London CYVlE 6BT. animals, by plotting receptive field positions for different recording sites. Results

Transcription

1 J. Physiol. (1978), 277, pp With 3 plates and 14 text-figures Printed in Great Britain THE TOPOGRAPHIC ORGANIZATION OF RHESUS MONKEY PRESTRIATE CORTEX BY D. C. VAN ESSEN* AND S. M. ZEKIt From the Department of Anatomy, University College London, London CYVlE 6BT (Received 31 March 1977) SUMMARY 1. The topographic organization of prestriate visual cortex in the rhesus monkey has been studied both anatomically, by determining the pattern of termination of fibres passing through the corpus callosum, and physiologically, in the same animals, by plotting receptive field positions for different recording sites. Results are displayed on two-dimensional, 'unfolded' maps of the cortex in the dorsal half of the occipital lobe. 2. Transcallosal fibres terminate in a narrow strip of cortex along the boundary between striate and prestriate areas and in a separate, broader, zone occupying much of the anterior bank of the lunate sulcus, the annectant gyrus, and the parietooccipital sulcus. The detailed pattern of inputs is highly complicated but shows considerable similarities from one animal to the next. 3. Physiological recordings confirmed earlier reports that regions where transcallosal fibres terminate correspond to representations of the vertical meridian in the visual field. This relationship is most precise along the striate-prestriate boundary and along the boundary of area V3 farthest from VI; it is less precise within area V4, where the pattern of transcallosal inputs is more complex. 4. A distinct, topographically organized visual area, named V3A, was found in the region between areas V3 and V4 in the lunate and parieto-occipital sulci. Area V3A differs from V2 and V3 in that both superior and inferior visual quadrants are represented in a single region of the dorsal occipital lobe. 5. The contralateral visual field is represented in a suprisingly complex fashion in areas V3A and V4. Within each area there are multiple representations of some, but perhaps not all, parts of the visual hemifield. It is unlcear whether V3A and V4 should be more appropriately considered as sets of distinct sub-areas, each representing only a portion of the hemifield, or as larger areas with complicated internal topographies. 6. Most cells in areas V2, V3 and V3A are orientation selective but not selective for stimulus colour or direction of movement. In contrast, area V4 contains a higher incidence of colour selective cells and a lower incidence of orientation selectivity. These results support the notion of a functional division of labour within the prestriate cortex. * Present address: Division of Biology, California Institute of Technology, Pasadena, California, U.S.A. t Head Fellow of the Royal Society. 7 PHY 277

2 194 D. C. VAN ESSEN AND S. M. ZEKI INTRODUCTION Although the analysis of visual information has long been accepted as the major function of the occipital lobe in primates, it is only in recent years that a clear understanding has begun to emerge of the functional organization of the portion of occipital cortex which lies outside the striate, or primary visual, cortex. Brodmann's original conception of the prestriate occipital cortex was that it consisted of two cytoarchitectonic areas, 18 and 19, arranged concentrically around area 17, the striate cortex (Brodmann, 195). It is clear, however, from the topography of projections from area 17 that there are at least four distinct visual areas (V2, V3, V4 and a striate-receptive area in the superior temporal sulcus) contained within Brodmann's areas 18 and 19 (Zeki, 1969, 1975; Cragg, 1969), each containing a separate representation of the visual fields. Moreover, physiological studies have shown that within these different visual areas specific types of functional analysis are emphasized, such as binocular disparity in V2 (Hubel & Wiesel, 197), colour in V4 (Zeki, 1973, 1975) and motion in the visual field in the striate-receptive area of the superior temporal sulcus (Zeki, 1974a, b). In order to pursue anatomical and functional studies of this type it is obviously important to know more about the over-all extent of each visual area and exactly where the boundaries between areas lie. This has been a difficult task in the rhesus monkey, partly because of the complex way in which the occipital cortex is folded into sulci and gyri and partly because there are no clear cyto-architectural boundaries in the prestriate cortex with which to delimit the borders of areas defined by visual function or by topographic mapping of the visual field. One way of overcoming the latter of these difficulties is to use the interhemispheric connexions running through the corpus callosum as an anatomical reference system over the extent of the prestriate cortex. If callosal fibres terminate selectively at regions where the vertical meridian of the visual field is represented, as is the case for areas VI, V2 and V3 (Whitteridge, 1965; Zeki & Sandeman, 1976), then the shape and arrangement of regions of the prestriate cortex receiving transcallosal projections should provide valuable information concerning the extent of each visual area and their boundaries. In the present study, we have undertaken a series of combined physiological and anatomical experiments, using the earlier approach (Zeki, 1975; Zeki & Sandeman, 1976) of mapping visual receptive fields during multiple electrode penetrations into the prestriate cortex following surgical transaction of the corpus callosum. In order to relate visual field topography to the over-all pattern of transcallosal projections, two-dimensional cortical reconstructions were made extending over a large portion of the dorsal prestriate cortex. The results have helped to establish the existence of a new visual area, V3A, and to clarify some of the topological relationships among different areas. METHODS The details of surgical, electrophysiological and histological techniques have been described elsewhere (Zeki, 197, 1974a). In each of ten rhesus monkeys, weighing between 2- and 2-5 kg, the posterior (splenial) part of the corpus callosum was transacted 6 days before the recording experiment, a survival time that is optimal for showing degenerating callosal fibres (Zeki, 197). For electrophysiological recording the animals were anaesthetized with sodium pento-

3 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 195 barbitone, paralysed with gallamine triethiodide (5 mg/kg. hr) and artificially respirated to maintain end-tidal CO:, at about 5%. Supplementary doses of anaesthetic were given during the experiment to maintain an adequate level of anaesthesia. To prevent cerebral oedema and consequent herniation, sodium decamethasone (.1 mg/hr) was administered intravenously. A region of cortex approximately 1 cm across and centred over the lateral portion of the lunate sulcus was exposed and covered with a chamber filled with 2-5 % agar-in-saline solution. Since our main objective was to study visual field topography over as large a region of cortex as possible, we used low-impedance (1-2,um exposed tip) tungsten-in-glass micro-electrodes (Merrill & Ainsworth, 1972) which enabled us to sample the activity of clusters of cells and, occasionally, single neurones. Most cell clusters could be driven easily by simple visual stimuli (stationary and moving slits, spots, and edges of light of variable orientation and colour) projected onto a tangent screen upon which the monkey's eyes had been focused by means of auxiliary lenses. All cell clusters were tested for colour preferences by interposing either Wratten or interference filters in the light path. Those which responded only to stimuli within a restricted portion of the visible spectrum were classified as colour-coded. Cells which responded to all colours and to white light but which showed a strong preference for a limited range of colours were classified as colour biased. Finally, cell clusters which responded equally well to red, green, blue and white stimuli were classified as non-colour-coded. The foveal projection of each eye onto the tangent screen was determined using an ophthalmoscope which could be rotated 18 after being centred on the fovea. We estimated the accuracy of this method to be within -5, based on repeated plots of foveal positions and comparisons of receptive field positions for the two eyes for binocularly driven cells. The foveal positions were checked frequently during the course of the experiment. Usually there was little drift (less than 2-3o) even over periods of many hours. Up to fifteen separate electrode penetrations were made in some experiments, and in order to identify individual tracks small electrolytic lesions were made at appropriate depths by passing 5-1,aA DC current through the electrode for 5-1 sec. At the end of the physiological experiment the animal was perfused through the heart with 4 % paraformaldehyde in phosphate buffer. Serial frozen sections were cut at 3 jum and stained for degeneration according to the procedure of Wiitanen (1969). Cortical reconstructions. The present study is concerned with the dorsal half of prestriate cortex, most of which is buried within a deep fissure that runs just anterior to the border of striate cortex and forms a semicircular arc around the upper part of the hemisphere (PI. 1A). Although the fissure runs continuously from the lateral to the medial surface of the hemisphere, it is labelled separately as the lunate sulcus laterally and the parieto-occipital sulcus medially. Because so little of the cortex in this region is exposed on the surface of the hemisphere, it was desirable to find some means of 'opening up '- the sulci and making their interior visible, so that continuous regions of degeneration seen in different sections could be traced on a single cortical map. The approach we used involved first making a set of drawings of the contour followed by the IVth cortical layer (where most callosal fibres terminate; Zeki, 197), for a series of sections spaced at regular intervals (usually -25 mm) through the region of interest. The drawings were made at a final magnification of 2 x, using a Zeiss drawing tube coupled to a 2-5 x microscope objective. The individual contours were then transferred to a single sheet with each section contour aligned so that it was spaced by the appropriate distances from neighbouring contours. This gave us a composite drawing of the entire region. The alignment of contours was relatively easy for those cortical regions where there were no sharp folds, such as the anterior and posterior banks of the lunate sulcus, but was much more difficult in regions of high curvature such as the bottom of the lunate sulcus and the annectant gyrus. To prevent contour lines from crossing over one another it was generally necessary to straighten out regions of high curvature. Text-fig. 1 illustrates the extent to which this was done in order to reconstruct a region covering most of the prestriate occipital cortex in the dorsal half of the hemisphere. On the left are drawings of two of the horizontal sections used in making the map, and on the right is the final composite map. The more dorsal of the two sections shown (A, upper left) lies dorsal to the annectant gyrus. At this level the continuity between the lunate and parietooccipital sulci can be seen, and the path followed by layer IV is divided into separate posterior and anterior routes (red and green lines). The other section (B, lower left) is only 4 mm ventral to the first, but it intersects the annectant gyrus, which is interposed between the lunate and parieto-occipital sulci. Consequently, the contour lines are broken into lateral and medial 7-2

