The function of visuospatial attention during retention of visual information in short term memory

The function of visuospatial attention during retention of visual information

in short term memory

Tijl van den Bos January 2016

There is an ongoing debate about the functional role of visuospatial attention when information needs to be maintained in visual short-term memory (VSTM). There is evidence that visuospatial attention supports VSTM by focusing on the location where stimuli were previously presented (attention-based rehearsal; Ahw and Vogel, 2006), whereas other evidence suggests that visuospatial attention has no function during retention (Belopolsky and Theeuwes, 2009). In the current study we investigated if visuospatial attention plays a functional role in supporting the maintenance of relevant information in VSTM during retention. We used a custom delayed match to-sample-task deploying targets and distracters. We elicited steady-state visual evoked potentials (SSVEPs) by tagging target and distracter locations with different frequencies allowing us to continuously and simultaneously track attention to these locations using electroencephalography (EEG). We manipulated the feature similarity of the distracters amongst targets to study whether this would affect the focus of attention during retention, and consequently memory performance. Our main analysis provided inconclusive results on the role of visuospatial attention during VSTM retention, and did not show an effect of feature similarity on attentional control during VSTM retention. However, a correlational analysis revealed that when distracters were difficult to discriminate, fast compared to slow responding individuals focused more on the target locations during retention. Therefore, we conclude that maintaining attention on the target location during retention might play a beneficial role for VSTM performance, particularly in demanding situations.

Introduction

Visual representations are subject to decay or interference by newly incoming visual input (Coltheart, 1980). Further processing is needed to consolidate the visual representations in a more durable format, referred to as visual short-term memory (VSTM; Jolicœur and Dell’Acqua, 1998). However, the transition from early visual representations to VSTM causes a substantial reduction of information. VSTM has an average storage limit of 3-4 objects (Cowan, 2001). To make sure relevant representations are held in memory, most models of visual working memory hypothesize a critical role for visuospatial attention. Visuospatial attention can serve as a kind of “gatekeeper” for working memory, by biasing the encoding of information toward the items that are most relevant (Leblanc et al., 2008).

The critical role of attention becomes apparent by studies investigating individual differences in short-term and working memory capacity. Short-short-term and working memory are often used interchangeably but are actually based on two slightly different theoretical backgrounds (Aben et al., 2012). Short-term memory refers

to the ability to maintain small pieces of information temporarily over seconds (Neath et al., 2005). Working memory on the other hand refers to the maintenance and controlled manipulation of a limited amount of information before recall (Baddeley, 1992). Research on attention and VSTM capacity shows that people with a high memory capacity are much more efficient in encoding relevant items in VSTM than people with a low memory capacity, who also encode irrelevant items that are presented (Vogel et al., 2005). Furthermore, the primary determinant of a high VSTM capacity appears to be the ability to suppress presented irrelevant information, instead of just focusing on the relevant information (Zanto and Gazzaley, 2009). Research on attention and visual working memory (VWM) showed similar results. The inability to suppress irrelevant stimuli is related to a lower VWM capacity (Gazzaley et al, 2008). Moreover, people with a high VWM capacity generally tend to suppress upcoming irrelevant information more than people with a lower VWM capacity who only focus on the relevant information, during response inhibition (e.g., Gulbinaite et al., 2014). Taken together, these studies illustrate that individuals that suppress presented irrelevant visual information are better able to encode and store task-relevant information, than individuals that do not suppress irrelevant information.

There is more research that supports the critical role of attention for VSTM capacity, although it propagates a slightly different view on the control of attention when information is encoded into VSTM (Fukuda and Vogel, 2011). In this study the authors suggest that individual differences in memory capacity are not related to the ability to suppress irrelevant stimuli, but to the ability to recover from distraction by irrelevant stimuli (recovery rate). This study indicates that participants with a high VSTM capacity were distracted evenly strong by irrelevant stimuli, but recovered faster from the distraction compared to the participants with a lower VSTM capacity. The proposition of recovery rate as an associable mechanism to memory capacity does not necessarily contradict the proposition of distracter suppression as a determinant for memory capacity. Both theories on the attentional strategies used for the encoding of relevant information support the close relation between visuospatial attention and memory capacity. People that effectively control their attention (whether it is suppressing or recovering from attentional capture by irrelevant information) show to have a higher VSTM capacity than people that do not effectively control their attention, stressing the importance of an efficient attentional filter for cognitive functioning.

An important factor in the control of attention is the detectability of relevant information and the distractibility of irrelevant information (guided search theory; Wolfe et al., 1989; Wolfe, 1994; Wolfe, 1998). When relevant information is distinguishable from irrelevant information by virtue of having a particular combination of shared sensory features (e.g. finding a red ‘T’ between a larger number of blue Ls), it can be detected more rapidly regardless of the number of distracting stimuli (Andersen et al., 2008; Wolfe and Horowitz, 2004). The additive facilitation of the features of the relevant information makes it stand out against the irrelevant information, guiding the deployment of attention (Wolfe and Horowitz, 2004). However, when the relevant and irrelevant information are only different in one feature, they are harder to distinguish from another (Andersen et al., 2008), implying that it is harder to find the relevant information. Consequently, the guided search theory may have implications for the encoding and storing of information. Namely, when the relevant information is harder to detect due to a lack of distinctiveness, the relevant information might not be encoded and stored into memory.

The previously reported research all focused on the critical role of attention during the encoding or selection of information. However, less research has been done on the function of visuospatial attention during maintenance of information in VSTM. Evidence suggests that visuospatial attention has no function during retention of information in VSTM, and that VSTM performance is not dependent on the focus of visuospatial attention (Belopolsky and Theeuwes, 2009). Specifically, if attention is withdrawn from the location previously containing encoded information, this goes without any negative consequences for memory performance (Belopolsky and Theeuwes, 2009, experiment 4), indicating that maintaining attention on the information locations does not improve memory performance (Belopolsky and Theeuwes, 2009, experiment 5). According to Belopolsky & Theeuwes (2009), spatial attention does not have a function during retention, but the mere presence of spatial attention on the target information location is an epiphenomenon of the maintenance of visual information in memory.

