|dc.description.abstract||Biological motion perception, defined as the ability to retrieve information from minimal displays of animal motion, has often been discussed as though it represents a single mechanism. However, depending on the task and details of the stimuli, there have been divergent results as to whether this ability is primarily based on motion or form processing, whether it relies more on local or global information, and whether the knowledge that is applied is learned or innate. These results can be reconciled by a multi-component framework, with five major components: local motion invariant processing, structure-from-motion, figure-ground segregation, action categorization, and style recognition. Several experiments are reported that are motivated by this framework.
To investigate the sensitivity of the visual system to local motion invariants, performance was tested on a direction discrimination and a detection task with a point-light walker based on real motion-capture data or a synthetic walker created by Cutting (Chapter 2). When the walkers were displayed normally, performance was equal for both stimuli. However, when the walkers were spatially scrambled, the direction could only be determined for the natural walker, demonstrating that the invariant is found in subtle features of the local motion trajectories.
Another experiment examined the difference in attentional distribution due to the task requirements (Chapter 3). Participants looked more often at the feet in a direction task than they did in a gender task, lending support to the idea that useful information can be derived from the local motion of the feet.
Finally, Chapter 4 describes a battery of psychophysical tests that assesses each of the components of biological motion perception as independently as possible. The fact that individual performance does not correlate between tests is evidence that they measure different underlying mechanisms, supporting the multi-component framework.
In the concluding chapter, multi-component computational models of biological motion perception are evaluated within this framework, and the evidence relating the components to activity in particular brain regions is reviewed.||en