Skip to main content
Brain Connectivity logoLink to Brain Connectivity
. 2020 May 14;10(4):170–182. doi: 10.1089/brain.2019.0710

Discordant Alpha-Band Transcranial Alternating Current Stimulation Affects Cortico-Cortical and Cortico-Cerebellar Connectivity

Claudia D Tesche 1,, Jon M Houck 1,2
PMCID: PMC7247042  PMID: 32216454

Abstract

Synchronization of oscillatory brain activity is believed to play a critical role in linking distributed neuronal populations into transient functional networks. Alpha-band alternating current stimulation (tACS) was applied over bilateral parietal cortex in a double-blind sham-controlled study to test the notion that widespread alpha mediates causal relationships in the gamma-band both within local neuronal populations and also across distant brain regions. Causal relationships of oscillatory alpha- and gamma-band activity were characterized during performance of a visual global/local attention task. Nonfocal and nonphase-locked tACS, discordant with endogenous oscillatory activity, was hypothesized to induce a performance deficit and differences in network-level causal relationships between both cortical and subcortical brain regions. Although modulation of fronto-parieto-cerebellar causal relationships was observed following stimulation, there was no evidence for a behavioral deficit. We propose that olivo-cerebellar circuits may have responded to the discordant tACS-induced currents as if they were “error signals” in the context of ongoing functional alpha-band brain dynamics. Compensatory cerebellar activity may have contributed to the lack of behavioral deficits and to differences in causal relationships observed following stimulation. Understanding a potential compensatory mechanism involving short-term plasticity in the cerebellum may be critical to developing potential clinical applications of tACS, particularly for disorders such as autism that are characterized by both atypical cortical and cerebellar dynamics.

Keywords: Granger analysis, MEG, tACS, visual attention

Introduction

Complex rhythmic activity has been observed across multiple levels of brain structure, from fluctuations of individual neuronal subthreshold membrane potentials to coherent oscillations of local field potentials between nodes of distributed brain networks. Oscillatory activity has been correlated with a wide variety of brain functions, including processing of sensory stimuli, attention, memory, and the generation of motor output, as well as with activity-induced synaptic plasticity (for a review, see Haegens et al., 2018). Oscillatory activity is believed to play a causal role in sculpting distributed functional networks, with low-frequency theta (4–8 Hz) and alpha (8–14 Hz) often associated with working memory and attentional control, and gamma (30–100 Hz) with the processing of sensory input and generation of motor output in local neuronal populations. In particular, alpha-band oscillations have been suggested to mediate top-down control over gamma-band processing of sensory input through phase/amplitude coupling (Palva and Palva, 2011; Rihs et al., 2007; van Kerkoerle et al., 2009). Importantly, atypical, reduced, and/or excessive oscillatory activity has been observed to correlate with, and may contribute to, numerous neurological and psychiatric disorders such as autism, schizophrenia, and Parkinson's disease (c.f. Buzsáki, 2006; O'Reilly et al., 2017).

Transcranial alternating current stimulation (tACS) is a noninvasive method to introduce weak oscillatory current flow into the brain and hence to test the causal role of oscillatory activity in brain function (for a review, see Antal and Paulus, 2013). The current literature suggests several hypotheses that may be tested utilizing tACS. Among these are that phase synchronization at lower frequencies (<25 Hz) may provide a timing mechanism for the temporal coherence of local processing in the gamma-band through a periodic hyperpolarization and depolarization of neuronal membrane potentials (Bonnefond et al., 2017; Florin and Baillet, 2015; Hyafil et al., 2015; Schroeder and Lakatos, 2009; see also Fries, 2005). A recent magnetoencephalographic (MEG) study of visual detection during alpha-band tACS has demonstrated that gamma-band amplitudes in the occipital cortex depended on the phase of the stimulation (Herring et al., 2019). We address a slightly different notion that widespread alpha mediates causal relationships in the gamma-band in the processing of visual stimuli both within local neuronal populations and also across distant brain regions.

The ability to parse the visual scene into local features within a global context is fundamental to human visual processing. This ability requires attention to both global and local features of the visual scene. In the present study, we combined tACS with MEG in a double-blind sham-controlled design to characterize causal relationships between brain regions in the alpha- and gamma-bands during performing a visual global/local attention task. We chose the well-known Navon global/local compound-letter task because task performance requires top-down attentional control to either global or local features of identical visual stimuli, thus reducing the confound of utilizing different stimuli for each task.

We utilized MEG to characterize the persistent effects on brain dynamics of 10-Hz tACS delivered through two nonfocal electrodes located over the left and right parietal cortex. The stimulation induced a complex pattern of current flow in the posterior cortex proximal to the electrodes. Importantly, the tACS was not phase-locked to the presentation of the visual stimuli. Given that the nonfocal and nonphase-locked 10-Hz stimulation would be discordant with endogenous alpha-band-mediated gamma-band processing of the visual stimuli, we hypothesized that there would be a subsequent deficit in performance and task-specific differences in causal relationships between both cortical and subcortical brain regions following stimulation.

Materials and Methods

Participants

The study conformed to the Declaration of Helsinki, and ethical approval was granted through the Human Research Protections Office, University of New Mexico, Albuquerque, New Mexico. All participants gave written informed consent before the experiment. Of the 10 participants who were recruited, 8 completed the study (5 male, age range: 21 − 45 years; mean age 23.4 years; SD = 2.65 years). One participant attended only one of two required experimental sessions and one participant left the study following an initial sham stimulation session due to anxiety. All participants were right-handed with no reported neurological issues and normal to corrected-normal vision.

Stimulation

In two separate double-blind experimental sessions, participants received either 10-Hz sinusoidal “verum” (1 mA peak-to-peak, 20 min) or “sham” (1 mA peak-to-peak, 30 sec) stimulation with no direct current (DC) offset (NeuroConn GmbH, Ilmenau, Germany). The verum and sham sessions were separated by 7 − 10 days and counterbalanced across participants. The current was applied through two 5 × 5 cm sponge electrodes located over the bilateral parietal cortex, centered at sites P3 and at P4 of the International 10 − 20 System and rotated from a vertical orientation by 45° (Okamoto et al., 2004). The location of the electrodes is illustrated in Figure 1A. The sponge electrodes fitted comfortably underneath the MEG helmet and were applied only during the stimulation epoch.

