IN 1965 CELESTE MCCOLLOUGH OF Oberlin College reported a puzzling phenomenon: a situation in which the human visual system imposes color on a black-and-white grating. You first study a grating of black stripes interspaced with a color. Later you look at a black-and-white grating identical in orientation and spacing with the first one. The grating's white stripes seem to be tinted with the complement of the color in the first grating.

The apparent color is surprising for several reasons. It appears only if the second grating has the same orientation and spacing as the first one. Although the apparent color may grow fainter with time, it appears even if the observer delays viewing the second grating for hours, days or weeks. (Its strength after a delay depends partly on the dietary and sleeping habits of the observer.)

A person who has normal color vision should be able to see the McCollough coloration by viewing the gratings below. Begin with the colored gratings on the left. Note that the black stripes of the horizontal gratings are interspaced with green and those of the vertical gratings are interspaced with magenta. Be sure the illustration is well illuminated. Do not fix your gaze but shift it so that you see each differently colored region for about the same amount of time. After five minutes or more examine the colorless gratings in Figure 3. The demonstration works better if the illumination is now low. You should perceive faint, unsaturated colors superposed on the white stripes.

The colors you perceive are associated with the orientation of the black stripes. Whereas previously the horizontal stripes were interspaced with green, they are now interspaced with magenta. The vertical stripes change color in precisely the opposite way. You can verify the fact that the colors are associated with the orientation of the stripes by rotating the illustration 90 degrees. Again magenta appears with horizontal stripes and green with vertical stripes. The shapes of the regions are not important; the association of the color is with orientation.

One should not mistake the McCollough effect for the phenomenon known as a negative afterimage. The color in that phenomenon is fleeting compared with the McCollough coloration. To demonstrate a negative after-image fix your gaze on a field of green color for about five minutes and then on a white, featureless surface. For a short time you will perceive magenta. If you delay viewing the white surface by 10 minutes or more, the afterimage fails to appear.

When you look in a normal way at a white surface, all the color sensors in the retina send signals of color to the brain. Somewhere along the way the signals are analyzed according to pairs of complementary colors that compete. If you see equal amounts of the colors in a pair, they result in a perception of no color. For example, green and magenta compete. If equally strong green light and magenta light reach the eye, the competition is a draw and you perceive white, that is, a colorless illumination.

The normal afterimage is often attributed to fatigue of some of the retinal color sensors. For example, if you view green, the sensors responsible for sending that signal to the brain come to be less responsive. Suppose you view a white surface just after the green sensors become fatigued. Although equally bright green and magenta enter the eye, the weak response by the green sensors allows the magenta to prevail. You perceive magenta in the white region. When the fatigue wears off, the competition between the two colors is again a draw and you see white.

The McCollough coloration also differs from the normal afterimage in that drugs can alter its duration, suggesting it depends on the neurotransmitters in the visual pathway that extends from the retina to the brain. C. C D. Shute of the University of Cambridge reported that caffeine accelerates the decay of the effect whereas fresh ginseng tea (not the instant variety) delays it. Bamidele O. Amure of the University of Cambridge reported that nicotine prolongs the effect. In addition D. M. MacKay and Valerie MacKay of the University of Keele in England demonstrated that the decay can be postponed by sleep and that the strength of the coloration can depend greatly on how well the observer sleeps before looking at the colored grating.

I did experiments with the McCollough effect by preparing several gratings of black stripes and photocopying them on a machine that retained the solid black character of the stripes. Next I colored the spaces between the stripes. Most often I tested a grating of horizontal black lines interspaced with green. I looked at the grating for about five minutes and then waited five minutes before looking at a colorless grating. The delay ensured that coloring due to a negative afterimage would not appear. The afterimage was normally not a problem anyway since I did not fix my gaze. I had more trouble with the need to allow the McCollough coloration to decay completely between experiments. On occasion the decay took as long as a day.

Does the McCollough effect require straight edges? I made a set of gratings identical with the previous ones except that the black stripes were replaced by rows of small black dots. No McCollough coloration appeared. I replaced the dots with rows of solid black circles but again found no McCollough coloration.

I thought perhaps the spacing of the grating altered the coloration. Conditioning my eyes to the original green-and-black grating, I then looked at a colorless grating that had the same oil orientation but on which the black-and-white stripes were scaled down to about half the original size. No coloration appeared.

I wondered how much of the field of view must be occupied by the original grating to condition the eyes properly for the color effect. Covering most of the original grating with opaque paper, I gazed at the small amount still visible and then examined the colorless grating without covering any of it. No coloration was present. Apparently one must see enough of the original grating to recognize it as a grating before the oil McCollough effect will work.

