edge of an occluding surface. The term
was ﬁrst introduced in a paper by J. J. Gibson, G. A. Kaplan, H. N. Reynolds,
and K. Wheeler (1969) on the various ways in which a thing can pass between
the state of being visible and the state of being invisible. The experiment will
be described in Chapter 11.
Occlusion arises because of two facts about the environment, both described
in Chapter 2. First, surfaces are generally opaque; and second, the basic environment,
the earth, is generally cluttered. As to the ﬁrst, if surfaces were as transFIGURE
5.5 Objects seen in the round and behind other objects.
Do you perceive covered surfaces as well as covering surfaces in this photograph?
(Photo by Jim Scherer.)
The Ambient Optic Array 71
parent as air, they would not reﬂect light at all and there would be no use for
vision. Most substances are nontransmitting (they reﬂect and absorb instead), and
therefore light is reﬂected back from the surface. A few substances are partially
transmitting or “translucent,” and hence a sheet of such a substance will transmit
part of the radiant light but will not transmit the structure of the ambient array;
it will let through photons but not visual solid angles. There can be an obstructing
of the view without obstructing of the light, although an obstructing of the light
will of course also obstruct the view. If we add the fact that surfaces are also
generally textured, the facts of opaque surfaces as contrasted with the surfaces of
semitransparent and translucent substances become intelligible.
The second fact is that the environment is generally cluttered. What I called
an open environment is seldom or never realized, although it is the only case in
which all surfaces are projected and none are unprojected. An open environment
has what we call an unobstructed view. But the ﬂat and level earth
receding unbroken to a pure linear horizon in a great circle, with a cloudless
sky, would be a desolate environment indeed. Perhaps it would not be quite as
lifeless as geometrical space, but almost. The furniture of the earth, like the
furnishings of a room, is what makes it livable. The earth as such affords only
standing and walking; the furniture of the earth affords all the rest of behavior.
The main items of the clutter (following the terminology adopted in Chapter 3)
are objects, both attached and detached, enclosures, convexities such as hills, concavities
such as holes, and apertures such as windows. These features of surface
layout give rise to occluding surfaces or, more exactly, to the separation of
occluding and occluded surfaces.
A surface is projected at a point of observation if it has a visual solid angle in the
ambient optic array; it is unprojected if it does not. A projected surface may become
unprotected in at least three ways—if its solid angle is diminished to a point, if the
solid angle is compressed to a line, or if the solid angle is wiped out. In the ﬁrst case
we say that the surface is too far away, in the second that it is turned so as not to
face the point of observation, in the third that the view is obstructed. The second
case, that of facing toward or away, is instructive. A wall or a sheet of paper has two
“faces” but only one can face a ﬁxed point. The relation between the occluding and
occluded surfaces is given by the relation of each to the point; the relation is not
merely geometrical but also optical. The relation is designated when we distinguish
between the near side and the far side of an object. (It is not, however, well
expressed by the terms front and back, since they are ambiguous. They can refer to
such surfaces as the front and the back of a house or the front and the back of a
head. Terms can be borrowed from ordinary language only with discretion!)
Going Out of and Coming Into Sight
A point of observation is to be thought of as moving through the medium to
and fro, back and forth, often along old paths but sometimes along new ones.
72 The Ecological Approach to Visual Perception
Displacements of this position are reversible and are reversed as its occupier
comes and goes, even as she slightly shifts her posture. Any face or facet, any
surface of the layout, that is progressively hidden during a displacement is
progressively unhidden during its reversal. Going out of sight is the inverse of
coming into sight. Hence, occluding and occluded surfaces interchange. The
occluding ones change into the occluded ones and vice versa, not by changing
from one entity to another but by a special transition.
The terms disappearance and its opposite, appearance, should not be used for this
transition. They have slippery meanings, like visible and invisible. For a surface
may disappear by going out of existence as well as by going out of sight, and the
two cases are profoundly different. A surface that disappears because it is no
longer projected to any point of observation, because it has evaporated, for
example, should not be confused with a surface that disappears because it is no
longer projected to a ﬁxed point of observation. The latter can be seen from
another position; the former cannot be seen from any position. Failure to distinguish
these meanings of disappear is common; it encourages careless observation
and vague beliefs in ghosts, or in the reality of the “unseen.” To disappear can also
refer to a surface that continues to exist but is no longer projected to any point of
observation because of darkness. Or we might speak of something disappearing
“in the distance,” referring to a surface barely projected to a point of observation
because its visual solid angle has diminished to a limit. These modes of so-called
disappearance are quite radically different. The differences between (1) a surface
that ceases to exist, (2) a surface that is no longer illuminated, (3) a surface that
lies on the horizon, and (4) a surface that is occluded are described in a paper by
Gibson, Kaplan, Reynolds, and Wheeler (1969) and are illustrated in a motion
picture ﬁlm (Gibson, 1968a). An experimental study of the perception of occlusion
using motion picture displays has been reported by Kaplan (1969).
The Loci of Occlusion: Occluding Edges
We must now distinguish an edge that is simply the junction of two surfaces
from an edge that causes one surface to hide another, an occluding edge. In the
proposed terminology of layout in Chapter 3, I deﬁned an edge as the apex of a
convex dihedral (as distinguished from a corner, which is the apex of a concave
dihedral). But an occluding edge is a dihedral where only one of the surfaces is
projected to the point of observation—an apical occluding edge. I also deﬁned a
curved convexity (as distinguished from a curved concavity), and another kind of
occluding edge is the brow of this convexity, that is, the line of tangency of the
envelope of the visual solid angle—a curved occluding edge. The apical occluding
edge is “sharp,” and the curved occluding edge is “rounded.” The two are illustrated
in Figure 5.6. The latter slides along the surface as the point of observation
moves, but the former does not. Note that an occluding edge always
requires a convexity of some sort, a protrusion of the substance into the medium.
The Ambient Optic Array 73
These two kinds of occluding edges are found in the ells of corridors, the
brinks of cliffs, the brows of hills, and the near sides of holes in the ground.
One face or facet or part of the layout hides another to which it may be
connected and which it may adjoin. This is different from what I called a
detached object, by which I mean the movable or moving object having a topologically
closed surface with substance inside and medium outside. The
detached object produces a visual solid angle in the optic array, as noted by
Euclid and Ptolemy, and yields a closed-contour ﬁgure in the visual ﬁeld, as
described by Edgar Rubin and celebrated by the gestalt psychologists under the
name of the “ﬁgure-ground phenomenon.” Occluding edges are a special case,
because not only does the near side of the object hide the far side but the object
covers a sector of the surface behind it, the ground, for example. The occluding
edges may be apical, as when the object is a polyhedron, or the locus of the
tangent of the envelope of the solid angle to the surface, as when the object is
curved. These are illustrated in Figure 5.7, where both the hiding of the far side
and the covering of the background are shown. The object is itself rounded or
FIGURE 5.6 The sharp occluding edge and the rounded occluding edge at a ﬁxed
point of observation.
The hidden portions of the surface layout are indicated by dotted lines.
FIGURE 5.7 Both the far side of an object and the background of the object are
hidden by its occluding edges.
Two detached objects are shown, one with sharp occluding edges and the other
with rounded occluding edges.
74 The Ecological Approach to Visual Perception
solid, and it is superposed on the ground, which is also continuous behind the
object. These two kinds of occlusion may be treated separately.
Self-occlusion and Superposition
An object, in the present terminology, is both voluminous and superposed. It
exists in volume and it may lie in front of another surface, or another object. In
short, an object always occludes itself and generally also occludes something
else. The effect of a moving point of observation is different in the two cases.
Projected and unprojected surfaces interchange as the point of observation
moves, but the interchange between parts of the object is not like that between
parts of the background. There is an interchange between opposite faces of the
object but an interchange of adjacent areas of the surface behind the object. For
the object, the near side turns into the far side and vice versa, whereas for the
background an uncovered area becomes covered and vice versa. The change of
optical structure in the former case is by way of perspective transformation,
whereas the disturbance of optical structure in the latter case is more radical, a
“kinetic disruption” being involved.
In Figure 5.7, as the point of observation moves each face of the facade of the
polyhedron undergoes transformation, for example, from trapezoid to square to
trapezoid. Ultimately, when the face is maximally foreshortened, it is what we
call “edge on,” that is, it becomes an occluding edge. The near face turns into
a far face by way of the edge. While this is happening at one edge, the other
edge is revealing a previously hidden face. A far face turns into a near face. The
two occluding edges in the diagram are perfectly reciprocal; while one is
converting near into far, the other is converting far into near. The width of the
polyhedron goes into depth, and the depth comes back into width. Width and
depth are thus interchangeable.
Similarly, one could describe the transformation of each facet of the textured
surface of the curved object. If the object is a sphere, the circular occluding
edge (the outline, in pictorial terminology) does not transform, but the optical
structure within it does. At one edge the texture is progressively turning from
projected into unprojected, from near into far, while at the other edge the
texture is progressively turning from unprojected into projected, from far into
near. The transition occurs at the limit of the slant transformation, the ultimate
of perspective foreshortening, but actually the optical texture reaches and goes
beyond this purported limit. It has to go beyond it because it comes from
beyond that limit at the other occluding edge.
Superposition
Now consider the separated background behind the objects in Figure 5.7, the
fact of superposition as distinguished from the fact of solidity. As the point of
The Ambient Optic Array 75
observation moves, the envelope of the visual solid angle sweeps across the
surface. The leading edge progressively covers the texture of the surface, while
the trailing edge progressively uncovers it. I have suggested metaphorically that
the texture is “wiped out” and “unwiped” at the lateral borders of the ﬁgure
(Gibson, 1966b, pp. 199 ff.). This was inspired by the metaphors used by A.
Michotte in describing experiments on what he called the “tunnel effect”
(Michotte, Thinès, and Crabbé, 1964). A somewhat more exact description of
this optical change will be given below. But note that if the texture that is
progressively covered has the same structure as the texture that is progressively
uncovered the unity of the surface is well speciﬁed.
The metaphor of “wiping” is inexact. A better description of the optical
transition was given by Gibson, Kaplan, Reynolds, and Wheeler (1969), and it
was also described by Kaplan (1969) as a “kinetic disruption.” There is a
disturbance of the structure of the array that is not a transformation, not even
a transformation that passes through its vanishing limit, but a breaking of its
adjacent order. More exactly, there is either a progressive decrementing of
components of structure, called deletion, or its opposite, a progressive incrementing
of components of structure, called accretion. An edge that is covering
the background deletes from the array; an edge that is uncovering the background
accretes to it. There is no such disruption for the surface that is covering
or uncovering, only for the surface that is being covered or uncovered. And
nondisruption, I suggest, is a kind of invariance.
The Information to Specify the Continuation of Surfaces
A surface always “bends under” an occluding edge, and another surface generally
“extends behind” it. These surfaces are connected or continuous. Is there information
in a changing optic array to specify the connnectedness or continuity?
Here is a tentative hypothesis for the continuous object surface:
Whenever a perspective transformation of form or texture in the optic
array goes to its limit and when a series of forms or textures are progressively
foreshortened to this limit, a continuation of the surface of an object
is speciﬁed at an occluding edge. This is the formula for going out of
sight; the formula is reversed for coming into sight.
Here is a tentative hypothesis for the continuous background surface:
Whenever there occurs a regular disturbance of the persistence of forms
and textures in the optic array such that they are progressively deleted at
a contour, the continuation of the surface of a ground is speciﬁed at an
occluding edge. This is for going out of sight; substituting accretion for
deletion gives the formula for coming into sight.
76 The Ecological Approach to Visual Perception
These two hypotheses make no assertions about perception, only about the
information that is normally available for perception. They do not refer to space,
or to the third dimension, or to depth, or to distance. Nothing is said about forms
or patterns in two dimensions. But they suggest a radically new basis for
explaining the perception of solid superposed objects, a new theory based not
on cues or clues or signs but on the direct pickup of solidity and superposition.
An object is in fact voluminous; a background is in fact continuous. A picture
or an image of an object is irrelevant to the question of how it is perceived.
The assumption for centuries has been that the sensory basis for the perception
of an object is the outline form of its image on the retina. Object perception
can only be based on form perception. First the silhouette is detected and
then the depth is added, presumably because of past experiences with the cues
for depth. But the fact is that the progressive foreshortening of the face of an
object is perceived as the turning of the object, which is precisely what the
transformation speciﬁes, and is never perceived as a change of form, which
ought to be seen if the traditional assumption is correct—that the silhouette is
detected and then the depth is added.
The two hypotheses stated above depend on a changing optic aray, and so far
the only cause of such change that has been considered is the moving point of
observation. The reader will have noted that a moving object will also bring
about the same kinds of disturbance in the structure of the array that have been
described above. A moving object in the world is an event, however, not a form
of locomotion, and the information for the perception of events will be treated
in Chapter 6.
The Case of Very Distant Surfaces
It is interesting to compare the occluding edges of objects and other convexities
on the surface of the earth with the horizon of the earth, the great circle dividing
the ambient array into two hemispheres. It is the limit of perspective miniﬁcation
for terrestrial surfaces, just as the edge-on line is the limit of perspective
compression (foreshortening) for a terrestrial surface. Objects such as railroad
trains on the Great Plains and ships on the ocean are said to vanish in the distance
as they move away from a ﬁxed point of observation. The line of the horizon in
the technology of pictorial perspective is said to be the locus of vanishing points
for the size of earth-forms and for the convergence of parallel edges on the earth.
The railroad train “vanishes” at the same optical point where the railroad tracks
“meet” in the distance. The horizon is therefore analogous to an occluding edge
in being one of the loci at which things go out of and come into sight. But going
out of sight in the distance is very different from going out of sight at a sharp or
a rounded edge nearby. The horizon of the earth, therefore, is not an occluding
edge for any terrestrial object or earth-form. It does not in fact look like an
occluding edge. It could only be visualized as an occluding edge for the lands
The Ambient Optic Array 77
FIGURE 5.8 Cartoon. (Drawing by S. Harris; © 1975 The New Yorker Magazine,
Inc.)
and seas beyond the horizon if the seemingly ﬂat earth were conceived as curved
and if the environment were thought of as a globe too vast to see.
It has long been a puzzle to human observers, however, that the horizon is in
fact visibly an occluding edge for celestial objects such as the sun and the moon.
Such objects undergo progressive deletion at a contour, as at sunset, and undergo
progressive accretion at the same contour, as at moonrise. This is in accordance
with the second hypothesis above. The object is obviously beyond the horizon,
more distant than the visible limit of earthly distance, and yet there is some
information for its being a solid surface. This conﬂicting information explains,
I think, the apparently enormous size of the sun and the moon at the horizon.
It also explains many of the ideas of pre-Copernican astronomy about heavenly
bodies. We should realize that the terrestrial environment was the only environment
that people could be sure of before Copernicus—the only environment
that could be perceived directly. Terrestrial objects and surfaces had affordances
for behavior, but celestial objects did not. More will be said about the perception
of objects on earth as distinguished from objects in the sky in Part III.
Summary: The Optics of Occlusion
1. In the ideal case of a terrestrial earth without clutter, all parts of the
surface are projected to all points of observation. But such an open environment
would hardly afford life.
2. In the case of an earth with furniture, with a layout of opaque surfaces
on a substratum, some parts of the layout are projected to any given ﬁxed point
of observation and the remaining parts are unprojected to that point.
78 The Ecological Approach to Visual Perception
3. The optically uncovered surface of an object is always separated from the
optically covered surface at the occluding edge. At the same time, it is always
connected with the optically covered surface at the occluding edge.
4. The continuation of the far side with the near side is speciﬁed by the
reversibility of occlusion.
5. Any surface of the layout that is hidden at a given ﬁxed point of observation
will be unhidden at some other ﬁxed point.
6. Hidden and unhidden surfaces interchange. Whatever is revealed by a
given movement is concealed by the reverse of that movement. This principle
of reversible occlusion holds true for both movements of the point of observation
and motions of detached objects.
7. We can now observe that the separation between hidden and unhidden
surfaces at occluding edges is best speciﬁed by the perspective structure of an array,
whereas the connection between hidden and unhidden surfaces at edges is speciﬁed
by the underlying invariant structure. Hence, probably, a pause in locomotion
calls attention to the difference between the hidden and the unhidden, whereas
locomotion makes evident the continuousness between the hidden and the
unhidden.
The seeming paradox of the perceiving or apprehending of hidden surfaces
will be treated further in Chapter 11.
How is Ambient Light Structured? A Theory
Let us return to the question of how ambient light is given its invariant structure,
the question asked at the beginning of this chapter but not answered
except in a preliminary way. Ambient light can only be structured by something
that surrounds the point of observation, that is, by an environment. It is
not structured by an empty medium of air or by a fog-ﬁlled medium. There
have to be surfaces—both those that emit light and those that reﬂect light.
Only because ambient light is structured by the substantial environment can it
contain information about it.
So far it has been emphasized that ambient light is made to constitute an
array by a single feature of these surfaces, their layout. But just how does the
layout structure the light? The answer is not simple. It involves the puzzling
complexities of light and shade. Moreover, the layout of surfaces is not the only
cause of the structuring of light; the conglomeration of surfaces makes a contribution,
that is, the fact that the environment is multicolored. The different
surfaces of the layout are made of different substances with different reﬂectances.
Both lighted or shaded surfaces and black or white surfaces make their
separate contributions to the invariant structure of ambient light. And how
light-or-shade can be perceived separately from black-or-white has long been a
puzzling problem for any theory of visual sense perception.
The Ambient Optic Array 79
I tried to formulate a theory of the structuring of ambient light in my last
book (Gibson, 1966b), asserting that three causes existed, the layout of surfaces,
the pigmentation of surfaces, and the shadowing of surfaces (pp. 208–216). But
the third of these causes is not cognate with the other two, and the interaction
between them was not clearly explained. The theory was static. Here, I shall
formulate a theory of the sources of invariant optical structure in relation to the
sources of variation in optical structure. What is clear to me now that was not
clear before is that structure as such, frozen structure, is a myth, or at least a
limiting case. Invariants of structure do not exist except in relation to variants.
The Sources of Invariant Optical Structure
The main invariants of the terrestrial environment, its persisting features, are the
layout of its surfaces and the reﬂectances of these surfaces. The layout tends to
persist because most of the substances are sufﬁciently solid that their surfaces are
rigid and resist deformation. The reﬂectances tend to persist because most of the
substances are chemically inert at their interfaces with the air, and their surfaces
keep the same composition, that is, the same colors, both achromatic and chromatic.
Actually, at the level of microlayout (texture) and microcomposition
(conglomeration), layout and reﬂectances merge. Or, to put it differently, the
layout texture and the pigment texture become inseparable.
Note once more that an emphasis on the geometry of surfaces is abstract and
oversimpliﬁed. The faces of the world are not made of some amorphous, colorless,
ghostly substance, as geometry would lead us to believe, but are made of
mud or sand, wood or metal, fur or feathers, skin or fabric. The faces of the
world are colorful as well as geometrical. And what they afford depends on
their substance as well as their shape.
The Sources of Variant Optical Structure
There are two regular and recurrent sources of changing structure in the
ambient light (apart from local events, which will be considered in the next
chapter). First, there are the changes caused by a moving point of observation,
and second, there are the changes caused by a moving source of illumination,
usually the sun. Many pages have been devoted to the former, and we must
now consider the latter. The motion of the sun across the sky from sunrise
to sunset has been for countless millions of years a basic regularity of nature.
It is a fact of ecological optics and a condition of the evolution of eyes in terrestrial
animals. But its importance for the theory of vision has not been fully
recognized.
The puzzling complexities of light and shade cannot be understood without
taking into account the fact of a moving source of illumination. For whenever
the source of light moves, the direction of the light falling on the surfaces of the
80 The Ecological Approach to Visual Perception
world is altered and the shadows themselves move. The layout and coloration of
surfaces persist, but the lightedness and shadedness of these surfaces do not. It is
not just that the optic array is different at noon with high illumination from
what it is at twilight with low illumination; it is that the optic array has a
different structure in the afternoon than it has in the morning.
Variants and Invariants with a Moving Source of Illumination
Just how does pure layout structure the ambient light? It is easy to understand
how a mosaic of black and white substances would structure the ambient light
but not how a pure layout would do so. For in this case the structuring would
have to be achieved wholly by differential illumination, by light and shade.
There are two principles of light and shade under natural conditions that seem
to be clear: the direction of the prevailing illumination and the progressive
weakening of illumination with multiple reﬂection.
The illumination on a surface comes from the sun, the sky, and other surfaces
that face the surface in question. A surface that faces the sun is illuminated
“directly,” a surface that faces away from the sun but still faces the sky is illuminated
less directly, and a surface within a semienclosure that faces only other
surfaces is illuminated still less directly. The more the light has reverberated,
the more of it is absorbed and the dimmer it becomes. Hence it is that surfaces
far from the mouth of a cave are more weakly illuminated than those near the
mouth. But within any airspace, any concavity of the terrain or any semienclosure,
there is a direction of the prevailing illumination, that is, a direction
from which more light comes than from any other.
The illumination of any face of the layout relative to adjacent faces depends
on its inclination to the prevailing illumination. Crudely speaking, the surface
that “faces the light” gets more than its neighbor. More exactly, a surface
perpendicular to the prevailing illumination gets the most, a surface inclined
to it gets less, a surface parallel to it gets still less, and a surface inclined away
from it gets the least. The pairs of terms lighted and shadowed or in light and in
shadow should not be taken as dichotomies, for there are all gradations of relative
light and shade. These two principles of the direction and the amount of illumination
are an attempt to distill a certain ecological simplicity from the
enormous complexities of analytical physical optics and the muddled practice
of illumination engineering.
A wrinkled surface of the same substance evidently structures the ambient
light by virtue of two facts: there is always a prevailing direction of illumination,
and consequently the slopes facing in this direction throw back more
energy than the slopes not facing in this direction. A ﬂat surface of different
substances structures the ambient light by virtue of the simple fact that the
parts of high reﬂectance throw back more energy than the parts of low
reﬂectance.
The Ambient Optic Array 81
Figure 5.9 shows an array from a wrinkled layout of terrestrial surfaces,
actually an aerial photograph of barren hills and valleys. The bare earth of
this desert has everywhere the same reﬂectance. The top of the photograph
is to the north of the terrain. The picture was taken in the morning, and
the sun is in the east. Some of the slopes face east, and some face west; the
former are lighted and the latter shaded. It can be observed that various inclinations
of these surfaces to the direction of the prevailing illumination determine
various relative intensities in the array; the more a surface departs from the
perpendicular to this direction, the darker is the corresponding patch in the
optic array.
Now consider what happens as the sun moves across the sky. All those
surfaces that were lighted in the morning will be shaded in the afternoon, and
all those that were shaded in the morning will be lighted in the afternoon.
There is a continual, if slow, process of change from lighted to shaded on
certain slopes of the layout and the reverse change on certain other slopes.
These slopes are related by orientation. Two faces of any convexity are related
in this way, as are two faces of any concavity. A ridge can be said to consist of
two opposite slopes, and so can a valley. The reciprocity of light and shade on
such surfaces might be described by saying that the lightness and the shadedness
exchange places. The underlying surfaces do not interchange of course, and their
colors, if any, do not interchange. They are persistent, but the illumination is
variable in this special reciprocal way.
In the optic array, presumably, there is an underlying invariant structure to
specify the edges and corners of the layout and the colors of the surfaces, and at
the same time there is a changing structure to specify the temporary direction
of the prevailing illumination. Some components of the array never exchange
places—that is, they are never permuted—whereas other components of the
array do. The former specify a solid surface; the latter specify insubstantial
shadows only. The surface and its color are described as opaque; the shadow is
described as transparent.
The decreasing of illumination on one slope and the increasing of illumination
on an adjacent slope as the sun moves are analogous to the foreshortening
of one slope along with the inverse foreshortening of an adjacent slope as
the point of observation moves. I suggest that the true relative colors of the
adjacent surfaces emerge as the lighting changes, just as the true relative shapes
of the adjacent surfaces emerge as the perspective changes. The perspectives of
the convexities and concavities of Figure 5.9 are variant with locomotion; the
shadows of these convexities and concavities are variant with time of day;
the constant properties of these surfaces underlie the changing perspectives and
the changing shadows and are speciﬁed by invariants in the optic array.
It is true that the travel of the sun across the sky is very slow and that the
correlated interchange of the light and the shade on surfaces is a very gradual
ﬂuctuation. Neither is as obvious as the motion perspective caused by loco-
82 The Ecological Approach to Visual Perception
motion. But the fact is that shifting shadows and a moving sun are regularities
of ecological optics whether or not they are ever noticed by any animal. They
have set the conditions for the perception of the terrain by terrestrial animals
since life emerged from the sea. They make certain optical information available.
