Conference on Cognitive Computational Neuroscience

We are constantly surrounded by people performing various actions. Although understanding them appears to come effortlessly, this task is complex and relies on multiple sources of information. Here we aimed to reveal dimensions underlying the organization of actions and the corresponding patterns of brain activity. First, we used principal component analysis (PCA) to reduce 44 action features identified in a recent study (Kabulska & Lingnau, accepted) to four dimensions related to general movements, arm movement kinematics, goal-directedness, and context. Second, representational similarity analysis (RSA) of fMRI data collected during the observation of 100 different actions revealed that three of the four dimensions evoked specific brain patterns in the lateral occipitotemporal cortex, while the Context dimension was located in the supramarginal gyrus. These results give novel insights into the representation of actions and suggest which types of information human participants extract to understand others’ actions.