Does excitatory fronto-extracephalic tDCS lead to improved working memory performance ?

Evidence suggests that excitatory transcranial direct current stimulation (tDCS) may improve performance on a wide variety of cognitive tasks. Due to the non-invasive and inexpensive nature of the method, harnessing its potential could be particularly useful for the treatment of neuropsychiatric illnesses involving cognitive dysfunction. However, questions remain regarding the efficacious stimulation parameters. Here, using a double-blind between-subjects design, we explored whether 1 mA excitatory (anodal) left dorsolateral prefrontal cortex stimulation with a contralateral extracephalic reference electrode, leads to enhanced working memory performance across two days, relative to sham stimulation. Participants performed the 3-back, a test of working memory, at baseline, and during and immediately following stimulation on two days, separated by 24-48 hours. Active stimulation did not significantly enhance performance versus sham over the course of the experiment. However, exploratory comparisons did reveal a significant effect of stimulation group on performance during the first stimulation phase only, with active stimulation recipients performing better than sham. While these results do not support the hypothesis that dorsolateral prefrontal cortex tDCS boosts working memory, they raise the possibility that its effects may be greatest during early learning stages. Referee Status:

Transcranial direct current stimulation (tDCS) has been utilised as a non-invasive brain stimulation methodology to improve performance on a variety of cognitive tasks in healthy volunteers, including decision-making 1 , planning 2 and working memory 3,4 .Due to the minimal risk profile, arising as function of the very low current delivered to the scalp, and the relatively inexpensive nature of the device, it has high potential as a clinical tool.Tentative evidence thus far suggests tDCS may be effective in ameliorating neurological and psychiatric illnesses associated with cognitive deficits.Prominent amongst these are recent developments in the study of the treatment of addiction, depression, schizophrenia and stroke 5 .However, questions remain over stimulation condition blinding 6 , optimal stimulation frequency 7 and appropriate electrode placement 8 as these parameters can strongly influence neuronal activity and the interpretation of effects on cognitive performance.Questions have also been raised about behavioural and stimulation controls 9 .
Many studies have reported enhanced task performance in healthy volunteers following application of the excitatory anodal electrode to a location of the scalp corresponding to a task-relevant region of the brain, including: the temporal parietal junction to enhance social cognition 10 ; inferior frontal cortex to enhance target detection 11,12 ; Wernicke's area to enhance visual picture naming 13 ; and dorsolateral prefrontal cortex (DLPFC) to enhance decision making 1,14 and working memory 3,4,7 .The application of the excitatory tDCS electrode to the scalp is thought to cause an increase in neuronal excitability in the stimulated area by altering the resting potential 15 .To complete the electrical circuit, the reference or inhibitory cathodal electrode must be placed somewhere on the head or body being stimulated.
The majority of studies exploring cognitive enhancement using tDCS have targeted DLPFC as their region of excitation while the inhibitory electrode has typically been placed on the contralateral supra-orbit (or DLPFC).For example, in a single blind investigation using this electrode montage, Ohn et al. 4 found that 30 minutes of 1 mA tDCS while participants performed the n-back task led to significant improvements in task performance over sham during stimulation only.However, placing the reference electrode on the scalp introduces a potential confound in the interpretation of any resulting behavioural effects: these could arise as a result of excitation, inhibition, or a combination of the two electrodes.The location of the reference electrode, whether extra or intra-cephalic, has been show to play a prominent role in the efficacy of the excitatory electrode 16 .
The n-back 17 is a cognitive task commonly used to assess aspects of executive function, and is thought to engage working memory in particular.Although criticisms have been made with respect to its construct validity 18 , the n-back has frequently been used in the context of functional neuroimaging experiments in both healthy volunteers and patients with psychiatric and neurological illnesses.The results of these studies consistently implicate a network of brain regions including parietal cortex and DLPFC, which are engaged with increasing load during n-back performance 19 .Importantly, altered DLPFC function is associated with several psychiatric conditions [20][21][22] .For example, research using the n-back has identified that the DLPFC is hyperactive in patients diagnosed with major depressive disorder (MDD) 23,24 , suggesting that MDD patients may need to use greater resources to achieve the same level of performance.Similar findings have been reported in schizophrenia 25 .
Working memory dysfunction, and executive function deficits more broadly, have been found across a number of psychopathologies, including attention deficit disorder, autistic spectrum disorder, traumatic brain injury, Alzheimer's disease, schizophrenia and depression [26][27][28] .Indeed, executive function performance has been identified as a tractable endophenotype to explore across neuropsychiatric illnesses due to its prevalence, particularly in depression and schizophrenia 29,30 .Thus, there is significant potential for non-invasive brain stimulation techniques such as tDCS to be applied clinically to ameliorate cognitive dysfunction.However, in order to establish whether tDCS has the potential to improve clinical conditions through modulatory effects on executive function, it is pertinent to first establish the effects of specific stimulation parameters in healthy volunteers.
A recent single-blind within-subjects investigation by Zaehle et al. 31 demonstrated that 15 minutes of 1 mA excitatory tDCS applied to left DLPFC during rest, with the inhibitory electrode placed ipsilaterally at the mastoid, resulted in enhanced post-stimulation performance on the 2-back task in comparison with cathodal, but not sham, stimulation.The position of the reference electrode can however affect the underlying cortical excitability 16 .Importantly, Zaehle et al. 31 utilized a fronto-extracephalic montage, which attenuates interpretational difficulties; but it remains unknown whether contralateral or ipsilateral positioned reference electrodes or bifrontal montages are superior in the stimulation of DLPFC.Furthermore, it remains to be determined whether stimulation during a task or while at rest is more beneficial.Andrews et al. 3 found tentative evidence to suggest that DLPFC tDCS applied concurrently with a cognitive task may provide more robust effects on subsequent working memory performance that stimulation during rest.
Here, we sought to build on prior research 3,4,31 by conducting a double-blind between-subjects experiment to examine whether excitatory DLPFC tDCS applied across two days would lead to enhancement of n-back performance during and post-stimulation.Specifically, we assessed whether excitatory fronto-extracephalic DLPFC tDCS, with the reference on the contralateral cheek, could improve performance on the 3-back in healthy volunteers across two stimulation days.We hypothesized that those receiving active stimulation would have greater task performance improvement, relative to baseline, in comparison with sham stimulation recipients.

