Abstract

Preparing to stop an inappropriate action requires keeping in mind the task goal and using this to influence the action control system. We tested the hypothesis that different subregions of prefrontal cortex show different temporal profiles consistent with dissociable contributions to preparing-to-stop, with dorsolateral prefrontal cortex (DLPFC) representing the task goal and ventrolateral prefrontal cortex (VLPFC) implementing action control. Five human subjects were studied using electrocorticography recorded from subdural grids over right lateral frontal cortex. On each trial, a task cue instructed the subject whether stopping might be needed or not (Maybe Stop [MS] or No Stop [NS]), followed by a go cue, and on some MS trials, a subsequent stop signal. We focused on go trials, comparing MS with NS. In the DLPFC, most subjects had an increase in high gamma activity following the task cue and the go cue. In contrast, in the VLPFC, all subjects had activity after the go cue near the time of the motor response on MS trials, related to behavioral slowing, and significantly later than the DLPFC activity. These different temporal profiles suggest that DLPFC and VLPFC could have dissociable roles, with DLPFC representing task goals and VLPFC implementing action control.

Introduction

Much evidence shows that stopping an action outright (i.e. completely, and to an external signal) implicates a predominantly right-hemisphere fronto-basal-ganglia network. This includes the ventrolateral prefrontal cortex (VLPFC), the presupplementary motor area (preSMA) and the basal ganglia (reviewed by Aron et al. 2007b; Nachev et al. 2008; Chambers et al. 2009; Chikazoe 2010; Levy and Wagner 2011). However, everyday life not only requires us to stop outright but also to prepare to stop. Here we investigate the neural correlates of preparing to stop using the high spatio-temporal resolution of electrocorticography (ECoG).

Studies with functional magnetic resonance imaging (fMRI) have recently investigated this issue by designing tasks where preparing to stop can be separated from outright stopping. These studies show preparing to stop recruits some of the same prefrontal regions as outright stopping, including the preSMA and the right VLPFC (specifically the inferior frontal gyrus), and in addition, the dorsolateral prefrontal cortex (DLPFC) (Vink et al. 2005; Chikazoe et al. 2009b; Jahfari et al. 2010; Zandbelt and Vink, 2010; Zandbelt et al. 2012). The DLPFC finding is consistent with a role for representing goals in working memory (Goldman-Rakic, 1990; Petrides, 2000; Olesen et al. 2004; Muller and Knight, 2006) and/or attentional monitoring (Corbetta and Shulman, 2002)—in this case for the stopping task goal. This motivates the hypothesis that the DLPFC and VLPFC play dissociable roles in preparing to stop. Specifically, the DLPFC may represent the task goal while the VLPFC may implement action control—by which we mean an active, although partial, suppression of the motor response (i.e. braking). A key aspect of this prediction is that the DLPFC and VLPFC activity should be temporally dissociable, with the DLPFC activity occurring early, around the time of a task goal cue (task cue) and the VLPFC activity occurring later, around the time when stopping is anticipated or perhaps around the time of the motor response itself. To test these predictions, we recorded ECoG in 5 subjects, each with electrode coverage of both DLPFC and VLPFC, while they performed a preparing-to-stop task.

On each trial, a stopping task cue was presented (“Maybe Stop” [MS] or “No Stop” [NS]), followed by a Go cue, and then, on the MS trials alone, a stop signal occurred half the time (Fig. 1). In a previous fMRI study with this task in healthy controls, we reported that the contrast of MS Go trials versus NS Go trials activated the DLPFC bilaterally and the VLPFC in the right hemisphere, as well as the preSMA (Swann et al. 2012). The preSMA is part of the stopping network, having functional and structural connectivity with the VLPFC and the subthalamic nucleus of the basal ganglia (Forstmann et al. 2010, 2012; Neubert et al. 2010; Swann et al. 2012), and all 3 of these regions show similar electrophysiological responses for outright stopping (Swann et al. 2009, 2012; Leventhal et al. 2012; Ray et al. 2012). However, in this study, we focused uniquely on the DLPFC and VLPFC (where we had coverage in all subjects), to test the hypothesis about their distinct functional roles in preparing to stop. Notably, we examined Go trials only, to evaluate the neural correlates of preparing to stop, in a context in which there was always a motor response. Specifically, we analyzed MS Go trials, NS Go trials, and the difference between them.

Figure 1.

Task: each trial begins with a task cue, either NS or MS, followed by an arrow pointing left or right. Subjects were instructed to respond with a button press to the arrows as quickly as possible. On half of the MS trials an auditory beep followed the arrow and indicated that subjects should attempt to stop the initiated response.

Figure 1.

Task: each trial begins with a task cue, either NS or MS, followed by an arrow pointing left or right. Subjects were instructed to respond with a button press to the arrows as quickly as possible. On half of the MS trials an auditory beep followed the arrow and indicated that subjects should attempt to stop the initiated response.

Although ECoG provides information about many different frequency bands of oscillatory activity, our present interest in the temporal profiles of DLPFC and VLPFC motivated a focus on high gamma activity (∼90–130 Hz). This high-frequency activity is a putative marker of local neural processing within a region (Edwards et al. 2005; Canolty et al. 2007; Miller et al. 2007a, 2007b; Ray et al. 2008; Flinker et al. 2010; Conner et al. 2011; Scheeringa et al. 2011), while lower-frequency activity (beta, alpha, and theta) is putatively important for long-distance communication between brain regions (Kopell et al. 2000, 2010; Fries, 2005). Thus, given our current goal of differentiating the temporal profiles of DLPFC and VLPFC, we focused only on high gamma activity as an index of local neural processing within a region.

Behaviorally, we expected that action control would be manifest as slower reaction time (RT) for MS Go versus NS Go trials (the “response delay effect,” RDE, cf. Chikazoe et al. 2009b; Jahfari et al. 2010; Zandbelt and Vink, 2010; Swann et al. 2012; Zandbelt et al. 2012). With ECoG we focused on 2 main aspects. First we tested the relative timing of activity in DLPFC versus VLPFC. We expected increased gamma activity in the DLPFC around the time of the task cue and increased gamma activity in the VLPFC following the go cue. This would support dissociable roles for these regions. Second, we more closely examined the VLPFC to better understand its role in preparing to stop. Whereas fMRI shows increased activity in VLPFC for MS Go compared with NS Go, it is unclear when this difference occurs in time. We expected that ECoG, with its higher temporal resolution, would clarify this issue, and thus provide novel information about the functional role of right VLPFC.

Methods

Participants

ECoG was collected in 7 subjects with pharmacologically resistant epilepsy who underwent surgery for resection of seizure foci. All subjects provided informed consent under the auspices of the Internal Review Board at University of Texas at Houston. Data from 2 subjects are not included in this report because of poor task performance (>30% omissions on Go trials). Thus, the analysis focuses on 5 subjects (3 female, mean age 34 years, SD = 9.8). All had right lateral frontal coverage that included the DLPFC and VLPFC and none of them had any evidence for right frontal abnormalities or underwent subsequent resection of these regions. One of these subjects (s2) was included in a previous report (Swann et al. 2012). However, here we focus on other aspects of this subject's data and also on the 4 additional subjects.

