The Anatomy of Mentalizing: Neural Representation of Events and Socio-Affective Concepts, and Mental Imagery of Agentic Scenes

Abstract

Humans navigate a complex social world by understanding others’ behaviors and inferring their mental states, such as emotion, intention and motivation. Dominant theories in social cognition propose that, as mental states are not directly observable and oftentimes ambiguous, people tend to use their own experiences as a reference point or initial guess, a process known as self-projection. However, it remains unclear how one’s conceptual knowledge of social events and mental states is represented in the brain and how such knowledge is activated and synthesized to generate behaviorally relevant outcomes. This dissertation investigates the building blocks of mentalizing by examining the neural representation of socio-affective concepts (events and mental states), division of labor within the Theory-of-Mind network, and the hierarchy of non-commitment in mental imagery of agentic events. Across three studies, this body of work combines neuroimaging, psycholinguistics, and behavioral experiments to unpack how we interpret others’ behaviors and mental states from verbal descriptions. Chapter 1 examines the neural representation of transitive socio-affective events (e.g., to love, to marry), and more specifically, the causal structure of such behavioral and mental activities. Results show that a region in the left temporal lobe encodes event information that distinguishes between event categories of different causal structures (behavioral vs. mental), specific to verbs but not semantically matched nouns. Chapter 2 dissociates conceptual representations of mental states from the neural mechanisms that infer them from behavior. Across two tasks, findings reveal a division of labor within the Theory-of-Mind network: right-lateralized regions (e.g., right temporoparietal junction and precuneus) preferentially support behavior-derived mental state inference, whereas left temporoparietal regions represent mental states independent of behavioral inputs. Chapter 3 shifts from the neural basis to behavioral experiments. A series of mental imagery tasks demonstrate that social scenes are constructed hierarchically, with a core skeleton of spatial physical properties are consistently rendered in imagery, while psychological properties of agents (emotion, goals, motivation) are preferentially included in social context. Together, these findings provide a unified framework for understanding mentalizing as emerging from structured representations of socio-affective concepts, flexibly deployed in context-dependent ways. By linking social semantic knowledge with neural mechanisms of mental state inference and mental imagery, this dissertation provides insights of how humans represent and reason about the social world.