Abstract
Illusory size perception is usually an outcome of complex processing that incorporates the object and its spatial context. The primary visual cortex (V1) has been found to reflect illusory perception; however, the stage of processing that leads to such projection is not fully understood. To test the involvement of higher-level processing in a custom realistic Ponzo-like size illusion, we first conducted a psychophysical experiment. As stimuli, we used one photorealistic road scene and three simplified versions of the same scene formed by 2-D shapes and lines, and their 180° rotated versions. Since the rotated versions consist of the same low-level image characteristics as the upright scene, any potential difference could only mean high-level involvement. Behavioral results showed that the illusion strength decreased significantly when the image was inverted and as the number of high-level components in the background decreased, as expected. Next, using functional MRI, we tested the effect of rotation on V1 activity. We presented the realistic background and its phase-scrambled version with the two rotation conditions. The stimuli that are subject to the size illusion flashed on and off on top of the backgrounds. We found a smaller extent of neuronal activation in V1 for the stimulus on the inverted than on the upright scene, in line with the perceived sizes. Then we compared the observed V1 response with the simulated responses to various physically larger and smaller stimuli. The stimulus size giving the maximum correlation between simulated and observed responses was larger for the upright than the rotated condition, showing a pattern similar to the perceived sizes. Taken together, our findings suggest that realistic Ponzo-like illusions involve high-level processing, which is being fed back to V1. These results provide support for the misapplied size constancy theory, which explains Ponzo-like illusions with respect to prior knowledge and expectations.