The Ponzo illusion and the perception of orientation

Illusions have long been considered to provide insights into normal visual processing. Thus Richard Gregory asked, “Can it be that illusions arise from informationprocessing mechanisms that under normal circumstances make the visible world easier to comprehend?” (Gregory, 1968, p. 66). A critical issue has been which normal information-processing mechanisms are reflected by specific illusions. The Ponzo illusion, illustrated in Figure 1, has been claimed to reflect many different fundamental visual processes, from spatial frequency filtering in cortical area V1 to the linear perspective and the inappropriate application of size constancy. We propose a new theory of the Ponzo illusion that is based on the mechanisms of orientation perception. Specifically, we propose that the Ponzo illusion is due to the misperception of orientation induced by local visual cues. A consequence of the misperception of orientation is that the bar near the apex (i.e., on the right) of the Ponzo illusion illustrated in Figure 1 appears longer than the bar near the base (i.e., the bar on the left). We will suggest that the mechanisms that normally help us perceive constant orientation despite changes in retinal orientation (tilt constancy) may underlie this illusion. In this report, we will first present this new account of the Ponzo illusion. Second, we will test our theory against theories based on linear perspective and size constancy (e.g., Gillam, 1980; Gregory, 1963), low-pass filtering (Ginsburg, 1984), assimilation (Girgus & Coren, 1982; Pressey & Epp, 1992), and relative size comparisons (Künnapas, 1955). Finally, we will extend our theory to account for other illusions, some of which are illustrated in Figure 1. The role of local visual cues in the perception of orientation was illustrated in a classic study by Gibson (1937). This effect is termed the tilt induction effect. Observers were presented a stimulus such as that in Figure 2A. The task was to set the shorter line to be vertical. Gibson found that the observers distorted their adjustments in the direction of the context lines. That is, the tilted context distorted the observer’s sense of vertical. Thus, with stimuli such as that in Figure 2A, the observers perceived the vertical line as tilted in a counterclockwise direction and therefore adjusted it too far in a clockwise direction to make it appear vertical. The opposite was true of stimuli like that shown in Figure 2B. When the context is presented before the line to adjust, the effect is called the tilt-after effect (Gibson & Radner, 1937). The Zöllner illusion has long been considered an example of the tilt induction effect (Day, 1972; Howard, 1982, pp. 156–157). In Figure 1 (Zöllner illusion), the horizontal lines appear slanted in the direction opposite the context lines. Note that the effect of context is local, so that in Figure 1, the context lines on the top determine the perceived orientation of the line on the top and those on the bottom determine the perceived orientation of the line on the bottom. We propose that the Ponzo illusion is caused by the same mechanism as the tilt induction effect and the Zöllner illusion. To understand the relation between the tilt induction effect and the Ponzo illusion consider Figure 3. In Figure 3A, because of the tilt induction effect, the vertical line will appear slanted in the direction opposite the slanted line (i.e., it will appear tilted in a counterclockwise direction). We have shown that this effect will cause the dot on the top to be perceived as shifted to the left, as compared with the dot on the bottom (Shimamura & Prinzmetal, 2000). In Figure 3B, the dot on the top will appear