Abstract

Filter mechanisms that prevent irrelevant information from consuming the limited storage capacity of visual STM are critical for goal-directed behavior. Alpha oscillatory activity has been related to proactive filtering of anticipated distraction. Yet, distraction in everyday life is not always anticipated, necessitating rapid, reactive filtering mechanisms. Currently, the oscillatory mechanisms underlying reactive distractor filtering remain unclear. In the current EEG study, we investigated whether reactive filtering of distractors also relies on alpha-band oscillatory mechanisms and explored possible contributions by oscillations in other frequency bands. To this end, participants performed a lateralized change detection task in which a varying and unpredicted number of distractors were presented both in the relevant hemifield, among targets, and in the irrelevant hemifield. Results showed that, whereas proactive distractor filtering was accompanied by lateralization of alpha-band activity over posterior scalp regions, reactive distractor filtering was not associated with modulations of oscillatory power in any frequency band. Yet, behavioral and post hoc ERP analyses clearly showed that participants selectively encoded relevant information. On the basis of these results, we conclude that reactive distractor filtering may not be realized through local modulation of alpha-band oscillatory activity.

INTRODUCTION

Visual STM (VSTM)—or the ability to temporarily keep information in mind—is fundamental for goal-directed behavior but is limited in capacity to simultaneous maintenance of only three to four objects (Cowan, 2001). Efficient use of VSTM storage capacity thus necessitates selective encoding of goal-relevant information, preventing irrelevant or distracting items from consuming capacity. A large body of research shows that individual VSTM capacity is indeed strongly related to one's ability to filter out goal-irrelevant, distracting information (Fukuda & Vogel, 2009; McNab & Klingberg, 2008; Vogel, McCollough, & Machizawa, 2005). Distractor filtering thus seems to serve as a functional bottleneck in VSTM, selectively routing goal-relevant information into VSTM (Awh & Vogel, 2008), thereby ensuring that its capacity is effectively used.

Although the idea that distractor filtering serves an important role in VSTM is widely accepted, the functional networks of brain regions that enable efficient distractor filtering are not yet precisely delineated (Gazzaley & Nobre, 2011). A limiting factor here is that, in the existing literature on the neural correlates of distractor filtering, the distinction between proactive versus reactive mechanisms of distractor filtering is largely ignored in the interpretation of results (Geng, 2014). Yet, these two types of interference control may recruit distinct mechanisms: Whereas proactive filtering or suppression entails prevention from interference (Bonnefond & Jensen, 2012), reactive interference control involves quick disengagement from irrelevant stimuli (Fukuda & Vogel, 2011). Everyday life situations usually demand a mixture of proactive and reactive distractor filtering (Peelen & Kastner, 2014): Whereas particular behavioral goals allow intended, proactive routing of neural processing according to these goals, of course unexpected but goal-irrelevant events may also occur. At present, it remains unclear to what extent these two types of distractor filtering rely on distinct neural mechanisms.

So far, many studies have related proactive distractor filtering to a local enhancement of alpha-band (8–12 Hz) oscillatory activity over regions processing task-irrelevant information before stimulus presentation (cf. Mazaheri et al., 2014; Payne, Guillory, & Sekuler, 2013; Mathewson et al., 2011; Snyder & Foxe, 2010; Klimesch, Sauseng, & Hanslmayr, 2007; Kelly, Lalor, Reilly, & Foxe, 2006; Thut, Nietzel, Brandt, & Pascual-Leone, 2006). This increase in prestimulus alpha power is generally interpreted in terms of proactive inhibition of irrelevant sensory processing (Mazaheri & Jensen, 2010; Klimesch et al., 2007), allowing incoming perceptual information to be selectively routed to task-relevant brain regions and networks. This notion is supported by many reports of increased prestimulus alpha power over sensory cortices that represent distracting sensory input (Bonnefond & Jensen, 2012; Foxe & Snyder, 2011; Händel, Haarmeier, & Jensen, 2011). Yet, some studies have observed enhanced alpha activity over irrelevant posterior regions even in conditions where no distracting information was present, indicating that local modulations of alpha power in task-irrelevant brain regions are not necessarily linked to proactive distractor suppression (Noonan et al., 2016; Myers, Walther, Wallis, Stokes, & Nobre, 2015; Rihs, Michel, & Thut, 2007). In other words, although proactive inhibition has been associated with enhanced alpha power over task-irrelevant sensory regions, it is unclear if this modulation reflects distractor inhibition per se or rather a more general attentional orienting toward task-relevant information.

Interestingly, findings from one EEG study suggest that alpha-band oscillatory activity may also play a role in reactive distractor filtering during VSTM retention (Sauseng et al., 2009). Specifically, this study found that during VSTM retention, alpha power scaled with the number of distractors (which could not be anticipated in advance) presented in the irrelevant visual hemifield, which was indicated by a cue preceding each stimulus display. These results suggest that alpha oscillations may not only serve as a proactive filter before encoding but can be flexibly modulated to filter out unanticipated distractors reactively. Yet, to the best of our knowledge, these results have never been replicated. Moreover, although participants did not know the exact number of distractors beforehand, distractor presence could be anticipated, as distractors were always present in the irrelevant hemifield in each trial (i.e., never in the same hemifield as the targets). This renders it unclear to what extent alpha oscillations play a role in reactive inhibition of distractors during VSTM retention in situations in which distractors may appear unexpectedly and/or in the attended hemifield.

Other studies that have so far examined the neural mechanisms involved in reactive distractor filtering, required when distractors are unpredicted, have mostly used ERPs and fMRI. This body of work suggests an important role for frontal brain regions and the BG in reactive distractor filtering. For example, a recent ERP study on distractor filtering and VSTM revealed a frontal distractor-sensitive signal during early stages of VSTM retention (Liesefeld, Liesefeld, & Zimmer, 2014). Interestingly, the amplitude of this signal was negatively related to later retention of irrelevant information, as reflected by the posterior contralateral delay activity (CDA; Vogel & Machizawa, 2004), in line with behavioral findings that suggest a close relationship between distractor filter ability and VSTM capacity. This pattern of findings also coarsely parallels earlier findings from fMRI studies, which indicate a critical role for frontal areas as well as the BG in protecting VSTM against irrelevant information and for more occipital and parietal regions in actual storage of the information (Baier et al., 2010; McNab & Klingberg, 2008; Jha, Fabian, & Aguirre, 2004). Yet, the precise oscillatory mechanisms involved in reactive distractor filtering, specifically the role of alpha oscillations herein, remain to be elucidated.

The aim of the current EEG study was to enhance our understanding of oscillatory mechanisms involved in proactive and reactive distractor filtering, in relation to VSTM. To this end, we used a typical lateralized change detection task that allowed for proactive distractor filtering based on an attentional cue, which indicated which hemifield would contain relevant, to-be-remembered stimuli (cf. e.g., Vogel & Machizawa, 2004). To examine the neural mechanisms underlying reactive filtering of unanticipated distractor load across hemifields, we varied the amount of distractors present in the relevant and irrelevant hemifield. Our main prediction was that alpha oscillatory activity, given its role in active inhibition, is involved in both proactive and reactive distractor filtering. First, we predicted to replicate previous studies showing greater alpha oscillatory activity over irrelevant versus relevant posterior scalp regions in anticipation of distraction (e.g., Händel et al., 2011). Second, we expected to replicate the previously reported load-dependent modulation of local alpha power over task-irrelevant scalp sites during suppression of distractors in the irrelevant hemifield (Sauseng et al., 2009). Our third and final prediction was that reactive suppression of distraction in the cued hemifield similarly relies on modulations of occipital alpha power. To this end, importantly, we used the spatially coarse lateralization response of alpha activity (contra- vs. ipsilateral posterior alpha power) to infer distractor-related alpha enhancement and target-related alpha suppression within one hemifield. By itself, these processes would be hard to disentangle, given the close spatial proximity of target and distractors within the cued hemifield (see Figure 1A, B) and the poor spatial resolution of EEG. However, we applied the logic that if measured lateral posterior alpha power contralateral to the cued visual hemifield reflects a summation of these two signals, then the resulting lateralized alpha response should be reduced in the presence of distractors in the cued hemifield, compared with a situation in which the cued visual hemifield contains targets only. Hereby, we aimed to shed further light on the neurophysiological basis of different types of distractor filtering that aid selective retention of goal-relevant information in VSTM, in particular alpha oscillatory activity.

Figure 1. 

VSTM task design and performance. (A) Example of a trial. In this example, participants had to remember the color of four targets (squares) presented among two distractors (circles) in the cued visual hemifield. (B) Overview of the nine experimental conditions. Dashed white rectangles (not visible in the actual experiment) indicate the attended hemifield (right cue trials). (C) The effect of distractors in the irrelevant visual hemifield on accuracy of performance. As can be seen, distractors in the irrelevant visual hemifield did not affect performance. (D) The effect of distractors in the cued visual hemifield on accuracy of performance. Distractors in the cued visual hemifield significantly impaired performance. This effect was dependent on memory load. (E) The relationship between filter costs (effect of distractors on performance; % performance decline) and VSTM capacity for two (black dots) and four (gray dots) distractors in the cued hemifield. Error bars: ±1 SEM.

Figure 1. 

