Oscillatory activity associated with feature-based selective attention and visuo-spatial short-term memory.
Isac Sehlstedt
Abstract
Visual input in everyday life is a mixture of disparate stimuli. Electroencephalography (EEG) research has shown that one difference between high and low working memory capacity individuals is that high working memory individuals filter out distracting information, whereas low do not (or to a lesser extent). Previous studies have found an increase of alpha power suppression for the area of the brain that processes irrelevant information. However, few or none have explored the oscillatory dynamics when relevant and irrelevant information is subsequently processed within the same area of the brain. Using EEG we show that both alpha power differences between occipital hemispheres, and absolute alpha power suppression within one occipital hemisphere seems to be unaffected by the amount of targets and
distracters being presented. Despite previous findings, our results suggest that alpha power suppression over occipital hemispheres does not occur in all situations where irrelevant information is presented. Additionally, we suggest that oscillatory activity patterns found in an established experimental paradigm may be sensitive to design modifications; however, all results can be considered unjustified given our small number of participants.
When searching for an object (e.g. keys to your apartment), it is important to keep track of the goal (finding your keys) and simultaneously update the mental list of places that have already been searched to ultimately reach the goal state. The concept attributed to allow this type of active manipulation and temporary storage of information is working memory (WM) and it is essential for a range of everyday tasks (Baddeley, 1992; Hönegger et al., 2011).
WM is an active memory system. Passive measurements - such as subjective behavioral reports that are collected after an event - are often a combination of successful manipulation of information in WM, and long term memory retrieval. Active measurement techniques (such as Electroencephalography) are able to convey neural events during WM processes. Therefore, Electroencephalography (EEG) is able to provide more clear measures of active WM processing.
Numerous EEG studies have shown correlations between aspects of visual WM and oscillatory activity (e.g. Sauseng et al., 2009; Snyder & Foxe, 2010; Vogel & Machizawa, 2004; Vogel, McCollough, & Machizawa, 2005). Many of these findings have come from the Change Detection paradigm (CDP), in which subjects are asked (i.e. cued) to remember stimuli in one part of their visual field (left or right) and ignore the other. For instance, comparisons of Event Related Potentials (ERPs) between hemispheres (i.e. lateralization) has been shown to correlate with individuals' ability to remember relevant information (Vogel & Machizawa, 2004). Furthermore, Vogel et al. (2005) compared high and low working
memory capacity individuals using a similar experimental paradigm to investigate differences in ability to ignore irrelevant information. They regularly included irrelevant information within the attended visual field and found that ability to ignore irrelevant information correlated with memory capacity. Therefore, WM- capacity seems to be limited by both the ability to remember relevant information (retention), and ignore irrelevant information (filtering).
The differences in retention and filtering found between individuals with low working memory capacity (LWC) and high working memory capacity (HWC) seems to be closely related to attentional control. Fukuda and Vogel (2009) showed that WM capacity was not related to proficiency at directing voluntary attention; however, compared to HWC
individuals, LWC had a reduced ability to overcome involuntary attention capture stemming from irrelevant information. This specific deficiency in ability to control attention gives an explanation to the retention and filtering differences seen between HWC and LWC.
Although, ERP research has resulted in crucial findings regarding WM, many interesting oscillatory dynamics become invisible when utilizing this type of analysis
(Makeig, Jung, Bell, Ghahremani, & Sejnowski, 1997; Cohen, 2011). ERPs are calculated by aligning EEG traces to the onset of a button press or stimulus presentation and averaging the oscillatory activity over time and frequency. Therefore, ERPs only reflect the oscillatory activity that is phase-locked to an event (i.e. evoked activity; Makeig, Debener, Onton, & Delorme, 2004). Time-frequency analysis is an alternative way to analyze EEG that provides richer information about the oscillatory dynamics underlying neurocognitive processes (Cohen, 2011). For instance, time-frequency (TF) analysis allows for investigation of non-phase-locked (i.e. induced) activity that cannot be identified using ERP analyses. In lack of time, the current study focused on EEG measures of filtering capability. Previous findings regarding TF and filtering are discussed below.
Alpha oscillations (10 Hz) were classically thought to be a measure of nil-work or "cortical idling" (Pfurtscheller, Stancak, & Neuper, 1996). This theory was recently refined and Jensen and Mazaheri (2010) neatly summarized research proposing that alpha oscillations is a specific EEG signature (or biomarker) for inhibition of task-irrelevant brain areas. For instance, Sauseng et al. (2009) adapted the CDP and found that alpha suppression increased steadily with increasing load. This alpha suppression was strongest for both the occipital
hemisphere processing irrelevant information, and independent of the amount of relevant information presented to participants. Additionally, ability to ignore irrelevant information (i.e. lateralization of alpha frequency band activity) correlated with individual WM-capacity.
