1
The Effects of Attention on
Steady-State Visual Evoked Potentials in
Response to 3 – 80 Hz Flicker
Diane Helena Maria Roozendaal
Abstract
Steady-State Visual Evoked Potentials (SSVEPs) are periodic brain responses to periodic visual stimuli (flicker). Although frequency of the brain responses is identical to the flicker frequency and is stable over time, the amplitude varies as a function of the flicker frequency and has local maxima at ~10, ~20 and ~40 Hz (resonance frequencies). Previous studies also found that SSVEP amplitude to attended stimuli is higher than for ignored. In this study we sought to quantify modulatory effects of spatial attention on SSVEP amplitude in a wide flicker frequency range (3-80 Hz), with a focus on resonance frequencies. Eighteen subjects participated in a covert spatial attention task while EEG was recorded. Subjects were instructed to detect targets while maintaining fixation and covertly attending or ignoring the flicker. Attentional modulation was defined as the difference in SSVEP amplitude between attended and ignored flicker trials, divided by its sum. In accordance with previous reports, SSVEP response was present up to 80 Hz, with increased SSVEP amplitude at resonance frequencies. Attentional modulation was most pronounced in alpha and low gamma frequency bands: With a positive attentional modulation in gamma, and a negative in alpha. Our findings suggest that attentional modulation is not constant across flicker frequencies, and is affected by interactions with endogenous oscillations. This is the first study to demonstrate significant attentional modulation above 30 Hz and therefore provides insights in the underlying mechanisms of spatial attention.
Student number: 10001424 Supervisor: Dr. Rasa Gulbinaite Co-assessor: Dr. Heleen Slagter
Department: Centre de Recherche Cerveau et Cognition & Université Paul Sabatier Research Master Brain and Cognitive Sciences, track Cognitive Neuroscience January 2016 – August 2016, 42 ECTS
2 Introduction
Steady-State Visual Evoked Potentials (SSVEP) are periodic brain responses to periodic visual stimuli (flicker). The frequency of the brain oscillations matches the flicker frequency and is phase-locked (Regan, 1977; Silberstein, 1995). SSVEPs are thought to be a superposition of event-related potentials (ERPs) in response to each flash of light (Capilla et al., 2011). The constant rate of stimulus presentation prevents the ERP to return to its baseline. As a result, neurons start synchronizing with the flicker frequency and the brain response becomes sinusoidal. However, alternative mechanisms have been proposed (Thut et al., 2011; Spaak et al., 2014).
SSVEPs in the power spectrum can be categorized in fundamental and harmonic components, in which the fundamental component refers to the Fourier-power component at flicker frequency. Second harmonics are the components twice the flickering frequency (Regan, 1989; Hermann, 2001; Di Russo et al., 2002; Norcia et al., 2015). Subharmonics are a smaller component of the flicker (i.e. half). Higher harmonics also exist (three, four, five times the flickering frequency), but their origin remains unclear (Kim et al., 2011).
SSVEP amplitude depends on the amount of attention paid to each stimulus (Toffanin et al., 2009). When several stimuli are presented simultaneously and flicker at different frequencies, the amplitude of SSVEP to the attended stimuli is higher compared to unattended stimuli. Therefore, it is often referred to as ‘frequency tagging’ because the processing of attended and unattended stimuli can be easily tracked (Regan, 1977; Tononi et al., 1998; Müller & Hillyard, 2000).
The frequency range in which SSVEPs are commonly experimentally elicited varies from 3 to 60 Hz (Vialatte et al., 2010). However SSVEPs with frequencies below 3 Hz and up to 100 Hz have been reported (Vialatte
et al., 2008; Elliot & Müller, 1998; Herrmann,
2001). In an elegant study, Hermann (2001) flickered stimuli from 1 to 100 Hz in steps of 1 Hz to investigate the interaction between flicker and endogenous brain oscillations, and
to characterize the amplitude spectrum of SSVEPs across a wide range of frequencies. He found that SSVEP amplitudes were increased relative to the surrounding frequencies when flickering in alpha (8-13 Hz), beta (13-25 Hz), and gamma (40-60 Hz) frequency bands. Thus, he demonstrated that the human visual system exhibits resonance phenomena at certain flicker frequencies.
Previous studies on attention reported attentional modulation for flicker frequencies only up to 30 Hz (Kashiwase et al., 2012; Kus et al., 2013). However, the effects of attention at different flicker frequency bands are divergent, e.g. positive and negative modulation have been reported. For example in alpha band frequency, numerous studies found attentional facilitation of SSVEP responses (Morgan et al., 1996; Toffanin et al., 2009; Lauritzen et al., 2010). However, other findings have been reported (Ding et al., 2006). Negative attentional modulation can be explained by the dominance of alpha in spontaneous EEG in rest. Several studies found a causal interaction between SSVEPs and endogenous alpha activity. When the flicker frequency was at alpha, endogenous alpha power was suppressed compared to resting state alpha (Mast & Victor, 1991; Rau
et al., 2002; Birca et al., 2006).
