Optimizing Transcranial Direct Current Stimulation Protocols to Promote Long-Term Learning

  • Jacky Au
  • Carley Karsten
  • Martin Buschkuehl
  • Susanne M. Jaeggi

DOI: 10.1007/s41465-017-0007-6

Cite this article as:
Au, J., Karsten, C., Buschkuehl, M. et al. J Cogn Enhanc (2017). doi:10.1007/s41465-017-0007-6


Transcranial direct current stimulation (tDCS) is a form of non-invasive brain stimulation that has the potential to induce polarity-specific changes in neural activity within targeted brain regions. There is growing interest in the use of this technology for the enhancement of higher cognitive functions, and application of tDCS directly before or concomitant with task performance has shown promise in modulating a range of behavioral outcomes, including motor skill acquisition, working memory performance, and implicit and explicit learning. The proposed mechanism for the observed enhancements is a temporary and targeted shift in the excitability of the cortical regions that subserve the relevant tasks, lasting from minutes up to about an hour after cessation of stimulation. Although empirical work does support at least a partial role for this mechanism, an arguably more potent but relatively underexplored phenomenon is thought to occur in the hours or days after stimulation—that is, a facilitation of consolidative processes. Here, we review the literature describing the nature of tDCS-enhanced consolidation and argue that some of the mixed results among the single-session studies that currently dominate the extant literature may be explained by a failure to take advantage of these potentially powerful offline effects. Accordingly, we further contend that the full potential of tDCS cannot be truly realized without a longitudinal design which allows for tDCS to act directly upon learning by promoting consolidation between sessions. Finally, we review preliminary evidence that these consolidation effects can be even further enhanced via strategically spaced out stimulation sessions, which take advantage of a long-held tenet in the literature that distributed learning produces better outcomes than massed learning. We conclude by proposing potential study designs to encourage the use of tDCS as more than merely a method to promote temporary enhancement, but also a technique to enhance long-term learning.


Consolidation Plasticity Meta-plasticity Distributed learning Spacing 


Transcranial direct current stimulation (tDCS) is a non-invasive form of brain stimulation that sends weak direct currents through the scalp and into the underlying cortex. tDCS is thought to alter the resting membrane potential of target neurons in a polarity-dependent manner such that the anode increases while the cathode decreases cortical excitability (e.g., Nitsche et al. 2003). Behaviorally, this has been shown to manifest in increased motor-evoked potentials (MEP; Horvath et al. 2015a), improved motor functioning (Hashemirad et al. 2016), enhanced working memory (WM) performance (Mancuso et al. 2016), as well as a plethora of other cognitive, physical, and emotional changes (e.g., Utz et al. 2010).

However, recent work has raised questions about the mechanisms and effects of tDCS. For example, since most of the current delivered at the scalp gets shunted away by the skin, skull, and cerebrospinal fluid before entering the brain, the electric field generated in the brain is orders of magnitude below that typically delivered by other methods such as TMS (Ruohonen and Karhu 2012). It has consequently been argued that the effects of tDCS are too weak to have any meaningful impact on membrane potential (Underwood 2016), furthering existing controversy surrounding the reliability of tDCS (Antal et al. 2015; Chhatbar and Feng 2015; Horvath et al. 2015b; Price and Hamilton 2015). However, this argument has not been empirically evaluated since direct evidence for membrane polarization is based on intracellular recordings in animal models where current is applied directly to cortical slices rather than transcranially (e.g., Purpura and McMurtry 1965). Nevertheless, most meta-analyses converge on small overall effects within healthy young adults, which become larger and more robust when studies with lower-performing populations such as clinical patients and the elderly are included (Dedoncker et al. 2016; Hill et al. 2016; Hsu et al. 2015; Mancuso et al. 2016). Therefore, it is imperative to develop a stronger theoretical understanding to account for the diverse effects and to optimize protocols for greater reliability and more meaningful results.