VSTM task design and performance. (A) Example of a trial. In this example, participants had to remember the color of four targets (squares) presented among two distractors (circles) in the cued visual hemifield. (B) Overview of the nine experimental conditions. Dashed white rectangles (not visible in the actual experiment) indicate the attended hemifield (right cue trials). (C) The effect of distractors in the irrelevant visual hemifield on accuracy of performance. As can be seen, distractors in the irrelevant visual hemifield did not affect performance. (D) The effect of distractors in the cued visual hemifield on accuracy of performance. Distractors in the cued visual hemifield significantly impaired performance. This effect was dependent on memory load. (E) The relationship between filter costs (effect of distractors on performance; % performance decline) and VSTM capacity for two (black dots) and four (gray dots) distractors in the cued hemifield. Error bars: ±1 SEM.

METHODS

Participants

Participants took part in the experiment in exchange for course credits or monetary compensation. Participants had normal or corrected-to-normal vision, were neuropsychologically healthy, and were not colorblind. We obtained written and informed consent from each participant before the start of the experiment. The ethics committee of the Psychology faculty at the University of Amsterdam approved the experiment.

In total, 44 participants were tested. One participant was excluded because of technical problems during data recording, and two participants were excluded because more than half of the trials in the experiment were contaminated by muscle artifacts and/or horizontal eye movements. Ten additional participants were excluded because there was still residual eye movement-related activity in the horizontal electro-oculogram (HEOG) after preprocessing (see EEG data collection and preprocessing section for details). Analyses are based on the remaining 31 participants (mean age = 22.4 years, SD = 2.01 years; 22 women, 9 men).

Experimental Task

We used a lateralized change detection task with distractors (cf. e.g., Vogel et al., 2005). Participants were instructed to memorize the color of the targets (defined as squares) in the cued visual hemifield and were explicitly told that the distractors (defined as circles) would not be relevant at any point throughout the experiment. Participants were asked to focus their eyes on fixation throughout a trial and blink as little as possible to prevent contamination of the EEG with blink-related artifacts. To be able to measure VSTM capacity, we manipulated the number of target stimuli (2 vs. 4 vs. 6). We also parametrically manipulated the number of distractors in the irrelevant visual hemifield (2 vs. 4) for different target loads (2 vs. 4) in separate conditions to investigate load dependence of distractor suppression. Thus, as in Sauseng et al. (2009), although participants could anticipate distractors in the irrelevant field, they had no knowledge about the number of distractors in these trials. Lastly, we parametrically manipulated the number of distractors (0 vs. 2 vs. 4; circles) in the cued visual hemifield to investigate reactive distractor filtering in the relevant hemifield under different memory loads (2 vs. 4 targets; squares). When manipulating the number of distractors in the cued visual hemifield, we kept the number of stimuli in the irrelevant hemifield identical to the number of stimuli in the cued visual hemifield. In total, our experiment contained nine conditions consisting of 100 trials each (see Figure 1B for an overview of the conditions with the numbers and types of stimuli). The task consisted of 25 blocks of 36 trials. Blocks were interspersed by self-paced breaks. Trials from the nine different conditions were administered in random order within each block.

Each trial started with a 200-msec presentation of an arrow-shaped cue at fixation that indicated the to-be-attended hemifield on that trial (0.5 degrees of visual angle [dva]; 50% left cue), followed by a 500-msec blank screen containing a central fixation cross (0.5 dva). Note that this interval allowed participants an additional 500 msec to orient their attention to the cued hemifield compared with the study by Sauseng et al. (2009), in which the memory display immediately followed the attention cue. Subsequently, the memory display was presented for 250 msec. This display was followed by a 1000-msec retention interval during which the fixation cross remained on the screen (see Figure 1A for a graphical display of the trial order). Next, a probe display was presented that was either identical to the memory display (50% of the trials, counterbalanced across cue sides) or was different because one of the target stimuli had changed color. Distractor stimuli could never change color. Participants needed to respond to the probe stimulus within 2000 msec. Sixteen participants responded using the keyboard (press Z key with the left index finger for no-change; press M key with the right index finger for change). Fifteen participants responded by pressing one of two response boxes placed on each arm rest (with the left index finger for no-change, with the right index finger for change). The fixation cross remained visible throughout the entire trial to aid participants to maintain fixation. Trials were separated by variable intertrial intervals, with a duration that was randomly jittered around 2000 msec (range = 1700–2300 msec).

The stimulus display consisted of a bilateral display with colored stimuli on a gray background (see Figure 1A). We positioned all stimuli within two square regions located left and right of fixation in the lower visual field (4.25 × 4.25 dva; positioned at 2 dva from the vertical meridian). Stimuli were presented in the lower visual field to obtain a relatively focal projection of stimulus-related oscillatory activity on the scalp (cf. Bahramisharif, Van Gerven, Heskes, & Jensen, 2010), such that changes in occipital activity related to stimulus processing in one hemifield should be reflected in the aggregate of activity measured at these focal contralateral scalp sites. Target stimuli were colored squares (0.75 dva); distractor stimuli were colored circles (0.85 dva). Colors of stimuli were randomly selected from 11 highly distinguishable colors on each trial. A color would never appear twice in one visual hemifield, and new target colors on change trials were always different from the colors of the stimuli presented in the cued visual field. Stimulus positions were constrained, such that individual stimuli would never touch or overlap.

Procedure

Participants were seated at a 90-cm distance from the screen. Before the start of the experiment, participants practiced two blocks of the task of 20 trials each (low load conditions only) to become familiar with the task and the response buttons. Participants were given extensive task instructions before the start of the practice task. During the first practice block, participants received immediate feedback on their response (correct/incorrect/late). In the second practice block, immediate feedback was no longer provided to prepare participants for the experimental phase, in which participants also did not receive feedback on their performance.

Behavioral Data Analysis

Behavioral analyses were based on trials in which participants responded in time (within 2000 msec), excluding trials containing fast, uninformed responses (reflected by RTs < 200 msec). Using participants' HEOG data, we discarded trials in which participants made an eye movement away from fixation (for details on eye movement detection, see the section below). All trials in which participants responded in time and maintained fixation were used for behavioral data analysis. This procedure yielded an average of 837 trials (93% of the total number of trials) per participant (SD = 58, range = 701–893).

The effects of memory load and distractor load in the cued and irrelevant visual hemifield on accuracy of performance were tested using repeated-measures ANOVAs. The corrected degrees of freedom (Greenhouse–Geisser correction) are reported when the assumption of sphericity was violated (tested using Mauchly's test of sphericity). VSTM capacity was computed using Pashler's index of VSTM capacity (Kp; Pashler, 1988), as the probe in our change detection task required whole-display recognition (Rouder, Morey, Morey, & Cowan, 2011). Correlations between VSTM capacity and the effect of distractors on performance were assessed using a Pearson correlation coefficient. On the basis of previous research (Fukuda & Vogel, 2009; Vogel et al., 2005), we a priori predicted an inverse relationship between the effect of distractors on performance and VSTM capacity and used a one-tailed test to assess the significance of the correlation (α = .05).

EEG Data Collection and Preprocessing

EEG data were recorded at 512 Hz from 64 channels placed according to the international 10–20 system using a Biosemi (Amsterdam, The Netherlands) setup. External electrodes were used to measure horizontal and vertical eye movements and were placed next to the right and left eye (HEOG), and below and above the left eye (vertical electro-oculogram). Reference electrodes were placed on the earlobes.

Offline, EEG data were rereferenced to the average earlobes. Data were high-pass filtered at 0.05 Hz and were epoched from −2.2 to +3.3 sec surrounding stimulus onset (to avoid edge artifacts resulting from wavelet filtering in the time–frequency analysis) and baseline-corrected by removing the average activity in the 200-msec time window preceding presentation of the cue for each channel and trial. The epoched data were visually inspected, and trials containing EMG activity or other artifacts on multiple electrodes that were not related to blinks were manually removed. We applied single-trial interpolations of electrodes that displayed artifactual activity at the single-trial level, for which interpolation of the channel for the entire experiment would have been too stringent. Electrodes that contained noisy data throughout the majority of the experiment were temporarily set to zero to be interpolated after independent component analysis (ICA; interpolation was postponed to avoid reducing the rank of the data before running ICA; see van Driel, Knapen, van Es, & Cohen, 2014). ICA decomposition was computed on the cleaned data set using EEGlab software (Delorme & Makeig, 2004). Components containing blinks or other artifacts that could be clearly distinguished from brain-driven EEG signals were subtracted from the data. As in the behavioral analyses, error trials and trials in which participants responded too early (<200 msec) or too late (>2000 msec) were excluded from the analysis. On the basis of manual inspection of the HEOG trace, trials containing horizontal eye movements were removed from further analysis. Only participants who displayed no residual HEOG activity after preprocessing were included in subsequent analyses. To determine if residual HEOG activity was present, we subtracted the average HEOG activity on left-cue trials from that on right-cue trials and examined the extent to which the trial average difference deviated from zero. Participants for whom the trial average difference between left and right cue trials showed an absolute difference of more than 3 mV for a consecutive period of 50 msec or more were excluded from the analyses (cf. Luck, 2014). Eleven participants were excluded this way. In the remaining group of participants, an average of 642 (SD = 76.8, range = 516–781) artifact-free trials (71% of total number of trials) per participant were included in further analyses. Next, the EEG data were transformed, such that, on all trials, the left hemisphere was contralateral to stimulus presentation. Hereto, we swapped data at symmetrical electrodes (e.g., F1 and F2) along the midline (Iz to FPz) in trials with a leftward cue, so that data in all trials reflected activity following a rightward attentional cue. After this transformation, the data were collapsed across cue directions (left vs. right).

