Research Articles, Systems/Circuits

Local Interactions between Steady-State Visually Evoked Potentials at Nearby Flickering Frequencies

Kumari Liza and Supratim Ray
Journal of Neuroscience 11 May 2022, 42 (19) 3965-3974; https://doi.org/10.1523/JNEUROSCI.0180-22.2022

Abstract

Steady-state visually evoked potentials (SSVEPs) are widely used to index top-down cognitive processing in human electroencephalogram (EEG) studies. Typically, two stimuli flickering at different temporal frequencies (TFs) are presented, each producing a distinct response in the EEG at its flicker frequency. However, how SSVEP responses in EEGs are modulated in the presence of a competing flickering stimulus just because of sensory interactions is not well understood. We have previously shown in local field potentials (LFPs) recorded from awake monkeys that when two overlapping full-screen gratings are counterphased at different TFs, there is an asymmetric SSVEP response suppression, with greater suppression from lower TFs, which further depends on the relative orientations of the gratings (stronger suppression and asymmetry for parallel compared with orthogonal gratings). Here, we first confirmed these effects in both male and female human EEG recordings. Then, we mapped the response suppression of one stimulus (target) by a competing stimulus (mask) over a much wider range than the previous study. Surprisingly, we found that the suppression was not stronger at low frequencies in general, but systematically varied depending on the target TF, indicating local interactions between the two competing stimuli. These results were confirmed in both human EEG and monkey LFP and electrocorticogram (ECoG) data. Our results show that sensory interactions between multiple SSVEPs are more complex than shown previously and are influenced by both local and global factors, underscoring the need to cautiously interpret the results of studies involving SSVEP paradigms.

SIGNIFICANCE STATEMENT Steady-state visually evoked potentials (SSVEPs) are extensively used in human cognitive studies and brain–computer interfacing applications where multiple stimuli flickering at distinct frequencies are concurrently presented in the visual field. We recently characterized interactions between competing flickering stimuli in animal recordings and found that stimuli flickering slowly produce larger suppression. Here, we confirmed these in human EEGs, and further characterized the interactions by using a much wider range of target and competing (mask) frequencies in both human EEGs and invasive animal recordings. These revealed a new “local” component, whereby the suppression increased when competing stimuli flickered at nearby frequencies. Our results highlight the complexity of sensory interactions among multiple SSVEPs and underscore the need to cautiously interpret studies involving SSVEP paradigms.

Introduction

Steady-state visually evoked potentials (SSVEPs) are stimulus-locked oscillatory electrical signals generated in response to a temporally periodic visual stimulus (Regan and Regan, 1988; Regan, 1989) with the response frequency strictly following the input stimulation frequency. SSVEP amplitude and phase are stable over time, have a high signal-to-noise ratio, and are relatively immune to artifacts (Vialatte et al., 2010; Norcia et al., 2015). Apart from the response at the stimulation frequency, harmonic and intermodulation (IM) components are also obtained (Zemon and Ratliff, 1984), which could convey important information about the nonlinear neural interactions in the visual system. In a noninvasive approach like the electroencephalogram (EEG), which captures the synchronous activity of many neurons, it is often difficult to isolate the responses because of the presentation of multiple stimuli. By flickering the stimuli at different temporal frequencies (TFs), SSVEPs present a convenient way to segregate the stimulus-specific responses. Many cognitive paradigms such as visual attention (Ding et al., 2006; Müller et al., 2006; Toffanin et al., 2009; Andersen et al., 2012), binocular rivalry (Wang et al., 2004), and working memory (Silberstein et al., 2001) extensively use multiple stimuli tagged with distinct frequencies while SSVEP responses are measured. But the interactions of SSVEP responses because of multiple concurrently presented flickering stimuli have not been thoroughly investigated.

In masking studies, two stimuli are simultaneously presented in the same visual space, leading to neural interactions between the competing stimuli (Legge and Foley, 1980). The response of the one grating, often called target grating, is attenuated because of another mask grating, even when the second grating fails to elicit any response when presented alone (Morrone et al., 1982; Foley, 1994; Boynton and Foley, 1999; Candy et al., 2001; Tsai et al., 2012). Such effects have often been explained using a normalization model (Heeger, 1992; Carandini and Heeger, 1994; Carandini et al., 1997) and provide a framework to study interactions between two competing flickering stimuli.

We recently studied interactions between SSVEPs in local field potential (LFP) signals recorded from the primary visual cortex (V1) of awake macaques (Salelkar and Ray, 2020); the possible interactions and previous findings are shown in Figure 1. The SSVEP response or “gain” profile (Fig. 1A) shows that stimuli flickering at ∼10 Hz produced the strongest response. According to the normalization model, this TF should produce the strongest suppression to a competing (target) TF if the normalization strength only depended on SSVEP gain (Fig. 1D; SSVEP gain specific). By presenting the target at 16 Hz and masks at nearby frequencies (Fig. 1, magenta circles), we confirmed this asymmetric suppression in which lower mask TFs produced greater suppression (Fig. 1D), but only when the constituent gratings were parallel (for orthogonal gratings, the suppression was nonspecific; Fig. 1C). However, this can also be explained in a model in which normalization is stronger at lower frequencies (Fig. 1E), potentially because of low-pass filtering in the normalization pool (Tsai et al., 2012). To distinguish these two possibilities, we presented the target at 8 Hz, for which the two hypotheses have opposite predictions, and found evidence in favor of low-frequency suppression (Fig. 1E).

Figure 1.
Figure 1.

Different mechanisms of target and mask frequency interactions. A, Averaged ECoG data from monkey 1 when a single counterphase grating presented at 25% contrast and varying temporal frequency was used to obtain the SSVEP response function. Curves indicate a polynomial fit (degree = 3). Cyan and magenta dots represent the two target frequencies used in the main study. B, Difference between the average SSVEP amplitudes at 9, 11, and 13 Hz, and 1, 3, and 5 Hz (marked as asterisks), under different hypotheses, as discussed in C–F. Cyan and magenta bars correspond to the conditions where the target frequency was 7 and 15 Hz, respectively. C, Nonspecific interaction: cyan and magenta curves depict 7 and 15 Hz target frequency responses as a function of the temporal frequency of mask grating. The green curve is the underlying suppression function, which is a horizontal line indicating that the suppression of target SSVEP response is independent of the mask frequency. Solid dots are the different mask frequencies used in this study, whereas the circled data are the frequencies used in our previous study (we had actually used 8 and 16 Hz target frequencies in the previous study, but shifted to 7 and 15 Hz). D, SSVEP gain-specific interaction: the suppression function is same as the SSVEP response function, indicating the strongest suppression for the temporal frequency range that elicits the maximum SSVEP response, which is ∼10 Hz for both the target frequencies (7 Hz, cyan curve; 15 Hz, magenta curve). E, Low-frequency tuned interaction: the suppression signal (green curve) has more strength at progressively lower frequencies, potentially because of a low-pass filtering action on the normalization signal. F, Target frequency-dependent interaction: the suppression function varies with the target frequency and is maximum in the vicinity of the target frequency.