FIG. 1.

FIG. 1.

(A) Locations of two tACS electrodes over left and over right bilateral parietal cortex (P3 and P4 of the 10:20 EEG montage). (B) Results of a numerical simulation of the magnitude of the electric field strength corresponding to a 1 mA peak current value for this electrode configuration. EEG, electroencephalography; tACS, transcranial alternating current stimulation.

The tACS induced a common spatial pattern of current flow in the brain that waxed and waned from positive to negative values throughout each cycle of the 10-Hz stimulation. The spatial extent of the magnitude of the induced electric field induced by a peak current of 1 mA was simulated using a forward model on the MNI-152 brain in the ROAST toolbox (Huang et al., 2018). Figure 1B shows the spatial pattern of the absolute value of the current flow in the brain. Figure 1B does not show the sign of the current flow, which is dependent on the phase of the stimulation and the local cortical morphology. The majority of current flow passed through the posterior parietal cortex, precuneus (PC) and middle occipital gyrus, with channeling of current also into posterior white matter areas. The maximum induced electric field in these regions ranged from 0.32 to 0.36 V/m. The simulation did not suggest any substantial direct effect on frontal areas (i.e., middle frontal gyrus 0.046 V/m, hippocampus [HC] 0.074 V/m, cerebellum 0.05 V/m).

Task

Participants performed a visual global/local attention task (Navon, 1977; c.f. Stone and Tesche, 2009). Visual stimuli were presented sequentially in blocks of four to eight compound letters, each preceded by a cue to attend to either the global “G” or local “L” features of the stimuli (Fig. 2A). The compound letters were either a “global” “H” or “S,” each composed of smaller “local” letters, either all “H” or all “S.” Both congruent and incongruent combinations were represented equally. A global attention block was followed by a local attention block and vice versa. All stimuli consisted of white letters on a black background and were back-projected onto a screen located within a magnetically shielded room. The visual angle span for the global compound letters was 3.6 × 2.1 degrees, for the individual letters within each compound letter 0.46 × 0.27 degrees, and for the target cues 1.5 × 1.1 degrees. The duration of compound letters and target cues was 80 ms. All interstimulus intervals ranged between 1200 and 1800 ms. Approximately 30 target cues and 150 compound letters were presented in each block. Participants were requested to press a button with the right hand index finger following perception of an “H” stimulus and with the right middle finger for perception of an “S” stimulus consistently for all global and local trials. All stimuli were presented and responses recorded using Presentation software. Eyes-open and eyes-closed data were also acquired (results not presented here).

FIG. 2.

FIG. 2.

(A) Participants performed a global/local visual attention task. Compound “Navon” letters were presented sequentially in blocks of 4 − 8 images. The letters were either a global “H” or “S” formed by smaller “local” letters, either all “H” or all “S.” Both congruent and incongruent combinations were presented. Each block was preceded by a cue to respond to either the global “G” or local “L” features of the stimuli. (B, C) There were two experimental sessions, one with verum stimulation and one with sham stimulation (double-blind, randomly assigned). Each session included acquisition of baseline MEG data, MEG data recorded during stimulation, and MEG recorded after stimulation. For the verum session, 10-Hz sinusoidal tACS was delivered at 1 mA amplitude (no DC offset) for a duration of 20 min. For the sham session, the tACS was applied only for 30 sec. MEG data acquisition included 3 min of eyes-open and 3 min of eyes-closed spontaneous activity in addition to performance of the task. DC, direct current; MEG, magnetoencephalographic.

Data acquisition

MEG data were recorded by a 306-channel array (Elekta Neuromag) at the Mind Research Network (Albuquerque, NM). Four head position coils (HPI) were placed on the scalp to track the location and orientation of the participant's head throughout the scan. Bipolar horizontal and vertical electro-oculogram (EOG) and electrocardiogram electrodes were utilized to monitor eye movement and cardiac activity. A Polhemus 3D position tracker was utilized before the MEG scan to record the relative positions of the HPI and EOG coils/electrodes with respect to bilateral preauricular fiducial points and the nasion. The Polhemus tracker was also used to record the shape of the scalp and facial features.

Participants were seated under the scanner inside a double-layered magnetically shielded room (Vacuumschmelze) and visually monitored throughout the scan. A brief practice session of two global/local blocks of stimuli preceded initiation of the recording of MEG and behavioral data. MEG data were sampled at 1200 Hz and band-pass filtered on-line at 0.1 − 330 Hz. Blocks of MEG data were acquired for 3 min of spontaneous activity with eyes closed, then with eyes-open, followed by task performance (Fig. 2 B, C). There were two experimental sessions, each with three blocks: baseline, stimulation, and after-stimulation.

Structural magnetic resonance image (sMRI) data were recorded with a 3T Siemens Trio TIM at the Mind Research Network as high-resolution T1-weighted anatomic images. Scan parameters for a multiecho magnetization prepared rapid gradient echo (MPRAGE) sequence were echo time (TE) = 1.64, 3.5, 5.36, 7.22, and 9.08 ms; repetition time (TR) = 2530 ms; inversion time (TI) = 1200 ms; flip angle = 7°; number of excitations = 1; slice thickness = 1 mm; field of view (FOV) = 256 mm; resolution = 256 × 256 mm. The sMRI scan session lasted <30 min. The average image (root-mean-square) across the five echoes of the MPRAGE was used for coregistration with the MEG data and for construction of an individualized realistic boundary element model (BEM) for source localization of the MEG data.

MEG data analysis

MaxFilter (Elekta) was utilized to downsample to 600 Hz, to compensate for head movement with signal-space separation, and to remove blink and cardiac artifacts (Taulu and Simola, 2006; Tesche et al., 1995; Uusitalo and Ilmoniemi, 1996). All subsequent analyses were performed on single trial data. The analysis of the verum and sham data was identical. Single trials were extracted from the MEG data for epochs from 500 ms before stimulus presentation to 1500 ms after stimulus presentation. Trials for the first two blocks of stimuli and all trials immediately following a “G” or “L” cue were excluded from further analysis to reduce initial accommodation to the paradigm and to reduce brain dynamics related to switching, rather than sustaining, global/local task performance.