Participants
Twenty-one (14 females, M = 23.09years, SD = 3.95) righthanded participants were recruited from the Psychology subject pool at University College London, UK.Participants self-reported no history of mental or neurological illness, current psychiatric medication use, no prior or current participation in another brain stimulation experiment within the previous 24 hours and had normal or corrected to normal vision.There were no significant age (t (19) = 0.211, P = 0.835) or gender (X 2 (1) = 1.527,P = 0.361) differences between the two stimulation groups (active and sham).All participants provided written informed consent and were compensated for their time.The study was approved by the UCL ethics committee.
Task.The n-back (Figure 1) consisted of a continuous sequence of 300 (150 for baseline) centrally presented consonants (500 ms) interleaved with fixation crosses (1500 ms).Participants were instructed to respond to every appearance of a letter, pressing the 'H' key only when the letter onscreen matched the letter 3-back, and pressing the 'F' key for all other instances.It is thought that this version of the n-back may afford increased sensitivity to working memory performance than versions that focus solely on hits 31,32 .Matches (one-fifth of all stimuli) and non-matches to 3-back stimuli were randomized in order but their ratio was fixed throughout the experiment.tDCS tDCS was administered continuously at 1 mA using the Neuroconn DC-Stimulator (Neuroconn, Germany) via a pair of rubber electrodes (7 cm × 5 cm) housed in small synthetic sponges dampened with salt water to increase conductivity.The excitatory (anodal) electrode was placed over F3 (Figure 2A), corresponding to the left DLPFC, while the reference (cathodal) electrode was placed on the contralateral cheek.F3 was located using a 10-20 electroencephalography cap and demarcated using a removable marker.
Left DLPFC was chosen as the anodal electrode position as this region has been consistently implicated in working memory paradigms 19 .Additionally as the task involved processing static letters, the left side of the brain was considered most appropriate 33 .Once the area was located, the electrodes were fastened in position using two headbands (a polyester hairband across the forehead and a rubber band beneath the jaw and around the circumference of the head; Figure 2B).
Before arrival, participants were randomized to one of two brain stimulation conditions using Matlab, active (N=10) or sham (N=11).Specific codes were selected from the tDCS device manual by an independent researcher not involved with the study and were assigned to each condition and randomized to each participant.Importantly, utilizing the 'study mode' of the device allowed the stimulation-administering researchers to remain blinded to the condition participants were in as the readout on the stimulation apparatus was identical for both active and sham stimulation.The administered current was applied for 10 minutes with an additional 15-second fade-in and fade-out ramping period to minimize discomfort and facilitate participant blinding.Sham stimulation was limited to small pulses of 100-200 μA every 400-550 ms between a 15-second fade-in and fade-out voltage ramp 34 .

