Abstract

Errors trigger changes in behavior that help individuals adapt to new situations. The dorsal anterior cingulate cortex (dACC) is thought to be central to this response, but more lateral frontal regions are also activated by errors and may make distinct contributions. We investigated error processing by studying 2 distinct error types: commission and timing. Thirty-five subjects performed a version of the Simon Task designed to produce large number of errors. Commission errors were internally recognized and were not accompanied by explicit feedback. In contrast, timing errors were difficult to monitor internally and were explicitly signaled. Both types of error triggered changes in behavior consistent with increased cognitive control. As expected, robust activation within the dACC and bilateral anterior insulae (the Salience Network) was seen for commission errors. In contrast, timing errors were not associated with activation of this network but did activate a bilateral network that included the right ventral attentional system. Common activation for both error types occurred within the pars operculari and angular gyri. These results show that the dACC does not respond to all behaviorally salient errors. Instead, the error-processing system is multifaceted, and control can be triggered independently of the dACC when feedback is unexpected.

Introduction

The ability to adapt behavior to new demands is a central feature of cognitive control. Errors frequently result in behavioral changes that allow individuals to adapt to new situations. For example, slowing responses after an error is a common strategy that generally makes future performance more accurate in speeded reaction time tasks (Rabbitt 1966). Previous work has focused on the contribution of the dorsal anterior cingulate cortex (dACC) to behavioral adaptation. This region is activated soon after errors of many types and appears to be a key part of the neural network mediating error responses (Dehaene et al. 1994; Ridderinkhof et al. 2004; Debener et al. 2005; Sharp et al. 2006). Electrophysiological work has revealed an error-related negativity (ERN). This appears likely to be generated by the dACC (Dehaene et al. 1994; Debener et al. 2005), although recent studies suggest that error-related responses are also produced by other cortical regions (Eichele et al. 2008; Agam et al. 2011).The dACC is tightly linked both structurally and functionally to the anterior insulae (Dosenbach et al. 2006; Seeley et al. 2007). Together these regions have been termed the Salience Network, which is thought to act as an interface between limbic and cognitive aspects of behavioral control (Dehaene et al. 1994; Critchley et al. 2004; Seeley et al. 2007). However, the error-processing system is likely to be multifaceted and the precise contribution of the dACC to error processing remains controversial (Jessup et al. 2010).

Reinforcement learning theory provides an important framework within which to understand error processing. One influential model proposes that the dACC indexes prediction error signals generated by midbrain dopaminergic cells in response to a mismatch between expected and actual outcomes (Holroyd and Coles 2002, 2008; Holroyd et al. 2004). In the initial formulation of this model, the dACC was thought to act as a generic error monitor, responding to all types of errors and signaling the need for a change in behavior. In favor of this general role, activation of the dACC had been observed in response to errors generated either internally, as a result of a mismatch between expected and actual outcomes, or externally, as the result of explicit feedback (Holroyd et al. 2004). However, more recent electrophysiological work suggests that the dACC is sensitive to external feedback only when it is perceived as being “learnable” or directly related to the subject's own behavior (Holroyd et al. 2009).

Neuropsychological studies provide converging evidence that the dACC is required for certain types of error processing (Swick and Turken 2002; di Pellegrino et al. 2007). However, some types of behavioral adaptation have been shown to occur after damage to the dACC, suggesting that involvement of the region is not always required to initiate cognitive control (Fellows and Farah 2005; Kennerley et al. 2006; Modirrousta and Fellows 2008). Parts of the lateral temporoparietal lobes and inferior frontal cortex have also been shown to respond to behaviorally salient events, particularly unexpected events, and have been labeled the ventral attentional network (Corbetta et al. 2002). This network provides an alternative route by which external information might engage cognitive control potentially independent of the dACC.

Here, we use functional magnetic resonance imaging (fMRI) to investigate 2 distinct types of behaviorally salient errors. A version of the Simon task designed to generate a large number of errors of different types was used. Commission errors involved responding with an inaccurate key press, whereas timing errors involved responding after an externally imposed time limit. During training, subjects were instructed that both incorrect and late responses were to be considered errors on the task, and this was reiterated during performance of the task in the scanner. Commission errors were internally recognized and involved no explicit feedback. In contrast, timing errors were difficult to monitor internally and were explicitly signaled on slow trials. Comparison of the neural response to these errors measured using fMRI allowed an analysis of the error-processing network associated with these distinct error types.

We specifically tested the hypothesis that the dACC generically signals the need for behavioral adaptation, by comparing neural responses with internally generated commission errors and to externally signaled timing errors on the same task. This work builds on previous studies of error monitoring, but importantly employs a task in which reward is not present. We reasoned that if the dACC operates as a generic error monitor, one would expect similar activation across distinct error types. However, if dACC activation reflects a subset of error processing, for example, only responding to errors perceived as emotionally salient or where it is possible to generate a prediction error signal, differential activation may be present in the dACC despite adaptive behavioral change being triggered.

Materials and Methods

Participants

Thirty-five subjects performed the main Simon Task paradigm (17 males, mean age 30.6 ± 8.6 years). A further 15 subjects performed a control variant of the Simon Task designed to further investigate the effect of explicit feedback (4 males, mean age 29.4 ± 6.9 years). Subjects gave written consent. The experiment was approved by the Hammersmith and Queen Charlotte's, and Chelsea Research ethics committee.

Simon Task Procedure

Cognitive control was investigated during performance of the Simon task. The Simon task is a stimulus/response compatibility task that uses incongruency between the salient and nonsalient features of a stimulus to generate response conflict (Simon 1969; Simon and Small 1969). Building on previous electrophysiological work (Christ et al. 2000), we used a version of the Simon task designed to produce large numbers of errors of both commission and timing. Subjects were presented with a colored cue to the right or left of a fixation cross (Fig. 1). Cue color determined the direction of the required response: red signified a right-hand response and blue a left-hand response. Spatial location and cue direction were either congruent or incongruent with respect to each other. In the incongruent condition, the prepotent response—to respond in the direction of the spatial location of the cue rather than the direction signaled by the color—must be inhibited. Errors of commission occur when the subject's direction of response is not that signaled by the color. In the main part of the experiment, explicit feedback about errors or commission was not provided. Hence, any neural response to these errors was internally generated. Errors of timing were generated by explicit feedback given when a subject responded outside the time limit. Feedback was presented in the form of the words “Speed up” displayed on the screen accompanied by a 400-Hz auditory tone, which emphasized the error. Visual and auditory feedback lasted 500 ms. Subjects were told at the start to perform the task as accurately and quickly as possible and were aware that slowing down would result in error signal about the timing of their responses. This was emphasized during training and also between runs of the paradigm performed in the scanner.

Figure 1.

Schematic of the Simon task paradigm. Subjects responded with a right or left finger press for red and blue cues, respectively. The precue, an empty square, appeared to either the left or right side of the fixation cross for 200 ms before filling in with the cue color (either red or blue). (a) On congruent trials, the spatial location of the cue corresponded to the side of the appropriate response press. (b) On incongruent trials, the spatial location conflicted with the side of the response.

Figure 1.

Schematic of the Simon task paradigm. Subjects responded with a right or left finger press for red and blue cues, respectively. The precue, an empty square, appeared to either the left or right side of the fixation cross for 200 ms before filling in with the cue color (either red or blue). (a) On congruent trials, the spatial location of the cue corresponded to the side of the appropriate response press. (b) On incongruent trials, the spatial location conflicted with the side of the response.

An adaptive staircase procedure was used to vary the response delay necessary to trigger feedback with the goal of producing errors of commission within a target range. At the start of the experiment, feedback was triggered if a subject's response was more than 500 ms after the presentation of the color stimulus. After the first 15 trials, the rate of commission errors was calculated after each trial, and the time limit was adjusted if this rate fell outside the target range. For congruent trials, a target commission error range of 8–12% was used. The time limit was increased by 50 ms if the error rate was high and decreased by the same amount if it was too low. Adaptation was performed separately for incongruent trials, aiming for a range of 17–25% commission errors. The time limit was adapted within a range of 400–1000 ms. For subsequent runs, the starting time limit was carried over from the previous run.

