6 The Physiology of Color Vision - University of Illinois at Chicago · 2016-01-13 · 6.1...

217

CHAPTER CONTENTS

The Physiology ofColor Vision

Peter LennieCenter for Neural Science, New York University, New York, NY10003, USA

6

6.1 Introduction 2186.1.1. Basic anatomy of the visual system 218

6.1.1.1 The retina 2186.1.1.2 Central projections 2216.1.1.3 Visual cortex 221

6.1.2 Exploring function 2236.1.2.1 Recording signals from individual

neurons 2236.1.2.2 Recording larger scale electrical

activity 2266.1.2.3 Direct imaging of activity 226

6.2 Photoreceptors 2276.2.1 Receptor signals 2276.2.2 Visual properties 227

6.2.2.1 Spectral sensitivity 2276.2.2.2 Light adaptation 2286.2.2.3 Dynamics 229

6.3 Intermediate retinal neurons 2306.3.1 Horizontal cells and bipolar cells 230

6.3.1.1 Connections to cones 2306.3.1.2 Functional organization 2306.3.1.3 Amacrine cells 231

6.4 Ganglion cells and LGN cells 2316.4.1 Structural issues 2326.4.2 Function organization 232

6.4.2.1 Receptive field organization 2336.4.2.2 Contrast sensitivity 234

6.4.3 Chromatic properties 2346.4.3.1 Chromatic adaptation 235

6.4.4 Candidate chromatic and achromatic pathways 236

6.5 Cortex 2376.5.1 Structural issues 2376.5.2 Functional organization 238

6.5.2.1 Striate cortex 2386.5.2.2 Extrastriate cortex 239

6.5.3 Chromatic properties 2396.5.3.1 Separation of chromatic and achromatic

signals 2396.5.3.2 Number of chromatic channels 2406.5.3.3 Spatial contrast effects 2406.5.3.4 Private pathways for color 241

6.6 Acknowledgments 242

6.7 Notes 242

6.8 References 242

The Science of Color Copyright © 2003 Elsevier LtdISBN 0–444–512–519 All rights of reproduction in any form reserved

6.1 INTRODUCTION

Most of what we know about color vision hasbeen learned from psychophysical investiga-tions, most of what we know about the underly-ing physiology has been discovered with theexplicit guidance of theories grounded in psy-chophysical observation, and mostly the physio-logical findings have confirmed expectations.One might therefore be forgiven for supposingthat to discuss the physiology of color vision ismerely to provide an account of the mechanicsof systems whose operating principles we under-stand well. To some extent that is true, particu-larly for the earliest stages of color vision, butmodern physiological investigations have alsorevealed an organization that could not be sus-pected from psychophysical observations.

This chapter first reviews briefly the grossanatomy of the visual pathway, from the retina to

the occipital cortex. Then it examines the physiol-ogy of the different stages, beginning with a look atthe techniques used to explore it. Relatively littleof this work has been undertaken on the humanvisual system, but a great deal has been done onthe visual system of the macaque monkey, which,because its structure is similar to that of thehuman, is widely thought to be a good model.

6.1.1 BASIC ANATOMY OF THE VISUALSYSTEM

6.1.1.1 The retinaThe image is formed on the retina, shown in ver-tical cross-section in Figure 6.1. This highlightsvery clearly the layers that comprise a structureless than 0.5 mm thick.

The general organization of the retina isbroadly the same in all vertebrates: there are

■ THE SCIENCE OF COLOR

218

Figure 6.1 Vertical section through the primate retina, showing its layered structure. Light enters the retinafrom the bottom of the picture, passing through all layers before being absorbed in the outer segments ofphotoreceptors.The three principal layers of cells are identified.The rods and the cones lie nearest the top ofthe figure, with their different parts identified. Bipolar cells and amacrine cells lie in the inner nuclear layer.Ganglion cells lie in the ganglion cell layer. (From Boycott and Dowling, 1969.)

Outer SegmentsInner Segments

PhotoreceptorCell Bodies

Bipolar andAmacrine Cells

Ganglion Cells

three vertical stages, with interconnecting hori-zontal pathways at the junctions between stages.Figure 6.2 shows this diagramatically. The pho-toreceptors, rods and cones, which form themost peripheral stage, lie farthest from the pupil,and light must pass through the thickness of theretina before being absorbed. Since the neuralretina is transparent, this is visually inconse-quential. The inverted organization seems to bean adaptation to the demands of the photore-ceptors – metabolically the most active cells inthe body – which derive their nutrients from thenearby choroid. Structurally, rods and cones aregrossly similar, consisting of two clearly definedparts, the inner and outer segments. The outersegment, nearest the choroid, contains the pho-topigment, and within it originate the light-

evoked signals. The inner segment contains thebiological support mechanisms.

Until relatively recently most anatomical workon the retina used vertical sections of the kindshown in Figure 6.1. These make clear the verti-cal strata and also some structural features suchas the fovea (Figure 6.3), which contains no rodsand where all neurons beyond the cones aredisplaced, forming a pit over the very denselypacked cones.

Although neuroanatomists working with ver-tical sections have been able to identify somesub-classes of the major neuron groups identi-fied in Figure 6.2 (principally through scrutiny ofthe levels at which their dendrites and axonsbranch), the clearest indications of different sub-classes have often emerged through examination

THE PHYSIOLOGY OF COLOR VISION ■

219

Figure 6.2 Diagram of the neurons and their principal connections in the primate retina. C, cone; R, rod; MB,midget bipolar cell; DB, diffuse bipolar cell; S, S cone bipolar cell; H, horizontal cell;A, amacrine cell; MG, midgetganglion cell; PG, parasol ganglion cell; SG, S cone ganglion cell. Bipolar cells and ganglion cells colored red areoff-center types; those colored green are on-center types. (Adapted from Rodieck, 1998, with corrections byR.W. Rodieck.)

of horizontal sections of retina – sometimeswhole-mounted retinas – in which one can viewthe retina in the plane of the image, and whichreveal the pattern and spread of a neuron’sdendritic field.

It is now clear that the primate retina containsseveral kinds of bipolar cells, several kinds ofhorizontal cells, several kinds of ganglion cells,and perhaps many kinds of amacrine cells. Wewill consider some of these in detail later, but forthe moment the important idea is that the exis-tence of several types among each major class ofcell suggests the retina forms multiple represen-tations of the image. This idea draws clear sup-port from an examination of ganglion cells,whose axons form the optic nerve. Anatomicalclasses of ganglion cells are characterized prin-cipally by the pattern, horizontal extent, anddepth of branching of their dendritic fields. Thecharacteristic dimensions vary, of course, withposition on the retina, but at any one spot cellsof different classes can be robustly distinguished.Ganglion cells of the different classes form quasi-regular mosaics of sampling elements, each ofwhich conveys a (presumably) different repre-sentation of the image to the brain. These differ-ent arrays project differently into the brain. Theirrelevance to color vision stems from the fact thatdifferent classes of ganglion cells are connectedin different ways to the cone photoreceptors.

In the primate retina there are two major andseveral minor classes of ganglion cells. The two

major classes, now widely known as P cells andM cells,1 together constitute about 90% of the1.25 million ganglion cells in each eye. P cellsalone probably constitute 80% of all ganglioncells. The most distinctive anatomical differencebetween them is size: at any one eccentricity theP cells have much smaller cell bodies, smallerdendritic fields, and smaller axons. P and M cellsare connected to cones through different kindsof bipolar cells (they both also make indirectconnections with rods, but these are not relevanthere). In and near the fovea each P cell contactsa single midget bipolar cell, which in turn contactsa single cone (Wässle and Boycott, 1991). In thefovea, there are two midget ganglion cells andtwo midget bipolar cells for every cone. Eachcone drives two midget bipolar cells, which inturn drive two P cells. These dual contacts madeby a single cone seem to be the origin of distinctpathways, for the two midget bipolar cells areanatomically different, and they in turn contacttheir counterpart P cells in different planes in theretina. As we shall see, these pathways have dif-ferent physiological properties. There are corre-sponding pathways feeding two kinds of M cells.Each M cell contacts a single diffuse bipolar cell,which in turn contacts several cones. Other,rarer, kinds of ganglion cells are clearly distin-guished anatomically (Rodieck et al., 1993), butexcept for a bistratified cell that is associatedwith signals from S cones (Dacey and Lee, 1994),little is known about their function, and their


220

Figure 6.3 Vertical section through the human fovea, showing the elongation of the cones, and the absence ofthe other neurons, which lie outside the center of the fovea and to which the cones are connected through longfibers. (Photograph courtesy of Anita Hendrickson.)

Cones

FovealCenter

Bipolar andAmacrine Cells

Ganglion Cells

central projections, discussed below, suggest thatmost have no substantial role in color vision.

The horizontal pathways in the retina, repre-sented by horizontal cells and (to some extent)amacrine cells, seem to be organized to permitthe infusion of remote signals into the direct ver-tical pathways. Horizontal cells, of which thereare at least two types, make connections directlywith cones at their junctions with bipolar cells,and near the fovea each horizontal cell makescontact with most of the cones that lie in its den-dritic field (Wässle et al., 1989; Ahnelt and Kolb,1994). Amacrine cells exist in many more dis-cernible forms than do other retinal neurons.Some have an identified special role in thetransmission of rod signals from bipolar cells toganglion cells, but beyond some involvement inshaping the responses of ganglion cells, the rolesof most are unclear.

6.1.1.2 Central projectionsGanglion cells project to several centers in thebrain, notably the superior colliculus (SC) in themidbrain and the lateral geniculate nucleus(LGN) in the thalamus (Figure 6.4). Each opticnerve branches at the optic chiasm, and approx-imately half of the nerve fibers (representingganglion cells in the nasal retina) cross to theother hemisphere. The representation of theretinal image is split vertically through the fovea,each half being represented in one hemisphere.The peculiarity of the projection results in theleft visual field being represented in the righthemisphere, and vice-versa. This brings the repre-sentation of visual space broadly into registerwith the representations provided by othersenses, which are all crossed.

The SC is phylogenetically older, and is themore important center in lower mammals; inprimates it receives inputs from only a smallfraction of the retinal ganglion cells. These cellsare of the rarer types found in the retina (i.e.,neither P nor M cells). The SC has clear roles indirecting eye movements and in intersensorylocalization (Sparks and Nelson, 1987), but allthat we know about it, including the phys-iology of neurons in it, suggests that it has noconsequential role in color vision.

The LGN in primates is a highly developed,laminated, structure to which the retinal P cellsand M cells project. The prototypical LGN in the

primate has six layers, organized in two distinctgroups (Figure 6.5).

The four dorsal, parvocellular, layers receiveinputs from the retinal P cells, with those fromthe left and right eyes being interleaved. The twoventral, magnocellular, layers receive inputs fromthe M cells, separately from each eye. There areno discernible connections among layers, andphysiological evidence suggests no function con-nections. The topography of the retinal image ispreserved on the LGN, each layer of which con-tains an orderly though distorted map. The dis-tortions in the map are broadly consistent witheach projecting ganglion cell occupying a fixedterritory in the LGN, so that the representationof the central visual field is large – a conveniencefor physiologists.

6.1.1.3 Visual cortexNeurons in the LGN project to striate cortex (alsoknown as primary visual cortex or V1), an anatom-ically distinctive cortical region in the occipital


221

Figure 6.4 Major visual centers in the brain.Theoptic nerve (ON) from each eye (E) divides at theoptic chiasm (OC), sending half its fibers into the optictract (OT) of each hemisphere, which projects to thelateral geniculate nucleus (LGB).The lateralgeniculate nucleus in turn projects to the striatecortex (STR).The projection into each hemisphereoriginates in the half-retina on that side of the head.(From Polyak, 1957.)

e

on

ocot

lgb

or

str

lobe, at the back of the brain. The cortex as awhole is a large, thin (~2 mm) sheet containingseveral clearly defined layers of neurons, and inthe human brain (and to a lesser extent themonkey brain) is deeply folded to fit into thecranial cavity. Cortex is functionally specialized,but, with a few exceptions, its structure provideslittle evidence of this and almost everywhere issimilar anatomically. Figure 6.6 shows the majorlaminar subdivisions of striate cortex. It differsfrom cortex elsewhere in the brain by having amuch thicker layer 4, to which the incomingfibers from the LGN project.

The general organization of cortex is as follows:inputs from lower levels of the visual pathwayarrive in layer 4, ascending outputs to higherlevels of the pathway arise in the upper layers(above layer 4), and descending (feedback) out-puts to lower levels arise in the lower layers(below layer 4).

