Horan, Mattias; (2025) Inference and Generalisation in the Hippocampal-Entorhinal Circuit. Doctoral thesis (Ph.D), UCL (University College London).

Text Horan_10208643_thesis_Redacted.pdf Access restricted to UCL open access staff until 1 June 2026. Download (9MB)

Abstract

Simple problems may have simple solutions, but facing a complex world, decision makers develop complex behaviours. Going beyond stimulus-response associations, model-based strategies allow flexibility through prediction and planning. Many tasks can be considered as states with common relational structures, allowing decision makers to predict outcomes in new situations with similar rules. Multiple theories hold that cognitive maps capture such relational structures, and exist in the hippocampalentorhinal circuit in both rodents and humans. Here, we use immersive virtual reality and long-term high-yield electrophysiology in mice to investigate the updating and application of such cognitive maps. First, we evaluated our method to achieve long-term electrophysiological recordings. Deployed among multiple researchers and laboratories, we showed that the technique was a feasible and flexible method for chronic recordings in rodents. We tested how relational structure can be learnt using predictions in an associative inference paradigm as mice navigated between virtual rooms. Online predictions bound environments beyond their virtual boundaries, including to infer the rooms’ relational structure, creating a coherent map of the combined environment. New transitions were preferentially replayed - and inferred transitions tested - offline during sleep. We asked how these maps could generalise and be applied in response to a change in an environment’s underlying structure. Mice learned to forage for rewards in a virtual 2D arena. We then changed the transition statistics of the environment by introducing “teleports” that linked remote locations across the virtual arena without changing its visual appearance. Mice freely chose to shortcut via these teleports. In response, spatial cells remapped near the teleports and made online predictions via the teleports. Offline, the new transition points were preferentially replayed. Taken together, hippocampal-entorhinal representations are consistent with mechanisms that could support model-based strategies using predictions of relational structure.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item