ADAPT YOUR EYES TO INDOOR darkness for 10 or 15 minutes and then set off a bright flash of light. Keep your gaze steady as the glare subsides. Within a few seconds you will see what resembles a snapshot of the room. Details lost in the dazzle of the flash are now readily apparent. This is an afterimage (a positive one because it reproduces the light and dark regions in the scene that was illuminated). It lasts for tens of seconds if you keep your eyes steady.

I first learned of this kind of afterimage from Peter Hasselbacher of the University of Louisville School of Medicine, who had seen it at parties when he was in college. One of the earliest reports of the afterimage was made in 1894 by Shelford Bidwell, a British physicist who did research on vision. When he illuminated an object with the flash from an electric discharge, he perceived the object in an afterimage about .2 second later. Sometimes the afterimage faded and reappeared several times.

In another experiment Bidwell made a gas-discharge tube rotate at about .4 revolution per second. The tube was periodically illuminated with a flash of light. When he viewed the rotation with a steady gaze instead of tracking it with his eyes, the tube appeared to be followed by a dim blue or violet ghost light. When the rotation stopped, the ghost light caught up with the discharge tube. The afterimages Bidwell saw in each of his experiments have since been called Bidwell's ghost.

Bidwell also shone a light on the hole of a rotating disk so that a spot of light circled a screen on the other side of the disk. The room was otherwise dim. He perceived a ghost light following the spot at a delay of about .2 second. When the original light was white, blue or green, the ghost light was violet. When the light was orange or yellow, the ghost light was blue or blue green. Red light, however, did not produce a ghost.

In 1934 William Stewart Duke-Elder, a British surgeon and ophthalmologist, described how an afterimage can be produced while an observer looks toward a spot of light with a card blocking the view. When the card is rapidly moved in and out of the line of vision, the observer perceives a positive afterimage of the light. "So vivid, indeed, may be the impression of the original afterimage," Duke-Elder wrote, "that the card appears transparent, and details which were not noted in looking at the-light are brought to the attention in the afterimage."

I investigated the afterimage with an electronic flash from my camera. After a few minutes in a dark room (well before my eyes were adapted to the dark) I opened a magazine on a table. The room was so dark that I could see nothing of the magazine. Having placed the flash to one side of my head so that it pointed toward the magazine, I triggered it. (I was careful not to point the flash unit directly into my eyes. At such close quarters the light might damage the retina.) The glare was dazzling, but I was able to make out the general outline of the magazine pages.

Within seconds a positive afterimage of the magazine appeared. (I call it a snapshot afterimage.) The image was bright enough for me to recognize photographs and read the headlines. I "saw" the smaller print but could not read it. The image soon blurred and faded, finally disappearing after about 15 seconds. The darker sections (the photographs) disappeared first and the lighter sections last.

What came next was a vaguer but much more persistent negative afterimage in which the bright and dark regions were reversed. I believe the negative afterimage results from a retinal fatigue caused by the snapshot afterimage. The bright portions of the illuminated scene create much activity in the regions of the retina where their images fall. When that activity dies out, those regions seem dark compared with the areas that had less activity. Hence the positive afterimage fades into a negative one.

I continued flashing the light and examining the snapshot afterimages for the next half hour. For the first 15 minutes they improved steadily in quality as my eyes continued to adapt to the darkness. Both the snapshot and the negative afterimages were always colorless even when the objects in view were colored. Once the snapshot afterimage was so vivid that I tried walking across the room as if I could really see. I quit when I walked into the edge of an open door that seemed to be distant.

Improving the quality and duration of the snapshot afterimage took some practice because the flash tended to make me blink. A blink or a movement of the eyes across the scene usually weakened or erased the snapshot afterimage. Moving my head did not seem to matter if I did not appreciably move my eyes with respect to my head.

I checked the steadiness of my gaze during the afterimage by looking at a dark television screen. The flash of light made the screen glow faintly. I then perceived the glowing screen superposed on the snapshot afterimage of the dark set. The two images slowly separated, indicating that in the darkness after the flash I had unwittingly let my eyes drift, apparently without affecting the afterimage, which was fixed in place on my retina.

Nearly all the literature on this type of Bidwell's ghost says the gaze must be kept steady if the afterimage is to be maintained. With a few trials I discovered that a steady gaze is usually not crucial in the early stages of the proceeding. If I move my eyes during the flash of light or in the next several seconds, the afterimage disappears momentarily but soon returns, apparently as sharp as it would have been if I had kept my eyes still. If I move my eyes later, the afterimage disappears immediately and permanently. No one has been able to explain how such a movement of the eyes erases the afterimage.

The negative afterimage was much less fragile, appearing even if I could not keep my gaze fixed. It usually persisted for several minutes. If I flashed the light in order to produce another snapshot afterimage, an old negative afterimage would sometimes be superposed on it.