In this study, we first tested whether these results also hold in human EEG recordings. Further, we characterized the response suppression profile by presenting masks over a much larger frequency range (1–29 Hz; Fig. 1, dots). Surprisingly, the suppression profile was found to be gain specific (Fig. 1D), not low frequency (Fig. 1E), for the higher target frequency (Fig. 1, magenta traces). Note that in all these models, the suppression has the same “global” profile (Fig. 1, green traces), regardless of the target frequency. However, it is possible that the suppression profile itself depends on the target frequency, as shown in Figure 1F. By presenting the target at a lower frequency as well (blue trace), we found evidence in favor of this hypothesis, in both human EEG and monkey LFP and electrocorticogram (ECoG) recordings.

Materials and Methods

Human EEG recordings.

Three experiments were conducted with 10, 10, and 11 subjects (total subjects, 31; 14 females and 17 males) recruited from the student community of the Indian Institute of Science, Bangalore (mean age, ∼23.1 years; age range, 21–28 years). All experimental procedures were approved by the Institute Human Ethics Committee of the Indian Institute of Science. Informed consent was taken from all subjects before the start of the experiment, and monetary compensation was provided for their voluntary participation. EEG signals were recorded from eight active electrodes (actiCAP) using the BrainAmp DC EEG acquisition system (Brain Products). The electrodes were placed in the parieto-occipital and occipital areas based on the international 10–10 system. The electrodes used were PO7, PO3, POz, PO4, PO8, O2, Oz, and O1. Raw signals were filtered between 0.016 Hz (first-order filter) and 1000 Hz (fifth-order Butterworth filter), sampled at 2500 Hz, and digitized at 16 bit resolution (0.1 µV/bit). The reference electrode was at FCz. The impedance was maintained at <20 kΩ throughout the recording session.

Animal recordings.

Animal experiments were approved by the Institutional Animal Ethics Committee of the Indian Institute of Science and were conducted in accordance with the guidelines approved by the Committee for the Purpose of Control and Supervision of Experiments on Animals. Appropriate measures were taken during the experiment to minimize pain and discomfort to the animals. Two adult female bonnet monkeys (Macaca radiata; weight, 3.3 and 3.4 kg) were used in this study. A titanium head post was implanted over the frontal region under general anesthesia before training. Monkeys were then trained for passive visual fixation tasks. The first monkey was the same as Monkey 1 in our previous report on SSVEP interactions (Salelkar and Ray, 2020), but with a different “hybrid” array implanted on the other hemisphere that consisted of 81 (9 × 9) microelectrodes (Utah array: 1 mm long and 400 µm apart) and 9 (3 × 3) ECoG electrodes (Ad-Tech Medical Instrument Corporation). More details of this custom-made hybrid array, including its implantation, receptive field location, and electrode selection (five of nine ECoG electrodes) are described in our previous study (Dubey and Ray, 2019, their monkey 3). At the time of recording, the microelectrode array had stopped working, so only ECoG data from five electrodes were used for analysis. For the second monkey, a 10 × 10 microelectrode array grid (Utah array: 1 mm long and 400 µm apart) with 96 active platinum electrodes (Blackrock Microsystems) was implanted in the primary visual cortex of right cerebral hemisphere (centered at ∼12 mm lateral from midline and ∼10 mm rostral from occipital ridge). Reference wires were put over the dura near the recording sites. As in our previous studies, only electrodes with reliable and stable receptive field centers across days and with impedances between 250 and 2500 KΩ were used for analysis, yielding 20–21 electrodes. Raw signals were recorded using a 128-channel Cerebus Neural Signal Processor (Blackrock Microsystems), bandpass filtered between 0.3 Hz (Butterworth filter, first order, analog) and 500 Hz (Butterworth filter, fourth order, and digital), sampled at 2000 Hz, and digitized at 16 bit resolution.

Experimental setup.

Visual stimuli were presented using an LCD monitor (1280 × 720 resolution, 100 Hz refresh rate; model XL2411, BenQ), gamma corrected, and calibrated to a mean luminance of 60 cd/m2. Human subjects sat in front of the monitor in a dark place covered by black curtains, with their head movement restricted using a chin rest placed in front of them. Monkeys sat on a primate chair with their head restrained inside a Faraday cage to reduce any external electrical noise. Human subjects viewed the monitor from 58 cm, such that the full-screen gratings covered the width and height of 46.8° and 27.2° of visual field. Monkeys viewed the monitor from 50 cm, such that the gratings covered 56° × 33° of visual field. Humans and monkeys were required to hold fixation within 2.5° and 2°, respectively, with a small spot of 0.1° in the center; trials in which fixation was broken were aborted immediately. Eyes were tracked using the EyeLink 1000 (sampled at 1000 Hz; SR Research) for humans and the ETL-200 Primate Eye Tracking System (sampled at 200 Hz; ISCAN) for monkeys. In human recordings, each trial began with the onset of fixation where they were required to hold their gaze for 2000 ms, after which a series of two to three stimuli appeared for 2500, 800, and 1500 ms with interstimulus intervals of 2500, 700, and 1500 ms in the three experiments, respectively (Movies 1, 2, 3). Each stimulus was either a grating or a plaid stimulus, as explained below. In monkey recordings, each trial started with the onset of fixation where they were required to hold their gaze for 2000 ms, after which two stimuli appeared for 1500 ms with interstimulus intervals of 1500 ms. Monkeys were rewarded with a drop of juice after every successful trial.

Movie 1.

Parallel condition: target grating is at 100% contrast, 4 cycles/° spatial frequency, 0° orientation, and 15 Hz temporal frequency. Mask grating is at 100% contrast, 4 cycles/° spatial frequency, 0° orientation, and 11 Hz temporal frequency.

Movie 2.

Orthogonal condition: Same as Movie 1 except that mask grating is at 90° orientation.

Movie 3.

One trial with three stimuli: target grating parameters are the same as in Movie 1. The orientation and temporal frequency of the three mask gratings are as follows: 0°, 11 Hz; 90°, 13 Hz; and 90°, 15 Hz. Each stimulus is presented for 800 ms, with 700 ms as the interstimulus duration.

Visual stimuli.

In all three experiments, two fully overlapping full-screen counterphase gratings were used to generate a plaid stimulus. Each of the constituent gratings could have a different orientation, contrast, and temporal frequency. If one of the constituent gratings had 0% contrast, the plaid stimulus was reduced to a grating stimulus.