4 196 D. C. VAN ESSEN AND S. M. ZEKI portions occupying, respectively, the lunate and parieto-occipital sulci (yellow and blue lines). In the cortical map on the right each pair of contour lines (red-green and yellow-blue) remains easily recognizable, since the transition from the sections to the map involves only the widening of angles and the straightening of bends, with no change of the length of individual contour lines. In relating landmarks on the cortical map to their positions in the intact hemisphere it is necessary to appreciate the manner in which the three co-ordinate axes are represented on the two-dimensional map. The posterior-anterior and medio-lateral axes are orientated in the figure in approximately the same way as in the individual horizontal sections, that is, leftright and up-down respectively, except that there are local variations in orientation according to the degree of unbending involved in laying out the map. The dorso-ventral axis, which represents the transition between sections, is orthogonal to the individual contour lines and thus takes on all possible orientations in one or another part of the map. For example, an arrow pointing from more dorsal to more ventral sections would be aimed towards the right in the posterior banks of the lunate and parieto-occipital sulci, towards the left in the anterior banks of the same sulci, and upwards and downwards, respectively, in the medial and lateral portions of the annectant gyrus. The distance between X and X' on the red line in A is identical to the distance between X and X' on the red line in the central reconstruction. To get from X to X', one would have to move from lateral to medial. Similarly, the distance between Y and Y' on the drawing of the horizontal section in B (yellow line) is identical to the same distance in the central reconstruction, but to get from Y to Y' one would have to go from medial to lateral (and slightly anteriorly) in the lunate sulcus. On the other hand, to get from X to Y, one would have to go from dorsal (X) to ventral (Y). Similarly to get from W to Z would entail going from a ventral level (W) to a more dorsal level (Z) in addition to going from medial to lateral. The accuracy with which sections could be aligned during construction of the cortical map was within a fraction of a millimetre. In regions where even greater accuracy was needed in order to follow complex patterns of degeneration, every section was drawn and aligned either with respect to sharply defined patches of degeneration or, if present, with respect to electrode tracks passing obliquely through the sections. It was necessary to make many trial drawings of each map before reaching a stage at which all of the sulci and gyri were represented without severe distortions. In the final maps only the contours for sections spaced 2 mm apart are actually drawn, although much closer intervals were used in making the map and for plotting degeneration. With respect to accuracy, where areas of cortical surface lie in a vertical plane, and therefore are at right angles to the horizontal plane of section, the situation is simple. Horizontal sections through such a plane will produce a series of contours which can be exactly represented as parallel, and separated by a distance scaled from the original distance between sections. When such vertical planes run mediolateral as in parts of the posterior wall of the lunate sulcus they can be represented on paper as spaced from above downwards and when they run anteroposterior, as in the lateral wall of the annectant gyrus, they can be represented from left to right. Unfortunately most areas of cortex form part of planes inclined to the primary axes of the cortex, and here the situation is more complicated. Even here, however, points on these contours can be represented by eye depending on the slopes of their original planes. This is a Text-fig. 1. A method of constructing a two-dimensional 'unfolded' cortical map of the prestriate cortex. Drawings of representative sections at two horizontal levels are shown in upper and lower left corners. In the more dorsal section (A, upper left) the contour followed by layer IV was divided into separate posterior (red) and anterior (green) sections. The shape of each contour was then changed slightly in order to fit into the composite cortical map (centre). In the more ventral section (B, lower left) layer IV contours are divided into separate lateral (yellow) and medial (blue) sections. The pattern of degeneration displayed on this same cortical map is shown in Text-fig. 2. LS, lunate sulcus; CS, calcarine sulcus; POS, parieto-occipital sulcus. The prestriate cortex of the right hemisphere has been reconstructed in this and subsequent Figures.

5 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 197 CO) CO~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Co ~~~~~~~~ N)~~~~~~~~~~~~~~C en~~~~~~~~~~~~~~oc oo ~~~~~~~~~~~~~~~~~~~~~~~ ow~~~~~~~~~~~~c co~ ~ ~~~.~

6 198 D. C. VAN ESSEN AND S. M. ZEKI process of opening out or flattening curves connecting up the rigorously defined representation of vertical planes. In addition, making this surface on a flat plane implies that the curved surface it represents can be flattened without stretching, and this is very unlikely to be true. Distortion has therefore to be introduced to flatten it. In spite of this, distances along contours of course remain exact, and distances across contours are not more than double the true value. In general, the distortions in the map are worst where contour lines are crowded together, and even in such regions one can compensate for most of the compression by counting the number of contour lines crossed rather than simply measuring the distance on the map. There is considerable flexibility in the choice of which regions are to be most accurately represented on the map, and in general we tried to minimize the distortions in the anterior bank of the lunate sulcus, where the pattern of degeneration is most complex. For portraying the pattern of interhemispheric connexions it was necessary to represent systematically the wide variations in density of degeneration seen in silver-stained sections. In each section there was a continuous gradation from very sparse to very dense levels of degeneration. We found, however, owing in part to the slight variability in the quality of staining of different sections, that under simple microscopic inspection it was difficult to grade the density of degeneration more finaly than into three levels: dense, moderate and sparse. Representative examples of each of the three grades of degeneration density are shown in PI. 2. Although degeneration was seen in all cortical layers (Zeki, 197), the individual patches generally were densest and most sharply defined in layer IV, and the plots of transcallosal inputs illustrated in the results refer specifically to terminations in this layer. RESULTS It is evident from an earlier study (Zeki, 197) that the transcallosal inputs to the prestriate visual cortex of the rhesus monkey are arranged in a highly complicated pattern. This is especially so in the anterior bank of the lunate sulcus, in the region of the fourth visual area, where the pattern of connexions varies considerably from one section to the next. Although much information can be obtained during sequential examination of single sections, it is difficult to appreciate the overall arrangement of transcallosal inputs without making reconstructions from closely spaced sections. For example, P1. 3B is a photographic montage showing the pattern of degeneration in a restricted portion of the anterior bank of the lunate sulcus made from a series of dark-field photomicrographs spaced at -25 mm intervals. P1. 3A shows the degeneration in a single section situated at the level of the arrows in PI. 3B. The complexity of the pattern in both the section and the montage is evident at a glance, with islands of heavy degeneration surrounded by areas containing little or no degeneration, and it immediately raises the question of how the pattern of interhemispheric connexions is related to visual field topography. If callosal fibres in this region terminate selectively at the vertical meridian representation, as is the case at the V1/V2 border, then it follows that the vertical meridian, and presumably the rest of the visual hemifield, must be represented in a more complex fashion than has been found for other visual areas. If, on the other hand, there is no longer a correlation between callosal connexions and the vertical meridian representation, then the significance of the enormous local variations in the density of callosal inputs becomes unclear. In order to examine these possibilities more closely, we reconstructed, using a somewhat different technique, the pattern of degeneration over a larger region of cortex than shown in P1. 3. These results are presented in Part I. In Part II we shall describe the results of extensive visual field mapping in the same hemispheres for which cortical reconstructions were

7 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 199 made, which allowed a direct comparison between anatomical and physiological patterns of organization. Part I. The pattern of transcallosal inputs to the prestriate cortex In this section we describe the distribution of callosal fibre terminations in a limited portion of prestriate cortex encompassing the dorsal half of Brodmann's area 18. The total area reconstructed represents a compromise between the desire to map as much cortex as possible and the difficulty in obtaining an adequate two-dimensional representation of a large area of cortex. Anteriorly, we reconstructed the degeneration up to the crown of the prelunate gyrus (Text-fig. 1). This choice was not entirely arbitrary. The callosal inputs to the anterior bank of the lunate sulcus spread over the posterior crown of the prelunate gyrus but are usually quite separate from the inputs in the anterior crown of the prelunate gyrus and the posterior bank of the superior temporal sulcus (Zeki, 197). Medially, we reconstructed in detail the first two major patches of callosal fibre terminations anterior to area 17. There are other cortical regions receiving callosal inputs, especially in the intraparietal sulcus, but these have not been reconstructed for this study. Although it seemed at first that photographic montages of the type illustrated in P1. 3 would be the most satisfactory way of constructing a cortical map, we found two major drawbacks to this approach. First, it was difficult to make a montage of regions where the cortex was curved, so that the regions which could be covered in a single map were relatively small. Secondly, the density of degeneration was not represented accurately in the low-power dark-field photomicrographs, in that regions with sparse or even moderate degeneration usually appeared as dark as regions where there was no degeneration at all. Because of these difficulties we turned to the method of reconstruction based on aligning contour drawings of individual sections as described in the Methods and illustrated in Text-fig. 1. Detailed cortical maps were made for three separate hemispheres, as illustrated in Text-figs Two of the maps extend from the dorsal rim of the occipital lobe approximately to the ventral terminations of the lunate and parieto-occipital sulci (Text-figs. 2 and 3). The third map (Text-fig. 4) contains a representation of only the ventral two-thirds of these sulci. Solid black areas in the maps indicate regions of dense degeneration; large and small filled circles indicate, respectively, where the degeneration is moderate and sparse. It is worth emphasizing both the basic similarities in the pattern of degeneration seen in all three maps and the distinct individual regional differences between one map and another. The most important features common to all of the maps are (a) a narrow band of degeneration that runs along the V1/V2 (striate/prestriate) boundary at the left-hand margin of the map, (b) a wide zone of cortex, occupying the posterior banks of the lunate and parietooccipital sulci, which is free of degeneration, and (c) a broad region, occupying the anterior banks of the lunate and parieto-occipital sulci, which contains a complex network of degeneration. On the basis of the evidence to be discussed in Part II we believe that this pattern of degeneration reflects the arrangement of the distinct visual areas within this part of the prestriate cortex, with a single large visual area, V2, occupying a wide zone of cortex adjacent to the striate cortex and areas V3,

8 2 D. C. VAN ESSEN AND S. M. ZEKI V3A and V4, which have progressively more complex internal organizations, occupying increasingly more anterior regions of the prestriate cortex. The simplest region of degeneration in the cortical maps is that at the V1/V2 boundary. Most of the degenerating terminals are confined to a narrow strip at Text-fig. 2. The pattern of degeneration produced by callosal transaction, as displayed on a two-dimensional cortical map. In this and the following two figures regions containing dense degeneration in layer IV are shaded black; regions of moderate and sparse degeneration are indicated by large and small dots, respectively. Same hemisphere as in Text-fig. 1. LS, lunate sulcus; AG, annectant gyrus; POS, parietooccipital sulcus. the V1/V2 border and extend.5-1 mm into VI and up to 2 mm into V2. In addition, there are occasional 'fingers' of degeneration protruding 1-8 mm further into V2. The continuity between the protrusions and the principal strip of degeneration at the VI-V2 border could be established in every case by examination of serial sections, even though in individual sections there sometimes appeared to be

