Abstract
PURPOSE: The human visual system often encounters situations in which moving objects disappear behind an occluding surface. In such situations it might be beneficial to predict where the object will emerge on the other side of the occluder. Here we investigate the information in the scene, and in the motion of the object, that is important to observers when performing such a task. METHODS: Observers were asked to track a small dot on the screen as it moved towards a vertical occluder, which was clearly visible. The point initially moved purely horizontally. At subsequent time steps the speed was held constant while the direction of the velocity was sampled from a unimodal probability distribution which was symmetric about the current direction. As soon as the dot reached the occluder it disappeared. Observers were asked to estimate where the dot would emerge on the other side of the occluder, as well as indicate a vertical interval around that point that they were certain would capture the dot as it emerged. Between trials the variance of the direction distribution was changed, increasing or decreasing the ‘randomness’ of the path. Occluder width and dot speed were also manipulated between trials. RESULTS: We found that observers increased the size of the capture region as the variance of the direction distribution increased. The capture region also increased in size with increasing occluder width, however, we found no effect of speed. CONCLUSIONS: Human observers extrapolate motion by integrating what is known about the characteristics of the path with information about the scene. We discuss and model this data in terms of a probabilistic account of motion perception.