Optimizing cognitive control through the interaction between stimulation intensity and duration in single-site and dual-site tDCS

Abstract

In the present study, we investigated the relationship between tDCS dosage and the effects of single-site and dual-site tDCS. In Experiment 1, two types of stimulation intensities (1 mA or 1.5 mA) were applied while participants performed a Flanker task. In Experiment 2, two different stimulation durations were manipulated while participants conducted the same task. Behavioural performance under different dosages was evaluated in the framework of both traditional reaction time measures and the drift diffusion model (DDM), respectively. Conventional reaction time analyses did not reveal any significant effect of stimulation intensity and stimulation duration on cognitive control task. In contrast, DDM analyses revealed that the detrimental effects of high-intensity tDCS diminished as tDCS continued. Additionally, these results showed that high-intensity tDCS in the single-site condition led to longer non-decision time, indicating an interruption to the perceptual and motor processes, while for dual-site tDCS, both the perceptual process and evidence accumulation process were interrupted with longer non-decision time and slower drift rates. This study underscores the importance of understanding the effects of tDCS from a temporal perspective, providing important implications for optimizing tDCS parameters and understanding the underlying mechanisms of using tDCS to enhance cognitive control.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-14509-8.

Introduction

Transcranial direct current stimulation (tDCS), as a non-invasive neuro-modulatory technique, has been commonly used to investigate the causal relationships between brain regions of interest and their underlying cognitive functions. In recent decades, an increasing body of tDCS studies have attempted to examine the neural correlates of cognitive functions by altering performances across a range of cognitive tasks on language processing (e.g^1–3, working memory (e.g⁴., and motor skill acquisition (e.g⁵). However, only a few studies have shed light on the effects of different tDCS parameters such as stimulation intensity or stimulation duration^6–8.

According to previous studies^8,9, tDCS modulates cortical excitability by regulating the resting membrane potential, which in turn, changes how likely they are to fire. Anodal tDCS typically enhances cortical excitability by depolarizing the neurons, which facilitates synaptic transmission, while cathodal tDCS tends to hyperpolarize the neurons, decreasing cortical excitability¹⁰. According to the physics of tDCS, the intensity of current flow in the brain exhibits a linear increase corresponding to the applied current¹¹. More specifically, the effects of tDCS depend on various parameters, including current intensity and duration, which has been theorized to determine the degree of excitability changes¹². Nevertheless, it remains inconclusive about the relationship between stimulation intensity and stimulation efficacy. For example, Karuza et al¹³. found that the Flanker interference effect was significantly reduced by stimulating the DLPFC with 1.5 mA of anodal tDCS. Conversely, Zmigrod and colleagues¹⁴ examined the effect of 2 mA anodal tDCS over the dorsolateral prefrontal cortex (DLPFC) on interference inhibition using the Flanker task, but found no significant effect. So far, few studies have directly compared the effect of different stimulation intensities. In a study using a sensory-motor decision-making task, Ehrhardt et al¹⁵. examined the impact of current intensity (i.e., 0.7 mA, 1 mA and 2 mA) on decision making and found significant improvements in both reaction time and accuracy after 1 mA tDCS but not 0.7 or 2 mA tDCS. Another study on motor learning suggested a significant enhancement in motor performance only after 1.5 mA tDCS but not 1 mA¹⁶. Overall, distinct stimulation intensities may yield various outcomes, but the findings thus far have been inconsistent.

Another potential factor that impacts the efficacy of tDCS is stimulation duration⁹. In a study using the stop signal task, Mansouri and colleagues¹⁷ examined the impact of 1.5 mA anodal tDCS over the DLPFC for 10 min on cognitive control, and found a significant decrease in stop signal reaction time, indicating enhanced cognitive control. Conversely, Stramaccia et al¹⁸. observed no significant improvement by administering 1.5 mA anodal tDCS to the DLPFC for 20 min before the same task. This discrepancy in behavioural outcomes emphasizes the need for further investigation into the influence of stimulation duration within a single study.

