The visual

association

cortex

Semir Zeki University The concept of visual studies, assorted cases speculation. A review review our

Current

College

Opinion

in Neurobiology

The concept of association cortex is essentially a somewhat vague functional concept inferred from fairly precise anatomical studies. In time, philosophical speculation came to make its contribution to the concept, especially in visual cortex. Not surprisingly, this served to confuse the concept rather than illuminate it. However that may be, the different approaches reinforced each other and led to a fundamentally flawed view of the cerebral processes involved in vision. This supposed that the function of visual association cortex was to ‘understand what was seen, seeing being a function of the primary visual cortex, area Vl. Any revision of the concept of vi sual association cortex thus entails a profound revision of our views not only of visual cortex and the processes it undertakes, but also of our philosophical approach to the problem of vision.

of visual ‘association’

UK

association cortex derives from early myelogenetic of so-called visual agnosia and much philosophical of the evidence suggests that it is perhaps time concept of the visual association cortex.

Introduction

Definition

London,

1993,

to

3:155-159

scious awareness? Or was association intended to signify the unification of the representation of different points, representing different regions of the body surface, in the topographically organized sensory areas? Neurologists, assuming them to have thought about the implications of their terminology, remained vague about what they meant by the term. Instead, they proposed a definition that was so general that it applied to all the above categories. They imagined that it was association cortex that gave visual ‘impressions’ their meaning and hence that it was visual association cortex which dealt with ‘higher’ functions.

cortex

The human cerebral cortex is not fully differentiated at birth. Some areas, which Flechsig [1,2] called ‘primordial’ and amongst which he numbered the primary visual cortex (area Vl), are myelinated at birth, though they occupy only a small fraction of the total cortical surface (Fig. 1). They are connected to the peripheral organs and are separated from each other by other, and much larger, cortical areas. The latter are not connected to the peripheral organs and become myelinated at various stages after birth, as if their myelination depends upon the acquisition of experience. Flechsig used the term ‘association’ cortex to describe these latter areas, believing them to be the Geistige Zentren or Cogitationszentren (mind centres or psychic centres). Hence, the commonly used altemative term for visual association cortex was the visuo-psychic cortex. Large cerebral areas, whose boundaries can be defined with fair precision, were thus inferred to have a function, that of association, though the function itself was not precisely delined. Was it an association of present with past records in a given modality, such as vision, or an association between different modalities that neurologists had in mind? Would it be the kind of association that dignifies vision with meaning and therefore con@Current

Biology

F&l. Flechsig’s diagram of the medial view of the human brain, to show the primordial areas (hatched and cross-hatched) and the association areas (in white). The visual association cortex was considered to surround the primary visual cortex.

Lissauer [ 31, whose speculations were to have a powerful influence on visual neurologists, anticipated the present talk about ‘lower’ and ‘higher’ levels of vision by using Leibniz’s term ‘apperception’ to speak of the function of Vl in high sounding terms. Apperception was “....the highest degree of perception, in which the consciousness accepts the sensory impression with maximal intensity” - a view which invests Vl with a critical role in consciousness. This was followed by the process of ‘association’, of “....connecting other conceptions with the content of the perception”, thus giving them their meaning. But no one specified what these conceptions may Ltd ISSN 0959-4388

155

156

Cognitive

neuroscience

be or what underlying neural mechanisms to look for. It was suficient that lesions within Vl led to blindness, while those in the ‘visuo-psychic’ cortex led to the syndrome of mind blindness (seelenblindheit), later termed agnosia, a condition in which a patient was deemed to be able to ‘see’ but not to ‘understand’ what was seen. This view, approved of by both Henschen and Holmes, was well summarized by Campbell [4] in 1905 when he spoke of two areas “ ...one specialized for the primary reception of visual sensations, and the other constituted for the final elaboration and interpretation of these sensations”. Flechsig had believed, though without compelling evidence, that the role of visual association cortex was to associate visual signals with signals derived from other sources, and endowed it with a certain level of consciousness. Other neurologists, including Campbell; Holmes and Henschen, had believed, again without much evidence, that it would associate the received visual ‘impressions’ with similar past ‘impressions’, thus leading to ‘understanding’. By the time neurophysiologists got hold of the idea, they perpetuated the earlier views without considering the evidence, which was in any case scant. Thus, Clare and Bishop [5] studied an area well removed from the primary visual cortex in the cat and “inferred [it] to comprise an association area relating optic and acoustic activity” although no acoustic activity was studied there. In summary, these views - however much they may have differed in detail - separated the process of ‘seeing’ from that of ‘understanding’ and attributed a separate cortical seat to each. It is this fundamental concept that more recent work on visual ‘association’ cortex has challenged.

