Mechanisms of flexible behaviour in complex environments

I aim to understand brain mechanisms of flexible behaviour in complex environments, focusing on how predictive models of the environment’s structure are used for action selection. The key questions are: - What is the computational structure of the model used for planning? Does it store local relationships between locations and use step-by-step simulation to evaluate options, or store long-range relationships allowing rapid action selection at the cost of reduced flexibility? - What is the differential contribution of, and interaction between, medial frontal cortex and hippocampus in model-based action? Do these regions represent different levels of hierarchically organised behaviour? - How does information in the model get translated into action? Do dopaminergic reward prediction errors, or short-term memory by recurrent cortical activity, store the output of model-based evaluations to guide choices? I will answer these questions using a novel behavioural assay for mice, in which they choose among goals, and plan routes, in a complex maze environment ideally suited to mathematically modelling planning computations. I will characterise activity in frontal cortex and hippocampus using high density silicon probes, and in the dopamine system using photometry, including simultaneous recordings to study interaction between regions, and optogenetic manipulations to test causality.