In addition, the interaction between tDCS intensity and duration is not necessarily cumulative. At higher intensities, extended stimulation may engage homeostatic plasticity changes, which could potentially reverse the intended excitatory effects. Thus, the effect of stimulation depends not only on each parameter individually but also on how the interaction between parameters to influence cortical excitability over time. Based on the physiological and empirical evidence stated above, we hypothesized that stimulation outcomes would differ depending on the interaction between intensity and duration.

Previous studies have provided inconsistent evidence regarding the effect of intensity and duration of tDCS. One possible reason is that most previous studies rely on conventional measures like reaction times (RT) and accuracy rate to examine overall performance. The impact of stimulation on cognitive performance may result in improvement in different subcomponents of cognitive processes. Therefore, it is essential to further decompose RT and accuracy data. The drift diffusion model (DDM), a well-established computational model of the decision-making processes in a forced two-choice problem, permits a quantitative latent variable framework to explain human decision-making behaviour^19–21. Under the DDM framework, the decision-making process is theorized into separate parameters including drift rate, and non-decision time (for a comprehensive review, see²². This attempt may be beneficial in uncovering more nuanced changes with varying tDCS parameters, which, in turn, contributes to more precise parameters of tDCS for both clinical and theoretical purposes.

Furthermore, most research has mainly applied tDCS over one site to establish causal relationships between brain regions and cognitive function (e.g^1,23–25. Contrary to the documented efficacy of single-site tDCS on the motor cortex²⁶, there have been inconsistent findings regarding the effect of single-site tDCS on more complex cognitive functions, such as language processing (e.g^27,28, and cognitive control (e.g^3,29). Given that complex cognitive functions depend on the interaction and integration of local and remote brain regions through highly connected networks³⁰, dual-site anodal tDCS may increase more cortical excitability than single-site³¹. Therefore, it is worth investigating the potentially different effects of single-site versus dual-site tDCS.

To this end, in the present study, we aimed to investigate the effects of different stimulation intensity and duration of anodal tDCS on cognitive processing by utilizing DDM alongside traditional behavioural measures. Additionally, we aimed to explore the differences in stimulation effects produced by single-site versus dual-site tDCS. This attempt could establish a quantitative relationship between tDCS dosage and behavioral performance. The Flanker task³², a commonly used paradigm to tap into cognitive control³³, was adopted. It has been shown that the preSMA/SMA is more strongly activated for incongruent trials compared with neutral trials^34–37, and the cerebellum is also involved in cognitive control associated with the Flanker task^38–41. A structural mapping between the cerebellum and the cerebral cortex has also been demonstrated⁴². Therefore, in the present study, the preSMA (single-site stimulation) and preSMA-cerebellum (dual-site stimulation) were selected as the target regions to investigate the effect of tDCS on the Flanker interference effect. Specifically, In Experiment 1, we examined the impact of 1 mA and 1.5 mA tDCS on RTs and DDM parameters including both drift rate and non-decision time^22,43. We chose the intensities of 1.0 mA as lower intensity and 1.5 mA as higher intensity based on previous literature documenting neuro-physiological and cognitive differences between these two intensities (e.g^16,44–46. For example, Khalil et al. (2023) investigated the effect of tDCS intensity on a perception task labelling 1.0 mA as low intensity and 1.5 mA as high intensity and found that performance in the task improved only after 1 mA but not 1.5 mA tDCS. In Experiment 2, we explored the effects of stimulation duration, and expected stronger effects with longer stimulation, observed through traditional measures and DDM parameters.

Methods

Participants

All methods in this study were carried out in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the Imaging Center for Brain Research of Beijing Normal University. Informed consent was obtained from all participants prior to their participation in the study. Each experiment involved a separate group of participants. All participants were right-handed, had normal or corrected-to-normal vision, and reported no history of neurological or psychiatric disorders. A priori power analysis using G*Power version 3.1.9.7⁴⁷ was conducted to determine appropriate sample sizes. For Experiment 1, a sample size of N = 40 was calculated to detect a medium effect size (0.3) with a significance of α = 0.05 and power = 0.80⁴⁸. To ensure sufficient power and allow for potential exclusions, we recruited 48 participants for each of Experiment 1a and 1b (i.e., 24 per intensity group in each experiment). For Experiment 2, a similar analysis indicated a minimum of N = 17 would be sufficient for a medium effect size using the same statistical criteria. 24 participants were recruited for each of Experiment 2a and 2b. It should be noted that participants in Experiment 2 were in the same cohort as the high intensity group in Experiment 1. Specifically, after completing the first block of the Flanker task (under stimulation conditions lasting 10 min), participants continued with a second block after extended stimulation (20 min total). Thus, Experiment 2 used a within-subject design of stimulation duration using the same sample. A modest monetary remuneration was paid after their participation. Descriptive statistics, including the mean age and standard deviation for each group, are provided in Table 1.

