Review of Martha Farah, Visual Agnosia

Philosophical Psychology, 7(1), 1994, 126-29

Martha Farah, Visual Agnosia: Disorders of Object Recognition and What They Tell Us about Normal Vision (MIT Press, Bradford Books, 1990).


Austen Clark
Department of Philosophy U-54
University of Connecticut
Storrs, CT 06269-2054


Common sense says that visual agnosia is impossible. It ought not exist. If an object like a safety pin or a bar of white soap is in full view, you see it, and you know what a "safety pin" or a "bar of soap" is, then you cannot fail to recognize what you see. If you identify the safety pin as "something silver and shiny like a watch or a nail clipper," or you identify the bar of white soap as "a piece of paper," then common sense would dictate that either you fail to see the object, or your knowledge is somehow deficient. But visual agnosics, who make such responses, clearly do--in some sense--"see" the objects in question. Often they can, for example, make accurate and recognizable drawings of what they see. Some lack measurable visual field defects; they have a full visual field. And they clearly know what a "safety pin" or a "bar of soap" is: if allowed to touch the object, or its use is pantomimed, correct identification is immediate. "I see it now," they may say. We get a failure specifically in visual recognition, even though sufficient sensory functions and cognitive functions are demonstrably intact. Such is the mystery of visual agnosia.

Martha Farah has cast considerable light into the darkness with her book Visual Agnosia: Disorders of Object Recognition and What They Tell Us about Normal Vision (MIT Press, Bradford Books, 1990. Page references to that volume.) This is, literally, a wonderful book: filled with perceptions of things marvellous and (apparently) inexplicable. It is invigorating to see how steadfast empiricism and contemporary cognitive neuroscience can clear up parts of the mystery. Aside from the intrinsic interest of the subject matter, the book provides a case-study of how cognitive neuroscientists can successfully go about their business.

The first step is to recognize that there are many different kinds of visual agnosia, and some of the claims made about some kinds are not true of other kinds. Farah scrupulously reviews the literature, and on that basis constructs a new taxonomy. Historically the visual agnosias have been classified as either "apperceptive"--entailing some degree of perceptual impairment, though not of the sort sufficient to prevent object recognition--or "associative"--lacking such impairment. While ultimately Farah challenges the coherence of this division, she uses it as an organizing principle for the book. A chapter is devoted to describing the various apperceptive agnosias, followed by a chapter that reviews theoretical interpretations and offers some new insights. A parallel pair of chapters describe and interpret the associative agnosias. The concluding chapter pulls together the new theoretical interpretations, summarizes the emerging model of the architecture of normal vision, and describes how failures in parts of that model could produce the observed syndromes.

Its contribution to taxonomy alone would make the book worthwhile. One hopes that every clinician who might encounter a visual agnosic will acquire a copy. Since common sense has long inspired skepticism in the very existence of agnosia, many traditional case reports were devoted more to a proof of the existence of the syndrome than to a careful cataloging of what capacities are or are not present. As a consequence, some venerable classifications included distinct clusters of symptoms, and clinicians did not always perform all the tests necessary to distinguish one syndrome from another. For example, "simultanagnosics" are traditionally described as those unable to perceive more than one thing at a time. Farah found that patients so described had distinct clusters of abilities, and fell into two distinct groups. Some can, and some cannot perform such tasks as counting the dots in a line, or localizing objects in the visual field. Purely on empirical grounds, she distinguishes two groups, labelling one "dorsal simultanagnosia" and the other "ventral". The defining characteristics and differences are carefully noted. This process yields some surprises. Farah notes that ventral simultanagnosia and pure alexia share most of their defining characteristics. She boldly proposes that the two are expressions of the same underlying impairment. This is in spite of the fact that pure alexia has traditionally been classed an "associative" and not apperceptive agnosia. Taxonomic shifts are not theoretically inert.

Other surprises abound. The existence of prosopagnosia, the inability visually to recognize faces, is widely acknowledged. But the specificity of the deficit is very much in question. Prosopagnosics often have difficulty recognizing other classes of stimuli, such as: animals within a class, plants, makes of automobile, or kinds of food. A dairy farmer who became prosopagnosic lost the ability to recognize the individual cows in his herd. A bird-watcher who became prosopagnosic thereafter reported, sadly, "all the birds look the same." That some sufferers lose the ability to discriminate Fords from Toyotas is startling. If prosopagnosia derives from damage to "modules" specifically devoted to face recognition, do we use those same modules to recognize automobiles? Or perhaps some class of higher-level shape representation is engaged both by faces and by automobiles, and it just happens that the most salient exemplar of that shape primitive is a face. In any case it is clear that researchers need to test for a wide variety of other stimulus classes to delimit the bounds of the deficit. Some patients can visually recognize such stimuli as coins, handwriting, personal items, and even facial expressions (happy vs. sad, etc.) but not facial identity. They might recognize that a person is sad, but not that that person is their spouse.