Task

We used a MS/NS task, (Swann et al. 2012; Fig. 1). Each trial began with a task cue presented for 1 s. The task cue was either “MS” (red text on a black background) or “NS ” (green text on a black background). Each cue was equally likely to occur. The task cue was followed by a go cue (white arrow), pointing to either the right or the left, to which subjects were instructed to respond with a button press. On half the MS trials an auditory beep followed the go cue, instructing the subjects to attempt to stop the motor response. In 3 subjects (s2, s4, and s5), the delay between the go cue and the stop signal (stop signal delay, SSD) was dynamically adjusted (for full details see Aron and Poldrack, 2006), with an initial delay set to 150 ms, whereas for the other 2 subjects (s1 and s3), SSDs were randomly selected from 3 individually determined fixed values spaced 100-ms apart based on an initial training session. We think it is unlikely that this procedural difference influenced the current results since the behavioral results for the go trials for subjects with the fixed SSD method did not differ systematically from the other subjects, and also because the fixed SSD method is a standard procedure (Verbruggen and Logan, 2009). After the go cue, there was a 1.5-s response window followed by a jittered intertrial interval (1–1.4 s). All subjects responded with their left hand (so that primary motor cortical responses could be recorded from the right lateral grids). The subjects performed either 6 or 7 blocks of the experiment, with 96 trials per block.

Behavioral Analysis

For each subject we calculated mean RT for MS and NS Go trials (with RTs less than 100 ms excluded), the difference between the 2 (the RDE), the average SSD, the percentage discrimination and omission errors on go trials, the percentage of successful stop trials, and the RT for failed stop trials. We also estimated stop signal reaction time (SSRT) for each subject. For the subjects whose SSDs were determined using the “tracking” method (s2, s4, and s5), SSRT was estimated with the “mean method” (Logan et al. 1997); for the other 2 subjects (s1 and s3) SSRT was estimated with the “integration method” (Verbruggen and Logan, 2009).

Electrode Localization

Electrode localization procedures were the same as our previous reports (Swann et al. 2009, 2012). In brief, a computed tomography scan was obtained after electrode implantation, and the scan was coregistered to a preimplantation MRI scan. These electrodes were then projected onto a cortical surface model generated in FreeSurfer (Dale et al. 1999). SUMA software was used to visualize the electrodes as spheres (Saad et al. 2003). Locations were verified with intraoperative photos.

ECoG Acquisition and Analysis

This was similar to our previous reports (Swann et al. 2009, 2012). Data were acquired with an electroencephalography (EEG) 1100 Neurofax clinical Nihon Koden acquisition system sampled at 1000 Hz. Trial onsets were sent to the EEG acquisition computer with transistor–transistor logic pulses.

The ECoG data were first digitally re-referenced to the common average, excluding channels with frequent artifacts or ictal contamination. Manual inspection was used to identify and exclude trials with either ictal activity that spread across channels or other sources of noise. Following artifact rejection and rejection of trials with incorrect behavior, there was on average 132 MS Go trials remaining per subject (SD = 12, range 116–148), and roughly twice as many NS Go trials. Channels identified as seizure onset channels were also excluded from the entire analysis.

Analysis proceeded as follows. First, the data were filtered with Gabor wavelets into separate analytic amplitudes at 128 different center frequencies (ranging from 2.5 to 250 Hz; see Canolty et al. 2007). Second, analytic amplitudes for each frequency were aligned to the task cue (for the main analysis), averaged across trials, and compared with an average baseline of 500 ms in the center of the intertrial interval (−750 to −250 ms prior to trial start). Third, the average analytic amplitude values were converted to z-scores using a permutation technique where a standard deviation of the data was derived for each frequency from a distribution generated by repeating the above procedure 10 000 times, time locking to random points in time (see Canolty et al. 2007). For calculations of between-condition differences (MS Go vs. NS Go) average analytic amplitude values for differences were directly subtracted (without a baseline correction applied), and z-scores were derived from an analogous permutation method as above, but now using 5000 iterations of a label-swapping procedure. Fourth, multiple comparisons were corrected for using the false discovery rate (FDR) method. Fifth, while some results show the full time–frequency spectrum, others are specifically focused on high gamma activity (∼90–130 Hz). This is the frequency band that has been most closely linked to local neuronal activity, and to the blood oxygen level dependent signal (Ray et al. 2008; Conner et al. 2011; Scheeringa et al. 2011). For these analyses data were extracted from this frequency band (center frequency: 110 Hz) and statistics were performed for this band separately.

Results

Behavior

All 5 subjects performed well, having less than 3% discrimination errors on Go trials (i.e. errors where the incorrect button was pressed) and 10% or less omission errors on Go trials (see Table 1). The speed of stopping (SSRT) was in a typical range for ECoG subjects (Swann et al. 2009). Three of the 5 subjects (s1, s2, and s3) showed a significant RDE, with MS RT being significantly longer than NS RT, as in healthy participants (Chikazoe et al. 2009b; Jahfari et al. 2010; Swann et al. 2012). However, the other 2 subjects (s4 and s5) showed no significant RDE, perhaps because they had difficulty keeping this additional rule in mind. Instead, they might have treated the task like a standard stopping task and thus anticipated stopping on all trials regardless whether they were in an MS or NS condition (such that a task goal and action control strategies were maintained for all trials, instead of for MS Go trials only). As these 2 subjects did perform the task well otherwise (i.e. their stopping behavior was typical), they are included in the ECoG analysis, and in fact the absence of the RDE is an informative control condition when examining their activity compared with the other subjects.

Table 1

Behavioral results

Subjects Mean MS RT (SD), ms Mean NS RT (SD), ms RDE, ms Proactive slowing Disc errors (%) Omission errors (%) Failed stop RT (SD), ms SSRT Percent inhibit (%) 
S1a 700 (168) 552 (125) 148 Yes 1.0 10 603 (114) 271b 56 
S2a 618 (225) 491 (138) 127 Yes 1.7 3.5 540 (172) 239 49 
S3a 735 (168) 535 (136) 200 Yes 2.2 2.4 653 (121) 124b 71 
S4 861 (325) 831 (313) 30 No 1.6 645 (160) 219 49 
S5 598 (140) 596 (139) No 0.2 0.8 524 (163) 469 39 
Subjects Mean MS RT (SD), ms Mean NS RT (SD), ms RDE, ms Proactive slowing Disc errors (%) Omission errors (%) Failed stop RT (SD), ms SSRT Percent inhibit (%) 
S1a 700 (168) 552 (125) 148 Yes 1.0 10 603 (114) 271b 56 
S2a 618 (225) 491 (138) 127 Yes 1.7 3.5 540 (172) 239 49 
S3a 735 (168) 535 (136) 200 Yes 2.2 2.4 653 (121) 124b 71 
S4 861 (325) 831 (313) 30 No 1.6 645 (160) 219 49 
S5 598 (140) 596 (139) No 0.2 0.8 524 (163) 469 39 

MS, Maybe Stop Go RT; NS, No Stop Go RT; RDE, response delay effect (mean MS RT—mean NS RT); Proactive slowing (whether or not there was a significant RDE); Disc. errors, percentage discrimination errors; omissions errors, percentage omission errors; failed stop RT, mean RT on failed stop trials; SSRT, stop signal reaction time; percent stop, percentage of stop trials with successful stopping.

aSignificant response delay effect (MS RT sig greater than NS RT).

bSubjects who were tested with a fixed SSD method.

