Dissociating Distractor-Filtering at Encoding and During Maintenance

Abstract

The effectiveness of distractor-filtering is a potentially important determinant of working memory capacity (WMC). However, a distinction between the contributions of distractor-filtering at WM encoding as opposed to filtering during maintenance has not been made and the assumption is that these rely on the same mechanism. Within 2 experiments, 1 conducted in the laboratory with 21 participants, and the other played as a game on smartphones (n = 3,247) we measure WMC without distractors, and present distractors during encoding or during the delay period of a WM task to determine performance associated with distraction at encoding and during maintenance. Despite differences in experimental setting and paradigm design between the 2 studies, we show a unique contribution to WMC from both encoding and delay distractor performance in both experiments, while controlling for performance in the absence of distraction. Thus, within 2 separate experiments, 1 involving an extremely large cohort of 3,247 participants, we show a dissociation between encoding and delay distractor-filtering, indicating that separate mechanisms may contribute to WMC. (PsycINFO Database Record (c) 2014 APA, all rights reserved).

Full text

Dissociating Distractor-Filtering at Encoding and During Maintenance
Fiona McNab
University of Birmingham, United Kingdom and University
College London, United Kingdom
Raymond J. Dolan
University College London, United Kingdom
The effectiveness of distractor-filtering is a potentially important determinant of working memory
capacity (WMC). However, a distinction between the contributions of distractor-filtering at WM
encoding as opposed to filtering during maintenance has not been made and the assumption is that these
rely on the same mechanism. Within 2 experiments, 1 conducted in the laboratory with 21 participants,
and the other played as a game on smartphones (n⫽3,247) we measure WMC without distractors, and
present distractors during encoding or during the delay period of a WM task to determine performance
associated with distraction at encoding and during maintenance. Despite differences in experimental
setting and paradigm design between the 2 studies, we show a unique contribution to WMC from both
encoding and delay distractor performance in both experiments, while controlling for performance in the
absence of distraction. Thus, within 2 separate experiments, 1 involving an extremely large cohort of
3,247 participants, we show a dissociation between encoding and delay distractor-filtering, indicating that
separate mechanisms may contribute to WMC.
Keywords: distraction, distractor filtering, working memory, working memory capacity
Working memory (WM) is important for a wide range of cog-
nitive functions such as reasoning and language (Sü␤, Oberauer,
Wittmann, Wilhelm & Schulze, 2002;Baddeley, 2003). WM ca-
pacity (WMC) is reduced in many psychiatric and neurological
disorders (Goldman-Rakic, 1994;Castellanos & Tannock, 2002)
as well as with normal aging (Bopp & Verhaeghen, 2005). A
possible functional architecture that determines the efficiency of
WMC relates to filtering, which enables the discarding of irrele-
vant and storage of relevant information. It has been reported that
storage-related parietal activity for distractors negatively correlates
with WMC, consistent with the idea that participants with low
WMC unnecessarily store distractors, whereas participants with
high WMC filter them effectively and remember relevant infor-
mation alone (Vogel, McCollough & Machizawa, 2005;McNab &
Klingberg, 2008).
In studies of WM, distractors are sometimes presented during
WM encoding (simultaneous with presentation of to be remem-
bered stimuli), whereas in others distractors are shown during the
delay period (when the WM stimuli are held in mind but no longer
displayed). Previous work has shown that WM performance im-
proves with sequential compared with simultaneous stimulus pre-
sentation, with the latter predicting improvement (Ihssen, Linden
& Shapiro, 2010). Here we ask whether the efficiency of distractor
filtering during the simultaneous presentation of targets and dis-
tractors differs from the efficiency of distractor filtering during the
sequential presentation of targets and distractors, and whether they
differ in terms of their contribution to WMC.
Distractor filtering processes at encoding and during delay have
sometimes been described differently, for example the former seen
as requiring a “selective gating mechanism” (Frank et al., 2001)
whereas maintaining information in WM in the face of delay
distraction is described as a process in which “the doors of per-
ception close” (Bonnefond & Jensen, 2012). However, to the best
of our knowledge a distinction between the two has not been
experimentally probed and in some cases it has been assumed that
encoding and delay distractor filtering rely on the same neural
mechanism. For example, it has been argued that enhanced resis-
tance to distraction seen in patients deficient in striatal dopamine
(with distractors presented during the delay period) is not consis-
tent with reports of striatal dopamine playing an important role in
This article was published Online First February 10, 2014.
Fiona McNab, School of Psychology, University of Birmingham, Bir-
mingham, United Kingdom and Wellcome Trust Centre for Neuroimaging,
University College London, London, United Kingdom; Raymond J. Dolan,
Wellcome Trust Centre for Neuroimaging, University College London,
London, United Kingdom.
We thank Emily De Leon and Marc Guitart Masip for their help with
participant recruitment and data collection for the laboratory study, Kimron
Shapiro and Howard Bowman for helpful discussion, and Peter Zeidman,
Richard Adams, Harriet Brown, Robb Rutledge, Peter Smittenaar, Neil
Millstone from White Bat Games, and The Wellcome Trust for making
“The Great Brain Experiment” possible. This work was supported by the
Wellcome Trust Research Career Development Fellowship 091826/Z/10/Z
(F.M.), The Wellcome Trust Centre for Neuroimaging is supported by core
funding from the Wellcome Trust 091593/Z/10/Z. R.J.D. holds a Well-
come Trust Senior Investigator Award (098362/Z/12/Z).
This article has been published under the terms of the Creative Com-
mons Attribution License (http://creativecommons.org/licenses/by/3.0/),
which permits unrestricted use, distribution, and reproduction in any me-
dium, provided the original author and source are credited. Copyright for
this article is retained by the author(s). Author(s) grant(s) the American
Psychological Association the exclusive right to publish the article and
identify itself as the original publisher.
Correspondence concerning this article should be addressed to Fiona
McNab, School of Psychology, University of Birmingham, Edgbaston,
Birmingham, B15 2TT, United Kingdom. E-mail: F.McNab@bham.ac.uk
Journal of Experimental Psychology:
Human Perception and Performance © 2014 The Author(s)
2014, Vol. 40, No. 3, 960–967 0096-1523/14/$12.00 DOI: 10.1037/a0036013
960