Furthermore, research regarding feature-based search (where appearance decide if stimuli are relevant or irrelevant) have shown, with color and motion stimuli, that suppression of irrelevant stimuli is associated with an increase of alpha power suppression over spatially separate brain areas (dorsal and ventral, respectively) that processed the specific features of the distracting stimuli (Snyder & Foxe, 2010).
Real-life situations most often require visual search to be focused on one area of the visual field (e.g. the table) whilst looking for specific features (e.g. the keys to the apartment). Studies combining spatial and feature based search may let the field advance towards a deeper understanding of oscillatory dynamics during real life situations. To our knowledge, there are no published studies that have combined both feature-based and spatially based attention when investigating oscillatory activity (e.g. alpha power differences between electrodes) within the same spatial area of the cortex. Therefore, further investigation of the oscillatory activity associated with distracter suppression in relation to visual working memory capacity when features of the stimuli decide their relevance (i.e. whether they are relevant or
irrelevant) is warranted.
Based on the findings mentioned above, we investigated alpha power suppression in relation to irrelevant visual information. We used a modified version of the CDP where the relevance of the stimuli was decided by the shape of the stimulus, and irrelevant information could be presented in both the attended and unattended visual field. Behavioral and oscillatory effects of relevant information (targets) and irrelevant information (distracters) was
investigated in one experiment divided into three principal parts (see table 1). The two first parts were used to replicate previous findings by Sauseng et al. (2009), and the last part was
an original contribution. For the behavioral analyses, increasing the amount of both relevant, and irrelevant information in either visual field was presumed to make the task more difficult. Therefore, we hypothesized that increase of relevant, and irrelevant information in either vi-sual field would decrease accuracy rates. The oscillatory analyses will be described now in more detail with consideration to each principle part described above.
First, the effect of relevant information was investigated using three conditions with increasing amount of targets (two, four, and six targets, respectively) and equal amount of uncued distracters. The motivation to include these conditions was that increase in memory load also resulted in an equal increase of irrelevant information (as in Sauseng et al., 2009). Therefore, as suppression of irrelevant information has been related to in an increase of alpha power suppression, we anticipated an increase of alpha power suppression ipsilateral to the attended visual field. Second, the effect of uncued distracters was discerned from that of targets by adding two conditions with varying amounts of relevant and irrelevant information in the attended and unattended visual field (2 targets and 4 uncued distracters, and 4 targets and 2 uncued distracters, respectively). As specified above, increase of irrelevant information is related to an increase of alpha power suppression. Therefore, we anticipated that alpha power suppression ipsilateral to the attended visual field would vary with the amount of unattended distracters, and not with the amount of targets (as in Sauseng et al., 2009). Last, two conditions were added in order to investigate the effect of cued distracters (including 2 or 4 cued distracters, and 2 cued targets). As with the previous two hypotheses, irrelevant
information should increase alpha power suppression for the hemisphere that processes the irrelevant information. If irrelevant information is present among relevant information in one visual field, alpha power suppression should increase with the amount of irrelevant
information in that visual field. Therefore, we predicted that an increase of irrelevant information in the attended visual field (i.e. cued distracters) would increase alpha power
suppression contralateral to the attended visual field.
Finally, one finding that we wanted to replicate was the finding of a close relationship between ability to ignore irrelevant information and WM-capacity. It has been shown that lateralization of alpha power suppression between conditions near peak memory capacity correlated with a measure of WM-capacity (Sauseng et al., 2009). We therefore hypothesized that there would be a correlation between alpha power lateralization measures and individual memory capacity.
Method
Participants
Twenty-two Dutch university students were recruited through a database of volunteers at the University of Amsterdam. Ten participants were excluded because of artifacts (N=7),
technical difficulties (N=2), and age (N=1), Twelve Dutch university students (9 women, mean age 23.4, age range 19-26, 2.5 SD) are included in the analysis. Participants received 15 Euros or course credit as compensation for their participation. The study was approved by the ethical committee at the University of Amsterdam.
Materials
The CDP (Vogel et al., 2005; Sauseng et al. 2009) task was adapted. It contained conditions in which distracters were presented within the cued visual field. Stimuli were colored squares (targets) and circles (distracters). All stimuli were presented below fixation and minimally 1.9 visual degrees vertically from fixation within two (3.66 visual degrees X 3.66 visual degrees) square regions. Stimulus locations were mirrored between the visual fields and the colors of the stimuli were randomized using predefined colors. All stimuli were placed on a gray background using predefined location templates created in Matlab with a minimum distance of at least 1.459 visual degrees between stimuli (center to center). Stimuli sizes of targets and
distracters were equal in their area. Squares were 1 visual degree wide, and circles had a radius of 0.65419 visual degrees.
