Abstract
The ability to categorize actions is critical for interacting in complex environments. Previous studies have examined the neural correlates of categorization using static stimuli. The goal of our study was to investigate the neural substrates that mediate learning of complex movement categories in the human brain. We used novel dynamic patterns that were generated by animation of an artificial skeleton model and presented as point-light displays. We created prototypical stimuli that differed in the spatial arrangement of their segments and their kinematics. Intermediate stimuli between the prototypes were generated by a weighted linear combination of the prototypical trajectories in space-time. We compared fMRI activations when the observers performed a categorization vs. a spatial discrimination task on the same stimuli. In the categorization task, the observers discriminated whether each stimulus belonged to one of four prototypical classes. In the spatial discrimination task, the observers judged whether each stimulus was rotated (or translated) leftwards vs. rightwards. These tasks were matched for difficulty based on the observers' performance during a practice session. We observed significantly stronger fMRI activations for the categorization than the spatial discrimination tasks in the dorsal, inferior parietal and the medial, inferior frontal cortex, consistent with previous findings on the categorization of static stimuli. Interestingly, we also observed actvations in visual motion areas (V3a, hMT+/V5), higher-order motion areas in the intraparietal sulcus (VOIPS, POIPS, DIPSM, DIPSA) and parieto-frontal areas (supramarginal gyrus, postcentral gyrus, ventral and dorsal premotor cortex) thought to be involved in action observation and imitation. These findings suggest that categorization of complex dynamic patterns may modulate processing in areas implicated in the analysis of visual motion and actions.