AN OPPONENT-PROCESS THEORY OF COLOR VISION · 2019-02-27 · color theory" remind us more and more of its chief theoretical rival, the Bering theory of three paired, opponent color

Psychological ReviewVol. 64, No. 6, 19S7

AN OPPONENT-PROCESS THEORY OF COLOR VISION

LEO M. HURVICH1 AND DOROTHEA JAMESON1

Eastman Kodak Company

The two major theoretical accountsof color vision are those classified asthe Young-Helmholtz and the Heringtypes of theories. For many years theformer has been judged by most work-ers in the field to provide the simplestexplanation of the way in which lightstimuli give rise to color sensations.The advantages that appear to favorthe Young-Helmholtz three-componenthypothesis are two: it is parsimonious,and its postulates are easily quantifiableand hence subject to precise experimen-tal test. In its parsimonious and easilyquantifiable form, the theory is simple:in addition to the rods which subservetwilight vision, the eye contains threekinds of cone photoreceptors; each typeof cone contains a differently selectivephotochemical substance; each is asso-ciated with its own specific nerve fiber;and each cone-photochemical-nerve fibersystem is correlated with one of thethree specific "fundamental" color sen-sations, namely, red, green, and blue(or violet). All sensations are consid-ered as compounded of varying amountsof these three excitatory systems, withwhite arising from equal and simultane-ous excitation of all three, and yellowfrom equal red and green excitations.

The Young-Helmholtz three-cone,three-nerve, three-sensation theory de-rives directly from the basic fact ofcolor mixture, namely, that all visiblehues can be matched by the mixture,in proper proportions, of only threephysical light stimuli. Based squarelyon this fact, the theory is readily quanti-fied in terms of the three measurablevariables of color mixture experiments.But the three measured variables, it

1 Now at New York University.

must be emphasized, are the threephysical light stimuli used in the colormixture experiments; they are not thepostulated three "fundamental" colorsensations, for with each different stimu-lus triad used for color matching a dif-ferent and equally valid triad of colormixture functions is obtained. Conse-quently, throughout some hundred yearssince the original formulation of theidea, a continued series of attempts hasbeen made to find the proper transfor-mation of the three measured color-mixture curves that will bridge the gapand yield the unique spectral distribu-tion curves of the desired physiologicalcorrelates of the three postulated "fun-damental" color sensations. An infinityof such transformations is available fortrial, and almost every serious adherentof the theory has proposed at leastone new set of "fundamental sensationcurves" (48, pp. 368-372). The search,however, continues, because serious de-fects have been found in every proposalmade thus far. When the explanatoryor predictive power of the theory in anygiven quantified form is tested it can-not handle more than a limited numberof facts satisfactorily (11, p. 805).

Moreover, some facts of color experi-ence seem unassimilable into the frame-work of the simple Young-Helmholtztheory with its three independent, funda-mental, process-sensation systems. Howcan this system of three independentprocesses be made to account, for ex-ample, for the apparent linkages thatseem to occur between specific pairs ofcolors as either the stimulus conditionsor the conditions of the human observerare varied? Why should the red andgreen hues in the spectrum predominate

384

OPPONENT-PROCESS THEORY OF COLOR VISION 385

at low stimulus levels, and the yellowand blue hue components increase con-comitantly as the spectrum is increasedin luminance (43)? Why, as stimulussize is greatly decreased, should dis-crimination between yellow and bluehues become progressively worse thanthat between red and green (4, 10)?Why should the hues drop out in pairsin instances of congential color defect,or when the visual system is impairedby disease (29, 31)? On the otherhand, since the sensation of white isgranted no special physiological processin this parsimonious theory, but occursas the fusion product of three equallylarge fundamental hue sensations, howaccount for the large degree of inde-pendence of white and chromatic quali-ties when the adaptation of the visualsystem is varied (37, 41)?

As more and more ad hoc hypothesesare added to the original Young-Helm-holtz formulation in order to answerthese and other problems forced by theincreasing accumulation of experimentaldata, we naturally find the formulationbecoming less and less precise and quan-tifiable, and obviously less parsimonious.We also find, however, that exactlythose phenomena that require modifica-tion and extension of the simple "three-color theory" remind us more and moreof its chief theoretical rival, the Beringtheory of three paired, opponent colorprocesses.

In view of this situation, it seemshighly desirable that we take a closesecond look at Bering's alternative ap-proach to an understanding of colorphenomena. The vast accumulation ofpsychophysical data for which any ade-quate theoretical proposal must accountrequires that the basic postulates of thetheory, as outlined qualitatively by Her-ing (13, 14), be restated in quantitativeterms for such a critical scrutiny to bemost meaningful. This paper will re-view our attempt to provide such a

quantitative restatement, and will sum-marize briefly some of the critical com-parisons between the theoretical de-ductions and relevant psychophysicaldata. (Detailed quantitative accountsare given in 21, 22, 23, 25, 26, 27.)

BASIC SCHEMA FOR THEBERING THEORY

The Three Variables

The Bering theory is like the Young-Belmholtz theory in that it, too, postu-lates three independent variables as thebasis for color vision, but the Beringvariables are three pairs of visual proc-esses directly associated with three pairsof unique sensory qualities. The twomembers of each pair are opponent,both in terms of the opposite nature ofthe assumed physiological processes andin terms of the mutually exclusive sen-sory qualities. These paired and op-ponent visual qualities are yellow-blue,red-green, and white-black.

The basic schema for the opponent-colors mechanism is shown diagram-matically in Fig. 1. The three pairedopponent response systems are labeledy-b, r-g, and w-bk. The convention ofpositive and negative signs is used to

Neural Responses

ErD

light mediating substance

FIG. 1. Basic schema for Bering opponent-colors mechanism.

386 LEO M. HUEVICH AND DOROTHEA JAMESON

indicate that each neural system is ca-pable of two modes of response thatare physiologically opponent in nature,and that the paired sensory qualitiescorrelated with these opposed modes ofresponse are also mutually opponent orexclusive. That is, we may experiencered-blues or green-blues but never yel-low-blues, and we see yellow-greens orblue-greens, but never red-greens, andso on. In the absence of any externalvisual stimulus, the state of the visualsystem is assumed to be a condition ofactive equilibrium, and this equilibriumcondition is associated with the neutral,homeogeneous "gray" sensation per-ceived after a long stay in completedarkness. This sensation is quite dif-ferent from the black experience of thewhite-black opponent pair. Blacknessarises neither by direct light stimulationnor in the simple absence of light, butrather by way of either simultaneous orsuccessive contrast during, or following,light stimulation of some part of theretina.