Next I reversed the procedure, looking at the full original grating and then examining only a small part of the colorless one, most of which was covered with opaque paper. The faint magenta appeared. Indeed, it seemed to be there even when I covered up all but a small segment of one or two black stripes. Some students of the McCollough effect have suggested that if the stripes h of a grating are widely spaced in one's view, the visual system focuses on the edges of the stripes. If instead the stripes are closely spaced in one's view, the system focuses on the periodicity of the stripes. When I examined the small part of the colorless grating, the McCollough coloration was probably associated with the edges rather than with the periodicity.

In 1974 Charles F. Stromeyer III of Stanford University reported that the McCollough coloration is perceived better if the colorless grating is examined in dim light. The coloration can be seen even when the illumination is so low that the retinal cones, which are responsible for color information, do not function. I tested these results by conditioning my eyes to the black-and-green grating in bright light and then dimming the illumination. As my eyes adjusted to the dim light the magenta coloration in the colorless grating became more pronounced even though I could not see color anywhere else in the room.

Stromeyer also reported that the spacing of the grating can determine the McCollough coloration. An observer first looks at a green-and-black grating with wide spacing. Then he views a closely spaced magenta-and-black grating. The orientations of the two gratings are identical. Some time later he looks at colorless gratings that have the same orientation as the first two. If a colorless grating has the same spacing as the first one, he perceives magenta. If instead it has the same spacing as the second one, he perceives green. Somehow the observer has stored color along with the information about spacing.

How is the McCollough effect produced? No one knows in detail, but a few elementary models have been suggested. One of them, which has been put forward by a number of students of the visual system, is shown in Figure 4. The model is crude in that details of how it works are not known. At the left is a section of retina illuminated with a horizontal grating of black-and-green stripes. The circuitry at the right represents the early stages of vision. Signals move toward the right to reach the higher levels of processing by the brain. Keep in mind that the illustration is only representative. I do not know where the analyzing sections of the visual system exist, if indeed they can even be localized. I also am uncertain about how they function.

Signals relating to the color of the green stripes are initiated by cone photoreceptors in the retina. The cones come in three types, each of which is sensitive to a different part of the visible spectrum. Once the cones are excited their signals are compared and other signals are relayed deeper into the visual system. These new signals indicate luminosity and the relative strengths of red v. green and of yellow v. blue. At some later stage, which is probably within the brain, the new signals are compared in terms of the competing complementary colors.

For example, the colors green and magenta compete in this later stage of analysis. Consider the competition in terms of numbers assigned to the signal strengths: green positive, magenta negative and an even mixture zero. When the eye sees green, the signal strength might be, say, +200. When it sees magenta, the strength might be -200. If equal amounts of green and magenta are detected (as when you see white), the strength of the color signal is zero.

The edges of the stripes are detected by groups of photoreceptors lying at or near the place where the image of the edge falls on the retina. The information from many edge detectors feeds into a grating detector, which sends deeper into the visual system a signal that a grating of a certain orientation and a certain periodicity is being viewed. No one is certain about how the grating detector works. Assume that it recognizes a grating by comparing it with some standard gratings. Once a match is made a signal is sent along an output line. The output line chosen depends on which of the standard gratings matches the one being viewed. If a horizontal grating of a certain spacing is viewed, a signal is sent along one of the output lines. If the grating is rotated to be vertical, the signal is sent along another line. If the spacing of the grating is varied enough to make a match with a different standard grating necessary, the signal is sent along a third line.

The McCollough effect might be explained in terms of this model. As you view a horizontal grating of black-and-green stripes, strong signals are sent along two lines. One signal concerns the green of the light and the other signal concerns the orientation and spacing of the grating. An inhibiting interconnection begins to build between the two lines, somewhat weakening the signal of green. The inhibition is not large enough for you to notice it in normal room light. Although the strength of the interconnection builds up within five minutes or so, it may last for hours or even longer. Suppose you look later at a colorless grating that has the same orientation and spacing as the first one. On the color line the signal strength is zero because the grating has white stripes. If the color signal continued to be zero, the perception would be of white. The inhibition imposed by the interconnection, however, makes the strength of the signal negative. The color signal then registers magenta, and you perceive magenta on the colorless grating.

Inhibition is imposed on the color line only if the grating detector actuates the output line that has the strengthened interconnection. If some other output line is actuated, the color , line is unaltered. Suppose you rotate the grating until it is vertical. The signal from the grating detector moves along another output line, the color line is not inhibited and the McCollough coloration disappears.

Although this model fits many of the experiments done with the McCollough effect, it is unsatisfactory. Why should the output line of a grating detector inhibit the color line? I am also troubled by the following experiment. If the model is correct, the inhibiting interconnection should be built up if you shift your gaze over a colorless, horizontal grating alongside a green region. Yet such an arrangement actually fails to produce a McCollough coloration.