The experiment was divided into two parts: A practice task that included five conditions with eight trials each, and the actual experiment that included four additional conditions. All nine conditions in the actual experiment had 100 trials each. Each condition presented individual amounts of cued targets, cued distracters, and uncued distracters, disregarding two conditions that were identical on all for all types of stimuli (see Methods, sub-headline EEG preprocessing). The actual experiment was divided into blocks (of 36 trials) that included two right cue (one change, one no-change) trials and two left cue trials (one change, one no-change) from each condition, resulting in a total of 25 blocks. Trials from different conditions were presented in a random order. The task was programmed using NBS Presentation.
EEG data was acquired using a BioSemi ActiveTwo amplifier from 64 scalp electrodes and four peri-occular electrodes.
Procedure
All participants performed two practice blocks of 20 trials before starting the actual
experiment. The participants were seated 90 cm away from the computer screen. They read through instructions in their own pace before starting the first practice block (and key
instructions were repeated before starting the actual experiment). The participants were asked to keep their eyes on a fixation cross and covertly attend (shifting attention without shifting gaze) to only one visual field which was determined by a cue at the start of every trial. The entire trial sequence is depicted in Figure 1. The first trial of every block was preceded by a 3000 ms presentation of a fixation cross. The fixation was followed by a 200 ms cue picture showing the relevant visual field. A memorize display with condition specific stimuli was presented 500 ms after the cue picture for 250 ms. A 1000 ms retention delay was then
followed by a 2000 ms probe where one of the rectangles in the cued visual field had, or had not changed color. Neither locations of the stimuli in either visual field, nor the colors of stimuli in the uncued visual field were changed during a trial. We used an inter trial interval (ITI) that was randomly determined and lasted between 1700-2300 ms. During half of the trials, one square in the cued visual field changed color. The participants were to report whether or not a square changed color in the cued visual field by means of a right hand (change) or left hand (no change) button box while the 2000 ms probe was displayed. During the first practice block participants got an extra 750 ms presentation of feedback after every trial with the text “Correct” for correct trials and “Wrong” for incorrect trials. All blocks started with a reminder to "keep eyes on fixation" throughout the coming block, and ended with a block summary on the condition average percentage correct. The participants were asked to take a self-paced break after each block had ended.
EEG preprocessing
EEG preprocessing was performed offline. Data was re-referenced using the average activity of the earlobes. The continuous EEG data was epoched into 5.5 second epochs (2.2 sec before to 3.3 sec after the onset of the memorize display). Data were stimulus-locked: time 0 represented the onset of the memorize display. A high pass filter of 0.05 Hz was applied in order to remove slow drifts from the data. The channel baseline correction for each trial was done with the average activity -200 to 0 ms preceding the onset of the cue display for every trial. All trials were visually inspected using EEGLAB and trials with artifacts were marked and removed. Blink artifacts were removed using Independent Component Analysis (ICA). HEOG channel data was visually inspected and trials with horizontal eye movements removed. Behaviorally incorrect trials were identified and removed.
Correction of between condition trial variability was performed to insure that no within-subject effects were confounded by differences in amount of trials between conditions.
We identified the condition with the smallest number of trials for each participant. All conditions for each participant were allowed a 20 trial difference compared to the condition with the least trials for that participant. If a condition for a participant exceeded this 20 trial limit, then we randomly extracted trials at our maximally accepted within-participant condition difference (20 trials more than the conditions with the least number of trials) from that condition. The averaged total amount of trials remaining before and after this correction are shown in Table 1. One condition was removed post experimentally because it presented incorrect amounts of targets and distracters, and a condition with 4 targets, 4 cued distracters, and 8 uncued distracters was not included in any further analysis.
Power (amplitude squared) was extracted from each condition separately using a family of 20 logarithmically spaced wavelets from 1 Hz to 40 Hz using in-house Matlab code (similar to Cohen, Ridderinkhof, Haupt, Elger, & Fell, 2008). After time-frequency
decomposition, power was converted to decibel-scale using the average activity between 250 and 50 ms preceding the memorize display, which enables comparison of power across frequency bands.
Evoked oscillatory activity was subtracted from the data in order to study the effects of the experimental manipulations on induced oscillatory activity (Sauseng et al., 2009).
Trials across cue-sides were collapsed. All left cue trials were flipped to right-cue trials, by swopping the data on left and right EEG channels to the mirrored position in the opposite hemisphere using in-house written Matlab code. Hence, after mirroring, all trials reflect right cue trials. This was done to maximize the statistical power, and was possible because none of our hypotheses were aimed at exploring differences between cue sides.
