Speaker
Description
Humans excel at rapidly learning and flexibly performing multiple tasks. This cognitive flexibility is thought to be the hallmark of human intelligence. Flexible learning requires retaining and utilizing previously learnt information while remaining sensitive to changes in the environment as well as rapid mental adjustments across different tasks. An open question remains how the brain balances utilizing learnt information while flexibly performing multiple tasks on the same set of perceptual features. We asked 1) whether this requires neural codes that are specialized for individual tasks or multipurpose codes that can exhibit context-dependent flexibility, and 2) if multi-purpose codes are utilized, where these flexible codes exist in the human brain – i.e., in the fronto-parietal network or already in sensory cortices. To this end, we conducted intracranial electrophysiological recordings from a wide range of brain areas in human epilepsy patients while they learn and perform three different tasks in a context-dependent manner, i.e., individuate, categorize and form abstract concepts over similar visual stimuli. By assessing representational geometries, we find that multi-purpose neural codes exist already in the temporal cortex where representational spaces are flexibly adjusted to perform the different tasks. Therefore, our results provide evidence against the traditional outlook that sensory neural codes in the temporal lobe simply represent the external world in a stable manner. Instead, we find that neural representations in the temporal cortex are multi-purpose, exhibit context-dependent flexibility based on task demand and can already be used to learn and perform multiple tasks.
