This is a very
fine book on the psychology of vision and visualizing.  In it Pylyshyn defends in detail the
information-processing account he has developed over a number of years on these
topics.  The first chapter introduces
his central thesis.  Vision does not
involve the creation of a picture, or visual image, in the head.  When we see the rich and stable
three-dimensional world of objects around us, there is no "inner
display"; rather, we have a visual percept of some distal object in the
external world.  To the extent that we
are lead to believe in the awareness of inner items, we are misled by the
phenomenology of vision, by our conscious experience of how things seem.

Pylyshyn does not deny that there is a subjective phenomenology of
seeing and visualizing, which is in some way linked to the neurological
processes in the brain.  But these
subjective experiences are little understood, and do not tell us about the
nature of the processes that lead us to visually represent the external objects
around us.  Whether he does adequate
justice to the phenomenology is a point to which I shall return.

The theory of vision is articulated in more detail in chapters two to
five.  In the final three chapters
Pylyshyn defends a related account of visualizing, arguing against what he
terms the ‘picture theory of mental imagery’: 
when someone engages in imaginative reasoning, there is no autonomous
quasi-spatial visual image with intrinsic features that are independent of the
act of imagining.  Inner images play no
essential role.  There is much of
importance and interest in these final chapters, and they extend and in places
amplify some of the claims made in the early parts of the book.  Here I will concentrate for the most part on
Pylyshyn’s claims about vision, and how the points he makes about the objective
nature of this process square with our understanding of the subjective
experience of seeing.

Chapters two and three deal with the relation of vision to the higher
cognitive processes.  Here Pylyshyn
attacks the "new look" tradition that denies any hard and fast
distinction between what we see and how we interpret things to be.  Such views were for a long time very
fashionable, and have been highly influential outside of psychology.  Much bad philosophy of science, for example,
has resulted from an overeager, and insufficiently considered, acceptance of
the idea that everything we see is dependent upon our conceptual framework —
thus making it hard to provide an adequate account of how scientists could ever
discover results that failed to fit in with the predictions of their theories.  But as Pylyshyn shows, there is considerable
evidence to show that, as he puts it, ‘early vision is both complex and
impervious to such cognitive influences as expectations, beliefs and desires….’
(p.91).  The early visual system, he
argues, is encapsulated, that is, not cognitively penetrable; its deliverances
are not responsive to the constraints of rationality.  This accounts not only for the fact that we experience illusions
of degree — as when we misjudge the relative lengths of lines in the familiar
Müller-Lyer illusion — but also that we seem to see impossible objects and
events — for example, in illusions where two solid objects appear to pass
through each other.

The early vision system, Pylyshyn argues, does not consist just of
operations that transduce the optical inputs. 
Nor does it involve processes that make use of explicit background
knowledge about the way the world is. 
Rather, it applies a variety of built-in assumptions, developed through
the course of evolution, which act as natural constraints on inputs, resulting
in perceptual beliefs that normally hold true in our physical environment.  So what of the experimental evidence, for
example, about how we organize experiences, evidence that appears to show that
what we see is somehow dependent upon cognitive processes?  Part of the problem is that we use the
concept of attention in a variety of ways. 
Pylyshyn argues convincingly that we need to distinguish three different
stages or general processes in perception. 
In one important sense, the mechanism of selective or focal attention
operates at what Pylyshyn terms, perhaps a bit misleadingly, a ‘pre-perceptual
stage’.  This attentional process is in
part responsive to voluntary control. 
It can be applied when we turn our visual attention to certain
objects.  Such attentional processes do
not interfere with the encapsulated operations of the early visual system.  These operations are not in themselves influenced
by our reasoning processes.  But
attention can also work as part of a post-perceptual process, by enhancing or
attenuating the availability of certain perceptual categories.  In this way, our attentional set is involved
in evaluating objects, and may lead us to be more prepared to classify objects
using certain concepts.

The architecture of the early vision system itself involves a number of
distinct operations.  As Pylyshyn
rightly argues, we cannot adequately explain how the visual system responds to
the information contained in the seen environment without postulating some kind
of transformations of representations within the perceiving subject. The result
is that at an early stage in the output from the visual system there are
representations of the spatial layout of visible surfaces, and also
computations of depth.  Clusters of
features are individuated as primitive visual objects, whose shape is
represented but not conceptually encoded.

The idea that attention operates at a pre-perceptual stage is linked
with the notion that part of what is delivered by the early visual system is a
rudimentary, or primitive, (spatially defined) object, and is the subject of
discussion in chapter four.  In the
following chapter, Pylyshyn provides a detailed account and defense of his
visual-index theory.  He argues that
objects in the environment are indexed directly by demonstratives, and not
conceived of as necessarily individuated via their properties. 