Alpha lateralization was computed by subtracting the alpha power activity of one channel from the alpha power activity in the mirrored channel in the opposite hemisphere. Both contralateral electrodes minus ipsilateral electrodes and ipsilateral electrodes minus
contralateral electrodes were calculated (see Figure 2). All lateralization analyses were then done using the new dataset.
Data analyses
All statistical analyses were performed using SPSS (Windows version 20). Each factor is explicitly defined concerning the amount of cued targets (T), cued distracters (D), and uncued distracters (UD) load. All statistical analyses used a significance level of p <.05.
Behavioral analyses
Basic calculations of accuracy rates were analyzed on a scale from 0-1 with 1 representing an error free performance (see Table 1). Increase of targets, cued distracters, and uncued
distracters were hypothesized to decrease accuracy rates Three separate analyses were calculated in order to investigate what effect all three types of stimuli had on accuracy rates. First, the effect of increasing target load (with the same number of uncued distracters) on accuracy rate was investigated using one-way repeated measures ANOVA with the factor TARGET_LOAD (2T, 4T, 6T). Second, the effect of different target and uncued distracter loads on accuracy rate was investigated using a two-way repeated measures ANOVA with the factor TARGETS (2T, 4T) and DISTRACTERS (2UD, 4UD). Last, the effect of increasing cued distracter load on accuracy rates was investigated using a one-way repeated measures ANOVA with the factor CUED_DISTRACTER_LOAD (0T, 2T, 4T).
Post-hoc analyses of error rates were performed. Misses (when there was a change in the cued visual field but the participants did not report that there was a change) and false alarms (when there was no change but participants reported a change) were analyzed in order to get a better understanding as to why errors were made (see Figure 3). We compared error rates (i.e. misses and false alarms) only for the conditions investigating effects of increasing target load (as in Sauseng et al., 2009). A two-way repeated measures ANOVA with factors ERROR_TYPE (false alarm, miss), and MEMORY_LOAD (2T, 4T, 6T) was calculated.
Behavioral data was used to compute the memory capacity measure Cowan's K (Cowan, 2001). The measure K was calculated for all participants individually for the three conditions that included equal amounts of targets and uncued distracters (see Table 1). The highest measure of K (K_max) for these conditions were used as a measure of memory capacity (Sauseng et al., 2009). The equation for calculation of K is, K = (Hit rate- False alarm rate) * Set size. Hit rate was computed per condition as the percentage of trials in which there was a change in the cued visual field and participants reported a change. False alarm rate was computed per condition, as the percentage of trials where participants reported a change in the cued visual field when there was none. Set size is the number of squares (targets) in the cued visual field.
EEG analyses
All EEG analyses were performed on the averaged activity over the delay period (1000 ms window between memorize display and probe; Sauseng et al., 2009). Averaged oscillatory activity at 8-12 Hz was used for all EEG analyses.
All power analyses were done on the channel flipped dataset and exclusively on induced oscillatory activity. The channels used for these analyses were the cued (average activity of P1, PO3, O1) and uncued (average activity of P2, PO4, O2) based on them having the most pronounced condition average activity (see Figure 2). In sum, increase of distracting information in either visual field was hypothesized to increase alpha power suppression in the contralateral occipital hemisphere.
In order to compare hemispheric alpha power differences with increasing target load in conditions with equal amounts of cued targets and uncued distracters, and replicate the
findings by Sauseng et al. (2009), a two-way repeated measures ANOVA with factors HEMISPHERE (cued vs. uncued), MEMORY_LOAD (2T, 4T, 6T) was performed.
activity between hemispheres was investigated (Sauseng et al., 2009). A three-way repeated measures ANOVA with factors TARGETS (2T, 4T), and DISTRACTERS (2UD, 4UD), and HEMISPHERE (cued vs. uncued) was calculated.
The outcome of increasing amount of cued distracters within one occipital hemisphere on alpha power (with a constant target load of two targets) was investigated using a one-way repeated measures ANOVA with the factor CUED_DISTRACTER_LOAD (0D, 2D, 4D for the cued visual field).
All lateralization analyses (aimed at investigating alpha power differences between hemispheres) were done on the induced oscillatory activity in the channel flipped dataset lateralization dataset. The channels used for these analyses were (average lateralization of P3, P5, PO3, PO7) based on them having the most pronounced condition average lateralization over the delay period (see Figure 2). Briefly stated, we predicted that differences in alpha power lateralization measures would correlate with individual memory capacity.
In order to replicate the finding of alpha power lateralization difference between a memory load of four targets and a memory load of two targets correlating with memory capacity (Sauseng et al., 2009), Pearson correlation analysis between K_max (behavioral measure of memory capacity) and occipital alpha power lateralization (lateralization for 4T-4UD minus lateralization for 2T-2UD) was calculated.
