We appreciate illusionistic art without being deluded; we know that what we are seeing is mere artifice; we experience illusion because we are in collusion with the artist. In contrast to illusionistic art, we appreciate trompe l'œil because we were initially deluded. Mental collusion has very little to do with our appreciation of these creations, which, if we appreciate them at all, are reminders of the fallibility of knowledge acquired through the senses.

Having discussed the nature of delusion in trompe l'œil and the nature of collusion in illusionism, we turn now to a third anomalous state of mind we sometimes experience when viewing a painting, namely, perceptual paradox. In the preceding chapter, we discussed the sorts of deformations we perceive in paintings despite the fact that in general perspective is robust. Although it seems paradoxical that, at one and the same time as one passes in front of a painting, the scene appears to turn and to remain the same, it is possible because not all aspects of our perception are processed by the same mechanisms; there is a division of labor that usually works so well that it is not noticed. The well-trained bureaucracy of the mind can deal with practically all the contingencies that occur in our environment. But when psychologists contrive devices that stimulate us in unusual ways, ways that are unlikely to arise in our environment, perception can be made to reveal the division of labor without which it could not function. The rules by which the bureaucracy has been accustomed to work may now lead to incompatible decisions.

For instance, take the waterfall illusion: On a screen, we display an unbroken series of horizontal black stripes moving downward. After a viewer stares at this stylized waterfall for a while, the motion is stopped, and he or she is asked to report what the display looks like. The display looks paradoxical: The stripes appear to be moving upward, but at the same time each stripe does not seem to be changing its position relative to the frame of the screen. This sort of perceptual decomposition has led to the hypothesis, now well-supported by experimental evidence, that motion and location in space are processed by different mechanisms Attneave (1974). No less interesting, though, is the following implication of the phenomenon: The visual system makes no attempt to reconcile these contradictory pieces of information about the world; we experience these unreconciled contradictories, this perceptual paradox, as illusion.

		Fig.6.19 The vertical-horizontal illusion.

It is important to keep in mind the distinction between illusion as perceptual error, which we have called delusion, and illusion as an awareness of perceptual error, which we have called collusion. As we have seen in our discussion of the vase-face illusion, most illusions do not provide us with the experience of illusion unless we are given an opportunity for collusion, an understanding of what we are to expect to experience. Take, for instance, the vertical-horizontal illusion (Figure 6.19). The vertical looks longer than the horizontal: That is a perceptual error. But it is only when you are put in a position to experience a perceptual dilemma - such as being told to rotate the drawing slowly, and becoming aware of the changes in the relative lengths of the two lines during the rotation, while realizing that the drawing itself is invariant that you may experience an illusion: This is a metaperceptual experience: It is an awareness of perceptions; the visual system does not try to reconcile the two experiences, and that non-reconciliation gives rise to the experience of illusion.

The impression of following in a painting is one of those rare instances where an object spontaneously gives rise to the experience of an illusion. My explanation of this phenomenon is schematically summarized in Figure 6.20. The experience of the picture turning stems from two perceptions: On the one hand, even though we are walking around the picture, we perceive the spatial layout of the represented scene as if it remains unchanged. This is what we have called in Chapter 5 the robustness of perspective (which we will discuss at length later). On the other hand, even though the spatial layout of the scene remains unchanged, we perceive our own motion in space as we walk past the picture. The experience of rotation of the painting is one way to resolve this dilemma: To perceive the scene as being invariant while we are walking past it, we must perceive the picture to be rotating.

Fig.6.20 The double dilemma of picture perception that leads to the experience that the turning of the picture, as we walk past it, is illusory.

As Gombrich has pointed out,¹¹ this resolution of the dilemma is reinforced in paintings that contain objects with a pronounced aspect such as a foreshortened gun-barrel, a pointing finger, a human eye, or a road receding into the distance from the center foreground to the horizon (such as the Rousseau painting discussed toward the end of the preceding chapter). These are objects that are represented in an orientation that is visually unstable: If you are looking down the barrel of a gun, you need to take only a very small step sideways in order not to be looking down the barrel of the gun. We say here that objects are represented in a visually unstable orientation by analogy with objects that are in a physically unstable equilibrium, such as a pyramid that has been balanced on its tip: You need to apply only a minuscule change to the forces exerted upon the pyramid to cause it to fall.¹² It is quite natural, therefore, that we perform the unconscious inference: The object is shown in a visually unstable orientation; I am moving enough to destabilize the view; the view is not destabilized; therefore, the object must be turning to follow me.

But that solution to the dilemma is, so to speak, shortsighted, because it gives rise to another dilemma: If the picture is turning, how is it that it looks so well attached to the wall? Why does its relation to the room not change? The experience of illusion stems from the visual system's inability to resolve this dilemma within a dilemma. The fact that the face is simultaneously transformed to become thinner at the same time is, apparently, not sufficiently noticeable to counteract the perceived following.

Although we have shown that some distortions do take place in the perception of paintings that are viewed by moving observers, the balance of perecptual interpretations tends to favor the robustness of perspective that emerges most clearly from our analysis. As we will see presently, it is this robustness that is probably the most important justification for not using Brunelleschi peepholes to view perspective paintings.

---

¹¹ See Gombrich's essay, "Perception and the Visual Deadlock," in Gombrich (1963); also see Gombrich (1973).

¹² This formulation is inspired by Shepard (1981, pp. 307-9), who refers to René Thom's (1975) catastrophe theory. A similar notion can be found in the work of Huffman (1971), who calls accidentals what we have called "visually unstable orientations." See also Draper (1980). Anstis, Mayhew, and Morley (1969) have shown that the position of the iris and pupil with respect to the eye socket and the eyelids is sufficient to determine the perceived direction of a gaze. If the iris and the pupil are centered, we feel that the person is looking directly at us. Hence, if we move and the gaze remains directed at us, we perceive the gaze to be following us.