Abstract
It has been reported that humans can accurately perceive heading during simulated eye rotation using information from optic flow alone. Here we investigated what roles spatial frequency content, surface structure, foreground motion, and expansion information play in heading perception during rotation. The display (110°H × 94°V) simulated an observer walking on a straight path over a ground plane (depth range: 1.4–50 m) at 2 m/s while fixating a target off to one side (mean R/T ratios: ±0.8, ±1.6, ±2.4) under six display conditions, in which (1) 50 random dots were uniformly distributed on the ground plane (2) the same number of dots was placed at each distance in depth to increase foreground motion (3) the dots were uniformly distributed on the image plane to remove static depth cues (4) the ground plane was replaced with the lower half of a 3D random-dot cloud that had the same velocity profile as (3) to remove the surface structure defined by dot motion in (3) (5) a ground similar to (3) but consisting of discs that expanded during the trial; and (6) a textured ground having the same spatial frequency power spectrum as (5). At the end of each 1s trial, observers used the mouse to indicate their perceived heading along a horizontal line in the center of the display. Mean heading bias was significantly smaller for (6) than for (5), indicating that spatial frequency content is not crucial for accurate heading perception during rotation. Mean heading bias was similar for (3), (4), and (5), indicating that expansion and surface structure information do not affect heading perception. Lastly, mean heading bias was borderline significantly smaller for (2) than (1), revealing that foreground motion plays a modest role in accurate heading perception. We conclude that dense motion parallax information is most important for accurate heading perception during rotation.
Supported by: Hong Kong Research Grant Council, HKU 7478//08H.