Electrophysiological correlates of response inhibition in children and adolescents with ADHD: influence of gender, age, and previous treatment history

Deficits in response inhibition may be at the core of the cognitive syndrome in ADHD. Here, inhibitory control mechanisms were studied in 36 ADHD‐combined type and 30 healthy children by exploring the event‐related brain activity during the Stop Signal task. The influence of age, gender, and previous treatment history was evaluated. The ADHD group showed reduced N200 wave amplitudes. For successful inhibitions, the N200 reduction was greatest over right inferior frontal scalp, and only the control group showed a success‐related enhancement of such right frontal N200. Source analysis identified a source of the N200 group effect in right dorsolateral prefrontal cortex. Finally, a late positive wave to failed inhibitions was selectively reduced only in treatment‐naïve ADHD children, suggesting that chronic stimulants may normalize late conscious error recognition. Both effects were independent of gender and age.

Keywords: Event‐related potentials; Stop signal task; ADHD; N200; NoGo‐P3; LPW; Stimulants

Attention Deficit‐Hyperactivity Disorder (ADHD) is a common behavioral syndrome, estimated to occur in 3%–7% of school‐aged children worldwide ([3]). Symptoms include low levels of attention and concentration and high levels of activity, distractibility, impulsivity, and the inability to inhibit actions ([2]). One of the most influential theoretical models of ADHD posits that deficits in inhibitory control are the core symptoms in ADHD ([3], [40]). The critical neuroanatomical substrate for inhibitory control would be a circuit connecting the basal ganglia to right prefrontal cortex, with its structural/functional integrity being impaired in ADHD ([7], [8]). Other theoretical models emphasize deficits in cognitive control mechanisms, including both conflict monitoring and error processing ([30]), a dysfunction in the regulation of motivation and reward, with a preference for immediate versus delayed rewards (delay aversion, e.g., [47]) or deficits in state/arousal regulation (cognitive‐energetic model, e.g. [45]).

In support of the inhibitory control model, impairment has been found in children with ADHD on laboratory tasks that tap into the inhibitory control symptoms of ADHD (impulsiveness and inattention), such as the Continuous Performance Test or CPT and the Stop Signal task or SST (for reviews, see [23], [27], [29]). Both the CPT and the SST are varieties of "Go/NoGo" tasks, requiring subjects to occasionally inhibit an ongoing action or response. In the C‐X version of the CPT, subjects are presented with rapid sequences of letters. Twenty percent of the stimuli are the letter C, which is followed 50% of the time by an X and 50% of the time by another letter (e.g., R, V, T, etc). Subjects are instructed to withhold their response when they see an X following a C ("NoGo stimulus") but to respond when any other letter follows the letter C (e.g, C‐R, C‐V, etc., "Go stimuli"; [3]). The SST imposes greater demands of inhibitory control, in that subjects have to be prepared to withdraw a response on each trial instead of only on 20% of the trials, and probability of the NoGo stimulus is 25% rather than 50%, creating the need of inhibiting a more prepotent response ([25], [40]). In a visual version of the SST (e.g., [34], [36]), subjects are presented with either the letter A or the letter B and are instructed to respond to the A with one button and the B with a second button. On 25% of the trials a Stop signal (the letter S) follows the A or B by a variable time interval (e.g., 200–600 ms, Stop signal intervals), and the subject must withhold his/her response on that trial. For typical healthy subjects, for the shorter Stop signal intervals (e.g., 200–400 ms) it is easier to inhibit responding, whereas at the longer Stop signal intervals (400–600 ms) it is substantially more difficult to do so, and the probability of inhibition is much lower. A number of studies have shown that individuals with ADHD perform more poorly on the SST, including showing a significantly different slope of the inhibitory function (i.e., the function of the probability of successfully inhibiting as a function of the Stop signal interval) when compared to age‐matched control subjects, due to a difficulty withholding a response regardless of the Stop signal latency (for a meta‐analyses of SST studies, see [23]).

Neural mechanisms underlying inhibitory processes can be studied with a high degree of temporal resolution by recording event‐related potentials (ERPs) from the scalp. In ERP studies of the CPT, the frontally maximal N200 wave, peaking around 200 ms, has been shown to have greater amplitude for NoGo relative to Go trials ([12], [19], [46]). It has been proposed that the NoGo N200 indexes an early mechanism of inhibitory control that is a reflection of a "red flag" signal generated in prefrontal cortex to trigger the inhibitory process ([17], [19]). Note, however the NoGo N200 in a Go/NoGo task has been recently interpreted as reflecting cognitive control rather then response inhibition ([30]). In the SST, a right frontal‐maximal N200 evoked by the Stop signal has been reported both in healthy children ([9], [36]) and young adults ([42]). A critical prediction of the inhibitory control model of ADHD is that if the frontal NoGo N200 indexes the initiation of the inhibitory process, impairment in individuals with ADHD should be seen. ERP studies of the CPT comparing children with ADHD and healthy children have produced mixed results. Children with ADHD have been found to have reduced amplitude ([32]; but only in children with comorbid ODD), increased amplitude ([46]), or the same amplitude ([10], [32]). This lack of consistency may be due to other attentional components (sustained attention, spatial orienting) being at play during the CPT. In the SST, a marked reduction of the right frontal N200 wave to the Stop signal in ADHD relative to healthy children has been found ([1], [9], [36]). A right inferior frontal generator has been suggested for the Stop signal N200 ([24]) consistent with functional magnetic resonance imaging (fMRI) studies showing activation of right middle‐inferior dorsolateral prefrontal cortex (DLPFC) in healthy adolescents and adults in Go/NoGo and the SST ([15], [20], [37]) and the absence of activation in this region in ADHD children and adolescents ([37], [38]).