Properties of Paired SystemsThe three pairs of visual response

processes are independent of each other;that is, they have different responsethresholds, they follow different laws ofincrease with increase in strength ofstimulation, and probably have differenttime constants. The achromatic systemis the most sensitive; that is, the amountof photochemical absorption necessaryto excite the achromatic white responseis less than the amount of photochemi-cal activity required to stimulate eitherthe y-b or r-g chromatic pairs. Thischaracteristic accounts for the existenceof the so-called achromatic interval, i.e.,the fact that spectral lines appear achro-matic at the absolute threshold for visi-bility (42, p. 167). Similarly, the red-green system has a lower threshold thanthe yellow-blue one. The failure of theyellow-blue system to respond at near-

threshold levels that are sufficient toactivate the red-green system exhibitsitself in the facts of so-called "smallfield dichromasy," in which the eye be-haves, with respect to stimuli that arevery small in area as well as of low in-tensity, in a manner similar to the con-genital tritanope, i.e., a specific type of"color blind" individual for whom yel-low and blue discriminations are im-possible and the only hues seen are redsand greens (4, 49).

With increase in level of stimulationthe different paired systems also showdifferences in rate of response increase,such that the athromatic response in-crease is probably the most rapid of thethree, with the result that at very highintensities all spectral stimuli show astrong whitening, or desaturation, rela-tive to their appearance at some inter-mediate luminance level (42, p. 168).Of the two chromatic pairs, the yel-low-blue system, although exhibiting ahigher threshold, shows a more rapidrate of increase in response with increasein luminance than does the red-greensystem. Thus, the mixed hues of thespectrum—the violets, blue-greens, yel-low-greens, and the oranges—all varysystematically with increase in spectralluminance, and all show a tendency tobe more blue or yellow, respectively, athigh luminances, and more red or greenat the lower luminance levels (theBezold-Brucke hue shift phenomenon).

The opponent systems show a tend-ency toward restoring the balanced equi-librium condition associated with theneutral "gray" sensation. Thus excita-tion, say, of the r process in the r-gsystem results in a decrease with timein r responsiveness, and in an increasein the responsiveness of the opponentg process. If we think of the r processas perhaps associated with the buildingup of an electrical potential in theneural system, and of the g process asassociated with the collapse of the po-


tential during impulse firing, then it iseasy to see that as the neural potentialis increased to higher values there willbe a tendency to resist further build up,and also an increased disposition of thetissue toward impulse firing in order torestore the potential to its normal equi-librium value. Although we are not atall ready to ascribe a specific neuralcorrelate of this sort to the postulatedopponent processes at this time, theneurophysiological parallels are usefulfor conceptualizing the opponent-proc-ess notion as a real biological phenome-non.

To return to our example, if the re-sponsiveness of the opponent g processtends to increase as r excitation is con-tinued, then when the r stimulus is re-moved we can expect g activity to bereleased, strongly at first, then moreslowly, and ultimately fading out asequilibrium is again approached. Thesensory correlate of this reversal of op-ponent activities with removal of stimu-lation is, of course, the familiar phe-nomenon of the complementary after-image. If the stimulus (of constantmagnitude) is not removed but con-tinues to act for a considerable lengthof time, then the r process, whose re-sponsiveness is being continuously de-creased, will eventually cease to respondfurther, and a' new equilibrium statewill be reached. The disappearance ofa sensory response with continued con-stant stimulation can be observed eitherby the Ganzfeld technique, in which thewhole retina is uniformly illuminated bydiffuse light (18), or by the "paintedimage" technique, in which opticalmeans are used to fix a well definedimage on the retina in such a way thatits retinal position remains constant andindependent of eye movements (39).By either method the eventual resultof continued steady stimulation is adisappearance of the visual experience:the light seems to have gone out in the

Ganzfeld situation, or, in the fixed-image situation, the perceived objectsimply fades out of view.

Not only are the visual responsesmodified by changes in time in the ex-citabilities of the opponent processes,but they are also importantly affectedby spatial interaction among the vari-ous elements of the visual field. Withincertain limits there is evidence of sum-mation of similar kinds of activity inadjacent elements, as in threshold re-sponses for small stimulus areas (5, pp.846-852). But perhaps more impor-tant for the over-all functioning of thevisual system are the antagonistic in-teractions, such that r activity in onearea induces g activity in adjacent areas,and similarly for the yellow-blue andwhite-black paired response systems.These opponent spatial induction effectsare evident in all the familiar color andbrightness contrast phenomena (35, pp.138-142). They are probably also pri-marily responsible for the great visual-image clarity that characterizes visionin spite of the fact that the optical sys-tem of the eye is obviously imperfect,and that consequently the light imageformed on the retinal surface lackssharply defined boundaries (17, pp.1S1-159). The spatial interaction caus-ing intensification of opponent qualitiesat adjacent surfaces would seem anideal crispening device to sharpen upthe initially blurred retinal image.

Photochemical PostulatesIn addition to the various temporal

and spatial induction effects, which areassumed to be based in the neuralvisual-response tissue, visual adaptationprobably also involves changes in thephotochemical activities that initiatethe neural responses, since a certainamount of photochemical bleaching isexpected to occur with continued ex-posure of the photosensitive materialsto a retinal light stimulus. In order

388 LEO M. HURVICH AND DOROTHEA JAMESON

Neural Responses

Photochemical Absorptions

Neural Responses

Photochemical Absorptions

y-b = k:(r-g -k2(o+y-2/s)

FIG. 2. Schematic diagram showing rela-tions between photosensitive materials a, /3,and 7 and neural opponent response processesy-b, r-g, and w-bk.

for the three paired opponent-responsesystems to be selectively stimulated,there must, of course, be more than onesubstance available for photochemicalmediation between the incident lightand the neural excitation. Whateverthe specific nature of the photosensitivematerials, they must form a link in thesystem of three independent variables,

and hence we have postulated threeindependent photosensitive materials,which we may call a, ft and y.

Our schematic model now takes theform shown in Fig. 2A or 2B. Thethree independent photosensitive mate-rials may be contained in discrete reti-nal units with complex interconnectionsto the neural response systems, as shownin Fig. 2A, or two or more of thesematerials may be combined in receptorunits having simpler connections to theneural response systems, as diagrammedin Fig. 2B. There is no way of dif-ferentiating these models in terms ofvisual behavior; and however the threephotochemicals may be segregated orcombined in the retina, and whateverthe number of different photoreceptorunits, there remain only three independ-ent photosensitive materials, and thetheory remains a three-variable, op-ponent-colors schema.

QUANTIFICATION OF OPPONENTSTHEORY

Since our aim is to present thisschema in quantitative terms, one ofthe first questions that has to be askedis this: Is it possible to obtain by psy-chophysical experiment direct measure-ments of the spectral distributions ofthe three basic response variables ofthe Hering theory?

Measures of Achromatic and ChromaticResponses

It can fairly be assumed that theachromatic, white response is closelyconnected with the distribution of thebrightness quality throughout the visiblespectrum, and Fig. 3 therefore showstwo functions (which we have measuredby a threshold technique) that give thewhiteness distribution of an equal en-ergy spectrum for two observers (20).The induced rather than directly stimu-lated black component of the achro-


WAVELENGTH-nyl

FIG. 3. Whiteness distribution of an equal en-ergy spectrum for two observers.

matic white-black response pair has thissame distribution, but of opposite sign,since the strength of the black contrastresponse is directly related to the mag-nitude of either the surrounding or thepreceding whiteness or brightness.

