TDCS/EEG addresses causal role of PFC in cognitive control

Title: TDCS/EEG addresses causal role of PFC in cognitive control

Name: RL van den Brink Supervisor: MX Cohen Co-assessor: TE Gladwin Student Number: 5749409 Date: 07-08-2012

Abstract

Goal-directed behavior requires the selection of appropriate sensory input in the face of distraction. When multiple competing alternatives are presented, conflict arises. Studies on the neural mechanisms that monitor and regulate behavior have implicated medal- and lateral frontal cortices (MFC and LPFC, respectively). However, causal evidence for the involvement of these structures in cognitive control remains sparse. Here, we tested the hypothesis that MFC and LPFC mediate cognitive control directly, using transcranial direct current stimulation (tDCS) whereby a low-voltage/amperage electric current is applied to manipulate cortical excitability. We show that facilitative and inhibitory stimulation of left dorsolateral prefrontal cortex (DLPFC) have opposing effects on behavioral adaptation following errors, and conflict resolution. Furthermore, theta band (4-8Hz) oscillatory activity over left central/frontal cortex was differentially affected by facilitative and inhibitory stimulation. Our results provide causal evidence for the involvement of DLPFC in mediating cognitive control, and indicate that tDCS can potentially be used to treat disorders where cognitive control is impaired.

1. Introduction

Goal-directed behavior requires the selection of appropriate sensory input in the face of distraction. When multiple competing alternatives are presented, conflict arises. Such conflict requires cognitive control and the active monitoring of performance, if behavior is to be adjusted accordingly. Studies on the neural mechanisms of cognitive control suggest that the medial frontal cortex (MFC) mediates performance monitoring by signaling e.g. reward prediction errors, response conflict, error likelihood, and response errors (Ridderinkhof et al., 2004b). For instance, invasive electrophysiological recordings in primates have shown that MFC neurons are active in response to errors, partial errors and the omission of rewards (Ito et al., Matsumoto et al., 2003, Stuphorn et al., 2000). Consistent with the notion that the MFC monitors performance, functional neuroimaging in humans have shown that the MFC is active during conditions of high response conflict (Carter et al., 1998). Furthermore, Kerns et al., (2004) showed that conflict-related MFC activity predicts behavioral adjustments in response to conflict.

Additionally, conflict- and error-related activity in MFC predicted PFC activation on subsequent trials (Kerns et al., 2004), suggesting that these performance monitoring signals in MFC adaptively engage behavioral adjustment mechanisms in lateral prefrontal cortex (LPFC) according to task demands (Botvinick et al., 2001, Ridderinkhof et al., 2004b). Theta band (4-8Hz) oscillatory activity is believed to underlie, at least in part, the adaptation of cognitive control (Luu & Tucker, 2001, Marco-Pallares et al., 2008) and the communication between these brain regions in frontal cortex that monitor and control behavior (Cavanagh et al., 2009, Cohen & Cavanagh, 2011). For example, Cavanagh, et al., (2009) showed medial and lateral frontal scalp electroencephalography (EEG) electrodes transiently synchronize during response errors, and the degree of synchronization predicts theta band power during these trials. Furthermore, the degree of behavioral adaptation on following trials (indexed by post-error slowing) was also predicted by both theta power and phase during error trials.

Taken together, these studies suggest that the MFC and LPFC actively monitor behavior, and implement behavioral adjustments when the need is detected. Furthermore, this top-down regulation is likely to be mediated by theta band oscillatory activity. Indeed, temporary inactivation of dorsal MFC via transcranial magnetic stimulation (TMS) has been shown to increase error rates on incongruent trials during an Eriksen Flanker task (Taylor et al., 2011). Furthermore, anterior cingulate cortex (ACC) lesions in humans can reduce the conflict effect, where participants are generally slower in response on incongruent trials following congruent trials (cI trials) compared to congruent trials following congruent trials (cC trials) on a Simon task, indicating conflict-dependent performance monitoring is disrupted (di Pellegrino et al., 2007).

