created hundreds of visual illusions that have been cunningly devised to help us explore the assumptions that drive perception. Illusions are fun to look at since they seem to violate common sense. But they have the same effect on a perceptual psychologist as the smell of burning rubber does on an engineer—an irresistible urge to discover the cause (to quote what biologist Peter Medawar said in a different context).
Take the simplest of illusions, foreshadowed by Isaac Newton and established clearly by Thomas Young (who, coincidentally, also deciphered the Egyptian hieroglyphics). If you project a red and a green circle of light to overlap on a white screen, the circle you see actually looks yellow. If you have three projectors—one shining red, another green, and another blue—with proper adjustment of each projector’s brightness you can produce any color of the rainbow—indeed, hundreds of different hues just by mixing them in the right ratio. You can even produce white. This illusion is so astonishing that people have difficulty believing it when they first see it. It’s also telling you something fundamental about vision. It illustrates the fact that even though you can distinguish thousands of colors, you have only three classes of color-sensitive cells in the eye: one for red light, one for green, and one for blue. Each of these responds optimally to just one wavelength but will continue to respond, though less well, to other wavelengths. Thus any observed color will excite the red, green, and blue receptors in different ratios, and higher brain mechanisms interpret each ratio as a different color. Yellow light, for example, falls halfway in the spectrum between red and green, so it activates red and green receptors equally and the brain has learned, or evolved to interpret, this as the color we call yellow. Using just colored lights to figure out the laws of color vision was one of the great triumphs of visual science. And it paved the way for color printing (economically using just three dyes) and color TV.
My favorite example of how we can use illusions to discover the hidden assumptions underlying perception is shape-from-shading (Figure 2.3). Although artists have long used shading to enhance the impression of depth in their pictures, it’s only recently that scientists have begun to investigate it carefully. For example, in 1987 I created several computerized displays like the one shown in Figure 2.3—arrays of randomly scattered disks in a field of gray. Each disk contains a smooth gradient from white at one end to black on the other, and the background is the exact “middle gray” between black and white. These experiments were inspired, in part, by the observations of the Victorian physicist David Brewster. If you inspect the disks in Figure 2.3, they will initially look like a set of eggs lit from the right side. With some effort you can also see them as cavities lit from the left side. But you cannot simultaneously see some as eggs and some as cavities even if you try hard. Why? One possibility is that the brain picks the simplest interpretation by default, seeing all of the disks the same way. It occurred to me that another possibility is that your visual system assumes that there is only a single light source illuminating the entire scene or large chunks of it. This isn’t strictly true of an artificially lit environment with many lightbulbs, but it is largely true of the natural world, given that our planetary system has only one sun. If you ever catch hold of an alien, be sure to show her this display to find out if her solar system had a single sun like ours. A creature from a binary star system might be immune to the illusion.
FIGURE 2.3 Eggs or cavities? You can flip between the two depending on which direction you decide the light is shining from, right or left. They always all flip together.
So which explanation is correct—a preference for the simpler interpretation, or an assumption of a single light source? To find out I did the obvious experiment of creating the mixed display shown in Figure 2.4 in which the top and bottom rows have different directions of shading. You will notice that in this display, if you get yourself to see the top row as eggs, then the bottom row is always seen as cavities, and vice versa, and it is impossible to see them all simultaneously as eggs or simultaneously as cavities. This proves it’s not simplicity but the assumption of a single light source.
FIGURE 2.4 Two rows of shaded disks. When the top row is seen as eggs, the bottom row looks like cavities, and vice versa. It is impossible to see them all the same way. Illustrates the “single light source” assumption built into perceptual processing.
FIGURE 2.5 Sunny side up. Half the disks (light on top) are seen as eggs and half as cavities. This illusion shows that the visual system automatically assumes that light shines from above. View the page upside down, and the eggs and cavities will switch.
It gets better. In Figure 2.5 the shaded disks have been shaded vertically rather than horizontally. You will notice that the ones that are light on top are nearly always seen as eggs bulging toward you, whereas the ones that are dark on top are seen as cavities. We may conclude that, in addition to the single-light-source assumption revealed in Figure 2.4, there is another even stronger assumption at work, which is that the light is shining from above. Again, makes sense given the position of the sun in the natural world. Of course, this isn’t always true; the sun is sometimes on the horizon. But its true statistically—and it’s certainly never below you. If you rotate the picture so it’s upside down, you will find that all the bumps and cavities switch. On the other hand, if you rotate it exactly 90 degrees, you will find that the shaded disks are now ambiguous as in Figure 2.4, since you don’t have a built-in bias for assuming light comes from the left or the right.
Now I’d like you to try another experiment. Go back to Figure 2.4, but this time, instead of rotating the page, hold it upright and tilt your body and head to the right, so your right ear almost touches your right shoulder and your head is parallel to the ground. What happens? The ambiguity disappears. The top row always looks like bumps and the bottom row as cavities. This is because the top row is now light on the top with reference to your head and retina, even though it’s still light on the right in reference to the world. Another way of saying this is that the overhead lighting assumption is head centered, not world centered or body-axis centered. It’s as if your brain assumes that the sun is stuck to the top of your head and remains stuck to it when you tilt your head 90 degrees! Why such a silly assumption? Because statistically speaking, your head is upright most of the time. Your ape ancestors rarely walked around looking at the world with their heads tilted. Your visual system therefore takes a shortcut; it makes the simplifying assumption that the sun is stuck to your head. The goal of vision is not to get things perfectly right all the time, but to do get it right often enough and quickly enough to survive as long as possible to leave behind as many babies as you can. As far as evolution is concerned, that’s all that matters. Of course, this shortcut makes you vulnerable to certain incorrect judgments, as when you tilt your head, but this happens so rarely in real life that your brain can get away with being lazy like this. The explanation of this visual illusion illustrates how you can begin with a relatively simple set of displays, ask questions of the kind that your grandmother might ask, and gain real insights, in a matter of minutes, into how we perceive the world.
Illusions are an example of the black-box approach to the brain. The metaphor of the black box comes to us from engineering. An engineering student might be given a sealed box with electrical terminals and lightbulbs studding the surface. Running electricity through certain terminals causes certain bulbs to light up, but not in a straightforward or one-to-one relationship. The assignment is for the student to try different combinations of electrical inputs, noting which lightbulbs are activated in each case, and from this trial-and-error process deduce the wiring diagram of the circuit inside the box without opening it.
In perceptual psychology we are often faced with the same basic problem. To narrow down the range of hypotheses about how the brain processes certain kinds of visual information, we simply try varying the sensory inputs and noting what people see or believe they see. Such experiments enable us discover the laws of visual function, in much the same way Gregor Mendel was able to discover the laws heredity by cross-breeding plants with