A method for determining the spectraldistributions of the paired chromatic re-sponses is implicit in the opponents the-ory itself. Since the two members ofeach hue pair are mutually opponent orexclusive, then a yellow response ofgiven strength should be exactly can-celed by a stimulus that, taken alone,elicits the same magnitude of blue re-sponse, and a similar relation shouldhold between red and green responses.Thus a null method, based on the an-tagonism of the two members of eachhue pair, can be used to measure thespectral distributions of the chromaticresponses. In brief, a wave length isfirst selected that evokes, say, a blue hueresponse. The observer then views, inturn, a series of spectral wave lengthsthat appear yellowish in hue (yellow-greens, yellow, and yellow-reds). Toeach of these yellow stimuli just enoughof the previously selected blue stimulusis then added exactly to cancel the yel-low hue without introducing any blue-ness. The observer simply reports whenthe test field appears neither yellow norblue; the hue remainder that he sees

may be green, neutral, or red, depend-ing on the test wave length. Knowingthe energies of the series of spectralyellow stimuli, and having determinedexperimentally the energy of the bluestimulus of fixed wave length that is re-quired for the hue cancellation in eachcase, we can now plot the distributionof the relative magnitudes of yellow hueresponse evoked by the various testwave lengths. The procedure is simplyreversed to obtain the distribution of

WAVELENGTH -m/i

UJ + 050 •

WAVELENGTH-m/i

FIG. 4A and 4B. Chromatic and achromaticresponse functions for equal energy spectrumfor two observers.

390 LEO M. HUEVICH AND DOROTHEA JAMESON

the blue component of the yellow-bluepair; that is, varying amounts of a fixedwave length of yellow hue are used tocancel the blue hue quality of a seriesof "blue" test wave lengths. By usinga red stimulus of fixed wave length andvariable energy to cancel the greens,and a green stimulus to cancel the reds,the spectral distribution of the red-green pair of chromatic responses issimilarly determined.

Two sets of paired chromatic responsevs. wave length functions that weremeasured in this way (25), togetherwith the achromatic spectral functionsshown in Fig. 3, are plotted in Fig. 4for an equal energy spectrum. The op-ponent members of each hue pair havebeen given arbitrary positive and nega-tive designations, to correspond withtheir opponent characteristics. Thusthe positive values of the red-greenfunction indicate redness, and the nega-tive values greenness. Similarly, thepositive values of the yellow-blue func-tion indicate yellowness, and the nega-tive values blueness.

These are the psychophysical func-tions that represent the spectral dis-tributions of the three independentvariables of the Hering opponent-colorstheory for these two observers. Theyare assumed to be directly correlatedwith the response activity of the visualnervous tissue (retina, optic nerve, andvisual centers), and should not be takenas photochemical absorption spectra,about which these data tell us nothing.

Brightness, Hue, and SaturationThe psychophysical opponent-response

functions shown in Fig. 4 provide a di-rect description of the appearance ofthe spectrum, for these observers, for aneutral condition of bright adaptationand at the moderate level of luminancefor which the functions were obtained.Thus, all wave lengths evoke somewhiteness as well as hue; the whiteness

and brightness of an equal energy spec-trum is relatively small at the two spec-tral extremes and relatively high at theintermediate wave lengths.

The short wave lengths appear asred-blue hues (violets); there is a nar-row band of pure or unique blue wherethe red-green function is equal to zero;then come the blue-greens, followed bya narrow band of unique green at thewave length where the yellow-blue func-tion is equal to zero; this is followedby the yellow-greens, and then pureyellow occurs at the second intersectionof the red-green function with the zeroordinate value; and finally the yellow-red hues appear in the long wave lengthregion (19). A quantitative expressionfor hue, a "hue coefficient," can be ob-tained by taking the value of one ofthe chromatic responses, say, the yellowvalue at 550 imt, relative to the total ofall chromatic responses at that wavelength, in this case, yellow plus green.

The saturation of the color dependson the relative amounts of chromaticand achromatic responses. At the twospectral extremes where the chromaticresponses are large relative to the whiteresponse, the spectral saturation is high.Where the reverse is true, spectral satu-ration is low. This can be expressedquantitatively in the form of a "satura-tion coefficient." To use the same ex-ample, the total of the yellow-plus-greenvalues relative to the white plus yellowplus green is relatively low at 550 m/t,and this wave length appears much lesssaturated than does, say, either 440 m/j.or 670 m/t.

Color MixtureSince color-mixture experiments sim-

ply involve matching the three per-ceived qualities evoked by one stimulusby the proper mixture of three otherstimuli, it is possible to determine thecolor-mixture relations that are inher-ent in the response curves of Fig. 4 for


any three arbitrarily selected mixtureprimaries. That is, the red-green value,the yellow-blue value and the whitevalue of the total visual response to anywave length of unit energy are matchedby the totals of the three correspondingvalues for the three mixture primarieswhen the latter stimuli are combined inthe proper ratios. On paper, the colorequations for most spectral matches re-quire the admission of negative valuesfor one of the mixture primaries. Inactual color-mixture experiments, thesenegative values are realized by remov-ing one of the mixture primaries fromthe matching field and adding it to thetest stimulus.

To calculate, for example, the amountsof energy required for a color match toa given wave length A by the mixture ofthe spectral primaries 460 m/*, 530 m/tand 650 imt, let a = the energy at 460m/u, b = the energy at 530 m/n, and c —the energy at 650 m/x. The three equa-tions to be solved for these three un-knowns a, b, and c are then:

= y\

The values for r (or for — r when the re-sponse function is negative, indicatingthat the hue is green rather than red),for y (or for — y when the response isblue rather than yellow), and for w arethen read from the response functions forunit energy for each wave length in ques-tion. (See Fig. 4.) The values r\, y\and WA represent the unit energy responsevalues for any spectral wave length forwhich a color-mixture equation is to becalculated. Solving this set of threeequations for the three unknowns a, b,and c, we then have a color-mixture equa-tion of the form

0460 + #630 + Cm = U

This equation, which is expressed in en-ergy units, may be converted to photo-

a(yw)

000

WAVELENGTH-HV*

FIG. S. Calculated color mixture functionsfor observers H and J and experimental colormixture functions for observers T and W (24,47).

metric units in the usual way by multi-plying each energy value by the relativeluminosity (given by the achromaticresponse function) at the given wavelength.

Color-mixture relations calculated inthis manner for wave lengths A from420 m/t through 700 m/i from smoothedvisual response data for two observersare shown in Fig. 5. The two addi-tional sets of color-mixture functions(for the same three mixture primaries)that are shown for comparison in thefigure are the results of actual color-mixture experiments by W. D. Wrightand L. C. Thomson (24, 47).

