Testing causal influence of alpha oscillations in processing of
relevant and irrelevant visual information
Tara van Viegen1, Michael X Cohen1,3 & Rasa Gulbinaite2
1 Department of Psychology, University of Amsterdam, Amsterdam, the Netherlands 2 Centre de Recherche Cerveau et Cognition, Université Paul Sabatier, Toulouse, France 3 Radboud University, Nijmegen, the Netherlands
Abstract
Flickering stimuli can be used to induce steady-state visual evoked potentials (SSVEPs) and are a powerful method to investigate attention, because the amplitude of the SSVEP measured with electro-encephalography (EEG) correlates with attention. Recent research however, indicates that the frequency of the flicker modulates endogenous oscillations. In addition, it is well known that alpha (~8-13 Hz) oscillations correlate with cognitive performance. Therefore, we tested whether we can interfere with processing of relevant and irrelevant information and behavior by inducing alpha oscillations with flickering stimuli. We hypothesized that behavior would be impaired when the relevant target stimulus flickered within the alpha range (longer reaction times (RTs) and higher error rates), accompanied by higher amplitudes in the SSVEPs compared to when the irrelevant flankers flickered in the alpha range. Additionally, we hypothesized that the difference between incongruent (target and flankers gave competing information) and congruent (target and flankers gave same information) trials would lead to more theta power over mid-frontal electrodes for incongruent than for congruent trials. Alpha flicker significantly led to longer RTs irrespective of whether relevant or irrelevant information flickered in the alpha range. EEG analyses revealed an attentional modulation effect: Larger signal-to-noise ratio (SNR) values when the target flickered within the alpha range compared to the flankers flickering within the alpha range. These results indicate that alpha oscillations induced with flickering stimuli hamper information processing by negatively influencing the processing network as a whole.
Introduction
The role of alpha oscillations (~8-13 Hz) in cognitive performance is well established (Klimesch, 1999). Moreover, alpha oscillations are hypothesized to actively inhibit task-irrelevant regions (Jensen & Mazaheri, 2010). For example, attending one hemifield increases alpha power over the irrelevant, ipsilateral visual cortex and decreases alpha power over the relevant, contralateral visual cortex (Kelly et al., 2006; Rihs et al., 2007). Increases in alpha power over contralateral regions are associated with slower reaction times and decreased accuracy of target detection (Thut et al., 2006; Kelly et al., 2009). Furthermore, both phase and power of alpha oscillations influence the detection of near-threshold targets (van Dijk et al., 2008; Busch
et al., 2009; Mathewson et al., 2009).
Flickering stimuli can be used to induce steady-state visual evoked potentials (SSVEPs) at different frequencies (Morgan et al., 1996). Different flicker frequencies can be used to “tag” different visual stimuli, allowing dissociation of these different stimuli (Keitel et al., 2013; Kim et
al., 2007), even in different modalities (Porcu et al., 2014). So far, frequency tagging was
thought to be a means by which visual processing could easily be examined, without interfering with stimulus processing (Keitel et al., 2014). However, recent research suggests that 10 Hz frequency tagging may additionally induce neural alpha-band oscillations (Spaak et al., 2014; de Graaf et al., 2013). The phase of entrained alpha oscillations was predictive of behavioral performance (Spaak et al., 2014). However, it remains to be elucidated how alpha entrainment
influences the processing of stimuli that are competing for processing resources, i.e. task-relevant and task-irtask-relevant (distracting) stimuli.
A recent study on conflict processing found differential processing of relevant and irrelevant stimuli depending on whether they were flickered at 10 or 12.5 Hz (Gulbinaite et al., 2014). Participants were significantly slower when the target flickered at 10 Hz compared to the 12.5 Hz flicker frequency. One possible interpretation for such findings is that alpha flicker entrained alpha oscillations, which, in turn, suppressed processing of the target stimulus when it flickered at 10 Hz. Alternatively, behavioral differences can be explained by suppressed processing of distractors when they flickered at 10 Hz. Here, we replicated and extended these findings by using control frequencies outside the alpha range. Specifically, we tested whether task performance is impaired by forcing the networks processing task-relevant information to oscillate in alpha.
