Does fronto-parietal phase synchrony in the
alpha-band affect visuo-spatial attention?
– a combined EEG – tACS approach
byLynn K. A. Sörensen
A research report submitted in partial fulfillment of the degree MSc in Brain and Cognitive Sciences, Track: Cognitive Neuroscience
University of Amsterdam 07.01.2016 – 15.08.2016 32 EC Lynn K. A. Sörensen Carolina MacGillavrylaan 2652 1098 XK Amsterdam Student number: 11116633 lynn.sorensen@student.uva.nl
Heleen Slagter, supervisor
Martine van Schouwenburg, co-assessor
Cognition and Plasticity Laboratory Department of Psychology
Abstract
Visual-spatial attention shapes human perception and the mechanism by which this selectivity is achieved is a matter of vivid scientific debate. Neuroimaging identified a group of cortical regions, the fronto-parietal attention network (FPAN), argued to orchestrate attentional selection. A concurrent line of research advances the idea that this selection could be achieved through phase synchronization of brain oscillations allowing to integrate and select information between remote cortical areas. Interestingly, there is indication of functional connectivity in the alpha-band between hubs close to the FPAN. Integrating these findings, the present study will ask whether communication in the FPAN can be modulated by chancing phase coherence in the alpha-band (8 – 13 Hz), resulting in changes in attentional selection. A 10-Hz-transcranial alternating current stimulation (tACS) was applied to a frontal and parietal site of the right FPAN. The phase relationship between those sites was varied systematically (0° and 180° relative phase difference) and behavioral effects were assessed in a spatial cueing paradigm. Simultaneous EEG-measurements allowed us to assess the aftereffects of stimulation in stimulation-free blocks. Behavioral analysis did not show any significant effect of stimulation as compared to a sham-group. The electrophysiological correlate of spatial attention, the alpha modulation index, did show a stimulation specific change in the right hemisphere, which was, however, independent of the stimulated phase relationship. Furthermore, the functional connectivity assessed in phase-locking-values between frontal and parietal sites was not changed as a result of stimulation, raising doubts about the success of the phase-manipulation. Implications of the observed dissociation between behavior, alpha modulation and functional connectivity are discussed.
Keywords: Visuo-spatial attention, transcranial alternating current stimulation, fronto-parietal attention network, phase coherence, alpha-band
Attention, the ability to focus selectively on task-relevant information and to resist irrelevant and distracting information (Posner, 1980) is a key concept for the understanding of human performance. However, while the world becomes more and more challenging by providing ongoing distractions via e.g. advertisements, the neural substrates and enabling mechanisms of attention remain elusive. The present study aims to clarify the neural underpinnings of directed spatial attention. More specifically, it examines the role of phase synchronization in alpha-band oscillations in the fronto-parietal attention network (FPAN) as a candidate mechanism for the direction of spatial attention.
The dynamics of visuo-spatial attention have been the aim of cognitive psychology and a large body of research describes the cost and gain in accuracy and speed that are modulated by attention. This benefit of visuospatial attention becomes clear in a spatial cueing paradigm (Posner, 1980), during which participants are instructed to direct their attention to the location of an upcoming target, leading to improved performance. However, this cue can also be misleading and guide the participants’ attention to an alternative location. Such a situation usually results in decreased performance in response to the target. Thus, the cue introduces a bias in perception, a phenomenon known as directed visuospatial attention.
Even though the behavioral aspects are well described, the neural mechanism by which such behavior can be enabled stand on less solid ground. Several studies suggest a functional link between the power of occipital alpha oscillations and spatial attention: a decrease in alpha-power contralateral to the attended hemifield after a spatial cue and, in addition, an ipsilateral increase to the attended site of space have been found consistently. This phenomenon is henceforth referred to as alpha-lateralization. This distinct suppression in anticipation of the target is assumed to be a physiological correlate for the shift in directed visuospatial attention (Sauseng et al., 2005; Worden, Foxe, Wang, & Simpson, 2000). Jensen and co-workers (2010) suggest that alpha-band might subserve as a gating mechanism for inhibition, allowing for pulsed inhibition of gamma-band, which is related to the active processing of sensory information, e.g. in the visual cortex (Jensen, Kaiser, & Lachaux, 2007). Accordingly, higher amplitudes in the alpha-band lead to stronger inhibition pulses and thus altered alpha power is interpreted as the amount of inhibition that is exerted over a certain part of cortex.
A different line of research capturing the network perspective by means of functional magnetic resonance imaging (fMRI) pinpoints to a set of cortical sites: Frontal eye field (FEF), dorsolateral prefrontal cortex (DLPFC), intraparietal sulcus (IPS) and inferior parietal lobe (IPL) are together referred to as the FPAN (Corbetta, Kincade, Ollinger, McAvoy, &
Shulman, 2000; Hopfinger, Buonocore, & Mangun, 2000; Kincade, Abrams, Astafiev, Shulman, & Corbetta, 2005; for a review see Ptak, 2012). This research links the FPAN to the exertion of top-down attentional control by dynamically integrating bottom-up and top-down attentional processes (Corbetta & Shulman, 2002). Especially, the FEF and IPS are hypothesized to adopt a prominent role for the directing of attention as they exhibit activity during the attention-shifting after both a spatial and a feature-based cue (Egner et al., 2008). Therefore, the collaboration between those two sites is argued to be a key link for the capacity of top-down attention allocation (Corbetta, Patel, & Shulman, 2008). There are several open questions related to this network perspective as for example how these distributed areas achieve information exchange to form a functional unit. Spatially precise measures as fMRI fail to capture the putatively temporal precise code which might be used to orchestrate the dynamic interplay as it is argued to happen in the FPAN.
This need has been recognized and multiple studies integrated both fMRI and electrophysiological recordings to achieve a both spatially and temporally accurate picture of the dynamics during attention allocation: Combining these concurrent lines of research reveals that there is a vast overlap in cortical loci that are both identified as the FPAN and the regions, showing functional connectivity in the alpha band (Doesburg, Bedo, & Ward, 2016; Sadaghiani et al., 2012). This crucial integration thus allows to speculate that the alpha-band might act as a currency of information used in the FPAN and that especially alpha-oscillations are related to top-down attentional control (Doesburg et al., 2016; Jensen & Mazaheri, 2010). Importantly, this extends the function of oscillations as discussed above as a local inhibition mechanism in the occipital cortex to a network-wide approach, in which alpha-band carries functional information throughout the FPAN.
This idea is supported by recent evidence indicating a link between the hubs of the FPAN and the engagement in alpha-lateralization during attention-shifting. More specifically, the frontal eye field (FEF) and the intraparietal sulcus (IPS) were reported to be implicated in this process by showing that transient lesions by means of rTMS led to changed alpha modulation after cue presentation alongside with an altered attentional bias (Capotosto, Babiloni, Romani, & Corbetta, 2009; Marshall, O’Shea, Jensen, & Bergmann, 2015; Sauseng, Feldheim, Freunberger, & Hummel, 2011). Note that alpha modulation is a standardized measures that compares the amount of alpha-power in response to a spatial cue for one hemisphere.
The idea that oscillations and, in particular, the phase relationship might serve as a means for communication throughout large networks like the FPAN was recently suggested
by several researchers (Sauseng & Klimesch, 2008; Womelsdorf & Fries, 2007). Some studies did indeed find a relationship between phase coherence in the alpha-band and directed spatial attention around regions in the FPAN (Sadaghiani et al., 2012; Sauseng et al., 2005; Siegel, Donner, Oostenveld, Fries, & Engel, 2008). Additional confirmation for the concept that a synchronous phase relationship represents an advantageous computation was found as well applicable to cognitive capacities such as working memory (Sauseng et al., 2005).