Since the relations between the meas-ured response functions and the color-mixture data are, as we have just seen,known for two individual observers, itis now also possible (by assumingspecific spectral loci for the uniquehues) to reverse the procedure andderive opponent-response functions fromthe color-mixture data for Wright andThomson, or for any other observerwhose color-mixture data are available.Since it seems preferable to develop ageneral theoretical model on the basisof a representative average, rather thana small number of individual observers,we have used for the model chromaticand achromatic response functions de-rived from the average color-mixturedata for the CIE international standardobserver (30). These derived functions


are shown in Fig. 6. (The details ofthe derivation are given in 22.) Theyare, of course, smoother and more regu-lar than the individual, measured func-tions, but in other respects they arequite similar (compare Fig. 4).

Photochemical DistributionsThe specific set of a, ft, and y photo-

sensitive absorption functions that havebeen assumed for the theoretical modelare shown in Fig. 7. These curves havenot been measured, and they have theparticular forms shown and the high de-gree of overlap exhibited because of thespecific interrelations that we have pos-tulated a priori between the photo-chemical events and the neural re-sponse activities of the visual opponentmechanisms. Once the photopigmentsactually present in the human retinahave been identified by the biochemists,the visual theorist will have no need tomake such a priori postulates, and thespecific interrelations required betweenthe identified photosensitive materialsand the neural processes underlying thecolor responses can easily be deduced.

I

WAVELENGTH -m/i

FIG. 6. Theoretical chromatic and achro-matic response functions for equal energy spec-trum for CIE average observer.

WAVELENGTH-m/i

FIG. 7. Spectral distribution curves for as-sumed photosensitive materials.

As matters now stand, however, thefunctions shown in Fig. 7 meet thebasic demands of the known facts, andany changes in these theoretical absorp-tion functions that will no doubt be re-quired by results of photochemical re-searches will not importantly affect anyof the basic postulates of the theoreticalmodel. The broadness and similarity ofshape of all three selective functionsthat we have assumed are characteristicof all visual pigments so far identifiedin a variety of animal species (2).

These assumed photopigment distribu-tions do not enter into the considerationof color phenomena, in normal vision,until we come to an examination ofsome of the phenomena of chromaticadaptation in which selective photo-chemical bleaching seems to act as oneof the important determinants. Theother determinants are, of course, thespatial and temporal induction effectsin the neural opponent response proc-esses that have been mentioned earlier.

Dependence of Hue and Saturation onBoth Wave Length and LuminanceWhat are the visual phenomena for

which the model can account? As we


have already indicated, the measuredchromatic and achromatic response func-tions provide a direct and quantifiabledescription of the color sensations evokedby any stimulus of specified wave-lengthcomposition (23). The achromatic,white function is taken as a direct ex-pression of spectral brightness. Spectralhue, which is determined by the chro-matic responses evoked by each wavelength, can also be expressed quantita-tively as a coefficient value relating themagnitude of response of one chromaticsystem to the total of all chromatic re-sponses at that wave length. An exam-ple of such a hue coefficient function fora moderate level of luminance is shownin Fig. 8. It is clear, from the varyingrate of change in the hue coefficientfunction from one region of the spec-trum to the next, that an observer'sability to discriminate among neighbor-ing wave lengths on the basis of huechanges alone will also differ for thedifferent regions of the spectrum. Thisdiscriminative capacity is obviously alsoquantifiable in terms of the amount ofwave-length change required to elicit athreshold change of fixed amount in thevalue of the hue coefficient. Withchange in the luminance at which thespectrum is presented, these coefficient

WAVELENGTH-m/i

Fic. 8. Spectral hue coefficients. Moderateluminance.

functions will be altered, in the sensethat the yellow-blue values will increaseat the higher luminances, and will bediminished at the lower luminances.This is so because, in accordance withthe different energy-vs.-response func-tion postulated for the yellow-blue sys-tem as compared with the red-greenone, as the excitation level is increased,the yellow and blue spectral responseswill be uniformly magnified relative tothe red and green ones at the higherlevels, and uniformly diminished at thelower levels. Although the exact dif-ferential between the two paired sys-tems is not known, under certain cir-cumstances an over-all difference in re-sponse magnitudes of approximately 20per cent seems to occur for a log unitchange in luminance. Thus, at somewave length for which, say, the red andyellow responses are equal at a lumi-nance of 10 mL, the yellow will beabout 20% greater than the red at 100mL, and about 20% less at a luminanceof only 1 mL. If we assume this 20%differential between y-b and r-g re-sponse magnitudes per log unit of lumi-nance change as a reasonable value, andcompute the spectral hue coefficients fora range of approximately three log unitsof luminance variation, then we canspecify the amount of hue shift associ-ated with a change in intensity of anywave length. Conversely, we can alsospecify the wave length changes neces-sary to maintain a constant hue sensa-tion (constant hue coefficient value) asthe luminance is increased or decreased.The latter procedure has been used toobtain the functions shown in Fig. 9,and the curves in the upper part of thefigure are functions measured by Purdyin an actual experiment of this sort(38).

These hue phenomena do not involvethe achromatic response pair at all, anddepend only on the two paired chro-matic response systems. Whatever the


chromatic response to a given stimulus,the perceived color saturation clearlywill also depend on the extent to whichthe white system is simultaneously re-sponding. For any given amount ofchromatic response, the color will obvi-ously appear less saturated if there is alarge magnitude of white response todilute the color, and more saturated ifthe white component of the total re-sponse is relatively small. The per-ceived saturation of the spectrum isalso expressed as a quantitative coeffi-cient function. (See Fig. 10.) Herethe value taken as the saturation coeffi-cient is the ratio of the total chromaticto the chromatic-plus-white responses ateach wave length. The relatively highvalues at the spectral extremes and theminimal value in the pure yellow regionare perfectly consistent both with quali-tative reports and with the experimentaldata on this problem (e.g., 28). Again,as in the hue functions, the rate ofchange of the saturation coefficient fromone spectral region to the next is in-dicative of a varying discriminative ca-pacity with respect to wave length;and, again, the form of the function asshown applies to a moderate luminancelevel and varies in a determinable man-

INWkRIANT POINTS

•THEORETICAL <

WAVELENCTH-m/i

FIG. 9. Constant hue contours as measured byPurdy (38) and as predicted by theory.

II

oo500 6OO 700

WAVELENGTH-m/i

FIG. 10. Spectral saturation coefficients.Moderate luminance.

ner with change in the level of excita-tion in accordance with the differentenergy-vs.-response rates of the threeindependent response systems.

In view of the variations in the hueand saturation functions with change inluminance, we should expect that dis-crimination functions that depend onchanges in these two color attributes,such as discrimination of one wavelength from the next in an equal bright-ness spectrum, would also reflect sucha dependence on luminance. Figure 11shows, in the upper half, a set of wave-length discrimination functions obtainedat two luminance levels by Weale (45).

500 600

WAVELENGTH-m/i

FIG. 11. Wave-length discrimination func-tions for two luminance levels as measured byWeale (45) and as predicted by theory.