Time–Frequency Analysis of the EEG Data

We subtracted the evoked activity from the EEG data before time–frequency analysis (cf. Sauseng et al., 2009). Subsequently, we applied a surface Laplacian on the data to increase topographical specificity and reduce the effects of volume conduction (Cohen, 2014; Srinivasan, Winter, Ding, & Nunez, 2007). We used a filter with a tenth-order Legendre polynomial and a smoothing parameter (λ) of 10−5. The units of the data after this transform are μV/cm2.

Frequency band-specific power was extracted using time–frequency decomposition of the EEG data for each channel and condition with in-house written Matlab (The MathWorks, Natick, MA) routines. Single-trial stimulus-locked data were convolved with a family of complex Morlet wavelets, defined as Gaussian-windowed complex sine waves:
formula
Here, i reflects the complex operator, t reflects time, and f reflects frequency. Frequencies increased from 1 to 40 Hz in 20 logarithmically spaced steps. The width of the wavelet for each frequency (σ) was set as n/2πf, where n is the number of wavelet cycles, scaled logarithmically from 3 to 10. Power (p) was extracted from the signal as the squared magnitude of the complex signal (Z(t)) resulting from the convolution (p(t) = real[z(t)]2 + imag[z(t)]2). After time–frequency decomposition, power was converted to a decibel scale (dB), which enables comparison of amplitude across frequency bands (dB power = 10 × log10 [power/baseline]), using a baseline time window between 500 and 200 msec preceding presentation of the cue. The data were downsampled to 40 Hz for computational purposes after wavelet convolution.

Electrode Selection and Statistical Analysis of Time–Frequency Data

To examine the oscillatory mechanisms underlying reactive filtering of distracting information, we tested whether lateralized occipital alpha power was sensitive to distractor load in the cued and/or irrelevant hemifield. To this end, we averaged the power of alpha (8–12 Hz) activity during the delay interval (350–1250 msec poststimulus) at occipital electrodes (PO3, PO7, O1 vs. PO4, PO8, O2 for contralateral vs. ipsilateral sites) across conditions. To study the potential time–frequency equivalent of the “frontal bias signal” located at frontal sites (Liesefeld et al., 2014), we selected time–frequency windows based on the peaks in the condition average activity at frontal channels (see Figure 3). These condition averages were then subjected to repeated-measures ANOVA's with the factors (i) distractor load (2 vs. 4 in the irrelevant hemifield, cf. Sauseng et al., 2009, or 0 vs. 2 vs. 4 in the relevant hemifield), (ii) memory load (2 vs. 4), and (iii) hemisphere (contralateral vs. ipsilateral). For all statistical tests, we report the corrected degrees of freedom (Greenhouse–Geisser correction) when the assumption of equal variances was violated (tested using Mauchly's test of sphericity). Correlations between behavioral variables of interest, such as VSTM capacity (Kp) and time–frequency power were tested using Spearman rank correlations, because power data are often not normally distributed (Cohen, 2014).

ERP Analyses

We also aimed to replicate previously reported effects of distractors presented among targets in the relevant hemifield on ERP components related to distractor storage (the CDA) and distractor signaling (the frontal bias signal; Liesefeld et al., 2014; Vogel et al., 2005). ERP analyses were performed using EEGLAB (Delorme & Makeig, 2004) and Matlab, using in-house written code to extract the CDA and the frontal bias signal. For computation of the CDA, data were first baselined using a 200-msec time window preceding stimulus presentation (cf. Spronk, Vogel, & Jonkman, 2012) and low-pass filtered using an 80-Hz low-pass filter (Hamming window sinc FIR filter, using the function pop_eegfiltnew). The CDA was computed as the difference between the average activity at posterior channels in the hemisphere contralateral (P3, P5, P7, P9, PO3, PO7, O1) and ipsilateral (P4, P6, P8, P10, PO4, PO8, O2) to target presentation, in the time window between 350 and 1250 msec after stimulus onset. An additional cleaning procedure was applied to the resulting difference wave to remove slow drifts resulting from our low cutoff frequency (0.05 Hz) of our high-pass filter from our ERP of interest, which gained strong influence on the trial averages, because of the long time windows used to compute the CDA. To minimize their influence, the data were trimmed by first sorting trials according to amplitude and subsequently removing the 10% most extreme values at each tail of the distribution for each condition (cf. Wilcox, 2012; Rousselet & Pernet, 2011). This way, the original central tendency of the condition-specific CDA waveforms was maintained, but trials in which the CDA obtained an extreme value because of drifting that may obscure condition differences were selectively removed.

We also focused our ERP analyses on the frontal bias signal observed at frontal sites early during VSTM retention. On the basis of the latency and topography of this ERP reported by Liesefeld et al. (2014) and visual inspection of our condition-averaged data, we computed the frontal bias signal at a set of frontal electrodes (F1, F3, AF3, Fz, Afz, AF4, F4, F2) in a time window between 240 and 290 msec poststimulus (Liesefeld et al., 2014).

We first examined if distractors could effectively be reactively prevented from entering VSTM storage. To this end, the effect of distractors on CDA amplitude was quantified by comparing the condition average activity in the condition with Memory Load 2 with and without distractors, for different levels of distractors (2T vs. 2T2D; 2T vs. 2T4D). Then, we assessed if individuals with higher VSTM capacity were generally more successful from keeping distractors from entering VSTM by correlating the distractor-related amplitude of the CDA with VSTM capacity (Kp) using a Pearson correlation test. Next, we aimed to replicate the previously reported effect of general distractor-presence on the frontal bias signal, and averaged the ERP in the time window of interest using the conditions containing targets only (2T; 4T; 6T), and compared this to the average ERP in the conditions that also contained distractors in the cued visual hemifield (2T2D; 2T4D; 4T2D; 4T4D). Finally, we examined if individuals with a stronger frontal bias signal generally displayed a reduced influence of distractors on CDA amplitude and correlated the distractor-related amplitudes of the CDA and the “frontal bias signal” for two distractors and Memory Load 2 (2T2D vs. 2T) using a Pearson correlation test. On the basis of previous research (Liesefeld et al., 2014; Vogel et al., 2005), we had strong expectations of an inverse relationship between the effect of distractors on the CDA and VSTM capacity as well as the distractor-related amplitude of the “frontal bias signal” and used one-tailed tests (α = .05) to assess the significance of the correlation coefficients. For plotting purposes, ERP waveforms were temporally smoothed (using a moving average method, with a 100-msec span). Statistical analyses were based on the nonsmoothed data sets. In all figures displaying ERP waveforms, negative voltages are plotted upward.

RESULTS

Behavior

VSTM Performance and Capacity

Accuracy of performance decreased with memory load (F(2, 60) = 304.58, p < .001): Planned contrasts showed that performance decreased when remembering four instead of two (82% vs. 96% correct; F(1, 30) = 213.69) as well as six instead of four items (71% vs. 82% correct; F(1, 30) = 375.84, p < .001; see Figure 1). In line with previous findings, the average value of Kp was 3.07 and ranged between 1.94 and 4.36 across participants, indicating an average VSTM storage capacity of about three objects and large interindividual differences in storage capacity.

Effects of Distractors on Behavior

We first examined the extent to which distractors in the irrelevant hemifield impaired task performance (see Figure 1C). A repeated-measures ANOVA with the factors Distractor load in the irrelevant hemifield (2 vs. 4) and Memory load (2 vs. 4) revealed no main effect of the factor Distractor load (F(1, 30) = .57, p = .457), but a significant interaction between Distractor load and Memory load (F(1, 30) = 5.15, p = .031). This interaction was driven by small distractor-related performance effects of opposite sign across memory loads: A high distractor load in the irrelevant visual hemifield tended to impair performance under high memory loads (−1.7%; t(30) = 1.564, p = .128) but improve performance under low memory loads (+.8%; t(30) = −1.954, p = .060). Thus, these individual post hoc contrasts as well as the main effect of distractor load were nonsignificant. This suggests that the effect of distractors in the irrelevant hemifield, which could be anticipated based on the attentional cue on performance, was very small or even negligible.

Next, we examined to what extent distractors presented among the targets in the relevant hemifield influenced performance, and if their effect depended on memory load. To this end, a repeated-measures ANOVA was conducted with the factors Distractor load in the cued hemifield (0 vs. 2 vs. 4) and Memory load (2 vs. 4; note that this analysis was performed on the subsample of participants for whom all required levels of distractor load were included in the task; n = 15). Results revealed a main effect of the Factor distractor load (F(1.4, 19.3) = 58.24, p < .001) as well as a main effect of Memory load (F(1, 14) = 282.30, p < .001). Planned contrasts showed that performance decreased with increasing numbers of distractors (−8.2% for 0 vs. 2 distractors; F(1, 14) = 50.27, p < .001; −8.2% for 2 vs. 4 distractors; F(1, 14) = 61.52, p < .001; see Figure 1D). We observed a trend toward an interaction between distractor load and memory load (F(1.5, 20.8) = 3.04, p = .082), where the performance decrease due to two instead of zero distractors was bigger under high memory loads (−10%) than low memory loads (−7%; t(14) = −2.12, p = .053), whereas the decrease due to four instead of two distractors was biggest under low memory loads (−11% vs. −5%; t(14) = .96; p = .355). This pattern of findings suggests that reactive filtering mechanisms of within-hemifield distractors are imperfect: The presence of distractors in the cued visual hemifield impaired memory performance.

