HD-tDCS mitigates the executive vigilance decrement only under high cognitive demands

Maintaining vigilance is essential for many everyday tasks, but over time, our ability to sustain it inevitably decreases, potentially entailing severe consequences. High-definition transcranial direct current stimulation (HD-tDCS) has proven to be useful for studying and improving vigilance. This study explores if/how cognitive load affects the mitigatory effects of HD-tDCS on the vigilance decrement. Participants (N = 120) completed a modified ANTI-Vea task (single or dual load) while receiving either sham or anodal HD-tDCS over the right posterior parietal cortex (rPPC). This data was compared with data from prior studies (N = 120), where participants completed the standard ANTI-Vea task (triple load task), combined with the same HD-tDCS protocol. Against our hypotheses, both the single and dual load conditions showed a significant executive vigilance (EV) decrement, which was not affected by the application of rPPC HD-tDCS. On the contrary, the most cognitively demanding task (triple task) showed the greatest EV decrement; importantly, it was also with the triple task that a significant mitigatory effect of the HD-tDCS intervention was observed. The present study contributes to a more nuanced understanding of the specific effects of HD-tDCS on the vigilance decrement considering cognitive demands. This can ultimately contribute to reconciling heterogeneous effects observed in past research and fine-tuning its future clinical application.

hypothesis) assert that the vigilance decrement occurs due to the consumption of attentional resources with time-on-task due to the demanding nature of vigilance tasks 31,32 , with the associated experience of stress [31][32][33][34] .Other accounts (underload theories) posit that the underwhelming nature of vigilance tasks, more associated with boredom 35,36 , ultimately leads to a gradually more mindless execution of the task 37,38 .These theories can be tested empirically by manipulating cognitive demands (i.e., the number of simultaneous tasks to perform and therefore, task instructions to hold in working memory).Overload theories pose that increasing task demands would lead to a greater vigilance decrement, which has indeed been observed under normal conditions [39][40][41] and found to be accentuated by sleep deprivation 42 .Underload theories, on the other hand, predict that lowering cognitive demands would lead to a less engaged and more mindless performance, steering thoughts away from the task's goal 43 , producing the vigilance decrement 44 .Further support for underload theories stems from selfreported high mindlessness predicting worse performance in a vigilance task where targets appear with low frequency 45 , reports of task-induced physiological disengagement (i.e., parasympathetic activation and reduced cardiac reactivity) 46 , and activation of DMN structures with time-on-task 47 .Given this disparity of results, Thomson et al. propose the resource-control account, wherein resources are constant, but executive control declines with time-on-task causing the progressive shift of attentional resources from task-related towards task-unrelated thoughts (mind-wandering) 48 .This account considers that other factors than task demand can modulate the vigilance decrement: observing results such as a mitigated vigilance decrement with increased perceptual variability of the task's target 49 , where higher difficulty demanding more resources is countered by higher engagement, possibly posing a smaller toll on executive control.Among other theories on the vigilance decrement (for a review see: Fortenbaugh et al. 50), some accounts represent passive fatigue and active fatigue 51 as two extremes on an inverse U-shaped function 52 between performance and cognitive load 53 or arousal 54 .These models incorporate both underload and overload as two extremes, between which we may attain a middle-ground of optimal performance 53,54 .As a case in point, Luna et al. created three load conditions (single task, dual task, and triple task) using the ANTI-Vea task 15,55 and observed that the single and triple task groups showed a significant EV decrement, which was mitigated in the dual task group 56 .This further reinforced the view that the EV decrement, present with under and over-demand, is mitigated with intermediate cognitive demands.
The current understanding of how cognitive demands affect the vigilance decrement is still unclear given the disparity of findings [39][40][41]44,46 , and the current lack of models that explain diverging results. This s further obscured by the contradictory findings when using tDCS to modulate these effects 11,57,58 .A better understanding of cognitive load-dependent effects and their interaction with tDCS effects is needed for a better translation of these results towards applied fields.Critically, a more systematic modulation of task demands and stimulation parameters is required in order to define (i) which conditions lead to a greater vigilance decrement, and (ii) critically, under which conditions the vigilance decrement can be mitigated or reduced.The potential impact of these results can branch into (i) providing a small step towards research parameters to follow for understanding and mitigating the vigilance decrement, shedding some light on the currently often contradictory findings, (ii) adapting real-life contexts to optimize performance in human factor applications where the potential negative consequences of the vigilance decrement are greatest (e.g., air traffic control or security screening 59,60 ), and (iii) provide the basis for constructing more efficient intervention or rehabilitation strategies for attention deficits such as those encountered in Attention Deficit and Hyperactivity Disorder (ADHD) 61 or as a sequelae of stroke 62 , with better informed decisions on when to use compensatory strategies (e.g., reduce task demands to adapt to a lower threshold of what would be considered overdemanding) or restitutive approaches (e.g.training program where threshold of overdemand is increased with tDCS) during rehabilitation.In order to obtain a better roadmap for these outlined applications, further replications and, specifically, more systematic manipulations of cognitive load and tDCS is needed, which was the objective of the present study.