In this study we systematically quantified the attentional modulation for SSVEP amplitude in a wide flicker frequency range from 3 – 80 Hz, and investigated the interaction between flicker frequency and endogenous brain rhythms. We employed a spatial attention task, in which participants covertly ignored and attended the flicker stimulus. The difference between the two SSVEP amplitudes, divided by its sum, was defined as attentional modulation. Given the selective preference of the brain for certain frequencies, we hypothesized that attentional modulation of SSVEP response in the entire frequency spectrum will resemble the pattern of resonance phenomena (Regan, 1977), i.e. increased attentional effects at ~10, ~20 and ~40 Hz compared to neighboring frequencies, with the highest modulation in alpha and the lowest in gamma band. To preview the results, we found SSVEP responses up to 80 Hz with
3 increased amplitudes at resonance
frequencies. Attentional modulation was most pronounced at alpha and low gamma frequency bands.
Methods
Subjects
Twenty healthy volunteers (mean age= 27.35, SD= 3.60, 5 females) participated in this experiment for financial compensation. All had normal or corrected-to-normal vision and were screened for photosensitivity and epilepsy. All subjects gave written informed consent and the study was approved by the local ethics committee at Centre National de la Recherche Scientifique (CNRS).
Experimental design
Subjects were seated in a dark room, approximately 66 cm from a custom made special light-emitted diode (LED) setup with their head placed in a chinrest (figure 1A). The setup (placed in front of a computer screen: 1280 x 1024 pixels; 85Hz refresh rate) consisted of one box with one yellow LED for fixation (75 x 75 x 30 millimeters; at the center of the screen) and two boxes with 12 LEDs (4 blue LEDs surrounded by 8 white LEDs) placed in each lower visual field (75 x 75 x 70 millimeters; eccentricity of 6.5° visual angle to the center of the screen).
A
B
Figure 1. Experimental design. A: The LED setup was placed in front of the computer screen. One fixation LED was place at the center of the screen. The other LEDs were placed in the lower visual field with an eccentricity of 6.5° visual angle to the center of the screen. B: Four blue LEDs (for target and distractor presentation) were surrounded by 8 white LEDs (for static and flicker presentation). C: Each trial started with a cue (80% validity) which indicated the to-be-attended location; a brief white flash of 100 milliseconds. We instructed subjects to maintain fixation at the yellow LED and covertly displace their attention to the cued location. After a short break of 400 milliseconds the left or right location flickered in a sine wave with a period varying from 3 – 80 Hz, the other location had a constant luminance of 50. Zero, one, two or three blue flashes could appear at both locations and subjects were instructed to count the number of blue flashes in the attended location. At the end of the trial, a grey square asked about the number of blue flashes seen at that particular location. An inter trial interval (ITI) of 1.0 – 1.5 seconds was introduced after the response.
4 0 10 20 30 40 50 60 70 80 90 1 7 13 19 25 31 37 Fl ic ke r fr e q u e n cy ( H z) Number of frequencies The task comprised a spatial attention
paradigm (figure 1C), in which subjects had to count the number of blue flashes (targets) at the cued location.Subjects were instructed to maintain fixation throughout the entire trial and covertly displace their attention to the cued location. A cue at the beginning of each trial (location randomized across trials) indicated the to-be-attended location. This cue had a validity of 80% and was generated by a brief flash of the white LEDs (100 milliseconds). After a 400 milliseconds break, either the left or the right location flickered for eight seconds while the other location was kept constant (static) at 100% of the maximum luminance.
The flicker stimuli were generated using a sinewave profile (Teng et al., 2011). In total, we used 41 frequencies from 3 till 80 Hz, with high resolution sampling of flicker frequencies in alpha band (figure 2). Each frequency was presented three times in four different conditions: attend right flicker, ignore right flicker, attend left flicker and ignore left flicker (counterbalanced and randomized across trials). During each trial 0, 1, 2 or 3 targets (cued location) or distractors (un-cued location) could appear with duration of 200 milliseconds. Presentation of the blue flashes was phase-locked to the peak of the periodic sine wave (Busch & VanRullen, 2010). The probability of the number of blue flashes was drawn from a gamma distribution (k: 7 and θ: 0.35; see supplementary materials figure S1a). The timing of the blue flashes was drawn from three separate truncated negative exponential distributions to achieve maximal unpredictability of the next stimulus presentation. The timing was as follows: one-flash trial (λ= 3; T= 1 – 8 seconds), two-one-flash trial (first flash occurred between 1 – 3 seconds with λ= 2; the second flash occurred between 1 second after the first flash and 8 seconds with λ= 1) and three-flash trial (first flash between 1 – 3 seconds with λ= 2; the second flash between 1 second after the second flash and 6 seconds with λ= 1; third flash between 1 second after the second flash and 8 seconds with λ= 1. See supplementary materials: figure S1b).
We kept performance accuracy constant at 80% by adjusting the target luminance with a QUEST adaptive staircase procedure (Watson & Pelli, 1983). The QUEST parameters were: number of trials for sliding window= 40; beta = 3.5; delta = 0.01; gamma = 0.05; grain = 0.01; range in which luminance could vary = 10 – 80 % of the maximum luminance. Luminance of blue LEDs was also adjusted based on flicker frequency, with a frequency window for QUEST procedure set to +-5 frequencies surrounding the frequency of interest. Separate QUEST sliding windows were defined for attend flicker and ignore flicker (attend static) trials. Luminance of distractor blue LEDs was defined based on the performance on attend flicker and attend static trials.
At the end of each trial a grey square behind the left or right location appeared on the monitor screen, asking how many blue flashes the subject counted at that particular side. The cue at the beginning of the trial had 80% validity, indicating that the cue matched the location of the grey box in 80% of the trials. Thus, 20% of the trials were invalid and the question was about the un-cued location. We instructed subjects to only pay attention to the cued location and maintain fixation.