ECoG

Prefrontal Activity When Preparing to Stop

Our hypothesis was that DLPFC represents the task goal (i.e. “this is [or is not] a trial on which stopping may be needed”) and that VLPFC implements action control, by which we mean an active, although partial, suppression of the motor response (i.e. braking). This predicts that DLPFC activity should occur in the task cue period, while VLPFC activity should occur later, perhaps following the go cue or around the time of the motor response.

First, to give a descriptive example of the neural time-course for the whole brain we examined high gamma amplitudes for the full time course for 1 subject, s2 (Fig. 2). Here the amplitude was averaged over 100 ms for each time-bin, and converted to a z-score (P< 0.05, FDR corrected for all electrodes and time-bins). We focus on s2 because this subject was unique in having extensive brain coverage, including of primary visual cortex and primary motor cortex, and also a typical pattern of activity for the task. Note that the DLPFC shows bursts of activity approximately 300 after the MS cue and approximately 200 ms after the go cue, whereas the VLPFC shows activity mainly after the go cue, peaking around the time of the motor response. There is also activity in occipital cortex around the time of the visual stimulus and in the precentral gyrus around the time of the motor response, providing validation of the ECoG processing stream, and of the electrode localization method.

Figure 2.

The full dynamic picture of activity for a single subject (s2) during a preparing to stop trial. DLPFC was active after both the MS cue and the go cue, and VLPFC was active later, most strongly around the time of the motor response. The colored electrodes denote those with significant high gamma activity (z-scores) for MS Go versus baseline for each time bin (P< 0.05, FDR corrected). The MS cue occurred at 0 ms, the go cue occurred at 1000 ms, and for this subject average RT was 618 ms (1618 ms relative to the MS cue).

Figure 2.

The full dynamic picture of activity for a single subject (s2) during a preparing to stop trial. DLPFC was active after both the MS cue and the go cue, and VLPFC was active later, most strongly around the time of the motor response. The colored electrodes denote those with significant high gamma activity (z-scores) for MS Go versus baseline for each time bin (P< 0.05, FDR corrected). The MS cue occurred at 0 ms, the go cue occurred at 1000 ms, and for this subject average RT was 618 ms (1618 ms relative to the MS cue).

Next, to formally test this hypothesized temporal difference for DLPFC and VLPFC, we proceeded as follows. We identified electrodes with significant high gamma amplitude increases for either MS Go trials versus baseline or for NS Go versus baseline for at least 50 ms, to exclude spurious events (P< 0.05, FDR corrected, for all electrodes and all time points). Then, for each of these electrodes, we examined if MS Go activity was greater than NS Go, for at least 50 ms (P< 0.05, FDR corrected for all electrodes and time-points examined). The above steps were done separately for the 2 trial periods: 1) the task cue period (defined as 1 s following the task cue) and 2) the go cue period (defined as 1 s following the go cue).

During the task cue period, for MS Go versus baseline, 3 out of 5 subjects (s1, s3, and s5), exhibited significantly increased high gamma band activity in the prefrontal cortex, mainly in the DLPFC, that is, in the middle frontal gyrus, or in and around the inferior frontal sulcus (Fig. 3A). (Note, s2 also had DLPFC activity but it did not pass the correction for multiple comparisons for this particular analysis, when correcting for all time points, but the subject did show significant activity when correcting for time-bins, see Fig. 2.) However, for MS Go versus NS Go, there were no significant effects in prefrontal regions for any subject. Only one subject showed any VLPFC activity during this period (s1), and this was for 2 electrodes slightly more rostral than the VLPFC electrodes that were active in the other subjects during either time period.

Figure 3.

(A and B). DLPFC and VLPFC activity in the task cue and go cue periods. (A) The task cue period (i.e. 1s following the MS cue). For most subjects, there is activity in and above the inferior frontal sulcus (i.e. DLPFC) for MS Go versus baseline, indicated with black arrows, but not for the comparison of MS Go versus NS Go. Note only 1 subject (s1) shows VLPFC during this period in 2 more rostral VLPFC electrodes. (B) The go cue period (i.e. 1 s following the go signal). In all subjects, there is activity in and around the precentral suclus (i.e. VLPFC) and there are also differences between MS Go and NS Go for the 3 subjects who had significant behavioral slowing (RDE), indicated with black arrows. Note that there is also DLFPC activity in most subjects during this period and differences between MS Go versus NS Go in 2 subjects (s2 and s3). Red electrodes denote significantly greater high gamma amplitude (110 Hz) for MS Go versus baseline and black circles denote significant MS Go versus NS Go trials (P< 0.05, FDR corrected). The dotted rectangles denote the 2 subjects who did not slow down in preparation for stopping. (C) DLPFC precedes VLPFC. Left panel: for all subjects the time of significant gamma in DLPFC occurred earlier than VLPFC. Zero millisecond is the time of the “MS” cue and 1000 ms is the time of the go signal. Right panel: Pooling across the channels and subjects in the left panel, a single trial analysis reveals that DLPFC activity increased after both the task cue and the go cue, while VLPFC had a tendency to increase mainly after the go cue and after DLPFC. Zero millisecond is the time of the MS cue and 1000 ms is the time of the go signal.

Figure 3.

(A and B). DLPFC and VLPFC activity in the task cue and go cue periods. (A) The task cue period (i.e. 1s following the MS cue). For most subjects, there is activity in and above the inferior frontal sulcus (i.e. DLPFC) for MS Go versus baseline, indicated with black arrows, but not for the comparison of MS Go versus NS Go. Note only 1 subject (s1) shows VLPFC during this period in 2 more rostral VLPFC electrodes. (B) The go cue period (i.e. 1 s following the go signal). In all subjects, there is activity in and around the precentral suclus (i.e. VLPFC) and there are also differences between MS Go and NS Go for the 3 subjects who had significant behavioral slowing (RDE), indicated with black arrows. Note that there is also DLFPC activity in most subjects during this period and differences between MS Go versus NS Go in 2 subjects (s2 and s3). Red electrodes denote significantly greater high gamma amplitude (110 Hz) for MS Go versus baseline and black circles denote significant MS Go versus NS Go trials (P< 0.05, FDR corrected). The dotted rectangles denote the 2 subjects who did not slow down in preparation for stopping. (C) DLPFC precedes VLPFC. Left panel: for all subjects the time of significant gamma in DLPFC occurred earlier than VLPFC. Zero millisecond is the time of the “MS” cue and 1000 ms is the time of the go signal. Right panel: Pooling across the channels and subjects in the left panel, a single trial analysis reveals that DLPFC activity increased after both the task cue and the go cue, while VLPFC had a tendency to increase mainly after the go cue and after DLPFC. Zero millisecond is the time of the MS cue and 1000 ms is the time of the go signal.

During the go cue period, for MS Go versus baseline, 4 of the 5 subjects (all except for s4) again had activity in the DLPFC. Two of the 5 subjects (s2 and s3) also showed differences for the MS Go versus NS Go comparison during this period (Fig. 3B). For the VLPFC, each of the 5 subjects had significant activity in the caudal VLPFC, that is, in the ventral inferior frontal gyrus or in and around the precentral sulcus (Fig. 3B). Additionally, for MS Go versus NS Go there was activity in the VLPFC in 3 out of 5 subjects (s1, s2, and s3), and notably all of these subjects had significant slowing (the RDE), whereas the 2 subjects who did not have a significant difference in activity (s4 and s5) had an RDE close to 0. This supports the notion that the MS versus NS activity difference in the VLPFC could relate to the behavioral slowing effect. We speculate that subjects without an RDE were perhaps exercising proactive control on all trials, and thus showed no difference between the 2 trial types in VLPFC, but still exhibited above-baseline activity in these regions for both trial types, and see below for more on this.