A positive‐polarity ERP component (NoGo‐P3, peaking around 300 ms), has also been found to be associated to response inhibition in the CPT and the SST, with greater amplitude over the midline central scalp region for NoGo than Go trials (e.g., [14]). In the SST, a frontocentral NoGo‐P3 has also been found to have greater amplitude for successful than failed inhibitions in healthy subjects ([24], [31], [42]). It has been therefore proposed that the frontocentral NoGo‐P3 reflects monitoring of the successful outcome of the inhibitory process by frontomedial cortical structures (e.g., [24]). A prediction of such functional interpretation of the NoGo‐P3 and of its possible relevance to ADHD is that its amplitude should be abnormal in ADHD. Indeed in Go/NoGo studies using the CPT comparing ADHD to healthy children, the NoGo‐P3 has been found to be reduced in children with ADHD ([10], [32]). In the SST, the success‐enhanced NoGo‐P3 was also diminished in children with ADHD relative to healthy children ([24], [31]). A dorsal anterior cingulate cortex (dACC) generator has been proposed for the NoGo‐P3 ([10]), and the success‐related NoGo‐P3 ([24]). Such candidate EEG sources are in agreement with activations in the dACC reported in healthy adults and adolescents during Go/NoGo tasks using fMRI ([22]) and are consistent with dACC general role in monitoring conflict and in cognitive control ([5]). Lack of normal activation of dACC in ADHD adolescents and children has been reported during the Counting Stroop Task ([6]) and in response to failed inhibitions in the SST ([35]). Taken together, the NoGo‐P3 evidence emphasizes the need for revising the inhibitory control model ([3], [36]) to incorporate the contribution of cognitive control operations at play during the SST ([24], [30]).

A third ERP component associated with response inhibition in the SST is a late positive wave (LPW, 500–700 ms) maximal over posterior scalp and with greater amplitude for failed than success inhibitions ([31], [42]). This component appears very similar to the error positivity (Pe), which has been associated with the process of late conscious error recognition (e.g., [11], [42]). A prediction of such functional interpretation of the LPW is that its amplitude may also be abnormal in individuals with ADHD. One study comparing ADHD to healthy children reported decreased amplitude of the LPW in children with ADHD ([31]). Possible sources of the LPW could be located in cingulate cortex, consistent with fMRI studies of the SST reporting activation of the mid‐dorsal and posterior cingulate gyri in response to failed inhibitions, and lack of activation in such structures in children ([35]) and adolescents ([38]) with ADHD.

The present study aimed at replicating and extending the previous findings of abnormal ERP activity during the SST in ADHD in a considerably larger cohort of children with ADHD‐combined type. Confounds of previous research addressed here were comorbidity, previous treatment status, gender and age of the participant, and level of intellectual functioning. First, published ERP studies of the SST in children with ADHD did not exclude comorbid symptoms (e.g., [9], [31]). Second, all previous ERP investigations of response inhibition in children with ADHD included cohorts formed exclusively or prevalently by subjects chronically treated with stimulants ([9], [24], [32], [31], [36], [46]), confounding the effects of long‐term stimulant treatment, and/or acute stimulant withdrawal. Furthermore, previous studies did not assess the effects of gender, having used only boys ([9], [10], [36], [46]). Gender differences in the neural correlates of inhibitory control may be important to evaluate, because behavioral/clinical manifestations and comorbidity of ADHD appear to vary between girls and boys (e.g., [21]). In addition, several previous studies had only included young children (8–11 years old; [9], [24], [31], [36], [46]). Finally, not all previous studies had carefully controlled their groups by level of intellectual functioning and may have included children with learning disorder ([36]).

To clarify the influence of comorbidity, two groups of individuals with ADHD were studied: those with at least 1 year of successful stimulant treatment and those who were treatment naïve. To evaluate gender differences, the sample included both male and female subjects. To assess the effect of age, the present study included both children and adolescents (age 9–15 year old). Finally, all subjects were screened for levels of intellectual abilities. Based on our prior studies ([24], [36]), we hypothesized that children with ADHD would show decreased right frontal N200 amplitude and an altered NoGo‐P3 in response to the Stop signal relative to controls, and such differences would not be affected by gender or history of prior stimulant treatment, being therefore ascribable to impaired response inhibition and cognitive control mechanisms in individuals with ADHD.

Participants were right‐handed children and adolescents aged 9 to 15 years of both genders. The study groups were healthy controls (n=30; 16 boys, 14 girls) and participants meeting criteria for ADHD‐Combined Type (ADHD‐C: n=36; 25 boys, 11 girls). These were subdivided in ADHD‐C with history of chronic stimulant treatment (ADHD‐RX: n=20) and ADHD‐C with no history of psychotropic medication treatment (ADHD‐TN: n=16). Written informed consent from a parent and assent from the child were obtained according to the Institutional Review Board of the Health Science Center at San Antonio.

Individuals with ADHD met Diagnostic Interview for Children‐Version IV‐Parent version (DISC‐IV‐P) criteria for ADHD‐C, could meet criteria for oppositional defiant disorder, but not meet criteria for conduct disorder or any anxiety, tic, or affective disorder. Healthy controls could not meet criteria for any psychiatric disorders or any history of past treatment with psychotropic medication. No participants in any group had a history of neurological conditions or symptoms, such as head injury, loss of consciousness, motor or sensory loss, nor had they a history of substance or alcohol abuse. Children with ADHD were not taking any medications for any condition other than ADHD at the time of the study. Participants in the ADHD‐RX group took a stimulant‐only medication (amphetamine or methylphenidate) for at least 1 year prior to the study (range of treatment 1–9 years, mean=5±3 years), and were withdrawn from treatment 24 h before the baseline assessment and again before the ERP session.