Previous research has shown that the extent to which within-hemifield distractors impair performance (or filter efficiency) predicts individual VSTM capacity (e.g., Fukuda & Vogel, 2009). In line with this observation, the correlation between Kp-max and the effect of distractors in the cued visual hemifield for Memory Load 2 reached trend level significance for two distractors (2T0D vs. 2T2D; r29 = −.25, p = .085) but was significant for four distractors (2T0D vs. 2T4D; r29 = −.30, p = .050). Previous research used Cowan's K (Kc-max) to index VSTM capacity (e.g., Vogel & Machizawa, 2004). Although Kp-max is considered a more proper index of VSTM capacity for change detection task requiring whole-display recognition (Rouder et al., 2011), we post hoc analyzed the relationship between Kc-max and the effect of distractors on performance to be able to directly compare our results to previous research on the relationship between VSTM capacity and distractor filtering. This revealed significant correlations between Kc-max and filter costs for two as well as four distractors (r29 = −.34, p = .030 and r29 = −.36, p = .023, respectively). Thus, replicating earlier findings, we found that distractors generally influenced performance less in individuals with a higher VSTM capacity, but more robustly using Kc-max than Kp-max.

Oscillatory Dynamics of Distractor Filtering

The Role of Occipital Alpha Lateralization in Distractor Filtering

We first examined if we could replicate the previously reported cue-induced lateralization in prestimulus alpha-band oscillatory activity. As can be seen in Figures 2B and 2C, alpha power lateralization developed well before memory display onset, after the presentation of the spatial cue. This effect reflected significantly stronger prestimulus alpha power activity over irrelevant versus relevant posterior regions (300–100 msec preceding stimulus presentation; t(30) = 8.70; p < .001), suggesting that participants may have proactively suppressed processing of distracting information in the irrelevant hemifield. To explore this possibility further, we examined whether the degree of cue-related alpha lateralization (300–100 msec preceding stimulus presentation) predicted performance by testing whether lateralization of prestimulus alpha power was predictive of the subsequent effect of distractors in the irrelevant hemifield on VSTM performance (cf. Händel et al., 2011). However, we found no evidence for a relationship between stronger lateralization and distractor load-related performance decline under low memory loads (2T; r29 = −.18, p = .342) or high memory loads (4T; r29 = .10, p = .609; Spearman rank correlations). To examine the possibility that individual differences in VSTM capacity may have masked this relationship, we post hoc repeated this analysis, but now controlling for VSTM capacity (Kp-max). However, the relationship between lateralized prestimulus alpha power and the effect of distractors in the irrelevant field on performance across individuals remained nonsignificant when controlling for VSTM capacity (r28 = −.118, p = .536 for trials with Memory Load 2; r28 = .047, p = .803 for trials with Memory Load 4). The lack of such a relationship in combination with the lack of an effect of irrelevant hemifield distractors on VSTM performance (reported in the previous section) may indicate that all participants effectively filtered the distracting information in the irrelevant hemifield. Possibly, the longer time interval between the cue and the memory display allowed all participants to proactively establish alpha lateralization and thereby gate these distractors from storage in VSTM.

Figure 2. 

Lateralized alpha oscillatory activity is not reactively modulated by distractor load. (A) Topographical distribution of condition average alpha activity (8–12 Hz) during the retention interval (350–1250 msec). Black electrodes depict the electrodes used for statistical analysis of the effects of the attention (cue) direction and distractor load. Note that, in this plot, the left hemisphere is the hemisphere contralateral to the cued hemifield, whereas the right hemisphere is ipsilateral to the cued hemifield. (B) Time–frequency representation of condition average lateralized activity throughout an entire trial at occipital electrodes contralateral (PO3, PO7, O1) versus ipsilateral (PO4, PO8, O2) to the cued hemifield. The white rectangles reflect the time–frequency windows used for (1) the correlation between prestimulus lateralized alpha power and behavior and (2) the effects of distractor load and memory load on alpha power during VSTM retention. (C) Time course of condition average lateralization of alpha activity (8–12 Hz; contralateral minus ipsilateral electrodes). Alpha lateralization increased before presentation of the stimulus and probe. (D) Time course of condition average occipital alpha activity (8–12 Hz) at relevant (contralateral; blue line) and irrelevant (ipsilateral; green line) electrodes. Alpha power followed a similar pattern across hemispheres, but with stronger suppression in the contralateral compared with the ipsilateral hemisphere. (E) Bar plots showing the effect of distractor load in the irrelevant hemifield (2 vs. 4) under different memory loads (2 vs. 4) on alpha power in the hemisphere contralateral (C) and ipsilateral (I) to the cued hemifield. Alpha power was lower in the hemisphere contralateral to the cued hemifield regardless of the number of distractors in the irrelevant hemisphere. (F) Bar plots showing the effect of distractor load in the cued hemifield (0 vs. 2 vs. 4) under different memory loads (2 vs. 4) on alpha power in the hemisphere contralateral (C) and ipsilateral (I) to the cued hemifield. Alpha power was lower in the hemisphere contralateral to the cued hemifield and not modulated by the number of distractors in the relevant hemifield. Error bars: ±1 SEM.

Figure 2. 

Lateralized alpha oscillatory activity is not reactively modulated by distractor load. (A) Topographical distribution of condition average alpha activity (8–12 Hz) during the retention interval (350–1250 msec). Black electrodes depict the electrodes used for statistical analysis of the effects of the attention (cue) direction and distractor load. Note that, in this plot, the left hemisphere is the hemisphere contralateral to the cued hemifield, whereas the right hemisphere is ipsilateral to the cued hemifield. (B) Time–frequency representation of condition average lateralized activity throughout an entire trial at occipital electrodes contralateral (PO3, PO7, O1) versus ipsilateral (PO4, PO8, O2) to the cued hemifield. The white rectangles reflect the time–frequency windows used for (1) the correlation between prestimulus lateralized alpha power and behavior and (2) the effects of distractor load and memory load on alpha power during VSTM retention. (C) Time course of condition average lateralization of alpha activity (8–12 Hz; contralateral minus ipsilateral electrodes). Alpha lateralization increased before presentation of the stimulus and probe. (D) Time course of condition average occipital alpha activity (8–12 Hz) at relevant (contralateral; blue line) and irrelevant (ipsilateral; green line) electrodes. Alpha power followed a similar pattern across hemispheres, but with stronger suppression in the contralateral compared with the ipsilateral hemisphere. (E) Bar plots showing the effect of distractor load in the irrelevant hemifield (2 vs. 4) under different memory loads (2 vs. 4) on alpha power in the hemisphere contralateral (C) and ipsilateral (I) to the cued hemifield. Alpha power was lower in the hemisphere contralateral to the cued hemifield regardless of the number of distractors in the irrelevant hemisphere. (F) Bar plots showing the effect of distractor load in the cued hemifield (0 vs. 2 vs. 4) under different memory loads (2 vs. 4) on alpha power in the hemisphere contralateral (C) and ipsilateral (I) to the cued hemifield. Alpha power was lower in the hemisphere contralateral to the cued hemifield and not modulated by the number of distractors in the relevant hemifield. Error bars: ±1 SEM.

Strikingly, lateralized alpha power was not affected by the number of distractors in the irrelevant hemifield during VSTM retention, in contrast to an earlier report by Sauseng et al. (2009). That is, a repeated-measures ANOVA with the factors Distractor load in the irrelevant visual field (2 vs. 4), Memory load (2 vs. 4), and Hemisphere (contralateral vs. ipsilateral to the cued hemifield) only revealed a main effect of Hemisphere (F(1, 30) = 55.89, p < .001), with lower alpha power at contralateral (M = −3.14 dB) as compared with ipsilateral sites (M = −2.18 dB), regardless of the number of ipsilateral distractors. Thus, contrary to our second prediction, ipsilateral alpha power during VSTM retention was not sensitive to distractor load (F(1, 30) = .05, p = .82; see Figure 2E) when participants were given sufficient time to orient attention before the memory display was presented.

Moreover, in contrast to our third prediction, lateralized occipital alpha power was also not reactively modulated by distractors in the cued visual hemifield during the retention interval (see Figure 2F; note that this analysis was again performed on the subsample of participants for whom all required levels of distractor load were included in the task; n = 15). There was no effect of Distractors in the cued visual field (0 vs. 2 vs. 4) under different memory loads (2 vs. 4) on occipital alpha power across both hemispheres (contralateral vs. ipsilateral; F(2, 28) = 2.36, p = .113) and also no interaction between the effect of Distractors and Hemisphere (F(1.46, 2.46) = .51, p = .605). The only significant effect was a main effect of Hemisphere (F(1, 14) = 32.94, p < .001), reflecting that alpha power was always lower over the hemisphere contralateral to the cued visual hemifield (M = −3.49 dB) than over the ipsilateral hemisphere (M = −2.40 dB). Furthermore, post hoc analyses revealed no evidence for a relationship between the effect of distractors in the cued hemifield on lateralized occipital alpha power during the delay period and their effect on accuracy of performance (comparing 2 vs. 0 and 4 vs. 0 distractors) or VSTM capacity (all ps > .07), indicating that the effect of distractors on lateral occipital alpha power did not vary as a function of performance. Thus, our findings do not provide evidence for a role for lateralized alpha oscillatory activity in reactive distractor filtering during VSTM retention.