Figure 2. Flicker frequency distribution. In total we used 41 flicker frequencies with a period varying from 3 – 80 Hz, with a focus at resonance frequencies: theta (green; N=4), alpha (red; N=13), beta (yellow; N=9) and gamma (blue; N=15).
5 30 50 70 90 Valid Invalid A cc u rac y ( 95% C .I . in te rv al ) Attend Flicker Ignore Flicker 30 40 50 60 70 80 90 Left Right A cc u rac y ( 95% C .I . in te rv al ) Hemifield The total experiment consisted of 12
blocks, with 41 trials each. Each block started with a 9-point custom position calibration and validation of the eye tracker (EyeLink 1000 plus version 5.04, SR Research, Ontario). Initially, 10 trials (3, 8, 11, 15, 20, 25, 30, 40, 55 and 65 Hz) with performance feedback were practiced. Luminance of the blue flashes was held constant at 70% of the maximum luminance and the cued location was randomized. Resting state alpha was recorded before starting the actual experiment. Subjects were instructed to keep their eyes open or closed for 60 seconds; each trial was presented twice and in randomized order. All experiments were written in MATLAB, using the Psychophysics Toolbox (Brainard, 1997; Pelli, 1997).
Stimuli
The custom made LED setup consisted of 3 boxes; one for fixation and two for the flicker stimuli. The stimuli contained four blue LEDs, for target and distractors presentation, surrounded by eight white LEDs for flicker and static presentation (figure 1B). All boxes were painted black and covered with transparent tracing paper for equally distributed luminance perception. To generate periodic sinewave stimuli, the LEDs were connected to a STM32F4-Discovery board, which embed an ARM CortexM4 micro-controller. Communication to the sinewave generator was controlled via MATLAB with a USB virtual com port protocol (programmed in C++) and an on-board JTAG programmer/debugger.
Luminance of the LED lights was adjusted with a high frequency Pulse Width Modulation (PWM) signal, and then linearized with a low-pass filter. Luminance of the sine wave stimuli (white LEDs) was fixed throughout the experiment and set to 70% of the maximum luminance, whereas the luminance of targets and distractors (blue LEDs) was variable and adjusted based on detection task performance. The period of the sine wave varied between 3 – 80 Hz. The maximum variation of the current in the LED is given by:
∆I = (Vdd-Vd)/(L*f)/nbLEDs
• Vdd = 5V, the power supply voltage • Vd = 3.2V @ 20mA, the LED voltage • L = 10mH, the inductance value
• nbLEDs = number of LEDs, 4 blue and 8 white LEDs
• f, the PWM switching frequency, 200kHz by default
Data acquisition
EEG was continuously recorded with a 64-channel ActiveTwo Biosemi system. Electrodes were placed according to the international 10/20 system, with two external additional electrodes for reference (CMS; common mode sense) and ground (DRL; driven right leg), and to create a feedback loop for amplifier reference. The data were sampled at 1024 Hz and online high-pass filtered at 0.16 Hz (by the amplifier).
A
B
Figure 3. Behavioural results. A: Accuracy as a function of validity for attended and ignored flicker. A significant main effect of validity was found, p < 0.001. The errors bars reflect the 95% confidence interval. B: Accuracy as a function of hemifield. A significant main effect of hemifield and flicker was found, as well as an interaction effect, p<0.05.
6 Three external electrodes were placed
around the eyes to record horizontal and vertical electrooculograms (EOG). Additionally, to control for fixation and sizable eye movements, eye positions were recorded with an eye-tracker.
Data preprocessing and analyses
All the data were analyzed using custom written scripts in MATLAB and EEGLAB Toolbox extensions were used for pre-processing (Delorme & Makeig, 2004). We used a high pass filter at 0.5 Hz and re-referenced to the average reference. Data were epoched from -0.5 till +9.0 seconds, relative to cue-onset. All trials were visually inspected for artefacts and marked for rejection. Eye blinks and muscle artefacts were not rejected, because SSVEPs have high signal to noise ratio and a narrow frequency band whereas artefacts have broad frequency spectra. On average, 0.15% of all trials per subject were discarded. Independent Component Analysis (ICA) decomposition was executed and the components were visually inspected to remove eye blink components. On average, 1 component per subject was rejected.
We analyzed the data in two different ways: at electrode level and using a spatial filtering method that combines information from all channels (Cohen & Gulbinaite, in
prep).
Figure 4. SNR comparison between RESS
filtered data and electrode PO4 of single subject for 11 Hz. SNR values for RESS filtered data is about 7 times higher than compared to SNR values at electrode PO4 for left hemifield flicker.
Electrode-level analyses
The first 1500 milliseconds (cue duration, post-cue time, and 500 milliseconds of the flicker) of each trial were discarded to remove cue-related activity, as well as ERPs related to the onset of the flicker. Thus, the time window included in all analyses was 1500 – 8000 milliseconds relative to the cue onset. We performed a Fast-Fourier Transform to extract power of the SSVEP amplitude for each flicker frequency. Since the stimuli were presented in lower hemifield, we choose PO3 for attend and ignore right flicker and PO4 for attend and ignore left flicker (Müller & Hillyard, 2000).
