Positive and negative reward coding by serotonin, acetylcholine, and dopamine network.

点击翻译按钮获取中文摘要

The strength of firing of midbrain dopaminergic neurons has been revealed that it carries reward prediction error signal in the framework of reinforcement learning where an agent chooses actions to maximize rewards. The dopamine neurons receive inputs from the pedunculopontine tegmental nucleus(PPTN), where many cholinergic neurons exist, and the dorsal raphe nucleus(DRN), where many serotonergic neurons exist. Thus we measured single neuronal activity of PPTN and DRN while monkeys performed eye movement tasks to obtain reward of different values and the Pavlovian conditioning task in which appetitive and aversive events were imposed. We found that while different PPTN neurons encoded predicted reward and actual rewards separately, DRN neurons often carried predicted and actual rewards by single neurons. Such different reward signals may be fed and integrated in the dopamine neurons to compute reward prediction error signal, which is used for the control of action and decision making.