Separate Cue- and Alpha-Related Mechanisms for Distractor Suppression
- Department of Neuroscience and Del Monte Institute for Neuroscience, University of Rochester, Rochester, New York 14627
-
Author contributions: Z.V.R. and I.C.F. designed research; Z.V.R. performed research; Z.V.R. and I.C.F. analyzed data; Z.V.R. and I.C.F. wrote the paper.
Abstract
Research on selective attention has largely focused on the enhancement of behaviorally important information, with less focus on the suppression of distracting information. Enhancement and suppression can operate through a push−pull relationship attributable to competitive interactions among neural populations. There has been considerable debate, however, regarding (1) whether suppression can be voluntarily deployed, independent of enhancement, and (2) whether voluntary deployment of suppression is associated with neural processes occurring prior to the distractor onset. Here, we investigated the interplay between pre- and post-distractor neural processes, while male and female human subjects performed a visual search task with a cue that indicated the location of an upcoming distractor. We utilized two established EEG markers of suppression: the distractor positivity (PD) and alpha power (∼8–15 Hz). The PD—a component of event-related potentials—has been linked with successful distractor suppression, and increased alpha power has been linked with attenuated sensory processing. Cueing the location of an upcoming distractor speeded responses and led to an earlier PD, consistent with earlier suppression due to strategic use of a spatial cue. In comparison, higher predistractor alpha power contralateral to distractors led to a later PD, consistent with later suppression. Lower alpha power contralateral to distractors instead led to distractor-related attentional capture. Lateralization of alpha power was not linked to the spatial cue. This observation, combined with differences in the timing of suppression—as indexed by earlier and later PD components—demonstrates that cue-related, voluntary suppression can occur separate from alpha-related gating of sensory processing.
- alpha
- attention
- distractor positivity
- distractor suppression
- electroencephalogram
- voluntary suppression
Significance Statement
Selective suppression of distracting information is important for survival. There are ongoing debates, however, regarding whether and how selective suppression can be voluntarily deployed independent of selective enhancement. Here, we recorded EEG while subjects performed a visual search task with cues that indicated the location of upcoming distractors. Behavioral and electrophysiological results revealed that foreknowledge of a distractor's location speeded suppression, facilitating target detection. The results further revealed that such cue-related suppression of a distractor can occur separate from alpha-related gating of sensory processing. That is, alpha-related suppression in the present task occurred independent of the spatially informative cue (i.e., on both cued and noncued trials) and later than cue-related suppression (as indexed by later and earlier distractor positivity components).
Introduction
The brain uses selective attention to preferentially enhance or suppress the processing of sensory information (Posner et al., 1980). Behaviorally important sensory information is enhanced based on stimulus salience, prior experience, and behavioral goals (Desimone and Duncan, 1995; Fecteau and Munoz, 2006). In comparison, there remains considerable debate regarding distractor suppression (Noonan et al., 2018; Geng et al., 2019; van Moorselaar and Slagter, 2020). This debate largely focuses on whether suppression (1) occurs independent of enhancement (Noonan et al., 2018; Chelazzi et al., 2019), (2) depends on previous experience and/or voluntary control (Theeuwes, 2010), and (3) occurs in anticipation of predictable distractors (Gaspelin and Luck, 2018a). Here, we used a spatially informative cue during a visual search task to investigate voluntary (i.e., goal-directed) distractor suppression, independent of enhancement. We specifically investigated the interplay between pre- and postdistractor neural processes associated with voluntary suppression.
There is growing evidence that neural processes occurring prior to the onset of predictable distractors can contribute to suppression (Noonan et al., 2016; van Zoest et al., 2021; Zhao et al., 2023), but it remains unclear whether anticipatory suppression is under voluntary control. Noonan et al. (2016), for example, presented distractors at a cued location, leading to faster target detection and electrophysiological evidence of distractor suppression (i.e., reduced P1 amplitude in the ERP); however, evidence of suppression was not observed when the cue changed locations from trial to trial. The authors therefore concluded that distractor suppression is mediated by predictive information derived from experience but not by voluntary control mechanisms like those linked to target enhancement.
Contrary to this conclusion, behavioral studies have shown that an informative spatial cue, presented at different locations on different trials, can promote distractor suppression (Munneke et al., 2008; Chao, 2010; Chang et al., 2019). A recent EEG study provided neural evidence linking the voluntary use of spatial cues with distractor suppression (van Zoest et al., 2021). Cueing the distractor location was associated with (1) higher anticipatory (i.e., predistractor) alpha power over contralateral visual cortex and (2) a lower-amplitude distractor–evoked N2pc (van Zoest et al., 2021). Higher alpha power has been repeatedly linked to attenuated sensory processing (Worden et al., 2000; Jensen and Mazaheri, 2010; Foxe and Snyder, 2011; however, see Foster and Awh, 2019), and the N2pc is an ERP component that reflects covert target enhancement (Luck and Hillyard, 1994). These findings therefore provide evidence that the voluntary use of a spatial cue can attenuate attentional capture by distractors. Complementing the N2pc component, the distractor positivity (PD) component reflects covert distractor suppression (Hickey et al., 2009). Cueing the location of an upcoming distractor has also been recently associated with the reduced PD amplitude (Heuer and Schubö, 2020; Zhao et al., 2023), an effect that similarly co-occurred with increased anticipatory alpha power contralateral to a spatial cue (Zhao et al., 2023). The authors interpreted these findings as evidence that anticipatory neural processes (i.e., the lateralization of alpha power prior to distractor onset) reduce the need for reactive distractor suppression (as indexed by a reduced PD component). That is, these findings provided evidence that voluntary suppression in response to spatial cues is driven by anticipatory changes in alpha power.
Here, we further investigated how the potential interplay between pre- and post-distractor neural processes influences distractor suppression. Previous research has more often focused on either pre-distractor neural processes (e.g., anticipatory changes in alpha power) or ERP evidence of successful suppression (e.g., the PD component), but the relationship between these neural markers of suppression is not well understood. The present results indicate that a spatially informative cue (i.e., a cue that indicated the upcoming location of a distractor) was associated with earlier suppression, as indexed by faster response times (RTs) and an earlier PD component. Higher alpha power contralateral to an upcoming distractor was also associated with successful suppression, as indexed by a later PD component. This later, alpha-related suppression, however, occurred despite no consistent, cue-related lateralization of alpha power. In comparison, lower alpha power contralateral to a subsequent distractor was associated with evidence of attentional capture, as indexed by a distractor-evoked N2pc component. These findings therefore demonstrate that cue-related, voluntary suppression can occur independent of alpha-related gating of sensory processing. That is, we demonstrate separate mechanisms associated with successful suppression of distracting information: (1) a cue-related, voluntary mechanism associated with an earlier PD component and (2) a cue-independent mechanism associated with both higher alpha power and a later PD component.
Materials and Methods
Subjects
Twenty-seven individuals with normal or corrected-to-normal vision and no history of neurological disorders participated in the experiment (21 females, 6 males; mean age, 25.3 years). All subjects provided informed consent. The study was conducted in accordance with protocols approved by the Research Subjects Review Board at the University of Rochester.