In addition to the DISC, the Conners Global Index was used. Criteria for inclusion were based on parent and teacher ratings of the Restless/Impulsive score (RI). For the healthy group, inclusion criteria were an RI equal or less than one standard deviation (SD) of the mean for child's age and gender. For the group of children with ADHD‐TN, inclusion criteria were an RI equal or greater than 1.5 SD of the mean for child's age and gender. Finally, for the group of children with ADHD‐RX, Teacher RI rating on stimulants were equal or less than 1 SD of the mean for child's age and gender, whereas Parent RI rating, measured in a period when the child was off medication in the last 6 months, was equal or greater than 1.5 SD of the mean for child's age and gender. Current cognitive functioning was measured with the Differential Abilities Scales (DAS). All participants were required to have a general conceptual ability (GCA) score >85. The Wechsler Individual Achievement Test Version II (WIAT‐II) was administered to assess reading, writing, and mathematics. To rule out learning disorders, all participants had to have reading and mathematic standard scores on the WIAT‐II that were within one standard deviation of their full scale IQ on the Differential Abilities Scale (see Table 1).

1   Demographics and Behavioral Data

1 Notes. For the neuropsychological variables , * ADHD‐TN** Controls *** ADHD‐TN >other two groups. 
 For the SST performance data, *Controls **Controls>ADHD groups. ADHD‐TN: ADHD Treatment‐Naïve group; ADHD‐RX: ADHD Chronic Treatment group.

Participants sat with their eyes 50 cm from a computer monitor presenting a visual version of the SST ([36]). They discriminated between the letters "A" or "B" (Go stimuli) by responding with the index finger of the left or right hand. Stimuli were flashed for 150 ms slightly above a fixation dot. The Go stimuli were followed by a nonfixed and unpredictable time delay randomly varying within one of the following equal probability subranges: 200–300, 300–400, 400–500, and 500–600 ms (Stop signal interval). Following that, either the letter "S" appeared on the screen slightly below the fixation dot for 150 ms (Stop Signal trials, 25%) or the fixation dot was present on the screen alone for 150 ms (Go trials, 75%). Subjects were instructed to inhibit their response when seeing an S. Each trial ended with a random intertrial interval (600–900 ms). Therefore the overall interval between successive Go stimuli varied between 950 and 1650 ms independent of an intervening button press.

Each experimental run included 72 Go and 24 Stop Signal trials; there was a total of 10 runs, each lasting about 3 min. After each run, subjects were given about 30 s to rest. After that, they were asked if they needed additional time, which was allowed as requested until they were ready to resume the task. In addition, a longer resting pause (about 5 min) was provided halfway through the session. Subjects received one to three practice runs before the acquisition of data to ensure they understood the task and were capable of performing above chance. Subjects were told that it was important both to respond accurately to the Go signal (and not miss very many) and not to slow down excessively. To discourage strategic slowing, an online adjustment of difficulty of inhibitory performance was implemented. For each subject and run, if the mean Go RT in a run was longer or shorter than 600 ms (for example, 670 ms), all Stop signal intervals in the following run were increased or decreased by the amount of time over or under 600 ms (in our example, +70 ms; see also [36]). This discouraged deliberate slowing to catch all the Stop signals.

The following behavioral parameters were measured: RT and percent error in the Go trials, probability of inhibition [p(I)] for each of the four 100‐ms SOA subranges, and Stop signal reaction time (SSRT) for each SOA subrange ([25], [36]). The [p(I)] was calculated for each of the four SOAs. For example, if a child successfully inhibited 25 of the 30 Stop signals with an SOA of 250 ms, the [p(I)] for that SOA would be.83. It is generally easy to inhibit a response when the Stop signal occurs close to the Go stimulus, and more difficult (resulting in a lower [p(I)]) when the Stop signal occurs far from the Go signal. A slope was then calculated using the [p(I)] values for the four Stop signal SOAs. The SSRT, which provides a measure of latency of the inhibitory process, was then determined according to the method of [25]. The SSRT is presumed to begin when the Stop signal occurs. It is calculated from the [p(I)] and the distribution of RT to the Go signal. The subject's reaction times to the Go signal are normally distributed. If a subject inhibits successfully 80% of the time at the 200‐ms SOA, it is assumed that the stop process is fast enough to interrupt all the Go responses between the onset of the Stop signal and the Go RT that is 80% of the distance from the longest RT. The RTs are rank‐ordered from the longest to the shortest, and the RT that is 80% down the list is selected. The SSRT is calculated by subtracting that SOA (200 ms) from that RT; this is done for each of the four Stop signal delays. The mean of these four values is the final SSRT for the subject.

A statistic (relative finishing time, or ZRFT) was also used to adjust for variation in Go reaction time, variation in SSRT, and probability of inhibition between the groups ([25]).

Brain electrical activity was recorded using a 64‐channel cap (Electrocap, Inc.) referenced to the right mastoid. Amplifier settings were bandpass=0.01–100 Hz, gain=104, sampling rate=400 Hz, impedances <5k. Epochs of the EEG contaminated by eyeblinks were rejected off‐line. ERP averages were obtained for successful inhibitions (SI) and failed inhibitions (FI), time‐locking to Stop signal onset. For the FI analysis, trials in which button presses preceded the Stop signal (premature responses) were excluded. All channels were re‐referenced to the average of the two mastoid electrodes, and smoothed with a nine‐point running average filter. Grand averages were calculated across subjects for each trial type and group. To help isolate effects of interest, within‐ and between‐group difference waves were calculated. To facilitate visualization, scalp voltage topographic distributions were obtained using spherical spline interpolation ([33]).

