Abstract
One hundred and fifty years ago, Helmholtz reported that when viewing the world through a telestereoscope (which increases the effective interocular distance), the scene appears as if the observer were looking “at a very exquisite and exact model”. In contrast, Glennerster et al (Current Biology 16 2006 428–432) claimed that observers “failed to notice” the changing size of a scene (altered by a factor of 4:1 in their virtual reality set-up) and they concluded that we “ignore motion and stereo cues in favour of a fictional stable world”. Our experiments were designed to resolve these apparently incompatible findings. Using a wide-field stereoscope (70 × 70 deg), observers viewed the captured stereo images of a richly furnished room containing many familiar objects of known size. The telestereoscopic image pairs were created using camera separations of between 0.5 and 4.0 times the usual interocular separation, so that the depicted size of the room could be varied from twice to one quarter the size of the original. Observers were asked to estimate the apparent size of the depicted rooms and the objects within. All observers reported clear differences in the apparent size of the room and the size of objects within the room with changes in camera separation. As a more powerful test of the role of stereo cues in our perception of the differently sized rooms, we asked observers to set a disparity-defined ridge surface so that the flanks appeared to meet at 90deg. Since the disparity gradients of a 90deg ridge surface in a half-sized room are doubled, observers ought to see the flanks as steeper, if no account is taken of room size. In fact, all our observers' settings showed a high degree of constancy (60–80%) and we find no evidence to support the claim that stereo cues are ignored.