Investigating the Neural Overlap of Actions and Physics

Abstract

Action and physics are closely intertwined in everyday life. We are remarkably good at inferring physical properties of a scene, predicting the outcome of physical events, and using this information to guide our own behaviors. Neuroimaging research has identified a set of frontoparietal and posterior regions, often titled the “intuitive physics network”, that are recruited during physical reasoning. Notably, these regions overlap with those commonly recruited by action planning and action understanding. However, the extent and meaning of this overlap remains unclear, as prior evidence largely relies on comparisons across separate studies using different paradigms and participant samples. This thesis investigates the relationship between intuitive physical reasoning and action at multiple levels of analysis, combining functional neuroimaging, multivoxel pattern analysis (MVPA), and neuropsychological methods. In Chapter 1, I show that some frontoparietal regions encode action goal (i.e., desired physical outcome) independently of the observed outcome. In contrast, representations in the lateral occipitotemporal cortex (LOTC) were sensitive to the observed outcome and did not generalize across successful and failed actions. In Chapter 2, using a neuropsychological approach, I present a behavioral dissociation between tool use and physical reasoning in left-hemisphere stroke patients with frontoparietal lesions, providing evidence that these abilities rely on partially distinct cognitive and neural mechanisms despite the apparent neural overlap. In Chapter 3, I show that information about object weight could be decoded from partially overlapping but task-dependent regions. Notably, the LOTC consistently encodes weight-related information across all contexts, including action planning, human action perception and object event perception, suggesting a shared representation of weight-related information in the posterior part of the brain. Together, these chapters refine our understanding of the relation between action and physical reasoning. The results support a division of labor between the frontoparietal vs. posterior regions and highlight the contribution of the LOTC, which has been relatively overlooked in the physical reasoning literature. More broadly, this thesis highlights the need to move beyond broad network-level characterizations toward a more precise account of the computations supported by individual brain regions.