"NEON SPREADING" is not a warning that gas is escaping from a Times Square signboard. Rather, the term was introduced in 1975 by Harrie F.J.M. van Tuijl of the University of Nijmegen in the Netherlands to describe a certain illusory spreading of color resembling the diffuse light that issues from a glowing neon tube.

The human visual system is remarkably skilled at reducing the bewildering amount of information received by the eyes to a recognizable mental picture of the external world. Yet in some circumstances the picture contains curious errors: visual illusions, which often offer a glimpse into otherwise hidden operations of the visual system. In some illusions the failure of the visual system to create a precise mental picture of a rather spare diagram can be quite perplexing.

One of van Tuijl's demonstrations of the neon-spreading illusion is a case in point: the illustration is simple, and yet we perceive it erroneously [see illustration at left]. To see the illusion, hold the page up-right, with the horizontal lines level, and avoid fixing your gaze on any part of the illustration. (It may also help if you hold the page at somewhat more than the usual reading distance.) Bright blue-tinted spots will appear to overlie the blue crosses, as though some of the blue ink from the crosses had seeped out over the surface of the page. The spots may also seem to connect to form slanted lanes. Why do the colored spots appear?

The neon-spreading illusion is related to another illusory appearance, called subjective contours, which was the subject of this department in January, 1988. You can produce an example of subjective contours by redrawing the grid with blank spaces where there are now blue crosses. In the blank spaces you will perceive bright circular spots with perceptible borders. The bright spots are not really there, of course; they are said to be subjective. In van Tuijl's illustration, the blue of the crosses spreads out and fills the circles. The illusory bright spots in the empty gaps and the neon color surrounding the colored crosses may be of similar origin, because they appear in the same kind of pattern.

Van TuiJl offered two other patterns, reproduced at the right, below, to demonstrate that the neon spreading is not ubiquitous in grid patterns. In the figure on the left the red bleeds into the framework of the design, but in the one on the right the red is confined to the lines. Interconnections, present on the left but not on the right, must be somehow important.

In 1979 van Tuijl and his colleague Charles M. M. de Weert examined additional grids, including some in which bright lines lay on a dark background. Three rules emerged from their work. If neon spreading is to appear, both the basic grid and the interrupting crosses must be either lighter or darker than the background. If they are both darker than the background, as in the illustration at the left, below, then the crosses must be less dark than the grid (blue rather than black, in this case). If they are both light on a dark background, the grid must be lighter than the crosses. In either case, then, the brightness of the crosses must be intermediate between the brightness of the grid and that of the background.

Although the rules allow one to predict whether a particular grid pattern will display neon spreading, they do not reveal the source of the illusion. According to van Tuijl and de Weert, 9 the illusion stems from the way the visual system interprets what is being h seen: the system apparently finds it more "efficient" to perceive faint, colored spots and dulled, murky crosses instead of the true scene. The argument was repeated in another 1979 paper by van Tuijl and E.L.J. Leeuwenberg, also a colleague.

The concept of efficiency here may need clarification. As an example, consider again that first illustration below. When you view it, information about the black lines and colored crosses is sent along the pathway from the eyes to the visual cortex of the brain. Somewhere along the route the system begins to try to make sense of what is being seen. It compares the relative brightness of background, grid and crosses, and it also considers the overall pattern, in particular its repetitive elements. The system might liken the pattern to what has been seen in the past. It might consider whether parts of the pattern may be hidden by something closer to the viewer. After these and other matters are weighed, an interpretation is made and a final picture is brought to the level of consclousness.

When you look at the first of van Tuijl's illustrations, your visual system presumably decides that the most probable interpretation of it is that parts of what is actually a complete grid of black lines are masked or obscured by projections of circular blue spots. Where the projections seem to fall on the grid, the blue glow dulls the contrast between the colored crosses and the surrounding white region by tinting the white.

Why should the visual system be attracted to such an interpretation? Perhaps it is because the system seeks to complete line segments that extend toward each other the way the black lines do from each side of a colored cross. That would relate the pattern to other grids or gridlike structures often seen before-a fence, say. It is normal that an object that is closer to the viewer than a gridlike structure would be brighter than the grid; the fact that the crosses are brighter than the black lines supports the notion that they are in front of the grid and block the view. A cross is not commonly seen precisely positioned to obscure segments of a grid, however, and so the visual system conjures up a spot, which is a more likely obstacle. If the crosses are darker than the grid, the whole interpretation falls flat because the crosses do not appear to be in front of the grid; the attempt at mentally completing the grid fails, and there is no need for obscuring spots or neon spreading.

