Agents, objects, and actions: Investigations into the neural representation of dynamic information

Abstract

Making sense of dynamic scenes is essential for navigating our world. Traditionally, research on dynamic scene processing has distinguished between animate agents and inanimate objects, often using the movement of inanimate objects as a baseline to reveal the unique processes and representations related to animate agents. In this thesis, I adopted an alternative approach by highlighting that animate agents are also physical entities governed by the same physical laws as inanimate objects. While animate agents have internal mechanisms that allow them to initiate their own movement or react to external physical forces, their movement dynamics can, at a certain level of analysis, be equated to those of inanimate objects. By mapping the shared neural representations that span both domains, this approach can better delineate the distinct neural representations that arise due to their inherent differences. Following these considerations, in Chapter I, I show that a set of frontoparietal and posterior temporal brain regions, commonly studied in relation to human action recognition, host a shared neural code for capturing structurally similar movements of humans and inanimate objects. In Chapter II, I replicate these findings and examine how agentive and physical forces behind motion events shape their neural representation. I find that these regions encode a shared neural code for the physics and kinematics of dynamic events, regardless of animacy or the nature of the forces driving them. I also find that regions such as the right posterior superior temporal sulcus and temporoparietal junction are more sensitive to actions of animate agents compared to movements of objects, even when the structural and kinematic properties of movement are matched across the two. Chapters I and II focus on the structural similarities between motion events involving animate agents and inanimate entities. These chapters identify a neural representation that capture both animate and inanimate movement dynamics within a wide array of frontoparietal and posterior temporal brain regions. In Chapter III, I investigate whether these regions have subcomponents with differential sensitivity to animate or inanimate movement by using data analytical approaches that can reveal differences in nearby anatomical structures, within individuals. Additionally, I use analyses of intrinsic functional connectivity to situate the neural responses to animate and inanimate movement within the brain's functional network architecture. Building on the findings of Chapters I and II, Chapter III finds that brain regions that are involved in analyzing the physical and kinematic aspects of movement establish a network of interconnected areas in precentral and postcentral structures and anterior lateral occipitotemporal cortex. In contrast, brain regions more attuned to agentive or animate dynamics of movement form a distinct network, often adjacent to regions involved in the general analysis of dynamic scenes. This network encompasses not only the right posterior temporal sulcus and temporoparietal junction, as often cited in the literature, but also includes a range of other frontoparietal regions. Collectively, this thesis contributes to our knowledge of the neural basis of dynamic scene understanding. By identifying neural activity patterns that capture both the shared and distinct aspects of dynamic scenes that involve animate and inanimate entities, it provides a unified framework for studying the complex neural processes that underlie the perception and interpretation of dynamic information.