For the fun of it I devised a different model for the McCollough effect. I do not know if this model has been studied before. It differs from the preceding one in that the grating detector is itself color-sensitive, sending a color signal into the analyzer that compares complementary colors. The analyzer sums the signal from the grating detector and the principal color signal coming directly from the retina. The strength of the color signal from the grating detector depends on the recognition of the grating.

The color sensitivity of the grating detector may be an unavoidable consequence of the fact that it and the edge detectors analyze signals from color-sensitive retinal cells. In part the detection of an edge may depend on the contrast of color on its two sides. Hence a signal from the grating detector about color is not surprising.

Suppose you view a horizontal grating of black-and-green stripes. After making a match with a standard grating, the grating detector sends a signal out along an output line concerning the orientation and periodicity of the grating. Because of the match, the detector sends to the color analyzer a signal about the green in the grating. Suppose the direct signal is +200 units and the green signal from the grating detector is +1O. The signal that emerges from the color analyzer is +210 units, which is perceived as green.

As you continue to examine the grating, the detector grows less responsive to both the grating and its color. After a while the detector begins to reduce its signal about the grating characteristics and you get a poorer perception of the grating. The detector also begins to inhibit the signal of green, sending instead a signal of magenta to the color analyzer. Suppose the strength of the magenta signal is -10 units. If the direct signal of green is again +200, the signal emerging from the color analyzer is +190: a slightly weaker green than before.

When you replace the colored grating with a colorless one of the same spacing and orientation, the grating detector recognizes the grating and again sends a signal of magenta to the color analyzer because of its inhibition to green. Note that at this point the color from the grating detector is triggered by the recognition of the grating, not by the strength of any color signal arriving from the edge detectors. If the magenta signal has a strength of -10 and the direct color signal is zero (because of the whiteness of the stripe being viewed), the color analyzer sends out a signal of -10: a faint magenta. Hence when you view the colorless grating, you perceive a faint magenta superposed on the white stripes. This coloration is the McCollough effect.

What if you look at the colorless grating in dim light? The coloration is more prominent not because of greater signal strength but because the illumination from the white stripes is weaker. You can then pick out the magenta coloring better. The coloration is still apparent even if the light is so dim that the cones no longer function. In such light the visual cells called rods provide vision. They send no color information, but they do serve to detect edges and thus feed a signal to the grating detector. Since the grating detector is still fatigued because you had looked at a grating earlier, the signal it sends to the color analyzer is still -10 units of magenta. You still perceive the coloration.

From the model I figured that under the proper lighting conditions a signal of magenta from a fatigued grating detector might cancel a small, direct signal of green. For example, if the magenta signal is -10 and the direct green signal is +10, they sum in the color analyzer as zero. A green stripe in the grating would then appear to be h gray: it is both colorless and dim. Does this graying actually take place? If it does, would the green reappear if I rotated the grating so that the stripes were vertical and the grating detector therefore was forced to carry out a new task of recognition? Suppose an isolated green mark lay in another part of my field of view. Would I still see it as green even if the green of the grating had turned gray?

I prepared for the experiment by l adding some extra marks just to one side of the usual green-and-black grating. Some of the marks were black and some were green. Some were small and others were stripes as long and wide as those in the grating. I lowered the room lights, waited for 15 minutes so that my eyes would adjust to the illumination and then examined the grating for another 20 minutes.

Although I could initially distinguish the green in the grating, the color soon began to fade into gray. The grating itself became harder to perceive. At this point the grating detector in my visual system must have become less responsive and its color sensitivity must have switched to an inhibition of green. Still, I wondered whether perhaps the gray resulted from a decrease in the direct signal of green because the cones were becoming fatigued by the color. I checked this possibility by shifting my gaze to the isolated marks of green on the paper. Their green was still perceptible. I also checked by rotating the grating until the stripes were vertical. The green of the stripes immediately reappeared. For these reasons the gray of the stripes in the initial orientation seems to be due to the inhibition of green within the grating detector.

Many additional experiments with the McCollough effect have been described in published work. Perhaps you can devise some of your own and can construct a better model than the ones I have considered. If you can, I should like to hear about your work. I would be particularly interested in experiments that disprove my model of a color-sensitive grating detector.

Bibliography

COLOR ADAPTATIONS OF EDGE-DETECTORS IN THE HUMAN VISUAL SYSTEM. Celeste McCollough in Science Vol.149, No.3688, pages 1115-1116, September 3, 1965.

FORM-SPECIFIC COLOUR AFTER EFFECTS IN SCOTOPIC ILLUMINATION. Charles F. Stromeyer III in Nature, Vol. 250, No. 5463, pages 266-268; July 19, 1974.

The Society for Amateur Scientists (SAS) is a nonprofit research and educational organization dedicated to helping people enrich their lives by following their passion to take part in scientific adventures of all kinds.