Lateral habenula neurons signal step-by-step changes of reward prediction.

In real life, multiple objects of different values are mixed in a variety of environments. To survive, animals need to find rewarding objects that may be located but hidden in particular contexts (e.g., environments) with bad objects that are unassociated with reward. Then, animals and humans pay attention to the enriched environment so that they can find the rewarding object vigorously. How can the brain initiate such behavior based on the context? We thus created a behavioral task for monkeys in which multiple contextual events (environment, action cue) sequentially occurred before objects appeared. We then studied the lateral habenula (LHb), which inhibit dopamine neurons (Matsumoto and Hikosaka, 2007). LHb neurons showed phasic responses in each event step-by-step across the sequential events, whose direction (excitation or inhibition) corresponded to the immediate change of the predicted value. Moreover, LHb neurons sequentially compared detailed prediction errors based on their significance in multiple contexts.