After cleaning up taxonomy, the second step to clarity is to discard the common sense model of visual recognition as a process of matching sensory inputs directly to concepts. The interpretive chapters in Visual Agnosia replace this with a more modern picture, hailing in part from the work of David Marr, but strongly influenced by contemporary cognitive psychology and connectionism as well. "Seeing" describes the collective results of layered and dissociable subprocesses. These processes generate a series of representations of retinal input. Later representations in the series are computed from earlier ones, are increasingly abstract, employ an object-centered coordinate scheme, use new primitives, and encode object-invariant properties such as shape and size. In such a system agnosia is possible. Brain damage knocking out just one of these intermediary representations could leave one selectively blind, but with a full visual field. For example, loss of grouping and segmentation processes would leave one with punctate visual impressions that are not organized as objects--perhaps accounting for "pure" apperceptive agnosia. If specialized modules are devoted to particular classes of stimuli--such as faces--their loss could leave one unable to recognize a face, even though all its features are fully visible.

The clusters of observed symptoms yield clues to the constitution and character of those sub-processes. The goal is a model of visual architecture which shows both how it can break down and how those break-downs yield the clinical syndromes observed. Farah makes considerable progress towards this goal. She carefully reviews the available theoretical interpretations of the visual agnosias. Some of them--disconnection and symbol-search accounts of associative agnosia, for example--are shown to be inadequate. Farah's proposals are developed within the context of what cognitive psychology and neuroscience know of visual architecture. She proposes that there are two fundamental capacities required to recognize objects: an ability rapidly to encode "structure descriptions" for each part of an object, and an ability to compute relations among those parts and derive an object description. Different types of associative agnosia are hypothesized to result from impairment in one or another of these capacities. Prosopagnosia represents a failure in the first sort of capacity; pure alexia the second sort. An implication of this account is that one will never find object agnosia without some degree of either prosopagnosia or alexia and that "agnosia without prosopagnosia is a more severe case of the same impairment that underlies pure alexia" (133). In a review of 99 case reports there is no clear disconfirming instance. Where before we had muddle and mystery, now we have a clear target. This is progress.

There are some loose ends. It is unclear what it means to say simultanagnosics can perceive just "one object" at a time. It is unsettling to find that dorsal simultanagnosics can perceive the shape of one object, and spatial relations among parts of one object, but not spatial relations among several objects. Some clinical syndromes remain paradoxical even after one absorbs the idea of "higher level" visual representations. Optic aphasia is the prime example. Patients specifically lack the ability to name visually presented objects. A sound or touch immediately yields the name. But they can pantomime the use of objects, and sort them by category. None of the models of this syndrome are satisfactory. As Farah puts it,

In order to name a visually presented object, I must see it clearly enough to be able to access some semantic information about it (i.e., to know what it is), and once I know what it is I must then retrieve and produce its name. Surprisingly, there is no part of this model that can be damaged to produce optic aphasia. Given that optic aphasics can gesture appropriately to visual stimuli they cannot name, and correctly sort or match visual stimuli according to semantic attributes, then their impairment cannot lie anywhere in vision, semantics, or the path between the two. Given that they can supply the appropriate name to verbal definitions, sounds, and palpated objects, then their impairment cannot lie anywhere in semantics, naming operations, or the path between the two. Note that all possible loci for damage in this simple model for visual naming have just been eliminated! (135)

She eventually plunks for the admittedly ugly hypothesis that the syndrome is produced by two lesions whose effects are super-additive.

A conceptual tidbit is left on the table, and is fit for the tender feasting of philosophers. How are we to characterize the perceptions of visual agnosics? In what sense do they or don't they "see" things normally? Between quantum catches at the retina and the eventual linguistically mediated judgements, we must postulate processes that are neither sensory nor cognitive, but something in between. In what categories are we to describe the results when some of these processes cease? In many visual agnosics, object recognition becomes a task of explicit inference; they add up impressions and attempt to deduce an answer. As one patient said, "I have to use my mind to interpret what I'm seeing. My eyes used to do that" (65). Another, looking at a U.S. flag, reports: "I see a lot of lines. Now I see some stars. When I see things like this, I see a lot of parts. It's like you have one part here and one part there, and you put them together to see what they make" (23). Ultimately the distinction between "apperceptive" and "associative" agnosia itself breaks down. As Farah says,

To some extent, the distinction between these two general types of agnosia is academic ... all the agnosias seem attributable to faulty perception. The more relevant distinction concerns the level of visual processing at which perception is impaired. (120)

Philosophically we have yet to digest the idea that there are various, higher "levels" of visual representation. Farah says it is a "neuropsychological truism" that vision is composed of distinct and dissociable subprocesses. Full absorption and assimilation of this idea will take some effort; philosophers need to chew on it for awhile. You are invited to the feast.


Return to Austen Clark's online papers .

Return to the Philosophy Department home page.