Repeated anodal tDCS of the lDLPFC during
a working memory task
Author:
Henryk A. Kroese
H.A.Kroese@student.uva.nl
Supervisor:
Lotte J. Talsma
Brain and Cognition
L.J.Talsma@uva.nl
September 18, 2014
Abstract
It has been shown that both cognitive training and transcranial Direct Current Stimulation (tDCS) can be used to alter and im-prove working memory. The aim of this study is to investigate if repeated tDCS can be used to improve working memory performance and whether these improvements transfer to other tasks. In this study, three tasks were used to look at the transfer effects of multiple ses-sions of tDCS in combination with cognitive training. No significant transfer effects of anodal tDCS on working memory performance were found for any of the tasks. This could be due to the relatively short cognitive training period of three days and/or the limited number of three tDCS sessions. In future studies a longer training period and more tDCS sessions could be implemented to further investigate the effect of multiple sessions of tDCS on working memory performance. Keywords: tDCS, Working Memory, Cognitive Training
1
Introduction
Working memory can be summarized as the system that is responsible for the storage, manipulation and updating of information, which is essential for complex tasks like learning and reasoning [1]. Previous studies have shown that the dorsolateral prefrontal cortex (DLPFC) is essential in these (working memory) tasks [2].
Trancranial Direct Current Stimulation (tDCS) is a non-invasive brain stimulation method [3] that can be used in double-blind sham-controlled studies [4] to investigate and alter working memory when applied to the DLPFC. It has been argued that tDCS can be used to affect the brain in two ways that may be related. Firstly, by applying anodal tDCS the resting state membrane potential of neurons can be lowered. In other words, neu-ronal excitability can be increased [5]. However, this is not a long-lasting effect. In previous studies tDCS has already been shown to improve all kinds of cognitive functions through this instant effect [6]. Secondly, and more rele-vant to the current study, it seems that anodal tDCS can facilitate long-term potentiation. In other words, it can induce a long-lasting strengthening of synapses between two neurons. By applying anodal tDCS you may therefore be able to increase the plasticity of the brain. Combining tDCS with prac-tice on a certain task may allow the task to grasp on better and result in improved long-term learning effects.
In past research cognitive training has been shown to successfully improve working memory, for example by practice with a (dual) N-back task [7] [8]. In other research the N-back task was combined with different kinds of tDCS. And indeed, in these studies working memory improved at a faster rate under the influence of tDCS than when applying cognitive training alone [1] [9] [10]. This faster rate of working memory improvement was found when applying anodal tDCS to the left DLPFC. This increase in working memory perfor-mance seems to continue when stimulation is applied for a longer period of time within one session. This increase in working memory performance seems to be linear to the stimulation duration up to a duration of 30 minutes [10]. In tDCS studies targeting the motor cortex, multiple sessions of tDCS (repeated tDCS) were shown to have more beneficiary effects than single session tDCS. For example, repeated tDCS was used to successfully reduce pain after spinal cord injuries [11] and it was suggested that repeated tDCS sessions could improve motor function in stroke patients beyond the point of single session tDCS [12]. Furthermore, in a motor improvement experiment
in healthy subjects it was found that the beneficial effect of repeated tDCS remained present at least until 3 months after training [13].
Combining these findings of on the one hand improvement in working memory as a result of tDCS to the lDLPFC and on the other hand more beneficial effects of repeated tDCS in studies on the motor cortex, we hy-pothesize that repeated tDCS to the lDLPFC could benefit working memory. To assess this, we will compare tDCS with sham tDCS during a working mem-ory task. We expect that tDCS as compared to sham stimulation shows a greater and faster learning effect. Moreover, we expect this learning effect to transfer to other working memory tasks that were not practiced during tDCS sessions, which would implicate that tDCS does not simply improve task per-formance on a specific task, but actually increases overall working memory. This study will help give insight into whether repeated tDCS can be used to functionally increase plasticity of the brain, when stimulating other areas than the motor cortex. Specifically, it will give insights into the cognitive processes of updating information (e.g. working memory). If findings are positive, repeated tDCS could potentially be used for educational or clinical purposes.
