Allocentric Perception of Space and Voluntary Head Movement

Although visual input depends on the position of the eye, at least some neural representations of space in mammals are allocentric, i.e., independent of the observer’s vantage point or motion. By comparing the visual perception of 3D object motion in actively and passively moving human subjects, I show that the motor command contributes to the perception of space in an observer-independent reference frame: observers executing active head movements are more likely to apply minimal-motion criteria relative to an allocentric frame of reference than are observers undergoing similar passive displacements. However, the bias towards an allocentric reference frame decreases or disappears unless self-motion matches the motor command. A fundamental feature of our visual system is that the viewpoint can move through space, either as a result of voluntary motor action on the observer’s part, or passively. While eye rotations result in almost uniform shifts of the entire 2D retinal image, and therefore generate no 3D information, head translations (leading to translations of the center of the eye) give rise to motion parallax, which does provide information on 3D structure and motion. At least two reference frames (rfs) are therefore possible for visual perception: an egocentric or subjective rf centered on the eye and therefore fixed to the head, and an allocentric or objective, earth-fixed rf. Although visual input is in an egocentric rf, it is frequently advantageous to perceive spatial information, such as 3D motion, in an allocentric rf: this would allow, for instance, to differentiate an approaching object from one that is being approached. Since motion in the two rfs differs by the movement of the observer’s eye in space, which in turn depends on head movement, in order to convert to an allocentric rf the visual system must compensate egocentric visual input by information about the observer’s head movement. For this compensation to be possible, information about head movements must be incorporated in spatial vision. To study whether such information is used, one must compare the performance of a moving observer (moving actively or passively) with that of an immobile observer, taking care to keep visual input constant in all conditions. Held and his colleagues compared actively moving to passively moved human observers, finding effects of voluntary action on visuomotor learning [1] (and on visual development in cats [2]). Though comparisons between moving and immobile observers have seemed to show no effect of self-motion on the precision of spatial vision [3], it was recently demonstrated that actively generated motion parallax can lead to different perceptions of 3D shape than the same visual stimulation experienced by an immobile observer [4]. The complement to the perception Date: 26 September 2001.