Abstract
Our daily visual environment is abound with motion – people walk around, snowflakes fall, and trains speed on by. Interacting with a moving object requires estimating its speed, and extrapolating its spatiotemporal position when the object is temporarily occluded. Experimental stimuli used to study speed perception are typically designed to omit spatiotemporal cues (e.g., drifting gratings or random dot motion), ensuring that spatial displacement, or the passing of time, cannot be independently used as a proxy for speed. However, such stimuli do not readily lend themselves to questions concerning motion extrapolation, which are important to ask, as most objects in the natural world are spatiotemporally bound. Here we examine if speed estimation differs for spatiotemporally bound and unbound motion stimuli. We measured speed recall in human participants over a wide range of speeds (2–32 º/s), and across several delays (1, 4, and 8s), for a single target dot or random dot motion. Both stimulus types were shown moving along a circular trajectory for 4–6s, and were recalled after the delay via method-of-adjustment. Our data show a strong logarithmic drop in behavioral performance with increasing speeds for both types of motion stimuli. We also observed a wholesale decrease in performance for random dots compared to a single target, implying an advantage for spatiotemporally bound objects when estimating speed. This advantage might arise from additional spatiotemporal cues, a narrower focus of spatial attention (which is more distributed during random dot motion), or anticipatory neural mechanisms that allow for motion extrapolation in spatiotemporally bound objects. We found little effect of delay duration. Together, our results provide a solid psychophysical basis for understanding human speed estimation, and potentially motion extrapolation.