long as it is still clearly recognizable?
We then used a second, more direct test called popout, which psychologists employ to determine whether an effect is truly perceptual (or only conceptual). If you look at Figure 3.1 you will see a set of tilted lines scattered amid a forest of vertical lines. The tilted lines stick out like a sore thumb—they “pop out.” Indeed, you can not only pick them out of the crowd almost instantly but can also group them mentally to form a separate plane or cluster. If you do this, you can easily see that the cluster of tilted lines forms the global shape of an X. Similarly in Figure 3.2, red dots scattered among green dots (pictured here as black dots among gray dots) pop out vividly and form the global shape of a triangle.
In contrast, look at Figure 3.3. You see a set of Ts scattered amid the Ls, but unlike the tilted lines and colored dots of the previous two figures, the Ts don’t give you the same vivid, automatic “here I am!” popout effect, in spite of the fact that Ls and Ts are as different from each other as vertical and tilted lines. You also cannot group the Ts nearly as easily, and must instead engage in an item-by-item inspection. We may conclude from this that only certain “primitive,” or elementary, perceptual features such as color and line orientation can provide a basis for grouping and popout. More complex perceptual tokens such as graphemes (letters and numbers) cannot do so, however different they might be from each other.
FIGURE 3.1 Tilted lines embedded in a matrix of vertical lines can be readily detected, grouped, and segregated from the straight lines by your visual system. This type of segregation can occur only with features extracted early in visual processing. (Recall from Chapter 2 that three-dimensional shape from shading can also lead to grouping.)
To take an extreme example, if I showed you a sheet of paper with the word love typed all over it and a few hates scattered about, you could not find the hates very easily. You would have to search for them in a more or less serial fashion. And even as you found them, one by one, they still wouldn’t segregate from the background the way the tilted lines or colors do. Again, this is because linguistic concepts like love and hate cannot serve as a basis for grouping, however dissimilar they might be conceptually.
FIGURE 3.2 Dots of similar colors or shading can also be grouped effortlessly. Color is a feature detected early in visual processing.
Your ability to group and segregate similar features probably evolved mainly to defeat camouflage and discover hidden objects in the world. For instance, if a lion hides behind a mottling of green foliage, the raw image that enters your eye and hits your retina is nothing but a bunch of yellowish fragments broken up by intervals of green. However, this is not what you see. Your brain knits together the fragments of tawny fur to discern the global shape, and activates your visual category for lion. (And from there, it’s straight on to the amygdala!) Your brain treats the probability that all those yellow patches could be truly isolated and independent from each other as essentially zero. (This is why a painting or a photograph of a lion hiding behind foliage, in which the patches of color actually are independent and unrelated, still makes you “see” the lion.) Your brain automatically tries to group low-level perceptual features together to see if they add up to something important. Like lions.
FIGURE 3.3 Ts scattered among Ls are not easy to detect or group, perhaps because both are made up of the same low-level features: vertical and horizontal lines. Only the arrangement of the lines is different (producing corners versus T-junctions), and this is not extracted early in visual processing.
Perceptual psychologists routinely exploit these effects to determine whether a particular visual feature is elementary. If the feature gives you popout and grouping, the brain must be extracting it early in sensory processing. If popout and grouping are muted or absent, higher-order sensory or even conceptual processing must be involved in representing the objects in question. L and T share the same elementary features in common (one short short horizontal and one short vertical line touching at right angles); the main things that distinguish them in our minds are linguistic and conceptual factors.
So let’s get back to Mirabelle. We know that real colors can lead to grouping and popout. Would her “private” colors be able to elicit the same effects?
To answer this question I devised patterns similar to the one shown in Figure 3.4: a forest of blocky 5s with a few blocky 2s scattered among them. Since the 5s are just mirror images of the 2s, they are composed of identical features: two vertical lines and three horizontal ones. When you look at this image, you manifestly do not get popout; you can only spot the 2s through item-by-item inspection. And you can’t easily discern the global shape—the big triangle—by mentally grouping the 2s; they simply don’t segregate from the background. Although you can eventually deduce logically that the 2s form a triangle, you don’t see a big triangle the way you see the one in Figure 3.5, where the 2s have been rendered in black and the 5s in gray. Now, what if you were to show Figure 3.4 to a synesthete who claims to experience 2s as red and 5s as green? If she were merely thinking of red (and green) then, just like you and me, she wouldn’t instantly see the triangle. On the other hand if synesthesia were a genuinely low-level sensory effect, she might literally see the triangle the way you and I do in Figure 3.5.
For this experiment we first showed images much like Figure 3.4 to twenty normal students and told them to look for a global shape (made of little 2s) among the clutter. Some of the figures contained a triangle, others showed a circle. We flashed these figures in a random sequence on a computer monitor for about half a second each, too short a time for detailed visual inspection. After seeing each figure the subjects had to press one of two buttons to indicate whether they had just been shown a circle or a triangle. Not surprisingly, the students’ hit rate was about 50 percent; in other words, they were just guessing, since they couldn’t spontaneously discern the shape. But if we colored all the 5s green and all the 2s red (in Figure 3.5 this is simulated with gray and black), their performance went up to 80 or 90 percent. They could now see the shape instantly without a pause or a thought.
The surprise came when we showed the black-and-white displays to Mirabelle. Unlike the nonsynesthetes, she was able to identify the shape correctly on 80 to 90 percent of trials—just as if the numbers were actually colored differently! The synesthetically induced colors were just as effective as real colors in allowing her to discover and report the global shape.2 This experiment provides unassailable proof that Mirabelle’s induced colors are genuinely sensory. There is simply no way she could fake it, and no way it could be the result of childhood memories or any of the other alternative explanations that have been proposed.