RESS filter
In order to increase the SNR of the SSVEP power, we created a spatial filter by taking all the electrodes into account. The spatial filter was designed as follows. First, we temporally filtered the data collapsed across attend and ignore flicker conditions using three Gaussian filters: one filter at flicker frequency with Full Width at Half Maximum (FWHM) = 0.5 and two filters -+2 Hz relative to flicker frequency with FWHM = 2 Hz. Second, we computed covariance matrices of all electrodes for the data filtered at flicker frequency (S: Signal) and both neighboring frequencies (R: Reference). Third, we performed eigenvalue decomposition and selected the eigenvector with the largest eigenvalue as a spatial filter:
𝑆𝑥 = λxR (1) S is the covariance matrix of the EEG data at flicker frequency, x is the eigenvector of the covariance matrix, λ is the scaling value of the eigenvector and R is the average of the two neighboring covariance matrices.
Given that different cortical areas oscillate at a characteristic rate, spatial filters were constructed separately for each flicker frequency and hemifield channels (Rosanova
et al., 2009; Hermann et al., 2015). Weights of
the spatial filters were multiplied by the EEG data to obtain time series that were analyzed using FFT.
7
SSVEP amplitude
To quantify frequency tagging, we extracted power of SSVEP amplitude at each flicker frequency with a fast-Fourier-transform. We standardized the Fourier-power to Signal-to-Noise (SNR) units, prior to combining across subjects. The SNR value of each flicker frequency (SNR(ff)) was computed using the
peak of the Fourier power relative to the mean of -+ 2 neighboring frequencies, with 0.1 Hz frequency resolution (2). 𝑆𝑁𝑅(𝑓𝑓) = 𝑓𝑓 1 𝑛∑ 𝑓𝑖+ 1 𝑛∑ 𝑓𝑖 30 𝑖=10 −30 𝑖=−10 (2)
To test whether tagging was significant, we compared SNR values against the noise. Noise SNR was calculated at all flicker frequencies using all conditions but the flicker frequency itself. All SNR spectra were smoothed (relative to -+ 2 neighboring frequencies) at the subject level prior to statistical analyses. To quantify attentional modulation, we analyzed SNR spectra for two conditions: attend and ignore flicker.
Statistical analyses
Three sets of statistical analyses were performed. Firstly, we tested up to what flicker frequency tagging was elicited, when averaged over electrodes PO3 and PO4.
At each flicker frequency, we performed a t-test between SSVEP SNR and noise SNR values. Second, based on previous findings (Corbetta, 1998; Keil et al., 2005), we compared frequency tagging between left and right hemifield. Therefore, we performed separate paired t-tests for left and right SSVEP SNR against noise SNR values and additionally a paired t-test between left and right hemifield SSVEP SNR values.
Figure 6. Response frequency as a function of
flicker frequency. SNR values of SSVEP amplitude (x-axis) are plotted against the flicker frequency (y-(x-axis). The diagonal shows that SSVEP responses follow flicker frequency, with most pronounced response at alpha flicker frequency. Second harmonics are clearly visible (the first line below the diagonal), as well as third and fourth harmonics. Endogenous alpha (~10 Hz) was always present, regardless of the flicker frequency.
Figure 5. Frequency tagging across all frequencies. Tagging was elicited up to 80 Hz. SNR values of SSVEP amplitude were highest in alpha band and lower at around 20 – 30 Hz and 40 Hz. SNR values of each flicker frequency (black curve) were tested against the noise (red curve) and were significant for each flicker frequency, p < 0.05 (uncorrected for multiple comparisons across frequencies). The shading of each curve represents the 95 % confidence interval.
8 Figure 7. SNR spectra for left and right hemifield. A:SNR spectra of SSVEP and noise for left and right hemifield. A paired t-test between each SSVEP SNR and noise SNR value revealed significant differences for all flicker frequencies, except 19 and 80 Hz for left hemifield. The shading curve represents the 95% confidence interval. B: Paired t-tests between each SSVEP SNR value of left and right hemifield revealed only a significant difference for 11.5 Hz, p<0.05 (uncorrected for multiple comparisons across frequencies).
To quantify attentional modulation across frequencies we computed attentional modulation index (MI). For each flicker frequency, we computed the difference between attend and ignored flicker conditions. We standardized the difference by dividing it with the sum of the two (3).
𝑀𝐼 = 𝑆𝑁𝑅(𝑓𝑝 𝑎𝑡𝑡𝑒𝑛𝑑)−𝑆𝑁𝑅(𝑓𝑝 𝑖𝑔𝑛𝑜𝑟𝑒)
𝑆𝑁𝑅(𝑓𝑝 𝑎𝑡𝑡𝑒𝑛𝑑)+𝑆𝑁𝑅(𝑓𝑝 𝑖𝑔𝑛𝑜𝑟𝑒) (3)
Finally, we tested the attentional modulation index for each flicker frequency against zero. Results
One subject was excluded from the analyses because of extensive muscle artefacts, the percentage of rejected trials deviated more
than 3 standard deviations from the mean of all rejected trials per subject. A second subject was excluded, because single trial SNR units deviated more than 3 standard deviations from the mean. In total, eighteen subjects (mean age= 27.61, SD= 3.65, 5 females) were included in the analyses.
Eye-tracker
We recorded eye-movements and fixation during the entire experiment. We analyzed the data offline with the Dataviewer tool of Eyelink. Each block was visually inspected for extensive eye-movements and broken fixation. No trials were rejected based on eye movement data inspection.
