Dissociating goal from outcome during action observation.

Abstract

Understanding the goal of an observed action requires computing representations that are invariant to specific instantiations of the action. For example, we can accurately infer the goal of an action even when the agent's desired outcome is not achieved. Observing actions consistently recruits a set of frontoparietal and posterior temporal regions, often labeled the "action observation network." While progress has been made in charting which regions of the action observation network are involved in understanding goals of observed actions, it is not clear where goals are represented independently of outcomes. We used functional magnetic resonance-based multivariate pattern analysis to identify such regions. Human participants (20 females, 12 males) watched videos of successful and failed attempts of actions with different goals involving two different object types. We found that the bilateral anterior inferior parietal lobe and the right ventral premotor cortex distinguished between object-specific action goals regardless of outcomes. The left anterior inferior parietal lobe encodes action goals regardless of both outcomes and object types. Our results provide insights into the neural basis of representing action goals and the different roles of frontoparietal and posterior temporal regions in action understanding.