That explanation seems to me to be incomplete. I do not think I fancy a completed grid more than I do one with color changes, which is actually more interesting. I am also bothered by the fact that once I realize the illusion is an error, I cannot make the error go away. Even when I examine I the illusion for a while, concentrating on the error, I still perceive the spread of color. If my visual system strives for efficiency and also compares a present view with learned information, surely after enough experience and conscious thought it should at least reduce, if not eliminate, the error of neon spreading. Could there be, instead, some stage in my visual system where a physical process introduces an error-one that is so well entrenched by the time it reaches the conscious level that neither visual experience nor willpower can shake it? I that is the case, where exactly does the erroneous processing take place?

A bountiful harvest of clues came in 1981, when Christoph Redies and Lothar Spillmann, then both of the University of Freiburg in West Germany, reported their studies of variations of the grid arrangements that give rise to neon spreading [see illustration at right]. In the top figure red crosses replace some grid crossings and blue crosses replace others. Red neon spreading adorns the red crosses, blue neon spreading decorates the blue ones and red or blue slanted lanes connect the colored spots. To see the illusion best, you should again hold the page so that the horizontal lines of the grid are level and adjust the distance between your eyes an the page. If you then rotate the page around your line of sight, the illusion weakens as the rotation reaches 45 degrees and reappears as the previous vertical grid lines become horizontal.

Now move back from the grid until the blue crosses grow dark enough to blend with the black lines they connect. Although you can no longer distinguish the color of the crosses, the still induce what are perceived as being blue spots. If the crosses were yellow instead of blue, moving back would cause them to fade into the background until they disappeared but yellow spots would still be recognizable. Either experiment demonstrates that although the color of the crosses may not survive to the conscious level, it can continue to trigger neon spreading.

When red or blue crosses are isolated from the grid lines, as in the second figure from the top, they fail to pr duce neon spreading at any angle orientation; the juxtaposition of the grid's black lines with a cross must then, be vital to the illusory spread of color. Notice also that an isolate cross is more sharply defined and contrasts more with the background that does one enmeshed in the grid: in previous examples, neon spreading makes the cross look dull and murky.

Two more variations are seen at the right. In the third figure from the top each cross is made up of one red line and one blue line. Colors spread from both colored lines but not as much as in the previous examples. In the bottom figure the black lines are eliminated, so that the grid consists solely of red or blue crosses. Yet the neon spreading of each color is robust and even yields slanted colored lanes. The result adds a new rule about brightness. In this case either the red or the blue crosses can be considered the "grid," and the crosses of the other color are taken to be the "crosses." Neon spreading can apparently appear when both elements are about equal in brightness but are darker than the background.

Redies and Spillmann investigated how the strength of neon spreading depends on the angular size of the colored crosses enmeshed in a black grating. When the pattern is on the direct line of sight (which is called foveal viewing), the angle occupied by a cross in an observer's field of view has to be between four and 35 minutes of arc if neon spreading is to appear. At the lower end of the range, the spreading is strongest and fills out a circle. At the upper end, it is weaker and takes on a diamond shape. When the angle is even larger, the spread of color retreats and lies just along the sides of the lines in the cross, an effect dubbed neon flanks. When the pattern is off the direct line of sight by several degrees (extrafoveal viewing), the upper and lower ends of the angular range for neon spreading shift to somewhat larger values. All of this indicates why the illusions on these pages often intensify if you adjust your distance from the illustrations: you thereby vary the angle subtended in your field of view by each colored line or cross, optimizing the illusion.

Redies and Spillmann also examined variations of a single unit in which one cross has black lines at each end [see Figure 4]. When the cross is disconnected from the black lines by a separation or by misalignment through displacement or rotation, neon spreading disappears. The spreading is also eliminated if the cross is circled. If the black lines are short, the illusion is strong; it may still appear, albeit weakly, even if there are only mere dots at the ends of the cross.

The most intriguing illustrations by Redies and Spillmann are the ones in which they strip the illusion to the bare essentials: single lines [see Figure 5]. If an isolated red line is viewed, there is no spread of its color, but if it connects two black lines so as to form a single, straight line, red neon flanks run along the red line. The spreading of color disappears if the black line on either side is removed.

With the results from these and other experiments, Redies and Spillmann conjectured that neon spreading is a result of the activity of visual cells that detect the presence and orientation of lines and that are sensitive to line lengths corresponding to the angular ranges given above. Apparently the neon spreading takes place only if the line segment has straight continuations at each end and if the continuations differ from the segment in brightness or color. If the continuations are misaligned, the averaging and spreading disappear; they also disappear if there is only one continuation. When a colored cross replaces an intersection in a black grid, the color spreads from both lines of the cross, but the spreading is more extensive than just the sum of the neon flanks along the lines. This result suggests that the visual cells detecting each arm of the cross may interact.

