Michael
Tye’s The Imagery Debate is an introduction to the issues in philosophy
and psychology that define the debate over the content and structure of mental
imagery.  The experience of visual
mental imagery, the act of visualizing something in your imagination, seems to
share phenomenal properties with ordinary visual experience.  For instance when I recollect childhood
Christmas experiences it is as if I can smell the freshly waxed kitchen floor
and see the clean crisp blanket of snow in the yard with my mind’s eye.  Furthermore, behavioral evidence from
studies of normal subjects, and patients suffering localized brain damage,
suggest that visual mental images have similar spatial structure to ordinary
visual representations.  Visual deficits
due to damage to the visual system are preserved in mental imagery and tasks
that require patients to scan mental images take the same amount of time as
similar visual scanning tasks.  Some
philosophers and cognitive scientists, most notably the neuropsychologist Stephen
Kosslyn, argue as a result that visual mental imagery is an internal,
non-perceptual visual experience caused either by recollecting or
conceptualizing something.  On this
account visual mental images are structurally analogous to visual representations,
and are caused, at least in part, by psychological processes shared with the
visual system.  Others, most notably the
cognitive psychologist Zenon Pylyshyn, argue that all thoughts, including
mental images, are propositional.  As a
result, any phenomenal or structural similarity reported by subjects between
mental images and perceptual representations is an illusion.

Tye
writes that the book is intended for a general audience and that no significant
background in either psychology of philosophy is presupposed in the
presentation of its material.  I found
Tye’s prose very readable.  But, whereas
his summary of the issues surrounding the psychological side of the debate are
clear and should be readily accessible to a general audience, his discussion of
the philosophical side of the debate is steeped in the technical jargon and
problems of the philosophy of mind.  A
reader unfamiliar with the general dialectic of these philosophical issues may
find it difficult to understand their relevance to the debate concerning the
structure and content of mental images. 
Nonetheless, despite the decade that has passed since its original
printing, the book serves as a good general summary of the imagery debate and a
nice example of the integration of empirical data and philosophical discussion
in cognitive science.

The
Imagery Debate divides
naturally into two main sections that bracket Tye’s discussion of his own
theory of imagery.  The first four
chapters introduce both traditional conceptions of imagery from the history of
philosophy and the contemporary state of the imagery debate in cognitive
science.  The last three chapters
discuss solutions to some philosophical problems that have plagued the notion
of mental imagery, e.g. how should one understand the phenomenal content of
imagery and how can a depictive theory of imagery account for phenomena like
image indeterminacy, and the effects of subjects’ beliefs on the content of
mental images.  Tye’s proposal is that
visual mental images should be treated as "interpreted symbol-filled
arrays."  This model is intended as
a hybrid of what Tye identifies as Pylyshyn and Kosslyn’s key insights into the
nature of imagery.  The gist of Tye’s
theory is that visual mental images are constituted of rudimentary abstract
visual representations of the surface geometry of scenes and objects whose
structure is provided by a sentential interpretation to which it is
affixed. 

Tye’s
theory draws on David Marr’s computational model of vision.  On these accounts the rich three dimensional
structure of visual representations is constructed top down in the visual
system.  Underdetermined visual arrays
are segmented relative to viewer’s tacit and explicit knowledge of the shapes
and functions of objects.  Tye argues,
like Kosslyn, that mental imagery exploits these same processes, but replaces
retinal input with top down conceptual input. 
Mental images of scenes and objects are generated from information
stored in long term memory concerning their general form, function, and
appearance.  Tye contends that, as a
result, the visual component of imagery, the initial array, is not by itself an
image.  Rather it is an abstract
representation of the visible properties of surfaces and locations seen from a
particular viewpoint .  Only by being
interpreted relative to particular conceptual information is this initial
visual array segmented into forms and objects. 
Therefore Tye’s model attributes to images both propositional and
depictive content.