We don’t identify objects by reference to the properties they have, or
by a fixed location.  Pylyshyn analyses
experiments on numerosity that suggest we have the ability to directly identify
objects as individuals, independently of how we conceive them.  Pylyshyn seems right about this, and some of
what he says here fits, in interesting ways, with what Wilfrid Sellars argued
in his attack on the "Myth of the Given".  The relation between what is present in experience at a lower
level and the subsequent conceptual uptake, the immediate beliefs about the
world, is not an inferential one; it is straightforwardly causal.  A process leads from the external objects in
the "field of view" to the formation of low-level representations of
clusters of properties.  This in turn
causes the individuation of primitive visual objects that are "detected",
or "registered", but not conceptually encoded.  Some of these primitive objects prompt the
employment of a limited number of indexicals (a maximum of four or five), which
we use in tracking the objects.  Since
such objects are not necessarily conceived as having any fixed properties or
locations, this allows for the fact that what is identified can change in
various ways.

The standard of argumentation is consistently high throughout.  Careful and balanced consideration is given
to the detailed experimental evidence, and one great virtue of Pylyshyn’s
discussion is the careful attention he pays to theoretical issues.  One worry I have, however, concerns the
initial claim that vision does not involve any inner spatial display, and how
this issue relates to the subjective aspect of experience.  In some places Pylyshyn is too readily
inclined simply to put the question of subjective conscious experience to one
side, as an unresolved issue, beyond our present understanding.  It is true that no-one has yet provided a
persuasive account of why certain neuronal configurations of the brain should
give rise to the kinds of conscious experiences they do; this is the
"explanatory gap" (as Joseph Levine famously described it), a gap in
our understanding that remains problematic, even if we hold to a materialist
view of the mind.  Nevertheless, even if
we don’t fully grasp why experiences ultimately take the form they do, we still
owe an account of their proper place in perception.  As Pylyshyn himself admits, we cannot ignore the contents of our
conscious experience; he rightly observes: 
‘visual experience remains the reference point against which we measure
what we mean by seeing‘ (p. 37).

Yet at other points he appears to equate the claim that seeing involves
a subjective experience with the theory he is concerned to criticize
throughout, that in seeing physical objects the mind constructs some kind of
internal screen or display.  In
summarizing his objections to the idea of an inner display at the end of
chapter one, he writes, ‘we will have to jettison the phenomenal image or
display and come to grips with the information-processing task that vision
carries out’ (p.47).  But several quite
different aspects of vision need to be carefully distinguished here.  Normally, when I see an object, my concepts
do not focus upon my own states. 
Pylyshyn rightly observes that what I see are people, tables and chairs,
and so on.  At the conscious level, I
directly represent the fact that there are different kinds of physical things
in my environment.  The important point,
however, is that the representational states involved in seeing are essentially
different from the representational states involved in merely thinking about an
object.  My experience of seeing is not
exhausted by its representational content. 
There is a non-representational aspect to visual experience.

This crucial matter is glossed over by Pylyshyn’s use of the term
‘percept’.  In line with a long
tradition, Pylyshyn helps himself to what is a problematic concept.  The "percept" is a mongrel
notion.  Percepts are supposed to
combine two quite different logical roles. 
A percept represents something other than itself and the subject.  So percepts normally represent physical
things and events in the environment of the subject, such as tables and chairs,
and even things that might not exist (as in hallucinations).  Yet percepts are also supposed to account
for the actual, intrinsic, qualitative nature of experience.  Thus in seeing the table top, I experience
it to be brown. 

Reflection suggests that the actual qualities I am aware of in
experience are, in actual fact, qualities that somehow belong to my inner
states.  Double vision is one way of
illustrating the point.  If I see a red
pen in front of me as double, I have two percepts of a pen.  In one sense, it is true that what is
represented is single physical object, the pen situated just in front of
me.  But I can distinguish and refer
separately to the two red expanses I have in my experience.  When I do so, I am conceptualizing these
expanses as I might conceptualize a hallucinatory image I experience, when, for
example, I have a visual aura accompanying a migraine.  There are good reasons for holding that such
qualitative aspects of my experience are inner states.  Neither of the images of the pen I see can
be identified with the actual pen; and they appear to supervene upon my
neurological states.  Now it is true
that the phenomena of hallucinatory experiences and double vision are remote
from normal visual experiences.  Yet it
is arguable, so far as the subjective experience is concerned, that they do not
differ in principle from the kinds of subjective experience that occur in
normal vision (although there is much philosophical controversy over this
point).   For such reasons I believe it
is important to note the role of phenomenal experiences in vision, not as objects we are aware of, but as the
essential inner accompaniments of our visual awareness of external
objects. 

It is important, therefore, to be clear about the place of subjective
experience in seeing.   The subjective
phenomenal aspects of experiences are, in the above sense, inner states.  Visual phenomenal experiences have a
quasi-spatial nature.  Yet acknowledging
this fact does not upset the general claims that Pylyshyn makes about the
information processes involved in vision. 
Considered from the third person, objective perspective, there are good
reasons for taking vision to be a process involving a series of
representational states, states that focus on the external world, as Pylyshyn
shows.  Pylyshyn’s book is to be
commended as a through and persuasive defense of the information-processing
approach to vision and visualizing.  It
should be essential reading for psychologists, cognitive scientists and
philosophers.

© 2005 Paul Coates

Paul Coates Philosophy, University of Hertfordshire

Categories: Psychology, Philosophical