Table 1.

Participants Demographics by Experimental Condition.

Condition	tDCS Intensity	n	Age (Mean ± SD)	Gender (Female)
Single-site	1.0 mA	24	22.29 ± 2.54	13
	1.5 mA	24	21.13 ± 2.73	16
Dual-site	1.0 mA	24	22.38 ± 2.65	10
	1.5 mA	24	21.29 ± 1.70	14
Single-site	1.5 mA	24	21.13 ± 2.73	10
Dual-site	1.5 mA	24	21.29 ± 1.70	14

Materials

Stimuli consisted of five horizontally arranged arrows or straight lines, with the central (target) arrow in the center of the screen. In congruent conditions, all arrows faced the same direction (e.g., < < < < < or > > > > >); in incongruent conditions, flankers faced the opposite direction as the target (e.g., < < > < < or > > < > >). Neutral trials had non-directional line flankers (e.g., - - < - - or - - > - -). The Flanker task included one block of 90 trials (30 per condition) in a pseudorandom order, with black stimuli (RGB: 0; 0; 0) against a white background (RGB: 255; 255; 255).

Procedure

All participants provided written informed consent and completed a standard tDCS safety screening prior to the experiment. They then received 20 min of transcranial direct current stimulation (tDCS) while performing the Flanker task, which was used to assess cognitive control. The task was administered in two blocks, each consisting of 90 trials, lasting approximately 4–6 min. The task was administered in two halves with a break of approximately 3 min between blocks. Only participants in the high-intensity group completed two blocks of the Flanker task with the first block starting 3 min after tDCS started and the second block starting after 10 min of tDCS. No other cognitive tasks were presented during this break, and participants remained seated quietly throughout the whole process. Stimulation was initiated 3 min prior to task onset and continued throughout task performance, in line with standard. For clarity, the first block is referred to as the first half and the second as the second half, which were analyzed and compared in Experiment 2.

Although all participants completed both blocks, only data from the first half were analyzed in Experiment 1 (intensity manipulation), while Experiment 2 (duration manipulation) included both the first half and second half in the analysis.

Each trial began with a fixation cross (“+”) presented for 300 ms, followed by a blank screen for 200 ms. The flanker stimulus was then displayed and remained on screen until the participant responded. After a 500 ms blank screen, the next trial began. Participants were instructed to indicate the direction of the central target arrow as quickly and accurately as possible using the ‘F’ key for left-facing targets and the ‘J’ key for right-facing targets, while ignoring the flanking arrows on either side. Stimuli were presented, and responses were recorded using E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA; https://pstnet.com/products/e-prime/). At the end of the experiment, participants completed a debriefing form regarding their experience. The entire session lasted approximately 30 min.

tDCS protocol

For the current study, direct current stimulation was delivered using a high-definition stimulator (Soterix Medical) with 12 mm AgCl electrodes mounted in an HD-Cap. The setup of high-definition tDCS (HD-tDCS) involves both the anodes and cathodes to deliver more precise currents to the targeted brain region. For anodal HD-tDCS used in the current study, the anode was used to increase cortical excitability at the targeted site, while the cathodes were used to complete the electrical circuit rather than produce significant inhibitory effects. Simulated electric fields distribution and montage placement are shown in the Supplementary Material Figure S1.

