Neural representations of compositional elements in flexible behaviour.

Flexible behaviour depends on the brain’s ability to extract structure from experience and apply it to novel situations. This is thought to rely on latent states – internal representations that capture hidden structure in the world and serve as behavioural frameworks for planning and decision-making. To support flexibility, these models must also be compositional – generalise to new scenarios by combining and rearranging their individual components. While growing evidence suggests that the hippocampus encodes abstract latent states and organises simple representations into sequences, the neural mechanisms underlying the development and compositional use of abstract latent states remain poorly understood. This project will investigate how latent states are constructed and used compositionally and uncover the learning algorithms and neural mechanisms that support their development. Using a complex spatial navigation task in mice, I will combine high-density electrophysiology, two- photon calcium imaging, and causal manipulations to track the emergence, organisation, and compositional use of latent state representations in the ventral hippocampus. I will test how hippocampal replay and dopamine signalling shape and update these representations. This work will establish a mechanistic account of how the brain abstracts experience into compositional elements to support flexible behaviour – core process of complex cognition often disrupted in mental illness.