An interesting epiphenomenon within the tDCS literature is that occasionally effects are observed some time after (hours to months), but not immediately during, stimulation. This suggests that changes in membrane potential as a result of tDCS are not exclusively driving behavioral improvements since the polarizing effects of tDCS should have largely washed out by the time these delayed effects emerge. Rather, what these studies suggest is that tDCS may play a role in consolidation. Although the effects of tDCS on neural membrane potential have been argued to be small and inconsequential, its effects on glial cells, which comprise ∼50% of brain cells and are sensitive to much smaller depolarizations than neurons, are estimated to fall within a biologically meaningful range (Ruohonen and Karhu 2012). Moreover, glial cells, which secrete many of the same molecules and transmitters as neurons such as calcium and glutamate, play a direct role in learning and synaptic plasticity, thereby providing a tenable mechanism for tDCS to act upon consolidation (see Ben Achour and Pascual 2010; Gibbs et al. 2008; Monai et al. 2016). Harnessing these latent effects could help mitigate the low reliability found among some studies, particularly the single-session studies that dominate the extant literature, which are not designed to capture delayed consolidation effects. Here, we overview the neurobiological foundation of tDCS’ putative effect on consolidation, and then review the experimental evidence in support of this phenomenon. Finally, we discuss the optimization of protocols to capitalize on these effects.

Neurobiology of Consolidation and tDCS

Consolidation refers to a process by which learning becomes increasingly resistant to interference over time and operates on two different time scales: a fast-acting synaptic consolidation on the order of minutes to hours that strengthens local synaptic transmission, and a slow-acting system consolidation on the order of days to weeks or years that anchors the memory trace into a long-term store across distributed brain regions (Born and Wilhelm 2012). Support for this model comes primarily from studies of declarative memory, but other memory domains such as those involved in skill learning are also thought to operate similarly (reviewed in Dudai et al. 2015). The administration of tDCS may enhance consolidation at both of these stages.

Synaptic consolidation involves gene expression changes and protein synthesis that result in higher levels of plasticity-related proteins at recently active synapses (Frey and Morris 1997; Steward et al. 1998). The after-effects of tDCS are known to be dependent on this ongoing protein synthesis that occurs during the stimulation period, as well as on increased functioning of NMDA receptors which facilitate synaptic transmission. Administration of both protein inhibitors in animal models and NMDA receptor antagonists in humans is able to nullify these after-effects (Gartside 1968; Nitsche et al. 2003). This suggests that online tDCS interacts with the brain’s normal plastic response, which allows for the continued manifestation of effects offline. In fact, tDCS has been demonstrated to upregulate the expression of BDNF, an important protein involved in long-term potentiation (LTP) and memory (Podda et al. 2016), and alter the balance of GABAergic and glutamatergic transmission, which can modulate cortical excitability and facilitate or obstruct activity-dependent LTP (Krause et al. 2013). See Stagg and Nitsche (2011) for a more thorough account of the cascade of cellular and molecular modifications that arise from tDCS.

Furthermore, in the minutes to hours following a learning event, cells which were engaged by the event undergo patterned reactivation (Diba and Buzsáki 2007; Foster and Wilson 2006). Such “replay” has been observed both during slow-wave sleep (Wilson and Mcnaughton 1994) and awake rest (Foster and Wilson 2006; Karlsson and Frank 2009) and is thought to play a prominent role in system consolidation (Sirota and Buzsaki 2005). In fact, disruption of neural replay has been demonstrated to prevent consolidation (Genzel and Robertson 2015). Therefore, it is conceivable that tDCS, with its putative effects on cortical excitability (which may operate via glial-neuron interactions rather than neural stimulation directly), could enhance neural replay and therefore enhance consolidation. Though a direct demonstration is lacking, several studies have applied tDCS during waking rest at time periods when replay is thought to occur (Javadi and Cheng 2013; Sandrini et al. 2014; Tecchio et al. 2010) and subsequently demonstrated greater consolidation. Similarly, the application of slow-oscillating tDCS during slow-wave sleep has been shown to enhance declarative memory (reviewed in Barham et al. 2016). These enhancements are also accompanied by an increase in slow-oscillating waves (<1Hz), a neural frequency band which is temporally associated with the onset of neural replay (Genzel and Robertson 2015).

Experimental Evidence

A number of animal studies have measured polarity-dependent cellular changes in LTP both in vitro (Ranieri et al. 2012; Ruohonen and Karhu 2012) and in vivo (Podda et al. 2016; Rohan et al. 2015) as a function of direct current stimulation. Analogously, human experiments have demonstrated on a behavioral level that tDCS over multiple days can lead to cumulative effects between sessions, suggesting the existence of continued offline processes between each bout of stimulation. These offline between-session effects have even been observed in some studies to be greater than the online within-session effects. For example, extending the classic MEP paradigm (an assessment of motor excitability) over 5 days led to step-wise increases in baseline amplitude each day (Alonzo et al. 2012; Galvez et al. 2013). However, despite the impressive between-session effects in these studies, the within-session responsivity to tDCS was comparable across most days and in fact did not even always differ significantly from baseline in every session (see also Horvath et al. 2016).