Two thirds of trials were “congruent” and one-third “incongruent”. All subjects performed 6 runs of 120 pseudorandomly ordered trials with interstimulus intervals of 2.25 s. To increase task difficulty, we introduced a precue in the form of empty rectangle that filled in after 200 ms with the color that indicated response direction. The precue increased the interference effect produced by a spatially incongruent color cue. The relative timing of the precue and color cues was designed to generate the maximum number of errors of commission (Christ et al. 2000). All subjects performed 120 trials (80 congruent and 40 incongruent) as training prior to scanning.

Behavioral Analysis

Mean error rates and reaction times were calculated separately for congruent and incongruent trials. In addition, behavior was investigated on trials surrounding errors. This allowed the effects of cognitive control to be observed as posterror slowing after commission errors and postfeedback speeding after timing errors. The accuracies of these trials were also assessed and compared with subject's accuracy across the entire run.

The trials before an error (N−1), the error trials (N), and up to 3 trials following an error (N+1, N+2 and N+3) were investigated. Reaction times were compared to a baseline performance on trials of that type, (i.e., congruent or incongruent), calculated from the mean of the last ten stable correct trials of a particular type. Stable correct trials were defined as correct timely trials that had also been preceded by a correct timely trial. Baseline performance was calculated this way firstly to avoid contamination from the effects of internal and external feedback on preceding trials, and also to control for slow fluctuations in subject's attention.

Image Acquisition

MRI data were obtained using a Philips (Best, The Netherlands) Intera 3.0 Tesla MRI scanner using Nova Dual gradients, a phased array head coil, and sensitivity encoding (SENSE) with an under sampling factor of 2. fMRI images were obtained using a T2*-weighted gradient–echo echoplanar imaging (EPI) sequence with whole-brain coverage (time repetition [TR]/time echo [TE] = 2000/30; 31 ascending slices with thickness 3.25 mm, gap 0.75 mm, voxel size 2.19 × 2.19 × 4 mm, flip angle 90°, field of view 280 × 220 × 123 mm, matrix 112 × 87). Quadratic shim gradients were used to correct for magnetic field inhomogeneities within the brain. The inferior part of the cerebellum was not included in our field of view. T1-weighted whole-brain structural images were also obtained in all subjects. Paradigms were programmed using Matlab Psychophysics toolbox (Psychtoolbox-3 www.psychtoolbox.org) and stimuli presented through an IFIS-SA system (In Vivo Corporation). Responses were recorded through a fiber optic response box (NordicNeurolab, Norway), interfaced with the stimulus presentation PC running Matlab. Sounds were presented using ear-defending headphones (MR Confon).

Whole-Brain fMRI Analysis

Data was analyzed with standard random effects general linear models using tools from the FSL library (FEAT version 5.98) (Smith et al. 2004). Image preprocessing involved realignment of EPI images to remove the effects of motion between scans, spatial smoothing using an 8-mm full-width half-maximum Gaussian kernel, prewhitening using FILM and temporal high-pass filtering using a cutoff frequency of 1/50 Hz to correct for baseline drifts in the signal. FMRIB's Linear Image Registration Tool was used to register EPI functional data sets into standard MNI space using the participant's individual high-resolution anatomical images. fMRI data were analyzed using voxelwise time-series analysis within the framework of the General Linear Model. To this end, a design matrix was generated with a synthetic hemodynamic response function and its first temporal derivative. For congruent and incongruent trials, a number of distinct event types were modeled: correct trials (appropriate response within the time limit), errors of commission (incorrect button press but within the time limit), errors of timing (correct button press but outside of the time limit), errors or both timing and commission, and confounded trials where timing or commission errors occurred directly after another timing or commission error. Commission and timing errors were contrasted with correct trials, as well as being directly contrasted with each other. Group effects analysis was carried out using FLAME (FMRIB's Local Analysis of Mixed Effects). Separate analyses were performed for congruent and incongruent trials. Errors of both commission and timing were insufficient to be analyzed. Final statistical images were thresholded using False Discovery Rate threshold of P < 0.05.

Region of Interest Analysis

An additional focused region of interest (ROI) analysis was performed using Featquery within FSL. A number of anatomically defined frontal lobe regions known to be involved in error processing were investigated. These consisted of left and right anterior cingulate cortices (ACC) and anterior insulae, as well as left and right pars operculari and pars triangulari within the inferior frontal gyri. Probabilistic masks were defined from the Harvard Cortical Atlas tool in FSL and were thresholded at 70%. The regions were all available in the atlas apart from the anterior insula. To generate the anterior insula mask, the insula atlas map was divided along the anterior–posterior dimension into 2 halves. For each axial slice, the image was divided at the midpoint along the anterior–posterior dimension. This produced anterior and posterior insula masks. Mean percentage signal change associated with contrasts of interest was calculated for all voxels falling within each ROI.

Time-Series Analysis

To investigate whether commission and timing errors were associated with different time courses, Perl Event-related Average Time-course Extraction was used to further characterize the neural responses within the dACC (www.jonaskaplan.com/peate/peate-tk.html). The mean time series from a 10-mm diameter spherical ROI within the dACC was extracted. To test whether the dACC responded to all types of error, we used an ROI centered on the peak activation for the contrast of errors of commission versus correct trials (x = 2, y = 32, z = 28). The response was examined for a period 2 s proceeding to 12 s following a subject's response.

Additional Investigation Controlling for the Presence of External Feedback

In the main version of the Simon task errors of timing were signaled by explicit “error” feedback, but errors of commission were not. To test whether this difference influenced dACC activation, we performed an additional study with a separate group of 14 subjects. Here, both types of error were accompanied by explicit feedback. The design of the experiment was the same as that described above except for the presence of audio–visual feedback for errors of commission. This took the form of the word “Wrong!” presented after an error of commission, accompanied by a 500 Hz tone. The timing and duration was the same as for error of timing described above.

Analysis of Strategic Differences in Task Performance: Comparison of High and Low Timing Error Groups

The number of timing errors was variable across individuals and runs. This is likely to reflect strategic differences in the perceived salience of timing feedback. During task performance, subjects could either focus on reducing errors of commission, reducing errors of timing, or they could adjust their speed-accuracy trade-off in a more complex way to achieve an optimum reduction in both error types. Therefore, the number of timing errors on a particular run is likely to reflect a subject's strategy on that run. To assess whether this variability affected neural activation within the cognitive control system, we analyzed the behavioral data from all subjects and compared runs with high numbers of timing errors to those with low numbers (defined by taking the upper and lower thirds of the distribution). To focus on the effects of variable strategy with respect to timing, we only included runs where the commission error rate fell approximately within the ranges that the paradigm was designed to produce (i.e., 3–17% congruent and 12–30% incongruent errors). This resulted in the potential inclusion of 169 runs (Fig. 8a). Three subjects had separate runs that would have been included in the High Timing error and others in Low Timing error groups. In these cases, the runs that were in the minority were removed from the analysis (i.e., if a subject had 3 runs included in the High and 1 run in the Low Timing error group then the run in the Low Timing error group would be excluded from the analysis). We did this to avoid mixing between-subject effects and within-subject effects. As a result, 3 single runs were removed from the analysis. The Low Timing error group consisted of 53 runs spread across 23 subjects (0.83–6.7% timing errors) and the High Timing group of 54 runs across 22 subjects (13.3–48.33% timing errors). Comparison of errors of timing against correct trials was made for the 2 groups using a standard mixed level analysis, and the resulting contrasts were then directly compared.