Lesion studies are however restricted to using clinical sub-populations that might not be representative for the general population. Moreover, naturally occurring lesions are often not restrained to the region of interest. Additionally, little attention has gone to the effect of causal manipulation of frontal activity on oscillatory dynamics, specifically in the theta band. In the current study, we aimed to find causal evidence for the involvement of the LPFC/MFC and theta band oscillatory activity in cognitive control, and the monitoring of performance.

Causal manipulation of PFC excitability in healthy humans can be established via transcranial direct current stimulation (tDCS), where a low voltage/ampere direct current is applied to the head (Jacobson et al., 2012, Utz et al., 2010). The placement of the anode and cathode determines the effect on cortical excitability. Whereas anodal stimulation (with the anodal electrode placed over a brain region of interest and the cathode placed over a functionally less relevant region) will lead to an increase in cortical excitability (i.e. a facilitative effect), cathodal stimulation will lead to a reduction (i.e. an inhibitory effect). If MFC/LPFC mediate cognitive control, manipulations of PFC excitability via tDCS should be reflected in behavioral measures of cognitive control. Specifically, anodal/cathodal stimulation of PFC should have opposing effects on behavioral adaptation following errors (post-error slowing, PES) and conflict (the difference between cC and cI trials). Furthermore, manipulations of PFC excitability should have consequences for the MFC-LPFC interaction in the theta band. If the MFC/LPFC mediate cognitive control, anodal and cathodal stimulation of PFC

should lead to differential effects on medial 1) frontal theta power, and 2) theta phase synchronization between LPFC and MPFC following errors and conflict.

In line with our expectations, post-error slowing (PES) and the conflict effect (response time difference between cC and cI trials) were differentially affected by facilitative and inhibitory tDCS stimulation. Furthermore, theta band power over left frontal/central cortex was differentially affected by tDCS stimulation. However, no effects on phase synchrony between medial and lateral frontal electrode sites were observed, presumably due to the small sample size. The observed behavioral effects, and effects on theta power implicate DLPFC causally in cognitive control. These findings indicate that tDCS can potentially be used to treat disorders where cognitive control is impaired, such as attention deficit and hyperactivity disorder (ADHD).

2. Materials and methods

2.1. Participants

Nine healthy right-handed male participants (mean age: 23.5 years) recruited from the general population participated in the study. Participants could only partake in the experiment if they: 1) Were between 18 and 30 years of age; 2) Did not use psychoactive medication; 3) Were neurologically healthy; 4) Did not have a family history of epilepsy; and 5) Did not have any skin conditions that cause irritable skin. Participants were compensated with 30 Euro’s for the first session, and 40 Euro’s for the second session. The experiment was approved by the University of Amsterdam department of psychology ethics committee, and was conducted in accordance with the declaration of Helsinki. All participants signed informed consent beforehand.

Testing took place on two sessions, with a minimum of one day in between. On each session, participants performed a Simon task twice: once prior to tDCS application, and once following it. During each task performance, brain activity was recorded with EEG. Prior to the beginning of the first session, participants were given 15 seconds (plus 8 seconds ramp-up and ramp-down) of 1mA tDCS stimulation so that they could experience the sensation of tDCS. After this trial-stimulation participants were given the option to withdraw from the experiment if they found the tDCS stimulation uncomfortable. None of the participants indicated that the sensation was uncomfortable enough to withdraw from the experiment.

2.3. Simon task

Conflict has been extensively studied using the Simon task (Simon & Rudell, 1967). In its most canonical form, the Simon task involves responding to a stimulus property (e.g. shape or color) while ignoring its presented location. Thus, when the property and location are incongruent (e.g. the color requires a ‘left’ response, but the stimulus is presented on the right side of the screen), spatial conflict is induced.

In our version of the task, participants were instructed to respond as quickly as possible to the color of a circle, presented at 6.4o degrees visual angle left or right from fixation, while ignoring its location. Color-response mapping were as follows: Blue/yellow – left; Red/green – right. Stimulus presentation side could either be congruent or incongruent, where presentation side and color-response association were the same or different, respectively. Because trial sequence effects (i.e. Gratton effect; Gratton et al., 1992) cause the conflict effect to be maximal following congruent trials, trial types were matched across conditions such that there were equal numbers of congruent and incongruent trials following congruent trials, and no identical (same color and presentation side) trials were presented in succession. Stimuli were presented for 150ms, and following the participant’s response (maximum response time: 1000ms), a 1000ms fixed inter-trial interval (ITI) was presented during which only a fixation cross was present on the screen (Figure 1a.). In total, the task consisted of 700 trials with a self-paced brake after every 100th trial, and lasted approximately 20 minutes.

