RETINAL PROCESSING of VISUAL DATA* Generated in the Retina. The

RETINAL PROCESSING OF VISUAL DATA* BY EDWARD F. MACNICHOL, JR.

DEPARTMENT OF BIOPHYSICS, THE JOHNS HO6KINS UNIVERSITY, BALTIMORE, MARYLAND The experiments of the psychophysicists have shown with great precision what the human visual system is capable of doing, as Dr.. Judd has so ably summarized in this symposium. However, I am sure that none of us will be satisfied until we learn in detail how the seeming miracle of vision is accomplished. Throughout recorded history, man has sought to explain the workings of the eye, and in the last 150 years, progress has been made ht an ever-accelerating pace. The visual process of course starts with the focusing of a picture of the'outside world on the retina. By the end of the 19th century the main outlines of image formation in the eye were well understood, and at the present time the details are almost entirely filled in. However, the functioning of the retina has not been nearly as easy to elucidate. This year is the hundredth anniversary of the use of objective methods of studying retinal function, for it was in 1865 that Holmgren' reported that there is an electrical response to illumination of the eye and was able to show that the response is generated in the retina. The study of the visual pigments, begun slightly later by Boll and by Ktihne, has elucidated many facts in regard to their chemical nature and the reactions in which they participate, much of the key work in this area having been done in the laboratory of one of the participants in this symposium, Pro- fessor Wald. The pigments responsible for color discrimination have, until very recently, presented us with an apparently insoluble problem. Although since the time of Thomas Young there has been overwhelming evidence of the trichromacy of normal human vision, the pigments responsible for color vision have still not been separated and identified, presumably due to their instability, chemical similarity, and the difficulty of getting them into solution. Yet two key questions have to be answered before the earliest step in retinal analysis of color information could be understood: Are there three pigments which absorb light best in different parts of the spectrum in human cones, and are these segregated in separate receptors? Alternatively, are there three pigments mixed in a single kind of receptor which somehow responds in different ways depending upon which pigment absorbs the most light? A third possibility exists that there is only one pigment which is somehow distributed in the receptor in such a way that it is excited differently by different wavelengths of light. All of these possible mechanisms could give results which are equivalent by any psychophysical test. Therefore, it was necessary to use techniques which permit measurements to be made on the receptors themselves. Two techniques, microspectrophotometry and electrophysiology, appear to have given definite and unequivocal qualitative answers to our question, though many quantitative details remain to be filled in. There are indeed three human cone pigments and each cone contains mainly, if not exclusively, one of these. Let us examine the evidence for this statement. That there is more than one photosensitive pigment in the human fovea was show a number of years ago by Rushton2 who, with Campbell, developed an in- 1331 Downloaded by guest on September 30, 2021 1332 MECHANISMS OF COLOR VISION PROC. N. A. S.

a~~~I S

-at,~

|- -,.Xa. ' e- ..- -I'- am

a's

all *a...... ~as

FIG. 1.-Plots of corrected bleaching-difference spectra of 28 isolated goldfish cones measured by passing light through the outer segments transversely to their axes. (From ref. 7.)

strument that analyzed the light reflected from the back of the eye of a living human subject after it had passed twice through the receptors. By measuring light of various wavelengths absorbed by the receptors before and after bleaching them with colored lights, he identified a green-absorbing pigment he called chlorolabe and a red-absorbing pigment, erythrolabe. He measured the pigment in protanopes, and showed that the erythrolabe was missing; and in deuteranopes, who were found to have no chlorolabe. Thus, he related color blindness to lack of pigment. Undoubtedly, the story of color blindness is not as simple as this, as Professor Wald's paper in this symposium makes evident; but Rushton clearly showed the existence of two different cone pigments and the lack of one or the other of them in some cases of color blindness. Similar experiments were performed on isolated excised foveas by Wald and Brown,I and by Ripps and Weale4 in the living eye with qualitatively similar results. Because of its small quantity and interference from the absorption spectrum of rhodopsin in neighboring rods, it was very difficult to demonstrate a blue-violet-absorbing pigment which is required by the trichromatic theory. Furthermore, experiments on populations of receptors could not answer the second part of the question: Are the pigments segregated into three kinds of receptors? Apparently, only measurements of absorption spectra or action of individual receptors could provide the answer. The problem was a formidable one because cone outer segments are very small and the pigments in them absorb atmostha few Downloaded by guest on September 30, 2021 VOL. 55, 1966 N. A. S. SYMPOSIUM: E. F. MAcNICHOL JR 1333

455 530 625

... . ._ ......

