[72-Science]

J. J. McCann,

. Rod-Cone Interactions:Different Color Sensations from Identical Stimuli",

Science, 176, 1255-1257, 1972.

Copyright AAAS

Reprinted from16June 1972, Volume 176, pp. 1255-1257 SCIENCE

Rod..Cone Interactions: Different Color

Sensations from Identical StinluliJohn l \feCann

Rod-Cone Interactions: Different Color

Sensations from Identical Stimuli

Abstract. Different color sensa/ions were generated by two areas in a complexscene, even though both areas sent to the eye the same 656-nanometer radiancethat excited the long-wave cones and the same 546-nanometer radiance thatexcited only the rods.

Land (1) set up an experiment inwhich one area appeared red and an­other area appeared green even thoughthe radiations coming from both areaswere identical. The experimental dis­play was a pastel chalk drawing byJeanne Benton, called the Street Scene,which included a green awning on theleft side and a red door on the rightside. An incandescent lamp with a long­wave filter was placed on the floor nearthe left side of the display and a secondlamp with a middle-wave filter wasplaced on the right side. The proceduremade use of the inverse square law tocontrol the relative amounts of lightthat fell on the various parts of thedisplay and hence the radiance thatcame to the observers' eyes from partic­ular areas. In Land'~ experiments andin those discussed in this report the fluxwas measured from the central portionof the particular area. In each case,the change of flux from side to sidewas small because the areas were small.Each area appeared to be uniform inlightness and color. The location of thelong-wave light was adjusted so thatthe same long-wave radiance came tothe eye from the center of the awningas from the center of the door. Similar­ly, the location of the middle-wave lightwas chosen so that the same 'middle­wave radiance came from the center ofthe door as from the center of the awn­ing. Since the long- and middle-wavelamps were the only sources of radia­tion, the total flux that came from thecenter of the awning was identi::al tothat from the center of the door. This ifurther implied that the retinal recep-'tors stimulated by tbose areas musthave had identical inputs, since the·radiance that came from those areaswas identical. The observen>, hcwever,reported that the awning and the doorproduced sensations that were far fromidentical in longo, in middle-, and incombined long-plus-middle-wave illumi­nations (2). In long-wave light the doorappeared very light, and the awningwas very dark, near black. In middle­wave light the door was very dark andthe awning was very light. In com­bined long- plus middle-wave illumi-

nation the awning was green and thedoor was red even though the radiationfrom the awning was identical to thatfrom the door.

Land, in his retinex theory, proposedthat the color of objects is determined bythe lightnesses in different wave bands(3). He proposed that the differentkinds of receptors acted as if they wereindependent sets to form different light­ness images for different regions of thevisible spectrum. These lightness im­ages are compared to produce colorsensations. In the Street Scene experi·ment, for example, the long-wave in­formation is processed by one retinexto form one lightness image and themiddle-wave information is processedby a different retinex k) form a secondlightness image (4). Areas that arelighter in long-wave and dark in middle­wave light look red regardless of theradiance-wavelength distdbution of theflux comi'ng from them. Areas that looklight in middle-wave and dark in long­wave light always look green.

Experiments have shown that colorsensations can be generated by the in-

Fig. 1. Photographs of the Street Scenetaken with the 656-nm (top) and 546-nm(bottom) filters in uniform illumination.

teractions of rods and cones (5, 6).McCann and Benton found that whena stimulus that excited only the rodswas combined with a stimulus that ex­cited the long-wave cones, observerssaw colored images, which, except forbrightness and sharpness, were indis­tinguishable from images seen on twowavebands entirely above cone thresh­old.

Duplicity theory (7) suggests thatthe rods and cones are completely dif­ferent systems: one for low-radiance,colorless vision and the other for high­radiance, color vision. The many dif­ferent properties of images produced byrods and cones have accumulated sub­stantial support for the idea that therod and cone processes are different (8).Nevertheless. MoCann and Benton'sexpcriments showed that the lightnessesproduced by these different processescan be compared to form color sensa­tions. Repeating Land's Street Sceneexperiment with only rods and long­wave cones provides a test of the hy­pothesis that color depends on thelightnesses produced by independentsystems. Is it possible, in a compleximage that excites only the rods, to pro­duce two significantly different light­nesses from areas simultaneously send­ing the same radiance to the eye? Is itpossible to combine such a rod imagewith an image secn by the long-wavecones, which has ditTerent lighmessesproduced by areas having the same ra­diance? Do these areas produce differ­ent color sensations? Are the sensationsconsistent with the hypothesis thatcolors are determined by the compari­son of lightness?