The task was presented on a 21” TFT monitor at 60Hz using Presentation software (Neurobehavioral systems). Participants were seated approximately 90cm away from the screen. To maximize conflict participants responded with the middle- and index finger of their right hand. Colors that required a right response corresponded to a left mouse-button click, and right responses corresponded to a right mouse-button click.

2.4. Transcranial direct current stimulation

On each session, participants received 20 minutes of either cathodal (-1mA) or anodal (+1mA) tDCS applied via 35cm2 electrodes enclosed in saline-soaked sponges. The duration of stimulation was chosen to ensure that the effect on cortical excitability lasted longer than the duration of the task (Utz et al., 2010). Additionally, sudden onset large voltage changes can lead to increased onset neuronal discharge effects (Creutzfeldt et al., 1964). Therefore, at the start of stimulation, current was gradually ramped up over a period of eight seconds, and ramped down for eight seconds at the end.

Electrodes were centered around position F3 of the international 10/20 system (i.e. over left DLPFC) and on the right orbital (forehead). The type of stimulation determined the effect on cortical excitability, with anodal

Figure 1. Behavioral task and results. a) Task overview. Stimuli were

presented for 150ms, and following the participant’s response (maximum response time: 1000ms), a 1000ms fixed inter-trial interval (ITI) was presented during which only a fixation cross was present on the screen. Participants were instructed to respond as quickly as possible to the color of a circle while ignoring its location. In congruent trials, the presentation side and color/response mapping were the same, while in incongruent trials they were different. b) Mean correct response time. c) Mean error rate. d) Behavioral adaptation following errors (PES). e) Conflict effect (cI versus cC trials) on response times. Asterisks indicate significant interaction effects.

stimulation leading to and up-regulation of DLPFC excitability. With cathodal stimulation, the polarity was reversed such that DLPFC excitability was down-regulated. Stimulation type on each session was determined in a double-blind randomized and counterbalanced order. To ensure that stimulation did not lead to skin leasoning and to provide the most comfortable experience for the subject, electrode impedances were kept below 50 k (and often did not exceed 5 k).

2.5. Data collection

Behavioral data were acquired with Presentation software (Neurobehavioral systems). EEG data were acquired with an EGI HydroCel Geodesic 256 electrode Sensor Net, at 1000Hz and referenced on-line to electrode Cz. Netstation sofware was used to record data and no on-line temporal or spatial filters were used.

2.6. Analysis

2.6.1. Behavioral data analysis

Prior to analysis, all trials in which the subject failed to respond within the response window were excluded. To test our predictions that manipulation of frontal cortical excitability should be reflected in behavioral measures, we used a 2x2 repeated measures analysis of variance (ANOVA) in IBM SPSS Statistics 19 with the factors ‘stimulation’ (cathodal and anodal tDCS) and ‘order’ (pre- and post tDCS measurement). We tested whether tDCS had an effect on mean correct response times (all reported response times are relative to stimulus offset), accuracy, behavioral adaptation, and conflict. The degree of behavioral adaptation was indexed by post-error slowing (PES). PES was defined by the response time (RT) difference between each middle trial of three consecutive correct trials, and correct trials following errors (van Driel et al., under review). The conflict effect was defined as the difference in RT (hereafter referred to as conflict RT) and accuracy between correct cC trials and cI trials. Additionally, to verify that any found effects were not driven by differences in

pre-tDCS performance across sessions, we compared these different behavioral measures between both pre-tDCS sessions for all subjects with two-tailed Student’s t-tests.