The perception of a snapshot afterimage often conflicted with my common sense. Once I illuminated my hand while it was in front of my face and then moved it behind my back. When the snapshot afterimage appeared, I had a clear perception that my hand was in front of me and an equally clear feeling that it was behind my back.

In another experiment I squatted, looking toward my feet and the floor while I triggered the light. Then I stood up, while continuing to look at the floor. The snapshot afterimage convinced me that my head was near my feet, but I knew that I was standing up.

The reverse arrangement was equally unsettling. I flashed the light toward the floor while I looked down from an upright position. Before the afterimage appeared I squatted. The contrast between what I perceived and what I felt was startling. Just as disconcerting was a snapshot afterimage of my face when I looked into a mirror during the flash of light.

By triggering the flash unit twice within a few seconds I could superpose two snapshot afterimages. In one experiment I held my hand in front of my face while I flashed the light. Immediately I moved my hand to the right and triggered the light again. When the snapshot afterimages appeared, I saw both positions of the hand. I knew the hands were mine, but neither image made sense because my hand was then at my side. If I held my hand with the palm toward me in one flash and away from me in the next, I saw a remarkable hand with 10 fingers, half curled toward me and half curled away.

I next sat with my right leg crossed over my left. With my head down I illuminated my legs with a flash, quickly crossed them the other way and illuminated them again. In the resulting afterimage I perceived a leg extending to the left and another one extending to the right.

The ability to superpose snapshot afterimages enabled me to stroboscopically freeze an object moving through my field of view. I held a coin in front of me, released it and then flashed the light twice while the coin was falling. In the afterimage I saw my hand and two positions of the coin.

I wondered if an afterimage from one eye could be superposed on an afterimage from the other. I triggered the light while only my left eye was open. Then I closed that eye, opened the other one and triggered the light again. The two views were indeed superposed.

If I moved my head between one flash and the next, the scene was usually too cluttered to be clearly recognizable. Sometimes I created two images that I could fuse to produce a weak stereoscopic effect. Normal stereopsis yields a perception of depth that is correct because the brain is accustomed to the usual separation of the eyes. In my experiment the views from the eyes were made with a different separation, giving rise to a different perception of depth.

I next experimented with several cellophane filters to check for color dependence in the snapshot afterimages. I covered the flash unit with a filter, but I could just as well have looked through the filter. With a green or blue filter the snapshot afterimages were vivid. With a red filter they were dim and brief or did not appear at all.

I then investigated the ghost image reported by Bidwell for a moving spot of light. I moved a small orange spot of light (the trigger button of my camera flash) rapidly across my field of view while I kept my gaze fixed. After a noticeable delay a dim blue spot raced along the same path, running into the orange spot where I had stopped it The trail behind the ghost spot glowed dimly (it may have been blue, but I was not sure) for tens of seconds. When I moved the orange light through an intricate pattern, the delayed ghost spot raced along the same path, leaving the entire path glowing slightly.

If the ghost image were due merely to the persistence of vision, it would appear simultaneously with the stimulus light, and the path just behind the stimulus would glow. Hence the ghost is an afterimage. As the spot of light is brought across the field of view, its image on the retina moves across photoreceptors that were previously unilluminated. The sudden supply of light provokes the afterimage, but only after a short delay. If you track the stimulus light, the same photoreceptors are continuously illuminated and no trailing ghost appears.

Returning to my camera flash, I illuminated a page of print with only my right eye open. I was careful to keep gazing toward the left side of the page. Looking at the snapshot afterimage with both eyes, I noticed that the image was poorer on the left side. In normal viewing an object looked at produces an image on the fovea, a small area of the retina that is densely packed with cones. The region lacks rods, the other type of photoreceptor. Because the snapshot afterimage of the page was unclear in the area on which my eye focused, I concluded that rods must produce the afterimage.

There should be a second unclear spot in an afterimage. Toward the temple side of the field of view the eye has a blind spot because that is the part of the retina where the nerves enter the eye. The area has no photoreceptors and cannot transmit an image to the brain. In normal viewing the blind spot is ignored because the region not seen by one eye is seen by the other. When I exposed only one eye to the briefly illuminated page, I expected the afterimage to have an indistinct area toward my temple. In some cases I thought I perceived such an area, but I was never quite certain.

Sometimes I replaced the flash unit with a standard stroboscope set to a low frequency. After a flash or two I turned off the stroboscope to examine the afterimage. (Do not look directly into a flashing stroboscope. The light might be intense enough to harm your eyes. If you are subject to seizures when viewing flashing lights, avoid a stroboscope.)

In one experiment I generated a snapshot afterimage with my camera flash unit, closed my eyes and then turned on the stroboscope for about 30 seconds. Light came through my eyelids with each flash, creating a diffuse red background (red because of the blood in the eyelids). Against the background I perceived a negative image of the scene that had been illuminated by the camera flash unit. When I turned off the stroboscope, I was surprised to find a snapshot afterimage long after it should have disappeared.