2
Method
2.1
Subjects
In this study a total of 37 healthy subjects (13 males and 24 females) partici-pated. One participant withdrew after the premeasurement because of other obligations, one because of tDCS unrelated illness and one participant was excluded due to a metal plate in the back of her head. Ten other participants were excluded for scoring at ceiling level at the start of the experiment. Ceil-ing level was assumed when participants had an accuracy of 85% or higher on the average of the 3- and 4-back letter N-back task on the premeasurement. This resulted in 24 included participants within age range 19-30 years (mean 22.4; SD 2.54). There were 13 participants in the final active group and 11 in the final sham group. All participants gave a written informed consent and the study was approved by the ethical committee of the University of Am-sterdam. The participants were either rewarded with ’proefpersoon punten’ or with a total of 120 euros (100 euro + 20 euro bonus) or with a combination of ’proefpersoon punten’ and euros.
Figure 1: Stimulus presentation for the letter and spatial N-back tasks. Both tasks consist of a stream of stimuli. The participants have to press the spacebar when the stimulus is identical to the stimulus N stimuli ago.
2.2
Design
Participants were pseudo-randomly divided into two groups (active vs. sham) matching the groups on gender, age, and premeasurement scores. Both the participants and the experimenter were unaware of which participant be-longed to which group (double-blind). In total the participants came to the lab for five days (premeasurement, 3 stimulation days, postmeasurement), each day at the same time (except for the premeasurement). Between the premeasurement and the first stimulation day there was a gap of 72 hours, between each of the stimulation days there was a gap of 24 hours and there was a 48 hour rest period between the last stimulation day and the post-measurement. This 48 hour rest period was implemented because it takes some time before the cell membranes return to their normal resting state membrane potential and we wanted to ensure that the temporary effects of tDCS have worn out.
During the stimulation days the active group received anodal tDCS and the sham group received sham tDCS, both for three consecutive days while doing a letter N-back task. On the pre- and postmeasurement no tDCS was applied and both groups did a letter N-back task with a different stimulus set, a spatial N-back and an O-span task (see 2.4). These tasks were used to look at possible transfer effects of repeated tDCS in combination with cognitive training.
2.3
tDCS
To successfully stimulate the lDLPFC, the anodal electrode was placed over F3 based on the EEG International 10-20 system. To ensure that electrode placement was the same each stimulation day, F3 was localized with a marker on the first stimulation day. The EEG-cap was placed over the participants head and through a hole in the EEG-cap F3 could be marked. The catho-dal electrode, that served as a reference, was placed over the contralateral supraorbital area.
A 1 mA direct current was applied to participants in the active group for 20 minutes using two saline-soaked surface sponge electrodes (5 x 7 cm). The current was delivered by a battery driven constant current stimulator (neuroConn DC-stimulator, M¨unich, Germany). In the sham condition the electrode placement was exactly the same but tDCS was turned on for only 60 seconds before turning it off again (excluding 20 sec ramp-up and 60 sec ramp-down which was the same in both the active and sham group).
2.4
Working memory tasks
On the stimulation days the participants did a letter N-back task. On the pre- and postmeasurement days the participants did a letter N-back task with a different stimulus set, a spatial N-back and an automated version of the O-span task. To prevent order effects half of the participants in each group (sham and active) started with the letter N-back task followed by the spatial N-back; the other half vice versa.
In the letter N-back task a stream of letters was presented (using presen-tation software) and the participant had to press the spacebar if the letter that appeared on the screen was identical to the letter that appeared N let-ters ago (fig. 1). This way, N determines the level of the task. Each stimulus was shown for 300 milliseconds with 1500 milliseconds between each stimulus. Trials consisted of a stream of 20 + N. On the pre- and postmeasurement the task consisted of 48 trials, on stimulation days 24x3 and 36x1 trials were used. Per trial there were either 6, 7 or 8 targets (this amount was balanced per 6 trials). On the pre-and postmeasurement N varied between 2 and 5; on the stimulation days N varied between 3 and 4. The participants performed 6 trials of a certain level N-back before moving to the next level N-back (thus doing every level N-back twice for 6 trials). Two stimulus sets were used in the letter N-back task in order for it to be used as a transfer task. One set
Figure 2: Stimulus presentation for the o-span task. The o-span task consists of an interchanging stream of mathproblems and letters. After the stream has ended the participant is asked to remember, in correct order, the letters that occured in the stream.
was used for the pre- and post-measurement and the other for the stimula-tion days). The first set contained the letters [a, b, c, d, e, f, g, h, j, k] and the second set contained the letters [k, m, n, o, p, r, s, t, u, w] both upper and lower case.