We used an Eriksen flanker task with the target and flankers flickering within or outside the alpha range. The Eriksen task evokes a conflict effect when the target and the flankers are incongruent (e.g. MMNMM), hence participants are slower and less accurate (Cavanagh et al., 2009; Nigbur et al., 2012). When the flankers (e.g. M) flickered at alpha, we expected reaction times to be faster and accuracy to increase, because inhibition of the distractors would help processing of the target stimulus. On the other hand, when the target (e.g. N) is flickered at alpha, we expected reaction times to be slower and accuracy to decrease, because alpha oscillations would be detrimental in task-relevant regions. When the target and the distractors are congruent (e.g. NNNNN), entrained alpha oscillations are not hypothesized to affect stimulus processing, because both the target and distractors require the same response. Finally, a null result may indicate that processing on a local scale (competing stimuli) is differentially modulated by alpha oscillations when compared to more global processing in different hemispheres as found by Spaak et al. (2014).
To our knowledge, the effects of entrained alpha oscillations when stimuli are competing for processing resources has not been reported before. Moreover, evoking alpha entrainment in task-relevant regions will increase our understanding of alpha oscillations because it is relatively unknown what happens when a network is forced to oscillate at alpha. This study allows extension and reproduction of the findings from Spaak et al. (2014) and Gulbinaite et al. (2014). In a broader context, the results of this study have implications for understanding the role of alpha oscillations in inhibiting neural processing within an information processing network, on a local scale.
Materials and Methods
Participants
Thirty-one participants volunteered in exchange for course credits or money (€15). Flickering stimuli have a slight chance of inducing photosensitive epilepsy; therefore participants with a first-degree family member with epilepsy or migraine were excluded from this study. Participants had no (history of) psychiatric diseases and normal or corrected-to-normal vision and were self-reported right-handed. The experiment was approved by the local ethics
committee of the University of Amsterdam and informed consent was obtained from all participants.
Data from eight participants were excluded: One participant was excluded due to technical issues, one participant was excluded because of poor behavioral performance (accuracy < 70%), one participant was excluded because SSVEP responses were not higher than the general noise level. Five participants were excluded because extensive movement artefacts led to rejecting over 30% of trials. During manual trial rejection, trials with more than one eye-blink or a prolonged eye-blink during the mask period (0-2000 ms) were rejected, because an eye-blink can last up to several hundreds of milliseconds which may hamper SSVEP entrainment. Thus, 23 participants were included in the analyses.
Task
An adapted version of the Eriksen-flanker task with a four-to-two mapping of stimuli to responses was used in this experiment (Wendt et al., 2007; Gulbinaite et al., 2014). The task at hand was previously used by Gulbinaite et al. (2014), with some minor adjustments. Trials with different stimuli were interleaved (e.g. MMNMM would be followed by EEFEE) to minimize congruency sequence effects (Weissman et al., 2014). The task was implemented in Matlab using the Psychtoolbox library (Brainard, 1997). Stimuli were displayed on a 23-in. CRT monitor with a resolution of 1920 × 1080 pixels and a refresh rate of 120 Hz. As in Gulbinaite et al. white stimuli comprised of a central target letter and four identical flanker letters were presented against a black background.
Flankers were displayed exactly below or above the horizontal meridian or exactly right or left of the vertical meridian (Fig. 1), in order to maximize the recorded SSVEP signal related to the processing of the flankers (Vanegas et al., 2013). The stimuli were presented in Sloan font, letters of which are equally discriminable and for which height equals width. Viewing distance was 90 cm; each letter subtended ~3.35° of visual angle, separated by ~0.65° visual angle. Participants used the mouse to respond, where four letter stimuli (E,F,M,N) were mapped to two response keys. When the central target was M or E participants replied with a left button press, with their left thumb and when the central target was N or F participants replied with a right button press, with their right thumb. The task consisted of only response congruent (e.g. EEEEE) or response incongruent (e.g. EEFEE) trials. Overall probability was equal for congruent and incongruent and left and right response trials. Participants completed a practice session (40 trials over 4 blocks), which was followed by the experimental session (12 blocks, with 56 trials per block). In order to keep participants motivated each block was followed by behavioral feedback. Participants were encouraged to respond as fast and accurate as possible, with emphasis on fast responses.
Trials started with a 2 s presentation of a mask, which consisted of hash marks. The central hash mark flickered within the alpha range (10 Hz) or outside the alpha range (7.5 or 15 Hz) and the flanking hash marks flickered outside the alpha range when the central mark flickered at alpha and vice versa. After 2 s the mask was replaced by the stimulus and stimulus presentation lasted until a response was made, or until the deadline of 1200 ms after the stimulus onset was exceeded. Tagging frequencies were changed on each block and blocks were randomly presented to participants. Participants were instructed to focus on a small green dot presented continuously on the screen, in order to prevent eye-movements.
