Abstract
Human heading perception based on optic flow is not only accurate (Warren & Hannon, 1988; van den Berg, 1992), but remarkably robust and stable. These qualities are especially apparent when observers walk or drive through environments containing moving objects: these objects generate motion that is uninformative about the observer’s heading direction and may occupy large portions of the visual field, yet heading judgments are only biased by several degrees. This suggests a mechanism that maintains a stable, temporally evolving heading estimate to tolerate intermittent disruptions to the optic flow field, such as the locally discrepant motion produced by moving objects. Stability, however, may come at the cost of sluggishness that impedes sensitivity to rapid and abrupt changes: heading estimates may persist and it may take time to incorporate emerging optic flow signals. We tested this in experiments where subjects judged their heading after viewing simulated self-motion produced by an observer moving along two consecutive, different heading directions. Consistent with a persistent representation of heading, we found that judgments lay in between the initial and final headings and heading error increased in proportion to the deviation between the heading directions. Bias increased toward the initial heading direction when the duration of the second heading period decreased from 500 ms to 250 ms, as would be expected if the duration of optic flow from the second period is insufficient to shift an earlier heading estimate. Consistent with this interpretation, introducing a blackout period in which the optic flow ceases immediately prior to the heading judgment yielded increased bias. Surprisingly, however, the bias was also elevated when the blackout appeared in between the optic flow corresponding to the initial and final headings. Our results may shed light into how the visual system balances robustness in heading perception with sensitivity to change.