Due to the short interval between the Go and Stop stimuli, the elicited ERP responses overlapped in time, distorting the final ERP averages ([50]). To correct for this differential overlap distortion problem, ERP subaverages for the successful and failed Stop signals were obtained for each of the four 100‐ms time‐delay subranges (200–300 ms, 300–400 ms, etc.) for each subject. Then, separately for each condition, these four subaverages were collapsed together in an equally weighted way (25% for each subaverage), thereby better equating the overlap from the Go‐event ERPs on the successful and failed Stop‐signal ERPs (see [36]).

Based on previous findings on N200, NoGo‐P3, and LPW in the SST ([24], [31], [36]) after inspection of grand‐average waveforms and scalp topography distributions for each trial type and various difference waves, time windows were selected around N200 (170–220 ms), NoGo‐P3 (280–350 ms), and LPW (480–700 ms), and mean voltage amplitudes in such component‐specific windows were used for statistical analysis.

Regions of interest (ROIs) were selected, by averaging together neighbor electrode sites. For the N200 analysis, four ROIs were chosen, two anterior inferior (sites F7P‐F8P, C5a‐C6a, T3′‐T4′ and PA1a‐PA2a) and two posterior inferior (P3i‐P4i, T35i‐T46i, TO1‐TO2, O1′‐O2′). For the NoGo‐P3 window, four ROIs were chosen, two frontocentral (F3S‐4S, FC1‐ FC2, C1a‐C2a and C3a‐C4a), whereas the posterior ROIs collapsed four parietal scalp sites (C1p‐C2p, P1′‐P2′, P3a‐P4a, and PO1‐PO2). For the LPW, two frontocentral ROIs were selected (sites FC1‐FC2, C1a‐C2a, F3i‐F4i, C3a‐C4a), along with two posterior ROIs (P1′‐P2′, TO1‐TO2, PO1‐PO2, O1′‐O2′ and O1i‐O2i).

For the first three time windows, mixed design repeated measures ANOVAs were used, with the between‐subjects factor being group (ADHD vs. Control) and within‐subjects factors being Trial type, Anterior‐Posterior, and Hemisphere. For all time windows, additional analyses were also carried out to assess the influence of intellectual status, age, gender, and previous stimulant history on each relevant ERP wave sampled. For all analyses, the critical p value was set at.05 (degrees of freedom were corrected with the Greenhouse–Geisser epsilon method; [16]). When appropriate, effect sizes were calculated for the Group effects using the Cohen d' method. In addition, all analyses were repeated using normalized voltage values (square root of the mean of squares method), to correctly interpret scalp topography differences between groups ([28]). Global analyses were followed by restricted between‐ and within‐group analyses for interpretation of significant findings.

The groups were similar in age and verbal IQ or non‐verbal IQ, but there was a significant difference in GCA, F(2,63)=4.4, p=.016. By Tukey test, the ADHD‐TN group had a lower GCA than the control group. There also was a main effect of diagnosis on WIAT Reading, F(2,63)=4.3, p=.017, and WIAT Math, F(2,63)=4.5, p=.015. For both variables, the Tukey test showed that the ADHD‐TN was lower than the control group. Although these scores statistically differ, they are all well within the average range and indicate adequate functioning in all areas. As such, these findings may not be clinically important. As expected, there were main effects of diagnosis on the Parent Conners R/I scale, F(2,63)=150, p=.0008, and Global scales, F(2,63)=149, p=.0009; for both scales, multiple comparisons showed that the control groups were significantly lower than the two ADHD groups. For the teacher ratings, there was a main effect of diagnosis on both the R/I, F(2,61)=53.6, p=.003, and the Global scale, F(2,61)=59.0, p=.0012, with Tukey tests showing that the ADHD‐TN group was higher than the other two groups. A summary of clinical and demographic variables is presented in Table 1.

The groups were not different in accuracy to the Go task. They did differ in mean RT to the Go signal, with the control group being slower than the ADHD groups, F(2,63)=4.1, p=.02. The two ADHD groups had a more variable RT to the Go stimuli than the controls, F(2,63)=4.7, p=.02. The ADHD groups had more failed inhibitions following the Stop signal than controls, F(2,63)=5.3, p=.030 (ADHD‐RX=28.4%, ADHD‐TN=29.1%, controls=24%). They also had a larger proportion of failed inhibitions preceding the Stop signal, F(2,63)=9.6, p=.003 (ADHD‐RX=28.21%, ADHD‐TN=29.1%, controls=19.2%). As a result of greater number of failed inhibitions, ADHD groups had a flatter slope of the inhibitory function, F(2,63)=5.1, p=.01. There was a marginally significant difference in SSRT between controls and ADHD groups, with shorter SSRT in controls (p=.06). The ZRFT score, taking into account group variability in global Go RT, SSRT, and slope of the inhibitory function, was significantly different between controls and ADHD groups, F(2,63)=4.18, p=.020. In summary, behavioral differences in inhibitory control measures were present between controls and the two ADHD subgroups, but not between ADHD‐RX and ADHD‐TN. A summary of behavioral findings in the SST is provided in lower part of Table 1.