FIGURE 1. Eriksen flanker task, exemplar incongruent trial with alpha at the target. Trials started
with the presentation of a mask consisting of hash marks for 2000 ms. The stimulus consisted of a centrally presented target, flickering within (10 Hz) or outside (7.5 or 15 Hz) the alpha range, while flanking stimuli would flicker in the opposite direction. Stimulus presentation lasted until a button press or until the deadline of 1200 ms was exceeded. Each trial was followed by an inter-trial interval (ITI) of 1000 ms.
Behavioral analyses
The goal of this analysis was to examine whether different flicker frequency differentially affected behavioral performance. We expected longer RTs and more errors on incongruent trials compared to congruent trials and we expected the same pattern for targets flickering within the alpha range compared to outside the alpha range. If alpha flicker affects stimulus processing on a very local scale, an interaction effect is to be hypothesized. Since, congruent trials still provide behaviorally relevant information for stimulus processing in the distracting information carried by the flankers. However, the absence of an interaction effect may be indicative of alpha entrainment influencing the processing network at large because both incongruent and congruent trials are affected by the flicker to the same extent.
To test these hypotheses, the first trial of each block was excluded from behavioral and EEG data analyses. For error rate analyses only errors of commission (incorrect button press) were included. And for reaction time (RT) analyses, trials faster than 150 ms, trials with RTs 3 standard deviations from the participant’s averaged mean, error trials and trials where no response was given were excluded. To test whether tagging frequency influenced behavioral performance a 2 × 2 repeated measures ANOVA was applied with factors: Congruency (congruent and incongruent trials) and tagging frequency at the target (within and outside the alpha range).
EEG preprocessing
EEG-data were acquired at 1024 Hz with 64 BioSemi scalp electrodes (http://www.biosemi.com), which were placed according to the international 10-20 system. Two additional electrodes were placed on the outer eye canthi to record horizontal eye movements. Offline data analyses were performed with EEGLAB toolbox and custom written matlab scripts. Data were high-pass filtered at 0.5 Hz and epoched -1.5 to 3.5 s around mask onset. Data were average referenced and noisy channels which were marked during the recording were excluded from this reference scheme. A baseline correction was applied for a baseline window of -200 ms to mask onset. Trials with incorrect responses or with extreme RT values, as described above, were rejected. Trials containing excessive blinks during the period of interest (mask presentation period), trials with muscle artefacts and trials with horizontal eye movements away from fixation were manually rejected. On average, 14,25% of trials per subject were rejected. After manual trial rejection independent component analysis was used to subtract eye blinks and noise from the data. Finally, bad electrodes, the ones marked during recording sessions, were interpolated and cleaned data from each participant was Laplacian transformed. The surface Laplacian is a spatial band-pass filter, also known as current source density, which attenuates activity from distant sources (Cohen, 2014). After Laplacian transformation the units of EEG amplitude are µV/cm2.
SSVEP ANALYSIS
The goal of the SSVEP analyses was to find a neural source for the behavioral results.
Electrode selection
First, a subject best-electrode procedure was applied to select the electrodes for the SSVEP analyses (Gulbinaite et al., 2014). Power (amplitude squared, µV2/cm2) was estimated
by taking the fast fourier transform (FFT) of the data from mask onset until 800 ms after stimulus onset and averaged over the different target frequencies. The occipital electrode showing the highest power value was selected for each participant and used for further analysis (Fig. 2).
Computation of signal-to-noise ratio for static SSVEPs
Second, signal-to-noise ratios (SNRs) were calculated, because although clear peaks in the frequency spectra were observed at all tagging frequencies (see Fig. 2B), peak power values are affected by the shape of the frequency spectrum, where lower frequencies contain more power than higher frequencies, and are subject to individual differences across participants. SNRs are a more sensitive measure and are more robust than peak amplitude values (Norcia et al., 2015). SNRs of power values, with the target or the flankers flickering at 10 (or 7.5, or 15) Hz, were calculated by taking the peak value at 10 (or 7.5, or 15) Hz of the best-electrode over the mean of the surrounding power values at 9.5 (or 7, or 14,5) and 10.5 (or 8, or 15,5) Hz. In addition to the stimulation frequencies in the frequency spectrum, we also examined the first harmonic of 10 Hz at 20 Hz (Müller-Putz et al., 2005). The SNRs were used to test for attentional modulation with a two-way repeated measures ANOVA with flicker location (target vs. flankers) as a factor. We applied different rmANOVAs to each hypothesis. The goal of this analysis was to examine attentional modulation of the SNRs, where we expected larger SNR values for targets compared to flankers.