Results

Behavior on the Simon Task

Behavioral performance was in keeping with previous studies (Christ et al. 2000). Subjects performing the main version of the Simon Task produced a total of 5992 errors. Relatively large numbers of errors of both commission (11.7% of all trials) and timing (12.1% of all trials) were committed. This was expected and was probably due to the substantial increase in overall task difficulty that the timing feedback adds, coupled with the fact that performances on congruent and incongruent trials are not independent. A significantly larger number of errors of commission were produced on incongruent than on congruent trials (t = 3.65, degrees of freedom [df] = 34, P = 0.001). Furthermore, as a percentage of the number of congruent and incongruent trials, commission errors were substantially more common on incongruent trials (t = 9.81, df 34, P < 0.0005) (Fig. 2). We performed an additional analysis to help clarify reason for the high error rate on congruent trials. We assessed how the congruency of the trial immediately before a trial affected subjects' accuracies. Correct congruent trials were more likely to have been preceded by another congruent trial than incorrect congruent trials (68.5% vs. 56.3%, df = 34, t = 25.29, P < 0.0005). Similarly, correct incongruent trials were more likely to have been preceded by another incongruent trial than incorrect incongruent trials (37.5% vs. 34%, df 34, t = 5.76, P < 0.0005). Although the effect size was differed between congruent and incongruent trials, indicating that the effect of previous trials was less for incongruent trials.

Figure 2.

Simon task behavioral performance. (a) Rate of commission errors shown for congruent and incongruent responses. (b) Average reaction times (millisecond) for correct trials and commission errors separated based upon the trial congruency. (c) Rate of timing errors for congruent and incongruent trials. (d) Average reaction times (millisecond) for congruent and incongruent timing errors.

Figure 2.

Simon task behavioral performance. (a) Rate of commission errors shown for congruent and incongruent responses. (b) Average reaction times (millisecond) for correct trials and commission errors separated based upon the trial congruency. (c) Rate of timing errors for congruent and incongruent trials. (d) Average reaction times (millisecond) for congruent and incongruent timing errors.

Most timing errors involved a button press with the correct hand (87.5 ± 1.7% of all timing error trials). Significantly more timing errors were produced on congruent than incongruent trials (t = 5.89, df = 34, P < 0.0005). However, as congruent trials were more frequent, the percentage of timing errors was similar for the 2 trial types (t = 1.26, df = 34, P = 0.216) (Fig. 2). A small number of timing errors were also errors of commission, that is, a subject responded late and with the wrong hand. The frequency of this type of error was similar for both congruent and incongruent trials (1.5 ± 0.28% and 1.2 ± 0.24% of congruent and incongruent trials, respectively). These were not analyzed further because of their low number.

Response speed had a major effect on the accuracy of incongruent trials (Fig. 2). A 2 × 2 ANOVA showed a significant interaction between trial type and accuracy (F = 127.2, P < 0.0005). This was due to subjects responding slower on correct incongruent trials than to correct congruent trials (t = 10.6, df = 34, P < 0.0001), whereas subjects responded faster to incorrect incongruent trials than incorrect congruent trials (t = 2.89, df = 34, P < 0.0001). This is consistent with a speed-accuracy trade-off operating for incongruent trials, where fast responses are more likely to be wrong because of a premature response to a spatially contradictory cue. Timing errors were, by definition, significantly slower than correct trials (t = 10.49, df = 34, P < 0.0005), and incongruent timing errors were slightly slower than congruent timing errors (t = 2.71, df = 34, P = 0.01) (Fig. 2).

Errors of Commission and Timing Both Produce Adaptive Changes in Behavior

We investigated adaptive behavior by studying the slowing of responses following errors of commission and the speeding of responses following feedback after timing errors (Fig. 3). In the first case, the adaptive “signal” is internally generated, as there is no explicit feedback. In the second, as we only included correct button presses in the analysis, the adaptive “signal” arises from the external feedback given after a late response.

Figure 3.

Adaptive changes following errors of commission and timing. (a) Posterror slowing on incongruent trials. Reaction time for trials around a commission error (C) relative to baseline performance. (b) Postfeedback speeding following errors of timing. Reaction times for trials around an error of timing (T) Relative to baseline performance. * indicates trials that significantly differed from baseline performance for the preceding 10 correct trials of a similar type. Columns represent the accuracy of the trial around either commission errors (a) or timing error (b). Columns are hashed if the trials accuracy significantly differs from subjects' normal performance. Error bars represent the standard error of the mean.

Figure 3.

Adaptive changes following errors of commission and timing. (a) Posterror slowing on incongruent trials. Reaction time for trials around a commission error (C) relative to baseline performance. (b) Postfeedback speeding following errors of timing. Reaction times for trials around an error of timing (T) Relative to baseline performance. * indicates trials that significantly differed from baseline performance for the preceding 10 correct trials of a similar type. Columns represent the accuracy of the trial around either commission errors (a) or timing error (b). Columns are hashed if the trials accuracy significantly differs from subjects' normal performance. Error bars represent the standard error of the mean.

Both types of feedback were behaviorally salient. For errors of commission, the behavioral effect was only present on incongruent trails (Fig. 3). Incongruent commission errors (C) were relatively fast compared with baseline (t = 15.66, df = 34, P < 0.0005) and posterror slowing was observed on the next trial (C + 1) (t = 8.11, df = 34, P < 0.0005). In contrast, errors on congruent trails were not abnormally fast and were not followed by posterror slowing. Fast responses on incongruent trials are very likely to result in errors; therefore, slowing response speed on the next trial is an effective strategy to improve performance. In contrast, errors on congruent trials are much less influenced by this type of speed-accuracy trade-off, so posterror slowing is far less adaptive in this context.

As expected, external feedback after a timing error resulted in speeding of subsequent responses (Fig. 3). In contrast to errors of commission, both trial types were associated with postfeedback speeding. Errors of timing (T) were by definition slower than average ([t = 24.43, df = 34, P < 0.0005] for congruent timing errors and [t = 10.37, df = 34, P < 0.0005] for incongruent timing errors). In the case of congruent timing errors, which were far more numerous, all 3 subsequent trials were faster than baseline (“T + 1,” t = −3.47, P = 0.001, “T + 2,” t = −6.91, P < 0.0005, “T + 3,” t = −4.04, P < 0.0005), whereas for incongruent timing errors, this was true only for the second and third trials after the timing error (“T + 2,” t = −3.79, P = 0.001, “T + 3,” t = −2.68, P = 0.012). The postfeedback speeding is not simply a product of responses getting quicker over the course of a run. In fact, the opposite trend is observed, with responses generally slowing, which is demonstrated by the first tertile being significantly quicker than the last (t = 9.91, df = 33, P < 0.0005). In addition to being faster, the trial immediately following a timing error (T + 1) is also more accurate than a subjects' average accuracy over the run (t = 4.02, df = 34, P < 0.0005), again suggesting that subjects engage greater cognitive control in response to timing feedback.

Neuroimaging

Errors of Commission Activate the dACC

The network of brain regions activated by errors of commission (Fig. 4) was consistent with previous work (Garavan et al. 2002; Ullsperger and von Cramon 2003; Hester et al. 2004, 2005). Extensive activation was seen in the dACC during errors of commission compared with correct trials. In addition, peaks of activation were observed in the superior frontal gyrus, bilateral anterior insulae, and pars operculari, as well as in the frontal poles and supramarginal and angular gyri (Table 1). There was also activation of subcortical structures, including the brainstem and bilateral thalami. Extensive activation in the ACC and insulae fell within what has been termed the Salience Network (Seeley et al. 2007). Similar patterns of activation were observed for errors of commission on congruent and incongruent trials. A direct contrast of errors on both trial types showed no significant differences in brain activation.