However, a substantial amount of literature supports the idea that visuospatial attention supports the maintenance of information in VSTM, by focusing on the spatial location where the encoded visual information was previously presented, during retention (attention-based rehearsal; Ahw and Vogel, 2006). The authors to first report attention-based rehearsal are Smyth & Scholey (1994). In their study, participants were required to learn a series of spatial locations, which had to be recalled after a delay. Memory performance was impaired if the participants saw visual targets or heard a tone during retention, and this impairment was increased if either a motor response or a categorical response was to be made during retention. Smyth & Scholey (1994) suggested that both the serial span memory task and the distracting tasks (e.g. seeing a visual target) required spatial attention, and therefore diverting spatial attention away from the memory task during retention impaired spatial memory performance. The theory of Smyth & Scholey (1994) was later supported by another study on spatial attention during retention of a working memory task (Ahw et al., 1998). In the study of Ahw et al. (1998), participants had to remember a spatial location, and had to recall if this location was the same after a delay. During the delay the participants had to identify the colour of a distracter stimulus presented on a different location, or they had to identify the colour of a stimulus covering all possible stimulus locations. The results showed that only the distracter stimulus on the different location impaired working memory performance. These studies strongly support the functional existence of spatial attention during retention during a working memory task.

More support for the attention-based rehearsal hypothesis comes from a functional magnetic resonance imaging (fMRI) study (Ahw et al., 1999). In this study attention-based rehearsal was shown to lead to enhanced activity in the visual cortex, contralateral to the memorized location. Participants were requested to indicate whether a stimulus was presented on the same location as before a delay (delayed match-to-sample task). The stimulus could be presented on the left or right side off a screen. During the delay, the left and right side of the screen was masked with a grid, rendering the stimulus invisible. The grid evoked enhanced activity in the contralateral side of the visual cortex, compared to the ipsilateral side, indicating that attention was directed to the stimulus location. Another study also showed evidence for the attention-based rehearsal hypothesis by making a direct comparison of event related potential (ERP) modulations observed during two separate

conditions (Ahw et al., 2000). The two conditions contained identical stimulus displays but required participants to perform a spatial attention task and a VSTM task. Similar to the fMRI study (Ahw et al., 1999), in the VSTM task the participants had to perform in a delayed match-to-sample task. In the spatial attention task the participants had to attend the stimulus location during the delay, where the stimulus was presented. During the delay a checkerboard was presented covering the visual field of the to-be-remembered location or the other visual field. Three ERP components (a P1, an anterior N1 and a posterior N1) were elicited by the checkerboard during the delay that were larger in amplitude when the checkerboard was superimposed on the to-be-remembered location than when the checkerboard was presented on the other visual field. These three ERP components were found when attention had to be focused on the stimuli (attention task) and when the stimuli had to be remembered (VSTM task), separately, indicating that attention was focused on the stimuli when the stimuli need to be maintained in VSTM. Because of the similar ERP components, Ahw et al. (2000) suggested that memorizing and attending visuospatial information rely on a common neural mechanism. There is additional evidence for a common neural mechanism from an fMRI study investigating spatial attention and VWM load by means of a dual task (Silk et al., 2010). In this study, participants were requested to do a visual attention search task during the delay of a delayed match-to-sample task. The results showed that increasing the visual attention search load negatively affected working memory performance. In addition, a wide network of prefrontal, premotor, and parietal regions showed increased activity with increased spatial working memory load. Of these areas, activity in part of the right supramarginal gyrus was mediated by both VWM and visual attention search load, indicating that part of the right supramarginal gyrus is associated with VWM and spatial attention. Ahw & Vogel (2006) argued that a common neural mechanism (for spatial attention and VSTM, or spatial attention and VWM) indicates that spatial attention is present during retention of a VSTM or VWM task, but they stress the importance of proving the functional role of spatial attention during retention.

In the current study, we aimed to investigate the functional role of spatial attention during VSTM retention. Hereto, we tracked the focus of visuospatial attention to locations previously containing targets and distracters, during retention in a VSTM task. We assessed the focus of attention during retention of a VSTM task by means of non-invasive scalp recordings of steady state visual-evoked potentials (SSVEPs). SSVEPs are oscillatory potentials that are generated by neurons in the visual cortex in response to flickering stimuli (Regan, 1977) and can be picked up using electroencephalography (EEG; Rager and Singer, 1998). Because the SSVEP oscillates at the same frequency as the driving stimulus, it is possible to record different SSVEPs which are driven by different stimuli flickering at unique frequencies (Reagan 1989). Importantly, the SSVEP signal increases in strength for the driving stimulus that is attended relative to the SSVEP signal of the driving stimulus that is unattended (Müller and Hübner, 2002; Andersen and Müller, 2010). For our study a custom-made visual delayed match-to-sample (VDMS) task was used in which subjects needed to memorize three targets and report after a delay if one of the targets was different. To divert attention away from the targets, three distracters were simultaneously presented with the targets. The targets were placed on placeholders that flickered at one frequency and the distracters were placed on placeholders that flickered at another frequency. Consequently, unique SSVEPs were elicited for the target and distracter locations. By keeping the flickering

placeholders present during the delay period, we were able to track attention to the target and distracter locations during VSTM retention, and thereby investigate the functional role of attention during retention.

Furthermore, we wanted to investigate if a reduced attentional focus on the target locations during retention would deteriorate the maintenance of the target information. To this end, we manipulated the discriminability of the distracters: in one condition the distracters were of the same colour (easy condition) and in the other condition the distracters were in a different colour (difficult condition). According to the guided search theory, we predicted that in the easy condition the feature facilitation would guide the focus of attention to the targets easier than in the difficult condition. In the difficult condition the feature facilitation was predicted to be less effective, making it more difficult to find the targets, which would deteriorate performance since time to encode the stimuli was limited. If we would follow the attention-based rehearsal hypothesis, the difference in separable features would lead to a difference in attention modulation during the retention period: in the easy condition subjects would know where to focus their attention on (target locations). In the difficult condition, subjects might not know where the targets were placed and would not intentionally focus on the target locations. Consequently, decreasing the number of separable features during encoding was predicted to lead to a deterioration of performance on the task since subjects would not know where to focus during retention.

To be able to relate the results of our current study to studies investigating the relationship between attention and working memory, we included a symmetry span task to measure working memory capacity (WMC). It was previously reported that high WMC individuals tend to suppress irrelevant information during encoding more than low WMC individuals (Gazzaley et al, 2008; Gulbinaite et al., 2014). We wanted to investigate if this difference in attention modulation between high and low WMC individuals would also be present during retention of information in VSTM. Hereby, we aim to extend earlier work on the role of attention control, by investigating the allocation, and more specifically the functional role, of attention during VSTM retention.