Study design
During the baseline session on day 1 (D1), participants were first trained on the n-back with a brief exposure to 1, 2 and finally 3-back.
Thereafter, participants completed a 5-minute version of the 3-back, which served as a baseline pre-stimulation measure of performance.Immediately after, tDCS was administered for 10 minutes while participants performed the 3-back task (D1 tDCS; Figure 2C).Following this, participants completed a further 10-minute session of the 3-back (D1 post-tDCS) without stimulation.Participants were instructed that continuation to day 2 was dependent on task performance but were not given feedback until the end of day 1.Continuation to day 2 was dependent on above chance performance, which was any positive d' value:

Statistical analyses
To assess the effect of active stimulation versus sham over time, we conducted a linear mixed model in SPSS, version 21 (IBM Corp New York 2012).The dependent variable was d'.Follow up models also were conducted using hit rate, correct rejection rate (1 -false alarm rate) and reaction time for both of these variables.The four testing sessions after baseline (D1 tDCS, D1 post-tDCS, D2 tDCS, D2 post-tDCS) were entered as a fixed effect of time; tDCS and sham stimulation were entered as a fixed effect of group; and their interaction (time-by-group) was also entered as a fixed effect.Participant number was entered as a random effect and baseline 1 participant (active group), approximately 40% through the task.These data were included in the model, and the participant completed a further post-tDCS test, which was used only to determine progress to day 2; these data were not included in the model.There was no significant difference in d' performance between the groups at baseline (t (19) = 1.044,P = 0.309; Figure 3).As expected, there was a significant main effect of time (F (3,36) = 7.669, P < 0.001) on d' performance, reflecting improvement across both groups with increasing exposure to the task.However, contrary to our hypothesis, no main effect of stimulation group was identified (F (1,16) = 2.228, P = 0.155) and there was no group × time interaction (F (3,36) = 1.339,P = 0.277).
Exploratory Bonferonni corrected pairwise comparisons were carried out to assess group performance differences at the four post-baseline time points.A significant difference between active performance was entered as a covariate.A heterogeneous first order autoregressive covariance structure was employed.Bonferonni corrected tests between the groups at each time point were conducted using linear contrasts to assess between-group differences.Follow up assessments of significant points were assessed using a general linear model with baseline performance entered as a covariate.Performance differences at baseline were assessed using an independent sample t-test.Based on our sample size we had 80% power to detect a large effect size (d=1.3)at P = 0.05 (two-tailed) between the stimulation groups.