FIGURE 2. Electrode selection. A participant’s best electrode was visually selected from the
topographical distribution averaged over different tagging conditions (A). Frequency spectrum of the EEG data showed clear peaks at the tagging frequencies and its harmonics (B). A and B are illustrative data from one participant. Electrode distribution over participants (C). The size of the black dots corresponds to the frequency the electrode that was selected as the best electrode over participants. Iz (N = 12), POz (N = 4), Oz (N = 4), PO8 (N = 1), O1 (N = 1), O2 (N = 1).
Dynamic changes in SSVEP power
Finally, SSVEP amplitude time-course analyses were performed by concatenating epoched, Laplacian transformed EEG data, whereafter a narrow-band pass Gaussian filter was applied with a mean of 7.5, 10 or 15 Hz respectively and a standard deviation of 0.5 Hz, as described previously by Gulbinaite et al. (2014). Filtering was performed by multiplying the FFT of the data with the exponential function:
e - 0.5 ( f - f0 )2 / s2
with f representing the mean frequency and f0 representing the frequency of interest, which was
7.5, 10 or 15 Hz here, and s represents the previously reported standard deviation. After the multiplication, the inverse FFT was obtained and the data were Hilbert transformed and squared to extract power. The trial structure was recreated and the power values were normalized to the average prestimulus baseline power at each frequency band. The data was decibel (dB) transformed [dB power = 10 × log10(power / baseline)], to counter scaling differences between
the different tagging frequencies and scaling differences over different participants. The selected baseline window was -400 to -100 ms before stimulus onset.
We expected more power when the target was flickering within the alpha range at 10 Hz, compared to flankers flickering within the alpha range. Moreover, we expected this difference to outlast after stimulus offset. Here, we hypothesized that behavioral differences would be accompanied by differences between processing networks, indicated by SSVEP differences at the electrode showing pronounced activity evoked by the flickering stimuli lasting over time. THETA BAND POWER ANALYSIS
Additionally, we performed EEG theta band power analysis. The goal of this analysis was to show that this paradigm showed traditional conflict processing at midfrontal electrodes, where we expected greater theta power for incongruent trials compared to congruent trials. To
this end, EEG data was transformed to the time-frequency domain by applying wavelet convolution to the FFT of the data. The FFT was taken from the Laplacian transformed data and multiplied by a family of complex Morlet wavelets for each recording electrode. Morlet wavelet:
ei2πft e-t^2 / (s^2)
with t representing downsampled time, f representing frequency, which started at 2 Hz and increased in 20 linear steps until 35 Hz, and s representing the width of the wavelet cycle, determined by n / 2πf, where n represented the amount of cycles, which ranged from 3 to 9 cycles in 20 logarithmically spaced steps. From the resulting analytical signal the power was obtained by squaring the magnitude of the signal z (power = real[z (t)]2 + imag[z (t)]2). All power
values were baseline normalized to the pre-mask period -400 to -100 ms before mask onset and dB normalized (dB power = log10 (power / baseline)), to counter scaling differences across
participants. STATISTICS
Statistics on dynamic changes in SSVEP power and the theta band conflict effect were performed with non-parametric permutation tests (Maris & Oostenveld, 2007). For the dynamic changes SSVEP power we tested the difference between 10 Hz stimulation at the target versus 10 Hz stimulation at the flankers by computing the average difference between these conditions and testing whether the t-values matching these differences were more or as extreme as values found under the null hypothesis. The permutation distribution was obtained by multiplying the data from a random amount of participants by -1, hence swapping the effect.
The same approach was applied to time-frequency data for each time-frequency point, but here the analysis tested the difference between high and low-conflict conditions (incongruent and congruent trials). All permutation tests reported used 500 iterations.
Results
Behavioral analysis
A typical congruency effect of reaction times (RTs) was observed in the data, even though the stimuli were larger than in previous designs (Gulbinaite et al., 2014; Wendt et al., 2007). Participants were significantly slower (F(1,22) = 8.03, p < 0.05) on incongruent (488.9 ms) than congruent trials (481.8 ms). Slower RTs (F(1,22) = 5.43, p < 0.05) were also observed when the target flickered at alpha (488.2 ms) compared to the target flickering outside the alpha range (482.4 ms). There was no interaction between congruency and flicker frequency (F(1,22) = 0.70, p > 0.05).
Consistent with previous reports, participants also made significantly more errors (F(1,22) = 19.97, p < 0.05) on incongruent (13,8%) than on congruent trials (11,8%). Flicker frequency, however, did not influence error rates significantly (F(1,22) = 0.92, p > 0.05) and the data showed no evidence for an interaction effect (F(1,22) = 0.42, p > 0.05). Moreover, these results indicate that flickering the target stimulus in the alpha range affected behavior for RTs, but not for error rates.