The present study
In the present study, we applied the task manipulations performed by Luna et al., measuring vigilance in a single and dual task 56 , in combination with HD-tDCS over the rPPC, following the same stimulation protocol as Hemmerich et al. 13 .Further comparisons were made with data from the original triple task studies (standard ANTI-Vea, of two previously collected samples 13,14 ).This will allow (i) the replication of prior findings of cognitive load-dependent effects on the vigilance decrement 56 , and (ii) further understanding of whether/how these are affected by HD-tDCS.Given the specificity of HD-tDCS on the EV and not the AV effects 13,14 , and the differences in EV decrements depending on cognitive load 56 , we preregistered the following hypotheses (osf.io/9wfbx) regarding behavioural outcomes: (i) we expected a mitigated EV decrement (significantly reduced linear decrement of hits across task blocks in EV trials) in the anodal HD-tDCS group compared to the sham group performing the single load task, replicating the findings from Luna et al. 56 in the sham group, and expecting the same beneficial effect of HD-tDCS in the anodal group that had been observed under higher cognitive load 13 , (ii) no EV decrement (no linear decrement) in the dual load task, expecting to replicate the findings from Luna et al. 56 , and therefore, no expected differences between stimulation conditions, and (iii) no modulation of AV performance (i.e., linear increment of SD of RT across blocks) in any load or stimulation group (replicating the specificity observed for the stimulation intervention for EV) 13,14 .

Participants
Participants (N = 120) were randomly assigned to perform a single or dual version of the ANTI-Vea task while receiving either sham or anodal HD-tDCS.The sample size of 30 participants per experimental condition matched those of prior studies with the standard ANTI-Vea with a priori estimated sample sizes 13,14 .See Table 1 for demographic data.All participants met the safety inclusion criteria for transcranial electrical stimulation

Behavioural measures
Participants performed modified versions of the ANTI-Vea Task (as shown in Fig. 1B), where all trials of the standard task 15 were presented, but task instructions and responses were coded differently.The ANTI-Vea task is an adapted version of the classical attentional networks task 65 , that includes independent measures of the executive and arousal vigilance components.For this purpose, the task is comprised of three types of trials (ANTI, EV, and AV) that are presented in pseudorandomized order.All ANTI-Vea versions used in this study were run for 7 blocks (560 trials in total).The ANTI trials (60% of total trials) allow measuring the functioning of the classical attentional networks (alerting, orienting, and executive control) 66,67 .These trials present a flanker task where the direction of the target (i.e., a central arrow) must be detected (pressing the c-key for left-pointing arrows, and m-key for right-pointing arrows) regardless of the direction of the flankers (i.e., surrounding arrows).The EV trials (20% of the total) prompt participants to detect an infrequent and large vertical displacement of the target of the flanker task, by giving an alternative response (pressing the space bar).This sub-task would be akin to signal-detection tasks such as the Mackworth Clock Test (MCT 68 ).Lastly, AV trials (remaining 20% of trials) feature a red countdown (instead of the stimuli from ANTI or EV trials), which has to be stopped as fast as possible by pressing any key from the keyboard, akin to the Psychomotor Vigilance Test (PVT 69 ).For a more detailed description of the standard task and its parameters, please refer to: Luna et al. 15 , and Luna et al. 55 .
General task instructions across the different load conditions were given for participants to keep their gaze on the fixation point (" + ") in the centre of the screen and to respond as fast and as accurately as possible.Then, instructions diverged according to the manipulation of cognitive load, to reflect the correct response for each type of trial as depicted in Fig. 1A.While maintaining perceptual load constant, the manipulation of task instructions and response coding resulted in: (i) a single task, which required participants to respond only to EV trials, and (b) a dual task, in which participants had to respond to both EV and AV trials.These two groups were then further compared with data from (iii) a triple task, where participants had to respond to ANTI, EV, and AV trials (standard ANTI-Vea), collected from two previous studies 13,14 (N = 120).