Results
One participant (sham group) scored a negative d' value for day 1 and did not participate in session 2, but their data from day 1 were included in the linear mixed model.Additionally, the testing computer malfunctioned during the day 1 post-tDCS assessment for and have also been targeted as potentially tractable endophenotypes in both schizophrenia 35 and depression 29 .Although tDCS has been suggested as a therapeutic tool for many cognitive and neurological impairments, very few tDCS studies have conducted double-blind assessments in either clinical or non-clinical populations.Furthermore, the specific ameliorative stimulation parameters, such as amplitude, frequency and electrode positioning, are largely undefined.Importantly, few electrode montages have been tested 1,31,36,37 , fewer studies yet have applied tDCS for greater than one session 38-40 and many of the variables in the available parameter space have not been subject to systematic manipulation 9 .Contrary to our hypothesis, no effect of tDCS stimulation was identified in this study.However, exploratory tests did suggest that active stimulation was associated with enhanced performance relative to sham stimulation during the first stimulation period on day 1 only.Our results may therefore indicate that the performance enhancement effects of excitatory tDCS may be limited to earlier stages of learning 40 .They also suggest that reports of improvements after one session of tDCS -the most common report in enhancement studies -may not translate to continual improvement with additional stimulation.
Our results do not support the hypothesis that excitatory tDCS applied to DLPFC results in post-stimulation improvement on the n-back task across multiple days.This result is consistent with some previous research; in comparison with sham stimulation, sham stimulation was identified at the day 1 tDCS time point (F (1,13.373)= 10.747,P = 0.006; controlling for baseline performance and Bonferonni corrected for multiple comparisons), indicating a large effect size (Cohen's d = 1.427, r 2 = 0.337).No other stimulation group differences in d' were found at other time points (all F < 1.2, P > 0.3; see Table 1 for group performance across sessions and task components).Further analyses of performance during stimulation during day 1 (including baseline as a covariate) revealed that both the hit rate (F (1,18) = 4.454, P = 0.049, ηp 2 = 0.198) and correct rejection rate (F (1,18) = 3.680, P = 0.071, ηp 2 = .170) of active stimulation recipients were significantly, or at trend level, better than sham.demonstrated significant performance enhancements on the n-back task immediately following excitatory DLPFC tDCS.Nevertheless, we did find evidence for a specific improvement in performance during stimulation on day 1 only, an outcome consistent with results from Ohn et al. 4 and others 41 .Andrews et al. 3 found that DLPFC excitatory tDCS applied during a working memory task (n-back) led to significant improvements in post-stimulation performance in comparison with baseline on an alternative working memory task (digit span forward but not backward).The improvements found 3 were not present for either sham stimulation in conjunction with task performance or stimulation without task performance.Behavioural data were not reported for the task during stimulation and an intracephalic reference electrode was used, limiting direct comparison with the present study.Furthermore, Hoy et al. 7 found that 1 mA excitatory tDCS applied to DLPFC at rest resulted in an enhancement in 2-back reaction times 40 minutes post-stimulation, but found no improvement in accuracy.However, other reports have found evidence for more enduring cognitive enhancement following tDCS 2,39 (but see Walsh 9 ).