And, although shifting shadows and a moving sun are too slow to be
noticed in daylight, a moving source of illumination and the resultant shadows
become more obvious at night. One has only to carry a light from place to place
in a cluttered environment in order to notice the radical shifts in the pattern of
the optic array caused by visibly moving shadows. And yet, of course, the layout
FIGURE 5.9 Hills and valleys on the surface of the barren earth.
The hills in this aerial photograph, the convexities or protuberances, can be
compared to the “humps” shown in Figure 5.1
The Ambient Optic Array 83
of surfaces and their relative coloration is visible underneath the moving
shadows.
How the differential colors of surfaces are speciﬁed in the optic array
separately from the differential illumination of surfaces is, of course, a great
puzzle. The difference between black and white is never confused with
the difference between lighted and shadowed, at least not in a natural environment
as distinguished from a controlled laboratory display. There are many
theories of this so-called constancy of colors in perception, but none of them
is convincing. A new approach to the problem is suggested by the above
considerations.
From an ecological point of view, the color of a surface is relative to the
colors of adjacent surfaces; it is not an absolute color. Its reﬂectance ratio is
speciﬁed only in relation to other reﬂectance ratios of the layout. For the natural
environment is an aggregate of substances. Even a surface is sometimes a conglomerate
of substances. This means that a range of black, gray, and white surfaces
and a range of chromatically colored surfaces will be projected as solid angles
in a normal optic array. The colors are not seen separately, as stimuli, but
together, as an arrangement. And this range of colors provides an invariant
structure that underlies both the changing shadow structure with a moving sun
and the changing perspective structure with a moving observer. The edges and
corners, the convexities and concavities, are thus speciﬁed as multicolored
surfaces, not as mere slopes; as speckled or grained or piebald or whatever, not
as ghostly gray shapes.
The experimental discoveries of E. H. Land (1959) concerning color perception
with what he calls a “complete image” as distinguished from color perception
with controlled patches of radiation in a laboratory are to be understood in
the above way, I believe.
Ripples and Waves on Water: A Special Case
It is interesting and revealing to compare the optical information for a solid
wrinkled surface as shown in Figure 5.9 and the information for a liquid wavy
surface, which the reader will have to visualize. Both consist of convexities and
concavities, but they are motionless on the solid surface and moving on the
liquid surface. In both cases the convexities are lighted on one slope and shadowed
on the other. In both cases the surface is all of the same color or reﬂectance.
The difference between the two arrays is to be found chieﬂy in the two
forms of ﬂuctuation of light and shade. In the terrestrial array, light and shade
exchange places slowly in one direction; they do not oscillate. In the aquatic
array, light and shade interchange rapidly in both directions; they oscillate. In
fact, when the sun is out and the ripples act as mirrors, the reﬂection of the sun
can be said to ﬂicker or ﬂash on and off. This speciﬁc form of ﬂuctuation is
characteristic of a water surface.
84 The Ecological Approach to Visual Perception
Summary
When ambient light at a point of observation is structured it is an ambient optic
array. The point of observation may be stationary or moving, relative to the
persisting environment. The point of observation may be unoccupied or occupied
by an observer.
The structure of an ambient array can be described in terms of visual solid
angles with a common apex at the point of observation. They are angles of
intercept, that is, they are determined by the persisting environment. And they
are nested, like the components of the environment itself.
The concept of the visual solid angle comes from natural perspective, which
is the same as ancient optics. No two such visual angles are identical. The solid
angles of an array change as the point of observation moves, that is, the
perspective structure changes. Underlying the perspective structure, however,
is an invariant structure that does not change. Similarly, the solid angles of an
array change as the sun in the sky moves, that is, the shadow structure changes.
But there are also invariants that underlie the changing shadows.
The moving observer and the moving sun are conditions under which
terrestrial vision has evolved for millions of years. But the invariant principle of
reversible occlusion holds for the moving observer, and a similar principle of
reversible illumination holds for the moving sun. Whatever goes out of sight
will come into sight, and whatever is lighted will be shaded.
8
THE THEORY OF AFFORDANCES
I have described the environment as the surfaces that separate substances from
the medium in which the animals live. But I have also described what the
environment affords animals, mentioning the terrain, shelters, water, ﬁre,
objects, tools, other animals, and human displays. How do we go from surfaces
to affordances? And if there is information in light for the perception of surfaces,
is there information for the perception of what they afford? Perhaps the composition
and layout of surfaces constitute what they afford. If so, to perceive them is
to perceive what they afford. This is a radical hypothesis, for it implies that the
“values” and “meanings” of things in the environment can be directly perceived.
Moreover, it would explain the sense in which values and meanings are external
to the perceiver.
The affordances of the environment are what it offers the animal, what it
provides or furnishes, either for good or ill. The verb to afford is found in the
dictionary, but the noun affordance is not. I have made it up. I mean by it something
that refers to both the environment and the animal in a way that no
existing term does. It implies the complementarity of the animal and the environment.
The antecedents of the term and the history of the concept will be
treated later; for the present, let us consider examples of an affordance.
If a terrestrial surface is nearly horizontal (instead of slanted), nearly ﬂat
(instead of convex or concave), and sufﬁciently extended (relative to the size of
the animal) and if its substance is rigid (relative to the weight of the animal),
then the surface affords support. It is a surface of support, and we call it a
substratum, ground, or ﬂoor. It is stand-on-able, permitting an upright posture
for quadrupeds and bipeds. It is therefore walk-on-able and run-over-able. It is
not sink-into-able like a surface of water or a swamp, that is, not for heavy
terrestrial animals. Support for water bugs is different.
120 The Ecological Approach to Visual Perception
Note that the four properties listed—horizontal, ﬂat, extended, and rigid—
would be physical properties of a surface if they were measured with the scales
and standard units used in physics. As an affordance of support for a species of
animal, however, they have to be measured relative to the animal. They are
unique for that animal. They are not just abstract physical properties. They
have unity relative to the posture and behavior of the animal being considered.
So an affordance cannot be measured as we measure in physics.
Terrestrial surfaces, of course, are also climb-on-able or fall-off-able or getunderneath-able
or bump-into-able relative to the animal. Different layouts
afford different behaviors for different animals, and different mechanical
encounters. The human species in some cultures has the habit of sitting as
distinguished from kneeling or squatting. If a surface of support with the four
properties is also knee-high above the ground, it affords sitting on. We call it a
seat in general, or a stool, bench, chair, and so on, in particular. It may be
natural like a ledge or artiﬁcial like a couch. It may have various shapes, as long
as its functional layout is that of a seat. The color and texture of the surface are
irrelevant. Knee-high for a child is not the same as knee-high for an adult, so
the affordance is relative to the size of the individual. But if a surface is horizontal,
ﬂat, extended, rigid, and knee-high relative to a perceiver, it can in fact
be sat upon. If it can be discriminated as having just these properties, it should
look sit-on-able. If it does, the affordance is perceived visually. If the surface
properties are seen relative to the body surfaces, the self, they constitute a seat
and have meaning.
There could be other examples. The different substances of the environment
have different affordances for nutrition and for manufacture. The different
objects of the environment have different affordances for manipulation. The
other animals afford, above all, a rich and complex set of interactions, sexual,
predatory, nurturing, ﬁghting, playing, cooperating, and communicating.
What other persons afford, comprises the whole realm of social signiﬁcance for
human beings. We pay the closest attention to the optical and acoustic information
that speciﬁes what the other person is, invites, threatens, and does.
The Niches of the Environment
Ecologists have the concept of a niche. A species of animal is said to utilize or
occupy a certain niche in the environment. This is not quite the same as the
habitat of the species; a niche refers more to how an animal lives than to where it
lives. I suggest that a niche is a set of affordances.
The natural environment offers many ways of life, and different animals
have different ways of life. The niche implies a kind of animal, and the animal
implies a kind of niche. Note the complementarity of the two. But note also
that the environment as a whole with its unlimited possibilities existed prior to
animals. The physical, chemical, meteorological, and geological conditions of
The Theory of Affordances 121
the surface of the earth and the pre-existence of plant life are what make animal
life possible. They had to be invariant for animals to evolve.
There are all kinds of nutrients in the world and all sorts of ways of getting
food; all sorts of shelters or hiding places, such as holes, crevices, and caves; all
sorts of materials for making shelters, nests, mounds, huts; all kinds of locomotion
that the environment makes possible, such as swimming, crawling,
walking, climbing, ﬂying. These offerings have been taken advantage of; the
niches have been occupied. But, for all we know, there may be many offerings
of the environment that have not been taken advantage of, that is, niches not yet
occupied.
In architecture a niche is a place that is suitable for a piece of statuary, a place
into which the object ﬁts. In ecology a niche is a setting of environmental
features that are suitable for an animal, into which it ﬁts metaphorically.
An important fact about the affordances of the environment is that they are
in a sense objective, real, and physical, unlike values and meanings, which are
often supposed to be subjective, phenomenal, and mental. But, actually, an
affordance is neither an objective property nor a subjective property; or it is
both if you like. An affordance cuts across the dichotomy of subjective-objective
and helps us to understand its inadequacy. It is equally a fact of the environment
and a fact of behavior. It is both physical and psychical, yet neither. An
affordance points both ways, to the environment and to the observer.
The niche for a certain species should not be confused with what some
animal psychologists have called the phenomenal environment of the species. This
can be taken erroneously to be the “private world” in which the species is
supposed to live, the “subjective world,” or the world of “consciousness.” The
behavior of observers depends on their perception of the environment, surely
enough, but this does not mean that their behavior depends on a so-called
private or subjective or conscious environment. The organism depends on its
environment for its life, but the environment does not depend on the organism
for its existence.
Man’s Alteration of the Natural Environment
In the last few thousand years, as everybody now realizes, the very face of the
earth has been modiﬁed by man. The layout of surfaces has been changed, by
cutting, clearing, leveling, paving, and building. Natural deserts and mountains,
swamps and rivers, forests and plains still exist, but they are being
encroached upon and reshaped by man-made layouts. Moreover, the substances
of the environment have been partly converted from the natural materials of
the earth into various kinds of artiﬁcial materials such as bronze, iron, concrete,
and bread. Even the medium of the environment—the air for us and the water
for ﬁsh—is becoming slowly altered despite the restorative cycles that yielded a
steady state for millions of years prior to man.
122 The Ecological Approach to Visual Perception
Why has man changed the shapes and substances of his environment? To
change what it affords him. He has made more available what beneﬁts him and
less pressing what injures him. In making life easier for himself, of course, he
has made life harder for most of the other animals. Over the millennia, he has
made it easier for himself to get food, easier to keep warm, easier to see at night,
easier to get about, and easier to train his offspring.
This is not a new environment—an artiﬁcial environment distinct from the
natural environment—but the same old environment modiﬁed by man. It
is a mistake to separate the natural from the artiﬁcial as if there were two
environments; artifacts have to be manufactured from natural substances.
It is also a mistake to separate the cultural environment from the natural
environment, as if there were a world of mental products distinct from the
world of material products. There is only one world, however diverse, and all
animals live in it, although we human animals have altered it to suit ourselves.
We have done so wastefully, thoughtlessly, and, if we do not mend our ways,
fatally.
The fundamentals of the environment—the substances, the medium, and
the surfaces—are the same for all animals. No matter how powerful men
become they are not going to alter the fact of earth, air, and water—the lithosphere,
the atmosphere, and the hydrosphere, together with the interfaces that
separate them. For terrestrial animals like us, the earth and the sky are a basic
structure on which all lesser structures depend. We cannot change it. We all ﬁt
into the substructures of the environment in our various ways, for we were all,
in fact, formed by them. We were created by the world we live in.
Some Affordances of the Terrestrial Environment
Let us consider the affordances of the medium, of substances, of surfaces and
their layout, of objects, of animals and persons, and ﬁnally a case of special
interest for ecological optics, the affording of concealmeant by the occluding
edges of the environment (Chapter 5).
The Medium
Air affords breathing, more exactly, respiration. It also affords unimpeded locomotion
relative to the ground, which affords support. When illuminated and
fog-free, it affords visual perception. It also affords the perception of vibratory
events by means of sound ﬁelds and the perception of volatile sources by means
of odor ﬁelds. The airspaces between obstacles and objects are the paths and the
places where behavior occurs.
The optical information to specify air when it is clear and transparent is not
obvious. The problem came up in Chapter 4, and the experimental evidence
about the seeing of “nothing” will be described in the next chapter.
The Theory of Affordances 123
The Substances
Water is more substantial than air and always has a surface with air. It does not
afford respiration for us. It affords drinking. Being ﬂuid, it affords pouring from
a container. Being a solvent, it affords washing and bathing. Its surface does not
afford support for large animals with dense tissues. The optical information for
water is well speciﬁed by the characteristics of its surface, especially the unique
ﬂuctuations caused by rippling (Chapter 5).
Solid substances, more substantial than water, have characteristic surfaces
(Chapter 2). Depending on the animal species, some afford nutrition and some
do not. A few are toxic. Fruits and berries, for example, have more food value
when they are ripe, and this is speciﬁed by the color of the surface. But the food
values of substances are often misperceived.
Solids also afford various kinds of manufacture, depending on the kind of
solid state. Some, such as ﬂint, can be chipped; others, such as clay, can be
molded; still others recover their original shape after deformation; and some
resist deformation strongly. Note that manufacture, as the term implies, was
originally a form of manual behavior like manipulation. Things were fabricated
by hand. To identify the substance in such cases is to perceive what can be
done with it, what it is good for, its utility; and the hands are involved.
The Surfaces and their Layouts
I have already said that a horizontal, ﬂat, extended, rigid surface affords support.
It permits equilibrium and the maintaining of a posture with respect to gravity,
this being a force perpendicular to the surface. The animal does not fall or
slide as it would on a steep hillside. Equilibrium and posture are prerequisite
to other behaviors, such as locomotion and manipulation. There will be more
about this in Chapter 12, and more evidence about the perception of the
ground in Chapter 9. The ground is quite literally the basis of the behavior
of land animals. And it is also the basis of their visual perception, their socalled
space perception. Geometry began with the study of the earth as
abstracted by Euclid, not with the study of the axes of empty space as abstracted
by Descartes. The affording of support and the geometry of a horizontal plane
are therefore not in different realms of discourse; they are not as separate as
we have supposed.
The ﬂat earth, of course, lies beneath the attached and detached objects on it.
The earth has “furniture,” or as I have said, it is cluttered. The solid, level, ﬂat
surface extends behind the clutter and, in fact, extends all the way out to the
horizon. This is not, of course, the earth of Copernicus; it is the earth at the
scale of the human animal, and on that scale it is ﬂat, not round. Wherever
one goes, the earth is separated from the sky by a horizon that, although it may
be hidden by the clutter, is always there. There will be evidence to show that
124 The Ecological Approach to Visual Perception
the horizon can always be seen, in the sense that it can be visualized, and that
it can always be felt, in the sense that any surface one touches is experienced in
relation to the horizontal plane.
Of course, a horizontal, ﬂat, extended surface that is nonrigid, a stream or
lake, does not afford support for standing, or for walking and running. There is
no footing, as we say. It may afford ﬂoating or swimming, but you have to be
equipped for that, by nature or by learning.
A vertical, ﬂat, extended, and rigid surface such as a wall or a cliff face is a
barrier to pedestrian locomotion. Slopes between vertical and horizontal afford
walking, if easy, but only climbing, if steep, and in the latter case the surface
cannot be ﬂat; there must be “holds” for the hands and feet. Similarly, a slope
downward affords falling if steep; the brink of a cliff is a falling-off place. It is
dangerous and looks dangerous. The affordance of a certain layout is perceived
if the layout is perceived.
Civilized people have altered the steep slopes of their habitat by building
stairways so as to afford ascent and descent. What we call the steps afford
stepping, up or down, relative to the size of the person’s legs. We are still
capable of getting around in an arboreal layout of surfaces, tree branches, and
we have ladders that afford this kind of locomotion, but most of us leave that
to our children.
A cliff face, a wall, a chasm, and a stream are barriers; they do not afford
pedestrian locomotion unless there is a door, a gate, or a bridge. A tree or a
rock is an obstacle. Ordinarily, there are paths between obstacles, and these
openings are visible. The progress of locomotion is guided by the perception
of barriers and obstacles, that is, by the act of steering into the openings and
away from the surfaces that afford injury. I have tried to describe the optical
information for the control of locomotion (Gibson, 1958), and it will be further
elaborated in Chapter 13. The imminence of collision with a surface during locomotion
is speciﬁed in a particularly simple way, by an explosive rate of magniﬁcation
of the optical texture. This has been called looming (e.g., Schiff, 1965).
It should not be confused, however, with the magniﬁcation of an opening
between obstacles, the opening up of a vista such as occurs in the approach to a
doorway.
The Objects
The affordances of what we loosely call objects are extremely various. It will be
recalled that my use of the terms is restricted and that I distinguish between
attached objects and detached objects. We are not dealing with Newtonian objects
in space, all of which are detached, but with the furniture of the earth, some
items of which are attached to it and cannot be moved without breakage.
Detached objects must be comparable in size to the animal under consideration
if they are to afford behavior. But those that are comparable afford an
The Theory of Affordances 125
astonishing variety of behaviors, especially to animals with hands. Objects
can be manufactured and manipulated. Some are portable in that they afford
lifting and carrying, while others are not. Some are graspable and other not.
To be graspable, an object must have opposite surfaces separated by a distance
less than the span of the hand. A ﬁve-inch cube can be grasped, but a ten-inch
cube cannot (Gibson, 1966b, p. 119). A large object needs a “handle” to afford
grasping. Note that the size of an object that constitutes a graspable size is
speciﬁed in the optic array. If this is true, it is not true that a tactual sensation
of size has to become associated with the visual sensation of size in order for the
affordance to be perceived.
Sheets, sticks, ﬁbers, containers, clothing, and tools are detached objects that
afford manipulation (Chapter 3). Additional examples are given below.
1. An elongated object of moderate size and weight affords wielding. If
used to hit or strike, it is a club or hammer. If used by a chimpanzee behind bars
to pull in a banana beyond its reach, it is a sort of rake. In either case, it is an
extension of the arm. A rigid staff also affords leverage and in that use is a lever.
A pointed elongated object affords piercing—if large it is is a spear, if small a
needle or awl.
2. A rigid object with a sharp dihedral angle, an edge, affords cutting and
scraping; it is a knife. It may be designed for both striking and cutting, and then
it is an axe.
3. A graspable rigid object of moderate size and weight affords throwing.
It may be a missile or only an object for play, a ball. The launching of missiles by
supplementary tools other than the hands alone—the sling, the bow, the catapult,
the gun, and so on—is one of the behaviors that makes the human animal
a nasty, dangerous species.
4. An elongated elastic object, such as a ﬁber, thread, thong, or rope, affords
knotting, binding, lashing, knitting, and weaving. These are kinds of behavior
where manipulation leads to manufacture.
5. A hand-held tool of enormous importance is one that, when applied to
a surface, leaves traces and thus affords trace-making. The tool may be a stylus,
brush, crayon, pen, or pencil, but if it marks the surface it can be used to depict
and to write, to represent scenes and to specify words.
We have thousands of names for such objects, and we classify them in many
ways: pliers and wrenches are tools; pots and pans are utensils; swords and pistols
are weapons. They can all be said to have properties or qualities: color, texture,
composition, size, shape and features of shape, mass, elasticity, rigidity, and
mobility. Orthodox psychology asserts that we perceive these objects insofar as we
discriminate their properties or qualities. Psychologists carry out elegant experiments
in the laboratory to ﬁnd out how and how well these qualities are discriminated.
The psychologists assume that objects are composed of their qualities. But
126 The Ecological Approach to Visual Perception
I now suggest that what we perceive when we look at objects are their affordances,
not their qualities. We can discriminate the dimensions of difference if
required to do so in an experiment, but what the object affords us is what we
normally pay attention to. The special combination of qualities into which an
object can be analyzed is ordinarily not noticed.
If this is true for the adult, what about the young child? There is much evidence
to show that the infant does not begin by ﬁrst discriminating the qualities
of objects and then learning the combinations of qualities that specify them.
Phenomenal objects are not built up of qualities; it is the other way around. The
affordance of an object is what the infant begins by noticing. The meaning is
observed before the substance and surface, the color and form, are seen as such.
An affordance is an invariant combination of variables, and one might guess
that it is easier to perceive such an invariant unit than it is to perceive all the
variables separately. It is never necessary to distinguish all the features of an
object and, in fact, it would be impossible to do so. Perception is economical.
“Those features of a thing are noticed which distinguish it from other things
that it is not—but not all the features that distinguish it from everything that it is
not” (Gibson, 1966b, p. 286).
TO PERCEIVE AN AFFORDANCE IS NOT TO CLASSIFY
AN OBJECT
The fact that a stone is a missile does not imply that it cannot be other things
as well. It can be a paperweight, a bookend, a hammer, or a pendulum bob.
It can be piled on another rock to make a cairn or a stone wall. These affordances
are all consistent with one another. The differences between them are
not clear-cut, and the arbitrary names by which they are called do not count
for perception. If you know what can be done with a graspable detached
object, what it can be used for, you can call it whatever you please.
The theory of affordances rescues us from the philosophical muddle of
assuming ﬁxed classes of objects, each deﬁned by its common features and
then given a name. As Ludwig Wittgenstein knew, you cannot specify the
necessary and sufﬁcient features of the class of things to which a name is
given. They have only a “family resemblance.” But this does not mean you
cannot learn how to use things and perceive their uses. You do not have to
classify and label things in order to perceive what they afford.
Other Persons and Animals
The richest and most elaborate affordances of the environment are provided by
other animals and, for us, other people. These are, of course, detached objects
The Theory of Affordances 127
with topologically closed surfaces, but they change the shape of their surfaces
while yet retaining the same fundamental shape. They move from place to
place, changing the postures of their bodies, ingesting and emitting certain
substances, and doing all this spontaneously, initiating their own movements,
which is to say that their movements are animate. These bodies are subject to the
laws of mechanics and yet not subject to the laws of mechanics, for they are not
governed by these laws. They are so different from ordinary objects that infants
learn almost immediately to distinguish them from plants and nonliving things.
When touched they touch back, when struck they strike back; in short, they
interact with the observer and with one another. Behavior affords behavior, and
the whole subject matter of psychology and of the social sciences can be thought
of as an elaboration of this basic fact. Sexual behavior, nurturing behavior,
ﬁghting behavior, cooperative behavior, economic behavior, political behavior—all
depend on the perceiving of what another person or other persons
afford, or sometimes on the misperceiving of it.
What the male affords the female is reciprocal to what the female affords the
male; what the infant affords the mother is reciprocal to what the mother
affords the infant; what the prey affords the predator goes along with what the
predator affords the prey; what the buyer affords the seller cannot be separated
from what the seller affords the buyer, and so on. The perceiving of these
mutual affordances is enormously complex, but it is nonetheless lawful, and it
is based on the pickup of the information in touch, sound, odor, taste, and
ambient light. It is just as much based on stimulus information as is the simpler
perception of the support that is offered by the ground under one’s feet. For
other animals and other persons can only give off information about themselves
insofar as they are tangible, audible, odorous, tastable, or visible.
The other person, the generalized other, the alter as opposed to the ego, is
an ecological object with a skin, even if clothed. It is an object, although it
is not merely an object, and we do right to speak of he or she instead of it. But
the other person has a surface that reﬂects light, and the information to
specify what he or she is, invites, promises, threatens, or does can be found in
the light.
Places and Hiding Places
The habitat of a given animal contains places. A place is not an object with deﬁnite
boundaries but a region (Chapter 3). The different places of a habitat may have
different affordances. Some are places where food is usually found and others
where it is not. There are places of danger, such as the brink of a cliff and the
regions where predators lurk. There are places of refuge from predators. Among
these is the place where mate and young are, the home, which is usually a partial
enclosure. Animals are skilled at what the psychologist calls place-learning. They
can ﬁnd their way to signiﬁcant places.
128 The Ecological Approach to Visual Perception
An important kind of place, made intelligible by the ecological approach to
visual perception, is a place that affords concealment, a hiding place. Note that it
involves social perception and raises questions of epistemology. The concealing
of oneself from other observers and the hiding of a detached object from other
observers have different kinds of motivation. As every child discovers, a good
hiding place for one’s body is not necessarily a good hiding place for a treasure.
A detached object can be concealed both from other observers and from the
observer himself. The observer’s body can be concealed from other observers
but not from himself, as the last chapter emphasized. Animals as well as children
hide themselves and also hide objects such as food.
