Abstract

Human alpha-band activity (8–12 Hz) has been proposed to index a variety of mechanisms during visual processing. Here, we distinguished between an account in which alpha suppression indexes selective attention versus an account in which it indexes subsequent working memory storage. We manipulated two aspects of the visual stimuli that perceptual attention is believed to mitigate before working memory storage: the potential interference from distractors and the size of the focus of attention. We found that the magnitude of alpha-band suppression tracked both of these aspects of the visual arrays. Thus, alpha-band activity after stimulus onset is clearly related to how the visual system deploys perceptual attention and appears to be distinct from mechanisms that store target representations in working memory.

INTRODUCTION

What brain processes are measured by human alpha-band activity has been one of the longest lived mysteries in neuroscience (Walter, 1938; Berger, 1929). One of the earliest hypotheses was that the strength of the 8- to 12-Hz activity evident in the raw EEG might measure the attentional state of the participant (Berger, 1929), with alpha power being strong when not attentive and weak when attending to a stimulus or train of thought. During the subsequent decades, alpha has been proposed to index everything from neural noise, to awareness, to working memory storage (Palva, Monto, Kulashekhar, & Palva, 2010; Woodman, 2010). However, modern versions of Berger's conjecture continue to propose that the alpha-band suppression that follows the onset of a stimulus array may measure the operation of a selective attention mechanism (e.g., Hakim, Adam, Gunseli, Awh, & Vogel, 2019; Wang, Rajsic, & Woodman, 2019; Klimesch, 2012). Our goal here was to distinguish between two competing views: One hypothesizes that alpha-band suppression is generated by a perceptual attention mechanism, and the competing hypothesis proposes that alpha suppression is because of later-stage working memory storage.

Neural oscillations have increasingly become the focus of neuroscientists trying to understand how the brain processes information through specialized neural circuits (Buzsáki & Freeman, 2015; Prescott, Ratté, De Koninck, & Sejnowski, 2008; Makeig et al., 2002). However, the mapping between cognitive mechanisms and the modulations of specific frequency bands is still hotly debated (Cohen, 2017; Hanslmayr & Staudigl, 2014). The 8- to 12-Hz alpha band has received much attention because of it being strongly suppressed during task performance and the signal itself effectively passing through the skulls of human participants (Nunez & Srinivasan, 2006; Cooper, Winter, Crow, & Walter, 1965). However, the relationship between alpha activity and different mechanisms of information processing continues to be debated (Fukuda, Mance, & Vogel, 2015; Jensen & Mazaheri, 2010; Hanslmayr, Spitzer, & Bäuml, 2009; Palva & Palva, 2007; Jensen, 2002). Thus, our goal here was to directly test competing predictions to distinguish between the accounts relating alpha activity to different information processing mechanisms in the brain.

Under the attentional hypothesis, alpha-band suppression indexes a mechanism of perceptual attentional selection that serves to filter out distractors and to focus attention to task-relevant objects (e.g., Jensen & Mazaheri, 2010). Under the working memory hypothesis, alpha-band suppression indexes a mechanism that stores task-relevant information in memory, after front-end perceptual attention has worked to filter distractors and focus the spotlight of attention on task-relevant inputs (Jensen, 2002). To distinguish between these competing hypotheses, we manipulated two characteristics of the visual stimuli that theories propose attentional selection handles before working memory storage. These are the presence of distracting stimuli (in Experiments 1 and 2) and the Spatial extent of the attended area (in Experiment 3).

For many decades, cognitive neuroscientists have studied how mechanisms of perceptual attention operate. This literature has shown that two characteristics of the visual input appear to be mitigated by visual attention, before the later storage of task-relevant information in memory. Specifically, these are factors that would seem to make task-relevant information more difficult to process, but these factors do not impair performance on a memory task because it appears that attention resolves these difficulties before memory storage. First, attention has been shown to effectively resolve competition from distracting stimuli (Wolfe, 1998; Luck, Chelazzi, Hillyard, & Desimone, 1997; Chelazzi, Miller, Duncan, & Desimone, 1993; Broadbent, 1957). Second, the aperture of visual attention has been shown to be adjustable so as to select a set of task-relevant stimuli that may extend across space, such that the aperture of attentional selection has been described as being like a zoom lens (Müller, Bartelt, Donner, Villringer, & Brandt, 2003; Eriksen & St. James, 1986). In both of these cases, it appears that front-end attentional selection mechanisms work to prevent distraction from task-irrelevant objects and to spatially spread attention out across space, such that subsequent mechanisms of information processing are shielded from these aspects of the visual input that could make processing slower or more error prone. We sought to manipulate these characteristics of the visual stimuli to determine whether alpha-band activity tracked our parametric manipulations, as would be expected of an attentional selection mechanism, or was insensitive to these manipulations, as would be expected of a working memory storage mechanism.

To test the hypothesis that alpha-band suppression indexes a mechanism of attentional selection that serves to prevent interference from distractors, we presented single target objects embedded in arrays of distractors and we varied the number of distracting objects (i.e., the set size) as well as the target–distractor similarity, as shown in Figure 1. These experiments were designed to hold target-processing demands constant within each array by presenting a single-colored square that participants needed to remember until the end of the trial, with each trial consisting of three targets that needed to be remembered on each trial. Across trials, we varied the number of distractors. The attention literature has shown that increasing the number of distractors taxes perceptual processing because of the need to filter out these task-irrelevant objects (Vogel, McCollough, & Machizawa, 2005; Wolfe, 1998, 2003; Luck et al., 1997), and this interference is greater when the distractors are more similar to targets in feature space (Woodman, Kang, Thompson, & Schall, 2008; Duncan & Humphreys, 1989; Green & Anderson, 1956). If the amplitude change of alpha-band activity after the onset of a stimulus is because of the brain filtering out task-irrelevant distractors to facilitate target processing (e.g., Jensen & Mazaheri, 2010), then we should see stronger alpha modulation when more distractors are presented (Experiments 1 and 2) and when those distractors are more similar to the targets (Experiment 1 vs. 2). In contrast, if alpha-band activity is because of the need to store targets in memory (e.g., Jensen, 2002), then we should see suppression increase with each additional target (i.e., with each additional array in Experiments 1 and 2), regardless of the nature or number of distractors.

Figure 1. 

Illustration of the experimental paradigm and example trials from Experiments 1 and 2. (Top) An example of the sequential working memory task of Experiment 1 from an individual trial with two distractors in the target hemifield. The cue arrows in the example trials point to the right side, instructing participants to attend to the upcoming items appearing in the right hemifield, keeping their eyes fixed on the center fixation. They were instructed to only remember the color of the squares presented in the cued hemifield while ignoring the white circles presented at the same hemifield and all the other items that appeared in the uncued hemifield. After three sequentially presented memory arrays, participants then pressed one key on the keyboard (f or j) to indicate whether one of the colored squares had changed its color or not (key being counterbalanced). In this example trial, the participant would compare their memories of the items presented in the right hemifield to the colored squares presented on the right side of the test array. Below the example trial, we show the possible Distractor set sizes that could occur in Experiment 1. Each memory array could contain 0, 1, 2, 3, or 4 same-color distractors but only one target item. (Bottom) An example of the sequential working memory task of Experiment 2 from an individual trial with two distractors. Similar to Experiment 1, participants were instructed to only remember the color of the squares presented in the cued hemifield while ignoring all the other items. The bottom panel illustrates all the possible Distractor set sizes that could occur during a trial in Experiment 2; each memory array could contain 1, 2, 3, or 4 different-color distractors but only one target item. The colors of the test arrays in both of the example trials were the same (no color changes).

Figure 1. 

Illustration of the experimental paradigm and example trials from Experiments 1 and 2. (Top) An example of the sequential working memory task of Experiment 1 from an individual trial with two distractors in the target hemifield. The cue arrows in the example trials point to the right side, instructing participants to attend to the upcoming items appearing in the right hemifield, keeping their eyes fixed on the center fixation. They were instructed to only remember the color of the squares presented in the cued hemifield while ignoring the white circles presented at the same hemifield and all the other items that appeared in the uncued hemifield. After three sequentially presented memory arrays, participants then pressed one key on the keyboard (f or j) to indicate whether one of the colored squares had changed its color or not (key being counterbalanced). In this example trial, the participant would compare their memories of the items presented in the right hemifield to the colored squares presented on the right side of the test array. Below the example trial, we show the possible Distractor set sizes that could occur in Experiment 1. Each memory array could contain 0, 1, 2, 3, or 4 same-color distractors but only one target item. (Bottom) An example of the sequential working memory task of Experiment 2 from an individual trial with two distractors. Similar to Experiment 1, participants were instructed to only remember the color of the squares presented in the cued hemifield while ignoring all the other items. The bottom panel illustrates all the possible Distractor set sizes that could occur during a trial in Experiment 2; each memory array could contain 1, 2, 3, or 4 different-color distractors but only one target item. The colors of the test arrays in both of the example trials were the same (no color changes).

Moreover, to validate the specificity of the alpha-band modulations to the demands of attentional filtering or working memory storage, we compared the pattern of alpha-band effects to a frontal ERP component (i.e., the formerly named “prefrontal bias signal”) that has been proposed to index distractor suppression (Liesefeld, Liesefeld, & Zimmer, 2014) and a posterior parietal component (i.e., the contralateral delay activity [CDA]) that has been shown to track working memory storage. Previous work showed that the prefrontal bias signal appears about 250–300 msec after item onset in the presence of distractors and has been proposed to reflect the filtering of distractors. The CDA is an established neural marker of storing targets in visual working memory (for a review, see Luria, Balaban, Awh, & Vogel, 2016). Its amplitude increases as more representations are held in this limited-capacity visual memory store (Vogel & Machizawa, 2004).

To determine if alpha-band suppression is sensitive to the scope, or Spatial extent, of the deployment of perceptual attention, we manipulated the distance between the target objects while keeping the number of targets constant in Experiment 3. If alpha suppression indexes attentional selection, then this signal should be stronger when larger arrays are shown, because the focus of attention has to zoom out to select all of the objects in the array. In contrast, if alpha suppression indexes working memory storage, alpha suppression would not be expected to be sensitive to the spatial layout of the array, because existing research has shown that visual working memory representations of objects are relatively divorced from spatial location (Xu & Chun, 2006; Phillips, 1974), such that observers can accurately detect changes of simple objects regardless of their spatial layout (Woodman, Vogel, & Luck, 2012). We also compared the pattern of alpha-band effects to those of the CDA component to cross-validate the alpha-band activity as an index of the spatial focus of attention versus working memory storage.

EXPERIMENTS 1 AND 2

To test the hypothesis that alpha-band suppression indexes a mechanism of attentional selection that serves to filter out distractors, we presented single target objects embedded in arrays of distractors that varied in the number of distracting objects (i.e., the set size) as well as the target–distractor similarity, as shown in Figure 1. Our experiments were designed to hold target-processing demands constant within each array by presenting a single-colored square that participants needed to remember until the end of the trial. Across trials, we varied the number of distractors.

Materials and Methods

Participants

Eighteen participants from Vanderbilt University and the surrounding community participated in Experiment 1 (17 women, Mage = 23.2 years, SDage = 5.2 years) and Experiment 2 (14 women, Mage = 21.9 years, SDage = 3.4 years). None of the participants in Experiment 2 participated in Experiment 1. All participants gave informed consent before experimental procedures approved by the Vanderbilt University institutional review board and received compensation of $15 per hour. All self-reported normal or corrected-to-normal visual acuity and normal color vision. Three participants' data in Experiment 1 were replaced because of excessive eye movements and muscular artifacts (described below in the EEG Analyses section). Seven participants' data were replaced because of the same reason in Experiment 2.

Stimuli and Procedures

Stimuli were presented using MATLAB (R2017b 9.3.0; MathWorks) and the Psychophysics Toolbox (Version 3.0.12; Brainard, 1997; Pelli, 1997) on a 24-in. LED gaming monitor (ASUS VG 248; 120-Hz refresh rate). Stimuli were presented on a gray background (x = 286, y = 384, L = 167 cd/m2, x and y define chromaticity, L represents luminance in the CIE xyY color space derived from the CIE 1931 color space). Participants were seated approximately 75 cm from the screen.

Figure 1 shows example trials from both Experiments 1 and 2. The stimuli in the two experiments were essentially identical with the exception that all the distractors were white in Experiment 1, whereas the color of each distractor was drawn from the same set of possible colors as the targets in Experiment 2. The purpose of Experiment 2 was to increase the similarity between the to-be-remembered targets and the distractors. We did this by replacing the white circles in Experiment 1 with colored circles in Experiment 2 (see Figure 1). This manipulation increased the difficulty of filtering out distractors in Experiment 2 relative to Experiment 1, based on prior literature (Duncan & Humphreys, 1989).

Each trial began with a display containing a black fixation cross (x = 290, y = 298, L = 0.66 cd/m2, 0.4° of visual angle) in the center of the screen for 500 msec, followed by a black arrow cue (1.3° of visual angle wide and 0.4° tall) presented 1.5° above the center fixation for 200 msec indicating the relevant side of the screen (right or left) to be attended. Participants were instructed to attend to the upcoming items appearing in the cued hemifield, keeping their eyes fixated at the center of the screen. After a postcue delay of 900 msec, we sequentially presented three memory arrays. Each hemifield had one colored square (the target) and either zero to four white circle distractors in Experiment 1 or one to four colored circle distractors in Experiment 2. Participants were instructed to only memorize the colored squares that appeared in the cued hemifield while ignoring the circles. Each of these arrays appeared on the screen for 200 msec, and the interval between each successive array was 500 msec, with a 1000-msec delay period after the last array. This presentation rate was selected to minimize the total duration of each trial because participants needed to not blink during the trial, while allowing us to measure the alpha-band modulations of interest. In summary, the targets were presented one at a time across three successive arrays, for three targets in total across each trial. Each target was presented once on the screen and disappeared rapidly. The number of distractors stayed the same across the three displays within each trial.

