Abstract
We have previously demonstrated that human observers are able to identify scene-relative object movement against the complex pattern of retinal motion that arises during self-movement ("flow parsing", Rushton & Warren, 2005, Current Biology). Here we examined the time-course of the parsing process. We placed a small probe to one side of a central fixation point and a hemi-field of twinkling dots to the other. On each trial, at predetermined times, the dots would move radially inwards or outwards, and the probe would briefly (35ms) appear and move, at one of three speeds, towards or away from fixation. Observers indicated whether they perceived the probe as moving inwards or outwards. From their responses we were able to estimate the magnitude (or presence) of the flow parsing effect: any discrepancy between the physical and perceived motion could be attributed to the parsing process. In a first experiment we manipulated the duration of the flow and established that parsing occurs with as little as 17ms of flow. In a second experiment we manipulated the asynchrony between the onset of the flow and the probe motion. We found that the parsing effect was maximal when the two appeared simultaneously. In a last experiment we assessed how rapidly the flow parsing process occurs. We first established how quickly (reaction times) observers can detect inward or outward movement of the probe. We then examined the impact of the movement of dots in the hemifield on the detection time. We found that by moving the dots in the opposite radial direction to the probe it was possible to reduce reaction times. This indicates a remarkable, very low-level, direct and essentially instantaneous, interaction between global (3D) and object motion processing. We might summarise the results as indicating that we perceive the world through a 3D motion filter.
Meeting abstract presented at VSS 2013