FIGURE 3. Reaction times (left) and Error rates (right) for congruent (blue) and incongruent (red) trials when the flankers or target flickered at 10 Hz. Reaction times (RTs) were larger for incongruent
than congruent trials. RTs were also larger when the target flickered at 10 Hz compared to flankers flickering at 10 Hz. Error rates were larger for incongruent than congruent trials, but error rate was not affected by the flickering of the target or flankers at 10 Hz. Error bars represent one standard error of the mean, * indicate significant effects (p < 0.05).
SSVEP analyses
Visual inspection of the topographical maps showed an occipital topography when averaged over the different tagging frequencies (see Fig. 2A for exemplar data of one participant). From this topography the participant’s best electrode was selected through visual inspection for further analyses of SSVEPs (Fig. 2C). Plotting the FFT of the selected channel showed clear spectral peaks at 7.5, 10 and 15 Hz and their harmonics, which was consistent with the tagging frequencies used in our paradigm (see Fig. 2B for exemplar data of one participant).
SSVEP amplitudes showed typical attention effects, where there was more power for target stimuli, than flanking stimuli at 7.5 and 15 Hz (Fig. 4A, left and middle plot). Contrary to our hypothesis this attention effect was flipped for the 10 Hz tagging frequency (Fig. 4A, right plot). However, when these amplitudes were transformed to signal to noise ratios (SNRs, see Materials and Methods), the SNRs showed the predicted pattern (Fig. 5). When stimuli were flickering at 10 Hz, dynamic SSVEPs were larger when stimulation was at the target location compared to stimulation at the flankers, but these differences were not significant (Fig. 4B).
When the target flickered within the alpha range SNRs were significantly larger (5.27 ± 1.1; mean ± SEM) than when the flankers (2.6 ± 0.4; mean ± SEM) flickered within the alpha range (p < 0.05, Fig. 5). Moreover, when we examined the peak surrounding 20 Hz, the harmonic of 10 Hz, which is considered common practice by some researchers (Müller-Putz et
al., 2005), because the effects are less affected by participants’ individual alpha peaks, the SNR
was significantly (p < 0.05) larger when the target flickered within the alpha range (5.6 ± 1,0; mean ± SEM) compared to when the flankers flickered at alpha (3.6 ± 0,7; mean ± SEM). The SNRs with a peak at 15 Hz were also significantly (p < 0.05) affected by the target flickering at
15 Hz (7.0 ± 1.4; mean ± SEM) compared to the flankers flickering at 15 Hz (3.5 ± 1.0; mean ± SEM). The SNRs with a peak at 7.5 Hz were not significantly affected (p > 0.05, Fig. 5).
FIGURE 4. Static and dynamic SSVEPs. Frequency spectra of the EEG data averaged over
participants’ best electrode for the different tagging frequencies at the target (blue) and at the flankers (red; A). The time-course of SSVEP amplitudes for the target (blue) and the flankers (red) when stimuli flickered at 10 Hz (B). Dynamic SSVEPs did not differ significantly.
FIGURE 5. Signal to noise ratios for static SSVEPs. SNRs for the different flicker frequencies and the
first harmonic of 10 Hz (top row). There was no difference between 7.5 and 15 Hz flankers when the target flickered at 10 Hz (bottom left) and there was also no difference between 7.5 and 15 Hz targets when the flankers flickered at 10 Hz (bottom right). Error bars represent standard error of the mean, * indicates significant result (p < 0.05).
The goal of these analyses was to relate the behavioral effect to the brain data. To exclude that the differences between SNRs were driven by the target flickering at alpha, instead of the target flickering at 7.5 or 15 Hz we performed post-hoc comparison of the target at those frequencies. Post-hoc analysis revealed that the effect with the SNR peak at 10 Hz was not caused by one of the conditions, but that the SNRs were generally larger when the target flickered within the alpha range compared to the flankers flickering within the alpha range (p > 0.05, Fig. 5).
Finally, we examined conflict modulation of theta oscillations over midfrontal electrodes. The topographical distribution of stimulus-locked theta power showed a clear fronto-central distribution, in line with previous results (Cohen & Donner, 2013). Over participants a spectral peak was observed at electrode Cz (Fig. 6A). Moreover, the time-frequency condition average of congruent and incongruent trials regardless of flicker frequency shows a large positivity 300-700 ms after stimulus onset in the theta range (3-7 Hz; Fig. 6B, top). The difference plot between incongruent and congruent trials shows more power for incongruent trials compared to congruent trials (Fig. 6B, bottom). Assessing the statistical differences between incongruent and congruent trials showed more power for incongruent trials, than congruent trials (thick black lines in Fig. 6B, bottom), but when cluster correction was applied to correct for multiple comparisons, the result did not survive.