Apparatus and stimuli
Subjects were seated in a comfortable chair in a sound- and light-attenuated chamber. Chair and table heights were adjusted so that the subject could comfortably use a padded chin and forehead rests. A foot stool was provided if needed. Subjects were instructed to remain as relaxed as possible during trials to minimize noise in the EEG attributable to muscle activity. Task contingencies were controlled with custom Presentation (Neurobehavioral Systems) scripts run on a Dell Precision 5,820 desktop computer with a 27 in Acer Predator XB2 LCD monitor (1,920 × 1,080 pixels at 240 Hz) positioned 57 cm from subjects' eyes. Figure 1 depicts the primary experimental display which consisted of a white (187 cd/m2) fixation square (0.5 × 0.5°) and a visual search array comprising four stimuli of varying shapes (circle, square, diamond, hexagon) drawn in red (49.5 cd/m2; x = 0.536; y = 0.342) or green (46.2 cd/m2; x = 0.309; y = 0.548) on a black background. Stimuli were positioned 4° from the fixation square (above, right, below, left). The circle had a diameter of 1.4°, the square had a width of 1.24°, and the diamond and hexagon had similar areas (all within 10 pixels). Each shape in the search array contained a black line (0.0625 × 0.75°) rotated 45° clockwise or counterclockwise from horizontal.
Experimental task. A peripheral cue indicated the location of a distractor that occurred on 50% of trials. The space bar was pressed to initiate a variable duration delay period (500–1,600 ms). The left or right mouse buttons were used to indicate the orientation of a tilted line within a target of fixed color and shape, regardless of the presence or absence of the distractor.
Procedure and design
Figure 1 illustrates the behavioral task. Each trial started with a cueing period in which a tone sounded after 500 ms to signal that subjects could continue the trial when ready. On 50% of trials, subjects saw only a white fixation square on a black background (noncued), while on the remaining 50% of trials, a cue was presented at one of the four stimulus locations (i.e., top, right, bottom, left), indicating where a distractor could appear in the search array (cued). This display remained in place for a minimum of 500 ms to give subjects sufficient time to process cues. After the 500 ms hold, a tone indicated that subjects could proceed when ready. On cued trials with a distractor, the distractor always occurred at the cued location (i.e., 100% cue validity). The distractor cue was an unfilled white circle with an outside diameter of 1.4° and an internal diameter of 1.3° (i.e., a line thickness of 0.1°). To isolate the effects of distractor suppression, target location always remained unknown prior to the array onset, preventing subjects from relying on anticipatory target enhancement (Wöstmann et al., 2022). When ready, subjects advanced the trial by pressing the space bar on a standard computer keyboard. That is, the trials were self-paced, and the spatial cue provided no temporal information about the array onset. The screen then displayed only the fixation square for a duration sampled randomly from a uniform distribution between 500 and 1,600 ms. Following this delay period, the search array was displayed and remained on screen until subjects responded. Subjects searched the array to find a target shape presented with either three randomly selected neutral shapes of the same color (20% of trials) or two randomly selected neutral shapes of the same color and one randomly selected color-singleton distractor shape of a different color (80% of trials). These conditions were randomized across trials within a block. Previous research has demonstrated different conditions under which a singleton distractor either can or cannot be successfully suppressed, as indexed by the PD component and behavioral performance. Here, we wanted to test the influence of voluntary attentional deployment on the successful suppression of distracting information. We therefore utilized an experimental design that has previously been shown to promote successful suppression of a color singleton. That is, target features and distractor color were held constant, and nontarget shapes were heterogeneous (Burra and Kerzel, 2013). Participants might therefore have used a combination of anticipatory selective processing based on spatial location (i.e., foreknowledge of the distractor location) and color (i.e., foreknowledge of the distractor color). Whereas participants were directed to use the spatial cue to suppress distractors, they were not directed to use color to suppress distractors. To test whether feature-based attention based on experience (Won and Geng, 2020; Addleman and Störmer, 2022, 2023) contributed to the suppression of distractors (i.e., given the consistent distractor color), we collected an additional behavioral dataset (n = 10) while allowing the color of the distractor to vary across trials. The behavioral task was otherwise identical. The behavioral results across the two tasks, with either consistent distractor color (i.e., the main task) or randomized distractor color, were highly consistent (Figs. 2, 3). These findings indicate that the use of a consistent distractor color in the main task did not contribute to distractor suppression. The target shape in both tasks was either a circle or a diamond drawn in either red or green. Target shape and color were counterbalanced across subjects. After identifying the target shape within the array, subjects discriminated the orientation of the line inside the target shape and responded by pressing the left or right mouse button. Correct responses triggered a feedback tone. For 600 ms after the response, the search array disappeared, and the cueing period for the next trial began. After every 20 trials, the error rate for those trials was displayed on the screen. Subjects were instructed to respond as quickly as possible while maintaining error rates of 10% or less. Each subject completed at least 20 practice trials followed by eight blocks of 140 trials for a total of 1,120 trials. Subjects were instructed to maintain fixation without blinking for the duration of each trial (i.e., the period beginning with the subject pressing the space bar and ending with their response). Fixation was monitored using an infrared eye-tracking camera (EyeLink 1,000 Plus, SR Research). If fixation was broken, the trial was aborted, and a warning message appeared, instructing subjects to press the space bar to begin the next trial. If trials were aborted in this way, additional trials were added to the block such that the total number of completed trials for each condition was consistent across subjects.
The distractor cue was associated with faster RTs. A, Grand average RTs for combinations of cue (present/absent) and distractor (present/absent) conditions. B, Mean accuracy for combinations of cue and distractor conditions. C, Grand average RTs for cued (orange) and noncued (pink) distractor trials within overlapping 200 ms bins. All error bars indicate ±1 SEM. Bolded light orange lines (above the RT data) indicate the time points at which RTs significantly differed between cued and noncued trials without correcting for multiple comparisons, while bolded dark orange lines indicate the time points at which RTs significantly differed after correcting for multiple comparisons (p < 0.05). D, Examples of stimulus arrays with targets (green diamonds in this example) in close and far positions relative to the singleton distractor. Dashed white lines are included for illustration and were not visible to subjects. E, Grand average RTs for combinations of distractor condition (cued/noncued) and proximity to targets (close/far). F, Grand average RTs by proximity to cues on trials without distractors.
Behavioral results with unpredictable singleton color. A, Example stimulus arrays depicting variable singleton colors. B, Grand average RTs for combinations of cue (present/absent) and distractor (present/absent) conditions. Error bars indicate ±1 SEM.
EEG recording and preprocessing
EEG data were collected using a BioSemi ActiveTwo system (BioSemi) from 128 active Ag/AgCl electrodes. Data were digitized at 2,048 Hz and downsampled to 512 Hz off-line. All electrode impedances were kept below 20 kΩ. To ensure consistent placement of the EEG cap, the vertex electrode (A1) was placed at 50% of the distance between the inion and the nasion and between the tragus on the left and the right ears. EEG data were analyzed using the FieldTrip toolbox (Oostenveld et al., 2011) and custom scripts written in MATLAB (R2021b). Online, the BioSemi ActiveTwo system uses a feedback loop between an active electrode (Common Mode Sense) and a passive electrode (Driven Right Leg) instead of ground electrodes. Off-line, the continuous EEG data were rereferenced to the average of all electrodes, high-pass filtered (fourth-order Butterworth with 0.5 Hz cutoff), and then segmented into epochs relative to the search array onset (−2.2–0.9 s). Bad channels were visually identified. Artifacts related to blinks and eye movements were excluded from the data by design, as task performance was contingent on maintaining fixation (i.e., trials with blinks and saccades were aborted). Data were visually inspected for each subject and channel to determine a voltage threshold for identifying noise transients within the delay period or the ERP window. Bad data were interpolated using a distance-weighted average of the four nearest good electrodes, but only if the average distance of these electrodes was <4.0 cm (otherwise data were excluded). On average 2.37% of data were interpolated and kept, while 0.67% of data were excluded (0.22% of data from the primary electrodes of interest, A10/B7, were excluded from ERP analyses).