Experiment 1 stimulation intensity

Stimulation lasted for 20 min with 30-second fade-in and fade-out periods. Participants were instructed to inform the researchers if they experienced any discomfort, but no one requested to stop the stimulation. In Experiment 1a (single-site), 1 mA or 1.5 mA was applied to the preSMA, with the anode placed at FCz and cathodes at Cz and FC2, according to the 10–10 EEG international system). For Experiment 1b (dual-site), stimulation targeted both the preSMA and the left cerebellum. One anode was placed at FCz (with cathodes at FC2 and Cz), and the second anode at PO9 (with cathodes at PO7 and O9) with 1 mA or 1.5 mA per anode, following the same electrode system.

Experiment 2 stimulation duration

In Experiment 2a, the stimulation setup was identical to Experiment 1a (single-site), but the intensity was fixed at 1.5 mA. Anodal stimulation was applied over FCz, with cathodes placed at Cz and FC2, following the 10–10 EEG system. In Experiment 2b, the setup matched that of Experiment 1b (dual-site), again with a fixed intensity of 1.5 mA per site. Anodes were placed at FCz and PO9, with corresponding cathodes at FC2/Cz and PO7/O9, respectively. In both 2a and 2b, stimulation lasted 20 min with 30-second fade-in and fade-out periods. Participants performed two blocks of the Flanker task during stimulation: the first block began after 3 min (the first half) and the second block after 10 min (the second half).

Data analysis

Data exclusion and analysis procedures were consistent across all experiments. Only correct-response reaction times (RTs) were included in the analysis. RTs shorter than 200 ms or longer than 1500 ms were removed, along with RTs exceeding 2.5 standard deviations from each participant’s mean. Accuracy was uniformly high across all experiments (above 90%), and error trials were excluded from the RT analyses. In Experiment 1a, 1.13% of trials were removed due to incorrect responses, 2.08% due to RTs exceeding 2.5 standard deviations, and no trials were removed based on absolute RT thresholds. In Experiment 1b, 1.93% of trials were excluded as RT outliers, with no trials excluded based on timing thresholds; error rates were not separately reported. For Experiment 2a, 1.20% of trials were excluded for incorrect responses, 1.99% as outliers, and 0.02% due to being too short or too long. In Experiment 2b, 1.65% of trials were incorrect, 2.03% were RT outliers, and none were removed due to absolute RT criteria. Only incongruent and neutral trials were analyzed to measure inhibitory control⁴⁹. Results including congruent trials for Experiment 1 and Experiment 2 are provided in Supplementary Material Figure S3 and S4, respectively. The HDDM toolbox in Python 3.8 was used to fit RTs and accuracy with DDM⁵⁰, analyzing drift rates and non-decision times²⁰. To extend the main analyses, we also included an analysis of the Flanker Effect for each study. Specifically, the Flanker Effect was calculated for response time, drift rate, and non-decision time to capture both the behavioural and computational indices of cognitive control. The full analysis and corresponding figure are presented in the Supplementary Material (see Supplementary Material Figure S2).

Experiment 1 Stimulation Intensity.

Experiment 1a. Single-site stimulation.

Results

A two-way ANOVA on RTs (Fig. 1a) showed a significant main effect of trial type, F(1, 46) = 29.46, p <. 001, η_p² = 0.39, with slower responses in incongruent (M = 423 ± 55 ms) than neutral trials (M = 408 ± 52 ms). Neither the main effect of intensity nor the interaction was significant, Fs(1, 46) < 1.

Fig. 1.

Mean RT, Drift Rate, and Non-decision Time for Single and Dual-site Stimulation Under Low and High Intensity.

For drift rate (Fig. 1b), there were no significant effects: F(1, 46) < 1 for trial type; F(1, 46) = 2.84, p =. 099, η_p² = 0.06 for intensity, and F(1, 46) = 2.99, p =. 091, η_p² = 0.06 for the interaction.