Results
Behavior
Behavioral outcomes were analyzed in three principal parts trying to discern the effects of relevant and irrelevant information in either visual field on accuracy rates. In short, we predicted that increased amount of either type of stimuli in either visual field would decrease accuracy rates.
First, accuracy rates were used in order to investigate the behavioral effects of increasing memory load (see Figure 4). A one-way repeated measures ANOVA revealed a significant main effect of memory load F(2,22) = 181.717. Contrasts revealed that accuracy rates were higher for a load of two targets, compared to a load of four targets F(1,11) = 116.463. Additionally, accuracy rates were higher for a load of four targets compared to a target load of six targets F(1,11) = 99.442. This suggests that increasing target affected accuracy rates.
Second, the effect of varying target and distracter load on behavior was investigated using accuracy rates (see Figure 3). A one-way repeated measures ANOVA showed a
significant main effect of targets F(1,11) = 164.541, with accuracy rates being higher for two targets compared to four targets. However, there was no main effect of uncued distracters F(1,11) = 0.203. Additionally, there was no significant interaction effect F(1,11) = 0.001. This suggests target load (but not distracter load) had an effect on accuracy rates.
Last, accuracy rates were used in order to investigate the behavioral effects of
increasing cued distracter load with a constant target load of two targets (see Figure 4). A one-way repeated measures ANOVA revealed a significant main effect of cued distracter load F(2,22) = 37.273. Contrasts revealed that accuracy rates were higher when no cued distracters were presented, compared to a load of two cued distracters F(1,11) = 44.701. Additionally, accuracy rates were higher for a load of two cued distracters compared to a target load of four cued distracters F(1,11) = 16.615. This suggests that increasing amount of cued distracter affects accuracy rates.
Additionally, post-hoc analyses of error rates (misses and false alarms) with increasing memory load were performed in order to get a better understanding why errors were made (Sauseng et al., 2009). A two-way repeated measures ANOVA showed a significant main effect of the type of error that was made F(1,11) = 24.550. There was also a significant main
effect of memory load F(2,22) = 101.326. Contrasts revealed that error rates were
significantly higher for a load of four targets F(1,11) = 67.498, compared to a load of two targets. Error rates were also significantly higher for a load of six targets F(1,11) = 36.828, compared to a load of four targets.
Additionally, there was a significant interaction effect between the type of error that was made and memory load F(1,11) = 19.262. This indicates that memory load had different effects on the error that was made. Miss rates and false alarm rates are comparable at a memory load of two targets; however, misses increased to approximately 3 times the amount of false alarms for the highest loads (see Figure 5). Therefore, increase of memory load had a larger effect on misses when compared to false alarms.
Analysis of within-subject peak memory capacity measure K (K_max) for conditions with the same amount of targets as uncued distracters gave an across subject average of 2.93. Therefore, analysis of K_max suggests an average memory capacity across participants close to three items.
Alpha power
Alpha power was analyzed in three principal parts investigating the different effects targets or distracters had on occipital alpha power suppression. In sum, we predicted that an increase of irrelevant information (distracters) presented in a visual field would increase alpha power suppression in the contralateral occipital hemisphere.
First, hemispheric alpha power differences with increasing target load using conditions with the same number of targets as uncued distracters was investigated (see Figure 6). The average alpha power activity (8-12 Hz) over the delay period for contralateral electrodes (P1, PO3, O1) processing targets in the cued visual field, and ipsilateral electrodes (P2, PO4, O2) processing distracters in the uncued visual field were compared (see Figure 2). A two-way repeated measures ANOVA for alpha power between hemispheres with increasing load
(2T,4T,6T) revealed that alpha power for the ipsilateral hemisphere was numerically higher when compared to the contralateral hemisphere for all loads; however, we found no main effect of hemisphere F(1, 11) = 1.324. There was no main effect of target load F(1, 11) = 0.062. Additionally, there was no significant interaction effect F(1, 11) = 0.178. These results suggest that alpha power suppression is unaffected by increasing target load.
Second, the effect of targets and uncued distracters on alpha power was further investigated using four conditions with two or four targets and two or four uncued distracters that also lacked cued distracters. The average alpha power activity (8-12 Hz) over the delay period for contralateral electrodes (P1, PO3, O1) processing targets in the cued visual field, and ipsilateral electrodes (P2, PO4, O2) processing distracters in the uncued visual field were compared (see Figure 7). A three-way repeated measures ANOVA revealed no main effect of targets F(1, 11) = 0.113, or main effect of distracters F(1, 11) = 0.006. There was a trend for main effect of hemisphere F(1, 11) = 3.529, p = 0.087. Additionally, no interaction effects were significant. These results suggest that alpha power suppression seems to be unaffected by both distracter, and target load.