Behavioral analyses
Only trials with RTs <2,000 ms were included in analyses. For analyses of RTs, only correct trials were used. Within subjects, trials were randomly subsampled without replacement from distractor-present conditions to match the average trial counts for distractor-absent conditions. Repeated-measure ANOVA was performed on RT and accuracy measures with distractor presence (present/absent) and cue presence (cued/noncued) as repeated-measure factors. Based on previous research indicating that suppression can spread from a distractor location to nearby target stimuli (Zhao et al., 2023), repeated-measure ANOVA was also performed on RT with cue presence (cued/noncued) and target-to-distractor proximity (close/far) as repeated-measure factors. Figure 2D illustrates the difference between close and far trials. For each subject, trials in the close condition were subsampled to match trial counts for the far condition. Consistent with previous findings (Abrams et al., 2012; Forschack et al., 2023), targets appearing on the horizontal meridian were detected significantly faster than targets appearing on the vertical meridian (p < 0.001). Subsampling of close trials was therefore performed separately for trials with targets appearing on the vertical and horizontal meridians, ensuring a balance of vertical and horizontal meridian trials between the close and far conditions. Post hoc statistical analyses were corrected for multiple comparisons (α = 0.0125).
For analyses of cueing effects on RT as a function of delay duration, all cued and noncued trials were used after averaging across distractor-present and distractor-absent conditions (i.e., there was no subsampling). For each subject, trials were binned by delay duration. Overlapping bins (200 ms width) were created at 50 ms increments (e.g., 500–700, 550–750, …, 1,400–1,600 ms), and RTs were averaged within these bins. Differences between cued and noncued trials at each of these bins were measured using a permutation-based approach to compensate for multiple comparisons (Luck, 2014). Briefly, mean RTs were compared for cued versus noncued trials within each bin using paired-sample t tests. For each subject, trials were randomly reassigned to cued or noncued conditions, and the process of calculating means and performing t tests for each bin was repeated. The maximum t value (tmax) across all bins was recorded. This process of shuffling the data and recalculating tmax across bins was repeated 1,000 times. The resulting distribution of tmax values was then used to determine the likelihood of observed t values given the null hypothesis of no difference between conditions.
ERP analyses
ERP analyses were limited to the time window from −100 to 400 ms relative to the array onset. Time series were baseline corrected using the period from −100 to 0 ms and averaged across trials. With a minimum cue–target delay of 500 ms and self-paced trials (i.e., trials were initiated with a button press after participants processed the spatially informative cue), ERPs associated with the cue did not overlap ERPs associated with the array. For each subject, ERPs in response to the array were calculated at electrodes A10/B7 (Fig. 4D,E), contralateral and ipsilateral to lateralized positions of interest. These electrodes correspond with electrodes PO7/8, which have been typically used to measure the PD component (Hickey et al., 2009; Sawaki and Luck, 2010; Sawaki et al., 2012b; Gaspelin and Luck, 2018b; Wang et al., 2019). Specifically, ERPs were calculated (1) relative to the position of the singleton distractor on distractor-present trials (i.e., cued/distractor-present and noncued/distractor-present trials) and (2) relative to the cued location on cued/distractor-absent trials. For each condition, difference waves were calculated for each subject by subtracting the ipsilateral ERP waveform (i.e., relative to the cue/distractor location) from the contralateral ERP waveform. To achieve this, electrode positions were flipped across hemispheres for trials with the cue/distractor on the right side. Data for these trials were then combined with data for trials with the cue/distractor on the left side. Only trials with midline targets (i.e., top or bottom) were analyzed, eliminating effects of target processing on lateralized ERP components (Woodman and Luck, 2003). For plotting, data were low-pass filtered with a cutoff of 30 Hz. The amplitudes of ERP components were calculated using signed area to minimize biases from the choice of measurement windows (Sawaki et al., 2012a; Luck, 2014). We selected windows for ERP components based on the observation of the grand-averaged waveform combined across both cued and noncued trials (Heuer and Schubö, 2020). For trials with a lateral distractor and midline target, the early PD amplitude was determined by calculating the positive area from 130 to 180 ms, the late PD amplitude was determined by calculating the positive area from 205 to 260 ms, and the distractor-evoked N2pc amplitude was determined by calculating the negative area from 265 to 330 ms. For trials with a lateral target and midline distractor, the target-evoked N2pc amplitude was determined by calculating the negative area from 200 to 270 ms. To test the significance of individual components, we used a nonparametric permutation approach that was developed by Sawaki et al. (2012a). For each of 1,000 iterations, contralateral and ipsilateral labels were randomized across trials, and ERPs were recalculated. The signed areas (positive for early and late PD components and negative for the N2pc component) were calculated within their respective windows for each of these iterations, resulting in a distribution of areas that would be expected if the null hypothesis were true (i.e., there was no component). Statistical significance was then determined using the percentage of permuted results that surpassed the observed signed area. For comparisons of component amplitudes between conditions (i.e., cued vs noncued), paired-sample t tests were performed on the observed signed areas.
The distractor cue was associated with earlier distractor suppression on trials with lateralized distractors and midline targets. A, Grand average ERPs recorded at electrodes contralateral (solid) and ipsilateral (dashed) to lateralized, cued distractors. B, Grand average ERPs relative to lateralized, noncued distractors. C, Contralateral-minus-ipsilateral difference waves relative to cued (blue) and noncued (red) distractors. Gray-shaded boxes represent analysis windows for early PD (130–180 ms), PD (205–260 ms), and N2pc (265–330 ms) components. Results of tests comparing the signed area between cued and noncued conditions are indicated in the corner of the corresponding shaded region. ns, not significant; *p < 0.05. PD latency is indicated by colored diamonds for cued (blue) and noncued (red) trials. ***p < 0.001. D, Topographic map depicting mean amplitude during observed peak early PD window for cued trials (130–180 ms). Average voltages for contralateral minus ipsilateral difference wave. E, Topographic map depicting mean amplitude during observed peak PD window for noncued trials (205–260 ms).
Cue-related distractor suppression was observed when analyses were limited to trials without microsaccades occurring prior to the search array onset. A, Grand average ERPs recorded at electrodes contralateral (solid) and ipsilateral (dashed) to lateralized, cued distractors. B, Grand average ERPs relative to lateralized, noncued distractors. C, Contralateral-minus-ipsilateral difference waves relative to cued (blue) and noncued (red) distractors.