I set up the Street Scene display andrepeated Land's experiment with onlythe rods and long-wave cones. I thenperformed a second similar experimentwith a different display and differentcolored areas. In both experiments Iused two projectors with nanowJbandinterference filters, one 656 nm, theother 546 nm, to illuminate the entiredisplay. I used a series of square pho­tographic transparencies called con­tinuous wedges, which increase indensity in one direction. These transpar­encies allowed the use of projectors assources of controlled nonuniform il­lumination. Variable transformers wereconnected to the projectors to controlthe brightness of the lamps. The trans­miss~on bands of the filters were sufli­cienlly narrow so that change in colortemperature of the lamp with changesin transformer settings did not signift-

Table 2. Munsell chips chosen by the ob·servers to match with the door and theawniog in Street Scene in experiment 1.

Table l. Radiance measurements in 656-nmand 546-nm illumination in experiment 1.

Radiances (watts sr' moo')

Observer Door Awning

tvlAW 5.0 R 4/14 7.5 G 3/4

JAH 5.0 R 4/12 10.0 GY 3/10

TT 5.0 R 4/9 5.0 G S/10JJS 7.5 R 3/10 7.5 G 3/2

JMC 5.0 R 3/12 5.0 G 3/6WW 5.0 R 4/18 5.0 G 5/10

cantly affect the wavelength distribu­tion of the light.

Figure 1 is a pair of photographs ofthe Street Scene taken with the 656-nmand 546-nm filters in uniform illumi­nation. The rectangular display sub­tended 36° by 32 0 and. the door andawning had diagonals that subtendedroughly 8° and 6°, respectively. Thedoor reflected a high percentage oflong-wave radiation and a low percent­age of middle-wave radiation. A wedgetransparency was chosen for one pro­jector so that the same radiance at 656run came to the eye from the awningand the door. Similarly, another wedgewas placed on the projector so that thesame radiance at 546 nm came to theeye from the awning and the door.Therefore, when both projectors wereturned on together, exactly the samecomposite stimulus was coming fromthe center of both the awning and thedoor.

The remaining problem was to ad­just the radiance from the display sothat it excited only the rods and long­wave cones. Since the relative radianceat various points in the display wascontrolled by the position of the wedgetransparencies, we could adjust theoverall output of the lamps withoutdisturbing the radiance identity of theawning and the door. In earlier work(6), McCann and Benton performedthree different experiments to deter­mine the upper limit of rod sensitivitythat was still below cone threshold. Thefirst experiment measured the recoveryof threshold sensitivity to 546-nm and450-nm light following light adaptation.The plot of threshold versus time inthe dark is called a dark adaptation

546 nm

1.0 X 10.....4.9 X It}-"6.9 X 10-"6.9 X 10-1

656nm

3.0 X 1()--71.6 X 10-"5.2 X to-7

5.2 X to-7

Observer Left Right

TT 2.5 R 7/5 5.0 R 3(3

JAH 2.5 R 8/4 5.0 R 2/6py 5.0 YR 7/8 7.5 R 3/6JMC 2.5 Y 7/8 10.0 R 3/6MAW 7.5 Y 8/4 to.O R 3/6

JJS 2.5 Y 8/6 2.5 YR 3/6

Cone thresholdBrightes t areaLeft areaRight area

Table 4. Munsell chips chosen by the ob·servers to match the left and the right areasof a "Mondrian" in experiment 2.

Table 3. Radiance measurements in 656-nmand 546-nm illumination in experiment 2.

Radiances (walls sri m....)

threshold response from the middle­wave cones. Table 1 shows for bothwavelengths the cone threshold radi­ances, the highest radiances from theentire scene, and the radiances fromthe awning and the door. These radi­ances show that the entire scene wasbelow cone threshold for 546-nm lightand slightly above long-wave conethreshold for 656 nm.

