Decision making has both cognitive and affective components, but previous neuroimaging studies in this domain predominantly have focused on affect and reward. The current study examined a decision-making paradigm that placed strong demands on cognitive control processes by making reward payoffs contingent upon decision-making history. Payoffs were maximized by choosing the option that, paradoxically, was associated with a lower payoff on the immediate trial. Temporal integration requirements were manipulated by varying, across conditions, the window of previous trials over which the reward function was calculated. The cognitive demands of the task were hypothesized to engage neural systems responsible for integrating and actively maintaining actions and outcomes over time and the top-down biasing of response selection. Brain activation was monitored with functional magnetic resonance imaging (fMRI) using a mixed-blocked and event-related design to extract both transient and sustained neural responses. A network of brain regions commonly associated with cognitive control functions, including bilateral prefrontal cortex (PFC), bilateral parietal cortex, and medial frontal cortex, showed selectively sustained activation during the task. Increasing temporal integration demands led to a shift from transient to sustained activity in additional regions, including right hemisphere dorsolateral and frontopolar PFC. These results demonstrate the contribution of cognitive control mechanisms to temporally extended decision-making paradigms and highlight the benefits of decomposing activation responses into sustained and transient components.