The anatomical distinctiveness of striate cortex(the thick layer 4 produces a texture visible tothe naked eye) makes it relatively easy to dis-cover that in each hemisphere it contains a

single map of the left or right half of the visualfield. As would be expected from the topographyof the earlier map in the LGN, this contains alarge representation of the central visual fieldand a progressively reduced representation ofthe peripheral field. Beyond this cortex lies agreat deal more that is intimately coupled tovision. Because it is not structurally distinctive,its organization has been less easy to discern, butmodern work, using both anatomical and physi-ological methods, shows unequivocally that itcontains multiple maps of the visual field. Thesemaps (at least twenty-two have now been iden-tified) cover the whole of the occipital lobe of thebrain, and substantial parts of the temporal andparietal lobes. The fraction of the brain theyoccupy, and their general arrangement, are mosteasily seen in representations that unfold thecortex and lay it flat. Figure 6.7 (Felleman andVan Essen, 1991) shows this done for themacaque monkey, and makes clear how large afraction (around 50%) of cortex is devoted tovisual analysis. In the human brain the fractionis considerably smaller, perhaps around 15%.

By tracing the connections among visualareas, and knowing from which layers in cortexthese connections originate and to which layersthey project, it becomes possible to discover ahierarchical structure that can accommodate,somewhat loosely, all the areas (Figure 6.8). Thishierarchy has its origin in striate cortex, ascendsthrough the second visual area, V2, and then onthrough perhaps seven further levels. BeyondV2 each level contains more than one area, sothe general picture that emerges is one of multi-ple parallel streams organized hierarchically.

The parallel organization of cortical streamsreflects to some degree, but by no means fully,the parallel organization of signals conveyed bythe M and P pathways through the LGN. The Mand P cells project to different subdivisions oflayer 4, and these in turn project to other layerswithin striate cortex. The separate identities ofthe pathways are generally not well-preservedbeyond this level, although a small but distinc-tive projection that appears to be dominated bysignals from the M pathway has been tracedfrom striate cortex to an extrastriate regionknown as the middle temporal area (MT).

Within the collection of extrastriate corticalareas concerned with vision, there appears to be


222

Figure 6.5 Section through the lateral geniculatenucleus (LGN), showing its laminated structure.The prototypical LGN has six layers – four dorsal(parvocellular) layers and two ventral (magnocellular)layers – though this arrangement is not alwayspresent.The magnocellular layers are thinner andcontain larger cells.The LGN contains a topographicalmap of the half of the retina that projects to it.The arrows in this figure mark the path of amicroelectrode that recorded the discharges ofindividual neurons. (From Wiesel and Hubel, 1966.)

a broad division into two groups. One contains aset of heirarchically connected areas that form aconduit from striate cortex, through area V2,into the parietal lobe of the brain. The other con-tains a set of areas that carry information fromstriate to the temporal lobe. Experimental andclinical studies of parietal cortex show it to haveimportant roles in spatial orientation and visuallocalization, and perhaps also with the analysisof self-movement, but there are no indicationsthat it is important for color vision. Temporalcortex is for object vision, and damage to it, or tothe visual areas that lead to it, can bring aboutprofound disruption of various aspects of objectvision, including color vision.

6.1.2 EXPLORING FUNCTION

Some useful inferences about function can bedrawn from our knowledge of the structure ofthe visual pathway (for example, one can inferthat P cells are probably important for visualacuity), but a great deal more can be learned byexamining activity in the nervous system.Physiologists have available to them a powerfularmament of methods. For the purposes ofunderstanding mechanisms of color vision, andparticularly how these give rise to perceptual

phenomena, most of the activity we need to pur-sue occurs on a time scale of tens of msec, andon a spatial scale that involves groups of neu-rons. The following paragraphs review brieflysome of the more important methods that arerelevant to exploration on these scales, and out-line their strengths and weaknesses.

6.1.2.1 Recording signals fromindividual neurons

Within an individual neuron signals are propa-gated electrically, and between neurons, chemi-cally, via neurotransmitters. In their resting statesmost neurons maintain a steady potential differ-ence of ~60 mv across the cell membrane. Areduction in this potential difference (depolar-ization) constitutes the neural signal. The mem-brane potential is controlled by neurochemicalevents at the synapses (junctions) between thecell and those from which it receives signals. Aneuron receives most of its synaptic connectionson its dendrites and soma (Figure 6.9).

Each neuron receives thousands of synapticconnections from other neurons, and eachsynapse, when active, gives rise to a brief event,a postsynaptic potential, that tends either to excite(depolarize) or inhibit (hyperpolarize) the neu-ron. The aggregate weight of the postsynaptic


223

Figure 6.6 The arrangement of layers in striate cortex, revealed by two stains that highlight different aspectsof the organization. (Right) A stain that reveals the locations of the cell bodies of neurons. (Left) A stain for theenzyme cytochrome oxidase, which labels densely the input layers in the middle of cortex, and also shows the‘blobs’ in layer 2/3 that are a distinctive feature of the upper layers.

1

2/3

4A4B

4Ca

4Cb

5

6

potentials within a certain integration time (andover a certain integration distance) determinesthe resulting change in the membrane potentialof the cell. In most neurons, if the cell becomessufficiently depolarized (taking the membranepotential from perhaps 60 mv to 50 mv) a largeimpulsive depolarization (action potential) is trig-gered and propagated rapidly along the mem-brane. When this action potential reaches theaxon terminals, it causes the release of neuro-transmitter at synapses making contact withother neurons. Activity at the synaptic connec-

tions between cells thus controls the propaga-tion and transformation of signals in the nervoussystem.

The impulsive nature of most neural commu-nication permits signals to be transmitted rapidlyand reliably (within broad limits the existence ofan action potential is the important event, ratherthan its size), and almost all neurons propagatesignals via action potentials. A drawback, how-ever, is that the dynamic range of a neuron islimited (the rates at which action potentials aredischarged vary from a few per second to, at most,


224

Figure 6.7 Diagram of the cortex unfolded from one hemisphere of the monkey, showing identified visualareas.The areas that are colored are recognized as being visual sensory areas, or areas closely associated withthem.V1 is the primary visual (striate) cortex, which receives input from the lateral geniculate nucleus.V1projects principally to area V2. Many of these visual areas contain topographically organized maps of the half ofthe retina that sends a projection to the hemisphere.Visual areas constitute about half of the monkey’s cortex.Corresponding areas in the human cortex undoubtedly account for a smaller fraction of the total. (FromFelleman and Van Essen, 1991.)

a few hundred per second). In rarer kinds ofneurons, found particularly in sensory systems,the changes in membrane potential induced by

synaptic events are propagated passively alongthe membrane, as relatively slowly changingpotentials. This mechanism of transmission is


225

Figure 6.8 Diagram showing the hierarchical organization of visual cortical areas and the known connectionsamong them.The areas at the highest level are at the top.The hierarchy is inferred from the polarities of theconnections between areas. Each element in the diagram corresponds to an area identified in Figure 6.7. (FromFelleman and Van Essen, 1991.)

used by all retinal neurons except ganglion cells.It provides neurons with a larger dynamic rangethan would be possible were action potentialsused, but at the cost of slower and less reliablesignal transmission over distances.

In many parts of the nervous system individ-ual neurons are large enough to withstand pen-etration by a microelectrode that can record the

potential difference across the membrane. Evenwhere this is not the case microelectrodes that sitjust outside the membrane can record (extracel-lularly) the local electrical disturbance producedby an action potential, though not any slowpotentials. Intracellular methods are usuallymost practicable in vitro, when the neurons arenot subject to movement by the pulse and othersmall body movements, but extracellular meth-ods are much more robust, and can record sig-nals from individual neurons in moving animals.The small size of neurons in the primate retinamakes it exceptionally difficult to record theiractivity with intracellular electrodes. Some spe-cial techniques have been developed for study-ing the signals generated by photoreceptors (seethe next section), but apart from photoreceptorsonly ganglion cells have been studied exhaus-tively in primates, because they generate actionpotentials that can be recorded with extracellu-lar electrodes. Much of what we know directlyabout the function of other retinal neurons hasbeen learned through studying reptiles and fish,in which cells are much larger. Elsewhere in theprimate’s visual system, extracellular recordingsprovide valuable information about the charac-teristics of individual neurons.

6.1.2.2 Recording larger scaleelectrical activity

Gross electrical activity evoked by light can berecorded from several places in the nervous sys-tem. The electroretinogram (ERG) is a compositesignal that can be recorded either with electrodesplaced on the retina, or with surface electrodeson the intact eye. Elements of the compositelight-evoked signal can be attributed to differentstages of signal analysis in the nervous system.The ERG has been particularly helpful in theanalysis of signals generated by photoreceptors.

Gross electrical signals evoked by visual stim-uli can be recorded with electrodes placed on thescalp, or on the surface of the cortex. Althoughlarge signals can be evoked by time-varyingchanges in color, they are often difficult to inter-pret, because the source of the signals is hard tolocalize.

6.1.2.3 Direct imaging of activityActive neurons consume more oxygen thaninactive ones, and as a result provoke local


226

Figure 6.9 Diagram of the general structure ofneurons, and their specializations. (Left) Bipolar cell inthe retina. (Right) Pyramidal cell in cortex.Theneuron accumulates signals from other cells mostlythrough synaptic connections on its dendrites.Theaction potential usually originates at the junction ofthe cell body and axon, and is propagated down theaxon to the axon terminal system where it causes therelease of neurotransmitter at synapses. (Adaptedfrom Kuffler and Nichols, 1976.)

Bipolar Cell

Pyramidal Cell

Dendrites

Dendrites

Dendrites

Axon

Axon

increase in the circulation of blood. Severalrecently developed techniques can record thesestimulus-induced local changes in the circula-tion, and have been exploited in attempts toreveal cortical structures that might be involvedin the analysis of color. In cortex that can beviewed directly following removal of the skull,one can record changes in the composition oflight reflected from the cortical surface to pro-duce a map that shows local regional specializa-tion (Roe and Ts’o, 1995). Positron emissiontomography can be used to measure localchanges in circulation in the human brain, andhas been used to identify regions that areengaged in the analysis of color. The attraction ofthis method, and similar methods that exploitmagnetic resonance imaging (Engel et al., 1997),is that they reveal activity on a scale of mil-limeters. The drawback is that they measurechanges in circulation occurring on a time scaleof seconds.

6.2 PHOTORECEPTORS

6.2.1 RECEPTOR SIGNALS

Many of the important steps in the transductionof light to electrical signals in the nervous systemare now understood, and have been clearlydescribed (see, for example, Baylor 1987; Torreet al., 1995). The general functional principlesare the same in rods and cones, although thereare important differences of detail betweenthem, notably in the speed of responses, and intheir time-courses. In the dark the receptormaintains a steady potential of ~ �40 mv acrossits membrane. When the receptor is illuminated,the potential becomes hyperpolarized to asmuch as �70 mv. This hyperpolarization by aneffective stimulus makes photoreceptors uniqueamong vertebrate neurons, which normally be-come depolarized when excited. Primate conesare too small to accommodate microelectrodesthat can measure change in membrane poten-tial, but Baylor and colleagues (Schnapf et al.,1990) have characterized the electrical responsesin vitro by drawing the outer segment of a coneinto a micropipette and measuring the currentflow through the outer segment. In darkness,there is a steady inward current (the dark current)

of 30–50 pA. This arises from the flow of Na�ions through the permeable membrane of theouter segment, and a flow of K� ions out of theinner segment. The current is maintained bypumps in the inner segment that drive Na�ions out and K� ions in. During illumination,the Na� permeability of the outer-segmentmembrane is reduced, and the current flowcorrespondingly reduced, resulting in hyper-polarization of the membrane. These light-induced changes in outer-segment membraneconductance result from a recently discoveredcascade of biochemical events inside the outersegment (Pugh and Lamb, 1993; Torre et al.,1995). In darkness, the photoreceptors releaseneurotransmitter continuously from their termi-nals, which contact bipolar cells and horizontalcells. The absorption of light and consequenthyperpolarization of the cell reduces the amountof transmitter released.

6.2.2 VISUAL PROPERTIES

Figure 6.10 shows the change in current flow-ing through the outer segment of a singlemonkey cone induced by a series of brief flashesof progressively increased intensity, at threewavelengths.

Several important principles can be inferred.First, over a substantial range of progressivelyincreasing flash intensities, the cone’s responseshave exactly the same shape, and can be super-imposed if scaled by the flash intensity. Thislinear relationship between light intensity andresponse breaks down at high flash intensities,where the responses saturate. Second, the conesgenerate sets of responses of identical shaperegardless of the wavelength of light that excitesthem – they are univariant. Third, the responsesare biphasic, and last considerably longer thanthe flash.