HD-tDCS setup
HD-tDCS was applied with a Starstim 8 ® device and hybrid NG Pistim Electrodes (Ag/AgCl, contact area: 3.14 cm 2 ) controlled through NIC v20.6 software (Neuroelectrics ® , Barcelona).Five of the electrodes, placed in a neoprene headcap, were set up in a 4 × 1 ring-like array, targeting the rPPC by placing the central anode over P4, and the four surrounding cathodes over CP2, CP6, PO4, and PO8 (see Fig. 1B and C).Using a single-blind procedure, anodal (1.5 mA) or sham (0 mA) HD-tDCS was applied according to random group allocation, from the 2nd to the 6th task block (see Fig. 1D).The sham protocol consisted of two ramps (30 s ramp-up and 30 s ramp-down) at protocol onset and offset.The anodal protocol consisted of an initial ramp-up (30 s) followed by active stimulation (~ 28 min), and a ramp-down (30 s) at offset.In this study, electroencephalographic (EEG) signal was recorded during the 1st task block serving as a baseline, and during the 7th block, serving as a poststimulation measure.Further details regarding this step are beyond the scope of this report as EEG data will not be presented.

Fatigue assessment
Subjective mental and physical fatigue ratings were assessed throughout the experiment: baseline, pre-task, and post-task (see procedure or Fig. 1D).Responses were recorded through a visual analogue scale: a horizontal line ranging from minimum (left side of the screen) to maximum fatigue (right side).The assessment order for fatigue type was counterbalanced across participants but kept constant for each participant's session, following the procedure of Luna et al. 56 .Table 1.Sample sizes and demographic data for each experimental condition.(~ 28 min), mainly focused on acquiring diffusion-weighted imaging data.This data is being collected as part of a larger research project and will not be covered in the present report.Participants then sat in a separate, dimly lit room to complete the experiment.First, participants completed the baseline fatigue assessment and the ANTI-Vea's practice blocks (adapted for each load condition).After electrode set-up, participants completed the pre-task fatigue assessment.Then the experimental task started, during which stimulation was applied from the 2nd to the 6th experimental block.Right after the completion of the last (i.e., 7th) experimental block, the post-task fatigue assessment and the tES Survey 70 were completed.
Figure 1.ANTI-Vea Task procedure, electrode setup and resulting E-field simulation, and experimental procedure.(A) ANTI, EV, and AV targets of the ANTI-Vea task.The bottom table shows which target(s) participants are instructed to respond to (with a check) for the single, dual, and triple tasks.Note that perceptual load is maintained constant across all task conditions, as only instructions and response coding are modified.Note that both hands are placed over the keyboard at all times, using the left hand to press the "C" key and the right hand for the "M" key, whilst the "spacebar" key and the key chosen by the participant for AV trials can be pressed by any finger/hand (and must thus not be necessarily held constant).(B) Electrode setup for HD-tDCS: the anode is placed over P4 (red), and the surrounding cathodes over CP2, CP4, PO4, and PO8 (green), following the same protocol as

