One of the key goals in neuroscience is to understand how neural activity generates behavior. Electrophysiological recordings have provided great insights into this relationship, but they cannot resolve the contribution of specific neuronal cell types. Moreover, it has been notoriously difficult to move beyond correlation and demonstrate the functional significance of neural firing for specific behaviors. Recently developed optogenetic techniques overcome these limitations by tagging specific neuronal populations for electrophysiological identification and manipulation in vivo. The experiments described in this proposal apply this approach to midbrain dopamine (DA) neurons in awake, behaving mice.DA neurons are key players in reward-processing. They respond to unexpected rewards and sensory cues predicting future rewards, while unexpected reward omission suppresses them. Thus, DA neurons seem to encode the discrepancy between predicted and actual reward, also called reward prediction errors. Based on formal theories of reinforcement learning, it has been proposed that DA acts as a teaching signal to mediate learning from reward. Here, we propose to evaluate the causal relationship between DA activation to reward-predicting cues and ongoing behavior, learning and extinction. Moreover, we will perform an analysis of the neuronal computations underlying dopamine’s role in reinforcement learning to understand how DA responses are generated at the circuit level. Since disturbances of reinforcement learning have been directly linked to human pathological conditions including schizophrenia, depression and autism as well as maladaptive behaviors in addiction, the research program put forward in this proposal has broad implications for improving our understanding of these pathological conditions and may ultimately support the development of novel interventional strategies.