Experiment 1 was aimed to replicate the findings of our previous study (Salelkar and Ray, 2020) in human EEG recordings. Here, the target grating was presented at an orientation of either 0° or 90°, with contrast of 0 or 50% and temporal frequency of 16 Hz. The mask grating was presented at a range of different orientations (n = 4; 0°, 30°, 60°, 90°), 50% contrast, and seven different temporal frequencies (10, 12, 14, 16, 18, 20, 22 Hz). When the target grating was absent (0% contrast), the SSVEP response was only because of the mask grating, which allowed us to get the SSVEP gain function (Fig. 1A). The spatial frequency of both target and mask grating was fixed at 2 cycles/°. The total number of stimulus conditions in experiment 1 was 56 (two target contrasts × four mask orientations × seven mask temporal frequencies). Each stimulus condition on average had ∼10 repeats for each subject.

In experiment 2, the temporal frequency of target grating was fixed at 15 Hz and the temporal frequency of mask grating ranged from 1 to 29 Hz (n = 15; 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27 and 29 Hz). We shifted to 15 Hz because it has fewer subharmonics as compared with 16 Hz, although none of the results depended on the choice of target frequency. The contrast, orientation and spatial frequency of target/mask grating were same as experiment 1. The total number of stimulus conditions in experiment 2 was 120 (2 target contrasts × 4 mask orientations × 15 mask temporal frequencies). Each stimulus condition on average had 10 repeats. Because of the large number of stimulus conditions, recordings were performed in two sessions, each lasting 2 h on average; stimulus repeats were pooled across the session for analysis.

Experiment 3 was conducted on both human and monkeys. Here, multiple target stimuli were used. To reduce the number of conditions, the number of mask orientations was reduced to two (parallel and orthogonal). Target grating was presented at 7 and 15 Hz in human recordings and at 7, 11, and 15 Hz in monkey recordings (recorded in three separate sessions, each with one of the three target frequencies). Mask grating was presented at 15 temporal frequencies (1–29 Hz; resolution, 2 Hz) in human recordings and at 29 temporal frequencies (1–29 Hz; resolution, 1 Hz) in monkey recordings. The contrast of the target/mask grating was 50% in human recordings and 25% in monkey recordings (to remain consistent with the previous study). Target grating orientation was fixed at 0°, and mask grating was fixed at 0° and 90°. The spatial frequency of target/mask grating was fixed at 4 cycles/°. In human recordings, we performed two back-to-back sets of recordings. First, we used one of the target frequencies and presented it at either 0% or 50%, yielding 60 stimulus conditions (2 target contrasts × 2 mask orientations × 15 mask temporal frequencies). Subsequently, we presented the second target frequency at 50% contrast, yielding 30 conditions (1 target contrast × 2 mask orientations × 15 mask temporal frequencies). This is because the 0% target contrast condition, which yielded the SSVEP gain function, could be estimated from the first set itself. For monkey recordings, we used 1 target frequency in each session, yielding 116 stimulus conditions (2 target contrasts × 2 mask orientations × 29 mask temporal frequencies) per session. Each stimulus condition on average was repeated 10, 10, and 15 times for human, monkey 1, and monkey 2, respectively.

Data analysis.

Data were analyzed using custom codes written in MATLAB (MathWorks). In human recordings, baseline period was defined as −500 ms to 0 (with 0 being the stimulus onset time), and the stimulus period was defined as 250–750 ms, yielding a frequency resolution of 2 Hz. This duration was chosen to reduce stimulus onset-related transients (for review of time–frequency power spectra produced by the presentation of a 16 Hz counterphasing stimulus recorded in human EEG, see Murty et al., 2020, their Fig. 6; Murty et al., 2021, their Fig. 8). Note that since counterphasing stimuli produced a response at twice the stimulus frequency, the response was always at an even frequency (e.g., a 1 Hz stimulus produced a peak at 2 Hz), and therefore 2 Hz resolution was sufficient to capture the response. Further, analyzing between 250 and 1250 ms for conditions in which stimuli were presented for 1500 or 2500 ms (experiments 1 and 3) yielded very similar results (data not shown). In monkey recordings, the baseline period was −1000 ms to 0 and the stimulus period was 250–1250 ms, yielding a frequency resolution of 1 Hz.

EEG data were subjected to a series of processing pipelines before using it for further analysis. First, bad stimulus repeats were removed using the pipeline described in our previous article (Murty et al., 2020, artifact rejection) based on which ∼20% of stimulus repeats were rejected. In this pipeline, any stimulus repeat in which either the raw waveform or power spectral density (PSD) was >6 standard deviations away from their mean across all repeats was tentatively labeled “bad.” Next, electrodes with >30% of all repeats labeled as bad were marked as bad electrodes and removed from analysis. A stimulus repeat was then marked bad when it was present in >10% of the remaining electrodes. Finally, any electrode for which the mean PSD had a slope <0 between 56 and 84 Hz in the baseline period (indicating a flat PSD) was also discarded. This procedure yielded a set of good electrodes for each subject and a set of good stimulus repeats that were common for all good electrodes.

Second, subjects or sessions with fewer than two repeats for each stimulus condition were also removed. Third, we computed SSVEP amplitude spectra by taking the fast Fourier transform of the averaged signal across all the repeats deemed good for a stimulus condition (equivalent to averaging the complex Fourier spectra across stimulus repeats) and rejected electrodes with target SSVEP responses of <0.5 µV in EEG and 2.5 µV in ECoG and LFP. Finally, human subjects with fewer than two good electrodes (of eight) were rejected. Overall, this yielded 8 (of 10), 8 (of 10), and 7 (of 11) good subjects for the three experiments, respectively. The amplitudes of the Fourier spectra were then averaged across subjects. In monkey recordings (experiment 3), we obtained usable data from 2–5 ECoG electrodes in the three sessions (one for each target frequency) from monkey 1 and 20–21 LFP electrodes from monkey 2 in all sessions. The counterphasing stimuli used in this study produced a prominent response at twice the stimulus frequency, since neurons respond to luminance changes in either the positive or negative direction (two responses per stimulus cycle). The amplitude difference was therefore calculated as AmpST(2*f)−AmpBL(2*f), where AmpST(2*f) and AmpBL(2*f) are the amplitudes at twice the target/mask frequency in the stimulus (ST) and baseline (BL) period, respectively.

Statistical analysis.

One-sample t tests were performed to quantify the significance level of differences in high- and low-frequency responses. We also calculated the Bayes factor (BF; https://klabhub.github.io/bayesFactor/) for one-sample t tests in our dataset, which is the ratio of the likelihood of an alternate hypothesis to the likelihood of a null hypothesis (Jarosz and Wiley, 2014). The magnitude of BF indicates the strength of evidence in favor of the alternate hypothesis: BF values between 1 and 3 suggest anecdotal evidence, values between 3 and 10 provide substantial evidence, and values >10 provide strong evidence (Jeffreys, 1998).

