Flexible and abstract neural representations of structural knowledge

Abstract Humans’ ability for generalisation is outstanding. Our previous research has demonstrated that humans can generalise over tasks that share the same statistical rules (structural form) that govern the relational structure between states within a task. Studies suggest that the brain represents non-spatial tasks’ structure similarly to the representation of spatial tasks and have highlighted the entorhinal cortex (EC) as a possible neural substrate for this structural generalisation. EC spatial representation generalises across environment’s size and shape. However, in non-spatial domains this was only shown across tasks with the exact same underlying graphs. Here we show with fMRI that EC representations generalise across complex non-spatial tasks that share a hexagonal grid structural form but differ in their size and sensory stimuli, i.e. their only shared feature is the rule governing their statistical structure.