Materials and methods Participants

We tested 21 participants in the current study. Data from eight participants were excluded: four participants did not maintain their gaze at the fixation cross (fixation violation) on more than 50% of the trials, one performed at chance level and one had already participated in a very similar experiment. Two additional participants were excluded because they did not show an SSVEP response in the stimulated frequencies that exceeded the general noise level (Gulbinaite et al., 2014; Fuchs et al., 2008). Finally, 13 participants were included in the analyses. Participants had normal or corrected-to- normal vision and were not colour blind. Participants did not suffer (and had no first relatives who suffered) from migraine or epilepsy. The study was approved by the local ethics committee of the University of Amsterdam, and informed consent was obtained from all participants.

Apparatus and Stimuli

Stimulus presentation and response registration were controlled using Psychtoolbox (Brainard, 1997) and in Matlab (MathWorks). Stimuli were displayed on a 21-inch monitor (1920 x 1080 pixels; 144 Hz refresh rate). We used eye tracking (Tobii X120; Tobii AB; Stockholm; Sweden) to make sure participants adhered to the instruction to fixate on the fixation cross throughout a trial. The eye tracker was positioned 55 cm in front of the participant (measured from the glabella of the participant to the lens of the eye tracker) on the table top. Participants were seated in a chair in a dimly lit room 90 cm from the monitor. Two buttons were fixed on the right armrest of the chair for participants to respond with.

We used a VDMS task to investigate how visuospatial attention would support the maintenance of visual information in short-term memory. Each trial of the experiment started with a 1000 ms presentation of a fixation cross on which the participant needed to remain their gaze throughout the whole trial. Subsequently, six black squares (placeholders) were added around the fixation cross to elicit the SSVEPs. The placeholders were tagged with a frequency by modulating the brightness of the placeholders by a sinewave (gradually changing the colour from black to white and back). Half of the placeholders were tagged with one frequency and the other half were tagged with another frequency (16 or 18 Hz), according to their future stimulus type. These placeholders remained present on screen throughout the rest of the trial. After 1000 ms, the targets were presented on one half of the placeholders tagged with one frequency. The distracters were presented on the other half of the placeholders tagged with the other frequency. The positioning of the target and distracter stimuli across possible spatial locations was counter balanced across trials, but targets were never placed in one separate hemi field. Targets and distracters were characterized by a unique shape (circles vs. squares, counterbalanced across participants). The targets and distracters were respectively presented as three circles or three squares in seven different colours: red, pink, purple, blue, spring green, cyan and yellow. The distracters were presented in the same colour as each other (easy condition), or were presented each in a different colour (difficult condition). The targets were always presented in a different colour as each other. The proportion of easy and difficult trials was equal. Target and distracter stimuli remained on screen for 500 ms. The participants were instructed to memorize the targets and ignore the distracters. During the following delay period (2000 ms) the targets and distracters were removed from the screen. After the delay period one of the targets (probe) was presented again on one of the placeholders. The participants needed to press the button under their right index finger if the probe was the same as before, or press the button under their right middle finger if the probe had a different colour. The probe was visible until a response was given, or until the response window (1000 ms) had passed. See Figure 1 for an illustration of a trial. The participants were instructed to respond as fast as possible and were motivated to react faster when accuracy levels reached 90%. This was done to avoid ceiling effects in performance and to keep the speed-accuracy trade off constant over the participants.

Working memory capacity was measured with a symmetry span task (Foster et al., 2014; Unsworth et al., 2005). Participants were given a sequence of to-be-remembered locations consisting of red squares in a 4×4 grid of potential locations. Participants also had to complete a distracter task, which was interleaved between

the presentations of the to-be-remembered locations. The distracter task required the participants to judge whether a displayed shape was symmetrical along its vertical axis. We used the partial score of the symmetry span task because it allows for better discrimination between high and low WMC individuals and therefore covers more variance between participants (Conway et al., 2005). The partial span score is the total number of locations recalled in the correct order of a sequence of to-be-remembered locations.

Procedure

Participants were seated in an upright position in the chair before the computer screen on which the stimuli were presented. The participants started with 64 practice trials divided in four blocks. The practice trials were followed by 384 experimental trials divided in twelve blocks. Feedback (mean reaction time, missed trials and accuracy) was provided after each practice and experimental block. Calibration of the eye tracker was performed before the start of the practice trials, before the start of the experimental trials, and was repeated each four experimental blocks, resulting in five calibrations per session. Working memory capacity was measured in a separate session on another day, to avoid fatigue from exertion.

Behavioural data analyses

No-response trials, trials on which the participant responded faster than 200 ms and trials on which participants made an eye movement (defined as a deviation from fixation exceeding 1.5 degrees visual angle) were excluded from the data analysis. Differences in mean reaction time (RT) and accuracy were analysed by means of a repeated measures ANOVA with the factors difficulty (easy and difficult) and target frequency (target on a 16 Hz placeholder or target on an 18 Hz placeholder). The relation between performance on the VDMS task and the WMC task was analysed by estimating the correlation between mean RT and accuracy per condition on the VDMS task and the partial score on the WMC task.

EEG recording and pre-processing

We recorded EEG data using 64 scalp EEG electrodes (Biosemi) positioned according to the international 10-20 system. Two reference electrodes were placed on the earlobes. Vertical and horizontal eye movements were recorded with external electrodes placed on the outer eye canthi and above and below the left eye. The data were recorded using an amplifier (Biosemi) and were sampled at 512 Hz. Offline pre-processing

7

2000 ms 1000 ms

1000 ms 1000 ms

500 ms

Figure 1. Trial design of the delayed match-to-sample task, showing an easy (top) and difficult (bottom) trial. From the placeholders (here presented in black) three were tagged with a 16 Hz frequency and three were tagged with an 18 Hz frequency.

and analysis of the EEG data was done using EEGLAB toolbox for Matlab (http://sccn.ucsd.edu/eeglab/) and custom written Matlab scripts.