Results

Asymmetric suppression shown previously in monkey recordings is also present in human EEG

We first tested whether the asymmetric suppression observed in macaque LFP responses (Salelkar and Ray, 2020) is also present in human EEGs. We recorded EEGs from parieto-occipital and occipital electrodes of ten healthy young adults (five females). As in our previous study, we presented two overlapping gratings (i.e., a plaid stimulus), with both constituent gratings counterphasing at different frequencies. One grating, which we call the “target,” always counterphased at 16 Hz and produced a strong SSVEP response at 32 Hz. The orientation of this target grating was either 0° (5 of 10 subjects) or 90° (remaining subjects). The second gratings (the “mask”) had variable temporal frequency (n =7; 10, 12, 14, 16, 18, 20, 22 Hz) and orientation (n = 4; 0°, 30°, 60°, 90°). The contrast of the mask frequency was fixed at 50%, while the target grating was either 0% or 50%.

Figure 2A shows the amplitude spectra of EEG responses of one subject, averaged over all electrodes that qualified the thresholding criteria (N = 8; for details, see Materials and Methods). The response of the plaid was evaluated as the change in amplitude between grating-only and plaid conditions (Fig. 2A, blue and color traces) at 32 Hz (Fig. 2B). These were generally negative, indicating target frequency SSVEP suppression by the mask. Positive responses were observed when both gratings had the same temporal frequency. In the parallel condition, we observed substantially stronger suppression of the amplitude response of plaids with mask frequencies <16 Hz target (e.g., compare 10 Hz mask frequency; row 1 in Fig. 2A) versus more (22 Hz; row 7 in Fig. 2A). We quantified the asymmetry by subtracting the average response of plaids with mask frequencies less than the target (10, 12, and 14) from the greater ones (18, 20, and 22), which showed stronger asymmetry in the parallel condition compared with orthogonal conditions (Fig. 2C), although the difference was significant at all orientations (N = 8; parallel: BF = 12.8, p = 0.004; 30° separation: BF = 52.5, p = 0.0007; 60° separation: BF = 208.8, p = 0.0001; orthogonal: BF = 5.4, p = 0.013).

Figure 2.
Figure 2.

EEG amplitude response suppression of subject 1. A, The amplitude spectra are averaged over the good electrodes (N = 8). Each row indicates a different temporal frequency of mask grating (indicated at the extreme right of the fourth column) that is superimposed with a grating with a temporal frequency of 16 Hz (the target grating); the columns show different orientations of mask grating relative to the target grating. The blue trace is the 16 Hz “grating-only” condition (same trace in all plots). The number in the inset is the difference in amplitude at the target frequency (i.e., 32 Hz) between the grating-only condition and the plaid condition (i.e., between the blue traces and traces of other colors at 32 Hz). B, Difference between the grating-only condition and the plaid condition, as described above, plotted as a function of mask frequency for different orientations. C, Average difference between amplitudes above (18, 20, and 22 Hz) and below (10, 12, and 14 Hz) the target frequency (16 Hz), for different orientation difference conditions. The error bars indicate the SEM along with the significance level. ***p < 0.001; **p < 0.01; *p < 0.05.

Results were consistent for averaged response across all good subjects (N = 8; Fig. 3A,B). Because these conditions were also tested in experiment 2 (except for a shift in the target frequency from 16 to 15 Hz), we combined the subjects from the two experiments (N = 16; Fig. 3C). Asymmetry in suppression decreased with orientation dissimilarity, although remained significant in all conditions (N = 16; parallel: BF = 35.1, p = 0.0011; 30° separation: BF = 20.0, p = 0.0021; 60° separation: BF = 215.2, p = 0.0001; orthogonal: BF = 8.9, p = 0.0056). These results are consistent with our previous study (Salelkar and Ray, 2020), and are consistent with both the gain-specific and low frequency-specific hypotheses (Fig. 1D,E).

Figure 3.
Figure 3.

Averaged EEG amplitude response suppression of all subjects. A, B, Same as Figure 2, A and B, but for data averaged over all usable subjects (N = 8). C, Same as B, but after adding data from eight good subjects who participated in experiment 2. The amplitude response is plotted as a function of delta frequency (mask frequency – target frequency), since for the second experiment the target frequency was 15 Hz. D, Same as Figure 2C, but for data averaged across 16 subjects.

Target frequency suppression is not low frequency tuned

In our previous study (Salelkar and Ray, 2020), we dissociated the two hypotheses by presenting the target at a different frequency (8 Hz). Here, we tested a different prediction. While keeping the target frequency fixed, if we increase the range of mask frequencies, the two hypotheses predict different behavior at low mask frequencies (Fig. 1, magenta traces, compare D, E). In experiment 2, we presented two overlapping counterphase gratings with target grating at 15 Hz at 0° (5 of 10 subjects) or 90° (remaining subjects) orientation and varied the temporal frequency (n =15; 1–29 Hz in steps of 2 Hz) and orientation (n = 4; 0°, 30°, 60°, 90°) of the mask grating. Human EEG was recorded from parieto-occipital and occipital electrodes of ten healthy young adults (five females).

Figure 4 depicts the averaged amplitude spectra of eight good subjects (same format as in Figs. 2, 3). In the orthogonal condition, symmetric suppression was observed (nonspecific hypothesis; Fig. 1C). Surprisingly, in the parallel condition, the suppression profile was more similar to the SSVEP gain-specific hypothesis (Fig. 1D) than the low frequency-specific hypothesis (Fig. 1E), which is inconsistent with our previous results (Salelkar and Ray, 2020). To quantify these results, we computed the difference between the mean amplitude response for mask frequencies 1, 3, and 5 Hz and the mean amplitude response for mask frequencies of 9, 11, and 13 Hz (these data points are shown as asterisks in Fig. 1, and as thick filled circles in Fig. 4B). This quantity is expected to be near zero for nonspecific suppression, negative for SSVEP-gain-specific hypothesis and positive for low-frequency hypothesis (Fig. 1B, magenta bars). The difference was significantly negative for parallel and 30° orientation difference (N = 8; parallel: BF = 4.6, p = 0.015; 30° separation: BF = 3.9, p = 0.019; 60° separation: BF = 1.4, p = 0.074; orthogonal: BF = 0.50, p = 0.35), consistent with the SSVEP gain-specific hypothesis.

Figure 4.
Figure 4.

Averaged EEG amplitude response suppression of all subjects for a wider range of mask frequencies. A, B, Same format as in Figure 3, A and B, but for eight usable subjects who participated in experiment 2, which had a wider range of mask frequencies and a target frequency of 15 Hz. C, The average difference between responses obtained for plaids with mask frequencies of 9, 11, and 13 Hz and mask frequencies of 1, 3, and 5 Hz (as indicated in solid circles; Fig. 1B magenta bars shows this value under different hypotheses).