With more experimentation I confirmed that this technique prolonged the snapshot afterimage. I do not know why, but I can offer two hypotheses. The diffuse light transmitted through the eyelids might boost the chemical changes induced by the original flash, making them persist longer. Alternatively, what I perceived after turning off the stroboscope might have been an afterimage that was the contrast of the negative one that would have been there anyway. Since the contrast to a negative image is positive, I perceived a snapshot afterimage when the stroboscope was turned off.

In some classic experiments done with other types of afterimages an observer can vary his judgment of the size of the afterimage if he sees it against a background that changes distance. Suppose he perceives something occupying one degree of arc in his field of view. His decision about the size of the object is determined in part from the apparent distance to the object. If enough clues convince him that the object is distant, he concludes it must be large. If the clues convince him that the object is close, he concludes it must be small.

I had a vague sensation of how the size of the snapshot afterimage can depend on such clues. In a dimly lighted room I illuminated a food can with my camera flash. Then I held a sheet of paper in front of my open eyes. The snapshot afterimage of the can was superposed on the dim outline of the page. When I moved the page away from me, the can did seem to be slightly larger. The experiment was difficult because I tended to focus my eyes on the page. The eye movement erased the afterimage.

Other research on the snapshot afterimage suggests that such a variation in the judged size of an object can arise in total darkness. Supposedly the observer alters the judged size as he moves his hand (unseen in the dark ness) from just in front of his eyes to full arm length. I suppose one imagines the hand to be a viewing plane for the afterimage even though the position of the hand can only be felt. I failed to create this illusion.

The positive afterimage following a brief flash of light has been recently investigated by Edward H. Adelson of RCA's David Sarnoff Research Center, drawing on work done independently by Barbara Sakitt of the Massachusetts Institute of Technology and Wilson S. Geisler III of the University of Texas at Austin. Adelson attributes the afterimage to the retinal rods. In his experiments an observer adapted to darkness was exposed to a green square on a circular red background. The flash of light lasted for .01 second. The square occupied 4.5 degrees in the observer's field of view; the background occupied 11 degrees. The patterns were centered about 15 degrees off the observer's line of sight, and so their images on the retina were away from the fovea.

The afterimage depended on the intensity of the light. Some of the results are depicted in Figure 1. The top row is a sequence of the observer's impressions when the intensity was low. The square was immediately visible as the glare of the flash subsided, and the afterimage was brief. The second row, a sequence resulting from intermediate intensity, indicates a brief delay before the square could be distinguished. This time the afterimage lasted longer.

The sequence in the bottom row is from high-intensity light. Several seconds went by before the square emerged from the background. The afterimage lasted longer than the others. It is evident that the observer's ability to distinguish parts of the afterimage is delayed when the intensity of the flash is increased. The increase also gives rise to a longer afterimage.

Adelson found in addition that both patterns may seem to brighten as the square emerges from the background. I saw all these results in my experiments with snapshot afterimages of magazine covers and pages.

Adelson suggested how the rod system can be saturated by the flash of light. When the outer segment of a rod absorbs light, a series of chemical reactions generate a substance blocking the flow of sodium ions through a membrane in the outer segment. When the sodium flow is totally blocked, the signal sent to the brain is at maximum strength and the rod is said to be saturated. During this stage additional light will not augment the signal.

Hence the initial flash of light can provokes the afterimage, but only after a short delay. If you track the stimulus light, the same photoreceptors are continuously illuminated and no trailing ghost appears.

Other research on the snapshot afterimage suggests that such a variation in the judged size of an object can arise in total darkness. Supposedly the observer alters the judged size as he moves his hand (unseen in the dark ness) from just in front of his eyes to full arm length. I suppose one imagines the hand to be a viewing plane for the afterimage even though the position of the hand can only be felt. I failed to create this illusion.

Hence the initial flash of light can saturate the rod system so that the signal from the rods stimulated by the square is as strong as the signal from the rods stimulated by the background. As the signals decay, however, they eventually differ in strength, enabling the observer to distinguish the square. When the flash of light is intense, the rod systems remain saturated longer, delaying the emergence of the square.

In one series of experiments Adelson tested the color dependence of the positive afterimage. The background was kept red-orange while various colors were chosen for the square. With each trial the observer adjusted the intensity of the square so that in the afterimage the square was barely distinguishable from the background. When the choices of intensity were plotted against the wavelength for the color of the square, the graph matched the wavelength dependence of rods, indicating that the rods were responsible for the afterimage.

Adelson also ran trials in which the color of the square was kept green while the background color was varied. This time the observer adjusted the intensity of the background until the square was just visible in the afterimage. Again the graph of the results matched the wavelength dependence of the rods.

