The authors develop a theoretical framework to explain how mesencephalic dopamine systems distribute signals representing future expectations to their targets. They propose that activity in the cerebral cortex can make predictions about future rewards, and fluctuations in the activity levels of neurons in diffuse dopamine systems above and below baseline levels represent errors in these predictions, which are then delivered to cortical and subcortical targets. The theory is consistent with physiological data from dopaminergic neurons in the ventral tegmental area and surrounding regions. The authors also present a model for how such errors could be constructed in a real brain and make testable predictions about human choice behavior in a simple decision-making task. The theory suggests that dopamine neuron output during learning tasks can be explained by a general learning principle, and it provides insights into how information about future events can be represented in ways more subtle than tonic firing during delay periods. The model's predictions are supported by experimental data from monkey conditioning tasks and human subjects in a card choice experiment. The authors discuss the implications of their theory for understanding the role of dopamine in learning, behavioral control, and decision-making.The authors develop a theoretical framework to explain how mesencephalic dopamine systems distribute signals representing future expectations to their targets. They propose that activity in the cerebral cortex can make predictions about future rewards, and fluctuations in the activity levels of neurons in diffuse dopamine systems above and below baseline levels represent errors in these predictions, which are then delivered to cortical and subcortical targets. The theory is consistent with physiological data from dopaminergic neurons in the ventral tegmental area and surrounding regions. The authors also present a model for how such errors could be constructed in a real brain and make testable predictions about human choice behavior in a simple decision-making task. The theory suggests that dopamine neuron output during learning tasks can be explained by a general learning principle, and it provides insights into how information about future events can be represented in ways more subtle than tonic firing during delay periods. The model's predictions are supported by experimental data from monkey conditioning tasks and human subjects in a card choice experiment. The authors discuss the implications of their theory for understanding the role of dopamine in learning, behavioral control, and decision-making.