Nucleus accumbens (NAc) is thought to contribute to motivated behavior by signaling the value of reward-predicting cues and the delivery of anticipated reward. NAc is subdivided into core and shell, with each region containing different populations of neurons that increase or decrease firing to rewarding events. While there are various and numerous theories of functions pertaining to these subregions and cell types, most are in the context of reward processing, with fewer considering that NAc might serve functions related to action selection more generally. We recorded from single neurons in NAc as rats of both sexes performed a STOP-change task that is commonly used to study motor control and impulsivity. In this task, rats respond quickly to a spatial cue on 80% of trials (GO) and must stop and redirect planned movement on 20% of trials (STOP). We found that the activity of reward-excited neurons signaled accurate response direction on GO, but not STOP trials, and that these neurons exhibited higher pre-cue firing after correct trials resulting in stronger firing during correct GO trials and errant STOP trials, while reward-inhibited neurons significantly represented response direction on STOP trials at the time of the instrumental response. Finally, the proportion of reward-excited to reward-inhibited neurons and the strength of pre-cue firing decreased as the electrode traversed NAc. We conclude that reward-excited cells (more common in core) promote proactive action selection, while reward-inhibited cells (more common in shell) contribute to accurate responding on STOP trials that require reactive suppression and redirection of behavior.Significant Statement The ability to appropriately adapt behavior is an important part of human cognition, and one that is disrupted by many neuropsychiatric disorders. Here we recorded from neurons in the nucleus accumbens (NAc) as rats performed a cognitive control task and found cell type- and subregion-specific firing patterns. Core and reward-excited cells track trial outcome history, proactively driving behavior to the first cue-a strategy that is appropriate for most trials. Conversely, shell and reward-inhibited neurons signal accurate response direction on trials requiring redirection of behavior. Together, these data suggest that NAc neuronal populations differentially contribute to action selection.