Abstract
Behavioral perception-action dissociations are widely used to test models of high-level vision, but debates concerning their interpretation have underestimated the role of multisensory mechanisms in such tests. Sensorimotor tasks engage multisensory processing in fundamentally different ways than perceptual tasks, and these differences can modulate dissociations in different ways based on task features. To test this idea, we compared perception and action using a well-understood size-contrast effect, the Uznadze illusion, and manipulated both unimodal and crossmodal stimulation as well as conditions that are known to favor or hinder multisensory integration. Results demonstrated that similar or dissociable effects on perception and action can be observed depending on factors that are known to affect multisensory processing. Specifically, such factors can cause a visual task to be affected by the illusion or remain fully unaffected, whereas a visuomotor task can be affected by the illusion, remain immune from the illusion, or, unexpectedly, even show a clear reverse effect. These findings provide a novel perspective on a long standing debate in behavioral cognitive neuroscience.