The test array consisted of all three colored squares simultaneously presented in one array. On 50% of trials, the color of one square in the cued hemifield of the test array was replaced with another color. On these change trials, one of the colored squares would change to a color not previously seen in that hemifield on that trial. The locations of all the colored squares in the test array stayed the same as they appeared in the memory arrays, including the changed one. Participants then pressed one key on the keyboard to indicate that one of the colors of the squares had changed and pressed another key to indicate that all colors of the squares had stayed the same. The change versus no-change keys were counterbalanced across participants, with the two keys used being f and j.

In Experiment 1, the colors of the squares were chosen from a pool of six highly discriminable colors: red (x = 592, y = 381, L = 39 cd/m2), green (x = 258, y = 664, L = 128 cd/m2), blue (x = 141, y = 66, L = 10 cd/m2), magenta (x = 314, y = 177, L = 427 cd/m2), yellow (x = 437, y = 516, L = 260 cd/m2), and black (x = 290, y = 298, L = 0.66 cd/m2). All the circles were white (x = 328, y = 350, L = 278 cd/m2). In Experiment 2, both the colors of the squares and the colors of the circles were chosen from a pool of seven highly discriminable colors, including all the colors from the color pool of Experiment 1 and white (x = 328, y = 350, L = 278 cd/m2). In both Experiment 1 and Experiment 2, sets of colors were randomly chosen from the color pool, without replacement for the squares in a given hemifield. Because of the limited number of colors, the colors of the circles in a given hemifield could have been replaced no more than twice in Experiment 2. The sizes of the squares and circles were approximately equal (squares: 0.65° of visual angle in width and 0.65° tall, circles: 0.78° of visual angle diameter). All the targets and distractors were placed along the circumference of one of three imaginary circles with different eccentricities (visual angles of 3.3°, 4.5°, and 5.7°), centered on fixation. Each imaginary circle had four locations on each side of the screen with a minimum difference of 20° of polar angle between neighboring locations. Therefore, on each trial, the position of the three sequentially presented targets was randomly chosen from the 12 locations (without replacement), and participants could not predict the position of any target. Any two objects presented on the same array could not share the same location, and the minimum distance between two items was at least 1.2° of visual angle. To prevent participants from knowing in advance the number of distractors in an upcoming trial, all trials with different numbers of distractors (0, 1, 2, 3, or 4 distractors in Experiment 1 and 1, 2, 3, or 4 distractors in Experiment 2) were randomly interleaved within a given block. All participants completed 240 trials with each distractor condition for 1200 trials in total for Experiment 1 and 960 trials in total for Experiment 2.

Change Detection Task Performance

Accuracy of remembering the target colors given the set size of distractors was calculated from the number of correct responses divided by the total number of responses (i.e., percent correct). The calculation of working memory capacity (K) for each distractor condition in the change detection task followed the formula K = Set Size × (Hit Rate − False Alarm Rate) / (1 − False Alarm Rate) (Rouder, Morey, Morey, & Cowan, 2011), where hit rate represents proportion of correct responses on change trials and false alarm rate represents proportion of incorrect responses on no-change trials.

EEG Acquisition

The EEG was recorded in an electrically shielded, soundproof booth from 30 active Ag/AgCl electrodes (Brain Products actiCHamp, Munich, Germany) mounted in an elastic cap positioned according to the International 10–20 system (Fp1, Fp2, Fz, F3, F4, F7, F8, FC5, FC6, C3, C4, Cz, T7, T8, P7, P8, P3, P4, Pz, PO7, PO8, PO3, PO4, POz, O1, O2, Oz). Impedance values were kept below 10 kΩ. The reference electrodes were affixed with stickers to the left and right mastoids, and a ground electrode was placed in the elastic cap at Fpz. Data were referenced online to the right mastoid and rereferenced offline to the algebraic average of the left and right mastoids. EOG activity was recorded with two horizontal electrodes placed ∼1 cm lateral to the outer canthi of the two eyes, and one vertical EOG electrode was placed below the right eye to detect eye movements and blinks. All channels were bandpass filtered from 0.01 to 100 Hz and recorded with a 250-Hz sampling rate.

EEG Analyses

Artifact rejection

We rejected trials with artifacts using a three-step procedure. We first applied a standard artifact-rejection function from the EEGLAB Toolbox (eegthresh.m; Delorme & Makeig, 2004) to the segmented raw data to exclude trials contaminated by blinks, amplifier saturation, or excessive noise. If the maximum voltage during the 2600-msec interval after the first memory array onset of a trial was greater than +100 μV or the minimum voltage was less than −100 μV, then it was marked as a trial with artifact and rejected.

The next two steps were performed on the remaining trials to detect segments of the EEG contaminated by artifacts. In particular, our methods were designed to maximize our ability to detect small horizontal eye movements that generate electrophysiological responses below the abovementioned threshold, because these eye movements could contaminate our lateralized measures. First, we performed a split-half sliding window approach (Adam, Robison, & Vogel, 2018; window size = 200 msec, step size = 20 msec, threshold = 20 μV) on the horizontal EOG (HEOG) difference signal. The HEOG difference signal was calculated by subtracting the ipsilateral HEOG electrode from the contralateral HEOG electrode of the cued hemifield. We slid a 200-msec time window in steps of 20 msec from the beginning to the end of the trial. If the change in HEOG difference voltage from the first half to the second half of the window was greater than 20 μV, then that trial was marked as having an eye movement and was rejected. Because of the limited signal-to-noise ratio of EOG recordings on individual trials, trials with small eye movements could not be reliably detected and rejected by the first or second step but were caught during the third step.

In the third step, we averaged the difference HEOG amplitude for the left and right cue trials for each participant (Woodman & Luck, 2003). If this averaged difference HEOG exceeded ±5 μV, then the participant was excluded from further analyses. We adopted a slightly higher threshold for rejecting participants than the 3 μV adopted by Woodman and Luck (2003) because the long trial durations needed for sequentially presented memory arrays were more prone to low-frequency noise pushing the waveforms near threshold (because of sweating, etc.). In Experiment 1, three participants were replaced because of excessive eye movements and artifacts. For the remaining participants, an average of 9.23% (SD = 6.25%) of trials were excluded. In Experiment 2, seven participants were replaced. For the remaining participants, an average of 9.39% (SD = 6.55%) of trials were excluded because of eye movements and artifacts.

Alpha power analysis.

To measure the alpha power change on each trial, the raw EEG data were first bandpass filtered by the eegfilt.m function and then Hilbert transformed by the hilbert.m function in MATLAB to obtain the instantaneous power values for the alpha band (8–12 Hz). Percent change in alpha power was then calculated relative to a baseline period before the onset of the cue (−400 to 0 msec relative to the cue onset, −1500 to −1100 msec relative to the first memory array onset) similar to previous studies' analyses of this frequency band (e.g., Wang et al., 2019; Adam et al., 2018). Next, the ipsilateral alpha power and the contralateral alpha power were calculated by averaging the percent change in alpha power for ipsilateral electrodes relative to the cued hemifield and contralateral electrodes relative to the cued hemifield, respectively. Global alpha power was calculated by averaging the percent change in alpha power for both ipsilateral electrodes and contralateral electrodes across the lateral-occipital and posterior-parietal electrodes (i.e., P3, P4, P7, P8, PO3, PO4, PO7, PO8, O1, and O2; Adam et al., 2018; Fukuda et al., 2015).

CDA analysis.

To relate the alpha-band activity that we measured to a canonical electrophysiological measure of visual working memory storage, we averaged the EEG data into ERPs to measure the CDA. Before calculating the CDA, the signal was baseline corrected to the mean amplitude of the ERPs from −200 to 0 msec relative to the onset of the first memory array. The CDA amplitude was then computed by subtracting the activity of the electrodes ipsilateral to the cued hemifield from the activity of the electrodes contralateral to the cued hemifield across the same lateral-occipital and posterior-parietal electrodes we used to measure alpha activity: P3, P4, P7, P8, PO3, PO4, PO7, PO8, O1, and O2. To ensure that the timing and amplitude of the CDA waveforms were not distorted by filtering, statistical analyses of the CDA were performed on the baseline-corrected but unfiltered data. For visualization purposes, trials were low-pass filtered with a two-way least squares finite impulse response filter function from the EEGLAB Toolbox (eegfilt.m; Delorme & Makeig, 2004) with the low pass at 30 Hz.

Prefrontal bias signal analysis.

The prefrontal bias signal is an ERP component that is evident 250–300 msec after the onset of the visual stimuli in the presence of distractors and has been proposed to reflect the initiation of distractor filtering (Liesefeld et al., 2014). To measure the prefrontal bias signal, we first baseline corrected the ERP signal to the mean amplitude of the ERPs from −200 to 0 msec relative to the onset of the first memory array. Because this component is not lateralized, its amplitude was then computed by averaging the activity of the frontal electrodes: Fp1, Fp2, F3, F4, F7, and F8 (Liesefeld et al., 2014). Statistical analyses of the prefrontal bias signal were also performed on the baseline-corrected but unfiltered data. Trials were low-pass filtered at 30 Hz for visualization purposes only.

Experimental Design and Statistical Analyses

To examine the influence of distractors on behavioral performance in Experiment 1, repeated-measures ANOVAs with the within-participant factor of Distractor set size (0, 1, 2, 3, or 4) were performed on the accuracy and estimated capacity (K). Separate statistical analyses were then performed on the alpha power percent change, the CDA amplitude, and the prefrontal bias signal amplitude using two-way repeated-measures ANOVAs with the within-participant factors of Serial position (first item vs. second item vs. third item) and Distractor set size (0 vs. 1 vs. 2 vs. 3 vs. 4 distractors). Preplanned pairwise comparisons with Bonferroni corrections for multiple comparisons were used across each neighboring Serial position and across each neighboring Distractor set size to verify the nature of the Serial position or Distractor set size effects. All these statistical analyses were performed in SPSS 19.0 (IBM Inc.).

All the statistical analyses for the measurements in Experiment 2 were the same as the analyses in Experiment 1 except that the factors of the Distractor set size in Experiment 2 are 1 versus 2 versus 3 versus 4 distractors instead of zero to four distractors.

Results

Behavior

We sequentially presented three memory arrays with a single target in each array and a variable number of distractors (zero to four white distractors in Experiment 1 and one to four different-colored distractors in Experiment 2). Our behavioral findings across both experiments are shown in Figure 2. We found that participants were able to use attention to filter out these distractors effectively such that their presence or set size did not significantly impair memory task performance. This is consistent with previous evidence that human participants can easily remember three colored squares (e.g., Vogel, Woodman, & Luck, 2001) and previews the main focus of this article, whether alpha suppression increased with each additional target or with each distractor, enabling this high-level performance.

Figure 2. 

Behavioral performance from Experiments 1 and 2. The mean visual working memory capacity (left) and the mean accuracy rate (right) estimates for the different numbers of distractors in Experiment 1 (red bars) and Experiment 2 (blue bars). D0, D1, D2, D3, and D4 represent arrays with zero to four distractors. Error bars indicate the SEMs.

Figure 2. 

Behavioral performance from Experiments 1 and 2. The mean visual working memory capacity (left) and the mean accuracy rate (right) estimates for the different numbers of distractors in Experiment 1 (red bars) and Experiment 2 (blue bars). D0, D1, D2, D3, and D4 represent arrays with zero to four distractors. Error bars indicate the SEMs.

Separate one-way repeated-measures ANOVAs on accuracy rate and capacity estimates (K) showed no effect of Distractor set size (F(4, 68) = 0.621, p = .649, ηp2 = .035, and F(4, 68) = 0.135, p = .969, ηp2 = 0.008, respectively, in Experiment 1; F(3, 51) = 0.581, p = .630, ηp2 = .033, and F(3, 51) = 1.988, p = .127, ηp2 = 0.105, respectively, in Experiment 2). Thus, our behavioral findings show that participants were able to effectively filter out the distractors in both experiments.

Next, we computed Bayes factors to determine how much more likely the null hypothesis, that there was no distractor effect on behavior measurements, was than the possibility that the presence of distractors actually did show an effect (Rouder, Morey, Verhagen, Swagman, & Wagenmakers, 2017). We found that, for the effect of Distractor set size on working memory capacity (K), the null hypothesis was 19.4 and 1.8 times more likely than the hypothesis that a Distractor set size effect existed in Experiment 1 and Experiment 2, respectively. Similarly, for the effect of Distractor set size on accuracy, the null hypothesis was 10.3 and 7.3 times more likely than the hypothesis that a Distractor set size effect existed in Experiment 1 and Experiment 2, respectively. These results demonstrate that participants were able to select the target objects and store them in memory regardless of the presence of distractors.

Next, we analyzed participants' response accuracy for each Serial position to verify that normal Serial position effects were observed. The results showed a clear recency effect. In both Experiment 1 and Experiment 2, participants' response accuracy for the third target was significantly higher than for the first and second targets. This pattern was evidenced by a marginal significant main effect of Serial position of the target on response accuracy in Experiment 1, F(2, 34) = 3.303, p = .079, ηp2 = .163, and a significant main effect in Experiment 2, F(2, 34) = 7.702, p = .002, ηp2 = .312, because of more accurate responses to tests of the third target compared to the first target (Experiment 1: F(1, 17) = 3.624, p = .074, ηp2 = .176; Experiment 2: F(1, 17) = 9.184, p = .008, ηp2 = .351) and the second target (Experiment 1: F(1, 17) = 3.517, p = .078, ηp2 = .171; Experiment 2: F(1, 17) = 12.591, p = .002, ηp2 = .426). Thus, participants' memory for the last target was better than that for the first two targets, demonstrating the classic recency effect in visual memory (Woodman et al., 2012).

EEG and ERPs

Alpha power suppression.

The alpha power suppression showed clear differences as a function of the number of distractors present (Figure 3A shows the time–frequency plot of power showing the profile of activity across frequency space) and demonstrating a dissociation from the participants' behavior that was largely invariant to the presence of distractors in Experiment 1. Figure 3B shows that we measured stronger global alpha power suppression as more white distractors appeared in the array with the colored target. In addition, we found that this alpha suppression did not increase as the number of targets that participants were storing in working memory increased across the trial, supporting a recent report in which alpha power was unrelated to the number of targets held in working memory (Wang et al., 2019). These observations about the pattern of the global alpha power in Experiment 1 were statistically verified with the following steps.

Figure 3. 