6.2.2.1 Spectral sensitivityGiven the univariance of a cone, demonstratedvery clearly by Baylor et al. (1987), its spectralsensitivity can be readily measured by finding, ateach of several wavelengths, the flash intensityrequired to evoke a response of a specified size.Figure 6.11 shows the result of essentially thiskind of measurement made on a sample of conesfrom the macaque retina (Baylor et al., 1987).


227

These measurements, remarkable for the rangeof wavelengths over which they can be made,show higher sensitivity at short wavelengthsthan is found in curves of cone fundamentalsmeasured psychophysically. Differences of thiskind would be expected from the different cir-cumstances under which measurements aremade – the psychophysically derived fundamen-tals include the effects of selective absorption oflight by passive filters in the eye, and of self-screening by photopigment. After correcting forthe effects of these other absorbing filters, alinear transformation of the cone spectral sensi-tivities describes well the shapes of the color-matching functions measured by Stiles andBurch (1959).

The measurements in Figure 6.11 are averagesfrom populations of cones that fall very tightlyinto three distinct groups. No electrophysiologicalmeasurements have revealed cones with anom-alous spectral sensitivities, neither have spec-trophotometric methods that measure directlythe absorption spectrum of the photopigment inthe outer segment, and which have been appliedto larger numbers of cones from retinas of old-world primates. Anomalous pigments have beenfound in the cones of new-world primates(Mollon et al., 1984).

6.2.2.2 Light adaptationThe small dynamic range of impulse-dischargingneurons requires that the retina map an enor-mous range of light levels on to a small range ofsignal amplitudes in optic nerve fibers. We knowsomething of the general principles that governthis relationship between output and input: theretina as a whole preserves little informationabout the absolute level of illumination, and atany moment the range of outputs is mapped on toa small part of the input range around the ambi-ent level of illumination; the input range repre-sented in the output is larger at higher levels ofillumination. The upshot is that the signal enter-


228

Figure 6.10 Current flow induced in the outer-segment of the three kinds of primate cones. Eachset of traces shows the responses to brief flashes(identified by the pulse at bottom) at a series ofintensities stepped by factors of 2.The responses oflow and middle amplitudes would be superimposed ifscaled by the flash intensity. (From Schnapf et al.,1990.)

30 pA

20 pA

10 pA

0 0.2 0.4 s

Blue

Green

Red

Figure 6.11 Spectral sensitivities of cones in theretina of the macaque monkey. From left to right thecurves show the averaged spectral sensitivities ofsmall samples of S, M, and L cones. (From Baylor et al.,1987.)

Log

rela

tive

sens

itivi

tyWavelength (nm)

�4

�5

�6

�1

�2

�3

0

400 500 600 700 800 900

ing the optic nerve conveys information aboutthe local contrast in the image. To achieve thisrequires two kinds of regulating mechanisms: asubtractive one to discard the signal about theambient level of illumination, and a multiplica-tive one to reduce the gain of signal transmission(Walraven et al., 1989). Cones contribute some-thing to the normalization of these signals.

Because each cone – at least in and near thefovea – appears to have an essentially privatepathway through the retina, and does not poolits signals with those from other cones, therewould appear to be some advantage in regulat-ing sensitivity at the receptors. Psychophysicalmeasurements demonstrate light-adaptation inmechanisms that have spectral sensitivities closeto those of individual cones (Stiles, 1939); physi-ological measurements, made by recording extra-cellular current flow across outer segments(Valeton and van Norren, 1983), in vivo, meas-urements of current flow in single cones in vitro(Schnapf et al., 1990), and measurements of thea-wave of the electroretinogram (Hood andBirch, 1993) show that steady illuminationreduces the sensitivity of cones through amultiplicative gain reduction, and possibly alsosome response compression. At high levels of illu-mination (above 104 td), photopigment bleach-ing contributes substantially to the reduction insensitivity. It is not clear from physiologicalmeasurements that the regulation of sensitivitywithin cones is sufficient to explain gain changesmeasured psychophysically; the latter are evi-dent at low levels of illumination that do notdesensitize cones, and implicate mechanismsbeyond cones, perhaps at the synaptic connec-tions between cones and bipolar cells.

There appears also to be some subtractivemechanism within the cones, for over much oftheir operating range the steady-state responses tostanding background illumination vary littlewith level of illumination. Figure 6.12, whichshows how flash responses vary with thestrength of background and flash, illustrates theeffects of both desensitizing gain changes andsubtractive changes that stabilize the responsesto standing backgrounds.

6.2.2.3 DynamicsA cone’s biphasic response to a flash (Figure 6.10)implies a band-pass frequency characteristic. The

Fourier transform of the response to a pulse canbe used to reveal a cone’s sensitivity to differenttemporal frequencies. When this is done to amoderately light-adapted cone (Figure 6.13) wesee that it has a pass-band that peaks near 5 Hz,with sensitivity falling on either side.

The extent to which cones limit the temporalresolving power of the visual system is unclear.The temporal characteristics of vision dependon the spatial (Robson, 1966), and chromatic


229

Figure 6.12 Responses of primate cones to a seriesof flashes of different intensity, recorded in thepresence of backgrounds at a range of intensities. Eachcurve represents the set of responses obtained on asingle background, whose intensity is marked by theshort horizontal bar on that curve. Points to the left ofthe bar represent responses to decremental flash,points to the right, responses to incremental flashes.The operating range of the cones is set by thebackground. (From Valeton and van Norren, 1983.)

Tota

l res

pons

e, V

(lV

)

Log intensity, I (log td)

800

600

400

200

0

�200

5 6 74321��

DA 222 3 4 5 6

Figure 6.13 Temporal modulation transfer functionof primate cones, obtained by Fourier transform ofthe impulse response. (Courtesy of W. Makous.)

Log

Rel

ativ

e Se

nsiti

vity

Frequency, Hz

�1

�0.5

0

�1.5

�2.0

�2.5100101

(Estévez and Spekreijse, 1974) attributes of stim-uli in ways that suggest that cones seldom limitthe temporal resolving power of the visual sys-tem; measurements of the late receptor potential(a component of the electroretinogram attrib-uted to photoreceptors) point to the same con-clusion (Boynton and Baron, 1975).

Thekineticsof thecone’sphotocurrentresponseto flashes vary little with level of light-adaptation(Schnapfetal.,1990),sotheadaptation-dependentchanges observed psychophysically (Kelly, 1961)apparently must reflect action at later sites.

6.3 INTERMEDIATE RETINALNEURONS

6.3.1 HORIZONTAL CELLS ANDBIPOLAR CELLS

6.3.1.1 Connections to conesHorizontal cells and bipolar cells make synapticcontacts with cones through a specialized con-nection (the ‘triad’) that seems designed to per-mit horizontal cells to regulate the transmissionof signals from cones to bipolar cells. The triadconsists of a central bipolar cell dendrite flankedby two horizontal cell dendrites. Each coneaccommodates 20–30 triads in invaginations inits pedicle (foot).

Two morphological classes of horizontal cells(H1, H2) have been identified in the primateretina; a third class (H3) might also be present(Kolb et al., 1994). The H2 cell contacts onlycones; the H1 horizontal cell contacts both rodsand cones: dendrites contact cones, axon termi-nals contact rods. The axon is long and thin andapparently provides rather poor communicationbetween the two ends of the cell, which areoften considered to operate relatively independ-ently. Each horizontal cell appears to contactevery L and M cone within its dendritic field (inthe fovea covering perhaps six or seven cones, inthe periphery perhaps two to three times asmany), but only H2 cells contact all S cones(Wässle et al, 1989; Ahnelt and Kolb, 1994). H1cells make very few contacts with S cones(Ahnelt and Kolb, 1994; Goodchild et al., 1996).

Two morphologically distinctive types of bipo-lar cells contact cones: diffuse bipolar cells and

midget bipolar cells (other types contact rods).Diffuse bipolars contact 5–7 cones (Boycott andWässle, 1991); a midget bipolar cell contacts asingle cone, everywhere in the retina (Wässleet al., 1994). Each class of bipolar cell appears toexist in two sub-types, named for the character-istic morphology of their cone connections. Aninvaginating bipolar cell inserts a dendritic termi-nal into the cone pedicle, forming the centralelement of a triad; a flat bipolar cell makes con-tacts with the surface of the pedicle. In the cen-tral retina each cone probably makes contactwith four bipolar cells, one of each kind (Wässleand Boycott, 1991).

6.3.1.2 Functional organizationPrimate horizontal cells have not been studiedin vivo, but single-unit recordings in cat, wherehorizontal cells resemble in form and connec-tions the H1 type in primate, show some mixingof rod and cone signals at levels of illuminationwhere both receptor types are active (Lankheetet al., 1991). In vitro recordings from H1 cells inprimate (Verweij et al., 1999) show that they tooreceive rod input.

The dense connections horizontal cells makewith the cones underlying their dendritic fieldsimplies mixing of signals from cones of differenttypes. In vitro recordings from horizontal cells inthe intact retina (Dacey, 1996; Dacey et al., 1996)show that H1 cells receive signals of the samepolarity from L cones and M cones, but not Scones; H2 cells receive signals from all coneclasses, with that from S cones being conspicuous.

Bipolar cells receive direct signals from conesand perhaps directly from horizontal cells. Eachbipolar cell is connected to a region of retina (or,equivalently, it views a region of visual space)within which light will evoke some physiologicalresponse. This is known as the receptive field.Receptive fields in vertebrates typically consist oftwo concentrically organized regions, a circularcenter and enclosing, overlapping, surround(Naka, 1976), and the primate is no exception(Dacey et al., 2000). Center and surround gener-ate signals of opposite polarity, so that whenboth are illuminated together, the bipolar cellresponds poorly. The receptive field surroundappears to originate in H1 horizontal cells, whichinfluence the bipolar cell either directly, or byregulating transmission of signals from cones


230

(Dacey et al., 2000). The midget bipolar cell inthe primate will therefore have a receptive fieldin which the center and surround have differentspectral sensitivities: the spectral sensitivity ofthe center is that of a single cone, while the spec-tral sensitivity of the surround is that of horizon-tal cells that accumulate signals from both L andM cones. The spectral sensitivities of the centerand surround of a diffuse bipolar cell are proba-bly more alike, because each mechanism drawsinputs from several cones.

L and M cones, and their associated pathways,are not distinguishable by modern histologicalmethods, but Kouyama and Marshak (1992)have identified an immunocytochemically dis-tinct class of invaginating midget bipolar cellsthat are connected only to S cones. In mostinstances each bipolar cell contacts a single cone,although each S cone contacts more than onebipolar cell (Mariani, 1984; Kouyama andMarshak, 1992).

There are two functional types of bipolar cells:one in which illumination of the center depolar-izes the cell and illumination of the surroundhyperpolarizes it (often known as ON-bipolarcells), another in which illumination of thecenter hyperpolarizes the cell, and illuminationof the surround depolarizes it (OFF-bipolarcells). These two fundamental types seem toexist in all retinas, and provide the first stages ofon- and off-pathways that remain distinct as faras visual cortex.

The on- and off- types of bipolar cell appear tobe the physiological expression of an anatomicaldistinction we noted earlier: on-bipolar cellsmake invaginating contacts with cones; off-bipolar cells make flat contacts with cones (Stellet al., 1977). In general, each cone therefore pro-vides a signal to two on-pathways (midget anddiffuse), and two off-pathways. The organizationof connections to S cones might be different: apresumed on-pathway is known to exist, butalthough a counterpart off-pathway might beinferred from the presence of some flat bipolarcell contacts that appear on S cones (Kouyamaand Marshak, 1992), the bipolar cells makingthese contacts have not been identified.

6.3.1.3 Amacrine cellsAmacrine cells (named by Cajal for their lack ofan axon) lie in the inner retina and make con-

nections with bipolar cells and ganglion cells.They exist in a wide variety of morphologicaltypes (Masland, 1988; Wässle and Boycott,1991). With rare exceptions, little is known abouttheir roles. Some amacrine cells might have littleto do directly with vision, instead controllingfunctions such as eye-growth (Schaeffel et al.,1995). One class of amacrine cell (AII) providesan essential link in the chain from rod photo-receptor to ganglion cell: it links the rod bipolarcell through an inhibitory synapse to a diffusecone off-bipolar cell, and through an excitatorysynapse to an on-bipolar cell. These cone bipolarcells in turn contact ganglion cells (Sterling et al.,1988). Amacrine cells that contact midget bipo-lar cells and midget ganglion cells make indis-criminate contacts with all midget bipolar cellswithin reach. These include bipolar cells drivenby both L and M cones (Calkins and Sterling,1996). Most types of amacrine cells are notdirectly interposed between bipolar and ganglioncells, and the general organization of theirconnections suggests that they modulate thetransmission of information from bipolar cellsto ganglion cells, or add components to thereceptive fields of ganglion cells.