Last, the effects of increasing cued distracter load (with constant target load) on alpha power was investigated (see Figure 9). The average alpha power activity (8-12 Hz) over the delay period for contralateral electrodes (P1, PO3, O1) was included in the analysis (see Figure 2). A one-way repeated measures ANOVA for the effect of increasing amount of cued distracters (0D, 2D, 4D) on alpha power with a constant target load of two targets showed no main effect of cued distracter load F(1, 11) = 0.48. This suggests that cued distracters do not affect alpha power suppression.
Alpha lateralization
Alpha power differences across occipital hemispheres (alpha lateralization; cf. Sauseng et al., 2009) and a behavioral measure of memory capacity (K_max) were correlated to investigate if one is predictive of the other (as in Sauseng et al., 2009). The average lateralization at four occipital electrodes (P3, P5, PO3, PO7) were used for this analysis. A bivariate correlation between occipital alpha power lateralization (lateralization for 4T-4UD minus lateralization for 2T-2UD) and K_max was calculated. This correlation was not significant (R = 0.285, p = 0.185, R2 = 0.081) which suggests that alpha power lateralization difference between a memory load of four targets and a memory load of two targets does not predict individual memory capacity (see Figure 8).
Discussion
In contrast to our predictions, alpha power suppression did not vary with the amount of irrelevant information presented in neither the unattended, nor attended visual field.
Additionally, we did not find any correlation between alpha power suppression and working memory capacity. However, at the behavior level both increasing target load and cued distracter load had an adverse effect on accuracy rates. Therefore, the methods used in the current study appear to be valid. In sum, we suspect that the current finding might be affected by both the way analyses were performed, and minor modifications to a standard
experimental design. The results are discussed in more detail below.
The lack of significant behavioral effect of distracters with regards to accuracy rates, and general null-finding of alpha power differences appears to be in line with the concept of attentional stimulus enhancement. Specifically, previous research suggest that, instead of suppressing irrelevant information, individuals may enhance attentional resources towards relevant information (Gulbinaite, Johnson, de Jong, Morey, & van Rijn, in press).
(Parks, Beck, & Kramer, 2013; Xu, Monterosso, Kober, Balodis, & Potenza, 2011). For instance, Parks et al. (2013) found that when a combination of the color and orientation of stimuli (and not only one of these features) decided their relevance, participants were more likely to enhance attentional recourses towards targets than to try and reduce attention to distracters. One of our minor design modifications to the CDP was implemented to make the area covered by individual relevant and irrelevant stimuli identical. Because of the relatively close resemblance in appearance between squares and circles, this minor modification might have promoted stimulus enhancement in a more pronounced manner than previous CDP studies. The discussion above highlights the need to further investigate the precise effects of attentional enhancement of targets and suppression of distracters on behavioral performance (Ho & Ester, 2012).
Lack of significant effects in the current study could be because of the current study's relatively larger modifications of the CDP. We presented all stimuli below fixation and the conditions included in our experiment had a larger variation in amount of relevant and irrelevant information that was presented in either visual field compared to previous
experiments using the CDP (e.g. Sauseng et al., 2009; Vogel & Machizawa, 2004; Vogel et al., 2005). This might be indicative of highly specific elements of the original paradigm being crucial for finding oscillatory differences between hemispheres. For instance, Couperus and Mangun (2010) cleverly demonstrated that ERP measures of distracter processing may be affected by ability to anticipate if distracters would be presented during a trial. Although, direct comparisons between ERP and time-frequency data is problematic because of
differences in analysis methods and ongoing controversy regarding the origin of both signals (for a review see Cohen & Gulbinaite, 2014), the current experiment did not provide
participants with any pre-trial information (e.g. if distracters would be present in the attended visual field). Distracters did appear in the attended visual field on randomly selected trials
throughout each experimental block in the current study. This almost certainly resulted in participants being unable to predict when distracters would be present in the attended visual field and might have affected our EEG measures.
A probable principal reason for the lack of findings in the current study might be interactions between Pre- and Post-Stimulus Alpha. Because of an added preparation period after the cue during each trial, and the way data was analyzed with regard to this preparation period, alpha oscillation differences otherwise found using the CDP might have become undetectable. Although, increase of alpha suppression with increasing memory load has been found without taking Pre-Stimulus Alpha (PSA) into account (Sauseng et al. 2009), PSA has been shown to impact subsequent visual processing (for a review see Foxe & Snyder, 2011). In addition, online rhythmic transcranial stimulations at alpha frequencies (10 Hz) and not at control frequencies (5 or 20 Hz) has been shown to affect visual processing by improving visual detection thresholds in one visual field when stimulating the ipsilateral occipito-parietal hemisphere, and worsening detection thresholds when stimulating the contralateral occipito-parietal hemisphere before stimulus presentation (Romei, Gross, and Thut, 2010).