Taken together, there is converging evidence from multiple imaging techniques and behavioral measures that alpha-phase coherence correlates to information transmission in the FPAN. The question, however, whether the phase of alpha oscillations carries functional value or is a mere side effect of other processes in the FPAN cannot be answered conclusively based on current research. In order to understand if there is a mechanistic relationship between alpha phase coherence and visuospatial top-down attention, it requires a technique to experimentally manipulate the phase coherence between different sites of the FPAN.
Novel stimulation techniques such as transcranial Alternating Current Stimulation (tACS) offer the opportunity to introduce causality in this discourse by specifically affecting phase coherence. tACS induces rhythmic changes in the excitability of cortical tissue, a dynamic comparable to an oscillation (Zaehle, Rach, & Herrmann, 2010). This technique was recently applied in a study on working memory aiming at inducing higher coherence in theta oscillations between two cortical sites. Indeed, the study could show that increased coherence resulted in enhanced memory performance (Polanía et al., 2012) and thus this research shows that it is suitable to tackle questions on coherence with tACS as in the present project.
This study will seek to contribute to the understanding whether alpha-band coherence might subserve communication in the FPAN. Furthermore, the experimental testing and the application in a directed attention task will add to the theoretical understanding of oscillation phases as a means of communication in neuronal networks (Sauseng & Klimesch, 2008; Womelsdorf & Fries, 2007). Specifically, we will try to understand if visuo-spatial attention performance can be improved by increasing phase coherence in the alpha-band between frontal and parietal sites. It will be assessed whether increased phase coherence in one hemisphere similarly affects visuospatial attention in the left, right or both hemifields as compared to a sham-stimulation. Also, this research tries to provide an answer if the coherence between the fronto-parietal stimulation sites will affect alpha-modulation in posterior areas during target anticipation as compared to sham-stimulation.
Enhanced phase synchrony in the alpha-band is hypothesized to improve performance in a spatial attention task and also to reinforce the alpha modulation observed in the respective
hemispheres as a neural correlate for attentional processing. The improvement will be assessed based on a sham-group, which does not receive any effective stimulation. Furthermore, the experimental setup will put emphasis on the distinction between mere stimulation effects and those that are specific to the phase synchrony by including a condition receiving the same stimulation setup without increased coherence, but instead decreased coherence. Thereby, effects of alpha-band stimulation and coherence manipulation can be disentangled.
Methods
Participants
Twenty-five right-handed participants (4 males) were examined for the present study. All of them had normal or corrected-to-normal vision. None reported any neurological or psychological disorders, took medication regularly or during the time the experiment was conducted. Two participant did not complete all sessions and their data was excluded from consecutive analysis, another participant had to be excluded due to excessive muscle artifacts during the electroencephalogram (EEG)-recording and two more because of lacking fixation to the center of the screen during the task. The final sample of twenty participants had an average age of 20.45 years (SD = 3.58).
After giving informed consent (see Appendix A), all participants were tested over three sessions and presented with one of three stimulation conditions (in-phase, out-of-phase, sham) on each session. The order was allocated randomly to each participant. After the third session, they were compensated with either research credits or money. Ethical approval has been granted by the ethics committee of the University of Amsterdam.
Behavioral Task & Materials
Spatial attention task. The computer-based spatial attention task was organized in
eight blocks, each consisting of 100 trials per block. The task was a cued attention paradigm (Posner, 1980) containing three types of possible trials: Valid, invalid and neutral trials. A trial started off with the presentation of a fixation cross (see figure 1). After a short delay a cue appeared for 250 ms indicating to the participant to direct attention either to the right or the left half of the screen (valid & invalid cues) or to spread attention equally over both halves of the screen in case of a neutral cue. Shortly after that (900 – 1500 ms), a target and a distractor were shown (67 ms) in the lower left and right visual field (Euclidian distance to fixation cross: 5.76 visual angle). The target and distractor were identical Gabor patches only differing in their orientation: A target pattern would either be horizontal or vertical, whereas all distractor patterns were diagonally tilted. The contrast for the target and distractor were
titrated to the participant’s individual performance level through an adaptive threshold procedure. Shortly after target/distractor presentation, two masks (checkerboards of the same size and spatial frequency as the targets and distractors) appeared in the same locations. The participant’s task was to determine whether the target was horizontal or vertical in a respective trial and to press the according buttons with both hands simultaneously. In case of a horizontal pattern, the lower buttons needed to pressed with both index fingers and for vertical patterns the upper buttons with both middle fingers. Participants were instructed to maintain fixation throughout the entire trial and to answer as fast and accurately as possible. After each block, participants received feedback on their accuracy and average reaction time during that block.
Visual analogue scale (VAS). The paper-based questionnaire was presented to the participant
at the start and end of each session. This questionnaire contains 16 word pairs assessing the participants’ mood by asking them to indicate their current state on a line ranging between
Figure 1. Experimental paradigm. Following a variable intertrial interval a visual cue was presented in the location of the fixation cross instructing the participant where to covertly direct attention. Left, neutral and right cues were presented randomly and would give information about the lighter shaded circles in the lower part of the screen. After a short or long interval during which fixation must be maintained, two stimuli appeared in those lower circles for a brief duration. Subject had to identify the orientation of the target Gabor patch (horizontal/ vertical) while ignoring the diagonal ones and respond accordingly with a button press.
two extreme adjectives (e.g. alert vs. drowsy, see Appendix B).
Four items yielded in a significant difference between the questionnaire filled in before the session and after the session: Participants of all stimulation conditions felt more feeble (F (1, 22) = 8.434, p = .008), more dreamy (F (1, 22) = 8.110, p = .009), more bored (F (1,22) = 5.808, p = .025) and tended to be more withdrawn (F (1, 22) = 3.271, p = .084) after the sessions. Furthermore, there was a significant difference of the level of happiness/sadness between the stimulation conditions (F (2, 44) = 5.252, p = .009): The out-of-phase-condition was the happiest (M = 2.51, SE = 0.26), followed by the sham-condition (M = 2.76, SE = 0.32) and the in-phase-condition (M = 3.23, SE = 0.31). Note that lower values signify happier judgements and accordingly higher values sadder judgments with a range form 0 – 10.6. Consequently, all stimulation groups were even though they showed significant differences still much closer to positive extreme of the scale.
tACS-Stimulation
Stimulation was applied through a DC-Stimulator Plus (NeuroConn) connected to two or three rubber patch electrodes for the respective experimental conditions. Two of those had a surface area of 3.2 x 3.2 cm2 (target electrodes) and a third of 5 x 7 cm2 (return electrode). The target electrodes were placed over F4 and P4 and in case of sham and in-phase stimulation, the return electrode was placed over Cz in the international 10 – 20 system (see figure 2A). F4 was chosen with respect to its proximity to the FEF while still avoiding contact to the bigger return electrode at Cz and similarly, P4 was chosen to approximate the location of the IPS. Please note that due to the nature of sinusoidal stimulation for an out-of-phase stimulation, the return electrode over Cz was not required but still was placed during preparation. A sinusoidal stimulation at 10 Hz with no DC-offset and a current of 1000/ 2000 µA anodal-to-cathodal-peak (out-of-phase-stimulation/ in-phase-stimulation, with the respective amount of electrodes) and a current density of 0.097 mA/cm2 per target electrode was administered throughout the four stimulation blocks of the main task (see figure 2B).