For non-decision time (Fig. 1c), results revealed a significant main effect of trial type, F(1, 46) = 34.07, p <. 001, η_p² = 0.43, with incongruent trials (M = 0.21 ± 0.03) yielding shorter non-decision time than neutral trials (M = 0.22 ± 0.08). Results also showed a significant main effect of intensity, F(1, 46) = 18.00, p <. 001, η_p² = 0.28, with high intensity (M = 0.20 ± 0.03) yielding shorter non-decision time compared with low intensity (M = 0.23 ± 0.03). The interaction was also significant, F(1, 46) = 50.36, p <. 001, η_p² = 0.52. Further analyses showed incongruent trials yielded shorter non-decision time than neutral trials only in the low intensity group, t(23) = 10.75, p <.001, but not in the high intensity group, t(23) < 1.

A two-sample t-test on the Flanker Effect in RTs revealed no significant difference between the low intensity group and the high intensity group.

For drift rate, a two-sample t-test comparing the Flanker Effect between the low intensity and high intensity group revealed no significant difference, t(46) = 1.73, p =.091, indicating that stimulation intensity did not significantly influence the difference in evidence accumulation between neutral and incongruent trials.

For non-decision time, the two-sample t-test showed that there was a significant difference between intensity groups, t(46) = 7.10, p <.001. Specifically, under low intensity stimulation, the flanker effect was negative.

Experiment 1b. Dual-site stimulation.

Results

ANOVA results of RTs (Fig. 1d) showed a significant main effect of trial type, F(1, 46) = 72.47, p <. 001, η_p² = 0.61, with slower responses in incongruent (M = 413 ± 55 ms) than neutral trials (M = 393 ± 50 ms). But no other significant effect was found, Fs(1, 46) < 1.

For drift rate (Fig. 1e), results revealed a significant main effect of trial type, F(1, 46) = 357.91, p <. 001, η_p² = 0.89, with incongruent trials (M = 7.27 ± 2.42) yielding lower drift rate than neutral trials (M = 10.22 ± 1.30). There was also a significant main effect of intensity, F(1, 46) = 96.87, p <. 001, η_p² = 0.46, indicating that high stimulation intensity (M = 7.74 ± 2.76) reduced drift rate compared with low intensity (M = 9.75 ± 1.51). The interaction was significant, F(1, 46) = 194.13, p <. 001, η_p² = 0.81, and further analyses showed that drift rate was significantly lower for incongruent conditions compared with neutral trials in both low, t(23) = 3.99, p <.001 and high stimulation intensity, t(23) = 21.03, p <.001, suggesting that increasing stimulation intensity resulted in a further reduced drift rate for incongruent trials compared with neutral trials.

For non-decision time (Fig. 1f), significant effect was found for trial type, F(1, 46) = 22.72, p <. 001, η_p² = 0.33, with incongruent trials (M = 0.22 ± 0.08) longer than neutral trials (M = 0.21 ± 0.03). There was also a significant main effect of intensity, F(1, 46) = 135.66, p <. 001, η_p² = 0.75, with high intensity (M = 0.26 ± 0.04) longer than low intensity (M = 0.17 ± 0.04). The interaction was significant, F(1, 46) = 465.74, p <. 001, η_p² = 0.91. Further analyses showed that incongruent trials had shorter non-decision times than neutral trials in the low intensity group, t(23) = 12.52, p <.001, but longer in the high intensity group, t(23) = 17.78, p <.001.

Discussion

In Experiment 1a, the RT analysis of single-site stimulation revealed a significant main effect of trial type, but no significant main effect of stimulation intensity or interaction. However, subsequent decomposition of RTs under the DDM framework revealed interesting results. The results of drift rate data showed that increasing stimulation intensity did not produce a significant interaction effect, suggesting that increasing intensity does not influence the evidence accumulation indexed by drift rate. Analysis for non-decision time revealed a significant interaction. Specifically, under low-intensity stimulation, incongruent trials exhibited shorter non-decision time compared to neutral trials, whereas this difference was not significant under high-intensity stimulation. This suggests low-intensity tDCS mitigates perceptual demands of incongruent conditions^43,51. Taken together, the present study found that for single-site tDCS, 1 mA is sufficient enough to enhance cognitive control, and that increase in the intensity does not produce additional benefits and may even be detrimental.