4 0 . .- ...... ^.-. ^ ,' . - . percentoftheincidenlight. Furthermore, the pigment is continu

bbu l ref~7.

re. 7.)** per cent of the inidn light. Futeroe the pimn is cotnal being

FIG.e2.-Averagenbleaching-differncelcurveshobtainedibyedividingshedcuoe ishwlinmig.1 wvereintostremengousaei o detectand plottpingbleaohabldetheargei pigmetsofteainpthesgroupntsotereac segmensgroup atoeacothewaveengh.cotes (Fre-om-o ithe carhop.eThe c minno a spectra.-Measure-t absorpn specta andbeudesir in- instrumentblderachdblylagradtheemeasrinspnot fficie eihsense.dusirin , thebttheexeasriethnt.clarly deosrtoedthathacus oeofcns.Hw rFat measraemts wloueingd epossibesit atnique puwnwr thed totrvs limit. t imtesuoreenrsoaf rotdinthine te oueate, sgmekn ts fh rograos wh gh.areo istrumentwsntsier lnatrumespeilyfortisinlntrad therepupse.7tieease toWithitheueMlarkswayaltomoidentifyconets.fththreeu difraen kindreensofwonedinte gosildfish,ancanimalmveas knwneromthigmandtbeshavitora stdistoitHow-e veavraheiriSmounedyearshagoreslets menouragein isolaionorfithsedfitultieasmalpicthemt limgprove theirnofat hiqeretnad verythe sensJapan,rwith theahopeposestaWithsaTohimprovedyunusuallyeasurint wel-evhiqelBopedel-ehiqonae eslBown,5spectra.iMeanwhileandittabiitdtL iciiebadiscriminatreneeableMaksandcolors.tolos makteraccrae,Furtermcrae, iesreetscoesar largersthanthosofprmtes, semakngtheWysB.measurementds,cmdesieasiehr.onthed in- Thderisruetscrrnlyingrnuseeorearier all veasry siilrandhonsisofacourceHof- vaiablrue-wavpelength moochisromatic ligthmicMroscoeo whichbl theireceptortcell

is mounted, either in isolation or in situ on a small piece of retina, a very sensitive photomultiplier tube, and an electronic recording system which compares the in- Downloaded by guest on September 30, 2021 13:34 MECHANISMS OF COLOR VISION PROCt. N. A. S.

tensity of a beam of light which passes through the receptor to that of a reference beam which does not. Figure 1 shows a number of bleaching difference spectra of single goldfish (cones. The curves were scaled to the same peak height, corrected for bleaching of the light-sensitive pigment during the measurement, and plotted by a computer. It is evident that all but one of these curves fall into three groups. The odd one is the composite spectrum of a pair of twin cones and was rejected in the analysis of the results. Figure 2 shows curves, obtained by averaging the members of each of the three groups. The outer curves are standard deviations of the points, and the large spots are the absorptions of hypothetical rhodopsinlike pigments (the Dartnall nomogram) having maximum absorption at the same wavelengths as the peaks of the receptor curves.- As one can see, the agreement is quite good, so that it is unlikely that the pigments are of a very different composition from rhodopsin, the rod pigment, about which so much is already known. Liebman has repeated and confirmed these experiments independently, using a somewhat different method in which an absorption spectrum rather than a bleaching-difference spectrum is plotted. In addition, Tomita,8 in Japan, has obtained electrically recorded action spectra from single receptors of the carp, a species closely related to the goldfish. By mounting the retina on a vibrating plate, he was able to impale individual receptors with very fine micropipette electrodes. Figure 3 shows his results. The responses to light are negative or hyperpolarizing. The first response is to illumination of a very small area, the second is to illumination at the same intensity of a much larger area. The lower record (b) shows the so-called "S" potentials recorded from deeper in the retina. Illumina- a _ _ _ - tion of a large area gives a larger response than illumination of a small area, indicating that the "S" potential summates the responses of many receptors. The receptor potentials, on the other hand, appear to be area-insensitive as long as the illuminated region is larger than one receptor, as one might expect. This test quickly distinguishes a receptor response from the more easily obtained "S" potential; and as we shall see, the responses to different wave- O _ lengths are quite different. The effect of changing wavelength is shown in Figure 4. The spectrum was scanned in opposite directions in the left and right FIG. 3.-Potentials recorded by columns. As the responses of vertebrate re- Tomita from the carp retina using ceptors are known from the work of Cone9 micropipette electrodes. (a) Receptor to be linearly proportional to the number of potentials. (b) "5" potentials. The first response in each record is to a small quanta absorbed in the photoreceptors, it spot of light projected on the retina; should be possible to obtain a true action spec- the second, to a large spot of the same inteiisity. (From ref. 8.) trum from these records by dividing the am- Downloaded by guest on September 30, 2021 \OL. 55, 1966 N. A. S.SYMPOSIUM: E. F. AMACWICHOL, JR. 1335

