How dopamine neurons devalue delayed rewards

The activity of neuronal cells that release the neurotransmitter dopamine is thought to encode differences between predicted and actual rewards. This ‘prediction error’ is key to reinforcement learning. It emerges that dopamine neurons differ in their timescales for temporal discounting, enabling the brain to implement sophisticated reinforcement-learning algorithms along multiple timescales. Reinforcement learning in the brain relies on different dopamine neurons discounting rewards at different rates.