Determination of subject-specific peak alpha frequency
Alpha frequency varies between individuals (Haegens et al., 2014); therefore, peak alpha frequencies were determined for each subject. All alpha power analyses used data from electrodes A10/B7, A15/A28, A16/A29, and A17/A30, a cluster that corresponds to electrodes PO3/4, PO7/8, and O1/2, which have been used in prior work studying the role of alpha in attentional selection (Kelly et al., 2006; Wang et al., 2019). Data for trials with delays longer than 850 ms were isolated from −500 to 0 ms relative to the array onset, and irregular resampling autospectral analysis (Wen and Liu, 2016) was used to obtain oscillatory power spectra at electrodes contralateral to cued locations. Data were padded with zeros to obtain a frequency resolution of 0.1 Hz. For each subject, local maxima in the alpha-band frequency (8–15 Hz) were identified after averaging spectra across trials and electrodes.
Effects of prearray alpha power on ERPs
For each subject, alpha power was measured immediately prior to the array onset. EEG data from each trial were convolved with
a two-cycle, complex Morlet wavelet at the subject-specific peak alpha frequency. The temporal extent of a wavelet is determined
by its frequency and width (e.g., a 9 Hz wavelet with two cycles is ∼222 ms wide). Observed alpha peak frequencies were >9 Hz
for all subjects; therefore, wavelets were centered at −111 ms relative to the array onset to avoid overlap with array-related–evoked
activity (i.e., the sensory response associated with the array onset). All trials had a cue–target delay of at least 500 ms,
such that estimates of prearray alpha power were also not influenced by cue-evoked activity. Lateralization indices were calculated
separately for the four conditions (i.e., cued distractor on the left, cued distractor on the right, noncued distractor on
the left, and noncued distractor on the right) using the following equation:<mml:math display="block"><mml:mrow><mml:mi mathvariant="normal">L</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">I</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">contra</mml:mi></mml:mrow><mml:mstyle displaystyle="false" scriptlevel="0"><mml:mtext>-</mml:mtext></mml:mstyle><mml:mrow><mml:mi
mathvariant="normal">ipsi</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mspace width=".1em"></mml:mspace><mml:mo>=</mml:mo><mml:mspace
width=".1em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:mrow><mml:mrow><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi
mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi mathvariant="normal">contra</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mrow><mml:mtext></mml:mtext><mml:mo>–</mml:mo></mml:mrow><mml:mrow><mml:mi
mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">ipsi</mml:mi></mml:mrow></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">)</mml:mo><mml:mo>/</mml:mo><mml:mrow><mml:mo
stretchy="false">(</mml:mo><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">contra</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mspace width=".1em"></mml:mspace><mml:mo>+</mml:mo><mml:mspace
width=".1em"></mml:mspace><mml:mrow><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">ipsi</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo>.</mml:mo></mml:math>LIcontra-ipsi=(Powercontra–Poweripsi)/(Powercontra+Poweripsi).
Lateralization indices for each condition were then averaged across electrode pairs, and trials were split based on the resulting
values (top and bottom, 33%). ERPs were then separately calculated, using data from electrodes A10/B7, for high- and low-alpha
lateralization trials. Comparisons of ERP component amplitudes between the high- and low-alpha lateralization conditions were
conducted with paired-sample t tests, using the signed area during the same windows used for prior ERP analyses (see above).
To analyze the relationship between prearray alpha lateralization and behavior, we split the trials within cued and noncued conditions by alpha lateralization (as described above), and paired-sample t tests were performed on RT. Based on our behavioral results showing speeded RTs only for targets presented far from the cued distractor location (Fig. 2E), trials with targets near the cued distractor location were excluded from this analysis. Based on these same behavioral results, and the fact that only trials with midline targets can be used to calculate the PD component, it was not possible to test for a potential link between the PD component and RTs. Such relationships have, however, been described by previous studies (Gaspelin et al., 2015; Gaspelin and Luck, 2018b).
Analyses of cue-related alpha power lateralization
For each subject, data were convolved with a two-cycle, complex Morlet wavelet of subject-specific alpha frequency, centered
−111 ms relative to the array onset. As contralateral and ipsilateral labels could not be assigned for noncued trials, we
calculated an alpha lateralization index using the following equation:<mml:math display="block"><mml:mrow><mml:mi mathvariant="normal">L</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">I</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">right</mml:mi></mml:mrow><mml:mstyle displaystyle="false" scriptlevel="0"><mml:mtext>-</mml:mtext></mml:mstyle><mml:mrow><mml:mi
mathvariant="normal">left</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mspace width=".1em"></mml:mspace><mml:mo>=</mml:mo><mml:mspace
width=".1em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:mrow><mml:mrow><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi
mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi mathvariant="normal">right</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mrow><mml:mtext></mml:mtext><mml:mo>–</mml:mo></mml:mrow><mml:mrow><mml:mi
mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">left</mml:mi></mml:mrow></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">)</mml:mo><mml:mo>/</mml:mo><mml:mrow><mml:mo
stretchy="false">(</mml:mo><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">right</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mspace width=".1em"></mml:mspace><mml:mo>+</mml:mo><mml:mspace
width=".1em"></mml:mspace><mml:mrow><mml:mi mathvariant="normal">Powe</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">r</mml:mi></mml:mrow><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">left</mml:mi></mml:mrow></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo>.</mml:mo></mml:math>LIright-left=(Powerright–Powerleft)/(Powerright+Powerleft).
Here, the right and left refer to the electrode positions. Lateralization indices were averaged across the subset of electrodes
used for analyses of ERP components based on alpha lateralization (see above). For each subject, noncued trials were subsampled
to match the average trial count for cue left and cue right conditions. For each condition, alpha lateralization indices were
tested against zero using one-sample t tests, and conditions were compared using one-way ANOVA.
Results
Behavioral evidence of voluntary suppression
We first tested whether the presence of the distractor and the cue (Fig. 2) influenced behavior (i.e., RT and accuracy). For the analysis of RTs, a repeated-measure ANOVA revealed no evidence of an interaction between distractor presence and cue presence (F(1,26) = 0.692; p = 0.413; ηp2 = 0.026; Fig. 2A), but there were statistically significant main effects for both distractor presence and cue presence. RTs were faster for distractor-present trials (734 ms) relative to distractor-absent trials (751 ms; F(1,26) = 11.883; p = 0.002; ηp2 = 0.314). Although it may at first seem counterintuitive that RTs were faster on distractor-present trials, it should be noted that a suppressed distractor decreases the number of stimuli in the array that need to be searched to identify the target (Vatterott and Vecera, 2012; Gaspelin et al., 2015, 2017; Gaspelin and Luck, 2018c). That is, on distractor-present trials, subjects only needed to search three locations (rather than four locations), leading to faster RTs. This same pattern of results has been described by previous studies with the same interpretation (Vatterott and Vecera, 2012; Gaspelin et al., 2015, 2017; Gaspelin and Luck, 2018c). Critically, RTs were also faster for cued trials (732 ms) relative to noncued trials (753 ms; F(1,26) = 21.358; p < 0.001; ηp2 = 0.451). Foreknowledge about the location of an upcoming distractor led to significantly faster RTs, consistent with voluntary suppression of the distractor location.
For analyses of accuracy, repeated-measure ANOVA revealed no evidence of an interaction between distractor presence and cue presence (F(1,26) = 1.802; p = 0.191; ηp2 = 0.065; Fig. 2B). There was also no evidence of either a main effect of distractor presence (F(1,26) = 0.492; p = 0.489; ηp2 = 0.012) or cue presence (F(1,26) = 1.538; p = 0.226; ηp2 = 0.019). For the remaining analyses, we therefore focused exclusively on RTs.