The alpha power suppression across arrays and Distractor set sizes in Experiment 1. (A) The mean event-related spectral perturbation (ERSP) changes in the frequency range of 2–50 Hz averaged across all Distractor set sizes observed at a representative parieto-occipital channel (PO8). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. (B) Left: the global alpha power suppression in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set size (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. (B) Right: the mean global alpha power suppression after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) Left: the ipsilateral alpha power suppression in Experiment 1 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (C) Right: the mean ipsilateral alpha power suppression after each memory item in Experiment 1. (D) Left: the contralateral alpha power suppression in Experiment 1 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (D) Right: the mean contralateral alpha power suppression after each memory item in Experiment 1. Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

Figure 3. 

The alpha power suppression across arrays and Distractor set sizes in Experiment 1. (A) The mean event-related spectral perturbation (ERSP) changes in the frequency range of 2–50 Hz averaged across all Distractor set sizes observed at a representative parieto-occipital channel (PO8). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. (B) Left: the global alpha power suppression in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set size (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. (B) Right: the mean global alpha power suppression after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) Left: the ipsilateral alpha power suppression in Experiment 1 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (C) Right: the mean ipsilateral alpha power suppression after each memory item in Experiment 1. (D) Left: the contralateral alpha power suppression in Experiment 1 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (D) Right: the mean contralateral alpha power suppression after each memory item in Experiment 1. Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

To quantify the global alpha effects in Experiment 1, we first calculated the mean global alpha suppression after each memory array. The two-way repeated-measures ANOVA of Serial position (first vs. second vs. third array) and Distractor set size (0 vs. 1 vs. 2 vs. 3 vs. 4 distractors) on the mean global alpha suppression revealed a significant main effect of Serial position, F(2, 34) = 4.486, p = .019, ηp2 = .209, because of a strongest alpha power suppression at the first memory array compared to the second memory array, F(1, 17) = 8.198, p = .011, ηp2 = .325, and the last memory array, F(1, 17) = 1.497, p = .238, ηp2 = .081, again showing that the alpha power modulations did not track the number of targets held in mind. Most importantly, for the hypothesis, we found a significant main effect of Distractor set size, F(4, 68) = 5.470, p = .001, ηp2 = .243. Finally, there was no interaction of Serial position × Distractor Set Size, F(8, 136) = 1.816, p = .079, ηp2 = .097, on the alpha power modulation because the number of distractors was qualitatively similar even as the depth of suppression weakened across the successive arrays of each trial.

Our measurements of global alpha power suggest that it provides an index that tracks the number of distractor objects that need to be filtered out. However, a number of recent studies have shown that the global alpha power across all posterior electrodes tracks the number of items stored in visual working memory (Adam et al., 2018; Fukuda, Kang, & Woodman, 2016; Fukuda et al., 2015). Meanwhile, other studies using tasks in which targets are presented on one side of the monitor, and distractors on the other, have found that alpha-band activity ipsilateral to the target hemifield increases with higher demands to suppress task-irrelevant distractors, whereas contralateral alpha activity to the target hemifield decreases with higher demands to attend to the targets (Bacigalupo & Luck, 2019; Jensen & Mazaheri, 2010). Because the calculation of global alpha power is performed by averaging alpha power from both ipsilateral and contralateral electrodes relative to the attended hemifield, the function of global alpha suppression in distractor filtering might be driven by either or both of the ipsilateral and contralateral alpha power. Therefore, we further examined the alpha-band activity contralateral to the cued hemifield compared to the ipsilateral alpha activity to test the conjecture that these lateralized signals measure different cognitive mechanisms. If only ipsilateral alpha suppression is related to the suppression of distractors, whereas contralateral alpha activity is related to attending to the targets, we would have seen that ipsilateral alpha was selectively modulated by the number of distractors, whereas the contralateral alpha was not modulated by the number of distractors and stayed constant with the sequential presence of targets. We calculated the average of lateral-posterior electrodes ipsilateral to the task-relevant hemifield and the electrodes contralateral to the task-relevant hemifield, respectively. The results of this lateralized averaging are shown in Figure 3C and D. As evidenced by these means, the patterns of alpha suppression were similar in both hemispheres, with both ipsilateral and contralateral hemispheres exhibiting the parametric modulation across the number of distractors present.

A three-way repeated-measures ANOVA with within-participant factors of Hemisphere (ipsilateral vs. contralateral), Serial position (first vs. second vs. third item), and Distractor set size (0 vs. 1 vs. 2 vs. 3 vs. 4) on the mean ipsilateral and contralateral alpha suppression revealed a significant main effect of Hemisphere, F(1, 17) = 10.564, p = .005, ηp2 = .383, because of the contralateral alpha suppression being stronger than the ipsilateral alpha suppression. Moreover, there was a significant main effect of Distractor set size, F(4, 68) = 5.470, p = .001, ηp2 = .243. However, we found neither an interaction of Hemisphere × Distractor Set Size, F(4, 68) = 0.756, p = .557, ηp2 = .043, nor an interaction of Hemisphere × Distractor Set Size × Serial position, F(8, 136) = 0.594, p = .782, ηp2 = .034. These findings appear to contradict predictions of an account in which contralateral alpha and ipsilateral alpha are related to different cognitive functions (such as attending to targets vs. suppressing distractors).

Recall that the goal of Experiment 2 was to increase the difficulty of distractor suppression by making the distractors colored like the targets relative to the all-white distractors in Experiment 1. Experiment 2 also serves to replicate and extend the pattern of results we observed in Experiment 1 (Figure 4A shows the time–frequency profile across trials). As shown in Figure 4B, we observed a nearly identical pattern of effects in Experiment 2 as we saw in Experiment 1. The primary difference appears to be that alpha power suppression may have been near ceiling in Experiment 2. This is shown in Figure 5, which overlays the alpha-power measures across both experiments. The percent change in alpha power with four distractors was similar in both experiments, but the effect of one colored distractor in Experiment 2 was closer to the peak suppression with four distractors that we found in Experiment 1. This overall pattern of alpha power across both experiments further supports the hypothesis that alpha power suppression tracks a distractor filtering mechanism that becomes taxed when targets are embedded in highly distracting contexts.

Figure 4. 

The alpha power suppression across arrays and Distractor set sizes in Experiment 2. (A) The mean event-related spectral perturbation (ERSP) changes in the frequency range of 2–50 Hz averaged across all Distractor set sizes observed at a representative parieto-occipital channel (PO8). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. (B) Left: the global alpha power suppression in Experiment 2 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. (B) Right: the mean global alpha power suppression after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) Left: the ipsilateral alpha power suppression in Experiment 2 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (C) Right: the mean ipsilateral alpha power suppression after each memory item in Experiment 2. (D) Left: the contralateral alpha power suppression in Experiment 2 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (D) Right: the mean contralateral alpha power suppression after each memory item in Experiment 2. Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

Figure 4. 

The alpha power suppression across arrays and Distractor set sizes in Experiment 2. (A) The mean event-related spectral perturbation (ERSP) changes in the frequency range of 2–50 Hz averaged across all Distractor set sizes observed at a representative parieto-occipital channel (PO8). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. (B) Left: the global alpha power suppression in Experiment 2 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. (B) Right: the mean global alpha power suppression after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) Left: the ipsilateral alpha power suppression in Experiment 2 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (C) Right: the mean ipsilateral alpha power suppression after each memory item in Experiment 2. (D) Left: the contralateral alpha power suppression in Experiment 2 averaged over electrode pairs: P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2. (D) Right: the mean contralateral alpha power suppression after each memory item in Experiment 2. Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

Figure 5. 

The global alpha suppression in Experiment 1 overlaid with Experiment 2. Solid lines and dashed lines represent the global alpha power suppression in Experiment 1 and Experiment 2, respectively. Black, yellow, magenta, red, and blue lines represent zero distractors (D0), one distractor (D1), two distractors (D2), three distractors (D3), and four distractors (D4), respectively. The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power.

Figure 5. 

The global alpha suppression in Experiment 1 overlaid with Experiment 2. Solid lines and dashed lines represent the global alpha power suppression in Experiment 1 and Experiment 2, respectively. Black, yellow, magenta, red, and blue lines represent zero distractors (D0), one distractor (D1), two distractors (D2), three distractors (D3), and four distractors (D4), respectively. The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power.

For the global alpha suppression analysis in Experiment 2, we entered the mean global alpha suppression after each memory array into a two-way repeated-measures ANOVA of Serial position (first vs. second vs. third item) and Distractor set size (1 vs. 2 vs. 3 vs. 4). This yielded a significant main effect of Distractor set size, F(3, 51) = 2.848, p = .047, ηp2 = .143, but no Serial position effect, F(2, 34) = 1.207, p = .312, ηp2 = .066. However, the ANOVA yielded an interaction of Serial position × Distractor Set Size, F(6, 102) = 2.892, p = .017, ηp2 = .145. Post hoc pairwise t tests across each neighboring Distractor set size in each Serial position showed that the interaction of Serial position × Distractor Set Size was driven by global alpha suppression being significantly stronger when four distractors were present than one distractor, specifically in the first Serial position, t(17) = 3.056, p = .043, not in any other Serial positions (ps > .083). We believe this interaction may also be a result of alpha power suppression being near ceiling, particularly when elicited by the second and third memory arrays.

As in Experiment 1, we next examined whether the strength of the ipsilateral alpha suppression and the contralateral alpha suppression showed different patterns, as would be expected if these neural measures reflect different functions. We analyzed the ipsilateral and contralateral alpha suppression in the same way as we did in Experiment 1 (see Figure 4C and D). The three-way repeated-measures ANOVA using the within-participant factors of Hemisphere (ipsilateral vs. contralateral), Serial position (first vs. second vs. third item), and Distractor set size (1 vs. 2 vs. 3 vs. 4) on the mean ipsilateral and contralateral alpha suppression revealed a significant main effect of Hemisphere, F(1, 17) = 7.043, p = .017, ηp2 = .293, because of the contralateral alpha suppression being stronger than the ipsilateral alpha suppression. Moreover, there was a significant main effect of Distractor set size, F(3, 51) = 2.848, p = .047, ηp2 = .143, and a significant interaction of Hemisphere × Distractor Set Size, F(3, 51) = 4.235, p = .010, ηp2 = .199. Follow-up post hoc pairwise comparisons revealed that this interaction was driven by the contralateral alpha suppression being stronger than the ipsilateral alpha suppression when there were two or four distractors, t(17) = 3.166, p = .006, and t(17) = 2.720, p = .015, respectively. However, there was no significant difference between the contralateral alpha suppression and the ipsilateral alpha suppression when there was one or three distractors, t(17) = 0.989, p = 0.336, and t(17) = 1.479, p = .158, respectively. Finally, we did not observe an interaction of Hemisphere × Distractor Set Size × Serial position, F(6, 102) = 0.151, p = .988, ηp2 = .009. In summary, whereas the effect of number of distractors was ubiquitous across experiments and set sizes, the differences in lateralized activity were not consistently observed.

To determine whether the Distractor set size has similar effects in Experiment 1 and Experiment 2, we entered the global alpha power measurements into a mixed model ANOVA with the between-participant factor of Experiment (1 vs. 2) and the within-participant factors of Distractor set size (1, 2, 3, or 4) and Serial position (first vs. second vs. third; see the global alpha suppression in Experiment 1 overlaid with Experiment 2 in Figure 5). This yielded no significant effect of Experiment, F(1, 34) = 0.010, p = .920, ηp2 = .000. Moreover, no interactions of Experiment × Serial position, F(2, 68) = 1.597, p = .210, ηp2 = .045, Experiment × Distractor set size, F(3, 102) = 0.250, p = .861, ηp2 = .007, or Experiment × Serial position × Distractor set size, F(6, 204) = 1.766, p = .108, ηp2 = .049, were found. We also compared the ipsilateral alpha suppression and the contralateral alpha suppression across two experiments, as in global alpha suppression, no differences were found between Experiments 1 and 2 (Fs < 1.838, ps > .093). We also computed Bayes factors for the alpha power suppression between Experiments 1 and 2 (Rouder et al., 2017). We found that the null hypothesis was 8.9, 7.5, and 9.4 times more likely than the hypothesis that a difference existed in global alpha power suppression, ipsilateral alpha suppression, and contralateral alpha suppression between Experiment 1 and Experiment 2, respectively. These null results support our conclusion that we observed the same general pattern of results in Experiment 2 and Experiment 1, as shown in Figure 5.

If global alpha suppression indexes an attentional mechanism that serves to filter out task-irrelevant distractors, its signal should be stronger on trials that they successfully filtered out the distractors and responded correctly, compared to trials with incorrect responses. To test this, we calculated the global alpha power suppression of trials in which participants made incorrect responses. As shown in Figure 6, the global alpha suppression was consistent with the idea that weaker alpha suppression is observed on incorrect response trials, presumably because of poor distractor filtering, at least on a proportion of trials. The difference in alpha power suppression between correct and incorrect trials appears larger in the first memory array. In Experiment 1, the paired-samples t tests between the averaged alpha power suppression on incorrect trials and the averaged alpha power suppression on correct trials at each Serial position revealed a significantly stronger alpha suppression on correct trials with three distractors and with four distractors, compared to trials with incorrect responses in the first memory array, t(17) = 2.251, p = .038, and t(17) = 2.387, p = .029, respectively. However, there was no statistically significant difference in alpha power suppression between correct and incorrect trials in the second and third memory arrays (ps > .066). In Experiment 2, none of the difference in alpha power suppression between correct and incorrect trials reached statistical significance (ps > .192). These findings might be only suggestive at this point because of performance being around 85% correct across both experiments, that is, the statistical analyses not reaching significance because of the small number of incorrect trials (i.e., approximately 15% of trials in Experiments 1 and 2). Thus, these analyses suggest an interesting potential to measure the efficacy of distractor filtering with the strength of alpha suppression.

Figure 6. 

The comparison of the global alpha power suppression on correct trials and incorrect trials across arrays and Distractor set sizes across Experiments 1 and 2. (A) Left: the global alpha power suppression in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set size (black, yellow, magenta, red, and blue solid lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively; the green dashed line represents incorrect trials). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. Right: the mean global alpha power suppression after each memory item in Experiment 1, separated by Distractor set sizes (green, black, yellow, magenta, red, and blue bars represent incorrect trials, zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (B) Left: the global alpha power suppression in Experiment 2. Right: the mean global alpha power suppression after each memory item in Experiment 2.

