Abstract
Objects are judged up to 8m farther in wide (40m) versus narrow (1.5m) rooms (for targets rendered at a distance of 37m; VSS 2017). These large effects emerge at relatively far distances, where the role of visual cues is poorly understood. The multiplicity of possible factors makes determining the cue bases a challenge. Here, we used rendered scenes and Amazon's Mechanical Turk to test several candidate hypotheses. Stimuli were empty 40m-deep rooms lined with irregularly-placed doors and an orange cone 4–37 m from the observer's viewpoint. 95 MTurk workers numerically judged the cone distance. We tested the role of linear perspective (by adding black lines to floor and ceiling in 40m-wide rooms), visibility of the near and far ground plane texture (by introducing transverse walls and hallway openings), and visibility of a larger outer environment (by removing walls from a 1.5m-wide hallway but leaving freestanding doors). We also included 40m-wide rooms and 1.5m-wide hallways. Results showed main effects of scene and distance, and a scene x distance interaction (all ps < .0001). Judgments for the 40m width and 40m width + linear perspective scenes were similar to each other (p>.05), and larger than in the other four scenes (by up to 8.7m for the 37m target distance). Judgments in the other four scenes did not differ from each other (all ps>.05). The results did not identify a single visual feature that drives the difference in judgments across room widths, although they did constrain the possibilities. Most interestingly, large occlusions of the near versus far ground plane yielded statistically indistinguishable judgments, and neither gave rise to the larger judgments associated with wide rooms. Some remaining possibilities for the effect of room width on distance judgments include total amount of visible ground texture and presence of boundaries or barriers to navigation.
Meeting abstract presented at VSS 2018