9 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 21 two completely separate patches of degeneration near the V1/V2 border. There was no obvious consistency from one brain to the next in the length, position or spacing of these fingers of degeneration. For example, the longest protrusion in any of the brains we examined was one located along the medial (upper) side of the hemisphere B A AB *@ *~~~~ Lunate Late~~~tmpral sulcus Anterior Text-fig. 3. Transcallosal inputs to dorsal prestriate cortex. displayed for a different hemisphere but in the same fashion as in Text-fig. 2. The degree of unfolding of the contour lines of section A (upper left, circles) can be judged by following the line marked with circles in the reconstruction. The decroe of unfolding of section B can be judged by following the line wsvith crosses. illustrated in Text-fig. 2, yet in the other two reconstructed hemispheres all of the prominent protrusions were confined to the lateral side of the hemisphere. At present we have no clear indication of the functional significance of these peculiar irregularities in the projections to the Vt/V2 border. Adjacent to the strip of degeneration along the Vt/V2 boundary is a large cortical

10 22 D. C. VAN ESSEN AND S. M. ZEKI region entirely free of transcallosal inputs. In the medial part of the hemisphere the width of this zone is about 8-12 mm. More laterally, in the region where the lunate sulcus becomes shallow and then terminates, the degeneration-free zone becomes much narrower and eventually disappears altogether. Although we have not constructed a map of the ventral half of the prestriate cortex, it is clear from earlier observations (Zeki, 197) that in the inferior occipital sulcus, ventral to the lunate sulcus, a gap once again opens up next to the V1/V2 strip of degeneration. Thus the region bounded by the two major zones of degeneration in the occipital LS Text-fig. 4. A third example of the pattern of transcallosal inputs to dorsal prestriate cortex, displayed in the same manner as for Text-figs. 2 and 3. This map covers only the portion of the dorsal occipital lobe between the regions indicated by heavy arrows in the drawing of the hemisphere (inset); thus it is less complete than the other two maps, which extend up to the most dorsal part of the occipital lobe. Thin interrupted lines in the cortical map indicate the region shown in the photographic montage of Pi. 2. LS, lunate sulcus; 13, inferior occipital sulcus; AG, annectant gyrus; POS, parieto-occipital sulcus; IPS, intraparietal sulcus. Conventions as in previous Figures.

11 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 23 lobe has a constricted segment laterally which opens up both dorsomedially and ventromedially. It is likely that area V2 has a similar configuration (Cragg, 1969; Zeki, 1969). In each of the three cortical maps illustrated, degenerating terminals are present in well over half of the anterior banks of the lunate and parieto-occipital sulci. This region can be subdivided into three major zones, as illustrated in Text-fig. 5. The Figure shows the same cortical map as in Text-fig. 3, except that, for simplicity, variations in the density of degeneration were ignored, and all regions where degeneration was present were shaded uniformly. Arrows point to several zones of degeneration that could be identified in all of the cortical maps. Zone 1 runs along the V1/V2 border, and was described above. Zones 2, 3 and 4 form a continuous band that almost completely encloses a degeneration-free region within the central part of the map. Our reasons for breaking this one broad strip into three separate zones are based not on the pattern of degeneration itself, in relation to which the subdivisions are somewhat arbitrary, but rather on the relationship of the map to the physiological results described in Part II. Zone 2 extends from the anterior bank of the lunate sulcus, across the annectant gyrus, and into the posterior bank of the parieto-occipital sulcus, where it is continuous with zone 3. Zone 3, in turn, extends from the parieto-occipital sulcus, across the intra-parietal sulcus, and merges with zone 4 in the anterior bank of the lunate sulcus. It is possible that zone 4 may have to be subdivided (see Zeki, 1978 a), but for the present we regard it as a single zone. We emphasize that, despite the lack of firm grounds for drawing precise boundaries between zones, this formulation nevertheless provides a convenient basis for relating the pattern of degeneration to the mapping of visual topography described below. It was obviously of interest to see whether the similarities in individual patterns of degeneration extended to finer details than just the general grouping into several broad zones. In each of the maps zones 3 and 4 are wider than zone 2 and have a more complex internal structure. It is difficult, however, to see any striking consistency from one hemisphere to the next in the sizes, shapes and positions of the various patches of degeneration that make up each zone. The only obvious pattern we have noticed is that in Text-figs. 3 and 4 most of the dense patches of degeneration are elongated and are orientated orthogonal to the contour lines and thus approximately along the dorso-ventral axis (2:-8: in zone 4 and 5:- 11: in zone 3). This feature is not evident in Text-fig. 2, however. The degeneration produced by callosal lesions extends into several regions of the hemisphere not included in the maps of Text-figs Specifically, zone 3 of degeneration in the intraparietal sulcus and dorsal part of the superior temporal sulcus (middle right in Text-figs. 2 and 3) is continuous with large callosal-recipient areas located farther anterior, in Brodmann's areas 7 and 19. The degeneration in the shallow ventro-lateral part of the lunate sulcus (zone 4, lower left) can be traced to the posterior lip of the inferior occipital sulcus. This continuity has been described previously (Zeki, 197) and is particularly evident in the maps of Textfigs. 2 and 4. This feature is of particular interest in view of the similarities of the two regions in terms of their anatomical inputs (Zeki, 1971) and functional organization (Zeki, 1978a). Although large-scale cortical reconstructions were not attempted for the other

12 24 D. C. VAN ESSEN AND S. M. ZEKI hemispheres available from the present series of experiments, it was clear from a few partial reconstructions and from examination of large numbers of sequential sections that the basic pattern of interhemispheric connexions illustrated in the most extensive maps was present in all the other hemispheres and that the degree of variability in the details of the pattern was generally in the range of that seen in Text-figs Medial view Zone 3 ai 1 Lateral view Text-fig. 5. A schematic illustration of the principal zones of degeneration produced by callosal transaction. The cortical map is the same as in Text-fig. 3, except that contour lines of individual sections have been omitted and that all regions where degeneration was present have been shaded uniformly. Zones 2-4 of degeneration are continuous with one another. but for reasons discussed in the text have been considered as separate regions. Part II Physiological recordings from prestriate cortex Although the distribution of transcallosal projection fibres provides a valuable set of reference marks within the prestriate cortex, the information does not by itself permit a complete description of the borders between prestriate visual areas. Therefore, we combined the callosal transaction procedure with physiological mapping experiments which involved recording from up to a hundred sites over a limited extent of the cortex. The specific goals were (a) to determine the relationship between transcallosal inputs and representations of the vertical meridian, (b) to compare the types of topographical organization and the precision of topographic

13 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 25 representations among different prestriate visual areas, and (c) to see whether, even using coarse electrodes and recording from many cells simultaneously, functional differences between these different prestriate areas would still be evident. Comparison of striate and extrastriate receptive fteld8 Anatomical studies of striate cortical projections (Zeki, 1969; Cragg, 1969) have shown that a small lesion in the lateral part of the striate cortex projects directly across the thin underlying layer of white matter to area V2 in the posterior bank of the lunate sulcus. This suggests that the visual field maps of VI and V2 should be approximately in register over the region where the two areas are closely opposed. Furthermore, a small lesion in VI leads to degeneration in V2 occupying a larger 1 2 [ ~~~V2 V3 4 at ~ # \/ ]141.~~~Iv~1 4 LS - '~42mm Text-fig. 6. Comparison of receptive fields in areas VI, V2 and V3. Recordings from fifteen successive sites were made in a single penetration parallel to the horizontal plane and at the level indicated in the drawing of the hemisphere. The drawing of the section indicates the relationship of the electrode track to regions containing transcallosal inputs (black dots). Receptive fields in V3 were larger than those in V2, which in turn were larger than those in Vi, even though they were at similar distances from the fovea. area than the initial lesion site, suggesting that receptive fields in V2 might be larger than those in VI (Zeki, 1969; Cragg, 1969). We were able to confirm both of these possibilities physiologically, since many of our electrode penetrations were aimed directly through the striate cortex in order to reach areas buried within the lunate sulcus. Text-fig. 6 shows one such penetration in which recordings were made at fifteen successive sites as the electrode passed through VI, V2 and V3 approximately orthogonal to the cortical surface. The recordings in Vi were made at a position close to the representation of the vertical meridian, about 3 from

14 26 D. C. VAN ESSEN AND S. M. ZEKI the fovea. In agreement with the findings of Hubel & Wiesel (1974), the receptive fields in VI were small, and the scatter in receptive field positions was similar to the dimensions of individual fields. As the electrodes advanced into V2, receptive field positions shifted only slightly towards the horizontal meridian, with little change in eccentricity. There was, however, a substantial increase in receptive field size, with all of the fields in V2 being larger than those in VI. As in Vi, the scatter in receptive field position was comparable to the individual field sizes, although a more extensive analysis would be necessary to determine whether the relationship of scatter to field size was exactly the same in the two areas. Text-fig. 6 also shows 5. 9 _ 8 5. MCI. 5. Lateral ~~~~~~~~~2 2 mm 1 Text-fig. 7. Visual topography in area V2. The positions of nine separate penetrations into V2 in a single experiment are shown in the L-shaped enlargement of part of the cortical map (lower left). On the right are receptive fields obtained from each of the penetrations. The gradual shift from receptive fields at the vertical meridian (penetrations 1-3) to fields increasingly distant from the vertical meridian (penetrations 4-9) is evident in the Figure. Same hemisphere as in Text-fig. 2. that as the electrode advanced from the posterior to the anterior bank of the lunate sulcus, from V2 into V3, there was a further increase in receptive field size accompanied by an increase in field scatter and a shift in receptive field eccentricity (from 3 to 5 ).