Exploratory Analyses: Oscillatory Mechanisms Involved in Reactive, Within-hemifield Distractor Filtering

Reactive distractor filtering may also rely on nonlateralized neural mechanisms in frequency bands other than alpha. Therefore, we next explored the involvement of other low-frequency oscillatory mechanisms in reactive distractor filtering. Here we focused on effects in the theta (3–7 Hz) and beta (14–22 Hz) frequency ranges in the entire retention interval, as we had no specific predictions on the frequency characteristics and timing of the effect. Relevant time–frequency windows to subject to statistical testing were based on visual inspection of the condition-averaged data. Note that this procedure is essentially orthogonal to the contrast of interest and therefore does not lead to biased results of later statistical tests. Topographical representations of the data revealed a positive theta and beta peak at AFz and Fz with a latency and topography resembling that of the frontal bias signal ERP reported by Liesefeld et al. (2014) (see Figure 3A and B). Therefore, these electrodes were selected to explore a relationship between oscillatory activity in these frequency bands and reactive distractor filtering. Figure 3C shows the time–frequency representation of condition average oscillatory power throughout the trial. On the basis of the peaks in the condition-averaged activity at AFz and Fz, we selected low theta activity (2–5 Hz) between 350 and 950 msec poststimulus, and low beta activity (12–18 Hz) between 550 and 1150 msec poststimulus, for statistical testing of the effect of distractors in the cued visual hemifield on these different frequency bands (see the black boxes in Figure 3C).

Figure 3. 

Effects of distractors on frontal theta and beta activity. (A) Topographical distribution of condition average theta activity (3–7 Hz) during the retention interval (350–1250 msec). (B) Topographical distribution of condition average low beta activity (12–20 Hz) during the retention interval (350–1250 msec). (C) Full time–frequency spectrum of the condition average power at electrodes AFz and Fz. Black rectangles reflect the time–frequency windows that were used for statistical analysis of the effects of distractors in the cued hemifield on theta and beta oscillatory power. (D) Time–frequency representation of the effect of two distractors in the cued hemifield for trials with Memory Load 2 (2T2D vs. 2T0D). (E) Time–frequency representation of the effect of four distractors in the cued hemifield for trials with Memory Load 2 (2T4D vs. 2T0D). (F) Bars depict the power across conditions with zero, two, or four distractors for the time–frequency windows in theta (left; 3–7 Hz, 350–950 msec) and beta (right; 12–18 Hz, 550–1150 msec) used for statistical testing. Distractor load did not modulate theta or beta power in the selected time windows. Error bars: ±1 SEM.

Figure 3. 

Effects of distractors on frontal theta and beta activity. (A) Topographical distribution of condition average theta activity (3–7 Hz) during the retention interval (350–1250 msec). (B) Topographical distribution of condition average low beta activity (12–20 Hz) during the retention interval (350–1250 msec). (C) Full time–frequency spectrum of the condition average power at electrodes AFz and Fz. Black rectangles reflect the time–frequency windows that were used for statistical analysis of the effects of distractors in the cued hemifield on theta and beta oscillatory power. (D) Time–frequency representation of the effect of two distractors in the cued hemifield for trials with Memory Load 2 (2T2D vs. 2T0D). (E) Time–frequency representation of the effect of four distractors in the cued hemifield for trials with Memory Load 2 (2T4D vs. 2T0D). (F) Bars depict the power across conditions with zero, two, or four distractors for the time–frequency windows in theta (left; 3–7 Hz, 350–950 msec) and beta (right; 12–18 Hz, 550–1150 msec) used for statistical testing. Distractor load did not modulate theta or beta power in the selected time windows. Error bars: ±1 SEM.

A repeated-measures ANOVA on the average theta activity in this time window with the factor Distractors in the cued hemifield (0 vs. 2 vs. 4), for trials with Memory Load 2 did not reveal an effect of Distractor load on theta power during this time window (F(1, 30) = .36, p = .698; see Figure 3DF). Frontal beta activity was also not modulated by distractor presence (F(1, 30) = .45, p = .639). Thus, these exploratory analyses did not provide evidence for the existence of a nonlateralized oscillatory correlate of reactive distractor filtering in service of VSTM.

Exploratory Analyses: Single-trial Regression to Elucidate Oscillatory Mechanisms Involved in Distractor Filtering

Given the above null findings, which may be somewhat surprising in light on the well-known role of alpha oscillations in inhibition of sensory processing, finally, we performed single-trial multiple regression on our EEG data. This allowed us to examine possible effects of the factors of interest in a more data-driven manner to make sure that we did not miss any distractor-related power modulations by restricting our statistical analyses to selected electrodes and time–frequency windows. Single-trial analysis was performed using robust regression (using iteratively reweighted least squares; cf. Cohen & Cavanagh, 2011; Wager, Keller, Lacey, & Jonides, 2005). We used the design matrix of our task as predictors, with the factors cue side, memory load, distractor load in the cued hemifield, and distractor load in the irrelevant hemifield. Additionally, we included possible interactions between the effects of cue side and memory load, distractor load in the cued visual field, and distractor load in the irrelevant field as predictors. The regression was performed using the following equation: Y = INT + b1CUE + b2ML + b3DLC + b4DLI + b5CUE.ML + b6CUE.DLC + b7CUE.DLI + E. Here, Y is the data vector with oscillatory power across trials. INT is the intercept (accounting for the Power Law scaling of frequencies; Cohen & Cavanagh, 2011), and E is unexplained variance. CUE is a vector that codes the side of the attentional cue (left or right). ML is a vector representing the memory load in the cued hemifield, DLC is a vector representing the distractor load in the cued hemifield, and DLI is a vector representing the distractor load in the irrelevant hemifield. The last three predictors represent interaction terms between cue side and the factors memory load, and distractor load in the cued (DLC) and irrelevant hemifield (DLI).

We included single-trial time–frequency representations of the data covering the time between presentation of the attentional cue and the probe (−700 to 1250 msec), 20 frequencies ranging from 1 to 40 Hz (see Methods section for details), and all channels, to compute the beta weights for each predictor for each subject separately. The single-trial power data were normalized (mean-centered and scaled by their standard deviation) before performing the regression. Note that we did not flip any of the trials based on cue direction in this analysis to be able to estimate the regression weights for the effect of cue side. The result of the regression was a subject by channel by time by frequency matrix with beta values for each predictor in our model. To be able to compare the estimated beta weights across subjects, we normalized the beta values by their standard error (Cohen & Cavanagh, 2011). Statistical significance of the beta weights was assessed at the group level and was done by testing the beta weights against zero for each Channel × Time × Frequency point (α = .01). Subsequently, the significant Channel × Time × Frequency points were cluster-corrected by removing data points that were not included in a cluster exceeding the size of the 99th percentile of maximum cluster sizes observed under the null hypothesis. The distribution of cluster sizes under the null hypothesis was estimated using 1000 permutations of the data on which we randomly changed the sign of the beta weights across participants, clustered the Channel × Time × Frequency points with significant beta weights, and extracted the maximum cluster size across the remaining clusters. Cluster level statistics were computed separately for each predictor included in the model.

We used the predictor cue side to assess whether the regression could reproduce the results from the trial average data showing lateral occipital alpha power being responsive to the direction of the attentional cue. Figure 4 depicts the topographical and time–frequency characteristics of the group-averaged standardized regression coefficients for the predictor cue side. Black outlines represent clusters of data points significantly affected by the predictor of interest. Robust regression of the factor cue that allowed proactive preparation for distractors in the irrelevant hemifield revealed four significant clusters in which bilateral alpha-band activity was affected by the direction of the cue. Two of these clusters developed following presentation of the spatial cue before stimulus presentation (see Figure 4A), whereas the other two clusters developed during VSTM retention (Figure 4B), before presentation of the probe. Critically, these results from our data-driven approach closely mirror the findings of lateralized alpha power as a mechanism involved in general attentional orienting depicted in Figure 2B and C.

Figure 4. 

Results from the data-driven single-trial regression analysis. Topoplots and time–frequency representations display the group average standardized beta weights for the predictors included in the single-trial regression. Clusters of activity significantly affected by factors of interest are depicted by black outlines. The data-driven analysis revealed greater alpha activity over ipsilateral versus contralateral scalp regions before stimulus (A) and probe (B) presentation. Moreover, both distractor load and target load in the relevant hemifield modulated early theta activity, as indicated by significant interactions between cue side and distractor load in the cued hemifield (C) and between cue side and memory load (D), respectively. (E) Distractor load in the irrelevant hemifield was not associated with modulations of lateralized (alpha) oscillatory activity, as the interaction between cue side and the number of distractors in the irrelevant hemifield did not reach significance in any time–frequency window across electrodes.

