The most frequently used placement of electrodes in studies on cerebellar tDCS is a lateralized position with the active electrode placed on the skin over one cerebellar hemisphere at 1–2 cm below and 3–4 cm lateral to the inion (Ferrucci et al., 2015) and the reference electrode over (a) the ipsilateral buccinator muscle, (b) the ipsilateral deltoid muscle, or (c) the ipsilateral forehead/supraorbital area (see Figure 1). Although most unilateral cerebellar stimulation setups made use of an ipsilateral reference electrode, some placed the reference electrode contralaterally. The possible differences in terms of effects between both setups have not been investigated yet. Bilateral stimulation of the cerebellum is also possible, but for bilateral stimulation the setups differ substantially (see Table 1 for the electrode placements in each study). It is assumed that a bilateral stimulation of the cerebellum would impact on both cerebellar hemispheres with a more diffuse effect upon the cerebral cortex. Stimulation of the vermis requires a placement of the active electrode in the midline.

Since electrodes' positions determine the direction of the current flow and the orientation of the electric field (Ferrucci et al., 2015), modeling studies have attempted to estimate the current density field distributions and electric fields induced in the nervous tissue by tDCS using computational methods to solve the Laplace equation (Priori et al., 2014). However, only a few modeling studies have specifically investigated the current flow in cerebellar tDCS (Parazzini et al., 2014b; Priori et al., 2014; Rahman et al., 2014; Rampersad et al., 2014). Parazzini et al. (2014b) modeled cerebellar tDCS using three virtual human head models of different ages and genders constructed of 77 different tissue types, segmented into a (hexahedral) voxel-based format (1 mm voxels). In this study a bilateral setup was used with the active electrode centered on the median line, 2 cm below the inion, and the reference electrode over the right arm (5 × 7 cm). A current intensity of 2 mA was used. The authors found that the highest electric field and current density was located below the stimulating electrode in the posterior cerebellum. Only a slight spread to other structures (e.g., occipital cortex) was found, unlikely sufficient to produce relevant functional effects. Parazzini et al. (2014b) also showed that no alteration of brainstem excitability occurred (this is particularly important given the numerous connections between brainstem nuclei and the cerebellum) and that there was only a very low current spread to the heart. There are, however, some slight differences between the models. In particular, differences in cerebrospinal fluid (CSF) distribution and/or skull thickness may influence the spread of the field amplitude toward the occipital region. Although Parazzini et al. (2014b) did not find a significant spread of the current to the brainstem in the child model, the use of cerebellar tDCS in children is still discouraged due to a possible spread of the current to this area (Priori et al., 2014; Ferrucci et al., 2015). Since the direction of the field within the cerebellum was not addressed in this study, Rahman et al. (2014) investigated this issue using four different electrode montages, varying the direction of the current flow (inward, outward, lateralized left, lateralized right) while only stimulating the cerebellar area. They found that in the four simulated setups, a current flow was induced and was largely uniform in direction, confirming the findings of Parazzini et al. (2014b) that the cerebellum can indeed be stimulated with cerebellar tDCS (Rahman et al., 2014). However, both studies focused on bilateral cerebellar stimulation, a setup that is not commonly used in experimental studies.

Rampersad et al. (2014) investigated six of the most frequently used setups in clinical and experimental cognitive research with finite element models. They used an MRI- and DTI-based model of a healthy 25-year-old man with 11 different tissue types, reconstructed as a mesh of tetrahedral elements. To simulate cerebellar stimulation they placed a square anode of 5 × 5 cm, 3 cm rightwards of the inion and a square cathode of 5 × 5 cm on the right buccinator muscle (cheek). Simulations were made for 1 mA tDCS. Results showed that during cerebellar stimulation, the actual maximum of the electric field is more inferior and medial to the targeted area due to the highly concave shape of the area. However, the high electric field also covered most of the inferior surface of the right cerebellar hemisphere, which makes it the most efficient setup of this modeling study. The study also showed that the maximum electric field strength values are much lower in the cerebellar setup. This is probably due to large amounts of shunting under the skull and through the skin. In all configurations only a small amount of the current enters the brain, but this was especially true for the cerebellar setup since the cerebellar electrode is placed on the back of the head. Most of the remaining current enters the gray matter perpendicularly. This might be more important than the mean or the maximum electric field strength (Rampersad et al., 2014). Overall, this study has validated most experimental setups applied in experiments with cerebellar tDCS.