Brainstorm (http://neuroimage.usc.edu/brainstorm), a documented and freely available software package downloaded online under a GNU general public license, was used for subsequent analysis of the MEG data. Emphasis was placed on exceedingly clean data. All of the single-trial data were visually inspected in Brainstorm. Trials were rejected that contained movement and blink artifacts, incorrect responses, single or multiple button presses occurring between 500 and 0 ms before stimulus presentation (baseline) and between 0 and 1500 ms after stimulus presentation. Since the data were combined across all participants for the causal analysis, it was important to have similar numbers of trials for all participants. For 7 of the participants, we retained the first clean 40 trials for each condition for further analysis. For 1 participant, only 29 (35) acceptable trials for global (local) task performance were identified and used in the subsequent analysis. No distinction was made in the subsequent analysis between congruent and incongruent compound letters. The percentage difference of congruent to incongruent trials for each condition was 3.2% (sham global), 3.8% (sham local), 7.4% (verum global), and 4.3% (verum local).

BrainSuite (http://brainsuite.org) was used to extract a surface tessellation of the cerebral and cerebellar cortex from the individual participant's MR images. Each surface was approximated by a grid of 7000 points. The cerebral and cerebellar cortical tessellations were merged to form an individualized BEM for MEG data inversion. Acceptable MR data were not available for one participant. A cortico-cerebellar tessellation constructed for the Colin 27 adult brain (Holmes et al., 1998) was warped to fit this participant's scalp based on the Polhemus data. Brainstorm was utilized to compute a weighted minimum norm estimate of brain activation as a function of time for each trial for all participants and conditions at each tessellation point. This inversion was conducted utilizing the participant-specific BEM. The brain-based waveforms then were projected into the tessellated Colin 27 brain.

Two sets of brain regions were identified for further analysis based on existing literature. The first set was related to goal-directed processing of visual stimuli (VP). The left and right lateral occipital (LO) cortex, middle temporal (MT) cortex, medial orbitofrontal (MO) cortex, and inferior parietal (IP) cortex, as well as cerebellum, have been identified from functional magnetic resonance imaging (fMRI) data as participating to processing of visual stimuli in the ventral stream (c.f. Buckner et al., 2011; Krienen and Buckner, 2009; although see Habas et al., 2009). The second set of brain regions were members of the “Rich Club.” The Rich Club has been identified in structural diffusion tensor imaging of the human connectome as a set of densely interconnected brain network hubs (Bullmore and Sporns, 2009; van den Heuvel and Sporns, 2011). Interestingly, the Rich Club includes nodes of multiple resting-state networks determined from fMRI data. The Rich Club has been proposed to play a central role in maintaining communication between widely separated brain regions (van den Heuvel et al., 2009). Thus, we anticipated that the Rich Club would play a role in shifting brain dynamics between networks involved in global and in local feature detection. The notion that widespread alpha-band activity plays a causal role in sculpting local gamma-band dynamics also motivated including Rich Club nodes. Looking across results from structural MRI, functional MRI, and MEG/electroencephalography (EEG) methodologies, it would seem reasonable that alpha-band activity might form a component of Rich Club dynamics. Thus, our expectation was that discordant alpha-band tACS would affect causality both within and between both Rich Club and VP regions.

Figure 3 shows the cortical regions for the Desikan/Killiany atlas as implemented in Brainstorm (Desikan et al., 2006). Since a functional atlas for the cerebellum is not available in Brainstorm, the locations of the brain regions in lateral left and right crus I/II of the cerebellum were defined in the Colin 27 brain from evoked-response waveforms averaged over all participants, trials, and conditions (data not shown).

FIG. 3.

FIG. 3.

Brain regions used in the present study. A source-space analysis was performed on the MEG sensor data. Waveforms were determined on a tessellation of cortical and subcortical surfaces. The waveforms were then averaged over cortical surfaces as defined in the Desikan/Killiany atlas. Waveforms were also determined from the tessellation for brain regions in left and right crus I/II of cerebellum.

Waveforms were determined for each VP and Rich Club node were determined as mean values at each sampled instant across all tessellation points in the respective brain regions. A bivariate spectral Granger analysis was performed in Brainstorm on the waveforms for each pair of brain regions for each participant, trial, and condition (Ding et al., 2006; Granger, 1969; Wen et al., 2013). Analysis procedures were identical for global and local trials. An epoch from 100 to 500 ms after stimulus onset was chosen to be long enough to permit evaluation of causality at 10 Hz and short enough to avoid baseline activity, activity before the earliest substantial MEG responses, and activity related to the evaluation of motor responses. Average values across the epoch were subtracted before the Granger analysis. Causal relationships were computed with a maximum frequency of 100 Hz, a frequency resolution of 2 Hz, and maximum Granger model order 10. Parametric p-values were returned based on Wald statistics (Geweke, 1982). A single tACS versus sham comparison was conducted on the set of all individual-trial Granger values for each pair of brain regions and task (global N = 309, local N = 315). A nonparametric permutation test for significance (two-tailed t-test, randomizations = 1000) was followed by a false discovery rate (FDR) correction for multiple comparisons (threshold p = 0.05). Results are presented at the drive frequency (10 Hz) and at frequencies characteristic of the processing of visual stimuli at low (30 and 40 Hz) and high (70 Hz) gamma.

Behavioral data analysis

The multivariate approach to repeated-measures analysis of variance was used to test the influences of task conditions and stimulation on behavioral performance, examined as both participant response time and as accuracy (percentage correct). Effects of interest included verum versus sham (“Stimulation”), global versus local (“Attention”), and prestimulation versus poststimulation (“Time”). Main effects were compared using univariate pairwise comparisons to explain the direction of effects.