In summary, across subjects and electrodes, for the DLPFC there were high gamma responses after the task cue, the go cue, or both. For the VLPFC, there were high gamma responses mainly only after the go cue, and more strongly on MS Go than NS Go trials (for the subjects who showed an RDE). For detailed information on each subject, see Supplementary Figures 1–5 for spectral plots.

Activity in the DLPFC Precedes the VLPFC

The comparison of task cue and go cue periods (Fig. 3A vs. B) suggests a timing difference in DLPFC and VLPFC. We formally tested this as follows. 1) For each subject, for the period from 500 ms before the task cue until 1 s after the go cue (i.e. a 2500 ms window), electrodes were identified in DLPFC and VLPFC which showed significant activity for MS Go versus baseline (P< 0.05, FDR corrected for all electrodes within each region of interest (ROI) and all time points). (Note that for s4 there was no DLPFC electrode in either time window that surpassed the FDR corrected threshold, so for this analysis, for s4 alone, the threshold was lowered to P< 0.01, uncorrected to select a DLPFC electrode.) 2) For all of these electrodes, we found the time point (relative to the MS cue) where the high gamma amplitude first became significantly greater than baseline (for subjects with more than one significant DLPFC or VLPFC electrode the onsets for all electrodes in each ROI were averaged). 3) We performed a paired t-test across the 5 subjects on the latency values for DLPFC versus VLPFC. The main result was that the DLPFC was active before the VLPFC (P= 0.01, t= 4.2, df = 4), and this was true in each subject (Fig. 3C). (Note that because there were no differences in average amplitude or standard deviation between the DLPFC and VLPFC [P∼0.9 for each], it is unlikely that the difference in onset times is an artifact of different signal-to-noise properties of the 2 regions.)

We also examined the DLPFC/VLPFC timing difference using a second method. We performed a single trial analysis for MS Go trials for all the DLPFC and VLPFC channels pooled across subjects that had significantly above baseline high gamma activity (on average 132 trials per subject, SD = 12). The dependent variable for each trial, in DLPFC or VLPFC, was the time of maximum high gamma amplitude. DLPFC showed a relatively narrow peak after the MS cue and another broader and slightly smaller one after the go cue (Fig. 3C). Thus, the DLPFC exhibited more variability, responding similarly to both the task cue and the go cue. The VLPFC, in contrast, while showing some responses to the task cue, had its largest peak after the go cue around the time of the response. This single trial analysis supports the above impressions (Fig. 3A,B) that the DLPFC is active after both the task cue and the go cue while the VLPFC is active primarily after the go cue.

To help characterize the data summarized in Figure 3, spectral plots are shown from an example DLPFC and VLPFC electrode from 1 subject (s3; Fig. 4A). Here, the DLPFC electrode had significant high gamma activity shortly after the task cue, and did not differentiate between MS Go versus NS Go. In contrast, the VLPFC was active only after the go cue, and significantly more for MS Go compared with NS Go (P< 0.05, FDR corrected for all frequencies, time points, and conditions). For comparison, we also show VLPFC activity from s4, who failed to show an RDE. Note that the MS Go activity pattern for this subject was very similar to s3. However, s4, did not show a significant difference between MS Go and NS Go, and likewise had no behavioral difference either (no RDE). See Supplementary Figures 1–5 for spectral plots of all DLPFC and VLPFC with significant MS Go activity relative to baseline (indicated in Fig. 3A,B) from all subjects.

Figure 4.

(A) Example spectral plots for DLPFC and VLPFC for preparing to stop. For s3 high gamma amplitude increases are visible for both DLPFC and VLPFC, with DLPFC preceding VLPFC. Note that only VLPFC shows a significant difference for MS Go compared with NS Go (P< 0.05, FDR corrected). For a subject that did not show behavioral slowing (s4), the high gamma increases relative to baseline are still present (and are similar in time to s3), but there is no difference for MS Go versus NS Go. Zero millisecond is the time of the task cue, 1000 ms is the time of the Go signal. Amplitude is expressed in color as a z-score. Significant values (P< 0.05, FDR corrected) are outlined (black indicates positive differences, red indicates negative differences). Electrode locations are indicated in blue for each subject and region. For similar plots from all the other significant DLPFC and VLPFC electrodes, for all subjects, see Supplementary Figures 1–5. (B) VLPFC shows consistent activity across subjects when preparing to stop. Spectral plots for MS Go versus baseline are shown for a VLPFC electrode from each subject time-locked to both the MS cue (left column), and the RT (middle column). For each subject the high gamma activity centers around the motor response. Zero millisecond is the time of the task cue and 1000 ms is the time of the go signal for the left column. Zero millisecond is the time of the RT for the middle column. (C) VLPFC activity is time-locked to the motor response on a single trial level. In 2 subjects (s2 and s5), the VLPFC response on MS Go versus baseline relates to the motor response more than the expected stop signal. The top 2 plots show single trial high gamma amplitude (110 Hz) plotted for all MS Go trials aligned to go cue (0 ms), and sorted by RT. The bottom 2 plots show the same data sorted by the expected SSD. The dark line indicates RT (top plots) or expected SSD (bottom plots) for each trial. Raw amplitude relative to baseline is shown in color. For similar plots, showing the single trial analysis sorted by RT, from other DLPFC, and VLPFC electrodes, for all subjects, see Supplementary Figures 1–5.

Figure 4.

(A) Example spectral plots for DLPFC and VLPFC for preparing to stop. For s3 high gamma amplitude increases are visible for both DLPFC and VLPFC, with DLPFC preceding VLPFC. Note that only VLPFC shows a significant difference for MS Go compared with NS Go (P< 0.05, FDR corrected). For a subject that did not show behavioral slowing (s4), the high gamma increases relative to baseline are still present (and are similar in time to s3), but there is no difference for MS Go versus NS Go. Zero millisecond is the time of the task cue, 1000 ms is the time of the Go signal. Amplitude is expressed in color as a z-score. Significant values (P< 0.05, FDR corrected) are outlined (black indicates positive differences, red indicates negative differences). Electrode locations are indicated in blue for each subject and region. For similar plots from all the other significant DLPFC and VLPFC electrodes, for all subjects, see Supplementary Figures 1–5. (B) VLPFC shows consistent activity across subjects when preparing to stop. Spectral plots for MS Go versus baseline are shown for a VLPFC electrode from each subject time-locked to both the MS cue (left column), and the RT (middle column). For each subject the high gamma activity centers around the motor response. Zero millisecond is the time of the task cue and 1000 ms is the time of the go signal for the left column. Zero millisecond is the time of the RT for the middle column. (C) VLPFC activity is time-locked to the motor response on a single trial level. In 2 subjects (s2 and s5), the VLPFC response on MS Go versus baseline relates to the motor response more than the expected stop signal. The top 2 plots show single trial high gamma amplitude (110 Hz) plotted for all MS Go trials aligned to go cue (0 ms), and sorted by RT. The bottom 2 plots show the same data sorted by the expected SSD. The dark line indicates RT (top plots) or expected SSD (bottom plots) for each trial. Raw amplitude relative to baseline is shown in color. For similar plots, showing the single trial analysis sorted by RT, from other DLPFC, and VLPFC electrodes, for all subjects, see Supplementary Figures 1–5.