Although modeling studies provide insights in the understanding of cerebellar tDCS, the results should be interpreted with much caution since little is known about tissue conductivity (Priori et al., 2014). Especially the values of muscle conductivity vary substantially in the literature (Rampersad et al., 2014). In their study, Rampersad et al. (2014) compared the results obtained with the largest values (as reported for the neck muscles) with the lowest reported values and found an increase of 11% in the mean field strength in the target volume. Efforts should be devoted to improve our knowledge about tissue conductivity. This would increase significantly the accuracy of the modeling studies (Rampersad et al., 2014).

It is still unclear whether the position of the reference electrode is critical or not. Grimaldi and Manto (2013), for instance, used a setup with the reference electrode on the contralateral supraorbital area. To exclude the possibility that the results were due to an inhibition of the prefrontal area, they repeated the experiment with the reference electrode on the ipsilateral shoulder. The results remained unchanged. In addition, the model of Parazzini et al. (2014b) showed that varying the position of the active electrode with ±1 cm only induced a small change in the field amplitude distributions, suggesting that the use of advanced neuronavigation systems is probably not needed to reliably perform cerebellar tDCS. The clinical evidence of studies using cerebellar tDCS in different setups seems to corroborate this view. However, more modeling and clinical studies are needed to systematically investigate the impact of electrode placement on the effects induced by cerebellar tDCS (Priori et al., 2014; Ferrucci et al., 2015).

There are two types of tDCS that can be used, depending on the direction of the current: anodal and cathodal. Anodal stimulation is frequently associated with enhanced neuronal excitability below the site of stimulation, whereas cathodal stimulation is thought to inhibit neuronal excitability (Rahman et al., 2013). However, this seems to be a simplification of the mechanisms of action. To understand which type of stimulation should be used, it is important to keep in mind the impact of tDCS on neurons.

Neurons, when inactive, remain at their resting electric potential due to the concentration gradient between the intra- and the extra-cellular medium. This electric potential can be estimated using the Goldman-Hodgkin-Katz voltage equation (Hodgkin and Katz, 1949). When tDCS is applied, a difference of electric potential is created between the stimulator's anode and cathode to allow a constant current flow between them. This difference of potential results in an electric field equal to the opposite of the potential gradient (according to Maxwell's law). The electric field induces a shift in the membrane electric potential of the neuron. This potential change can influence neuronal activity but is not strong enough to induce action potentials inside the neurons. As a result, tDCS can only modulate excitability in active neurons and has little or no impact on resting neuronal populations (Woods et al., 2016).

A positive (cathodal) extra-cellular field hyperpolarizes the membrane and lowers the action potential firing rate (i.e., lower excitability), whereas a negative (anodal) extra-cellular field depolarizes the membrane and increases the action potential firing rate (i.e., hyperexcitability; McIntyre and Grill, 1999; Liebetanz et al., 2002; Bikson et al., 2004). However, these physiological mechanisms are not always operational and depend on the orientation of neuronal structures. Whether an electric field has an excitatory or inhibitory effect depends on the axonal orientation relative to the field (parallel vs. perpendicular, current flow from soma to dendrites vs. from dendrites to soma; Kabakov et al., 2012; Rahman et al., 2014). In addition, it remains unclear which compartments (soma, dendrites, axons) are involved in modulation through electrical stimulation and whether depolarization or hyperpolarization is responsible for enhancing synaptic efficacy (Rahman et al., 2013).

tDCS also causes polarity-dependent physiological changes in the neurons that can last for a few hours after the end of a stimulation session (so called after-effects), depending on the intensity and duration of stimulation (Manto et al., 2011). One of the mechanisms that might be responsible for these long-lasting after-effects is a change in the ionic gradient (due to a change in membrane potential) on the extra-cellular side (Ardolino et al., 2005; Priori et al., 2014), or at the synaptic level by N-methyl-D-aspartate (NMDA) receptors (Liebetanz et al., 2002; Ardolino et al., 2005). It is suspected that anodal tDCS may change the intra-cellular Ca²⁺ level, leading to an NMDA receptor-mediated augmentation of synaptic strength, while cathodal membrane hyperpolarization may lead to a depression of synaptic strength (Woods et al., 2016). In addition, it has been shown that an externally applied electric field causes redistribution of membrane proteins and migration of the acetylcholine receptors (Ardolino et al., 2005). Both the protein redistribution and the channel migration may affect the propagation of neuronal activity and change neuronal plasticity (Debanne et al., 2003). tDCS has also been reported to change the acid-base balance due to water electrolysis from constant current. This mechanism may affect membrane, receptor, and cell function (Ardolino et al., 2005). If recommended stimulation duration and intensity are respected, these changes are only temporary and no stable functional or structural cortical modifications have been observed after tDCS (Nitsche et al., 2003b). However, since the physiological effects of electrical stimulation are studied using computer models based on the Hodgkin-Huxley and cable models (McIntyre and Grill, 1999; McIntyre, 2004; Manola et al., 2005; Molaee-Ardekani et al., 2013; Rahman et al., 2013; Dougherty et al., 2014; Parazzini et al., 2014a), novel studies on animals and/or humans are needed to better understand the complex effects of tDCS on brain function.