Table 1

Cluster analysis for contrasts of different error types

Anatomical region False discovery rate (q value) MNI coordinates 
X Y Z 
Commission errors > correct trials 
    Anterior cingulate gyrus <0.05 32 28 
    Left frontal pole <0.05 −30 50 20 
    Right frontal pole <0.05 28 52 22 
    Left insula cortex <0.05 −38 16 −14 
    Right insula cortex <0.05 58 14 −2 
    Left supramarginal gyrus <0.05 −62 −46 28 
    Right supramarginal gyrus <0.05 64 −48 28 
    Left thamamus <0.05 −8 −14 
    Right thalamus <0.05 −16 
    Brain stem <0.05 −28 −10 
Timing errors > correct trials 
    Left inferior frontal gyrus (pars triangularis) <0.05 −52 23 16 
    Right inferior frontal gyrus (pars triangularis) <0.05 60 −44 −6 
    Frontal pole (midline) <0.05 58 28 
    Precuneus <0.05 −48 12 
    Left caudate <0.05 −8 11 
    Right caudate <0.05 
    Left lateral occipital cortex <0.05 −30 −92 −20 
    Right lateral occipital cortex <0.05 28 −92 −26 
    Left middle temporal gyrus <0.05 −64 −44 −6 
    Right middle temporal gyrus <0.05 57 −18 −20 
Commission errors > timing errors 
    Anterior cingulate gyrus <0.05 32 28 
    Left insula cortex <0.05 −38 12 −8 
    Right insula cortex <0.05 44 10 −8 
    Left frontal pole <0.05 −32 48 16 
    Right frontal pole <0.05 28 54 26 
    Left supramarginal gyrus <0.05 −52 −40 38 
    Right supramarginal gyrus <0.05 56 −46 34 
    Left thamamus <0.05 12 −13 
    Right thalamus <0.05 −3 −11 −2 
    Brain stem <0.05 −22 −8 
Timing errors > commission errors 
    Left inferior frontal gyrus (pars triangularis) <0.05 −56 24 10 
    Right inferior frontal gyrus (pars triangularis) <0.05 58 30 14 
    Left middle temporal gyrus <0.05 −60 −20 −12 
    Right middle temporal gyrus <0.05 64 −15 12 
    Precuneus <0.05 −53 32 
    Left superior frontal gyrus <0.05 −7 56 34 
    Left lateral occipital lobe <0.05 −30 −92 −22 
    Right lateral occipital lobe <0.05 30 −92 −16 
Areas common to commission and timing errors > correct trials 
    Left inferior frontal gyrus (pars opercularis)  −47 18 19 
    Right inferior frontal gyrus (pars opercularis)  50 15 14 
    Left supramarginal gyrus  −63 −48 27 
    Right supramarginal gyrus  62 −44 20 
    Brain stem  −29 −10 
    Left thamamus  −10 −1 
    Right thalamus  −1 
    Left temporal pole  −40 10 −25 
    Right temporal pole  41 10 −25 
Anatomical region False discovery rate (q value) MNI coordinates 
X Y Z 
Commission errors > correct trials 
    Anterior cingulate gyrus <0.05 32 28 
    Left frontal pole <0.05 −30 50 20 
    Right frontal pole <0.05 28 52 22 
    Left insula cortex <0.05 −38 16 −14 
    Right insula cortex <0.05 58 14 −2 
    Left supramarginal gyrus <0.05 −62 −46 28 
    Right supramarginal gyrus <0.05 64 −48 28 
    Left thamamus <0.05 −8 −14 
    Right thalamus <0.05 −16 
    Brain stem <0.05 −28 −10 
Timing errors > correct trials 
    Left inferior frontal gyrus (pars triangularis) <0.05 −52 23 16 
    Right inferior frontal gyrus (pars triangularis) <0.05 60 −44 −6 
    Frontal pole (midline) <0.05 58 28 
    Precuneus <0.05 −48 12 
    Left caudate <0.05 −8 11 
    Right caudate <0.05 
    Left lateral occipital cortex <0.05 −30 −92 −20 
    Right lateral occipital cortex <0.05 28 −92 −26 
    Left middle temporal gyrus <0.05 −64 −44 −6 
    Right middle temporal gyrus <0.05 57 −18 −20 
Commission errors > timing errors 
    Anterior cingulate gyrus <0.05 32 28 
    Left insula cortex <0.05 −38 12 −8 
    Right insula cortex <0.05 44 10 −8 
    Left frontal pole <0.05 −32 48 16 
    Right frontal pole <0.05 28 54 26 
    Left supramarginal gyrus <0.05 −52 −40 38 
    Right supramarginal gyrus <0.05 56 −46 34 
    Left thamamus <0.05 12 −13 
    Right thalamus <0.05 −3 −11 −2 
    Brain stem <0.05 −22 −8 
Timing errors > commission errors 
    Left inferior frontal gyrus (pars triangularis) <0.05 −56 24 10 
    Right inferior frontal gyrus (pars triangularis) <0.05 58 30 14 
    Left middle temporal gyrus <0.05 −60 −20 −12 
    Right middle temporal gyrus <0.05 64 −15 12 
    Precuneus <0.05 −53 32 
    Left superior frontal gyrus <0.05 −7 56 34 
    Left lateral occipital lobe <0.05 −30 −92 −22 
    Right lateral occipital lobe <0.05 30 −92 −16 
Areas common to commission and timing errors > correct trials 
    Left inferior frontal gyrus (pars opercularis)  −47 18 19 
    Right inferior frontal gyrus (pars opercularis)  50 15 14 
    Left supramarginal gyrus  −63 −48 27 
    Right supramarginal gyrus  62 −44 20 
    Brain stem  −29 −10 
    Left thamamus  −10 −1 
    Right thalamus  −1 
    Left temporal pole  −40 10 −25 
    Right temporal pole  41 10 −25 
Figure 4.

Errors of commission and timing. (a) Areas of significant brain activation associated with errors of commission compared with correct trials (red–yellow). (b) Areas of significant brain activation associated with errors of timing compared with correct trials (light—dark blue). Results are superimposed on the MNI 152 T1 1-mm brain template.

Figure 4.

Errors of commission and timing. (a) Areas of significant brain activation associated with errors of commission compared with correct trials (red–yellow). (b) Areas of significant brain activation associated with errors of timing compared with correct trials (light—dark blue). Results are superimposed on the MNI 152 T1 1-mm brain template.

Errors of Timing Activate Lateral Prefrontal and Superior Frontal Regions but Not the dACC

Compared with errors of commission, errors of timing were associated with activation in a distinct and only partially overlapping network. Relative to correct trials, errors of timing were associated with activation in the pars opercularis, which extended forward into the pars triangularis bilaterally, and also in the anterior part of the medial superior frontal gyrus (Fig. 4). More posteriorly, activation was seen in superior temporal regions extending into the inferior parietal lobe bilaterally. This included activation of the angular and supramarginal gyri bilaterally. Activation was not observed within the ACC and only marginally spread into the anterior insulae from the overlying inferior frontal gyrus. Significant subcortical activation was seen in the brain stem and bilateral caudate nuclei.

A direct contrast of commission versus timing errors demonstrated that activation was significantly higher in the Salience Network for errors of commission. Peaks for this contrast were seen in the ACC and the bilateral insulae, as well as in the frontal poles and angular gyri (Fig. 5). The reverse contrast showed greater activation bilaterally within the pars triangulari for timing errors, as well as increased activation within the posterior cingulate cortex (Fig. 5).

Figure 5.

Common and distinct activation for errors of commission and timing. (a) A conjunction analysis showing common activation for both types of error (green). The direct contrast of error types shows brain regions more activated by errors of commission (red) or timing (blue). Results are superimposed on the MNI 152 T1 1-mm brain template. (b) ROI analysis using anatomically derived masks. Significant differences from baseline are shown by *. Right (R), left (L), anterior (ant), and anterior cingulate cortex (ACC).

Figure 5.

Common and distinct activation for errors of commission and timing. (a) A conjunction analysis showing common activation for both types of error (green). The direct contrast of error types shows brain regions more activated by errors of commission (red) or timing (blue). Results are superimposed on the MNI 152 T1 1-mm brain template. (b) ROI analysis using anatomically derived masks. Significant differences from baseline are shown by *. Right (R), left (L), anterior (ant), and anterior cingulate cortex (ACC).