Suppression profile is dependent on the target frequency

The three hypotheses described in Figure 1C–E all have a global suppressive function, as shown in green traces in Figure 1. Another hypothesis, not tested previously, is that the suppression also has a “local” component, arising from local interactions between target and mask frequencies. This would lead to different suppression profiles that are dependent on the target frequency, as shown in Figure 1F, and could reconcile our previous results.

We tested this prediction in both monkey and human recordings, including one monkey who was part of the previous study (monkey 1; only ECoG data could be recorded because the microelectrode array had stopped working). We used two overlapping full-screen counterphase gratings with target frequencies set to 7, 11, or 15 Hz (7 and 15 Hz for humans) and mask frequencies that varied from 1 to 29 Hz (n = 29 for monkeys; n = 15 for humans; for more details, see Materials and Methods).

Figure 5A shows the amplitude response as a function of mask frequency for the two monkeys (Fig. 5A, columns 1 and 2) and human EEG, in the parallel condition. Consistent with the new hypothesis, the suppression profile changed systematically depending on the target frequency. The difference between the responses from plaids with mask frequencies higher versus lower than 7 Hz (i.e., average between 8–13 and 1–6 Hz; shown in solid circles in Fig. 5A,B) was positive when the target was at 7 Hz (Fig. 5C, blue traces), consistent with our previous report (Salelkar and Ray, 2020), but became progressively negative as the target frequency shifted to 15 Hz. This effect was largely abolished in the orthogonal condition (Fig. 5B,C). To quantify this, we computed the difference between the values for 7 and 15 Hz (Fig. 5C, cyan and magenta bars) and tested whether this difference was significantly greater than zero using a one-tailed one-sample t test. We used common electrodes across 7 and 15 Hz target frequencies, yielding 2 and 19 electrodes for monkey 1 ECoG and monkey 2 LFP, respectively. The difference was highly significant for the parallel condition (monkey 1 ECoG: N = 2, BF = 8.06, p = 0.003; monkey 2 LFP: N = 19, BF = 2.37 × 106, p = 5.23 × 10−9) and human EEG (N = 7, BF = 5.14, p = 0.017). However, the difference was not salient for the orthogonal condition (monkey 1 ECoG: N = 2, BF = 0.59, p = 0.44; monkey 2 LFP: N = 19, BF = 0.60, p = 0.20) and human EEG (N = 7, BF = 1.0, p = 0.15).

Figure 5.
Figure 5.

Suppression profile for different target frequencies. A, SSVEP amplitude suppression for target frequencies 7, 11, and 15 Hz in monkey recordings, and 7 and 15 Hz in human recordings as a function of mask temporal frequency in the parallel condition. B, Same as A, but for the orthogonal condition. C, For 7 and 15 Hz target frequencies, the average difference between responses obtained for plaids with mask frequencies of 8–13 Hz and mask frequencies of 1–6 Hz in monkey recordings. In human recordings, the average difference between responses obtained for plaids with mask frequencies of 9, 11, and 13 Hz and mask frequencies of 1, 3, and 5 Hz. For 11 Hz target frequency in monkey recordings, the average difference between responses obtained for plaids with mask frequencies of 8–14 Hz (except at 11 Hz) and with mask frequencies of 1–6 Hz (as indicated in solid circles). M1: monkey 1; M2: monkey 2.

Discussion

We used two overlapping counterphasing gratings at varying relative orientations and temporal frequencies to comprehensively map the interactions between the two SSVEPs in human EEG. We confirmed that the low-frequency suppression (in which lower temporal frequencies cause larger suppression) for parallel but not orthogonal gratings, as described recently in monkey LFPs (Salelkar and Ray, 2020), was also observed in human EEGs. However, this low-frequency suppression gradually diminished as the difference between the target and mask frequency increased, inconsistent with the hypothesis proposed previously (Salelkar and Ray, 2020). Instead, we found that the overall suppression profile was dependent on the target frequency in both human EEG and monkey LFP/ECoG data, which reconciled previous results. Together, these results provide a comprehensive account of the suppressive effects of two SSVEP tags along temporal frequency and relative orientation domains.

Previous studies

Previous psychophysical studies measuring contrast sensitivity as a function of mask frequency provide important clues about the mechanisms that can potentially explain our results. These studies have reported the presence of at least two temporal frequency masking channels in the visual system: low pass and bandpass (Anderson and Burr, 1985; Hess and Snowden, 1992; Boynton and Foley, 1999; Cass and Alais, 2006). The frequency cutoff of the low-pass masking channel was observed to be 10 Hz. The bandpass masking channel response was observed in the frequency range of 7–13 Hz with a peak at ∼10 Hz, which closely resembled our SSVEP response function or the global profile (Fig. 1A). The two channels might interact with each other and engage different inhibitory mechanisms, depending on the target frequency. Low and high target frequency have shown to involve low-pass and bandpass masking mechanisms, respectively (Anderson and Burr, 1985; Cass and Alais, 2006). In our experiments with parallel masks, for a target frequency of 15 Hz, the suppression profile was indeed bandpass and similar to the global profile, while it was low pass for 7 Hz target frequency, consistent with the idea that different masking mechanisms were involved at different target frequencies. However, not all results in these psychophysics studies were consistent with our results. For example, some studies have suggested asymmetric suppression where high mask frequency suppressed low target frequency but not vice versa (Allison et al., 2001; Cass and Alais, 2006), but we observed asymmetry in the opposite direction. It should be noted that there are important differences between the stimuli used: we used counterphasing gratings to generate SSVEPs, while many of these previous studies used drifting gratings (a counterphase grating can be decomposed into two drifting gratings moving in opposite directions). Further, these studies measured behavior (e.g., masking threshold), as opposed to a neural measure (SSVEP amplitude) reported here. If some of these temporal frequency channels have noncortical origins, then the neural responses from a cortical area may differ from behavior. For example, Cass and Alais (2006) suggested that bandpass channel is precortical, while the low-frequency channel has cortical origins.

In previous studies with EEG recordings, Regan (1983) studied SSVEP responses while presenting target and mask gratings at varying spatial and temporal frequencies. The target response was heavily suppressed by masks that had similar spatial frequency as the target, indicating the presence of multiple narrow-bandwidth spatial frequency-tuned mechanisms (Fiorentini et al., 1983), consistent with psychophysics as well (Anderson and Burr, 1985). For temporal frequency, the interaction (recorded from a single human subject) was consistent with our results, so our findings generalize these results to a larger pool of subjects and over multiple scales of electrophysiological signals. In addition, that study did not report the effect of the relative orientation of the two gratings. Evidence of asymmetrical suppression in the spatial frequency domain has been shown by Legge (1979, their Figs. 6, 7). An SSVEP study that looked at the effect of orthogonal versus parallel masks found stronger suppression with orthogonal masks, contrary to our results (Burr and Morrone, 1987). However, this study used a variety of different spatial frequencies and contrasts as well, while we used a fixed spatial frequency and contrast. Given the dependence of masking strength on such parameters, it is difficult to directly compare these results.