The in-phase condition targeted F4 and P4 via 10-Hz sinusoidal stimulation and set those areas in 0° relative phase difference to each other (see figure 2B). The return electrode was placed equally over both motor cortices centered around Cz, which is a setup shown to not affect the excitability of the primary motor cortices (Feurra et al., 2011). The out-of-phase stimulation yielded in a 180° relative phase difference between F4 and P4 (see figure 2B). The sham stimulation followed the setup of an in-phase-stimulation but entailed only 20 seconds of 10-Hz-stimulation (100 cycles of fade–in and –out, respectively) at the start of each block. The stimulation blocks contained 3000 cycles added to 100 cycles for fade-in and
–out and thus the stimulation duration in these blocks was 320 seconds and administered while participants performed a block of the main task.
Both stimulation conditions (in- & out-of-phase) had an intermixed order of real and sham stimulation (see figure 2B). In those conditions, the second, fourth, fifth and seventh block would contain stimulation and the remaining blocks would be a sham-stimulation. For a sham-condition, all eight blocks would consist of sham stimulation.
At the end of each session, participants were asked to report the perceived side effects during stimulation on a scale from 1 (not present) to 5 (extremely noticeable, see Appendix C). The assessed items were headache, neck pain, nauseousness, muscle contractions in face or neck, tingling/itching sensation under the electrodes, burning sensation under the electrodes, sleepiness, mood changes and uncomfortable feeling (unspecific). Only two items showed a significant difference between stimulation sessions, as was assessed by means of a
Figure 2. Overview of stimulation setup. (A) Three rubber patch electrodes were centered around F4, P4 and Cz with the first two featuring a smaller surface and thus resulting in a higher current density compared to the return electrode at Cz. In case of out-of-phase stimulation the return electrode was not used but still placed in the usual position. The current was adjusted so that both in-phase and out-of-phase stimulations entailed the same current density at the target electrode. The left and right regions of interest (ROI) refer to the electrodes used for the analysis of the alpha modulation index. (B) Comparison of stimulation properties for the respective conditions at F4 and P4. A flat blue lines signifies a sham-block. The curved blue lines show alternating current and thus stimulation blocks. The difference between in- and out-phase stimulation in the phase difference and thus the relationship between the lines depicted at F4 and P4. For all sessions all eight block of the main tasks are shown and note that sham blocks were used during later EEG-analysis.
non-parametric Friedman-Test for repeated measures: First, headache was reported most for in-phase-stimulation (Mdn = 2.00), followed by out-of-stimulation (Mdn = 1.00) and sham-stimulation (Mdn = 1.00) and this difference was meaningful (χ2(23) = 7.396, p = .025). Second, burning sensation under the electrode occurred differently for the stimulation session (χ2(23) = 15.318, p < .001). This side-effect was equally prominent during in-phase-stimulation (Mdn = 2.00) and sham-stimulation (Mdn = 2.00), and less pronounced during out-of-phase stimulation (Mdn = 1.00).
EEG data collection
EEG data was collected with a BioSemi ActiveTwo 64 Ag–AgCl channel setup (BioSemi, Amsterdam, The Netherlands) at 512 Hz from all accessible channel during the main task. The setup enabled 64 electrodes placed according to the international 10 – 20 system, however, channels at F4, F6, P4, P6, Cz, CPz were omitted in most of the participants due to the stimulation electrodes. Additional electrodes were placed as horizontal and vertical electrooculogram (h/vEOG), which were used to detect eye movements and blinks, as well as on the earlobes, which were later used as a reference. Electrode impedances were kept below 30 kΩ.
Experimental procedure
Each session was identical in terms of procedure: Firstly, participants were welcomed in the lab and excluding criteria (see appendix A) were checked. If none were applicable, the informed consent (see Appendix A) was signed and the setup started with a trial stimulation. Thereafter, if the participant decided to continue with the experiment, the EEG electrodes were prepared.
During the task, participants were seated in a comfortable armchair and asked to position both arms on the armrests and place their index and middle fingers on the integrated buttons in the armrests. The subjects sat in a distance of approximately 90 cm to the screen and the target size was adjusted based on screen distance.
During a first practice phase with only valid trials (100 trials divided into three blocks), the subjects were acquainted to the task by receiving auditory feedback after every trial, as well as an overview of their average performance after every block. It was ensured that they would at least achieve a performance of 70 % accuracy on the last practice block before they would continue. This served to allow for a stable performance during the contrast titration blocks. The contrast for the target presentation was titrated to the participant’s performance level: An adaptive staircase procedure adjusted the contrast intensity to converge to an intensity yielding at a 70 % accuracy on valid trials (see Marshall et al., 2015 for a
similar approach). Note that no trial-wise auditory feedback was given during this phase. For left and right stimuli the same contrast was used and the thresholding blocks consisted in total of 200 valid trials (divided into two blocks). The average presented contrast did not differ between the stimulation sessions (F (2, 40) = 0.896, p = .416, η2 = 0.43; sham: M = 0.15, SE = 0.03; in-phase: M = 0.20, SE = 0.04; out-of-phase: M = 0.21, SE = 0.04).
After the estimation of the optimal threshold, a second practice phase (80 trials divided into three blocks) followed again with a trial-wise auditory feedback. This practice occasion introduced neutral and invalid cues and the ratio of trial types (valid, invalid, neutral) was identical to the main task.
Upon completion of those three pre-tasks, the eight blocks of the actual experiment started with a 15-minute break after half of the blocks. Each block had a length of about 5 min and stimulation was started by the experimenter prior to the start of each block. A quarter of all presented trials contained neutral cues. The remaining trials were in 25 percent invalid (25 % neutral cues, 56.23 % valid cues, 18.77 % invalid cues of all trials). Throughout the main blocks, eye tracking data were collected to control for a successful fixation during the trials. The participants were supervised for another 60 min after the last administration of stimulation and asked to fill out a questionnaire on the side effects of the stimulation and the VAS.
Data Analysis
Behavioral data. Behavioral data was acquired with Psychophysics Toolbox
extensions (Brainard, 1997) implemented in MATLAB 2012b (Mathworks, Natic, MA) and analyzed in SPSS (Version 23, IBM Corp.). Only trials during which fixation was maintained were submitted to data analysis. A trial was discarded, if it contained a sample with more than 1.5 visual angle deviation from the center of a cluster of all trials in the respective block. This way it was possible to counteract the measurement noise of the eyetracker. Systematic deviation induced by the cue was identified by comparing the average horizontal deviation of left- and rightward cues for all trials in a session. Subjects were excluded if the mean difference between the two types of cues exceeded 0.5 visual angle in an entire session (see participants). On average, 789.30 trials (SE = 2.89) of all 800 trials were submitted to the analysis per session.
The structure of the remaining repeated-measures data required a linear mixed model approach of the following form:
where Yi represents a vector of the dependent variable for the ith participant, Xi is a covariate matrix for the respective predictors p for the ith participant, β is a vector of fixed-effect parameter for each predictor in Xi for the ith participant , Zi is a covariate matrix of q random effect predictors for the ith participant, γ represents a random-effect parameter for each predictor in Zi, and εi represents a vector of residuals, that is, the model fit error, which signifies the difference between the model prediction for each observation from the ith participant and the actual value of that observation.
