Dissociating the Role of the Dorsolateral Prefrontal and Anterior Cingulate Cortex in Cognitive Control

See allHide authors and affiliations

Science  09 Jun 2000:
Vol. 288, Issue 5472, pp. 1835-1838
DOI: 10.1126/science.288.5472.1835


Theories of the regulation of cognition suggest a system with two necessary components: one to implement control and another to monitor performance and signal when adjustments in control are needed. Event-related functional magnetic resonance imaging and a task-switching version of the Stroop task were used to examine whether these components of cognitive control have distinct neural bases in the human brain. A double dissociation was found. During task preparation, the left dorsolateral prefrontal cortex (Brodmann's area 9) was more active for color naming than for word reading, consistent with a role in the implementation of control. In contrast, the anterior cingulate cortex (Brodmann's areas 24 and 32) was more active when responding to incongruent stimuli, consistent with a role in performance monitoring.

Cognitive control has long attracted the attention of philosophers and psychologists interested in how the human brain carries out the higher functions of awareness, memory, and language. The concept of control generally refers to a resource-limited system that guides voluntary, complex actions. Solving difficult, novel, or complex tasks, overcoming habitual responses, and correcting errors all require a high degree of cognitive control. Cognitive control has frequently been operationalized as the provision of top-down support for task-relevant processes; for example, a representation of the attentional demands of the task can be used to bias processing in favor of task-relevant stimuli and responses and thereby establish the appropriate stimulus-response mapping (1–3). Other work suggests that a second component is required to provide ongoing feedback indicating whether control is being allocated effectively (4, 5).

Studies using functional neuroimaging techniques have related cognitive control to activity in the dorsolateral prefrontal cortex (DLPFC) [Brodmann's area (BA) 9 and BA 46] and the anterior cingulate cortex (ACC) (BA 24 and BA 32). For example, both the DLPFC and ACC activate when participants are required to hold increasingly long sequences of items in working memory or when two tasks are performed at once, compared to when they are performed one at a time (6). From these data, it is impossible to dissociate the respective roles of the DLPFC and ACC because, as the tasks become more difficult, there are increased demands both for strategic processes, such as top-down support, and for evaluating the output of the system.

A number of neuroimaging studies have reported relative dissociations for these regions. DLPFC activity in the absence of ACC activity has been found for tasks that require maintenance and manipulation of information in working memory (7). For example, the DLPFC is active when a simple cue has to be maintained over a delay (8). ACC activity has been more consistently observed than DLPFC activity when tasks require divided attention, novel or open-ended responses, or the overcoming of a prepotent response (9). For example, the traditional Stroop task involves naming the ink color of colored words. Sometimes the word and ink color are congruent (“RED” printed in red ink), and sometimes they are incongruent (“RED” printed in blue ink). Because participants automatically read the word, they are slower to name the color in the incongruent condition, which is also when greater ACC activation is observed (10).

These relative dissociations led us to hypothesize that the DLPFC may be involved in representing and maintaining the attentional demands of the task. In contrast, the ACC may be involved in evaluative processes, such as monitoring the occurrence of errors or the presence of response conflict, which occurs when two incompatible responses are both compelling. Monitoring such occurrences is necessary to provide feedback as to when strategic processes must be more strongly engaged to adapt ongoing behavior. However, because these studies report single dissociations, this interpretation is tenuous; that is, both the DLPFC and ACC may be engaged for many of these tasks, but either because of the demands of the subtractive condition or a lack of power, activity in only one region crosses a statistical threshold. As a result, uncertainty remains about how these two frontal regions contribute to cognitive control processes.

To test for a hypothesized double dissociation between the functions of the DLPFC and ACC, we used a modified version of the Stroop paradigm (Fig. 1), in which subjects were given an instruction before each trial indicating whether to read the word (a more automatic response) or name the color (requiring greater control). Following a delay, the stimulus was presented. Thus, the task temporally separated instruction-related strategic processes, including those responsible for representing and maintaining the attentional demands of the task (color naming versus word reading) from response-related, including evaluative, processes.

Figure 1

Functional magnetic resonance imaging (fMRI) activity across the course of a trial in the left DLPFC (L. DLPFC) (BA 9; Talairach coordinates x = –41, y = 18,z = 28; maximum F = 7.14; 36 pixels) and ACC (BA 24 and BA 32; Talairach coordinates x = 4,y = 1, z = 43; maximumF = 7.98; 10 voxels). At the beginning of each 25-s trial, subjects were given an instruction to either read the word or name the color of the following stimulus. A colored word was presented 12.5 s after the beginning of each trial. Significant differences between conditions were detected in the left DLPFC across scans 1 to 5 (lower left quadrant) and in the ACC across scans 7 to 10 (upper right quadrant).