2.6.2 EEG analysis

All EEG data were analyzed in MATLAB 2010a, using the EEGLAB toolbox (Delorme & Makeig, 2004) and custom in-house code. Data were high-pass filtered at 0.5 Hz to remove drifts, notch-filtered at 50Hz to remove line-noise, and down sampled to 250Hz to speed up computation. The continuous data were epoched in bins ranging from -1 to 1.5s centered around stimulus onset and baseline corrected by subtracting the average offset during the -400 to -100ms pre-stimulus window. Next, trials with artifacts and eye-movements were manually rejected from analysis, and channels with poor signal quality were interpolated. Eye blinks and electrocardiogram (EKG) artifacts were identified using Jader independent component analysis (ICA) and the corresponding components were removed from the data.

Spectral power and phase dynamics were extracted via Morlet wavelet deconvolution. Wavelet deconvolution involves convolving the data with a set of Gaussian windowed complex sine waves, here defined as:

𝜓𝑓𝑤 = 𝑒2𝑖𝜋𝑓𝑤𝑡∙ 𝑒 −𝑡2

2𝑆𝑤2 ₍₁₎

where fw denotes frequency, which ranged from two to 60 Hz with 25 logarithmically spaced steps. Time is denoted by t, and sw determines the width of the Gaussian window, and thus the tradeoff between time and frequency resolution. Because higher frequency bands span a wider range of frequencies and power/phase dynamics change more rapidly, a better temporal resolution is preferred. We therefore linearly increased the Gaussian width with frequency, such that:

𝑠_𝑤 = 𝑐𝑤

2𝜋𝑓𝑤 (2)

where cw denotes the number of cycles, which ranged from three to 12.

P(t) = ReM(t)2 + iM(t)2 (3)

Where ReM and iM denote the magnitude of the real and imaginary components of the convolution result, respectively. To enable comparisons across frequency bands, power was converted to decibel (dB) scale and baseline corrected using a -300 to -100 ms baseline window.

Frequency-specific instantiations phase (ϕ) is given by the angle of the convolution result (arctangent of iM over ReM) at time t, and ranges between zero and one. Phase coherence (p) at time t can be calculated using:

p(t)

=

| n

𝑒

|

where ҡ is the difference in phase angles between electrode pairs (ϕ1 – ϕ2) to compute inter-electrode phase coherence. Spuriously high correlations in phase synchrony between electrodes can be found due to volume conduction effect. Therefore, prior to the analysis of inter-electrode phase synchrony, a current source density (CSD) transformation was applied to effectively remove shared activity between electrodes due to volume conduction (Kayser & Tenke, 2006).

2.6.3 Electrode, time, and frequency selection

To test our hypotheses about theta power, electrodes were selected based on group-averaged data. The electrodes where the conflict (the difference between cI and cC trials) and the error effect (here defined as the difference between correct trials and error trials) was maximal over all pre- and post-stimulation conditions in theta power were chosen. Data were averaged across the time-frequency window with maximal theta power for each subject, and the resulting values were entered into a 2x2 repeated measures ANOVA, with ‘stimulation’ and ‘order’ as within-subjects variables, similar to the analysis of behavioral data. To visualize interactions (see results), the contrast between cathodal and anodal stimulation was used, corrected for tDCS theta power ([andoal post-tDCS vs anodal

pre-tDCS] vs [cathodal post-tDCS vs cathodal pre-pre-tDCS]). Because any pre-tDCS differences are subtracted out, this contrast controlled for learning and cognitive fatigue effects.

For inter-electrode synchrony, seed electrodes were selected based on peak theta power effects, similar to power analysis. Target electrodes were selected based on peak group-averaged theta phase synchronization between the seed electrode, and electrodes over left LPFC. For target selected target electrodes, the same contrasts were generated as for power analysis (see above). All reported time-points (see results) are relative to stimulus onset.

3. Results

3.1. Behavioral results

One participant performed at chance level on the behavioral task, and was therefore excluded from the analysis. Of the remaining 8 participants, the mean number of non-response trials was 2.97 (SD 2.00), the mean error rate was 11.23% (SD 4.56%) and the mean response time on correct trials was 476.22ms (SD 39.04ms) (Figure 1b-c).