More recently, masking phenomena have been explained using normalization, a neural computation observed throughout the brain describing response modulation because of multiple stimuli (Carandini and Heeger, 2011). According to the model, the excitatory drive because of one stimulus is normalized by inputs from the surrounding neuronal population having broader tuning properties (Heeger, 1992; Carandini and Heeger, 1994; Carandini et al., 1997). The normalization model initially proposed for single-unit recordings has also been used to explain temporal frequency masking effects in human EEG studies (Boynton and Foley, 1999; Candy et al., 2001; Tsai et al., 2012). These models could represent target, mask, as well as IM response profiles as a function of target/mask contrasts by manipulations in the way excitatory and inhibitory inputs were summed and passed through simple nonlinearities (e.g., rectification or squaring). We also recently used a tuned normalization model to capture the signal contributing to target SSVEP response suppression as a function of mask frequency (Salelkar and Ray, 2020). The low-frequency suppression was modeled by essentially assuming a particular type of suppression function [Fig. 1C–E, green traces (assumed to be a downward sloping linear function); Salelkar and Ray, 2020].

Neural mechanisms underlying local interactions

In the previous article (Salelkar and Ray, 2020), we discussed a potential reason that could lead to such a low-frequency suppression profile (Fig. 1E), and here we further speculate how some changes in this model can explain the local interactions between target and mask gratings as reported in this article (Fig. 1F). To explain the low-frequency suppression profile, we first consider a variant of the normalization model proposed previously (Foley, 1994; Baker and Wade, 2017). In these models, two input sinusoids each for target and mask frequency (sinωTt and sinωMt) act as the excitatory drive, which is the numerator of the response. The denominator (normalization) part of the model is the inhibitory drive, which is the combined responses of target and mask frequency after passing through some nonlinearity. For example, consider a response of the following form: R=(sinωTt + sinωMt)2/[sin(ωTt)2 + sin(ωMt)2+σ]. By adding a squaring nonlinearity in the numerator, the response has strong responses not only at 2ωT and 2ωM but also at the intermodulation terms |ωT + ωM| and |ωT−ωM|, as observed in data (Figs. 2–4; Foley, 1994). Further, by changing whether the nonlinearity is applied before or after summation, or the exponent of the nonlinearity, the behavior of the model can be modified.

Simulating the response as per the model described above shows that the presence of the rapidly varying sinusoidal signals in the denominator lead to spurious peaks at many frequencies beyond ωT,ωM and their harmonics and sum/differences (Regan and Regan, 1988; Baker and Wade, 2017). However, such peaks were not observed in our dataset. This issue can be addressed by assuming that the normalization signal varies slowly over time, which can be achieved by low-pass filtering the normalization signal (Tsai et al., 2012). However, this procedure also makes the normalization strength stronger when the mask frequency is low, in effect generating a low-frequency suppression profile as shown in Figure 1E. A simple way to achieve target frequency-specific suppression (Fig. 1F) in this framework is to sum the sinusoids before squaring and low-pass filtering the normalization term (i.e., by assuming that the normalization signal is low-pass filter{(sinωTt+sinωMt)2+σ}). This is because now expanding the sinusoidal terms produces a |ωT−ωM| IM component, which has very low frequency when target and mask frequencies are close to each other and therefore is preserved after low-pass filtering, while other frequencies are filtered out. This model explains the strong suppression of target SSVEP response when mask frequency is in the vicinity of the target frequency, leading to different suppression profiles depending on the position of the target frequency (Fig. 1F). Further, the observed asymmetry around the target, where masks below the target frequency produce stronger suppression, can also be explained because slower mask frequencies in general get less attenuated by the low-pass filter compared with higher mask frequencies (this is the original low-frequency suppression argument discussed previously).

Although in our simulations with this model we were able to reproduce the target-specific suppression, we have refrained from developing a formal model here because the model is highly unconstrained (similar to our previous results; Salelkar and Ray, 2020). The model parameters are fitted for each target frequency separately (Salelkar and Ray, 2020), so, to adequately constrain such models, multiple contrast levels of mask and target frequencies are needed for each combination, as was done, for example, in the studies by Foley (1994) and Boynton and Foley (1999). In our experiments, we used only one contrast level of target and mask gratings because of the large number of stimulus conditions that led to long recording hours (∼3 h for experiment 3). We noticed that human SSVEP responses sometimes showed minor variations when recorded on different days, and therefore all recordings were performed in a single day. Along with multiple contrast levels for each target and mask frequency, multiple spatial frequencies also need to be tested to fully characterize the interactions between two flickering stimuli, for which monkey recordings with chronically implanted arrays may be the ideal recording platform. It would also be interesting to see how these results relate to the spiking unit activity. In the current monkey recordings, we used full-screen flickering stimuli to be comparable to the human EEG recordings, which led to a small number of stable spiking units in monkey 2.

Implications for attention studies

Insights about interactions between SSVEP tags are important in properly interpreting the results of attention studies, where two or more simultaneously presented stimuli are tagged with different flickering frequencies and subjects attend to one of the stimuli. Typically, attention increases the amplitude of the SSVEP of the attended stimulus and reduces the SSVEP of the unattended stimulus, consistent with the spotlight theory of attention (Morgan et al., 1996). Our results suggest that the reciprocal effect of attention on the attended versus unattended SSVEP frequencies could be because of normalization, whereby an increase in response strength at the attended location would increase the overall normalization signal, which in turn would reduce the response at the unattended location. More importantly, the magnitude of this attentional facilitation or suppression will critically depend on the two SSVEP frequencies as well as on the properties of the stimuli, such as their relative orientations.

However, we note that the stimulus interactions described here may be applicable only for a subset of attention studies. Attention studies often use small stimuli that occupy different locations in the visual field, either in the same or in opposite hemifields (Morgan et al., 1996; Störmer and Alvarez, 2014). It is unclear whether our results extend to these conditions, since stimuli need to be sufficiently close to each other to engage normalization mechanisms. Indeed, it is reassuring that even for fully overlapping stimuli, the suppression is largely nonspecific when the stimuli are well separated in feature space (here, orthogonal orientations). Our results are more relevant for feature attention studies that used fully overlapping random dot patterns (RDPs), with different stimuli shown in different colors (Müller et al., 2006). More studies are needed to determine whether the interactions observed with plaids are comparable to RDPs, and whether the usage of different orientations is comparable to the usage of different colors.