Figure 6. 

The comparison of the global alpha power suppression on correct trials and incorrect trials across arrays and Distractor set sizes across Experiments 1 and 2. (A) Left: the global alpha power suppression in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set size (black, yellow, magenta, red, and blue solid lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively; the green dashed line represents incorrect trials). The cyan bar and black bars on the time axis represent the onset and duration of the cue and each memory array. The gray intervals represent the time window of the global alpha power. Right: the mean global alpha power suppression after each memory item in Experiment 1, separated by Distractor set sizes (green, black, yellow, magenta, red, and blue bars represent incorrect trials, zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (B) Left: the global alpha power suppression in Experiment 2. Right: the mean global alpha power suppression after each memory item in Experiment 2.

In summary, the findings from Experiment 2 replicated the findings of Experiment 1 using a new group of participants and distractors that were composed of different features (i.e., shape and randomly colored). We found strong effects of Distractor set size, as expected if alpha-band activity is measuring a distractor filtering mechanism, but without systematic power changes related to adding another target representation from each array to the target set.

CDA.

In the preceding text, we based our conclusion that alpha power indexes distractor suppression, and not target selection, on the fact that this neural activity was sensitive to the number of distractors present on the screen, but not the number of targets that needed to be processed on each trial. To strengthen these conclusions, we wanted to examine a different neural measure from our data that should track the processing of the targets. To make this comparison, we measured the CDA, which is an established neural marker of storing targets in visual working memory (for a review, see Luria et al., 2016). Its amplitude increases as more representations are held in memory (Vogel & Machizawa, 2004).

As shown in Figure 7, we found that the CDA amplitude increased with each additional to-be-remembered target, unlike our measures of alpha suppression in Experiment 1 (see Figures 3 and 4). Overall, we found that the distractors had a minimal effect on CDA amplitude, consistent with the behavioral data suggesting that the distractors were successfully shielded from memory. This was clearest at the first two Serial positions where CDA amplitude was not measurably influenced by the presence of distractors. However, when memory demands approached working memory capacity, at the third Serial position, the CDA amplitude increased slightly in the presence of distractors compared to arrays with no distractors, as if a distractor occasionally broke through the alpha filter when working memory was filled to capacity. These observations about the pattern of the CDA effects were supported by the following statistical analyses.

Figure 7. 

The CDA activity change with Serial position and Distractor set sizes in Experiment 1 and Experiment 2. (A) The CDA waveforms in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the CDA. (B) The mean CDA amplitude after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) The CDA waveforms in Experiment 2 averaged over the same electrode pairs as in Experiment 1, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). (D) The mean CDA amplitude after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

Figure 7. 

The CDA activity change with Serial position and Distractor set sizes in Experiment 1 and Experiment 2. (A) The CDA waveforms in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the CDA. (B) The mean CDA amplitude after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) The CDA waveforms in Experiment 2 averaged over the same electrode pairs as in Experiment 1, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). (D) The mean CDA amplitude after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). Error bars indicate the SEMs. Asterisk indicates the significant difference between Distractor set sizes (*p < .05).

We calculated the mean amplitude of the CDA after each memory array. Then, we entered these means into a two-way repeated-measures ANOVA of Serial position (first vs. second vs. third item) and Distractor set size (0 vs. 1 vs. 2 vs. 3 vs. 4). This yielded a significant main effect of Serial position, F(2, 34) = 35.106, p < .001, ηp2 = .674, and a significant interaction of Serial position × Distractor set size, F (8, 136) = 2.762, p = .023, ηp2 = .140, because of the zero-distractor CDA amplitude being smaller than the CDA amplitude when distractors were present at Serial Position 3, and confirmed in pairwise t tests of a significantly larger CDA amplitude with one distractor than the CDA amplitude with no distractors in the third Serial position, t(17) = 3.564, p = .024 (with Bonferroni correction for multiple comparisons), and no differences in the other pairwise comparisons (ps > .160). The preplanned pairwise comparisons of the mean CDA amplitude across each neighboring Serial position revealed that the CDA amplitude increased from the first item to the second item, F(1, 17) = 19.424, p < .001, ηp2 = .533. Consistent with participants' K scores, it kept increasing from the second item to the third item, F(1, 17) = 15.146, p = .001, ηp2 = .471. However, there was no main effect of Distractor set size, F(4, 68) = 1.672, p = .182, ηp2 = .090, in sharp contrast to the alpha-power pattern observed across both experiments. A Bayes factor analysis was also conducted on the CDA amplitude to determine how much more likely the null hypothesis was than the hypothesis that a Distractor set size effect existed (Rouder et al., 2017). We found that the null hypothesis was 4.3 times more likely than the hypothesis that a Distractor set size effect on the CDA amplitude existed, demonstrating the reliability of the null Distractor set size effect in Experiment 1.

In Experiment 2, the CDA amplitude also increased with the accumulation of each additional target. However, the presence of distractors in Experiment 2 significantly influenced the CDA amplitude for all Serial positions. These observations about the pattern of CDA were verified statistically in the following steps. First, we calculated the mean CDA amplitude after each memory array (averaged across the same time window as in Experiment 1). Then, the two-way repeated-measures ANOVA of Serial position (first vs. second vs. third item) and Distractor set size (1 vs. 2 vs. 3 vs. 4) on the mean CDA amplitude after each memory array showed a significant Serial position effect, F(2, 34) = 23.221, p < .001, ηp2 = .557, and a significant Distractor set size effect, F(3, 51) = 6.651, p = .001, ηp2 = .281. However, there was no interaction of Serial position × Distractor set size, F(6, 102) = 1.250, p = .287, ηp2 = .069. The preplanned pairwise comparisons of the mean CDA amplitude across each neighboring Serial position revealed that the CDA amplitude increased from the first item to the second item, F(1, 17) = 37.438, p < .001, ηp2 = .688. In contrast to Experiment 1, it stopped increasing from the second item to the third item, F(1, 17) = 1.541, p = .231, ηp2 = .083. Next, the preplanned pairwise comparison of the mean CDA amplitude across each neighboring Distractor set size showed that the CDA amplitude increased from one distractor to two distractors, F(1, 17) = 7.315, p = .015, ηp2 = .301, and did not significantly increase from two to three distractors, F(1, 17) = 2.365, p = .142, ηp2 = .122, or from three to four distractors, F(1, 17) = 3.105, p = .096, ηp2 = .154.

These CDA results appear to support our conclusion that the alpha-power distractor suppression mechanism was being overwhelmed as the number of targets approached capacity, particularly in Experiment 2 with highly similar distractors, as we concluded above based solely on the magnitude of the alpha signal. Here, we can see that, in Experiment 2, the colored distractors may have gotten through the perceptual filter and were encoded into working memory, based on the CDA amplitude. Interestingly, it seems that the impact of this leakage on behavioral performance was fairly minimal across Distractor set sizes. Our between-experiment analyses of behavioral performance showed no differences of working memory capacity (K) or accuracy rate between two experiments, verified by the null effects of Experiment (K: F(1, 34) = 0.199, p = .658, ηp2 = .006; accuracy rate: F(1, 34) = 0.506, p = .482, ηp2 = .015) and no evidence of an interaction of Experiment × Distractor set size on these measurements (K: F(3, 102) = 1.911, p = .133, ηp2 = .053; accuracy rate: F(3, 102) = 0.855, p = .467, ηp2 = .025). Bayes factors showed that the null hypothesis was 3.9 and 2.2 times more likely than the hypothesis that a difference existed in working memory capacity (K) and response accuracy between Experiment 1 and Experiment 2, respectively, demonstrating that we did not find a reliable difference of behavior performance between two experiments. Thus, our results show a case in which alpha activity was strongly modulated, apparently to prevent impact on behavior that could have compromised task performance.

Prefrontal bias signal.

Finally, to validate the sensitivity of the alpha-band modulations to the demands of distractor filtering, we compared the pattern of alpha-band effects to a frontal ERP component that previous researchers have proposed indexes distractor suppression (Liesefeld et al., 2014). Previous research has termed this the “prefrontal bias signal,” which appears about 250–300 msec after the memory item onset in the presence of distractors and has been proposed to reflect the filtering of distractors in working memory. Our analyses on the prefrontal bias signal revealed that its amplitude became more negative when participants were shown an array with any distractors compared to no distractors (see Figure 8A and B). Moreover, the comparisons of the prefrontal bias signal across each neighboring Distractor set size showed that the prefrontal bias signal was not sensitive to the number of distractors, suggesting that the prefrontal bias signal might reflect the triggering of an all-or-none mechanism related to suppression but not index the amount of work that a suppression mechanism needed to perform. These observations about the pattern of the prefrontal bias signal were statistically verified by both the results from Experiment 1 and from Experiment 2.

Figure 8. 

The prefrontal bias signal change with Serial position and Distractor set sizes in Experiment 1 and Experiment 2. (A) The prefrontal bias signal waveforms in Experiment 1 averaged over electrode pairs, namely, Fp1/Fp2, F3/F4, and F7/F8, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the prefrontal bias signal. (B) The mean prefrontal bias signal amplitude after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) The prefrontal bias signal waveforms in Experiment 2 averaged over the same electrode pairs as in Experiment 1, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). (D) The mean prefrontal bias signal amplitude after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). Error bars indicate the SEMs.

Figure 8. 

The prefrontal bias signal change with Serial position and Distractor set sizes in Experiment 1 and Experiment 2. (A) The prefrontal bias signal waveforms in Experiment 1 averaged over electrode pairs, namely, Fp1/Fp2, F3/F4, and F7/F8, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the prefrontal bias signal. (B) The mean prefrontal bias signal amplitude after each memory item in Experiment 1, separated by Distractor set sizes (black, yellow, magenta, red, and blue bars represent zero distractors [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). (C) The prefrontal bias signal waveforms in Experiment 2 averaged over the same electrode pairs as in Experiment 1, separated by Distractor set sizes (yellow, magenta, red, and blue lines represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). (D) The mean prefrontal bias signal amplitude after each memory item in Experiment 2, separated by Distractor set sizes (yellow, magenta, red, and blue bars represent one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively) and Serial positions (empty, dots, and gradient filled bars represent the first, second, and third array, respectively). Error bars indicate the SEMs.

In Experiment 1, the mean amplitude of the prefrontal bias signal was entered into a two-way repeated-measures ANOVA of Serial position (first vs. second vs. third item) and Distractor set size (0 vs. 1 vs. 2 vs. 3 vs. 4). This showed a significant Distractor set size effect, F(4, 68) = 6.618, p < .001, ηp2 = .280. However, the analysis revealed no Serial position effect, F(2, 34) = 0.003, p = .997, ηp2 = .000, or any interaction of Serial position × Distractor set size, F(8, 136) = 1.156, p = .330, ηp2 = .064. The preplanned pairwise comparisons of the mean prefrontal bias signal amplitude across each neighboring Distractor set size revealed that the prefrontal bias signal became more negative from zero distractors to one distractor, F(1, 17) = 6.874, p = .018, ηp2 = .288. However, its amplitude did not change across the other neighboring Distractor set sizes (one vs. two distractors, F(1, 17) = 0.813, p = .380, ηp2 = .046; two vs. three distractors, F(1, 17) = 0.491, p = .493, ηp2 = .028; three vs. four distractors, F(1, 17) = 0.043, p = .837, ηp2 = .003). Moreover, the prefrontal bias signal elicited by zero distractors was significantly more positive than all the other Distractor set sizes (Fs > 6.874, ps < .018).

We also computed Bayes factors to determine how much more likely it is that the null hypothesis is true than the hypothesis that the difference exists across the neighboring Distractor set sizes (more than zero distractors; Rouder et al., 2017). We found that the null hypothesis was, on average, 3.4 times more likely (Bayes factors range from 0.8 to 6.4, mean = 3.4, SD = 2.1) than the hypothesis that a difference existed across all of the pairwise comparisons of prefrontal bias signal amplitude in the conditions with greater than zero distractors present, demonstrating the pattern that prefrontal bias signal amplitude did not change across the other neighboring Distractor set sizes (i.e., when more than zero distractors were presented).

The pattern of effects on the prefrontal bias signal we observed in Experiment 1 could be because of the same shape and color of the distractors (all white, all circles), which may encourage participants to group them and reject the entire group in one processing step (Bundesen, 1990; Duncan & Humphreys, 1989), leading to the insensitivity of the prefrontal bias signal to the number of distractors. In Experiment 2, the distractors are colored circles, making it more difficult for participants to process them as a group. However, the prefrontal bias signal showed the same pattern as in Experiment 1, as shown in Figure 8C and D; that is, the amplitude of the prefrontal bias signal was not strongly modulated by number of distractors, and instead, this component appears to signal whether the array contains any potentially distracting information.

The mean prefrontal bias signal data from Experiment 2 were entered into a two-way repeated-measures ANOVA of Serial position (first vs. second vs. third item) and Distractor set size (1 vs. 2 vs. 3 vs. 4), and this yielded no Distractor set size effect, F(3, 51) = 2.496, p = .070, ηp2 = .128, and no interaction of Serial position × Distractor set size, F(6, 102) = 1.812, p = .104, ηp2 = .096. However, there was a significant Serial position effect, F(2, 34) = 3.661, p = .036, ηp2 = .177. The preplanned pairwise comparisons of the mean prefrontal bias signal amplitude across each neighboring Serial position revealed that the amplitude of the prefrontal bias signal decreased from the second array to the third array, F(1, 17) = 6.313, p = .022, ηp2 = .271, whereas no significant difference was found from the first array to the second array, F(1, 17) = 1.497, p = .238, ηp2 = .081. Moreover, the preplanned pairwise comparisons of the mean prefrontal bias signal amplitude across each neighboring Distractor set size showed that there were no differences between each neighboring distractor conditions (one vs. two distractors, F(1, 17) = 0.475, p = .500, ηp2 = .027; two vs. three distractors, F(1, 17) = 0.829, p = .375, ηp2 = .046; three vs. four distractors, F(1, 17) = 0.650, p = .431, ηp2 = .037). When we computed Bayes factors, we found that the null hypothesis was 9.4 times more likely than the hypothesis that a Distractor set size effect on prefrontal bias signal existed, demonstrating that the prefrontal bias signal amplitude was not influenced by the number of distractors.