The higher values of difference thresholdobtained at the low luminance level maybe explained by a general reduction ofdiscriminative capacity in dim light.The shift of the midspectral maximumtoward lower wave lengths, and therelatively greater heightening of theminimum in the yellow region, cannot,however, be attributed to such a gen-eralized reduction in discriminatory ca-pacity. The selectively greater loss inyellow and blue responses at the low-luminance level that is one of the pos-tulates of our model does, however, ac-count for changes of exactly this sort inthe form of the function. This is shownby the two theoretical functions com-puted from pairs of spectral hue andsaturation functions that are associatedwith the two specified luminance levels.Since brightness is kept constant in suchexperiments, only the hue and satura-tion variables need be considered in ouranalysis of these functions (22).

Chromatic AdaptationThe phenomena that we have treated

thus far all refer to the individual withnormal color vision in a neutral state ofadaptation. What of his color percep-tion after the visual system has beenexposed for some time to a stronglycolored illuminant? For analytical pur-poses, the simplest situation of this sortis the one in which the eye has beenexposed to a large surround field ofgiven color and luminance, and the teststimuli are viewed at the same level ofluminance as the surround. Under thesecircumstances, the three photochemicalreceptor substances will probably haveundergone some selective bleaching, andbecause of the similar brightness of thesurround and test fields, spatial induc-tion effects in the neural response proc-esses will probably be fairly constant.To simplify the treatment for these par-ticular conditions, therefore, we mayignore the constant neural inductions

and consider the photosensitive changesas exercising a controlling influence onthe response systems.

We know that under these circum-stances the color-mixture data do notchange. That is, with uniform chro-matic adaptation, any change in theperceived color of one side of a bipartitecolor-mixture field will also occur onthe other side, and to exactly the sameextent. Thus a color equation that hasbeen made with the eye adapted to aneutral white light will also be a validequation when the eye is adapted to acolored illuminant (15). These impor-tant constancies of color equations meanthat whatever photochemical changesoccur with adaptation must occur in avery specific way. That is, the spectraldistribution functions representing thethree selective photochemicals may beselectively multiplied or reduced by aconstant factor, but no one of them canchange its form (44, pp. 211-212). Inother words, any single substance can-not lose a greater percentage of its ab-sorption at one wave length than it losesat another wave length. Thus, exposureto a colored light can cause any one ofthe postulated photochemical functionsshown in Fig. 7 to be multiplied or di-vided by a constant amount, but this isthe only alteration in the photosensitivefunctions that is consistent with the factthat color equations are invariant withchromatic adaptation.

The extent to which the three sub-stances are selectively attenuated as aresult of exposure to colored light isclearly controlled by the light stimulusitself. That substance which initiallyabsorbs most of the adapting light willsuffer the greatest relative bleaching,and the substance which absorbs rela-tively little of the adapting light willbe relatively little affected by it. Thus,by determining their relative absorp-tions of the adapting light, we can com-pute the relative changes in the heights


of the photosensitive distribution func-tions for the three photopigments thatwe have postulated. Since the excita-tions of the opponent response systemsdepend on these photochemical lightabsorptions (see Fig. 2), we can nowalso determine the forms and magni-tudes of the chromatic and achromaticresponse functions for the new condi-tion of adaptation. In spite of theclose overlap of the photosensitive func-tions that we have postulated, the"adapted" chromatic response functionsdetermined in this way change in strik-ing fashion relative to the functions forthe neutral adaptation condition. Theachromatic function changes too, butrelatively very little. These theoreti-cally computed adaptation changes areconsistent with the kinds of changeknown to occur in situations of thissort. If the eye that has been adaptedto white light is exposed for some timeto a saturated red equal in brightnessto the white, the normally red end ofthe spectrum does not become exces-sively dark, but the amount of rednessseen is strongly reduced, and the greensbecome greatly supersaturated (3, pp.133-137). Also, the wave length thatformerly appeared pure yellow is nowstrongly greenish, and this is also truefor the wave length that formerly ap-peared pure blue. These changes canbe determined from the functions shownin Fig. 12 that have been computed fora given red adaptation, in comparisonwith the functions for the neutral statethat were given in Fig. 6.

From this new set of "adapted" op-ponent response functions the hue andsaturation coefficients and the discrimi-nation data for this new state can alsonow be determined (26).

These "adapted" response functionsare specified, as we said above, fora circumscribed set of conditions forwhich the photochemical adaptationchanges could be taken as primary. As

100 500 600 700

WAVELENGTH -m/i

FIG. 12. Chromatic and achromatic visualresponse functions for red adaptation as pre-dicted by theory.

soon as the relative luminance condi-tions are altered, however, then theneural inductions enter importantly intothe complex picture. For example, ifa test stimulus seen within a stronglycolored (say, red) surround looks neu-tral when its luminance is the same asthat of the surround, then it will prob-ably appear somewhat reddish at ahigher luminance, and the complemen-tary green at a lower luminance (12).(The test stimulus is assumed also tobe of predominantly long wave lengthcomposition.) In terms of opponent in-ductions this phenomenon is readily un-derstood. If the red process excited bythe red surround induces an opponentgreen process in the test area, then atan intermediate luminance this greeninduction is just strong enough to cancelthe red-process activity aroused by thetest stimulus itself. When the teststimulus is made brighter and the redresponse to it increases, the unchanged


green induction from the surround be-comes inadequate to cancel completelythe increased red response to the strongertest stimulus, and the red test hue isnow seen. At a much lower luminanceof test stimulus, the red process is ac-tivated to a much lesser extent, andthe green induction from the surround,which is still unchanged in strength, isnow sufficient to impart a green hue tothe perceived test area. These phe-nomena are not only consistent with theopponent induction postulate, but theyalso make it clear why attempts to treatthe problem of chromatic adaptation ex-clusively as a matter of photochemi-cal bleaching are foredoomed to failure(e.g., 1, 33).

Color Anomalies and Color BlindnessWhen we come to consider individu-

als who do not have normal color visionwe find that their color vision can de-part from the normal in two generalways. Their color perceptions may bedistorted relative to the normal, or theymay exhibit specific color weaknesses orlosses. Also, they may show both typesof these deviant characteristics at thesame time. By distorted color percep-tions we mean, for example, the percep-tions of the particular type of anomalousindividual who has the following char-acteristics: he sees a distinct orange inthe spectral region described normallyas pure yellow or nearly so; he needsthree stimuli for color mixture; hemakes color matches with high pre-cision but uses quite different propor-tions of the mixture stimuli than doesthe normal observer. An individual ofthis type does not seem to have lostany of the efficiency of his neural visualresponse processes, and it seems reason-able to assume that his color distortionshave their basis in the photochemicalcomplex responsible for selective lightabsorption.