The challenge cortex

to the concept

of association

Perhaps the first step in the revision of this view came from the demonstration that the visual association cortex, far from being the single area which its uniform architecture and relatively late myelination had implied, consists in fact of multiple visual areas (for a review, see [6] ). This demonstration raised the possibility that the cerebral processes involved in vision are much more complex than the one implied in the dual concept of the early neurologists. The fundamental turning point came, however, with the demonstration that the visual areas of the ‘association’ cortex undertake different tasks, not the same task at ever-increasing levels of complexity, as was implicit in the earlier doctrine of exclusive hierarchies [7,8]. Thus, area V5 is specialized for visual motion [9], while area V4 is specialized for cblour and form in association with colour [l&14]. V3, by contrast, is specialized for dynamic form [ 15,161. A functional specialization is also characteristic of the prestriate cortex of man [ 171, but this is not to imply that the specializations enumerated above are the only functions of these areas. That the areas of the visual association cortex (now better referred to as the prestriate cortex) receive parallel inputs from Vl [ 111, not only served to emphasize the fact that

the cortex undertakes several visual operations in parallel to construct the visual image in the brain, raising the fundamental question of how the specialized visual areas interact to provide the unitary visual image in the brain, but strongly suggested that the functional segregation evident in prestriate cortex would be mirrored somehow in area Vl itself [ 111, even if the compelling evidence for such a supposition was to come only several years later [ 181. No one has been able to show that the cells in the above prestriate areas are influenced to any extent by stimuli belonging to another modality, say, olfactory or auditory. Equally, no one has yet been able to show that the memory of past visual experiences or stimulation is crucial for the activation of cells in these areas, which is not the same thing as saying that such influences may not be found to be crucial in the future. Thus, the speculations of the early neurologists, whether of Flechsig or of Holmes, concerning the ‘understanding’ cortex do not gain much support from the current physiological profile of the visual areas of prestriate cortex.

Does visual association fundamentally different

cortex use a strategy?

What is it that distinguishes the visual areas of the prestriate cortex from area Vl? Is it a qualitative difference or a quantitative one? The single most striking feature of prestriate visual areas is their specialization. While anyone wanting to explore the functional organization of area Vl or area V2 (which surrounds it) with an electrode will encounter cells with many different properties [l’$-211, even if cells dealing with a given attribute are grouped together, the properties of cells in individual prestriate areas are more homogeneous. The cells of area V5, for example, are overwhelmingly directionally selective and uninterested in colour, whereas those of area V4 are overwhelmingly wavelength selective [ 12-15,21,22]. The initial temptation therefore would be to suppose that the role of prestriate cortex is to segregate or ‘dissociate’ signals rather than associate them. But the segregation of visual signals belonging to different sub-modalities of vision is not a radically new strategy employed by the presmate cortex, even if it was first demonstrated there. More recent studies show that visual signals are also segregated into sub-compartments in area Vl [IS], from which the specialized areas receive their cortical input, as well as in area V2, which surrounds area Vl and projects to the same specialized visual areas [ 23-251. Hence, the segregation of visual signals in the prestriate cortex is not a novel strategy but a continuation of the strategy employed at earlier levels of the visual pathways. The next striking feature of the prestriate areas is that, compared with the striate cortex, cells in the former have larger receptive fields. This is almost certainly the consequence of the need to collect information from larger parts of the field of view. But this is not a strategy developed in, or unique to, the visual areas of the prestriate cortex. Indeed it is a hallmark of the visual pathways in general. The simple cells of the striate cor-