Discussion

In Experiments 1 and 2, we found that the magnitude of alpha-band suppression tracked the number of distractors we presented, as well as how similar those distractors were to the targets. A strength of our approach was that the presence and set size of distractors are factors that front-end attention mechanisms filter out before working memory encoding (Vogel et al., 2005), such that their presence does not impair task performance. Consistent with this, we found that the amplitude of the CDA component of participants' ERPs was largely unaffected by these distractors. Thus, our findings support the predictions of the attentional hypothesis in which alpha suppression indexes the operation of attention, not working memory.

In our next experiment, we tested these hypotheses in a different way to provide converging evidence for our conclusions. Specifically, another characteristic of visual inputs that attention is believed to resolve before working memory encoding is the selection of task-relevant objects across space. In Experiment 3, we tested the prediction of the attentional hypothesis that alpha suppression should track the size of the region of space that needed to be selected.

EXPERIMENT 3

Our logic in this study was to manipulate factors that mechanisms of attention resolve before the subsequent visual working memory stage of processing to determine the locus of the alpha suppression effects. In Experiments 1 and 2, this was done by manipulating the potency of distractors that attention filters before working memory storage. In Experiment 3, we manipulated the Spatial extent of attention by varying the distance between multiple objects while keeping their numbers constant in low- and high-memory set size conditions. If alpha suppression indexes the work performed by attention selection, then we should see an increase in the strength of alpha suppression as the distance between the items increases and spreading attention across these items becomes more difficult. The competing working memory hypothesis predicts that alpha-band suppression will not track the Spatial extent of the to-be-remembered array, because behavioral experiments have shown that remembering the colors of objects is largely insensitive to the spatial layout of the remembered objects (Woodman et al., 2012; Phillips, 1974).

Materials and Methods

Participants

Eighteen participants from Vanderbilt University and the surrounding community participated in Experiment 3 (14 women, Mage = 22.5 years, SDage = 4.0). All participants gave informed consent before experimental procedures approved by the Vanderbilt University Institutional Review Board and received compensation of $15 per hour. All self-reported normal or corrected-to-normal visual acuity and normal color vision. Twelve participants' data in Experiment 3 were replaced because of excessive eye movements and muscular artifacts (described below in the EEG Analyses section). This is a higher rate of artifact rejection than we typically observe in human electrophysiology (Woodman, 2010); however, the difficulty of selecting the cued arrays of items appears to have been great enough that observers broke fixation to aid in the apprehension of these items. We analyzed the data of the replaced participants to determine if the pattern of results generalized to the trials in which they could maintain fixation. We found that the alpha activity and ERPs were the same as we show below.

Stimuli and Procedures

Stimuli were presented using MATLAB (R2017b 9.3.0; MathWorks) and the Psychophysics Toolbox (Version 3.0.12; Brainard, 1997) on a CRT monitor contained in a Faraday cage. Stimuli were presented on a gray background (x = 283, y = 314, L = 55.1 cd/m2, x and y define chromaticity, L represents luminance in the CIE xyY color space derived from the CIE 1931 color space). Participants were seated approximately 75 cm from the screen.

Figure 9 shows an example trial from Experiment 3. Each trial began with a display containing a black fixation cross (x = 349, y = 313, L = 0.11 cd/m2, 0.75° of visual angle) in the center of the screen for 500 msec, followed by a black arrow cue (1.3° of visual angle in width and 0.4° tall) presented 1.5° above the center fixation for 100 msec indicating the relevant side of the screen (right or left) to be attended. Participants were instructed to attend to the upcoming items appearing in the cued hemifield, keeping their eyes fixated at the center of the screen. After a postcue delay of 900 msec, we presented the memory arrays with either two or four colored squares. To prevent a sensory confound because of unilateral stimulation of the visual system (Woodman, 2010), we presented an equal number of items in both hemifields (i.e., both the task-relevant hemifield and task-irrelevant hemifield). Participants were instructed to only memorize the colored squares that appeared in the cued hemifield while ignoring the squares that presented in the uncued hemifield. The memory array appeared on the screen for 100 msec, which was followed by a 900-msec delay period before the appearance of the test array. The test array consisted of all colored squares presented in the memory array. On 50% of trials, the color of one square in the cued hemifield of the test array was replaced with another color. On these change trials, one of the colored squares would change to a color not previously seen in that hemifield on that trial. The locations of all the colored squares in the test array stayed the same as they appeared in the memory arrays, including the changed one. Participants then pressed one key on the keyboard to indicate that one of the colors of the squares had changed and pressed another key to indicate that all colors of the squares had stayed the same. The change versus no-change keys were counterbalanced across participants, with the two keys used being f and j.

Figure 9. 

Illustration of the experimental paradigm and conditions from Experiment 3. (Top) An example of the change detection task of Experiment 3 from an individual trial with four colored squares in the target hemifield. The cue arrow in the example trial points to the right side, instructing participants to attend to the upcoming items appearing in the right hemifield, keeping their eyes fixed on the center fixation. After a short delay, participants then pressed one key on the keyboard (f or j) to indicate whether one of the colored squares that appeared in the right hemifield of the test array had changed its color or not (key being counterbalanced) compared to the colored squares presented on the right side of the memory array. This example trial is a change one. (Bottom) The possible set sizes and the distances between the adjacent squares that could occur in Experiment 3. Each memory array could contain two or four colored squares, and the distance between each adjacent colored square could be 1°, 3°, or 5° of visual angle, with small, medium, or large Spatial extent, respectively.

Figure 9. 

Illustration of the experimental paradigm and conditions from Experiment 3. (Top) An example of the change detection task of Experiment 3 from an individual trial with four colored squares in the target hemifield. The cue arrow in the example trial points to the right side, instructing participants to attend to the upcoming items appearing in the right hemifield, keeping their eyes fixed on the center fixation. After a short delay, participants then pressed one key on the keyboard (f or j) to indicate whether one of the colored squares that appeared in the right hemifield of the test array had changed its color or not (key being counterbalanced) compared to the colored squares presented on the right side of the memory array. This example trial is a change one. (Bottom) The possible set sizes and the distances between the adjacent squares that could occur in Experiment 3. Each memory array could contain two or four colored squares, and the distance between each adjacent colored square could be 1°, 3°, or 5° of visual angle, with small, medium, or large Spatial extent, respectively.

In Experiment 3, the colors of the squares were chosen from a pool of seven highly discriminable colors: red (x = 631, y = 328, L = 11.8 cd/m2), green (x = 276, y = 593, L = 48.1 cd/m2), blue (x = 145, y = 66, L = 5.86 cd/m2), magenta (x = 284, y = 140, L = 16.5 cd/m2), yellow (x = 390, y = 507, L = 72.8 cd/m2), white (x = 281, y = 314, L = 76.2 cd/m2), and black (x = 349, y = 313, L = 0.11 cd/m2). Sets of colors were randomly chosen from the color pool, without replacement for the squares in a given hemifield. To manipulate the spatial distribution of the to-be-remembered objects, the colored squares (0.65° of visual angle in width and 0.65° tall) were always presented in a line (horizontal or vertical) at Set Size 2 or in a square at Set Size 4 with a distance of 1°, 3°, or 5° of visual angle between the center of each two adjacent colored squares (as illustrated in Figure 9). In a given trial, the distance between the colored squares was always the same in both the left and right hemifields. To control for other perceptual factors, such as saccades that could influence the function of attention, the matrices of the colored squares (a line at Set Size 2 or a square at Set Size 4) were placed inside the same space area, with the furthest colored square being placed along the circumference of an imaginary circle with the eccentricity of 7° of visual angle to the center of the fixation. To prevent participants from knowing in advance the number of to-be-remembered items and the distance between them in an upcoming trial, all trials with different set sizes (two or four colored squares) and different distances between each adjacent colored square (1°, 3°, or 5° of visual angle) were randomly interleaved within a given block. All participants completed 192 trials with each condition for 1152 trials in total for Experiment 3.

Change Detection Task Performance

We calculated the accuracy of remembering the colors and the working memory capacity in Experiment 3 using the same methods as in Experiments 1 and 2.

EEG Acquisition

The EEG was recorded in an electrically shielded, soundproof booth from a 20-channel cap (Electro-Cap International), embedded with tin electrodes that make contact with the skin through electrode gel. The 20 electrodes were positioned according to the International 10–20 system (F3, F4, C3, C4, P3, P4, PO3, PO4, O1, O2, PO7, PO8, T3, T4, T5, T6, Fz, Cz, and Pz). Impedance values were kept below 4 kΩ. The reference electrodes were affixed with stickers to the left and right mastoids, and a ground electrode was placed in the elastic cap at Fpz. Data were referenced online to the right mastoid and rereferenced offline to the algebraic average of the left and right mastoids. EOG activity was recorded with two HEOG electrodes placed ∼1 cm lateral to the outer canthi of the two eyes, and one vertical EOG electrode was placed below the right eye to detect eye movements and blinks. All channels were bandpass filtered from 0.01 to 100 Hz and recorded with a 250-Hz sampling rate.

EEG Analyses

Artifact rejection.

Trials were rejected using the same three-step procedure used in Experiments 1 and 2 with two exceptions: First, the rejection interval of the segmentations is from the onset of the memory array to 1000 msec (instead of 2600 msec in Experiments 1 and 2) after it because there was only single memory array in Experiment 3. Second, we used a threshold of ±3 μV (Woodman & Luck, 2003) to reject trials contaminated by HEOG because of a relatively short display time of the memory array in Experiment 3, unlike the three sequentially presented memory arrays in Experiments 1 and 2. After the rejection, participants were replaced when their data did not contain enough trials in each cell for analyses. In Experiment 3, 12 participants were replaced because of excessive eye movements and artifacts. For the remaining participants, an average of 2.04% (SD = 2.64%) of trials were excluded. It appears our sample included a run of participants who found fixation particularly difficult. We note that the same patterns of results were observed when we focused our analyses on the usable data from the replaced participants.

Alpha power analysis.

To measure the alpha power change on each trial, the raw EEG data were first bandpass filtered by the eegfilt.m function and then Hilbert transformed by the hilbert.m function in MATLAB to obtain the instantaneous power values for the alpha band (8–12 Hz). Percent change in alpha power was then calculated relative to a baseline period before the onset of the cue (−400 to 0 msec relative to the cue onset, −1400 to −1000 msec relative to the memory array onset) similar to previous studies' analyses of this frequency band (e.g., Wang et al., 2019; Adam et al., 2018). Next, the ipsilateral alpha power and the contralateral alpha power were calculated by averaging the percent change in alpha power for ipsilateral electrodes relative to the cued hemifield and contralateral electrodes relative to the cued hemifield, respectively. Global alpha power was calculated by averaging the percent change in alpha power for both ipsilateral electrodes and contralateral electrodes across the same lateral-occipital and posterior-parietal electrodes we used in Experiments 1 and 2 (i.e., P3, P4, T7, T8, PO3, PO4, PO7, PO8, O1, and O2; Adam et al., 2018; Fukuda et al., 2015).

CDA analysis.

To relate the alpha-band activity that we measured to a canonical electrophysiological measure of visual working memory storage, we also calculated the CDA in Experiment 3. Before calculating the CDA, the signal was first low-pass filtered with a two-way least squares finite impulse response filter function from the EEGLAB Toolbox (eegfilt.m; Delorme & Makeig, 2004) with the low pass at 30 Hz to remove the high-frequency noise of the EEG signal. After filtering, trials were baseline corrected to the mean amplitude of the ERPs from −200 to 0 msec relative to the onset of the memory array. The CDA amplitude was then computed by subtracting the activity of the electrodes ipsilateral to the cued hemifield from the activity of the electrodes contralateral to the cued hemifield across the lateral-occipital and posterior-parietal electrodes: P3, P4, T7, T8, PO3, PO4, PO7, PO8, O1, and O2.

Experimental Design and Statistical Analyses

To examine the influence of set size and Spatial extent on behavioral performance in Experiment 3, separate two-way repeated-measures ANOVAs with the within-participant factor of Set size (2 or 4) and Spatial extent (small, medium, or large) were performed on the behavioral estimates (accuracy and K), the alpha power percent change, and the CDA amplitude. Preplanned pairwise comparisons with Bonferroni corrections for multiple comparisons were used across each neighboring Spatial extent to verify the nature of the Spatial extent effects. All these statistical analyses were performed in SPSS 19.0 (IBM Inc.).

Results

Behavior

Figure 10 shows our behavioral findings in Experiment 3. We found that our manipulation of the distance between the adjacent colored squares had no impact on participants' ability to remember either two or four targets presented on each trial.

Figure 10. 

Behavioral performance of Experiment 3. The mean visual working memory capacity (left) and the mean accuracy rate (right) estimates for different set sizes and Spatial extents. The x axis represents Spatial extent (small, medium, or large), and the y axis represents visual working memory capacity (K; left) or accuracy rate (right). Red and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs.

Figure 10. 

Behavioral performance of Experiment 3. The mean visual working memory capacity (left) and the mean accuracy rate (right) estimates for different set sizes and Spatial extents. The x axis represents Spatial extent (small, medium, or large), and the y axis represents visual working memory capacity (K; left) or accuracy rate (right). Red and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs.

Participants' capacity estimates (K) and change-detection accuracy were entered into separate two-way repeated-measures ANOVAs with within-participant factors of Set size (2 or 4) and Spatial extent (small, medium, or large). The analysis of capacity estimates (K) revealed a significant Set size effect, with a significant increase of WM capacity from Set Size 2 to Set Size 4, F(1, 17) = 65.910, p < .001, ηp2 = .795. However, there was no significant Spatial extent effect, F(2, 34) = 0.779, p = .467, ηp2 = .044, nor any interaction between Set size and Spatial extent, F(2, 34) = 0.590, p = .560, ηp2 = .034. Similarly, the analysis on raw accuracy yielded a significant Set size effect, with a significant decrease of the response accuracy from Set Size 2 to Set Size 4, F(1, 17) = 34.882, p < .001, ηp2 = .672. However, there was no significant Spatial extent effect, F(2, 34) = 0.045, p = .956, ηp2 = .003, nor any interaction between Set size and Spatial extent, F(2, 34) = 0.137, p = .873, ηp2 = .008. Thus, our behavioral findings from Experiment 3 suggest that presenting the same amount of targets across different Spatial extents had no influence on the storage of these colors in visual working memory.