No physiological recordings have been ob-tained in vivo from amacrine cells in primates, butwhere recordings have been obtained from otherspecies amacrine cells have often exhibited com-plex behaviors (for example giving depolarizingresponses to both light onset and light offset) thatsuggest an important role in the formation ofsome distinctive nonlinear behaviors in kinds ofganglion cells that project to the superior collicu-lus (see below). Where receptive fields have beencharacterized (Kaneko, 1973) they are circular,but have no center-surround organization.

6.4 GANGLION CELLS ANDLGN CELLS

Neurons in LGN are, in their general behavior,almost indistinguishable from the ganglion cellsthat drive them, so it is convenient to considerboth kinds together. Because different classes ofLGN neuron are segregated in different layers ofthe nucleus, and can be selected for study, LGNneurons are more often studied than ganglioncells.


231

6.4.1 STRUCTURAL ISSUES

In primate retina the principal classes of bipolarcells, midget and diffuse, make specific connec-tions in the inner plexiform layer (IPL) with thetwo kinds of ganglion cells known as parasol andmidget cells. A midget bipolar cell in the fovea(Calkins et al., 1994), and to eccentricities of per-haps 6�, makes exclusive contact with a singlemidget ganglion cell, providing each cone with aprivate pathway through the retina. A parasolcell in the fovea receives inputs, via diffuse bipo-lar cells, from perhaps 30–50 cones (Grünertet al., 1993).

The on- and off- types of each of the majorclasses of bipolar cell contact their counterpartganglion cells at different depths within theretina, permitting the identification of on- andoff-center ganglion cells in anatomical sections(Perry et al., 1984; Watanabe and Rodieck,1989). The on-bipolars contact on-center gan-glion cells in a stratum of the IPL closer to theganglion cells (Famiglietti and Kolb, 1976).

Recent anatomical evidence points to reliablestructural differences among midget pathwaysthat might be associated with the L and M cones.Calkins et al. (1994) found that (within each ofthe major classes, on- and off-) most midgetbipolar cells in the monkey retina made one oftwo different kinds of synaptic connections withmidget ganglion cells: either the synapse hadaround 50 synaptic ribbons, or it had around 30,with no overlap between the distributions. Thereis a corresponding disparity in the numbers ofsynaptic ribbons each kind of bipolar cell makeswith the cone that drives it. The bipolar cellsmaking the two kinds of connections were ran-domly distributed on the retina. Among thebipolar-ganglion connections studied, 44% hadthe larger number of synaptic ribbons. These dif-ferent synaptic connections might reflect differ-ences between the pathways conveying signalsfrom L and M cones, although it is not knownwhich kind of cone might be associated witheach pathway.

It is not known if any midget ganglion cellsconvey S cone signals, but these signals areknown to be carried by another, relatively rare,kind of ganglion cell described by Rodieck(1991) and Dacey (1993). This small bistratifiedcell has dendrites that branch in two planes in

the IPL, one in the region where off-bipolarscontact midget ganglion cells, and the other inthe region where on-bipolars contact midgetganglion cells. The span of the bistratified cell’sdendrites matches that of the parasol cell, so inthe fovea might cover 30–50 cones, and the cellmight receive signals from 3–5 S cones. Smallbistratified cells constitute less than 2% of gan-glion cells in and near the fovea (Dacey, 1994),and probably provide a mosaic that is too coarseto account for the spatial resolving power ofthe S cone system measured psychophysically(Williams and Collier, 1983). This and other evi-dence to be reviewed below suggests that theretina must contain additional S cone pathways.

Parasol and midget cells project respectively tothe magnocellular and parvocellular divisions ofthe LGN (Perry et al., 1984); small bistratifiedcells project to the parvocellular layers (Rodieck,1991). The internal organization of the LGN islocally complex with afferent fibers from gan-glion cells making contact with relay neuronsthat convey signals to cortex, and with interneu-rons. The LGN also receives projections fromseveral other parts of the brain. Nothing thatwe know of the internal organization helps usunderstand color vision.

6.4.2 FUNCTION ORGANIZATION

The action potentials discharged by a ganglioncell or LGN cell can be recorded with an elec-trode that sits outside the neuron, and thismakes it possible to record from individual neu-rons in an intact animal, and to study their visualsensitivities. Where the visual properties of gan-glion cells have been compared with those of theLGN neurons to which they project, they havebeen found to be essentially identical, exceptperhaps for differences in sensitivity that proba-bly reflect the action of regulatory mechanismsin the LGN (Sherman, 1996). In the followingdiscussion ganglion cells and LGN neurons aretherefore distinguished only when their proper-ties are known to differ. The physiological coun-terparts of the anatomically identified parasoland midget cells are named for their projectionsto magnocellular and parvocellular layers in theLGN, and are known as M and P cells, respec-tively. This terminology provides some potentialfor confusion, for it does not accommodate


232

the small bistratified cells that project to theparvocellular layers.

6.4.2.1 Receptive field organizationBoth P (parvocellular) and M (magnocellular)cells have receptive fields organized into twoconcentric antagonistic regions: a center (on- oroff-) and a surrounding region of opposite sense.This arrangement is common in vertebrates. Thereceptive fields of small bistratified cells appearto lack clear center–surround organization(Dacey and Lee, 1994).

The distributions of sensitivity within centerand surround mechanisms are usually repre-sented by Gaussian profiles of different extentsand opposite polarity (Figure 6.14A; Rodieck,1965); the properties of these mechanisms arecommonly inferred from a neuron’s spatial mod-ulation transfer function (or contrast-sensitivityfunction (Enroth-Cugell and Robson, 1966))measured with grating patterns whose lumi-nance is modulated sinusoidally about a con-stant mean (see Figure 6.14B).

Neurons respond to moving or counterphase-flickering gratings with a modulated dischargethat reflects the time-varying changes of lumi-nance locally within the receptive field. Contrastsensitivity is typically measured by finding, as afunction of the spatial frequency of a grating,the contrast required to evoke a modulated dis-charge of specified amplitude (Figure 6.14C).Such measurements show that, to a close approx-imation, both center and surround accumulatecone signals linearly, and the cell responds to thesum of the signals aggregated from the two mech-anisms (Kaplan and Shapley, 1982; Derringtonand Lennie, 1984). It is important to note that thislinear behavior is observed when neurons arestably adapted to the average luminance of thegrating and are driven by modulations about thismean; changing mean luminance leads to theexpression of nonlinearities (see below).

As one would expect from the anatomy oftheir connections, M cells have larger receptivefields than P cells. Although anatomical evidencepoints to the center of a P cell receptive fieldoriginating in a single cone, conventional meas-urements of contrast sensitivity only rarely hintat this (Derrington and Lennie, 1984; Blakemoreand Vital-Durand, 1986). Measurements madewith interference fringes formed directly on the


233

Figure 6.14 (A) Diagram showing the distributionof sensitivity within the center and surroundmechanisms of a ganglion cell’s receptive field.Thepeak sensitivity of the center greatly exceeds that ofthe surround, but because the center is much smallerthan the surround, the center only slightly dominateswhen both regions of the receptive field areilluminated. (B) Sinusoidal grating patterns of thekind used to explore the spatiotemporal transfercharacteristics of visual neurons. (C) Contrastsensitivity function obtained from a parvocellularneuron in LGN. (From Derrington and Lennie, 1984.)

Con

tras

t se

nsiti

vity

Spatial frequency (cycles deg�1)

1

10

100

0.1 1 10 100

(C)

(B)

(A)

retina (by-passing the optics of the eye) are con-sistent with the center receiving signals from asingle cone (McMahon et al., 2000).

6.4.2.2 Contrast sensitivityM and P cells differ in their sensitivities to con-trast, M cells having several times greater con-trast sensitivity to achromatic gratings (Kaplanand Shapley, 1982; Derrington and Lennie,1984). Over a range of low to moderate con-trasts, the responses of both M and P cells tostimuli of optimal spatial frequency grow lin-early with contrast, but the greater sensitivityof M cells leads to their giving responses ofsaturating amplitude to high-contrast stimuli(Derrington and Lennie, 1984). Moreover, thetemporal frequency at which sensitivity peaks isaround 20 Hz in M cells and around 10 Hz inP cells (Derrington and Lennie, 1984). This,coupled with the higher contrast sensitivity ofM cells, results in M cells responding to a rangeof high frequency signals that are apparentlynever seen by P cells.

Signals from the center and surround of areceptive field travel through different pathwaysto reach a neuron, and are subject to differentdelays and attenuation, so it is not surprisingthat the interaction between center and sur-round varies with the temporal frequency ofvisual stimulation. As temporal frequency israised, center–surround antagonism is appar-ently reduced, so the spatial contrast sensitivitycurve (Figure 6.14C) tends to lose its distinctiveband-pass shape, becoming instead low-pass.This behavior is consistent with surround signalslagging center signals by several milliseconds(Derrington and Lennie, 1984).

6.4.3 CHROMATIC PROPERTIES

In the M cell’s receptive field, center and sur-round generally have different spectral sensitivi-ties, although the differences are not oftenconspicuous. The center has a spectral sensitivityclose to that of Vk (see Chapter 2), thereforedrawing its inputs from L and M cones, while thesurround draws on some different mix of conesignals (Derrington et al., 1984; Kaiser et al.,1990), possibly even having some local chromat-ically opponent structure within it (Schiller andColby, 1983; Lee et al., 1989).

In the P cell’s receptive field, center and sur-round have overtly different spectral sensitivities(Wiesel and Hubel, 1966; Gouras, 1968; Dreheret al., 1976; Derrington et al., 1984). The differentspectral sensitivities of center and surroundendow the P cell with curious properties: thespatial contrast sensitivity curve measured withachromatic gratings has a characteristic bandpassshape that reflects that antagonistic interactionbetween center and surround (Figure 6.14), butthe curve measured with chromatic gratings(sinusoidal modulation of chromaticity ratherthan luminance) has a lowpass shape, so thatsensitivity to chromatic modulation is highestwhen a spatially uniform field of light is modu-lated in time. For this reason a spatially uniformfield is now often used to characterize the chro-matic properties of receptive fields. The lumi-nance and/or chromaticity of this field areusually modulated in time about some pointnear the center of the chromaticity diagram.

P cells fall into two chromatic classes, looselyred–green and yellow–blue (De Valois et al.,1966; Wiesel and Hubel, 1966; Gouras, 1968).The analysis of cone inputs reveals one class thatreceives opposed inputs from L and M conesonly, and a second class that receives inputs fromS cones opposed to some unspecified combina-tion of signals from L and M cones (Derringtonet al., 1984; Figure 6.15).

The techniques now used to characterize thechromatic properties of P cells show that theneurons receive opposed inputs from differentcone classes, but do not reveal the spatial distri-bution of these inputs. Until recently, it was pre-sumed that the different kinds of cones whosesignals were opposed were cleanly segregated incenter and surround of the receptive field (bear-ing in mind possible exceptions for cells drivenby S cones), but several lines of evidence haverecently prompted a closer look at alternatives.First, if indeed the P cell receives its center inputfrom a single cone, it will have a chromaticallyopponent receptive field even if its surrounddraws indiscriminately on all cone classes(Lennie, 1980). Second, the recently discoveredsimilarity of the genes that encode L and M conepigments, and the very similar structures of thecone pigments themselves (Nathans et al., 1992),raise the possibility that the visual system mightnot distinguish them during development; third,


234

the connections horizontal cells make withcones (Wässle et al., 1989), and the physiologicalsignals recorded from horizontal cells (Dacey,1996) suggest that they receive mixed signalsfrom the different classes of cones. Fourth, theamacrine cells that contact midget bipolar cellsand midget ganglion cells make indiscriminatecontact with the bipolar cells driven by bothkinds of cones (Calkins and Sterling, 1996).Lennie et al. (1991) showed that the propertiesof L–M opponent neurons were consistent witheither pure cone input to the surround, or mixedcone input from L and M cones, but not S cones.Reid and Shapley (1992) have argued that the Pcell’s surround receives inputs from a single classof cone. Given the mixing of cone signals presentin horizontal cells and amacrine cells, this wouldrequire some elaborate organization that segre-gates cone signals again. In vitro recordings frommidget ganglion cells in the periphery showmixed cone input to both center and surround(Dacey and Lee, 1999); no recordings have yetbeen made from central retina.