The repeated measures ANOVA revealed a main effect of order (F(1,7) = 13.10, p<0.01**)) on response times, indicating a significant learning effect, but no effect of stimulation or interaction between the two (all p>0.05). Interestingly, and in line with our prediction, no main effect of stimulation or order on PES and conflict RT were found, but a significant interaction between the two was present on both PES (F(1,7) = 7.21, p<0.05*), and conflict RT (F(1,7) = 18.861, p<0.01**). This indicated that tDCS had differential effects on the behavioral adaptation and conflict resolution, with anodal tDCS leading to faster response times following errors, and during conflict (figure 1d-e), while cathodal stimulation had the opposite effect on response times. Although PES and conflict RT were lower on the pre-cathodal tDCS than on the pre-anodal tDCS measurements, these differences were non-significant (t(14) = 0.60, p=0.50 and t(14) = 0.70, p=0.53, respectively) and thus driven by post-tDCS differences. No significant effects on accuracy were found (all p>0.05).

3.2. Results theta power

The results of the time frequency analysis showed that the difference in theta power between correct and error trials was maximal between 4-6Hz, in the time window 470-680ms post stimulus over medial frontal electrode 9 (figure 2a). The conflict effect (cI versus cC trials) was maximal between 5-8Hz in the time window 345-500ms post-stimulus (Figure 2a) over medial frontal electrode 186. Contrary to our expectations, no main effects or interaction effects between stimulation and order were present (all p>0.05) on conflict or error versus correct trials.

The topographical distribution of theta power in the contrast between anodal and cathodal stimulation, corrected for pre-tDCS theta power, did however indicate a left lateralized effect. Given that tDCS stimulation was applied over left DLPFC, it is reasonable to expect lateralized effects in theta power. We therefore selected one additional electrode post-hoc, based on the maximal difference

Figure 2. Results theta power. a) Group averaged theta power following conflict (cI versus cC trials), and errors (error trials versus correct

trials). b) Conflict related theta power for different stimulation conditions. The left panel shows the interaction effect of tDCS on theta power for an electrode selected post-hoc. The right panels show time-frequency plots for individual pre- and post-stimulation conditions for this same electrode. preA: pre anodal tDCS; postA: post anodal tDCS; preC: pre cathodal tDCS; postC: post cathodal tDCS.

in theta power between the cathodal and anodal stimulation (figure 2b). For this electrode, no main effects of ‘stimulation’ or ‘order’ were present, but a significant interaction effect on power between 5-8Hz, from 280 to 400 ms was present on conflict (F(1,7)=11,72, p<0.05*), indicating differential effects of stimulation on theta power. When compared by t-test, the pre-cathodal and pre-anodal tDCS session did not differ from each other in theta power for the selected time-frequency window (t(14) = 1.39, p = 0.17).

3.3. Results inter-electrode theta phase synchrony

Based on group-averaged power results, two electrodes were selected as seeds for inter-electrode phase synchronization analysis: Electrode 186 and inter-electrode 9, for the conflict effect and error-related theta phase synchronization, respectively. Group-averaged topographical plots of phase synchronization with seeded electrodes (figure 3a) showed maximal synchronization with LPFC at

Figure 3. Results inter-electrode theta phase synchrony. Time-frequency plots are shown for target electrodes. a) Group averaged theta

phase synchrony following conflict (cI versus cC trials), and errors (error trials versus correct trials), seeded from electrode 186, and electrode 9, respectively. b) Conflict and error-related related theta phase synchrony, contrasted between stimulation conditions: preA: pre anodal tDCS; postA: post anodal tDCS; preC: pre cathodal tDCS; postC: post cathodal tDCS; synch: synchronization, which indicates phase angle differences between electrodes and conditions. Note that raw synchronization values vary between zero and one, but because only contrasts between trial-types/stimulation conditions are shown, synchronization values reflect phase angle differences.

electrodes 3 and 2 for conflict- and error-related theta at times 630 to 730ms and 400 to 650 ms between 4 to 5Hz, respectively.

Although inter-electrode synchrony was increased for anodal- compared to cathodal stimulation, corrected for pre-tDCS differences (figure 3b), repeated measures ANOVA did not reveal any significant effects in the selected time-points (all p>0.05).