To provide stronger inferential power regarding the null results of the Spatial extent effect on behavior performance, we also computed Bayes factors to determine how much more likely the null hypothesis than the possibility that different Spatial extent did show an effect (Rouder et al., 2017). We found the null hypothesis was 11.1 and 11.5 times more likely than the hypothesis that a Spatial extent effect on working memory capacity (K) and response accuracy existed, demonstrating that we did not find a reliable modulation of different Spatial extents on behavior performance.

EEG and ERPs

Alpha power suppression.

In contrast to the participants' behavior that was invariant across Spatial extent, the alpha power suppression was modulated by both the number of targets and the Spatial extent. Figure 11A shows the pattern of the alpha power suppression changes with set size and the Spatial extent. We measured stronger global alpha power suppression as more targets were simultaneously presented in a memory array, consistent with recent findings that varied memory set size using multiobject arrays (Fukuda et al., 2015; Sauseng et al., 2009). In addition, we found that this alpha power suppression increased as the Spatial extent increased, supporting our hypothesis that alpha-band suppression indexes a mechanism of attentional selection that serves to control the Spatial extent of attention. These observations about the pattern of the global alpha power in Experiment 3 were statistically verified with the following steps.

Figure 11. 

The alpha power suppression across set sizes and Spatial extents in Experiment 3. (A) Left: the global alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. The solid lines and dashed lines represent Set Size 2 and Set Size 4, respectively. Blue, red, and green lines represent small, medium, and large Spatial extent, respectively. The cyan bar and the black bar on the time axis represent the onset and duration of the cue and the memory array. The gray interval represents the time window of the global alpha power. Right: the mean global alpha power suppression after the onset of the memory array, separated by set size and Spatial extent. The x axis represents Spatial extent (small, medium, or large), and the y axis represents the mean global alpha power suppression. Red bars and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs. (B) Left: the ipsilateral alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. Right: the mean ipsilateral alpha power suppression after the onset of the memory array, separated by set size and Spatial extent. (C) Left: the contralateral alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. Right: the mean contralateral alpha power suppression after the onset of the memory array, separated by set size and Spatial extent.

Figure 11. 

The alpha power suppression across set sizes and Spatial extents in Experiment 3. (A) Left: the global alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. The solid lines and dashed lines represent Set Size 2 and Set Size 4, respectively. Blue, red, and green lines represent small, medium, and large Spatial extent, respectively. The cyan bar and the black bar on the time axis represent the onset and duration of the cue and the memory array. The gray interval represents the time window of the global alpha power. Right: the mean global alpha power suppression after the onset of the memory array, separated by set size and Spatial extent. The x axis represents Spatial extent (small, medium, or large), and the y axis represents the mean global alpha power suppression. Red bars and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs. (B) Left: the ipsilateral alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. Right: the mean ipsilateral alpha power suppression after the onset of the memory array, separated by set size and Spatial extent. (C) Left: the contralateral alpha power suppression in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. Right: the mean contralateral alpha power suppression after the onset of the memory array, separated by set size and Spatial extent.

To quantify the global alpha effects in Experiment 3, we first calculated the mean global alpha suppression after the memory array (averaged over 300–700 msec after the presentation of the memory array) for each set size and Spatial extent. The two-way repeated-measures ANOVA of Set size (2 vs. 4) and Spatial extent (small vs. medium vs. large) on the mean global alpha suppression revealed a significant Set size effect, F(1, 17) = 8.146, p = .019, ηp2 = .209, with a significantly stronger alpha power suppression at Set Size 4 than Set Size 2. Most importantly, for the hypothesis, the analysis also revealed a significant Spatial extent effect, F(2, 34) = 20.714, p < .001, ηp2 = .549. The preplanned pairwise comparisons across each neighboring Spatial extent show that the global alpha power suppression significantly increased from the small Spatial extent to the medium Spatial extent, F(1, 17) = 5.347, p = .034, ηp2 = .239, and kept increasing from the medium Spatial extent to the large Spatial extent, F(1, 17) = 17.943, p = .001, ηp2 = .513. Finally, there was no interaction of Set Size × Spatial extent, F(2, 34) = 0.805, p = .455, ηp2 = .045.

The findings that the global alpha suppression changes with set size and Spatial extent support our hypothesis that this signal indexes an attention mechanism that serves to control the size of the focus of attention. In parallel with our analyses from Experiments 1 and 2, we next compared the alpha activity contralateral to the target objects with the alpha activity ipsilateral to the target objects to test the conjecture that these lateralized signals measure different cognitive mechanisms. To measure these lateralized signals, we calculated the average alpha suppression of the lateral-posterior electrodes ipsilateral to the task-relevant hemifield and the electrodes contralateral to the task-relevant hemifield, respectively. The results of this lateralized averaging are shown in Figure 11B and C. As evidenced by these means, the patterns of alpha suppression were similar in both hemispheres, with both ipsilateral and contralateral hemispheres exhibiting the parametric modulation across set sizes and Spatial extents.

A three-way repeated-measures ANOVA with within-participant factors of Hemisphere (ipsilateral vs. contralateral), Set size (2 vs. 4), and Spatial extent (small vs. medium vs. large) on the mean ipsilateral and contralateral alpha suppression revealed a significant main effect of Hemisphere, F(1, 17) = 10.146, p = .005, ηp2 = .374, because of the contralateral alpha suppression being stronger than the ipsilateral alpha suppression. Moreover, there was a significant Set size effect, F(1, 17) = 8.146, p = .011, ηp2 = .324, with a significantly stronger alpha power suppression at Set Size 4 than Set Size 2. The analysis also revealed a significant Spatial extent effect, F(2, 34) = 20.714, p < .001, ηp2 = .549. The preplanned pairwise comparisons across each neighboring Spatial extent show that both the contralateral and ipsilateral alpha power suppression significantly increased from the small Spatial extent to the medium Spatial extent, F(1, 17) = 5.347, p = .034, ηp2 = .239, and kept increasing from the medium Spatial extent to the large Spatial extent, F(1, 17) = 17.943, p = .001, ηp2 = .513. However, we found no interaction of Hemisphere × Set Size, F(1, 17) = 2.215, p = .155, ηp2 = .115, no interaction of Hemisphere × Spatial extent, F(2, 34) = 0.717, p = .496, ηp2 = .040, and no interaction of Hemisphere × Set Size × Spatial extent, F (2, 34) = 0.572, p = .570, ηp2 = .033. These results suggest that contralateral and ipsilateral alpha activity behave similarly as a function of memory set size and the Spatial extent of attention.

In summary, in Experiment 3, we found that, when we spread the to-be-remembered items across a larger Spatial extent, alpha power suppression increased, confirming the predictions of the attentional hypothesis that alpha power suppression measures the operation of a perceptual attention mechanism that operates early in the flow of information processes, before working memory storage. In Experiment 3, we also replicated previous studies showing that the alpha power suppression is stronger when larger arrays of task-relevant objects are presented.

CDA.

We next wanted to compare the pattern of alpha power to an established measure of working memory storage, the CDA (for a review, see Luria et al., 2016). We had based the logic of our manipulation on the fact that Spatial extent does not seem to influence how colored squares are stored in visual working memory (e.g., Woodman et al., 2012). However, by measuring the CDA, we can validate this assumption and strengthen our conclusions that alpha is tracking the operation of attention and not working memory storage per se. As shown in Figure 12, we found that the CDA amplitude increased as the working memory set size increased. Unlike alpha suppression, the Spatial extent had no influence on the CDA amplitude. These observations about the pattern of the CDA effects were supported by the following statistical analyses.

Figure 12. 

The CDA activity changes with set size and Spatial extent in Experiment 3. (A) The CDA waveforms in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. The solid lines and dashed lines represent Set Size 2 and Set Size 4, respectively. Blue, red, and green lines represent small, medium, and large Spatial extent, respectively. The black bar on the time axis represents the onset and duration of the memory array. The gray interval represents the time window of the CDA. (B) The mean CDA amplitude after the onset of the memory array, separated by set size and Spatial extent. The x axis represents Spatial extent (small, medium, or large), and the y axis represents the mean CDA amplitude (averaged across the gray area in A). Red bars and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs.

Figure 12. 

The CDA activity changes with set size and Spatial extent in Experiment 3. (A) The CDA waveforms in Experiment 3 averaged over electrode pairs, namely, P3/P4, T7/T8, PO3/PO4, PO7/PO8, and O1/O2, separated by set size and Spatial extent. The solid lines and dashed lines represent Set Size 2 and Set Size 4, respectively. Blue, red, and green lines represent small, medium, and large Spatial extent, respectively. The black bar on the time axis represents the onset and duration of the memory array. The gray interval represents the time window of the CDA. (B) The mean CDA amplitude after the onset of the memory array, separated by set size and Spatial extent. The x axis represents Spatial extent (small, medium, or large), and the y axis represents the mean CDA amplitude (averaged across the gray area in A). Red bars and blue bars represent Set Size 2 and Set Size 4, respectively. Error bars indicate the SEMs.

We calculated the mean amplitude of the CDA after the memory array (400–700 msec after the memory array onset) across different set sizes and Spatial extents. The two-way repeated-measures ANOVA of Set size (2 vs. 4) and Spatial extent (small vs. medium vs. large) on the mean CDA amplitude revealed a significant Set size effect, F(1, 17) = 34.959, p < .001, ηp2 = .673, with a more negative CDA amplitude at Set Size 4 than Set Size 2. However, there was no significant Spatial extent effect, F(2, 34) = 2.179, p = .129, ηp2 = .114, nor a significant interaction of Set Size × Spatial extent, F(2, 34) = 0.355, p = .704, ηp2 = .020. We found the null hypothesis was 2.6 times more likely than the hypothesis that a Spatial extent effect on CDA amplitude existed, supporting our observation that we did not find a reliable modulation of different Spatial extents on CDA amplitude. These findings are consistent with our observation of the behavioral patterns of the capacity (K) estimates, supporting a selective role of the CDA in tracing the number of the target objects instead of the Spatial extent.

Discussion

In Experiment 3, we found that the strength of alpha suppression tracked the spatial distribution of task-relevant items in the arrays we presented to participants. Classic research in cognitive psychology indicates that the focus of attention is adjustable, such that attention works like a zoom lens to select arrays of task-relevant stimuli that vary in their spatial distribution (Eriksen & St. James, 1986). Here, we showed that alpha tracked this spread of attention across space, with stronger alpha suppression when attention needed to cover more areas of the visual field. Whereas much of the research measuring alpha-band activity has manipulated the characteristics of task-relevant stimuli to see how alpha is influenced, our findings in all three experiments show that alpha is sensitive to these factors of physical stimulus array characteristics that are handled by front-end perceptual attention mechanisms.

GENERAL DISCUSSION

The goal of this study was to test the competing hypotheses that the alpha-band activity elicited by visual stimulus arrays measures a mechanism of attention versus a mechanism of working memory storage. We focused on two aspects of the visual stimuli that perceptual attention is believed to mitigate before working memory storage. Specifically, two of the classic functions that perceptual attention has been proposed to mitigate is the potential interference from distractors and the size of the focus of attention. To this end, we measured the alpha-band suppression in question while we simultaneously manipulated the target load, the distractor load, and the spacing of the items in the arrays. Consistent with previous behavioral work, we found that neither the distractors nor spacing between items influenced performance in the color change-detection task. However, we found that the magnitude of alpha-band suppression tracked both of these aspects of the visual arrays.

The careful reader might be worried about the fact that the participants in all experiments exhibited behavioral performance that was essentially at ceiling and indifferent to the presence of the distractors or their spatial arrangement. However, these high levels of performance fit well with our goal of examining the role of alpha-band activity in attention mechanisms that resolve perceptual difficulties before working memory storage and are consistent with previous behavior findings that attention could effectively resolve competition from distracting stimuli (Wolfe, 1998; Luck et al., 1997; Chelazzi et al., 1993; Broadbent, 1957). More specifically, the distractors were likely easy to filter in the current study because all of the distractors were the same shape and color (all white circles in Experiment 1 and all circles in Experiment 2). This grouping may have encouraged participants to group them and reject the entire group in one processing step (Bundesen, 1990; Duncan & Humphreys, 1989). Clearly, the job performed by alpha occurs before these distractors are rejected. Thus, our study shows a clear dissociation between the sensitivity of alpha-band power to distractors and spacing, compared to its relative insensitivity at the behavioral level.

We contrasted the parametric sensitivity of the alpha index to the demands of distractor suppression with the modulations of the simultaneously measured ERP components that previous work had associated with the storage of targets (i.e., the CDA) and the need to suppress or filter distracting information (the prefrontal bias signal; Liesefeld et al., 2014). We found that the CDA increased with each new target that needed to be encoded into working memory but was relatively unaffected by the distractors that strongly modulated the alpha-band activity, until working memory was full of targets, at which point it appears that filter may break down, as previously predicted (Vogel et al., 2005). Moreover, whereas the prefrontal bias signal was modulated by the presence of any distractors at all, the alpha signal was parametrically manipulated by the number of distractors in each array.

