To make reward-guided decisions we must form predictions about future reward, generate action plans, and update behavior when reward predictions are violated. I will present neural correlates of these functions in rats deciding between two rewarding options. In these studies value was manipulated by varying reward size and delay to reward (i.e., time-discounting). I will describe data demonstrating that single unit firing of neurons in orbitofrontal cortex reflects reward predictions, whereas firing of midbrain dopamine neurons represents errors in reward prediction. Furthermore, I will show that reward prediction errors signals in the midbrain are dependent on orbitofrontal cortex. Signals from these regions are likely translated into behavior via projections to ventral and dorsal striatum. Neural activity in ventral striatum is modulated by value and motivation, whereas encoding in dorsal striatum is highly associative (e.g., stimulus-response associations). These two striatal systems appear to run in parallel because disruption of one region does not disrupt encoding in the other. Finally, I will describe data from substantia nigra pars reticulata, an area more closely tied to motor output and response bias. Thus, I will describe neural correlates from several nodes in a circuit necessary for transforming sensory information into to motivated behavior.