Similarly, pairing tDCS with a behavioral task has also been shown to improve performance offline, leading to performance increases between consecutive sessions. For example, Reis et al. (2015, 2009) demonstrated greater skill learning using a visuomotor task with concurrent tDCS over motor cortex. Importantly, the respective contributions of online and offline effects were systematically evaluated by starting stimulation only after the first block of training each day. This allowed comparing the offline improvement between this first block and the last block of the previous day, with the online performance gains within a session. Their analysis suggested that the majority of learning occurred offline1 (cf. Prichard et al. 2014). Furthermore, our previous work in the cognitive domain (Au et al. 2016) showed similar effects using a WM intervention combined with online tDCS. As with motor skill, we also observed a higher rate of learning in our stimulated group, relative to sham. Although we did not systematically evaluate the relative contributions of online and offline learning, several lines of evidence suggest a substantial role of offline consolidation in our study. First, there was no hint of a between-group difference after the first training session, suggesting a minimal or non-existent role of online stimulation, at least for the first session. However, differences became increasingly pronounced over the course of training, after offline consolidation had a chance to occur. Furthermore, we demonstrated that the strongest between-session effects occurred after a weekend break, in accordance with predictions from the learning-consolidation literature (Ebbinghaus 1885), and also demonstrated long-term maintenance of training effects up to a year later (Katz et al. in press), suggesting that increased consolidation led to stronger long-term retention.

Altogether, we note that although limitations can sometimes exist in the immediate manifestation of online effects at the behavioral level, a complementary mechanism occurs offline that can further consolidate learning gains beyond the stimulation period. In further support of this notion, delayed effects have been demonstrated even in the absence of immediate effects. For example, there have been rapid consolidation effects where group differences only begin to emerge in the minutes or hours after stimulation, or become stronger/more robust with time (e.g., Clark et al. 2012; Ehsani et al. 2016; Hoy et al. 2014; Hsu et al. 2015; Javadi and Cheng 2013; Penolazzi et al. 2013; Reis et al. 2015). Similarly, overnight consolidation has been enhanced when performance is measured the next day, despite a lack of group differences on day 1 (Koyama et al. 2015; Martin et al. 2014), and even cognitive training studies that failed to show immediate tDCS-related enhancements have still demonstrated greater tDCS-related retention a couple months later (Jones et al. 2015; Martin et al. 2013; Stephens and Berryhill 2016).

Optimizing tDCS Protocols

With the many degrees of freedom that tDCS protocols offer, a pertinent goal is to hone in on the parameters that can optimize the benefits derived from electrical stimulation. Based on the evidence discussed herein, we make recommendations for future research to explore with respect to three important parameters: (1) timing of stimulation, (2) number of sessions, and (3) spacing of sessions.

Timing of Stimulation

One issue that should be given careful consideration in future studies is the timing of stimulation—that is, whether it should be done concurrent with a task or offline. Online stimulation seems to be theoretically preferable to offline stimulation in that it potentiates task-relevant networks rather than resting state networks, a distinction which may prove crucial for later consolidation of task performance. In line with this hypothesis, online and offline stimulation were directly contrasted with a WM task, and while both methods led to similar immediate performance, only the online condition promoted greater performance the next day (Martin et al. 2014). Furthermore, as discussed earlier, many studies using online stimulation have successfully demonstrated greater learning consolidation after a delay, but no study to our knowledge has done the same with offline stimulation prior to task performance. The potential delayed benefits of online, rather than offline, stimulation then is an important consideration given that meta-analyses suggest both forms of stimulation provide comparable immediate benefits2 (Dedoncker et al. 2016; Hill et al. 2016).