One seemingly possible explanation for the Distractor set size effects on alpha power suppression that we found in Experiments 1 and 2 is that these effects were only a result of more visual stimuli being presented. In other words, the alpha power suppression may index the total sensory input instead of some kind of attention mechanism. Consistent with this view, we found that the amplitude of the N1 component showed the same pattern of Distractor set size effects that the subsequent alpha suppression showed (see Figure 13 for an example from Experiment 1), and the visual N1 component is believed to be generated by the sensory processing of visual inputs (Woodman, 2010). However, it is unlikely that alpha suppression is because of a purely sensory mechanism for the following reasons. If alpha power suppression simply reflects the total amount of visual input, then we should see its amplitude only increase as the number of distractors increases. However, we found larger alpha suppression in Experiment 2 than Experiment 1, because of Experiment 2 having higher target–distractor similarity, although the number of stimuli was the same across Distractor set sizes in both experiments. In addition, experiments that have manipulated set sizes across a large range have found that alpha suppression stops increasing even as the total amount of sensory stimulation continues to increase (e.g., Fukuda & Woodman, 2017). Moreover, the results of Experiment 3 showed that the strength of alpha power suppression was modulated by the Spatial extent of the array of task-relevant items, even when the number of the task-relevant items was kept constant. Obviously, this rules out the simple idea that alpha power suppression simply reflects the total energy of visual input. Thus, our findings of alpha power suppression modulated by Distractor set size, similarity, and spacing suggest that it reflects an attentional mechanism instead of simply being a lingering sensory response, although alpha suppression does mirror the early ERP responses in much of the present data set.

Figure 13. 

The N1 component changes with Distractor set sizes in Experiment 1. The ERP waveforms in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractor [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the N1 component, whose amplitude shows a trend of linear increase with the increase of Distractor set sizes.

Figure 13. 

The N1 component changes with Distractor set sizes in Experiment 1. The ERP waveforms in Experiment 1 averaged over electrode pairs, namely, P3/P4, P7/P8, PO3/PO4, PO7/PO8, and O1/O2, separated by Distractor set sizes (black, yellow, magenta, red, and blue lines represent zero distractor [D0], one distractor [D1], two distractors [D2], three distractors [D3], and four distractors [D4], respectively). The black bars on the time axis represent the onset and duration of each memory array. The gray intervals represent the time window of the N1 component, whose amplitude shows a trend of linear increase with the increase of Distractor set sizes.

In this study, we tested competing predictions of an early attentional account of alpha suppression with an account that proposes these signals arise because of later working-memory-related processes. However, within the theoretical space of an attentional account of alpha, two competing views have been proposed with regard to the function of alpha-band activity: distractor suppression or target enhancement. Empirical studies using spatial cueing tasks and lateralized working memory tasks have shown that alpha-band activity increases ipsilateral to the cued locations and decreases contralateral to the cued locations, with distractor suppression accounts proposing that higher alpha-band power reflects increased suppression of irrelevant stimuli (e.g., Payne & Sekuler, 2014; Händel, Haarmeier, & Jensen, 2011; Jensen & Mazaheri, 2010). The target enhancement account emphasizes that the ipsilateral alpha power increase accompanies the contralateral alpha power decrease to the cued hemifield. Under this perspective, the lower alpha power reflects an increase in target signal enhancement (see a recent review by Foster & Awh, 2019). However, dissociating these two accounts is difficult because enhancing targets in the cued hemifield and suppressing the task-irrelevant distractors in the uncued hemifield may happen at the same time. The present findings are potentially relevant to this debate because we increased the number of distractors while keeping the number of targets constant in each memory array. Because alpha suppression increased as the number of distractors increased, as well as their similarity to the memory targets, it would be intuitive to conclude that alpha-band activity tracks distractor suppression, as we may have implied in the present article in our description of distractor interference. However, it could be that increasing the number of distractors causes a target enhancement mechanism to work harder. For example, if you were talking to a friend in a room that was growing increasingly loud because of the arrival of more loud-talking people, then a natural countermeasure would be to talk more loudly so you can be heard. Although this is fundamentally about increasing the intensity of the target signal, it is done in reaction to the distractors. In the same way, it is possible that the present findings reflect signal enhancement, not distractor suppression, by an early attention mechanism (Noonan et al., 2016; also see a recent review by Foster & Awh, 2019). Thus, we believe that we can firmly conclude that alpha tracks an early attentional selection mechanism, with the flavor of that selection mechanism still to be definitively determined.

Mirroring the sensitivity of alpha to the presence and number of distractors, we found that the strength of alpha-band activity was also sensitive to the Spatial extent of an array of task-relevant items. Again, we found that manipulating the Spatial extent of the memory array had no influence on participants' behavioral performance, showing another example in which attention mitigates a factor of the stimulus input that shields downstream processes like memory storage. This effect would seem to provide a potential alternative explanation for previous findings in which the set size of task-relevant information was varied (Fukuda & Woodman, 2017). That is, previous reports that interpreted set-size-dependent alpha suppression as being an index of the processing of the targets themselves might instead be largely because of inherent confound in which presenting arrays with larger memory set sizes takes up more space in the visual field and requires the participant to select a larger or smaller area of the visual field with attention. Thus, the present findings may require us to reconsider the theoretical meaning of previous empirical reports. In addition to demonstrating the utility of alpha-band suppression to study the ability of the brain to deploy attention, our findings impact our theoretical understanding of attention.

Theories of visual attention have long debated whether the selection of targets and the suppression of distractors are performed by a unitary mechanism or are separable attentional subroutines (Hickey, Di Lollo, & McDonald, 2009; Ruff & Driver, 2006; Carrasco, Ling, & Read, 2004; Awh, Matsukura, & Serences, 2003; Lu, Lesmes, & Dosher, 2002; Carrasco, Penpeci-Talgar, & Eckstein, 2000; Dosher & Lu, 2000; Cheal & Gregory, 1997; Hawkins et al., 1990). Unitary models have been favored by some because of parsimony and the fact that the same neurons in the visual system show increased firing rates when a target is in its receptive field and show decreased firing rates when a distractor appears in its receptive field, particularly when the distractor flanks the target in the visual field (e.g., Thompson, Hanes, Bichot, & Schall, 1996). However, recent work in human electrophysiology has supported the proposal that separate mechanisms perform target selection and distractor suppression (Sawaki & Luck, 2014; Sawaki, Geng, & Luck, 2012; Hickey et al., 2009). The present findings appear problematic for models that propose a unitary attention mechanism that is used to keep multiple target representations active in mind and to select new perceptual inputs from potentially distracting information from the visual field (Cowan, 2001, 2012; Jonides et al., 2008; Oberauer, 2002). Our observation that alpha-band suppression is sensitive to the distractor load, but not the target load that accumulates across the trial, is contrary to predictions of a unitary attention mechanism model, as we should have observed both factors increase the magnitude of suppression if both operations are performed by the same mechanism.

Of further potential theoretical impact, we observed independence between the alpha-band suppression signal and other components that have been proposed to specialize in target processing. The CDA is believed to provide a sensitive index of the number of objects simultaneously maintained in visual working memory (Vogel et al., 2005; Vogel & Machizawa, 2004). Consistent with this idea, we found that this ERP component did track the number of colored targets as the trial unfolded. It was relatively insensitive to the presence of distractors or the spatial layout of the targets. The exception to this was when the number of targets approached capacity (i.e., three objects; Luck & Vogel, 2013; Vogel et al., 2001). After the presentation of the third target, we did observe that the CDA was larger in amplitude when distractors were present than when they were not. Previous work has suggested that, when we are tasked with an informational load that taxes working memory capacity, attentional filtering may breakdown (Vogel et al., 2005). Our simultaneous measurements of alpha and the CDA suggest that we can observe the dynamics of this breakdown as working memory is sequentially loaded with items in the face of distraction.

Although our manipulations in this study were done to isolate the potential contribution of variation in the number of distractors and spatial deployment of attention on alpha, these experiments leave open an important question. It is possible that alpha-band suppression might index a mechanism that serves to not only filter distracting objects and control the focus of attention but also filter distracting features within target objects themselves. We do not yet know whether this alpha suppression mechanism operates on fine-grained representations. Does it operate to filter irrelevant features from our representations of task-relevant information, as researchers have previously proposed is possible (e.g., Woodman & Vogel, 2008)? One useful avenue for future research would be to manipulate the nature of the targets that need to be selected. It is possible that, when targets are defined by a single feature (e.g., color), the presence of additional irrelevant object features could modulate alpha suppression, with each additional distracting feature increasing alpha suppression (e.g., shape, orientation, the presence of a bar). In contrast, it is possible that the alpha suppression mechanism only operates on grouped object representations (Vecera, Behrmann, & McGoldrick, 2000; Vecera & Farah, 1994), so alpha would not vary in this way.

Our findings stand in contrast to recent work that attempted to test a similar hypothesis about alpha indexing attentional selection versus distractor suppression (Bacigalupo & Luck, 2019). That study manipulated the distance between a target and a distractor and reported a nonlinear effect in which lateralized alpha suppression was strongest when the distance between a target and a distractor was in a sweet spot. Here, we show that the relationship between alpha suppression and distractors may be more straightforward in that it tracks the spatial deployment of attention necessary to filter out distracting information. However, in our experiments, the minimum distance between objects was outside the range studied in this previous article, and the increases in alpha suppression are difficult to account for with such a crowding explanation alone. We believe that experiments are needed that contrast spatial proximity with the number of distractors presented, with such experiments addressing the spatial version of the granularity question raised in our previous paragraph.

It is also worth noting that alpha power suppression, sometimes referred to as alpha power desynchronization, functions differently from alpha power synchronization. Previous studies have shown stronger alpha power synchronization (i.e., an increase in alpha power) across electrode sites ipsilateral to task-irrelevant items presented in an uncued hemifield (Sauseng et al., 2009) and increased alpha power synchronization elicited by the presence of a task-irrelevant distractor compared to the presence of task-relevant items in a sequentially presented memory task (Payne, Guillory, & Sekuler, 2013). Thus, the present findings observed across our three experiments appear to be surprising in not clearly showing this same pattern of alpha power increases as was observed in these two other experiments, although the present arrays used much denser arrays of distractors and heavier memory demands. Future studies are needed to verify the functional difference of alpha power synchronization relative to the desynchronization measures of attentional filtering and distractor suppression that we observed here.

In conclusion, the present findings challenge previous proposals about the nature of the cognitive mechanism that is indexed by alpha suppression after the onset of a visual array. Recent work had proposed that alpha might index attending to task-relevant information (Bacigalupo & Luck, 2019; Fukuda et al., 2016), storing that information in working memory (Fukuda & Woodman, 2017; Fukuda et al., 2015), or even be related to action selection (Deiber et al., 2012). Our findings suggest that alpha-band activity after stimulus onset is related to how the visual system deploys attention to handle potentially distracting objects and appears to be distinct from mechanisms that select target representations or store them in working memory.

Author Contributions

Sisi Wang: Conceptualization; Data curation; Formal analysis; Investigation; Project administration. Emma E. Megla: Data curation; Formal analysis; Investigation; Project administration. Geoffrey F. Woodman: Conceptualization; Formal analysis; Funding acquisition; Supervision; Writing-review & editing.

Funding Information

National Eye Institute (http://dx.doi.org/10.13039/100000053), Grant number: P30-EY08126, R01-EY019882, T32-EY007135. National Institute of Mental Health (http://dx.doi.org/10.13039/100000025), Grant number: R01-MH110378.

Acknowledgments

The present work was supported by grants from the National Institutes of Health (R01-EY019882, R01-MH110378, P30-EY08126, and T32-EY007135) to G. F. W. and a China Scholarship Council scholarship (201706140083) to S. W.

Reprint requests should be sent to Geoffrey F. Woodman, Department of Psychology, Vanderbilt University, 2301 Vanderbilt Place, Nashville, TN 37240, or via e-mail: geoffrey.f.woodman@vanderbilt.edu.