As expected, the effects of tDCS critically depend on (a) the previous neuronal physiological state and (b) the structure orientation relative to the electric field direction (Bikson et al., 2004; Manola et al., 2005; Woods et al., 2016). Neurons of the cerebellum are not identically orientated and even follow complex anatomical distributions over the numerous folia. This will cause a hyperpolarization in some compartments while others will be depolarized at the same time (Figure 2). Therefore, the global effects of tDCS on the cerebellum remain difficult to simulate (Woods et al., 2016). The linking function of parallel fibers in the cerebellar cortex and the peculiar disposition of the 10 lobules of the cerebellum surrounded by CSF and vessels render the simulation even more difficult.

There is currently a lack of information about the specific effects of tDCS on the various cerebellar neurons (Purkinje neurons, inhibitory interneurons of the cerebellar cortex, granule cells, nuclear neurons) and on the afferent pathways (mossy fibers and climbing fibers). Several studies using cerebellar tDCS have compared both anodal and cathodal stimulation with a sham condition. Varying results were obtained in these studies. Most of the studies reported a different effect for anodal and cathodal tDCS. Chen et al. (2014), Galea et al. (2009), Herzfeld et al. (2014), Jayaram et al. (2012), Yavari et al. (2016), and Zuchowski et al. (2014) reported an increased cerebellar brain inhibition (CBI) following anodal stimulation applied over the cerebellar cortex. By contrast, cathodal stimulation reduced CBI. Only two studies found the reverse effect (Bocci et al., 2015; Panouillères et al., 2015b). The remaining studies found an effect of either cathodal (Pope and Miall, 2012; Picazio et al., 2015) or anodal (Macher et al., 2014; Cantarero et al., 2015; Bocci et al., 2016; Wessel et al., 2016) stimulation. A few studies did not find any effect after both stimulation types (Jayaram et al., 2012; Sadnicka et al., 2013; Chen et al., 2014; Bocci et al., 2016). Heterogeneous effects of cerebellar stimulation in different tasks have been observed. In general, anodal stimulation is believed to enhance motor and cognitive functions, whereas cathodal stimulation typically inhibits functioning. However, in several studies, cathodal stimulation was associated with a neurobehavioral effect in agreement with an enhanced function of the cerebellar cortex (Galea et al., 2011; Ferrucci et al., 2012; Pope and Miall, 2012; Shah et al., 2013; Bersani et al., 2015; Bocci et al., 2015; Bradnam et al., 2015; Minichino et al., 2015; Panouillères et al., 2015b). On the other hand some studies have shown that anodal stimulation may impair cerebellar function (Ferrucci et al., 2008; Foerster et al., 2013; Dutta et al., 2014; Macher et al., 2014; Bocci et al., 2015; Doeltgen et al., 2015; Panouillères et al., 2015b; Chothia et al., 2016). For instance, Panouillères et al. (2015b) showed that cathodal cerebellar tDCS facilitates saccadic adaptation, while anodal tDCS disrupted the adaptation. The effect of anodal and cathodal stimulation also seems to depend on the behavioral task that is considered. Bradnam et al. (2015) reported an overall positive effect of anodal cerebellar tDCS on handwriting with a reduced average pen pressure, but also a slower mean stroke frequency, which could indicate a worsening of the handwriting function. Interestingly, the same effects were found with cathodal tDCS for handwriting, but the improvements in cyclic drawing (decreased pen pressure, increased average speed) were only apparent for anodal tDCS.

These findings demonstrate that more information is required about the specific impacts of tDCS on cerebellar neurons to reliably predict the outcome of a given cerebellar stimulation.

