Flexible task representations for intelligent behavior (360G-Wellcome-210849_Z_18_Z)

A long research tradition in cognitive neuroscience has investigated the role of the prefrontal cortex (PFC) in flexible goal-directed behaviour. This work has yielded important insights, e.g., that the PFC’s coding properties flexibly adapt when task demands change to prioritise goal-relevant information. However, it relies primarily upon the use of familiar and explicitly instructed tasks that provide poor fit to human behaviour in naturalistic and open-ended environments, and likely miss essential aspects PFC function. I therefore propose to study the representations that form within PFC as humans learn to perform complex categorisation tasks with only minimal external instruction. I hypothesise that during learning the PFC will form abstract representations of the task’s latent structure, that support not only performance in the given task, but also provide provides a scaffold for learning in similar environments. Using representational similarity analyses, in combination with multiple brain recording techniques (fMRI, MEG, ECoG), I will measure how stimulus-evoked neural responses change adaptively during learning, and test if these changes generalise to novel task environments with similar structure. These findings will provide important new insights about the coding mechanisms by which PFC supports flexible behaviour in complex, naturalistic environments.