Abstract
The leaning tower illusion occurs when two identical images of the Leaning Tower of Pisa are placed side-by-side to each other. When this occurs, the two towers appear to be leaning at two different angles. Discovered by Kingdom, Yoonessi & Gheorghui (2007), they proposed the illusion works because our visual system tends to group two side-by-side objects as being the same scene. As both tower images have identical parallel outlines, yet cannot be parallel in the physical world, our knowledge of perspective compensates, causing the perception of the two towers diverging. Here I wanted to explore if our knowledge of perspective is the cause of the illusion. Before testing the hypothesis, a measure of illusion strength needed to be defined. This led to the thought: could increasing the distance between the two images cause the illusion to weaken? 20 observers performed a method of adjustment task, where they increased or decreased the distance between the two images. Each participant performed 100 trials (50 up/50 down) per condition. On average, participant still reported the illusion with a gap of 8.7° (sd = .57) between the two images. Next, a CGI version of the tower was created, allowing for manipulations of perspective and image structure. The CGI image results in a similar gap size for the illusion. However, when the sense of perspective is enhanced by reducing the convergence point, then illusion strength is reduced, requiring a smaller distance (6.8°, sd = .38, d = 3.9) between the two images to maintain the illusion. A similar result is also obtained when the perspective is reduced by increasing the convergence point (7.2°, sd = .48, d = 2.8). Other image manipulations (low & high-pass filtering, inverting image, distorted horizontal and vertical contours) had no effect on the illusion strength. These data suggest that the leaning tower illusion requires our knowledge of perspective to cause the illusion.
Research is funded by NSERC.