Abstract
Recent studies have argued that trajectories of motion-in-depth stimuli are perceived quite inaccurately (e.g. Harris & Drga 2005). The perceptual errors could have been caused by the absence of appropriate changes in image size (looming), the absence of appropriate changes in focus cues (blur and accommodation), and the use of an estimation procedure that is subject to the response-mapping problem. To better understand the causes of these reported errors, we manipulated the availability of disparity, looming, and focus cues, and used a response measure that should not be subject to the mapping problem. Observers viewed stimuli that moved in elliptical paths in depth and judged whether the path was too compressed or stretched to be circular. Real world motion of a LED served as a condition in which all trajectory cues were consistent. In a second condition, an equivalent stimulus was shown on a computer display; in this case focus cues specified a frontoparallel path while other cues were veridical. In a third condition, the computer-displayed stimulus had constant angular size; in this case, looming and focus cues specified a frontoparallel path. With the real moving LED, responses indicated very accurate percepts with binocular viewing and reasonably accurate percepts with monocular viewing. With the computer-displayed target with appropriate looming, responses with binocular and monocular viewing were somewhat less accurate than for the corresponding real motion. With fixed angular size, responses with binocular viewing were less accurate than the other binocular conditions and responses with monocular viewing were very inaccurate. As cues to depth were removed observers perceived circular paths as increasingly compressed in depth. Thus, the visual system uses many cues to estimate 3D trajectories, so accurate percepts occur only when all cues specify the same path.