Summary/Abstract: Project 3, Internal Brain States Recent work from our collaboration has revealed that sensory decision-making in rodents relies on time-varying internal states, with distinct decision-making strategies employed in different states. We have developed a statistical model for identifying these states from decision-making behavior alone, and find that mice switch among strategies on the timescale of tens to hundreds of task trials. This finding presents a major challenge to traditional models of decision-making, which assume that trained animals rely on a single evidence-accumulation strategy that is relatively fixed within a session. Furthermore, we found that inactivation of the striatum affects the animal’s choices only in some states; in other states, decisions are not affected by striatal inactivation, suggesting that mice rely on distinct neural circuits for making decisions in different states. This project will follow up on this startling discovery in order to investigate the neural mechanisms underlying internal states throughout the brain. Aim 1 will focus on characterizing the neural basis for the internal states governing sensory decision-making and working memory. We will use causal perturbations and large-scale neural recordings to characterize how population activity varies across states, and use closed-loop optogenetic inactivation experiments to examine how different brain regions contribute to decision-making in different states. Aim 2 will look inside the brain to identify internal states from the dynamics of neural activity. We will develop new models to characterize how internal states evolve on the timescale of single trials using spike train data. We will then use these models to characterize state-dependent communication between brain regions in large-scale multi-region electrophysiological recordings. Aim 3 will focus on determining how the cognitive decision-making states identified by our model relate physiologically-defined internal states such as thirst, hunger, and arousal. This will allow us to connect our findings about decision-making strategies to the extensive literature on physiological internal states. To assay arousal, we will measure pupil diameter and use fiber photometry to measure activity of noradrenergic neurons in the locus coeruleus and cholinergic neurons in basal forebrain. As a neural readout of hunger and thirst, we will measure activity in relevant hypothalamic neuron populations. These measurements will be compared to internal states derived from our models. We expect the experiments and modeling efforts in this project to substantially advance two priority areas of the BRAIN Initiative: demonstrating causality and identifying fundamental principles.