The Role of the VLPFC in Preparing to Stop

These different temporal activation patterns for DLPFC versus VLPFC are consistent with different roles in preparing to stop action. The DLPFC was active in the task cue and go cue periods, consistent with a role in representing task goals, which could include any of encode, maintain, rehearse, attend, monitor, or retrieve. The VLPFC, in contrast, was consistently active later in the go phase around the time of the response, consistent with a role in either action control (Aron et al. 2003; Chambers et al. 2006, 2007; Aron et al. 2007b; Jahfari et al. 2010; Neubert et al. 2010), attentional monitoring for the stop signal (Chao et al. 2009; Duann et al. 2009; Hampshire et al. 2010; Sharp et al. 2010; Boehler et al. 2011; Chatham et al. 2012), both of these, although perhaps in dissociable subregions (Chikazoe et al. 2009a; Chikazoe, 2010; Verbruggen et al. 2010; Levy and Wagner, 2011), as well as other possible accounts such as a violation of expectancies about stopping (Zandbelt et al. 2012). We speculated that if the VLPFC is important for action control, that is, as a brake, than its activity might relate to the time of the motor response (especially in the “executive” setting of MS Go trials).

To investigate this, we plotted spectral data for MS Go trials from a VLPFC electrode for each subject time-locked to the MS cue (Fig. 4B, left column) and to the reaction time (Fig. 4B, middle column). For s1–s3, the electrode is the one that showed the MS Go greater than NS Go difference (see Fig. 3B). (Note that s2 and s3 had 2 neighboring electrodes with this difference. Although all showed a similar pattern, we show results from the one that was more clearly within the VLPFC for each subject, see Supplementary Figures 1–3 for plots of data from the other electrodes). For subjects s4 and s5, no electrodes had an MS Go versus NS Go difference, therefore we show the VLPFC electrode with an MS Go versus baseline difference (shown in Fig. 3B). For all subjects the VLPFC activity was roughly centered around the time of the motor response, usually starting well before, and continuing well after. These results point to a relationship between the VLPFC and the motor response (especially on MS Go trials). Although this is consistent with a role for the VLPFC in action control, the fact that the activity was spread across time could also be consistent with other accounts, such as attentional monitoring for the stop signal. If this is the case, then its engagement should be most prominent around the time of the expected stop signal (i.e. the SSD) rather than the time of the motor response. To test this, we conducted a single trial analysis sorted by either RT, or a proxy for expected SSD (based on the SSD from the last stop trial). For s2 and s5, the VLPFC electrode shown in Figure 4B had high gamma activity that corresponded more strongly with RT than with the time of expected SSD (Fig. 4C). For the other 3 subjects the single trial analysis did not reveal any clear patterns in VLPFC when sorted by either variable. (See Supplementary Figures 1–5 for additional examples of the single trial analysis sorted by RT for other VLPFC and DLPFC electrodes.)

Discussion

We recorded ECoG from right lateral subdural grids in 5 subjects while they performed the MS/NS task. The analysis mainly focused on high gamma activity in the task cue period and following the go cue. In most subjects, there was DLPFC activity around the task cue and/or around the go cue with some electrodes responding to both, while others responded to one or the other. By contrast, in all subjects, there was activity in the VLPFC more consistently after the go cue, especially around the time of the motor response on MS trials, and significantly later than the DLPFC activity. For the VLPFC, the activity on MS trials was more closely time-locked to the motor response than it was to the anticipated time of the stop signal, and it was greater, relative to NS trials, in subjects who slowed down in anticipation of stopping compared with those who did not. These results suggest that the DLPFC and VLPFC could play dissociable roles. One interpretation is that the DLPFC represents the task goal and that the VLPFC implements action control, perhaps specifically by “braking” the initiated motor response when stopping is anticipated. This speaks to current theories of action control as well as the debate about the functional specialization of the prefrontal cortex.

Functional Role of the DLPFC

Preparing to stop a response requires having a stopping task goal and also using this information to control the response tendency. This “task goal” could encompass functions related to encoding the rules of the task, retrieving these rules, holding the rules in working memory, and/or attending to or monitoring for task-relevant stimuli. A strong candidate region for representing this task goal is the DLPFC, based on much evidence implicating it in working memory (Goldman-Rakic, 1990; Fuster, 1997; Petrides, 2000; Olesen et al. 2004; Muller and Knight, 2006) and attention (Corbetta and Shulman, 2002). Consistent with this, functional MRI shows greater activity in DLPFC for preparing to stop (Chikazoe et al. 2009b; Jahfari et al. 2010; Zandbelt and Vink, 2010; Swann et al. 2012; Zandbelt et al. 2012), although in most studies it was not possible to establish when this difference occurred, that is, during the task cue or go cue phase (but see Zandbelt et al. 2012). Here, we used ECoG to show that DLPFC was active for MS Go versus baseline, and moreover, we could elucidate the timing. We found that DLPFC was active following the task cue, the go cue, or both. One interpretation of this pattern could be that the DLPFC encodes the task goal following the task cue and then shows a reactivation following the go cue, perhaps reflecting attention to the cue or task goal retrieval, although other accounts are possible. Future studies using tasks that modulate different cognitive processes (e.g. by adjusting working memory load) would be useful to more clearly define the functional role of DLPFC in preparing to stop. At present, with the current task, we can only suggest that the pattern of activity that we observe is consistent with a role for the DLPFC in “representing the task goal,” which could relate to working memory, attention, other executive functions, or a combination of these.

The fact that significant DLPFC activity was observed around the go cue and not the task cue for some subjects could reflect a statistical thresholding issue and/or stronger reactivation or retrieval of the stopping rule as the response period draws near (i.e. when that information is most temporally pertinent). Notably, activity in the DLPFC preceded VLPFC in all subjects, which is consistent with different functional roles for these regions.

Functional Role of the VLPFC

Activity in the VLPFC had several features: 1) it occurred mainly after the go cue and more specifically around the time of the response; 2) it was later than activity in the DLPFC for all subjects; 3) it was greater for MS Go than NS Go for the 3 subjects who showed a behavioral RDE and not for the 2 who did not; 4) in 2 subjects, activity was more tightly related to RT than the expected SSD on a single-trial level. This pattern of data provides novel information about the temporal activations of the VLPFC in preparing to stop, and speaks to a debate about the functional role of the VLPFC in cognitive control. Whereas the right VLPFC is critical for the outright stopping of action (Aron et al. 2003; Chambers et al. 2006, 2007; Verbruggen et al. 2010), its underlying functional role could reflect the active suppression of a response (entirely or partially), that is, “action control” (Aron et al. 2007a; Jahfari et al. 2010), attentional monitoring (Hampshire et al. 2010; Sharp et al. 2010; Chatham et al. 2012), violation of a stopping expectancy (Zandbelt et al. 2012), or some combination of these (Chambers et al. 2009; Boehler et al. 2010; Chikazoe, 2010; Aron, 2011; Levy and Wagner, 2011).

The action control account proposes that there is slowing down on MS Go trials because of partial suppression of the emitted response, and that VLPFC implements this. This is analogous to a brake that is applied as movement occurs. This predicts VLPFC activity around the time of the motor response. In contrast, the attentional monitoring account proposes that the VLPFC is monitoring for the stop signal, which predicts earlier activity, around the time of SSD. Yet, we show that the VLPFC activity in all subjects is centered around the motor response, and in two subjects, the single trial analysis clearly showed that this activity was much more tightly locked to RT than the estimated SSD. Thus, the results are more compatible with the action control account.