distractor filtering (referring to research in which distractors are
presented during encoding, e.g., McNab & Klingberg, 2008;Cools
et al., 2010). An involvement of striatal dopamine in encoding
but not delay distractor filtering could account for such appar-
ent inconsistency. Within this study we directly compare en-
coding and delay distractor filtering to challenge an assumption
that the two processes are indistinct, and highlight a need to
draw a distinction between encoding and delay distractor fil-
tering in future research.
With two separate experiments we directly compare the contri-
bution of encoding and delay distractor filtering to WMC to
determine whether they each make a unique contribution to WMC.
Such a finding would not only distinguish the two processes, and
suggest that separate neural mechanisms may be involved, but also
indicate that they represent two potential bases for WMC. Al-
though their contribution to WMC has not been directly compared,
research suggests that both encoding (e.g., Vogel et al., 2005;
McNab & Klingberg, 2008) and delay distractor filtering predict
WM performance. For example, participants with high WMC are
reported to show greater accuracy for trials that include delay
distractors (pictures of faces) compared with those with low WMC
(Minamoto, Osaka & Osaka, 2010). Furthermore, greater activity
seen in the fusiform gyrus in response to face distractors in
participants with low compared with high WMC suggests distrac-
tors were not filtered as effectively.
Our two experiments differed in a number of substantive ways.
For the laboratory experiment measures of WMC and distractor
cost were obtained by presenting stimuli of a fixed set-size and
estimating the number of items successfully remembered using the
hit rate and false alarm rate (the K value; Vogel et al., 2005;
Cowan, 2001). In the smartphone game the number of items to
remember (WM load) in each condition increased in line with
performance until the participant failed two successive trials. Per-
formance was determined by the maximum load the participant
successfully completed. The smartphone experiment also involved
a much larger sample than the laboratory study, but time spent on
the task had to be kept to a minimum and the environment in which the
game was played could not be controlled. We hypothesized that,
despite these differences, in both experiments distractor filtering at
encoding and at delay would uniquely predict WMC.
Laboratory Study
Method
Participants. Twenty-three healthy participants gave in-
formed consent to participate in the laboratory study, which was
approved by the University College London Ethics Committee.
Two participants failed to complete the study, leaving 21 (8
females, ages 20–29, right-handed).
Experimental design and task. Data were collected during
the placebo condition of a drug study, the results of which are not
reported here. Participants performed the experiment twice, 7 days
apart. On one day they received levodopa (150 mg ⫹37.5 mg
benserazide), and on the other day they received placebo (500 mg
ascorbic acid). The order was counterbalanced such that 11 re-
ceived levodopa on their first visit and 10 received levodopa on
their second visit. The study was double-blind. Participants began
the WM tasks approximately 2 hours and 15 minutes after receiv-
ing either levodopa or placebo, and after completing a MEG study
of decision-making, the results of which are not reported here.
Only data from the placebo condition are considered here, and as
we used a repeated measures design we consider that differences in
previous exposure to the task could not have affected the results.
The experiment consisted of the following computer-based WM
tasks. The experiment was divided into three sections, with the
order of the three sections counterbalanced across participants. The
three sections are described below.
Working memory capacity (WMC). To obtain a measure of
WMC participants were asked to remember the positions of five
red circles (targets) presented simultaneously on a circular grid
(Figure 1a).
Distractor-filtering at encoding and during the delay.
Measures of accuracy with and without distraction were obtained
from a task in which 3 red circles (target stimuli) were presented
simultaneously. In the “no distraction” condition (Figure 1b)no
yellow distractor circles were shown. In the “encoding distraction”
condition (Figure 1c) two yellow distractor circles were presented
together with the three target circles. In the “delay distraction”
condition (Figure 1d) two yellow distractor circles were presented
after the three target circles had disappeared. Participants were
asked to remember the positions of the red target circles, ignore the
yellow distractor circles, and indicate with a button press whether
the probe was in one of the target positions. Trials were presented
in two blocks, with 60 trials in each. The trials in each block were
divided equally between the three conditions, and presented in a
pseudorandom order. Half of the trials in each condition required
a “yes” response, and in both the encoding and delay distraction
conditions, half of the trials that required a “no” response had the
probe presented in the position of a distractor.
Forward and backward span tasks. Participants performed
one backward span task and two forward span tasks, as described
previously (McNab et al., 2009), the results of which are not
reported here.
Data analysis. WMC was estimated with the Kvalue, esti-
mating how much information can be stored in working memory,
using a standard formula (Vogel et al., 2005;Cowan, 2001); K⫽
S(H⫺F), where Kis the WMC, Sis the array size, His the
observed hit rate, and Fis the false alarm rate. This uses the false
alarm rate to correct for guessing and assumes that if Kitems can
be held in WM, from an array of Sitems, the probed item would
have been one of those held in memory on K/Sof trials, so that
performance will be correct on K/Sof the trials.
For each participant we calculated four Kvalues; one for the
WMC task, one for the no distraction condition, one for the
encoding distraction condition, and another for the delay distrac-
tion condition.
To determine whether the effects of encoding and delay distrac-
tion on performance uniquely predict WMC, we performed a
hierarchical regression analysis. In this hierarchical regression,
performance in the no distractor condition was used to predict
WMC, and then performance in both the encoding and delay
distractor conditions was added to the model. R
2
change between
the two models was used to assess the variability in WMC that
could be explained by distractor filtering in general. Once all
predictors were in the model we examined the unique contribu-
tions of encoding and delay distraction while controlling for WM
performance in the other conditions.
961
DISSOCIATING DISTRACTOR-FILTERING