Dacey and Lee (1994) showed that the small

bistratified ganglion cell that projects to parvo-cellular LGN receives ‘on’ (depolarizing) signalsfrom S cones and ‘off’ signals from L and Mcones. Neurons that receive strong ‘off’ (hyper-polarizing) signals from S cones, and ‘on’ signalsfrom L and M cones, also exist, although theyare encountered less frequently (de Monasterioand Gouras, 1975; Derrington et al., 1984;Valberg et al., 1986). Their anatomical substratehas not yet been found. As was noted in an ear-lier section, there are probably yet other kinds ofneurons that carry signals from S cones, for inand near the fovea the sampling density of thesmall bistratified system is too low to account forthe visual acuity of the S cone system.

Just as the different temporal characteristics ofcenter and surround make the spatial propertiesof receptive fields depend on the temporal fre-quency of visual stimulation, so too do they affectthe chromatic properties. At high temporal fre-quencies the phase difference between centerand surround signals is reduced, so the mecha-nisms tend to act synergistically rather thanantagonistically. In P cells this leads to some lossof chromatic opponency (Gouras and Zrenner,1979; Derrington et al, 1984; Smith et al., 1992).However, up to frequencies of 20 Hz or more theeffect is small (Derrington et al., 1984; Lee et al.,1990) – too small to contribute significantly to therapid decline in sensitivity to chromatic flickerfound psychophysically (Wisowaty and Boynton,1980), which must therefore originate in cortex.

6.4.3.1 Chromatic adaptationThe two-color adaptation methods developed byStiles (1939) to isolate chromatic mechanismspsychophysically have been used to isolate thechromatic mechanisms in ganglion cells andLGN cells (Wiesel and Hubel, 1966; Gouras,1968), but have not been employed to exploredetails of chromatic adaptation. We know that achange in mean illumination causes a ganglioncell’s sensitivity to decline approximately in pro-portion to the level of illumination (Purpuraet al., 1990; Lee et al., 1990). The mechanismalmost certainly lies before the ganglion cellsthemselves, for the gain change proceeds inde-pendently in rod and cone pathways that con-verge on the same ganglion cells. Less attentionhas been paid to the changes in ganglion cellbehavior that result from changing the mean


235

Figure 6.15 Distribution of weights that P cells inLGN attach to inputs from the different classes ofcones.The weights attached to signals from L and Mcones are represented explicitly; cells that receiveinputs from only L and M cones are represented bypoints that lie on the unit diagonals; cells that receiveinputs from S cones are represented by points insidethe diagonals.The unsigned S cone weight can be readfrom the internal scale. (From Derrington et al., 1984.)

R cone weight

G c

one

wei

ght

1�1 �0.5

�0.5

�1

1

0.5

0

0.25

0.75

0.5

chromaticity of illumination. DePriest et al.(1991) found that, at constant luminance, amodest step change in the chromaticity of abackground brought about a long-lasting changein the maintained discharge of a P cell, a step inthe ‘on’ color direction increasing the discharge,a step in the ‘off’ color direction decreasing thedischarge. A step in luminance that provided asimilar change in cone excitation had little effectupon the maintained discharge. Yeh et al. (1996)described similar behavior. These long-lastingchanges in discharge provide cortex with a per-sisting signal about the ambient chromaticity.

In addition to changing the maintained dis-charge, the step change in background color per-turbs the cell’s sensitivity, although differentinvestigations do not fully agree on the nature ofthe change. DePriest et al. (1991) found a rapid,paradoxical, increase in sensitivity to chromaticprobe stimuli following a change in backgroundthat would have been expected to reduce sensi-tivity. Yeh et al. (1996), who used larger back-ground steps, found prolonged disturbances ofsensitivity, possibly reflecting saturation of anopponent site. These experiments point to possi-bly complex mechanisms of sensitivity regula-tion involving sites at which signals from thedifferent classes of cones have converged.

6.4.4 CANDIDATE CHROMATIC ANDACHROMATIC PATHWAYS

The two different types of chromatically opponentP cells, a ‘red–green’ one and a ‘blue–yellow’one, together with M cells, whose spectral sensi-tivities are close to that of Vk, provide plausiblesubstrates of the three kinds of post-receptoralmechanisms postulated by psychophysicists.

There seems little doubt that the P cells pro-vide signals to two chromatically opponentvisual channels, but the role of M cells in thethird channel is less secure. Modern psy-chophysical work attributes to this mechanismhigh spatial and temporal resolution, and a spec-tral sensitivity close to that of Vk (for a review,see Lennie et al., 1993). The temporal resolvingpower of M cells (and also of P cells) consider-ably exceeds that measured psychophysically,but the spatial resolving power of the mosaic ofM cells is demonstrably too low to explainpsychophysical performance (Lennie, 1993).

Moreover, having a Vk-like spectral sensitivitydoes not necessarily implicate M cells as the thirdpost-receptoral mechanism. Lennie et al. (1993)argue that although M cells are a plausible sub-strate of luminous efficiency functions measuredwith heterochromatic flicker photometry, similarfunctions can be obtained from linear transfor-mations of the signals carried by P cells, and thisprovides a more probable account of luminousefficiency functions derived by other means,such as acuity measurements.

The above considerations encourage one toexplore the possibility that the P pathway carriessignals for all three chromatic dimensions ofvision. Several observations, puzzling whenconsidered in isolation, become more intelligiblein that context. First, around 90% of P cells(about 80% of all ganglion cells) are of the‘red–green’ type that receive opponent signalsfrom L and M cones – many more than areneeded to account for visual acuity for coloredobjects, but the right number to explain visualacuity for achromatic objects. Second, the center–surround organization of their receptive fieldsensures that they respond well to chromaticchanges at low spatial frequencies and to achro-matic stimuli at high spatial frequencies. Simplyput, a single P cell is equally capable of convey-ing information about the chromatic and achro-matic content of the image, albeit in differentfrequency bands. Its response is ambiguous, butthis ambiguity can be resolved by comparing,at some higher level, signals from differentneurons (Lennie and D’Zmura, 1988). This kindof account leaves M cells little role in colorvision, save possibly in heterochromatic flickermatches. Lee (1996) attributes to M cells alarger role in object vision.

6.5 CORTEX

6.5.1 STRUCTURAL ISSUES

In primates, V1 is the only cortical area thatreceives projections from LGN. Those from mag-nocellular and parvocellular layers arrive in dif-ferent anatomical subdivisions of layer 4: M cellsproject principally to layer 4Ca, and P cells proj-ect mainly to layers 4Cb and 4A (see Figure 6.6).The incoming fibers from the LGN are segregated


236

by eye of origin into strips about 0.5 mm wide.These strips, known as ocular dominance columns,can be readily identified anatomically by exam-ining the uptake in cortex of the radioactiveamino acid proline injected into one eye. This istransported to cortex and deposited in layer 4(Figure 6.16). This anatomical segregation ofsignals from the two eyes is not well maintainedin layers above and below layer 4, and oculardominance columns are less sharply defined.

V1 contains a topographically organized mapof half of the visual field, with a large represen-tation of the fovea and a much smaller represen-tation of the periphery. Most of the distortion inthe map can be explained by the variation in thedensity of retinal ganglion cells from fovea toperiphery; each ganglion cell projects, via LGN,to a roughly constant volume of cortex. Thearrangement and size of distinctive anatomicalfeatures such as ocular dominance columns doesnot vary from place to place in the map.

Within the overall pattern established by theocular dominance columns, other repeatingstructures can be discerned in V1. Cortex stainedfor the presence of the enzyme cytochrome oxi-dase displays a regular pattern of patches wherethe enzyme is concentrated (Figure 6.17). These‘blobs,’ which are more prominent above and

below layer 4, lie securely within the separatedomains of left- and right-eye columns.

The existence of repeating structures in V1suggests a modular organization, assembled fromunits of a certain size. It is widely believed thatthe fundamental organizing unit spans the widthof a pair of adjacent strips of input from left andright eyes, and therefore occupies about 1 mm2

of cortical surface and has a depth of about2 mm. This unit, known as a hypercolumn (Hubeland Wiesel, 1974), contains perhaps 200 000cells. Within a hypercolumn there are furtherorganizational regularities that can be discernedphysiologically but not anatomically.

Signals from different subdivisions of layer 4are delivered to different places within V1. Layer4Ca, which receives its input from M cells,makes a distinctive projection to layer 4B, whichsends an equally distinctive projection out ofstriate cortex to extrastriate regions that appearto be particularly concerned with the analysis ofimage movement (Zeki, 1974; Newsome et al.,1985), and which ultimately send major projec-tions to the parietal lobe. Neurons in layers 4Cband 4A, which receive input from P cells, projectprincipally to the upper layers of striate cortex.


237

Figure 6.16 The organization of ocular dominancecolumns in striate cortex of the macaque monkey.The cortex here has been unfolded as a flat sheet.Radioactive tracer injected into one eye is depositedin striate cortex, in regions to which that eye projects.This results in the distinctive pattern of stripes. (FromHubel and Wiesel, 1977.)

5 mm

Figure 6.17 Distribution of cytochrome oxidase‘blobs’ in the unfolded striate cortex of the macaquemonkey.The blobs are spaced almost uniformlythroughout striate cortex, and fall within the separateocular dominance columns that define the territoriesdominated by the two eyes. (From Livingstone andHubel, 1984.)

1 cm

Neurons in the upper layers of striate cortexsend projections principally to area V2.

6.5.2 FUNCTIONAL ORGANIZATION

6.5.2.1 Striate cortexThe earliest studies showed that receptive fieldsof V1 neurons often differ profoundly from thoseof neurons in retina and LGN. The most promi-nent differences lay in the spatial organizationof receptive fields, which instead of being con-centrically organized, were often elongated,sometimes having no distinct excitatory andinhibitory regions (Hubel and Wiesel, 1977). Theresult is that most neurons in V1 are selective forthe orientation of visual stimuli that fall withintheir receptive fields; they are also often sharplyselective for the size of the stimulus (Figure6.18), and for its direction of movement. Theuniformity of receptive field properties in retinaand LGN thus gives way to great heterogeneityin cortex.

Hubel and Wiesel (1962, 1968) drew a broaddistinction, generally upheld by subsequentwork, between two major kinds of cortical cell

that they called simple and complex. These havesimilar orientation, spatial, and directional selec-tivities, but differ sharply in the form of theirresponses. A simple cell generates a responsethat reflects the quasi-linear addition of signals(excitatory or inhibitory) arising in differentparts of the receptive field. A map of the excita-tory and inhibitory regions in a simple receptivefield provides a reasonable guide to the visualselectivity of the cell. A complex cell is inher-ently nonlinear: the receptive field generallycannot be parsed into distinct excitatory andinhibitory regions, and the cell will respondwith an increased discharge to either a localizedincrease or decrease in illumination. The behav-ior of the complex cell represents an importantchange in the way the visual system analyzes theimage, for, unlike the simple cell or a neuronat an earlier stage in the pathway, it gives aresponse from which the image cannot be recon-structed. It is unclear whether or not complexcells derive their inputs from simple cells, orreceive the same inputs as simple cells.

Cells with different properties tend to be con-centrated at different depths within cortex.Complex cells are most often found in layers2/3 and in layer V. Simple cells are found inlayers 2/3 and in layer 4. Neurons with con-centrically organized receptive fields are foundin layer 4, where inputs from LGN arrive. Aboveand below layer 4, neurons can often be excitedby stimulation through either eye, althoughrarely are both eyes equally effective. If a cell canbe driven binocularly, it is often sensitive to therelative position of the stimuli presented to thetwo eyes, responding well when the stimuli fallon corresponding points, much less well other-wise. This makes neurons sensitive to binoculardepth (Poggio and Talbot, 1981).

Neurons with different visual preferences areplaced in an orderly arrangement in cortex. Fromany point on the cortical surface, extendingperpendicularly through the depth of the cortex,neurons prefer stimuli of the same orientation.On an adjacent, parallel, trajectory neuronsprefer a slightly different orientation. A set of‘orientation columns,’ covering the range oforientations around the circle, is containedwithin the width of a pair of ocular dominancecolumns (a hypercolumn; Hubel and Wiesel,1977).


238

Figure 6.18 Diagram of the receptive fieldorganization often found in cells in striate cortex.Crosses represent excitatory regions giving ‘on’responses; triangles represent inhibitory regions giving‘off ’ responses. (A) and (B) illustrate the arrangementof regions in concentrically-organized receptive fieldstypical of ganglion cells and LGN cells, but relativelyrare in cortex. (C–G) represent arrangements ofexcitatory and inhibitory regions in different simplecells. (From Hubel and Wiesel, 1962.)