Thus, while attentional monitoring is implicit in the stop signal paradigm, since one could not stop unless one detected a signal to stop, and while it is possible that VLPFC plays a role in such attentional monitoring (Corbetta and Shulman, 2002), the current data argue that VLPFC's role is not merely one of attentional monitoring. Instead, there is likely a sector (perhaps the more ventral and caudal aspect) that implements action control (Chikazoe et al. 2009a; Verbruggen et al. 2010; Levy and Wagner, 2011).

It is unclear as to how these results speak to the violation of stopping expectancy account. Based on fMRI, Zandbelt et al. (2012) proposed that VLPFC activity on go trials, in which stopping was not needed, although it was expected, reflected a violation of stopping expectancy. Such an interpretation could be consistent with our results if it is the motor response that instantaneously triggers the realization that the stop signal did not occur.

In any event, other data argue for a role for VLPFC in action control. For example, 2 studies using paired-pulse transcranial magnetic stimulation, with 1 coil on right VLPFC and the other on M1, showed that VLPFC had a suppressive effect on M1 for response control tasks (Buch et al. 2010; Neubert et al. 2010).

Thus, the VLPFC activity may reflect action control that is applied as the response is being made (i.e. braking). More specifically, the VLPFC, which is critical for stopping outright at the behavioral level, and putatively plays a role in implementing inhibitory control by projecting to the preSMA and the basal ganglia to cancel motor output (Aron et al. 2007b), may also be recruited proactively, when stopping is anticipated. Accordingly, the proactive recruitment of the VLPFC might have the consequence that if a response is initiated it is also dampened/slowed. Future studies could bolster this interpretation by, for example, using paired-pulse TMS to reveal if right VLPFC has inhibitory effects over the motor system when preparing to stop, or using electrical stimulation during task performance (Filevich et al. 2012).

Limitations

First, the data are from epilepsy subjects and may not generalize well to the healthy population. However, some of the findings clearly were consistent across all or most of the 5 subjects. Further, the findings corresponded well with extant fMRI studies of preparing to stop (Chikazoe et al. 2009b; Jahfari et al. 2010; Zandbelt and Vink, 2010; Swann et al. 2012; Zandbelt et al. 2012). Further, all subjects had epileptic foci outside of the reported regions. Second, the sample size, while acceptable by the standard of ECoG studies, is small, particularly for examining differences between individuals (i.e. whether or not behavioral slowing, the RDE, was present). Further, an important result, that single-trial gamma activity in the VLPFC tracked RT more tightly than expected SSD, could only be established in 2 subjects (probably because single-trial analyses are noisy, particularly in prefrontal regions due to increased variability in the timing and consistency of the neural response in this region compared with primary motor/sensory regions, for instance). This limits the strength of the conclusion that can be drawn. Third, for 2 of the subjects there was no slowing down (RDE) for MS versus NS Go trials, running against a common observation in normative populations (Chikazoe et al. 2009b; Jahfari et al. 2010; Zandbelt and Vink, 2010; Swann et al. 2012; Zandbelt et al. 2012). We suggest they were preparing to stop on every trial, and so were not differentially sensitive to the MS versus NS cues. However, their stopping and other task performance was otherwise satisfactory, and the absence of an RDE along with a null ECoG difference for MS Go versus NS Go, provided a nice control. Fourth, given the simplicity of our task design we could not specify the role of the DLPFC with any functional specificity beyond representing task goals. Fifth, some of the electrodes with similar functional properties were in different locations across subjects with reference to gyral/sulcal landmarks. Yet, it well established that there is high variability in the gyrification and cytoarchitecture of prefrontal regions (Amunts et al. 1999). Sixth, our current focus was on Go trials only, and not on stop trials. While it is theoretically interesting to examine DLPFC and VLPFC on outright stop trials, our previous report showed activity in both of these regions prior to the stop signal (see Fig. 6 Swann et al. 2012). This, combined with the fact that the go and stop signals occur in close temporal succession, and with a varying delay on different trials, makes it difficult to delineate different functional roles for these brain regions on stop trials. Hence, the focus on Go trials can better delineate the temporal profiles of these regions, as well as clarifying how they contribute to preparing-to-stop.

Conclusions and Implications

Our results speak to a long-standing debate about the functional specialization of the prefrontal cortex. Whereas some have argued that the primary functional role of the prefrontal cortex is working memory (with different sectors processing different kinds of information) (Goldman-Rakic, 1996), others have argued that different sectors perform different functions such as working memory, attention, and inhibitory control (Fuster, 1997). The current results argue for functional specialization, at least to a degree, with DLPFC representing the task goal, and VLPFC implementing action control.

Within VLPFC, we more specifically observed that the activity was around the time of the motor response, rather than around the time of the estimated stop signal. This, along with other results (Buch et al. 2010; Jahfari et al. 2010; Neubert et al. 2010) speaks in favor of a role for VLPFC in actively braking response tendencies. This action control mechanism over response tendencies, implemented via the VLPFC, could operate in concert with the basal ganglia and/or the primary motor cortex (Vink et al. 2005; Jahfari et al. 2010; Zandbelt and Vink, 2010; Zandbelt et al. 2012).

If our interpretation is correct it would suggest that task goal processing (represented in DLPFC) is used to implement action control (in the VLPFC). This predicts a causal influence of DLPFC over VLPFC, which could be tested in a future study of inter-regional communication. However, communication between these regions may not be direct. One possibility is that the preSMA may mediate communication between the 2. This account is supported by ECoG data from our previous report where the preSMA was active between the early DLPFC activity and the later VLPFC activity in one unique subject with subdural coverage in all these areas (Swann et al. 2012). This observed timing pattern is consistent with a paired-pulse TMS study using a response control task which showed preSMA's influence on primary motor cortex preceded VLPFC, and that VLPFC's influence on primary motor cortex depended on preSMA (Neubert et al. 2010). It is possible that the preSMA plays a role in task configuration (Rushworth et al. 2004), one aspect of which is translating the stopping rule into the action system. Thus, the preSMA may mediate the putative transfer of information from DLPFC to VLPFC when preparing to stop. Clearly, future research is required to develop a richer understanding of the functional roles and neural communication of different regions within this overall putative response control network.

In conclusion, the current data provide a more mechanistic understanding of human self-control. They point to a dissociation between the task goal and the implementation of action control in different sectors of prefrontal cortex, and they provide insight into how people respond more cautiously when they might have to stop.

Supplementary Material

Supplementary material can be found at:http://www.cercor.oxfordjournals.org/

Funding

Funding is gratefully received from NIH grant DA026452 to A.R.A. N.S. was supported by an NSF Graduate Student Fellowship, an NIMH training grant via the Institute for Neural Computation at UCSD, and an NSF Gk-12 fellowship. N.T. was supported by a Clinical and Translational Award KL2 RR0224149 from the National Center for Research Resources and by the Vivian Smith Foundation.

Notes

The authors thank John Serences for his helpful comments on the manuscript. We also thank our patients for their co-operation with the study, the nurses and the EEG technicians at the epilepsy monitoring unit at Memorial Hermann Hospital for facilitating patient data collection, and Drs. Jeremy Slater, Giridhar Kalamangalam, and Omotola Hope for contributing patients to this study. Conflict of Interest: None declared.