15 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 27 Recordings in area V2. In penetrations throughout the lateral portion of the hemisphere we confirmed that there was relatively little difference in receptive field positions between points in VI and V2 directly opposite one another. One would therefore expect that the 'magnification factor' (cf. Daniel & Whitteridge, 1961) with which the visual field is represented in the cortex should be similar in corresponding parts of VI and V2. An indication of the magnification factors in the foveal and parafoveal portions of V2 is provided in Text-fig. 7, which shows receptive fields for fourteen recording sites in nine penetrations into V2 in a single experiment. The positions of the individual penetrations have been plotted on the enlarged L-shaped portion of the cortical map (lower left); on the right are the receptive fields encountered in each of the nine penetrations. In the three penetrations closest to the V1/V2 boundary (penetrations 1-3) receptive fields overlapped the vertical meridian and were centred about the fovea. In penetrations 4-7, several mm medial and dorsal to the first three, receptive fields were on the horizontal meridian, 2-3 from the fovea. Finally, in penetrations 8 and 9 receptive fields were near the horizontal meridian and 4-6 from the fovea. Thus, the magnification factor in V2 is approximately 3-5 mm/degree at eccentricities of 1-5. The relatively large scatter in receptive field position within individual penetrations made it difficult to determine more precisely the magnification factor as a function of eccentricity. Nevertheless the values we obtained for V2 are similar to those which have been reported for parafoveal regions of striate cortex (Daniel & Whitteridge, 1961; Hubel & Wiesel, 1974). One consequence of the increase in receptive field size and scatter in V2 without a change in magnification factor (relative to Vi) is that a greater distance within the cortex must be tranversed before a net shift in receptive field position can be detected. In Text-fig. 7, for example, there is no significant shift in average receptive field position over the 4 mm distances between penetrations 1 and 3 or between penetrations 6 and 8. In contrast, a measurable change in receptive field position is detectable within about 1 mm in VI (Hubel & Wiesel, 1974). In other words, the visual field map is represented more coarsely in V2 even though the map occupies comparable areas of cortex in Vi and V2. With regard to this point it is noteworthy that the sharpness of the band of transcallosal inputs along the V1/V2 boundary gives a misleading impression of the sharpness of visual topography in V2. For example, receptive fields for the first three penetrations in Text-fig. 7 all overlap the vertical meridian, even though the third penetration is 5 mm away from the nearest region of degeneration. Thus, while the presence of transcallosal inputs to V2 signifies a region of vertical meridian representation, the converse is not always true, in that the vertical meridian representation extends well beyond the region of transcallosal inputs. Other interesting aspects of visual topography are revealed in Text-fig. 7. One is the presence of receptive fields situated 1-2 in the ipsilateral hemifield (e.g. penetration 1). Ipsilateral receptive fields were seen occasionally, but only in penetrations close to the V1/V2 boundary. Their occurrence is somewhat surprising, since the interhemispheric projections to V2, a logical source for an ipsilateral representation, should have been eliminated completely by transaction of the splenial portion of the corpus callosum (Pandya, Karol & Heilbronn, 1971). Nevertheless,

16 28 D. C. VAN ESSEN AND S. M. ZEKI we are confident that the ipsilateral positions of the receptive fields were genuine, because the foveal projections were checked for each eye and because the receptive fields were in corresponding places for the two eyes. Presumably, the source of the ipsilateral input comes either from the strip along the nasal temporal division of the retina that projects to both hemispheres (Stone, Leicester & Sherman, 1973), or else from a non-geniculate thalamic input such as the pulvinar (Clark & Northfield, 1937; Cragg, 1969). Despite the occasional presence of ipsilateral fields in V2, Lateral 2 mm Text-fig. 8. Visual topography in area V3. The enlarged portion of the cortical map (lower right) shows the positions of six penetrations within or close to V3 in the anterior bank of the lunate sulcus. Receptive fields for penetrations 1-3 were near the horizontal meridian and therefore along the V2/V3 border, while penetrations 5 and 6 were within a region of degeneration and had receptive fields close to the vertical meridian. Same experiment as in Text-fig. 7. there does not appear to be a continuous representation of a strip of the ipsilateral hemifield along the V1/V2 border, since in some penetrations right at this boundary receptive fields abutted, but did not cross over the vertical meridian. Another surprising finding was that within the dorsal part of V2 (i.e. the portion within the lunate sulcus) some receptive fields were distinctly superior to the horizontal meridian (e.g. Text-fig. 7, penetrations 2, 1, 6). Only a small portion of the

17 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 29 superior contralateral quadrant was represented in dorsal V2, however, as we never encountered V2 receptive fields in the lunate sulcus more than 1-2 superior to the horizontal meridian. Since we did not examine the organization of the ventral portion of V2 (in the inferior occipital sulcus) (Zeki, 1969), we were unable to determine whether there is a dual representation of a strip along the horizontal meridian in both the dorsal and ventral divisions of V2 or, alternatively, whether the split between the two divisions-occurs superior to, rather than exactly at the horizontal meridian representation. Recordings from area V3. Although there are no terminations of transcallosal fibres to demarcate the border between areas V2 and V3, it was possible to determine the location of this border physiologically by noting the progression of receptive field positions in recordings at increasing distances from the striate cortex. The V2/V3 border was taken to be the place at which there was a reversal from a progression of receptive fields away from the inferior vertical meridian to an orderly progression back towards the midline. We found that over part of its extent this border was situated close to the fundus of the lunate sulcus (i.e. its medial side, as seen in the horizontal sections). This is illustrated in Text-fig. 8 which shows receptive fields encountered in six penetrations within a restricted portion of the same hemisphere shown in Text-fig. 2. In penetrations 1-3, situated 1-2 mm medial to zone 2 of degeneration (cf. Text-fig. 5), receptive fields were all at or near the horizontal meridian. Receptive fields in penetration 4, less than 1 mm from zone 2 degeneration, were slightly inferior to the horizontal meridian, but still about 2 from the vertical meridian. Penetrations 5 and 6 actually traversed the region of degeneration; as expected, receptive fields came up to or even crossed the vertical meridian. Thus, by the criterion of visual topography, penetrations 1-3 were at the V2/V3 boundary and penetrations 4-6 were within V3. An independent indication of this transition is provided by the sudden jump in receptive field size between fields 1-3 and 4-6, similar to that illustrated in Text-fig. 6. These results confirm that the horizontal meridian (or a line slightly superior to it, cf. above) is represented at the V2/V3 boundary (Zeki, 1969; Cragg, 1969), and that the other boundary of V3, lateral and anterior to the first, is at the representation of the vertical meridian in zone 2 of transcallosal projections (Zeki & Sandeman, 1976). They suggest in addition that area V3, at least over part of its extent is only a few mm wide, in contrast to V2, which occupies much of the posterior bank of the lunate sulcus and is 8-1 mm wide in some regions (see also Zeki & Sandemann, 1976). In accordance with this the cortical magnification factor was found to be much smaller in V3 than in V2 or Vt. Although the large sizes of receptive fields in V3 prevented a very accurate estimate, it was clear from numerous penetrations, including those shown in Text-figs. 8 and 11, that the magnification factor in V3 is only a fraction of a mm per degree even for regions representing the central 5. This is in sharp contrast to areas VI and V2 where, as discussed above, the magnification factors are at least several mm per degree at comparable eccentricities. The finding that one of the borders of V3 coincides with zone 2 of degeneration raises an important question concerning the organization of V3, given that zone 2 terminates abruptly within the ventral part of the lunate sulcus. This suggests that the dorsal division of V3 might also terminate within the lunate sulcus and

18 21 D. C. VAN ESSEN AND S. M. ZEKI not be continuous with the ventral division of V3 in the inferior occipital sulcus. Our physiological results are consistent with this interpretation, and it seems the simplest way of accounting for our results. On the other hand, the borders of V2 and V3 in the region of foveal representation were difficult to determine precisely, owing to uncertainties in the exact positions of the foveal projections for each eye, and we could not exclude the alternative that a thin wedge of V3 extends ventrally and laterally past the end of zone 2. Further analysis of striate cortical projections to V2 and V3 may provide a more suitable way of resolving this issue (cf. Zeki, 1978a). A separate question concerns the extent of V3 in the medial part of the hemisphere, in the representation of the visual periphery. Our recordings indicate that V3 extends from the lunate sulcus, over the annectant gyrus, and into the parietooccipital sulcus, a result which is consistent with anatomical evidence for a projection from V1 to V3 in the parieto-occipital sulcus (Zeki, 1977a). We did not explore this region extensively enough, however, to determine precisely where V3 ends or how the region medial and anterior to V3 is organized topographically. Recordings in V3A. Recordings from the region bounded by zones 2, 3 and 4 of degeneration (cf. Text-fig. 5) showed that this part of the cortex differs from areas V2 and V3, in containing a representation of both superior and inferior visual quadrants within the dorsal half of the hemisphere. A sequence of recordings illustrating this difference is shown in Text-fig. 9. The sequence, part of which was shown in the preceding figure, begins at the V2/V3 border (sites 1-3) with receptive fields at an eccentricity of about 3 along the horizontal meridian. Sites 4-6 ran along the vertical meridian representation of V3, starting within a few degrees of the fovea (site 4) and extending to an eccentricity of 15-2' at sites within the annectant gyrus (site 6). Recording sites 6-13 traversed the gap between zones 2 and 3 of degeneration. During this 5 mm progression receptive fields moved in a semicircular arch from the inferior vertical meridian, past the horizontal meridian, and up to the superior vertical meridian. Thus, instead of a double representation of the inferior quadrant, as was found between zones 1 and 2, separate representations of the inferior and superior quadrants were found between zones 2 and 3. Because this region is bounded by V3 on its medial side and V4 on its lateral side, because it has a topographic organization intermediate in complexity between V3 and V4 and because, just like V3, it has a heavy concentration of orientation selective cells (Zeki, 1978b) we refer to it as V3A. Examination of receptive fields at different positions along the V3/V3A border (zone 2 of degeneration) revealed an unexpected reversal in the progression of fields along the vertical meridian. From Text-fig. 9, sites 4-6, it was clear that during movement medially along the V3/V3A border, away from its termination in the lunate sulcus, receptive fields moved away from the fovea. Yet in Text-fig. 1 (from the same experiment) receptive fields 2-4, encountered close to the transition from zone 2 to zone 3, were shifted back towards the fovea. Similar reversals in the visual representation along the V3/V3A border were seen in other experiments. We did not encounter receptive fields farther than 2-25' from the fovea in this region, but the penetrations were never spaced closely enough in the vicinity of the reversal to be sure that the actual reversal did not take place at a greater eccentricity.

