Perceived heading biased by a moving object: Effects of disparity and object position

Abstract

Purpose: We have previously reported that perceived direction of heading can be biased by the presence of a moving object that covers the focus of expansion (ARVO 94, 95). The heading error induced by opaque objects can even be greater than those with transparent objects, despite clear dynamic occlusion boundaries for segmentation. This suggests that the scene is not segmented on the basis of motion information prior to determining heading. In the present experiments, we added stereo as a segmentation cue. Since disparity is known to modulate responses of motion cells in area MT, it could act to reduce the heading bias. We also examined effects of initial object position with respect to the focus of expansion. Method: Displays simulated observer translation toward a frontal background plane with an independently moving opaque object closer in depth. Subjects adjusted a probe to indicate their perceived heading with respect to the background, ignoring the moving object. In the Stereo condition, both disparities and optic flow specified object boundaries and motion, while in the No Stereo condition, motion alone did so. In a second experiment, the initial position of the object with respect to the focus of expansion and the object's direction of motion in depth were parametrically varied. Results: Similar heading errors were observed in both the Stereo and No Stereo condition, indicating that disparity information does not reduce the bias. Effects of initial position will also be reported. Conclusion: Our results provide further evidence that the scene is not segmented prior to determining heading, such that heading perception appears to be based on global pooling of optic flow. We describe a heading model based on properties of MT and MST that can account for the data.