Abstract
Stereoscopic depth perception can be as fast as the perception of luminance (Caziot et al 2015 JEP:HPP) and disparity perturbations have fast effects on reaching (Greenwald, Knill & Saunders 2005 Vis Res). Yet stereo can also be slow: depth was tracked much more slowly than horizontal or vertical position by the eyes (Mulligan, Stevenson & Cormack 2013 Hum Vis and Electronic Imaging) or hands (Bonnen, Huk & Cormack 2015 J Neurophys), and modulation in depth is not visible above 5 to 10 Hz (Kane, Guan & Banks, 2014, J Neurosci). What accounts for these extraordinary discrepancies? We hypothesized that stereoscopic depth is estimated very quickly when an object first appears, but that as it persists, its stereoscopic depth is integrated over time with the long time constants measured previously. We employed continuous target-tracking to measure latency in the response to changes of position in 3D. Non-stereo depth cues were not informative. Targets followed a 120-sec random walk in 3D (normally distributed steps in x, y, and z, SD 3mm, 90 Hz). Large discrete jumps were introduced randomly every 1-4 sec. The purpose of the jump was to re-start depth integration, so we predicted that latencies would be shorter for jumps than during random-walk motion. We did not confirm the hypothesis: instead, some participants were very fast to track depth, as fast as for horizontal or vertical (peak correlation between stimulus and tracker at 220-270 ms), while others were selectively slow for depth (450 ms or more). These differences were similar for continuous tracking and jumps within an individual. However, the instructions mattered: when asked to be accurate in depth, latencies decreased for depth tracking. We conclude that the use of disparity to track targets in depth is not universally slow.
Meeting abstract presented at VSS 2018