a b --g FIG.. 4.-Potentials recorded by Tomita from single cones of the carp. Each response was to a flash of light of a different wavelength as indicated by the scale at the top of the figure. Traces (a), (b), and (c) are for three receptors having different wavelengths of maximum response. (From ref. 8.) plitude at each wavelength by the relative number of quanta in the stimulus. Un- fortunately, Tomita's preliminary records are too noisy to permit an accurate comparison; but he will almost certainly have better data in the near future. As might be expected, Tomita identified three different receptor types with absorptions in the red, green, and blue regions of the spectrum. Responses of three typical receptors are shown in Figure 4. Responses from a single receptor are shown on each line and wavelengths of each pair of records were changed in opposite directions. It is evident that each of the three receptors is maximally sensitive in a different region of the spectrum. Another kind of receptor potential, known as the early receptor potential (ERP), ,which has extremely short latency, was first reported by Brown and _Murakami10 using extracellular microelectrodes, and was found by Cone9 in the gross electroretinogram of the whole eye illuminated by brilliant flashes of light of short duration. The early receptor potentials have a latency of less than 50 /Asec, are linear with intensity, and are resistant to freezing and anoxia." They seem to be signs of processes that take place very early in the excitation of receptors. Their study appears to be opening up a new and exciting chapter in visual physiology. However, this is far from the subject of this paper. The psychologists may rightly ask: There may indeed be three kinds of receptors in the goldfish, but what has this fact to do with human color vision? Fortunately, it has been possible to record absorption spectra from single human cones in the parafovea where the ellipsoids of the receptors are much larger than the outer segments so that these are kept well separated and can be measured in isolation. Marks and Dobelle,"2 in our laboratory, and Brown and Wald" have made such measurements on the cones of humans and of rhesus monkeys. Although our own measurements have been very few to date, a sufficient number have been obtained to permit the plot- ting of fairly smooth average curves. These are shown in Figure 5. The peak absorptions occur at 447, 540, and 577 nm. The black spots are points on the curves for hypothetical (Dartnall nomogram) pigments having maximum absorption at Downloaded by guest on September 30, 2021 1336 MECHANISMS OF COLOR VISION PROC. N. A S.

these three wavelengths. It is evident that the fit is reasonably good, though not perfect. Although the measurements have been few and the individual spectra noisy, we believe the average curves represent closely the true absorption spectra of the receptor pigments. The fact that the curves are narrower than the hypothetical Dartnall curves makes it likely that there is only one pigment in each receptor. Appreciable contamination with the other pigments would be expected to broaden the recorded curves. However, par- ,- A >;tial0bleaching' experiments are needed

75 _- X_\A to settle this question conclusively, and alco AS/ \ we have not yet had the opportunity of > doing them. are X AidHaving\ established that there 2 -' ------three kinds of receptors, each having a