Figure 4. 

Results from the data-driven single-trial regression analysis. Topoplots and time–frequency representations display the group average standardized beta weights for the predictors included in the single-trial regression. Clusters of activity significantly affected by factors of interest are depicted by black outlines. The data-driven analysis revealed greater alpha activity over ipsilateral versus contralateral scalp regions before stimulus (A) and probe (B) presentation. Moreover, both distractor load and target load in the relevant hemifield modulated early theta activity, as indicated by significant interactions between cue side and distractor load in the cued hemifield (C) and between cue side and memory load (D), respectively. (E) Distractor load in the irrelevant hemifield was not associated with modulations of lateralized (alpha) oscillatory activity, as the interaction between cue side and the number of distractors in the irrelevant hemifield did not reach significance in any time–frequency window across electrodes.

Interestingly, our data-driven approach also revealed an early interaction between cue direction and distractor load in the cued hemifield (Figure 4C). Specifically, between ±100 and 500 msec poststimulus presentation, contralateral theta band (2–5 Hz) activity increased as a function of number of distractors in the relevant hemifield (please note that the topography of this effect differs from the topography of the theta band activity displayed in Figure 3F; see Figure 3A). Although one could interpret this as an effect of distractor load, contralateral theta band activity over similar scalp regions also increased in this same time window as a function of the number of targets presented, as indicated by an interaction between cue direction and memory load (Figure 4D). These observations together indicate that early contralateral theta activity may have increased as a function of the number of visual stimuli presented in the relevant hemifield, regardless of whether they were targets or distractors. In line with this conclusion, the beta weights of the two clusters were strongly correlated across participants (r29 = .866, p < .0001). No other clusters displaying an interaction between cue direction and distractor load in the relevant hemifield were observed. It is possible that individual differences in VSTM capacity masked effects of distractors at the group level. To address this possibility, we post hoc computed the significance of the regression weights for low and high VSTM capacity subjects separately (subgroups defined using a median split of Kp-max). However, there were no clusters for which the interaction between cue side and number of distractors in the cued hemifield reached significance in either VSTM subgroup. Thus, our more data-driven approach also did not reveal an oscillatory mechanism selectively involved in reactive suppression of distraction in the relevant hemifield.

Similarly, no significant interaction was observed between the number of distractors in the irrelevant visual field and the cued hemifield during VSTM retention. This indicates that the above-reported nonreplication of a role for lateralized occipital alpha activity in reactive filtering of distractors in the irrelevant hemifield (as reported by Sauseng et al., 2009) cannot just be attributed to our choice of the electrodes and time–frequency window used in that analysis. Inspection of the uncorrected beta weights (Figure 4E) also did not reveal any evidence for the involvement of ipsilateral occipital alpha activity in distractor filtering of distractors in the irrelevant visual hemifield. Thus, although we observed a clear effect of spatial attention on lateralized occipital alpha activity, we found no evidence for reactive modulation of lateralized alpha activity during VSTM in the presence of distractors using a data-driven approach either.

We also inspected the results of the regression for main effects of distractor load in the cued and irrelevant field to identify oscillatory mechanisms that may underlie distractor detection or suppression. However, the presence of distractors in the cued as well as the irrelevant field only seemed to modulate early visual processing following stimulus presentation but did not involve activity outside visual areas.

ERP Components Related to (Failed) Distractor Filtering

Contrary to our predictions, time–frequency analyses did not reveal any distractor filtering-related modulations of (alpha) oscillatory power during VSTM retention. Yet, previous ERP studies have shown robust effects of distractors presented among targets in the relevant hemifield on ERP components related to distractor storage (the CDA) and distractor signaling (the frontal bias signal; Liesefeld et al., 2014; Vogel et al., 2005). We therefore examined if we could replicate these findings and observe ERP correlates of (failed) distractor filtering.

The amplitude of the CDA scaled with the number of to be remembered items (F(2, 60) = 8.41; p < .01) but leveled off at VSTM capacity (cf. e.g., Vogel & Machizawa, 2004; Figure 5B). Moreover, in line with previous reports (e.g., Liesefeld et al., 2014; Vogel et al., 2005), VSTM capacity predicted the extent to which distractors were prevented from being stored in VSTM across individuals, as indicated by a trend toward a negative relationship between individual VSTM capacity and the effect of distractors on CDA amplitude, both for the effect of two distractors (2T0D vs. 2T2D; r29 = −.26, p = .076; see Figure 5C) as well as for the effect of four distractors (2T0D vs. 2T4D; r29 = −.28, p = .061). Repeating this analysis with the estimate of VSTM capacity used in previous research (Cowan's K instead of Pashler's K), yielded similar results (r29 = −.28, p = .065 for 2T0D vs. 2T2D; r29 = −.27, p = .075 for 2T0D vs. 2T4D).

Figure 5. 

ERP components captured (failed) reactive distractor filtering. (A) Scalp topography of the condition average CDA (contralateral minus ipsilateral activity) in the time window 350–1250 msec postpresentation of the memory array. Black dots represent the electrodes used to compute the CDA. Contralateral electrodes are depicted over the left hemisphere. (B) CDA waveforms for the Memory Load 2 conditions with zero, two, and four distractors. (C) Correlation between individual VSTM capacity (Kp) and the efficiency of distractor filtering (excluding distractors from VSTM storage) as indicated by the effect of two (black data points) and four (gray data points) distractors on the amplitude of the CDA. (D) Scalp topography of the condition average frontal bias signal between 240 and 290 msec poststimulus. Black dots represent the electrodes used to compute the frontal bias signal. (E) FBS waveform during the first 300 msec following stimulus presentation (left), and ERP waveforms in the same time window for the conditions without distractors (no reactive filtering) and with distractors (reactive filtering) that were used to compute the FBS (right). (F) Distractor signaling, as reflected by the distractor-related amplitude of the frontal bias signal, was negatively related to unnecessary storage of distractors in VSTM, as indicated by the increase in CDA amplitude when two versus no distractors were present at Memory Load 2 (2T2D vs. 2T0D) across participants.

Figure 5. 

ERP components captured (failed) reactive distractor filtering. (A) Scalp topography of the condition average CDA (contralateral minus ipsilateral activity) in the time window 350–1250 msec postpresentation of the memory array. Black dots represent the electrodes used to compute the CDA. Contralateral electrodes are depicted over the left hemisphere. (B) CDA waveforms for the Memory Load 2 conditions with zero, two, and four distractors. (C) Correlation between individual VSTM capacity (Kp) and the efficiency of distractor filtering (excluding distractors from VSTM storage) as indicated by the effect of two (black data points) and four (gray data points) distractors on the amplitude of the CDA. (D) Scalp topography of the condition average frontal bias signal between 240 and 290 msec poststimulus. Black dots represent the electrodes used to compute the frontal bias signal. (E) FBS waveform during the first 300 msec following stimulus presentation (left), and ERP waveforms in the same time window for the conditions without distractors (no reactive filtering) and with distractors (reactive filtering) that were used to compute the FBS (right). (F) Distractor signaling, as reflected by the distractor-related amplitude of the frontal bias signal, was negatively related to unnecessary storage of distractors in VSTM, as indicated by the increase in CDA amplitude when two versus no distractors were present at Memory Load 2 (2T2D vs. 2T0D) across participants.

Replicating previous findings (Liesefeld et al., 2014), we observed a distractor-sensitive frontal bias signal centered on electrode AFz (incorporating F1, F3, AF3, Fz, AFz, AF4, F4, and F2; see Figure 5D), with a peak between 240 and 290 msec after stimulus presentation, of which its amplitude differed between conditions with and without distractors (t(30) = 2.18, p = .037; see Figure 5E). Moreover, further in line with previous findings, we found a negative cross-subject correlation between the amplitude of this frontal distractor detection signal and unnecessary storage of distractor stimuli, as reflected in distractor-related amplitude of the CDA and frontal bias signal (2T0D vs. 2T2D; r29 = −.45, p = .006; see Figure 5F). Thus, our ERP analyses confirmed that participants did recruit reactive filtering mechanisms in response to the presence of irrelevant distractor stimuli in our study.

DISCUSSION

In this EEG study, we examined the functional characteristics and neural mechanisms of proactive and reactive distractor filtering. Behaviorally, distractors presented among targets in the relevant hemifield impaired VSTM performance, but distractors in the irrelevant hemifield did not. As the presence of distractors in the irrelevant hemifield could be predicted, it is possible that proactive suppression of information processing in the irrelevant hemifield prevented them from influencing target encoding. Indeed, time–frequency analysis of our EEG data showed that anticipated distraction was accompanied by alpha modulations over lateral occipital areas, with stronger alpha activity over the hemisphere processing distractors compared with the hemisphere processing the targets (prestimulus alpha lateralization). Also, the degree of proactive distractor suppression (measured as prestimulus alpha lateralization) was not related to the effect of subsequent distractors in the irrelevant hemifield on behavior across individuals, possibly suggesting that participants were generally well capable of preventing irrelevant-hemifield distractors from influencing task performance. Furthermore, and in contrast to our prediction, we did not observe a reactive adjustment of alpha lateralization that scaled with the number of distractors in the irrelevant hemifield, in contrast to earlier findings reported by Sauseng et al. (2009). One explanation for this discrepancy in findings is that, in our study, participants had more time (700 msec after cue onset) to proactively prepare for upcoming distraction compared with the study by Sauseng et al., in which the stimulus display almost immediately followed the attentional cue (200 msec after cue onset). Thus, in our study, proactive filtering of distractors presented at an irrelevant location was associated with a distractor load-independent lateralization in alpha-band activity both in the prestimulus interval and during VSTM retention. This may suggest that alpha activity in sensory cortices primarily serves to tune the brain toward processing of goal-relevant information and thereby impedes encoding of anticipated irrelevant information into VSTM even when distractor load is high.