The visual

tex have larger receptive fields than those of the lateral geniculate nucleus from which they receive input, and the complex cells have larger fields still [ 71. The strategy is continued well beyond the prestriate cortex, for cells in the visual areas of the inferior temporal cortex and the pati& cortex have yet larger fields (for examples, see [ 26,271). There is, next, the question of complexity in the cellular responses - itself a consequence of the enlargement of receptive fields and the collecting of information from large parts of the field of view. For example, the responses of cells in V4 correlate with the perception of colours, whereas the responses of their counterparts in ~1 do not 1221; the generation of colour is itself a more complex process than the registering of the presence and intensity of different wavelengths [ 281, and to this extent the responses of V4 cells are more complex than those of Vl. Equally, the cells of area V5, or at least some of them, respond to the coherent motion of an entire object, whereas their counterparts in Vl (from which V5 receives its input) respond only to the components of which the whole is made [29]. The consequence of this is that the relevant cells of Vl may signal a direction of motion which is not identical to the direction of motion of the entire object. But this complexity is not a new departure; instead it is the continuation of a process which starts in the retina itself, to the extent that the photoreceptors have simpler receptive fields and simpler responses than the ganglion cells into which they feed. In a continuation of this process, the orientation selective cells of Vl have more exigent requirements than the cells of the lateral geniculate nucleus,

The effects of cortical

lesions on vision

If, therefore, the anatomical and functional profile for the prestriate cortex that we have built up over the past two decades does not suggest a radical functional departure from the kind of functional organization found in area vl, is there any plausible reason to suppose that seeing is vested in Vl and understanding in the surrounding cortex, apart from the fact that lesions in Vl lead to nearly complete blindness, whereas those in visual association cortex do not? Insights into this problem may be gained by a renewed study of the effects of cortical lesions on vision. I do not refer here to the carefully controlled experimental lesions in monkeys, which have been the single worst guide to the organization of the visual cortex imaginable, but to the natural uncontrolled lesions in human brains produced by gunshot wounds or cerebral accidents which, paradoxically, have been a lot more informative. That lesions in Vl, and possibly also V2 (see [30] ), should cause a total blindness is relatively easy to explain - such lesions do not usually spare a given subcompartment, for example, the blobs of Vl or the thin stripes of V2 in which cells concerned with colour are concentrated, but instead involve all compartments, The consequence is that cells dealing with all the attributes

association

cortex

Zeki

of vision are destroyed, leading to blindness. Naturally, much the same thing would happen if all the specialized visual areas were to be destroyed. But this could only be the consequence of a lesion that is so large that it would amount to an hemispherectomy, assuming it not to have led to death. The consequence of the more common type of lesion, one restricted to one of the specialized visual areas, is a blindness for the corresponding visual attribute. A striking example is provided by the syndrome of cerebral achromatopsia, or acquired cortical colour blindness following lesions in area V4, which is located in the fusiform gyrus (for a review, see [31]). No less specific is the syndrome of cerebral akinetopsia or motion imperception [17]. This is the consequence of lesions in area V5 [32], which is located laterally and ventrally in man and, perhaps surprisingly, in a zone (Feld 16) that Flechsig considered to have been myelinated at birth and therefore a primordial area. But is the consequence of such lesions a radically new kind of syndrome, concerned with ‘understanding’ motion or colour, or is it, as the physiology suggests, a more complex example of the same phenomenon? Can meaning and understanding be attached to vision only through the participation of the visual areas of prestriate cortex, or can areas Vl and V2 contribute explicitly to both seeing and understanding the visual world? Examination of the so-called agnosic patients, as well as patients with akinetopsia and achromatopsia, leads one to a general theory of residual vision [6]. This supposes that each visual area contributes explicitly to visual perception (that is in a way that requires no further processing) and that the patient is able to see and to understand in exact relationship to that contribution and no more. Moreover, the theory supposes that the ability to see a particular visual attribute, for example motion, is not dependent upon the integrity of the entire visual pathway, up to V5 and possibly beyond. Instead, it supposes that, if the latter is destroyed, the patient will be able to see visual motion in proportion to the direct and explicit contribution to vision that the intact parts of the system, including areas Vl and V2 which feed V5, are able to make. When area V5 is destroyed in an akinetopsic patient, the directionally selective cells of area Vl that feed it are not; consequently, the patient is aware of the presence of motion but cannot make much of it (see [33]). Equally, an achromatopsic patient is able to discriminate between different wavelengths, but is unable to combine this information to construct colours (see [6] ), a deficit remarkably similar to the consequence of lesions in macaque area Vd [34]. Indeed, the orientation selective cells of Vl, which are able to respond to pure chromatic borders [35,36], would endow an achromatopsic patient with an intact, or partially intact, Vl, with the ability to detect the orientation of the boundary between two equiluminant surfaces of different colour, even though the colours on either side of the boundary remain identical to him [ 371, Additional support for this view comes from comparing the nature of the so-called agnosia in carbon monoxide