One of the laws of the ambient optic array (Chapter 5) is that at any ﬁxed
point of observation some parts of the environment are revealed and the
remaining parts are concealed. The reciprocal of this law is that the observer
himself, his body considered as part of the environment, is revealed at some
ﬁxed points of observation and concealed at the remaining points. An observer
can perceive not only that other observers are unhidden or hidden from him
but also that he is hidden or unhidden from other observers. Surely, babies
playing peek-a-boo and children playing hide-and-seek are practicing this kind
of apprehension. To hide is to position one’s body at a place that is concealed at
the points of observation of other observers. A “good” hiding place is one that
is concealed at nearly all points of observation.
All of these facts and many more depend on the principle of occluding edges
at a point of observation, the law of reversible occlusion, and the facts of opaque
and nonopaque substances. What we call privacy in the design of housing, for
example, is the providing of opaque enclosures. A high degree of concealment
is afforded by an enclosure, and complete concealment is afforded by a complete
enclosure. But note that there are peepholes and screens that permit seeing
without being seen. A transparent sheet of glass in a window transmits both
illumination and information, whereas a translucent sheet transmits illumination
but not information. There will be more of this in Chapter 11.
Note also that a glass wall affords seeing through but not walking through,
whereas a cloth curtain affords going through but not seeing through. Architects
and designers know such facts, but they lack a theory of affordances to encompass
them in a system.
Summary: Positive and Negative Affordances
The foregoing examples of the affordances of the environment are enough to
show how general and powerful the concept is. Substances have biochemical
offerings and afford manufacture. Surfaces afford posture, locomotion, collision,
manipulation, and in general behavior. Special forms of layout afford
shelter and concealment. Fires afford warming and burning. Detached objects—
tools, utensils, weapons—afford special types of behavior to primates and
The Theory of Affordances 129
humans. The other animal and the other person provide mutual and reciprocal
affordances at extremely high levels of behavioral complexity. At the highest
level, when vocalization becomes speech and manufactured displays become
images, pictures, and writing, the affordances of human behavior are staggering.
No more of that will be considered at this stage except to point out that
speech, pictures, and writing still have to be perceived.
At all these levels, we can now observe that some offerings of the environment
are beneﬁcial and some are injurious. These are slippery terms that should
only be used with great care, but if their meanings are pinned down to biological
and behavioral facts the danger of confusion can be minimized. First,
consider substances that afford ingestion. Some afford nutrition for a given
animal, some afford poisoning, and some are neutral. As I pointed out before,
these facts are quite distinct from the affording of pleasure and displeasure in
eating, for the experiences do not necessarily correlate with the biological
effects. Second, consider the brink of a cliff. On the one side it affords walking
along, locomotion, whereas on the other it affords falling off, injury. Third,
consider a detached object with a sharp edge, a knife. It affords cutting if
manipulated in one manner, but it affords being cut if manipulated in another
manner. Similarly, but at a different level of complexity, a middle-sized metallic
object affords grasping, but if charged with current it affords electric shock.
And fourth, consider the other person. The animate object can give caresses or
blows, contact comfort or contact injury, reward or punishment, and it is not
always easy to perceive which will be provided. Note that all these beneﬁts and
injuries, these safeties and dangers, these positive and negative affordances are
properties of things taken with reference to an observer but not properties of the
experiences of the observer. They are not subjective values; they are not feelings of
pleasure or pain added to neutral perceptions.
There has been endless debate among philosophers and psychologists as to
whether values are physical or phenomenal, in the world of matter or only in
the world of mind. For affordances as distinguished from values, the debate
does not apply. Affordances are neither in the one world or the other inasmuch
as the theory of two worlds is rejected. There is only one environment, although
it contains many observers with limitless opportunities for them to live in it.
The Origin of the Concept of Affordances: A Recent History
The gestalt psychologists recognized that the meaning or the value of a thing
seems to be perceived just as immediately as its color. The value is clear on the
face of it, as we say, and thus it has a physiognomic quality in the way that the
emotions of a man appear on his face. To quote from the Principles of Gestalt
Psychology (Koffka, 1935), “Each thing says what it is . . . . a fruit says ‘Eat me’;
water says ‘Drink me’; thunder says ‘Fear me’; and woman says ‘Love me”’
(p. 7). These values are vivid and essential features of the experience itself.
130 The Ecological Approach to Visual Perception
Koffka did not believe that a meaning of this sort could be explained as a pale
context of memory images or an unconscious set of response tendencies. The
postbox “invites” the mailing of a letter, the handle “wants to be grasped,” and
things “tell us what to do with them” (p. 353). Hence, they have what Koffka
called “demand character.”
Kurt Lewin coined the term Aufforderungscharakter, which has been translated
as invitation character (by J. F. Brown in 1929) and as valence (by D. K. Adams in
1931; cf. Marrow, 1969, p. 56, for the history of these translations). The latter
term came into general use. Valences for Lewin had corresponding vectors, which
could be represented as arrows pushing the observer toward or away from the
object. What explanation could be given for these valences, the characters of
objects that invited or demanded behavior? No one, not even the gestalt theorists,
could think of them as physical and, indeed, they do not fall within the
province of ordinary physics. They must therefore be phenomenal, given the
assumption of dualism. If there were two objects, and if the valence could not
belong to the physical object, it must belong to the phenomenal object—to
what Koffka called the “behavioral” object but not to the “geographical”
object. The valence of an object was bestowed upon it in experience, and
bestowed by a need of the observer. Thus, Koffka argued that the postbox has
a demand character only when the observer needs to mail a letter. He is attracted
to it when he has a letter to post, not otherwise. The value of something was
assumed to change as the need of the observer changed.
The concept of affordance is derived from these concepts of valence, invitation,
and demand but with a crucial difference. The affordance of something
does not change as the need of the observer changes. The observer may or may
not perceive or attend to the affordance, according to his needs, but the affordance,
being invariant, is always there to be perceived. An affordance is not
bestowed upon an object by a need of an observer and his act of perceiving it.
The object offers what it does because it is what it is. To be sure, we deﬁne what
it is in terms of ecological physics instead of physical physics, and it therefore
possesses meaning and value to begin with. But this is meaning and value of a
new sort.
For Koffka it was the phenomenal postbox that invited letter-mailing, not the
physical postbox. But this duality is pernicious. I prefer to say that the real
postbox (the only one) affords letter-mailing to a letter-writing human in a
community with a postal system. This fact is perceived when the postbox is
identiﬁed as such, and it is apprehended whether the postbox is in sight or out
of sight. To feel a special attraction to it when one has a letter to mail is not
surprising, but the main fact is that it is perceived as part of the environment—
as an item of the neighborhood in which we live. Everyone above the age of six
knows what it is for and where the nearest one is. The perception of its affordance
should therefore not be confused with the temporary special attraction it
may have.
The Theory of Affordances 131
The gestalt psychologists explained the directness and immediacy of the
experience of valences by postulating that the ego is an object in experience
and that a “tension” may arise between a phenomenal object and the phenomenal
ego. When the object is in “a dynamic relation with the ego” said Koffka,
it has a demand character. Note that the “tension,” the “relation,” or the
“vector” must arise in the “ﬁeld,” that is, in the ﬁeld of phenomenal experience.
Although many psychologists ﬁnd this theory intelligible, I do not. There is an
easier way of explaining why the values of things seem to be perceived immediately
and directly. It is because the affordances of things for an observer are
speciﬁed in stimulus information. They seem to be perceived directly because
they are perceived directly.
The accepted theories of perception, to which the gestalt theorists were
objecting, implied that no experiences were direct except sensations and that
sensations mediated all other kinds of experience. Bare sensations had to be
clothed with meaning. The seeming directness of meaningful perception was
therefore an embarrassment to the orthodox theories, and the Gestaltists did
right to emphasize it. They began to undermine the sensation-based theories.
But their own explanations of why it is that a fruit says “Eat me” and a woman
says “Love me” are strained. The gestalt psychologists objected to the accepted
theories of perception, but they never managed to go beyond them.
The Optical Information for Perceiving Affordances
The theory of affordances is a radical departure from existing theories of value
and meaning. It begins with a new deﬁnition of what value and meaning are.
The perceiving of an affordance is not a process of perceiving a value-free physFIGURE
8.1 The changing perspective structure of a postbox during approach by an
observer.
As one reduces the distance to the object to one-third, the visual solid angle of the
object increases three times. Actually this is only a detail near the center of an
outﬂowing optic array. (From The Perception of the Visual World by James Jerome
Gibson and used with the agreement of the reprint publisher, Greenwood Press, Inc.)
132 The Ecological Approach to Visual Perception
ical object to which meaning is somehow added in a way that no one has been
able to agree upon; it is a process of perceiving a value-rich ecological object.
Any substance, any surface, any layout has some affordance for beneﬁt or injury
to someone. Physics may be value-free, but ecology is not.
The central question for the theory of affordances is not whether they exist
and are real but whether information is available in ambient light for perceiving
them. The skeptic may now be convinced that there is information in light for
some properties of a surface but not for such a property as being good to eat.
The taste of a thing, he will say, is not speciﬁed in light; you can see its form
and color and texture but not its palatability; you have to taste it for that. The
skeptic understands the stimulus variables that specify the dimensions of visual
sensation; he knows from psychophysics that brightness corresponds to intensity
and color to wavelength of light. He may concede the invariants of structured
stimulation that specify surfaces and how they are laid out and what they are
made of. But he may boggle at invariant combinations of invariants that specify
the affordances of the environment for an observer. The skeptic familiar with
the experimental control of stimulus variables has enough trouble understanding
the invariant variables I have been proposing without being asked to
accept invariants of invariants.
Nevertheless, a unique combination of invariants, a compound invariant, is
just another invariant. It is a unit, and the components do not have to be
combined or associated. Only if percepts were combinations of sensations
would they have to be associated. Even in the classical terminology, it could be
argued that when a number of stimuli are completely covariant, when they
always go together, they constitute a single “stimulus.” If the visual system is
capable of extracting invariants from a changing optic array, there is no reason
why it should not extract invariants that seem to us highly complex.
The trouble with the assumption that high-order optical invariants specify
high-order affordances is that experimenters, accustomed to working in the
laboratory with low-order stimulus variables, cannot think of a way to measure
them. How can they hope to isolate and control an invariant of optical structure
so as to apply it to an observer if they cannot quantify it? The answer
comes in two parts, I think. First, they should not hope to apply an invariant to
an observer, only to make it available, for it is not a stimulus. And, second, they
do not have to quantify an invariant, to apply numbers to it, but only to give it
an exact mathematical description so that other experimenters can make it
available to their observers. The virtue of the psychophysical experiment is
simply that it is disciplined, not that it relates the psychical to the physical by a
metric formula.
An affordance, as I said, points two ways, to the environment and to the
observer. So does the information to specify an affordance. But this does not in
the least imply separate realms of consciousness and matter, a psychophysical
dualism. It says only that the information to specify the utilities of the environ-
The Theory of Affordances 133
ment is accompanied by information to specify the observer himself, his body,
legs, hands, and mouth. This is only to reemphasize that exteroception is accompanied
by proprioception—that to perceive the world is to coperceive oneself.
This is wholly inconsistent with dualism in any form, either mind-matter
dualism or mind-body dualism. The awareness of the world and of one’s
complementary relations to the world are not separable.
The child begins, no doubt, by perceiving the affordances of things for her,
for her own personal behavior. She walks and sits and grasps relative to her own
legs and body and hands. But she must learn to perceive the affordances of
things for other observers as well as for herself. An affordance is often valid for
all the animals of a species, as when it is part of a niche. I have described the
invariants that enable a child to perceive the same solid shape at different points
of observation and that likewise enable two or more children to perceive the
same shape at different points of observation. These are the invariants that
enable two children to perceive the common affordance of the solid shape despite
the different perspectives, the affordance of a toy, for example. Only when each
child perceives the values of things for others as well as for herself does she
begin to be socialized.
Misinformation for Affordances
If there is information in the ambient light for the affordances of things, can
there also be misinformation? According to the thoery being developed, if
information is picked up perception results; if misinformation is picked up
misperception results.
The brink of a cliff affords falling off; it is in fact dangerous and it looks
dangerous to us. It seems to look dangerous to many other terrestrial animals
besides ourselves, including infant animals. Experimental studies have been
made of this fact. If a sturdy sheet of plate glass is extended out over the edge it
no longer affords falling and in fact is not dangerous, but it may still look
dangerous. The optical information to specify depth-downward-at-an-edge is
still present in the ambient light; for this reason the device was called a visual
cliff by E. J. Gibson and R. D. Walk (1960). Haptic information was available to
specify an adequate surface of support, but this was contradictory to the optical
information. When human infants at the crawling stage of locomotion were
tested with this apparatus, many of them would pat the glass with their hands
but would not venture out on the surface. The babies misperceived the affordance
of a transparent surface for support, and this result is not surprising.
Similarly, an adult can misperceive the affordance of a sheet of glass by
mistaking a closed glass door for an open doorway and attempting to walk
through it. He then crashes into the barrier and is injured. The affordance of
collision was not speciﬁed by the outﬂow of optical texture in the array, or it
was insufﬁciently speciﬁed. He mistook glass for air. The occluding edges of
134 The Ecological Approach to Visual Perception
the doorway were speciﬁed and the empty visual solid angle opened up
symmetrically in the normal manner as he approached, so his behavior was
properly controlled, but the imminence of collision was not noticed. A little
dirt on the surface, or highlights, would have saved him.
These two cases are instructive. In the ﬁrst a surface of support was mistaken
for air because the optic array speciﬁed air. In the second case a barrier was
mistaken for air for the same reason. Air downward affords falling and is
dangerous. Air forward affords passage and is safe. The mistaken perceptions
led to inappropriate actions.
Errors in the perception of the surface of support are serious for a terrestrial
animal. If quicksand is mistaken for sand, the perceiver is in deep trouble. If a
covered pitfall is taken for solid ground, the animal is trapped. A danger is
sometimes hidden—the shark under the calm water and the electric shock in
the radio cabinet. In the natural environment, poison ivy is frequently mistaken
for ivy. In the artiﬁcial environment, acid can be mistaken for water.
THINGS THAT LOOK LIKE WHAT THEY ARE
If the affordances of a thing are perceived correctly, we say that it looks like
what it is. But we must, of course, learn to see what things really are—for
example, that the innocent-looking leaf is really a nettle or that the helpfulsounding
politician is really a demagogue. And this can be very difﬁcult.
A wildcat may be hard to distinguish from a cat, and a thief may look
like an honest person. When Koffka asserted that “each thing says what it is,”
he failed to mention that it may lie. More exactly, a thing may not look like
what it is.
Nevertheless, however true all this may be, the basic affordances of the
environment are perceivable and are usually perceivable directly, without an
excessive amount of learning. The basic properties of the environment that
make an affordance are speciﬁed in the structure of ambient light, and hence
the affordance itself is speciﬁed in ambient light. Moreover, an invariant variable
that is commensurate with the body of the observer himself is more easily picked
up than one not commensurate with his body.
Summary
The medium, substances, surfaces, objects, places, and other animals have
affordances for a given animal. They offer beneﬁt or injury, life or death. This
is why they need to be perceived.
The Theory of Affordances 135
The possibilities of the environment and the way of life of the animal go
together inseparably. The environment constrains what the animal can do, and
the concept of a niche in ecology reﬂects this fact. Within limits, the human
animal can alter the affordances of the environment but is still the creature of
his or her situation.
There is information in stimulation for the physical properties of things, and
presumably there is information for the environmental properties. The doctrine
that says we must distinguish among the variables of things before we can learn
their meanings is questionable. Affordances are properties taken with reference
to the observer. They are neither physical nor phenomenal.
The hypothesis of information in ambient light to specify affordances is
the culmination of ecological optics. The notion of invariants that are related
at one extreme to the motives and needs of an observer and at the other
extreme to the substances and surfaces of a world provides a new approach to
psychology.
Direct perception is what one gets from seeing Niagara Falls, say, as distinguished
from seeing a picture of it. The latter kind of perception is mediated. So
when I assert that perception of the environment is direct, I mean that it is not
mediated by retinal pictures, neural pictures, or mental pictures. Direct perception
is the activity of getting information from the ambient array of light. I call this
a process of information pickup that involves the exploratory activity of looking
around, getting around, and looking at things. This is quite different from the
supposed activity of getting information from the inputs of the optic nerves,
whatever they may prove to be.
The evidence for direct visual perception has accumulated slowly, over
many years. The very idea had to be developed, the results of old experiments
had to be reinterpreted, and new experiments had to be carried out. The next
two chapters are devoted to the experimental evidence.
The experiments will be considered under three main headings: ﬁrst, the
direct perception of surface layout; second, the direct perception of changing
surface layout; and third, the direct perception of the movements of the self.
This chapter is devoted to the direct perception of surface layout.
Evidence for the Direct Perception of Surface Layout
Some thirty years ago, during World War II, psychologists were trying to
apply the theory of depth perception to the problems of aviation, especially
the problem of how a ﬂier lands an airplane. Pilots were given tests for depth
perception, and there was controversy as to whether depth perception was
learned or innate. The same tests are still being given, and the same disagreement
continues.
9
EXPERIMENTAL EVIDENCE
FOR DIRECT PERCEPTION
Persisting Layout
140 The Ecological Approach to Visual Perception
The theory of depth perception assumes that the third dimension of space is
lost in the two-dimensional retinal image. Perception must begin with form
perception, the ﬂat patchwork of colors in the visual ﬁeld. But there are supposedly
cues for depth, which, if they are utilized, will add a third dimension to the
ﬂat visual ﬁeld. A list of the cues for depth is given in most psychology textbooks:
linear perspective, apparent size, superposition, light and shade, relative motion,
aerial perspective, accommodation (the monocular cues), along with binocular
disparity and convergence (the binocular cues). You might suppose that adequate
tests could be made of a prospective ﬂier’s ability to use these cues and that experiments
could be devised to ﬁnd out whether or not they were learned.
The trouble was that none of the tests based on the cues for depth predicted
the success or failure of a student pilot, and none of the proposals for improving
depth perception by training made it any easier to learn to ﬂy. I was deeply
puzzled by this fact. The accepted theory of depth perception did not work. It did
not apply to problems where one might expect it to apply. I began to suspect that
the traditional list of cues for depth was inadequate. And in the end I came to
believe that the whole theory of depth perception was false.
I suggested a new theory in a book on what I called the visual world (Gibson,
1950b). I considered “the possibility that there is literally no such thing as a
perception of space without the perception of a continuous background surface”
(p. 6). I called this a ground theory of space perception to distinguish it from the
air theory that seemed to underlie the old approach. The idea was that the world
consisted of a basic surface with adjoining surfaces, not of bodies in empty air.
The character of the visual world was given not by objects but by the background
of the objects. Even the space of the airplane pilot, I said, was determined
by the ground and the horizon of the earth, not by the air through which
he ﬂies. The notion of space of three dimensions with three axes for Cartesian
coordinates was a great convenience for mathematics, I suggested, but an
abstraction that had very little to do with actual perception.
I would now describe the ground theory as a theory of the layout of surfaces.
By layout, I mean the relations of surfaces to the ground and to one another, their
arrangement. The layout includes both places and objects, together with other
features. The theory asserts that the perception of surface layout is direct. This
means that perception does not begin with two-dimensional form perception.
Hence, there is no special kind of perception called depth perception, and the
third dimension is not lost in the retinal image since it was never in the environment
to begin with. It is a loose term. If depth means the dimension of an object
that goes with height and width, there is nothing special about it. Height
becomes depth when the object is seen from the top, and width becomes depth
when the object is seen from the side. If depth means distance from here, then it
involves self-perception and is continually changing as the observer moves about.
The theory of depth perception is based on confusion and perpetuated by the
fallacy of the retinal picture.
Experimental Evidence for Direct Perception 141
I now say that there is information in ambient light for the perception of the
layout of surfaces but not that there are cues or clues for the perception of depth.
The traditional list of cues is worthless if perception does not begin with a ﬂat
picture. I tried to reformulate the list in 1950 as “gradients and steps of retinal
stimulation” (Gibson, 1950b, pp. 137 ff.). The hypothesis of gradients was a
good beginning, but the reformulation failed. It had the great handicap of
being based on physiological optics and the retinal image instead of ecological
optics and the ambient array.
Such is the hypothesis of the direct perception of surface layout. What is the
evidence to support it? Some experiments had been carried out even before
1950, outdoor experiments in the open air instead of laboratory experiments
with spots of light in a darkroom, but they were only a beginning (Gibson,
1947). Much more experimental evidence has accumulated in the last twentyﬁve
years.
The Psychophysics of Space and Form Perception
The studies to be described were thought of as psychophysical experiments at
the time they were performed. There was to be a new psychophysics of perception
as well as the old psychophysics of sensation. For I thought I had discovered
that there were stimuli for perceptions in much the same way that there were
known to be stimuli for sensations. This now seems to me a mistake. I failed
to distinguish between stimulation proper and stimulus information, between
what happens at passive receptors and what is available to active perceptual
systems. Traditional psychophysics is a laboratory discipline in which physical
stimuli are applied to an observer. He is prodded with controlled and systematically
varied bits of energy so as to discover how his experience varies correspondingly.
This procedure makes it difﬁcult or impossible for the observer to
extract invariants over time. Stimulus prods do not ordinarily carry information
about the environment.
What I had in mind by a psychophysics of perception was simply the emphasis
on perception as direct instead of indirect. I wanted to exclude an extra process
of inference or construction. I meant (or should have meant) that animals and
people sense the environment, not in the meaning of having sensations but in
the meaning of detecting. When I asserted that a gradient in the retinal image
was a stimulus for perception, I meant only that it was sensed as a unit; it was not
a collection of points whose separate sensations had to be put together in the
brain. But the concept of the stimulus was not clear to me. I should have asserted
that a gradient is stimulus information. For it is ﬁrst of all an invariant property
of an optic array. I should not have implied that a percept was an automatic
response to a stimulus, as a sense impression is supposed to be. For even then I
realized that perceiving is an act, not a response, an act of attention, not a triggered
impression, an achievement, not a reﬂex.
142 The Ecological Approach to Visual Perception
So what I should have meant by a “psychophysical” theory of perception in
1950 and by perception as a “function of stimulation” in the essay I wrote in
1959 (Gibson, 1959) was the hypothesis of a one-stage process for the perception
of surface layout instead of a two-stage process of ﬁrst perceiving ﬂat forms
and then interpreting the cues for depth.
I now believe that there is no such thing as ﬂat-form perception, just as there
is no such thing as depth perception. (There are drawings and pictures, to be
sure, but these are not “forms,” as I will explain in Part IV. The theory of form
perception in psychology is no less confused than the theory of depth perception.)
But this was not clear when I wrote my book in 1950, where I promised
not only a psychophysics of space perception in Chapter 5 but also a psychophysical
approach to form perception in Chapter 10. This sounded promising and
progressive. Visual outline forms, I suggested, are not unique entities. “They
could be arranged in a systematic way such that each form would differ only
gradually and continuously from all others” (Gibson, 1950b, p. 193). What counts
is not the form as such but the dimensions of variation of form. And psychophysical
experiments could be carried out if these dimensions were isolated.
Here was the germ of the modern hypothesis of the distinctive features of
graphic symbols. It also carries the faint suggestion of a much more radical
hypothesis, that what the eye picks up is a sequential transformation, not a
form. The study of form discrimination by psychophysical methods has ﬂourished
in the last thirty years. W. R. Garner, Julian Hochberg, Fred Attneave,
and others have achieved the systematic variation of outline forms and patterns
in elegant ways (e.g., Garner, 1974). My objection to this research is that it tells
us nothing about perceiving the environment. It still assumes that vision is
simplest when there is a form on the retina that copies a form on a surface facing
the retina. It perpetuates the fallacy that form perception is basic. It holds back
the study of invariants in a changing array. But the hypothesis that forms are
directly perceived does not upset the orthodoxies of visual theory as does the
hypothesis that invariants are directly perceived, and hence it is widely accepted.
The psychophysical approach to surface perception is much more radical
than the psychophysical approach to form perception, and it has not been widely
accepted over the last twenty-ﬁve years. Has its promise been fulﬁlled? Some
experiments can be summarized, and the evidence should be pulled together.
Experiments on the Perception of a Surface as Distinguished
from Nothing
Metzger’s Experiment
Is tridimensional space perception based on bidimensional sensations to which
the third dimension is added, or is it based on surface perception? The ﬁrst
experiment bearing on this issue is that of W. Metzger in 1930. He faced the
Experimental Evidence for Direct Perception 143
eyes of his observer with a large, dimly lighted plaster wall, which rendered the
light coming to the visual system unfocusable. Neither eye could accommodate,
and probably the eyes could not converge. The total ﬁeld (Ganzfeld) was, as he
put it, homogeneous. Under high illumination, the observer simply perceived the
wall, and the outcome was so obvious as to be uninteresting. But under low
illumination, the ﬁne-grained texture of the surface was no longer registered
by a human eye, and the observer reported seeing what he called a fog or haze
or mist of light. He certainly did not see a surface in two dimensions, and
therefore Metzger was tempted to conclude that he saw something in three
dimensions; that is, he was perceiving “space.”
