Abstract
Spatial context is critical for telling how big a visual object is, although it may also cause the perceived size to diverge dramatically from the true dimensions. Interestingly, responses in the primary visual cortex (V1) mirror such illusory perception; however, the stage of processing that leads to such neural correlates remains unknown. Here, we tested the involvement of higher level processing in a Ponzo-like illusion, by quantifying the effect of manipulating depth cues and inversion of the whole scene. We report a stronger illusion for realistic compared with simpler backgrounds, and for upright compared with inverted scenes (except for scenes where the target objects appeared on the ceiling or in the sky). Next, using functional MRI, we tested the effect of inversion on V1 responses. Inverted scenes elicited a smaller extent of activation in V1 compared with upright scenes, consistent with their perceived sizes. Taken together, since the inversion should disrupt the high-level processing while keeping the low-level features intact, our findings demonstrate that Ponzo-like illusions involve high-level processes that integrate contextual depth cues and visual experience, thereby modulating the object's neural representation in V1.