Results

Behavioral results

There was no evidence of a significant stimulation × attention interaction [F(1, 7) = 4.277, p = 0.077] or of a stimulation × time × attention interaction [F(1, 7) = 1.777, p = 0.224]. The values in seconds of the response times as a function of condition are shown in Table 1. Persistent changes in motor responses to the visual stimuli shown in Figure 2 included significant main effects on response time for time [F(1, 7) = 9.189, p = 0.019, ηP2 = 0.568] and for attention [F(1, 7) = 38.602, p < 0.001, ηP2 = 0.846] conditions. There was a significant interaction between time and attention [F(1, 7) = 8.692, p = 0.021, ηP2 = 0.554] conditions such that at baseline only, the mean response time for local stimuli was longer than that for global stimuli. There was no evidence of a significant stimulation × attention interaction [F(1, 7) = 4.277, p = 0.077, ηP2 = 0.379] or of a stimulation × time × attention interaction [F(1, 7) = 1.777, p = 0.224, ηP2 = 0.202]. The values in seconds of the response times as a function of condition are shown in Table 1.

Table 1.

Mean Response Times in Seconds and Accuracy as a Function of Condition

Condition Response time (sec)
Accuracy (% correct)
Mean (SD)
Mean (SD)
Verum Sham Verum Sham
Local
 Pre 0.628 (0.093) 0.642 (0.096) 0.964 (0.028) 0.967 (0.025)
 Post 0.558 (0.054) 0.552 (0.062) 0.974 (0.037) 0.977 (0.02)
Global
 Pre 0.511 (0.102) 0.554 (0.113) 0.968 (0.028) 0.977 (0.02)
 Post 0.498 (0.059) 0.498 (0.059) 0.976 (0.02) 0.982 (0.02)

Effects for accuracy (percentage correct) in responding to the visual stimuli included only a significant main effect of time [F(1, 7) = 9.429, p = 0.018, ηP2 = 0.574], such that accuracy was higher poststimulation than prestimulation. There were no significant effects of stimulation or for attention (all p > 0.25), nor any interactions (all p > 0.74). Descriptive statistics for accuracy as a function of condition are shown in Table 1.

tACS induced widespread differences in cortical/subcortical causal relationships

Figure 4 shows results from a verum versus sham comparison across all trials of the bivariate spectral Granger causal relationships for the epoch from 100 to 500 ms following stimulus presentation. The magnitude and direction of the differences between the bivariate spectral Granger values following verum and those following sham stimulation are shown for each participant in Figure 5. Figure 6 illustrates the results shown in Figure 4 as a function of network membership.

FIG. 4.

FIG. 4.

Results from a verum versus sham comparison of the bivariate spectral Granger values for (A) global and (B) local task performance. The analysis was conducted on all of the individual-trial brain-based waveforms for global (N = 309) and for local (N = 315) task performance for each pair of brain regions shown in Figure 3. The analysis was performed on epochs from 100 − 500 ms after presentation of the compound letter stimuli and evaluated at 10, 30, 40, and 70 Hz. A nonparametric permutation test for significance (two-tailed t-test, randomizations = 1000) was followed by an FDR correction for multiple comparisons (threshold p = 0.05). The colored squares indicate the t-values for the comparisons, with red (blue) squares corresponding to an increase (decrease) of causality following verum compared with sham stimulation. The grid from left to right (top to bottom) is organized according to network. L and R Rich Club: SF cortex, SP cortex, PC, and HC. L and R stimulus processing areas (VP): LO cortex, MT cortex, IP cortex, MF cortex, and CZ. HC, hippocampus; IP, inferior parietal; L, left; LO, lateral occipital; MF, medial orbitofrontal; MT, middle temporal; PC, precuneus; R, right; SF, superior frontal; SP, superior parietal.

FIG. 5.

FIG. 5.

Individual differences in the bivariate spectral Granger causal relationships for (A) global and (B) local task performance. The analysis was performed on epochs from 100 to 500 ms after presentation of the compound letter stimuli and evaluated at 10, 30, 40, and 70 Hz. The individual participants are identified by the color of the bars. The horizontal extent of each of the bars represents the difference between the bivariate spectral Granger values following verum and those following sham stimulation for each participant. The pairs of brain regions for each combined set of of horizontal bars and the direction of the interaction (i.e., LSF:LPC for the causal relationship from LSF to LPC) is indicated on the right. Bars located to the left (right) side of the individual vertical axes represent a decrease (increase) of causality following verum compared with sham stimulation. LPC, left PC; LSF, left SF.

FIG. 6.

FIG. 6.

FIG. 6.

(A–E) Schematic representation of the bivariate spectral Granger causal relationships between pairs of brain regions as a function of task and frequency. Rich Club brain regions are indicated by red and VP by white circles. Rose (blue) connections indicate increased (decreased) causality following verum compared with sham stimulation. VP, visual processing network.

Differences in causal relationships at the drive frequency

The bilateral parietal stimulation induced significant differences in the spectral values not only between pairs of brain regions located in posterior cortex directly under the stimulation electrodes, but also more widely between posterior, frontal, and subcortical brain regions for both global and local task performances. At the tACS drive frequency of 10 Hz, differences in causality were seen not only within the VP and Rich Club but also between nodes of these two networks. Interestingly, the stimulation had a modest effect on causality between superior parietal (SP), IP, and PC, regions that received the most actual current stimulation. The only significant difference was for the pair left IP onto left SP for global trials. Moreover, all but one of the differences corresponded to decreased causality following verum compared with following sham stimulation.

For the VP network, differences in causality for global task performance included a decrease from right CZ onto left IP and left LO, from right CZ onto right LO, and from right CZ onto left CZ. Causality decreased from left CZ onto left IP, from left LO onto left IP, and from right LO onto left MT. For local task performance, causality decreased from right CZ onto right IP, and from right LO onto left CZ. The only increase in causality was seen in the right hemisphere from IP onto MT.

For Rich Club and global task performance, both left and right superior frontal (SF) influence onto right HC and left SF onto left PC decreased. There were decreases from right HC onto left SF and left SP. Causality from left (right) HC onto right (left) SP also decreased. For local task performance, causality was reduced from left SF onto left PC and onto right HC and from left HC onto left SF.