Similarly, in dual-site stimulation, RT analysis showed a significant main effect of trial type but no significant main effect or interaction. DDM analysis revealed a significant interaction for drift rate: low-intensity tDCS improved cognitive control, reducing drift rate differences, while high-intensity tDCS further reduced drift rate for incongruent trials. Non-decision time analysis showed low-intensity tDCS reduced time for incongruent trials, but high-intensity increased it, indicating added attentional demand. This suggests that low-intensity dual-site tDCS enhances cognitive control, while increasing intensity worsens it.

Whether these detrimental effects vary with changes in stimulation duration was examined in Experiment 2.

Experiment 2 Stimulation Duration.

Experiment 2a. Single-site stimulation.

Results

A two-way ANOVA for trial type and stimulation duration on RTs (Fig. 2a) showed a significant main effect of trial type, F(1, 46) = 41.99, p <. 001, η_p² = 0.48, with slower responses in incongruent trials (M = 426 ± 48 ms) than neutral trials (M = 408 ± 44 ms). No significant effects for duration or interaction were found, Fs(1, 46) < 1.

Fig. 2.

Mean RT, Drift Rate, and Non-decision Time for Single and Dual-site Stimulation During the First and Second Half of Stimulation.

For drift rate data (Fig. 2b), there was an insignificant effect of trial type, F(1, 46) < 1, a significant main effect of duration, F(1, 46) = 11.71, p =. 001, η_p² = 0.20, with longer duration (M = 8.16 ± 1.21) reducing drift rate compared to the first half (M = 9.33 ± 1.36). The interaction was also significant, F(1, 46) = 5.81, p =. 020, η_p² = 0.12. Further analyses showed that no difference was found between incongruent and neutral trials in the first half, t(23) = −1.06, p =.302, but incongruent trials had faster drift rate in the second half, t(23) = 2.37, p =.026.

For non-decision time (Fig. 2c), significant effects were found for trial type, F(1, 46) = 63.23, p <. 001, η_p² = 0.58, with incongruent trials (M = 0.20 ± 0.03) yielding shorter times than neutral trials (M = 0.23 ± 0.04). Duration had a significant effect, F(1, 46) = 18.24, p <. 001, η_p² = 0.28, with the second half (M = 0.23 ± 0.04) longer than the first half (M = 0.20 ± 0.03). There was also a significant interaction, F(1, 46) = 82.75, p <. 001, η_p² = 0.64, indicating that no difference in non-decision time was found in the first half, t(23) < 1, but incongruent trials had shorter times in the second half, t(23) = 12.41, p <.001.

Experiment 2b. Dual-site stimulation.

Results

A two-way ANOVA for trial type and stimulation duration on RTs (Fig. 2d) showed a significant main effect of trial type, F(1, 46) = 62.14, p <. 001, η_p² = 0.58, with slower responses in incongruent trials (M = 408 ± 31 ms) than neutral trials (M = 389 ± 25 ms). No significant main effect of stimulation intensity or interaction was found, Fs(1, 46) < 1.

For drift rate (Fig. 2e), significant main effects were found for trial type, F(1, 46) = 188.89, p <. 001, η_p² = 0.80, with lower drift rates in incongruent trials (M = 7.75 ± 2.79) than neutral trials (M = 9.92 ± 1.18), and for duration, F(1, 46) = 115.68, p <. 001, η_p² = 0.59, with longer duration tDCS increasing drift rate (M = 9.92 ± 1.18) compared to short duration (M = 7.74 ± 2.76). A significant interaction was found, F(1, 46) = 346.44, p <. 001, η_p² = 0.88. Drift rate was lower for incongruent conditions in the first half, t(23) = 21.03, p <.001, but higher in the second half, t(23) = 3.81, p <.001.

For non-decision time (Fig. 2f), significant main effects were found for trial type, F(1, 46) = 9.25, p <. 001, η_p² = 0.17, with longer non-decision time in incongruent trials (M = 0.22 ± 08) than neutral trials (M = 0.21 ± 03), and for duration, F(1, 46) = 110.50, p <. 001, η_p² = 0.71, with longer duration yielding shorter non-decision time (M = 0.20 ± 0.05) than short duration (M = 0.26 ± 0.04). A significant interaction was found, F(1, 46) = 854.39, p <. 001, η_p² = 0.95. Non-decision time was longer for incongruent trials in the first half, t(23) = 17.78, p <.001, but shorter in the second half, t(23) = 23.86, p <.001.

