O'Leary, Alice; (2023) Learning to Learn, an Optical Imaging Study of Hippocampal Representations During Learning and Generalization. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text O'Leary__thesis.pdf - Other Download (8MB) | Preview

Abstract

The hippocampus has long been thought to provide a map of spatial location. However, recent experiments suggest that the same circuits also map task-relevant continuous features such as temporal delays, auditory-frequency space, and discrete relations such as odour-place pairings. Thus, the hippocampus may, plausibly, provide a cognitive map of abstract task space, which may generalize across contexts to facilitate flexible problem-solving. Entorhinal grid cells are thought to facilitate this generalization as they maintain their firing phase offsets across environments, while hippocampal place cells remap. However, in CA1 a subset of cells maintain their firing fields relative to salient landmarks and reward zones across environments. The aim of the current experiment is to investigate whether CA1 representations of abstract rules might also generalize. GCaMP6f mice (n=6, additional pilot n=20) ran laps of a cue-rich 7m virtual linear track with a fixed goal. On each trial two additional cues appeared at pseudorandom locations. The reward-cue predicted an invisible reward zone, located +70cm from its centre. The distractor-cue did not predict reward. On each trial animals licked within reward zones associated with the moving reward-cue and fixed goal to receive rewards. After mice were expert in this task (~3 weeks), they were exposed to the same task in 2 additional environments, using different cues. Learning of the cue task was accelerated in environments 2 and 3. Two-photon imaging was used to record from CA1 across learning and generalization. Alongside place cells, a subset of neurons had stable fields locked to the reward- and distractor-cues. Reward-cue cells were not purely reward-anticipation cells. However, while place cells remapped across environments, reward-cue cells maintained their firing fields. Thus, despite previous evidence that spatial context drives complete remapping, when animal’s flexibly transfer previous learning to a new spatial context, CA1 representations of task rules are maintained.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item