Statistical analyses
Following the preregistered plan of analysis, we analysed EV and AV data from baseline (1st block) to the final active or sham stimulation block (6th), following prior HD-tDCS studies 13,14 .Following the standard approach to ANTI-Vea scores 15 , we computed EV indices [Hits (percentage of correct responses), False Alarms (FA), Sensitivity (A'), and Response Bias (B")] and AV indices [mean RT and standard deviation of RT (SD of RT)].For EV data, we compared baseline differences in EV indices between stimulation groups using an ANOVA.Then, each index was included in an ANOVA as a dependent variable, with Blocks (1st-6th) as a within-participant factor and Stimulation Group (anodal or sham HD-tDCS) and Task Load (single or dual) as between-participant factors, followed up by partial ANOVAs for each Task Load level.Polynomial contrasts were used to analyse the linear component of each index across Stimulation Group for each Task Load level.Then, the single and dual task data, combined as a not-triple condition, were re-analysed jointly with triple-task data 13,14 , combined as a triple condition, repeating the above-described analyses (with Updated Task Load) on two balanced samples (n triple = 120, n not-triple = 119).Lastly, results for AV data are reported first considering only low-load conditions (i.e., only dual task) and then comparing low and high-load conditions (i.e., dual vs triple task, using data from the present study and data from Hemmerich et al. 13 to achieve comparable sample sizes in each group).
Note that for all reported ANOVAs, degrees of freedom are reported with Greenhouse-Geisser correction when the sphericity assumption was violated (i.e., p > 0.05 in Mauchly's test).Additionally, across results, equivalent Bayesian tests are reported to further test the validity of our inferences, as a supplement to non-significant frequentist results.Note Bayes Factors in favour of the null hypothesis (BF 01 ) provided for polynomial contrasts on the linear decrement correspond to independent or one sample t-tests completed on the Slope across Blocks (1st-6th).
Methods and Results for Subjective Mental and Physical Fatigue are reported in Appendices E-H of the Supplementary Material.

Blinding efficacy
The total amount of self-reported discomfort/sensations associated with stimulation 70 was significantly different between the Stimulation Groups, U = 2190, p = 0.037, with higher discomfort reported in the sham (M = 2.43, SD = 2.08) than in the anodal (M = 1.68,SD = 1.85) group.This difference seems to be mainly driven by the significantly higher intensity reported for pinching in the sham group (M = 0.38, SD = 0.80) than in the anodal group (M = 0.03 SD = 0.18), U = 2166, p = 0.001, without any differences for the remaining sensations (all p's > 0.136, see Appendix A of the Supplementary Material for further statistical details).The higher discomfort reported in the sham group likely led to a higher estimation of belonging to the active stimulation group in the sham (62%) than in the anodal group (42%).However, the guessed active group allocation was not statistically different between Stimulation Groups, χ 2 (2, N = 120) = 4.85, p = 0.088.Taken together with the evidence for group differences in total discomfort (BF 10 = 1.07) and pinching (BF 10 = 0.93) being anecdotal 71 at most, leads us to conclude that blinding was still effective in the present study.
The ANOVA performed on Hits in EV trials with Blocks as a within participants variable and Stimulation Group and Updated Task Load (triple/not-triple) as between-participant factors, reflected a main effect of Block, F(4.31, 513.24) = 21.42,p < 0.001, ƞ p 2 = 0.15, which interacted significantly with Stimulation Group, F(4.31, 513.24) = 3.69, p = 0.005, ƞ p 2 = 0.03.Importantly, the three-way Blocks × Stimulation Group × Updated Task Load interaction was significant, F(4.24, 999.51) = 2.97, p = 0.017, ƞ p 2 = 0.01 (For transparency, to complement the pre-registered analyses over Blocks 1-6, repeating the same analyses across Blocks 1-7, yields the same results: the main effect of Block, F(5.01, 1181.07)= 44.78,p < 0.001, ƞ p 2 = 0.16, and the critical three-way Block × Stim × Updated Task Type interaction remain significant, F(5.01, 1181.07)= 2.91, p = 0.013, ƞ p 2 = 0.01).Polynomial contrasts completed on the grouped (triple vs. not-triple) data showed that the linear decrement between the anodal and sham conditions was not different for the not-triple condition, F < 1 (BF 01 = 4.09), whereas it was for the triple task condition, F(1, 119) = 8.62, p = 0.004, ƞ p 2 = 0.07.Bayesian analyses further showed that there was moderate evidence (BF 10 = 5.66) for this mitigated EV decrement in the triple task anodal group, as can be seen in Fig. 3 (right), compared to extreme evidence (BF 01 = 145.25)against a significant interaction in the not-triple task condition, as shown in Fig. 3 (left).Lastly, there was a significant difference in the linear decrement observed between sham conditions between the not-triple and triple tasks, F(1, 117) = 7.99, p = 0.006, ƞ p 2 = 0.06, reflecting, the significantly greater EV decrement under high compared to lower load conditions.In contrast, the Finally, an ANOVA performed on SD of RT, contrasting the dual and triple conditions, showed a significant AV decrement (increment of SD of RT) across Blocks, F(4.02, 466.