The data were offline re-referenced to the average activity recorded at the two earlobes and high-pass filtered at 0.5 Hz. Next, the EEG recordings were epoched (-2200 to 5500 ms after stimulus onset) and baseline corrected to the time window from -1350 to -1050 ms before onset of the SSVEP. By manual inspection of the data, trials with eye blink artefacts and extensive muscle activity in the pre-stimulus, encoding and delay period were removed. Components that were repeatedly present in the data and were not generated by the brain (e.g., eye blinks) were removed by use of an independent component analysis (Delorme and Makeig, 2004). After pre-processing, the average proportion of trials per subject included in the analysis was 0.82 (SD = 0.1). We applied a Laplacian transformation on the pre-processed data (algorithms from Perrin et al., 1989 and implemented in Matlab according to Cohen, 2014) to increase topographical specificity, by filtering out the low spatial frequencies.

Figure 2. Power frequency and topographical plots of the power at 16 and 18 Hz frequency for participant 18. (A) Power frequency plots showing the power at the electrode with the highest power increase (top) and the power at an electrode with little to no power increase (bottom), at the stimulating frequencies. (B) Topographical plots showing the signal-to-noise ratio (SNR) of the power at 16 and 18 Hz frequencies. The red indicates the highest power increase relative to the mean of the neighbouring frequencies. Per example: a SNR of 2.0 means that the power increased by twofold, relative to the mean of the neighbouring frequencies.

Computation of the SSVEP Amplitude

For the analysis of the static SSVEP amplitudes during the delay period, each epoched trial was tapered from 500 to 2500 ms after stimulus onset using a Taylor window, to avoid edge artefacts. The Taylor window had two nearly constant-level side lobes and a peak side lobe level of -30 dB relative to the main lobe peak (Costa and Pacheco, 2011; Carrara et al., 1995; Brookner, 1991). In order to extract SSVEP amplitudes, the power (amplitude squared) for the frequencies of interest was extracted by taking the single closest frequency bin to each stimulated frequency (using a frequency resolution of 0.1). The FFT was calculated for the 16 and 18 Hz frequencies per trial. Subsequently, we averaged the resulting power over all trials in each condition for 16 and 18 Hz separately, for each electrode and each individual subject. Since we compared SSVEP power within the same frequency band across conditions, we did not normalize power and analysed the absolute amplitude of our SSVEP signal.

Electrode selection

For each subject, the electrode with the highest average of the SSVEP amplitudes at the stimulated frequencies, during the delay period (500 to 2500 ms after stimulus onset) was selected for the SSVEP analysis. To this end, the signal-to-noise ratio (SNR) of the power was calculated by dividing the stimulated frequency by the average of four neighbouring frequencies (-0.8, -0.4, +0.4, +0.8 Hz neighbouring the stimulated frequency), for each stimulating frequency. Next, the electrode with the highest SNR - averaged over conditions and the stimulating frequencies - was selected. Figure 2 shows the electrode with the best effect of participant 18 and another electrode not selected for the SSVEP analysis.

Statistical analyses of the effect of attention on the SSVEP

For the statistical analysis of the attention effect we applied a nonparametric permutation test. Our SSVEP data was not normally distributed and therefore could not be analysed with a parametric statistical test (Maris and Oosterveld, 2007). We examined if participants focused their attention on the target locations during VSTM retention, by estimating the difference in power between the target and distracter tagging frequency (collapsed over difficulty levels). Secondly, we examined the effect of distracter similarity on the influence of attention to the distracters during VSTM retention, by estimating the difference in power between the target and distracter tagging frequency, between the easy and difficult condition. Here, we performed nonparametric permutation tests for the 16 Hz and 18 Hz SSVEPs separately. See Figure 3 for an illustration of the analysis and our expected results.

In order to test if attention was focused on the target locations during retention using the nonparametric permutation testing procedure, we carried out three steps (for each frequency separately). First, we ran paired samples t-tests comparing the power difference between the frequency tagging the targets and distracters. Second, a null hypothesis distribution was created by randomly re-assigning the power values, per subject, over the conditions and re-computing the t-values. The second step was repeated a 1000 times.

target distracter 0 50 100 150 200 250 300

A

B

target distracter 0 50 100 150 200 250 300

C

stimulus type in the easy condition

p o w e r (μ V 2 ) target distracter 0 50 100 150 200 250 300

D

stimulus type in the difficult condition

p o w e r (μ V 2 )

Figure 3. Example of the analysis with simulated power values at 16 and 18 Hz according to our expectations. (A) Illustrates the mean power difference between the frequencies tagging the targets and distracters. By subtracting the difference in power (target - distracter) in the difficult condition (D) from the difference in power (target - distracter) in the easy condition (C), the power difference between the easy and difficult condition is acquired (B). * Indicates a simulated significant difference between the power for the tagged stimulus types of the 16 and 18 Hz separately. ** Indicates a simulated significant difference between the power between the conditions for the 16 and 18 Hz separately. We did not expect a significant difference between power for the target and distracter locations tagged by the 16 and 18 Hz separately (n.s.).

Third, the difference in power was statistically significant if the t-value (from step 1) was greater than 95% of the t-values under the null hypothesis (p < 0.05, one tailed). In order to test if the distracters that were different in colour would negatively affect the focus of attention on the target locations during retention, we used the same three steps described above (again for each frequency separately). Only this time we compared the power differences (acquired by subtracting the distracter from the target SSVEPs, per difficulty level) between the easy and difficult condition.

Results

Behavioural performance

To test whether participants would perform better when the distracters had the same colour compared to a different colour, we used a repeated measures ANOVA analysing the difference in reaction time and accuracy between conditions. However, responses were not faster (F(1,12) = 0.389, p = 0,545) and not more accurate (F(1,12) = 0.645, p = 0.438) in the easy condition (RT, m = 567 ms; Acc, m = 76 %) than in the difficult condition (RT, m = 562 ms; ACC, m = 77 %), as is shown in Figure 4. These results indicate that distracter similarity (same vs different colours) had no effect on memory performance.

0 100 200 300 400 500 600 700

A

B

Figure 4. Behavioural results. Bars show the mean reaction time (A) and mean accuracy (B) as a function of condition. The error bars reflect one standard error from the mean. No significant difference was found between the conditions (indicated with n.s.).

