According to contemporary learning theories, the discrepancy, or error, between the actual and predicted reward determines whether learning occurs when a stimulus is paired with a reward. The role of prediction errors is directly demonstrated by the observation that learning is blocked when the stimulus is paired with a fully predicted reward. By using this blocking procedure, we show that the responses of dopamine neurons to conditioned stimuli was governed differentially by the occurrence of reward prediction errors rather than stimulus±reward associations alone, as was the learning of behavioural reactions. Both behavioural and neuronal learning occurred predominantly when dopamine neurons registered a reward prediction error at the time of the reward. Our data indicate that the use of analytical tests derived from formal behavioural learning theory provides a powerful approach for studying the role of single neurons in learning.Classic theories assume that predictive learning occurs whenever a stimulus is paired with a reward or punishment 1,2 . However, more recent analyses of associative learning argue that simple temporal contiguity between a stimulus and a reinforcer is not suf®cient for learning and that a discrepancy between the reinforcer that is predicted by a stimulus and the actual reinforcer is also required 3±6 . This discrepancy can be characterized as a`prediction error' 7 . Presentations of surprising or unpredicted reinforcers generate positive prediction errors, and thereby support learning, whereas omissions of predicted reinforcers generate negative prediction errors and lead to reduction or extinction of learned behaviour. Expected reinforcers do not generate prediction errors and therefore fail to support further learning even when the stimulus is consistently paired with the reinforcer. Modelling studies have shown that neuronal messages encoding prediction errors can act as explicit teaching signals for modifying the synaptic connections that underlie associative learning 8±17 .Current research suggests that one of the principal reward systems of the brain involves dopamine neurons 18±21 . Both psychopharmacological manipulations and lesions of the dopamine system impair reward-driven behaviour of animals 18,21 , and drugs of abuse, which provide strong arti®cial rewards, act via dopamine neurons 19,20 . Neurobiological investigations of associative learning have shown that dopamine neurons respond phasically to rewards in a manner compatible with the coding of prediction errors 22±28 , whereas slower dopamine changes are involved in a larger spectrum of motivating events 18,28,29 . Dopamine neurons show short-latency, phasic activations when rewards occur unpredictably, they are not modulated by fully predicted rewards and show phasically reduced activity when predicted rewards are omitted. During initial learning, when rewards occur unpredictably, dopamine neurons are activated by rewards. They gradually lose the response as the reward becomes increasingly predicted 24,25,27 . The co...