The Two-Wrongs model explains perception-action dissociations for illusions driven by distortions of the egocentric reference frame

Paul Dassonville, Scott A. Reed

Several studies have demonstrated a dissociation of the effects of illusion on perception and action, with perception generally reported to be susceptible to illusions, while actions are seemingly immune. These findings have been interpreted to support Milner and Goodale's Two Visual Systems model, which proposes the existence of separate visual processing streams for perception and action. However, an alternative interpretation suggests that this type of behavioral dissociation will occur for any illusion that is caused by a distortion of the observer's egocentric reference frame, without requiring the existence of separate perception and action systems that are differently affected by the illusion. In this scenario, movements aimed at illusory targets will be accurate if they are guided within the same distorted reference frame used for target encoding, since the error of motor guidance will cancel with the error of encoding (hence, for actions, two wrongs do make a right). We further test this Two-Wrongs model by examining two illusions for which the hypothesis makes very different predictions: the rod-and-frame illusion (which affects perception but not actions) and the simultaneous-tilt illusion (which affects perception and actions equally). We demonstrate that the rod-and-frame illusion is caused by a distortion of the observer's egocentric reference frame suitable for the cancellation of errors predicted by the Two-Wrongs model. In contrast, the simultaneous-tilt illusion is caused by local interactions between stimulus elements within an undistorted reference frame, precluding the cancellation of errors associated with the Two-Wrongs model such that the illusion is reflected in both perception and actions. These results provide evidence for a class of illusions that lead to dissociations of perception and action through distortions of the observer's spatial reference frame, rather than through the actions of functionally separate visual processing streams.