Although distractors in the irrelevant hemifield had no effect on VSTM performance, distractors in the relevant hemifield did impair VSTM performance. Moreover, the degree to which these distractors impaired performance was related to individuals' VSTM capacity. Our ERP data mirrored these behavioral findings: We observed a prefrontal filter signal, which determined subsequent unnecessary parietal storage of distracting information, as indicated by the amplitude of the CDA. Although the ERP data clearly revealed the presence of within-hemifield distractor filtering mechanisms, using time–frequency analyses, we could not find evidence for oscillatory power modulations involved in within-hemifield distractor filtering, even when taken a data-driven single-trial regression approach. This analysis revealed that, although early contralateral theta activity scaled with the number of distractors in the cued hemifield, it similarly scaled with the number of targets in the cued visual hemifield, arguing against an interpretation of this activity as a distractor-selective mechanism. On the basis of these results, we conclude that, although participants clearly tried to selectively encode the relevant information into VSTM, as indicated by our behavioral and ERP findings, selective filtering of irrelevant information is not evidently reflected in power modulations of low-frequency cortical oscillatory activity. Strikingly, a recent EEG study by Noonan et al. (2016) on the neural mechanisms involved in attentional modulation of target and distractor processing also found no evidence for a role for alpha oscillations in distractor suppression, even though their ERP and behavioral findings showed evidence of distractor suppression. Furthermore, the absence of reactive alpha power modulations by unanticipated distractors in our study is in line with recent studies showing that, in the absence of distractors, alpha power modulations bias sensory activity before anticipated goal-relevant events but do not play a role during subsequent sensory processing of information (Bauer, Stenner, Friston, & Dolan, 2014; van Ede, Szebényi, & Maris, 2014). These and our findings suggest that alpha power modulations may not unequivocally be involved in reactive distractor suppression.

We investigated reactive distractor filtering using targets and distractors that were spaced closely together in the cued hemifield, resembling distraction in more realistic situations (Peelen & Kastner, 2014). However, this may have two potentially problematic consequences. First, the distractor and target stimuli may have been processed in (partially) overlapping receptive fields (Kastner & Ungerleider, 2001), preventing or complicating selective modulation of distractor processing. Second, smearing of oscillatory brain activity at the level of the scalp may have masked possible focal effects of processing of target and distractor stimuli. Yet, this issue would likely be most influential in sensory cortex, where initial processing of the targets and distractors may take place in a retinotopic fashion. Moreover, our ERP findings indicate that we were able to capture distractor-dependent changes in cortical activity using our experimental paradigm. Yet, it is important to establish whether distractor-related local modulations of alpha power can be detected using electrophysiological methods characterized by a higher spatial resolution, such as electrocorticography (cf. de Pesters et al., 2016; Harvey et al., 2013).

Another consideration is that our mixed design (in which trials with and without distractors were randomly intermixed in a block) may have generated ambiguity about the actual presence and precise number of within-hemifield distractors across trials. It has recently been shown that people become less efficient in filtering of distractors when they need to switch between different filter settings across trials (Noonan et al., 2016; Jost & Mayr, 2015). The need to flexibly adjust filter settings on each trial may thus have caused participants to adopt a more cautious, filter-like strategy across trials, regardless of the presence and precise number of distractors in the cued hemifield (see also Marini, Demeter, Roberts, Chelazzi, & Woldorff, 2016; Marini, Chelazzi, & Maravita, 2013). Yet, this leaves unexplained why we did observe clear effects of distractors on ERP components and VSTM performance. Future studies are necessary to elucidate the possible differential effects of blocked versus mixed presentation of distractor-present and distractor-absent trials on distractor filtering and associated oscillatory mechanisms. Nonetheless, our current findings indicate that, in situations in which (the amount of) distraction is not always fully predictable, alpha oscillatory power is not involved in reactive distractor suppression.

Note that it is possible that distractors affected oscillatory dynamics that were not investigated in this study, as we only looked at oscillatory power. Perhaps within-hemifield distractor filtering is more robustly reflected in other indices of neural oscillatory activity such as long-distance phase synchronization, which has been suggested to play an important role in top–down attention (Fries, 2015; Siegel, Donner, Oostenveld, Fries, & Engel, 2008; Womelsdorf & Fries, 2007). VSTM retention and top–down attentional control have specifically been shown to be subserved by interregional phase synchronization of alpha and beta band activity (e.g., Womelsdorf & Everling, 2015; Palva, Monto, Kulashekhar, & Palva, 2010; Zanto & Gazzaley, 2009). Furthermore, there is evidence for the involvement of phase–amplitude coupling in attentional modulation of sensory processing: Coupling of gamma power to the phase of alpha oscillations results in rhythmic inhibition of sensory processing (Bonnefond & Jensen, 2015; Jensen, Bonnefond, Marshall, & Tiesinga, 2015; Jensen, Gips, Bergmann, & Bonnefond, 2014). The potential role of interregional phase synchronization as well as phase–amplitude coupling in reactive suppression of irrelevant information needs further attention.

Nonetheless, it is intriguing that we did not observe any power modulations associated with of reactive distractor filtering, in particular given that we did observe effects of distracting information on VSTM performance and ERP indices related to filtering and VSTM storage. We conclude that reactive filtering of distracting information during VSTM retention may not rely on local modulations of occipital alpha oscillatory power.

Acknowledgments

M. E. Vissers, H. A. Slagter, and this study are supported by a VIDI grant from the Netherlands Organisation for Scientific Research (NWO) awarded to H. A. Slagter. We would like to thank Isac Sehlstedt and Bob Bramson for their assistance in data collection and Mike X. Cohen for helpful advice on the robust regression.

Reprint requests should be sent to Marlies E. Vissers, Department of Psychology, Brain and Cognition, University of Amsterdam, Nieuwe achtergracht 129-B, Amsterdam, Netherlands, 1018 WS, or via e-mail: M.E.Vissers@uva.nl.

