CENTER FOR RESEARCH IN
COGNITION & NEUROSCIENCES/ConnexConnex

Human subjects can learn efficiently even in stochastic and changing environments. This capability relies on the dynamic adjustment of a flexibility-stability tradeoff in learning but its neural mechanism remains debated. In Bayesian learning models, surprise tracks unexpected outcomes arising from stochasticity or changes while confidence gates the updating process, promoting flexibility when a change is suspected. I will first present an investigation of different types of surprise across a variety of subcortical nuclei notably neuromodulator centers using novel dedicated fMRI methods. Then, I will focus on preliminary results testing the involvement of several neuromodulators (noradrenaline, acetylcholine, dopamine, serotonin) in the coding of both surprise and confidence during adaptive learning. Finally, I will discuss these results with respect to confidence in decision making and will briefly present future projects.