Abstract
Recent evidence suggests that observers have difficulty perceiving path curvature from optic flow when curvature is not accompanied by simulated view rotation. In this study, we investigated whether scene structure could reduce such biases. Landmarks provide reference objects, and a familiar scene might allow self-motion to be perceived from multiple static views. We tested three scenes: textured ground with no landmarks, with landmarks in a fixed configuration, or with landmarks in randomly varied configuration. In the fixed landmark condition, observers were pre-trained to learn the configuration. Observers viewed 1s displays of simulated self-motion along circular paths, with various curvature, and adjusted a pole to lie on their perceived future path. Two pole distances were tested to assess perceived path curvature. Across all conditions, judgments showed errors consistent with underestimation of path curvature, and a bias toward the center of the screen. The presence of landmarks did not reduce these biases, nor improve precision of judgments. In contrast to some previous studies, we found no benefit from rich scene structure. We argue that the biases observed here are due to use of instantaneous optic flow, which provides insufficient information in these conditions even when a scene has rich structure.