N200 wave (170–220 ms). ERP waveforms and topographic maps illustrating the N200 effects are shown in Figures 1 and 2. In the global analysis, the main effect of Group was significant, F(1,64)=5.0, p=.029, with greater N200 amplitudes in the control than ADHD group (−2.44±0.32 μV vs. −1.47±0.29 μV). This effect was qualified by the significance of the interactions Group × Hemisphere, F(1,64)=11.5, p=.001, and Group × Trial × Hemisphere, F(1,64)=4.02, p=.049. In addition, the interaction Group × Trial Type × Anterior Posterior Topography was marginally significant, F(1,63)=3.5, p=.067. To explain these interactions, restricted within‐group and between‐groups ANOVAs were carried out. In the control group, the interaction Trial Type × Hemisphere was significant, F(1,29)=11.82, p=.002. This was explained by the fact that for successful inhibitions, N200 amplitudes were greater over the right than the left hemisphere (−3.26±0.47 vs. −.94±0.32 μV), whereas there was no hemispheric difference for failed inhibitions (−.35±0.67 vs. −.21±0.50 μV). Furthermore, the interaction Trial Type × Anterior Posterior Topography was significant, F(1,29)=15.04, p=.001. This was explained by greater N200 amplitudes for SI than FI trials over anterior scalp (−.28±0.47 vs. −.8±0.58 μV, p=.003) and greater N200 amplitudes for FI trials over posterior than anterior scalp (−.77±0.66 vs. −.8±0.58 μV, p=.00008). In the ADHD group, the main effect of Hemisphere was significant, F(1,35)=15.7, p=.0007, with overall greater N200s over the left than right hemisphere (−.06±0.28 vs. 0.88±0.31 μV). This was qualified by the Hemisphere × Anterior Posterior Topography interaction, F(1,35)=4.5, p=.041. A strong N200 asymmetry was significant only over posterior scalp (LH=−.95±0.46 μV; RH=−.3±0.52 μV), F(1,35)=18.53, p=.00007.

Graph: 1 Grandaverage ERPs over 15 representative scalp sites for successful inhibitions (left) and failed inhibitions (right). Black solid line: control group; Black dashed line: ADHD‐TN group; Gray solid line: ADHD‐RX group. Labels indicate main N200, NoGo‐P3, and LPW group differences. ADHD‐TN: ADHD treatment‐naïve group; ADHD‐RX: ADHD chronic treatment group.

Graph: 2 Left: N200 amplitude for successful inhibitions (top) and failed inhibitions (bottom) over left (L) and right (R) inferior frontal and temporoparietal scalp sites. In purple: control group; in blue: ADHD group. Right: F value topographical maps showing the scalp location where the N200 amplitude (170–220 ms) was statistically greater (p<.05, uncorrected) in control than the ADHD group for successful inhibitions (top) and failed inhibitions (bottom). Note the different maxima over the right inferior prefrontal scalp for the former and the right temporo‐parietal scalp for the latter.

Results of the restricted between‐groups analyses are detailed in Table 2A and illustrated in Figure 2. N200 amplitudes for successful inhibitions were significantly greater in the control than ADHD group over the right anterior region (controls=−.20±0.59 μV; ADHD=−.93±0.54 μV), F(1,64)=8.14, p=.006, effect size=.67, and the right posterior region (controls=−.31±0.63 μV; ADHD=−.44±0.56 μV), F(1,64)=5.04, p=.028, effect size=.54. For failed inhibitions, N200 amplitudes were significantly greater for controls than ADHD only in the right posterior region (controls=−.75±0.71 μV; ADHD=−.16±0.65 μV), F(1,64)=7.32, p=.009, effect size=.64, whereas they were similar for all other ROIs. Results of restricted within‐group analyses are detailed in Table 2B. For the control group, N200 amplitudes were significantly greater for successful than failed inhibitions over the right anterior region, F(1,29)=20.55, p=.00007 (SI=−.20±0.63 μV; FI=−.95±0.67 μV), whereas they were similar in the other three regions. Within the ADHD group, N200s were of similar amplitude for successful and failed inhibitions over all four regions.

2   N200 and NoGo‐P3 Results by Type of Trial, Scalp Region, and Group

2   N200 and NoGo‐P3 Results by Type of Trial, Scalp Region, and Group

2   N200 and NoGo‐P3 Results by Type of Trial, Scalp Region, and Group

2   N200 and NoGo‐P3 Results by Type of Trial, Scalp Region, and Group

2 *p<.05 ; **p<.01 ; ***p<.0001 ; ∼ p<.10.

N200 source analysis. To gain insight into the possible brain generator(s) of the N200 group difference for successful inhibitions, we applied dipole source modeling using the BESA 5.1 software ([41]; see Figure 3). We used the following approach. First, we hypothesized that the main source of the N200 group difference would correspond to the fMRI coordinates of a right DLPFC activation observed in the contrast of success inhibitions and Go trials in a companion study of the SST performed in a subset of the same subjects ([35]). We proceeded to seed the right frontal source in the source model (x=40; y=+5; z=+30; Brodmann Area 9), with the orientation only left to vary. At the peak of the N200 activity, this yielded a solution with a residual variance (RV) of 29%. To capture the residual scalp activity in the N200 range, we then added a second unconstrained dipole that fit in left temporo‐occipital cortex, (x=−3; y=−6; z=+6), leading to an improved peak residual variance of 18%. This second dipole possibly reflects a prominently left posterior N200 component in the ADHD group (see above). The same two‐dipole model was then applied to the failed inhibition N200 group difference, with the location fixed and the orientation only left to vary. Source localization in this case yielded a totally unsatisfactory solution, with 72% unexplained variance (65% at the N200 peak).

Graph: 3 BESA source analysis of the N200 group difference wave for success inhibitions. Left top: Time course of the two source waveforms (right prefrontal cortex in red and left temporo‐occipital cortex in blue). Left bottom: Minimum norm solution for the N200 group difference wave for success inhibitions. Note the distinct right frontal focus. Right: scalp average reference scalp topography for the observed data (top) and modeled data (bottom). Note the similarity of distributions.