REFERENCES

REFERENCES
Awh
,
E.
, &
Vogel
,
E. K.
(
2008
).
The bouncer in the brain
.
Nature Neuroscience
,
11
,
5
6
.
Bahramisharif
,
A.
,
Van Gerven
,
M.
,
Heskes
,
T.
, &
Jensen
,
O.
(
2010
).
Covert attention allows for continuous control of brain-computer interfaces
.
European Journal of Neuroscience
,
31
,
1501
1508
.
Baier
,
B.
,
Karnath
,
H.-O.
,
Dieterich
,
M.
,
Birklein
,
F.
,
Heinze
,
C.
, &
Müller
,
N. G.
(
2010
).
Keeping memory clear and stable—The contribution of human basal ganglia and prefrontal cortex to working memory
.
Journal of Neuroscience
,
30
,
9788
9792
.
Bauer
,
M.
,
Stenner
,
M.-P.
,
Friston
,
K. J.
, &
Dolan
,
R. J.
(
2014
).
Attentional modulation of alpha/beta and gamma oscillations reflect functionally distinct processes
.
Journal of Neuroscience
,
34
,
16117
16125
.
Bonnefond
,
M.
, &
Jensen
,
O.
(
2012
).
Alpha oscillations serve to protect working memory maintenance against anticipated distracters
.
Current Biology
,
22
,
1969
1974
.
Bonnefond
,
M.
, &
Jensen
,
O.
(
2015
).
Gamma activity coupled to alpha phase as a mechanism for top–down controlled gating
.
PLoS One
,
10
,
e0128667
.
Cohen
,
M. X.
(
2014
).
Analyzing neural time series data
.
Cambridge, MA
:
The MIT Press
.
Cohen
,
M. X.
, &
Cavanagh
,
J. F.
(
2011
).
Single-trial regression elucidates the role of prefrontal theta oscillations in response conflict
.
Frontiers in Psychology
,
2
,
30
.
Cowan
,
N.
(
2001
).
The magical number 4 in short-term memory: A reconsideration of mental storage capacity
.
The Behavioral and Brain Sciences
,
24
,
87
114
;
discussion 114–85
.
de Pesters
,
A.
,
Coon
,
W. G.
,
Brunner
,
P.
,
Gunduz
,
A.
,
Ritaccio
,
A. L.
,
Brunet
,
N. M.
, et al
(
2016
).
Alpha power indexes task-related networks on large and small scales: A multimodal ECoG study in humans and a non-human primate
.
Neuroimage
,
134
,
122
131
.
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
.
Foxe
,
J. J.
, &
Snyder
,
A. C.
(
2011
).
The role of alpha-band brain oscillations as a sensory suppression mechanism during selective attention
.
Frontiers in Psychology
,
2
,
1
13
.
Fries
,
P.
(
2015
).
Rhythms for cognition: Communication through coherence
.
Neuron
,
88
,
220
235
.
Fukuda
,
K.
, &
Vogel
,
E. K.
(
2009
).
Human variation in overriding attentional capture
.
Journal of Neuroscience
,
29
,
8726
8733
.
Fukuda
,
K.
, &
Vogel
,
E. K.
(
2011
).
Individual differences in recovery time from attentional capture
.
Psychological Science
,
22
,
361
368
.
Gazzaley
,
A.
, &
Nobre
,
A. C.
(
2011
).
Top–down modulation: Bridging selective attention and working memory
.
Trends in Cognitive Sciences
,
16
,
129
135
.
Geng
,
J. J.
(
2014
).
Attentional mechanisms of distractor suppression
.
Current Directions in Psychological Science
,
23
,
147
153
.
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
.
Harvey
,
B. M.
,
Vansteensel
,
M. J.
,
Ferrier
,
C. H.
,
Petridou
,
N.
,
Zuiderbaan
,
W.
,
Aarnoutse
,
E. J.
, et al
(
2013
).
Frequency specific spatial interactions in human electrocorticography: V1 alpha oscillations reflect surround suppression
.
Neuroimage
,
65
,
424
432
.
Jensen
,
O.
,
Bonnefond
,
M.
,
Marshall
,
T. R.
, &
Tiesinga
,
P.
(
2015
).
Oscillatory mechanisms of feedforward and feedback visual processing
.
Trends in Neurosciences
,
38
,
192
194
.
Jensen
,
O.
,
Gips
,
B.
,
Bergmann
,
T. O.
, &
Bonnefond
,
M.
(
2014
).
Temporal coding organized by coupled alpha and gamma oscillations prioritize visual processing
.
Trends in Neurosciences
,
37
,
357
369
.
Jha
,
A. P.
,
Fabian
,
S. A.
, &
Aguirre
,
G. K.
(
2004
).
The role of prefrontal cortex in resolving distractor interference
.
Cognitive, Affective, & Behavioral Neuroscience
,
4
,
517
527
.
Jost
,
K.
, &
Mayr
,
U.
(
2015
).
Switching between filter settings reduces the efficient utilization of visual working memory
.
Cognitive, Affective & Behavioral Neuroscience
,
16
,
207
218
.
Kastner
,
S.
, &
Ungerleider
,
L. G.
(
2001
).
The neural basis of biased competition in human visual cortex
.
Neuropsychologia
,
39
,
1263
1276
.
Kelly
,
S. P.
,
Lalor
,
E. C.
,
Reilly
,
R. B.
, &
Foxe
,
J. J.
(
2006
).
Increases in alpha oscillatory power reflect an active retinotopic mechanism for distracter suppression during sustained visuospatial attention
.
Journal of Neurophysiology
,
95
,
3844
3851
.
Klimesch
,
W.
,
Sauseng
,
P.
, &
Hanslmayr
,
S.
(
2007
).
EEG alpha oscillations: The inhibition-timing hypothesis
.
Brain Research Reviews
,
53
,
63
88
.
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
.
Luck
,
S. J.
(
2014
).
An introduction to the event-related potential technique
.
Cambridge, MA
:
MIT Press
.
Marini
,
F.
,
Chelazzi
,
L.
, &
Maravita
,
A.
(
2013
).
The costly filtering of potential distraction: Evidence for a supramodal mechanism
.
Journal of Experimental Psychology. General
,
142
,
906
922
.
Marini
,
F.
,
Demeter
,
E.
,
Roberts
,
K. C.
,
Chelazzi
,
L.
, &
Woldorff
,
M. G.
(
2016
).
Orchestrating proactive and reactive mechanisms for filtering distracting information: Brain-behavior relationships revealed by a mixed-design fMRI study
.
Journal of Neuroscience
,
36
,
988
1000
.
Mathewson
,
K. E.
,
Lleras
,
A.
,
Beck
,
D. M.
,
Fabiani
,
M.
,
Ro
,
T.
, &
Gratton
,
G.
(
2011
).
Pulsed out of awareness: EEG alpha oscillations represent a pulsed-inhibition of ongoing cortical processing
.
Frontiers in Psychology
,
2
,
1
15
.
Mazaheri
,
A.
, &
Jensen
,
O.
(
2010
).
Rhythmic pulsing: Linking ongoing brain activity with evoked responses
.
Frontiers in Human Neuroscience
,
4
,
177
.
Mazaheri
,
A.
,
van Schouwenburg
,
M. R.
,
Dimitrijevic
,
A.
,
Denys
,
D.
,
Cools
,
R.
, &
Jensen
,
O.
(
2014
).
Region-specific modulations in oscillatory alpha activity serve to facilitate processing in the visual and auditory modalities
.
Neuroimage
,
87
,
356
362
.
McNab
,
F.
, &
Klingberg
,
T.
(
2008
).
Prefrontal cortex and basal ganglia control access to working memory
.
Nature Neuroscience
,
11
,
103
107
.
Myers
,
N. E.
,
Walther
,
L.
,
Wallis
,
G.
,
Stokes
,
M. G.
, &
Nobre
,
A. C.
(
2015
).
Temporal dynamics of attention during encoding versus maintenance of working memory: Complementary views from event-related potentials and alpha-band oscillations
.
Journal of Cognitive Neuroscience
,
27
,
492
508
.
Noonan
,
M. 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
.
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
.
Pashler
,
H.
(
1988
).
Familiarity and visual change detection
.
Perception & Psychophysics
,
44
,
369
378
.
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
.
Peelen
,
M. V.
, &
Kastner
,
S.
(
2014
).
Attention in the real world: Toward understanding its neural basis
.
Trends in Cognitive Sciences
,
18
,
242
250
.
Rihs
,
T. A.
,
Michel
,
C. M.
, &
Thut
,
G.
(
2007
).
Mechanisms of selective inhibition in visual spatial attention are indexed by alpha-band EEG synchronization
.
European Journal of Neuroscience
,
25
,
603
610
.
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
.
Rousselet
,
G. A.
, &
Pernet
,
C. R.
(
2011
).
Quantifying the time course of visual object processing using ERPs: It's time to up the game
.
Frontiers in Psychology
,
2
,
107
.
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
.
Siegel
,
M.
,
Donner
,
T. H.
,
Oostenveld
,
R.
,
Fries
,
P.
, &
Engel
,
A. K.
(
2008
).
Neuronal synchronization along the dorsal visual pathway reflects the focus of spatial attention
.
Neuron
,
60
,
709
719
.
Snyder
,
A. C.
, &
Foxe
,
J. J.
(
2010
).
Anticipatory attentional suppression of visual features indexed by oscillatory alpha-band power increases: A high-density electrical mapping study
.
Journal of Neuroscience
,
30
,
4024
4032
.
Spronk
,
M.
,
Vogel
,
E. K.
, &
Jonkman
,
L. M.
(
2012
).
Electrophysiological evidence for immature processing capacity and filtering in visuospatial working memory in adolescents
.
PloS One
,
7
,
e42262
.
Srinivasan
,
R.
,
Winter
,
W. R.
,
Ding
,
J.
, &
Nunez
,
P. L.
(
2007
).
EEG and MEG coherence: Measures of functional connectivity at distinct spatial scales of neocortical dynamics
.
Journal of Neuroscience Methods
,
166
,
41
52
.
Thut
,
G.
,
Nietzel
,
A.
,
Brandt
,
S. A.
, &
Pascual-Leone
,
A.
(
2006
).
Alpha-band electroencephalographic activity over occipital cortex indexes visuospatial attention bias and predicts visual target detection
.
Journal of Neuroscience
,
26
,
9494
9502
.
van Driel
,
J.
,
Knapen
,
T.
,
van Es
,
D. M.
, &
Cohen
,
M. X.
(
2014
).
Interregional alpha-band synchrony supports temporal cross-modal integration
.
Neuroimage
,
101
,
404
415
.
van Ede
,
F.
,
Szebényi
,
S.
, &
Maris
,
E.
(
2014
).
Attentional modulations of somatosensory alpha, beta and gamma oscillations dissociate between anticipation and stimulus processing
.
Neuroimage
,
97
,
134
141
.
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
.
Wager
,
T. D.
,
Keller
,
M. C.
,
Lacey
,
S. C.
, &
Jonides
,
J.
(
2005
).
Increased sensitivity in neuroimaging analyses using robust regression
.
Neuroimage
,
26
,
99
113
.
Wilcox
,
R. R.
(
2012
).
Introduction to robust estimation and hypothesis testing
.
Cambridge, MA
:
Academic Press
.
Womelsdorf
,
T.
, &
Everling
,
S.
(
2015
).
Long-range attention networks: circuit motifs underlying endogenously controlled stimulus selection
.
Trends in Neurosciences
,
38
,
682
700
.
Womelsdorf
,
T.
, &
Fries
,
P.
(
2007
).
The role of neuronal synchronization in selective attention
.
Current Opinion in Neurobiology
,
17
,
154
160
.
Zanto
,
T. P.
, &
Gazzaley
,
A.
(
2009
).
Neural suppression of irrelevant information underlies optimal working memory performance
.
Journal of Neuroscience
,
29
,
3059
3066
.