A
9 Figure 8. Attentional modulation
across all frequencies. A: SNR spectra for attend and ignored flicker conditions, smoothed relative to +- 2 Hz neighboring
frequencies. B: Attentional modulation
was quantified with an attentional modulation index (MI): SNR attend flicker – SNR ignore flicker / SNR attend flicker +
ignore flicker. Significant attentional
modulation was visible in alpha (9.5 and 10 Hz), and low gamma (29, 31, 33, 37, 39, 41, 45 and 47 Hz) band, p < 0.05 (uncorrected
for multiple comparisons across
frequencies). In alpha band, attentional modulation was negative.
Behavioral results
A 2 (validity: valid and invalid) x 2 (flicker: attend and ignore) within subjects repeated measures ANOVA was performed (figure 3A). A main effect of validity was found (F(1,17) = 111.39 p<0.001), meaning that accuracy was higher on valid cue trials.
We tested whether the number of to-be-detected targets affected task performance. A 2 (flicker: attend and ignore) x 4 (number of targets) repeated measures ANOVA for valid trials only, revealed no significant main effects or interactions, p>0.05.
In order to investigate whether subjects had a bias towards one hemifield or flicker condition, we performed a 2 (hemifield: left and right) x 2 (flicker: attend and ignore) within subjects repeated measures ANOVA (figure 3B).
We found a significant main effect of hemifield F(1,17)= 8.332, p=0.01, a main effect of flicker F(1,17)=14.782, p=0.001 and an interaction between hemifield x flicker F(1,17)=10.432, p=0.005. Thus, subjects performed better when targets appeared in the left hemifield, i.e. a bias towards left hemifield. Performance was increased for attend compared to ignored flicker in left hemifield. However, an inversed attention effect was found for right hemifield flicker, i.e. performance was higher for ignore right flicker compared to attend flicker.
RESS filtered data
We pre-processed all RESS filtered data and found 3 – 4 times enlarged SNR compared to analyses at electrode level (figure 4). However, in the interest of time, we performed our final analyses at electrode level PO3 and PO4.
A
10
Frequency-tagging
Figure 5 shows SNR values of SSVEP amplitude for each flicker frequency, when averaged across electrodes (PO3 and PO4) and flicker conditions (attend and ignore). We found frequency tagging up to 80 Hz and therefore replicated previous report (Hermann, 2001). The SNR spectrum resembled the pattern found by Regan (1977), with larger peak at alpha band and smaller peaks around 20 and 40 Hz.
We tested the SSVEP SNR value of each flicker frequency against the noise SNR and found significant differences for each flicker frequency, p < 0.05 (uncorrected for multiple comparisons across frequencies).
The profile of SSVEP response as a function of flicker frequency is shown in figure 6. Along the diagonal, the response frequency perfectly followed the flicker frequency, i.e. the fundamental frequency. A consistent presence of endogenous alpha (response frequency around 10 Hz) was visible throughout the experiment (the vertical line at around ~10 Hz across all flicker frequencies), regardless of the flicker frequency. In addition, second harmonics were clearly visible for all frequencies, and third and fourth harmonics were marginally visible.
Left and right hemifield
Enhanced responses in right hemisphere compared to left hemisphere have been previously reported (Morgan et al., 1996; Keil et al., 2005). We therefore investigated the SNR spectra of SSVEPs response for left and right hemifield separately (figure 7A). Firstly, within each hemifield we tested if SSVEP SNR values were significantly different from the noise SNR. Paired t-tests between SNR values of signal and noise revealed significant differences for almost all frequencies within each hemifield (p<0.05 uncorrected for multiple comparisons). To test whether left and right hemifield responses differed from each other, we performed a paired t-test between each SNR value (figure 7B). Left and right hemifield responses statistically did not differ for all flicker frequencies, except for 11.5 Hz, p< 0.05
(uncorrected for multiple comparisons across frequencies).
Attentional Modulation
Figure 8A shows the SNR spectra for attend and ignored flicker conditions, when pooled across hemifield. For each flicker frequency, we defined attentional modulation of SSVEP amplitude using an attentional modulation index (3). We collapsed the data across hemifield and found significant attentional modulation in alpha (9.5 and 10 Hz), and low gamma (29, 31, 33, 37, 39, 41, 45 and 47 Hz) band, p<0.05 uncorrected for multiple comparisons across frequencies (figure 8B). Thus, attentional modulation is not constant across frequencies and is most pronounced in alpha and low gamma band: With positive modulation in low gamma band and negative modulation in alpha band frequency.
To further explore negative attentional modulation in alpha, we visualized the difference in SNR spectra between attend and ignore flicker conditions (figure 9). Endogenous alpha was increased for ignored flicker as compared to attend flicker condition, regardless of flicker frequency. In addition, SSVEP amplitude for ignored flicker was increased compared to attend flicker. This
Figure 9. SSVEP response for attend and ignored flicker conditions as a function of flicker frequency. Endogenous alpha and SSVEP responses for alpha flicker were increased in ignored compared to attend flicker.
11 Figure 10. Attentional modulation for left and right hemifield. We found significant modulation for left hemifield flicker in alpha (9.5 and 10.5 Hz), and gamma band (35, 39, 41 and 43 Hz). For right hemifield flicker, we found only significant effects in beta band (31, 33, 35, 37, 53, 60 and 65 Hz). Paired t-tests between left and right attentional modulation revealed only a significant difference for 60 Hz, p< 0.05 (uncorrected for multiple comparisons across frequencies).
indicates interaction between alpha flicker and endogenous alpha. However, further analyses are necessary to disentangle the effects of attention on endogenous alpha versus alpha flicker.