We then tried a second approach, with a two‐dipole model fitting two unconstrained sources to the 170–220‐ms interval. This led to a solution with a primary dipole in more anterior and superior R DLPFC (x=29; y=36, z=+49) and a secondary dipole in L temporo‐occipital cortex (x=−2; y=−7, z=+10). This model had an improved peak RV of 13.3%. See in Figure 2 the similarity of the topographical distributions for the original scalp data (above) and the modeled source data (below) at the peak (195 ms) of the N200 difference wave for this second model. Finally, a Minimum Norm solution was calculated for the N200 SI group difference, which is an unconstrained estimate of current density emanating from a distributed source. Note the MN solution nicely identifies a single focus in right anterior DLPFC (see Figure 3, bottom left).

NoGo‐P3 effects (280–350 ms). ERP waveforms and topographic maps illustrating the NoGo‐P3 effects are shown in Figure 4. There was a significant Group × Anterior‐Posterior interaction, F(1,63)=6.25, p=.015. Controls' NoGo‐P3 was greater than ADHD's only over frontocentral scalp (controls=8.41±1.5 μV; ADHD=3.89±1.38 μV), F(1,63)=4.87, p=.031, effect size=.53. In addition to these group effects, the main effects of Trial Type, F(1,63)=17.7, p=.00008, attained significance, whereas there was no Trial Type × Group interaction, indicating that overall there was greater amplitude for successful than failed inhibitions, independent of diagnostic group. Focal ANOVAs between and within groups are presented in Table 2C, D.

Graph: 4 Left: NoGo‐P3 amplitude for successful inhibitions (top) and failed inhibitions (bottom) over left (L) and right (R) frontocentral and parietal scalp sites. In purple: control group; in blue: ADHD group. Right: F value topographical maps showing the scalp location where the NoGo‐P3 amplitude (280–350 ms) was statistically greater (p<.05, uncorrected) for control than the ADHD group.

LPW effects (480–700 ms). The main effects of Trial, AP Topography, and Hemisphere were significant, F(1,64)=11.8, p=.001, F(1,64)=9.5, p=.003, and F(1,64)=13.2, p=.002, respectively. LPW was greater in amplitude for failed than successful inhibitions, for the posterior than the anterior ROI, and over the right than the left hemisphere. The main effects were qualified by the significance of the interactions of Trial × AP Topography and Trial × AP Topography × Hemisphere, F(1,64)=16.4, p=.00008, and F(1,64)=4.3, p=.032. These were explained by greater LPW amplitude for failed inhibitions over the posterior scalp, more so over the right hemisphere. No main effect of Group, F(1,64)=2.12, p=.15, nor interactions involving Group (ADHD vs. controls) approached significance (for all, F<2.5, n.s.). However, see below for significant LPW effects when the ADHD group was subdivided according to treatment status.

Given the significant difference in GCA between controls and the ADHD‐TN group, it was crucial to ascertain the ERP findings were not due to differences in level of intellectual functioning (measured by the GCA score) among groups. To achieve this goal, the global ANOVAs were repeated in each time window adding GCA score as a covariate. When the effect of GCA score was partialled out, a very similar pattern of results was present, with all the relevant Group effects still significant. Furthermore, no effects including IQ approached significance either in the global analyses or in the breakdown into focal F tests.

ANOVAs restricted to the two ADHD subgroups were carried out for each time window. For the N200 and NoGo‐P3 analyses, there was no significance of the Treatment main effect (for all, F<1.1, n.s.) or for any of the interactions including Treatment (for all, F<1.6, n.s.).

In contrast, for the LPW, there was a significant main effect of Treatment, with smaller overall LPW amplitudes in ADHD‐TN (1.55±.62 μV), compared to both the ADHD‐RX group (3.77±.70 μV), F(1,34)=7.1, p=.012, effect size=.82, and the control group (3.82±.51 μV), F(1,44)=6.2, p=.017, effect size=.73. In contrast, there was no difference in LPW amplitude between ADHD‐RX and controls, F(1,48)=0.01, p=.95. To better understand the LPW effects, focal analyses for each trial type and ROI showed that in the contrast between ADHD‐RX and ADHD‐TN, the LPW reduction was most pronounced for failed inhibitions, over the left hemisphere, and for the posterior ROI (see Figure 1 and Table 3A).

3   LPW Results by Type of Trial, Scalp Region, and Group

3   LPW Results by Type of Trial, Scalp Region, and Group

3   LPW Results by Type of Trial, Scalp Region, and Group

3 *p<.05 ; **p<.001 ; ∼uncorrected.

A secondary analysis was also carried out to substantiate the visual impression that LPW amplitude for failed inhibitions was actually greater for ADHD‐RX than controls over more inferior left temporo‐occipital scalp, a region including site TI1 and TO1 (see Figure 1, right). This difference proved statistically significant, F(1,48)=6.51, p=.014, effect size=.70 (ADHD‐RX=3.26±0.81 μV; controls=0.60±0.66 μV).

Age was added to the global ANOVAs in each time window as a between‐subject factor (younger=9–12.5 years, older=12.6–15.5 years). In the N200 analysis, no main effect of Age or interactions with Group approached significance, whereas the Group × Hemisphere interaction was still significant. In contrast, Age had an influence on both NoGo‐P3 and the LPW, but did not modify the significant findings described in each respective analysis. In the NoGo‐P3 window, there was a main effect of Age, F(1,63)=4.3, p=.043, qualified by the significant interaction of Trial Type × Age, F(1,63)=5.71, p=.02 (which was independent of diagnosis). This was due to greater amplitudes for successful than failed inhibitions in the older subjects, F(1,32)=24.9, p<.0001, but not in the younger subjects. Furthermore, although successful inhibitions yielded greater amplitude in older than younger subjects, F(1,63)=7.07, p=.01, failed inhibitions were not dissimilar as a function of age. For the LPW, the main effect of Age, F(1,63)=10.76, p=.002, and the interaction Hemisphere × Age, F(1,63)=6.0, p=.017, were significant. The LPW was greater for older than younger subjects over the left parietal scalp, but again independent of diagnosis. This effect was greater for successful inhibitions.

