Abstract
In general, the literature suggests that visual information alone is insufficient to control rotational self-motion accurately. Typically, subjects misperceive simulated self-rotations when no vestibular or proprioceptive feedback is available (see Bakker et al., 1999; 2001 — these studies were done with HMDs). On the other hand, Riecke et al. (2002) found nearly perfect turning performance when a curved, half-cylindrical projection screen with a large FOV of 180 was used. So far, no study has systematically looked at the effect of screen curvature on ego-motion perception.
To investigate whether screen curvature influences turning performance, we had 14 participants perform visually simulated ego-rotations either using a flat projection screen (FOV 86 ∞ 64 ) or a curved projection screen (radius 2m) with the same FOV in a within-subject repeated-measures design. Subjects saw a “star field” of limited lifetime dots without any landmarks, and they used a joystick to control instructed turn angles between 45 and 270 (steps of 45 ). No feedback about accuracy was provided. A repeated-measures ANOVA revealed a significant effect of screen curvature, and also an interaction between curvature and turn angle: While target angles were undershot on the curved screen (gain factor 0.84), a surprising overshoot was observed for the flat screen (gain factor 1.12). Subjects' verbal reports indicate that on the curved screen, the simulated self-rotations looked more realistic than on the flat screen. This may have led them to overestimate turns on the curved screen (thus undershoot turn angles) and to underestimate turns on the flat screen (thus overshoot turn angles). A possible explanation is that rotational lamellar flow on the flat screen was misperceived as translational flow rather than as rotational flow. Results indicate that screen curvature is a critical parameter to be considered for ego-motion simulation and vection studies.