Tye
argues that his theory resolves objections like Pylyshyn’s "beaker
experiment."  Pylyshyn reports that
4 year old children, first shown an inclined beaker half full of water, and
then asked to draw it from memory, typically draw the fluid level perpendicular
to the sides of the beaker.  He takes
this to entail that they are not bringing up an image of the beaker from memory
and copying it.  Rather their mental
images are influenced by what they believe about beakers and fluid.  They are "cognitively
penetrable."  Pylyshyn argues that
the content of perception is not so cognitively penetrable.  Thoughts are influenced by beliefs,
perceptions are not.  The children
reports about their ordinary perceptions of beakers demonstrate that they
represent the fluid correctly as level with the ground.  Therefore the structure of visual mental
images is not analogous to the structure of visual representations.

But,
on a constructivist model of vision like Marr’s, this does not follow.  Consider the following example.  It is well known that practical knowledge
interferes with seeing in artistic contexts. 
Adults in beginning drawing classes, when asked to draw, for instance, a
chair from a side view just below their line of sight, have a tendency to
elevate their point of view.  One
explanation of these phenomena is that they draw it as if they were looking
down at it from a higher perspective in order to resolve discrepancies between
perceptual foreshortening and their canonical concept of the shape of the
chair.  One can argue, as a result that
the content of their visual experience is an interpolation between what they
"see" and what they know. 
They see, in a sense, what they expect to.  This can be corrected by showing them their error and giving them
a new framework with which to conceive the visual structure of scenes.  In part, this is the role of beginning
drawing classes. 

If
conceptual influences interfere with one’s ability to discern, and so one’s
ability to construct, correct visual relations in ordinary perceptual
experience, it is no surprise that they do so in tasks that require one to
recall visual relations from memory.  On
this account the 4 year olds are drawing what they would expect to see. These
expectations are based on their limited knowledge of the world.  Their visual mental images, constructed from
their concept of what they would expect to see, are not counterexamples to, but
rather are consistent with the constructivist account of vision.  Tye’s model for mental imagery appeals to
these same top down processes. 
Therefore the argument from cognitive impenetrability is not an
objection to a depictivist account of mental imagery. 

Tye’s
argument entails that the structure and content of mental images is constrained
by the content of the semantic knowledge used to generate them.  Tye generalizes this argument, using facts
about the conceptual contribution to the content of ordinary perception to
resolve philosophical objections to mental imagery concerning image
indeterminacy and the phenomenal content of images.  But there is a difficulty for Tye’s theory.  One might ask why Tye believes that the
visual arrays that account for the depictive content of imagery require
interpretation.  In ordinary vision
these arrays are constructed bottom up, from underdetermined retinal inputs,
relative to sets of perceptual principles tacitly encoded in the early visual
system.  The product of these processes
is an undifferentiated, 2 ½ D sketch of the surface geometry of an image.  It is often argued that such images do not
yet represent objects in the visual field because they are consistent with many
possible interpretations.  The
structure, and so content, of visual representations is provided in a second,
top down process by which the visual system segments arrays relative to
knowledge of the general shapes and functions of types of objects.  It is only in being transformed relative to
such semantic knowledge that an array comes to represent something
visually.  But, imagery is not generated
in the same manner.  The initial inputs
responsible for image generation are the top down processes that implement
image segmentation.  This entails that
images are generated from the structural information employed to segment
ordinary perceptual arrays, and so carry the structural information Tye argues
they lack.  As a result, the later
interpretive process employed by Tye to accommodate the insights of
propositional accounts of imagery is computationally otiose.

© 2003 Bill Seeley

Bill Seeley is an Instructor in the Department of Philosophy at
Brooklyn College.  In the Fall of 2003
he will be a Lecturer in Yale College at Yale University where he will offer a
seminar on the role that neurophysiology and cognitive science can play in
explanations of art and aesthetic experience. 
He received an M.A. in Philosophy from CUNY — The Graduate Center in
New York City, and is currently a doctoral candidate there A.B.D. in the
Program in Philosophy and Concentration in Cognitive Science.  His research is focused on the
neurophysiology of vision and aesthetic experience.  He also has an M.F.A. in sculpture from Columbia University.  His welded steel constructions and mobiles
have been exhibited in New York City, Tokyo, and at The Addison Gallery of
American Art in Massachusetts.  In 1992
he had a solo exhibition of outdoor work at Yale University.  He is represented by Studio Facchetti in
Brooklyn, New York.

Categories: Psychology, Philosophical