different kind of pigment, in the eyes 400 450 s/,AIk' 500 550 600 650 of fishes and primates, and that the re- WAVELENGTH (MILLIMICRONS) ceptors respond by means of a slow FIG. 5.-Average bleaching-difference spectra electrical polarization rather than by of primate cones obtained by Marks from his best microspectrophotometer records of single cones. the discharge of impulses, the question Circles, dots, and triangles represent points on the arises as to how the color information is absorption curves of hypothetical pigments obey- ing Dartnall's law having maximum absorption further processed in the retina. The at 447, 540, and 577 nm. (Reproduced from anatomists have found that there are MacNichol, Sci. Am., December 1964.) many more receptors than optic nerve fibers. If this were not so, the optic nerve would probably be so bulky that we should be unable to move our eyes. To be able to transmit the large amount of detailed information in this highly convergent pathway, a very efficient code must be used which somehow uses the same optic nerve fibers to transmit both wavelength and pattern information. Furthermore, the limited range of frequencies at which nerve fibers can respond requires a great compression in the magnitude of the data. The visual system can function efficiently over at least 7 decades of light intensity so that adaptation to any level of ambient illumination is an important aspect of retinal data processing. This adaptation appears to be in part photochemical and in part neural. It is quite clear that dark adaptation is closely, though far from linearly, related to the regeneration of photopigment (Rushton, ref. 14). On the other hand, light adaptation (increment thresholds) does not appear to be related to bleaching at all. The detailed mechanisms where- by adaptation is produced are scarcely __* I Iunderstood and furnish an important FIG. 6.-"S" potentials recorded from the and most interesting subject for future isolated retina of the mullet. The responses research. were to flashes of constanit energy at wave- . lengths indicated by the scale at the bottom of The kkey to the retinal processing of the figure. Upper record, "luminosity" or "L"- color and pattern information appears to type response; lower record, "chromatic" or "C"-type response. (From ref. 17.) be the comparison in the neural retina Downloaded by guest on September 30, 2021 VOL. 55, 1966 N. A. S. SYMPOSIUM: E. F. MACANICHOL, JR. 1337

of the output signals from the receptors. The magnitudes of the responses of the individual receptors are not transmitted directly to the brain, but rather the responses of groups of receptors having different locations and wavelength sensitivities are compared. Signals which are functions of the differences in excitation of the receptors are transmitted along the optic nerve fibers. Thus, although the finding that there are three color receptors is in accord with the Helmholtz theory of color vision, the neural retina appears to operate in a manner reminis- cent of the opponent-colors theory for- - -- mulated by Hering. Let us examine some of the evidence for this statement. __ Some years ago Svaetichin'5 recorded slow potential changes, unlike any poten- FIG. 7.-Chromatic "S" potentials recorded tials previously reported, from the iso- as a function of time. The numbers in the lower records refer to the responses recorded lated retinas of fishes by means of fine simultaneously on a wavelength scale shown in micropipette electrodes. Two kinds of the upper record. Major divisions on time trace responses are shown in Figure 6. One of 0.1 sec. (From ref. 17.) these, termed the luminosity or "L" response, arose from a variable resting potential of about -10 to -40 mv and became more negative, or hyperpolarizing, as the light intensity was increased. It had a broad spectrum and showed summation of stimuli delivered over a large area. It also appeared to arise from a large structure since

FIG. 8.-Effects of steady background illumination upon "L" and "C"-type responses. Tri- angles indicate the wavelength of the background illumination. Arrows in each record indicate the d-c potential shift caused by the background illumination. (From ref. 17.) Downloaded by guest on September 30, 2021 1338 MECHANISM'S OF COLOR VISION PRoc. N. A. S.

the microelectrode could be moved a considerable distance without changing the response. By means of a dye-marking technique, Svaetichin and 116 were able to show that this response arises in giant horizontal cells which are prominent in the retinas of fishes. The other type of response, called by Svaetichin the chromatic or "C" response, is hyperpolarizing in some parts of the spectrum and depolarizing in others. It is evidently the result of the opposition of two antagonistic processes which are sub- tracted algebraically. Evidence that this is so is provided by Figures 7 and 8. The first shows the responses to flashes of light on an expanded time scale. Each lower record corresponds to a number in the upper figure. At short wavelengths the response is negative and rises and falls rapidly. At long wavelengths, it is positive and rises and falls more slowly. At intermediate wavelengths the potentials are of different amplitudes and subtract from one another, giving short transients at the beginning and end of the response due to the more rapid rise and fall of the short- wavelength-sensitive negative process. In the steady state the response is either negative or positive, depending upon whether the wavelength is shorter or longer than a well-defined neutral point, at which steady amplitudes of the two processes cancel exactly. Figure 8 shows the effects of red and blue background illumination upon the magnitude of the responses. As shown in the first three traces, blue light en- hances the red-sensitive depolarizing component and depresses the blue-sensitive hyperpolarizing component. The next two traces show that red light has just the opposite effect. The fifth trace is a control at the end of the experiment. The remaining traces show that blue and red background lights have equivalent effects in decreasing the amplitude of an "L"-type response, in which antagonistic processes are apparently not present. It has been suggested that the "L" response is the sign of an integrative process that ties together large aeras of the retina. Perhaps