Gender was added to the global ANOVAs in each time window as a between‐subjects factor (boys vs. girls). In the N200 and LPW analyses, no main effect of Gender or interactions with Group approached significance. In contrast, Gender had an influence on the NoGo‐P3. There were significant interactions of Trial Type × Gender, F(1,63)=4.13, p=.046, and Trial Type × Anterior‐Posterior × Gender, F(1,63)=5.42, p=.023. Independent of diagnosis, NoGo‐P3 amplitudes were greater for girls relative to boys over frontocentral scalp, but the effect was restricted to failed inhibitions.

In this study of the evoked response to the Stop signal, several ERP waves were abnormal in the ADHD group relative to the control group, and differential effects of age, gender, and treatment history were found on such ERP components.

The N200 wave elicited by the Stop signal was significantly reduced in the ADHD subjects, but the scalp distribution of such group difference varied depending on the outcome of the inhibitory process. In response to successful stops, N200 amplitude was significantly reduced over right inferior frontal scalp and to a lesser extent over right temporoparietal scalp, whereas in response to failed stops, N200 amplitude was significantly reduced only over right temporoparietal scalp.

In addition, over right inferior frontal scalp there was a success‐specific N200 enhancement (SI greater than FI) in healthy children, which was absent in children with ADHD. No failure‐specific N200 modulation (FI greater than SI) was present in either group. Dipole source analysis of the N200 group difference for successful inhibitions modeled such an effect to right DLPC, whereas an unsatisfactory solution was obtained using the same dipole model for failed inhibitions. Critically, the N200 effect was not affected by IQ, previous treatment history, age, or gender.

The present findings confirm and extend previous results of abnormally reduced right inferior frontal N200 waves in smaller samples of chronically treated ADHD boys using various versions of the SST ([1], [9], [36]). Most notably, they report for the first time a success enhancement of the right frontal N200 in healthy children and an absence of such a modulation in ADHD. Taken together, the results of the present study confirm and strengthen the notion that the frontal N200 plays a crucial role as an early inhibitory control mechanism in healthy individuals by triggering and modulating the efficiency of response inhibition. Critically, NoGo‐N200 reduction and the lack of success‐related modulation may represent central features of impaired inhibitory control in ADHD. It has been recently argued that the NoGo‐N200 in a Go/NoGo task reflects conflict monitoring rather than response inhibition. This interpretation is in part based on the N200 source localization in dorsal ACC rather than right inferior frontal cortex and its colocalization with the error‐related negativity elicited in the same task ([30]). It appears that the N200 origin may be task specific, perhaps reflecting greater demands on lateral PFC response inhibition mechanisms in the case of the SST and greater requirements on aspects of cognitive control (such as task switching, conflict monitoring) in the case of the Go/NoGo task. Further studies employing the two paradigms in the same subjects are warranted to resolve this issue.

The N200 interpretation in terms of response inhibition is consistent with findings of involvement of right middle‐inferior frontal gyri in inhibitory control in fMRI studies of Go/NoGo and SST in healthy adults ([15], [20], [37], [38]), and the reduction of such activation in ADHD children and adolescents ([37], [38]). It is worth noting that in an event‐related fMRI study of the SST in a subsample of the same cohort of patients, we found that although the right middle frontal gyrus (BA9) was activated in response to successful Stop trials relative to Go trials, there was no BOLD signal reduction in the ADHD group. A possible interpretation of this discrepancy is a selection bias in the fMRI study favoring higher functioning ADHD children with greater DLPFC activation and greater ability to hold still in the scanner, whereas lower functioning ADHD children were included in the present ERP study, given that head motion is much less of an issue in EEG recordings.

The amplitude of the NoGo‐P3 evoked by the Stop signal was significantly reduced in the ADHD relative to the control group over frontocentral scalp. Furthermore, successful inhibitions elicited greater NoGo‐P3 amplitude than failed inhibitions, independent of diagnosis. Such effects were not affected by previous treatment history or IQ. Abnormal reduction of the NoGo‐P3 in ADHD relative to healthy children was previously reported in the CPT ([10], [32]). In the SST, the success‐enhanced NoGo‐P3 was previously found diminished in children with ADHD relative to healthy children ([24], [31]). In the present study, there was no difference between groups in success‐related NoGo‐P3 enhancement. The reasons for the discrepancy are not clear; although the samples of this and previous studies ([24], [31]) differed in mean age and gender distribution, these factors did not interact with diagnosis in terms of the NoGo‐P3 findings. The results of the present study corroborate previous evidence that the SST may also tap into cognitive control mechanisms, in particular the monitoring of the outcome of the inhibitory process and its efficiency ([24], [30]). This interpretation is supported by EEG source localization of NoGo‐P3 effects, with activity generated in the dACC (e.g., [10]), confirming fMRI activations in the dACC in healthy subjects during Go/NoGo tasks ([22], [37]). Lack of normal dACC activation in ADHD adolescents has been reported during fMRI studies of the Counting Stroop Task ([6]) and the SST ([35]). In addition, the error‐related negativity has been found reduced in ADHD boys showing right frontal N200 reduction ([24]). The abnormality of NoGo‐P3 amplitude in ADHD may therefore reflect a deficit in cognitive control operations affecting overall performance monitoring, consistent with the dACC's role in monitoring conflict and error processing in cognitive control ([5], [30]).

