Abstract
Perception of non-rigid motion of water, clouds, fire, bird flocks etc. is challenging. Their optic flow patterns deform over time, lack clear borders, have independently moving parts, and can even be inconsistent with their motion. We investigated whether non-rigid motion can be perceived via the dynamics it causes in the optic flow. We developed three types of rotating pattern stimuli that contain different flow properties. Each single frame of a stimulus consisted of a random dot distribution devoid of any structural information. Across frames, the dots were rotated around the center of the pattern. The speed of the rotation depended on the distance to the center. In the first pattern, dot speed increased with distance up to a maximum, while dots further away stood still, creating a sharp boundary. In the second pattern, dot speed decreased with distance like in a water vortex. In the third pattern, dot speed was constant throughout the screen. The patterns themselves then moved over the screen. The dots did not follow this motion, only the center of the rotation moved, so this motion was independent from the instantaneous optic flow. Subjects were well able to report the final position and travelled path for all three patterns, showing that perception did not rely on borders or speed differences in the optic flow but on its dynamics. Adding a coherent global motion did not hamper performance, but adding local noise did, indicating that the visual system derives the dynamic pattern on a local scale. In a subsequently speed discrimination experiment, subjects judged the non-rigid patterns only slightly slower than a rigid control stimulus, revealing that speed perception was not just based on the instantaneous optic flow. We propose that the visual system can use the dynamics in the optic flow for motion perception of non-rigid objects.