Many psychiatric disorders, including schizophrenia and autism spectrum disorder, are associated with behaviors that are consistent with a breakdown in the ability to predict the consequences of one’s own actions. Consequently, models based on reinforcement learning, which describe how one knows the relationship between one’s own actions and sensory percepts, can account for behavioral features of many psychiatric illnesses. It is far less clear to what extent these phenomenological models relate to real physiological processes in the brain. In close collaboration with experimentalists, we build models that link physiology and behavioral output to shed light on these disorders.

In many brain regions, the stimulus tuning of neurons is stable on a timescale of hours but not on a timescale of weeks, a phenomenon often called ‘representational drift’. For example, in piriform cortex, which is commonly considered primary olfactory cortex, the cells responsive to a given odor are completely uncorrelated with those activated by the same odor a few weeks later. This would seem to imply that piriform cortex, like other brain regions whose activity appears to drift, is useless for the retrieval of associative memories learned several weeks prior. We study how representations change over time in both models and data in order to understand which properties of these representations are stable over time and how this could be leveraged by downstream areas.

Advances in many experimental techniques in neuroscience, spanning imaging, electrophysiological, and molecular tools, are enabling experimental access to questions that were impossible not long ago. Fully leveraging the potential of these experimental tools requires the accompanying development of new analysis methods. We are working with collaborators to address some of these challenges. Two areas of particular interest are identifying structure on multiple scales in large scale recordings and aligning datasets from multiple experimental modalities.

When an animal moves, neurons across the brain become more active. This is true not just across many parts of cortex in a mouse, but even across the brain of a fly, which is typically thought of as being composed of highly specialized microcircuits. This implies that neural computations, across brain areas and species, operate within the context of an animal’s own movement. Why is the entire brain seemingly informed about what the animal is doing? Computational models can give invaluable insight into why the representation of behavior is brain-wide and how behavioral state impacts local circuit function. By recapitulating and understanding these phenomena in models, we are building concrete theories for why the representation of behavior is brain-wide and how behavioral state impacts local circuit function.