Although our study involved only a passive fixation task in which attention was uncontrolled and therefore could be on either one of the two gratings of the plaid, it is unlikely to influence our main results. This is because all the results shown in EEGs were also observed in LFP/ECoG recordings from V1 (Fig. 5), where the effect of attention is modest. In general, attention only leads to a small increase in the center frequency of gamma oscillations, and a modest reduction in gamma and alpha power (Chalk et al., 2010; Das and Ray, 2018). In a stimulus configuration very similar to ours, Lima et al. (2010) showed that gamma oscillations in V1 are attenuated by plaids compared with gratings. They trained one monkey to switch attention between the two components of the plaids and found only a small change in firing rates, gamma coherence, and gamma center frequency (Lima et al., 2010, their Fig. 10).

Apart from classic attention studies, SSVEP paradigms have become popular in brain–computer interface (BCI) applications, where two or more simultaneously presented stimuli are tagged with different flickering frequencies, and a relative change in SSVEP response when one stimulus is attended versus unattended is used to control an application (Ding et al., 2006). A better understanding of the interactions between multiple flickering stimuli at a neural level holds promise for improvement in such BCI applications and also enhances our understanding of how the visual system represents such competing stimuli.

Footnotes

  • This work was supported by Department of Biotechnology (DBT)/Wellcome Trust India Alliance (Grant IA/S/18/2/504003; Senior Fellowship to S.R.) and the DBT-IISc Partnership Program.

  • The authors declare no competing financial interests.