Attentional modulation for left and right hemifield
We calculated the attentional modulation index for left and right hemifield (figure 10). For left hemifield flicker, we found significant modulation in alpha (9.5 and 10.5 Hz), and gamma band (35, 39, 41 and 43 Hz). However, for right hemifield flicker we found only significant modulation in gamma (31, 33, 35, 37, 53, 60 and 65 Hz). A paired t-test between left and right hemifield revealed only a difference in attentional modulation for 60 Hz flicker, p<0.05 (uncorrected for multiple comparisons across frequencies).
Discussion
We systematically investigated the effects of attention on SSVEP responses across a wide flicker frequency range (3 – 80 Hz), with a focus at resonance frequencies:
~10, ~20, ~40 Hz. In a spatial attention task, subjects covertly attended or ignored flicker stimuli presented in each hemifield. Consistent with previous reports, we found significant frequency tagging up to 80 Hz (Hermann, 2001), with increased responses at resonant frequencies: ~10 Hz, ~20 Hz and ~40 Hz. We also found that attentional modulation (difference in SSVEP amplitude for attend and ignore flicker, divided by its sum) is not constant across frequencies. Attentional modulation was most pronounced in alpha and low gamma band, whereas modulation was negative in alpha band and positive in low gamma band frequency.
“Alpha dip”
Several studies reported positive modulation in alpha band when using alpha-band flicker (Morgan et al., 1996; Toffanin et
al., 2009; Lauritzen et al., 2010), which are in
contrast with our findings. However, one study reported negative attentional modulation for flicker in lower alpha-band frequency (Ding et al., 2006). There are
12 several potential explanations to this pattern
of results.
First, “alpha dip” in attentional modulation could be related to increased presence of endogenous alpha in response to flickers. If one actively ignores irrelevant information, endogenous alpha contralateral to the stimulus side increases (Frey et al., 2015). Alpha band activity was enhanced for the hemisphere contralateral to the distractors, indicating the inhibition of irrelevant information (Foxe et al., 1998; Sauseng et al., 2005; Kelly et al., 2006; Rihs et
al., 2009; Jensen et al., 2012). Increased
endogenous alpha could have enhanced the SSVEP response for ignore flicker condition compared to attend flicker condition. As a result, the dominance of endogenous alpha for ignore flicker could have driven the negative attentional modulation effect. This suppressive effect of alpha on irrelevant stimuli was proposed and expanded by numerous spatial attention studies.
Second, a causal interaction between alpha flicker and endogenous alpha has been previously reported: Task-related endogenous alpha power as compared to the resting state alpha power was suppressed more at alpha flicker frequencies (Mast & Victor, 1991; Birca
et al., 2006). Together these findings might
explain why attentional modulation was negative specifically in alpha band frequency. However, additional studies are necessary to disentangle the effects of attention on alpha-band flicker versus the effects of attention on endogenous alpha by separating the two.
Second harmonics
In this study we only focused our analyses on fundamental frequencies, i.e. the Fourier-power component at flicker frequency. However, findings regarding attentional effects at second harmonics (double the fundamental frequency) and even higher harmonics have previously been reported (for a review see Norcia et al., 2015). Kim et al., (2011) found increased posterior contralateral localization for second harmonics, whereas fundamental frequencies were more medial posterior focused. Furthermore, attention only modulated the second harmonics, but
not the fundamental frequency. According to these findings, it might be preferred to quantify attentional modulation at second harmonics instead of fundamental frequencies. Electrodes Frequency left hemifield Frequency right hemifield PO9 27 51 I1 27 38 PO7 14 76 PO3 25 90 O1 8 47 Iz 32 43 Oz 63 82 POz 105 150 PO10 42 34 I2 34 24 PO8 106 31 PO4 163 46 O2 92 26
Figure 11. Frequency and distribution of
subject-specific best-picked electrodes for left and right hemifield. A. For each frequency, each hemifield and each subject we computed the frequency of best picked electrode, i.e. the electrode that showed highest Fourier-power. We restricted our area of interest to the 13 most occipital electrodes. B: Left hemifield flickers are more lateralized than right hemifield. The size of the dot represents the frequency of that selected electrode.
Analyses at electrode level
We investigated which subject-specific electrodes showed highest Fourier power for each flicker frequency (figure 11). We found clear lateralization for left hemifield flicker (i.e. highest power at electrodes contralateral to the stimulus; right
A
13 hemisphere). In contrast, for right hemifield
flicker, this effect lacked. The distribution of best-picked electrode was more centered at electrodes POz and Oz. The lateralization for left hemifield flicker, but not for right hemifield is supported by previous studies (Corbetta, 1998; Keil et al., 2005). Right hemisphere appears to dominate in a spatial attention task, as compared to left hemisphere activation. Since we restricted our analyses to only PO3 and PO4, it could be possible that these electrodes not always captured optimal SSVEP responses.
Limitations and future directions
Previous studies found maximal SSVEP amplitudes for parietal-occipital electrodes contralateral to the stimulus side (Müller & Hillyard, 2000). Therefore, we restricted our analyses to electrode PO3 for flickers in the right hemifield and electrode PO4 for left hemifield flicker. However, channels which show maximal SSVEP response differ across flicker frequencies (Rosanova et al., 2009; Hermann et al., 2015). A subject-specific approach might be preferred, in which (for each flicker frequency) the electrode that shows the maximum Fourier power will be selected for further analyses.