Testing Times
This discrepancy between results may reflect the different tasks, electrode montages, stimulation parameters, sample sizes and study designs used.For example, it is possible that the payment schemes that served as a performance motivator here limited the potential to observe performance enhancing effects of tDCS.As there was no monetary motivation during the stimulation phase on day 1, participants may not have exerted themselves fully and thus the effects of stimulation may have had greater leverage; while on day 2, participants in both groups may have reached a level whereby any potential for further enhancement of performance through tDCS was limited.
Whilst the sample size used here is low for a between-subjects study, few tDCS studies have thus far been conducted using large sample sizes, and future studies should address the issue of stimulation parameter optimization using large sample sizes.Nevertheless, our results suggest that tDCS may be particularly sensitive to earlier stages of learning 40 .
In theory, the beneficial effects of tDCS may be most pronounced in poorer performers.Indeed, there is some evidence that tDCS may be particularly useful as a cognitive enhancer with lower performing individuals 42 .As the population utilised here primarily comprised students from University College London, between-group differences arising as a function of tDCS may have been attenuated due to high initial baseline ability.Finally, evidence suggests that individual differences in genotype may play a large part in susceptibility to the plasticity enhancing capabilities of tDCS.Fritsch et al. 43 found that tDCS was more efficacious in both mice and humans possessing the homozygous Val/Val genotype of the brain-derived neurotrophic factor polymorphism (rs6265), than Met carriers, though we did not have a sufficiently large sample to explore such moderators in the current study.
The results of this experiment require careful replication and extension to validate the potential role for tDCS in executive function enhancement.In particular, the evaluation of result specificity represents a prominent hole in the current literature.Few studies thus far have contrasted active stimulation results in comparison with control tasks and active stimulation of control site locations on the scalp 9 ; such measures would be beneficial in assessing the findings here and across the field.Additionally, it could be fruitful to replicate this experiment without the monetary incentive.Testing a larger and more representative sample including non-university students would also be informative.Furthermore, while performance improvements under stimulation are important, the clinical In conclusion, our results do not support the hypothesis that excitatory tDCS applied to the left DLPFC using a contralateral fronto-extracephalic electrode reference produces consistent improvements in executive function beyond the period of stimulation.Nonetheless, we found a beneficial effect of tDCS during task performance only when the task was relatively novel, which could be interpreted as indicating that this particular electrode montage, stimulation voltage and study design may be best suited to early stages of learning.
The statistical analyses conducted have been well detailed and I believe are appropriate for the questions of interest.The limitations of the study have also been adequately discussed.The authors rightly point out that the way in which the monetary incentive for participation was managed could have differentially affected performance between the two testing days, thus possibly obscuring any enhancing effects of tDCS.As also noted, the fact that the sample was primarily comprised of university students may have lessened the likelihood of observing tDCS effects, although Martin .similarly recruited a university et al student sample.Perhaps utilising a more difficult task, for example a dual n-back task or adjusting the n value based on participant performance on the previous trial block, may have produced greater differences in performance and allowed any potential effects of tDCS to be borne out.
Regarding other methodological aspects, the authors applied quite an unusual placement for the reference electrode on the contralateral cheek; one that I have never encountered before.This would not be considered an extracephalic montage as conventionally, an extracephalic electrode is one that has been placed off the head (e.g., deltoid).I am curious as to why the authors chose this montage and if there is any modelling data to support this.Also, were participants asked to guess their stimulation condition after each session?I note that a double-blind design was used but the success of blinding should still be tested.
In short, I commend the authors in investigating the potential uses of tDCS as a clinical tool but given the above limitations, changes in methodology would need to be implemented before drawing firmer conclusions.We thank Dr. Alonzo for the detailed review, kind comments and suggestions.We have altered the title of the article to better reflect the electrode locations used here.Furthermore, we have attempted to address concerns surrounding the electrode montage by adding references to articles that have used the contralateral cheek as a reference, and added a paragraph to the discussion section detailing the issue of the reference electrode in tDCS experiments.As astutely indicated by Dr. Alonzo, the montage used is somewhat novel and current modelling studies comparing this configuration to other more established montages have yet to be performed.no effect of tDCS was identified I have read this submission.I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