Common Activation for Errors of Commission and Timing Is Seen within Pars Operculari

A conjunction analysis demonstrated brain regions commonly activated by errors of commission and timing (Fig. 5). Common activation was observed in the pars operculari bilaterally, as well as within the inferior parietal lobes, and the anterior part of the medial superior frontal gyrus, the anterior thalami and temporal poles bilaterally. A ROI analysis using anatomically defined frontal masks confirmed the presence of distinct patterns of activation for the 2 types of error, as well as common activation within the pars operculari (Fig. 5). Errors of commission resulted in significant activation of the ACC and bilateral anterior insulae (right ACC [t = 4.96, df = 34, P < 0.0005]; left ACC [t = 4.37, df = 34, P < 0.0005]; right anterior insula [t = 4.04, df = 34, P < 0.0005]; and left ACC [t = 6.04, df = 34, P < 0.0005]), whereas timing errors were associated with activation of the pars triangulari bilaterally (right t = 6.56, df = 34, P < 0.0005 and left t = 4.17, df = 34, P < 0.0005). There was no significant activation of the ACC in timing errors compared with correct trials. Common activation relative to baseline was seen in the pars opercularis bilaterally for both error types. Timing errors showed significantly greater activation in the left (t = 5.09, df = 34, P < 0.0005) and right pars opercularis (t = 3.99, df = 34, P < 0.0005) compared with baseline; as did commission errors (left pars opercularis [t = 4.31, df = 34, P < 0.0005] and right pars opercularis [t = 3.75, df = 34, P = 0.001]).

Delayed Activation of the dACC Is not Present after Errors of Timing

Although errors of timing occurred slightly later than errors of commission (152 ms average difference), this did not account for the differences we observed in dACC activation. This small difference in timing is highly unlikely to produce difference in activation between the 2 types of error, and small variations in the hemodynamic response function were modeled using temporal derivatives. In addition, we performed a time-course analysis to confirm that delayed dACC activation was not present (Fig. 6). dACC activation was observed for errors of commission, which peaked around 6 s after the response was made. Following a commission error, activation of the dACC was significantly greater than following either correct trials or timing errors from 3 to 9 s after the response. In contrast, timing errors caused no statistically significant signal change compared with correct trials at any time point.

Figure 6.

Perl Event-related Average Time-course Extraction (PEATE) analysis. Graph of the averaged BOLD signal change across subjects within the ACC for the 12 s following correct trials, commission error trials, and timing error trials. ACC activity was sampled from a region around the peak of activation associated with errors of commission versus correct trials. Error bars represent the standard error of the mean.

Figure 6.

Perl Event-related Average Time-course Extraction (PEATE) analysis. Graph of the averaged BOLD signal change across subjects within the ACC for the 12 s following correct trials, commission error trials, and timing error trials. ACC activity was sampled from a region around the peak of activation associated with errors of commission versus correct trials. Error bars represent the standard error of the mean.

Controlling for the Presence of External Feedback on Errors of Timing

We also investigated whether the absence of external feedback after errors of commission could have influenced the difference in brain activation that we observed within the dACC. In a new version of the task, errors of commission were also signaled by explicit external feedback in a similar way to timing errors. Overall, the behavioral results for the control experiment were similar to the main experiment. Comparing the 2 experiments, there were no significant differences in overall reaction times (351 and 340 ms, P = 0.149) or number of late responses per run (14.5 and 13.2, P = 0.684). Similar behavioral adaptation was also observed after errors of commission and timing (Fig. 7). There were slightly fewer overall errors in the control experiment (11.7 ± 0.4% vs. 7.8 ± 0.9%), which was due to both fewer congruent errors (5.3 ± 0.4% vs. 3.1 ± 0.4%) and incongruent errors (6.4 ± 0.4% vs. 4.7 ± 0.7%). The neuroimaging results were also similar. The contrasts of commission and timing errors with timely correct responses showed similar activation to the main version of the task (Fig. 7). The direct contrast of commission and timing errors again confirmed increased dACC activation for commission errors using a small volume correction with a 10 mm diameter sphere centered around the peak of the activation difference between error types in the main analysis (X = 2, Y = 32, Z = 26).

Figure 7.

Errors of commission and timing in small group of subjects using an alternative Simon Task paradigm with additional external feedback after commission errors. (a) Areas of significant brain activation associated with errors of commission compared with correct trials (red–yellow). (b) Areas of significant brain activation associated with errors of timing compared with correct trials (light—dark blue). Results are superimposed on the MNI 152 T1 1-mm brain template.

Figure 7.

Errors of commission and timing in small group of subjects using an alternative Simon Task paradigm with additional external feedback after commission errors. (a) Areas of significant brain activation associated with errors of commission compared with correct trials (red–yellow). (b) Areas of significant brain activation associated with errors of timing compared with correct trials (light—dark blue). Results are superimposed on the MNI 152 T1 1-mm brain template.

Strategic Differences in Task Performance: Comparison of High and Low Timing Error Groups

Runs with low and high numbers of timing errors were compared, as we reasoned that they involve different performance strategies. Low numbers of timing errors suggest that subjects performed the task as requested, maintaining a generally fast response speed. In contrast, high numbers suggest that subjects paid less attention to the feedback, which is likely to be the result of a strategic decision to optimize accuracy over timing. Despite this strategic difference, timing feedback still had the effect of changing behavior in both groups in the immediate period following an error (Fig. 8). An ANOVA was performed using group assignment as one factor (High or Low Timing error) and time relative to the timing error trial (T) as a second factor with 5 levels (T − 1, T, T + 1, T + 2, and T + 3). There was no interaction between group type and time, demonstrating the similarity of the short-term response to timing feedback. In addition, there was no group difference in the average reaction times for correct trials (338 ± 5 ms for the High Timing group and 325 ± 5 ms for the Low Timing group), indicating that nonspecific differences in factors such as arousal level were not present.

Figure 8.

Strategy analysis. (a) Graph illustrating the timing and error rates of runs used in the analysis. Timing error rates for separate runs plotted against the commission error rate for that run (averaged for congruent and incongruent errors). Runs were used from the upper or lower third of the distribution. (b) Graph of reaction times on errors of timing (T) and on trials around this (T − 1 and T + 1 to 3). Behavioral data is plotted separately for groups with high and low timing errors. c) Brain regions activated by the contrast of timing errors compared with correct trials in the Low Timing error group. (d) Brain regions activated by the contrast of timing errors compared with correct trials in the High Timing error group. Results are superimposed on the MNI 152 T1 1-mm brain template.

Figure 8.

Strategy analysis. (a) Graph illustrating the timing and error rates of runs used in the analysis. Timing error rates for separate runs plotted against the commission error rate for that run (averaged for congruent and incongruent errors). Runs were used from the upper or lower third of the distribution. (b) Graph of reaction times on errors of timing (T) and on trials around this (T − 1 and T + 1 to 3). Behavioral data is plotted separately for groups with high and low timing errors. c) Brain regions activated by the contrast of timing errors compared with correct trials in the Low Timing error group. (d) Brain regions activated by the contrast of timing errors compared with correct trials in the High Timing error group. Results are superimposed on the MNI 152 T1 1-mm brain template.

The contrast of timing error and correct trials in both groups showed a pattern of activation similar to that seen for the overall effect of timing errors (Fig. 8). Activation was observed in the right superior frontal gyrus, bilateral pars triangulari, operculari, and the supramarginal gyri. No regions showed significant differences in activation when directly comparing the 2 groups. In addition, neither group showed any significant activation of the dACC.