To ensure that findings are really due to stimulation and not to a placebo effect or practice, the data are often compared with the measuring before and after a sham session. A sham session usually consists of ramping up the current to the same intensity as used in the active sessions, followed by immediately ramping down the current. In most studies this procedure results in an active stimulation of 30–60 s (Galea et al., 2009; Jayaram et al., 2012; Block and Celnik, 2013; Sadnicka et al., 2013; Chen et al., 2014; Dutta et al., 2014; Hardwick and Celnik, 2014; Herzfeld et al., 2014; Zuchowski et al., 2014; Benussi et al., 2015; Doeltgen et al., 2015; Martin et al., 2015; Panouillères et al., 2015a,b; Chothia et al., 2016; Van Wessel et al., 2016; Wessel et al., 2016). To effectively blind the subjects for sham or active stimulation, the electrodes remain on the head after ramping down the current to obtain the impression of comparable session lengths. Studies have shown that ramping down the current during cerebral stimulations does not elicit perceivable sensations, while sensations of turning on the current usually fade out in the first 30 s (Gandiga et al., 2006). As a result, sham sessions with at least 30 s of active stimulation with a ramping down of the current may effectively blind subjects (Russo et al., 2013). However, a study of Kessler et al. (2012) showed that the level of discomfort is higher during active sessions of tDCS than during sham sessions, making the implicit experience of the two conditions different. Kessler et al. (2012) speculated that the difference in discomfort is dependent on: (1) the duration of the ramp up and ramp down times, (2) the duration of the sham session, and (3) the intensity of the current. Adjusting these parameters might help to reduce the difference in (dis)comfort between sham and active stimulation. The following suggestions have been made:

Since current intensities of 2 mA are typically used in cerebellar tDCS, it seems important to use proper ramp times and maybe even a short period of active stimulation during sham sessions to ensure proper blinding of both the experimenter and the participant for a double-blind design. Practically, it may be recommended to use ramp times of a minimum of 30 s and to actively stimulate for at least 30 s during sham sessions, longer if the total amount of charge delivered is high.

Multiple sessions are considered to have a cumulative effect and are needed to induce reliable and/or long-lasting after effects (Brunoni et al., 2012). The repetition rate seems to play a crucial role to induce cortical plasticity (Brunoni et al., 2012). In various cerebral stimulation trials, daily sessions have proved to be more effective than weekly sessions (Boggio et al., 2007) or sessions given every other day (Alonzo et al., 2012). In addition, Monte-Silva et al. (2010) showed that stimulating during ongoing after-effects of previous stimulation (during the MEP amplitude spike; 20 min break) resulted in prolonged and enhanced tDCS-induced effects. When the second stimulation was administered after remission of after-effects (normal MEP amplitude; 3 or 24 h break), the initial effects were first abolished or attenuated but then re-established after one (3 and 24 h break) to 2 h (24 h break only). Interestingly, no prolongation of the after-effects was observed if the stimulations were only 3 min apart. It has to be noted that this study only investigated cathodal tDCS in a healthy population. The optimal repetition rate and inter-stimulation interval has still to be determined for cortical tDCS (Brunoni et al., 2012).

Monte-Silva et al. (2010) also reported that stimulation duration (18 min instead of 9 min) has a beneficial impact on the duration of the after-effects. In comparison with a single session of 9 min, a single continuous stimulation of 18 min prolonged the after-effects from 60 to 90 min. However, this prolongation is less marked than the ones reported in a study of Nitsche et al. (2003c). These authors investigated the after effects of stimulation durations of 5, 7, and 9 min, indicating that there might be a ceiling effect of cathodal tDCS (Nitsche et al., 2003c).

The number of repetitions remains a matter of debate as well. Lindenberg et al. (2012) examined the effects of two 5-day intervention periods of bihemispheric cortical tDCS in a patient group. They showed that the second 5-day intervention also resulted in an increase of motor function, but a significantly lower one than after the first 5-day intervention. Therefore, repetitive sessions of tDCS do not necessarily induce a linearly cumulative result.

With regard to cerebellar tDCS, we lack systematic studies in which the effect of multiple sessions is studied or in which the effect of stimulation duration on cortical excitability is investigated. Most studies employing cerebellar tDCS have used a single session of 15–25 min of stimulation and reported variable outcomes (Ferrucci et al., 2015). The few studies based on multiple sessions are clinical studies. Most of these studies used multiple 5-day interventions (1–4 weeks) and administered one (Gironell et al., 2014; Ho et al., 2014; Minichino et al., 2015; Ferrucci et al., 2016) or two stimulation sessions a day (Bation et al., 2016). Bradnam et al. (2014), on the other hand, stimulated twice a week for a period of 12 weeks. The patients' conditions and outcome measures varied greatly across these studies.