157

158

Cognitive

neuroscience

patients, and in patients with large lesions in the prestriate cortex. It is almost certain that, in the former, area Vl is much a&ted, although of course the areas of the prestriate cortex are probably also compromised. The consequence is a profound defect in form vision, with patients hardly able to recognize or copy simple geometric figures such as triangles or squares [38]. The interpretation that I have given to this syndrome is that even the elementary kind of integration and association which Vl is responsible for - the generation of cells especially responsive to straight lines - becomes compromised. The nature of the agnosia in patients with lesions in the prestriate cortex is markedly different. Now the patients can even draw complex figures, without being able to recognize the final product! Yet how is it that these patients draw? There is good agreement in the literature that the drawing is piecemeal, small segments of the picture, or of its outline segments that the patient can see and understand - being drawn, one after another. Once drawn, the patient can still only recognize small segments of his drawing and not its entirety. It is the simple components of a figure that the patients are able to see and to understand because the integrative mechanisms necessary to construct simple forms, such as lines, are intact, while those needed for more complex forms are compromised. The intact area Vl makes a direct and explicit contribution to perception, but the destroyed prestriate areas are not able to perform their function and hence associate the various elements and organize into a larger whole. The patient consequently sees, understands, and is able to draw only in proportion to what his intact Vl allows him to do.

Vl is as important and understanding

for conscious

Conclusion I have concentrated in this brief review on the conceptual doctrine that we have inherited about visual ‘association’ cortex, and hence about vision, and tried to show that the separation between seeing and understanding what is seen - a concept deeply tied to that of association cortex - is not easy to achieve. One can make a fair argument - from ordinary visual perception, from anatomy and physiology, and from the study of the dam aged brain - that seeing and understanding are part of the same process, though no one would wish to deny that there are many instances in which we see things that we do not properly understand. The concept of visual association cortex, in the sense intended by the early neurologists, is therefore perhaps now best abandoned. But, in doing so, we must acknowledge that those who originated the concept and speculated about it have a high and honourable place in the history of our subject. If the concepts that they fought for with such conviction and passion have turned about to be false, we must reflect that the concepts that we today fight for, with no less conviction and passion, may equally turn out to be ilawed speculations in the vast ocean of the unknowns that the visual brain still is.

References 1.

FLECHSIG

2.

FLECHSIG

3.

LISSAUER H:

4.

CA~~PBELL AW:

experience

There is, finally, the evidence from blindsight (for a review, see [39]), when subjects blinded by lesions in area Vl are nevertheless able to discriminate some visual stimuli correctly, even though they commonly have no conscious awareness of having seen anything at all. It is very likely that when such blindsight subjects detect the direction of motion, signals relating to motion reach area V5 directly, in the direct pathway from the lateral geniculate nucleus to V5 [40,41]. Direct physiological evidence has shown that cells in ~5 can maintain the property of direction selectivity, although in a compromised state, after ablation of V5. Yet the fact that blindsight patients have no conscious awareness of the visual stimulus, and hence no understanding of it, suggests either that preprocessing in Vl is a necessary step for stimulation to gain consciousness, or else that it is necessary for the operations performed by V5 to be referred back to Vl in the diffuse return projection linking V5 to Vl [42]. Of course, it is possible that both processes are necessary. Whatever the case, the results suggest that Vl is an essential part of the process that allows conscious experience and hence understanding of vision.