Discussion

In single-site tDCS, RT analysis revealed only a significant main effect of trial type, with no significant effects of stimulation duration or interactions. However, DDM analysis provided insights into how the negative impact on non-decision time with increasing stimulation intensity, as observed in Experiment 1a, varies with stimulation duration. Specifically, analysis for non-decision time revealed a significant interaction between stimulation duration and trial type. Initially, there was no difference between incongruent and neutral trials. However, incongruent trials induced shorter non-decision reaction times with longer stimulation, suggesting reduced attentional demand in the perceptual process. Additionally, analysis for drift rate further revealed that drift rate for incongruent trials was improved with increased duration, resulting in no significant interference effect.

In dual-site tDCS, similar patterns were obtained. RTs showed only a main effect of trial type, but DDM analysis revealed an interaction between stimulation duration and trial type. Initially, drift rates were lower for incongruent trials, but increased with stimulation duration. These results indicate longer stimulation duration improves individuals’ inhibitory control ability. Likewise, such improvement was also evident in non-decision time such that a significant reduction in non-decision time with longer stimulation duration for incongruent trials compared to neutral ones. These findings suggest that participants are able to overcome the additional attentional demand posed by the incongruent condition in the perceptual process.

In conjunction with the findings from Experiment 1, increasing stimulation intensity had negative effects in both single- and dual-site tDCS, but these effects diminished with longer stimulation duration.

General discussion

In the current study, we investigated the influence of stimulation intensity and duration on behavioural performance with both single-site and dual-site stimulation. By varying these parameters, our results revealed an important interaction between tDCS parameters, particularly with increasing duration. Using both traditional RT analysis and DDM analysis allowed us to quantify different cognitive processes in cognitive control, uncovering distinct patterns for single-site and dual-site tDCS with varying intensity and duration.

One of the most noteworthy findings of this study is that the detrimental effects caused by increasing tDCS intensity were found to disappear with increasing duration. Specifically, this reduction in detrimental effects was observed in the high intensity group (1.5 mA). Increased tDCS duration improved drift rates and reduced non-decision time for incongruent trials, which involve higher cognitive demands. This finding may reflect adaptive changes in the brain’s cognitive control functions. Specifically, in Experiment 1, we found that high-intensity tDCS resulted a decrease in behavioural performance, which may suggest overstimulation to specific brain circuits, disrupting normal cognitive processing and leading to a decrease in behavioural performance, which is consistent with existing literature that has shown limited additional benefits with increasing stimulation intensity⁵². Experiment 2 of the present study extended this investigation by exploring the temporal dynamics of this detrimental effect. Our initial hypothesis was that longer duration would lead to larger effects. However, the results showed that, as tDCS continued, such detrimental effects disappeared, which indicates that the brain may engage in compensatory mechanisms, such as synaptic plasticity, eventually allowing it to readjust to reduce disruption. To our knowledge, the present study is the first to explore the changes brought about by increasing stimulation intensity from a temporal perspective. While there is a mismatch between our initial hypothesis and the observed results, this likely reflects the novelty of applying DDM in the current study. Since no prior studies have examined tDCS duration using drift diffusion model (DDM), our predictions were primarily based on findings from reaction time studies, which do not capture the latent cognitive components that DDM could reveal. Because of this, we acknowledge that some aspects of our findings are exploratory and can open up new directions for future research to further explore how the brain responds to tDCS over time.

While previous research (e.g., Jamil et al., 2017) have documented the absence of significant changes in cortical excitability regardless of tDCS intensity (i.e., 0–2 mA) when applying tDCS to motor cortex, the current study focused on brain regions involved in cognitive control using a cognitive control task, which may contribute to this discrepancy. It is possible that intensity-dependent effects observed in cognitive domains differ from those documented in motor areas, which could explain why we found notable differences between 1.0 mA and 1.5 mA in DDM parameters. Furthermore, in additional to traditional reaction time analysis, we applied the DDM to decompose reaction time and accuracy data, which enabled us to capture subtle changes in the underlying cognitive processes such as the evidence accumulation process (i.e., drift rate) and the motor preparation process (i.e., non-decision time) that RT analyses might not reveal.