The particular assumptions that we

have made concerning the kinds of de-viation that the photosensitive mate-rials may exhibit stem from a generali-zation made by Dartnall (2), on thebasis of his researches concerned withthe identification of visual photopig-ments in a variety of lower organisms.Dartnall has found that when the ab-sorption curves of the various visualpigments are plotted as a function ofthe vibration frequency of the incidentlight (the reciprocal of the more usualwave-length specification), all the ab-sorption curves have very nearly thesame shape, and they can be made tocoincide simply by shifting the curvesso that they all reach an absorptionmaximum at the same frequency. Inother words, a single template repre-senting amount of absorption as ordi-nate, against frequency of radiant en-ergy as abscissa, can be used to fit theabsorption function of any visual pig-ment, whatever the locus of its absorp-tion maximum. It seems reasonable toexpect that this same generalization willapply to the photosensitive distributionsof anomalous individuals with respectto the population of observers with nor-mal color responses. We have conse-quently assumed that, in congenital ab-normalities of the visual system, the nor-mal photopigments can undergo changesthat result in a uniform shift of theentire set of photosensitive distributionfunctions as a group along the fre-quency scale. These shifts are assumedto occur in either of two directions: to-ward higher frequencies (shorter wavelengths) resulting in the type of anoma-lous color vision identified as prota-nomaly, or toward lower frequencies(longer wave lengths) relative to thenormal absorption loci, resulting in thesecond major type of anomalous colorvision known as deuteranomaly. Theamount of these displacements may alsovary in different degrees of congenitalanomaly.


Since the absorption of light by thephotosensitive materials provides thestimulus for the neural chromatic andachromatic response systems, the visualresponse functions thus controlled bythe deviant photosensitive materials willnecessarily be altered, too, and in a sys-tematic manner. Examples of theoreti-cally derived anomalous response func-tions based on these assumptions aregiven in Fig. 13. The set of functionsin the center block are those for the ob-server with normal photosensitive ma-terials; those in the upper block are fora protanomalous type whose visual pig-ment absorptions are assumed to beshifted toward the shorter wave lengthsby an amount equal to about 15 mju,

UJin

>

$

0.0

1.0

1.0

6.0

iX>

1.0

O.O

400 500 600 700

WAVELENGTH -mfl

FIG. 13. Theoretical chromatic and achro-matic response functions for equal energyspectrum. For observers with protanomalous,normal, and deuteranomalous photoreceptorsystems and with normal strength visual re-sponse processes.

from the normal peak of about 550 m/*.This type of individual will have a"luminosity function (described by theachromatic, white response function)that peaks at a shorter wave lengththan the normal and will show consid-erable loss of luminosity at the red endof the spectrum (48, Ch. 25). Thespectral hues will also be altered, witha distinctly reddish yellow occurringwhere the normal individual sees aunique or pure yellow, whereas theprotanomalous observer's pure yellowoccurs at a wave length described bythe normal as quite greenish. In makingcolor matches, such as a match between589 mju, on one side of a bipartite fieldand a mixture of 530 rm<, and 670 m^on the other, this observer will requirea much greater proportion of 670 m^u. inthe mixture than will the average ob-server with normal color vision (27, 46).This particular match, the Rayleighequation, is the earliest and best knowndiagnostic test for anomalous color vi-sion. In this same test, the anomalousindividual whose response functions areshown in the lower block in Fig. 13 willdeviate from the normal in the oppositeway; that is, he will require a muchgreater proportion of 530 m/j, in themixture for the Rayleigh equation (46).This type of anomalous individual (deu-teranomalous) is assumed to have pho-topigment absorptions that are shiftedtoward the longer wave lengths, and hewill see greenish-yellows where the nor-mal sees yellow, yellows where the nor-mal sees orange, etc. Since the neuralresponse processes of both types ofanomalies of this sort are assumed tobe operating at the normal efficiency,these individuals will show high pre-cision in making their distorted colormatches, and their discriminatory ca-pacities will also be good. As a matterof fact, anomalous individuals of thissort have understandably high conn-

OPPONENT-PROCESS THEORY OF COLOR VISION 399i.O

3>K +_J

>UlP 1.0

^^

J

/^\

/_ ^,/

*N*

§

+

1 /}

J^\ D

WAVELENGTH-m/i

FIG. 14. Theoretical chromatic and achro-matic response functions for equal energyspectrum. For observers with protanomalous,normal, and deuteranomalous photoreceptorsystems, and with impaired red-green responseprocesses.

dence in their own color capability, andthey are extremely resistant toward ac-cepting the results of diagnostic testswhich indicate that their visual capaci-ties are deviant from (with the impli-cation of "inferior to") those of thenormal population (36, pp. 235-238).

Not all anomalous individuals are asfortunate as the types shown in Fig. 13,however. Many give evidence of realcolor weakness, in addition to distor-tions of the kinds already discussed(40). These color-weak individualsseem to have been deprived of someof the efficiency of the neural responseprocesses, particularly of the red-greenopponent pair, and their systems may

be represented in terms of the theoryby the kinds of response functions givenas examples in Fig. 14. The visual pig-ments of these three types of individualsare taken to be the same as those shownin the preceding figure, respectively, butthe red-green paired system is reducedto one-tenth of the normal strength.Such observers have real losses in colordiscrimination in addition to possiblecolor distortions, and their color matchesare imprecise as well as deviant. Indi-viduals with congenitally abnormal colorsystems are frequently of this generaltype, and cases of acquired color blind-ness caused by degenerative disease in-variably show this kind of color weak-ness at some stage in the developmentof the neural disorder (31).

When the weaknesses become extreme,whether in congenital or acquired dis-orders, the red-green system seems tobe entirely lost to normal function, anda condition of dichromasy, or so-called"color-blindness," results. That is, thevisual system becomes a two-variable

WAVEUENGTH-m/i

FIG. IS. Theoretical chromatic and achro-matic response functions for equal energyspectrum. For observers with nonfunctioningred-green response processes.


one, as shown in Fig. IS. Here theyellow-blue and the white-black neuralsystems remain intact and functioning,but there is no red-green response func-tion. If the red-green loss occurs with-out changes in the visual pigments, theremaining yellow-blue and white-blackresponse functions are like those of thenormal individual; but, since there isno red-green system, the spectrum isdivided into only two hue sectionsfor these individuals. The short wavelengths which normally vary from violetthrough blue and blue-green to puregreen all appear as blue, but of varyingsaturations, with a neutral region wherethe normal pure green occurs. Beyondthis wave length the remainder of thespectrum appears yellow, in varyingsaturations, out to the extreme long-wave limit of visibility. The luminosityfunction is the same as for the observerwith normal color vision. Individualswho fit this response pattern would beclassified as deuteranopes (29). If thevisual pigments are altered, so as toproduce an absorption shift toward theshort wave lengths in addition to thecomplete red-green neural loss, then thespectrum is again divided into a short-wave blue and a long-wave yellow sec-tion, but the neutral region that dividesthe spectrum into the two major huesoccurs at a shorter wave length than forthe deuteranopes. The luminosity func-tion is also displaced in this type ofdichromasy, as it is for the anomalousindividuals with similar photopigmentchanges, and the type of "color-blind"vision associated with this pattern iscalled protanopia (29).