  • Correspondence should be addressed to Supratim Ray at sray{at}iisc.ac.in

References

  1. ↵
    2. Allison JD,
    3. Smith KR,
    4. Bonds AB
    (2001) Suppression in the cat striate cortex. Vis Neurosci 18:941–948. pmid:12020085
    OpenUrlPubMed
  2. ↵
    2. Andersen SK,
    3. Müller MM,
    4. Martinovic J
    (2012) Bottom-up biases in feature-selective attention. J Neurosci 32:16953–16958. pmid:23175846
    OpenUrlAbstract/FREE Full Text
  3. ↵
    2. Anderson SJ,
    3. Burr DC
    (1985) Spatial and temporal selectivity of the human motion detection system. Vision Res 25:1147–1154. pmid:4071994
    OpenUrlCrossRefPubMed
  4. ↵
    2. Baker DH,
    3. Wade AR
    (2017) Evidence for an optimal algorithm underlying signal combination in human visual cortex. Cereb Cortex 27:254–264. pmid:28031176
    OpenUrlCrossRefPubMed
  5. ↵
    2. Boynton GM,
    3. Foley JM
    (1999) Temporal sensitivity of human luminance pattern mechanisms determined by masking with temporally modulated stimuli. Vision Res 39:1641–1656. pmid:10343857
    OpenUrlCrossRefPubMed
  6. ↵
    2. Burr DC,
    3. Morrone MC
    (1987) Inhibitory interactions in the human vision system revealed in pattern-evoked potentials. J Physiol 389:1–21. pmid:3681721
    OpenUrlCrossRefPubMed
  7. ↵
    2. Candy TR,
    3. Skoczenski AM,
    4. Norcia AM
    (2001) Normalization models applied to orientation masking in the human infant. J Neurosci 21:4530–4541. doi:10.1523/JNEUROSCI.21-12-04530.2001
    OpenUrlAbstract/FREE Full Text
  8. ↵
    2. Carandini M,
    3. Heeger DJ
    (1994) Summation and division by neurons in primate visual cortex. Science 264:1333–1336. pmid:8191289
    OpenUrlAbstract/FREE Full Text
  9. ↵
    2. Carandini M,
    3. Heeger DJ
    (2011) Normalization as a canonical neural computation. Nat Rev Neurosci 13:51–62. pmid:22108672
    OpenUrlCrossRefPubMed
  10. ↵
    2. Carandini M,
    3. Heeger DJ,
    4. Movshon JA
    (1997) Linearity and normalization in simple cells of the macaque primary visual cortex. J Neurosci 17:8621–8644. pmid:9334433
    OpenUrlAbstract/FREE Full Text
  11. ↵
    2. Cass J,
    3. Alais D
    (2006) Evidence for two interacting temporal channels in human visual processing. Vision Res 46:2859–2868. pmid:16684555
    OpenUrlCrossRefPubMed
  12. ↵
    2. Chalk M,
    3. Herrero JL,
    4. Gieselmann MA,
    5. Delicato LS,
    6. Gotthardt S,
    7. Thiele A
    (2010) attention reduces stimulus-driven gamma frequency oscillations and spike field coherence in V1. Neuron 66:114–125. pmid:20399733
    OpenUrlCrossRefPubMed
  13. ↵
    2. Das A,
    3. Ray S
    (2018) Effect of stimulus contrast and visual attention on spike-gamma phase relationship in macaque primary visual cortex. Front Comput Neurosci 12:66. doi:10.3389/fncom.2018.00066
    OpenUrlCrossRef
  14. ↵
    2. Ding J,
    3. Sperling G,
    4. Srinivasan R
    (2006) Attentional modulation of SSVEP power depends on the network tagged by the flicker frequency. Cereb Cortex 16:1016–1029. doi:10.1093/cercor/bhj044
    OpenUrlCrossRefPubMed
  15. ↵
    2. Dubey A,
    3. Ray S
    (2019) Cortical electrocorticogram (ECoG) is a local signal. J Neurosci 39:4299–4311. doi:10.1523/JNEUROSCI.2917-18.2019 pmid:30914446
    OpenUrlAbstract/FREE Full Text
  16. ↵
    2. Fiorentini A,
    3. Pirchio M,
    4. Spinelli D
    (1983) Electrophysiological evidence for spatial frequency selective mechanisms in adults and infants. Vision Res 23:119–127. pmid:6868386
    OpenUrlCrossRefPubMed
  17. ↵
    2. Foley JM
    (1994) Human luminance pattern-vision mechanisms: masking experiments require a new model. J Opt Soc Am A Opt Image Sci Vis 11:1710–1719. doi:10.1364/josaa.11.001710 pmid:8046537
    OpenUrlCrossRefPubMed
  18. ↵
    2. Heeger DJ
    (1992) Normalization of cell responses in cat striate cortex. Vis Neurosci 9:181–197. doi:10.1017/s0952523800009640 pmid:1504027
    OpenUrlCrossRefPubMed
  19. ↵
    2. Hess RF,
    3. Snowden RJ
    (1992) Temporal properties of human visual filters: number, shapes and spatial covariation. Vision Res 32:47–59. pmid:1502811
    OpenUrlCrossRefPubMed
  20. ↵
    2. Jarosz AF,
    3. Wiley J
    (2014) What are the odds? A practical guide to computing and reporting bayes factors. J Probl Solving 7:2–9. doi:10.7771/1932-6246.1167
    OpenUrlCrossRef
  21. ↵
    2. Jeffreys H
    (1998) The theory of probability. Oxford, UK: Oxford UP.
  22. ↵
    2. Legge GE
    (1979) Spatial frequency masking in human vision: binocular interactions. J Opt Soc Am 69:838–847. doi:10.1364/josa.69.000838 pmid:490227
    OpenUrlCrossRefPubMed
  23. ↵
    2. Legge GE,
    3. Foley JM
    (1980) Contrast masking in human vision. J Opt Soc Am 70:1458–1471. pmid:7463185
    OpenUrlCrossRefPubMed
  24. ↵
    2. Lima B,
    3. Singer W,
    4. Chen N-H,
    5. Neuenschwander S
    (2010) Synchronization dynamics in response to plaid stimuli in monkey V1. Cereb Cortex 20:1556–1573. pmid:19812238
    OpenUrlCrossRefPubMed
  25. ↵
    2. Morgan ST,
    3. Hansen JC,
    4. Hillyard SA
    (1996) Selective attention to stimulus location modulates the steady-state visual evoked potential. Proc Natl Acad Sci U|S|A 93:4770–4774. doi:10.1073/pnas.93.10.4770
    OpenUrlAbstract/FREE Full Text
  26. ↵
    2. Morrone MC,
    3. Burr DC,
    4. Maffei L
    (1982) Functional implications of cross-orientation inhibition of cortical visual cells. I. Neurophysiological evidence. Proc R Soc Lond B Biol Sci 216:335–354. pmid:6129633
    OpenUrlCrossRefPubMed
  27. ↵
    2. Müller MM,
    3. Andersen S,
    4. Trujillo NJ,
    5. Valdés-Sosa P,
    6. Malinowski P,
    7. Hillyard SA
    (2006) Feature-selective attention enhances color signals in early visual areas of the human brain. Proc Natl Acad Sci U|S|A 103:14250–14254. pmid:16956975
    OpenUrlAbstract/FREE Full Text
  28. ↵
    2. Murty DVPS,
    3. Manikandan K,
    4. Kumar WS,
    5. Ramesh RG,
    6. Purokayastha S,
    7. Javali M,
    8. Rao NP,
    9. Ray S
    (2020) Gamma oscillations weaken with age in healthy elderly in human EEG. NeuroImage 215:116826. pmid:32276055
    OpenUrlPubMed
  29. ↵
    2. Murty DVPS,
    3. Manikandan K,
    4. Kumar WS,
    5. Ramesh RG,
    6. Purokayastha S,
    7. Nagendra B,
    8. Ml A,
    9. Balakrishnan A,
    10. Javali M,
    11. Rao NP,
    12. Ray S
    (2021) Stimulus-induced gamma rhythms are weaker in human elderly with mild cognitive impairment and Alzheimer's disease. Elife 10:e61666. doi:10.7554/eLife.61666
    OpenUrlCrossRef
  30. ↵
    2. Norcia AM,
    3. Appelbaum LG,
    4. Ales JM,
    5. Cottereau BR,
    6. Rossion B
    (2015) The steady-state visual evoked potential in vision research: a review. J Vis 15(6):4, 1–46. doi:10.1167/15.6.4 pmid:26024451
    OpenUrlAbstract/FREE Full Text
  31. ↵
    2. Regan D
    (1983) Spatial frequency mechanisms in human vision investigated by evoked potential recording. Vision Res 23:1401–1407. doi:10.1016/0042-6989(83)90151-7 pmid:6666040
    OpenUrlCrossRefPubMed
  32. ↵
    2. Regan D
    (1989) Human brain electrophysiology: evoked potentials and evoked magnetic fields in science and medicine. Amsterdam: Elsevier.
  33. ↵
    2. Regan MP,
    3. Regan D
    (1988) A frequency domain technique for characterizing nonlinearities in biological systems. J Theor Biol 133:293–317. doi:10.1016/S0022-5193(88)80323-0
    OpenUrlCrossRef
  34. ↵
    2. Salelkar S,
    3. Ray S
    (2020) Interaction between steady-state visually evoked potentials at nearby flicker frequencies. Sci Rep 10:5344. doi:10.1038/s41598-020-62180-y pmid:32210321
    OpenUrlCrossRefPubMed
  35. ↵
    2. Silberstein RB,
    3. Nunez PL,
    4. Pipingas A,
    5. Harris P,
    6. Danieli F
    (2001) Steady state visually evoked potential (SSVEP) topography in a graded working memory task. Int J Psychophysiol 42:219–232. doi:10.1016/S0167-8760(01)00167-2
    OpenUrlCrossRefPubMed
  36. ↵
    2. Störmer VS,
    3. Alvarez GA
    (2014) Feature-based attention elicits surround suppression in feature space. Curr Biol 24:1985–1988. doi:10.1016/j.cub.2014.07.030 pmid:25155510
    OpenUrlCrossRefPubMed
  37. ↵
    2. Toffanin P,
    3. de Jong R,
    4. Johnson A,
    5. Martens S
    (2009) Using frequency tagging to quantify attentional deployment in a visual divided attention task. Int J Psychophysiol 72:289–298. doi:10.1016/j.ijpsycho.2009.01.006
    OpenUrlCrossRefPubMed
  38. ↵
    2. Tsai JJ,
    3. Wade AR,
    4. Norcia AM
    (2012) Dynamics of normalization underlying masking in human visual cortex. J Neurosci 32:2783–2789. doi:10.1523/JNEUROSCI.4485-11.2012
    OpenUrlAbstract/FREE Full Text
  39. ↵
    2. Vialatte F-B,
    3. Maurice M,
    4. Dauwels J,
    5. Cichocki A
    (2010) Steady-state visually evoked potentials: focus on essential paradigms and future perspectives. Prog Neurobiol 90:418–438. pmid:19963032
    OpenUrlCrossRefPubMed
  40. ↵
    2. Wang R,
    3. Gao X,
    4. Gao S
    (2004) A study on binocular rivalry based on the steady state VEP. In: EMBC 2004: 26th annual international conference of the IEEE Engineering in Medicine and Biology Society: conference proceedings: linkages for innovation in biomedicine: 1-5 September, 2004, San Francisco, California (Dumont G, Hudson DL, Liang Z-P, eds), pp 259–262. Piscataway, NJ: IEEE.
  41. ↵
    2. Zemon V,
    3. Ratliff F
    (1984) Intermodulation components of the visual evoked potential: responses to lateral and superimposed stimuli. Biol Cybern 50:401–408. pmid:6487677
    OpenUrlCrossRefPubMed
Back to top

In this issue

Email
Print
View Full Page PDF
Citation Tools
Respond to this article
Local Interactions between Steady-State Visually Evoked Potentials at Nearby Flickering Frequencies
Kumari Liza, Supratim Ray
Journal of Neuroscience 11 May 2022, 42 (19) 3965-3974; DOI: 10.1523/JNEUROSCI.0180-22.2022
Twitter logo Facebook logo Mendeley logo

Jump to section

Keywords

Responses to this article

Respond to this article

Jump to comment:

No eLetters have been published for this article.