People are capable, at will, of trading speed for accuracy when performing a task; they can focus on performing accurately at the cost of being slow, or emphasize speed at the cost of decreased accuracy. Here, we used functional magnetic resonance imaging to investigate the neural correlates of this ability. We show increased baseline activity during speed emphasis in a network of areas related to response preparation and execution, including the premotor areas of the frontal lobe, the basal ganglia, the thalamus, and the dorsolateral prefrontal and left parietal cortices. Furthermore, speed emphasis was associated with reduced transient response-related activation in several of these structures, suggesting that because of the greater baseline activity under speed emphasis, less activation is needed in these structures to reach response threshold, consistent with the assumptions of several computational theories. Moreover, we identify the dorsolateral prefrontal cortex as providing the top-down control signal that increases this baseline activity.

The anterior cingulate cortex (ACC) has been shown to respond to conflict between simultaneously active, incompatible response tendencies. This area is active during high-conflict correct trials and also when participants make errors. Here, we use the temporal resolution of high-density event-related potentials (ERPs) in combination with source localization to investigate the timing of ACC activity during conflict and error detection. We predicted that the same area of the ACC is active prior to high-conflict correct responses and following erroneous responses. Dipole modeling supported this prediction: The frontocentral N2, occurring prior to the response on correct conflict trials, and the ERN, occurring immediately following error responses, could both be modeled as having a generator in the caudal ACC, suggesting the same process to underlie both peaks. Modeling further suggested that the rostral area of the ACC was also active following errors, but later in time, contributing to the error positivity ( P E ), and peaking at 200–250 msec following the ERN peak. Despite the inherent limitations of source localization, these data may begin to shed light on the timing of action-monitoring processes. First, the time course of caudal ACC activity follows the time course as predicted by the conflict theory of this region. Second, caudal ACC activity might be temporally dissociated from rostral ACC activity during error trials, which possibly reflects a separate, affective component of the evaluative functions of the ACC.