FIGURE 6. Relative theta power. Topographical distribution of theta (3-7 Hz) power averaged over
conditions for 300-700 ms after stimulus onset (A). Electrode Cz indicated with a black dot. Time-frequency representation for electrode Cz averaged over participants and congruent and incongruent trials (top) and time-frequency difference plot of incongruent versus congruent trials (bottom; B). Black thick lines indicate significant results, before cluster correction.
Discussion
In this study, we asked what happened to the information processing of relevant and irrelevant information when alpha oscillations were induced in the processing network with flickering stimuli. The traditional conflict effect was present in the data, such that participants were slower and made more errors on incongruent compared to congruent trials. Participants were also slower when the target flickered within the alpha range, compared to the target flickering outside the alpha range. EEG analyses revealed an attentional modulation effect: Larger SNR values when the target flickered within the alpha range compared to the flankers flickering within the alpha range.
Behavior
Next to the traditional conflict effect we expected reaction times to be affected by the flicker frequency, reflected in the interaction between congruency and flicker frequency of the target stimuli. We hypothesized that processing of the target stimuli on incongruent trials as compared to congruent trials would be more affected by the alpha flicker. Congruent trials still carry helpful information in the flankers. However, we found that reaction times were longer for both congruent and incongruent trials when the target was flickering within the alpha range. This suggests that the processing network is influenced by alpha entrainment more globally than expected. As such, the congruency and flicker frequency of the target stimuli do not interact. Entrained alpha oscillations at the target are detrimental for both congruent and incongruent trials.
Error rates, however were not significantly affected by our manipulation, even though the traditional flanker effect was present in our data. The discrepancy between reaction time and error rate measures can be explained firstly by the sensitivity of the measures. Reaction times in general are a more sensitive measure to differences between conditions than error rates. Second, a speed-accuracy trade-off is known to be present in these kinds of tasks (Heitz & Schall, 2012). The speed-accuracy trade-off predicts that participants are more accurate when reaction times increase, which may explain the results reported here.
EEG
From the EEG data we expected larger peaks in the spectral representations of the static SSVEPs for attended (target) compared to unattended (flankers) information. Visually, the data for the flicker frequencies at 7.5 and 15 Hz followed this hypothesis, but at 10 Hz the effect was flipped. To our knowledge there is one paper where it was previously shown that attention effects flip at 10 Hz stimulation (Ding et al., 2006). However, as explained previously, EEG peak values are highly sensitive to differences between participants. Therefore, we employed signal to noise ratios (SNRs) in further analyses. The SNRs showed a significant effect in the hypothesized direction for stimulation at 10, 15 and 20 Hz. SNRs were calculated by dividing the peak at 10 (or 15, or 20) Hz by the mean value of the power at 9.5 (or 14.5, or 19.5) Hz and 10.5 (or 15.5, or 20.5) Hz.
However, for dynamic SSVEPs we did not find significant effects when comparing 10 Hz stimulation at the target to 10 Hz stimulation at the flankers. A possible explanation can be found in the sensitivity of this measure, because it’s similar to the above-mentioned EEG peak amplitude measure. Moreover, when a participant’s individual alpha peak is close to 10 Hz, the peak at this frequency may be very broad, overshadowing any effects caused by the manipulation in this range. Variability in individual alpha peak is large over participants (Haegens et al., 2014) and this could have obscured the data for the dynamic SSVEPs.
Finally, we examined the effect of flicker frequency on theta power over midfrontal electrodes. In line with previous experiments (Cohen & Donner, 2013), we expected more power for incongruent compared to congruent trials. Again, the results were in the right direction but did not survive multiple comparisons correction. Cross-frequency coupling between alpha and theta oscillations may obscure the traditional conflict effect on theta power, but more thorough analyses are necessary to interpret the absence of a result.
Limitations
We found a behavioral effect of our manipulation on reaction times where participants were slower when the target flickered in the alpha range. This slowing may be related to increased alpha oscillations in the processing network, because the participants’ best electrode also showed higher SNRs when the target flickered at alpha compared to the flankers flickering at alpha. Together these findings suggest that induced alpha oscillations in the processing network hamper information processing. However, the data may also be explained by a relative speeding when the target is flickering at 15 Hz. Participants are faster when the target is flickering outside the alpha range and targets at 15 Hz show high SNRs. The design does not allow to distinguish between these possibilities, because 7.5 and 15 Hz were not presented simultaneously. In this design there was always stimulation of alpha present (at the target or the
flankers). However, incorporating trials where 7.5 and 15 Hz targets and flankers are paired would create many more trials.