In the spatial N-back task participants had to remember the locations of squares that appeared on screen and had to press the spacebar to indicate whether the square appeared on the same spot as N stimuli ago (fig. 1). Task design was equal to the letter N-back task (see previous paragraph). As the spatial N-back task was only presented on the pre- and postmeasurement, only one stimulus set was used for this task and 48 trials were used with N varying between 2 and 5.
The O-span task was done on the pre- and postmeasurement. On this task, participants had to remember sequences of letters while solving simple math problems (fig. 2). The O-span task is considered a good indicator of working memory. It can also be used to measure working memory capacity, instead of only performance, as it has good test-retest reliability. The O-span task consisted of 15 trials. The level of the task depended on the number of letters that had to be remembered; over trials it was varied between 3 and 7 letters. The speed at which the participants had to respond was determined separately for each individual and separately for the pre- and postmeasurement by using the response times in a block of practice trials (for further details [14]).
Figure 3: Accuracy on the letter N-back task: active vs. sham and pre- vs. postmeasurement (n = 24). Accuracy is represented on the y-axis; different level of N-back (2-5) on the x-axis. These results indicate that the task gets more difficult at higher level N and that the accuracy of the subjects improved between the pre- and postmeasurement.
2.5
Analysis
The primary outcome measures for the N-back tasks were number of hits, false alarms, misses and reaction time (RT). With these primary outcome measures a weighted accuracy measure was calculated1. For the O-span the primary outcome measures were the total amount of correct recalled letters, math errors and math time (the O-span task also yields an absolute score for the amount of recalled letters which was not analyzed in the current study). The results were analyzed using the statistical software SPSS. To look at whether there was an order effect between the active and sham group a two-way repeated measures of analysis of variance (ANOVA) was performed. On the two N-back tasks an ANOVA was used with between factor group (active vs. sham) and within factors day (pre and postmeasurement) and level (level of n). For the O-span task a one-way repeated measures ANOVA was used with between factor group (active vs. sham) and within factor day (pre- vs. posttest). If applicable, post-hoc analyses were performed to further clarify the results.
3
Results
3.1
Participant reports
There were no side-effects of tDCS or sham stimulation, except for some light headaches. All participants could finish the stimulation and none of them complained about pain. There were some reports of light itchiness during ramp-up. When participants were asked to force guess whether they were in the sham or active group they were at chance level.
3.2
N-back
On the letter N-back task a repeated measures ANOVA (between factor of group and within factors of day and level) showed a main effect of level on accuracy (F(2.4,20) = 85.151; P < 0.05). When running a post-hoc contrast
analyses (repeated) on level we find that every level differs significantly from the next level (2 vs. 3: F(1,22) = 45.116; P < 0.001; 3 vs. 4: F(1,22) = 22.783;
1Formula: ((correct rejections / hits) * (hits + correct rejections)) / ((correct rejections
Figure 4: The accuracy for the spatial N-back task: active vs. sham and pre- vs. postmeasurement (n = 24). Accuracy is represented on the y-axis; different level of N-back (2-5) on the x-axis. These results indicate that the task gets more difficult at higher level N and that the accuracy of the subjects improved between the pre- and postmeasurement.
P < 0.001; 4 vs. 5: F(1,22) = 25.381; P < 0.001). Which confirms that the
task gets more difficult with every increase in N (fig. 3). Furthermore, a main effect of day on accuracy (F(1,22) = 38.348; P < 0.05) was observed which
shows that subjects improved performance over days.
On the letter N-back task the only interaction effect on accuracy was found on day*level (F(2.5,20) = 4.785; P = 0.007). To further investigate
this interaction effect multiple tests of simple mean effects were done on day and level, which all show a significant effect2. This showed that on all levels participants scored significantly higher on the second day and that
2Premeasurement: 2 vs. 3: F
(1,22)= 57.490; P < 0.001; 3 vs. 4: F(1,22) = 13.316; P =
0.001; 4 vs. 5: F(1,22) = 9.185; P = 0.006. Postmeasurement: 2 vs. 3: F(1,22)= 11.783;
P = 0.002; 3 vs. 4: F(1,22) = 16.047; P = 0.001; 4 vs. 5: F(1,22) = 24.108; P < 0.001.