Six observers were asked to viewthe display first in long-wave, then inmiddle-wave illumination. The observ­ers were asked to describe the light­nesses of the awning and door aswhite, light gray, medium gray, darkgray, or black. In 656-nm illuminationthe observers were asked to describethe lightnesses as grays even thoughthe entire display appeared to be cov­ered with a red wash. In 546-000 il­lumination the entire display appearedcolorless grays, since it was below conethreshold. The observers reported thatthe awning was dark gray in 656-nmillumination and light gray in 546-nmillumination. They further described thedoor as being light gray in 656 nmand dark gray in 546 nm. The observ­ers were then asked to describe thecolor sensations of the' two areas in thecombined long- plus middle-wave il­lumination. They reported that the sen­sations from those two areas were dif­ferent from each other-green andred-even though tbe centers of theareas sent identical physical stimuli tothe eye. Just as in Land's experimentsentirely above cone threshold, the simul­taneously identical sets of radiancesproduce very different color sensations.Again, the apparent lightnesses of theawning-light gray in middle-wave,

curve. The test area was a 4° squareplaced in the center of a Mondrian-likedisplay. The rest of the display wascovered by a curtain so that the ob­servers saw only the 4° square thathad the highest reflectance in the entiredisplay. The transition from cones torods, shown by the break in the darkadaptation curve, was at 1 X 1O-5 wattper stradian (sr) per square meter for546 nm. This result was checked in thesecond experiment by asking the ob­server to look for the rapid change insharpness and the change in color sen­sation associated with the transitionfrom rods to cones at 546 nm. Thisapparent change agreed closely with thebreak in the dark adaptation curves.Using the value for cone threshold de­termined by the first two experiments,the third test combined the 546- and450-nm images both above cone thresh­old. The scene produced many differ­ent color sensations: blues, greens, yel­lows, and browns. The 546-nm and the450-nm illuminations were separatelyreduced to below the measured conethresholds. The combination of thesetwo illuminations produced no varia­tion of color, just a small increase ofapparent brightness. This is what wewould expect if both the images wereexciting only the rods.

In the experiments for this report, Iused the 546-nm radiance determinedby those three different techniques. Ialso performed a control experiment toinsure that radiances below I X 10- 5

watt sr- 1 m- 2 were below middle­wave-cone threshold in the conditionsdescribed in this report. Three of thesix observers remeasured the recoveryof threshold sensitivity after lightadaptation, using procedures describedearlier (6). They looked at an 8° squarearea that was illuminated by the 546­nm light used in the experiment. TheStreet Scene was removed from thelarge black easel and a piece of whitepaper was placed in the middle of theeasel. This insured that the illumina­tion falling on the white paper wasidentical to that falling on the centralareas of the Street Scene. For two ob­servers the break in the dark adaptationcurve was at 1 X 10-5 watt sr- 1 m- 2 .

For the third observer the break ap­peared at 7.0 X 10-0 watt sr- 1 m-2•

The brightest area in the Street Scenein 546-noo light was 4.6 X 10-6 wattsr- 1 m- 2• The awning and the doorhad a radiance of 1.4 X 10- 6 watt sr- 1

m-2. In other words, the awning and thedoor sent only one-seventh the amountof 546-nm light necessary to obtain a

546nm

1.0 X 10-64.9 X 10....1.4 X It}-"1.4 X 10.....

656 nm

3.0 X 10-7

2.2 X I()-<'6.9 X 10-7

6.9 X 10-7

Cone thresholdBrightest areaDoorAwning

dark gray in long-wave flux-eorre­spond to the sensation green. The light­nesses of the door-light gray in long­wave, dark gray in middle-wave flux­correspond to the sensation red.

To provide a better description ofthe color sensations, the observers wereasked to study the awning and the door,and then to select from the MunsellBook of Color (9) a chip that mostclosely matched their memory of theawning and the door. The MunsellBook of Color was viewed with a Stan­dard Illuminant C in an otherwise dark­ened room. If the observers could notfind a suitable chip, they were i..n:­structed to estimate the designation ofa chip that would match the awning orthe door. The Munsell chips chosenare listed in Table 2.

In the second experiment the StreetScene was replaced by a multicoloredpaper display called a "Mondrian." Thisdisplay was made of matte-surface pa­pers, cut and overlapped to formsquares, rectangles, polygons, and othershapes. There were 23 papers and theentire display subtended roughly 36 0 ona side. The areas studied were rectangu­larand their diagonals subtended rough­ly 8°. New wedge transparencies werechosen so that the same flux came tothe eye from these two areas in 656-nrnlight and 546-nm light. As in the firstexperiment, the entire display was be­low cone threshold for 546 om. Table3 shows that the two areas studied bythe observers were one-sixteenth theamount of light necessary to obtain athreshold respanse from the middle­wave cones.

