Anterior Cingulate Cortex, Error Detection, and the Online Monitoring of Performance

See allHide authors and affiliations

Science  01 May 1998:
Vol. 280, Issue 5364, pp. 747-749
DOI: 10.1126/science.280.5364.747


An unresolved question in neuroscience and psychology is how the brain monitors performance to regulate behavior. It has been proposed that the anterior cingulate cortex (ACC), on the medial surface of the frontal lobe, contributes to performance monitoring by detecting errors. In this study, event-related functional magnetic resonance imaging was used to examine ACC function. Results confirm that this region shows activity during erroneous responses. However, activity was also observed in the same region during correct responses under conditions of increased response competition. This suggests that the ACC detects conditions under which errors are likely to occur rather than errors themselves.

It has been proposed that the ACC plays a prominent role in the executive control of cognition (1). This hypothesis is based, in part, on functional neuroimaging studies that show ACC activity during tasks that engage selective attention, working memory, language generation, and controlled information processing (2). Disturbances in this brain region have been reported in disorders associated with cognitive impairment, including schizophrenia and depression (3). This account of ACC function is consistent with the rich anatomical connectivity of this region with association, limbic, and motor cortices (4). However, it is lacking in detail regarding the precise contribution of the ACC to cognitive control.

To date, the most explicit hypothesis regarding ACC function comes from event-related brain potential (ERP) studies during speeded response tasks. These studies have reported an error-related negativity (ERN), peaking 100 to 150 ms after a person shows electromyographic evidence of initiating an incorrect response. Dipole modeling suggests that the ERN has a medial frontal generator, possibly the ACC (5). These and other characteristics of the ERN have led to the hypothesis that the ACC is involved in monitoring and compensating for errors. Specifically it has been proposed that this involves a comparator process in which a representation of the intended, correct response is compared to a representation of the actual response (5, 6).

We propose that rather than implementing a comparator process, the ACC monitors competition between processes that conflict during task performance. For example, response competition arises when a task elicits a prepotent but inappropriate response tendency (manifested as activity in the incorrect response channel) that must be overcome to perform correctly. These conditions are also more likely to elicit incorrect responses, possibly accounting for the relationship of ACC activity to errors. However, our hypothesis predicts that response competition will activate the ACC even when a correct response is made. The present study employed event-related functional magnetic resonance imaging (fMRI) to accomplish two goals. First, we sought to test the hypothesis, suggested by ERN studies, that the ACC shows error-related activity. More important, we also sought to test our alternative hypothesis regarding the functional significance of the ERN and the performance-monitoring function of the ACC.

Thirteen people underwent fMRI (7) while performing variants of the Continuous Performance Test (AX-CPT) (8) that were designed both to increase error rates and to manipulate response competition (Fig. 1) (9). We observed a transient increase in ACC activity (10) occurring during incorrect responses (11) (Fig.2). However, as predicted by our hypothesis, greater ACC activity was also seen during correct responses, under conditions that elicited greater response competition (12). We interpret these results as suggesting that the contribution of the ACC to performance monitoring may be the detection of response competition rather than detection of errors per se. Sensitivity to response competition should also produce error-related effects, because error trials are often likely to be those in which activity in the incorrect response channel competes with, and prevails over, activity in the correct response channel.

Figure 1

The AX-CPT task. Stimuli were presented centrally on a visual display projected into the scanner [duration of cue and probe, 0.5 s; duration of interstimulus interval (ISI), 9.5 s; and duration of intertrial interval (ITI), 9.5 s]. Trials occurred as sequences of cue-probe pairs. Stimuli were either degraded or nondegraded letters, presented as cue-probe pairs. Eight time-locked scans were acquired during each 20-s trial, one every 2.5 s, and included sufficient time for the MRI signal to decay to baseline after transient response-related activity.

Figure 2

The location of the region of the ACC (BA 24/32) and its associated effects. On the left, the functional image from the confirmatory analysis of the 39-voxel ACC ROI (yellow) is rendered onto a three-dimensional structural MRI image for anatomic visualization. On the right, the temporal dynamics of activity in this region are shown for both the error and the trial-type effects, plotted as percent change from the mean MRI signal for each condition. Error bars indicate the standard error of the mean. The error effect (observed as an error × scan interaction during the degraded condition) occurred during the time of response as a transient increase in activity on incorrect trials. The competition effect (observed as a cue × probe × scan interaction) is shown only for correct trials and also occurred during the time of response. A transient increase in activity was found for trials with high response competition (AY and BX) relative to trials with low competition (AX and BY).

To determine whether error and competition effects were specific to the ACC, we used two additional analyses. First, we examined error- and competition-related effects in three other regions of interest (ROIs) previously shown to be activated by stimulus degradation (two right inferior frontal regions and a right striatal region) (8). We observed error-related activity in one right frontal region (BA 44/45), but neither this nor the two other ROIs showed competition effects (13). Next, we examined error and competition effects in an exploratory analysis of all brain regions scanned (14). Three regions in addition to the ACC showed significant transient increases in activity associated with errors. These were the right (BA 9) and left (BA 46/9) dorsolateral prefrontal cortex and the left premotor cortex (BA 6). However, none showed significant effects of response competition.

Our reconceptualization of ACC activity as being related to response competition, rather than errors per se, has a number of important implications. First, it links the wealth of literature concerning the role of this region in higher level cognition [including the hypothesis that it is involved with late selection or “attention to action” (1)] with the ERP literature suggesting that it is responsive to errors. In particular, it reconciles the observation that reliable ACC activation is observed in some tasks that are associated with low error rates, such as the Stroop task and verbal fluency (15, 16). Second, it is computationally parsimonious. A comparator function requires the ACC to have access to representations of intended states (for example, the correct action) as well as the outcome of processing, whereas detection of competition requires only information about the state of the response system.

This study focused on ACC activity associated with response competition. The question remains open as to whether the ACC is responsive to competition only at this level or is also responsive to competition earlier in processing. Whichever is correct, the present findings demonstrate how error-related activity can occur without the need for a comparator and how such activity might represent one instance of a more general performance-monitoring function of the ACC.


View Abstract

Navigate This Article