Discussion
This study aimed at investigating the influence of cognitive load and HD-tDCS, as well as their interaction, on the EV decrement.To this end, we manipulated task load (single or dual) and HD-tDCS application over the rPPC (sham vs. active).Contrary to our preregistered hypotheses, we observed no differences between the EV decrement in the single and dual task conditions and no modulation of this decrement by HD-tDCS.As expected, neither cognitive load nor HD-tDCS modulated the AV decrement.Importantly, when contrasted with prior results using a triple task, we are able to expand evidence on the specific effect of rPPC HD-tDCS on the executive component of vigilance 13,14 : the mitigatory effect of HD-tDCS is only evident under conditions of high cognitive demand.
Against our pre-registered hypothesis, we did not replicate the findings of Luna et al. 56 , as the single and dual load conditions both showed a significant EV decrement with time-on-task, without any differences across load conditions.Some studies report similar null effects comparing single and dual tasks 32,72 , or no vigilance decrement at all regardless of the load condition 73,74 .However, most of the literature is either skewed towards underload (observing larger decrements with lower task demands 44 or higher engagement 75 ) or overload theories (observing greater vigilance decrements with increased task demands by adding a secondary task [39][40][41] or increasing instruction complexity 72 ), without any clear consensus.One possible explanation for our diverging results is that single and dual tasks yielded conditions that were qualitatively not sufficiently different and therefore processed similarly.Under these low to medium load conditions, available resources may suffice to (somewhat successfully) complete the task and mind-winder in parallel (maintaining the same level of performance across slightly differing demand conditions).This could be explained by the resource-control account, as executive control decreases with time-on-task, gradually tipping the balance from task-related towards task-unrelated thoughts 48 .The single and dual tasks may operate at a relatively low "tipping point".Importantly, the EV decrement has been recently linked with the loss of executive control with time-on-task in the standard ANTI-Vea (triple task) 76 .Future research systematically manipulating task demands in a within-participants design could explore: (i) whether executive control measures and the EV decrement are related when task demands are reduced, and (ii) how each load level influences the presence of task-unrelated thoughts.
Contrary to the expected mitigated EV decrement in the single group receiving active HD-tDCS and no effect of HD-tDCS on EV performance in the dual group, we observed no mitigatory -or detrimental-effect of stimulation in either the single or dual task condition.Similar results have been observed with the Sustained Attention to Response Task (SART) comparable to our single task condition: prefrontal tDCS did not affect target accuracy 57 , and anodal or cathodal tDCS over the right inferior parietal cortex (rIPL) did not affect error rates or RTs 77 .Similarly, another study reports null effects of anodal tDCS over the left PFC in a dual working memory task 58 .However, there are also some reports of detrimental effects of higher doses of both anodal and cathodal tDCS over the rIPL on accuracy in the SART 78 , and beneficial effects on accuracy with anodal HD-tDCS over the left dorsolateral prefrontal cortex (DLPFC) regardless of the task demand condition of a standard and a modified SART 79 .Lastly, it has been suggested that prefrontal tDCS may modulate sustained attention by affecting its higher-order sub-processes, rather than simple target detection 7 , which could partially explain the absence of effects of tDCS in low demanding conditions.In contrast to the null effect of HD-tDCS on the EV decrement in the low and medium load conditions, the mitigatory effect of rPPC HD-tDCS was only observed in the most demanding condition (triple task).The EV decrement in the sham triple-task condition was more pronounced than under single and dual load, which was mitigated in the HD-tDCS condition.Similar results have been observed with anodal tDCS over the right DLPFC, leading to improved accuracy under the highest load condition of a working memory task 80 , and anodal tDCS over the left DLPFC leading to delayed beneficial effects on multitasking but not on single task performance 81 .Other studies also suggest that tDCS over right prefrontal or parietal areas can lead to detrimental effects on task performance under objective 11 and subject-specific high load conditions 82 .In contrast, some studies have reported beneficial effects of cathodal tDCS for maintaining or improving performance in high load conditions 83,84 .Studies on the intersection of cognitive load and tDCS are still rather scarce and yield no clear conclusions.While the inconsistencies across the existing literature are partially explained by the variability between stimulation procedures, cognitive processes studied, and tasks used across these different studies, a crucial factor to consider is the conceptualization of cognitive load and how its levels are established.Roe et al. argue that "[…] using a load level that overtaxes cognitive capacity, as well as making use of a wider range of load levels (i.e., more than two), is preferable if one's goal is to investigate the interaction between tDCS and cognitive load" 11 .Precisely, the high load condition of our study, although complex and demanding, is not overtaxing, as was the case for the high load condition of studies reporting detrimental effects of anodal tDCS 11,82 .The effects of tDCS on the vigilance decrement are likely to depend less on the externally imposed and conceptualized levels