Abstract
When combined with information about distance, binocular disparity can be used to judge shape. It is usually assumed that disparity information is used veridically, with observed systematic errors attributed to mis-estimation of viewing distance. Here we present data that challenges this assumption, demonstrating perceptual biases that appear to be due to an object?s elevation in a scene.
Observers viewed a simple scene consisting of a pair of dots arranged one above the other, 5cm apart. Depth was conveyed via binocular disparity using a Wheatstone stereoscope. The dots were presented at different depths, on average 10cm apart (21 min arc at D = 1m). For example, in one configuration the upper dot was as far behind the fixation plane as the lower dot was in front. Using a 2AFC method, observers were asked in which interval the distance in depth between the dots was longer. We compared conditions in which a test stimulus was presented in the same configuration as the fixed standard stimulus (e.g. test and standard: upper dot far, lower dot near), with conditions in which the configurations were opposite (e.g. test: upper far, lower near; standard: upper near, lower far).
Points of subjective equality (PSE) were collected from fitted psychometric functions. When configurations were the same for test and standard, the PSE was close to the value of the standard, as expected. When the configurations were opposite, some observers showed biases consistent with target separation being perceived as relatively larger for the upper-far, lower-near configuration. These results suggest that relative distance perception can vary depending on whether the same disparity information is presented in the upper or lower part of the scene. This could be linked to the known bias in the distributions of disparities present in real scenes (Hibbard & Bouzit, 2005, Spatial Vision).