Finally, there were numerous differences in causality that involved nodes from VP onto Rich Club and vice versa. For global task performance, Rich Club nodes of left (right) SF and HC had reduced causality onto VP targets of right (left) MT, as well as left HC onto MT. Causality was reduced for left HC onto left and right LO. For VP, causality was reduced from right CZ onto right HC and left PC, right HC onto left SF, and right MO onto left PC. In the left hemisphere, reductions occurred from left MT onto left SF and from left IP to left SP. For local task performance, Rich Club nodes of left and right SF had reduced causality onto left IP, and right HC onto right MT, as well as left HC onto left LO. For VP, causality was reduced from right CZ onto left SP and from left CZ onto left SF, as well as from right MT onto right SP and right SF.

Differences in causal relationships at gamma-band frequencies

Causal relationships were evaluated for low (30 and 40 Hz) and for high (70 Hz) gamma. The 30-Hz results are quite interesting, in that significant differences were observed only for local task performance. There were no significant differences at 40 Hz. Increased causality involved the VP nodes of right CZ onto left MO, and the left MO onto left IP, left MT, and right MO. The only increase in the Rich Club was from left HC onto right HC. For mixed connections, causality increased from right MT onto right PC and from right PC onto left IP. All of the decreases involved the SF cortex: in the Rich Club from left SF onto left PC and from left HC onto left SF. For mixed pairs, causality decreased from left CZ onto left SF, and from left and right SF onto left IP.

Significant differences in causal relationships were seen at 70 Hz, and, in contrast to the 10-Hz results, most of the differences corresponded to increased causality following verum compared with following sham stimulation. For global trials, within VP causality increased for left MO onto left LO and right MO, and for left MT onto right MO. For the Rich Club, increases were seen for left and right HC onto right PC, although left SP onto left SF increased. For mixed pairs, causality increased from left (right) HC onto right (left) MT. Left HC onto left and right MO also increased, as well as right MO onto left HC. For mixed pairs, increases included from right MT to right PC, from right PC to left IP, and from right HC to left MT. The only decrease was from left SP to right CZ.

Discussion

Behavioral measures: entrainment and stochastic resonance

There is intense interest in the possibility of utilizing transcranial current stimulation to induce persistent change in brain function and behavior in both healthy and clinical populations. Our expectation was that the discordant tACS would diminish, rather than enhance, performance of the global/local task. No evidence for a decrease in accuracy or difference in response times due to stimulation was observed in the present study. We suggest that the lack of a behavioral deficit may be due to the absence of a specific temporal relationship between the 10-Hz tACS and the presented stimuli. Previous studies have shown that neural entrainment to exogenous stimuli modulates sensory processing (for a review, see Haegens and Golumbic, 2018). Entrainment and stochastic resonance also occurs when the tACS frequency is matched to the participant's occipital alpha-band activity in eyes-open/eyes-closed conditions. The time to entrainment occurs over an epoch from ∼1.6 to 2.1 sec (Ali et al., 2013; Fehér et al., 2017). In the present study, the tACS frequency was not matched to an individual participant's eyes-closed alpha-band activity, nor to a spectral peak observed within a specific brain region during task performance, although there was considerable overlap between the stimulation frequency and task-related alpha-band activity observed in posterior cortical and subcortical areas. Importantly, the compound letter stimuli were presented at randomized interstimulus intervals. Even though entrainment may have been initiated in the baseline epoch, there was no fixed phase relationship between the applied tACS and the onset of each of the visual stimuli, reducing the potential for phase-dependent mechanisms to have a negative effect on behavior. Limitations on the behavioral results include the participants' performance near ceiling (accuracy ∼97%). This facilitated testing for behavioral deficits and also reduced the potential for the inclusion of error-related dynamics in the MEG analysis, but also compromised the potential to detect subtle increases in performance, as did the small sample size. In addition, the trial selection process introduced a slight imbalance between the number of congruent and incongruent trials across conditions in both the behavioral and MEG results.

Persistent differences in alpha-band dynamics

Previous studies of the aftereffects of tACS on brain dynamics have revealed frequency-specific increases in the power of endogenous oscillatory activity following stimulation (Neuling et al., 2013; Vossen et al., 2015; Zaehle et al., 2010). Increased posterior alpha-band activity and coherence following stimulation have been reported in tACS/EEG studies involving visual stimuli, although these scalp-EEG data were not analyzed to reveal differences in neuronal activity within specific brain regions (Kasten and Herrmann, 2016, 2017). In an MEG study of mental rotation, a beamformer approach revealed on-line and persistent increases in alpha induced by alpha-band tACS over parieto-occipital cortex (Kasten et al., 2018). Not discussed was the potential for a difference in posterior alpha to cause corresponding differences in causal relationships in the alpha-band between posterior, frontal, and subcortical structures.

Differences in brain dynamics following stimulation have been suggested to result from short-term neural plasticity, rather than to an entrainment mechanism per se (Vossen et al., 2015). We first discuss persistent differences in causality observed between brain regions following 10-Hz tACS. Looking across both tasks, all nodes in the VP and Rich Club, except right PC and left MO, experienced at least one difference following verum compared with sham stimulation, including nodes distant to the stimulation electrodes. This result evidences persistent and widespread differences in causal alpha-band dynamics following verum compared with sham tACS. Were the differences task-specific and hence dependent on the state of the brain? There is considerable evidence to support lateralization differences in the processing of global compared with local features of visual stimuli. Local features tend to elicit a left hemisphere and global a right hemisphere processing bias (Ivry and Robertson, 1998: Martin, 1979; Robertson and Ivry, 2000). Behavioral, patient, and fMRI studies of Navon compound-letter tasks show lateralization of brain activity in posterior visual processing regions, including temporo-parietal junction, parieto-occipital cortex, and superior temporal gyrus (for a review, see Flevaris et al., 2016). EEG studies have demonstrated a correlation between faster response times to local (global) features with higher baseline alpha amplitudes in right (left) centro-parietal cortex, with inhibition of alpha evidencing top-down control of stimulus processing (Volberg et al., 2009).