2-Back: pre vs. post: F(1,22)= 47.042; P < 0.001. 3-Back: pre vs. post: F(1,22)= 24.064;
P < 0.001. 4-Back: pre vs. post: F(1,22) = 28.297; P < 0.001. 5-Back: pre vs. post:
they scored significantly lower on higher level N-back tasks. No significant day*level*group interaction effect was found for either reaction time (F(2.1,20)
= 0.768; P = 0.477) or accuracy (F(2.5,20) = 0.268; P = 0.816). This shows
that there was no significant difference between the active and sham group. Results for the spatial N-back task resembled those of the letter N-back task. When performing a repeated measures ANOVA (between factor of group and within factors of day and level) a main effect of level on accuracy (F(2.2,20) = 143.393; P < 0.05) and a main effect of day on accuracy were
observed (F(1,22) = 15.224; P < 0.001). When running a post-hoc contrast
analyses (repeated) on level we find that every level differs significantly from the next level (2 vs. 3: F(1,22) = 70.670; P < 0.001; 3 vs. 4: F(1,22) = 49.881;
P < 0.001; 4 vs. 5: F(1,22) = 36.932; P < 0.001). This showed that the
spatial N-back also gets more difficult at higher level N and that the subjects improved over days.
Here too, the only interaction effect on accuracy was found on day*level (F(2.5,20) = 4.770; P < 0.007). To further investigate this interaction effect
multiple test of simple mean effects were done on day and level, which all show a significant effect3. Again this showed that on all of the level N-back tasks participants scored significantly higher on the second day and that they scored significantly lower on higher level N-back tasks. No significant day*level*group interaction effect was found for reaction time (F(2.6,20) =
0.472; P = 0.681) or accuracy (F(2.5,20) = 1.873; P = 0.153), showing that
there is no significant difference between the active and sham group.
However, in the spatial N-back task large standard deviations in accuracy are observed in both the active and the sham group. During premeasurement the standard deviation of the sham group seemed bigger than that of the active group; the opposite is true for the postmeasurement (fig. 4). For this reason we looked at individual differences by adding the covariates age, gender and baseline performance(the mean of the 3- and 4-back trials of the premeasurement). The day*level*group interaction effect on accuracy remained non-significant after correction for both age (F(2.5,19) = 1.711; P =
0.183) and gender (F(2.5,19) = 1.701; P = 0.185). But after adding a covariate
3Premeasurement: 2 vs. 3: F
(1,22)= 111.395; P < 0.001; 3 vs. 4: F(1,22) = 22.681; P
< 0.001; 4 vs. 5: F(1,22)= 4.891; P = 0.038. Postmeasurement: 2 vs. 3: F(1,22)= 26.520;
P < 0.001; 3 vs. 4: F(1,22) = 43.117; P < 0.001; 4 vs. 5: F(1,22) = 40.744; P < 0.001.
2-Back: pre vs. post: F(1,22) = 5.587; P = 0.027. 3-Back: pre vs. post: F(1,22) = 24.578;
P < 0.001. 4-Back: pre vs. post: F(1,22) = 16.054; P = 0.001. 5-Back: pre vs. post:
Figure 5: The mean math time: active vs. sham and pre- vs. postmeasure-ment (n = 22). These results indicate that there was an improvepostmeasure-ment in the time it took the participants to solve the mathproblems between the pre- and postmeasurement. However, this was seen in both groups and there seems to be no difference between the active and sham group.
thats represents baseline performance, a trend on accuracy was seen (F(2.6,19)
= 2.285; P = 0.098). i.e. when accounting for the individual differences in performance on the premeasurement a trend is seen on the day*level*group interaction effect on accuracy.
To further investigate this trend the effect of the separate levels was looked at, using a full-factorial repeated measurements ANOVA with the factors day and group. This showed that no trend was found on the level 3-back (F(1,21) = 0.006; P = 0.941), 4-back (F(1,21) = 0.11; P = 0.743) or
5-back (F(1,21) = 0.094; P = 0.762). Only the level 2-back had some indication
of causing the trend (F(1,21) = 21; P = 0.122). It appears that on the 2-back
the active group was already on ceiling level on the premeasurement and the sham group was not. On the postmeasurement both groups were at ceiling level on the 2-back, thus creating the trend in the interaction effect.