Unfortunately, the baseline used in the current study (average activity of the 250 to 50 ms preceding memorize display) might have been affected by PSA. Therefore, individual differences in ability to modulate PSA might have affected the current results. Moreover, when comparing error rates with Sauseng et al. (2009), we find that the current study had an opposite pattern of error rates, with misses being more frequent than false alarms. As
discussed above, an increase of pre-stimulus may affect subsequent visual processing and might partially explain this behavioral finding.
A significant loss of participants due to experimental issues, and therefore small sample size might have affected the results. More participants would have increased statistical power and may have unveiled differences that could not be found in the current study.
However, it is important to note that previous studies with similar experimental designs and relatively low sample sizes have found significant effects (e.g. Sauseng et al., 2009).
Researchers performing EEG studies struggle with inferring the exact origins of recorded signals and current spread/volume conduction (CS/VC) might have had an effect on the results. Laplacian or current source density (CSD) corrections are usually applied to reduce this confounding effect (Cohen, 2011). Even though Sauseng et al., (2009) applied a CS/VC correction, the current study analyzed induced alpha power on a non-corrected dataset. Therefore, the current study might have had oscillatory activity from one hemisphere spreading to the opposite hemisphere leading to a confounding interaction between
hemispheres.
Humans are almost constantly presented with a stream of information and our ability to maintain attention on visually relevant information while ignoring irrelevant information is paramount in many aspects of life. The current study suggests that our ability to detect biological markers (e.g. alpha power suppression over occipital hemispheres related to the ability of successfully ignoring irrelevant information) in established experimental paradigms might be sensitive to minor paradigm modifications. Therefore, further research is warranted in order to find guidelines to achieve optimal real-life-resemblance in experiments while still being able to identify reliable biological markers.
References
Baddeley, A. (1992). Working memory. Science, 255(5044), 556–559.
Cohen, M. X. (2011). It’s about Time. Frontiers in Human Neuroscience, 5(January), 1-15.
Cohen, M. X., & Gulbinaite, R. (2014). Five methodological challenges in cognitive electrophysiology. NeuroImage, 85(2), 702–10.
Cohen, M. X., Ridderinkhof, K. R., Haupt, S., Elger, C. E., & Fell, J. (2008). Medial frontal cortex and response conflict: evidence from human intracranial EEG and medial frontal cortex lesion. Brain Research, 1238, 127–42.
Couperus, J. W., & Mangun, G. R. (2010). Signal enhancement and suppression during visual – spatial selective attention. Brain Research, 1359, 155–177.
Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. The Behavioral and Brain Sciences, 24(1), 87-114; discussion 114-85.
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(July), 1-13.
Fukuda, K., & Vogel, E. K. (2009). Human variation in overriding attentional capture. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 29(27), 8726–33.
Gulbinaite, R., Johnson, A., Jong, R. De, Morey, C. C., & Rijn, H. Van. (2014). Dissociable mechanisms underlying individual differences in visual working memory capacity. NeuroImage, (In press).
Ho, T. C., & Ester, E. F. (2012). Target enhancement and distractor suppression in naturalistic visual search. The Journal of Neuroscience, 32(47), 16539–40.
Hönegger, C., Atteneder, C., Griesmayr, B., Holz, E., Weber, E., & Sauseng, P. (2011). Neural correlates of visuo-spatial working memory encoding--an EEG study. Neuroscience Letters, 500(2), 118–22.
Jensen, O., & Mazaheri, A. (2010). Shaping functional architecture by oscillatory alpha activity: gating by inhibition. Frontiers in Human Neuroscience, 4(November), 1-8.
Makeig, S., Debener, S., Onton, J., & Delorme, A. (2004). Mining event-related brain dynamics. Trends in Cognitive Sciences, 8(5), 204 – 210.
Makeig, S., Jung, T. P., Bell, A. J., Ghahremani, D., & Sejnowski, T. J. (1997). Blind
separation of auditory event related brain responses into independent components. Proc. Natl. Acad., 94(Sep), 10979–84.
Parks, N. A., Beck, D. M., & Kramer, A. F. (2013). Enhancement and suppression in the visual field under perceptual load. Frontiers in Psychology, 4(May), 1–8.
Pfurtscheller, G., Stancak, A., & Neuper, C. (1996). Event-related synchronization (ERS) in the alpha band - an electrophysiological correlate of cortical idling: A review.
International Journal of Psychophysiology, 24, 39–46.