In the present study, the Navon letters were presented to the center of the visual field and the transcranial stimulation was bilateral, reducing the expectation that the tACS would induce task-dependent differences in the lateralization of brain dynamics. Nevertheless, differences in causality depended on task, hemisphere, and network membership. For global, but not local, trials, the majority of differences within the VP network involved a decrease of causality of the right crus I/II of cerebellum (CZ) onto nodes in the left hemisphere (right LO was an exception). Since right lateral cerebellum projects to left cortical targets, this result is consistent with relatively stable dynamics within right cortical VP. The results for nodes between VP and Rich Club and within Rich Club were similar. Although there were multiple differences involving SF cortex, HC, and CZ, there were few differences in causal relationships for global trials that involved right LO, IP, SP, and PC. The lack of a causal impact on alpha in the right cerebral cortex may have contributed to the lack of behavioral deficits seen in global task performance. Although less numerous, a similar pattern held for differences observed for local task performance. For pairs within VP, no nodes in the left cerebral cortex showed altered causality for local trials, again consistent with a lack of behavioral deficit.

Differences in causality for pairs of nodes within Rich Club and between VP and Rich Club were seen for local as well as for global task performance. Rich Club has been proposed to play a central role in maintaining communication between widely separated brain regions (van den Heuvel et al., 2009). We included Rich Club in the analysis with the expectation that this densely interconnected set of brain network hubs would play a role in a task that required top-down shifting of attention and response between global and local features. Task-dependent differences induced within the Rich Club and between Rich Club and VP lend support to the notion that alpha-band activity may mediate Rich Club involvement in maintaining communication between widely separated brain regions.

Persistent differences in gamma-band dynamics

Previous studies have demonstrated that alpha/gamma phase amplitude is feature of local neuronal population dynamics (Palva and Palva, 2011; Rihs et al., 2007; van Kerkoerle et al., 2014). Previous studies also have demonstrated that gamma-band synchrony can extend across widely separated brain regions (Gollo et al., 2014; Traub et al., 1996). The hypothesis under consideration is that widespread alpha mediates not just local phase/amplitude coupling between alpha and gamma, but also causal relationships in the gamma-band across distant brain regions. Thus, we expect that persistent differences induced by tACS in causal relationships between pairs of regions in the alpha-band should correspond to similar differences in causality observed in the gamma-band.

Were the differences in causality observed for individual pairs of regions at 10 Hz also observed at gamma frequencies? For global task performance, there were no significant differences at 30 Hz nor at 40 Hz, and hence, no evidence that the differences in causality at 10 Hz were related to differences in causality for low gamma. In addition, for global trials, there was only one significant difference at 70 Hz for VP nodes (right CZ to right LO) that was seen also at 10 Hz. There were no common pairs within the Rich Club, nor for pairs between Rich Club and VP. What about local trials? At 30 Hz, within the VP network, there were no common pairs. Within Rich Club (left HC to left SF and left SF to left PC) and between Rich Club and VP (left CZ to left SF, left SF to left IP, and right SF to left IP) were represented also at the drive frequency. At 70 Hz, causality decreased from right CZ to left SP.

Although both increases and decreases in causal relationships were seen at 30 and at 70 Hz, pairs of nodes with significant differences at both the drive and gamma frequencies experienced only decreases in causality. For local trials, SF cortex was common for all pairs. IP cortex was also a target. CZ onto cortical targets was involved for both global and local task performance, suggesting engagement of a fronto-parieto-cerebellar network by the 10-Hz stimulation.

Discordant alpha-band tACS as an “error signal” for cerebellum

The traditional role of cerebellum in motor learning has been expanded to include cerebellar participation to a broad spectrum of sensory, cognitive, and affective functions, including processing of temporal information in the subsecond range (Molinari et al., 2007; Schmahmann, 1991; Schmahmann and Pandya, 1997; for a review, see Baumann et al., 2015). Given that afferents to lateral cerebellar cortex include somatosensory, motor, prefrontal, posterior parietal, and temporal cortex, the cerebellum has been proposed to contribute to the timing and synchronization of activity within and between distributed brain regions (Keele and Ivry, 1990; Ivry, in Baumann et al., 2015; for a review, see Baumann et al., 2015).

Whole-scalp MEG provides an opportunity to record oscillatory activity in the human cerebellum (Tesche and Karhu, 2000). The dominant source of the MEG signals is believed to originate from postsynaptic activity generated by climbing fiber synapses from inferior olivary neurons onto the extensive dendritic processes of Purkinje cells in the cerebellar cortex (Okada and Nicholson, 1988). Alpha-band activity has been observed both as subthreshold oscillations in the olivary nucleus and as local field potentials in crus II of cerebellum. These oscillations are believed to drive task-dependent spatiotemporal patterning of Purkinje cell populations, with a bias for temporal events separated by 100 ms (for a review, see Courtemanche et al., 2013). Lateral neocerebellum projects through the dentate nucleus to cortical frontal and posterior association areas, providing a justification for a modulation of causal influences of the CZ nodes onto fronto-parietal nodes SF and IP following the 10-Hz tACS.

We propose that fronto-parieto-cerebellar circuits may have been elicited to compensate for the discordant 10-Hz stimulation. Brain dynamics following stimulation may have reflected rebound brain dynamics similar to those that contribute to compensatory movement errors following removal of prism glasses (c.f. Calzolari et al., 2015; Panico et al., 2018; Redding et al., 2005). Compensation for the 10-Hz tACS currents during stimulation not only may have been evidenced by differences in cortico-causal relationships following stimulation but also by the lack of behavioral deficits seen during and following stimulation.

Conclusion

Synchronization of oscillatory brain activity is believed to play a critical role in linking distributed neuronal populations into transient functional networks. Phase synchronization at lower frequencies (<25 Hz) provides a timing mechanism for the temporal coherence of local processing in the gamma-band through a periodic hyperpolarization and depolarization of neuronal membrane potentials. Measures of phase/amplitude coupling are often used in MEG/EEG and animal studies to test this hypothesis within local neuronal populations. The present study addressed a slightly different notion that widespread alpha mediates causal relationships in the gamma-band not just within local neuronal populations but also across distant brain regions. 10-Hz tACS was applied over bilateral parietal cortex during performance of a visual global/local attention task. We hypothesized that nonfocal and nonphase-locked tACS would be discordant with endogenous oscillatory activity, leading to a performance deficit and to differences in network-level causal relationships between both cortical and subcortical brain regions. Although persistent differences in fronto-parieto-cerebellar causal relationships were observed following stimulation, there was no evidence for a behavioral deficit. We propose that olivo-cerebellar circuits may have responded to the discordant tACS-induced currents as if they were “error signals” in the context of ongoing functional alpha-band brain dynamics. Compensatory cerebellar activity may have contributed to the lack of behavioral deficits and to differences in causal relationships observed following stimulation. Understanding a potential compensatory mechanism involving short-term plasticity in the cerebellum may be critical to developing potential clinical applications of tACS, particularly for disorders such as autism that are characterized by both atypical cortical and cerebellar dynamics.