The Focus of Attention During Retention

To test whether participants focused their attention on the target location during retention we used a pair of nonparametric tests. Here, we compared the power of the frequencies tagging the targets with the power of the frequencies tagging the distracters, as is illustrated in Figure 5A. However, there was no difference in power at the 16 Hz tagging frequency between the targets (m = 94.922 µV2_{) and the distracters (m = 96.865}

µV2_{) during retention (at p > 0.05). Neither could we find a difference in power at the 18 Hz tagging frequency}

between the targets (m = 66.291 µV2_{) and the distracters (m = 72.140 µV}2_{) during retention (at p > 0.05). To test}

whether distracter similarity affected attentional allocation to the target locations during VSTM retention we ran a second pair of nonparametric permutation tests, testing the difference in SSVEP power between the easy and difficult condition, as is illustrated in Figure 5B. We expected the participants to focus more on the target locations during retention when the distracters were in the same colour than when they were in a different colour. However, the power difference (target vs distracter SSVEP power) for the 16 Hz tagging frequency, was not larger in the easy condition (m = 0.595 µV2_{) than in the difficult condition (m = -4.535 µV}2_{) during retention}

(at p > 0.05). Neither could we find a larger power difference (target vs distracter SSVEP power) for the 18 Hz frequency, in the easy condition (m = -5.338) than in the difficult condition (m = -2.396 µV2_{) during retention (at}

p > 0.05). In summary, our results indicate that the participants did not focus their attention more on the target locations during retention, than the distracter locations. Furthermore, the results indicate that the focus of the attention during retention was not affected by distracter similarity (same vs different colours).

16 Hz 18 Hz 0 20 40 60 80 100 120

A

B

Figure 5. The focus of attention represented in power. (A) Static SSVEP power for 16 and 18 Hz tagging frequencies, separate for the targets and distracters. (B) Differences in SSVEP power between frequencies tagging the targets and distracters, for the easy and difficult condition. E.g. in the easy condition the difference in power at the 16 Hz frequency is very small. However, the power at the 18 Hz frequency in the easy condition is negative. A negative power difference indicates that the power for the distracters was higher, implying that, on average, the participants focused more on the distracter than the target locations. Error bars represent one standard error from the mean. N.s. indicates that there was no significant difference, analysed for each frequency separately.

To estimate if there was a functional relationship between the focus of attention and behavioural performance, we ran an additional analysis. We were interested to see if participants who paid more attention to the target locations during retention had a shorter reaction time and/ or higher accuracy than the participants who did not pay attention to target locations. We ran a cross-subject correlation to estimate the association between the power differences during retention, between the SSVEP signal at the target and distracter locations, reaction time and accuracy, for the easy and difficult condition. Because the differences in power between the target and distracter tagging frequencies were not normally distributed we used Spearman’s rank-order correlation (Bischara and Hittner, 2015). Here, we observed that the power difference between the frequencies tagging the target and distracter locations during retention in the difficult condition correlated negatively with reaction time in the difficult condition, when a tagging frequency of 18 Hz was used (rs(11) = -.571, p = 0,041). Thus, across participants we observed that a bigger positive power difference

between the frequency tagging the targets and distracters correlated with faster reaction times, indicating that participants that had a stronger focus of attention on the target locations during retention of visual information, also reacted faster, as is shown in Figure 6A. However, the association between reaction time and power was only found when targets and distracters were difficult to discriminate, and an 18 Hz tagging frequency was used. We did not observe any correlation between accuracy and power, which indicates that the

focus of attention during retention does not related to the accuracy with which visual information can be retrieved from short-term memory.

The relationship between WMC, reaction time, and the control of attention during VSTM

We aimed to investigate if individual differences in the capability to suppress irrelevant information between high and low WMC individuals (Gulbinaite et al., 2014; Gazzaley et al, 2008), would also be present during retention in a VSTM task. First, we wanted to see if participants who scored higher on the WMC task reacted faster and were more accurate in retrieving the target information during our VDMS task, than participants who scored lower on the WMC task. To this end we estimated a correlation between reaction time, accuracy and WMC of the participants, by using the Pearson correlation coefficient. The partial scores on the WMC task correlated positively with reaction time averaged over conditions (r(11) = -,782, p = 0,002. Figure 6B). The partial score on the WMC task also correlated with reaction time in the easy (r(11) = -,756, p = 0,003) and difficult (r(11) = -,773, p = 0,002) condition. When a participant had a longer reaction time, the participant had a lower WMC. These results indicate that participants who are able to retrieve visual information from VSTM more quickly, also have a high working memory capacity as measured on the symmetry span task.

Second, we wanted to see if participants with a higher WMC paid more attention to the target locations, during retention in the VDMS task, than participants with a lower WMC. To this end, we correlated the power differences between the SSVEP signals at the target and distracter locations during retention in the easy and difficult condition, with the partial score on the symmetry span task. We found that the power difference (SSVEP power for targets vs. distracters) in the difficult condition correlated with the partial WMC score (rs(11)

= .698, p = 0,008. Figure 6C), when a tagging frequency of 18 Hz was used. Across participants we observed that a bigger positive power difference between the frequency tagging the targets and distracters correlated with higher WMC scores, indicating, that participants with a higher compared to a lower WMC directed more attention to the target than distracter locations during retention when distracters were different in colour. However, this was only the case when a tagging frequency of 18 Hz, but not 16 Hz, was used. No correlation

-20 -15 -10 -5 0 5 10 15 400 450 500 550 600 650 700 750

A

power diference (target - distracter) at 18Hz (μV2)

R e ac ti o n T im e 400 450 500 550 600 650 700 750 0 5 10 15 20 25 30 35 40 45

B

C

power diference (target - distracter) at 18 Hz (μV2)

W M p ar ti al s co re

Figure 6. cross-subject correlations between reaction time, the focus of attention and working memory capacity. (A) Across participants we observed that a stronger attention to the target locations, during retention, was related to the ability to retrieve the target information faster, when distracters were presented in different colours. Stronger attention to the target locations was reflected as a positive power difference between the frequencies tagging the target and distracter locations (also applies to C). Stronger attention to the distracter locations was related to a slower retrieval time of the target information, during retention. The higher level of focus on the target locations was reflected as positive power difference between the frequencies tagging the target and distracter locations (also applies to C). This association was only significant (indicated with a ‘*’) when targets where tagged with a 18 Hz frequency and distracters were different in colour. (B) A higher WMC was related to a shorter reaction time. Visa versa, a lower WMC was related to a longer reaction time. Shown here

is the average reaction time over conditions. (C) A higher WMC was related to stronger attention to the target locations. A lower WMC was related to stronger attention to the distracter locations.

was found between accuracy and WMC, indicating that the accuracy with which participants can retrieve information from VSTM was not related to their WMC. Time needed to retrieve information from VSTM and respond to the probe was related to WMC.

