Abstract
In the absence of looming cues, observers ignore large changes in the size of a virtual scene around them and have large biases in size judgments as a result (Glennerster et al, 2006, Current Biology 16, 428–432; Rauschecker et al, 2006, Journal of Vision, 6, 1471–1485). At first sight, these results suggest that observers have poor access to information about absolute distance from stereo and motion parallax. However, we show here that under similar conditions observers have low thresholds when asked to detect the expansion/contraction of the scene. In an immersive virtual reality environment, subjects viewed a static scene binocularly and with motion parallax from free head movement. The scene changed size after a blank ISI. The centre of expansion was the cyclopean point, so all objects in the scene remained the same angular size. The task was to identify whether the target object (and the rest of the scene) was closer or farther away in the second interval. Thresholds for this task (Weber fractions 10–33%) were best when the target was close to the observer and were invariably lower than when the target was presented alone, without the scene. We also measured thresholds for detecting changes in the location of an object relative to the scene, while its absolute distance remained constant. As expected, these relative thresholds improved with the proximity of the target to neighbouring objects. In a separate experiment, subjects judged the change in distance of an object presented in two intervals while, independently, the scene changed in size by a factor of between 0.25 and 4. The pattern of biases could be predicted from the absolute and relative thresholds measured in the previous experiments, showing that observers did not use absolute and relative cues independently.
Supported by University of Reading and The Wellcome Trust.