The Neural Computation of Internal Affective States

Citation

Nair, Aditya (2025) The Neural Computation of Internal Affective States. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/wryk-rb16. https://resolver.caltech.edu/CaltechTHESIS:10162024-031702349

Abstract

The study of neural computation has long concentrated on our cognitive abilities, with extensive research dissecting the mechanisms of memory, decision-making, and navigation. In contrast, the realm of social innate behavior and emotion has often been treated as a simpler problem, overlooking the immense complexity and biological significance it entails. This thesis aims to bring neural computation into the domain of emotional or affective states, employing data-driven modeling methods that approximate neural activity as dynamical systems. The application of these methods has uncovered brain representations that encode key qualities of persistence and escalation associated with aggressive states, formalized as line attractors. These emergent features of neural circuits arise from the complex interplay of connectivity and network dynamics, challenging long-held notions of subcortical computation. This discovery led us to rigorously test various key properties of line attractor dynamics. Through closed-loop modeling and holographic neural activation, we demonstrate that the line attractor is intrinsic to the mammalian hypothalamus, providing some of the first causal evidence of this property for any continuous attractor. These experiments also suggest that functional connectivity within the hypothalamus underpins the stability of this attractor. Furthermore, using a new cell-type-specific gene-editing system, we show that the implementation of this line attractor depends on neuropeptides, indicating a non-canonical mechanism that contributes to the robustness of this innate attractor. Finally, we reveal that line attractors encode emotional states beyond aggression, including states of sexual receptivity in the female hypothalamus. Longitudinal recordings of neural data across the estrus cycle show that the line attractor disappears during non-estrus states, suggesting long-timescale modulation of attractor dynamics by hormones. Together, these studies present a new paradigm for understanding subcortical computation underlying internal states and suggest a canonical motif that the brain reuses to encode diverse internal affective states.

Thesis Files