Twelve participants completed the switching Stroop task (11) during a functional magnetic resonance imaging scanning session (12). Instruction-related activity was examined for the first five scans of each trial. As illustrated in Fig. 1, activity was observed within the DLPFC in response to instructions to name the color but not read the word. This pattern was only observed within the DLPFC and is consistent with the increased requirement for top-down control in the color-naming task and the role of the DLPFC in representing and maintaining task demands needed for such control. No instruction-related activity was observed in the ACC. We hypothesized that if the DLPFC implements a top-down control function, more activity in this region should lead to less conflict (i.e., smaller reaction time interference effects) when responding to incongruent colored words. Consistent with this prediction, individuals who showed the most activation in the left DLPFC after the color-naming instruction showed the smallest Stroop interference effect (correlation coefficientr = −0.63, P = 0.02, one-tailed).

Response-related activity was examined for the last four scans of each trial. As illustrated in Fig. 1, within the right ACC, greater activity was observed for incongruent, compared to congruent, color-naming trials, consistent with a role in conflict monitoring (13). Although the DLPFC was active during the response, it was no more active during incongruent than during congruent color-naming trials. If the ACC monitors conflict, then high conflict (i.e., larger reaction time interference effects) should be associated with more ACC activation. Consistent with this prediction, individuals who showed the largest Stroop interference effect tended to have more ACC activation, although this effect was not significant for the number of subjects tested (r = 0.38, P = 0.12, one-tailed).

These findings suggest a dissociation between the left DLPFC and the ACC. The left DLPFC was selectively engaged during the preparatory period, more for color naming than for word reading. This is consistent with a role in the implementation of control, by representing and actively maintaining the attentional demands of the task. In contrast, the ACC was selectively activated during the response period, more for incongruent than for congruent color-naming trials. This is consistent with a role in conflict monitoring. Thus, these regions appear to have distinct, complementary roles in a neural network serving cognitive control.

The association between the left DLPFC and strategic control processes is consistent with findings from neuropsychological, neurophysiological, and lesion studies, as well as from computational analyses of the function of the prefrontal cortex. For example, patients with injuries to this region have shown a great deal of difficulty on the Stroop task, as well as with other tasks that require the representation and maintenance of the attentional demands of the task (14). Sustained neural activity during a delayed response task in a homologous region of the primate frontal cortex has been interpreted as strategic processing related to control, and lesions to this region result in an inability to use complex information stored in working memory (15). Computational modeling of the Stroop and related paradigms suggested that the DLPFC may be responsible for maintaining and representing context information, including the attentional demands of the task (16).

ACC activation during responses to incongruent colored words on the Stroop task has been found in previous studies using blocked paradigms and is frequently interpreted as evidence that the ACC is involved in implementing control [e.g., (10, 17)]. The current dissociation is not consistent with the interpretation that the ACC is implementing control by representing and maintaining the attentional demands of the task. Another possibility is that the ACC performs a more transient form of control, preferentially engaged by attentionally demanding (e.g., incongruent) stimuli. However, a number of considerations weigh against this interpretation of the observed ACC activity. Regions implementing control should be negatively correlated with the amount of conflict (i.e., more control, less conflict). This pattern was observed in the DLPFC, but the correlation between ACC activity and conflict was in a positive direction. This finding is consistent with significant, positive correlations found in two previous studies of ACC activity and conflict (18). This previous work has also shown that ACC activity increases when top-down control is low, whether control is reduced from trial to trial or across a number of trials (18). Further, computational modeling work suggests that many behavioral effects related to Stroop interference can be explained by representing and maintaining the attentional demands of the task, without the need for stimulus-dependent transient control (12). Therefore, it seems unlikely that the ACC activity observed in the current study reflects a transient form of top-down control, although if it does, the mechanism of this control must be quite different from attentional activity associated with the DLPFC.

One parsimonious interpretation of these results is that control and evaluative processes are linked in a negative feedback loop as part of a network to maintain optimal performance. The controversial role of the ACC in control highlights the problem of parsing the components of such a network; if evaluative components signal when more control is required, activity in regions serving these two functions will be correlated. In the past, the limited temporal resolution of functional neuroimaging relative to the underlying neural events made it difficult to discern whether increases in ACC activity were coincident with or produced increases in DLPFC activity. In a previous study that took advantage of the faster time scale of the event-related potential, increased error-related negativity magnitude (thought to index ACC activity) was followed by reduced error commission and an increased latency for the next correct response (19), suggesting that control is increased in response to the ACC signal. In recent modeling work (15), it was demonstrated that a simple feedback loop, incorporating a conflict-monitoring mechanism that regulates the strength of task-demand representations, can account for a variety of strategic adjustments in control that have been observed in simple response choice and selective attentional tasks, including the Stroop [e.g., (20)].

Taken together, these data suggest that cognitive control is a dynamic process implemented in the brain by a distributed network that involves closely interacting, but nevertheless anatomically dissociable, components. Within this system, the DLPFC provides top-down support of task-appropriate behaviors, whereas other components, such as the ACC, are likely to be involved in evaluative processes indicating when control needs to be more strongly engaged. We look forward to future studies that further detail these complementary functions and the role that the DLPFC and ACC play in the regulation of cognition and behavior.

  • * To whom correspondence should be addressed. E-mail: cartercs{at}


View Abstract

Stay Connected to Science

Navigate This Article