(A)

(B)

(E)

(C)

(F)

(D)

(G)

Beyond layer 4 of striate cortex the distinctidentities of the incoming M and P pathways arenot well preserved. Neurons in layer 4B tend tohave higher contrast sensitivities than neuronselsewhere (Hawken and Parker, 1984), presum-ably reflecting their close association with inputsfrom the sensitive M cells, but neurons in otherlayers give few explicit indications of the sourcesof their driving signals.

Neurons in striate cortex respond best tocontrast modulations at temporal frequenciesdistributed around 12 Hz, a lower and morevariable frequency than characterizes neurons inLGN (Hawken et al., 1996), but one that bettermatches psychophysical measurements. As tem-poral frequency is lowered below the peak, thesensitivity of a cortical neuron often declinessharply.

6.5.2.2 Extrastriate cortexThe visual characteristics of neurons in regionsbeyond striate cortex have been less thoroughlystudied, but some general characteristics are wellknown. Among the collection of visual areas thatlead toward the temporal lobe, and are the mostlikely to be important for color vision, receptivefields of neurons grow progressively larger asone moves towards the temporal lobe, andnearly all neurons can be driven by inputsthrough either eye, but the general properties ofreceptive fields – spatial selectivity, orientation,and direction selectivity – change little from areato area. There is some physiological evidencethat different areas are specialized for differentkinds of analysis; this issue is taken up in a latersection.

6.5.3 CHROMATIC PROPERTIES

In exploring the properties of cortical neurons,investigators have concentrated on aspects ofcolor vision that cannot be easily explained bythe known properties of neurons at earlier stagesin the pathway.

6.5.3.1 Separation of chromatic andachromatic signals

It was noted earlier that P cells driven by signalsfrom L and M cones must be the substrate of theachromatic pathway that supports visual acuity,and presumably also a red–green opponent

pathway, but that in any individual cell the chro-matic and achromatic signals are confounded. Incortex one might expect to find mechanismswhose behavior better reflects what is observedpsychophysically, namely substantial independ-ence of chromatic and achromatic mechanisms.

There is a profound change in the general chro-matic characteristics of cells as one moves fromLGN to cortex. The ubiquitous color-opponentP cells in LGN give way to a dearth of overtlycolor-opponent neurons in V1. Hubel and Wiesel(1968) commented on this in their first investi-gation of primate visual cortex, and it has beenconfirmed regularly since (Gouras, 1974).

Neurons with receptive fields in and near thefovea often have sharply defined preferences forstimuli containing relatively high spatial fre-quencies – high enough that chromatic aberra-tion removes much of the chromatic contrastfrom the image. This preference for high spatialfrequencies undoubtedly accounts for the weakresponses of many neurons to stimuli containingspatio-chromatic contrast, and as a result, mostinvestigations of the chromatic properties ofneurons concentrate on cells that prefer stimulicontaining low spatial frequencies.

Overtly color-opponent neurons are found inall layers of striate cortex, but more often inlayer 4 than elsewhere. In layer 4 and layer 6particularly, the most responsive neurons often,but by no means always, have receptive fieldswith poorly defined orientation selectivity, andlow-pass spatial frequency tuning (Lennie et al.,1990). These receptive fields are reminiscent ofreceptive fields in LGN, with the followingimportant difference: their chromatic character-istics depend little upon the spatial configurationof the stimulus. Cortical neurons therefore donot confound different dimensions of stimulusvariation in the way that LGN neurons do.

In the upper layers of cortex, where simpleand complex cells predominate, few neurons areovertly color-opponent. Most of those that canbe excited by colored patterns can also be excitedby patterns defined by brightness contrast; cellsthat respond best to isoluminant chromatic con-trasts are rare (Gouras and Krüger, 1979; Thorellet al., 1984; Lennie et al., 1990). When a neuron’sreceptive field properties can be established withstimuli defined by either color or brightness con-trast, the spatial and orientational selectivities


239

are generally similar (Thorell et al., 1984; Lennieet al., 1990).

The general picture to emerge from studies ofstriate cortex is that for the first time there appearin large numbers neurons whose receptive fieldproperties are indifferent to the chromatic com-position of the stimulus – they may be conceivedas analyzing spatial structure with little regard forthe spectral composition of the stimulus. Most ofthese neurons respond best when stimulus pat-terns are defined by brightness rather than colorcontrast. In this sense cortex might be viewedas having formed an achromatic pathway. Thestanding of independent chromatic pathways isless clear.

Neurons that respond preferentially to stimulidefined by color rather than brightness contrastare also found in areas V2, V3, and V4, and inparts of the temporal lobe.

6.5.3.2 Number of chromaticchannels

Psychophysical evidence reviewed in Chapter 3grants a special status to two chromatic axesin color space (loosely a red–green axis and ayellow–blue one), but at the same time that evi-dence is not consistent on the locus of thesespecial (cardinal) directions. Moreover, studiesthat have examined aftereffects of habituating tomodulations of chromaticity along differentdirections in color space (Krauskopf et al., 1986;Webster and Mollon, 1991) reveal more mecha-nisms than just the two tuned to the cardinaldirections – mechanisms tuned to intermediatedirections exist too.

P cells in LGN fall neatly into just two chro-matic classes. Moreover, their sensitivities arenot changed by prolonged exposure to habitu-ating stimuli of the kind used to reveal mul-tiple chromatic mechanisms in psychophysicalexperiments. These multiple mechanisms mustoriginate in cortex.

Several studies have examined the distribu-tion of chromatic preferences among cells instriate cortex. Although they do not agree ondetails, these investigations do agree that chro-matic preferences are not clustered in twogroups. Vautin and Dow (1985), in recordingsmade from awake monkeys, found that, whenchromatic preferences were explored withmonochromatic lights, neurons in layer 4 fell

into four loosely defined clusters (blue, green,yellow, red), identified by their wavelengths ofpeak excitability. Thorell et al. (1984) and Lennieet al. (1990) found a similar modest tendency forcells’ chromatic preferences to be aligned alongthe red–green and yellow–blue axes (properly anaxis of exclusively L and M cone modulation andone of exclusively S cone modulation) of colorspace. In the upper layers of cortex, chromaticpreferences are broadly distributed.

To the extent that cells’ chromatic preferencesare broadly distributed, neurons in striate cortexare a candidate substrate for the multiple chro-matic mechanisms inferred from psychophysicalexperiments. However, it is hard to tell fromphysiological observations which of the neuronsthat respond to colored stimuli are important forcolor vision – that is, which neurons might havea role in perceptual judgments about color.Attempts to identify the relevant neurons havegenerally sought to demonstrate physiologicalproperties that mirrored those of the psy-chophysical mechanisms. One such attempt, toexplore habituation to chromatically modulatedstimuli, showed that although some corticalneurons habituate (and neurons in LGN neverdo), others become more responsive followingprolonged stimulation (Lennie et al., 1994).Another approach has been to examine whetheror not color-opponent neurons are concentratedin particular places – pathways specialized forthe analysis of color.

6.5.3.3 Spatial contrast effectsSome of the most distinctive phenomena of colorvision – for example, the uniform color inducedin a neutral patch by enclosing it in a coloredregion – depend upon spatial interactionsbetween nearby regions of visual field. Theseremote influences have a longer reach than canplausibly be associated with any mechanisms inretina or LGN, and have generally been thoughtto originate in cortex. In area V1, long-range con-trast effects have been observed in the domainsof orientation and direction of movement(Blakemore and Tobin, 1972; Knierim and VanEssen, 1992; Lamme, 1995), but not consistentlyin the domain of color. In their first discussion ofneurons in striate cortex, Hubel and Wiesel (1968)described a rare ‘double-opponent’ receptivefield, in which a core region with co-extensive


240

color-opponent mechanisms (for example, red-on, green-off) was enclosed by an outer regioncontaining color-opponent mechanisms of theopposite sense. Neurons with such propertiescould perhaps account for induced color phe-nomena (Daw, 1984), but they are rare, andtheir receptive fields are small. Livingstone andHubel (1984) and Ts’o and Gilbert (1988) laterstudied neurons, perhaps differing from thoseexplored by Hubel and Wiesel, in which theregion enclosing the core of the receptive fieldwas not itself color-opponent, but always sup-pressed the response to a stimulus applied to thecore region. It is unclear what such cells mightdo for color vision.

In seeking physiological accounts of colorinduction, physiologists have looked most care-fully at area V4 (the fourth visual cortical area).Zeki (1973, 1977) discovered that this containsan unusually large proportion of cells with sharpchromatic selectivities. Moreover, Zeki (1983)later found that the response of a V4 cell to a col-ored stimulus in the middle of its receptive fielddepended on the color of light falling in sur-rounding regions, in a manner that was corre-lated with the colored appearance the stimulusto a human observer. Schein and Desimone(1990) later described a mechanism that mightbe responsible for this behavior: the receptivefield of a V4 cell is enclosed by a region that hasthe same spectral characteristics as the receptivefield proper, but when illuminated always sup-presses the response to a stimulus falling withinthe receptive field.

6.5.3.4 Private pathways for colorTwo kinds of evidence bear on the question ofwhether there exist cortical pathways specializedfor color. The first comes from physiologicalstudies that have looked for special concentra-tions of color-opponent cells in different parts ofcortex. The second comes from studies that haveattempted to localize color centers in humancortex.

All investigators seem to agree that area V4contains neurons with interesting chromaticproperties, but there is much less agreement onwhether V4 is an area specialized for analysis ofthe chromatic attributes of objects. V4 is theprincipal conduit of information from lowervisual cortical regions to the temporal lobe, a

region crucially involved in all aspects of objectvision, not just the analysis of color (Heywoodand Cowey, 1987; Heywood et al., 1992). Ittherefore seems unlikely that V4 could bedevoted exclusively to the analysis of color.Nevertheless, subdivisions of V4 might havedifferently specialized functions, undertakinganalyses of different attributes of the image; oneof these might be the analysis of color.

This issue has been explored in studies thathave traced pathways projecting from earlierstages of visual cortex to V4. Most of the workthat has attempted to trace a specialized colorpathway has concentrated on the projections thatoriginate in the cytochrome oxidase ‘blobs’ in V1.Livingstone and Hubel (1984) first drew attentionto properties of cells in blobs, finding that theytended to have concentrically organized receptivefields, more than half being color-opponent – amuch higher proportion than is typically foundelsewhere in V1. Neurons in the blobs project tothe second visual area, V2. When stained forcytochrome oxidase, V2 shows a pattern not ofblobs, but of three alternating stripes, often calledthick, thin, and pale (Tootell et al., 1983). Theprojections from blobs in V1 end preferentially inthe thin stripes (Livingstone and Hubel, 1983).Neurons in the thin stripes and the pale stripesproject to V4 (Shipp and Zeki, 1985; DeYoe andVan Essen, 1985), where their terminations seemto be segregated (DeYoe et al., 1994).

Physiological explorations of the pathwaythat originates in blobs are equivocal on thequestion of it being specialized for color. Ts’oand Gilbert (1988) corroborated Livingstoneand Hubel’s (1984) finding of a concentration ofcolor-opponent cells in blobs, but Lennie et al.(1990) and Leventhal et al. (1995) found noconcentrations of this kind. In area V2, Hubeland Livingstone (1987) described a concentra-tion of color-opponent cells in the thin stripesto which blobs project. Later studies in which cellswere characterized quantitatively (Levitt et al.,1994; Gegenfurtner et al., 1996) have foundonly a slight tendency for cells with differentproperties to be clustered in different stripes.

Although work on monkeys provides no firmpointers to the existence of a specialized colorpathway, studies of people with cortical lesions(usually resulting from stroke) that cause selec-tive impairment of color vision, without (or with


241

little) concomitant impairment of other dimen-sions of vision, suggest the existence of a corticalregion specialized for the analysis of color (Zeki,1990). However, it is often unclear from thesestudies whether the impairment affects thecapacity to identify colors, or the capacity to dis-tinguish surfaces of different color. Where thisquestion has been examined (Mollon et al., 1980;Victor et al., 1989; Barbur et al., 1994), achro-matopsic subjects often are able to use chro-maticity to segment surfaces, and sometimeshave near normal hue discrimination, but aremuch impaired in naming colors and grouping ofitems of similar color.

Recently developed methods can identify incortex the local changes in blood flow and bloodvolume associated with increased neural activity.Two of these kinds of measurement, positronemission tomography (PET) and functional mag-netic resonance imaging (fMRI), have been usedto investigate whether or not mechanisms ofcolor vision are localized in particular corticalregions (Zeki et al., 1991; Corbetta et al., 1991).The results show that a region in the fusiformgyrus, which lies between the occipital and tem-poral lobes, seems to be unusually active duringthe perception of colored patterns, though thedegree to which it is specialized for the analysisof color has not been fully explored.