4. Discussion

The current study tested the hypothesis that LPFC and MFC are causally involved in cognitive control, using facilitative (anodal) and inhibitory (cathodal) tDCS. In line with this hypothesis, we showed differential effects of anodal and cathodal PFC stimulation on PES, and the conflict effect. Whereas anodal tDCS reduced RTs following errors and conflict, cathodal tDCS increased them. Interestingly, no interactions were found on mean correct RT, and error rates, indicating the observed behavioral effects were specific to conflict and error resolution rather than generalizing to changes in overall task performance following tDCS. Furthermore, our findings that errors and conflict result in increased medial frontal theta power fit well with an already established body of literature (Cavanagh

et al., 2009, Cohen & Cavanagh, 2011, Luu & Tucker, 2001, Marco-Pallares et al., 2008, van Driel et al., under review).

Cognitively, a reduced conflict effect, and reduced PES can be interpreted in two ways, with opposing implications. First, they may reflect less implementation of behavioral adaptation when errors are detected, and when conflict is detected, and thus, reduced cognitive control. Second, they may reflect more efficient implementation of behavioral adaptation, and thus, increased cognitive control. Given that our post-hoc analysis showed that anodal stimulation resulted in increased left-lateralized theta power following conflict compared to cathodal stimulation, and theta band oscillatory activity is thought to underlie the communication between the brain regions that mediate cognitive control (Cavanagh et al., 2009, Cohen & Cavanagh, 2011), it seems likely that the behavioral effects

observed here reflect more efficient behavioral adaptation following conflict and errors. Thus, the most parsimonious explanation for a reduced conflict effect and reduced PES following anodal tDCS would be an increase in cognitive control. This finding lends support to the hypothesis that DLPFC is causally involved in mediating cognitive control. Moreover, to our knowledge, this is the first demonstration of a causal manipulation of PFC function in healthy humans leading to alterations conflict- and error-specific behavioral adaptation. Additionally, the behavioral effects were complemented by stimulation-type specific alterations in the neural signatures of cognitive control. Specifically, frontal theta dynamics were differentially altered by cathodal tDCS and anodal tDCS in a similar manner as behavioral conflict- and error adaptation. This further strengthening the link between frontal theta oscillatory activity and cognitive control (Cavanagh et al., 2009, Cohen & Cavanagh, 2011, Luu & Tucker, 2001, Marco-Pallares et al., 2008, van Driel et al., under review).

It should be noted that the absence of effects in phase synchronization between medial and lateral frontal electrodes limits the conclusions that can be drawn about the relevance of communication between MFC and LPFC as a mechanism for implementing cognitive control. Given our small sample size, and the fact inter-electrode synchrony was increased for anodal- compared to cathodal stimulation, it is possible that this effect could reach statistical significance with a larger subject group. Future studies using a similar paradigm will have to be conducted in order to shed more light on the dynamic interaction between MFC and LPFC. Furthermore, it should be noted that the all-male subject sample might not be representative for the general population. Future studies will have to take this into account.

Additionally, it should be mentioned that the observed effect of tDCS on frontal theta power was not significant for the electrodes selected a-priori. However, due to our left-lateralized stimulation set-up (the relevant stimulation electrode was placed over left DLPFC), it is reasonable to expect left-lateralized effects in the predicted oscillatory range. This finding does call for a confirmation of the result. By using both left- and right-lateralized stimulation, it could be shown that the observed changes in theta power following tDCS are specific to the cortical location underlying the stimulating electrode.

In summary, the current study shows that facilitative and inhibitory tDCS have opposing effects on behavioral measures of cognitive control, as well as the neural signatures of cognitive control. Our results provide tentative, yet much needed, causal evidence for the involvement of DLPFC in mediating cognitive control. Given that we have successfully demonstrated alteration both the brain signals and behavioral markers of cognitive control, our findings indicate that tDCS can potentially be used to treat disorders where cognitive control is impaired, such as ADHD. Whether the behavioral changes as observed here are still observable after prolonged periods of time, will have to be investigated in future studies.

5. Acknowledgements

The author would like to thank Mike X Cohen and Thomas Gladwin for their advice and supervision, and Jochem van Kempen, Daan van Es, and Rosanne van Diepen for their help with collecting the data.

6. References

Botvinick, M. M., Braver, T. S., Barch, D. M. Carter, C. S. & Cohen, J. D. (2001). Conflict monitoring and cognitive control. Psychological Reviews, 108:624-652.