P: Developmental (Myelogenetic) Localisation of the Cerebral Cortex in the Human Subject. Lancet lM1, 2:1027-1029. P: Gehirnphysiologie und Willenstbeorien. 1905. Translated by Von Bonin G. In Some Papers on the Cerebral Cortex, Fifth International Psychology Congress. Springfield, Illinois: 1905: 73-89.

Ein FaII Beitrage zur Theorie 1890, 21:221-270.

Histological Studies on the LocaIisation of Function. Cambridge: Cambridge University Press;

Cerebral 1905. 5.

Cm

von Seelenblindheit Nebst Einem Derseiben. Arch Pqhbtr Nervenkr

MH,

BISHOP

GH:

Area Secondarily Activated physol 1954, 17~271-277. 6.

S: A

ZEKI

Vision

Responses from an Association from Optic Cortex. J Neuro

of the Bruin.

Oxford:

Blackwell

Scientific;

1993. 7.

DH, WIESEL TN: Receptive Fields, Binocular tion and Functional Architecture In the Cat’s Visual J P@wol (Land) 1962, 160:106154.

8.

HUBEL

9.

ZEKI SM: Functional

10

ZEKI SM: Colour

HUBEL

DH, WIESEL TN: Receptive Fields and Functional Architecture in Two Nonstriate Visual Areas (18 and 19) of the Cat. J Neurq0hyiol 1965, 28:229-289.

Organization of a Visual Area in the Posterior Bank of the Superior Temporal Sulcus of the Rhesus Monkey. J PLysiol (Land) 1974, 236:54’+573. tex.

11

InteracCortex.

Brain

Coding in Rhesus Res 1973, 53~422-427.

ZEKI SM: The

Monkey

Prestriate

COr-

Functional Organization of Projections from Striate to Prestriate Visual Cortex in the Rhesus Monkey. Cold Smim Hat-b Ouant Biol 1975. 40:591a.

The

visual

association

The Retinex Theory Gt Br 1974, 47:23-25.

of Colour

12.

Z,EKI S: The Distribution of Wavelength and Orientation Selective Cells in Different Areas of Monkey Visual Cortex. Proc R Sot Lond [Biol] 1983, 217:44+470.

13.

DESIMONE R, SCHEIN SJ: Visual

14.

Properties of Neurons Sensitivity to Stimulus Form.

V4 of the Macaque: pbysiol 1987, 57:835i868. Macaque.

J Neurosci

2~x1 SM: Functional the Rhesus Monkey.

16.

ZEKI S, SHIPP S: The Functional tions. Nature 1988, 335:311-317.

18.

Specialization in the Visual Nature 1978, 274:423-428. Logic

in Human

LMNGSTONE

Color

Visual

MS, H~REL

System

DH:

in the Primate

Cortex

of Cortical

ZEKI S, WAIXON JDG, LUECK CJ, FR~STON FRACKOWW( RSJ: A Direct Demonstration

Specialization 11641-649.

29.

Cortex.

KJ,

30.

of

ConnecKENNARD

Anatomy and Visual Cortex.

Physiology J Neurosci

1988,

BA~ZER JS, RORINSON DL, Dow

18 Neurons 1977, 21.

Binocular

in Awake,

ZIHL

34.

in Rhesus

Monkey

22.

in the

Cerebral Cortex

tion of Cells in Monkey Visual Colours. Neuroscience 1983, 9:741-765. 23.

LMNGSTONE

Stereopsis

in a Subregion

1987, 24.

Field 1985,

Segregation Properties

of Efferent Connecin Visual Area V2 of

Handbook

Functions of Infero-Temporal Cortex. In of Sensory P&siology, vol 7, pan 3. Edited by Jung

Springer;

1972.

MOUNTCASTLE VB, MOLTER BC, STEINMETZ MA, DUFN

Gall

WE,

Cowan

WM.

New

York:

Wiley;

1984:

MAI

ZIHL

Field

N:

Brain

Selective

Brain

J: The

Defects. 1990,

Disturbance

Damage.

Brain

of

1983,

‘Motion-Blind’ Patient: Filters. J Neurcuzi 1989,

Temporal

WALSH V, CARDEN D, BUTLER SR, KUUKOWSKI

on the Visual Abilities and Colour Constancy.