Studies based on the same study populations to examine the effects of anodal, cathodal, and sham cerebellar tDCS usually separate the different stimulation conditions by at least 3 days (usually 1 week), with only two exceptions (1 day: Cantarero et al., 2015; 48 h: Foerster et al., 2013). Another way to avoid cross-over effects from previous stimulations is to start the experiment with the sham condition (Grimaldi et al., 2014b). Since long-lasting effects of tDCS-induced cerebellar excitability have not been investigated, it is recommended to adhere to long inter-session intervals (several days) or to use different groups of subjects/patients.

tDCS can be administered in two different conditions: online or offline. If the effects of tDCS are measured during the application of tDCS, or if tDCS is administered simultaneously with another intervention (such as physical/cognitive therapy, or a training session), the study applies an online approach. An offline protocol uses tDCS in between two measurements without any practice and/or therapy during the stimulation (Brunoni et al., 2012). Because of an increased cortical excitability during stimulation, the online approach may be of utmost importance in rehabilitation settings of patients with neurological conditions (Priori et al., 2009). One of the major advantages of tDCS is that the mobility of the patient is unaffected during stimulation, making online application very interesting (Priori et al., 2009). However, the precise effect of online vs. offline stimulation has not yet been studied in detail (Monti et al., 2013) and is probably dependent on the intended outcome and the targeted area (Pirulli et al., 2013). Pirulli et al. (2013) tested the effect of timing of tDCS on the outcome in a visuo-perceptual learning experiment (stimulating the visual cortex). They found that, in contrast to motor learning (Nitsche et al., 2003d; Kuo, 2007; Stagg and Nitsche, 2011), offline administration of tDCS induces a greater effect than online stimulation, suggesting that the timing of tDCS has to be investigated for each area separately.

The literature is only scantly documented with regard to the difference between online and offline stimulation in cerebellar tDCS. One study used both online and offline cerebellar tDCS in two different experiments and showed that both applications may induce cortical excitability changes (Dutta et al., 2014). Cerebellar tDCS studies using an online protocol applied tDCS during a variety of tasks: in the adaptation phase of a motor learning protocol (Galea et al., 2009; Jayaram et al., 2012; Block and Celnik, 2013; Hardwick and Celnik, 2014; Herzfeld et al., 2014; Avila et al., 2015; Calzolari et al., 2015; Panouillères et al., 2015a,b; Yavari et al., 2016) during the learning/(mental) practicing of a task (Foerster et al., 2013; Shah et al., 2013; Dutta et al., 2014; Cantarero et al., 2015; Van Wessel et al., 2016; Wessel et al., 2016) or during the acquisition phase of a conditioned response (Zuchowski et al., 2014). Most studies, however, used an offline application of cerebellar tDCS, especially when clinical study populations were involved (Grimaldi and Manto, 2013; Bradnam et al., 2014; Gironell et al., 2014; Grimaldi et al., 2014b; Ho et al., 2014; Benussi et al., 2015; Bersani et al., 2015; Bradnam et al., 2015; Minichino et al., 2015; Bation et al., 2016; Ferrucci et al., 2016). In only one study, performed by Calzolari et al. (2015), online cerebellar stimulation was used in a single patient, applying cerebellar tDCS during a prism adaptation task. Martin et al. (2015), on the other hand, used online measurements to assess the effect of cathodal stimulation of the right cerebellum on working memory in patients with bipolar disorders.

Future studies are needed to define the effect of online and offline cerebellar tDCS. The efficiency of tDCS strongly depends on the timing of tDCS and may vary depending on the stimulated area (Pirulli et al., 2013). Therefore, the mechanisms by which tDCS affects cerebellar neurons, a critical step as already mentioned, have to be identified to address these issues properly.

Modeling studies based on models of different gender and age have demonstrated that current density distributions vary among individuals according to anatomy (Parazzini et al., 2014b). As clearly established in case of cerebral cortical tDCS, the subject's response to stimulation may also depend on age, gender, brain state, hormonal levels, and pre-existing regional excitability (Kuo, 2007; Krause and Cohen Kadosh, 2014). These factors should be taken into account when comparing different studies.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.