of cognitive load, but rather on the specific demand they impose on each individual, and the specific neural state they induce 85 .Therefore, as illustrated in Fig. 5, high but manageable cognitive demands could lead to beneficial effects of anodal tDCS, as observed in the present study, where increasing neural excitability may further excite task-relevant processes.However, we hypothesize that when further increasing demands to a level where task performance cannot be maintained, the effects of anodal tDCS would be detrimental, as increasing the excitability of overtaxed neural circuits is likely to disrupt task performance.This might also explain facilitatory effects of cathodal tDCS in tasks with high demand 83,84 , where inhibitory processes could reduce over-demand.Lastly, in the lower load conditions (single and dual task), a ceiling effect of the modulatory effects of HD-tDCS may be taking place.
Another relevant result of the present study is the finding that performance gains, namely, the improved accuracy in target detection for EV trials, were due to improved sensitivity (i.e., ability to discriminate signal from noise), and not due to shifts in the response bias (i.e., the adoption of a more liberal response criterion, which would merely increase hits at the cost of increasing false alarms).While some studies do report similar results 86,87 , signal detection theory measures are not discussed in most studies exploring the effect of tDCS on vigilance, and opposite findings have also been reported showing greater sensitivity declines in less demanding tasks 88 .Thus, whilst requiring further replication, for now, our results highlight that when HD-tDCS mitigates the EV decrement (in high demand conditions), it does so by improving performance in a precise manner.
Taken together, our results further point to the fact that underlying mechanisms driving EV performance are not being properly explored with the tools at hand.As suggested above, a better understanding of what is causing Figure 5. Observed and hypothesized interaction of cognitive demands and HD-tDCS over rPPC on the accuracy performance with time-on-task (TOT), with lower values depicting a greater EV decrement.(A) Beneficial effect of active HD-tDCS over the rPPC, mitigating the EV decrement, as observed in the present study.(B) Further increasing task demands to a level that is overtaxing, would potentially lead to even worse EV performance, which could be further deteriorated by the application of active tDCS -as conceptualized and observed by Roe et al. 11 .the vigilance decrement, as would do, for example, collecting thought probes throughout the task, would help further understand the present results.Although future challenges still lie in the fact that the presence of mindwandering is not a fool-proof sign of underload, as the presence of mind-wandering does not always predict performance costs 89 , nor does the manipulation of task demands always lead to different mind-wandering rates 79 .Future research could bridge this gap by including, not only thought-probes in vigilance tasks but also including other more objective measures of engagement, such as eye movements 90 .Finally, given that the vigilance decrement can be shaped by a myriad of different factors 91 , future research should refine their approach in studying cognitive load dependent effects on vigilance, in which considering individual differences should be a key factor.
However, despite the above-mentioned limitations and open questions, the present findings can tentatively inform future decisions in research and clinical settings.The cognitive-load dependent effects of HD-tDCS on the EV decrement as observed in the present study underline the importance of considering cognitive load as an essential factor in: (i) predicting stimulation outcomes, and (ii) tailoring the interactions of demands and tDCS individually.Regarding the first point, whilst future research is needed to understand the generalizability of these results, our data suggests that in areas where a tDCS intervention is to be applied but cognitive demands cannot be modified or adapted, a prediction (based on behavioural data) could be made as of how successful a tDCS intervention would actually be.If the task is overdemanding, the intervention is likely to not adequately induce plastic changes towards the desired outcomes, whereas, if the task is under-demanding, a ceiling effect might hamper any real efficacy of the stimulation as well.While prior to such applications, further research would be needed, this consideration could be a first step in more precisely delineating the intervention and, potentially, offer a broad guideline that could avoid devoting resources to null findings.Regarding the second point, when the cognitive demands can be individually assessed and adjusted to an optimal level, the efficacy of interventions focused on the rehabilitation of attentional functions could be greatly improved.In a clinical setting, attention deficits such as those elicited by ADHD 61 or as a sequelae of a stroke 62 , could lead to the subjective and individual experience of high cognitive demands or even result in an over-taxing of resources in context that are considered to be of low demand under normal circumstances.Given that the threshold of what is considered overdemanding is not even uniform among healthy participants 58,82 , it will likely be even more heterogenous in these clinical populations.Therefore, instead of externally imposing a fixed demand, individually tailoring demand levels of cognitive training tasks to individual capacity 58,82 and gradually increasing task demands, for online use in a tDCS intervention may ensure that the neuroplastic effect of tDCS actually reinforces effective task-resolution and learning processes 85 as a restitutive approach to regain attentional functioning.