The present design foresaw four categorical predictors stimulation condition (in-phase-, out-of-phase, sham), stimulation mode (on-, offline), attention cue (valid, neutral, invalid) and target side (left, right) and consequently, interactions of those. The dependent variables accuracy, reaction time (RT) in correct trials and inverse efficiency, that is, RT divided by accuracy, were assessed in separate models with the same fixed predictors. As the reported side-effects of ‘headache’ and ‘burning sensation under the electrode’ showed a significant difference between conditions (see tACS-stimulation), they were included in the models as covariates that could differ between subjects and sessions. In all models, each subject was treated as a random variable. Planned paired-sample t-tests were performed for post-hoc analyses and p-values are reported two-tailed and corrected with Bonferroni’s method.
The first blocks of each of the three session was analyzed to exclude potential differences before manipulation by means of a repeated-measures ANOVA with the factors stimulation condition (in-phase, out-of-phase, sham), attention cue (valid, neutral, invalid), and target side (left, right) and accordingly data for the main analysis (described before) only stem from block 2 – 8. If sphericity was violated, the reported p- values were Greenhouse-Geisser corrected, degrees of freedom are reported for assumed sphericity for the sake of readability. All post-hoc tests were two-tailed and reported Bonferroni-corrected.
EEG data. The preprocessing was performed using Fieldtrip (Oostenveld, Fries,
Maris, & Schoffelen, 2010) implemented in MATLAB 2015b (Mathworks, Natic, MA). First, data was high-pass filtered at 0.5 Hz and cut into epochs from 1.5 sec before cue and 2.5 sec after target presentation. All data unaffected by stimulation was inspected for muscle and non-physiological artifacts and respective trials were rejected manually. The remaining trials were submitted to an Independent Component Analysis and components containing eye blinks and/or other artifacts that clearly did not stem from physiological signal (frequency composition, event-related potentials, individual trials of each component were carefully
inspected for this purpose) were subtracted from the data. Of note, only trials with successful fixation and from block three, six and eight were used for the analysis.
The time-frequency representation of the preprocessed trials was obtained by multiplying the frequency representation with Hanning windows covering 4 cycles for each frequency of interest and a sliding step of 0.01 sec for the inspected time window of 0.8 to 0.1 sec before target. The inspected frequency range was the individual alpha frequency (IAF) ± 2 Hz. The IAF was assessed by means of a Fast-Fourier-transformation of all three concatenated sessions and thereby we obtained one value per subject over all three sessions, which was used for all later analyses.
Alpha modulation index. To evaluate the effects of the stimulation on preparatory
activity during the cue-target-interval, we calculated the alpha modulation index (AMI), which is a normalized measure capturing the relationship between the attended hemifield and oscillatory activity from parieto-occipital sites in each hemisphere (Marshall et al., 2015; Thut, Nietzel, Brandt, & Pascual-Leone, 2006). The AMI left cue – right cue is captured in the following formula:
𝐴𝑀𝐼$%&' )*%+,-./' )*% = α 2*% $%&'− α 2*% ,-./' α 2*% $%&'+ α 2*% ,-./'
where α is the averaged alpha power of the respective trials with a right or left Cue. Positive values of the AMI left cue – right cue signify that the respective hemisphere features more alpha power on trials with left cues than on those with right cues and the opposite, that is, more alpha-power in response to right cues compared to left cues for negative values. A well established finding is an increase ipsilateral and a decrease of alpha power contralateral to the attended hemifield (Sauseng et al., 2005). Accordingly, we expect negative values for the AMI left cue – right cue in the right hemisphere and positive values for the left hemisphere as a function of successful attentional modulation in response to the cues. For statistical assessment, AMI left cue – right cue – values were submitted to a three-way repeated-measures ANOVA with the factors of stimulation condition (in-phase, out-of-phase, sham), hemisphere (left, right) and attentional cue (valid, neutral, invalid).
Furthermore, we were interested in the development of the alpha modulation index as a function of time. Therefore, we assessed the AMI left cue – right cue in four bins of 200 ms after cue onset and in separated time bins just before target presentation for both long (1.5 sec) and short trials (0.9 sec). The averaged values of each of these bins were passed to a four-way repeated-measures ANOVA examining for the factors of stimulation condition (in-phase, out-of-phase, sham), hemisphere (left, right), attentional cue (valid, neutral, invalid) and time bin (1 – 200 ms, 201 – 400 ms, 401 – 600 ms, 601 – 800 ms).
Phase-locking values. The manipulation intended to alter the fronto-parietal interplay
by means of enhanced or degraded phase synchrony. By consequence, it is of high interest to examine if the phase synchrony was indeed altered between stimulation conditions. This was assessed through Phase-Locking values between several frontal sites and the two occipito-parietal areas. Every frontal electrode (F1, F2, F7, F8) was tested for the phase-synchrony with both an ipsi – and a contralateral site (PO3, PO4, see figure 7 for an overview). Phase-locking values are a measure for phase synchrony between two signals over time and such a synchrony is interpreted as an indication for functional connectivity. The computation entails:
𝑃𝐿𝑉8,:,' =
1
𝑁 𝑒- >? ' + >@ (')
C
where the difference in phase values (φj – φk) from two signal (j, k) at a point in time t are relativized over N epochs (e.g. trials). Accordingly, values close to 1 signify a high similarity in phase synchrony between the inspected signals and values close to 0 a high dissimilarity (Lachaux, Rodriguez, Martinerie, & Varela, 1999). We expect that in-phase stimulation will increase the PL-values for fronto-parieto-occipital connections in the right hemisphere and that in turn out-of-phase stimulation decreases PL-values between those channels as compared to sham stimulation, which serves here as a baseline. The time of interest will be again -0.8 sec to -0.1 sec prior to target presentation and the analyzed frequencies are the IAF ± 2 Hz. The obtained PL-values for each channel pair (eight pairs in total) will be assessed with an ANOVA entailing the factors frontal electrode (F1, F7, F2, F8), parieto-occipital electrode (PO3, PO4), attentional cue (left, neutral, right) and stimulation condition (in-phase, out-of-phase, sham).
Results Behavioral data
Main analyses. The linear mixed models with the predictors stimulation condition
(in-phase, out-of-(in-phase, sham), stimulation mode (on-, offline), attention cue (valid, neutral, invalid) and target side (left, right) showed a similar pattern for all three dependent variables (see figure 3):
Accuracy. The model for accuracy showed a main effect for the attention cue
(F (2, 688) = 58.98, p < .001) with valid cues yielding the highest accuracy (Mean (M) = 0.80, Standard Error (SE) = 0.01), followed by neutral cues (M = 0.77, SE = 0.01) and invalid cues (M = 0.67, SE = 0.01). Pairwise post-hoc comparisons revealed that all levels were significantly different from each other (all p < .036). Next, the predictor target side showed a significant difference (F (1, 688) = 46.88, p < .001) in so far that subjects were more accurate
for stimuli cued on the right (M = 0.78, SE = 0.01) compared to those cued on the left side (M = 0.71, SE = 0.01). Lastly, there was a trending main effect for stimulation condition (F (2, 688) = 2.75, p = .065), which was moderated by an enhanced accuracy in the out-of-phase condition (M = 0.76, SE = 0.01) as compared to sham-stimulation (M = 0.72, SE = 0.01, p = .005). The in-phase session was similar to both sham- and out-of-phase and in between those in terms of accuracy (M = 0.74, SD = 0.01, all p > .298). The covariate Headache explained a significant amount of variance in the data (F (1, 688) = 9.38, p = .002) and a pearson-correlation confirmed that reported headache was negatively correlated to accuracy r(718) = -.085, p = .023. Of note, the factor of stimulation mode did not result in a significant difference (F (1, 688) = 1.77, p = .184) and hence measures of accuracy are comparable between blocks of on- and offline stimulation.
