Neural Representations of Observed Actions Generalize across Static and Dynamic Visual Input.

摘要: People interact with entities in the environment distinct and categorizable ways (e.g., kicking is making contact foot). We can recognize these action categories across variations actors, objects, settings; moreover, we them from both dynamic static visual input. However, neural systems that support recognition perceptual differences are unclear. Here, used multivoxel pattern analysis of fMRI data to identify brain regions categorization a format-independent way. Human participants were scanned while viewing eight interactions pulling) depicted two formats: (1) visually controlled videos interacting actors (2) varied photographs selected internet involving different settings. Action category was decodable formats bilateral inferior parietal, occipitotemporal, left premotor, middle frontal cortex. In most regions, representational similarity consistent subjects formats, property contribute common understanding actions among individuals. These results suggest identified codes important for understanding.SIGNIFICANCE STATEMENT Humans tend interpret observed others terms invariant incidental features: whether girl pushes boy or button see it real-time single snapshot, still pushing investigated facilitate such differences. Using fMRI, several areas cortex exhibit similar photographs. Our provide strong evidence involvement recognizing way people physically objects other people.