Item

Abstract

We conducted triangle completion experiments in virtual environments to investigate the role of optic flow in human homing performance. Ego-motion was visually simulated on an half-cylindrical 180-degree-projection screen of 7m diameter using the mouse buttons as input device. Subjects had to return to the origin after moving outwards along two prescribed segments of the triangle. To exclude scene-matching as a homing strategy, subjects were "teleported" to a different, however similar environment for the return path ("scene-swap condition"). Experiments were performed in two simulated environments: A cloud-of-dot-like environment and a photorealistic town. Only the latter contained landmarks and explicit scaling information. We found systematic errors in distances traveled, but not in turns performed. Homing based on optic flow alone in the cloud-of-dot environment was possible and led to similar performance as navigation in the town. Omitting scene-swap in a control experiment led to almost perfect homing performance in the town, suggesting that scene-matching (whenever possible) plays the dominant role in homing accuracy. A comparison with results from Loomis et al. (JEP, 1993), who studied triangle completion based on path integration using proprioceptive cues showed that optic flow information in our experiments led to considerably smaller systematic errors. Using scene-swap and virtual environments proved to be a successful paradigm to disentangle the role of two major information sources in spatial orientation: optic flow (path integration) versus landmarks (piloting).