A second option to improve SNR of SSVEP responses is the use of a spatial filter. We found that filtering the data with the RESS filter increased SSVEP SNR values 3-4 times compared to analyses at electrodes PO3 and PO4 (figure 4). These increased SNR values could have resulted in significant attentional modulation of other frequency bands, e.g. beta band.
In addition, we only used three trials per flicker frequency when analyzing the data within each hemifield. Taken all together, these explanations could support the discrepancies we found within left and right hemifield. For example, for left hemifield flickers we did not find a significant difference between SSVEP amplitude and noise for 19 and 80 Hz. If SSVEP response was not maximal at electrode PO4 for one of the three trials, the frequency tagging could have been diminished.
In summary, we replicated the study of Hermann (2001) by eliciting SSVEP responses up to 80 Hz, even though the flicker was imperceptible. In accordance with Regan (1977), we found increased SSVEP responses at resonance frequencies (~10-, ~20 and ~40 Hz), with most pronounced SSVEP peak at the alpha-band flicker. To the best of our knowledge, this is the first study that investigated attentional modulation using flicker above 30 Hz. We provided evidence for significant attentional modulation in low gamma band (30 – 50 Hz). Previous studies typically used low flicker frequencies in high alpha or low beta for highest SNR. However, low flicker frequencies are subjectively uncomfortable to look at. Here, we provide evidence for attentional modulation using flicker frequencies in the gamma band where flicker is imperceptible. The latter finding is specifically important in the application of SSVEP in Brain-Computer-Interface (BIC), in which flicker is used by the computer system to identify and execute the intention of the user (Kus et al., 2013). Finally, the systematic investigation of attentional modulation in 3 – 80 Hz flickers expands the possibility of SSVEP application in electrophysiology, and contributes to a better understanding of the neural correlates of attention.
Acknowledgements
We thank Ludovic Tyack for designing the circuit of the LED setup and programming the communication interface.
References
Birca, A., Carmant, L., Lortie, A., & Lassonde, M. (2006). Interaction between the flash evoked SSVEPs and the spontaneous EEG activity in children and adults. Clinical
neurophysiology, 117, 279-288.
Brainard, D.H. (1997). The psychophysics toolbox. Spatial Vision, 10, 443-446.
Busch, N., & VanRullen, R. (2010). Pre-stimulus EEG oscillations reveal periodic sampling of visual attention. Journal of
14 Buschman, T. J., & Kastner, S. (2015). From
behavior to neural dynamics: an integrated theory of attention. Neuron, 88, 127-144. Capilla, A., Pazo-Alvarez, P., Darriba, A., Campo, P., & Gross, J. (2011). Steady-state visual evoked potentials can be explained by temporal superposition of transient event-related responses. PLoS one, 6, e14543. Corbetta, M. (1998). Frontoparietal cortical networks for directing attention and the eye to visual locations: Identical, independent, or overlapping neural systems? Proceedings of
the National Academy of Sciences, 95,
831-838.
Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of neuroscience
methods, 134, 9-21.
Di Russo, F., Teder-Sälejärvi, W. A., & Hillyard, S. A. (2002). Steady-state VEP and attentional visual processing. The cognitive electrophysiology of mind and brain, 259-274.
Ding, J., Sperling, G., & Srinivasan, R. (2006). Attentional modulation of SSVEP power depends on the network tagged by the flicker frequency. Cerebral cortex, 16, 1016-1029. Elliott, M. A., & Müller, H. J. (1998). Synchronous information presented in 40-Hz flicker enhances visual feature binding. Psychological Science, 9, 277-283. Foxe, J. J., Simpson, G. V., & Ahlfors, S. P. (1998). Parieto‐occipital ∼10Hz activity reflects anticipatory state of visual attention mechanisms. Neuroreport, 9, 3929-3933. Frey, J. N., Ruhnau, P., & Weisz, N. (2015). Not so different after all: The same oscillatory processes support different types of attention. Brain research, 1626, 183-197. Jensen, O., Bonnefond, M., & VanRullen, R. (2012). An oscillatory mechanism for
prioritizing salient unattended stimuli. Trends
in cognitive sciences, 16, 200-206.
Herrmann, C. S. (2001). Human EEG responses to 1–100 Hz flicker: resonance phenomena in visual cortex and their potential correlation to cognitive phenomena. Experimental brain
research, 137, 346-353.
Herrmann, C. S., Strüber, D., Helfrich, R. F., & Engel, A. K. (2015). EEG oscillations: From correlation to causality. International Journal
of Psychophysiology, 103, 12-21.
Kashiwase, Y., Matsumiya, K., Kuriki, I., & Shioiri, S. (2012). Time courses of attentional modulation in neural amplification and synchronization measured with steady-state visual-evoked potentials. Journal of cognitive
neuroscience, 24, 1779-1793.
Keil, A., Moratti, S., Sabatinelli, D., Bradley, M. M., & Lang, P. J. (2005). Additive effects of emotional content and spatial selective
attention on electrocortical
facilitation. Cerebral Cortex, 15, 1187-1197. 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, 3844-3851.