But I did not see depth in the “mist of light.” Another way to get a homogeneous
ﬁeld is to confront the eyes with a hemisphere of diffusing glass highly
illuminated from the outside (Gibson and Dibble, 1952). A better way is to
cover each eye with a ﬁtted cap of strongly diffusing translucent material worn
like a pair of goggles (Gibson and Waddell, 1952). The structure of the entering
light, the optical texture, can thus be eliminated at any level of intensity. What
my observers and I saw under these conditions could better be described as
“nothing” in the sense of “no thing.” It was like looking at the sky. There was
no surface and no object at any distance. Depth was not present in the experience
but missing from it. What the observer saw, as I would now put it, was an
empty medium.
The essence of Metzger’s experiment and its subsequent repetitions is not the
plaster wall or the panoramic surface or the diffusing glass globe or the eyecaps.
The experiment provides discontinuities in the light to an eye at one
extreme and eliminates them at the other. The purpose of the experiment is to
control and vary the projective capacity of light. This must be isolated from the
stimulating capacity of light. Metzger’s experiment points to the distinction
between an optic array with structure and a nonarray without structure. To the
extent that the array has structure it speciﬁes an environment.
A number of experiments using a panoramic surface under low illumination
have been carried out, although the experimenters did not always realize what
they were doing. But all the experiments involved more or less faint discontinuities
in the light to the eye. What the observers said they saw is complex and
hard to describe. One attempt was made by W. Cohen in 1957, and the other
experiments have been surveyed by L. L. Avant (1965). It is fair to say that there
are intermediate perceptions between seeing nothing and seeing something as the
discontinuities become stronger. These are the polar opposites of perception
that are implied by Metzger’s experiment, not the false opposites of seeing in
two dimensions and seeing in three dimensions.
The confusion over whether there is or is not “depth” in Metzger’s luminous
fog is what led me to think that the whole theory of depth, distance, the third
dimension, and space is misconceived. The important result is the neglected
one that a surface is seen when the array has structure, that is, differences in
144 The Ecological Approach to Visual Perception
different directions. A perfectly ﬂat surface in front of the eyes is still a layout,
that is, a wall. And that is all that “seeing in two dimensions” can possibly
mean.
The Experiment with Translucent Eye-Caps
Eliminating optical texture from the light entering the eye by means of translucent
diffusing goggles is an experiment that has been repeated many times.
The observer is blind, not to light, for the photoreceptors are still stimulated,
but to the environment, for the ocular system is inactivated; its adjustments are
frustrated. The observer cannot look at or look around, and I shall devote a
chapter to this activity later. The eye-caps have also been adapted for experiments
on the development of vision in young animals. It was known that when
diurnal animals such as primates were reared from birth in complete darkness
they were blind by certain criteria when brought into an illuminated environment
(although this was not true of nocturnal animals whose ancestors were
used to getting around in the dark). Now it was discovered that animals
deprived of optical structure but not of optical stimulation were also partly
blind when the eye-caps were removed. Crudely speaking, they could not use
their eyes properly. Anatomical degeneration of the photoreceptors had not
occurred, as with the animals reared in the dark, but the exploratory adjustments
of the visual system had not developed normally. The experiments are
described in Chapter 12 of Perceptual Learning and Development by Eleanor J.
Gibson (1969).
Experiments with a Sheet of Glass
It is fairly well known that a clean sheet of plate glass that projects no reﬂections
or highlights to the observer’s eye is, as we say, invisible. This fact is not selfexplanatory,
but it is very interesting. It means that one perceives air where a
material surface exists, because air is speciﬁed by the optic array. I have seen
people try to walk through plate-glass doors to their great discomﬁture and
deer try to jump through plate-glass windows with fatal results.
A perfectly clear sheet of glass transmits both light considered as energy and
an array of light considered as information. A frosted or pebbled sheet of glass
transmits optical energy but not optical information. The clear sheet can be seen
through, as we say, but the frosted sheet cannot. The latter can be seen, but the
former cannot. An imperceptible sheet of glass can be made increasingly
perceptible by letting dust or powder fall on it or by spattering it. Even the
faintest specks can specify the surface. In this intermediate case, the sheet transmits
both the array from the layout behind the glass and the array from the glass
itself. We say that we see the farther surface through the glass surface. The optical
structure of one is mixed or interspersed with the optical structure of the other.
Experimental Evidence for Direct Perception 145
The transparency of the near surface, more properly its semitransparency, is
then perceived (Gibson, 1976). One sees two surfaces, separated in depth, in
the same direction from here or, better, within the same visual solid angle of
the ambient array. At least one sees them separated if the interspersed structures
are different, or if the elements of one move relative to the elements of the other
(E. J. Gibson, Gibson, Smith, and Flock, 1959).
Many of the above assertions are based on informal experiments that have
not been published. But the reader can check them for himself with little
trouble. I conclude that a surface is experienced when the structural information
to specify it is picked up.
Experiments with a Pseudotunnel
In the case of a sheet of glass, a surface may exist and go unperceived if it is not
speciﬁed. In the next experiment, a surface may be nonexistent but may be
perceived if it is speciﬁed. The pseudosurface in this case was not ﬂat and
frontal but was a semienclosure, a cylindrical tunnel viewed from one end. I
called it an optical tunnel to suggest that the surface was not material or substantial
but was produced by the light to the eye. Another way of describing it
would be to say that it was a virtual but not a real tunnel.
The purpose of the experiment was to provide information for the perception
of the inside surface of a cylinder without the ordinary source of this
information, the inside surface of a cylinder. I would now call this a display. The
fact that the perception was illusory is incidental. I wanted to elicit a synthetic
perception, and I, therefore, had to synthesize the information. It was an experiment
in perceptual psychophysics, more exactly, psycho-optics. The observers
were fooled, to be sure, but that was irrelevant. There was no information in
the array to specify that it was a display. This situation, I shall argue, is very rare.
My collaborators and I (Gibson, Purdy, and Lawrence, 1955) generated a
visual solid angle of about 30° at the point of observation. This array consisted
of alternating dark and light rings nested within one another, separated by
abrupt circular contours. The number of rings and contours from the periphery
to the center of the array could be varied. At one extreme there were thirty-six
contours, and at the other seven.
Thus the mean density of the contrasts in the array was varied from ﬁne to
coarse. The gradient of this density could also be varied; normally the density
increased from the periphery toward the center.
The source of this display, the apparatus, was a set of large, very thin, plastic
sheets, each hiding the next, with a one-foot hole cut in the center of each.
They were indirectly illuminated from above or below. The contours in the
array were caused by the edges of the sheets. The texture of the plastic was so
ﬁne as to be invisible. Black and white sheets could be hung in alternation one
behind another, or, as a control, all-black or all-white surfaces could be
146 The Ecological Approach to Visual Perception
FIGURE 9.1 The optic array coming to the eye from the optical tunnel.
There are nine contrasts in this cross-section of the array, that is, nine transitions of
luminous intensity. The next ﬁgure shows a longitudinal section. The point of observation
for the ﬁgure on the left is centered with the tunnel, whereas the point of observation
for the ﬁgure on the right is to the right of center. (From J. J. Gibson, J. Purdy,
and L. Lawrence: “A Method of Controlling Stimulation for the Study of
Space Perception: The Optical Tunnel,” Journal of Experimental Psychology, 1955, 50,
1–14. Copyright 1955 by the American Psychological Association. Reprinted by
permission.)
FIGURE 9.2 A longitudinal section of the optical tunnel shown in Figure 9.1.
Nine plastic sheets are shown, black and white alternating, with the cut edges of
the nine holes aligned. The increase in the density of the contrasts from the periphery
to the center of the array is evident. (From J. J. Gibson, J. Purdy, and L.
Lawrence: “A Method of Controlling Stimulation for the Study of Space Perception:
The Optical Tunnel,” Journal of Experimental Psychology, 1955, 50, 1–14. Copyright
1955 by the American Psychological Association. Reprinted by permission.)
Experimental Evidence for Direct Perception 147
displayed. The observers looked into these holes from a booth, and extreme
precautions were taken to prevent them from having any preconception of
what they would see.
The principal result was as follows. When all-black or all-white surfaces
were used, the observers saw nothing; the area within the ﬁrst hole was
described as a hazy or misty fog, a dark or light ﬁlm, without obvious depth. At
the other extreme, when thirty-six dark and light rings were displayed, all
observers saw a continuous striped cylindrical surface, a solid tunnel. No edges
were seen, and “a ball could be rolled from the far end to the entrance.”
When nineteen contrasts were displayed, two-thirds of the observers
described a solid tunnel. When thirteen contrasts were displayed, half did so;
and when seven contrasts were displayed, only one-third did so. In each case,
the remainder said they saw either segments of surface with air in between or a
series of circular edges (which was, of course, correct). With fewer contrasts,
the experience became progressively less continuous and substantial. The proximity
of these contours had proved to be crucial. Surfaciness depended on their
mean density in the array.
What about the cylindrical shape of the surface, the receding layout of the
tunnel? This could be altered in a striking way and the tunnel converted into a
ﬂat surface like an archery target with rings around a bull’s-eye simply by
rearranging the sheets in the way illustrated. The gradient of increasing proximity
toward the center of the array gives way to an equal proximity. But the
target surface instead of the tunnel surface appeared only if the observer’s head
was ﬁxed and one eye was covered, that is, if the array was frozen and single. If
the head was moved or the other eye used, the tunnel shape was again seen. The
frozen array speciﬁed a ﬂat target, but the dual or transforming array speciﬁed
a receding tunnel. This is only one of many experiments in which perception
with monocular ﬁxed vision is exceptional.
Conclusion
These experiments with a dimly lighted wall, with translucent eye-caps, with
a sheet of glass, and with a pseudotunnel seem to show that the perception of
surfaciness depends on the proximity to one another of discontinuities in the
optic array. A surface is the interface between matter in the gaseous state and
matter in the liquid or solid state. A surface comes to exist as the matter on one
side of the interface becomes more substantial (Chapter 2). The medium is
insubstantial. Mists, clouds, water, and solids are increasingly substantial. These
substances are also increasingly opaque, except for a substance like glass, which
is rare in nature. What these experiments have done is to vary systematically
the optical information for the perception of substantiality and opacity. (But see
the next chapter on the perception of coherence.)
148 The Ecological Approach to Visual Perception
The experiment with the pseudotunnel also seems to show that the perception
of a surface as such entails the perception of its layout, such as the frontfacing
layout of a wall or the slanting layout of a tunnel. Both are kinds of
layout, and the traditional distinction between two-dimensional and threedimensional
vision is a myth.
Experiments on the Perception of the Surface of Support
The ground outdoors or the ﬂoor indoors is the main surface of support. Animals
have to be supported against gravity. If the layout of surfaces is to be substituted
for space in the theory of perception, this fundamental surface should get ﬁrst
consideration. How is it perceived? Animals like us can always feel the surface of
support except when falling freely. But we can also see the surface of support
under our feet if we are, in fact, supported. The ground is always speciﬁed in the
lower portion of the ambient array. The standing infant can always see it and can
always see her feet hiding parts of it. This is a law of ecological optics.
The Glass Floor
A ﬂoor can be experimentally modiﬁed. When the “visual cliff” was being
constructed for experiments with young animals by E. J. Gibson and R. D.
FIGURE 9.3 An arrangement that provides an array with a constant density of
contrasts from periphery to center.
Only the ﬁrst seven apertures are shown. The observer does not see a tunnel with this
display but a ﬂat surface with concentric rings, something like an archery target, so
long as the head is immobile and one eye is covered. (From J. J. Gibson, J. Purdy, and
L. Lawrence: “A Method of Controlling Stimulation for the Study of Space Perception:
The Optical Tunnel,” Journal of Experimental Psychology, 1955, 50, 1–14. Copyright
1955 by the American Psychological Association. Reprinted by permission.)
Experimental Evidence for Direct Perception 149
Walk (1960), observations were made with a large sheet of glass that was horizontal
instead of vertical, a glass ﬂoor instead of a glass wall. The animal or
child can be put down on this surface under two conditions: when it is visible,
by virtue of textured paper placed just under the glass, and when it is invisible,
with the paper placed far below the glass. The glass affords support under both
conditions but provides optical information for support only under the ﬁrst. There
is mechanical contact with the feet in both cases but optical information for
contact with the feet only in the ﬁrst.
The animals or babies tested in this experiment would walk or crawl normally
when they could both see and feel the surface but would not do so when they
could only feel the surface; in the latter case, they froze, crouched, and showed
signs of discomfort. Some animals even adopted the posture they would have
when falling (E. J. Gibson and Walk, 1960, pp. 65–66). The conclusion seems
to be that some animals require optical information for support along with the
inertial and tactual information in order to walk normally. For my part, I should
feel very uncomfortable if I had to stand on a large observation platform with a
transparent ﬂoor through which the ground was seen far below.
The optical information in this experiment, I believe, is contradictory to the
haptic information. One sees oneself as being up in the air, but one feels oneself
in contact with a surface of support and, of course, one feels the normal pull of
gravity in the vestibular organ. In such cases of contradictory or conﬂicting
information, the psychologist cannot predict which will be picked up. The
perceptual outcome is uncertain.
Note that the perception of the ground and the coperception of the self are
inseparable in this situation. One’s body in relation to the ground is what gets
attention. Perception and proprioception are complementary. But the
commonly accepted theories of space perception do not bring out this fact.
The Visual Cliff
The visual cliff experiments of E. J. Gibson, R. D. Walk, and subsequently
others are very well known. They represented a new approach to the ancient
puzzle of depth perception, and the results obtained with newborn or darkreared
animals were surprising because they suggested that depth perception
was innate. But the sight of a cliff is not a case of perceiving the third dimension.
One perceives the affordance of its edge. A cliff is a feature of the terrain, a
highly signiﬁcant, special kind of dihedral angle in ecological geometry, a
falling-off place. The edge at the top of a cliff is dangerous. It is an occluding
edge. But is has the special character of being an edge of the surface of support,
unlike the edge of a wall. One can safely walk around the edge of a wall but not
off the edge of a cliff. To perceive a cliff is to detect a layout but, more than that,
it is to detect an affordance, a negative affordance for locomotion, a place where
the surface of support ends.
150 The Ecological Approach to Visual Perception
An affordance is for a species of animal, a layout relative to the animal and
commensurate with its body. A cliff is a drop-off that is large relative to the size
of the animal, and a step is a drop-off that is small relative to its size. A falling-off
edge is dangerous, but a stepping-down edge is not. What animals need to
perceive is not layout as such but the affordances of the layout, as emphasized in
the last chapter. Consider the difference between the edge of a horizontal
surface and the edge of a vertical surface, the edge of a ﬂoor and the edge of a
wall. You go over the former whereas you can go around the latter. Both are
dihedral angles, and both are occluding edges. But the meanings of the two
kinds of “depth” are entirely different.
Gibson and Walk (1960; Walk and Gibson, 1961) constructed a virtual cliff
with the glass-ﬂoor apparatus. They tested animals and babies to determine
whether or not they would go forward over the virtual cliff. Actually, they
provided two edges on either side of a narrow platform, one a falling-off edge
and the other a stepping-down edge appropriate to the species of animal being
tested. The animals’ choices were recorded. Nearly all terrestrial animals chose
the shallow edge instead of the deep one.
The results have usually been discussed in terms of depth perception and the
traditional cues for depth. But they are more intelligible in terms of the perception
of layout and affordances. The separation in depth at an edge of the surface
of support is not at all the same thing as the depth dimension of abstract space.
FIGURE 9.4 The invisibly supported object.
The real object is held up in the air by a hidden rod attached to a heavy base. The
virtual object appears to be resting on the ground where the bottom edge of the
real object hides the ground, so long as vision is monocular and frozen. One sees a
concave corner, not an occluding edge. Because the virtual object is at twice the
distance of the real object, it is seen as twice the size.
Experimental Evidence for Direct Perception 151
As for innate versus learned perception, it is much more sensible to assume an
innate capacity to notice falling-off places in terrestrial animals than it is to
assume that they have innate ideas or mental concepts of geometry.
An Object Resting on the Ground
I suggested that one sees the contact of his feet with the ground. This is equally
true for other objects than feet. We see whether an object is on the ground or
up in the air. How is this contact with or separation from the ground perceived?
The answer is suggested by an informal experiment described in my book on
the visual world (Gibson, 1950b, Fig. 72, pp. 178 ff.), which might be called the
invisibly-supported-object experiment. I did not clearly understand it at the time,
but the optics of occluding edges now makes it more intelligible.
A detached object can be attached to a long rod that is hidden to the observer.
The rod can be lowered by the experimenter so that the object rests on the
ground or raised so that it stands up in the air. The object can be a cardboard
rectangle or trapezoid or a ball, but it must be large enough to hide the rod and
its base. An observer who stands at the proper position and looks with two eyes,
or with one eye and a normally moving head, perceives a resting object as
resting on the surface of support and a raised object as raised above the surface
of support. The size and distance of the object are seen correctly. But an
observer who looks with one eye and a ﬁxed head, through a peephole or with
a biting board, gets an entirely different perception. A resting object is seen
correctly, but a raised object is also seen to be resting on the surface. It is seen
at the place where its edge hides the texture of the surface. It appears farther away and
larger than it really is.
This illusion is very interesting. It appears only with monocular arrested
vision—a rare and unnatural kind of vision. The increments and decrements of
the texture of the ground at the edges of the object have been eliminated, both
those of one eye relative to the other and those that are progressive in time at
each eye. In traditional theory, the cues of binocular and motion parallax are
absent. But it is just these increments and decrements of the ground texture that
specify the separation of object from ground. The absence of this accretion/deletion
speciﬁes contact of the object with the ground. A surface is perceived to
“stand up” or “stand out” from the surface that extends behind it only to the
extent that the gap is speciﬁed. And this depends on seeing from different
points of observation, either two points of observation at the same time or
different points of observation at different times.
A ﬂat surface that “goes back to” or “lies ﬂat on” the ground will seem to
have a different size, shape, and even reﬂectance than it has when it stands forth
in the air. This feature of the illusion is also very interesting, and I have demonstrated
it many times. The ﬁrst published study of it is that of J. E. Hochberg
and J. Beck (1954).
152 The Ecological Approach to Visual Perception
Experiments with the Ground as Background
Investigators in the tradition of space perception and the cues for depth have
usually done experiments with a background in the frontal plane, that is, a
surface facing the observer, a wall, a screen, or a sheet of paper. A form in this
plane is most similar to a form on the retina, and extension in this plane might
be seen as a simple sensation. This follows from retinal image optics. But investigators
of environment perception do experiments with the ground as background,
studying surfaces instead of forms, and using ecological optics. Instead
of studying distance in the air, they study recession along the ground. Distance
as such cannot be seen directly but can only be inferred or computed. Recession
along the ground can be seen directly.
Distance and Size Perception on the Ground
Although the linear perspective of a street in a painting had been known since
the Renaissance, and the converging appearance of a parallel alley of trees in a
landscape had been discussed since the eighteenth century, no one had ever
studied the perception of a naturally textured ground. Linear perspective was
an obvious cue for distance, but the gradient of density or proximity of the
texture of the ground was not so obvious. E. G. Boring has described the old
experiments with artiﬁcial alleys (1942, pp. 290–296), but the ﬁrst experiment
with an ordinary textured ﬁeld outdoors, I believe, was published at the end of
World War II (Gibson, 1947). A plowed ﬁeld without furrows receding almost
to the horizon was used. No straight edges were visible. This original experiment
required the judgment of the height of a stake planted in the ﬁeld at some
distance up to half a mile. At such a distance the optical size of the elements of
texture and the optical size of the stake itself were extremely small.
Up until that time the unanimous conclusion of observers had been that
parallel lines were seen to converge and that objects were seen to be smaller “in
the distance.” There was a tendency toward “size constancy” of objects, to be
sure, but it was usually incomplete. The assumption had always been that size
constancy must “break down.” It was supposed that an object will cease to be
even visible at some eventual distance and that presumably it ceases to be visible
by way of becoming smaller. (See Gibson, 1950b, p. 183, for a statement of this
line of reasoning.) With the naive observers in the open ﬁeld experiment,
however, the judgments of the size of the stake did not decrease, even when it
was a ten-minute walk away and becoming hard to make out. The judgments
became more variable with distance but not smaller. Size constancy did not break
down. The size of the object only became less deﬁnite with distance, not smaller.
The implication of this result, I now believe, is that certain invariant ratios were
picked up unawares by the observers and that the size of the retinal image went
unnoticed. No matter how far away the object was, it intercepted or occluded the
Experimental Evidence for Direct Perception 153
same number of texture elements of the ground. This is an invariant ratio. For any
distance the proportion of the stake extending above the horizon to that extending
below the horizon was invariant. This is another invariant ratio. These invariants
are not cues but information for direct size perception. The observers in this
experiment were aviation trainees and were not interested in the perspective
appearance of the terrain and the objects. They could not care less for the patchwork
of colors in the visual ﬁeld that had long fascinated painters and psychologists.
They were set to pick up information that would permit a size-match
between the distant stake and one of a set of nearby stakes. The perception of the
size and distance of an object on the ground had proved to be unlike the perception
of the size and distance of an object in the sky. The invariants are missing in
the latter case. The silhouette of an airplane might be a ﬁfty-foot ﬁghter at a onemile
altitude or a hundred-foot bomber at a two-mile altitude. Airplane spotters
could be trained to estimate altitude, but only by the method of recognizing the
shape, knowing the size by having memorized the wingspan, and inferring the
distance from the angular size. Errors were considerable at best. This kind of inferential
knowledge is not characteristic of ordinary perception. Baron von Helmholtz
called it “unconscious” inference even in the ordinary case, but I am skeptical.
Comparison of Stretches of Distance Along the Ground
The size of an object on the ground is not entirely separable from the sizes of
the objects that compose the ground. The terrain is made of clods and particles
of earth, or rocks and pebbles, or grass clumps and grass blades. These nested
objects might have size constancy just as much as orthodox objects. In the next
set of experiments on ground perception, the very distinction between size and
distance breaks down. What had to be compared were not stakes or objects but
stretches of the ground itself, distances between markers placed by the experimenter.
In this case distances between here and there could be compared with
distances between there and there. These open-ﬁeld experiments were conducted
by Eleanor J. Gibson (Gibson and Bergman, 1954; Gibson, Bergman, and
Purdy, 1955; Purdy and Gibson, 1955).
Markers could be set down and moved anywhere in a level ﬁeld of grass up
to 350 yards away. The most interesting experiment of the series required the
observer to bisect a stretch of distance, which could extend either from his feet
to a marker or from one marker to another (Purdy and Gibson, 1955). A mobile
marker on wheels had to be stopped by the observer at the halfway point. The
ability to bisect a length had been tested in the laboratory with an adjustable
stick called a Galton bar but not with a piece of ground on which the observer
stood.
All observers could bisect a stretch of distance without difﬁculty and with
some accuracy. The farther stretch could be matched to the nearer one, although
the visual angles did not match. The farther visual angle was compressed
154 The Ecological Approach to Visual Perception
relative to the nearer, and its surface was, to use a vague term, foreshortened.
But no constant error was evident. A stretch from here to there could be equated
with a stretch from there to there. The conclusion must be that observers were
not paying attention to the visual angles; they must have been noticing information.
They might have been detecting, without knowing it, the amount of
texture in a visual angle. The number of grass clumps projected in the farther
half of a stretch of distance is exactly the same as the number projected in the
nearer half. It is true that the optical texture of the grass becomes denser and
more vertically compressed as the ground recedes from the observer, but the
rule of equal amounts of texture for equal amounts of terrain remains invariant.
This is a powerful invariant. It holds for either dimension of the terrain, for
width as well as for depth. In fact, it holds for any regularly textured surface
whatever, that is, any surface of the same substance. And it holds for walls and
ceilings as well as for ﬂoors. To say that a surface is regularly textured is only to
assume that bits of the substance tend to be evenly spaced. They do not have to
be perfectly regular like crystals in a lattice but only “stochastically” regular.
The implications of this experiment on fractionating a stretch of the ground
are radical and far-reaching. The world consists not only of distances from here,
my world, but also of distances from there, the world of another person. These
intervals seem to be strikingly equivalent.
The rule of equal amounts of texture for equal amounts of terrain suggests
that both size and distance are perceived directly. The old theory that the
perceiver allows for the distance in perceiving the size of something is unnecessary.