We next measured the difference between RTs on cued and noncued trials as a function of delay length (Fig. 2C), characterizing how cueing effects evolved over time. RTs for cued trials were significantly faster (relative to noncued trials) between 850–1,200 and 1,300–1,400 ms after the delay onset (p < 0.05). After corrections for multiple comparisons, differences remained significant between 950 and 1,100 ms. These results suggest that the effects of a spatially informative cue on distractor suppression were not immediate in the present behavioral task but instead developed after hundreds of milliseconds.
Previous research has shown that suppression can spread from a distractor location to nearby target stimuli (Zhao et al., 2023). We therefore performed a two-way ANOVA to explore proximity effects between the target and distractor stimuli (Fig. 2D,E), revealing a significant interaction between cueing and proximity (F(1,26) = 6.95; p = 0.014; ηp2 = 0.211). Post hoc analyses revealed a significant effect of proximity in the cued condition (close, 740 ms; far, 724 ms; t(26) = 2.778; p = 0.01; d = 0.114), but not in the noncued condition (close, 738 ms; far, 742 ms; t(26) = 0.633; p = 0.532; d = 0.026). There was a significant effect of cueing when the target was further from the distractor (cued, 724 ms; noncued, 742 ms; t(26) = 2.75; p = 0.011; d = 0.125), but not when the target was in close proximity to the distractor (cued, 740 ms; noncued, 738 ms; t(26) = 0.574; p = 0.571; d = 0.016). These results further support the observation that the proximity of targets to a distractor can eliminate the behavioral benefit of a spatially informative cue (Zhao et al., 2023). We also tested whether proximity to the cue influenced response times when no distractor was presented (Fig. 2F). Here, trials counts were low, as only 50% of trials included a cue and only 20% of those trials did not include a distractor. RTs were again faster at locations further from the cue, but the difference in RTs between close and far trials (without a distractor) was not significant (close, 749 ms; far, 738 ms; t(26) = 0.95; p = 0.351; d = 0.183).
Because the color of the distractor in the experimental task (Fig. 1) was consistent across trials, we tested whether implicit feature-based learning (Won and Geng, 2020; Addleman and Störmer, 2022, 2023)—in addition to spatial cueing effects—might have played a role in distractor suppression. We therefore collected an additional behavioral dataset (n = 10) using a nearly identical experimental task but with a variable distractor color (Fig. 3A). The behavioral results from this additional dataset followed the same patterns as those from the primary task (Fig. 2). There was no evidence of an interaction between distractor presence and cue presence (F(1,9) = 0.046; p = 0.834; ηp2 = 0.005; Fig. 3B), and RTs were significantly faster on distractor-present trials (756 ms) relative to distractor-absent trials (791 ms; F(1,9) = 8.674; p = 0.016; ηp2 = 0.491) and on cued trials (764 ms) relative to noncued trials (782 ms; F(1,9) = 6.262; p = 0.034; ηp2 = 0.41). These results indicate that the use of a consistent distractor color in the primary task did not influence suppression.
ERP evidence of voluntary suppression
We next measured the PD—an ERP component associated with successful distractor suppression (Hickey et al., 2009; Sawaki and Luck, 2010; Gaspar and McDonald, 2014; Gaspelin and Luck, 2018b)—and determined whether it was affected by distractor cueing. Figure 4A depicts the ERP waveforms contralateral and ipsilateral to cued distractors on the horizontal meridian. Importantly, the target was on the vertical meridian in each of these trials and therefore did not contribute to lateralized ERPs. Figure 4B shows the corresponding ERP waveforms relative to noncued distractors. Figure 4C shows the contralateral-minus-ipsilateral difference waves for cued and noncued distractors. The area under the difference wave within the early PD window was significantly positive for cued trials (p < 0.001) but did not reach statistical significance for noncued trials (p = 0.089). There was a significant difference in early PD amplitude between cued trials (20.44 µVms) and noncued trials (11.14 µVms; t(26) = 2.688; p = 0.012; d = 0.68). The area under the difference wave within the later PD window was significantly positive for both cued trials (p = 0.004) and noncued trials (p < 0.001). The difference in PD amplitude between cued trials (17.8 µVms) and noncued trials (27.33 µVms), however, did not reach statistical significance (t(26) = 1.723; p = 0.097; d = 0.493). While subjects were able to successfully suppress the distractor on both cued and noncued trials (as indexed by the early and late PD components), foreknowledge of the distractor location was associated with stronger distractor suppression in the earlier window (i.e., we observed a higher-amplitude PD component for cued relative to noncued trials), consistent with faster RTs on cued trials (Fig. 2). This pattern was reversed in the later window (i.e., we observed a higher-amplitude PD component on noncued trials relative to cued trials) but did not reach statistical significance (p = 0.097). Finally, we confirmed that PD latency (Fig. 4C) was significantly earlier on cued trials (171 ms) relative to noncued trials (210 ms; t(26) = 5.833; p < 0.001; d = 1.12). To control for the possibility that cue-induced response preparation could be responsible for effects on PD latency, we used a linear mixed-effect model with RT included as a covariate. After controlling for RT, there was still a significant effect of cue on PD latency (t(51) = 5.935; p < 0.001).
We next sought to confirm this pattern of results using only trials without microsaccades occurring within 500 ms of the search array onset (28.45% of cued trials and 28.6% of noncued trials). Despite reduced power (Fig. 5), the positivity in the difference wave within the early PD window was found to be marginally significant for cued trials (p = 0.064). This was not the case for noncued trials (p = 0.915). In contrast, the positivity within the later PD window was found to be significant for noncued trials (p = 0.003) and not for cued trials (p = 0.993). Finally, we confirmed that PD onset latency was significantly earlier on cued trials (175 ms) relative to noncued trials (208 ms) (t(26) = 5.656; p < 0.001; d = 1.09).
Topographies in Figure 4D,E depict mean amplitudes of the difference waves within the early PD window for cued trials and within the later PD window for noncued trials, respectively. To test for differences between these topographies, we used a cluster-based permutation test. Mean amplitudes were normalized across electrodes to focus the analysis on spatial differences rather than differences in the amplitude. This test revealed no evidence that these topographies were distinct (p = 0.137). Further studies will be necessary to definitively establish whether the earlier and later PD components have the same or different neural generators.
A closer look at Figure 4, suggests that the early and late PD components might have been followed by an N2pc component, potentially indicating selection of the distractor. The area under the difference wave within the N2pc window was significantly negative for cued trials (p < 0.001) but not for noncued trials (p = 0.169). There was a significant difference in N2pc amplitude between cued trials (27.38 µVms) and noncued trials (16.21 µVms) (t(26) = 2.577; p = 0.016; d = 0.473). As ERPs are generated by averaging trials, these components could be associated with different sets of trials (Gaspelin and Luck, 2018a). That is, the PD components could have occurred on trials when subjects successfully suppressed the distractor, while the N2pc component could have occurred on different trials, when the distractor instead captured attention (i.e., subjects failed to suppress the distractor). We later present results that support this interpretation (Fig. 8).