Additionally, although previous studies have compared single-site and dual-site tDCS, these studies have mainly used cortical excitability as the outcome variable^{31,51,53–55}, without further exploring the impact of these two stimulation paradigms on cognitive functions. Findings of the current results uncovered that single-site tDCS primarily affected perceptual processes (non-decision time), while dual-site tDCS affected both perceptual (non-decision time) and evidence accumulation processes (drift rate). This may suggest that dual-site tDCS could affect a broader range of brain circuits, potentially enhancing communication between brain regions. These results, combined with previous research³⁰, suggest that complex cognitive functionscould result from interactions between brain regions rather than an isolated brain region.

Finally, we found that, compared with conventional reaction time and accuracy analyses, drift diffusion model could provide new insights for revealing the effect of anodal tDCS. By translating behavioural accuracy, reaction times, and reaction time distributions into latent cognitive components²⁰, DDM offers a more comprehensive account for decision-making process. Specifically, in the RT analysis, single-site and dual-site tDCS exhibited similar patterns of effects; however, through the decomposition by DDM, distinct patterns for single-site and dual-site tDCS were uncovered. The discrepancy between RTs results and DDM results is probably due to the fact that RTs reflect the sum of multiple cognitive sub-processes, including evidence accumulation, perpetual encoding, and motor execution. If tDCS only influences one of these components, or influences multiple sub-processes in different directions, the overall effect on RTs may not become observable, while DDM is able to decompose these processes, making it more sensitive to subtle changes²¹. This study is the first to use DDM to explore the effects of increasing tDCS dosage on cognitive control and the complex differences between single-site and dual-site stimulation, highlighting the importance of using DDM to capture the subtle differences in the effects of tDCS on cognitive processes missed by RT analysis.

However, it is important to note that our study did not include a sham group due to methodological and practical constraints. Given the primary goal of this study is to compare multiple tDCS parameters in both single-site and dual-site tDCS simultaneously, adding perfectly-matched sham groups would have significantly increased the complexity of comparisons and potentially reduced statistical power. Thus, while not having a sham group limits our ability to draw definitive conclusions regarding the absolute tDCS efficacy compared to baseline, the primary goal was focused on comparing differences in stimulation effects between intensities, durations and montage placements rather than establishing causal effects of stimulation versus no stimulation. Future studies aiming to further extend our findings by exploring the absolute effects of tDCS along with the relative effects of varying tDCS parameters would benefit from adding sham groups in their experimental design.

Conclusions and implications

In conclusion, for the first time, we comprehensively investigated the effects of different tDCS intensity and duration for both single-site and dual-site tDCS within the same study by combining both the traditional RT measures and DDM framework. Our findings suggest that stimulation duration may modulate the detrimental effects caused by higher tDCS intensity, indicating a potential interaction between tDCS parameters. However, given the exploratory nature of the current study, this interaction should be interpreted with caution and warrants further investigation. Different findings regarding tDCS intensity and duration underscore the need for a nuanced understanding of the underlying mechanism of tDCS on modulating cognitive processes, suggesting distinct patterns for single-site and dual-site tDCS. According to these findings, future studies may further explore individual differences and task-specific factors to optimize the application of tDCS for cognitive enhancement and therapeutic interventions.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CRediT authorship contribution statement X. L.: Conceptualization, Data curation, Investigation, Methodology, Writing – original draft, Writing – review and editing. T. G.: Investigation, Methodology, Writing – review and editing. Q. C.: Data curation, Investigation. J. C.: Data curation, Investigation. Q. D.: Data curation, Investigation. P. L.: Data curation, Investigation. T. G.: Conceptualization, Formal analysis, Funding acquisition, Methodology, Investigation, Writing – review and editing.

Data availability

The data supporting the findings of this study will be made available by the authors upon request. Please contact Taomei Guo at guotm@bnu.edu.cn to request the data.

Declarations

Competing interests

The authors declare no competing interests.