The assumption that the cues for distance compensate for the sensed smallness
of the retinal image is no longer persuasive. Note that the pickup of the
amount of texture in a visual solid angle of the optic array is not a matter of
counting units, that is, of measuring with an arbitrary unit. The other experiments
of this open-ﬁeld series required the observers to make absolute judgments,
so-called, of distances in terms of yards. They could learn to do so readily
enough (E. J. Gibson and Bergman, 1954; E. J. Gibson, Bergman, and Purdy,
1955), but it was clear that one had to see the distance before one could apply a
number to it.
Observations of the Ground and the Horizon
When the terrain is ﬂat and open, the horizon is in the ambient optic array. It
is a great circle between the upper and the lower hemisphere separating the sky
and the earth. But this is a limiting case. The farther stretches of the ground are
usually hidden by frontal surfaces such as hills, trees, and walls. Even in an
enclosure, however, there has to be a surface of support, a textured ﬂoor. The
maximum coarseness of its optical texture is straight down, where the feet are,
and the density increases outward from this center. These radial gradients
projected from the surface of support increase with increasing size of the ﬂoor.
Experimental Evidence for Direct Perception 155
The densities of texture do not become inﬁnite except when there is an inﬁnitely
distant horizon. Only at this limit is the optical structure of the array
wholly compressed. But the gradients of density specify where the outdoors
horizon would be, even in an enclosure. That is, there exists an implicit horizon
even when the earth-sky horizon is hidden.
EVEN SPACING
The fact that the parts of the terrestrial environment tend to be “evenly
spaced” was noted in my early book on the visual world (Gibson, 1950b,
pp. 77–78). This is equivalent to the rule of equal amounts of texture for
equal amounts of terrain. The fact can be stated in various ways. However
stated, it seems to be a fact that can be seen, not necessarily an intellectual
concept of abstract space including numbers and magnitudes. Ecological
geometry does not have to be learned from textbooks.
The concept of a vanishing point comes from artiﬁcial perspective, converging
parallels, and the theory of the picture plane. The vanishing limit of optical
structure at the horizon comes from natural perspective, ecological optics, and
the theory of the ambient optic array. The two kinds of perspective should not
be confused, although they have many principles in common (Chapter 5).
The terrestrial horizon is thus an invariant feature of terrestrial vision, an
invariant of any and all ambient arrays, at any and all points of observation. The
horizon never moves, even when every other structure in the light is changing.
This stationary great circle is, in fact, that to which all optical motions have
reference. It is neither subjective nor objective; it expresses the reciprocity of
observer and environment; it is an invariant of ecological optics.
The horizon is the same as the skyline only in the case of the open ground or
the open ocean. The earth-sky contrast may differ from the true horizon
because of hills or mountains. The horizon is perpendicular to the pull of
gravity and to the two poles of the ambient array at the centers of the two hemispheres;
in short, the horizon is horizontal. With reference to this invariant, all
other objects, edges, and layouts in the environment are judged to be either
upright or tilted. In fact, the observer perceives himself to be in an upright or tilted
posture relative to this invariant. (For an early and more complex discussion of
visual uprightness and tilt in terms of the retinal image, see Gibson, 1952, on the
“phenomenal vertical.”)
The facts about the terrestrial horizon are scarcely mentioned in traditional
optics. The only empirical study of it is one by H. A. Sedgwick (1973) based on
ecological optics. He shows how the horizon is an important source of invariant
information for the perception of all kinds of objects. All terrestrial objects, for
156 The Ecological Approach to Visual Perception
example, of the same height are cut by the horizon in the same ratio, no matter
what the angular size of the object may be. This is the “horizon ratio relation”
in its simplest form. Any two trees or poles bisected by the horizon are the same
height, and they are also precisely twice my eye-height. More complex ratios
specify more complex layouts. Sedgwick showed that judgments of the sizes of
objects represented in pictures were actually determined by these ratios.
The perceiving of what might be called eye level on the walls, windows,
trees, poles, and buildings of the environment is another case of the complementarity
between seeing the layout of the environment and seeing oneself in
the environment. The horizon is at eye level relative to the furniture of the
earth. But this is my eye level, and it goes up and down as I stand and sit. If I
want my eye level, the horizon, to rise above all the clutter of the environment,
I must climb up to a high place. The perception of here and the perception of
inﬁnitely distant from here are linked.
Experiments on the Perception of Slant
Experiments on the direct perception of layout began in 1950. From the beginning,
the crucial importance of the density of optical texture was evident. How
could it be varied systematically in an experiment? Along with the outdoor
experiments, I wanted to try indoor experiments in the laboratory. I did not
FIGURE 9.5 The base of each pillar covers the same amount of the texture of the
ground.
The width of each pillar is that of one paving stone.The pillars will be seen to have the
same width if this information is picked up.The height of each pillar is speciﬁed by a
similar invariant, the “horizon-ratio” relation, described later.
Experimental Evidence for Direct Perception 157
then understand ambient light but only the retinal image, and this led me to
experiment with texture density in a window or picture. The density could be
increased upward in the display (or downward or rightward or leftward), and
the virtual surface would then be expected to slant upward (or downward or
whatever). The surface should slant away in the direction of increasing texture
density; it should be inclined from the frontal plane at a certain angle that
corresponded to the rate of change of density, the gradient of density. Every
piece of surface in the world, I thought, had this quality of slant (Gibson,
1950a). The slant of the apparent surface behind the apparent window could be
judged by putting the palm of the hand at the same inclination from the frontal
plane and recording it with an adjustable “palm board.” This appeared to be a
neat psychophysical experiment, for it isolated a variable, the gradient of density.
The ﬁrst experiment (Gibson 1950a) showed that with a uniform density
over the display the phenomenal slant is zero and that with increases of density
in a given direction one perceives increasing slant in that direction. But the
apparent slant was not proportional to the geometrically predicted slant. It was
less than it should be theoretically. The experiment has been repeated with
modiﬁcations by Gibson and J. Cornsweet (1952), J. Beck and J. J. Gibson (1955),
R. Bergman and J. J. Gibson (1959), and many other investigators. It is not a neat
psychophysical experiment. Phenomenal slant does not simply correspond to the
gradient. The complexities of the results are described by H. R. Flock (1964,
1965) and by R. B. Freeman (1965).
What was wrong with these experiments? In consideration of the theory of
layout, we can now understand it. The kind of slant studied was optical, not
geographical, as noted by Gibson and Cornsweet (1952). It was relative to the
frontal plane perpendicular to the line of sight, not relative to the surface of
the earth, and was thus merely a new kind of depth, a quality added to each
of the ﬂat forms in the patchwork of the visual ﬁeld. I had made the mistake of
thinking that the experience of the layout of the environment could be
compounded of all the optical slants of each piece of surface. I was thinking of
slant as an absolute quality, whereas it is always relative. Convexities and concavities
are not made up of elementary impressions of slant but are instead unitary
features of the layout.
The impression of slant cannot be isolated by displaying a texture inside a
window, for the perception of the occluding edge of the window will affect
it; the surface is slanted relative to the surface that has the window in it. The
separation of these surfaces is underestimated, as the experimental results
showed.
The supposedly absolute judgment of the slant of a surface behind a window
becomes more accurate when a graded decrease of velocity of the texture across
the display is substituted for a graded increase of density of the texture, as demonstrated
by Flock (1964). The virtual surface “stands back” from the virtual
window. It slants away in the direction of decreasing ﬂow of the texture but is
158 The Ecological Approach to Visual Perception
perceived to be a rigidly moving surface if the ﬂow gradient is mathematically
appropriate. But this experiment belongs not with experiments on surface
layout but with those on changing surface layout, and these experiments will be
described later.
Is There Evidence Against the Direct Perception of
Surface Layout?
There are experiments, of course, that seem to go against the theory of a direct
perception of layout and to support the opposite theory of a mediated perception
of layout. The latter theory is more familiar. It asserts that perception is mediated
by assumptions, preconceptions, expectations, mental images, or any of a
dozen other hypothetical mediators. The demonstrations of Adelbert Ames,
once very popular, are well known for being interpreted in this way, especially
the Distorted Room and the Rotating Trapezoidal Window.
These demonstrations are inspired by the argument from equivalent conﬁgurations.
A diagram illustrating equivalent conﬁgurations is given in Figure 9.7.
The argument is that many possible objects can give rise to one retinal image
and that hence a retinal image cannot specify the object that gave rise to it.
FIGURE 9.6 The invariant horizon ratio for terrestrial objects.
The telephone poles in this display are all cut by the horizon in the same ratio.The
proportion differs for objects of different heights.The line where the horizon cuts the
tree is just as high above the ground as the point of observation, that is, the height of
the observer’s eye. Hence everyone can see his own eye-height on the standing objects
of the terrain.
Experimental Evidence for Direct Perception 159
But the image, according to the argument, is all one has for information. The
perception of an object, therefore, requires an assumption about which of
the many possible objects that could exist gave rise to the present image (or to
the visual solid angle corresponding to it). The argument is supposed to apply
to each of a collection of objects in space.
A distorted room with trapezoidal surfaces can be built so as to give rise to
a visual solid angle at the point of observation identical with the solid angle
from a normal rectangular room. Or a trapezoidal window with trapezoids for
windowpanes can be built and made to rotate so that its changing visual solid
angle is identical with the changing solid angle from a rectangular window
slanted 45° away from the real distorted window. The window is always oneeighth
of a rotation behind itself, as it were. A single and stationary point of
observation is taken for granted. An observer who looks with one eye and a
stationary head misperceives the trapezoidal surfaces and has the experience of
a set of rectangular surfaces, a “virtual” form or window, instead of the actual
plywood construction invented by the experimenter. Anomalies of perception
result that are striking and curious. The eye has been fooled.
The explanation is that, in the absence of information, the observer has presupposed
(assumed, expected, or whatever) the existence of rectangular surfaces
causing the solid angles at the eye. That is reasonable, but it is then concluded that
presuppositions are necessary for perception in general, since a visual solid angle
cannot specify its object. There will always be equivalent conﬁgurations for any
solid angle or any set of solid angles at a point of observation.
The main fallacy in this conclusion, as the reader will recognize, is the
generalization from peephole observation to ordinary observation, the assumption
that because the perspective structure of an optic array does not specify the
surface layout nothing in the array can specify the layout. The hypothesis of
invariant structure that underlies the perspective structure and emerges clearly
when there is a shift in the point of observation goes unrecognized. The fact is
that when an observer uses two eyes and certainly when one looks from various
points of view the abnormal room and the abnormal window are perceived for
what they are, and the anomalies cease.
The demonstrations do not prove, therefore, that the perception of layout
cannot be direct and must be mediated by preconceptions, as Adelbert Ames
and his followers wanted to believe (Ittelson, 1952). Neither do the many other
demonstrations that, over the centuries, have purported to prove it.
The diagram of equivalent conﬁgurations illustrates one of the perplexities
inherent to the retinal image theory of perception: if many different objects can
give rise to the same stimulus, how do we ever perceive an object? The other half
of the puzzle is this: if the same object can give rise to many different stimuli, how
can we perceive the object? (Note that the second question implies a moving object
but that neither question admits the fact of a moving observer.) Koffka was perplexed
by this dual puzzle (1935, pp. 228 ff.) and many other experimenters have tried to
160 The Ecological Approach to Visual Perception
resolve it, but without success (e.g., Beck and Gibson, 1955). The only way out, I
now believe, is to abandon the dogma that a retinal stimulus exists in the form of a
picture. What speciﬁes an object are invariants that are themselves “formless.”
Summary
The experiment of providing either structure or no structure in the light to an
eye results in the perception of a surface or no surface. The difference is not
between seeing in two dimensions and seeing in three dimensions, as earlier
investigators supposed.
The closer together the discontinuities in an experimentally induced optic
array, the greater is the “surfaciness” of the perception. This was true, at least,
for a 30° array having seven contours at one extreme and thirty-six at the other.
Optical contact of one’s body with the surface of support as well as mechanical
contact seem to be necessary for some terrestrial animals if they are to stand
and walk normally.
Perceiving the meaning of an edge in the surface of support, either a
falling-off edge or a stepping-down edge, seems to be a capability that animals
develop. This is not abstract depth perception but affordance perception.
Experiments on the perception of distance along the ground instead of
distance through the air suggest that such perception is based on invariants in
the array instead of cues. The rule of equal amounts of texture for equal
amounts of terrain is one such invariant, and the horizon ratio relation is
another. On this basis, the dimensions of things on the ground are perceived
FIGURE 9.7 Equivalent conﬁgurations within the same visual solid angle.
This perspective drawing shows a rectangle and three transparent trapezoids,all of which
ﬁt within the envelope of the same visual solid angle. Thus all four quadrangles are
theoretically equivalent for a single eye at a ﬁxed point of observation.They are,however,
ghosts, not surfaces.
Experimental Evidence for Direct Perception 161
directly, and the old puzzle of the constancy of perceived size at different
distances does not arise.
The fact of the terrestrial horizon in the ambient array should not be
confused with the vanishing point of linear perspective in pictorial optics.
A series of experiments on the perception of the slant of a surface relative to
the line of sight did not conﬁrm the absolute gradient hypothesis. The implication
was that the slants of surfaces relative to one another and to the ground,
the depth-shapes of the layout, are what get perceived.
Experiments based on the argument from equivalent conﬁgurations do not
prove the need to have presuppositions in order to perceive the environment,
since they leave out of account the fact that an observer normally moves about.
13
LOCOMOTION AND
MANIPULATION
The theory of affordances implies that to see things is to see how to get about
among them and what to do or not do with them. If this is true, visual perception
serves behavior, and behavior is controlled by perception. The observer
who does not move but only stands and looks is not behaving at the moment, it
is true, but he cannot help seeing the affordances for behavior in whatever he
looks at.
Moving from place to place is supposed to be “physical” whereas perceiving
is supposed to be “mental,” but this dichotomy is misleading. Locomotion is
guided by visual perception. Not only does it depend on perception but perception
depends on locomotion inasmuch as a moving point of observation is
necessary for any adequate acquaintance with the environment. So we must
perceive in order to move, but we must also move in order to perceive.
Manipulation is another kind of behavior that depends on perception and
also facilitates perception. Let us consider in this chapter how vision enters into
these two kinds of behavior.
The Evolution of Locomotion and Manipulation
Support
Animals, no less than other bodies, are pulled downward by the force of gravity.
They fall unless supported. In water the animal is supported by the medium,
which has about the same density as its body. But in air the animal must have a
substantial surface below if it is not to become a Newtonian falling body.
Locomotion has evolved from swimming in the sea to crawling and walking
on land to clinging and climbing on the protuberances that clutter up the land
214 The Ecological Approach to Visual Perception
and, ﬁnally, to ﬂying through the air, the most rapid kind of locomotion but the
most risky. Fish are supported by the medium, terrestrial animals by a substantial
surface on the underside, and birds (when they are not at rest) by airﬂow,
the aerodynamic force called lift. Zoologists sometimes classify animals as
aquatic, terrestrial, or aerial, having in mind the different ways of getting about
in water, on land, or in the air.
Visual Perception of Support
A terrestrial animal must have a surface that pushes up on its feet, or its
underside. The experiments reported in Chapter 4 with the glass ﬂoor apparatus
suggest that many terrestrial animals cannot maintain normal posture
unless they can see their feet on the ground. With optical information to specify
their feet off the ground, they act as if they were falling freely, crouching
and showing signs of fear. But when a textured surface is brought up under
the glass ﬂoor, the animals stand and walk normally (E. J. Gibson, 1969,
pp. 267–270).
This result implies that contact of the feet with the surface of support as
against separation of the feet from the surface is speciﬁed optically, at the
occluding edges of the feet. The animal who moves its head or uses two eyes
can perceive either no separation in depth between its feet and the ﬂoor or the
kind of separation it would see if it were suspended in air. Contact is speciﬁed
both optically and mechanically.
Note that a rigid surface of earth can be distinguished from a nonrigid
surface of water by its color, texture, and the absence or presence of ripples.
A surface of water does not afford support for chicks, but it does for
ducklings. The latter take to the water immediately after hatching; the former
do not.
Manipulation
Manipulation presumably evolved in primates, along with bipedal locomotion
and the upright posture, by the conversion of the forelimbs from legs into arms
and of the forepaws into what we call hands. Walking on two legs, it is sometimes
said, leaves the hands free for other acts. The hands are speciﬁed by “ﬁvepronged
squirming protrusions” into the ﬁeld of view from below (Chapter 7).
They belong to the self, but they are constantly touching the objects of the
outer world by reaching and grasping. The shapes and sizes of objects, in fact,
are perceived in relation to the hands, as graspable or not graspable, in terms of
their affordances for manipulation. Infant primates learn to see objects and
their hands in conjunction. The perception is constrained by manipulation, and
the manipulation is constrained by perception.
Locomotion and Manipulation 215
The Control of Locomotion and Manipulation
Locomotion and manipulation, like the movements of the eyes described in the
last chapter, are kinds of behavior that cannot be reduced to responses. The
persistent effort to do so by physiologists and psychologists has come to a dead
end. But the ancient Cartesian doctrine still hangs on, that animals are reﬂex
mahines and that humans are the same except for a soul that rules the body by
switching impulses at the center of the brain. The doctrine will not do.
Locomotion and manipulation are not triggered by stimuli from outside the
body, nor are they initiated by commands from inside the brain. Even the classiﬁcation
of incoming impulses in nerves as sensory and outgoing impulses as
motor is based on the old doctrine of mental sensations and physical movements.
Neurophysiologists, most of them, are still under the inﬂuence of dualism,
however much they deny philosophizing. They still assume that the brain is the
seat of the mind. To say, in modem parlance, that it is a computer with a
program, either inherited or acquired, that plans a voluntary action and then
commands the muscles to move is only a little better than Descartes’s theory,
for to say this is still to remain conﬁned within the doctrine of responses.
Locomotion and manipulation are neither triggered nor commanded but
controlled. They are constrained, guided, or steered, and only in this sense
are they ruled or governed. And they are controlled not by the brain but by
information, that is, by seeing oneself in the world. Control lies in the animalenvironment
system. Control is by the animal in its world, the animal itself
having subsystems for perceiving the environment and concurrently for getting
about in it and manipulating it. The rules that govern behavior are not like laws
enforced by an authority or decisions made by a commander; behavior is regular
without being regulated. The question is how this can be.
WHAT HAPPENS TO INFANT PRIMATES DEPRIVED OF
THE SIGHT OF THEIR HANDS?
Monkeys reared from birth in a device that kept them from seeing the hands
and body but not from feeling them move and touching things were very
abnormal monkeys. When freed from the device, they acted at ﬁrst as if they
could not reach for and grasp an object but must grope for it. An opaque
shield with a cloth bib ﬁtted tightly around the monkey’s neck had eliminated
visual kinesthesis and had thus prevented the development of visual
control of reaching and grasping. So I interpret the results of an experiment
by R. Held and J. A. Bauer (1974). See my discussion of the optical information
for hand movement in Chapter 7.
216 The Ecological Approach to Visual Perception
The Medium Contains the Information for Control
It should be kept in mind that animals live in a medium that, being insubstantial,
permits them to move about, if supported. We are tempted to call the
medium “space,” but the temptation should be resisted. For the medium, unlike
space, permits a steady state of reverberating illumination to become established
such that it contains information about surfaces and their substances.
That is, there is an array at every point of observation and a changing array at
every moving point of observation. The medium, as distinguished from space,
allows compression waves from a mechanical event, sound, to reach all points
of observation and also allows the diffusion ﬁeld from a volatile substance,
odor, to reach them (Gibson, 1966b, Ch. 1). The odor is speciﬁc to the volatile
substance, the sound is speciﬁc to the event, and the visual solid angle is the
most speciﬁc of all, containing all sorts of structured invariants for perceiving
the affordance of the object. This is why to perceive something is also to
perceive how to approach it and what to do about it.
Information in a medium is not propagated as signals are propagated but is
contained. Wherever one goes, one can see, hear, and smell. Hence, perception
in the medium accompanies locomotion in the medium.
Visual Kinesthesis and Control
Before getting into the problem of control, we should be clear about the difference
between active and passive movement, a difference that is especially
important in the case of locomotion. For animal locomotion may be uncontrolled;
the animal may be simply transported. This can happen in various
ways. A ﬂow of the medium can transport the animal, as happens to the bird in
a wind and the ﬁsh in a stream. Or an individual may be transported by another
animal, as happens to a monkey clinging to its mother or a baby carried in a
cradleboard. Or the observer may be a passenger in a vehicle. In all these cases,
the animal can see its locomotion without initiating, governing, or steering it.
The animal has the information for transportation but cannot regulate it. In my
terminology, the observer has visual kinesthesis but no visual control of the
movement. This distinction is essential to an understanding of the problem of
control. The traditional theory of the senses is incapable of making it, however,
and followers of the traditional theory become mired in the conceptual confusion
arising from the slippery notion of feedback.
Visual kinesthesis speciﬁes locomotion relative to the environment, whereas
the other kinds of kinesthesis may or may not do so. The control of locomotion
in the environment must therefore be visual. Walking, bicycling, and driving
involve very different kinds of classical kinesthesis but the same visual kinesthesis.
The muscle movements must be governed by vision. If you want to go
somewhere, or to know where you are going, you can only trust your eyes. The
Locomotion and Manipulation 217
bird in a wind even has to ﬂy in order to stay in the same place. To prevent
being carried away, it must arrest the ﬂow of the ambient array.
Before we can hope to understand controlled locomotion, therefore, we
must answer several preliminary questions about the information in ambient
light. I can think of four. What speciﬁes locomotion or stasis? What speciﬁes
an obstacle or an opening? What speciﬁes imminent contact with a surface?
What speciﬁes the beneﬁt or the injury that lies ahead? These questions must
be answered before we can begin to ask what the rules are for starting and stopping,
for approaching and retreating, for going this way or that way, and so on.
The Optical Information Necessary for Control of Locomotion
For each of the four questions above, I shall list a number of assertions about
optical information. I will try to put together what the previous chapters have
established.
What Speciﬁes Locomotion or Stasis?
1. Flow of the ambient array speciﬁes locomotion, and nonﬂow speciﬁes stasis. By
ﬂow is meant the change analyzed as motion perspective (Gibson, Olum, and
Rosenblatt, 1955) for the abstract case of an uncluttered environment and a
moving point of observation. A better term would be ﬂow perspective, or streaming
perspective. It yields the “melon-shaped family of curves” illustrated in Figure 13.1
and is based on rays of light from particles of the terrain, not on solid angles from
features of the terrain. Thus, it has the great advantages of geometrical analysis
but also has its disadvantages. Nevertheless, the ﬂow as such speciﬁes locomotion
and the invariants specify the layout of surfaces in which locomotion occurs.
2. Outﬂow speciﬁes approach to and inﬂow speciﬁes retreat from. An invariant
feature of the ambient ﬂow is that one hemisphere is centrifugal and the other
centripetal. Outﬂow entails magniﬁcation, and inﬂow entails miniﬁcation.
There is always both a going-to and a coming-from during locomotion. A
creature with semipanoramic vision can register both the outﬂow and the
inﬂow at the same time, but human creatures can sample only one or the other,
by looking “ahead” or by looking “behind.” Note that a reversal of the ﬂow
pattern speciﬁes a reversal of locomotion.
3. The focus or center of outﬂow speciﬁes the direction of locomotion in the environment.
More exactly, that visual solid angle at the center of outﬂow speciﬁes the
surface in the environment, or the object, or the opening, toward which the
animal is moving. This statement is not analytical. Because the overall ﬂow is
radial in both hemispheres, the two foci are implicit in any sufﬁciently large
sample of the ambient array, and even humans can thus see where they are going
without having to look where they are going. The “melon-shaped family of
curves” continues outside the edges of the temporary ﬁeld of view.
218 The Ecological Approach to Visual Perception
4. A shift of the center of outﬂow from one visual solid angle to another speciﬁes a
change in the direction of locomotion, a turn, and a remaining of the center within the
same solid angle speciﬁes no change in direction. The ambient optic array is here
supposed to consist of nested solid angles, not of a bundle of lines. The direction
of locomotion is thus anchored to the layout, not to a coordinate system.
The ﬂow of the ambient array can be transposed over the invariant structure of the
array, so that where one is going is seen relative to the surrounding layout. This
unfamiliar notion of invariant structure underlying the changing perspective
structure is one that I tried to make explicit in Chapter 5; here is a good example
of it. The illustrations in Chapter 7 showing arrows superposed on a picture of
FIGURE 13.1 The ﬂow velocities in the lower hemisphere of the ambient optic array
with locomotion parallel to the earth.
The vectors are plotted in angular coordinates, and all vectors vanish at the horizon.
This drawing should be compared with Figure 7.3 showing the motion perspective
to a ﬂying bird. (From Gibson, Olum, and Rosenblatt, 1955. © 1955 by the Board
of Trustees of the University of Illinois. Reprinted by permission of the University
of Illinois Press.)