~~2~~rct-~~~~~~T~~600V1

FIG. 9.-Responses of a retinal ganglion cell (goldfish) to illumination at different wavelengths. Bottom trace indicates 0.5-sec illumination. Stimullus presented at 1.5-sec intervals. Spikes occurring before start of stimulus are responses to a previous stimulus. (From ref. 21.) Downloaded by guest on September 30, 2021 VOL. 55, 1966 N. A. S. SYMPOSIUM: E. F. AlAcNICHOL, JR. 1339

560 _

l a k \ '\ OFF ON ~~ ..~~~~.*.I..I... S~~~~~~~ FIG 10 500 700TO WAVELENGTH IN MILLIMICRONS FIG. 10 (Left).-Response of same cell l rrecordedr in Fig. 9 to 10 nm change in wave- T- length. (From ref. 21.) FIG. 1 (Above).-Relativeenergy thresh- old for single "on" and "off" responses of a retinal ganglion cell (goldfish) at different wavelengths. it is part of a feedback mechanism that decreases the sensitivity of the ganglion cells in the presence of ambient light, and is thus involved in the neural component of light adaptation. We do not yet know. Svaetichin and I tried to identify the structures in which the antagonistic "C" responses arise. We were able only to show that they arise in the layer of bipolar cells further toward the vitreous side of the retina than the giant horizontal cells. The responses were abolished by slight movement of the electrode so that the structures involved are presumably smaller than the giant horizontal cells. Mitarail7 was able to localize the electrode tip more precisely in the tissue and tentatively identified the structures involved as the Muller fibers surrounding the bipolar cells. However, his slides do not present unequivocal evidence in support of this hypothesis. The stain marks he has demonstrated could probably be in the bipolar cells themselves or possibly in amacrine cells. The fact that the giant horizontal cells and Muller fibers were identified histologically as glial elements has led Svae- tichin and his co-workers to some rather elaborate theories of neuron-glial cell in- teraction. However, some recent anatomical findings indicate that a re-evaluation of these hypotheses is necessary. W. Stell, by means of a combination of the classic Golgi staining technique with electron microscopy, has shown that the giant horizontal cells in the fish retina have dendrites that synapse with the receptors in a manner identical with the bipolar cells.18 Thus, they appear to be neurons and not glia. Dowling and Boycott'9 have reached similar conclusions by similar methods with regard to the smaller horizontal cells of monkeys. If the "C" responses in fact arise in the bipolar or amacrine cells, the glial cell hypothesis is quite un- necessary. What has prevented this conclusion from being reached a long time ago is the peculiar electrophysiology of the structures giving rise to the "S" potentials. Unlike ordinary neurons, they have a small and variable resting potential, a comparatively small membrane resistance, and do not discharge all-or- nothing propagated impulses. Instead, they hyperpolarize -and depolarize in a graded fashion. Since conduction over great distances is not required, this is probably sufficient to cause the release of transmitter substance at their synapses. If the "C" responses do in fact arise in the bipolar cells and are thus on the direct pathway of information transfer in the retina, one would expect the discharge re- Downloaded by guest on September 30, 2021 1340 MECHANISMS OF COLOR VISION PROC. N. A. S.