All statistical analyses were performed using IBM SPSS Statistics
20. For all correlations, pvalues were determined with two-tailed
analyses.
Results
Values for WMC for each of the conditions in the two studies
are shown in Table 1, and zero-order correlations are shown in
Table 2. As shown in Table 1, although Kvalues for the measure
of WMC (load 5, no distractors) were lower than for each of the
load 3 conditions, when comparing performance across the load 3
conditions, performance declined with the inclusion of distractors.
As shown by Table 3, all predictors together significantly pre-
dicted WMC (adjusted R
2
⫽0.39, F
3,17
⫽5.29, p⫽.009). Over
and above the no-distractor condition, an additional 28% of the
variability (p⫽.025) in WMC was explained by adding both
encoding and delay distractor performance to the model in which
no distraction performance was used to predict WMC. Both K
values at encoding and delay distraction conditions significantly
and uniquely predicted WMC while controlling for performance in
the other two conditions (encoding distraction: standardized ␤⫽
0.616, p⫽.038; delay distraction: standardized ␤⫽0.615, p⫽
.025, Figure 2a and 2b). The positive associations indicate that
greater distractor filtering performance is associated with greater
WMC. Furthermore, no significant correlation was seen between
encoding distraction and delay distraction performance (partial
correlation controlling for WM no distraction, r⫽⫺0.211, p⫽
.372), in keeping with separate mechanisms being involved with
distractor-filtering at encoding and delay.
Figure 1. a) The figure illustrates the task used to obtain an estimate of WMC in the absence of distraction.
Participants were asked to remember the positions of five red target circles displayed for 1 s followed by a delay
period of 3 s (during which a white fixation cross was shown) and respond to a probe stimulus (a question mark
presented for 2 s) which asked participants to indicate whether a red target circle had been shown in the position
indicated (a yes/no response). Targets were positioned such that no more than two targets were in adjacent
positions. The probe was shown either in or adjacent to one of the target positions. Forty trials were given, with
half of the probes presented in a target position and half presented in a position adjacent to a target position. b
through d) The three conditions used to obtain estimates of distractor filtering ability at encoding and delay.
Target circles were displayed for 1 s, with no more than two of the targets in adjacent positions. The circular grid
remained on the screen throughout each trial. A probe (a white question mark) was shown either in or adjacent
to one of the target positions for 2 s, 3 s after the target stimuli had disappeared. b) In the “no distraction”
condition participants were required to remember the locations of three red target circles, and no distractors
appeared. c) In the “encoding distraction” condition two yellow distractor circles were shown together with the
three red target circles. One of the yellow distractor circles was always in a position adjacent to a target position.
d) In the “delay distraction” condition two yellow distractor circles were shown during the delay period, 0.5 s
after the red target circles had disappeared. One of the yellow distractor circles was always in a position adjacent
to a target position. Participants were asked to remember the positions of only the red target circles, and give a
yes/no response to indicate whether the probe was in a position that had been occupied by a red target circle.
962 MCNAB AND DOLAN