Locomotion and Manipulation 219
the terrain were supposed to suggest this invariance under change but, of
course, it cannot be pictured.
5. Flow of the textured ambient array just behind certain occluding protrusions into
the ﬁeld of view speciﬁes locomotion by an animal with feet. If you lower your head
while walking, a pair of moving protrusions enters the ﬁeld of view from its
lower edge (Chapter 7), and these protrusions move up and down alternately.
A cat sees the same thing except that what it sees are front feet. The extremities
are in optical contact with the ﬂowing array at the locus of maximal ﬂow and
maximally coarse texture. They occlude parts of the surface, but it is seen to
extend behind them. Convexities and concavities in the surface will affect the
timing of contact, and therefore you and the cat must place your feet with
regard to the footing.
What Speciﬁes an Obstacle or an Opening?
I distinguish two general cases for the affording of locomotion, which I will call
obstacle and opening. An obstacle is a rigid object, detached or attached, a surface
with occluding edges. An opening is an aperture, hole, or gap in a surface, also
with occluding edges. An obstacle affords collision. An opening affords passage.
Both have a closed or nearly closed contour in the optic array, but the edge of
the obstacle is inside the contour, whereas the edge of the opening is outside the
contour. A round object hides in one direction, and a round opening hides in
the opposite direction. The way to tell the difference between an obstacle and
an opening, therefore, is as follows.
ON LOOKING AT THE ROAD WHILE DRIVING
It must be admitted that when I turn around while driving our car and reply
to my wife’s protests that I can perfectly well see where I am going without
having to look where I am going because the focus of outﬂow is implicit, she
is not reassured.
6. Loss (or gain) of structure outside a closed contour during approach (or retreat)
speciﬁes an obstacle. Gain (or loss) of structure inside a closed contour during approach
(or retreat) speciﬁes an opening. This is the only absolutely trustworthy way to tell
the difference between an obstacle and an opening. In both cases the visual solid
angle goes to a hemisphere as you approach it, but you collide with the obstacle
and enter the opening. Magniﬁcation of the form as such, the outline, does not
distinguish them. But as you come up to the obstacle it hides more and more of
the vista, and as you come up to the opening it reveals more and more of the
vista. Deletion outside the occluding edge and accretion inside the occluding
220 The Ecological Approach to Visual Perception
edge will distinguish the two. Psychologists and artists alike have been confused
about the difference between things and holes, surfaces and apertures. The
ﬁgure-ground phenomenon that so impressed the gestalt psychologists and that
is still taken to be a prototype of perception is misleading. A closed contour as
such in the optic array does not specify an object in the environment.
What speciﬁes the near edge of an opening in the ground, a hole or gap in the
surface of support? This is very important information for a terrestrial animal.
7. Gain of structure above a horizontal contour in the ambient array during approach
speciﬁes a brink in the surface of support. A brink is a drop-off in the ground, a step,
or the edge of a perch. It is the essential feature of the experiments on the visual
cliff that were described in Chapter 9 (for example, E. J. Gibson and Walk,
1960). It is depth downward at an occluding edge, and depending on the
amount of depth relative to the size of the animal, it affords stepping-down or
falling-off. The rat, chick, or human infant who sees its feet close to such an
occluding edge needs to take care. The experimental evidence suggests that the
changing occlusion at the edge, not the abrupt increase in the density of optical
texture, is the effective information for the animal.
This formula applies to a horizontal contour in the array coming from the
ground. What about a vertical contour in the array coming from a wall?
8. Gain of structure on one side of a vertical contour in the ambient array during
approach speciﬁes the occluding edge of a barrier, and the side on which gain occurs is the
side of the edge that affords passage. This is the edge of a house, the end of a wall,
or the vertical edge of a doorway, often loosely called a corner. On one side of
the edge the vista beyond is hidden, and on the other side it is revealed; on one
side there is potential collision, and on the other potential passage. The trunk
of a tree has two such curved edges not far apart. To “go around the corner” is
to reveal the surfaces of the new vista. Rats do it in mazes, and people do it in
cities. To ﬁnd one’s way in a cluttered environment is to go around a series of
occluding edges, and the problem is to choose the correct edges to go around
(see Figure 11.2).
What Speciﬁes Imminent Contact with a Surface?
In an early essay on the visual control of locomotion (Gibson, 1958), I wrote:
Approach to a solid surface is speciﬁed by a centrifugal ﬂow of the texture
of the optic array. Approach to an object is speciﬁed by a magniﬁcation
of the closed contour in the array corresponding to the edges of the
object. A uniform rate of approach is accompanied by an accelerated rate of
magniﬁcation. At the theoretical point where the eye touches the object,
the latter will intercept a visual angle of 180°. The magniﬁcation reaches
an explosive rate in the last moments before contact. This accelerated
expansion . . . speciﬁes imminent collision.
Locomotion and Manipulation 221
This was true enough as far as it went. I was thinking of the problem of how
a pilot lands on a ﬁeld or how a bee lands on a ﬂower. The explosive magniﬁcation,
the “looming” as I called it, has to be canceled if a “soft” landing is to be
achieved. I never thought of the entirely different problem of steering through
an opening. The optical information provided by various kinds of magniﬁcation
is evidently not as simple as I thought in 1958.
The complexities were not clariﬁed by the empirical studies of Schiff,
Caviness, and Gibson (1962) and Schiff (1965), who provided the optical
information for the approach of an object in space instead of the information for
approach to a surface in the environment. They displayed an expanding dark
silhouette in the center of a luminous translucent screen, as described in
Chapter 10. No one saw himself being transposed; everyone saw something
indeﬁnite coming toward them, as if it were in the sky. The display consisted
of an expanding single form, a shadow or silhouette, not the magnifying of a
nested structure of subordinate forms that characterizes approach to a real
surface. The magnifying of detail without limit was missing from the display.
9. The magniﬁcation of a nested structure in which progressively ﬁner details keep
emerging at the center speciﬁes approach of an observer to a surface in the environment.
This formula emphasizes the facets within the faces of a substantial surface,
such as that of an obstacle, an object, an animate object, or a surface of rest that
the observer might encounter. In order to achieve contact without collision,
the nested magniﬁcation must be made to cease at the appropriate level instead
of continuing to its limit. There seems to be an optimal degree of magniﬁcation
for contact with a surface, depending on what it affords. For food one
moves up to eating distance; for manipulating one moves up to reaching distance;
for print one moves up to reading distance.
What Speciﬁes the Beneﬁt or Injury that Lies Ahead?
Bishop Berkeley suggested in 1709 that the chief end of vision was for animals
“to foresee the beneﬁt or injury which is like to ensue upon the application of
their own bodies to this or that body which is at a distance.” What the philosopher
called foresight is what I call the perception of the affordance. To see at a
distance what the object affords on contact is “necessary for the preservation of
an animal.”
I differ from Bishop Berkeley in assuming that information is available in the
light to the animal for what an encounter with the object affords. But I agree
with him about the utility of vision.
10. Affordances for the individual upon encountering an object are speciﬁed in the
optic array from the object by invariants and invariant combinations. Tools, food, shelter,
mates, and amiable animals are distinguished from poisons, ﬁres, weapons, and hostile
animals by their shapes, colors, textures, and deformations. The positive and negative
affordances of things in the environment are what makes locomotion through
222 The Ecological Approach to Visual Perception
the medium such a fundamental kind of behavior for animals. Unlike a plant,
the animal can go to the beneﬁcial and stay away from the injurious. But it must
be able to perceive the affordances from afar. A rule for the visual control of
locomotion might be this: so move as to obtain beneﬁcial encounters with
objects and places and to prevent injurious encounters.
Rules for the Visual Control of Locomotion
I suggested at the beginning that behavior was controlled by information about
the world and the self conjointly. The information has now been described.
What about the control?
I asserted that behavior was controlled by rules. Surely, however, they are not
rules enforced by an authority. The rules are not commands from a brain; they
emerge from the animal-environment system. But the only way to describe
rules is in words, and a rule expressed in words is a command. I am faced with
a paradox. The rules for the control of locomotion will sound like commands,
although they are not intended to. I can only suggest that the reader should
interpret them as rules not formulated in words.
The rules that follow are for visual control, not muscular, articular, vestibular,
or cutaneous control. The visual system normally supersedes the haptic system
for locomotion and manipulation, as I tried to explain in The Senses Considered
as Perceptual Systems (Gibson, 1966b). This means that the rules for locomotion
will be the same for crawling on all fours, walking, running, or driving an
automobile. The particular muscles involved do not matter. Any group of
muscles will sufﬁce if it brings about the relation of the animal to its environment
stated in the rule.
Standing. The basic rule for a pedestrian animal is stand up; that is, keep the
feet in contact with a surface of support. It is also well to keep the oval boundaries
of the ﬁeld of view normal with the implicit horizon of the ambient array;
if the head is upright the rest of the body follows.
Starting, stopping, going back. To start, make the array ﬂow. To stop, cancel the ﬂow.
To go back, make the ﬂow reverse. According to the ﬁrst two formulas listed in
the previous pages, to cause outﬂow is to get closer and to cause inﬂow is to get
farther away.
Steering. To turn, shift the center of outﬂow from one patch in the optic array to another,
according to the the third and fourth formulas. Steering requires that openings
be distinguished from barriers, obstacles, and brinks. The rule is: To steer, keep the
center of outﬂow outside the patches of the array that specify barriers, obstacles, and brinks
and within a patch that speciﬁes an opening (sixth, seventh, and eighth formulas).
Following this rule will avert collisions and prevent falling off.
Approaching. To approach is to magnify a patch in the array, but magniﬁcation
is complicated (formulas two and six). There are many rules involving magniﬁcation.
Here are a few. To permit scrutiny, magnify the patch in the array to such a
Locomotion and Manipulation 223
degree that the details can be looked at. To manipulate something graspable, magnify the
patch to such a degree that the object is within reach. To bite something, magnify the patch
to such an angle that the mouth can grasp it. To kiss someone, magnify the face-form, if
the facial expression is amiable, so as almost to ﬁll the ﬁeld of view. (It is absolutely
essential for one to keep one’s eyes open so as to avoid collision. It is also wise
to learn to discriminate those subtle invariants that specify amiability.) To read
something, magnify the patch to such a degree that the letters become distinguishable. The
most general rule for approach is this: To realize the positive affordances of something,
magnify its optical structure to that degree necessary for the behavioral encounter.
Entering enclosures. An enclosure such as a burrow, cave, nest, or hut affords
various beneﬁts upon entry. It is a place of warmth, a shelter from rain and
wind, and a place for sleep. It is often a home, the place where mate and
offspring are. It is also a place of safety, a hiding place affording both concealment
from enemies and a barrier to their locomotion. An enclosure must have
an opening to permit entry, and the opening must be identiﬁed. The rule seems
to be as follows: to enter an enclosure, magnify the angle of its opening to 180° and
open up the vista. Make sure that there is gain of structure inside the contour and not loss
outside, or else you will collide with an obstacle (formulas six and nine).
Keeping a safe distance. The opposite of approach is retreat. Psychologists
have sometimes assumed that the alternative to approach is retreat. Kurt Lewin’s
theory of behavior, for example, was based on approach to an object with a
positive “valence” and retreat from an object with a negative “valence.” This
ﬁts with a theory of conﬂict between approach and retreat, and a compromise
between opposite tendencies. But it is wrong to assume that approach and
retreat are alternatives. There is no need to ﬂee from an obstacle, a barbed-wire
fence, the edge of a river, the edge of a cliff, or a ﬁre. The only need is to
maintain a safe distance, a “margin of safety,” since these things do not pursue
the observer. A ferocious tiger has a negative valence, but a cliff does not. The
rule is this, I think: To prevent an injurious encounter, keep the optical structure of
the surface from magnifying to the degree that speciﬁes an encounter (formulas two
and ten).
For moving predators and enemies, ﬂight is an appropriate form of action
since they can approach. The rule for ﬂight is, so move as to minify the dangerous
form and to make the surrounding optic array ﬂow inward. If, despite ﬂight, the form
magniﬁes, the enemy is catching up; if it miniﬁes, one is getting away. At the
predator’s point of observation, of course, the rule is opposite to that for the prey:
so move as to magnify the succulent form by making the surrounding array ﬂow outward
until it reaches the proper angular size for capturing.
Rules for the Visual Control of Manipulation
The rules for the visual control of the movements of the hands are more
complex than those for the control of locomotion. But the human infant who
224 The Ecological Approach to Visual Perception
watches these squirming protuberances into his ﬁeld of view is not formulating
rules and, in any case, complexity does not seem to cause trouble for the nervous
system. I am unable to formulate the rules in words except for a few easy cases.
Locomotor approach often terminates in reaching and grasping. Reaching is
an elongation of the arm-shape and a miniﬁcation of the ﬁve-pronged handshape
until contact occurs. If the object is hand-size, it is graspable; if too large
or too small, it is not. Children learn to see sizes in terms of prehension: they
see the span of their grasp and the diameter of a ball at the same time (Gibson,
1966b, ﬁg. 7.1, p. 119). Long before the child can discriminate one inch, or two,
or three, he can see the ﬁt of the object to the pincerlike action of the opposable
thumb. The child learns his scale of sizes as commensurate with his body, not
with a measuring stick.
The affordance of an elongated object for pounding and striking is easily
learned. The skill of hammering or striking a target requires visual control,
however. It involves what we vaguely call aiming. I will not try to state the rules
for aiming except to suggest that it entails a kind of centering or symmetricalizing
of a diminishing form on a ﬁxed form.
Throwing as such is easy. Simply cause the visual angle of the object you have
in your hand to shrink, and it will “zoom” in a highly interesting manner. You
have to let go, of course, and this is a matter of haptic control, not visual
control. Aimed throwing is much harder, as ballplayers know. It is a sort of
reciprocal of steered locomotion.
Tool-using in general is rule governed. The rule for pliers is analogous to that
for prehending, the tool being metaphorically an extension of the hand. The
use of a stick as a rake for getting a banana outside the cage was one of the
achievements of a famous chimpanzee (Köhler, 1925).
Knives, axes, and pointed objects afford the cutting and piercing of other
objects and surfaces, including other animals. But the manipulation must be
carefully controlled, for the observer’s own skin can be cut or pierced as well as
the other surface. The tool must be grasped by the handle, not the point; that
is, the rule for reaching and the rules for maintaining the margin of safety must
both be followed. Visual contact with one part of the surface is beneﬁcial but
with another part is injurious, and the “sharp” part is not always easy to discriminate.
The case is similar to that of walking along a cliff edge in this respect:
one must steer the movement so as to skirt the danger.
The uses of the hands are almost unlimited. And manipulation subserves
many other forms of behavior of which it is only a part, eating, drinking, transporting,
nursing, caressing, gesturing, and the acts of trace-making, depicting,
and writing, which will concern us in Part IV.
The point to remember is that the visual control of the hands is inseparably
connected with the visual perception of objects. The act of throwing complements
the perception of a throwable object. The transporting of things is part
and parcel of seeing them as portable or not.
Locomotion and Manipulation 225
Conclusion about manipulation. One thing should be evident. The movements
of the hands do not consist of responses to stimuli. Manipulation cannot be
understood in those terms. Is the only alternative to think of the hands as
instruments of the mind? Piaget, for example, sometimes seems to imply that
the hands are tools of a child’s intelligence. But this is like saying that the hand
is a tool of an inner child in more or less the same way that an object is a tool
for a child with hands. This is surely an error. The alternative is not a return to
mentalism. We should think of the hands as neither triggered nor commanded
but controlled.
Manipulation and the Perceiving of Interior Surfaces
Finally, it should be noted that a great deal of manipulation occurs for the sake of
perceiving hidden surfaces. I can think of three kinds of such manipulation:
opening up, uncovering, and taking apart. Each of these has an opposite, as one would
expect from the law of reversible occlusion: closing, covering, and putting together.
Opening and closing apply to the lids and covers of hollow objects and also to
drawers, compartments, cabinets, and other enclosures. Children are fascinated
by the act of opening so as to reveal the interior and closing so as to conceal it.
They then come to perceive the continuity between the inner and the outer
surfaces. The closed box and the covered pot are then seen to have an inside as
well as an outside.
Covering and uncovering apply to a cloth, or a child’s blanket, or to revealing
and concealing by an opaque substance, as in a sandbox. The movement of the
hand that conceals the object is not always so clearly the reverse of the movement
that reveals it as it is in the case of closing-opening, however. The
perceiving of hidden surfaces may well be more difﬁcult in this case.
Taking apart and putting together apply to an object composed of smaller objects,
that is, a composite that can be disassembled and assembled. There are toys of this
sort. Blocks that can be ﬁtted together make such a composite object. Taking apart
is usually a simpler act of manipulation than putting together. Children need to see
what is inside these compound objects, and it is only to be expected that they
should take them apart, or break them apart if need be. After such visual-manual
cooperation, they can perceive the interior surfaces of the object together with the
cracks, joins, and apertures that separate them. This is the way children come to
apprehend a mechanism such as a clock or an internal combustion engine.
Summary
Active locomotor behavior, as contrasted with passive transportation, is under
the continuous control of the observer. The dominant level of such control is
visual. But this could not occur without what I have called visual kinesthesis, the
awareness of movement or stasis, of starting or stopping, of approaching or
226 The Ecological Approach to Visual Perception
retreating, of going in one direction or another, and of the imminence of an
encounter. Such awarenesses are necessary for control.
Also necessary is an awareness of the affordance of the encounter that will
terminate the locomotor act and of the affordances of the openings and obstacles,
the brinks and barriers, and the corners on the way (actually the occluding edges).
When locomotion is thus visually controlled, it is regular without being a
chain of responses and is purposive without being commanded from within.
Manipulation, like active locomotion, is visually controlled. It is thus
dependent on an awareness of both the hands as such and the affordances for
handling. But its regularities are not so easy to formulate.
14
THE THEORY OF INFORMATION
PICKUP AND ITS CONSEQUENCES
In this book the traditional theories of perception have been abandoned.
The perennial doctrine that two-dimensional images are restored to threedimensional
reality by a process called depth perception will not do. Neither
will the doctrine that the images are transformed by the cues for distance and
slant so as to yield constancy of size and shape in the perception of objects. The
deep-seated notion of the retinal image as a still picture has been abandoned.
The simple assumption that perceptions of the world are caused by stimuli
from the world will not do. The more sophisticated assumption that perceptions
of the world are caused when sensations triggered by stimuli are supplemented by
memories will not do either. Not even the assumption that a sequence of stimuli
is converted into a phenomenal scene by memory will do. The very notion of
stimulation as typically composed of discrete stimuli has been abandoned.
The established theory that exteroception and proprioception arise when
exteroceptors and proprioceptors are stimulated will not do. The doctrine of
special channels of sensation corresponding to speciﬁc nerve bundles has been
abandoned.
The belief of empiricists that the perceived meanings and values of things
are supplied from the past experience of the observer will not do. But even
worse is the belief of nativists that meanings and values are supplied from the
past experience of the race by way of innate ideas. The theory that meaning is
attached to experience or imposed on it has been abandoned.
Not even the current theory that the inputs of the sensory channels are subject
to “cognitive processing” will do. The inputs are described in terms of information
theory, but the processes are described in terms of old-fashioned mental
acts: recognition, interpretation, inference, concepts, ideas, and storage and
retrieval of ideas. These are still the operations of the mind upon the deliverances
228 The Ecological Approach to Visual Perception
of the senses, and there are too many perplexities entailed in this theory. It will
not do, and the approach should be abandoned.
What sort of theory, then, will explain perception? Nothing less than one
based on the pickup of information. To this theory, even in its undeveloped
state, we should now turn.
Let us remember once again that it is the perception of the environment that
we wish to explain. If we were content to explain only the perception of forms
or pictures on a surface, of nonsense ﬁgures to which meanings must be
attached, of discrete stimuli imposed on an observer willy-nilly, in short, the
items most often presented to an observer in the laboratory, the traditional
theories might prove to be adequate and would not have to be abandoned. But
we should not be content with that limited aim. It leaves out of account the
eventful world and the perceiver’s awareness of being in the world. The laboratory
does not have to be limited to simple stimuli, so-called. The experiments
reported in Chapters 9 and 10 showed that information can be displayed.
What is New About the Pickup of Information?
The theory of information pickup differs radically from the traditional theories
of perception. First, it involves a new notion of perception, not just a new
theory of the process. Second, it involves a new assumption about what there is
to be perceived. Third, it involves a new conception of the information for
perception, with two kinds always available, one about the environment and
another about the self. Fourth, it requires the new assumption of perceptual
systems with overlapping functions, each having outputs to adjustable organs as
well as inputs from organs. We are especially concerned with vision, but none
of the systems, listening, touching, smelling, or tasting, is a channel of sense.
Finally, ﬁfth, optical information pickup entails an activity of the system not
heretofore imagined by any visual scientist, the concurrent registering of both
persistence and change in the ﬂow of structured stimulation. This is the crux of
the theory but the hardest part to explicate, because it can be phrased in
different ways and a terminology has to be invented.
Consider these ﬁve novelties in order, ending with the problem of detecting
variants and invariants or change and nonchange.
A Redeﬁnition of Perception
Perceiving is an achievement of the individual, not an appearance in the theater
of his consciousness. It is a keeping-in-touch with the world, an experiencing of
things rather than a having of experiences. It involves awareness-of instead of
just awareness. It may be awareness of something in the environment or
something in the observer or both at once, but there is no content of awareness
independent of that of which one is aware. This is close to the act psychology of
The Theory of Information Pickup and its Consequences 229
the nineteenth century except that perception is not a mental act. Neither is it a
bodily act. Perceiving is a psychosomatic act, not of the mind or of the body but
of a living observer.
The act of picking up information, moreover, is a continuous act, an activity
that is ceaseless and unbroken. The sea of energy in which we live ﬂows and
changes without sharp breaks. Even the tiny fraction of this energy that affects
the receptors in the eyes, ears, nose, mouth, and skin is a ﬂux, not a sequence.
The exploring, orienting, and adjusting of these organs sink to a minimum
during sleep but do not stop dead. Hence, perceiving is a stream, and William
James’s description of the stream of consciousness (1890, Ch. 9) applies to it.
Discrete percepts, like discrete ideas, are “as mythical as the Jack of Spades.”
The continuous act of perceiving involves the coperceiving of the self. At
least, that is one way to put it. The very term perception must be redeﬁned to
allow for this fact, and the word proprioception must be given a different meaning
than it was given by Sherrington.
A New Assertion About What is Perceived
My description of the environment (Chapters 1–3) and of the changes that can
occur in it (Chapter 6) implies that places, attached objects, objects, and
substances are what are mainly perceived, together with events, which are
changes of these things. To see these things is to perceive what they afford. This
is very different from the accepted categories of what there is to perceive as
described in the textbooks. Color, form, location, space, time, and motion—
these are the chapter headings that have been handed down through the
centuries, but they are not what is perceived.
Places
A place is one of many adjacent places that make up the habitat and, beyond that,
the whole environment. But smaller places are nested within larger places.
They do not have boundaries, unless artiﬁcial boundaries are imposed by
surveyors (my piece of land, my town, my country, my state). A place at one
level is what you can see from here or hereabouts, and locomotion consists of
going from place to place in this sense (Chapter 11). A very important kind of
learning for animals and children is place-learning—learning the affordances of
places and learning to distinguish among them—and way-ﬁnding, which
culminate in the state of being oriented to the whole habitat and knowing
where one is in the environment.
A place persists in some respects and changes in others. In one respect, it
cannot be changed at all—in its location relative to other places. A place cannot
be displaced like an object. That is, the adjacent order of places cannot be
permuted; they cannot be shufﬂed. The sleeping places, eating places, meeting
230 The Ecological Approach to Visual Perception
places, hiding places, and falling-off places of the habitat are immobile. Placelearning
is therefore different from other kinds.
Attached Objects
I deﬁned an object in Chapter 3 as a substance partially or wholly surrounded by
the medium. An object attached to a place is only partly surrounded. It is a
protuberance. It cannot be displaced without becoming detached. Nevertheless,
it has a surface and enough of a natural boundary to constitute a unit. Attached
objects can thus be counted. Animals and children learn what such objects are
good for and how to distinguish them. But they cannot be separated from the
places where they are found.
Detached Objects
A fully detached object can be displaced or, in some cases, can displace itself.
Learning to perceive it thus has a different character from learning to perceive
places and attached objects. Its affordances are different. It can be put side by
side with another object and compared. It can therefore be grouped or classed
by the manipulation of sorting. Such objects when grouped can be rearranged,
that is, permuted. And this means not only that they can be counted but that an
abstract number can be assigned to the group.
It is probably harder for a child to perceive “same object in a different place”
than it is to perceive “same object in the same place.” The former requires that
the information for persistence-despite-displacement should have been noticed,
whereas the latter does not.
