Nuances of action: understanding action categories and their brain representation through naturalistic fMRI

The nature of human action remains a debated issue, with no definitive resolution. This problem can be approached from two perspectives. The first is theoretical: refining our understanding of what constitutes an action and developing a taxonomy that effectively captures its essential features can provide a more structured framework for investigating its neurobiological basis. The second is empirical: to fully grasp the complexity of human action, research must incorporate actions occurring in ecologically valid settings, such as naturalistic stimuli (e.g., movies). This thesis addresses both dimensions. On the theoretical side, I propose new high-level models for interpreting action, beginning with fundamental definitions and examining the minimal conditions required for an effective taxonomy. I highlight the inherent limitations and unavoidable incompleteness in any attempt to comprehensively map the action phenomenon. Additionally, I review existing empirical approaches in neuroscience that have sought to identify the key features representing actions in the brain. On the empirical side, I employ functional magnetic resonance imaging (fMRI) with naturalistic stimuli to investigate how actions are encoded in the human brain. Using a theoretically grounded action model, I examine the stability of action representations across different movies, demonstrating that action coding remains consistent despite variations in narrative and sensory complexity. I then conduct a detailed analysis of a single movie (101 Dalmatians), exploring how model fit varies across different sensory modalities. The findings reveal that certain action features have a stable impact on brain activity regardless of the modality through which they are perceived. By integrating theoretical modeling with empirical validation in naturalistic settings, this work advances our understanding of action representation in the brain and underscores the importance of studying actions in contexts that closely mirror real-world experiences.