Conclusions
According to our results, the EV decrement does not seem to be modulated by cognitive load under relatively undemanding conditions (towards improved performance in the dual load group, as was reported by Luna et al. 56 ).Indeed, both single and dual load conditions showed a similar vigilance decrement across time.Under these conditions (single and dual cognitive load), additionally, HD-tDCS does not affect EV performance.However, under conditions with higher demand (i.e., triple task) there is a steeper vigilance decrement compared to lower load conditions, which was mitigated via anodal HD-tDCS over the rPPC.This study highlights the fact that task demands should be an important factor in considering the efficacy of a tDCS intervention on vigilance performance.This will allow a better understanding of the vigilance decrement in itself and facilitate a more effective translation of these results into clinical settings.

Figure 2 .
Figure 2. (A)Mean % of Hits in EV trials across Blocks for single and dual cognitive load conditions.A linear decrement across Blocks was observed across all conditions.(B) Sensitivity (A') in EV trials across Blocks for the single and dual cognitive load conditions.An effect of Blocks on A' is observed regardless of the stimulation condition, although the linear component was not significant, whilst the single task condition shows a lower mean A' (averaged across Blocks) in the anodal compared to the sham condition.Note.The dashed vertical line represents the onset of the stimulation protocol.The dotted line represents the offset of the stimulation protocol.The shaded ribbons represent the standard error of the mean (SEM).

Figure 4 .
Figure 4. AV decrement (increment of of RT with time-on-task) as a function of stimulation condition for the dual task (left) and the triplet ask condition (right).No differences between the linear increment of SD of RT across Blocks were observed between Stimulation Groups of either task condition.Note.The dashed vertical line represents the onset of the stimulation protocol.The dotted line represents the offset of the stimulation protocol.The shaded ribbons represent the SEM. https://doi.org/10.1038/s41598-024-57917-y Hemmerich et al. (C) Simulated voltage field obtained from the stimulation protocol from a top and right-hemisphere view.(D) Experimental procedure, where the bottom arrow shows the exact or approximate (preceded with a tilde) duration of each step, in minutes.Each fatigue assessment took less than a minute.