Summary
Dopamine (DA) released from VTA neurons in response to external cues or rewards may represent a learning signal. Here we introduce a spatial task where a reward-predicting cue instructs mice to navigate to a specific location. Simultaneous in vivo single-unit recordings revealed that DA neurons multiplexed an internal representation of the animal’s goal-directed actions (locomotion, licking, distance) with phasic responses to cue and reward. Neuronal activity discriminated between rewarded and failed trials, generating an error signal even in the absence of external cues. Following a contingency change, mice readily learned to move to a different location, which became impossible, if the internal error signal was jammed by optogenetic stimulation. We conclude that a multiplexed internal representation of the task modulates VTA DA neuron activity, engaging a learning process that leads to the behavioral adaptation of goal directed action.