Discussion

In the current study we investigated the focus of attention when visual information needed to be maintained in VSTM. To this end, we used a VDMS task simultaneously presenting targets and distracters, and tracked attention to the target and distracter locations by eliciting SSVEPs for each stimulus location. Furthermore, we manipulated feature similarity of the distracters to investigate if this would affect the focus of attention during retention, and memory performance. According the attention-based rehearsal hypothesis (Ahw and Vogel, 2006) we expected the participants to focus their attention on the locations previously holding the relevant information, during retention. Additionally, we expected participants to have a stronger focus of attention on the locations previously holding the relevant versus the irrelevant information during VTSM retention, during retention, when the relevant and irrelevant information was distinguishable on one feature versus a conjunction of features, as proposed by the guided search theory (Wolfe et al., 1989; Wolfe, 1994; Wolfe, 1998). Consequently, we expected participants to respond faster and more accurate to the probe when they had a stronger focus of attention on the locations previously holding the relevant information, during VTSM retention. In contrast to our expectations, analyses of the SSVEP data did not reveal a difference in the focus of attention on the locations previously holding the relevant and irrelevant information, indicating that attention was focused on the irrelevant information locations as much as on the relevant information locations , during retention. In addition, participants responded equally fast and accurate when the relevant and irrelevant information was distinguishable on one versus a conjunction of features. With an exploratory analysis we did find that participants that responded faster, focused more on the locations previously holding the relevant than the irrelevant information, during retention, in situations where relevant and irrelevant stimuli were distinguishable on only one feature. However, this association was only found between reaction time and power at the 18 Hz frequency. We conclude that visuospatial attention during retention might play a functional role since it might be beneficial to attend the locations previously holding the relevant information to improve VSTM performance.

First, we examined whether the focus of attention during retention in a VSTM task supports VSTM maintenance as proposed by the attention-based rehearsal theory (Awh and Vogel, 2006). We compared the focus of attention during retention on locations previously containing the relevant versus the irrelevant information. Contrary to our expectations, our results indicate that attention was not focused more strongly on the relevant than the irrelevant information locations, during retention. This indicates that it was unnecessary to have a stronger focus on the relevant locations during retention to maintain a representation held in VSTM. Thus, our findings remain somewhat inconclusive. We cannot support earlier evidence demonstrating that not attending the locations previously containing the relevant information deteriorates memory performance (Ahw

et al., 1998). However, we also cannot conclude that VSTM does not benefit from attention-based rehearsal at all since the locations holding the relevant information were attended during retention. The possibility remains that attending all the information locations (relevant as well as irrelevant) was enough to support the maintenance of the relevant information in VSTM, and that additionally attending the irrelevant information locations during retention might not have interfered with memory performance.

Second, we investigated the effect of feature similarity on attention when relevant information needed to be maintained in VSTM. We predicted that it would be more difficult to detect the relevant information when relevant and irrelevant information were only distinguishable on a single feature compared to a conjunction of features (guided search theory; Wolfe et al., 1989; Wolfe, 1994; Wolfe, 1998). Consequently, in the condition with less discriminable distracters, participants might not have been able to focus their attention intentionally on the locations previously holding the relevant information to support VSTM, during retention. However, our results show that attention was focused on the irrelevant locations as much as on the relevant locations during retention, in spite of the manipulation of distracter similarity. Likewise, memory performance was unaffected by the manipulation of distracter similarity. Thus, our results show that manipulating the feature similarity of the distracters did not affect the focus of attention during retention or memory performance. This might indicate that the benefit of having a conjunction of features to guide the deployment of spatial attention (Wolfe and Horowitz, 2004) does not aid the maintenance and/or retrieval of the memory representation in VSTM. These findings are in line with results from a study on the role of feature binding for VSTM showing that VSTM operates according to different principles than visual attention (Logie et al., 2010). In their study, Logie et al. (2010) used a delayed match-to-sample task to test if VSTM performance was affected by feature-binding of irrelevant stimulus features. Hereto, they examined whether VSTM was affected if the task-irrelevant features of memorized stimuli were changed on recall. They found that after a delay of 200 ms, this change negatively affected memory performance. What is more interesting, is that they found that changing the delay to 1500 ms negated this negative effect. Logie et al. (2010) argued that the influence of task-irrelevant features on attention decay over time, and that VSTM performance is more reliant on bindings formed only between task-relevant features in VSTM. The conclusion of Logie et al. (2010) might also apply to our results. Since we used a retention interval of 2000 ms, task-irrelevant feature bindings might not have influenced VSTM performance. As a result, visuospatial attention might have been affected by the manipulation of feature similarity when information was first presented, but not when information needed to be maintained. The distracting, task-irrelevant feature binding (e.g. the colour and location of task-irrelevant information) might have decayed over the retention period. Consequently, whether the irrelevant information was presented in the same or a different colour did not affect attention during VSTM retention, since feature bindings that were task-irrelevant might have decayed during the retention interval used in our task.

Interestingly, our exploratory analysis showed an association between reaction time and the focus of attention during retention reflected as a difference in the SSVEP signal elicited by the relevant and the irrelevant information. We observed that the participants that responded faster also paid more attention to the locations previously containing the relevant information during retention, compared to the participants that

responded slower. This association was only present when irrelevant information was hard to distinguish from the relevant information, and a 18 Hz tagging frequency was used. The lack of an association when the irrelevant information was presented in the same colour might be caused by the difficulty of the task. When same coloured irrelevant information was presented it might not have been necessary to attend the relevant locations during retention to make a fast response. However when differently coloured irrelevant information was presented, attention to the relevant locations during retention might have been needed to make a fast response. The association we found therefore might support the attention-based rehearsal hypothesis in such a way that increased task difficulty makes it beneficial to focus attention on the relevant locations during retention to support maintenance of visuospatial information, during a VSTM task. The association we found between power in the 18 Hz frequency and WMC of the individuals supports this idea: participants with a higher compared to a lower WMC focused more on the relevant, and less on the irrelevant locations during VSTM retention, when relevant and irrelevant stimuli were hard to distinguish from another. This result indicates that participants with a higher WMC used their attention more efficiently by attending the relevant information locations during retention when task difficulty was increased. In addition, we found an association between reaction time and WMC, when not taking task difficulty into account, indicating that high WMC participants also reacted faster. Taken together, our results might imply that although task-irrelevant feature bindings are not stored in VSTM after a long delay (Logie et al., 2010), it might still be beneficial to attend the locations where relevant information was previously presented during retention when task difficulty is increased.

