Accurate perception of visual space from live-video in a head-mounted display

Abstract

Even when geometrically accurate displays are used in virtual reality (VR), most labs report substantial compressions of perceived distance. That is, it is hard to effectively simulate large-scale spaces in VR; they look small. There have been reports that pre-exposure to the space that is being simulated (e.g., in a VR rendering of a lab space) produces accurate scaling. We sought to measure whether pre-exposure to our experimental hallway had an effect on locomotor distance judgments when that hallway was viewed in a head-mounted display (HMD) with a correctly scaled live video image. The task was to walk, without visual or auditory feedback, to a pre-viewed target. Subjects had never been to the experimental hallway before. Distances were in the range of 3 to 7 meters. There were three conditions of the experiment. In the normal vision condition, subjects simply wore the HMD as a hat (to match any postural influences) and viewed each target directly before closing their eyes and walking to it. In the live-video-only condition, subjects performed the same task while wearing an HMD without any pre-exposure to the hallway environment. In the pre-exposure condition, subjects were allowed to view the hallway directly from a single vantage point prior to donning the HMD and beginning the experiment using live video. Surprisingly, although the average proportional distance (walked/actual) with pre-exposure did not, in fact, differ from that with normal vision, proportional distances walked in the live-video-only condition were actually greater (M = 1.00) than those in the other two conditions (M = 0.87), t(18) = 2.13, p [[lt]] .05. While it is possible that subjects were actively compensating for a mismatch between perceived size and expected size of a generically familiar scene, the HMD itself is probably not primarily responsible for distance compression.