References
The experiment appears to be very well done and the writing clearly explains their process.My primary concern is that the authors conducted this experiment using a very small sample size (which they acknowledge, and provide power estimates that are quite reasonable).However, tDCS effects in healthy young university students tend to be quite modest.Ideally, the authors would replicate the experiment with a larger sample and/or with a secondary WM task (apart from the baseline covariate) to identify any heterogeneities in the population that could be obscuring equal and opposite effects.More importantly, to provide greater generalizability in their data, it would be good for them to compare the cheek site with the commonly used orbital site.This would allow greater comparison between the current study and the existing research, and strengthen their point that different paradigms may elicit different effects.
My second point is that the authors do not really return to the issue of the second electrode site as a potential explanation for their modest effects.One possible explanation would be that the previous work stimulating the contralateral supraorbital region provided stimulation elsewhere in the PFC which contributed to the WM benefit reported by others.Here, current modelling would be particularly helpful to compare the flow of current between the L. DLPFC-R.cheek vs. L-DLPFC-R.supraorbitalmontages.However, it is clear that the full parameter-space of tDCS effects requires increased clarity and a forum to avoid the 'file-drawer problem' associated with neurostimulation and small effect sizes.Overall therefore, these data will be of interest to the tDCS research community.
I have read this submission.I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
No competing interests were disclosed.Competing Interests: d' = Z(hit rate) -Z(false alarm rate) (1) where hit and false alarm rate are the number of correct or incorrect 'H' responses, respectively, divided by the total number of opportunities (1/5 or 4/5 of total stimulus letters) and Z is the inverse of the cumulative Gaussian distribution.Participants received £10 for their participation on day 1 irrespective of task performance.Day 2 (24-48 hours later) consisted of one 10-minute task run with stimulation (D2 tDCS) and one post-stimulation (D2 post-tDCS).Participants were told that performing better than the test phase of day 1 would result in a bonus of £10 on top of the £5 basic payment on day 2.

Figure 1 .
Figure 1.Schema of 3-back task.Stimuli (consonants) were presented centrally for 500 ms and followed by a fixation cross for 1500 ms.Participants were instructed to respond to every stimulus, indicating whether the stimulus matched the letter 3-back ('H') or not ('F').

Figure 2 .
Figure 2. tDCS electrode montage and study design.A, B) The excitatory anodal electrode (red) was positioned using a 10-20 standard electroencephalography electrode cap under F3, which corresponds to dorsolateral prefrontal cortex.The inhibitory cathodal electrode (blue) was positioned on the contralateral cheek.C) Timeline of events in the study.Participants performed the 3-back on five separate occasions, once at baseline on day 1, twice during and twice following stimulation on both days.

Figure 3 . 3 -
Figure 3. 3-back d' performance (mean values) across testing times and days.The active stimulation group always performed better than the sham group but only statistically significantly so during stimulation on day 1 (D1 tDCS), denoted by an asterisk (*).Baseline performance did not differ between the groups but was included in the model as a covariate.Error bars represent ±1 standard error of the mean.
Finally, we thank Dr. Alonzo for the suggested article ( of Clinical Neurophysiology, Göttingen University, Göttingen, Germany Introduction: The authors cite previous studies (e.g., Zaehle ., 2011) as an example that the applied et al electrode montage makes the interpretation of the effect of previous tDCS results difficult, due to the confounding factor coming from the cephalic or extracephalic position of the return electrode.However, the present study addresses this question relatively shortly, limiting the question to the extracephalic reference electrode position.Please rewrite this sentence.Why did you include 21 subjects?Did the task last for exactly the same time for each participant Method: (irrespectively of the individual differences in the reaction time)?Why did the participants receive performance-dependent extra monetary reward on Day 2 only?It makes the comparison between Day 2 and the previous session rather difficult.

3-back scores of participants before, during and after tDCS stimulation to their left dorsolateral prefronatal cortex 124
Data Files http://dx.doi.org/10.6084/m9.figshare.818974Discussion Deficits in executive function, including working memory, have been implicated in many neurological and psychiatric conditions neither Zaehle et al. 31 nor Ohn et al.

Table 1 . Means and standard deviations of each 3-back session per group
. D1 = day 1, HR = hit rate, CRR = correct rejection rate, RT = reaction time, CR = correct rejection.

have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
1. Martin DM, Liu R, Alonzo A: Can transcranial direct current stimulation enhance outcomes from cognitive training?A randomized controlled trial in healthy participants..