Six observers were once again usedfor this experiment. They were askedto describe the lightnesses of the twoareas as white, light gray, middle gray,dark gray, or black, first in long-waveand then in middle-wave light. As inthe previous variation of the experi­ment the observers were asked to findthe Munsell chips that most closelymatched the two areas. The Munsellchips chosen are shown in Table 4. Theobservers reported that the area onthe left in 656-nm illumination waslight gray and that in 546-nm illumi­nation was also light gray. The area onthe right was middle gray in 656-nmlight and dark in 546-nm light. In com­bined 656- and 546-nm light the ob­servers reported that the area on theleft side was yellow and the area onthe right was dark brown. Again theidentical sets of radiances would notpredict this outcome, namely, that the

color sensations will be different. How­ever, above cone threshold, yellow areasalways appear light in both long- andmiddle-wave flux. Brown objects alwaysappear dark in both wave bands, butsomewhat lighter in long-wave flux.

In two similar experiments the il­lumination was controlled so that thedisplays were below cone threshold for546-nm light and slightly above conethreshold for 656-nm light. The observ­ers saw a variety of colors from theinteraction of rods and long-wavecones. In each wave band the illumina­tion was controlled so that the sameamount of light came from the particu­lar low-reflectance objects that ap­peared dark and high-reflectance ob­jects that appeared light. When the rodsand long-wave cone images were com­bined, the observer reported differentcolor sensations-red versus green andyellow versus brown-from areas thatwere identical physical stimuli. The colorsensations produced by these rods andlong-wave cone interactions support thehypothesis that lightnesses on independ­ent systems determine color sensations.

JOHN J. MCCANNVision Research Laboratory,Polaroid Corporation,Cambridge, Massachusetts 02139

Refereaces llIId Nole.

J. E. H. Land, WUliam James Lature, p~sented al Harvard University, 2 November 1966.

2. For discussion of mechanisms responsible foridentical radiances producing different light­nesses. see E. H. Land and J. J. McCaIlII,J. Opt. Soc. A mer. 61, 1 (t971).

3. E. H. Land, Ame,. Sci. 51. 247 (1964).4. In these experiments I always worked with

two wave bands of iDumination and henceI only discuss two lightness scales. Land'sretlnex theory propOses that there are threeor four independent lightness images. Inother experiments Land has used three pro­jectors with the Street Scene experiment. Theposition of the short-wave footlight waschosen so that the same short·wave flux camefrom both areas. Both the awning and thedoor appeared dark In short-wave light.

5. H. Blackwell and 01. Blackwell, VLrlo" Res.I, 62 (t961); B. Stabell, Scand. 1. Psycho!. 8,132 (1967).

6. J. J. McCann and J. Benton, J. Opt. Soc.Amer. 59. 103 (1969).

7. M. PJrenne, In The Eye, H. Davson, Bd.(AcademJc Press, New York, 1962), vol. 2,p. 24.

8. The rods and cones have dl1ferent shapes anddistnllution over the retina [M. Schultze,Ad,·an. aphtha/mol. 9. I (l866H. differentrales of dark adaptation [5. Hecht, Physlo/'Rev. 17, 239 (1937)1. dlfferenl spectral sensi­tivities [G. Wald, Scie"ce 101, 653 (1945)J,different acuitY properties [5. Shlaer, J. Gen.Physio/. 11, 165 (l937)J, different ftickcr £U.sian properties [5. Hecht and C. Vernjp,P,oc. NOI. A cad. Sci. U.S. 19, 522 (1933)),and independetl sensltlzing and desensltlzingrelinal interaet!ons 1M. Alpern, J. Physio/.London 176. 462 (1964); G. Westheimer. ibid.206, 109 (1969»).

9. MWlseU Book of Co/a, (Munsell Color Co.,Baltimore, 1942).

10. I thank E. H. Land for comments and adviceand Marie Watson for help with the experI­ments and the repOrt.

29 November 1971

Copyright© 1972 by the American Association for the Advancement of Shence