Abstract
The visual motion system can use disparity to segment a signal from noise (Snowden, Rossiter 1999). Here we use a stereomotion coherence task to examine how properties of segmentation vary as a function of disparity. Experiments: Subjects performed a left/right motion discrimination task. In Exp 1, a plane of signal dots was at fixation while the depth of a noise plane varied (0'–12'); In Exp 2 the signal plane was placed at +/−10'; in Exp 3 the noise plane was replaced by a 3D cloud of variable depth. Results: Exp 1: As expected, subjects' thresholds improved dramatically as disparity separation increased. Exp 2: For crossed disparities, motion discrimination thresholds were higher for noise in front of the signal, indicating that segmentation was more difficult. Exp 3: For signal planes at a fixed distance from the cloud thresholds remained constant, suggesting that the distance to the cloud, but not the depth of the cloud, had a strong effect on performance. The results of Experiments 2 and 3 show that the perception of motion when noise was a 3D cloud produced lower thresholds than when the noise was 2D (both were 2' from the signal plane). Conclusions: Taken together these results suggest a channel model which weights each noise dot by its distance from the signal. The model allows for estimates of disparity channel tuning widths as a function of depth. However, the near/far asymmetry we report requires a low-pass channel for crossed disparities without a corresponding channel for uncrossed disparities.
Supported by NIH grant R01EY007861-15 to L.M.V.