Author ORCID Identifier

https://orcid.org/0000-0002-2831-2941

Date of Award

Summer 7-21-2026

Document Type

Thesis (Ph.D.)

Department or Program

Psychological & Brain Sciences

First Advisor

Katherine Nautiyal

Abstract

The brain reliably supports learned behaviors over long periods of time despite continual biological changes that alter the activity of individual neurons, and the connections between them. For learned behaviors to persist, the information stored in neural populations must remain accessible even when the activity of individual neurons changes. Understanding how the brain achieves this is a fundamental challenge in neuroscience. The dorsomedial striatum (DMS) is critical for instrumental behavior, encoding information about actions, outcomes, and their relationships. Although extensive work has characterized the role of the DMS during learning and performance, less is known about how task-related information is organized across DMS populations and how their neural representations evolve over time. This dissertation investigated both questions using longitudinal 1-photon calcium imaging recordings of DMS neurons in mice performing a cue-guided instrumental task.

Chapter 1 characterizes how task-related information is organized in DMS populations. Individual neurons exhibited mixed selectivity, responding to multiple task epochs including trial initiation, cue presentation, reward delivery, and post-reward periods. Trial outcome information was distributed across the population rather than localized to a small subset of highly specialized neurons. In Chapter 2, the same neurons were tracked across recording sessions to examine the stability of these representations over time. Withinsession population decoding confirmed that trial outcome information could be reliably extracted from DMS activity in all sessions. However, individual neuronal responses changed ii both within and across sessions, and decoding accuracy of trial outcome declined as the time between the train and test sets increased. Despite these changes, several features of population organization remained stable. The proportion of task-modulated neurons was maintained across sessions, and low-dimensional population geometrywas preserved over time such that alignment using Procrustes transformations recovered cross-session and cross-animal decoding performance.

Together, these findings show that DMS representations are simultaneously dynamic and stable: individual neurons exhibit ongoing changes in their activity patterns, while the overall organization and relational structure of population activity remains preserved. These results provide insight into how neural systems may preserve information relevant for learned behavior despite continual changes in the activity of individual neurons.

Download

Share

COinS