We next examined responses to nonsalient distractors presented at cued locations. Figure 6A depicts the ERP waveforms contralateral and ipsilateral to cued nonsalient distractors on the horizontal meridian. Figure 6B shows the contralateral-minus-ipsilateral difference waves. In contrast with responses to salient distractors, we found no evidence for positivity in the difference wave within the early PD window (p = 0.734).
ERPs revealed no evidence of suppression when nonsalient distractors were presented at the cued location. A, Grand average ERPs recorded at electrodes contralateral (solid) and ipsilateral (dashed) to lateralized cues on trials without a salient distractor. B, Contralateral-minus-ipsilateral difference waves. The gray-shaded box represents the analysis window for the early PD component (130–180 ms). The results of the test of signed areas are indicated in the corner of the shaded region. ns, not significant.
Figure 7 shows the ERP waveforms contralateral and ipsilateral to targets presented on the horizontal meridian when either cued distractors (Fig. 7A) or noncued distractors (Fig. 7B) were presented on the vertical meridian (i.e., lateralized target trials). Here, we focused on the target-evoked N2pc, an ERP component associated with selective enhancement of sensory processing. Figure 7C shows the contralateral-minus-ipsilateral difference waves, isolating the N2pc component. In agreement with previous studies (Luck and Hillyard, 1994), we observed a significant N2pc contralateral to the target. The area under the difference wave within the N2pc window was significantly negative for both cued trials (p < 0.001) and noncued trials (p < 0.001). There was no difference in the N2pc amplitude when distractor location was cued (39.67 µVms) relative to noncued trials (42.62 µVms; t(26) = 0.502; p = 0.62; d = 0.07). Combined with ERP analyses of lateralized distractor trials (Fig. 4), these results indicate that the spatially informative cue influenced suppression of the distractor but not target selection (i.e., enhancement). That is, cue-related speeding of RTs and a cue-related increase in the amplitude of the early PD component (i.e., a proxy for successful suppression) occurred in the absence of cue-related enhancement effects (as indexed by the target-evoked N2pc component).
The distractor cue did not lead to differences in target selection on trials with lateralized targets and midline distractors. A, Grand average ERPs recorded at electrodes contralateral and ipsilateral to targets on trials with cued distractors. B, Grand average ERPs recorded at electrodes contralateral and ipsilateral to targets on trials with noncued distractors. C, Contralateral-minus-ipsilateral difference waves relative to targets on cued trials (blue) and noncued trials (red). The gray-shaded box represents the analysis window for the N2pc (200–270 ms) component. Results of the test comparing the signed area between cued and noncued conditions are indicated in the corner of the shaded region. ns is not significant.
Alpha power lateralization prior to the array onset affects ERP components
Increased alpha power over cortical regions processing behaviorally irrelevant stimuli has been repeatedly associated with sensory suppression (Worden et al., 2000; Kelly et al., 2006; Thut, 2006; Rihs et al., 2007; Foxe and Snyder, 2011). Here, we tested whether the magnitude of alpha lateralization prior to the presentation of the array (i.e., the distractor) was associated with successful distractor suppression. Two of the twenty-seven subjects did not have clear alpha peaks in both the cued and noncued conditions and were therefore excluded from alpha power analyses. The average alpha peak frequency across subjects and conditions was 11.3 Hz with an SD of 0.9 Hz. Alpha peak frequency was significantly higher in the cued condition (11.4 Hz) relative to the noncued condition (11.2 Hz; t(24) = 3.508; p = 0.002; d = 0.7). Individual alpha peak frequencies ranged from 9.4 to 14.2 Hz in the cued condition and from 9.2 to 13.1 Hz in the noncued condition. Subject-specific alpha frequencies for further analyses were based on each subject's peak frequency for the cued and noncued conditions, respectively.
Figure 8A depicts ERP waveforms and difference waves for trials with either high or low prearray alpha power lateralization (i.e., split based on the top and bottom 33% of trials) relative to cued distractor locations. High-alpha trials had higher alpha power contralateral to the upcoming distractor location, while low-alpha trials had lower alpha power contralateral to the upcoming distractor location. On cued trials, the area under the difference wave within the early PD window was significantly positive for high-alpha trials (p < 0.001) but did not reach significance for low-alpha trials (p = 0.077). There was no difference in the amplitude of the early PD when comparing high-alpha and low-alpha trials (high alpha, 27.13 µVms; low alpha, 20.08 µVms; t(24) = 1.272; p = 0.215; d = 0.31). The area under the difference wave within the late PD window was significantly positive for high-alpha trials (p = 0.001) but not for low-alpha trials (p = 0.414). The late PD amplitude was significantly higher on high-alpha trials relative to low-alpha trials (high alpha, 29.33 µVms; low alpha, 15.45 µVms; t(24) = 2.307; p = 0.03; d = 0.507).
Higher prearray alpha lateralization was associated with successful distractor suppression. A, B, Grand average ERPs recorded at electrodes contralateral and ipsilateral to cued (A) and noncued (B) distractors on trials with higher prearray alpha power contralateral to the distractor location (top). Grand average ERPs on trials with lower prearray alpha power (middle). Contralateral-minus-ipsilateral difference waves (bottom). Gray-shaded boxes represent analysis windows for early PD (130–180 ms), PD (205–260 ms), and N2pc (265–330 ms) components. Results of tests comparing the signed area between high- and low-alpha conditions are indicated in the corner of the corresponding shaded region. ns, not significant; *p < 0.05; **p < 0.01.
Figure 8B depicts ERP waveforms and difference waves for trials with high or low prearray alpha power lateralization relative to noncued distractor locations. The area under the difference wave within the early PD window did not reach significance for either high-alpha trials (p = 0.08) or low-alpha trials (p = 0.817). Similar to cued trials, early PD amplitude on noncued trials did not differ between high- and low-alpha lateralization (high alpha, 18.17 µVms; low alpha, 18.40 µVms; t(24) = 0.047; p = 0.963; d = 0.008). The area under the difference wave within the late PD window was significantly positive for high-alpha trials (p < 0.001) but not for low-alpha trials (p = 0.225). Also similar to cued trials, the late PD amplitude was significantly higher on high-alpha trials relative to low-alpha trials (high alpha, 40.28 µVms; low alpha, 21.14 µVms; t(24) = 2.757; p = 0.011; d = 0.701). These results, combined by those in Figure 8A, indicate that higher alpha contralateral to the upcoming distractor location generally led to a higher-amplitude PD component during the later window. This is in contrast to cueing effects on the PD component, which occurred during the earlier window (Fig. 4).
We also tested whether differences in alpha lateralization influenced the distractor-evoked N2pc component (Fig. 8A,B). The area under the difference wave within the N2pc window did not reach significance for high-alpha trials on either cued (p = 0.062) or noncued trials (0.726) but was significantly negative for low-alpha trials on both cued (p < 0.001) and noncued trials (p = 0.003). The amplitude of the N2pc component was significantly reduced on high-alpha trials relative to low-alpha trials on both cued (high alpha, 24.86 µVms; low alpha, 47.76 µVms; t(24) = 2.318; p = 0.029; d = 0.677) and noncued trials (high alpha, 15.58 µVms; low alpha, 39.81 µVms; t(24) = 2.92; p = 0.008; d = 0.711). That is, lower-alpha lateralization relative to the upcoming distractor location generally led to a stronger distractor-evoked N2pc component. Although the distractor-evoked N2pc component did not reach significance on cued high-alpha trials, it was trending toward significance (p = 0.062) relative to noncued high-alpha trials (p = 0.726). This pattern of results could reflect a subset of trials when foreknowledge of the distractor location led to enhancement at the cued location rather than suppression (i.e., an attentional “white bear” effect; Tsal and Makovski, 2006).