Although offline stimulation in the traditional sense (i.e., before task performance) may arguably not interact with consolidative processes, offline stimulation after task performance may present a more viable route as it can coincide directly with consolidation or reconsolidation periods. For example, stimulation immediately after, but not during, training of a finger-tapping task enhanced subsequent performance 30 min later, presumably by facilitating early consolidation of procedural memory (Tecchio et al. 2010). Similarly, reactivation of a previously learned word list hours or a day later led to better retention if stimulation was administered directly during this reactivation/reconsolidation period (Javadi and Cheng 2013; Sandrini et al. 2014). Furthermore, stimulation during sleep seems especially fruitful for declarative memory enhancement if timed during the appropriate consolidation period during slow-wave sleep (Barham et al. 2016). Therefore, it is possible that while online stimulation may promote LTP-related protein synthesis at the synapse (Gartside 1968), offline stimulation after task completion or during sleep may directly enhance learning-associated neural replay and system consolidation for long-term retention. Future research should systematically evaluate the mechanisms and relative efficacy of online and offline (after task) stimulation.

Number of Sessions

We have made the argument that the greater potential of tDCS may lie in its role in augmenting consolidation, rather than the online enhancement of membrane excitability which in of itself may arguably produce smaller or less reliable effects. Accordingly, we recommend future studies contain multiple sessions in order to capitalize on these delayed effects. These additional sessions may be stimulated or unstimulated, as either approach should allow consolidation effects from the previous session to manifest. However, we note that additional stimulation sessions, as commonly done in cognitive or motor training studies, may be particularly effective as these sessions offer the opportunity to simultaneously boost reactivation and neural replay from previous learning (e.g., see Javadi and Cheng 2013; Sandrini et al. 2014), as well as potentiate learning from the current session.

Spacing of Stimulation

If studies are to incorporate multiple sessions, a relevant question arises concerning the optimal spacing of these sessions in order to maximize consolidation and performance on subsequent sessions. Many intervention studies employ once-daily schedules, mostly out of convenience and convention rather than any empirically derived model. However, there is no reason to assume that this is the most optimal schedule. For example, several lines of evidence indicate that even shorter spacing protocols might be beneficial (reviewed in Goldsworthy et al. 2015). Monte-Silva et al. (2010) demonstrated that two bouts of cathodal stimulation separated by 20 min resulted in greater and more prolonged MEP depression than a continuous bout of cathodal stimulation for the same total duration. A shorter break of 3 min produced similar, but more muted effects, while longer breaks of 3 or 24 h were detrimental. Similar effects were replicated with anodal stimulation, in which breaks of 3 or 20 min were effective in prolonging and enhancing MEP amplitudes even into the next day, while breaks of 3 or 24 h showed similar suppression of tDCS effects (Monte-Silva et al. 2013). The short-term spacing of stimulation is thought to induce meta-plastic processes in the cortex, so-called because the first round of stimulation alters the plastic response from the second. Such an approach is also supported by animal models. Repeated trains of high-frequency electrical stimulation, usually spaced apart by intervals of less than an hour, are capable of inducing late-phase LTP that lasts for periods of weeks or longer, whereas a single train of stimulation is capable only of inducing early-phase LTP lasting hours or less (Goldsworthy et al. 2015). What appears to be critical to induce meta-plasticity with tDCS is to initiate the second round of stimulation during the after-effects of the first, a window of time which typically lasts up to an hour after stimulation (Nitsche et al. 2003). Behaviorally, meta-plastic protocols have also accentuated or prolonged effects on both motor learning (Bastani and Jaberzadeh 2014; Christova et al. 2015) and WM (Carvalho et al. 2015). In all these cases, it appears that inter-session breaks of approximately 20–30 min produce stronger effects than much shorter (3–5 min) or much longer (3–24 h) breaks, suggesting the existence of a non-linear optimum.