6.6 ACKNOWLEDGMENTS

This work was supported by NIH grants EY04440and EY01319.

6.7 NOTES

1 Other terms are sometimes used. The cell typeswere originally distinguished by Polyak (1941),who called them ‘midget’ and ‘parasol’ cells. Perryet al. (1984) have called them Pb and Pa cells.

6.8 REFERENCES

Ahnelt, P. and Kolb, H. (1994) Horizontal cells andcone photoreceptors in primate retina: a Golgi-lightmicroscope study of spectral connectivity. Journal ofComparative Neurology, 343, 387–405.

Barbur, J.L., Harlow, A.J., and Plant, G.T. (1994)

Insights into the different exploits of colour in thevisual cortex. Proceedings of the Royal Society of LondonB, 258, 327–34.

Baylor, D.A. (1987) Photoreceptor signals and vision.Investigative Ophthalmology and Visual Science, 28,34–49.

Baylor, D.A., Nunn, B.J., and Schnapf, J.L. (1987) Spec-tral sensitivity of cones of the monkey Macaca fascicu-laris. Journal of Physiology (London), 390, 145–60.

Blakemore, C. and Tobin, E.A. (1972) Lateral inhibi-tion between orientation detectors in the cat’s visualcortex. Experimental Brain Research, 15, 439–40.

Blakemore, C. and Vital-Durand, F. (1986) Effects ofvisual deprivation on the development of the mon-key’s lateral geniculate nucleus. Journal of Physiology(London), 380, 493–512.

Boycott, B.B. and Dowling, J.E. (1969) Organizationof the primate retina: light microscopy. PhilosophicalTransactions of the Royal Society of London B, 255,109–94.

Boycott, B.B. and Wässle, H. (1991) Morphologicalclassification of bipolar cells of the primate retina.European Journal of Neuroscience, 3, 1069–88.

Boynton, R.M. and Baron, W.S. (1975) Sinusoidalflicker characteristics of primate cones in responseto heterochromatic stimuli. Journal of the OpticalSociety of America, 65, 1091–100.

Calkins, D.J., Schein, S.J., Tsukamoto, Y., and Sterling,P. (1994) M and L cones in macaque fovea connectto midget ganglion cells by different numbers ofexcitatory synapses. Nature, 371, 70–2.

Calkins, D.J. and Sterling, P. (1996) Absence of spec-trally specific lateral inputs to midget ganglion cellsin primate. Nature, 381, 613–15.

Corbetta, M., Miezin, F.M., Dobmeyer, S., Schulman,G.L., and Petersen, S.E. (1991) Selective and dividedattention during visual discriminations of shape,color and speed: functional anatomy by positronemission tomography. Journal of Neuroscience, 11,2383–402.

Dacey, D.M. (1993) Morphology of a small-field bi-stratified ganglion cell type in the macaque andhumn retina. Visual Neuroscience, 10, 1081–98.

Dacey, D.M. (1994) Physiology, morphology and spa-tial densities of identified ganglion cell types in pri-mate retina. In Higher-order Processing in the VisualSystem. Chichester: Wiley, pp. 12–28.

Dacey, D.M. (1996) Circuitry for color coding in theprimate retina. Proceedings of the National Academy ofSciences (USA), 93, 582–8.

Dacey, D.M. and Lee, B.B. (1994) The blue-on oppo-nent pathway in the primate retina originates froma distinct bistratified ganglion cell. Nature, 367,731–5.

Dacey, D.M. and Lee, B.B. (1999) Functional architec-ture of cone signal pathways in the primate retina.In K.R. Gegenfurtner and L.T. Sharpe (eds), ColorVision: From Genes to Perception. Cambridge:Cambridge University Press, pp. 181–202.

Dacey, D.M., Lee, B.B., Stafford, D.K., Pokorny, J.,and Smith, V.C. (1996) Horizontal cells of the


242

primate retina: cone specificity without spectralopponency. Science, 271, 656–9.

Dacey, D., Packer, O.S., Diller, L., Brainard, D.,Peterson, B., and Lee, B. (2000) Center surroundreceptive field structure of cone bipolar cells in pri-mate retina. Vision Research, 40, 1801–11.

Daw, N.W. (1984) The psychology and physiology ofcolour vision. Trends in Neuroscience, 7, 330–35.

de Monasterio, F.M. and Gouras, P. (1975) Functionalproperties of ganglion cells of the rhesus monkeyretina. Journal of Physiology (London), 251, 167–95.

De Valois, R.L., Abramov, I., and Jacobs, G.H. (1966)Analysis of response patterns of LGN cells. Journal ofthe Optical Society of America, 56, 966–77.

DePriest, D.D., Lennie, P., and Krauskopf, J. (1991)Slow mechanisms of chromatic adaptation inmacaque. Investigative Ophthalmology and VisualScience, 32 (Suppl.), 1252.

Derrington, A.M., Krauskopf, J., and Lennie, P.(1984) Chromatic mechanisms in lateral geniculatenucleus of macaque. Journal of Physiology (London),357, 241–65.

Derrington, A.M. and Lennie, P. (1984) Spatial andtemporal contrast senstivities of neurones in lateralgeniculate nucleus of macaque. Journal of Physiology(London), 357, 219–40.

DeYoe, E.A., Felleman, D.J., Van Essen, D.C., andMcClendon, E. (1994) Multiple processing streamsin occipitotemporal visual cortex. Nature, 371,151–4.

DeYoe, E.A. and Van Essen, D.C. (1985) Segregation ofefferent connections and receptive field properties invisual area V2 of the macaque. Nature, 317, 58–61.

Dowling, J.E. and Boycott, B.B. (1966) Organizationof the primate retina: electron microscopy. Pro-ceedings of the Royal Society of London B, 166, 80–111.

Dreher, B., Fukuda, Y., and Rodieck, R.W. (1976)Identification, classification and anatomical segrega-tion of cells with X–like and Y–like properties in thelateral geniculate nucleus of old-world primates.Journal of Physiology (London), 258, 433–52.

Engel, S., Zhang, X., and Wandell, B. (1997) Colortuning in human visual cortex measured with func-tional magnetic resonance imaging. Nature, 388,68–71.

Enroth-Cugell, C. and Robson, J.G. (1966) The con-trast sensitivity of retinal ganglion cells of the cat.Journal of Physiology (London), 187, 517–52.

Estévez, O. and Spekreijse, H. (1974) A spectral com-pensation method for determining the flicker char-acteristics of the human colour mechanisms. VisionResearch, 14, 823–30.

Famiglietti, E.V., Jr and Kolb, H. (1976) Structuralbasis for on- and off-center responses in retinal gan-glion cells. Science, 194, 193–5.

Felleman, D.J. and Van Essen, D.C. (1991) Distributedhierarchical processing in the primate cerebral cor-tex. Cerebral Cortex, 1, 1–47.

Gegenfurtner, K.R., Kiper, D.C. and Fenstemaker, S.B.(1996) Processing of color, form, and motion inmacaque area V2. Visual Neuroscience, 13, 161–72.

Goodchild, A.K., Chan, T.L. and Grünert, U. (1996)Horizontal cell connections with short wavelengthsensitive cones in macaque monkey retina. VisualNeuroscience, 13, 833–45.

Gouras, P. (1968) Identification of cone mechanismsin monkey ganglion cells. Journal of Physiology(London), 199, 533–47.

Gouras, P. (1974) Opponent-colour cells in differentlayers of foveal striate cortex. Journal of Physiology(London), 238, 583–602.

Gouras, P. and Krüger, J. (1979) Response of cells infoveal visual cortex of the monkey to pure colorcontrast. Journal of Neurophysiology, 42, 850–60.

Gouras, P. and Zrenner, E. (1979) Enhancement ofluminance flicker by color-opponent mechanisms.Science, 205, 587–9.

Grünert, U., Greferath, U., Boycott, B.B., and Wässle,H. (1993) Parasol (Pa) ganglion-cells of the primatefovea: immunocytochemical staining with antibodiesagainst GABAA-receptors. Vision Research, 33, 1–14.

Hawken, M.J. and Parker, A.J. (1984) Contrast sensi-tivity and orientation selectivity in lamina IV of thestriate cortex of old world monkeys. ExperimentalBrain Research, 54, 367–72.

Hawken, M.J., Shapley, R.M., and Grosof, D.H. (1996)Temporal-frequency selectivity in monkey visualcortex. Visual Neuroscience, 13, 477–92.

Heywood, C.A. and Cowey, A. (1987) On the role ofcortical visual area V4 in the discrimination ofhue and pattern in macaque monkeys. Journal ofNeuroscience, 7, 2601–16.

Heywood, C.A., Gadotti, A., and Cowey, A. (1992)Cortical area V4 and its role in the perception ofcolor. Journal of Neuroscience, 12, 4056–65.

Hood, D.C. and Birch, D.G. (1993) Human conereceptor activity: the leading edge of the a-wave andmodels of receptor activity. Visual Neuroscience, 10,857–71.

Hubel, D.H. and Livingstone, M.S. (1987) Segregationof form color and stereopsis in primate area 18.Journal of Neuroscience, 7, 3378–415.

Hubel, D.H. and Wiesel, T.N. (1962) Receptive fields,binocular interactions, and functional architecturein the cat’s visual cortex. Journal of Physiology(London), 160, 106–54.

Hubel, D.H. and Wiesel, T.N. (1968) Receptive fieldsand functional architecture of monkey striate cor-tex. Journal of Physiology (London), 195, 215–43.

Hubel, D.H. and Wiesel, T.N. (1974) Sequence regu-larity and geometry of orientation columns in themonkey striate cortex. Journal of ComparativeNeurology, 158, 267–94.

Hubel, D.H. and Wiesel, T.N. (1977) The Ferrier lec-ture. Functional architecture of macaque monkeyvisual cortex. Proceedings of the Royal Society of LondonB, 198, 1–59.

Kaiser, P.K., Lee, B.B., Martin, P.R., and Valberg, A.(1990) The physiological basis of the minimally dis-tinct border demonstrated in the ganglion cells ofthe macaque retina. Journal of Physiology (London),422, 153–83.


243

Kaneko, A. (1973) Receptive field organization ofbipolar and amacrine cells in the goldfish retina.Journal of Physiology (London), 235, 133–53.

Kaplan, E. and Shapley, R.M. (1982) X and Y cells inthe lateral geniculate nucleus of the macaque mon-key. Journal of Physiology (London), 330, 125–44.

Kelly, D.H. (1961) Visual responses to time-dependentstimuli. II. Single-channel model of the photopicvisual system. Journal of the Optical Society of America,51, 747–54.

Knierim, J.J. and Van Essen, D.C. (1992) Neuronalresponses to static texture patterns in area V1 of thealert macaque monkey. Journal of Neurophysiology,67, 961–80.

Kolb, H., Fernandez, E., Schouten, J., Ahnelt, P.,Linberg, K.A., and Fisher, S. K. (1994) Are therethree types of horizontal cell in the human retina?Journal of Comparative Neurology, 343, 370–86.

Kouyama, N. and Marshak, D.W. (1992) Bipolar cellsspecific for blue cones in the macaque retina.Journal of Neuroscience, 12, 1233–52.

Krauskopf, J., Williams, D.R., Mandler, M.B., andBrown, A.M. (1986) Higher order color mecha-nisms. Vision Research, 26, 23–32.

Kuffler, S.W. and Nicholls, J.G. (1976) From Neuron toBrain. Sunderland: Sinauer.

Lamme, V.A.F. (1995) The neurophysiology of fig-ure–ground segregation in primary visual cortex.Journal of Neuroscience, 15, 1605–15.

Lankheet, M.J.M., van Wezel, R.J.A., and van deGrind, W.A. (1991) Effects of background illumina-tion on cat horizontal cell responses. Vision Research,31, 919–32.

Lee, B.B. (1996) Receptive field structure in the pri-mate retina. Vision Research, 36, 631–44.

Lee, B.B., Martin, P.R., and Valberg, A. (1989)Nonlinear summation of M– and L–cone inputs tophasic ganglion cells of the macaque. Journal ofNeuroscience, 9, 1433–42.

Lee, B.B., Pokorny, J., Smith, V.C., Martin, P.R., andValberg, A. (1990) Luminance and chromatic mod-ulation sensitivity of macaque ganglion cells andhuman observers. Journal of the Optical Society ofAmerica A, 7, 2223–6.

Lennie, P. (1980) Parallel visual pathways: a review.Vision Research, 20, 561–94.