Another weakness of the design may lie in the flicker frequencies chosen outside the alpha range. 7.5 Hz may be too close to the lower bound of alpha oscillations in some participants. However, even if for some participants 7.5 Hz is very close to their individual alpha peak, it does not explain the results obtained here. Moreover, the flicker frequencies are determined by the refresh rate of the screen and lowering the lower frequency may entrain theta oscillations over occipital areas.
In this task a participant’s best electrode approach was applied, which has been done in previous research (Gulbinaite et al., 2014). This approach is widely used and has the advantage of accounting for small changes in topography across participants. A disadvantage might be the decreased dimensionality of the data, because there is only one electrode with one peak value or SNR per participant.
Possible future directions
Future research is necessary to examine whether power or phase of alpha oscillations are the main contributors to the behavioral manipulation observed here. The present design was not time-locked, which makes it impossible to distinguish between the effect of alpha power and/or phase. However, the contribution of alpha power and phase to information processing remain controversial (compare Samaha et al., 2015 and Van Diepen et al., 2015). Phase-locking the stimuli from this paradigm would allow event related and intertrial coherence analyses, possibly elucidating the differential contribution of alpha power and alpha phase.
Another interesting line of research may be to adjust the flicker frequency to the individual alpha peak frequency, because large inter- and intravariability in alpha peak frequency exists (Haegens et al., 2014). On line determination of the participant’s individual alpha peak frequency may allow adjustment of the flicker frequency increasing the likelihood of successful entrainment.
Interpretations & Conclusion
Evidence is heaping up in favor of the hypothesis that the presentation of flickering stimuli influences subsequent information processing by entraining intrinsic rhythms (Keitel et
al., 2014). In line with previous results we found an effect caused by the different flicker
frequencies (Gulbinaite et al., 2014; Spaak et al., 2014). Although the limitations of our design do not directly allow us to ascribe the directionality of the effects to one of the flicker frequencies, attentional modulation was observed when the flicker was within the alpha range. This attentional modulation of SNR values suggests alpha entrainment that negatively influenced information processing. These data support the inhibitory role of alpha oscillations (Jensen & Mazaheri, 2008). However, it remains to be elucidated whether alpha power or phase is responsible for these effects.
In summary, entrainment of alpha oscillations led to slower responses for both congruent and incongruent trials. This behavioral effect was accompanied by larger SNRs when the target was flickering within the alpha range, compared to outside the alpha range. These findings suggest that alpha entrainment influences information processing on a larger scale than
hypothesized by us. Alpha oscillations are detrimental for the processing network as a whole without differentiating between relevant and irrelevant information.
Literature
Brainard, D.H. (1997). The Psychophysics Toolbox. Spatial Vision, 10, 4, 433-436.
Busch, N.A., Dubois, J., & VanRullen, R. (2009). The phase of ongoing EEG oscillations predicts visual perception. Journal of Neuroscience, 29, 24, 7869-7876.
Cavanagh, J.F., Cohen, M.X., & Allen, J.J. (2009). Prelude to and resolution of an error: EEG phase synchrony reveals cognitive control dynamics during action monitoring. Journal of
Neuroscience, 29, 1, 98-105.
Cohen, M.X. (2014). Analyzing neural time series data: theory and practice. MIT Press, Cambridge.
Cohen, M.X., & Donner, T.H. (2013). Midfrontal conflict-related theta-band power reflects neural oscillations that predict behavior. Journal of Neurophysiology, 110, 12, 2752-2763.
van Diepen, R.M., Cohen, M.X., Denys, D., & Mazaheri, A. (2015). Attention and temporal expectation modulate power, not phase, of ongoing alpha oscillations. Journal of Cognitive
Neuroscience, 27, 8, 1573-1586.
van Dijk, H., Schoffelen, J.M., Oostenveld, R., & Jensen, O. (2008). Prestimulus oscillatory activity in the alpha band predicts visual discrimination ability. Journal of Neuroscience, 28, 8, 1816-1823.
Ding, J., Sperling, G., & Srinivasan, R. (2006). Attentional modulation of SSVEP power depends on the network tagged by the flicker frequency. Cerebral Cortex, 16, 7, 1016-1029.
de Graaf, T.A., Gross, J., Paterson, G., Rusch, T., Sack, A.T., & Thut, G. (2013). Alpha-band brain rhythms in visual task performance: phase-locking by rhythmic sensory stimulation. PLoS
One, 8, 3, 1-12.