Discussion

We investigated the neural response to errors using event-related fMRI and provide evidence for a multifaceted error-processing system. A modified version of the Simon Task was used to generate large numbers of errors of both commission and timing. This allowed detailed analysis of the patterns of brain activity associated with these errors. While being quite different in nature, both types of error were associated with behavioral adaptation on subsequent trials, suggesting the engagement of cognitive control. As expected, commission errors activated the dACC and other parts of the Salience Network, and led to an adaptive behavioral change in the form of posterror slowing after incongruent trials. In contrast, timing errors were not associated with increased dACC activation, despite being explicitly signaled by feedback and leading to behavioral change. Strategic differences in task performance did not explain this result. The lack of dACC activation associated with timing errors demonstrates that cognitive control processes that affect behavior can be triggered by errors without an increase in activation of the dACC. In contrast, timing errors were associated with extensive activation elsewhere in the brain, including within parts of the ventral attentional system, with specific activation observed in the pars triangularis. This provides an alternative anatomical route by which behavioral control may be engaged, which is potentially independent of the dACC.

Cognitive control links performance monitoring to subsequent task performance, and the dACC has been placed at its heart (Badre and Wagner 2004; Botvinick et al. 2004; Kerns et al. 2004; Ridderinkhof et al. 2004; Rushworth et al. 2004; di Pellegrino et al. 2007). Our results show that involvement of this region depends on the nature of the error. In keeping with previous work, the region was extensively activated during commission errors (Falkenstein et al. 2000; Ridderinkhof et al. 2004; Debener et al. 2005). In contrast, timing errors were not associated with increased dACC activity. During performance of the task, it was repeatedly emphasized that pressing the wrong button and responding late should both be considered to be an error, and evidence that subjects considered this to be the case is provided by the adaptive changes in behavior observed. Commission errors on incongruent trials produced typical posterror slowing, which is often taken to indicate that cognitive control has been engaged. In contrast, timing errors were followed by postfeedback speeding. On this particular task, the speeding of responses after an error is distinct from the general trend of responses to slow as each run progresses and is also associated with a better than average accuracy for the subsequent response. Hence, this adaptive change in response speed is also indicative of increased cognitive control.

Models of cognitive control frequently propose that the dACC is involved in signaling a need for increased control, which leads to a change in behavior (Ridderinkhof et al. 2004). The amount of dACC activation can be related to the magnitude of posterror slowing (Gehring et al. 1993; Kerns et al. 2004), although this is not always observed (Li et al. 2008). In keeping with a role in monitoring but not implementing behavioral change, our results show that high dACC activation is not necessarily associated with behavioral adaptation, as congruent commission errors strongly activated the region but were not followed by posterror slowing (Fig. 3). A key question is what does the dACC monitor and over what time scale does it operate? A feedback-related negativity (fERN) has been demonstrated with similarities to the ERN (Miltner et al. 1997). Similarly, certain types of feedback are associated with dACC activation demonstrated with fMRI (Holroyd and Coles 2002; Holroyd et al. 2004; Ullsperger et al. 2007), although differences in reward may influence dACC activity (Bush et al. 2002). However, our results are not compatible with the proposal that the dACC signals all behaviorally salient errors (Holroyd and Coles 2002; Holroyd et al. 2004). We observed no increase in dACC activation after timing errors, despite explicit external feedback that changed behavior. Work in nonhuman primates shows the preservation of rapid responses to errors but impaired strategic learning in animals with ACC lesions (Kennerley et al. 2006), also suggesting that rapid cognitive control can be engaged without dACC involvement.

Subtle changes in timing are often difficult to perceive (Allan 1978; Miltner et al. 1997; Luu et al. 2000; Grondin 2010), and an important factor in explaining our results is likely to be the unpredictability of our external feedback. A number of factors make it highly unlikely that subjects were able to accurately judge the timing of their responses. First, the cutoff for timing feedback varied from trial to trial; second, separate timing limits were used for congruent and incongruent trials; and finally, the trial types were randomly intermixed, making it very difficult to learn the timing rules for the 2 trial types. The magnitude of the ERN depends on whether feedback is predictable (Hajcak et al. 2005; Holroyd et al. 2009) providing evidence that the ACC is involved in cognitive control only when action-outcome contingencies can be learned (Holroyd et al. 2009). This has lead to a reformulation of the reinforcement learning theory in which the ACC is involved in cognitive control only when monitoring involves production of an internally generated prediction error. Our results are in keeping with this distinction, as unpredictable feedback was not associated with dACC response.

Variation in the predictability of feedback could explain differences between our work and previous electrophysiological studies in this area (Miltner et al. 1997; Luu et al. 2000). For example, Miltner and colleagues show the presence of an fERN when subjects were provided with feedback 600 ms after an error on a time estimation task. Although the criterion for success was adaptively varied, there remained a relatively predictable relationship between performance and feedback. Hence, the fERN could be interpreted in the context of the generation of a prediction error. In a further study, Yuu and colleagues studied a cognitively demanding flanker task with a reaction time limit that triggered negative feedback when breached (Luu et al. 2000). This studies design was similar in many ways to ours, but in this case, timing errors were associated with the generation of an ERN. However, in contrast to our study, monetary reward was used to increase motivation, which confounds interpretation of the findings. In addition, the ERN amplitude was shown to increase as responses became later. This is compatible with prediction errors increasing in magnitude as the presence of a timing error became more predictable.

A further potential explanation for the absence of dACC activity during timing errors is low levels of response conflict on these trials. The dACC is activated in situations where response conflict is high, which arises when 2 or more response processes are simultaneously activated (Botvinick et al. 2001, 2004) Conflict can occur before a response, for example, on incongruent trials when there is conflict between the spatial location and the color response cue, or after a response, for example, as a result of attempts to immediately correct an erroneous response (Yeung et al. 2004). Timing errors are likely to have low levels of both pre- and postresponse conflict, because they are predominantly congruent and cannot be corrected with an alternative response, potentially explaining the absence of increased dACC activity. In contrast, incongruent commission errors are associated with preresponse conflict because of conflicting spatial and color information, and both congruent and incongruent errors of commission are associated with postresponse conflict, which may actually be greater for congruent errors, potentially explaining the similar levels of dACC activation on these 2 trial types (Yeung et al. 2004).

Our results clearly show distinct patterns of activation within the lateral frontal region. The dACC is tightly linked both structurally and functionally to the anterior insulae, forming what has been termed the Salience Network (Dosenbach et al. 2006; Seeley et al. 2007). The anterior insulae show robust activation in response to errors of many types, and may specifically respond to the conscious awareness of errors (Klein et al. 2007; Ullsperger et al. 2010). Our results are compatible with such a role following internally generated errors. However, in a similar way to the dACC, externally signaled timing errors were not associated with anterior insulae activation, suggesting that this structure is not always necessary for behavioral adaptation to occur.

In contrast, common activation was seen within the pars operculari and the supramarginal gyri. In the right hemisphere, these regions form part of what has been termed the ventral attentional network (Corbetta and Shulman 2002). This network responds to behaviorally relevant events, particularly when they are unexpected or salient (Corbetta et al. 2002; Sharp et al. 2010), and appears important for reorienting attention. Our results are compatible with this role in response to errors, and suggest that both internal and external events trigger similar activity within the network. The ventral attentional network is usually considered to be a right lateralized system, which complicates this interpretation as we observe bilateral inferior frontal and inferior parietal activation. However, unexpected behaviorally relevant events such as “odd ball” stimuli have previously been shown to produce bilateral activation in these regions (Stevens et al. 2000; Fichtenholtz et al. 2004). The bilateral activation we observe may be due to additional cognitive processing required to process the behavioral significance of the errors in our task.

Moving more anteriorly along the inferior frontal gyrus, a striking change in activation pattern was observed. In contrast to the common pars opecularis activity, greater activation was observed for timing errors in the pars triangulari. This is important as it provides evidence that timing errors were not simply less internally salient and so associated with less increase in neural activity. The increase in pars triangulari activity may be explained by the rostral–caudal organization of the lateral prefrontal cortex initially proposed by Koechlin et al. (2003). In this model, cognitive control processes are organized in a cascade, with the sensory control involved in selecting motor actions supported by lateral premotor regions and higher level control processes supported by more anterior regions. It is proposed that the pars triangularis is involved in the episodic control of behavior, guiding stimulus-response mapping on the basis of either past events or future plans. The key difference between timing and commission errors in this respect may be that the former are not perceived as being causally linked to the current behavioral episode because of their unpredictability, and hence engage control processes that allow behavioral adaptation over an extended time period.