corded from the retinal ganglion cells and their axons, the optic nerve fibers, to re- flect the opponent "C" responses. Some time ago, Wagner, Wolbarsht, and J20 obtained some results that indicated that this was indeed the case. We recorded impulses from single, retinal ganglion cells of the goldfish and found that when they were illuminated by '/2-sec flashes of white light, most of them were of the classic "on-off" type first described many years ago by Hartline.2' However, when we used monochromatic light of different wavelengths, some cells responded only during illumination with short-wavelength light, and firing was inhibited when the light was turned off. When light of long wavelength was used, the discharge was inhibited during illumination and there was a vigorous response as soon as the light was turned off. This phenomenon is shown in Figure 9. Each record is taken from a series of responses to flashes of '/2-sec duration separated by 11/2 sec. Responses occurring before the flash, which is indicated by the bottom trace, are "off" responses to the previous flash. It is evident that there is a change from "on" to "off" between 500 and 600 nnm. In fact, as shown in Figure 10, the change was complete in only 10 nm, indicating a very precise mechanism of wavelength discrimination. It was possible to plot the sensitivities of the "on" and "off" com- ponents by adjusting the intensity at each wavelength to give a constant response of one or more impulses. Such curves for a cell responding to "on" at short wavelengths and to "off" at long wavelengths is shown in Figure 11. Similar responses were found earlier by DeValois22 in the lateral geniculate body of the monkey and have been studied in detail by Hubel and Wiesel.23 They have also been recorded from the optic nerve of the ground squirrel by AIlichael.24 If the "on-off" discharges result from the opposed effects of stimulating the various classes of receptors, it might be expected that interactions of all possible pairs would be represented. This appears to be the case in the goldfish. One of the graduate students in our department, Lewis Bicking,25 in the course of working out a mathematical method of representing the firing of ganglion cells at high intensities of illumination, obtained the wavelengths of maximum on- and off-sensitivity of 55 ganglion cells. His results are summarized below. On Off Number 660 540 19 540 660 27 480 540 4 540 480 1 480 660 3 660 480 1 It is evident that all possible pairs of opponent responses are represented, though those involving blue are much fewer in number than those having red and green maxima. The wavelengths of maximum response do not correspond exactly to those determined spectrophotometrically by Marks, but this is not surprising for several reasons: to save time, determinations were made at 40-nm intervals, so that the exact wavelength of maximum sensitivity was not found. In addition, the curves are functions of the differences between the responses of two receptor types rather than responses of single types. Downloaded by guest on September 30, 2021 VOL. 55, 1966 N. A. S. SYMPOSIUM: E. F. MACNICHOL, JR. 1341

S o t 6

400 500 600 700 40 5000056000700 r X/ j * > ~~~~OFF t 2 ~~~~~~~X ON x A.650 ON

W 0 -I O IF O lX~ 'I 9

OFF S 5-500 10

MILLIMETERS -I bNLUMElERS~~~0+1 FIG. 12.-Thresholds for impulse discharge of two goldfish retinal ganglion cells as a function of position in receptive field. Insets show wavelength sensitivities of the "on" and "off" compo- nents of discharge of each cell. The wavelengths of stimulation were chosen to excite either the "on" or 'off" component separately. (From ref. 21.)

i X,

FIG. 13.-Schematic drawing of the synaptic connections of the primate retina (from Dowling and Boycott, ref. 19). Top layer of cells: receptors (rods and cones); second layer: horizontal cells; third layer: bipolar cells; fourth layer: amacrine cells; fifth layer: ganglion cells. Downloaded by guest on September 30, 2021 1342 AIECHANISAIS OF COLOR VISION PROC. N. A. S.