To summarize, results from the cued and the noncued conditions indicate that higher lateralization of alpha power contralateral to the upcoming distractor location did not modulate the earlier PD component; however, it did modulate the later PD component. This is in contrast to cue-related effects on the PD component, which occurred during the earlier window. Lower-alpha lateralization contralateral to the upcoming distractor location, in contrast, was associated with attentional capture by the distractor (as indexed by the distractor-evoked N2pc component). The level of prearray alpha lateralization, relative to the distractor location, was therefore an indicator of subsequent distractor suppression or attentional capture by the distractor. These results provide evidence that the PD and the N2pc components, which are both observable in Figure 4, occurred on different trials. That is, on a given trial, the distractor was either suppressed or selected (not suppressed, then selected).
We next tested whether there was a relationship between prearray alpha lateralization and RTs (Fig. 9). Our previous behavioral analysis (Fig. 2E) revealed that RTs were faster at the target location furthest from the cued distractor location (relative to target locations near the cued distractor location), indicating proximity effects (Zhao et al., 2023). We therefore limited the present analysis to trials with targets that occurred at the location furthest from the distractor (Fig. 2D). The results demonstrate that greater alpha lateralization prior to the array onset (i.e., higher alpha power contralateral to the upcoming distractor location) speeded RTs on noncued trials (high alpha, 669 ms; low alpha, 694 ms; t(24) = 2.739; p = 0.012; d = 0.194) but not on cued trials (high alpha, 660 ms; low alpha, 650 ms; t(24) = 1.365; p = 0.185; d = 0.082). The absence of a relationship between alpha lateralization and RTs on cued trials might reflect the cue-related occurrence of a separate mechanism for suppression, associated with an early PD (Fig. 4) and faster RTs (Fig. 2). In comparison, prearray alpha lateralization was primarily associated with a later PD (Fig. 8). That is, higher-alpha lateralization relative to an upcoming distractor does not appear to further increase behavioral benefits associated with the cue.
The relationship between alpha lateralization and response speed depends on both cueing and the proximity between distractor and target stimuli. Grand average RTs for high (blue) and low (red) alpha lateralization within cued and noncued conditions using only trials on which the target was further from the distractor location. The inset depicts an example of this array configuration. Dashed white lines are included for illustration and were not visible to subjects. Results of significant post hoc comparisons are indicated using a horizontal black bar. *p < 0.05 (corrected for multiple comparisons).
Prearray alpha power relative to the cued location
We next tested whether alpha power after the cue and before the array onset was consistently higher contralateral to the cued distractor location (van Zoest et al., 2021; Zhao et al., 2023). Figure 10 depicts alpha power lateralization prior to the array onset for cue left, cue right, and noncued trials. Alpha power lateralization was not different from zero for either cue left (LI = −0.011; t(24) = 0.464; p = 0.647; d = 0.093), cue right (LI = 0.021; t(24) = 0.958; p = 0.348; d = 0.192), or noncued trials (LI = −0.0008; t(24) = 0.037; p = 0.971; d = 0.008). There was no difference in alpha lateralization between cue left, cue right, and noncued trials (F(2,48) = 0.419; p = 0.563; ηp2 = 0.017). These results indicate that the spatially informative cue was not reliably associated with a relative increase in anticipatory (i.e., prearray) alpha power contralateral to the upcoming distractor location, consistent with other recent studies (van Moorselaar et al., 2021; Ferrante et al., 2023). See below for a detailed discussion of how we interpret these findings.
No cue-related differences in prearray alpha power lateralization. The line within each box represents the median, while the upper and lower boundaries of each box represent the upper and lower quartiles, respectively. Lines outside the box represent maximum and minimum data points excluding outliers. Outliers were >1.5 times the interquartile range outside of upper and lower quartiles and are denoted with an open circle.
Discussion
Here, we demonstrated evidence of suppression, independent of target enhancement, occurring through two mechanisms: (1) a cue-related, voluntary mechanism associated with earlier suppression (as indexed by an early PD component) and (2) a cue-independent mechanism associated with both higher alpha power and relatively later suppression (as indexed by a late PD component). Combined, these findings provide evidence that spatial cueing can influence voluntary suppression separately from anticipatory changes in alpha power. We used the PD component, which occurs following the presentation of a search array, as a proxy for successful distractor suppression (Hickey et al., 2009). Our results reveal a shift in the amplitude of the PD component from an earlier window on cued trials (130–180 ms) to a later window on noncued trials (205–260 ms). Whereas an earlier PD component has been interpreted as reflecting a preemptive suppressive process that prevents attentional allocation toward a distractor, a later PD component has been interpreted as a reactive suppressive process that terminates attentional allocation toward a distractor (Wang et al., 2019; van Moorselaar et al., 2021; Gaspelin et al., 2023). The present findings are only somewhat consistent with this interpretation. The PD component was absent on trials without a singleton distractor (Fig. 6), suggesting that suppression of a stimulus occurring at the cued location was not entirely preemptive. Rather, the occurrence of an earlier PD component on cued trials with a singleton distractor seems to reflect an interaction between preemptive processes (or cue-related) and processes that generate the PD component. In comparison to cue-related effects, higher anticipatory alpha power contralateral to the location of an upcoming distractor was associated with a higher-amplitude PD component in the later window (205–260 ms), occurring on both cued and noncued trials (i.e., independent of the spatially informative cue). That is, there was no relationship between alpha power and the earlier, cue-related PD component. The present findings therefore suggest that cue-related suppression in the present task occurred separately from alpha-related gating of sensory processing (Jensen and Mazaheri, 2010; Foxe and Snyder, 2011).
