Effect of scene information on path perception and steering control

Abstract

Human observers can perceive self-motion from optic flow for both circular paths and straight paths. However, when view does not rotate during travel on a circular path, observers often show significant bias in path judgments from optic flow. A proposed explanation is that human observers use the velocity field of optic flow for perception of self-motion and assume that path curvature is accompanied by body rotation.

Use of a scene-based strategy, as opposed to optic flow, could potentially reduce the bias in judgments of circular path with no view rotation. It has been shown that human observers can judge heading direction from a sequence of static pictures of a scene, suggesting that observers can perceive self-motion from change in observer position relative to a scene. Another previous study found that showing the same configuration of landmarks repeatedly improved performance in a steering task, suggesting that familiarity of a scene may further improve perception of self-motion and steering control.

In this thesis, I explored the effect of scene information on judgments of circular path and steering control. Experiments 1 and 2 tested path judgments for simulated travel along a circular path. Observers viewed 1 s presentations of optic flow and adjusted a response pole to lie on their future path, i.e. where they would intercept the pole if they keep travelling on the current path. Precision and accuracy of judgments were compared across conditions that varied in the presence and consistency of landmarks. I found no effect of landmarks on precision and accuracy, and curvature underestimation remained present in all conditions.

Experiment 3 tested whether familiarity with a scene can improve steering control. I used the same steering task by Andersen and Enriquez (2006). During a trial, heading direction was perturbed horizontally in a smooth, pseudo-random manner, as if unexpectedly wind perturbed a car and caused it to slide to the left and right while driving straight ahead. Observers were instructed to counteract these perturbations and steer to maintain a straight path of travel. Contrary to previous findings, I found no effect of scene familiarity on steering control.

I found no benefit from landmarks in any of the conditions tested in this thesis. These results suggest that scene information has a limited role for path judgments and steering control.