2025 Theses Doctoral

Representation of mental states in the primate brain

Hashim, Rahim

A fundamental goal of neuroscience has been to study sensory perception, internal states, and observable behavior simultaneously. Traditionally, however, studies of the brain have often examined these processes separately, either by studying how individual neurons respond to specific sensory stimuli, or by tracking neural activity during overt movement.

What these approaches often fail to consider are mental states---that is, the underlying neural mechanisms of emotions and homeostatic needs, such as motivation, arousal, thirst and hunger---that influence perception and bias behavior. Even for affective neuroscientists studying emotions, longstanding limitations in both electrophysiological recording systems and accessible compute have made it difficult, if not impossible, to analyze the neural representation of mental states across multiple brain regions in a single experiment, while also holistically characterizing behavior using objective and analytical approaches.

Recent advances in electrode manufacturing, GPU architecture, automated spike-sorting methods, and computer vision models have ushered in a new era of neuroscience, in which mind, brain, and behavior can be studied in a more unified framework. Here I describe a new approach leveraging these advances to bridge the gap between studies of internal states and observable behavior in the primate brain.

I utilize state-of-the-art Neuropixels Non-Human Primate (NHP) probes to record from over one thousand neurons in multiple cortical and subcortical regions in a single experimental session, generating across many sessions the largest known dataset of simultaneous high-density recordings across primate amygdala, insula, and IT cortex. I synchronize the neural activity to multiple high-resolution video cameras pointed at the face and body, and train neural network models on the video data to identify key body parts in order to relate recorded neural activity preceding and during overt behavioral motifs.

I investigate mental states that are observable only in behavioral data, or only in neural activity, i.e. hidden from behavioral observation. Using analysis techniques rooted in both representational and dynamical frameworks, this dissertation argues that recording high-dimensional neural data alongside high-dimensional behavioral data is required to reveal the source and expression of emotions, emotion-based decision-making, and ultimately, cognition more broadly.