Abstract
Introduction: Successful interaction with the environment requires a moving observer to maintain and update a map of egocentric location for objects in the world. We hypothesise that optic flow is used to solve this problem. To provide evidence for this we consider the induced motion illusion: an observer fixating a static target will perceive that it moves oppositely to the net direction of a moving array of dots. We construct a stimulus which in one spatial configuration gives rise to such an illusion, but does not when reconfigured to be consistent with radial flow, in spite of identical net motion information. Methods: Stimuli consisted of 2D limited lifetime dot motion patterns, briefly displayed on a CRT screen. The basic motion stimulus was consistent with expansion due to forward translation. This pattern was systematically varied in two ways. Firstly, while fixing the total number of dots in the display, the numbers on the left and right hand sides of the field were varied (in the ratios 1:9, 3:7, 5:5), causing different biases in net motion. Secondly, a variable proportion (0%, 50%, 100%) of the 2D start locations of the dots were shuffled. Thus, throughout its lifetime, the local motion of each shuffled dot was identical to that in the un-shuffled case, but the trajectory started in a location which made the motion inconsistent with radial flow. Observers reported the direction of motion of a central probe dot, travelling linearly in one of five directions arranged symmetrically about vertical. Results: For all observers and for both leftward and rightward net motion biases, there was a significant decrease in induced probe motion as the stimulus became more consistent with radial flow. Conclusions: These results suggest that induced motion is dependent upon first parsing the motion field to extract components consistent with flow. We propose that a primary role for optic flow is to update egocentric object position during self motion.