It is necessary to extend our sample size to validate the discrepancy between the findings from our main analysis and the findings from our exploratory analysis. Currently, the results from our main analysis do not directly support a functional role for attention-based rehearsal, while our exploratory analysis does. By extending our sample size we can group individuals together based on their behavioural performance and compare the focus of attention during retention between the groups (cf. Gulbinaite et al., 2014). Currently, our sample size does not allow for grouping of participants. Furthermore, with an increased sample size we can establish if the association between reaction time and power is not sole dependent on the 18 Hz frequency, but is a substantial finding. Studies investigating attention by eliciting SSVEPs using tagging frequencies very close to the tagging frequencies we used in the current study, reported that the effect of attention on the modulation of SSVEPs is independent of the tagging frequency used (Kim et al., 2007; Kelly et al., 2006; Toffanin et al., 2009). Accordingly, it will be important to establish whether we can replicate these findings by showing an attention effect on both tagging frequencies used in the current study. Therefore, increasing the sample size to validate our current results is essential.

If increasing our sample size will not give is more clear results, we should also investigate to what extent attending the locations previously holding the task-irrelevant information, during retention, can interfere with VSTM performance. As was earlier stated, our main analysis indicated that attention was spread evenly over the locations previously containing the relevant and irrelevant information, during retention. With these results we cannot conclude that visuospatial attention has no effect, or has an effect on maintenance in VSTM during

retention, since attending the irrelevant as well as the relevant information locations might still have aided VSTM to some extent. Therefore, it is important to investigate if attending the irrelevant together with the relevant information locations during retention also boosts memory performance, compared to diverting attention completely away from all stimulus locations during retention. To this end, it is again required to track the focus of attention with EEG by eliciting unique SSVEPs for the relevant and irrelevant stimulus locations during retention. If attending the relevant as well as the irrelevant information locations during retention still improves memory performance compared to not attending any information locations, we can conclude that VSTM really has benefit from attention-based rehearsal during retention, and that visuospatial attention towards task-irrelevant locations does not necessarily interfere with the maintenance of information. Furthermore, since Logie et al. (2010) proposed that task-irrelevant feature bindings are not stored in VSTM with a long retention period, it might be interesting to investigate if this is the case in our current study. To this end, we could use the same current task but change the shape of the probe on recall. In addition, we should have a short and a long retention period. According to Logie et al. (2010), we should find a deterioration in memory performance with the short retention period because the different shape interferes with the retrieval of the information. Additionally, if we find that memory performance is not affected when the longer retention period is used, we can conclude that, in the current study, the lack of a difference in the focus of attention on the locations holding the relevant versus the irrelevant information was due to the decay of irrelevant feature bindings.

In summary, we investigated the function of visuospatial attention during retention of information in VSTM. Our results showed no difference in the focus of attention on locations previously holding the relevant versus the irrelevant information, in spite of manipulating the feature similarity of the task-irrelevant information. From these results we cannot conclude that visuospatial attention was necessary or unnecessary to support the maintenance of information in VSTM. However, when we related the SSVEP signal during VSTM retention to the response time, we did find that fast and slow responders during our VDMS task, focused their attention on different locations during retention. This association may indicate that attention to previous stimulus locations helps to retain visual information in VSTM when the information has already been selected and encoded. Thereby, our study extends previous work on the role of attentional control by revealing a subtle, but functional role of attention during retention of information in VSTM.

References

Aben, B., Stapert, S., & Blokland, A. (2012). About the distinction between working memory and short-term memory. Frontiers in psychology, 3.

Andersen, S. K., Fuchs, S., & Müller, M. M. (2011). Effects of feature-selective and spatial attention at different stages of visual processing. Journal of Cognitive Neuroscience, 23(1), 238-246.

Andersen, S. K., Hillyard, S. A., & Müller, M. M. (2008). Attention facilitates multiple stimulus features in parallel in human visual cortex. Current Biology, 18(13), 1006-1009.

Andersen, S. K., & Müller, M. M. (2010). Behavioral performance follows the time course of neural facilitation and suppression during cued shifts of feature-selective attention. Proceedings of the National Academy of Sciences, 107, 13878-13882.

Andersen, S. K., Müller, M. M., & Martinovic, J. (2012). Bottom-up biases in feature-selective attention. The Journal of Neuroscience, 32, 16953-16958.

Awh, E., Anllo-Vento, L., & Hillyard, S. (2000). The role of spatial selective attention in working memory for locations: evidence from event-related potentials. Journal of Cognitive Neuroscience, 12(5), 840-847. Awh, E., Jonides, J., Reuter-Lorenz, P.A. (1998). Rehearsal in spatial working memory. Journal of Experimental

Psychology, 24(3), 780-790.

Awh, E., Jonides, J., Smith, E. E., Buxton, R. B., Frank, L. R., Love, T., Wong, E.C., & Gmeindl, L. (1999). Rehearsal in spatial working memory: Evidence from neuroimaging. Psychological Science, 10(5), 433-437.

Awh, E., & Vogel, E. K. (2008). The bouncer in the brain. Nature neuroscience, 11(1), 5-6.

Awh, E., Vogel, E. K., & Oh, S. H. (2006). Interactions between attention and working memory. Neuroscience, 139(1), 201-208.

Baddeley, A. (1992). Working memory. Science, 255, 556–559

Belopolsky, A. V., & Theeuwes, J. (2009). No functional role of attention-based rehearsal in maintenance of spatial working memory representations. Acta psychologica, 132(2), 124-135.

Bishara, A. J., & Hittner, J. B. (2015). Reducing Bias and Error in the Correlation Coefficient Due to Nonnormality. Educational and Psychological Measurement, 75(5), 785-804.

Brainard, D.H., 1997. The Psychophysics Toolbox. Spatial Vision. 10, 433–436.

Cohen, M.X., 2014. Analyzing Neural Time Series Data: Theory and Practice. Cambridge, MA: MIT Press. Coltheart, M. (1980). Iconic memory and visible persistence. Perception & psychophysics, 27(3), 183-228.

Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–185.

Delorme, A., Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods, 134(1), 9-21.

Dell’Acqua, R., Sessa, P., Toffanin, P., Luria, R., & Jolicœur, P. (2010). Orienting attention to objects in visual short-term memory. Neuropsychologia, 48(2), 419-428.