Acknowledgments

We thank Megan Schendel and the staff at the Mind Research Network for expertise in MEG data acquisition.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported by a joint grant from the Mind Research Network, Albuquerque, NM, and a Research Allocation Grant from the University of New Mexico (index #765301 fund #2U0005). JMH's effort on this work was supported by the National Institute on Alcohol Abuse and Alcoholism under award number K01AA021431. The funding agencies had no role in the design of the study, interpretation of the data, or in the decision to submit the work for publication. The MEG and behavioral data were acquired at the Mind Research Network.

References

  1. Ali MM, Sellers KK, Fröhlich F. 2013. Transcranial alternating current stimulation modulates large-scale cortical network activity by network resonance. J Neurosci 33:11262–11275 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Antal A, Paulus W. 2013. Transcranial alternating current stimulation (tACS). Front Hum Neurosci 7:317. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baumann O, Borra RJ, Bower JM, Cullen KE, Habas C, Ivry RB, et al. 2015. Consensus paper: the role of the cerebellum in perceptual processes. Cerebellum 14:197–220 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bonnefond M, Kastner S, Jensen O. 2017. Communication between brain areas based on nested oscillations. eNeuro 4: ENEURO.0153–16.2017 [DOI] [PMC free article] [PubMed]
  5. Buckner RL, Fenna KM, Castellanos A, Diaz JC, Yeo BTT. 2011. The organization of the human cerebellum estimated by intrinsic functional connectivity. J Neurophysiol 106:2322–2345 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bullmore E, Sporns O. 2009. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci 10:186–198 [DOI] [PubMed] [Google Scholar]
  7. Buzsáki G. 2006. Rhythms of the Brain. Oxford: Oxford University Press [Google Scholar]
  8. Calzolari E, Bolognini N, Casati C, Marzoli SB, Vallar G. 2015. Restoring abnormal after effects of prismatic adaptation through neuromodulation. Neuropsychologia 74:162–169 [DOI] [PubMed] [Google Scholar]
  9. Courtemanche R, Robinson JC, Aponte DI. 2013. Linking oscillations in cerebellar circuits. Front Neural Circuits 7:125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, et al. 2006. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage 31:968–980 [DOI] [PubMed] [Google Scholar]
  11. Ding M, Chen Y, Bressler S. 2006. Granger causality: basic theory and application to neuroscience. In: Schelter, B, Winterhalder, M, Timmer J (eds.), Handbook of Time Series Analysis. Weinheim, Germany: Wiley-VCH Verlag pp. 451–474. [Google Scholar]
  12. Fehér KD, Nakataki M, Morishima Y. 2017. Phase-dependent modulation of signal transmission in cortical networks through tACS-induced neural oscillations. Front Hum Neurosci 11:471. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Flevaris AV, Robertson LC. 2016. Spatial frequency selection and integration of global and local information in visual processing: a selective review and tribute to Shlomo Bentin. Neuropsychologia 83:192–200 [DOI] [PubMed] [Google Scholar]
  14. Florin E, Baillet S. 2015. The brain's resting-state activity is shaped by synchronized cross-frequency coupling of neural oscillations. Neuroimage 111:26–35 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fries P. 2005. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends Cogn Sci 9:474–480 [DOI] [PubMed] [Google Scholar]
  16. Geweke J. 1982. Measurement of linear dependence and feedback between multiple time series. J Am Stat Assoc 77:304–313 [Google Scholar]
  17. Gollo LL, Mirasso C, Sporns O, Breakspear M. 2014. Mechanisms of zero-lagsynchronization in cortical motifs. PLoS Comput Biol 10:e1003548. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Granger CWJ. 1969. Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37:424–438 [Google Scholar]
  19. Habas C, Kamdar N, Nguyen D, Prater K, Beckmann CF, Menon V, Greicius MD. 2009. Distinct cerebellar contributions to intrinsic connectivity networks. J Neurosci 29:8586–8594 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Haegens S. Golumbic EZ. 2018. Rhythmic facilitation of sensory processing: a critical review. Neurosci Biobehav Rev 86:150–165 [DOI] [PubMed] [Google Scholar]
  21. Herring JD, Esterer S, Marshall TR, Jensen O, Bergmann TO. 2019. Low-frequency alternating current stimulation rhythmically suppresses gamma-band oscillations and impairs perceptual performance. Neuroimage 184:440–449 [DOI] [PubMed] [Google Scholar]
  22. Holmes CJ, Hoge R, Collins L, Woods R, Toga AW, Evans AC. 1998. Enhancement of MR images using registration for signal averaging. J Comput Assist Tomogr 22:324–333 [DOI] [PubMed] [Google Scholar]
  23. Huang Y, Datta A, Bikson M, Parra LC. 2018. ROAST: An open-source, fully-automated, realistic volumetric-approach-based simulator for TES. Conf Proc IEEE Eng Med Biol Soc 2018:3072–3075 [DOI] [PubMed] [Google Scholar]
  24. Hyafil A, Giraud AL, Fontolan L, Gutkin B. 2015. Neural cross-frequency coupling: Connecting architectures, mechanisms, and functions. Trends Neurosci 11:725–740 [DOI] [PubMed] [Google Scholar]
  25. Ivry RB, Robertson LC. 1998. The Two Sides of Perception. Cambridge, MA: The MIT Press [Google Scholar]
  26. Keele SW, Ivry R. 1990. Does the cerebellum provide a common computation for diverse tasks? A timing hypothesis. Ann N Y Acad Sci 608:179–207. discussion 207–211. [DOI] [PubMed] [Google Scholar]