In 1984 Redies, Spillmann and their associate Kristian Kunz continued the study of neon flanks. Two of their figures appear in Figure 7. In the one at the left, notice that near the center of the star the blue lines are close enough for the color to spread between them, making a blue ring; farther out the blue lines provide only neon flanks. Neon spreading can also be seen in the figure at the right wherever the blue segments either are lined up or are shifted from one another in an orderly way. Where the positions of the segments are scrambled, you see only neon flanks.

Following clues from other studies of vision, Redies, Spillmann and Kunz proposed that the "end-stopped" cells in the visual cortex are responsible for the dulling of a colored line when neon spreading or neon flanks appear, and that they may figure in the entire illusion. Most cells in the visual cortex have the assignment of detecting lines, but the end-stopped cells have the particular assignment of detecting short lines. The visual field of an end-stopped cell has an elongated central region with inhibitory zones at each end [see Figure 6]. If both end zones are activated, they inhibit the activity of the central region. The design renders the cells sensitive to short lines, making them promising candidates in explanations of neon spreading.

When a line excites a cell, the line is said to have been "projected" onto it. (To be sure, the only true optical image lies on the retina only electrical signals reach the brain. But the idea of projection simplifies the description of how a cell works.) Suppose that a short red line on a white background lies on your direct line of sight and that its projection is on one of the end-stopped cells that monitor foveal viewing. If the projection is skewed with respect to the long axis of the cell, the cell ignores it, but if the projection is aligned with the axis, the cell sends a signal deeper into the visual system. When the projection occupies only the central region of the cell, the signal is maximum, conveying the information that a dark line lies on a brighter background. What you then perceive is a short red line that contrasts well with the background.

Now suppose that the red line has black continuations at each end and that the continuations pass through the end zones of the cell. The end zones inhibit the discharge triggered by activation of the central region, and so the output signal is less than maximum, which indicates that the line is not as dark as it was when it stood alone. What you now perceive is a red line that has less contrast with the background: it is dulled, as is the case when either neon spreading or neon flanks appear in the illusory illustrations.

An end-stopped cell may dull a line, but can it also spread the color perpendicularly to the line? No one knows yet, but a few months ago Redies wondered whether an end-stopped cell might interact with other, adjacent cells to spread the color of the line onto the adjacent background. One check will be to determine how large an angle a line can occupy before its projection exceeds the central region of an end-stopped cell. Does the angle match the upper limit of about 35 minutes that Redies and Spillmann reported for neon spreading in foveal viewing?

Other questions also remain. When a black grid with colored crosses is tilted, why does neon spreading weaken? Does the result reveal that end-stopped cells respond primarily to horizontal and vertical lines? Why does the spreading also disappear if a cross is encircled with a black line? And what visual cells are responsible for subjective contours?

It seems to me that the brightness rules for neon spreading can be explained by the end-stopped-cell model, if one assumes that the extent of the spreading is related to the amount of inhibition by the end zones. For example, recall that for the illusion to appear on a bright background, the colored lines must not be as dark as the black lines. To see why, again consider a short red line with black continuations. The fact that the continuations are quite dark means that the end zones diminish the output from the cell to a great extent, which may mean that the neon spreading and the dulling of the red line are enough to be perceptible. Now consider how a short black line with red continuations affects a cell. The central line, being quite dark, tends to make the cell fire strongly, and the continuations, being lighter than the central line, produce only a small inhibition- and so also less neon spreading and less dulling of the central line. Although this line of reasoning is interesting, it has not yet been experimentally tested by neurophysiologists.

Bibliography

A NEW VISUAL ILLUSION: NEONLIKE COLOR SPREADING AND COMPLEMENTARY COLOR INDUCTION BETWEEN SUBJECTIVE CONTOURS. H. E. J. M. van Tuijl in Acta Psychologica, Vol. 39, pages 441-445; 1975.

THE NEON COLOR EFFECT IN THE EHRENSTEIN ILLUSION. Christoph Redies and Lothar Spillmann in Perception, Vol. 10, pages 667-681; 1981.

COLORED NEON FLANKS AND LINE GAP ENHANCEMENT. C. Redies, L. Spillmann and K. Kunz in Vision Research, Vol. 24, No. 10, pages 1301-1309; 1984.

CORTICAL DYNAMICS OF THREE-DIMENSIONAL FORM, COLOR, AND BRIGHTNESS PERCEPTION: 1. MONOCULAR THEORY. Stephen Grossberg in Perception & Psychophysics, Vol. 41, No. 2, pages 87-116; 1987.

DISCONTINUITIES ALONG LINES: PSYCHOPHYSICS AND NEUROPHYSIOLOGY. Christoph Redies in Neuroscience and Biobehavioral Reviews, Vol. 13, No. 1, pages 17-22; Spring, 1989.

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.