19 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 211 Receptive fields 5 and 6 in Text-fig. 1 were encountered in zone 3 of degeneration, which forms the boundary between area V3A and an as yet unidentified area in the parieto-occipital and intraparietal sulci. Note that receptive field 6 was well away from the vertical meridian even though the recording site was in sparse degeneration. A similar situation was seen in other experiments, but in all cases Lateral 2 mm Text-fig. 9. Visual topography in areas V3 and V3A. The positions of 13 penetrations into parts of the lunate and parieto-occipital sulci are shown in the enlarged portion of the cortical map (lower right). Receptive fields for each recording site are shown in the visual field map on the left. Receptive field positions shifted from the horizontal meridian at the V2/V3 border (sites 1-3) to the inferior vertical meridian at the V3/V3A border (sites 4-6) to the horizontal meridian representation within V3A (sites 8, 9) and to the superior vertical meridian at the antero-medial boundary of V3A. Same experiment as in Text-figs. 7 and 8. Recordings 1-4 correspond to individual receptive fields from penetration 1, 2, 3 and 5, respectively, in Text-fig. 8. a progression towards denser degeneration brought receptive fields closer to the vertical meridian. As recording sites progressed more antero-dorsally in zone 3, receptive fields moved from central to peripheral. We did not, however, make detailed enough recordings in this region to determine whether there was a reversal in the progression of receptive fields as was seen along the V3/V3A border.

20 212 D. C. VAN ESSEN AND S. M. ZEKI Further information on the organization of area V3A came from comparing receptive field positions along the V3A/V4 and V3/V3A borders (zones 4 and 3 respectively) in the lunate sulcus. In the experiment illustrated in Text-fig. 11 a fortuitously well aligned sequence of seven parallel penetrations was made into areas V2, V3, V3A and V4. Most of the recordings were made from the anterior bank of the lunate sulcus, where the thirty-two recordings sites were at close enough intervals to provide a nearly continuous monitoring of receptive field positions 2X 7~~~~~~~~~ 2 mm Text-fig. 1. Central receptive fields in the parieto-occipital sulcus. Receptive fields 2-4, encountered near the juncture of zones 2 and 3 of degeneration (cf. Text-fig. 5) in the parieto-occipital sulcus, were within 5-1 of the fovea, whereas those encountered more anteriorly (sites 5, 6), laterally (sites 7, 8) and postero-laterally (site 1 and Text-fig. 9, site 6) were all 1-25from the fovea. It is not clear whether this region of 'central' representation should be considered as part of area V3A or as part of a separate, unidentified visual area. Same experiment asq in Text-figs Sites 5 and 8 correspond to sites 12 and 9, respectively, in Text-fig. 9. Over an 8 mm length of cortex. In order to make the numerical sequence correspond to the medio-lateral order of recording sites, the lowest number in each penetration was assigned to the most medial recording site. Penetration A passed directly through the V3/V3A border. The precision of visual topography in this region is indicated by the fact that receptive fields 36, encountered within the zone of degeneration, were all within 1 of the vertical meridian, whereas receptive fields 1, 2, 7 and 8 were distinctly further from the mid line even though they were recorded less than 1 mm from the zone of degeneration. Recording sites 7-2 were all from V3A, starting at the V3 border and extending to the V4 border; the associated

21 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 213 receptive fields followed a semicircular trajectory from the inferior to the superior vertical meridian, a progression similar to that illustrated in Text-fig. 9 except at a lesser eccentricity. The remaining penetrations in this sequence were all from area V4 and will be discussed below. Anterior A Va Lateral 15 1A ~ Area 17 -t 17 2' I-aJ 1 19 F 2 26 F 5 ID1 A 25 G B 28-6 ~~~~~~2 27 C 21 D 7 (C) C ~ F d ~~G 8 6 ~~~~~~~E 23 G from B9 t29 1 mm P cm (C Text-fig. 11. Comparison of receptive fields in areas V2, V3, V3A and V4. Seven parallel electrode penetrations, all in approximately the same horizontal plane indicated in the drawing of the hemisphere, were made into the lunate sulcus. For clarity electrode tracks found in a number of closely spaced sections were all superimposed on to the same drawing. Receptive fields a-f (lower left), all from V2, were small and close to the vertical meridian. Recording sites 1-32 formed a long, orderly sequence extending from the medial part of the lunate sulcus to the prelunate gyrus and traversing areas V3, V3A, and V4. Continuous lines connect receptive fields to the positions of the fovea, which have been shifted varying amounts along the vertical axis. Preferred orientations of orientation-selective cells are indicated by thin lines through the receptive fields. Receptive fields at site 3 in V3 showed a clear binocular disparity (continuous outline, contralateral eye: dashed line, ipsilateral eye). No other clear disparity units were seen in this sequence, and all receptive fields are drawn for the contralateral eye. Receptive fields 21 and 31 were colour-coded. Same hemisphere as in Text-fig. 4. Although our understanding of visual topography in V3A is clearly incomplete, at this stage we can point to several basic features of V3A organization, each of which has been established in two or more separate experiments. (1) Receptive fields in V3A are large relative to areas V1-V2, but visual topography is clearly preserved in the sense that distinct shifts in receptive field positions are evident during movements of 1 mm or less through the cortex. (2) The inferior vertical meridian is represented along the V3/V3A border, the superior vertical meridian 1 3 C

22 214 D. C. VAN ESSEN AND S. M. ZEKI along the V3A/V4 border and along zone 3 of degeneration in the parieto-occipital sulcus, and the horizontal meridian within the degeneration-free portion of V3A. (3) Central visual fields may be represented at least twice in V3A: once in lunate sulcus (in the gap between degeneration zones 2 and 4), where receptive fields are reasonably small, and again in the fundus of the parieto-occipital sulcus (at the confluence of zones 2 and 3), where receptive fields are large even when they overlap the fovea. The major uncertainties remaining concern the exact nature of the visual field map within V3A and the over-all extent of V3A, in particular whether it occupies all of the region bounded by degeneration zones 2-4 and whether it abuts on V2 along the fundus of the lunate sulcus laterally. Recordings in area V4. The most complex region of transcallosal projections was that labelled zone 4 (cf. Text-fig. 5) which falls within V4. This area occupies much of the lateral portion of the anterior bank of the lunate sulcus (Zeki, 1971). Textfig. 11 shows a series of recordings from this region, a continuation of the sequence already used to illustrate the organization of V3A in the medial part of the sulcus. At this particular horizontal level the zone 4 degeneration consisted of three dense patches separated by regions with moderate or sparse degeneration, a situation which allowed for a test of whether changes in the density of degeneration in V4 were associated with substantial shifts in receptive field positions. Between sites 21 (penetration D) and 23 (penetration E) receptive fields remained close to the vertical meridian but shifted from the superior to the inferior visual quadrant, suggesting that a transition had occurred from V3A to V4. Without a better understanding of visual topography in both areas, especially V4, it is difficult to state precisely where the V3A/V4 boundary should be placed. For the present we provisionally assume that the border runs along the medial edge of zone 4 of degeneration, so that most of zone 4 lies within area V4. For the recording sites of Text-fig. 11 that were clearly within V4, as defined above, it was difficult to discern any clear relationship between receptive field position and density of degeneration. Receptive fields and 3-32, associated with sparse or moderate degeneration, were marginally, if at all, farther from the vertical meridian than receptive fields 26-29, associated with patches of dense degeneration. Some form of topographic mapping in V4 clearly must be present, since receptive fields were still far from quadrantic in their dimensions, and since significant shifts in receptive field position occurred during electrode movement parallel to the cortical surface, but not in radially oriented penetrations. In order to pursue this problem we prepared detailed maps of receptive fields and recording sites from V4 for the two experiments in which the largest number of V4 recordings had been made (Text-figs. 12 and 13). The centre of each figure contains an enlargement of the regions in and around V4 from the respective cortical maps and shows all of the recording sites in relation to the pattern of degeneration. Unfortunately, owing to the fact that electrode penetrations were aimed 'blindly' into a buried, irregularly shaped sulcus, the recording sites do not form an orderly grid covering V4 at a uniform density. Nevertheless enough points in V4 were explored to permit a determination of how visual topography was correlated with the pattern of degeneration. Both experiments covered a region of V4 representing mostly the inferior contralateral quadrant from to 5-1O in eccentricity. In some parts of the map