Figure 6: The mean of the correct recalled letters (maximum is 75): active vs. sham and pre- vs. postmeasurement (n = 24). These results indicate that there was no improved performance between the pre- and postmeasure-ment and that there was no difference between the active and sham group performance.
3.3
O-span
When performing a repeated measures ANOVA (between factor of group and within factor of day) there was no main effect of day on the amount of recalled letters (F(1,22) = 0.577; P = 0.456) or the amount of math errors (F(1,22)
= 0.045; P = 0.834). In other words there was no improved preformance between the pre- and postmeasurement. However, a trend in the main effect of day on math time was observed (F(1,20) = 9.035; P = 0.007). This entails
that participants seemingly solved the math problems faster the second time they had to do the O-span task (fig. 54). Thus, performance improvement
was only seen on the interference task and not in the transfer task (which was remembering the letters).
No significant day*group interaction effect was observed for the amount of recalled letters (F(1,22) = 0.138; P = 0.714) (fig. 6), the amount of math
errors (F(1,22) = 0.913; P = 0.35) or math time (F(1,20) = 0.172; P = 0.682).
This entails that there was no significant difference in performance between the active and the sham group for the O-Span task.
4
Conclusion and Discussion
As stated in the introduction we hypothesized that working memory could benefit from repeated tDCS on the lDLPFC. In this paper we looked specif-ically at whether the effects of repeated tDCS in combination with cognitive training would result in transfer effects.
Our results showed that for both letter and spatial N-back tasks the task got more difficult at higher level N and that the participants performed better on the second day. However, our results also indicate that there are no differences between the active tDCS and the sham tDCS group on the different N-back tasks on the pre- and postmeasurement. Only for the spatial N-back task a trend was observed for the interaction effect of day, level and group after adding a baseline performance covariate. This trend in the interaction effect on the spatial N-back task seemed to be solely driven by the level 2-back accuracy scores. In other words, all participants improve performance on the N-back transfer task, but the active tDCS group did not perform better in comparison to the sham tDCS group in either of the tasks
4n = 22, because the math time was not recorded for two participants, one in each
and thus there seems to be no extra effect of repeated tDCS on cognitive training on the transfer N-back tasks.
In the O-span task we see no difference in performance between the active tDCS and sham tDCS group, or between the two days, in any of the task scores. Therefore there seems to be no transfer effect of repeated tDCS in combination with cognitive training on this task. The O-span is a consis-tent working memory capacity measurement and these findings support this. However, we do see a trend in the main effect of day on math time; partic-ipants appear to be faster in solving math problems on postmeasurement. We chose for the math time to be calculated twice and this seems to be the right decision, because the interference task should be of the same difficulty on both the pre- and postmeasurement, which is is the case when the math time is calculated separately for both the pre- and postmeasurement.
In conclusion, we can say that at the current settings there appears to be no effect of repeated tDCS on any of the tasks, because the active group did not perform better on any of the tasks than the sham group. A follow-up analysis, which is not reported in this paper, on the current data shows that sham and active groups did differ in accuracy on the first stimulation day, which is consisted with previous studies [1]. This finding suggests that the lack of results in the current study was not due to improper use of tDCS or other elements of our research design. A possible explanation for the lack of repeated tDCS-induced effects is that three sessions is not sufficient. In cognitive training extensive sessions are needed to get results [8] and the idea is that tDCS could be used to speed up that process. Our results suggest that three tDCS sessions are probably not sufficient to achieve working memory improvement.
Therefore, future research should include more sessions and be spread out over a longer cognitive training period. Another important issue is that there are a lot of individual differences; although we analyzed some in this study, a lot could be gained by exploring these individual differences further. For example, one could determine the extent to which participants benefit from tDCS during the stimulation sessions and use that as a covariate in the analysis. A limitation of this study is that the effect of cognitive training is not accounted for. To change this in future studies an adaptive task could be used to account for the effects of cognitive training. Another suggestion for future research is to repeat this study in combination with electroencephalog-raphy to study the neural effects of tDCS [15]. Furthermore, a more difficult N-back task and perhaps even a dual N-back task could be used to prevent
exclusion of participants due to ceiling effects on the premeasurement. Fi-nally, other transfer tasks could be used as the O-span task generally does not allow for a big learning effect, which makes it an unsuitable transfer task for this study, as we look at the difference in learning between sham and active tDCS groups.