Effects of Hue DisRes 1993,

JJ: Tbe

of Macaques; Bebav Brain

P, KRUGER J: Responses

Goup&

of the

Monkey

to Pure

42:850&O.

THORELL

LG, DEVAL~IS

HEY~~~D

CA, Cowxu

in a Cortically

RL, ALBRECHT

HUMPHREYS GW,

Study

of Visual

Blind

DG:

Spatial

and Luminance

A, NEWCOMBE

Colour

ciates;

1987.

R~DWCH

Aggnosia.

L: Blindsight.

YUKIE M, IWAI E: Direct

Geniculate Monkeys.

42.

Color

Mapping of Stimuli.

F: Chromatic

Observer.

EurJ

Discrimi-

Neurosci

3:802-812.

40.

41.

of Cells in Foveaf Visual CorColor Contrast. J Neur@ysiol

MJ: To See but Not Hillsdale, New Jersey:

Oxford:

Clarendon

to See. A Case Erlbaum

Press;

Asso-

1986.

Projection from the Dorsal Lateral Nucleus to the Prestriate Cortex in Macaque J Comp Neural 1981, 201:81-97.

FRIES W: The

Projection cleus to the Prestriate Proc R Sot Lond {Biol]

from Cortex 1983,

SHIPP S, ZEKI S: The Organization Areas V5 and Vl in Macaque J Neurosci 1989, 1:30’+332.

the Lateral Geniculate Nuof the Macaque Monkey. 21317380.

of Connections Between Monkey Visual Cortex. Eur

CJ: Look-

ing and Seeing. The Visual Functions of the Parietal Lobe. In Dynamic RFpects of Neocortical Function. Edited by Edeiman GM,

Visual

Achromatopsia

Bilateral

and

WEISKRANTZ

317:58-61.

GROSS CG: Visual

CL,

Spatial

39.

315~322-325.

DE YOE EA, VAN ESSEN DC:

R. Berlin: 27.

38.

of Pathways Leading from Area V4 and V5 of Macaque Monkey Visual Cortex.

BAKER

Low-Level 9:1628-1640.

1991,

of Form, Color and Area 18. J Neurosci

SHIPP S, ZEK~ S: Segregation

tions and Receptive the Macaque. Nature 26.

of Priiate

HESS RH,

nation

Reacand

7:33783415.

V2 to Areas Nature 1985, 25.

MS: Segregation

HUBEL DH,

Cortex: the to Wavelengths

159194.

R

Gtzt

Monkey Vl Cells with Pure Vision Res 1984, 24:751l769.

and Diversity of Structure and Function Prestriate Cortex. J Pbysiol (‘Land) 1978,

Coding

D,

after

1979,

37.

ZEKI S: Colour

CRAMON

Vision

tex

Visual Responses of Area Monkey. J Neuropbysiol

2773273-2900.

Proc

in press.

36.

ZEKI SM: Uniformity

J, VON

Vision.

Patterns.

of Cerebral

Movement 106:313-340.

V4 Lesions crimination 35.

40:10241037.

Visual

JC, HOYT WF: Quadrantic 1991, 114:1703-1718.

32.

33.

EH,

159

Zeki

MS, NEWSOME WT: The In Pattern Recognition by Chagas C, Gattass R, Gross CG. PonCitta de1 Vaticano; 1984: 117-151.

ZEKI S: A Century 113:1721-1777.

Mechanisms and Dispar-

8:4531-4550.

BM:

Behaving

Correlation

ADDLESON

31.

C,

of a 1984,

JA,

Academy:

HORTON

Brain

of Functional J Neurosci 1991,

POGGIO GF, GONZALEZ F, KRAUSE F: Stereoscopic

in Monkey Visual Cortex, ity Selectivity. J Neurosci 20.

MOVSHON

Analysis of Moving Mechanisms Edited

4:30%356. 19.

EH:

tifical

Properties of V4 Neurons 1990, lo:33693389.

15.

17.

IAND

Instn

in Area J Neuro

SCHE~N SJ, DESU~ONE R: Spectral

in the

28.

cortex

S Zeki, Department of Anatomy College London, Gower Street,

and Developmentai Biology, London, WClE GBT, UK.

University