Rewards satisfy the needs of biological agents for survival and gene propagation that result in evolutionary fitness. To do so, rewards elicit learning (positive reinforcement), approach behaviour, economic choice and emotions like pleasure, happiness and desire. Dopamine neurons signal reward prediction errors (RPE) that are suitable for learning according to reinforcement learning theory, including Pavlovian learning and temporal difference learning. As biological needs vary between individual agents, the value of reward is subjective, which can be assessed using principles of economic choice theory. Theory-driven tests ensure that animals make meaningful choices for inferring subjective reward value. Dopamine neurons code RPEs in the subjective value formalized as utility. Concepts of Revealed Preference Theory allow to extend these tests to rewards that typically contain multiple components. Neuronal responses in monkey orbitofrontal cortex follow preferences for multi-component rewards, even when preferred rewards have smaller individual components. These responses reflect the transformation of vectorial multi-component reward into scalar subjective value. In sum, concepts from animal learning theory and economic choice theory inform experimental designs of neurophysiological reward experiments that transcend ad-hoc situations and validate, extend and ultimately reduce theories of reinforcement learning and economic choice.