Alpha-band activity (∼8–15 Hz) plays a key role in selective sensory processing, potentially inhibiting neural populations that are representing task-irrelevant information (Worden et al., 2000; Jensen and Mazaheri, 2010; Foxe and Snyder, 2011) and/or aligning neural activity across brain regions to facilitate communication (Klimesch et al., 2007; Jensen and Mazaheri, 2010; Saalmann et al., 2012; Fiebelkorn et al., 2019). Changes in alpha power can track the deployment of covert spatial attention, with power increasing contralateral to unattended locations (i.e., ipsilateral to attended locations; Worden et al., 2000; Sauseng et al., 2005; Kelly et al., 2006; Thut, 2006; Rihs et al., 2007; Foxe and Snyder, 2011; Händel et al., 2011). Such findings suggest that increased alpha power is associated with the suppression of unattended locations to prevent distraction, but much of the existing evidence to support the role of alpha-related neural processes in distractor suppression comes from studies in which suppression was conflated with target enhancement (Foster and Awh, 2019; Schneider et al., 2022; Wöstmann et al., 2022). That is, anticipatory changes in alpha power might be related to mechanisms of voluntary enhancement rather than mechanisms of voluntary suppression. Several studies, however, have now measured changes in alpha power while manipulating distractor suppression orthogonally to target enhancement, confirming anticipatory modulation associated with expectation suppression (Wang et al., 2019) and voluntary suppression (Wöstmann et al., 2019; van Zoest et al., 2021). Here, we did not observe consistent increases in alpha power contralateral to cued distractor locations. These results are in agreement with other recent reports in which no alpha power increases were seen contralateral to the location of an expected distractor (van Moorselaar et al., 2021; Ferrante et al., 2023). Further research will be needed to parse the specific conditions under which a spatially informative distractor cue leads to a consistent lateralization of alpha power. While we did not observe a consistent lateralization of alpha power relative to the cued location (i.e., the cue was not consistently associated with a contralateral increase in alpha power), we did observe a relationship between prearray alpha lateralization and successful distractor suppression. Regardless of cueing (i.e., on both cued and noncued trials), the amplitude of a later PD component increased when prearray alpha power was greater contralateral to the upcoming distractor. These results therefore support a link between prearray alpha power and subsequent distractor suppression but contrast with a couple of recent studies that have reported a link between ERP markers of attentional capture or suppression (i.e., the N2pc or PD components) and cue-related changes in alpha power under sustained, spatially specific, and proactive control (van Zoest et al., 2021; Zhao et al., 2023). Those studies showed a consistent relationship between the location of distractor cues and anticipatory changes in alpha power, likely reflecting voluntary suppression. In contrast, cue-independent changes in alpha power within the present task might reflect enhancement-related neural processes (Schneider et al., 2022). Given the uncertainty in the present task as to where the target would occur (i.e., we cued the distractor location, and there were multiple potential target locations), target-related neural processes—including potential enhancement-related changes in alpha power—would be expected to be independent of the spatially informative cue.
While prearray lateralization of alpha power in the present task was not associated with the spatially informative cue, it did dissociate trials that resulted in either distractor suppression or attentional capture. The N2pc component is typically associated with attentional selection (rather than suppression); however, previous studies have reported a distractor-evoked N2pc following the PD component (Kiss et al., 2012; Sawaki et al., 2012b; Feldmann-Wüstefeld and Schubö, 2013, 2016; Hilimire and Corballis, 2014; Feldmann-Wüstefeld et al., 2015, 2016). There is an ongoing debate regarding the significance of this so-called P-N flip that occurs with the use of small search arrays [see Gaspelin et al. (2023) for review]. Some propose that the positivity (i.e., the PD component) reflects the initial shifting of attention to the position opposite the rejected singleton distractor and that the later negativity (i.e., the N2pc component) reflects the subsequent capture of attention by the initially rejected singleton (Kerzel and Burra, 2020). Others have suggested that the P-N flip reflects initial suppression of the singleton distractor followed by suppression of the opposite nonsingleton distractor (Drisdelle and Eimer, 2021). The P-N flip, however, could also be explained by these components occurring on separate trials that are merged together in the grand-averaged ERP (Gaspelin and Luck, 2018a). That is, trials characterized by successful distractor suppression contribute to the PD component, while trials characterized by attentional capture contribute to the distractor-evoked N2pc component. The present results support this latter interpretation, as the PD and N2pc components observed in response to singleton distractors were dissociated after splitting trials based on the prearray lateralization of alpha power. Trials with higher alpha power contralateral to an upcoming distractor were associated with a weaker N2pc component and a stronger PD component, consistent with successful suppression. Trials with lower alpha power contralateral to the upcoming distractor were associated with a stronger N2pc component and a weaker PD component, consistent with attentional capture (i.e., a failure to suppress the distractor).
Current best practices for the study of distractor suppression encourage demonstration of suppression relative to a predefined baseline (Wöstmann et al., 2022). While this is not possible for measures such as the PD, the assumption that the PD reflects successful suppression is supported by past studies that demonstrated an association between the PD and below-baseline suppression of behavioral measures (Gaspelin et al., 2015; Gaspelin and Luck, 2018b). For the present experimental task, proximity effects between the distractor and nearby targets (see Fig. 2 and the Materials and Methods) prevented us from directly testing whether the amplitude of the PD component was linked to RTs. The present results did, however, demonstrate that cued trials were associated with both an earlier PD component and faster RTs, indirectly linking these neural and behavioral measures.
Subjects also responded faster when a distractor was presented. This counterintuitive speeding of responses on trials with distractors—reported here and previously (Vatterott and Vecera, 2012; Gaspelin et al., 2015, 2017; Gaspelin and Luck, 2018c)—has been posited to reflect successful suppression of the distractor location, thereby reducing the number of positions that must be searched to locate a target (Gaspelin and Luck, 2018a). Importantly, responses were fastest when the potential distractor location was cued in advance, indicating that subjects strategically used the cue to more effectively suppress distractors. Post hoc analyses further revealed that cueing effects on RTs were limited to targets that were presented furthest from the distractor location, suggesting that suppression spread from the distractor to nearby targets (Hilimire et al., 2009; Zhao et al., 2023). The present findings thus add further support to accumulating evidence that suppression can be voluntarily deployed during selective spatial processing, leading to behavioral benefits (Munneke et al., 2008; Chao, 2010; Chang et al., 2019; van Zoest et al., 2021). It should be noted, however, that several studies have reported no behavioral benefits of cueing a distractor location (Noonan et al., 2016; Wang and Theeuwes, 2018), including during a similar visual search task (Wang and Theeuwes, 2018). The factors contributing to whether distractor locations can be voluntarily suppressed include cue validity (Wang and Theeuwes, 2018), similarities between targets and distractors (Folk et al., 1992), prior engagement of attention at the to-be-suppressed location (Moher and Egeth, 2012; Sawaki and Luck, 2013; Beck et al., 2018), learning of distractor features (Munneke et al., 2008; Chao, 2010; Chang et al., 2019; van Zoest et al., 2021), target and distractor proximity (van Zoest et al., 2021), relative salience of targets and distractors (Wang and Theeuwes, 2020), and similarities between target and distractor features (Barras and Kerzel, 2017).
The present findings provide further behavioral and electrophysiological evidence of voluntary distractor suppression, independent of target enhancement. Although we observed a link between higher prearray alpha power contralateral to an upcoming distractor and successful suppression (as indexed by a later PD component), alpha power was not consistently elevated relative to the location of the spatially informative cue. Observed cue-related effects on distractor suppression (i.e., faster response times and an earlier PD) were therefore separate from alpha-related neural processes. Whereas the lateralization of alpha power prior to the array onset was primarily associated with the amplitude of a later PD component, the presence of a spatially informative cue was primarily associated with the amplitude of an earlier PD component. The present findings thus provide evidence that voluntary suppression, independent of both target enhancement and alpha-related gating of sensory processing, can contribute to selective attention, helping to determine which aspects of the environment will receive preferential processing.
Footnotes
- Received July 28, 2023.
- Revision received May 2, 2024.
- Accepted May 6, 2024.
-
This work was supported by grants from the National Science Foundation (2120539), the National Institutes of Health (R01EY033726), and the Searle Scholars Program to I.C.F.
-
The authors declare no competing financial interests.
- Correspondence should be addressed to Zach V. Redding at zachary_redding{at}urmc.rochester.edu or Ian C. Fiebelkorn at ian_fiebelkorn{at}urmc.rochester.edu.
- Copyright © 2024 the authors