Gulbinaite, R., Johnson, A., de Jong, R., Morey, C.C., van Rijn, H. (2014). Dissociable mechanisms underlying individual differences in visual working memory capacity. NeuroImage, 99, 197-206.
Haegens, S., Cousijn, H., Wallis, G., Harrison, P.J., & Nobre, A.C. (2014). Inter- and inta-individual variability in alpha peak frequency. NeuroImage, 92, 46-55.
Heitz, R.P., & Schall, J.D. (2012). Neural mechanisms of speed-accuracy tradeoff. Neuron, 76,
3, 616-628.
Jensen, O., & Mazaheri, A. (2010). Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in Human Neuroscience, 4, 186.
Keitel, C., Andersen, S.K., Quigley, C., & Müller, M.M. (2013). Independent effects of attentional gain control and competitive interactions on visual stimulus processing. Cerebral Cortex, 23, 4, 940-946.
Keitel, C., Quigley, C., & Ruhnau, P. (2014). Stimulus-driven brain oscillations in the alpha range: entrainment of intrinsic rhythms or frequency-following response? Journal of
Neuroscience, 34, 31, 10137-10140.
Kelly, S.P., Lalor, E.C., Reilly, R.B., & Foxe, J.J. (2006). Increases in alpha oscillatory power reflect an active retinotopic mechanism for distracter suppression during sustained visuospatial attention. Journal of Neurophysiology, 95, 6, 3844-3851.
Kelly, S.P., Gomez-Ramirez, M., & Foxe, J.J. (2009). The strength of anticipatory spatial biasing predicts target discrimination at attended locations: a high-density EEG study. European
Journal of Neuroscience, 30, 11, 2224-2234.
Kim, Y.J., Grabowecky, M., Paller, K.A., Muthu, K., & Suzuki, S. (2007). Attention induces synchronization-based response gain in steady-state evoked potentials. Nature Neuroscience, 10, 1, 117-125.
Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Research Reviews, 29, 2-3, 169-195.
Mathewson, K.E., Gratton, G., Fabiani, M., Beck, D.M., & Ro, T. (2009). To see or not to see: prestimulus α phase predicts visual awareness. Journal of Neuroscience, 29, 9, 2725-2732.
Morgan, S.T., Hansen, J.C., & Hillyard, S.A. (1996). Selective attention to stimulus location modulates the steady-state visual evoked potential. PNAS, 93, 10, 4770-4774.
Müller-Putz, G.R., Scherer, R., Braunels, C., & Pfurtscheller, G. (2005). Steady-state visual evoked potential (SSVEP)-based communication: impact of harmonic frequency components.
Journal of Neural Engineering, 2, 123.
Nigbur, R., Cohen, M.X., Ridderinkhof, K.R., & Stürmer, B. (2012). Theta dynamics reveal domain-specific control over stimulus and response conflict. Journal of Cognitive Neuroscience, 24, 5, 1264-1274.
Norcia, A. M., Appelbaum, L.G., Ales, J.M., Cottereau, B.R., & Rossion, B. (2015). The steady-state visual evoked potential in vision research: A review. Journal of Vision, 15, 6.
Porcu, E., Keitel, C., & Müller, M.M. (2014). Visual, auditory and tactile stimuli compete for early sensory processing capacities within but not between senses. NeuroImage, 97, 224-235.
Rihs, T.A., Michel, C.M., & Thut, G. (2007). Mechanisms of selective inhibition in visual spatial attention are indexed by alpha-band EEG synchronization.
Samaha, J., Bauer, P., Cimaroli, S., & Postle, B.R. (2015). Top-down control of the phase of alpha-band oscillations as a mechanism for temporal prediction. Proceedings of the National
Academy of Sciences, 112, 27, 8439-8444.
Spaak, E., de Lange, F.P., & Jensen, O. (2014). Local entrainment of ⍺ oscillations by visual stimuli causes cyclic modulation of perception. Journal of Neuroscience, 34, 10, 3536-3544. Thut, G., Nietzel, A., Brandt, S.A., & Pascual-Leone, A. (2006). Alpha-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection. Journal of Neuroscience, 26, 37, 9494-9502.
Vanegas, M.I., Blangero, A., & Kelly, S.P. (2013). Exploiting individual primary visual cortex geometry to boost steady state visual evoked potentials. Journal of Neural Engineering, 10, 3. Wendt, M., Heldmann, M., Münte, T.F., & Kluwe, R.H. (2007). Disentangling sequential effects of stimulus- and response-related conflict and stimulus-response repetition using brain potentials. Journal of Cognitive Neuroscience, 19, 7, 1104-1112.