Cavanagh, J. F., Cohen, M. X & Allen, J. B. (2009). Prelude to and resolution of an error: EEG phase synchrony reveals cognitive control dynamics during action monitoring. The Journal of Neuroscience, 29(1):98-105.

Carter, C. S., Braver, T. S., Barch, D. M., Botvinick, M. M., Noll, D. & Cohen, J. D. (1998). Anterior cingulate cortex, error detection, and the online monitoring of performance. Science, 280:747-749.

Cohen, M. X & Cavanagh, J. F. (2011). Single-trial regression elucidates the role of prefrontal theta oscillations in response conflict. Frontiers in Psychology, 2:30.

Creutzfeldt, O.D., Fromm, G.H., & Kapp, H., (1962) Influence of transcortical d-c currents on cortical neuronal activity. Experimental Neurology, 5(6):436-452.

Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. The Journal of Neuroscience Methods, 134:9-21

di Pellegrino, G., Ciaramelli, E. & Ladavas, E. (2007). The Regulation of Cognitive Control following Rostral Anterior Cingulate Cortex Lesion in Humans. Journal of Cognitive Neuroscience, 19(2):275– 286.

Gratton, G., Coles, M. G., & Donchin, E. (1992). Optimizing the use of information: Strategic congrol of activation of responses. Journal of experimental psychology: General, 121(4):480–506.

Ito, S., Stuphorn, V., Brown, J. W & Schall, J. D. (2003). Performance monitoring by the anterior cingulated cortex during saccade countermanding. Science, 302:120-121.

Jacobson, L., Koslowsky, M. & Lavidor, M. (2012). tDCS polarity effects in motor and cognitive domains: a meta-analytical review, Experimental Brain Research, 216:1-10.

Kayser, J., & Tenke, C. E. (2006). Principal components analysis of Laplacian waveforms as a generic method for identifying ERP generator patterns: I. Evaluation with auditory oddball tasks. Clinical

Kerns, J. G., Cohen, J. D., MacDonaald III, A. W., Cho, R. Y., Stenger, V. A. & Carter, C. S. (2004). Anterior cingualte conflict monitoring and adjustments in control. Science, 303:1023-1026.

Luu, P. & Tucker, D. M (2001). Regulating action: alternating activation of midline frontal and motor cortical networks. Clinical Neurophysiology, 112:1295-1306.

Marco-Pallares, J., Camara, E., munte, T. F. & Rodrigues-Fornells, A. (2008). Neural mechanisms underlying adaptive actions after slips. Journal of Cognitive Neuroscience, 20:1595-1610.

Matsumoto, K, Suzuki, W. & Tanaka, K. (2003). Neuronal correlates of goal-based motor selection in the prefrontal cortex. Science, 301:229-232.

Ridderinkhof, K. R., Ullsperger, M., Crone, E. A. & Nieuwenhuis, S. (2004). The role of the medial frontal cortex in cognitive control. Science, 306:443-447.

Simon, J. R. & Rudell, A. P. (1967). Auditory S-R compatibility: the effect of an irrelevant cue on information processing. Journal of applied psychology, 51:300-304.

Stuphorn, V., Taylor, T. S. & Schall, J. D. (2000). Performance monitoring by the supplementary eye field. Nature, 408:857-860.

Taylor, P. C. J., Nobre, A. C. & Rushworth, M. F. S. (2011). Subsecond changes in top-down control exerted by human medial frontal cortex during conflict and action selection: a combined transcranial magnetic stimulation-electroenceponalography study. The Journal of Neuroscience, 27(42):11343-11353.

Utz, K. S., Dimova, V., Opperlander, K. & Kerkhoff, G. (2010). Electrified minds: transcranial direct current stimulation (tDCS) and Galvanic Vestibular Stimulation (GVS) as methods of non-invasive

brain stimulation in neuropsychology - a review of current data and future implications.

Neuropsychologia, 48:2789–2810.

Van Driel, J., Ridderinkhof, K. R. & Cohen, M. X. (under review). Not all errors are alike: Theta and alpha EEG dynamics relate to differences in error-processing dynamics. The Journal of Neuroscience