These two theoretically assumed kindsof deviation from the normal system—i.e., photopigment changes and neurallosses or weaknesses of the paired red-green response system—permit us toassemble a systematic picture of themany various manifestations of abnor-

mal red-green vision that defy under-standing in terms of any model of thevisual system that assumes a one-to-one correspondence between light ab-sorption in the retinal receptors andthe resulting color sensations (22, 27).

Defects or losses may also occur inthe yellow-blue neural response system,although such defects seem to be muchmore rare than the red-green defects.Again, these yellow-blue neural lossesmay take place either with or withoutchanges in the photosensitive materialsin the retina. Examples of the remain-ing red-green and white-black responsefunctions in two types of yellow-blueblindness are given in Fig. 16. In eachtype of this disorder, the yellow-blueneural response function is missing, andthe total gamut of colors for these in-dividuals includes only neutral and redsand greens of various saturations. Ifthere is no simultaneous photopigmentdisorder, there are two neutral points inthe spectrum, one in the region wherethe normal sees a pure yellow, and an-

WAVELENGTH-m/i

FIG. 16. Theoretical chromatic and achro-matic response functions for equal energyspectrum. For observers with nonfunctioningyellow-blue response processes.


other in the region where the normalsees a pure blue. Yellow-blue blindnessof this sort is called tetartanopia, andonly a few cases of it have been re-ported in the literature (e.g., 34, pp.68-92). Slightly more common is thesecond type of yellow-blue blindness,known as tritanopia (49), in which notonly the neural yellow-blue system islost, but also the short-wave photopig-ment seems to be missing. Observersof this type have a neutral point inthe normally yellow-green region of thespectrum, but there is no second neu-tral point, and the green hues extendinto the short-wave region that appearsviolet to the person with normal colorvision.

For all these types of deviant colorvision, calculation from the theoreticalspectral response functions of discrimi-nation curves, color mixture equations,and other psychophysical relations arein good agreement with the experimentaldata that are available for the variouskinds of defective color systems (22,27).

Opponents-Theory and NeurophysiologyThe conceptual model for the oppo-

nent-colors theory as originally presentedby Hering drew its sharpest criticism onthe grounds of being bad physiology.Some of this criticism was based onan erroneous interpretation of Hering'sviews, an interpretation that incorrectlyassigned the opponent processes to thephotochemical activities in the retinalcells. Hering's own concept of mutu-ally opponent neural processes, eachcapable of being activated by exter-nal stimulation, was also, however, farahead of the knowledge of neurophysi-ology at the time it was proposed (16).But this concept now turns out to beperfectly consistent with the picture ofneural function that is only just re-cently beginning to build up from elec-

trophysiological studies of the visualneural apparatus.

It has become clear that nerves donot simply respond or fail to respondwhen a stimulus is presented to the ap-propriate end-organ. Rather, they mayrespond according to any of a numberof quite specific patterns. For example,a nerve fiber may (a) discharge at theonset of stimulation and subsequentlygradually become quiet; (b) dischargeat both onset and cessation of stimula-tion with a quiet period in between; or(c) cease any spontaneous activity whenfirst stimulated and during continuedstimulation, but respond with a burst ofelectrical impulses when the stimulusceases to act (7). The on- and off-phases of discharge are mutually in-hibitory processes, they are associatedwith slow electrical potentials of oppo-site sign, and they cancel each otherwhen the experimental conditions areso manipulated as to cause both on-and off-discharges to impinge simultane-ously on the same ganglion cell (6). InGranit's opinion (6), the evidence fromelectrophysiology provides a "belatedvindication of Hering's view" that thevisual system is characterized by mutu-ally opponent neural processes.

The concept of mutual interactionamong the various elements of thephysiological field is also basic to thetheory and is critical to an understand-ing of both areal effects and simultane-ous contrast phenomena. Here again,we find the researches in electrophysi-ology indicating that individual nerveelements never act independently, andthat visual function must be thought ofin terms of the integrated action of allthe units of the neural visual system(8). Hartline (9) has found that, evenin the very simple Limulus eye, the dis-charge of impulses in any one opticnerve fiber depends not only upon thestimulus to the specific receptor unit


from which that fiber arises but alsoupon the stimulation over the entirepopulation of mutually interacting ele-ments. Both excitatory and inhibitoryinteractions of the sort to be expectedby theory have actually been demon-strated in the neural responses of thevertebrate visual system by Hartline(8), Kuffler (32), and Granit (6).

The way in which the postulatedthree independent systems of pairedopponent processes (y-b, r-g, w-bk)are differentiated neurally is still amatter for conjecture. Hering thoughtit was a matter of process specificity,but was willing to use the concept ofmaterial, or structural, specificity, whichhe guessed would be more readily com-prehended by most interested readers ofhis views at the time. Our own theo-retical preference at this time is theconjecture that a particular color qual-ity is more probably determined by aparticular state of the nervous tissuethan by activity of a particular struc-tural element in the nervous network.Thus, we would be inclined to look fora difference between yellow-blue vs. red^green processes, rather than toward iso-lation of yellow-blue or red-green fibersor nerve cells.

SUMMARY

This paper has presented a summaryof our progress to date in providing aquantitative formulation for the Heringopponent-colors theory, and in relatingthe postulated visual mechanism to spe-cific problems of color sensation, colormixture and color discrimination; to thedependence of these functions on thephysical variables of both stimulus wavelength and energy level; to their fur-ther dependence on adapting and sur-round stimulation; and to the changesin these functions that occur in variouskinds of abnormal color vision. It is

our conclusion that the opponent-colorstheory serves as a fruitful working hy-pothesis by bringing a systematic co-herence to the mass of isolated colorphenomena that have been reported andsubjected to quantitative experimentthroughout the years. The physiologi-cal concepts basic to the theory arealso shown to be consistent with recentfindings in neurophysiology.

REFERENCES

1. BREWER, W. L. Fundamental responsefunctions and binocular color matching./. Opt. Soc. Amer., 1954, 44, 207-212.

2. DARTNALL, H. J. A. The interpretationof spectral sensitivity curves. Brit,med. Bull., 1953, 9, 24-30.

3. EVANS, R. M. An introduction to color.New York: Wiley, 1948.

4. FARNSWOETH, D. Tritanomalous vision asa threshold function. Die Farbe, 19SS,4, 185-196.

5. GRAHAM, C. H. Vision: III. Some neuralcorrelations. In C. Murchison (Ed.),A handbook of general experimentalpsychology. Worcester: Clark Univer.Press, 1934. Pp. 829-879.

6. GRANIT, R. Receptors and sensory percep-tion. New Haven: Yale Univer. Press,1955.

7. HARTLINE, H. K. The response of singleoptic nerve fibers of the vertebrate eyeto illumination of the retina. Amer. J.PhysioL, 1938, 121, 400-415.

8. HARTLINE, H. K. The neural mechanismsof vision. Harvey Lectures, 1941-42,37, 39-68.

9. HARTLINE, H. K., WAGNER, H. G., & RAT-LIFF, F. Inhibition in the eye of limu-lus. /. gen. PhysioL, 1956, 39, 651-673.

10. HARTRIDGE, H. The polychromatic the-ory. Docttmenta Ophthal., 1949, 3,166-193.

11. HECHT, S. Vision: II. The nature of thephotoreceptor process. In C. Murchison(Ed.), A handbook of general experi-mental psychology. Worcester: ClarkUniver. Press, 1934. Pp. 704-828.