Reaction times. Similarly to accuracy, the linear mixed model revealed a significant
main effect for both the attention cue (F (2, 688) = 36.05, p < .001) and target side (F (1, 688) = 20.30, p < .001). Participants were fastest for valid cues (M = 646.50 ms, SE = 7.18), ensued by neutral (M = 699.96 ms, SE = 7.18) and slowest in case of invalid cues (M = 731.01 ms, SE = 7.18) with all kind of cues being highly significantly different from each other (all p > .007). Reactions times were faster for targets on the right side (M = 674.08 ms, SE = 5.87) in comparison with those on the left side (M = 710.89 ms, SE = 5.87). The covariate Burning Sensation contributed significantly to the explanation of the present data, F (1, 688) = 15.13, p < .001 and there was a positive correlation between the amount of reported burning sensation and the reaction time, so that the more reported burning under the electrode went along with longer reaction times, r (718) = .14, p < .001. Furthermore, there was neither an effect of stimulation condition (F (2, 688) = 0.94, p = .391),
Figure 3. Similar pattern of effects for all dependent variables: All bar graphs show a main effect of attention cue (valid, neutral, invalid, shown on the respective y-axis) and a main effect of target side (blue and red bars). Note that whereas for accuracy (A) a high performance is expressed through high values, in both reaction time (B) and inversed accuracy (C) a good performance translated to low values.
nor of stimulation mode (F (1, 688) = 0.55, p = .461).
Inverse Efficiency. The third dependent variable attempts to assess the speed-accuracy
trade-off more systematically by using a quotient of reaction time and accuracy. Lower values signify higher efficiency and vice versa. In a similar vein as in the case of accuracy and reaction times, we see a clear pattern in terms of attention cue (F (2, 688) = 78.91, p < .001) and target side (F (1, 688) = 36.77, p < .001). Valid cues were associated with the most efficient behavior (M = 825.61, SE = 19.71), neutral cues were less efficient (M = 946.18, SE = 19.71) and the least efficient trade-off between reaction and accuracy was observed for invalid cued trials (M = 1169.51, SE = 19.71). Overall, right target trials were reacted to more efficiently (M = 912.45, SE = 16.10) than to left target trials (M = 1048.42, SE = 16.10). The covariate Headache contributed to the explanation of the data (F (1, 688) = 5.318, p = .021) but there was no underlying linear relationship between the values of these variables (r (718) = .06, p = .100). Furthermore, there was neither effect of stimulation condition (F (1, 688) = 0.21, p = .479) nor of stimulation mode (F (1, 688) = 1.83, p = .643).
Control analyses. Data of the first blocks (sham-blocks in all stimulation sessions)
were submitted to a repeated-measures ANOVAs for all three dependent variables. Note that the linear model approach was not granted anymore since the side-effects of stimulation do not need to be included as a covariate for this analysis.
Accuracy. During the first block, there was no effect of stimulation session
(F (2, 38) = 1.31, p = .283, ηp2 = .06). However, there was a main effect for side of the target
(F (1, 19) = 5.70, p = .027, ηp2 = .23) with right targets evoking more accurate behavior
(M = 0.80, SE = 0.02) compared to left-sided targets (M = 0.74, SE = 0.02). Also, the attention cue (F (2, 38) = 34.71, p < .001, ηp2 = .65) yielded in systematic differences: Invalid cued
trials were associated with the lowest accuracy (M = 0.66, SE = 0.03) and significantly different from both neutrally cued (M = 0.82, SE = 0.02) and valid cued trials (M = 0.84, SE = 0.01, both p < .001). Only valid and neutral cue trials did not differ significantly (p = .778).
Reaction times. The analysis did not show an effect of stimulation session
(F (2, 38) = 0.21, p = .813, ηp2 = .01). There was a main effect of attention cue
(F (2, 38) = 21.36, p < .001, ηp2 = .53) with valid cued trials resulting in the fastest responses
(M = 663.65, SE = 16.10), followed by neutral cued trial (M = 723.99, SE = 20.91) and invalid cued trials (M = 762.92, SE = 26.72). Valid cues were significantly different from both other cues (all p < .001), but neutral and invalid cues did only feature a trending difference (p = .090). Moreover, the analysis revealed an advantage of right targets (M = 699.34,
SE = 20.18) as compared to left targets (M = 734.36, SE = 21.93, F (2, 38) = 5.88, p = .025, ηp2 = .24). Alongside with those main effects, an interaction between attention cue and target
side was significant (F (2, 38) = 4.026, p = .026, ηp2 = .17). Separate post-hoc tests for right-
and left target trials indicate that whereas for the right targets significant differences can be found between all types of attention cues (all p < .037), for the left targets only the valid trials were significantly faster compared to both other cued trials (both p < .001).
Inverse efficiency. Also, the third dependent variable did not show an effect of
stimulation session (F (2, 38) = 0.15, p = .811, ηp2 = .01). The attention cue was associated
Figure 4. Time-frequency-representations (TFR) and topographies of the alpha-modulation indices (AMI). For all stimulation conditions the AMI for both hemispheres and the topography is shown for the time of interest from -0.8 to -0.1 sec before target presentation. The left hemisphere (A) shows the parieto-occipital ROI consisting of PO3, P3, P7 and the right hemisphere (C) is a pooled representation of PO4, P4 and P8. The topographies (B) depict only the alpha range (8 – 13 Hz). Note that the same color scale is used in all plots. All plots support that in all stimulation conditions alpha-power was indeed modulated according to the cue. More extreme values show a stronger anticipatory alpha modulation. Whereas the pattern of alpha modulation looks similar for the left hemisphere (A) in all stimulation conditions, anticipatory alpha appears to modulate to a lesser amount in the right hemisphere (C) for the sham-stimulation as compared to both other conditions.
with a different degree of efficiency (F (2, 38) = 23.32, p < .001, ηp2 = .55) with the valid
cued trials showing the lowest inverse efficiency (M = 812.49, SE = 19.51), followed by the neutral cues (M = 931.38, SE = 46.56) and then invalid cues with the highest inverse efficiency (M = 1363.84, SE = 110.39) all being significantly different from each other (all p < .021). The target side (F (1, 19) = 2.124, p = .161, ηp2 = .101) did not show a statistically
significant pattern.
Thus, the control analyses exclude any prior difference in the behavior of the stimulation conditions. The effect of attentional cues is already present in this early stage and indicates that the participants followed the task requirements and the task featured the capacity to produce an attentional trade-off as usually found in Posner-task (Posner, 1980).
EEG data
Alpha modulation index. Prior to statistical analysis, the topography and
time-frequency representation for the later analyzed time window (-0.8 - -0.1 s to target presentation) and regions of interests were assessed for the group average and can be inspected in figure 4.