Despite the evidence in support of short, within-day spacing intervals, a number of successful intervention studies have been carried out using a once-daily approach (e.g., Alonzo et al. 2012; Au et al. 2016; Reis et al. 2015; Reis et al. 2009; Stephens and Berryhill 2016). Moreover, our own work demonstrated with a WM intervention that participants showed the greatest improvement after a weekend break (Au et al. 2016), suggesting in addition to the meta-plasticity protocols that longer spacing intervals of several days can actually be beneficial for consolidation as well. How do we reconcile these disparate results? One thing to consider is that any consolidation effect that arises is naturally an interaction between the effects of electrical stimulation on brain tissue and task-related neural activation, both of which individually are likely to involve different spacing parameters that optimize LTP-induction. Although current evidence suggests that about half an hour may be an optimum spacing interval for tDCS to synergistically engage meta-plastic processes, consolidation of task-related activity (independent of tDCS) may require longer intervals. Moreover, this consolidation is also likely to be task specific. For example, one meta-analysis summarizing the cumulative spacing research over the past century suggests that tasks with higher mental but lower physical complexity require longer spacing intervals than tasks with higher physical but lower mental complexity (Donovan and Radosevich 1999). This should be given careful consideration when piloting and designing intervention studies, as this suggests that relying on the parameters of MEP experiments to guide research, as is commonly done, may underestimate the optimum spacing interval required for more complex motor learning tasks, and even more so for cognitive skill learning. For instance, cognitive training studies (without tDCS) have systematically evaluated this phenomenon, finding that once-daily training promoted greater learning and transfer than shorter spacing schedules that involved multiple training sessions per day (Arthur et al. 2010; Wang et al. 2014). Moreover, it has been demonstrated that the optimal spacing interval increases in correspondence with longer retention intervals. For example, Cepeda et al. (2008) systematically varied the retention interval of a declarative memory task as well as the time gap between two study sessions. They found that a study gap of 1 day produced optimal retention when test sessions were separated by 1 week, whereas a gap of 3 weeks was optimal for retention 1 year later. Therefore, for the purposes of using cognitive or motor training to build lasting skills, it may be beneficial to space sessions out across days or longer.

Although the relative efficacy of short-term spacing on the order of minutes or longer-term spacing on the order of days should be systematically evaluated in future research, one enticing avenue we propose is to combine the benefits of meta-plasticity with longer-term spacing of learning. This can be accomplished by spacing sessions out over a couple days, but using a repeated, meta-plastic protocol within each session. This may optimize both the delivery of electrical stimulation as well as the timing of task learning for long-term retention.

Conclusion and Future Directions

We have reviewed the role of tDCS in promoting consolidation of learning and argued that in some cases, this mechanism may present a stronger and more reliable source of the tDCS effect than the commonly touted online effects on neural excitability. This is not to suggest that such online effects are weak or non-existent. On the contrary, many studies have documented worthwhile immediate enhancements both on the level of behavior and functional brain connectivity (e.g., Fregni et al. 2005; Keeser et al. 2011; Lindenberg et al. 2013; Meinzer et al. 2013; Polania et al. 2011; Price and Hamilton 2015 and others). Rather we point out that these online effects can have impressive downstream consequences that should be explored in order to gain a more comprehensive understanding of tDCS-induced cognitive enhancement.

Accordingly, we urge future research to include multiple sessions in order to capitalize on these consolidation effects, and to systematically evaluate the parameters that lead to these effects. For example, stimulation during task as well as directly during consolidation or reconsolidation periods afterwards both appear beneficial, but their relative efficacy is unknown. Also, the optimal spacing schedule of stimulation is unknown. Short spacing intervals on the order of minutes appear to effectively engage meta-plastic processes, but there is also evidence to suggest that longer spacing intervals of greater than a day might also be beneficial, especially for more complex cognitive tasks. As a final note, we caution that the evidence we have presented herein stems largely from healthy young adult populations and a few animal studies. The extent to which these lessons can be extrapolated to clinical studies and vulnerable populations is unknown and should be approached with care (e.g., Perceval et al. 2016).


In both studies, the first session was the only one that showed within-session effects, suggesting that online effects may saturate early on in an intervention, or may not be reliable. Additionally, we note that a similar paradigm (Prichard et al. 2014) with a different motor task showed predominantly online, but not offline, effects, suggesting some task-specificity in the degree of consolidation.


Hill et al. (2016) actually found a significant tDCS effect only with offline stimulation, but not online, in healthy young adults. However, the effect sizes are comparable and not different from each other. Also, sample sizes are trhree to six times greater in the offline studies compared to online, thus biasing interpretations based on significance alone. We interpret the data to suggest no difference between online and offline stimulation.


Compliance with Ethical Standards


This work was supported by the National Science Foundation Graduate Research Fellowship Grant No. DGE-1321846 to J.A. J.A. and M.B. are employed at the MIND Research Institute, whose interest is related to this work and S.M.J. has an indirect financial interest in MIND Research Institute. No other conflicts of interests or sources of funding are declared.

Copyright information

© Springer International Publishing 2017

Authors and Affiliations

  1. 1.Department of Cognitive SciencesUniversity of California, IrvineIrvineUSA
  2. 2.MIND Research InstituteIrvineUSA
  3. 3.Department of Anatomy and NeurobiologyUniversity of California, IrvineIrvineUSA
  4. 4.School of EducationUniversity of California, IrvineIrvineUSA