Different types of error can vary in their salience, and the pattern of errors we observed suggest that subjects may vary in their motivation for avoiding the 2 error types. As motivational factors affect the magnitude of the ERN (Gehring et al. 1993), this might influence the response of the dACC. We investigated this possibility by capitalizing on the large variability in the numbers of errors of timing across different runs and across individuals to investigate this possibility. The number of timing errors was used as a marker of the internal saliency of the feedback, that is, how motivated individuals were to consistently provide rapid responses. Our analysis showed that runs with high and low numbers of timing errors showed similar patterns of brain activation. In particular, the Salience Network was not activated in either situation, suggesting that variability in internal salience was not the reason for a difference in dACC activation.

The difference in dACC activation was unlikely to be because of methodological issues. First, the absence of external feedback during errors of commission did not in any way influence the dACC result, as demonstrated by our control experiment. Second, the frequency of error occurrence was not an important factor as errors of timing and commission accounted for similar proportions of the total number of trials. Finally, subtle timing differences between error types were not an important factor. Errors of timing occurred on average around 150 ms after errors of commission. We explicitly modeled this in our event-related design and added temporal derivatives of the error timings (Smith et al. 2004), which has the effect of correcting for intersubject temporal differences in the generation of the BOLD response. In addition, although the physiological response to errors of timing might be expected to occur around 200 ms after than the response to errors of commission (Miltner et al. 1997), our fMRI analysis is highly unlikely to be sensitive to this temporal difference. We observed widespread activation of other brain regions to timing errors, which provides evidence that the slight delay in timing error was not enough to explain the lack of dACC activation. Furthermore, we performed an additional time-series analysis of neural activation within the dACC, which confirmed that there was no evidence of late ACC activation following errors of timing.

Taken together, our results suggest that the dACC does not respond to all types of behaviorally salient error. In keeping with neuropsychological work (Modirrousta and Fellows 2008), the results suggest that the performance monitoring system is multifaceted, and cognitive control can be triggered independently of the dACC. Differences in dACC activity associated with errors of timing or commission may be due to different levels of feedback predictability or conflict present in the 2 situations. An alternative route exists to engage cognitive control, and this involves a bilateral frontoparietal network, including regions thought of being part of the ventral attentional system. The common activation of the pars opercularis across both types of error is in keeping with this region signaling, a general requirement for a shift in cognitive set.

Funding

Medical Research Council (UK) (Clinician Scientist Fellowship to D.J.S.); Imperial College Healthcare Charity (to D.J.S.) and Pfizer (Investigator Led Award to D.J.S.); and Hammersmith Hospitals Trustees' Research Committee.

This paper is dedicated to the memory of our friend and colleague, neuroscientist Dr Thomas Schofield. Tom tragically lost his life in a road-traffic accident in November 2010.

We would also like to thank Anna Joffe who helped with the development of the paradigm. Conflict of Interest: None declared.

References

Agam
Y
Hamalainen
MS
Lee
AK
Dyckman
KA
Friedman
JS
Isom
M
Makris
N
Manoach
DS
Multimodal neuroimaging dissociates hemodynamic and electrophysiological correlates of error processing
Proc Natl Acad Sci U S A
 , 
2011
, vol. 
108
 (pg. 
17556
-
17561
)
Allan
LG
Comments on current ratio-setting models for time perception
Percept Psychophys
 , 
1978
, vol. 
24
 (pg. 
444
-
450
)
Badre
D
Wagner
AD
Selection, integration, and conflict monitoring; assessing the nature and generality of prefrontal cognitive control mechanisms
Neuron
 , 
2004
, vol. 
41
 (pg. 
473
-
487
)
Botvinick
MM
Braver
TS
Barch
DM
Carter
CS
Cohen
JD
Conflict monitoring and cognitive control
Psychol Rev
 , 
2001
, vol. 
108
 (pg. 
624
-
652
)
Botvinick
MM
Cohen
JD
Carter
CS
Conflict monitoring and anterior cingulate cortex: an update
Trends Cogn Sci
 , 
2004
, vol. 
8
 (pg. 
539
-
546
)
Bush
G
Vogt
BA
Holmes
J
Dale
AM
Greve
D
Jenike
MA
Rosen
BR
Dorsal anterior cingulate cortex: a role in reward-based decision making
Proc Natl Acad Sci U S A
 , 
2002
, vol. 
99
 (pg. 
523
-
528
)
Christ
S
Falkenstein
M
Heuer
H
Hohnsbein
J
Different error types and error processing in spatial stimulus-response-compatibility tasks: behavioural and electrophysiological data
Biol Psychol
 , 
2000
, vol. 
51
 (pg. 
129
-
150
)
Corbetta
M
Kincade
JM
Shulman
GL
Neural systems for visual orienting and their relationships to spatial working memory
Cogn Neurosci
 , 
2002
, vol. 
14
 (pg. 
508
-
523
)
Corbetta
M
Shulman
GL
Control of goal-directed and stimulus-driven attention in the brain
Nat Rev Neurosci
 , 
2002
, vol. 
3
 (pg. 
201
-
215
)
Critchley
HD
Wiens
S
Rotshtein
P
Ohman
A
Dolan
RJ
Neural systems supporting interoceptive awareness
Nat Neurosci
 , 
2004
, vol. 
7
 (pg. 
189
-
195
)
Debener
S
Ullsperger
M
Siegel
M
Fiehler
K
von Cramon
DY
Engel
AK
Trial-by-trial coupling of concurrent electroencephalogram and functional magnetic resonance imaging identifies the dynamics of performance monitoring
J Neurosci
 , 
2005
, vol. 
25
 (pg. 
11730
-
11737
)
Dehaene
S
Posner
MI
Tucker
DM
Localization of a neural system for error-detection and compensation
Psychol Sci
 , 
1994
, vol. 
5
 (pg. 
303
-
305
)
di Pellegrino
G
Ciaramelli
E
Ladavas
E
The regulation of cognitive control following rostral anterior cingulate cortex lesion in humans
J Cogn Neurosci
 , 
2007
, vol. 
19
 (pg. 
275
-
286
)
Dosenbach
NU
Visscher
KM
Palmer
ED
Miezin
FM
Wenger
KK
Kang
HC
Burgund
ED
Grimes
AL
Schlaggar
BL
Petersen
SE
A core system for the implementation of task sets
Neuron
 , 
2006
, vol. 
50
 (pg. 
799
-
812
)
Eichele
T
Debener
S
Calhoun
VD
Specht
K
Engel
AK
Hugdahl
K
von Cramon
DY
Ullsperger
M
Prediction of human errors by maladaptive changes in event-related brain networks
Proc Natl Acad Sci U S A
 , 
2008
, vol. 
105
 (pg. 
6173
-
6178
)
Falkenstein
M
Hoormann
J
Christ
S
Hohnsbein
J
ERP components on reaction errors and their functional significance: a tutorial
Biol Psychol
 , 
2000
, vol. 
51
 (pg. 
87
-
107
)
Fellows
LK
Farah
MJ
Is anterior cingulate cortex necessary for cognitive control?
Brain
 , 
2005
, vol. 
128
 (pg. 
788
-
796
)
Fichtenholtz
HM
Dean
HL
Dillon
DG
Yamasaki
H
McCarthy
G
LaBar
KS
Emotion-attention network interactions during a visual oddball task
Brain Res Cogn Brain Res
 , 
2004
, vol. 
20
 (pg. 
67
-
80
)
Garavan
H
Ross
TJ
Murphy
K
Roche
RA
Stein
EA
Dissociable executive functions in the dynamic control of behavior: inhibition, error detection, and correction
Neuroimage
 , 
2002
, vol. 
17
 (pg. 
1820
-
1829
)
Gehring
W
Goss
B
Coles
M
A neural system for error detection and compensation
Psychol Sci
 , 
1993
, vol. 
4
 (pg. 
385
-
390
)
Grondin
S
Timing and time perception: a review of recent behavioral and neuroscience findings and theoretical directions
Atten Percept Psychophys
 , 
2010
, vol. 
72
 (pg. 
561
-
582
)
Hajcak
G
Holroyd
CB
Moser
JS
Simons
RF
Brain potentials associated with expected and unexpected good and bad outcomes
Psychophysiology
 , 
2005
, vol. 
42
 (pg. 
161
-
170
)
Hester
R
Fassbender
C
Garavan
H
Individual differences in error processing: a review and reanalysis of three event-related fMRI studies using the GO/NOGO task
Cereb Cortex
 , 
2004
, vol. 
14
 (pg. 
986
-
994
)
Hester
R
Foxe
JJ
Molholm
S
Shpaner
M
Garavan
H
Neural mechanisms involved in error processing: a comparison of errors made with and without awareness
Neuroimage
 , 
2005
, vol. 
27
 (pg. 
602
-
608
)
Holroyd
CB
Coles
MG
The neural basis of human error processing: reinforcement learning, dopamine, and the error-related negativity
Psychol Rev
 , 
2002
, vol. 
109
 (pg. 
679
-
709
)
Holroyd
CB
Coles
MG
Dorsal anterior cingulate cortex integrates reinforcement history to guide voluntary behavior
Cortex
 , 
2008
, vol. 
44
 (pg. 
548
-
559
)
Holroyd
CB
Krigolson
OE
Baker
R
Lee
S
Gibson
J
When is an error not a prediction error? An electrophysiological investigation
Cogn Affect Behav Neurosci
 , 
2009
, vol. 
9
 (pg. 
59
-
70
)
Holroyd
CB
Nieuwenhuis
S
Yeung
N
Nystrom
L
Mars
RB
Coles
MG
Cohen
JD
Dorsal anterior cingulate cortex shows fMRI response to internal and external error signals
Nat Neurosci
 , 
2004
, vol. 
7
 (pg. 
497
-
498
)
Jessup
RK
Busemeyer
JR
Brown
JW
Error effects in anterior cingulate cortex reverse when error likelihood is high
J Neurosci
 , 
2010
, vol. 
30
 (pg. 
3467
-
3472
)
Kennerley
SW
Walton
ME
Behrens
TE
Buckley
MJ
Rushworth
MF
Optimal decision making and the anterior cingulate cortex
Nat Neurosci
 , 
2006
, vol. 
9
 (pg. 
940
-
947
)
Kerns
JG
Cohen
JD
MacDonald
AW
3rd
Cho
RY
Stenger
VA
Carter
CS
Anterior cingulate conflict monitoring and adjustments in control
Science
 , 
2004
, vol. 
303
 (pg. 
1023
-
1026
)
Klein
TA
Endrass
T
Kathmann
N
Neumann
J
von Cramon
DY
Ullsperger
M
Neural correlates of error awareness
Neuroimage
 , 
2007
, vol. 
34
 (pg. 
1774
-
1781
)
Koechlin
E
Ody
C
Kouneiher
F
The architecture of cognitive control in the human prefrontal cortex
Science
 , 
2003
, vol. 
302
 (pg. 
1181
-
1185
)
 