Are rods also responsible for the ghost light trailing a moving stimulus light? I think they are, but I am puzzled by the color of the trailing ghost.

I can suggest a possible solution. The color of the ghost may come about because rods interfere with the color signal from cones during the afterimage. (Rods cannot directly signal color but cones can.) Bidwell's observation that a red stimulus light does not generate a ghost is probably attributable to the fact that rods are not excited by deep red. Therefore such a stimulus does not create the chemical activity in rods that generates a ghost.

A stimulus light of any other color excites the rods so that they generate an afterimage. Their activity inhibits the color signal (then weak but still present) from the cones that were excited by the stimulus light. Such an inhibition sends a signal to the brain that the cones are intercepting the complementary color of the stimulus light.

Suppose the stimulus color is yellow. It will excite the rods and those cones designed to detect yellow. As the stimulus light passes over a region of these cones yellow is perceived.

Within about .2 second the rods create an afterimage. Their activity inhibits the signal of yellow from the cones; a signal of blue (the complementary color of yellow) is sent to the brain. The trailing ghost created by the rods is colored blue by their interference with the cones. That subtle coloring is lost if the rods are saturated, as they are when a snapshot afterimage is made. I checked this conclusion by searching for a trailing ghost from an orange stimulus light just after I created a snapshot afterimage. I saw the orange light but no ghost.

Daniel H. Kriegman and Irving Biederman of the State University of New York at Buffalo have recently investigated the extent of information available in the snapshot afterimage. Earlier investigators had tested an observer's recall after glimpsing a display of 12 letters under normal lighting conditions. Usually an observer can recall only three or four letters. The rest are lost because the visual image rapidly fades and because the memory is inadequate.

Kriegman and Biederman wondered if recall would improve when the observer was adapted to darkness and was then shown an array illuminated with a flash of light triggered by the observer. The brightness was adjusted by filters of neutral density placed in front of the flash unit.

The arrays were prepared by typing 14 consonants and numerals on a blank index card. The observer recorded his observations by speaking into a tape recorder. The five vowels and the letter Y were avoided to make for unambiguous pronunciation. G, Q and zero were also eliminated because they could be visually confused. The letter B was eliminated because it might sound like P.

The arrays of the displays were typed in three rows. The middle row had four figures, the top and bottom rows five each. The blank at the center of the middle row was a point of fixation for the observer. In that spot he saw a point of dim red light made by a small flashlight on the other side of the card. This light was too feeble to illuminate any of the letters on the card.

Six observers took part in the experiments. Each person sat about a foot away from the array and adapted to the dark for about 12 minutes. Then a trial of exposure and recall was done every 30 seconds. The first 10 trials of the session and the first three trials with each filter were for practice. Fifty trials were recorded for each observer. In some sequences the observer began with the dimmest filter and progressed to the brightest. In others the progression was reversed.

The best recalls were made when the light from the flash was brightest. Better recall also resulted from the sequence beginning with the dimmest filter. With the brightest filter the observers were able to recall an average of 12.4 figures out of the 14 in the array. In nine trials at the highest brightness the observers gave perfect reports. (Some of the success may, of course, have been due to increasing familiarity with the figures.)

Apparently accurate and full reporting requires some practice. An observer must keep his gaze steady in order to maintain the afterimage. He must also ignore variations in the illumination of the card that might otherwise interfere with the reading of the figures. With proper concentration observers were able to read the figures in the afterimage even though they did not see them well in the initial flash.

When I repeated this experiment with a similar arrangement, I found that reading the letters in the afterimage was quite strange. Normally one reads by moving one's eyes across the page. With the afterimage I had to gaze steadily to avoid erasing the picture. I do not understand how I then read the letters off to the side. Somehow I thought about what was there and could then recognize the figures. I moved my concentration instead of my eyes.

Bidwell's ghost would still repay investigation. Is the trailing ghost actually related to the snapshot afterimage? What causes the colors in the ghost? Why are the snapshot afterimages colorless? Why are they weakened or erased if your eyes move appreciably? Why are they prolonged if the eyes view a diffuse, flickering field? I would enjoy hearing about what you find to further explain this phenomenon.

Bibliography

HOW MANY LETTERS IN BIDWELL'S GHOST? AN INVESTIGATION OF THE UPPER LIMITS OF FULL REPORT FROM A BRIEF VISUAL STIMULUS. Daniel H. Kriegman and Irving Biederman in Perception & Psychophysics, VoL 28, No. 1, pages 82-84; l980.

WORKSHOPS IN PERCEPTION. R. P. Power, S. Hausfeld and A. Gorta. Routledge & Kegan Paul, 1981.

THE DELAYED ROD AFTERIMAGE. Edward H. Adelson in Vision Research, Vol. 22, pages 1313-1328; 1982.

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.