Pretarget interval (-0.8 sec - -0.1 sec). The statistical analysis of the alpha modulation
indices for both hemisphere and all stimulation conditions attempts to assess the possible
Figure 5. Interaction of stimulation condition and hemisphere. Bar graphs for the respective hemispheres (blue and red bars) over the different stimulation condition (x-axis) showing the AMI on the y-axis. The brackets refer to the post-hoc comparison and ** labels refer to a significance of p < .001 and * to p < .05. The modulation is more distinct for both in-phase- and out-of-phase-stimulation.
physiological effects of the different kind of stimulation during the preparation for a target stimulus. Firstly, there was no significant overall difference between any of the stimulation conditions (F (2, 38) = 0.82, p = .445, ηp2 = .04). As expected, we saw a differential reaction
of both hemispheres to the respective cues, resulting in a main effect of hemisphere (F (1, 19) = 14.36, p < .001, ηp2 = .43). The left hemisphere showed on average a positive
AMI (M = 0.28, SE = 0.01), which translates to a stronger increase in alpha-power in response to left cues as compared to right cues. In turn, the right hemisphere featured rather negative values (M = -0.32, SE = 0.01) indicating that in the right hemisphere there was a stronger increase in response to right cued stimuli as compared to left cues. Interestingly, there was an interaction between the effect of stimulation condition and the hemisphere (F (2, 38) = 3.71, p = .034, ηp2 = .16, see figure 5). Separate post-hoc analyses for all stimulation conditions
showed that whereas for in- and out-of-phase stimulation left and right hemisphere differed highly significantly from each other (both p = .001), this difference is less pronounced, but still statistically significant, in the sham-condition (p = .023).
Post-cue development. In order to disentangle the interplay in anticipatory oscillatory
activity between the different stimulation conditions, we assessed the development of the Figure 6. Alpha modulation develops similarly in all stimulation conditions. For each stimulation condition both AMI (y-axis) of the left (dashed line) and right hemisphere (bold line) are plotted over time. The boxes indicate the time bins used during statistical analysis. The time bin significantly interacts with the difference between the hemisphere, indicating a development of modulation. The significance indication refers to the post-hoc test of this interaction. T stands for a p-value < .1 and ** for p < .001. This plot shows that respective modulation develops in the first 400 ms and is fully pronounced as time progresses. Note that here the data is shown cue-locked and that the first possible cues could arrive after 0.9 sec.
modulation indices after cue presentation: The three-way repeated-measure ANOVA of the factors time bin (0 – 200 ms, 201 – 400 ms, 401 – 600 ms, 601 – 800 ms), stimulation condition (in-phase, out-of-phase, sham) and hemisphere (left vs. right) showed a main effect for hemisphere (F (1, 19) = 17.986, p < .001, ηp2 = .49). The right hemisphere showed
systematically more negative modulation (M = -0.35, SE = 0.02) compared to the left hemisphere (M = -0.05, SE = 0.02). Furthermore, the difference between hemispheres changed over time as was expressed in an interaction between hemisphere and time bin, F (1, 19) = 4.173, p = .025, ηp2 = .18, see figure 6. Post-hoc analyses per time bin revealed
that whereas both hemispheres are only trending to be different in the first two time bins (p1st bin = .081, p 2nd bin = .070), the modulation of them becomes significantly distinct in the
last two time bins (both p’s < .001).
Furthermore, as the design of the present study entailed two trial lengths, if indeed there is a temporal development in alpha-modulation that exceeds the short trial length, it should be expressed in a significant difference in the amount of modulation between those two types. To this end, we compared the time window, -0.3 sec to -0.1 sec before target presentation between those two trials lengths. A three-way ANOVA with the factors trial length, stimulation condition and hemisphere showed that there was no significant amount of change between the short and the long trials, F (1, 19) = 0.002, p = .967, ηp2 = .01. In line
with our prior analyses, there was no effect of stimulation condition (F (1, 19) = 0.89, p = .421, ηp2 = .05) and a significant difference in modulation between hemispheres
(F (1, 19) = 8.74, p = .008, ηp2 = .32).
Phase-locking values. In terms of functional connectivity, the main analysis with the
factors frontal electrode (F1, F7, F2, F8), parieto-occipital electrode (PO3, PO4), attentional cue (left, neutral, right) and stimulation condition (in-phase, out-of-phase, sham) revealed firstly a main effect of frontal electrode (F (3, 57) = 72.59, p < .001, ηp2 = .79): All electrodes
were significantly different from each other regarding their phase-locking (all p < .019) with F2 showing the strongest synchronization to parieto-occipital regions (M = .39, SE = 0.02), followed by F1 (M = .33, SE = 0.02), F8 (M = .29, SE = 0.02) and finally F7 (M = .21, SE = 0.01), see figure 7. Secondly, both parieto-occipital sites engaged differently synchronized with the frontal sites (F (1, 19) = 37.52, p < .001, ηp2 = .66): PO4 in the right
hemisphere was more phase-locked to the frontal sites (M = .38, SE = 0.02) than PO3 in the left hemisphere (M = .23, SE = 0.01). There was neither a main effect of attentional cue (F (2, 38) = 0.63, p = .540, ηp2 = .03), nor of stimulation condition (F (2, 38) = 0.004,
synchronized to a different extent with both parieto-occipital regions depending on the attentional cue resulting in a significant three-way interaction (F (6, 114) = 2.41, p = .032, ηp2
= .11). Post-hoc analyses revealed that F1 synchronized significantly more with PO3 for right cues (M = .27, SE = 0.02) than for left cues (p = .040, M = .26, SE = 0.02). Similarly, F2 had more phase-locking with PO3 after right cues (M = .29, SE = 0.02) as compared to left cues (p = .007, M = .28, SE = 0.02), neutral cues evoked more synchronization in a trending dimension (p = .067, M = .29, SE = 0.02) in comparison to left cues.
Figure 7. Phase-locking-values between frontal and parieto-occipital channels. Each frontal channel’s (dark blue circles) phase-locking with the parieto-occipital channel (lighter blue circles) is shown. The grey squares are to show the position of the stimulation electrodes in respect to the assessed phase-locking values. In terms of fronto-parietal connection, the strongest functional connectivity can be observed between F2 and PO4. Note that this is the averaged connectivity data of all stimulation conditions, since there was no meaningful difference between them. The yellow-black dots shows the EEG-electrode setup.
Discussion
A fronto-parietal alpha-band stimulation attempted to alter the phase relationship between the frontal and parietal stimulation site. The in-phase stimulation was intended to improve performance as compared to a sham-stimulation. This was not confirmed by our observations in the current study: Besides a trending difference in terms of accuracy between out-of-phase and sham stimulation, there was no effect of the applied stimulations on the behavioral performance. In a similar vein, there were no spatially selective effects of different stimulation settings. Instead, the data showed an overall rightward bias in all assessed behavioral measures for all stimulation groups. The examined neural correlates to spatial attention shifting, the alpha modulation index, did however feature a reaction to the stimulation manipulation: In addition to an overall stronger modulation in the right hemisphere compared to the left equivalent for all stimulation conditions, an interaction between hemispheric modulation and stimulation condition became apparent. Both in-phase and out-of-phase stimulation showed stronger alpha modulation in the right hemisphere as compared to sham stimulation. This is partly in line with the hypothesis which argued that the stimulation would have an effect on the anticipatory alpha modulation. The findings of this neural measure suggest that the modulation in the right hemisphere was responsive to the alpha band stimulation and independent of the phase synchrony between frontal and parietal sites. To assess in how far the distinctly synchronized stimulation techniques altered physiological coherence as an after-effect of stimulation, we analyzed phase-locking values between sites in the proximity of our frontal and parietal stimulation electrodes. To complement the behavioral findings, there was no indication of altered fronto-parietal phase coherence between the different stimulation conditions.