Foster, J. L., Shipstead, Z., Harrison, T. L., Hicks, K. L., Redick, T. S., & Engle, R. W. (2014). Shortened complex span tasks can reliably measure working memory capacity. Memory & cognition, 43(2), 226-236.

Fuchs, S., Andersen, S. K., Gruber, T., & Müller, M. M. (2008). Attentional bias of competitive interactions in neuronal networks of early visual processing in the human brain. NeuroImage, 41(3), 1086-1101.

Fukuda, K., & Vogel, E. K. (2009). Human variation in overriding attentional capture. The Journal of Neuroscience, 29, 8726-8733.

Fukuda, K., & Vogel, E. K. (2011). Individual differences in recovery time from attentional capture. Psychological Science, 22(3), 361-368.

Gazzaley, A., Clapp, W., Kelley, J., McEvoy, K., Knight, R. T., & D'Esposito, M. (2008). Age-related top-down suppression deficit in the early stages of cortical visual memory processing. Proceedings of the National Academy of Sciences, 105, 13122-13126.

Gulbinaite, R., Johnson, A., de Jong, R., Morey, C. C., & van Rijn, H. (2014). Dissociable mechanisms underlying individual differences in visual working memory capacity. Neuroimage, 99, 197-206.

Jha, A.P. (2002). Tracking the time-course of attentional involvement in spatial working memory: an event-related potential investigation. Cognitive Brain Research, 15, 61-69.

Jolicœur, P., & Dell'Acqua, R. (1998). The demonstration of short-term consolidation. Cognitive psychology, 36(2), 138-202.

Kelly, S.P., Lalor, E.C., Reilly, R.B., Foxe, J.J., Jun 2006. Increases in alpha oscillatory power reflect an active retinotopic mechanism for distracter suppression during sustained visuospatial attention. Journal of Neurophysiology, 95, 3844–3851.

Kim, Y.J., Grabowecky, M., Paller, K.A., Muthu, K., Suzuki, S., 2007. Attention induces synchronization-based response gain in steady-state visual evoked potentials. Nature Neuroscience, 10, 117–125.

Leblanc, É., Prime, D. J., & Jolicoeur, P. (2008). Tracking the location of visuospatial attention in a contingent capture paradigm. Journal of Cognitive Neuroscience, 20(4), 657-671.

Lepsien, J., & Nobre, A. C. (2006). Cognitive control of attention in the human brain: Insights from orienting attention to mental representations. Brain research, 1105(1), 20-31.

Lepsien, J., Thornton, I., & Nobre, A. C. (2011). Modulation of working-memory maintenance by directed attention. Neuropsychologia, 49(6), 1569-1577.

Logie, R. H., Brockmole, J. R., & Jaswal, S. (2011). Feature binding in visual short-term memory is unaffected by task-irrelevant changes of location, shape, and color. Memory & cognition, 39(1), 24-36.

Maris, E., & Oostenveld, R. (2007). Nonparametric statistical testing of EEG-and MEG-data. Journal of neuroscience methods, 164(1), 177-190.

Müller, M. M., & Hübner, R. (2002). Can the spotlight of attention be shaped like a doughnut? Evidence from steady-state visual evoked potentials. Psychological Science, 13(2), 119-124.

Müller, M. M., Malinowski, P., Gruber, T., & Hillyard, S. A. (2003). Sustained division of the attentional spotlight. Nature, 424, 309-312.

Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental psychology, 49(2), 270.

Neath, I., Brown, G. D., Poirier, M., and Fortin, C. (2005). Short-term and working memory: past, progress, and prospects. Memory 13, 225–235.

Postle, B.R., Ahw, E., Jonides, J., Smith, E.E., & Esposito, M.D. (2004). The where and how of attention-based rehearsal in spatial working memory. Cognitive Brain Research, 20, 194-205.

Regan, D. (1977). Steady-state evoked potentials. Journal of the Optical Society of America, 67(11), 1475-1489. Reagan, D. (1989). Human brain electrophysiology. Evoked Potentials and Evoked Magnetic Fields in Science and

Medicine. New York: Elsevier.

Rager, G., & Singer, W. (1998). The response of cat visual cortex to flicker stimuli of variable frequency. European Journal of Neuroscience, 10(5), 1856-1877.

Silk, T. J., Bellgrove, M. A., Wrafter, P., Mattingley, J. B., & Cunnington, R. (2010). Spatial working memory and spatial attention rely on common neural processes in the intraparietal sulcus. Neuroimage, 53(2), 718-724.

Sternberg, S. (1966). High-speed scanning in human memory. Science, 153, 652-654.

Smyth, M., & Scholey, K.A. (1994). Interference in immediate spatial memory. Memory & Cognition, 22(1), 1-13. Theeuwes, J., Belopolsky, A., & Olivers, C. N. (2009). Interactions between working memory, attention and eye

movements. Acta psychologica, 132(2), 106-114.

Toffanin, P., de Jong, R., Johnson, A., & Martens, S. (2009). Using frequency tagging to quantify attentional deployment in a visual divided attention task. International Journal of Psychophysiology, 72(3), 289-298. Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W. (2005). An automated version of the operation span task.

Behavior research methods, 37(3), 498-505.

Vogel, E. K., McCollough, A. W., & Machizawa, M. G. (2005). Neural measures reveal individual differences in controlling access to working memory. Nature, 438, 500-503.

Waters, A. M., Farrell, L. J., Zimmer-Gembeck, M. J., Milliner, E., Tiralongo, E., Donovan, C. L., ... & Ollendick, T. H. (2014). Augmenting one-session treatment of children's specific phobias with attention training to positive stimuli. Behaviour research and therapy, 62, 107-119.

Wolfe, J. M. (1994). Guided search 2.0 a revised model of visual search. Psychonomic bulletin & review, 1(2), 202-238.

Wolfe, J. M., Cave, K. R., & Franzel, S. L. (1989). Guided search: an alternative to the feature integration model for visual search. Journal of Experimental Psychology: Human perception and performance, 15(3), 419.

Wolfe, J. M., & Horowitz, T. S. (2004). What attributes guide the deployment of visual attention and how do they do it? Nature Reviews Neuroscience, 5(6), 495-501.

Zanto, T. P., & Gazzaley, A. (2009). Neural suppression of irrelevant information underlies optimal working memory performance. The Journal of Neuroscience, 29, 3059-3066.