23 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 215 there appeared to be a definite correlation between receptive field position and density of degeneration. For example, in Text-fig. 12, receptive fields 9 and 1 (group B) were encountered in a band of dense degeneration, and their borders extended up to the vertical meridian, whereas receptive fields and 23-28, encountered in regions of moderate degeneration, were situated 2-4 from the mid line. Similarly, in Text-fig. 13, the sequence of recording sites proceeded from moderate degeneration to a degeneration-free zone, and the associated receptive fields (group H) started at the vertical meridian and progressed away from it. On the other hand, it was possible to find sequences of recordings in which no such correlation was evident, such as for sites (groups E-G) in Text-fig. 13. In order to decide whether a genuine correlation existed between visual topography and transcallosal inputs, statistical analyses were carried out for the two experiments in which twenty or more recordings clearly within V4 had been made. For the V4 recordings in Text-fig. 12 (all except 1-6, 29 and 3) sites within regions of dense degeneration had receptive fields whose closest approach to the vertical meridian was (standard error of the mean, n = 6). Those encountered in moderate degeneration were significantly farther from the mid line ( , n = 11), although no further shift occurred for fields associated with sparse or no degeneration ( , n = 5). A similar correlation was found for the experiment illustrated in Text-fig. 13. The significance of this correlation did not depend upon whether cells along the V4 border were included, or whether receptive field centres rather than closest approaches to the mid line were used as the basis of the calculations. Thus, these results verify our impression that the pattern of transcallosal inputs to V4 provides a meaningful index of local visual topography. Nevertheless, it is equally noteworthy that a statistical analysis of a large number of recordings was needed to establish the point firmly, whereas in all other areas we examined the relationship was evident at a glance. Moreover, the correspondence between degeneration and vertical meridian representation in V4 was sufficiently weak that some cells well away from degeneration had receptive fields crossing the mid line, while other cells within sparse or even moderate degeneration had fields many degrees from the mid line, thus confirming earlier anatomical conclusions (Zeki, 1971) that, within V4, interhemispheric connexions may not necessarily be restricted to areas of vertical meridian representation. Another striking feature of visual topography in V4 was the multiple representation of the same part of the visual field at distant points within the cortex. For example, in Text-fig. 12, sites 9 and 1 and 26-28, which were separated by about 6 mm, had receptive fields 1-2 from the fovea and close to the horizontal meridian; sites in between these two clusters also had receptive fields in a similar position. Another example can be seen in Text-fig. 13 where receptive fields cover the same region as receptive fields 33-36, recorded approximately 1-cm away. On the other hand, large shifts in receptive field position sometimes occurred over distances of 1-2 mm, as with sites 9-15 and in Text-fig. 12, and sites in Text-fig. 13. The presence of multiple representation of at least some parts of the visual field rules out the simplest type of visual organization of V4, namely a one-to-one mapping of the visual hemifield onto the cortical surface. There are three principal models to be considered as ways of accounting for our results. One possibility is

24 216 D. C. VAN ESSEN AND S. M. ZEKI Ftt **~Ee S CS4 * % I E S*Oi

25 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 217 I- El% lo ID cu E. 4kp. (1) ts 4 I C4 V)

26 218 D. C. VAN ESSEN AND S. M. ZEKI that area V4 actually consists of a number of distinct sub-areas, with each sub-area containing a map of only a portion of the visual hemifield. Our results are consistent with such a scheme, although they do not provide a very clear idea of how many sub-areas there might be in V4 or of what fraction of the visual map each might represent. The principal attractiveness of the hypothesis is that different sub-areas might have different functions and/or different inputs and outputs (cf. Zeki, 1977b). Our results are also consistent with the alternative idea that V4 should be considered as a single visual area in which the normal mode of representing visual space has been replaced by a more complex type of mapping although, on the face of it, such an idea would not be entirely consistent with the pattern of anatomical inputs to V4 (Zeki, 1971, 1978a). In a sense, the difference between the two schemes is a semantic one that depends upon exactly what should be meant by a visual area in regions of unorthodox visual topography. Nevertheless, the distinction could be a meaningful one, depending upon the degree of homogeneity of anatomical connexions and visual functions of V4. A third possible explanation of our results is that several 'conventional' areas, each containing a complete map of the visual hemifield (or of a visual quadrant at least), might exist within what we have called V4. Such an arrangement would imply that between each representation, say, of a given place on the vertical meridian, there should be an orderly (but not necessarily linear) progression of receptive fields out to the horizontal meridian and back. This scheme seems unlikely, as our recordings suggested that less than a complete quadrant was represented between each repetition of the same point in visual space. Conceivably, though, we might have missed such an arrangement, owing to an insufficient density of recording sites within V4. Before leaving V4, we emphasize the presence of relatively large receptive fields in V4 (see Text-figs. 12 and 13) and the invasion of the superior visual quadrants, in harmony with the predictions based on a study of the anatomical inputs to V4 (Zeki, 1971). A composite map of visual areas. Based on the combined results of all our anatomical and physiological observations, as well as on antecedent results (Zeki, 1969, 1971; Cragg, 1969), it is possible to construct a schematic map which contains the known borders between visual areas and indicates where the positions of borders remain uncertain (Text-fig. 14). Continuous and dashed lines mark the borders along which the vertical and horizontal meridians, respectively, are represented. Although the borders are not complete enough to permit an accurate estimate of the size of individual areas, it appears that V2 is the largest of the prestriate areas, V3A is intermediate in size, and V3 is the smallest. We do not know where the lateral border of V4 lies, and anteriorly it may extend on to the superior temporal sulcus (see Zeki, 1978b). Neither are we certain whether V4 extends dorsally beyond the level at which the annectant gyrus disappears (see Zeki, 1978a). Consequently we are not able at present to make even a rough guess of its size. The insets of Text-fig. 14 show the arrangement of areas as they appear on the medial and lateral surfaces of the hemisphere. Note that areas V3 and V3A are completely submerged within buried sulci and are not, as far as we are aware, visible anywhere on the surface in the dorsal half of the hemisphere. Neuronal function in different areas. Our principal objective in analysing cell

27 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 219 function in these experiments was to look for obvious, qualitative differences in physiological properties among cell populations in different areas. The most consistent difference we found was in receptive field size, as mentioned earlier and illustrated in Text-fig. 5. For any given retinal eccentricity examined, receptive fields in V2 were on average considerably larger than in VI and those in V3 were larger than in V2. Receptive fields were comparable in size to each other in V3A and V4, and they tended to be larger still than those in V3, especially at eccentricities in excess of 5. V2 vi Medial view vi S9. vi Lateral view Text-fig. 14. A schematic map of the arrangement of visual areas in dorsal prestriate cortex. In the central reconstruction, vertical and horizontal meridians are represented by continuous and dashed lines respectively. Regions where major uncertainties remain concerning the borders between areas are indicated by question marks. The cortical map is the same as that in Text-figs. 3 and 5, but results from all experiments have been considered in formulating the arrangement of areas. Of greater significance were the regional differences in more specialized aspects of visual analysis. Table 1 shows the incidence of selectivity with respect to stimulus orientation, direction, and colour for our recordings in each of the four areas where our investigations were concentrated. Included in the sample are all recordings which could be confidently assigned to a particular area and for which at least one aspect of functional organization (e.g. orientation selectivity) was adequately characterized. For a number of recordings a successful characterization was made

28 22 D. C. VAN ESSEN AND S. M. ZEKI for only one or two of the three aspects listed in the table; consequently, the number of recordings listed for any one area is not the same for each category. The most striking regional variations in cell function are (1) the high incidence of colour sensitivity in area V4 (32 %) vs. other areas (4, and 3 % in areas V2, V3 and V3A, respectively), and (2) the lower incidence of orientation selectivity in V4 (42 %) vs. other areas (72, 86 and 71 % in V2, V3, and V3A respectively). Other noteworthy features revealed in the Table are (1) the low percentage of directionally selective cells in all of the areas examined (in contrast to their high incidence in the striatereceptive area of the superior temporal sulcus; Zeki, 1974), and (2) the lack of significant differences among areas V2, V3 and V3A with regard to the incidence of selectivity for orientation, direction, or colour (see Zeki, 1978 c). TABLE 1. Functional properties in different areas No. of Orientation Directionally Colour-coded Binocularly Area recordings selective selective or biased driven V /64 (72%) 5/62 (8%) 3/69 (4%) 67/71 (94%) V /43 (86%) 5/4 (12%) /35 (%) 47/47 (1%) V3A 13 76/17 (71%) 8/1 (8%) 2/77 (3%) 129/13 (99%) V4 1 31/74 (42%) 2/74 (3%) 21/65 (32%) 1/1 (1%) Each entry shows the incidence in one of the four areas studied of recordings selective for a specific functional parameter (stimulus orientation, direction of movement, colour, or binocularity). Recordings for which an incomplete or inadequate characterization of a particular property were made have been excluded from the statistics relating to that property. The number of such cases ranged up to 9 in V2, 12 in V3, 53 in V3A and 34 in V4. The higher incidence of inadequately characterized receptive fields in V3A and V4 reflects in part the greater difficulty in driving cells vigorously in these areas. Also excluded from the sample are fifty-two recordings which could not reliably be assigned to a particular visual area. Of the twenty-six recordings in which colour selectivity was seen, fifteen were classified as colourcoded (see Methods) and eleven as colour-biased. Fourteen of the colour-coded recordings were from V4, and one was from V2. An important aspect of visual function not contained in Table 1 is the binocular interactions that could play a role in stereopsis. Almost all of the recordings we made were from cells or cell clusters that could be driven independently through each eye. For most recordings we detected no significant disparity in receptive field positions for the two eyes, although occasionally clear disparities of up to 1 were seen. The incidence of disparity cells would presumably have been higher had we been able to detect genuine differences of.5 or less in receptive field position and had we tested more carefully for cells requiring simultaneous binocular stimulation at a specific disparity (e.g. Hubel & Wiesel, 197; Pettigrew, Nikara & Bishop, 1968). A cautionary note concerning the specific percentages listed in Table 1 is advisable, given that most of our recordings were made from cell clusters rather than individual cells, and that our analyses of functional properties were, for reasons cited above, necessarily brief and qualitative. Such factors presumably could account for the low estimate of colour-coding in V4 relative to values reported in other studies (Zeki, 1975, 1978c). Despite these uncertainties concerning absolute percentages, there seems little doubt that these data reflect genuine regional variations in cell function in the prestriate cortex.