In addition to "on-off" coding of color information, the "on-off" code is also used to indicate spatial location within the receptive field of a given ganglion cell. This type of coding was first reported by Kuffler et al. to occur in the retinal ganglion cells of the cat.26 The goldfish cells show similar behavior. But because the "oin" and "off" processes have different wavelength sensitivities, it is possible to study them in isolation. When this is done, rather than being "on-center" "off-surround" or vice versa, as they appear to be at high intensities with white light, both processes are maximally sensitive in the center of the receptive field. But one is more sensitive in the center and the other has larger spatial extent as shown in Figure 12, which plots the sensitivity of two different cells as a small spot of light was traversed across their receptive fields. Thus, position and color appear to be coded in the same way by a given ganglion cell. If this is so, how can the two sense modalities be distinguished? From the output of a single ganglion cell there is of course no way of doing this; but when the responses of a number of different ganglion cells, each with different coding such as "on-center-red" and "off-periphery-blue," or "off-center-red" and "on-center-blue," are taken together, there is probably enough information to decode the message. Future advances in our understanding of the retinal encoding process will probably come from two lines of work: multiple unit analysis to determine the details of the code itself; and an analysis of the synaptic connections in the retina of the type used so beautifully in the laboratories of Eccles27 and Kuffler. The synaptic connections of the retina are many and complicated, as shown ill Figure 13, which is a schematic drawing of part of the retina prepared by Dowling'9 after analysis of his many electron micrographs. There is ample opportunity for con- vergence, excitation, and inhibition at all levels. Further careful electrophysio- logical and electron microscopic study of the synaptic organization should eventu- ally provide us with a complete "wiring diagram" of the retina. In closing, it is appropriate to discuss briefly the significance of the "on-off" code. The "off" response appears to be associated with a strong inhibition during illumination. The "on" response likewise appears to be associated with inhibition of the spontaneous activity which usually occurs in the dark. Thus, the paired excitatory and inhibitory effects have been thought to be mechanisms for en- hancing contrasts in color and pattern. The bursts of activity that most ganglion cells exhibit during transient changes in illumination also should enhance temporal contrast. As one's eyes are continually in motion, the "oIn" and "off" responses must be continually generated as the image of the outside world moves over the retinal mosaic. The now well-known deterioration of visual acuity when the retinal image is artificially stabilized, which was investigated by Riggs and Ratliff,28 and by Ditchburn,29 furnishes proof of the importance of eye movements in vision. Thus, the "on" and "off" responses appear to be aii important mechanism in the vision of intact animals and not just artifacts generated under laboratory conditions.

* This paper is essentially that presented orally at the One Hunidred and Second Annual MIeet- ing of the National Academy of Sciences, Washington, D.C., during the Symposium on Mechanisms of Color Vision, on April 28, 1965. Some new references were added at the time the paper was re- vised for publication. Support by NIH and NSF is gratefully acknowledged. I Holmgren, F., "Method att objectivera effecten av ljusintryck ph retina," Uppsala LdkFbr. Forh., 1, 177-191 (1865). Downloaded by guest on September 30, 2021 VOL. 55, 1966 N. A. S. SYMPOSIUM: E. F. MAcNICHOL, JR. 1343

