PREDICTIVE REWARD SIGNAL OF DOPAMINE NEURONS

The effects of lesions, receptor blocking, electrical self-stimulation , and drugs of abuse suggest that midbrain dopamine systems are involv ed in processing reward information and learning approach behavior. Mo st dopamine neurons show phasic activations after primary liquid and f ood rewards and conditioned, reward-predicting visual and auditory sti muli. They show biphasic, activation-depression responses after stimul i that resemble reward-predicting stimuli or are novel or particularly salient. However, only few phasic activations follow aversive stimuli . Thus dopamine neurons label environmental stimuli with appetitive va lue, predict and detect rewards and signal alerting and motivating eve nts. By failing to discriminate between different rewards, dopamine ne urons appear to emit an alerting message about the surprising presence or absence of rewards. All responses to rewards and reward-predicting stimuli depend on event predictability. Dopamine neurons are activate d by rewarding events that are better than predicted, remain uninfluen ced by events that are as good as predicted, and are depressed by even ts that are worse than predicted. By signaling rewards according to a prediction error, dopamine responses have the formal characteristics o f a teaching signal postulated by reinforcement learning theories. Dop amine responses transfer during learning from primary rewards to rewar d-predicting stimuli. This may contribute to neuronal mechanisms under lying the retrograde action of rewards, one of the main puzzles in rei nforcement learning. The impulse response releases a short pulse of do pamine onto many dendrites, thus broadcasting a rather global reinforc ement signal to postsynaptic neurons. This signal may improve approach behavior by providing advance reward information before the behavior occurs, and may contribute to learning by modifying synaptic transmiss ion. The dopamine reward signal is supplemented by activity in neurons in striatum, frontal cortex, and amygdala, which process specific rew ard information but do not emit a global reward prediction error signa l. A cooperation between the different reward signals may assure the u se of specific rewards for selectively reinforcing behaviors. Among th e other projection systems, noradrenaline neurons predominantly serve attentional mechanisms and nucleus basalis neurons code rewards hetero geneously. Cerebellar climbing fibers signal errors in motor performan ce or errors in the prediction of aversive events to cerebellar Purkin je cells. Most deficits following dopamine-depleting lesions are not e asily explained by a defective reward signal but may reflect the absen ce of a general enabling function of tonic levels of extracellular dop amine. Thus dopamine systems may have two functions, the phasic transm ission of reward information and the tonic enabling of postsynaptic ne urons.