References

Amunts
K
Schleicher
A
Burgel
U
Mohlberg
H
Uylings
HB
Zilles
K
Broca's region revisited: cytoarchitecture and intersubject variability
J Comp Neurol
 , 
1999
, vol. 
412
 (pg. 
319
-
341
)
Aron
AR
From reactive to proactive and selective control: developing a richer model for stopping inappropriate responses
Biol Psychiatry
 , 
2011
, vol. 
69
 (pg. 
e55
-
e68
)
Aron
AR
Behrens
TE
Smith
S
Frank
MJ
Poldrack
RA
Triangulating a cognitive control network using diffusion-weighted magnetic resonance imaging (MRI) and functional MRI
J Neurosci
 , 
2007a
, vol. 
27
 (pg. 
3743
-
3752
)
Aron
AR
Durston
S
Eagle
DM
Logan
GD
Stinear
CM
Stuphorn
V
Converging evidence for a fronto-basal-ganglia network for inhibitory control of action and cognition
J Neurosci
 , 
2007b
, vol. 
27
 (pg. 
11860
-
11864
)
Aron
AR
Fletcher
PC
Bullmore
ET
Sahakian
BJ
Robbins
TW
Stop-signal inhibition disrupted by damage to right inferior frontal gyrus in humans
Nat Neurosci
 , 
2003
, vol. 
6
 (pg. 
115
-
116
)
Aron
AR
Poldrack
RA
Cortical and subcortical contributions to stop signal response inhibition: role of the subthalamic nucleus
J Neurosci
 , 
2006
, vol. 
26
 (pg. 
2424
-
2433
)
Boehler
CN
Appelbaum
LG
Krebs
RM
Chen
LC
Woldorff
MG
The role of stimulus salience and attentional capture across the neural hierarchy in a stop-signal task
PLoS One
 , 
2011
, vol. 
6
 pg. 
e26386
 
Boehler
CN
Appelbaum
LG
Krebs
RM
Hopf
JM
Woldorff
MG
Pinning down response inhibition in the brain—conjunction analyses of the stop-signal task
Neuroimage
 , 
2010
, vol. 
52
 (pg. 
1621
-
1632
)
Buch
ER
Mars
RB
Boorman
ED
Rushworth
MFS
A network centered on ventral premotor cortex exerts both facilitatory and inhibitory control over primary motor cortex during action reprogramming
J Neurosci
 , 
2010
, vol. 
30
 (pg. 
1395
-
1401
)
Canolty
R
Soltani
M
Dalal
S
Edwards
E
Dronkers
N
Nagarajan
S
Kirsch
H
Barbaro
N
Knight
RT
Spatiotemporal dynamics of word processing in the human brain
Front Neurosci
 , 
2007
, vol. 
1
 (pg. 
185
-
196
)
Chambers
CD
Bellgrove
MA
Gould
IC
English
T
Garavan
H
McNaught
E
Kamke
M
Mattingley
JB
Dissociable mechanisms of cognitive control in prefrontal and premotor cortex
J Neurophysiol
 , 
2007
Chambers
CD
Bellgrove
MA
Stokes
MG
Henderson
TR
Garavan
H
Robertson
IH
Morris
AP
Mattingley
JB
Executive brake failure following deactivation of human frontal lobe
J Cogn Neurosci
 , 
2006
, vol. 
18
 (pg. 
444
-
455
)
Chambers
CD
Garavan
H
Bellgrove
MA
Insights into the neural basis of response inhibition from cognitive and clinical neuroscience
Neurosci Biobehav Rev
 , 
2009
, vol. 
33
 (pg. 
631
-
646
)
Chao
HH
Luo
X
Chang
JL
Li
CS
Activation of the pre-supplementary motor area but not inferior prefrontal cortex in association with short stop signal reaction time—an intra-subject analysis
BMC Neurosci
 , 
2009
, vol. 
10
 pg. 
75
 
Chatham
CH
Claus
ED
Kim
A
Curran
T
Banich
MT
Munakata
Y
Cognitive control reflects context monitoring, not motoric stopping, in response inhibition
PLoS One
 , 
2012
, vol. 
7
 pg. 
e31546
 
Chikazoe
J
Localizing performance of go/no-go tasks to prefrontal cortical subregions
Curr Opin Psychiatry
 , 
2010
, vol. 
23
 (pg. 
267
-
272
)
Chikazoe
J
Jimura
K
Asari
T
Yamashita
K
Morimoto
H
Hirose
S
Miyashita
Y
Konishi
S
Functional dissociation in right inferior frontal cortex during performance of go/no-go task
Cereb Cortex
 , 
2009a
, vol. 
19
 (pg. 
146
-
152
)
Chikazoe
J
Jimura
K
Hirose
S
Yamashita
K
Miyashita
Y
Konishi
S
Preparation to inhibit a response complements response inhibition during performance of a Stop-Signal Task
J Neurosci
 , 
2009b
, vol. 
29
 (pg. 
15870
-
15877
)
Conner
C
Ellmore
TM
Pieters
TA
Disano
M
Tandon
N
Variability of the relationship between electrophysiology and BOLD-fMRI across cortical regions in humans
J. Neurosci.
 , 
2011
, vol. 
31
 (pg. 
12855
-
12865
Forthcoming
Corbetta
M
Shulman
GL
Control of goal-directed and stimulus-driven attention in the brain
Nat Rev Neurosci
 , 
2002
, vol. 
3
 (pg. 
201
-
215
)
Dale
AM
Fischl
B
Sereno
MI
Cortical surface-based analysis. I. Segmentation and surface reconstruction
Neuroimage
 , 
1999
, vol. 
9
 (pg. 
179
-
194
)
Duann
JR
Ide
JS
Luo
X
Li
CS
Functional connectivity delineates distinct roles of the inferior frontal cortex and presupplementary motor area in stop signal inhibition
J Neurosci
 , 
2009
, vol. 
29
 (pg. 
10171
-
10179
)
Edwards
E
Soltani
M
Deouell
LY
Berger
MS
Knight
RT
High gamma activity in response to deviant auditory stimuli recorded directly from human cortex
J Neurophysiol
 , 
2005
, vol. 
94
 (pg. 
4269
-
4280
)
Filevich
E
Kuhn
S
Haggard
P
Negative motor phenomena in cortical stimulation: implications for inhibitory control of human action
Cortex
 , 
2012
Flinker
A
Chang
EF
Kirsch
HE
Barbaro
NM
Crone
NE
Knight
RT
Single-trial speech suppression of auditory cortex activity in humans
J Neurosci
 , 
2010
, vol. 
30
 (pg. 
16643
-
16650
)
Forstmann
BU
Anwander
A
Schafer
A
Neumann
J
Brown
S
Wagenmakers
EJ
Bogacz
R
Turner
R
Cortico-striatal connections predict control over speed and accuracy in perceptual decision making
Proc Natl Acad Sci USA
 , 
2010
, vol. 
107
 (pg. 
15916
-
15920
)
Forstmann
BU
Keuken
MC
Jahfari
S
Bazin
P-L
Neumann
J
Schäfer
A
Anwander
A
Turner
R
Cortico-subthalamic white matter tract strength predicts interindividual efficacy in stopping a motor response
Neuroimage
 , 
2012
, vol. 
60
 (pg. 
370
-
375
)
Fries
P
A mechanism for cognitive dynamics: neuronal communication through neuronal coherence
Trends Cogn Sci
 , 
2005
, vol. 
9
 (pg. 
474
-
480
)
Fuster
JM
The prefrontal cortex : anatomy, physiology, and neuropsychology of the frontal lobe
 , 
1997
Philadelphia
Lippincott-Raven
Goldman-Rakic
PS
Cellular and circuit basis of working memory in prefrontal cortex of nonhuman primates
Prog Brain Res
 , 
1990
, vol. 
85
 (pg. 
325
-
335
Discussion 335–326
Goldman-Rakic
PS
The prefrontal landscape: implications of functional architecture for understanding human mentation and the central executive
Philos Trans R Soc Lond B Biol Sci
 , 
1996
, vol. 
351
 (pg. 
1445
-
1453
)
Hampshire
A
Chamberlain
SR
Monti
MM
Duncan
J
Owen
AM
The role of the right inferior frontal gyrus: inhibition and attentional control
NeuroImage
 , 
2010
, vol. 
50
 (pg. 
1313
-
1319
)
Jahfari
S
Stinear
CM
Claffey
M
Verbruggen
F
Aron
AR
Responding with restraint: what are the neurocognitive mechanisms?
J Cog Neurosci
 , 
2010
, vol. 
22
 (pg. 
1479
-
1492
)
Kopell
N
Ermentrout
GB
Whittington
MA
Traub
RD
Gamma rhythms and beta rhythms have different synchronization properties
Proc Natl Acad Sci USA
 , 
2000
, vol. 
97
 (pg. 
1867
-
1872
)
Kopell
N
Kramer
MA
Malerba
P
Whittington
MA
Are different rhythms good for different functions?
Front Hum Neurosci
 , 
2010
, vol. 
4
 pg. 
187
 