Romei, V., Gross, J., & Thut, G. (2010). On the role of prestimulus alpha rhythms over
occipito-parietal areas in visual input regulation: correlation or causation? The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 30(25), 8692–7.
Sauseng, P., Klimesch, W., Heise, K. F., Gruber, W. R., Holz, E., Karim, A. A, … Hummel, F. C. (2009). Brain oscillatory substrates of visual short-term memory capacity. Current Biology, 19(21), 1846–52.
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. The Journal of Neuroscience : The Official Journal of the Society for
Neuroscience, 30(11), 4024–32.
Vogel, E. K., & Machizawa, M. G. (2004). Neural activity predicts individual differences in visual working memory capacity. Nature, 428(6984), 748–51.
Vogel, E. K., McCollough, A. W., & Machizawa, M. G. (2005). Neural measures reveal individual differences in controlling access to working memory. Nature, 438(7067), 500–3.
Xu, J., Monterosso, J., Kober, H., Balodis, I. M., & Potenza, M. N. (2011). Perceptual Load-Dependent Neural Correlates of Distractor Interference Inhibition. PloS One, 6(1), 1–9.
Table 1
Number of trials left before and after minimal trial rejection for the three principal parts of behavioral and oscillatory analyses.
Memory load (Cued - Uncued) 2T - 2UD 4T - 4UD 6T - 6UD 2T - 4UD 4T - 2UD 2T2D - 4UD 2T4D - 6UD Before rejection (nr) 77,33 64,75 57,00 77,75 68,00 71,17 65,83 After rejection (nr) 71,75 64,75 57,00 71,83 66,92 68,92 65,25 Average accuracy (%) 94,75 82,00 68,42 95,25 81,42 87,58 81,58
Figure 1. The figure illustrates the task sequence that was used in the actual experiment. Targets (squares) and distracters (circles) were always presented below fixation. Stimuli in the figure are not to scale. The left visual field is cued in this figure. The 250 ms Memorize (and 2000 ms Probe) depicts a change-trial with two cued targets, two cued distracters, and four uncued distracters.
Figure 2. The figure illustrates electrode selections for the two types of EEG analyses performed on the cue-side collapsed data simulating a constant right cue. A: Electrodes used for power analyses plotted on a topoplot of condition averaged induced alpha power activity over the 1000 ms delay period. The dots on the left are contralateral to the cued visual field, and dots on the right are ipsilateral to the cued visual field. B: Electrodes used for alpha power lateralization analyses plotted on a topoplot of condition averaged alpha power lateralization over the 1000 ms delay period. Lower values on the topoplot reflect stronger alpha power suppression for that hemisphere.
Figure 3. Plot depicts interactions associated with varying target and distracter load.
Comparable accuracy rates for conditions with same amount of targets and uncued distracters seem to have the same effect for both target loads. T =Cued targets, D = Uncued Distracters.
Figure 4. Both target load (to the left) and cued distracter load (to the right) had an effect on accuracy rates. However, when comparing the plots it seems that effect of target load exceeds the effect of cued distracters with 4 targets having roughly the same accuracy rate as 4 cued distracters.
Figure 5. Plot illustrating behavioral error rates. Misses increased more than false alarms with increasing target load. FA = false alarm rate, Misses = miss rates, T = cued targets, D = Cued distracters.
Figure 6. Plots of condition average alpha power for conditions with increasing memory load. Topoplots reflect a constant right cue. Dots in the topoplot specify channels that were
averaged and included in the analysis. Alpha power is comparable for all conditions even if the 6T condition shows a relatively large increase of alpha power supression for the ipsilateral hemisphere. Ipsilateral = processing uncued distracters, Contralateral = processing targets . T = cued targets.
Figure 7. Plots of condition average alpha power for conditions with varying memory and distracter load. Topoplots reflect a constant right cue. Topoplots show that alpha power activity over the electrodes included in the analysis (marked with a black dot) is comparable between the four combinations of target and distracter loads. T = Cued targets, UD = Uncued distracters.
Figure 8. Lateralized alpha power differences between memory load of two targets to a memory load of four targets does not seem to predict individual short-term memory capacity. Topoplots of lateralization mark the electrodes that were averaged and compared between the two conditions. Higher positive values in the topoplots reflect stronger ipsilateral alpha suppression. T = Cued targets, UD = Uncued distracters.
Figure 9. Plots of condition average alpha power for conditions with increasing cued distracter load and topoplots of average activity in the 8-12 Hz frequency range for the relevant conditions. Topoplots reflect a constant right cue. Dots in the topoplot specify channels that were averaged and included in the analysis. Alpha power suppression does not seem to increase with cued distracter load. D = Cued distracters.