Lennie, P. (1993) Roles of M and P pathways. InR.M. Shapley and D.M.K. Lam (eds), ConstrastSensitivity.Cambridge, MA: MIT Press, pp. 201–13.

Lennie, P. and D’Zmura, M. (1988) Mechanisms ofcolor vision. CRC Critical Reviews in Neurobiology, 3,333–400.

Lennie, P., Haake, P.W., and Williams, D.R. (1991) Thedesign of chromatically opponent receptive fields.In M.S. Landy and J.A. Movshon (eds), Compu-tational Models of Visual Processing. Cambridge, MA:MIT Press, pp. 71–82.

Lennie, P., Krauskopf, J., and Sclar, G. (1990) Chro-matic mechanisms in striate cortex of macaque.Journal of Neuroscience, 10, 649–69.

Lennie, P., Lankheet, M.J.M., and Krauskopf, J.

(1994) Chromatically selective habituation in mon-key striate cortex. Investigative Ophthalmology andVisual Science (Supplement), 35, 1662.

Lennie, P., Pokorny, J., and Smith, V. C. (1993) Lumi-nance. Journal of the Optical Society of America A, 10,1283–93.

Leventhal, A.G., Thompson, K.G., Liu, D., Zhou, Y.,and Ault, S. J. (1995) Concomitant sensitivity toorientation, direction, and color of cells in layers 2,3, and 4 of monkey striate cortex. Journal ofNeuroscience, 15, 1808–18.

Levitt, J.B., Kiper, D.C., and Movshon, J.A. (1994)Receptive fields and functional architecture ofmacaque V2. Journal of Neurophysiology, 71, 2517–42.

Livingstone, M.S. and Hubel, D.H. (1983) Specificityof cortico-cortical connections in monkey visualsystem. Nature, 304, 531–4.

Livingstone, M.S. and Hubel, D.H. (1984) Anatomyand physiology of a color system in the primatevisual cortex. Journal of Neuroscience, 4, 309–56.

Mariani, A.P. (1984) Bipolar cells in monkey retinaselective for the cones likely to be blue-sensitive.Nature, 308, 184–6.

Masland, R.H. (1988) Amacrine cells. Trends in Neuro-science, 11, 405–10.

McMahon, M.J., Lankheet, M.J.M., Lennie, P., andWilliams, D. (2000) Fine structure of parvocellularreceptive fields in the primate fovea revealed bylaser interferometry. Journal of Neuroscience, 20,2043–53.

Mollon, J.D., Bowmaker, J.K., and Jacobs, G.H.(1984) Variations of colour vision in a New Worldprimate can be explained by a polymorphism ofretinal photopigments. Proceedings of the Royal Societyof London B, 222, 373–99.

Mollon, J.D., Newcombe, F., Polden, P.G. and Ratcliff,G. (1980) On the presence of three cone mechanismsin a case of total achromatopsia. In G. Verriest (ed.),Colour Vision Deficiencies. Bristol: Hilger, pp. 130–5.

Naka, K.I. (1976) Functional organization of the cat-fish retina. Journal of Neurophysiology, 40, 26–43.

Nathans, J., Merbs, S.L., Sung, C.H., Weitz, C.J., andWang, Y. (1992) Molecular genetics of human visualpigments. Annual Review of Genetics, 26, 403–24.

Newsome, W.T., Wurtz, R.H., Dürsteler, M.R., andMikami, A. (1985) Deficits in visual motion pro-cessing following ibotenic acid lesions of the middletemporal visual area. Journal of Neuroscience, 5,825–40.

Perry, V.H., Öhler, R., and Cowey, A. (1984) Retinalganglion cells that project to the dorsal lateral genic-ulate nucleus in the macaque monkey. Neuroscience,12, 1101–23.

Poggio, G.F. and Talbot, W.H. (1981) Mechanisms ofstatic and dynamic stereopsis in foveal cortex of therhesus monkey. Journal of Physiology (London), 315,469–92.

Polyak, S.L. (1941) The Retina. Chicago: University ofChicago Press.

Polyak, S.L. (1957) The Vertebrate Visual System.Chicago: University of Chicago Press.


244

Pugh, E.N., Jr and Lamb, T.D. (1993) Amplificationand kinetics of the activation steps in phototrans-duction. Biochim Biophys Acta, 1141, 111–49.

Purpura, K., Tranchina, D., Kaplan, E., and Shapley,R.M. (1990) Light adaptation in the primate retina:analysis of changes in gain and dynamics of monkeyretinal ganglion cells. Visual Neuroscience, 4, 75–93.

Reid, R.C. and Shapley, R.M. (1992) Spatial structureof cone inputs to receptive fields in primate lateralgeniculate nucleus. Nature, 356, 716–18.

Robson, J.G. (1966) Spatial and temporal contrastsensitivity functions of the visual system. Journal ofthe Optical Society of America, 56, 1141–2.

Rodieck, R.W. (1965) Quantitative analysis of catretinal ganglion cell response to visual stimuli.Vision Research, 5, 583–601.

Rodieck, R.W. (1991) Which cells code for color? InA. Valberg and B.B. Lee (eds), From Pigments toPerception. New York: Plenum Press, pp. 83–94.

Rodieck, R.W. (1998) The First Steps in Seeing.Sunderland: Sinauer Associates.

Rodieck, R.W., Brening, R.K., and Watanabe, M.(1993) The origin of parallel visual pathways. In R.Shapley and D. M.-K. Lam (eds), Contrast Sensitivity.Cambridge, MA: MIT Press, pp. 117–47.

Roe, A.W. and Ts’o, D.Y. (1995) Visual topography inprimate V2: multiple representation across func-tional stripes. Journal of Neuroscience, 15, 3689–715.

Schaeffel, F., Bartmann, M., Hagel, G., and Zrenner, E.(1995) Studies on the role of the retinal dopamine/melatonin system in experimental refractive errorin chickens. Vision Research, 35, 1247–64.

Schein, S.J. and Desimone, R. (1990) Spectral proper-ties of V4 neurons in the macaque. Journal ofNeuroscience, 10, 3369–89.

Schiller, P.H. and Colby, C.L. (1983) The responses ofsingle cells in the lateral geniculate nucleus of therhesus monkey to color and luminance contrast.Vision Research, 23, 1631–41.

Schnapf, J.L., Nunn, B.J., Meister, M., and Baylor,D.A. (1990) Visual transduction in cones of themonkey Macaca fascicularis. Journal of Physiology(London), 427, 681–713.

Sherman, S.M. (1996) Dual response modes in lateralgeniculate neurons: mechanisms and functions.Visual Neuroscience, 13, 205–13.

Shipp, S. and Zeki, S.M. (1985) Segregation of path-ways leading from area V2 to areas V4 and V5 ofmacaque monkey visual cortex. Nature, 315, 322–5.

Smith, V.C., Lee, B.B., Pokorny, J., Martin, P.R., andValberg, A. (1992) Responses of macaque ganglioncells to the relative phase of heterochromaticallymodulated lights. Journal of Physiology (London), 458,191–221.

Sparks, D.L. and Nelson, J.S. (1987) Sensory andmotor maps in the mammalian superior colliculus.Trends in Neuroscience, 10, 312–17.

Stell, W.K., Ishida, A.T., and Lightfoot, D.O. (1977)Structural basis for on- and off-center responses inretinal bipolar cells. Science, 198, 1269–71.

Sterling, P., Freed, M.A., and Smith, R.G. (1988)

Architecture of rod and cone circuits to the ON-betaganglion cell. Journal of Neuroscience, 8, 623–42.

Stiles, W.S. (1939) The directional sensitivity of theretina and the spectral sensitivities of the rods andcones. Proceedings of the Royal Society of London B, 127,64–105.

Stiles, W.S. and Burch, J.M. (1959) NPL colour-matching investigation: final report. Optica Acta, 6,1–26.

Thorell, L.G., De Valois, R.L. and Albrecht, D.G. (1984)Spatial mapping of monkey V1 cells with pure colorand luminance stimuli. Vision Research, 24, 751–69.

Tootell, R.B.H., Silverman, M.S., De Valois, R.L., andJacobs, G.H. (1983) Functional organization of thesecond cortical visual area (V2) in the primate.Science, 220, 737–9.

Torre, V., Ashmore, J.F., Lamb, T.D., and Menini, A.(1995) Transduction and adaptation in sensoryreceptor cells. Journal of Neuroscience, 15, 7757–68.

Ts’o, D.Y. and Gilbert, C.D. (1988) The organization ofchromatic and spatial interactions in the primatestriate cortex. Journal of Neuroscience, 8, 1712–27.

Valberg, A., Lee, B.B., and Tigwell, D.A. (1986)Neurones with strong inhibitory S-cone inputs inthe macaque lateral geniculate nucleus. VisionResearch, 26, 1061–4.

Valeton, M.J. and van Norren, D. (1983) Light adap-tation of primate cones: an analysis based on extra-cellular data. Vision Research, 23, 1539–47.

Vautin, R.G. and Dow, B.M. (1985) Color cell groupsin foveal striate cortex of the behaving macaque.Journal of Neurophysiology, 54, 273–92.

Verweij, J., Dacey, D.M., Peterson, B.B., and Buck,S.L. (1999) Sensitivity and dynamics of rod signalsin H1 horizontal cells of the macaque monkeyretina. Vision Research, 39, 3662–72.

Victor, J.D., Maiese, K., Shapley, R., Sidtis, J., andGazzaniga, M.S. (1989) Acquired central dyschro-matopsia: analysis of a case with preservation of colordiscrimination. Clinical Vision Sciences, 4, 183–96.

Walraven, J., Enroth-Cugell, C., Hood, D.C.,MacLeod, D.I.A., and Schnapf, J. (1989) The controlof visual sensitivity: receptoral and postreceptoralprocesses. In L. Spillmann and J. Werner (eds),Visual Perception: The Neurophysiological Foundations.New York: Academic Press, pp. 53–101.

Wässle, H. and Boycott, B.B. (1991) Functional archi-tecture of the mammalian retina. PhysiologicalReviews, 71, 447–80.

Wässle, H., Boycott, B.B., and Röhrenbeck, J. (1989)Horizontal cells in the monkey retina: cone connec-tions and dendritic network. European Journal ofNeuroscience, 1, 421–35.

Wässle, H., Grünert, U., Martin, P.R., and Boycott,B.B. (1994) Immunocytochemical characterizationand spatial distribution of midget bipolar cells inthe macaque monkey retina. Vision Research, 34,561–79.

Watanabe, M. and Rodieck, R.W. (1989) Parasol andmidget ganglion cells of the primate retina. Journalof Comparative Neurology, 289, 434–54.


245

Webster, M.A. and Mollon, J.D. (1991) Changes incolour appearance following post-receptoral adap-tation. Nature, 349, 235–8.

Wiesel, T.N. and Hubel, D.H. (1966) Spatial and chro-matic interactions in the lateral geniculate body ofthe rhesus monkey. Journal of Neurophysiology, 29,1115–56.

Williams, D.R. and Collier, R.J. (1983) Consequencesof spatial sampling by a human photoreceptormosaic. Science, 221, 385–7.

Wisowaty, J.J. and Boynton, R.M. (1980) Temporalmodulation sensitivity of the blue mechanism:measurements made without chromatic adaptation.Vision Research, 20, 895–909.

Yeh, T., Lee, B.B., and Kremers, J. (1996) The timecourse of adaptation in macaque retinal ganglioncells. Vision Research, 36, 913–31.

Zeki, S.M. (1973) Colour coding in rhesus monkeyprestriate cortex. Brain Research, 53, 422–7.

Zeki, S.M. (1974) Functional organization of a visualarea in the posterior bank of the superior temporalsulcus of the rhesus monkey. Journal of Physiology(London), 236, 549–73.

Zeki, S.M. (1977) Colour coding in the superiortemporal sulcus of rhesus monkey visual cortex.Proceedings of the Royal Society of London B, 197,195–223.

Zeki, S.M. (1983) Colour coding in the cerebral cortex:the responses of wavelength-selective and colour-coded cells in monkey visual cortex to changes inwavelength composition. Neuroscience, 9, 767–81.

Zeki, S.M. (1990) A century of cerebral achromatop-sia. Brain, 113, 1721–77.

Zeki, S.M., Watson, J.D.G., Lueck, C.J., Friston, K.J.,Kennard, C., and Frackowiak, R.S.J. (1991) A directdemonstration of functional specialization inhuman visual cortex. Journal of Neuroscience, 11,641–9.


246

6 The Physiology of Color Vision - University of Illinois at Chicago · 2016-01-13 · 6.1...

Documents

Transcript of 6 The Physiology of Color Vision - University of Illinois at Chicago · 2016-01-13 · 6.1...