Kim, Y. J., Grabowecky, M., Paller, K. A., & Suzuki, S. (2011). Differential roles of frequency-following and frequency-doubling visual responses revealed by evoked neural harmonics. Journal of cognitive neuroscience, 23, 1875-1886.
Kuś, R., Duszyk, A., Milanowski, P., Łabęcki, M., Bierzyńska, M., Radzikowska, Z. et al. (2013). On the quantification of SSVEP frequency responses in human EEG in realistic BCI conditions. PloS one, 8.
Lauritzen, T. Z., Ales, J. M., & Wade, A. R. (2010). The effects of visuospatial attention measured across visual cortex using
source-15 imaged, steady-state EEG. Journal of vision, 10,
39.
Mast, J., & Victor, J. D. (1991). Fluctuations of steady-state VEPs: interaction of driven
evoked potentials and the
EEG. Electroencephalography and clinical neurophysiology, 78, 389-401.
Morgan, S. T., Hansen, J. C., & Hillyard, S. A. (1996). Selective attention to stimulus location modulates the steady-state visual evoked potential. Proceedings of the National
Academy of Sciences, 93, 4770-4774.
Müller, M. M., & Hillyard, S. (2000). Concurrent recording of steady-state and transient event-related potentials as indices of visual-spatial selective attention. Clinical
Neurophysiology, 111, 1544-1552.
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, 4-4. Pelli, D.G. (1997). The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial Vision, 10, 437-442.
Posner, M. I., Snyder, C. R., & Davidson, B. J. (1980). Attention and the detection of signals. Journal of experimental psychology:
General, 109, 160.
Rau, R., Raschka, C., & Koch, H. J. (2002). Uniform decrease of alpha-global field power induced by intermittent photic stimulation of healthy subjects. Brazilian journal of medical
and biological research, 35, 605-611.
Regan, D. (1977). Steady-state evoked potentials. JOSA, 67, 1475-1489.
Regan, D. (1989). Human brain evoked potentials and evoked magnetic fields in science and medicine. Elsevier, New York (1989).
Reynolds, J. H., & Chelazzi, L. (2004). Attentional modulation of visual processing. Annual Review Neuroscience, 27, 611-647.
Rihs, T. A., Michel, C. M., & Thut, G. (2009). A bias for posterior α-band power suppression versus enhancement during shifting versus
maintenance of spatial
attention. Neuroimage, 44, 190-199.
Rosanova, M., Casali, A., Bellina, V., Resta, F., Mariotti, M., & Massimini, M. (2009). Natural frequencies of human corticothalamic circuits. The Journal of Neuroscience, 29, 7679-7685.
Sauseng, P., Klimesch, W., Stadler, W., Schabus, M., Doppelmayr, M., Hanslmayr, S.,
et al. (2005). A shift of visual spatial attention
is selectively associated with human EEG alpha activity. European Journal of Neuroscience, 22, 2917-2926.
Silberstein, R. B., & Pipingas, A. (1995). Steady-state visually evoked potential topography during the Wisconsin card sorting test. Electroencephalography and Clinical
Neurophysiology/Evoked Potentials Section, 96, 24-35.
Spaak, E., de Lange, F. P., & Jensen, O. (2014). Local entrainment of alpha oscillations by visual stimuli causes cyclic modulation of perception. The Journal of Neuroscience, 34, 3536-3544.
Teng, F., Chen, Y., Choong, A. M., Gustafson, S., Reichley, C., Lawhead, P., & Waddell, D. (2011). Square or sine: Finding a waveform with high success rate of eliciting SSVEP.
Computational intelligence and neuroscience, 2011, 2.
Thorpe, S., Deng, S., Garcia, J., Lee, R., Huang, M., & Srinivasan, R. (2011). Spatial attention enhances steady state visual evoked responses in the gamma band. International
16 Thut, G., Schyns, P., & Gross, J. (2011).
Entrainment of perceptually relevant brain oscillations by non-invasive rhythmic stimulation of the human brain. Frontiers in
psychology, 2, 170.
Toffanin, P., de Jong, R., Johnson, A., & Martens, S. (2009). Using frequency tagging to quantify attentional deployment in a visual divided attention task. International Journal of
Psychophysiology, 72, 289-298.
Tononi, G., Srinivasan, R., Russell, D. P., & Edelman, G. M. (1998). Investigating neural correlates of conscious perception by frequency-tagged neuromagnetic responses. Proceedings of the National Academy of Sciences, 95, 3198-3203.
Vialatte, F. B., Maurice, M., Dauwels, J., & Cichocki, A. (2008). Steady state visual evoked potentials in the delta range (0.5-5 Hz). In
International Conference on Neural Information Processing, Springer Berlin Heidelberg, 400-407.
Vialatte, F. B., Maurice, M., Dauwels, J., & Cichocki, A. (2010). Steady-state visually evoked potentials: focus on essential paradigms and future perspectives. Progress
in neurobiology, 90, 418-438.
Watson, A. B., & Pelli, D. G. (1983). QUEST: A Bayesian adaptive psychometric method. Perception & psychophysics, 33, 113-120.
17 Supplementary materials
Figure S1a. Number of blue flashes. At the left side; targets (blue flashes in attended side) and at the right side distractors (blue flashes in unattended side). The probability of blue flashes was drawn from a gamma distribution (k: 7 and θ: 0.35).
Figure S1b. Timing of blue flashes. For maximal unpredictability of the next presentation, the timing of the blue flashes was drawn from three separate truncated negative exponential distributions.