The LPW amplitude was significantly greater for failed than successful inhibition trials, confirming previous evidence of failure modulation of this component ([31], [42]). The LPW appears to have the same timing and scalp topography of the error positivity (Pe), which has been associated with the process of late conscious error recognition (e.g., [11], [13], [42]).

More importantly, the LPW elicited by failed stops was selectively reduced in amplitude in the ADHD group which was treatment naïve, whereas the LPW amplitude in the chronically treated group was not only comparable to that of the control group, but even exceeded the control group over inferior temporo‐occipital regions. These findings suggest that late error monitoring processes, that is, conscious error recognition, normalize as a result of chronic treatment with stimulants in ADHD‐combined type. Chronic stimulants may facilitate the coming on line of strategy adjustments in the late conscious evaluation of errors ([49]), which may compensate for early abnormalities in inhibitory control and cognitive conflict (N200 and NoGo‐P3 reductions) that appear to persist in ADHD subjects in chronic stimulant treatment.

Importantly, the LPW finding cannot be attributed to the effect of acute stimulant medication: The ADHD children chronically treated were off stimulants at the time of testing. Although several ERP papers have reported acute effects of stimulant treatment (for reviews, see [18], [44]), this is, to our knowledge, the first time that ERP effects of chronic stimulant treatment history have been identified.

A previous study of the SST reported a reduced LPW in ADHD relative to healthy boys. In that study, all ADHD subjects were chronically treated ([31]). Other factors differed in their ADHD sample (age, gender, comorbidity), possibly contributing to the discrepancy with the present findings. Some recent studies are consistent with the interpretation that a deficit in late error processing may contribute to impaired SST performance in ADHD. First, the Pe has been recently found to be reduced in ADHD children ([49]). In addition, a recent study found reduced slowing (post‐error adjustment) following a failed inhibition in ADHD children relative to controls ([39]). Important corroborative evidence on the existence of a failure‐specific deficit in treatment‐naïve ADHD subjects comes from a recent event‐related fMRI of the SST ([38]). In the fMRI evoked response to failed inhibitions, their sample of adolescents with ADHD who were treatment naïve showed a selective reduction in activity in posterior cingulate cortex and precuneus relative to healthy adolescents, also interpreted as a deficit in error‐processing mechanisms. Given the posterior (parietal and occipital) distribution of the LPW effect in our study, we suggest that the source generator of the LPW effect may be in posterior cingulate cortex and adjacent precuneus ([38]). Note however that no source analysis was carried out in the present study to directly support this contention, which should therefore be taken with caution.

The results of the present study found no differential gender difference in the ERP correlates of impaired inhibitory control in our ADHD cohort. This finding is in line with recent large‐scale studies finding no significant gender differences in various domains of executive functioning in ADHD children (e.g., [43]). In the present study, however, gender influenced the NoGo‐P3 size independently of ADHD, being of greater amplitude in girls than boys across both groups. The NoGo‐P3 gender effect was seen only in response to failed inhibitions. This finding may be consistent with a meta‐analysis of studies exploring gender differences in inhibitory behavior (across social, behavioral, and cognitive domains) in typically developing children ([4]), showing an overall advantage for developing girls. Our results suggest a plausible mechanism for such a gender difference, with greater inhibitory control in developing girls possibly explained by better online processing of failure/error signals to compensate and guide future behavior.

The NoGo‐P3 displayed greater amplitude in older children, and such an effect was limited to successful inhibitions. Similarly, the LPW was increased in older subjects, more so for successful stops. There may be a developmental trajectory in the coming on line of success‐related components of conflict monitoring, with greater activity in older individuals, but this developmental change was not significantly affected by ADHD. Age‐related increases in the frontocentral NoGoP3 are consistent with findings from fMRI showing age‐related increases in prefrontal cortex activation in children during response inhibition in a Go/NoGo task ([48]). Lack of differential age differences in the ADHD group in the present study is consistent with the results of a large scale study finding no significant age differences in various domains of executive functioning in ADHD children ([43]).

The present study did not employ a tracking algorithm to adjust Stop signal intervals after each Stop trial (e.g., [26]), it rather adjusted the Stop signal intervals for the entire subsequent block based on mean Go RT in the previous block. This approach did not turn out to be as effective as the online stepwise correction method ([26]) in preventing strategic slowing in the control subjects. However, other behavioral measures differed in a way suggestive of impaired inhibitory performance in the ADHD groups, including the SSRT and the slope of the inhibitory function. It is also worth nothing that the present study focused on characterizing electrophysiological differences of the evoked response to Stop signals in ADHD and controls, above and beyond inherent speed/accuracy trade‐offs in the two groups.

ERPs to a Stop signal during the Stop signal task are altered in ADHD children and adolescents relative to controls in a manner suggesting a multifaceted deficit of inhibitory control, conflict monitoring, and error recognition mechanisms, rather than reflecting a simple inhibitory control deficit as originally postulated ([3], [36]). This interpretation is consistent with fMRI findings of BOLD signal reductions in ADHD patients during the Stop signal task in a wider network, including left middle‐inferior PFC, middorsal and posterior cingulate cortex, right parietal cortex, and other structures, and not only right DLPC ([36], [38]). The effect of treatment history on the LPW component elicited by to the Stop signal suggests that long‐term stimulant treatment may have a normalizing effect on late error processing in the ADHD children, and these effects persist even when the child is acutely withdrawn. Further studies of both the acute and chronic effects of stimulant treatment are warranted, but these results are reassuring that those treated chronically with stimulants show brain function more similar to controls than those who are treatment naïve.