Leventhal
DK
Gage
GJ
Schmidt
R
Pettibone
JR
Case
AC
Berke
JD
Basal ganglia beta oscillations accompany cue utilization
Neuron
 , 
2012
, vol. 
73
 (pg. 
523
-
536
)
Levy
BJ
Wagner
AD
Cognitive control and right ventrolateral prefrontal cortex: reflexive reorienting, motor inhibition, and action updating
Ann NY Acad Sci
 , 
2011
, vol. 
1224
 (pg. 
40
-
62
)
Logan
GD
Schachar
RJ
Tannock
R
Impulsivity and inhibitory control
Psychological Science
 , 
1997
, vol. 
8
 (pg. 
60
-
64
)
Miller
KJ
denNijs
M
Shenoy
P
Miller
JW
Rao
RPN
Ojemann
JG
Real-time functional brain mapping using electrocorticography
Neuroimage
 , 
2007a
, vol. 
37
 (pg. 
504
-
507
)
Miller
KJ
Leuthardt
EC
Schalk
G
Rao
RP
Anderson
NR
Moran
DW
Miller
JW
Ojemann
JG
Spectral changes in cortical surface potentials during motor movement
J Neurosci
 , 
2007b
, vol. 
27
 (pg. 
2424
-
2432
)
Muller
NG
Knight
RT
The functional neuroanatomy of working memory: contributions of human brain lesion studies
Neuroscience
 , 
2006
, vol. 
139
 (pg. 
51
-
58
)
Nachev
P
Kennard
C
Husain
M
Functional role of the supplementary and pre-supplementary motor areas
Nat Rev Neurosci
 , 
2008
, vol. 
9
 (pg. 
856
-
869
)
Neubert
FX
Mars
RB
Buch
ER
Olivier
E
Rushworth
MF
Cortical and subcortical interactions during action reprogramming and their related white matter pathways
Proc Natl Acad Sci USA
 , 
2010
, vol. 
107
 (pg. 
13240
-
13245
)
Olesen
PJ
Westerberg
H
Klingberg
T
Increased prefrontal and parietal activity after training of working memory
Nat Neurosci
 , 
2004
, vol. 
7
 (pg. 
75
-
79
)
Petrides
M
The role of the mid-dorsolateral prefrontal cortex in working memory
Exp Brain Res
 , 
2000
, vol. 
133
 (pg. 
44
-
54
)
Ray
NJ
Brittain
JS
Holland
P
Joundi
RA
Stein
JF
Aziz
TZ
Jenkinson
N
The role of the subthalamic nucleus in response inhibition: evidence from local field potential recordings in the human subthalamic nucleus
Neuroimage
 , 
2012
, vol. 
60
 (pg. 
271
-
278
)
Ray
S
Crone
NE
Niebur
E
Franaszczuk
PJ
Hsiao
SS
Neural correlates of high-gamma oscillations (60–200 Hz) in macaque local field potentials and their potential implications in electrocorticography
J Neurosci
 , 
2008
, vol. 
28
 (pg. 
11526
-
11536
)
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
)
Saad
ZS
Ropella
KM
DeYoe
EA
Bandettini
PA
The spatial extent of the BOLD response
Neuroimage
 , 
2003
, vol. 
19
 (pg. 
132
-
144
)
Scheeringa
R
Fries
P
Petersson
KM
Oostenveld
R
Grothe
I
Norris
DG
Hagoort
P
Bastiaansen
MC
Neuronal dynamics underlying high- and low-frequency EEG oscillations contribute independently to the human BOLD signal
Neuron
 , 
2011
, vol. 
69
 (pg. 
572
-
583
)
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
 , 
2010
, vol. 
107
 (pg. 
1
-
6
)
Swann
N
Tandon
N
Canolty
R
Ellmore
TM
McEvoy
LK
Dreyer
S
DiSano
M
Aron
AR
Intracranial EEG reveals a time- and frequency-specific role for the right inferior frontal gyrus and primary motor cortex in stopping initiated responses
J Neurosci
 , 
2009
, vol. 
29
 (pg. 
12675
-
12685
)
Swann
NC
Cai
W
Conner
CR
Pieters
TA
Claffey
MP
George
JS
Aron
AR
Tandon
N
Roles for the pre-supplementary motor area and the right inferior frontal gyrus in stopping action: electrophysiological responses and functional and structural connectivity
Neuroimage
 , 
2012
, vol. 
59
 (pg. 
2860
-
2870
)
Verbruggen
F
Aron
AR
Stevens
MA
Chambers
CD
Theta burst stimulation dissociates attention and action updating in human inferior frontal cortex
Proc Natl Acad Sci USA
 , 
2010
, vol. 
107
 (pg. 
13966
-
13971
)
Verbruggen
F
Logan
GD
Models of response inhibition in the stop-signal and stop-change paradigms
Neurosci Biobehav Rev
 , 
2009
, vol. 
33
 (pg. 
647
-
661
)
Vink
M
Kahn
RS
Raemaekers
M
van den Heuvel
M
Boersma
M
Ramsey
NF
Function of striatum beyond inhibition and execution of motor responses
Hum Brain Mapp
 , 
2005
, vol. 
25
 (pg. 
336
-
344
)
Zandbelt
BB
Bloemendaal
M
Neggers
SFW
Kahn
RS
Vink
M
Expectations and violations: delineating the neural network of proactive inhibitory control
Hum Brain Mapp
 , 
2012
Zandbelt
BB
Vink
M
On the role of the striatum in response inhibition
PLoS One
 , 
2010
, vol. 
5
 pg. 
e13848