  27. Kasten FH, Herrmann CS. 2017. Transcranial alternating current stimulation (tACS) enhances mental rotation performance during and after stimulation. Front Hum Neurosci 11:2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kasten FH, Maess B, Herrmann CS. 2018. Facilitated event-related power modulations during transcranial alternating current stimulation (tACS) revealed by concurrent tACS-MEG. eNeuro 5. DOI: 10.1523/ENEURO.0069-18.2018 1–15 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Krienen FM, Buckner RL. 2009. Segregated fronto-cerebellar circuits revealed by intrinsic functional connectivity. Cereb Cortex 19:2485–2497 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Martin M. 1979. Hemispheric specialization for local and global processing. Neuropsychologia 17:33–40 [DOI] [PubMed] [Google Scholar]
  31. Molinari M, Leggio MG, Thaut MH. 2007. The cerebellum and neural networks for rhythmic sensorimotor synchronization in the human brain. Cerebellum 6:18–23 [DOI] [PubMed] [Google Scholar]
  32. Navon D. 1977. Forest before trees: precedence of global features in visual perception. Cogn Psychol 9:353–383 [Google Scholar]
  33. Neuling T, Rach S, Herrmann CS. 2013. Orchestrating neuronal networks: sustained aftereffects of transcranial alternating current stimulation depend upon brain states. Front Hum Neurosci 7:161. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Okada YC, Nicholson C. 1988. Magnetic evoked field associated with transcortical currents in turtle cerebellum. Biophys J 53:723–731 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Okamoto M, Dan H, Sakamoto K, Takeo K, Shimizu K, Kohno S, et al. 2004. Three-dimensional probabilistic anatomical cranio-cerebral correlation via the international 10–20 system oriented for transcranial functional brain mapping. Neuroimage 21:99–111 [DOI] [PubMed] [Google Scholar]
  36. O'Reilly C, Lewis JD, Elsabbagh M. 2017. Is functional brain connectivity atypical in autism? A systematic review of EEG and MEG studies. PLoS One 12:e0175870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Palva S, Palva JM. 2011. Functional roles of alpha-band phase synchronization in local and large-scale cortical networks. Front Psychol 2:204. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Panico F, Sagliano L, Grossi D, Trojano L. 2018. Bi-cephalic parietal and cerebellar direct current stimulation interferes with early error correction in prism adaptation: toward a complex view of the neural mechanisms underlying visuomotor control. Cortex 109:226–233 [DOI] [PubMed] [Google Scholar]
  39. Redding GM, Rossetti Y, Wallace B. 2005. Applications of prism adaptation: a tutorial in theory and method. Neurosci Biobehav Rev 29:431–444 [DOI] [PubMed] [Google Scholar]
  40. Rihs TA, Michel CM, Thut G. 2007. Mechanisms of selective inhibition in visual spatial attention are indexed by alpha-band EEG synchronization. Eur J Neurosci 25:603–610 [DOI] [PubMed] [Google Scholar]
  41. Robertson LC, Ivry RB. 2000. Hemispheric asymmetries Attention to visual and auditory primitives. Curr Direct Psychol Sci 9:59–63 [Google Scholar]
  42. Schmahmann JD. 1991. An emerging concept. The cerebellar contribution to higher function. Arch Neurol 48:1178–1187 [DOI] [PubMed] [Google Scholar]
  43. Schmahmann JD, Pandya DN. 1997. The cerebrocerebellar system. Int Rev Neurobiol 41:31–60 [DOI] [PubMed] [Google Scholar]
  44. Schroeder CE, Lakatos P. 2009. Low-frequency neuronal oscillations as instruments of sensory selection. Trends Neurosci 32:9–18 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Stone DB, Tesche CD. 2009. Transcranial direct current stimulation modulates shifts in global/local attention. Neuroreport 20:1115–1119 [DOI] [PubMed] [Google Scholar]
  46. Taulu S, Simola J. 2006. Spatiotemporal signal space separation method for rejecting nearby interference in MEG measurements. Phys Med Biol 51:1759–1768 [DOI] [PubMed] [Google Scholar]
  47. Tesche CD, Karhu JJ. 2000. Anticipatory cerebellar responses during somatosensory omission in man. Hum Brain Mapp 9:119–142 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Tesche CD, Uusitalo MA, Ilmoniemi RJ, Huotilainen M, Kajola M, Salonen O. 1995. Signal-space projections of MEG data characterize both distributed and well-localized neuronal sources. Electroencephalogr Clin Neurophysiol 95:189–200 [DOI] [PubMed] [Google Scholar]
  49. Traub RD, Whittington MA, Stanford IM, Jefferys JG. 1996. A mechanism for generation of long-range synchronous fast oscillations in the cortex. Nature 383:621–624 [DOI] [PubMed] [Google Scholar]
  50. Uusitalo MA, Ilmoniemi RJ. 1996. Signal-space projection method for separating MEG or EEG into components. Med Biol Eng Comput 35:135–140 [DOI] [PubMed] [Google Scholar]
  51. van den Heuvel M, Sporns O. 2011. Rich-club organization of the human connectome. J Neurosci 31:15775–15786 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. van Kerkoerle T, Self MW, Dagnino B, Gariel-Mathis MA, Poort J, van Der Volberg G, et al. 2009. EEG alpha oscillations in the preparation for global and local processing predict behavioral performance. Hum Brain Mapp 30:2173–2183 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Volberg G, Kliegl K, Hanslmayr S, Greenlee MW. 2009. EEG alpha oscillations in the preparation for global and local processing predict behavioral performance. Hum Brain Mapp 30:2173–2183 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Vossen A, Gross J, Thut G. 2015. Alpha power increase after transcranial alternating current stimulation at alpha frequency (alpha-tACS) reflects plastic changes rather than entrainment. Brain Stimul 8:499–508 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Wen X, Rangarajan G, Ding M. 2013. Multivariate Granger causality: an estimation framework based on factorization of the spectral density matrix. Philos Trans A Math Phys Eng Sci 371:20110610. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Zaehle T, Rach S, Herrmann CS. 2010. Transcranial alternating current stimulation enhances individual alpha activity in human EEG. PNAS 5:e13766. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Brain Connectivity are provided here courtesy of Mary Ann Liebert, Inc.

RESOURCES