Abstract
Prior work has shown that perceived elevation relative to straight ahead is exaggerated with a gain of 1.5 (e.g., Durgin & Li, 2011), but that this expansion is rooted to the gravitational framework specified by a visible ground plane – both for sideways observers in the real world and for upright observers in immersive VR representing a sideways world (Klein et al., 2016; Li & Durgin, 2016). Here we compared the perception of elevation relative to straight ahead in a groundless virtual environment and one with a ground plane. In Experiment 1, 24 observers viewed a stereoscopic back-projected VE from a meter away. Half saw a blue background above a ground plane with the horizon at eye level. Half saw just a blue background. A single ball represented straight ahead. In phase 1, participants made forced choice judgments about whether a second ball was more or less than 30° above straight ahead. The mean PSE was 21.7° ±2.9°(95%CI), which was reliably less than 30°, but did not differ reliably from 20° (1.5 gain). In phase 2, numeric angle estimates for elevations 6-36° above and below straight ahead were recorded for both groups. The overall mean slope was 1.57±0.11. In no condition did the mean slope differ reliably from 1.5. In Experiment 2, we repeated the numeric estimation phase with new participants who were lying on their backs looking up so that straight-ahead was along the gravitational vertical. Despite the absence of both a visual and a gravitational ground direction reference frame, participants in the groundless VE still provided estimates with a mean slope no different from 1.5. These new findings regarding perceived elevation differentiate elevation from azimuth. Durgin and Keezing (2018) found that removing the ground in their VE reduced angular expansion in azimuth from 1.25 to 1.05.