29 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 221 DISCUSSION In the present study we have attempted to decipher the overall plan of organization of part of the prestriate cortex of the rhesus monkey by employing a combination of anatomical and physiological techniques. It was clear from the start that one of the major obstacles to the elucidation of this problem was the extensive degree of fissuration of the occipital prestriate cortex. This difficulty was largely circumvented by devising a method for constructing a two-dimensional, 'unfolded' map of the cortex. A second problem concerned the absence of clear architectural borders between topographically defined areas. This was partially overcome by using the connexions carried through the corpus callosum as an anatomical reference system. From these experiments it was possible to confirm and extend earlier observations on the visual topography of areas V2 and V3 (Zeki, 1969; Zeki & Sandeman, 1976), to establish the existence of a previously undescribed area, V3A, and to work out some aspects of the complex visual topography in V4. One interesting outcome of these explorations was the finding of an increasingly more complicated form of visual topography in the progression of areas from VI to V4. The simplest form of visual mapping is in Vi, where adjacent points in the contralateral visual hemifield are always represented in adjacent small regions of the cortex. This corresponds to the conventional notion of an orderly topographic projection from one two-dimensional structure onto another. The next area, V2, differs in visual topography from VI in having a split representation of upper and lower visual quadrants (Cragg, 1969; Zeki, 1969) with a dual representation of part of the horizontal meridian along the border farthest away from VI. This type of arrangement is also found for the second visual areas of other mammals, including the owl monkey (Allman & Kaas, 1971) and the cat. The principal difference between the old world rhesus and new world owl monkeys in this regard is in the place at which the representation of the horizontal meridian bifurcates. In the owl monkey the transition occurs at a site within the cortex several mm from the edge of striate cortex, corresponding to a visual representation approximately 6 from the fovea (Allman & Kaas, 1974a). In the rhesus monkey, the site of the horizontal meridian split appears to be within 1-2 mm of the V1/V2 border which, given the high magnification factor in V2, would be well within the region of foveal representation. Although it seems very likely, we have no direct evidence to suggest that the two subidivisions of V2, representing the superior and inferior hemiquadrants, are continuous with one another. The same is true of V3, an area which, topographically, has an organization that is similar to V2 in the sense of being divided into distinct superior and inferior quadrantic representation (Cragg, 1969; Zeki, 1969). However, our results suggest that there is a greater likelihood of the two quadrantic representations of V3 - one in the lunate, the other in the inferior occipital sulcus - being completely separate, since the band of callosal degeneration defining the anterior border of V3 terminates abruptly in the lunate sulcus. We are not certain whether the upper part of V3 (representing lower visual fields) continues ventral to this level, as mentioned earlier. Area V3A differs from both V2 and V3 in two major respects. First, both inferior and superior visual quadrants are represented within a single region in the dorsal

30 222 D. C. VAN ESSEN AND S. M. ZEKI occipital lobe, rather than in separate dorsal and ventral subdivisions. Thus, in topological terms, V3A resembles VI more than V2 or V3. A second difference is related to the finding of a secondary representation of central visual fields in a region which we had expected would represent the far periphery for V3A. Our recordings were not complete enough to distinguish between two alternative explanations for this result: either the secondary central representation forms a part of area V3A, in which case visual topography in V3A is complex to a degree not found in VI, V2 or V3; or else the secondary central representation forms part of an unidentified additional visual area, in which case our recordings would suggest that in V3A there is little or no representation of the visual periphery further than 2-25o from the fovea. A further departure from conventional visual topography appears to take place in V4, in the sense that multiple representations of the same place in the visual field occur at widely spaced points within V4. Once again, however, we were unable to determine whether the V4 regions should be considered as a set of separate areas each representing a small portion of the visual field or as a single entity in which there is no longer a point-to-point relationship between the visual field and the cortical surface. The distinction between these two possibilities is in part a semantic issue that depends upon exactly what is meant by a visual area. In principle there are a number of different ways of subdividing the visual cortex into separate areas. Historically, the earliest schemes were based on regional differences in cyto- and myelo-architecture (e.g. Brodmann, 199; Vogt & Vogt, 1919; von Economo & Koskinas, 1925). The idea that each visual area might contain a single retinal map awaited anatomical and physiological methods for tracing visual projections (e.g. Polyak, 1933; Talbot & Marshall, 1941). In many situations, notably area 17 in a variety of mammals (e.g. Daniel & Whitteridge, 1961; Cowey, 1964; Hall, Kaas, Killackey & Diamond, 1971; and Allman & Kaas, 1971 b), and areas 18 and 19 in the cat (Hubel & Wiese], 1965), topographically defined areas have been shown to be coextensive with architectonically defined ones, a result which strengthens confidence in both approaches. On the other hand, other cases are known for which a discrepancy exists between topographically and architectonically defined borders. A striking example which we have already emphasized is the disparity in the rhesus monkey between areas V2 and V3, defined topographically, and Brodmann's areas 18 and 19, defined cytoarchitectonically. There is no a priori reason, though, to expect that every visual area must have a uniquely distinctive achitecture. In addition, there is little doubt that separate topographic representations of the visual fields are more significant features than subtle variations in cortical architecture as a basis for subdividing the cortex. The absence of clearly defined architectural differences between areas presents an especially serious difficulty for the analysis of topographically complex areas such as V3A and V4, where it is difficult to define areas completely on the basis of visual topography alone. It then becomes particularly important to look for other types of information that might help in distinguishing among areas. Obviously, it would be helpful to find regional differences in cellular receptive field properties or in patterns of afferent and efferent connexions. We found, for example, substantial

31 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 223 regional differences in average receptive field size that made it possible to distinguish among areas V2, V3 and V3A (except for uncertainties in the region of foveal representation), and regional variations in the distributions of colour coded and orientation selective cells that permitted a functional distinction between area V4 on the one hand and areas V2, V3 and V3A on the other. However, in the present study we were not able to carry out these approaches in sufficient detail to resolve the more difficult questions concerning the internal organization of prestriate areas, in particular whether V3A and V4 should each be considered as several small, functionally distinct sub-areas or as a pair of larger, more complex areas. Comparisons with other species. In recent years numerous physiological and anatomical studies on visual areas outside the striate cortex have been undertaken in a variety of primates and other mammals. The most complete characterizations have been for the rhesus monkey (an Old World primate), the owl monkey (a New World primate), and the cat. The basic features common to all three species are (1) a striate cortex that contains a complete representation of the contralateral visual hemifield (Talbot & Marshall, 1941; Daniel & Whitteridge, 1961; Allman & Kaas, 1971 b), (2) a second visual representation, variously called VII or V2, which partially surrounds the striate area and which has a split representation of the horizontal meridian (Hubel & Wiesel, 1965; Zeki, 1969), and (3) a number of additional areas in the cortex surrounding visual II. In the rhesus monkey there are at least four such areas, V3, V3A and V4 and the striate receptive area of the superior temporal sulcus (Zeki, 1969, 1971; see above). In the owl monkey there are at least six additional areas, the middle temporal (MT), dorsolateral (DL), dorsointermediate (DI), dorsomedial (DM), medial (M) and tentorial (T) areas (Allman & Kaas, 1971a, 1974b, 1975, 1976). All but the middle temporal area adjoin V2 along part of their length and form what Allman & Kaas (1975) call a 'third tier' of visual areas. In the cat, Hubel & Wiesel (1965) have described areas VI-VIII. There also appear to be ten additional representations in surrounding cortical regions (Tusa, Palmer & Rosenquist, 1975). Not all of these visual maps in the cat represent the entire contralateral hemifield, however, a result which differs from the published findings for areas in the owl monkey but which appears similar to what we have found for the rhesus monkey. Given the existence of a multiplicity of visual areas in the occipital cortex of several mammalian species, it is natural to wonder which areas, if any, can be considered as homologous, in the strict sense of having a common embryological origin, or which can be thought of as closely related, in the sense of having a similar internal organization and carrying out similar visual functions. Too little is known about the embryonic development of the cerebral cortex to be sure of any strict homologies except perhaps that of the striate cortex, and even to establish any genuinely close relationships between areas is difficult with the limited knowledge presently available. For example, it is entirely plausible to suggest that the second visual areas found in different mammals are directly related to one another, on the basis of similarities in their internal visual topography and in the arrangement with respect to the striate cortex. On the other hand, it is known that there are significant differences in the pattern of major connexions of visual II in different species (cf. Wilson & Cragg, 1967; Glickstein, King, Miller & Berkley, 1967; Garey

32 224 D. C. VAN ESSEN AND S. M. ZEKI & Powell, 1971; Wong-Riley, 1974, 1976; Hubel, Wiesel & Le Vay, 1975; Gilbert & Kelly, 1975; Zeki, 1971; Tigges, Spatz & Tigges, 1973, 1974). Furthermore, it is not yet clear whether similar types of visual analysis take place in visual II of different species (Hubel & Wiesel, 1965, 197; Pettigrew, 1973; Tretter, Cynader & Singer, 1975). Another potentially significant similarity is between the striate-receptive portion of the superior temporal sulcus in the rhesus monkey and the middle temporal areas of the owl monkey, an area which also receives a direct input from the striate cortex and which lies in roughly the same region of the hemisphere. Once again, however, it seems inadvisable to come to any firm conclusions concerning these areas until more is known about neural function and anatomical connexions in the owl monkey and about visual topography in the rhesus monkey. Areas V3, V3A and V4 in the rhesus monkey occupy a broad region of cortex in the same general part of the hemisphere as areas M, DM, DI and DL of the owl monkey. Yet what is known about the visual topography, functions and connexions of these two sets of areas is not sufficient to pinpoint any strong similarities between particular areas in the two species. Although further investigations may well lead to the demonstration of many such similarities, the prospect nevertheless remains that major differences exist in the broad aspects of visual cortical organization of species even within the same order. Such a principle would emphasize the importance of studying the visual cortex of a variety of species and of relating the pattern of functional organization to the visual capacities of each species. This work was supported by the Science Research Council. D. C. Van Essen was a Helen Hay Whitney Postdoctoral Research Fellow. We are greatly indebted to Ms Pamela Jacobs and Ms Brenda Crane for their excellent histological assistance. We would like to record our thanks to Professor J. Z. Young and Professor D. Whitteridge for their critical reading of this manuscript. REFERENCES ALLMAN, J. M. & KAAs, J. H. (1971 a). A representation of the visual field in the caudal third of the middle temporal gyrus of the owl monkey (Aotus trivirgatue). Brain Res. 31, ALLMAN, J. M. & KAAS, J. H. (1971 b). Representation of the visual field in striate and adjoining cortex of the owl monkey (Aotu8 trivirgatus). Brain Re8. 35, ALLTMAN, J. M. & KAs, J. H. (1974a). The organization of the second visual area (VII) in the owl monkey: a second order transformation of the visual hemifield. Brain Res. 76, ALLMAN, J. M. &IKAAS, J. H. (1974b). A crescent-shaped cortical visual area surrounding the middle temporal area (MT) in the owl monkey (Aotuw trivirgatus). Brain Res. 81, ALLMAN, J. M. & KAAs, J. H. (1975). The dorsomedial cortical visual area: a third tier area in the occipital lobe of the owl monkey (Aotus trivirgatus). Brain Res. 1, ALLMAN, J. M. & KAAs, J. H. (1976). Representation of the visual field on the medial wall of occipital-parietal cortex in the owl monkey. Science, N.Y. 191, BRODMANN, K. (195). BeitrAge zur histologischen Localisation der Grosshirnrinde. 3. Mitteilung Die Rindenfelder der niederen Affen. J. Psychol. Neurol (Lpz). 4, BRODMANN, K. (199). Vergleichende Lokalisationslehre der Grosshirnrinde. Leipzig: Barth. CLARK, W. E. Le GRos & NORTHFIELD, D. W. C. (1937). The cortical projection of the pulvinar in the macaque monkey. Brain 6, COwEY, A. (1964). Projection of the retina on to striate and prestriate cortex in the squirrel monkey (Saimiri sciureus). J. Neurophysiol. 27, CRAaG, B. G. (1969). The topography of the afferent projections in circumstriate visual cortex of the monkey studied by the Nauta method. Vision Res. 9,