12. HELSON, H. Fundamental problems incolor vision. I. The principle 'govern-ing changes in hue, saturation, andlightness of non-selective samples in


chromatic illumination. /. exp. Psy-chol., 1938, 23, 439-476.

13. BERING, E. Zur Lehre vom Lichtsinne.Berlin, 1878.

14. HERINO, E. Zur Erklarung der Farben-blindheit aus der Theorie der Gegen-farben. Lotos, Jb. f. Naturwiss., 1880,1, 76-107.

15. HERING, E. Ueber Newton's Gesetz derFarbenmischung. Lotos, Jb. f. Natur-wiss., 1887, 7, 177-268.

16. HERING, E. Zur Theorie der Vorgange inder lebendigen Substanz. Prague: 1888.(English translation by F. Welby, inBrain, 1897, 20, 232-2S8.)

17. HERING, E. Grundziige der Lehre vomLichtsinn. Berlin: Springer, 1920.

18. HOCHBERG, J. E., TRIEBEL, W., & SEAMAN,G. Color adaptation under conditionsof homogeneous visual stimulation(Ganzfeld). /. exp. Psychol, 1951,41, 153-159.

19. HURVICH, L. M., & JAMESON, DOROTHEA.The binocular fusion of yellow in rela-tion to color theories. Science, 1951,114, 199-202.

20. HURVICH, L. M., & JAMESON, DOROTHEA.Spectral sensitivity of the fovea. I.Neutral adaptation. /. Opt. Soc. Amer.,1953, 43, 485-494.

21. HURVICH, L. M., & JAMESON, DOROTHEA.A quantitative theoretical account ofcolor vision. Trans. N. Y. Acad. Sci.,1955, 18, 33-38.

22. HXJRVICH, L. M., & JAMESON, DOROTHEA.Some quantitative aspects of an op-ponent-colors theory. II. Brightness,saturation, and hue in normal and di-chromatic vision. /. Opt. Soc. Amer.,1955, 45, 602-616.

23. HURVICH, L. M., & JAMESON, DOROTHEA.Some quantitative aspects of an op-ponent-colors theory. IV. A psycho-logical color specification system. /.Opt. Soc. Amer., 1956, 46, 416-421.

24. ISHAK, I. G. H. Determination of thetristimulus values of the spectrum foreight Egyptian observers and one Brit-ish observer. /. Opt. Soc. Amer., 1952,42, 844-849.

25. JAMESON, DOROTHEA, & HURVICH, L. M.Some quantitative aspects of an op-ponent-colors theory. I. Chromatic re-sponses and spectral saturation. /. Opt.Soc. Amer., 1955, 45, 546-552.

26. JAMESON, DOROTHEA, & HURVICH, L. M.Some quantitative aspects of an op-

ponent-colors theory. III. Changes inbrightness, saturation, and hue withchromatic adaptation. /. Opt. Soc.Amer., 1956, 46, 405-415.

27. JAMESON, DOROTHEA, & HURVICH, L. M.Theoretical analysis of anomalous colorvision. /. Opt. Soc. Amer., 1956, 46,1075-1089.

28. JONES, L. A., & LOWRY, E. M. Retinalsensibility to saturation differences. J.Opt. Soc. Amer., 1926, 13, 25-34.

29. JUDD, D. B. Current views on colourblindness. Documenta Ophthal., 1949,3, 251-288.

30. JUDD, D. B. Basic correlates of the visualstimulus. In S. S. Stevens (Ed.), Hand-book of experimental psychology. NewYork: Wiley, 1951. Pp. 811-867.

31. KOIXNER, H. Die Storungen des Parben-sinnes. Berlin: S. Karger, 1912.

32. KUEFLER, S. W. Discharge patterns andfunctional organization of mammalianretina. /. Neurophysiol., 1953, 16, 37-68.

33. MACADAM, D. L. Chromatic adaptation.J. Opt. Soc. Amer., 1956, 46, 500-513.

34. MUIXER, G. E. Darstellung und Erklarungder verschiedenen Typen der Farben-blindheit. Gottingen: Vandenhoeck andRuprecht, 1924.

35. PARSONS, J. H. An introduction to thestudy of colour vision. (2nd ed.)Cambridge: Cambridge Univer. Press,1924.

36. PICKFORD, R. W. Individual differences incolour vision. London: Routledge andKegan Paul, 1951.

37. PIERON, H. La dissociation de 1'adapta-tion lumineuse et de 1'adaptation chro-matique et ses consequences theoriques.Annie psychol., 1939, 40, 1-14.

38. PURDY, D. M. The Bezold-Briicke phe-nomenon and contours for constant hue.Amer. J. Psychol, 1937, 49, 313-315.

39. RIGGS, L. A., RATinT, F., CORNSWEET,JANET C., & CORNSWEET, T. N. Thedisappearance of steadily fixated visualtest objects. /. Opt. Soc. Amer., 1953,43, 495-501.

40. ROSMANIT, J. Anleitung zur Feststellungder Farbentuchtigkeit. Leipzig: Deu-ticke, 1914.

41. TROLAND, L. T. Apparent brightness: itsconditions and properties. Trans. Ilium.Engr. Soc., 1916, 11, 957-966.

42. TROLAND, L. T. The principles of psycho-physiology. Vol. 2. Sensation. NewYork: D. Van Nostrand, 1930.


43. VON BEZOLD, W. Ueber das Gesetz derFarbenmischung und die physiologischenGrundfarben. Ann. Phys. u. Chem.,1873, 150, 221-247.

44. VON KRIES, J. Die Gesichtsempfindungen.In W. Nagel (Ed.), Handbuch derPhysiologic des Menschen. Brunswick:Vieweg, 1905. Pp. 109-282.

45. WEALE, R. A. Hue-discrimination in para-central parts of the human retina meas-ured at different luminance levels. /.Physiol., 1951, 113, 115-122.

46. WILLIS, MARION P., & FARNSWORTH, D.Comparative evaluation of anomalo-

scopes. Med. Res. Lab. Rep. No. 190,1952, 11, No. 7, 1-89.

47. WRIGHT, W. D., & PITT, F. H. G. Thecolour-vision characteristics of two tri-chromats. Proc. Phys. Soc. (London),1935, 47, 205-217.

48. WRIGHT, W. D. Researches on normaland defective colour vision. St. Louis:Mosby, 1947.

49. WRIGHT, W. D. The characteristics oftritanopia. J. Opt. Soc. Amer., 1952,42, 509-521.

(Received for early publication July 1, 1957)

Documents

AN OPPONENT-PROCESS THEORY OF COLOR VISION · 2019-02-27 · color theory" remind us more and more of its chief theoretical rival, the Bering theory of three paired, opponent color