Inanimate detached objects, rigid or nonrigid, natural or manufactured, can
be said to have features that distinguish them. The features are probably not
denumerable, unlike the objects themselves. But if they are compounded to
specify affordances, as I argued they must be, only the relevant compounds
need to be distinguished. So when it comes to the natural, nonrigid, animate
objects of the world whose dimensions of difference are overwhelmingly rich
and complex, we pay attention only to what the animal or person affords
(Chapter 8).
Persisting Substances
A substance is that of which places and objects are composed. It can be vaporous,
liquid, plastic, viscous, or rigid, that is, increasingly “substantial.” A substance,
together with what it affords, is fairly well speciﬁed by the color and texture of
its surface. Smoke, milk, clay, bread, and wood are polymorphic in layout but
invariant in color-texture. Substances, of course, can be smelled and tasted and
palpated as well as seen.
The Theory of Information Pickup and its Consequences 231
The animal or child who begins to perceive substances, therefore, does so in
a different way than one who begins to perceive places, attached objects, and
detached objects. Substances are formless and cannot be counted. The number
of substances, natural compositions, or mixtures is not ﬁxed. (The number of
chemical elements is ﬁxed, but that is a different matter.) We discriminate
among surface colors and textures, but we cannot group them as we do detached
objects and we cannot order them as we do places.
We also, of course, perceive changes in otherwise persisting substances, the
ripening of fruit, and the results of boiling and baking, or of mixing and
hardening. But these are a kind of event.
Events
As I used the term, an event is any change of a substance, place, or object, chemical,
mechanical, or biophysical. The change may be slow or fast, reversible or
nonreversible, repeating or nonrepeating. Events include what happens to
objects in general, plus what the animate objects make happen. Events are nested
within superordinate events. The motion of a detached object is not the prototype
of an event that we have been led to think it was. Events of different sorts
are perceived as such and are not, surely, reducible to elementary motions.
The Information for Perception
Information, as the term is used in this book (but not in other books), refers to
speciﬁcation of the observer’s environment, not to speciﬁcation of the observer’s
receptors or sense organs. The qualities of objects are speciﬁed by information;
the qualities of the receptors and nerves are speciﬁed by sensations.
Information about the world cuts right across the qualities of sense.
The term information cannot have its familiar dictionary meaning of knowledge
communicated to a receiver. This is unfortunate, and I would use another term
if I could. The only recourse is to ask the reader to remember that picking up
information is not to be thought of as a case of communicating. The world does
not speak to the observer. Animals and humans communicate with cries,
gestures, speech, pictures, writing, and television, but we cannot hope to
understand perception in terms of these channels; it is quite the other way
around. Words and pictures convey information, carry it, or transmit it, but the
information in the sea of energy around each of us, luminous or mechanical or
chemical energy, is not conveyed. It is simply there. The assumption that
information can be transmitted and the assumption that it can be stored are
appropriate for the theory of communication, not for the theory of perception.
The vast area of speculation about the so-called media of communication had
a certain discipline imposed on it some years ago by a mathematical theory of
communication (Shannon and Weaver, 1949). A useful measure of information
232 The Ecological Approach to Visual Perception
transmitted was formulated, in terms of “bits.” A sender and receiver, a channel,
and a ﬁnite number of possible signals were assumed. The result was a genuine
discipline of communications engineering. But, although psychologists promptly
tried to apply it to the senses and neuropsychologists began thinking of nerve
impulses in terms of bits and the brain in terms of a computer, the applications
did not work. Shannon’s concept of information applies to telephone hookups
and radio broadcasting in elegant ways but not, I think, to the ﬁrsthand perception
of being in-the-world, to what the baby gets when ﬁrst it opens its eyes. The
information for perception, unhappily, cannot be deﬁned and measured as
Claude Shannon’s information can be.
The information in ambient light, along with sound, odor, touches, and
natural chemicals, is inexhaustible. A perceiver can keep on noticing facts about
the world she lives in to the end of her life without ever reaching a limit. There
is no threshold for information comparable to a stimulus threshold. Information
is not lost to the environment when gained by the individual; it is not conserved
like energy.
Information is not speciﬁc to the banks of photoreceptors, mechanoreceptors,
and chemoreceptors that lie within the sense organs. Sensations are
speciﬁc to receptors and thus, normally, to the kinds of stimulus energy that
touch them off. But information is not energy-speciﬁc. Stimuli are not always
imposed on a passive subject. In life one obtains stimulation in order to extract
the information (Gibson, 1966b, Ch. 2). The information can be the same,
despite a radical change in the stimulation obtained.
Finally, a concept of information is required that admits of the possibility of
illusion. Illusions are a theoretical perplexity in any approach to the study of
perception. Is information always valid and illusion simply a failure to pick it up?
Or is the information picked up sometimes impoverished, masked, ambiguous,
equivocal, contradictory, even false? The puzzle is especially critical in vision.
In Chapter 14 of The Senses Considered as Perceptual Systems (Gibson, 1966b)
and again in this book I have tried to come to terms with the problem of
misperception. I am only sure of this: it is not one problem but a complex of
different problems. Consider, ﬁrst, the mirage of palm trees in the desert sky,
or the straight stick that looks bent because it is partly immersed in water.
These illusions, together with the illusion of Narcissus, arise from the regular
reﬂection or refraction of light, that is, from exceptions to the ecological optics
of the scatter-reﬂecting surface and the perfectly homogeneous medium. Then
consider, second, the misperception in the case of the shark under the calm
water or the electric shock hidden in the radio cabinet. Failure to perceive the
danger is not then blamed on the perceiver. Consider, third, the sheet of glass
mistaken for an open doorway or the horizontal sheet of glass (the optical cliff)
mistaken for a void. A fourth case is the room composed of trapezoidal surfaces
or the trapezoidal window, which look normally rectangular so long as the
observer does not open both eyes and walk around. Optical misinformation
The Theory of Information Pickup and its Consequences 233
enters into each of these cases in a different way. But in the last analysis, are they
explained by misinformation? Or is it a matter of failure to pick up all the available
information, the inexhaustible reservoir that lies open to further scrutiny?
The misperceiving of affordances is a serious matter. As I noted in Chapter 8,
a wildcat may look like a cat. (But does he look just like a cat?) A malevolent man
may act like a benevolent one. (But does he exactly?) The line between the
pickup of misinformation and the failure to pick up information is hard to draw.
Consider the human habit of picture-making, which I take to be the devising
and displaying of optical information for perception by others. It is thus a means
of communication, giving rise to mediated apprehension, but it is more like
direct pickup than word-making is. Depiction and its consequences are deferred
until later, but it can be pointed out here that picture-makers have been experimenting
on us for centuries with artiﬁcial displays of information in a special
form. They enrich or impoverish it, mask or clarify it, ambiguate or disambiguate
it. They often try to produce a discrepancy of information, an equivocation
or contradiction, in the same display. Painters invented the cues for depth
in the ﬁrst place, and psychologists looked at their paintings and began to talk
about cues. The notions of counterbalanced cues, of ﬁgure-ground reversals, of
equivocal perspectives, of different perspectives on the same object, of “impossible”
objects—all these come from artists who were simply experimenting
with frozen optical information.
An important fact to be noted about any pictorial display of optical information
is that, in contrast with the inexhaustible reservoir of information in an
illuminated medium, it cannot be looked at close up. Information to specify the
display as such, the canvas, the surface, the screen, can always be picked up by
an observer who walks around and looks closely.
The Concept of a Perceptual System
The theory of information pickup requires perceptual systems, not senses.
Some years ago I tried to prove that a perceptual system was radically different
from a sense (Gibson, 1966b), the one being active and the other passive. People
said, “Well, what I mean by a sense is an active sense.” But it turned out that they
still meant the passive inputs of a sensory nerve, the activity being what occurs
in the brain when the inputs get there. That was not what I meant by a perceptual
system. I meant the activities of looking, listening, touching, tasting, or
snifﬁng. People then said, “Well, but those are responses to sights, sounds,
touches, tastes, or smells, that is, motor acts resulting from sensory inputs. What
you call a perceptual system is nothing but a case of feedback.” I was discouraged.
People did not understand.
I shall here make another attempt to show that the senses considered as special
senses cannot be reconciled with the senses considered as perceptual systems.
The ﬁve perceptual systems correspond to ﬁve modes of overt attention. They
234 The Ecological Approach to Visual Perception
have overlapping functions, and they are all more or less subordinated to an
overall orienting system. A system has organs, whereas a sense has receptors. A
system can orient, explore, investigate, adjust, optimize, resonate, extract, and
come to an equilibrium, whereas a sense cannot. The characteristic activities of
the visual system have been described in Chapter 12 of this book. The characteristic
activities of the auditory system, the haptic system, and the two related
parts of what I called the “chemical value system” were described in Chapters 5–8
of my earlier book (Gibson, 1966b). Five fundamental differences between a
sense and a perceptual system are given below.
1. A special sense is deﬁned by a bank of receptors or receptive units that
are connected with a so-called projection center in the brain. Local stimuli at
the sensory surface will cause local ﬁring of neurons in the center. The adjustments
of the organ in which the receptors are incorporated are not included
within the deﬁnition of a sense.
A perceptual system is deﬁned by an organ and its adjustments at a given level
of functioning, subordinate or superordinate. At any level, the incoming and
outgoing nerve ﬁbers are considered together so as to make a continuous loop.
The organs of the visual system, for example, from lower to higher are
roughly as follows. First, the lens, pupil, chamber, and retina comprise an organ.
Second, the eye with its muscles in the orbit comprise an organ that is both
stabilized and mobile. Third, the two eyes in the head comprise a binocular
organ. Fourth, the eyes in a mobile head that can turn comprise an organ for the
pickup of ambient information. Fifth, the eyes in a head on a body constitute a
superordinate organ for information pickup over paths of locomotion. The
adjustments of accommodation, intensity modulation, and dark adaptation go
with the ﬁrst level. The movements of compensation, ﬁxation, and scanning go
with the second level. The movements of vergence and the pickup of disparity
go with the third level. The movements of the head, and of the body as a whole,
go with the fourth and ﬁfth levels. All of them serve the pickup of information.
2. In the case of a special sense, the receptors can only receive stimuli, passively,
whereas in the case of a perceptual system the input-output loop can be
supposed to obtain information, actively. Even when the theory of the special
senses is liberalized by the modern hypothesis of receptive units, the latter are
supposed to be triggered by complex stimuli or modulated in some passive fashion.
3. The inputs of a special sense constitute a repertory of innate sensations,
whereas the achievements of a perceptual system are susceptible to maturation
and learning. Sensations of one modality can be combined with those of
another in accordance with the laws of association; they can be organized or
fused or supplemented or selected, but no new sensations can be learned. The
information that is picked up, on the other hand, becomes more and more
subtle, elaborate, and precise with practice. One can keep on learning to
perceive as long as life goes on.
The Theory of Information Pickup and its Consequences 235
4. The inputs of the special senses have the qualities of the receptors being
stimulated, whereas the achievements of the perceptual systems are speciﬁc to
the qualities of things in the world, especially their affordances. The recognition
of this limitation of the senses was forced upon us by Johannes Müller with
his doctrine of speciﬁc “nerve energies.” He understood clearly, if reluctantly,
the implication that, because we can never know the external causes of our
sensations, we cannot know the outer world. Strenuous efforts have to be made
if one is to avoid this shocking conclusion. Helmholtz argued that we must
deduce the causes of our sensations because we cannot detect them. The hypothesis
that sensations provide clues or cues for perception of the world is similar.
The popular formula that we can interpret sensory signals is a variant of it. But
it seems to me that all such arguments come down to this: we can perceive the
world only if we already know what there is to be perceived. And that, of
course, is circular. I shall come back to this point again.
The alternative is to assume that sensations triggered by light, sound, pressure,
and chemicals are merely incidental, that information is available to a
perceptual system, and that the qualities of the world in relation to the needs of
the observer are experienced directly.
5. In the case of a special sense the process of attention occurs at centers
within the nervous system, whereas in the case of a perceptual system attention
pervades the whole input-output loop. In the ﬁrst case attention is a consciousness
that can be focused; in the second case it is a skill that can be educated. In
the ﬁrst case physiological metaphors are used, such as the ﬁltering of nervous
impulses or the switching of impulses from one path to another. In the second
case the metaphors used can be terms such as resonating, extracting, optimizing, or
symmetricalizing and such acts as orienting, exploring, investigating, or adjusting.
I suggested in Chapter 12 that a normal act of visual attention consists of
scanning a whole feature of the ambient array, not of ﬁxating a single detail of
the array. We are tempted to think of attention as strictly a narrowing-down
and holding-still, but actually this is rare. The invariants of structure in an
optic array that constitute information are more likely to be gradients than
small details, and they are scanned over wide angles.
The Registering of Both Persistence and Change
The theory of information pickup requires that the visual system be able to
detect both persistence and change—the persistence of places, objects, and
substances along with whatever changes they undergo. Everything in the world
persists in some respects and changes in some respects. So also does the observer
himself. And some things persist for long intervals, others for short.
The perceiving of persistence and change (instead of color, form, space,
time, and motion) can be stated in various ways. We can say that the perceiver
separates the change from the nonchange, notices what stays the same and what
236 The Ecological Approach to Visual Perception
does not, or sees the continuing identity of things along with the events in
which they participate. The question, of course, is how he does so. What is the
information for persistence and change? The answer must be of this sort: The
perceiver extracts the invariants of structure from the ﬂux of stimulation while
still noticing the ﬂux. For the visual system in particular, he tunes in on the
invariant structure of the ambient optic array that underlies the changing
perspective structure caused by his movements.
The hypothesis that invariance under optical transformation constitutes
information for the perception of a rigid persisting object goes back to the movingshadow
experiment (Gibson and Gibson, 1957). The outcome of that experiment
was paradoxical; it seemed at the time that a changing form elicited the perception
of a constant form with a changing slant. The solution was to postulate invariants
of optical structure for the persisting object, “formless” invariants, and a particular
disturbance of optical structure for the motion of the object, a perspective transformation.
Separate terms needed to be devised for physical motions and for the
optical motions that speciﬁed them, for events in the world and for events in the
array, for geometry did not provide the terms. Similarly, different terms need to
be invented to describe invariants of the changing world and invariants of the
changing array; the geometrical word form will not do. Perhaps the best policy is
to use the terms persistence and change to refer to the environment but preservation
and disturbance of structure to refer to the optic array.
The stimulus-sequence theory of perception, based on a succession of
discrete eye ﬁxations, can assume only that the way to apprehend persistence is
by an act of comparison and judgment. The perception of what-it-is-now is
compared with the memory of what-it-was-then, and they are judged same.
The continuous pickup theory of perception can assume that the apprehension
of persistence is a simple act of invariance detection. Similarly, the snapshot
theory must assume that the way to apprehend change is to compare what-itis-now
with what-it-was-then and judge different, whereas the pickup theory
can assume an awareness of transformation. The congruence of the array with
itself or the disparity of the array with itself, as the case may be, is picked up.
The perception of the persisting identity of things is fundamental to other
kinds of perception. Consider an example, the persisting identity of another
person. How does a child come to apprehend the identity of the mother? You
might say that when the mother-ﬁgure, or the face, is continually ﬁxated by the
child the persistence of the sensation is supported by the continuing stimulus.
So it is when the child clings to the mother. But what if the mother-ﬁgure is
scanned? What if the ﬁgure leaves and returns to the ﬁeld of view? What if the
ﬁgure goes away and comes back? What is perceived when it emerges from the
distance or from darkness, when its back is turned, when its clothing is changed,
when its emotional state is altered, when it comes back into sight after a long
interval? In short, how is it that the phenomenal identity of a person agrees so
well with the biological identity, despite all the vicissitudes of the ﬁgure in the
optic array and all the events in which the person participates?
The Theory of Information Pickup and its Consequences 237
The same questions can be asked about inanimate objects, attached objects,
places, and substances. The features of a person are invariant to a considerable
degree (the eyes, nose, mouth, style of gesture, and voice). But so are the
analogous features of other things, the child’s blanket, the kitchen stove, the
bedroom, and the bread on the table. All have to be identiﬁed as continuing, as
persisting, as maintaining existence. And this is not explained by the constructing
of a concept for each.
We are accustomed to assuming that successive stimuli from the same entity,
sensory encounters with it, are united by an act of recognition. We have
assumed that perception ceases and memory takes over when sensation stops.
Hence, every fresh glimpse of anything requires the act of linking it up with
the memories of that thing instead of some other thing. The judgment, “I have
seen this before,” is required for the apprehension of “same thing,” even when
the observer has only turned away, or has only glanced away for an instant. The
classical theory of sense perception is reduced to an absurdity by this requirement.
The alternative is to accept the theory of invariance detection.
THE EFFECT OF PERSISTING STIMULATION
ON PERCEPTION
We have assumed that perception stops when sensation stops and that
sensation stops when stimulation stops, or very soon thereafter. Hence, a
persisting stimulus is required for the perception of a persisting object. The
fact is, however, that a truly persisting stimulus on the retina or the skin
speciﬁes only that the observer does not or cannot move his eye or his limb,
and the sense perception soon fades out by sensory adaptation (Chapter 4).
The persistence of an object is speciﬁed by invariants of structure, not by the
persistence of stimulation.
The seeing of persistence considered as the picking up of invariants under
change resolves an old puzzle: the phenomenal identity of the spots of a
retinal pattern when the image is transposed over the retina stroboscopically.
The experiments of Josef Ternus ﬁrst made this puzzle evident. See Gibson
(1950, pp. 56 ff.) for a discussion and references.
I used to think that the aftereffects of persisting stimulation of the retina
obtained by the prolonged ﬁxation of a display could be very revealing.
Besides ordinary afterimages there are all sorts of perceptual aftereffects, some
of which I discovered. But I no longer believe that experiments on so-called
perceptual adaptation are revealing, and I have given up theorizing about
them. The aftereffects of prolonged scrutiny are of many sorts. Until we know
more about information pickup, this ﬁeld of investigation will be incoherent.
238 The Ecological Approach to Visual Perception
The quality of familiarity that can go with the perception of a place, object,
or person, as distinguished from the quality of unfamiliarity, is a fact of experience.
But is familiarity a result of the percept making contact with the traces of
past percepts of the same thing? Is unfamiliarity a result of not making such
contact? I think not. There is a circularity in the reasoning, and it is a bad theory.
The quality of familiarity simply accompanies the perception of persistence.
The perception of the persisting identity of places and objects is more fundamental
than the perception of the differences among them. We are told that to
perceive something is to categorize it, to distinguish it from the other types of
things that it might have been. The essence of perceiving is discriminating.
Things differ among themselves, along dimensions of difference. But this leaves
out of account the simple fact that the substance, place, object, person, or whatever
has to last long enough to be distinguished from other substances, places,
objects, or persons. The detecting of the invariant features of a persisting thing
should not be confused with the detecting of the invariant features that make
different things similar. Invariants over time and invariants over entities are not
grasped in the same way.
In the case of the persisting thing, I suggest, the perceptual system simply
extracts the invariants from the ﬂowing array; it resonates to the invariant structure
or is attuned to it. In the case of substantially distinct things, I venture, the
perceptual system must abstract the invariants. The former process seems to be
simpler than the latter, more nearly automatic. The latter process has been
interpreted to imply an intellectual act of lifting out something that is mental
from a collection of objects that are physical, of forming an abstract concept
from concrete percepts, but that is very dubious. Abstraction is invariance
detection across objects. But the invariant is only a similarity, not a persistence.
Summary of the Theory of Pickup
According to the theory being proposed, perceiving is a registering of certain
deﬁnite dimensions of invariance in the stimulus ﬂux together with deﬁnite
parameters of disturbance. The invariants are invariants of structure, and the
disturbances are disturbances of structure. The structure, for vision, is that of
the ambient optic array.
The invariants specify the persistence of the environment and of oneself.
The disturbances specify the changes in the environment and of oneself. A
perceiver is aware of her existence in a persisting environment and is also aware
of her movements relative to the environment, along with the motions of
objects and nonrigid surfaces relative to the environment. The term awareness is
used to imply a direct pickup of the information, not necessarily to imply
consciousness.
There are many dimensions of invariance in an ambient optic array over
time, that is, for paths of observation. One invariant, for example, is caused by
The Theory of Information Pickup and its Consequences 239
the occluding edge of the nose, and it speciﬁes the self. Another is the gradient
of optical texture caused by the material texture of the substratum, and it speciﬁes
the basic environment. Equally, there are many parameters of disturbance
of an ambient optic array. One, for example, is caused by the sweeping of the
nose over the ambient optic array, and it speciﬁes head turning. Another is the
deletion and accretion of texture at the edges of a form in the array, and it
speciﬁes the motion of an object over the ground.
For different kinds of events in the world there are different parameters of
optical disturbance, not only accretion-deletion but also polar outﬂow-inﬂow,
compression, transformation, substitution, and others. Hence, the same object
can be seen undergoing different events, and different objects can be seen
undergoing the same event. For example, an apple may ripen, fall, collide, roll,
or be eaten, and eating may happen to an apple, carrot, egg, biscuit, or lamb
chop. If the parameter of optical disturbance is distinguished, the event will be
perceived. Note how radically different this is from saying that if stimulusevent
A is invariably followed by stimulus-event B we will come to expect B
whenever we experience A. The latter is classical association theory (or conditioning
theory, or expectancy theory). It rests on the stimulus-sequence
doctrine. It implies that falling, colliding, rolling, or eating are not units but
sequences. It implies, with David Hume, that even if B has followed A a thousand
times there is no certainty that it will follow A in the future. An event is
only known by a conjunction of atomic sensations, a contingency. If this recurrent
sequence is experienced again and again, the observer will begin to anticipate,
or have faith, or learn by induction, but that is the best he can do.
The process of pickup is postulated to depend on the input-output loop of a
perceptual system. For this reason, the information that is picked up cannot be
the familiar kind that is transmitted from one person to another and that can be
stored. According to pickup theory, information does not have to be stored in
memory because it is always available.
The process of pickup is postulated to be very susceptible to development
and learning. The opportunities for educating attention, for exploring and
adjusting, for extracting and abstracting are unlimited. The increasing capacity
of a perceptual system to pick up information, however, does not in itself
constitute information. The ability to perceive does not imply, necessarily, the
having of an idea of what can be perceived. The having of ideas is a fact, but it
is not a prerequisite of perceiving. Perhaps it is a kind of extended perceiving.
The Traditional Theories of Perception: Input Processing
The theory of information pickup purports to be an alternative to the traditional
theories of perception. It differs from all of them, I venture to suggest, in
rejecting the assumption that perception is the processing of inputs. Inputs mean
sensory or afferent nerve impulses to the brain.
240 The Ecological Approach to Visual Perception
Adherents to the traditional theories of perception have recently been
making the claim that what they assume is the processing of information in a
modern sense of the term, not sensations, and that therefore they are not bound
by the traditional theories of perception. But it seems to me that all they are
doing is climbing on the latest bandwagon, the computer bandwagon, without
reappraising the traditional assumption that perceiving is the processing of
inputs. I refuse to let them pre-empt the term information. As I use the term, it
is not something that has to be processed. The inputs of the receptors have to
be processed, of course, because they in themselves do not specify anything
more than the anatomical units that are triggered.
All kinds of metaphors have been suggested to describe the ways in which
sensory inputs are processed to yield perceptions. It is supposed that sensation
occurs ﬁrst, perception occurs next, and knowledge occurs last, a progression
from the lower to the higher mental processes. One process is the ﬁltering of
sensory inputs. Another is the organizing of sensory inputs, the grouping of
elements into a spatial pattern. The integrating of elements into a temporal
pattern may or may not be included in the organizing process. After that, the
processes become highly speculative. Some theorists propose mental operations.
Others argue for semilogical processes or problem-solving. Many theorists
are in favor of a process analogous to the decoding of signals. All theorists
seem to agree that past experience is brought to bear on the sensory inputs,
which means that memories are somehow applied to them. Apart from ﬁltering
and organizing, the processes suggested are cognitive. Consider some of them.
Mental Operations on the Sensory Inputs
The a priori categories of understanding possessed by the perceived, according
to Kant
The perceiver’s presuppositions about what is being perceived
Innate ideas about the world
Semilogical Operations on the Sensory Inputs
Unconscious inferences about the outer causes of the sensory inputs, according
to Helmholtz (the outer world is deduced)
Estimates of the probable character of the “distant” objects based on the
“proximal” stimuli, according to Egon Brunswik (1956), said to be a quasirational,
not a fully rational, process
Decoding Operations on the Sensory Inputs
The interpreting of the inputs considered as signals (a very popular analogy
with many variants)
The Theory of Information Pickup and its Consequences 241
The decoding of sensory messages
The utilizing of sensory cues
The understanding of signs, or indicators, or even clues, in the manner of a
police detective
The Application of Memories to the Sensory Inputs
The “accrual” of a context of memory images and feelings to the core of sensations,
according to E. B. Titchener’s theory of perception (1924).