Although a start was made in the use of repeated tDCS on the lDLPFC in combination with cognitive training a lot could still be improved. We found no effects of repeated tDCS on cognitive training, but with the suggested alterations to the research design this could change. Hopefully this will result in useful clinical or educational applications of repeated tDCS to the lDLPFC, as has already been observed in repeated stimulation of the motor cortex [11] [12] [13].
5
Acknowlegdments
I would like to thank Wouter Boekel for his help with and company during testing.
References
[1] Fregni, F., Boggio, P.S., Nitsche, M., Bermpohl, F., Antal, A., Fere-does E., Marcolin, M.A., Rigonatti, S.P., Silva, M.T.A., Paulus, W. & Pascual-Leono, A., (2005). Anodal transcranial direct current stimula-tion of prefrontal cortex enhances working memory. Experimental brain research: 166(1), 23-30.
[2] D’Espostio, M., Aguirre, G. K., Zarahn, E., Ballard, D., Shin, R. K. & Lease, J., (1998). Functional MRI studies of spatial and nonspatial working memory. Brain Research, 7, 1-13.
[3] Poreisz, C., Boros, K., Antal, A & Paulus, W., (2007). Safety aspects of transcranial direct current stimulation concerning healthy subjects and patients. Brain Research Bulletin, 72, 208-214.
[4] Gandiga, P. C., Hummel, F. C. & Cohen, L. G., (2006). Transcranial DC stimulation (tDCS): A tool for double-blind sham-controlled clinical studies in brain stimulation. Clinical Neurophysiology, 117, 845-850.
[5] Paulus, W., (2011). Transcranial electrical stimulation. Neuropsycholog-ical rehabilitation, 21(5), 602-617.
[6] McKinley, R.A., Bridges, N., Walters, C.M. & Nelson, J., (2012). Modu-lating the brain at work using noninvasive transcranial stimulation. Neu-roimage, 59, 129-137.
[7] Klingberg, T., (2010). Training and plasticity of working memory. Trends in cognitive sciences, 14, 317-324.
[8] Jaeggi, S.M., Buschkuehl, M., Jonides, J. & Perrig, W.J., (2008). Im-proving fluid intelligence with training on working memory. PNAS, 108, 10081-10086.
[9] Andrews, S.C., Hoy, K.E., Enticott, P.G., Daskalakis, Z.J. & Fitzgerald, P.B., (2011). Improving memory: the effect of combining cognitive activity and anodal transcranial direct current stimulation to the left dorsolateral prefrontal cortex. Brain Stimulation: 4, 84-9.
[10] Ohn, S.H., Park, C., Yoo, W., Ko, M., Choi, K.P. & Kim, G., (2008). Time-dependent effect of transcranial direct current stimulation on the enhancement of working memory. Neuroreport, 19(1), 43-47.
[11] Fregni, F., Boggio, P.S., Lima, M. C., Ferreira, M. J. L., Wagner, T., Rigonatti, S. P., Castro, A. W., Souza, D. R., Riberto, M., Freedman, S. D., Nitsche, M. A. & Pascuel-Leone, A., (2006). A sham-controlled, phase II trial of transcranial direct current stimulation for the treatment of central pain in traumatic spinal cord injurt. Pain, 122, 197-209. [12] Boggio, P. S., Nunes, A., Rigonatti, S. P., Nitsche, M. A.,
Pascuel-Leonne, A. & Fregni, F., (2007). Repeated sessions of noninvasive brain DC stimulation in associated with motor function improvement in stroke patients. Restorative Neurology and Neuroscience, 25, 123-129.
[13] Reis, J., Schambra, H. M., Cohen, L. G., Buch, E. R., Fritsch, B., Zarahn, R., Celnik, P. A. & Krakauer, J. W., (2009). Nonivasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation. PNAS, 106(5), 1590-1595.
[14] Unsworth, N., Heitz, R.P., Schrock, J.C. & Engle, R.W. (2005). An automated version of the operation span task. Behavior research methods, 37(3), 498-505.
[15] Zaehle, T., Sandmann, P., Thorne, J. D., J¨ancke, L. & Hermann, C. S., (2011). Transcranial direct current stimulation of the prefrontal cor-tex modulates working memory performance: combined behavioural and electrophysiological evidence. BMC Neuroscience, 12(2).