REFERENCES

REFERENCES
Adam
,
K. C. S.
,
Robison
,
M. K.
, &
Vogel
,
E. K.
(
2018
).
Contralateral delay activity tracks fluctuations in working memory performance
.
Journal of Cognitive Neuroscience
,
30
,
1229
1240
.
Awh
,
E.
,
Matsukura
,
M.
, &
Serences
,
J. T.
(
2003
).
Top–down control over biased competition during covert spatial orienting
.
Journal of Experimental Psychology: Human Perception and Performance
,
29
,
52
63
.
Bacigalupo
,
F.
, &
Luck
,
S. J.
(
2019
).
Lateralized suppression of alpha-band EEG activity as a mechanism of target processing
.
Journal of Neuroscience
,
39
,
900
917
.
Berger
,
H.
(
1929
).
Über das Elektrenkephalogramm des Menschen
.
Archiv für Psychiatrie und Nervenkrankheiten
,
87
,
527
570
.
Brainard
,
D. H.
(
1997
).
The Psychophysics Toolbox
.
Spatial Vision
,
10
,
433
436
.
Broadbent
,
D. E.
(
1957
).
A mechanical model for human attention and immediate memory
.
Psychological Review
,
64
,
205
215
.
Bundesen
,
C.
(
1990
).
A theory of visual attention
.
Psychological Review
,
97
,
523
547
.
Buzsáki
,
G.
, &
Freeman
,
W.
(
2015
).
Editorial overview: Brain rhythms and dynamic coordination
.
Current Opinion in Neurobiology
,
31
,
v
ix
.
Carrasco
,
M.
,
Ling
,
S.
, &
Read
,
S.
(
2004
).
Attention alters appearance
.
Nature Neuroscience
,
7
,
308
313
.
Carrasco
,
M.
,
Penpeci-Talgar
,
C.
, &
Eckstein
,
M.
(
2000
).
Spatial covert attention increases contrast sensitivity across the CSF: Support for signal enhancement
.
Vision Research
,
40
,
1203
1215
.
Cheal
,
M. L.
, &
Gregory
,
M.
(
1997
).
Evidence of limited capacity and noise reduction with single-element displays in the location-cuing paradigm
.
Journal of Experimental Psychology: Human Perception and Performance
,
23
,
51
71
.
Chelazzi
,
L.
,
Miller
,
E. K.
,
Duncan
,
J.
, &
Desimone
,
R.
(
1993
).
A neural basis for visual search in inferior temporal cortex
.
Nature
,
363
,
345
347
.
Cohen
,
M. X.
(
2017
).
Where does EEG come from and what does it mean?
Trends in Neurosciences
,
40
,
208
218
.
Cooper
,
R.
,
Winter
,
A. L.
,
Crow
,
H. J.
, &
Walter
,
W. G.
(
1965
).
Comparison of subcortical, cortical and scalp activity using chronically indwelling electrodes in man
.
Electroencephalography and Clinical Neurophysiology
,
18
,
217
228
.
Cowan
,
N.
(
2001
).
The magical number 4 in short-term memory: A reconsideration of mental storage capacity
.
Behavioral and Brain Sciences
,
24
,
87
114
.
Cowan
,
N.
(
2012
).
An embedded-processes model of working memory
. In
A.
Miyake
&
P.
Shah
(Eds.),
Models of working memory: Mechanisms of active maintenance and executive control
(pp.
62
101
).
Cambridge, UK
:
Cambridge University Press
.
Deiber
,
M.-P.
,
Sallard
,
E.
,
Ludwig
,
C.
,
Ghezzi
,
C.
,
Barral
,
J.
, &
Ibañez
,
V.
(
2012
).
EEG alpha activity reflects motor preparation rather than the mode of action selection
.
Frontiers in Integrative Neuroscience
,
6
,
59
.
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
.
Dosher
,
B. A.
, &
Lu
,
Z.-L.
(
2000
).
Noise exclusion in spatial attention
.
Psychological Science
,
11
,
139
146
.
Duncan
,
J.
, &
Humphreys
,
G. W.
(
1989
).
Visual search and stimulus similarity
.
Psychological Review
,
96
,
433
458
.
Eriksen
,
C. W.
, &
St. James
,
J. D.
(
1986
).
Visual attention within and around the field of focal attention: A zoom lens model
.
Perception & Psychophysics
,
40
,
225
240
.
Foster
,
J. J.
, &
Awh
,
E.
(
2019
).
The role of alpha oscillations in spatial attention: Limited evidence for a suppression account
.
Current Opinion in Psychology
,
29
,
34
40
.
Fukuda
,
K.
,
Kang
,
M.-S.
, &
Woodman
,
G. F.
(
2016
).
Distinct neural mechanisms for spatially lateralized and spatially global visual working memory representations
.
Journal of Neurophysiology
,
116
,
1715
1727
.
Fukuda
,
K.
,
Mance
,
I.
, &
Vogel
,
E. K.
(
2015
).
α power modulation and event-related slow wave provide dissociable correlates of visual working memory
.
Journal of Neuroscience
,
35
,
14009
14016
.
Fukuda
,
K.
, &
Woodman
,
G. F.
(
2017
).
Visual working memory buffers information retrieved from visual long-term memory
.
Proceedings of the National Academy of Sciences, U.S.A.
,
114
,
5306
5311
.
Green
,
B. F.
, &
Anderson
,
L. K.
(
1956
).
Color coding in a visual search task
.
Journal of Experimental Psychology
,
51
,
19
24
.
Hakim
,
N.
,
Adam
,
K. C. S.
,
Gunseli
,
E.
,
Awh
,
E.
, &
Vogel
,
E. K.
(
2019
).
Dissecting the neural focus of attention reveals distinct processes for spatial attention and object-based storage in visual working memory
.
Psychological Science
,
30
,
526
540
.
Händel
,
B. F.
,
Haarmeier
,
T.
, &
Jensen
,
O.
(
2011
).
Alpha oscillations correlate with the successful inhibition of unattended stimuli
.
Journal of Cognitive Neuroscience
,
23
,
2494
2502
.
Hanslmayr
,
S.
,
Spitzer
,
B.
, &
Bäuml
,
K.-H.
(
2009
).
Brain oscillations dissociate between semantic and nonsemantic encoding of episodic memories
.
Cerebral Cortex
,
19
,
1631
1640
.
Hanslmayr
,
S.
, &
Staudigl
,
T.
(
2014
).
How brain oscillations form memories—A processing based perspective on oscillatory subsequent memory effects
.
Neuroimage
,
85
,
648
655
.
Hawkins
,
H. L.
,
Hillyard
,
S. A.
,
Luck
,
S. J.
,
Mouloua
,
M.
,
Downing
,
C. J.
, &
Woodward
,
D. P.
(
1990
).
Visual attention modulates signal detectability
.
Journal of Experimental Psychology: Human Perception and Performance
,
16
,
802
811
.
Hickey
,
C.
,
Di Lollo
,
V.
, &
McDonald
,
J. J.
(
2009
).
Electrophysiological indices of target and distractor processing in visual search
.
Journal of Cognitive Neuroscience
,
21
,
760
775
.
Jensen
,
O.
(
2002
).
Oscillations in the alpha band (9–12 Hz) increase with memory load during retention in a short-term memory task
.
Cerebral Cortex
,
12
,
877
882
.
Jensen
,
O.
, &
Mazaheri
,
A.
(
2010
).
Shaping functional architecture by oscillatory alpha activity: Gating by inhibition
.
Frontiers in Human Neuroscience
,
4
,
186
.
Jonides
,
J.
,
Lewis
,
R. L.
,
Nee
,
D. E.
,
Lustig
,
C. A.
,
Berman
,
M. G.
, &
Moore
,
K. S.
(
2008
).
The mind and brain of short-term memory
.
Annual Review of Psychology
,
59
,
193
224
.
Klimesch
,
W.
(
2012
).
α-band oscillations, attention, and controlled access to stored information
.
Trends in Cognitive Sciences
,
16
,
606
617
.
Liesefeld
,
A. M.
,
Liesefeld
,
H. R.
, &
Zimmer
,
H. D.
(
2014
).
Intercommunication between prefrontal and posterior brain regions for protecting visual working memory from distractor interference
.
Psychological Science
,
25
,
325
333
.
Lu
,
Z.-L.
,
Lesmes
,
L.-A.
, &
Dosher
,
B. A.
(
2002
).
Spatial attention excludes external noise at the target location
.
Journal of Vision
,
2
,
312
323
.
Luck
,
S. J.
,
Chelazzi
,
L.
,
Hillyard
,
S. A.
, &
Desimone
,
R.
(
1997
).
Neural mechanisms of spatial selective attention in areas V1, V2, and V4 of macaque visual cortex
.
Journal of Neurophysiology
,
77
,
24
42
.
Luck
,
S. J.
, &
Vogel
,
E. K.
(
2013
).
Visual working memory capacity: From psychophysics and neurobiology to individual differences
.
Trends in Cognitive Sciences
,
17
,
391
400
.
Luria
,
R.
,
Balaban
,
H.
,
Awh
,
E.
, &
Vogel
,
E. K.
(
2016
).
The contralateral delay activity as a neural measure of visual working memory
.
Neuroscience & Biobehavioral Reviews
,
62
,
100
108
.
Makeig
,
S.
,
Westerfield
,
M.
,
Jung
,
T. P.
,
Enghoff
,
S.
,
Townsend
,
J.
,
Courchesne
,
E.
, et al
(
2002
).
Dynamic brain sources of visual evoked responses
.
Science
,
295
,
690
694
.
Müller
,
N. G.
,
Bartelt
,
O. A.
,
Donner
,
T. H.
,
Villringer
,
A.
, &
Brandt
,
S. A.
(
2003
).
A physiological correlate of the “zoom lens” of visual attention
.
Journal of Neuroscience
,
23
,
3561
3565
.
Noonan
,
M. A. P.
,
Adamian
,
N.
,
Pike
,
A.
,
Printzlau
,
F.
,
Crittenden
,
B. M.
, &
Stokes
,
M. G.
(
2016
).
Distinct mechanisms for distractor suppression and target facilitation
.
Journal of Neuroscience
,
36
,
1797
1807
.
Nunez
,
P. L.
, &
Srinivasan
,
R.
(
2006
).
Electric fields of the brain: The neurophysics of EEG
(2nd ed.).
New York
:
Oxford University Press
.
Oberauer
,
K.
(
2002
).
Access to information in working memory: Exploring the focus of attention
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
28
,
411
421
.
Palva
,
J. M.
,
Monto
,
S.
,
Kulashekhar
,
S.
, &
Palva
,
S.
(
2010
).
Neuronal synchrony reveals working memory networks and predicts individual memory capacity
.
Proceedings of the National Academy of Sciences, U.S.A.
,
107
,
7580
7585
.
Palva
,
S.
, &
Palva
,
J. M.
(
2007
).
New vistas for α-frequency band oscillations
.
Trends in Neurosciences
,
30
,
150
158
.
Payne
,
L.
,
Guillory
,
S.
, &
Sekuler
,
R.
(
2013
).
Attention-modulated alpha-band oscillations protect against intrusion of irrelevant information
.
Journal of Cognitive Neuroscience
,
25
,
1463
1476
.
Payne
,
L.
, &
Sekuler
,
R.
(
2014
).
The importance of ignoring: Alpha oscillations protect selectivity
.
Current Directions in Psychological Science
,
23
,
171
177
.
Pelli
,
D. G.
(
1997
).
The VideoToolbox software for visual psychophysics: Transforming numbers into movies
.
Spatial Vision
,
10
,
437
442
.
Phillips
,
W. A.
(
1974
).
On the distinction between sensory storage and short-term visual memory
.
Perception & Psychophysics
,
16
,
283
290
.
Prescott
,
S. A.
,
Ratté
,
S.
,
De Koninck
,
Y.
, &
Sejnowski
,
T. J.
(
2008
).
Pyramidal neurons switch from integrators in vitro to resonators under in vivo-like conditions
.
Journal of Neurophysiology
,
100
,
3030
3042
.
Rouder
,
J. N.
,
Morey
,
R. D.
,
Morey
,
C. C.
, &
Cowan
,
N.
(
2011
).
How to measure working memory capacity in the change detection paradigm
.
Psychonomic Bulletin & Review
,
18
,
324
330
.
Rouder
,
J. N.
,
Morey
,
R. D.
,
Verhagen
,
J.
,
Swagman
,
A. R.
, &
Wagenmakers
,
E.-J.
(
2017
).
Bayesian analysis of factorial designs
.
Psychological Methods
,
22
,
304
321
.
Ruff
,
C. C.
, &
Driver
,
J.
(
2006
).
Attentional preparation for a lateralized visual distractor: Behavioral and fMRI evidence
.
Journal of Cognitive Neuroscience
,
18
,
522
538
.
Sauseng
,
P.
,
Klimesch
,
W.
,
Heise
,
K. F.
,
Gruber
,
W. R.
,
Holz
,
E.
,
Karim
,
A. A.
, et al
(
2009
).
Brain oscillatory substrates of visual short-term memory capacity
.
Current Biology
,
19
,
1846
1852
.
Sawaki
,
R.
,
Geng
,
J. J.
, &
Luck
,
S. J.
(
2012
).
A common neural mechanism for preventing and terminating the allocation of attention
.
Journal of Neuroscience
,
32
,
10725
10736
.
Sawaki
,
R.
, &
Luck
,
S. J.
(
2014
).
How the brain prevents and terminates shifts of attention
. In
G. R.
Mangun
(Ed.),
Cognitive electrophysiology of attention: Signals of the mind
(pp.
16
29
).
New York
:
Elsevier
.
Thompson
,
K. G.
,
Hanes
,
D. P.
,
Bichot
,
N. P.
, &
Schall
,
J. D.
(
1996
).
Perceptual and motor processing stages identified in the activity of macaque frontal eye field neurons during visual search
.
Journal of Neurophysiology
,
76
,
4040
4055
.
Vecera
,
S. P.
,
Behrmann
,
M.
, &
McGoldrick
,
J.
(
2000
).
Selective attention to the parts of an object
.
Psychonomic Bulletin & Review
,
7
,
301
308
.
Vecera
,
S. P.
, &
Farah
,
M. J.
(
1994
).
Does visual attention select objects or locations?
Journal of Experimental Psychology: General
,
123
,
146
160
.
Vogel
,
E. K.
, &
Machizawa
,
M. G.
(
2004
).
Neural activity predicts individual differences in visual working memory capacity
.
Nature
,
428
,
748
751
.
Vogel
,
E. K.
,
McCollough
,
A. W.
, &
Machizawa
,
M. G.
(
2005
).
Neural measures reveal individual differences in controlling access to working memory
.
Nature
,
438
,
500
503
.
Vogel
,
E. K.
,
Woodman
,
G. F.
, &
Luck
,
S. J.
(
2001
).
Storage of features, conjunctions, and objects in visual working memory
.
Journal of Experimental Psychology: Human Perception and Performance
,
27
,
92
114
.
Walter
,
W. G.
(
1938
).
Critical review: The technique and application of electro-encephalography
.
Journal of Neurology, Neurosurgery and Psychiatry
,
1
,
359
385
.
Wang
,
S.
,
Rajsic
,
J.
, &
Woodman
,
G. F.
(
2019
).
The contralateral delay activity tracks the sequential loading of objects into visual working memory, unlike lateralized alpha oscillations
.
Journal of Cognitive Neuroscience
,
31
,
1689
1698
.
Wolfe
,
J. M.
(
1998
).
Visual search
. In
H.
Pashler
(Ed.),
Attention
(pp.
13
71
).
London, UK
:
Psychology Press
.
Wolfe
,
J. M.
(
2003
).
Moving towards solutions to some enduring controversies in visual search
.
Trends in Cognitive Sciences
,
7
,
70
76
.
Woodman
,
G. F.
(
2010
).
A brief introduction to the use of event-related potentials in studies of perception and attention
.
Attention, Perception, & Psychophysics
,
72
,
2031
2046
.
Woodman
,
G. F.
,
Kang
,
M.-S.
,
Thompson
,
K.
, &
Schall
,
J. D.
(
2008
).
The effect of visual search efficiency on response preparation: Neurophysiological evidence for discrete flow
.
Psychological Science
,
19
,
128
136
.
Woodman
,
G. F.
, &
Luck
,
S. J.
(
2003
).
Serial deployment of attention during visual search
.
Journal of Experimental Psychology: Human Perception and Performance
,
29
,
121
138
.
Woodman
,
G. F.
, &
Vogel
,
E. K.
(
2008
).
Selective storage and maintenance of an object's features in visual working memory
.
Psychonomic Bulletin & Review
,
15
,
223
229
.
Woodman
,
G. F.
,
Vogel
,
E. K.
, &
Luck
,
S. J.
(
2012
).
Flexibility in visual working memory: Accurate change detection in the face of irrelevant variations in position
.
Visual Cognition
,
20
,
1
28
.
Xu
,
Y.
, &
Chun
,
M. M.
(
2006
).
Dissociable neural mechanisms supporting visual short-term memory for objects
.
Nature
,
440
,
91
95
.