2Rushton, W. A. H., "Kinetics of cone pigments measured objectively on the living human fovea," Ann. N. Y. Acad. Sci., 74, 291-304 (1958). 3 Brown, P. K., and G. Wald, "Visual pigments in human and monkey retinas," Nature, 200, 37-43 (1963). 4Ripps, H., and R. A. Weale, "Cone pigments in the normal human fovea," Vision Res., 3, 531--543 (1963). 5 Brown, P. K., "A system for microspectrophotometry employing a commercial recording microspectrophotometer," J. Opt. Soc. Am., 51, 1000-1008 (1961). 6 Liebman, P. A., "In situ microspectrophotometric studies on the pigments of single retinal rods," Biophys. J., 2, 161--178 (1962). Marks, W. B., doctoral dissertation, "Difference spectra of the visual pigments in single gold fish cones," Department of Biophysics, The Johns Hopkins University, 1963; Marks, W. B., "Visual pigments of single goldfish cones," J. Physiol., 178, 14-32 (1965). 8 Tomita, T., "Relation between the electroretinogram (ERG) and single cell activities in the retina," U.S. Army Research and Development Group (9984) (Far East), contract no. CRD-AG- S92-544-63-G15, June 18, 1963 to June 17, 1964. 9 Cone, R. A., "Early receptor potential of the vertebrate retina," Nature, 204, 736--739 (1964). 10Brown, K. T., and M. Murakami, "A new receptor potential of the monkey retina with no detectable latency," Nature, 201, 626-628 (1964). 11 Pak, W. L., and T. G. Ebrey, "Visual receptor potential observed at sub-zero temperatures," Nature, 205, 484-486 (1965). 12 Marks, W. B., W. H. Dobelle, and E. F. MacNichol, Jr., "Visual pigments of single primate cones," Science, 143, 1181-1183 (1964). 13 Brown, P. K., and G. Wald, "Visual pigments in single rods and cones of the human retina," Science, 144, 45-51 (1964). 14Rushton, W. A. H., "Bleached rhodopsin and visual adaptation," J. Physiol., 181, 645-655 (1965). 15 Svaetichin, G., "Spectral response from single cones," Acta Physiol. Scand., Suppl. 39, 134, 17-46 (1956). 16 MacNichol, E. F., Jr., and G. Svaetichin, "Electric responses from the isolated retinas of fishes," Am. J. Ophthalmol., 46, No. 3, pt. II, 26-39 (1958). 17 Svaetichin, G., M. Laufer, G. Mitarai, R. Fatechand, E. Vallecalle, and J. Villegas, "Glial control of neuronal networks and receptors," in The Visual System: Neurophysiology and Psycho- physics, ed. R. Jung and H. Kornhuber (Berlin: Springer, 1961), pp. 445-456. 18 Stell, W. K., "Correlation of retinal cytoarchitecture and ultrastructure in Golgi prepara- tions," Anat. Rec., 153, 389-398 (1965); Stell, W. K., Ph.D. thesis, University of Chicago, 1966. 19 Dowling, J. E., and B. Boycott, "Organization of the vertebrate retina, electron microscopy," submitted to Proc. Roy. Soc. (London), 1966. N Wagner, H. G., E. F. MacNichol, Jr., and M. L. Wolbarsht, "The response properties of single ganglion cells in the goldfish retina," J. Gen. Physiol., 43, 45-62 (1960), suppl. 6, part 2; Wolbarsht, M. L., H. G. Wagner, and E. F. MacNichol, Jr., "The origin of "on"- and "off"- responses of retinal ganglion cells; receptive fields of retinal ganglion cells," in The Visual System: Neurophysiology and Psychophysics, ed. R. Jung and H. Kornhuber (Berlin: Springer, 1961), pp. 163-175; Wagner, H. G., E. F. MacNichol, Jr., and M. L. Wolbarsht, "Functional basis for "on"-center and "off"-center receptive fields in the retina," J. Opt. Soc. Am., 53, 66-70 (1963). 21 Hartline, H. K., "The response of single optic nerve fibers of the vertebrate eye to illumination of the retina," Am. J. Physiol., 121, 400-415 (1938); Hartline, H. K., "The receptive field of the optic nerve fibers," Am. J. Physiol., 130, 690 (1940). 22 DeValois, R. L., C. J. Smith, S. T. Kitai, and A. J. Karoly, "Responses of single cells in different layers of the primate lateral geniculate nucleus to monochromatic light," Science, 127, 238 (1958); DeValois, R. L., and A. E. Jones, "Single cell analysis of the organization of the primate color-vision system," in The Visual System: Neurophysiology and Psychophysics, ed. R. June and H. Kornhuber (Berlin: Springer, 1961), pp. 178-191; DeValois, R. L., G. H. Jacobs, and A. E. Jones, "Responses of single cells in primate red-green color vision system," Optik, 20, 87-98 (1962). Downloaded by guest on September 30, 2021 1344 MECHANISMS OF COLOR VISION PROC. N. A. S.

23 Hubel, D. H., and T. N. Wiesel, "Receptive fields of optic nerve fibers in the spider monkey," J. Physiol., 154, 572-580 (1960). 24 Michael, C., Ph.D. thesis, "Receptive fields of single optic nerve fibers in the ground squirrel," Harvard University, 1965; Michael, C., "Receptive fields of directionally selective units in the optic nerve of the ground squirrel," submitted to Science, 1966; Michael, C., "Receptive fields of opponent color units in the optic nerve of the ground squirrel," submitted to Science, 1966. 25 Bicking, L. A., Ph.D. thesis, "Some quantitative studies on retinal ganglion cells," Johns Hopkins University, 1965. 28Kuffler, S. W., "Discharge patterns and functional organization of mammalian retina," J. Neurophysiol., 16, 37 (1953); Barlow, H. B., R. Fitzhugh, and S. W. Kuffler, "Change of organization in the receptive fields of the cat's retina during dark adaptation," J. Physiol. (London), 137, 338 (1957). 27 Eccles, J. C., The Physiology of Synapses (New York: Academic Press, 1964). 28 Riggs, L. A., F. Ratliff, J. C. Cornsweet, and T. N. Cornsweet, "The disappearance of steadily fixated test objects," J. Opt. Soc. Am., 43, 495-501 (1953). 29 Ditchburn, R. W., and B. L. Ginsborg, "Vision with a stabilized retinal image." Nature, 170. 36 (1952). Downloaded by guest on September 30, 2021