This last hypothetical process is perhaps the most widely accepted of all, and the
most elaborated. Perceptual learning is supposed to be a matter of enriching the
input, not of differentiating the information (Gibson and Gibson, 1955). But the
process of combining memories with inputs turns out to be not at all simple when
analyzed. The appropriate memories have to be retrieved from storage, that is,
aroused or summoned; an image does not simply accrue. The sensory input must
fuse in some fashion with the stored images; or the sensory input is assimilated to
a composite memory image, or, if this will not do, it is said to be assimilated to a
class, a type, a schema, or a concept. Each new sensory input must be categorized—
assigned to its class, matched to its type, ﬁtted to its schema, and so on. Note that
categories cannot become established until enough items have been classiﬁed but
that items cannot be classiﬁed until categories have been established. It is this difﬁculty,
for one, that compels some theorists to suppose that classiﬁcation is a priori
and that people and animals have innate or instinctive knowledge of the world.
The error lies, it seems to me, in assuming that either innate ideas or acquired
ideas must be applied to bare sensory inputs for perceiving to occur. The fallacy
is to assume that because inputs convey no knowledge they can somehow be
made to yield knowledge by “processing” them. Knowledge of the world must
come from somewhere; the debate is over whether it comes from stored knowledge,
from innate knowledge, or from reason. But all three doctrines beg the
question. Knowledge of the world cannot be explained by supposing that
knowledge of the world already exists. All forms of cognitive processing imply
cognition so as to account for cognition.
FIGURE 14.1 The commonly supposed sequence of stages in the visual perceiving of
an object.
242 The Ecological Approach to Visual Perception
All this should be treated as ancient history. Knowledge of the environment,
surely, develops as perception develops, extends as the observers travel, gets
ﬁner as they learn to scrutinize, gets longer as they apprehend more events, gets
fuller as they see more objects, and gets richer as they notice more affordances.
Knowledge of this sort does not “come from” anywhere; it is got by looking,
along with listening, feeling, smelling, and tasting. The child also, of course,
begins to acquire knowledge that comes from parents, teachers, pictures, and
books. But this is a different kind of knowledge.
The False Dichotomy between Present and Past Experience
The division between present experience and past experience may seem to be
self-evident. How could anyone deny it? Yet it is denied in supposing that we
can experience both change and nonchange. The difference between present
and past blurs, and the clarity of the distinction slips away. The stream of experience
does not consist of an instantaneous present and a linear past receding into
the distance; it is not a “traveling razor’s edge” dividing the past from the future.
Perhaps the present has a certain duration. If so, it should be possible to ﬁnd out
when perceiving stops and remembering begins. But it has not been possible.
There are attempts to talk about a “conscious” present, or a “specious” present,
or a “span” of present perception, or a span of “immediate memory,” but they
all founder on the simple fact that there is no dividing line between the present
and the past, between perceiving and remembering. A special sense impression
clearly ceases when the sensory excitation ends, but a perception does not. It
does not become a memory after a certain length of time. A perception, in fact,
does not have an end. Perceiving goes on.
Perhaps the force of the dichotomy between present and past experience
comes from language, where we are not allowed to say anything intermediate
between “I see you” and “I saw you” or “I am seeing you” and “I was seeing
you.” Verbs can take the present tense or the past tense. We have no words to
describe my continuing awareness of you, whether you are in sight or out of
sight. Language is categorical. Because we are led to separate the present from the
past, we ﬁnd ourselves involved in what I have called the “muddle of memory”
(Gibson, 1966a). We think that the past ceases to exist unless it is “preserved” in
memory. We assume that memory is the bridge between the past and the present.
We assume that memories accumulate and are stored somewhere; that they are
images, or pictures, or representations of the past; or that memory is actually
physiological, not mental, consisting of engrams or traces; or that it actually
consists of neural connections, not engrams; that memory is the basis of all
learning; that memory is the basis of habit; that memories live on in the unconscious;
that heredity is a form of memory; that cultural heredity is another form
of memory; that any effect of the past on the present is memory, including hysteresis.
If we cannot do any better than this, we should stop using the word.
The Theory of Information Pickup and its Consequences 243
The traditional theories of perception take it for granted that what we
see now, present experience, is the sensory basis of our perception of the
environment and that what we have seen up to now, past experience, is added to
it. We can only understand the present in terms of the past. But what we see
now (when it is carefully analyzed) turns out to be at most a peculiar set of
surfaces that happen to come within the ﬁeld of view and face the point of
observation (Chapter 11). It does not comprise what we see. It could not
possibly be the basis of our perception of the environment. What we see now
refers to the self, not the environment. The perspective appearance of the world
at a given moment of time is simply what speciﬁes to the observer where he is
at that moment. The perceptual process does not begin with this peculiar
projection, this momentary pattern. The perceiving of the world begins with
the pickup of invariants.
Evidently the theory of information pickup does not need memory. It does
not have to have as a basic postulate the effect of past experience on present
experience by way of memory. It needs to explain learning, that is, the improvement
of perceiving with practice and the education of attention, but not by an
appeal to the catch-all of past experience or to the muddle of memory.
The state of a perceptual system is altered when it is attuned to information
of a certain sort. The system has become sensitized. Differences are noticed that
were previously not noticed. Features become distinctive that were formerly
vague. But this altered state need not be thought of as depending on a memory,
an image, an engram, or a trace. An image of the past, if experienced at all,
would be only an incidental symptom of the altered state.
This is not to deny that reminiscence, expectation, imagination, fantasy, and
dreaming actually occur. It is only to deny that they have an essential role to
play in perceiving. They are kinds of visual awareness other than perceptual.
Let us now consider them in their own right.
A New Approach to Nonperceptual Awareness
The redeﬁnition of perception implies a redeﬁnition of the so-called higher mental
processes. In the old mentalistic psychology, they stood above the lower mental
processes, the sensory and reﬂex processes, which could be understood in terms
of the physiology of receptors and nerves. These higher processes were vaguely
supposed to be intellectual processes, inasmuch as the intellect was contrasted
with the senses. They occurred in the brain. They were operations of the mind.
No list of them was ever agreed upon, but remembering, thinking, conceiving, inferring,
judging, expecting, and, above all, knowing were the words used. Imagining,
dreaming, rationalizing, and wishful thinking were also recognized, but it was not
clear that they were higher processes in the intellectual sense. I am convinced
that none of them can ever be understood as an operation of the mind. They will
never be understood as reactions of the body, either. But perhaps if they are
244 The Ecological Approach to Visual Perception
reconsidered in relation to ecological perceiving they will begin to sort themselves
out in a new and reasonable way that ﬁts with the evidence.
To perceive is to be aware of the surfaces of the environment and of oneself
in it. The interchange between hidden and unhidden surfaces is essential to this
awareness. These are existing surfaces; they are speciﬁed at some points of
observation. Perceiving gets wider and ﬁner and longer and richer and fuller as
the observer explores the environment. The full awareness of surfaces includes
their layout, their substances, their events, and their affordances. Note how this
deﬁnition includes within perception a part of memory, expectation, knowledge,
and meaning—some part but not all of those mental processes in each
case.
One kind of remembering, then, would be an awareness of surfaces that
have ceased to exist or events that will not recur, such as items in the story of
one’s own life. There is no point of observation at which such an item will
come into sight.
To expect, anticipate, plan, or imagine creatively is to be aware of surfaces
that do not exist or events that do not occur but that could arise or be fabricated
within what we call the limits of possibility.
To daydream, dream, or imagine wishfully (or fearfully) is to be aware of
surfaces or events that do not exist or occur and that are outside the limits of
possibility.
These three kinds of nonperceptual awareness are not explained, I think, by
the traditional hypothesis of mental imagery. They are better explained by some
such hypothesis as this: a perceptual system that has become sensitized to certain
invariants and can extract them from the stimulus ﬂux can also operate without
the constraints of the stimulus ﬂux. Information becomes further detached from
stimulation. The adjustment loops for looking around, looking at, scanning,
and focusing are then inoperative. The visual system visualizes. But this is still
an activity of the system, not an appearance in the theater of consciousness.
Besides these, other kinds of cognitive awareness occur that are not strictly
perceptual. Before considering them, however, I must clarify what I mean by
imaginary or unreal.
The Relationship between Imagining and Perceiving
I assume that a normal observer is well aware of the difference between surfaces
that exist and surfaces that do not. (Those that do not have ceased to exist, or
have not begun to, or have not and will not.) How can this be so? What is the
information for existence? What are the criteria? It is widely believed that
young children are not aware of the differences, and neither are adults suffering
from hallucinations. They do not distinguish between what is “real” and what
is “imaginary” because perception and mental imagery cannot be separated.
This doctrine rests on the assumption that, because a percept and an image both
The Theory of Information Pickup and its Consequences 245
occur in the brain, the one can pass over into the other by gradual steps. The
only “tests for reality” are intellectual. A percept cannot validate itself.
We have been told ever since John Locke that an image is a “faint copy” of a
percept. We are told by Titchener (1924) that an image is “easily confused with
a sensation” (p. 198). His devoted student, C. W. Perky, managed to show that a
faint optical picture secretly projected from behind on a translucent screen is
sometimes not identiﬁed as such when an observer is imagining an object of the
same sort on the screen (Perky, 1910). We are told by a famous neurosurgeon that
electrical stimulation of the surface of the brain in a conscious patient “has the
force” of an actual perception (Penﬁeld, 1958). It is said that when a feeling of
reality accompanies a content of consciousness it is marked as a percept and when
it does not it is marked as an image. All these assertions are extremely dubious.
I suggest that perfectly reliable and automatic tests for reality are involved in
the working of a perceptual system. They do not have to be intellectual. A surface
is seen with more or less deﬁnition as the accommodation of the lens changes; an
image is not. A surface becomes clearer when ﬁxated; an image does not. A
surface can be scanned; an image cannot. When the eyes converge on an object
in the world, the sensation of crossed diplopia disappears, and when the eyes
diverge, the “double image” reappears; this does not happen for an image in the
space of the mind. An object can be scrutinized with the whole repertory of
optimizing adjustments described in Chapter 11. No image can be scrutinized—
not an afterimage, not a so-called eidetic image, not the image in a dream, and
not even a hallucination. An imaginary object can undergo an imaginary scrutiny,
no doubt, but you are not going to discover a new and surprising feature of the
object this way. For it is the very features of the object that your perceptual
system has already picked up that constitute your ability to visualize it. The most
decisive test for reality is whether you can discover new features and details by
the act of scrutiny. Can you obtain new stimulation and extract new information
from it? Is the information inexhaustible? Is there more to be seen? The imaginary
scrutiny of an imaginary entity cannot pass this test.
A related criterion for the existence of a thing is reversible occlusion.
Whatever goes out of sight as you move your head and comes into sight as you
move back is a persisting surface. Whatever comes into sight when you move
your head is a preexisting surface. That is to say, it exists. The present, past, or
future tense of the verb see is irrelevant; the fact is perceived without words.
Hence, a criterion for real versus imaginary is what happens when you turn and
move. When the infant turns her head and creeps about and brings her hands
in and out of her ﬁeld of view, she perceives what is real. The assumption that
children cannot tell the difference between what is real and what is imaginary
until the intellect develops is mentalistic nonsense. As the child grows up, she
apprehends more reality as she visits more places of her habitat.
Nevertheless, it is argued that dreams sometimes have the “feeling” of reality,
that some drugs can induce hallucinations, and that a true hallucination in
246 The Ecological Approach to Visual Perception
psychosis is proof that a mental image can be the same as a percept, for the patient
acts as if he were perceiving and thinks he is perceiving. I remain dubious (Gibson,
1970). The dreamer is asleep and cannot make the ordinary tests for reality. The
drug-taker is hoping for a vision and does not want to make tests for reality. There
are many possible reasons why the hallucinating patient does not scrutinize what
he says he sees, does not walk around it or take another look at it or test it.
There is a popular fallacy to the effect that if you can touch what you see it
is real. The sense of touch is supposed to be more trustworthy than the sense of
sight, and Bishop Berkeley’s theory of vision was based on this idea. But it is
surely wrong. Tactual hallucinations can occur as well as visual. And if the
senses are actually perceptual systems, the haptic system as I described it
(Gibson, 1966b) has its own exploratory adjustments and its own automatic
tests for reality. One perceptual system does not validate another. Seeing and
touching are two ways of getting much the same information about the world.
A New Approach to Knowing
The theory of information pickup makes a clear-cut separation between
perception and fantasy, but it closes the supposed gap between perception and
knowledge. The extracting and abstracting of invariants are what happens in
both perceiving and knowing. To perceive the environment and to conceive it
are different in degree but not in kind. One is continuous with the other. Our
reasons for supposing that seeing something is quite unlike knowing something
come from the old doctrine that seeing is having temporary sensations one after
another at the passing moment of present time, whereas knowing is having
permanent concepts stored in memory. It should now be clear that perceptual
seeing is an awareness of persisting structure.
Knowing is an extension of perceiving. The child becomes aware of the world
by looking around and looking at, by listening, feeling, smelling, and tasting, but
then she begins to be made aware of the world as well. She is shown things, and
told things, and given models and pictures of things, and then instruments and
tools and books, and ﬁnally rules and short cuts for ﬁnding out more things.
Toys, pictures, and words are aids to perceiving, provided by parents and teachers.
They transmit to the next generation the tricks of the human trade. The labors
of the ﬁrst perceivers are spared their descendants. The extracting and abstracting
of the invariants that specify the environment are made vastly easier with these
aids to comprehension. But they are not in themselves knowledge, as we are
tempted to think. All they can do is facilitate knowing by the young.
These extended or aided modes of apprehension are all cases of information
pickup from a stimulus ﬂux. The learner has to hear the speech in order to pick
up the message; to see the model, the picture, or the writing; to manipulate the
instrument in order to extract the information. But the information itself is
largely independent of the stimulus ﬂux.
The Theory of Information Pickup and its Consequences 247
What are the kinds of culturally transmitted knowledge? I am uncertain,
for they have not been considered at this level of description. Present-day
discussions of the “media of communication” seem to me glib and superﬁcial.
I suspect that there are many kinds merging into one another, of great
complexity. But I can think of three obvious ways to facilitate knowing, to aid
perceiving, or to extend the limits of comprehension: the use of instruments,
the use of verbal descriptions, and the use of pictures. Words and pictures work
in a different way than do instruments, for the information is obtained at second
hand. Consider them separately.
Knowing Mediated by Instruments
Surfaces and events that are too small or too far away cannot be perceived. You
can of course increase the visual solid angle if you approach the item and put
your eye close to it, but that procedure has its limits. You cannot approach the
moon by walking, and you cannot get your eye close enough to a drop of pond
water to see the little animals swimming in it. What can be done is to enlarge
the visual solid angle from the moon or the water drop. You can convert a tiny
sample of the ambient optic array at a point of observation into a magniﬁed
sample by means of a telescope or a microscope. The structure of the sample is
only a little distorted. The surfaces perceived when the eye is placed at the
eyepiece are “virtual” instead of “real,” but only in the special sense that they
are very much closer to the observer. The invariants of structure are nearly the
same when a visual angle with its nested components is magniﬁed. This description
of magniﬁcation comes from ecological optics. For designing the lens
system of the instrument, a different optics is needed.
The discovery of these instruments in the seventeenth century enabled men
to know much more about very large bodies and very small bodies than they
had before. But this new knowledge was almost like seeing. The mountains of
the moon and the motions of a living cell could be observed with adjustments
of the instrument not unlike those of the head and eyes. The guarantees of
reality were similar. You did not have to take another person’s word for what
he had seen. You might have to learn to use the instrument, but you did not
have to learn to interpret the information. Nor did you have to judge whether
or not the other person was telling the truth. With a telescope or a microscope
you could look for yourself.
THE UNAIDED PERCEIVING OF OBJECTS IN THE SKY
Objects in the sky are very different from objects on the ground. The heavenly
bodies do not come to rest on the ground as ordinary objects do. The
248 The Ecological Approach to Visual Perception
rainbow and the clouds are transient, forming and dissipating like mists on
earth. But the sun, the moon, the planets, and the stars seem permanent,
appearing to revolve around the stationary earth in perfect cycles and
continuing to exist while out of sight. They are immortal and mysterious.
They cannot be scrutinized.
Optical information for direct perception of these bodies with the unaided
eye is lacking. Their size and distance are indeterminate except that they rise
and set from behind the distant horizon and are thus very far away. Their
motions are very different from those of ordinary objects. The character of
their surfaces is indeﬁnite, and of what substances they are composed is not
clear. The sun is ﬁery by day, and the others are ﬁery at night, unlike the
textured reﬂecting surfaces of most terrestrial objects. What they afford is not
visible to the eye. Lights in the sky used to look like gods. Nowadays they
look like ﬂying saucers.
All sorts of instruments have been devised for mediating apprehension.
Some optical instruments merely enhance the information that vision is ready
to pick up; others—a spectroscope, for example—require some inference; still
others, like the Wilson cloud chamber, demand a complex chain of inferences.
Some measuring instruments are closer to perception than others. The
measuring stick for counting units of distance, the gravity balance for counting
units of mass, and the hourglass for time are easy to understand. But the complex
magnitudes of physical science are another matter. The voltmeters, accelerometers,
and photometers are hard to understand. The child can see the pointer
and the scale well enough but has to learn to “read” the instrument, as we say.
The direct perception of a distance is in terms of whether one can jump it. The
direct perception of a mass is in terms of whether one can lift it. Indirect knowledge
of the metric dimensions of the world is a far extreme from direct perception
of the affordance dimensions of the environment. Nevertheless, they are
both cut from the same cloth.
Knowing Mediated by Descriptions: Explicit Knowledge
The principal way in which we save our children the trouble of ﬁnding out
everything for themselves is by describing things for them. We transmit information
and convey knowledge. Wisdom is handed down. Parents and teachers
and books give the children knowledge of the world at second hand. Instead of
having to be extracted by the child from the stimulus ﬂux, this knowledge is
communicated to the child.
It is surely true that speech and language convey information of a certain
sort from person to person and from parent to child. Written language can even
be stored so that it accumulates in libraries. But we should never forget that this
The Theory of Information Pickup and its Consequences 249
is information that has been put into words. It is not the limitless information
available in a ﬂowing stimulus array.
Knowledge that has been put into words can be said to be explicit instead of
tacit. The human observer can verbalize his awareness, and the result is to make
it communicable. But my hypothesis is that there has to be an awareness of the
world before it can be put into words. You have to see it before you can say it.
Perceiving precedes predicating.
In the course of development the young child ﬁrst hears talk about what she
is perceiving. Then she begins herself to talk about what she perceives. Then she
begins to talk to herself about what she knows—when she is alone in her crib,
for example. And, ﬁnally, her verbal system probably begins to verbalize silently,
in much the same way that the visual system begins to visualize, without the
constraints of stimulation or muscular action but within the limits of the invariants
to which the system is attuned. But no matter how much the child puts
knowledge into words all of it cannot be put into words. However skilled an
explicator one may become one will always, I believe, see more than one can say.
Consider an adult, a philosopher, for example, who sees the cat on the mat.
He knows that the cat is on the mat and believes the proposition and can say it,
but all the time he plainly sees all sorts of wordless facts—the mat extending
without interruption behind the cat, the far side of the cat, the cat hiding part
of the mat, the edges of the cat, the cat being supported by the mat, or resting
on it, the horizontal rigidity of the ﬂoor under the mat, and so on. The so-called
concepts of extension, of far and near, gravity, rigidity, horizontal, and so on,
are nothing but partial abstractions from a rich but unitary perception of caton-mat.
The parts of it he can name are called concepts, but they are not all of
what he can see.
Fact and Fiction in Words and Pictures
Information about the environment that has been put into words has this disadvantage:
The reality testing that accompanies the pickup of natural information
is missing. Descriptions, spoken or written, do not permit the ﬂowing stimulus
array to be scrutinized. The invariants have already been extracted. You have
to trust the original perceiver; you must “take his word for it,” as we say. What
he presents may be fact, or it may be ﬁction. The same is true of a depiction as
of a description.
The child, as I argued above, has no difﬁculty in contrasting real and
imaginary, and the two do not merge. But the factual and the ﬁctional may do
so. In storytelling, adults do not always distinguish between true stories and
fairy stories. The child herself does not always separate the giving of an account
from the telling of a story. Tigers and dragons are both fascinating beasts, and
the child will not learn the difference until she perceives that the zoo contains
the former but not the latter.
250 The Ecological Approach to Visual Perception
Fictions are not necessarily fantasies. They do not automatically lead one
astray, as hallucinations do. They can promote creative plans. They can permit
vicarious learning when the child identiﬁes with a ﬁctional character who
solves problems and makes errors. The “comic” characters of childhood, the
funny and the foolish, the strong and the weak, the clever and the stupid,
occupy a great part of children’s cognitive awareness, but this does not interfere
in the least with their realism when it comes to perceiving.
The difference between the real and the imaginary is speciﬁed by two
different modes of operation of a perceptual system. But the difference between
the factual and the ﬁctional depends on the social system of communication
and brings in complicated questions. Verbal descriptions can be true or false as
predications. Visual depictions can be correct or incorrect in a wholly different
way. A picture cannot be true in the sense that a proposition is true, but it may
or may not be true to life.
Knowing and Imagining Mediated by Pictures
Perceiving, knowing, recalling, expecting, and imagining can all be induced by
pictures, perhaps even more readily than by words. Picture-making and pictureperceiving
have been going on for twenty or thirty thousand years of human
life, and this achievement, like language, is ours alone. The image makers can
arouse in us an awareness of what they have seen, of what they have noticed, of
what they recall, expect, or imagine, and they do so without converting the information
into a different mode. The description puts the optical invariants into words.
The depiction, however, captures and displays them in an optic array, where
they are more or less the same as they would be in the case of direct perception.
So I will argue, at least. The justiﬁcation of this theory is obviously not a simple
matter, and it is deferred to the last chapters of this book, Part IV.
The reality-testing that accompanies unmediated perceiving and that is
partly retained in perceiving with instruments is obviously lost in the kind of
perceiving that is mediated by pictures. Nevertheless, pictures give us a kind of
grasp on the rich complexities of the natural environment that words could
never do. Pictures do not stereotype our experience in the same way and to the
same degree. We can learn from pictures with less effort than it takes to learn
from words. It is not like perceiving at ﬁrst hand, but it is more like perceiving
than any verbal description can be.
The child who has learned to talk about things and events can, metaphorically,
talk to himself silently about things and events, so it is supposed. He is said
to have “internalized” his speech, whatever that might mean. By analogy with
this theory, a child who has learned to draw might be supposed to picture to
himself things and events without movement of his hands, to have “internalized”
his picturemaking. A theory of internal language and internal images
might be based on this theory. But it seems to me very dubious. Whether or not
The Theory of Information Pickup and its Consequences 251
it is plausible is best decided after we have considered picturemaking in its own
right.
Summary
When vision is thought of as a perceptual system instead of as a channel for
inputs to the brain, a new theory of perception considered as information
pickup becomes possible. Information is conceived as available in the ambient
energy ﬂux, not as signals in a bundle of nerve ﬁbers. It is information about
both the persisting and the changing features of the environment together.
Moreover, information about the observer and his movements is available, so
that self-awareness accompanies perceptual awareness.
The qualities of visual experience that are speciﬁc to the receptors stimulated
are not relevant to information pickup but incidental to it. Excitation and
transmission are facts of physiology at the cellular level.
The process of pickup involves not only overt movements that can be measured,
such as orienting, exploring, and adjusting, but also more general activities,
such as optimizing, resonating, and extracting invariants, that cannot so
easily be measured.
The ecological theory of direct perception cannot stand by itself. It implies a
new theory of cognition in general. In turn, that implies a new theory of noncognitive
kinds of awareness—ﬁctions, fantasies, dreams, and hallucinations.
Perceiving is the simplest and best kind of knowing. But there are other
kinds, of which three were suggested. Knowing by means of instruments
extends perceiving into the realm of the very distant and the very small; it also
allows of metric knowledge. Knowing by means of language makes knowing
explicit instead of tacit. Language permits descriptions and pools the accumulated
observations of our ancestors. Knowing by means of pictures also extends
perceiving and consolidates the gains of perceiving.
The awareness of imaginary entities and events might be ascribed to the
operation of the perceptual system with a suspension of reality-testing.
Imagination, as well as knowledge and perception, can be aroused by another
person who uses language or makes pictures.
These tentative proposals are offered as a substitute for the outworn theory
of past experience, memory, and mental images.