Effect of stimulation on alpha modulation indices. The present pattern of results
affords several observations: If the effects observed for the alpha modulation index in the right hemisphere are related to the physiological mechanism associated to top-down attention allocation, it is surprising that this effect occurs in isolation. For instance, there was no concurrent effect in behavior. As a consequence, either this modulation did not yet reach behavioral significance or the stimulated frequency of 10 Hz was not the suitable frequency to cause a functional difference. A third possibility would be that the power of oscillations has no functional significance, rendering alpha modulation and lateralization to a mere side-product of some other physiological mechanism enabling spatial attention allocation. In the light of former research attempting to probe the causal relationship between preparatory alpha
and visual perception (Helfrich et al., 2014; Rihs, Michel, & Thut, 2007), it does not seem appropriate to discard alpha modulation as an epiphenomenon based on our results.
With regard to the relevance of IAF, the current state of research is conflicting. On the one hand, there is a study reporting that parieto-occipital regions can be engaged in more alpha-power by means of tACS in the IAF at the occipital cortex (Zaehle et al., 2010). Another study showed successful entrainment of alpha-band with a 10 Hz-tACS stimulation in the occipital cortex (Helfrich et al., 2014). Interestingly, they report that the stimulation causes a shift of the IAF towards the stimulated frequency and that the alpha power increase resulted from a synchronization with the stimulation frequency. Helfrich and co-workers (2014) also observed that the efficacy of the stimulation was not predicted by the IAF. An important difference between those two studies is that the first reported off-stimulation entrainment, while the latter assessed online EEG-measurements. In the current study, we were only able to assess offline EEG-data. However, we did assess the behavioral data for potential difference of the stimulation mode (on-/ offline stimulation) and there was no indication for significant changes in terms of behavior. Future research needs to assess the role of IAF for posterior alpha-band stimulation and further shed light on the possibly interesting dynamics between induced and endogenous alpha band and the longevity of those effects.
tACS stimulation is a novel tool and, as such, information on its effectiveness is still lacking. Similar to the present experiment, a recent study also deploying a 10 Hz-stimulation on the P6 electrode did also find only trending behavioral differences in a similar attentional task (Hopfinger, Parsons, & Fröhlich, 2016). This indicates that the present data do not stand out in terms of size of effects on behavior and gives an impression of behavioral outcomes of alpha-band-stimulation on this regions of cortical tissue. Note that in the present study, for most participants the location of P6 was covered as well with the stimulation electrode and given the conductivity of the cortical tissue, a comparison can be made for this reason.
A surprising feature of the observed interaction between hemisphere and stimulation condition is that alpha power seems to be altered only for trials with certain cues (left or right cues). As the modulation index is robust to changes in the overall amount of alpha-power, the here observed pattern for the stimulation groups could have been either created by a stronger engagement in alpha power for right cues or for a more pronounced disengagement for left cued trials. While right cues would require a boost in alpha-power, left cues afford to suppress alpha-band in the right hemisphere. Here it is important to disentangle the notion of oscillations as a local and a network feature. Local alpha effects have been reported in a wide
range of human and animal research to act as a local inhibitor in the sensorimotor cortices (Haegens, Nácher, Luna, Romo, & Jensen, 2011). With regard to the interregional information transmission it is, however, less clear, if this locally observed circuit in dynamics between gamma- and alpha-band, has more remote effects: As we induced more alpha-power in both the frontal and parietal site, we might have promoted top-down inhibition in excitability in the right occipital cortices, thus preferably supporting processing in left cues. A conclusion on how alpha-power acts in a network and thus in a more subtle interplay of excitation and inhibition of areas cannot be reached on the basis of current research.
The similarity of in-phase and out-of-phase stimulation regarding their effect on the modulation–index combined with the unaffected functional connectivity between frontal and parietal sites suggests the setup of stimulation did not introduce the desired change in phase synchrony. Considering the vicinity of the parietal stimulation electrode and the right region of interest, an influence on the local alpha power as a result of the stimulation appears likely though. Siegel and colleagues (2008) reported a particular strength in attentional modulations observed in alpha band for the IPS, which was absent in assessed frontal regions as the FEF. Unfortunately, the current scenario does not allow to disentangle between contributions of the frontal and parietal stimulation electrode on the effects in alpha-modulation.
Instead of comparing right cues to left cues, as was done in the current analysis, another common way to look at the alpha modulation is to compare ipsilateral versus contralateral cues respective to the evaluated hemisphere. The current approach allowed a more intuitive understanding of the pattern of results, especially with regard to the topographies and time-frequency-representations of the modulation, and was chosen for this reason.
Rightward bias in visuo-spatial attention. The finding of a rightward bias for targets
in all behavioral measures is rather uncommon in the research on spatial attention. Instead, a leftward bias was repeatedly reported alongside with research advancing the concept of a dominance of the right-hemisphere in spatial attention (Gitelman et al., 1999; Mesulam, 1981; Nobre et al., 1997).
Similar versions of the task used in the current study did not yield such a bias and thus this excludes an influence of the task arrangement and stimulus material (Gould, Rushworth, & Nobre, 2011; Marshall et al., 2015). The most prominent difference to these studies appears to be the unilaterality of the stimulation setup. Every block entailed at least a short amount of stimulation even though for the sham-blocks the stimulation only lasted about 20 seconds of the total five minutes per block. It is thus conceivable that the aftereffects of the stimulation
sensation might have served as an additional spatial cue to the participant’s attention. Indeed, there is indication that spatial attention integrates between modalities and also the effects of tactile cueing were shown to bear effects on visuo-spatial attention (Macaluso, Frith, & Driver, 2002). Interestingly, Marshall and colleagues (2015) used a bilateral rTMS setup and their data show a similar pattern in terms of behavior in so far that the attention shows a bias towards the stimulation side. However, in the current study we lack a non-lateralized and left-sided control group to decide on the nature of these observations.
Altered ipsi- and contralateral phase synchrony after right cues. The analysis of
functional connectivity during the cue-target interval showed increased phase-locking of both F1 and F2 with the left posterior-occipital site, PO3, after right cues as compared to left cues. With regard to the perspective that alpha-oscillations might serve as a local inhibitor (Jensen & Mazaheri, 2010), one would expect increased alpha power ipsilateral to the attended hemifield, which is what we found in the analysis of the alpha-modulation. This reinforced functional connectivity in terms of phase coherence contralateral to the attended hemifield here suggests a dissociation between possibly local inhibition via power and alpha-phase in a network as an information carrier as we see a dissociation between increased coherence and increased alpha power. This effect emphasizes the different roles alpha-oscillations might adopt in smaller and larger cell assemblies. Fries (2015) suggested that coherence in the alpha/beta-band is the main top-down control mediator and there is preliminary experimental work in animals that suggests that enhanced top-down influences might result in fortified bottom-up processing (Bosman et al., 2012; Grothe, Neitzel, Mandon, & Kreiter, 2012). In a similar vein, Fries (2015) argued that alpha oscillation with their specific relationship to the gamma rhythm might enable the maintenance of local representation and thus the alpha-band takes a specific role in terms of information conservation in a network. Future research needs to assess the function of alpha-band beyond the notion of inhibition and suppression and explore their possible value to information maintenance and transmission from a top-down-perspective. Particularly with regard to the current study, it is unclear in how far the increase functional connectivity to right cues in contrast to left is related to the observed behavioral rightward bias.