Li CS, Huang C, Yan P, Paliwal P, Constable RT, Sinha R. 2008. Neural Correlates of Post-error Slowing during a Stop Signal Task: A functional Magnetic Resonance Imaging Study. J Neurosci. 20: 1021–1029
Luu
P
Flaisch
T
Tucker
DM
Medial frontal cortex in action monitoring
The J Neurosci
 , 
2000
, vol. 
20
 (pg. 
464
-
469
)
Miltner
WHR
Braun
CH
Coles
MGH
Event-related brain potentials following incorrect feedback in a time-estimation task: evidence for a “generic” neural system for error detection
J Cogn Neurosci
 , 
1997
, vol. 
9
 (pg. 
788
-
798
)
Modirrousta
M
Fellows
LK
Dorsal medial prefrontal cortex plays a necessary role in rapid error prediction in humans
J Neurosci
 , 
2008
, vol. 
28
 (pg. 
14000
-
14005
)
Rabbitt
PM
Errors and error correction in choice-response tasks
J Exp Psychol
 , 
1966
, vol. 
71
 (pg. 
264
-
272
)
Ridderinkhof
KR
Ullsperger
M
Crone
EA
Nieuwenhuis
S
The role of the medial frontal cortex in cognitive control
Science
 , 
2004
, vol. 
306
 (pg. 
443
-
447
)
Rushworth
MF
Walton
ME
Kennerley
SW
Bannerman
DM
Action sets and decisions in the medial frontal cortex
Trends Cogn Sci
 , 
2004
, vol. 
8
 (pg. 
410
-
417
)
Seeley
WW
Menon
V
Schatzberg
AF
Keller
J
Glover
GH
Kenna
H
Reiss
AL
Greicius
MD
Dissociable intrinsic connectivity networks for salience processing and executive control
J Neurosci
 , 
2007
, vol. 
27
 (pg. 
2349
-
2356
)
Sharp
DJ
Bonnelle
V
De Boissezon
X
Beckmann
CF
James
SG
Patel
MC
Mehta
MA
Distinct frontal systems for response inhibition, attentional capture, and error processing
Proc Natl Acad Sci U S A
 , 
2010
, vol. 
107
 (pg. 
6106
-
6111
)
Sharp
DJ
Scott
SK
Mehta
MA
Wise
RJ
The neural correlates of declining performance with age: evidence for age-related changes in cognitive control
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
1739
-
1749
)
Simon
JR
Reactions toward the source of stimulation
J Exp Psychol
 , 
1969
, vol. 
81
 (pg. 
174
-
176
)
Simon
JR
Small
AM
Jr
Processing auditory information: interference from an irrelevant cue
J Appl Psychol
 , 
1969
, vol. 
53
 (pg. 
433
-
435
)
Smith
SM
Jenkinson
M
Woolrich
MW
Beckmann
CF
Behrens
TE
Johansen-Berg
H
Bannister
PR
De Luca
M
Drobnjak
I
Flitney
DE
, et al.  . 
Advances in functional and structural MR image analysis and implementation as FSL
Neuroimage
 , 
2004
, vol. 
23
 
Suppl 1
(pg. 
S208
-
S219
)
Stevens
AA
Skudlarski
P
Gatenby
JC
Gore
JC
Event-related fMRI of auditory and visual oddball tasks
Magn Reson Imaging
 , 
2000
, vol. 
18
 (pg. 
495
-
502
)
Swick
D
Turken
AU
Dissociation between conflict detection and error monitoring in the human anterior cingulate cortex
Proc Natl Acad Sci U S A
 , 
2002
, vol. 
99
 (pg. 
16354
-
16359
)
Ullsperger
M
Harsay
HA
Wessel
JR
Ridderinkhof
KR
Conscious perception of errors and its relation to the anterior insula
Brain Struct Funct
 , 
2010
, vol. 
214
 (pg. 
629
-
643
)
Ullsperger
M
Nittono
H
von Cramon
DY
When goals are missed: dealing with self-generated and externally induced failure
Neuroimage
 , 
2007
, vol. 
35
 (pg. 
1356
-
1364
)
Ullsperger
M
von Cramon
DY
Error monitoring using external feedback: specific roles of the habenular complex, the reward system, and the cingulate motor area revealed by functional magnetic resonance imaging
J Neurosci
 , 
2003
, vol. 
23
 (pg. 
4308
-
4314
)
Yeung
N
Cohen
JD
Botvinick
MM
The neural basis of error detection: conflict monitoring and the error-related negativity
Psychol Rev
 , 
2004
, vol. 
111
 (pg. 
931
-
959
)

Author notes

Timothy Ham and Xavier de Boissezon both contributed jointly to the authorship of this paper.