33 TOPOGRAPHY OF MONKEY PRESTRIATE CORTEX 225 DANIEL, P. M. & WHITTERIDGE, D. (1961). The representation of the visual field on the cerebral cortex in monkeys. J. Physiol. 159, GAREY, L. J. & POWELL, T. P. S. (1971). An experimental study of the termination of the lateral geniculo-cortical pathway in the cat and monkey. Proc. R. Soc. B 179, GILBERT, C. D. & KELLY, J. P. (1975). The projections of cells in different layers of the cat's visual cortex. J. comp. Neurol. 163, GLICKSTEIN, M., KING, R. A., MILLER, J. & BERKLEY, M. (1967). Cortical projections from the dorsal lateral geniculate nucleus of cats. J. comp. Neurol. 13, HALL, W. C., KAAs, J. H., KILLACKEY, H. & DIAMOND, I. T. (1971). Cortical visual areas in the grey squirrel (Sciurus carolinensis): a correlation between cortical evoked potential maps and architectonic subdivisions. J. Neurophysiol. 34, HUBEL, D. H. & WIESEL, T. N. (1965). Receptive fields and functional architecture in two non-striate visual areas (18 and 19) of the cat. J. Neurophysiol. 28, HUBEL, D. H.& WIESEL, T. N. (197). Cells sensitive to binocular depth in area 18 of the macaque monkey cortex. Nature, Lond. 225, HUBEL, D. H. & WIESEL, T. N. (1974). Uniformity of monkey striate cortex: a parallel relationship between field size, scatter, and magnification factor. J. comp. Neurol. 158, HUBEL, D. H., WIESEL, T. N. & LE VAY, S. (1975). Functional architecture of area 17 in normal and monocularly deprived macaque monkeys. Cold Spring Harb. Symp. quaint. Biol. 4, MERRILL, E. G. & AINSWORTH, A. (1972). Glass-coated platinum-plated tungsten microelectrodes. Mled. biol. Engng 1, PANDYA, D. N., KAROL, E. A. & HEILBRONN,. (1971). The topographical distribution of interhemispheric projections in the corpus callosum of the rhesus monkey. Brain Res. 32, PETTIGREW, J. D. (1973). Binocular neurones which signal change of disparity in area 18 of cat visual cortex. Nature, Lond. 241, PETTIGREW, J. D., NIKARA, T. & BISHOP, P.. (1968). Binocular interaction on single units in cat striate cortex: simultaneous stimulation by single moving slit with receptive fields in correspondence. Expl Brain Res. 6, POLYAK, S. D. (1933). A contribution to the cerebral representation of the retina. J. comp. Neurol. 57, STONE, J., LEICESTER, J. & SHERMAN, S. M. (1973). The naso-temporal division of the monkey's retina. J. comp. Neurol. 15, TALBOT, S. A. & MARSHALL, WV. H. (1941). Physiological studies of neural mechanisms of visual localization and discrimination. Am. J. Ophthal. 24, TIGGES, J., SPATZ, W. B. & TIGGES, M. (1973). Recripocal point-to-point connections between parastriate and striate cortex in the squirrel monkey (Saimiri). J. comp. Neurol. 148, TIGGES, J., SPATz, WV. B. & TIGGES, M. (1974). Efferent cortico-cortical fiber connections of area 18 in the squirrel monkey (Saimiri). J. comp. Neurol. 158, TRETTER, F., CYNADER, M. & SINGER, W. (1975). Cat parastriate cortex: a primary or secondary visual area? J. Neurophysiol. 38, TUSA, R. J., PALMER, L. A. & ROSENQUIST, A. C. (1975). The retinotopic organization of the visual cortex in the cat. Soc. Neurosci. Abstr. 5, 52. VOGT, C. & VOGT,. (1919). Allgemeinere Ergebnisse unserer Hirnforschung. J. Psychol. 25, VON EcoNoMo, C. & KOSEiNAS, G. N. (1925). Cytoarchitecktonit der Grosshirnrinde des erwachsen.mfenchen. Berlin u. WVien: Springer. WHITTERIDGE, D. (1965). Area 18 and the vertical meridian of vision. In Functions of the Corpus Callosum, ed. ETTLINGER, E. G. Churchill: London. WITTANEN, J. T. (1969). Selective silver impregnation of degenerating axons and axon terminals in the central nervous system of the monkey (Macaca mulatta). Brain Res. 14, XVILSON, M. E. & CRAGG, B. G. (1967). Projections from the lateral geniculate nucleus in the cat and monkey. J. Anat. 11, XWONG-RILEY, M. T. T. (1974). Demonstration of geniculocortical and callosal projection neurons in the squirrel monkey by means of retrograde axonal transport of horseradish peroxidase. Brain Res. 79, WONG-RILEY, M. T. T. (1976). Projections from the dorsal lateral geniculate nucleus to prestriate cortex in the squirrel monkey as demonstrated by retrograde transport of horseradish peroxidase. Brain Res. 19, P HY 277

34 226 D. C. VAN ESSEN AND S. M. ZEKI ZEKI, S. M. (1969). Representation of central visual fields in prestriate cortex of monkey. Brain Re8. 14, ZEKI, S. M. (197). Interhemispheric connections of prestriate cortex in monkey. Brain Re8. 19, ZEKI, S. M. (1971). Cortical projections from two prestriate areas in the monkey. Brain Res. 34, ZEKi, S. M. (1973). Colour coding in rhesus monkey prestriate cortex. Brain Re8. 53, ZEKI, S. M. (1974a). Functional organization of a visual area in the posterior bank of the superior temporal sulcus of the rhesus monkey. J. Phy~siol. 236, ZEKI, S. M. (1974 b). Cells responding to changing image size and disparity in the cortex of the rhesus monkey. J. Physiol. 242, ZEKI, S. M. (1975). The functional organization of projections from striate to prestriate visual cortex in the rhesus monkey. Cold Spring Harb. Symp. quant. Biol. 4, ZEKI, S. M. (1977a). Simultaneous anatomical demonstration of the representation of the vertical and horizontal meridians in areas V2 and V3 of rhesus monkey visual cortex. Proc. R. Soc. B 195, ZEKI, S. M. (1977 b). Colour coding in the superior temporal sulcus of rhesus monkey visual cortex. Proc. R. Soc. B 197, ZEKI, S. M. (1978a). The cortical projections of foveal striate cortex in the rhesus monkey. J. Physiol. 277, ZEKI, S. M. (1978b). The third visual complex of rhesus monkey prestriate cortex. J. Physiol. 277, ZEKI, S. M. (1978c). Uniformity and diversity of structure and function in rhesus monkey prestriate visual cortex. J. Physiol. 277, ZEKI, S. M. & SANDEMAN, D. R. (1976). Combined anatomical and electrophysiological studies on the boundary between the second and third visual areas of rhesus monkey cortex. Proc. R. Soc. B 194, EXPLANATION OF PLATES PLATE 1 A, lateral view of the brain of the rhesus monkey. B, medial view. C, dorsal view. The lunate sulcus extends from the lateral side of the hemisphere to its medial side, where it is continuous with the parieto-occipital sulcus. It also opens up, dorsally, into the intraparietal sulcus. In D, the lunate sulcus has been pried open to expose the annectant gyrus, buried within the lunate sulcus. l8, lunate sulcus; pos, parieto-occipital sulcus; c8, calcarine sulcus; ips, intraparietal sulcus; ag, annectant gyrus. PLATE 2 Examples of different densities of degenerating terminals within the prestriate cortex following callosal transaction. A, sparse degeneration; B, moderate degeneration; C, heavy degeneration. PLATE 3 The pattern of degeneration in part of the anterior bank of the lunate sulcus following surgical transaction of the corpus callosum. A, dark field photomicrograph of a horizontal section through the lunate sulcus. Lateral is to the right and anterior is upward in the photomicrograph. Regions where degeneration is dense appear bright under dark field illumination. Other regions, such as the V1-V2 boundary (continuous line), do not appear brighter than background even though a sparse or even moderate density of degenerating terminals may be present. B, a photcmicrographic montage of a limited region of the anterior bank of the lunate sulcus (approximately 4 mm mediolaterally and 1 mm dorsoventrally). The reconstruction was made from a series of sections spaced at -25 mm intervals. Lateral is to the right and dorsal upwards in the montage. The level of the individual section shown above is indicated by arrows. The montage was made by cutting out the strip occupied by layer IV in each micrograph and aligning the strips for adjacent sections. The resultant picture is that which would be obtained by viewing layer IV face-on from within the lunate sulcus. The same region, illustrated in a different fashion, is outlined by dashed lines in Text-fig. 4.

35 The Journal of Physiology, Vol. 277 A Plate 1 D. C. VAN ESSEN AND S. M. ZEKI (Facing p. 226)

36 The Journal of Physiology, Vol. 277 Plate 2 D. C. VAN ESSEN AND S. M. ZEKI

37 The Journal of Physiology, Vol. 277 A I Plate 3 D. C. VAN ESSEN AND S. M. ZEKI