A second aspect worthwhile to discuss is the observed equal interhemispheric interplay. We saw that both frontal channels showed increased connectivity to a similar extent. With regard to research on visuo-spatial attention, the allocation of attention was mostly described to be executed by the contralateral hemisphere respective to attended hemifield and to be achieved through a mainly interhemispheric process presumed based on
white matter tractography (de Schotten et al., 2011). More recent research raises the discussion whether this process should be seen as a joint effort and dynamic interplay of both hemispheres. There are multiple studies reporting that, for instance, the FEF has not only intra- but also interhemispheric connections and is engaged more dynamically as previously thought (Marshall et al., 2015; Vossel, Weidner, Driver, Friston, & Fink, 2012). For the current data, both F1 and F2 are in the vicinity of the FEF and the observed pattern of connectivity is in favor of a bilateral contribution during directed attentional engagement.
Concluding remarks. The presented research sheds light on the limited knowledge
and caveats of tACS-research with lacking knowledge on efficacy and the relationship between stimulation parameters, neural correlates and behavioral effects. Nonetheless, this investigation showed a couple of interesting dissociations between the different levels of measurements, as for example, that the increased alpha modulation in a hemisphere did not translate to behavior. Only the combination of electrophysiology and stimulation affords such an observation and future research needs to disentangle this relationship and continue the discourse on the translation between different level of measurement. Similarly, the interaction between the IAF and externally induced alpha rhythms needs to be addressed more systematically so as to fully understand the findings of the present study.
Taken together, on the basis of the presented evidence it is not possible to come to a conclusion on the role of phase synchronization in the alpha-band on visuospatial attention and related questions on the mechanism by which attention allocation could be achieved remain unanswered.
References
Bosman, C. A., Schoffelen, J.-M., Brunet, N., Oostenveld, R., Bastos, A. M., Womelsdorf, T., … Fries, P. (2012). Attentional Stimulus Selection through Selective Synchronization between Monkey Visual Areas. Neuron, 75(5), 875–888. http://doi.org/10.1016/j.neuron.2012.06.037
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10, 433–436. Capotosto, P., Babiloni, C., Romani, G. L., & Corbetta, M. (2009). Frontoparietal Cortex Controls Spatial Attention through Modulation of Anticipatory Alpha Rhythms. The Journal of Neuroscience, 29(18), 5863–5872. http://doi.org/10.1523/JNEUROSCI.0539-09.2009
Corbetta, M., Kincade, J. M., Ollinger, J. M., McAvoy, M. P., & Shulman, G. L. (2000). Voluntary orienting is dissociated from target detection in human posterior parietal cortex. Nature Neuroscience, 3(3), 292–297. http://doi.org/10.1038/73009
Corbetta, M., Patel, G., & Shulman, G. L. (2008). The Reorienting System of the Human Brain: From Environment to Theory of Mind. Neuron, 58(3), 306–324. http://doi.org/10.1016/j.neuron.2008.04.017
Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. Nature Reviews. Neuroscience, 3(3), 201–215. http://doi.org/10.1038/nrn755
de Schotten, M. T., Dell’Acqua, F., Forkel, S. J., Simmons, A., Vergani, F., Murphy, D. G. M., & Catani, M. (2011). A lateralized brain network for visuospatial attention. Nature Neuroscience, 14(10), 1245–1246. http://doi.org/10.1038/nn.2905
Doesburg, S. M., Bedo, N., & Ward, L. M. (2016). Top-down alpha oscillatory network interactions during visuospatial attention orienting. NeuroImage, 132, 512–519. http://doi.org/10.1016/j.neuroimage.2016.02.076
Egner, T., Monti, J. M. P., Trittschuh, E. H., Wieneke, C. A., Hirsch, J., & Mesulam, M.-M. (2008). Neural Integration of Top-Down Spatial and Feature-Based Information in Visual Search. The Journal of Neuroscience, 28(24), 6141–6151. http://doi.org/10.1523/JNEUROSCI.1262-08.2008
Feurra, M., Bianco, G., Santarnecchi, E., Testa, M. D., Rossi, A., & Rossi, S. (2011). Frequency-Dependent Tuning of the Human Motor System Induced by Transcranial Oscillatory Potentials. The Journal of Neuroscience, 31(34), 12165–12170. http://doi.org/10.1523/JNEUROSCI.0978-11.2011
& Mesulam, M.-M. (1999). A large-scale distributed network for covert spatial attention. Brain, 122(6), 1093–1106. http://doi.org/10.1093/brain/122.6.1093
Gould, I. C., Rushworth, M. F., & Nobre, A. C. (2011). Indexing the graded allocation of visuospatial attention using anticipatory alpha oscillations. Journal of Neurophysiology, 105(3), 1318–1326. http://doi.org/10.1152/jn.00653.2010
Grothe, I., Neitzel, S. D., Mandon, S., & Kreiter, A. K. (2012). Switching Neuronal Inputs by Differential Modulations of Gamma-Band Phase-Coherence. The Journal of Neuroscience, 32(46), 16172–16180. http://doi.org/10.1523/JNEUROSCI.0890-12.2012
Haegens, S., Nácher, V., Luna, R., Romo, R., & Jensen, O. (2011). α-Oscillations in the monkey sensorimotor network influence discrimination performance by rhythmical inhibition of neuronal spiking. Proceedings of the National Academy of Sciences, 108(48), 19377–19382. http://doi.org/10.1073/pnas.1117190108
Helfrich, R. F., Schneider, T. R., Rach, S., Trautmann-Lengsfeld, S. A., Engel, A. K., & Herrmann, C. S. (2014). Entrainment of brain oscillations by transcranial alternating current stimulation. Current Biology: CB, 24(3), 333–339. http://doi.org/10.1016/j.cub.2013.12.041
Hopfinger, J. B., Buonocore, M. H., & Mangun, G. R. (2000). The neural mechanisms of top-down attentional control. Nature Neuroscience, 3(3), 284–291. http://doi.org/10.1038/72999
Hopfinger, J. B., Parsons, J., & Fröhlich, F. (2016). Differential effects of 10-Hz and 40-Hz transcranial alternating current stimulation (tACS) on endogenous versus exogenous
attention. Cognitive Neuroscience, 0(0), 1–10.
http://doi.org/10.1080/17588928.2016.1194261
Jensen, O., Kaiser, J., & Lachaux, J.-P. (2007). Human gamma-frequency oscillations associated with attention and memory. Trends in Neurosciences, 30(7), 317–324. http://doi.org/10.1016/j.tins.2007.05.001
Jensen, O., & Mazaheri, A. (2010). Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in Human Neuroscience, 4, 186.
Kincade, J. M., Abrams, R. A., Astafiev, S. V., Shulman, G. L., & Corbetta, M. (2005). An Event-Related Functional Magnetic Resonance Imaging Study of Voluntary and Stimulus-Driven Orienting of Attention. The Journal of Neuroscience, 25(18), 4593–4604. http://doi.org/10.1523/JNEUROSCI.0236-05.2005
Lachaux, J.-P., Rodriguez, E., Martinerie, J., Varela, F. J., & others. (1999). Measuring phase synchrony in brain signals. Human Brain Mapping, 8(4), 194–208.
