Purchase this article with an account.
Peter Scarfe, Andrew Glennerster; Optimal estimates of distance and direction are mutually inconsistent. Journal of Vision 2015;15(12):1101. doi: 10.1167/15.12.1101.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: We used judgements of distance and direction to examine whether observers combine cues to construct a single unified representation of the environment as they move within it. Methods: Observers moved in a virtual reality simulation of a room with brick-textured walls and a checkerboard-textured floor. Upon entering a view-zone to the left of the room, a red sphere appeared 1.6m in front of the observer (Interval-1). Observers judged the position of the sphere and pressed a button triggering a half-second inter-stimulus-interval. Unknown to the observers, when the room reappeared (Interval-2) it had uniformly scaled in size around their cyclopean point. Experiment 1: observers walked rightward across the room, entering a second view-zone, upon which a second red sphere appeared in front of view-zone one. Observers judged whether the visual direction of the sphere in Interval-2 was to the left or right of the sphere in Interval-1. Experiment 2: observers remained in view-zone one, and a second ball appeared in front of them. They judged whether this ball was closer or further from them than the ball in Interval-1. In both experiments, we measured thresholds for “texture-based” cues (those unaffected by room expansion e.g. texture) and “physical-based” cues (those that could signal metric object position e.g. stereo and motion parallax) in isolation and used these to predict performance in a “combined-cues” condition where both cues varied. Results: For both distance and direction, observers combined cues “optimally”, such that performance in the combined-cues condition was well predicted by the single cue sensitivities. However, these “optimal” estimates were mutually inconsistent due to (1) differential weighting of physical- and texture-based cues and (2) different biases in the judgements made in the two tasks. Conclusion: In our experiment optimal cue combination was incompatible with observers constructing a single unified representation of the environment.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only