Smartphone Study
Method
Participants. Data from participants aged 18–29 years were
considered. Data were excluded from participants who failed at the
easiest level of any of the conditions (i.e., failed two consecutive
trials of WM load 2). Following these exclusions, data from 3,247
participants remained for analysis.
Experimental design and task. This smartphone game
formed part of the “The Great Brain Experiment,” which is funded
by the Wellcome Trust (http://thegreatbrainexperiment.com). The
working memory game involved six conditions, four of which are
described here. Participants were asked to remember the positions
of red circles that appeared on a 4 ⫻4 grid for 1 s. At the end of
each trial they were presented with an empty grid and asked to
press on the grid positions in which red circles had appeared. In
three of the conditions there was a delay period of 1 s during which
time an empty grid was shown, after the red circles had disap-
peared and before participants could make their response. In one of
these conditions (“no distraction”) only red circles were displayed.
In another (“encoding distraction), two yellow distractor circles
were shown together with the red circles. In the third of these
conditions (“delay distraction”), two yellow distractor circles were
displayed during the delay period. In the “short delay” condition,
only red circles were shown and participants could begin making
their response immediately after the circles had disappeared. For
each condition the number of red circles (WM load) increased with
performance (one red circle added each time two successive trials
were answered correctly) until the participant failed two succes-
sive trials of a condition (from which point the game continued
without that condition) or the game timed-out.
Data analysis. Although participants were able to play the
game multiple times, we only considered data from their first play,
to ensure that practice effects could not influence the results.
Performance in each condition was measured as the last WM load
at which two successive trials were answered correctly. To deter-
mine whether encoding and delay distraction performance
uniquely predicted WMC, when controlling for performance in the
absence of distractors, we again performed a regression analysis.
We used the score from the “short delay” condition as our measure
of WMC (the dependent variable) because it did not involve any
distractors. With hierarchical regression, performance in the no
distractor condition was used to predict WMC, and then perfor-
mance in both the encoding and delay distractor conditions was
added to the model. Again, R
2
change between the two models was
used to assess the variability in WMC that could be explained by
distractor filtering in general. Once all predictors were in the
model we examined the unique contributions of encoding and
delay distraction while controlling for WM performance in the
other conditions.
Results
Values for WMC for each of the conditions in the two studies
are shown in Table 1, and zero-order correlations are shown in
Table 2. Performance was lower for both the encoding and delay
distraction conditions compared with both our measure of WMC
and the no distraction condition.
As anticipated, the estimates of WMC are very different for the
two studies, most likely because of the very different ways in
which the measures were obtained (Kvalues for the laboratory
study, maximum level reached for the smartphone study).
Table 1
Mean Working Memory Capacity (K) for Each Condition for Both the Laboratory and
Smartphone Studies
Study WMC
measure No
distraction Encoding
distraction Delay
distraction
Laboratory study 2.13 (⫾0.84) 2.79 (⫾0.21) 2.59 (⫾0.47) 2.68 (⫾0.38)
Smartphone study 9.15 (⫾1.14) 9.26 (⫾1.05) 9.12 (⫾1.15) 8.74 (⫾1.42)
Note. For the laboratory study, “WMC measure” refers to the task used to estimate WMC in the absence of
distraction (load 5). For the smartphone study, “WMC measure” refers to the “short delay” condition.
Table 2
Results of Pearson Correlations (R) Between Each Variable in Each of the Two Studies
Study WMC measure No distraction Encoding distraction
Laboratory study
No distraction 0.40
ⴱⴱ
——
Encoding distraction 0.45
ⴱⴱ
0.46
ⴱⴱ
—
Delay distraction 0.43
ⴱⴱ
0.42
ⴱⴱ
0.45
ⴱⴱ
Smartphone study
No distraction 0.45
ⴱ
——
Encoding distraction 0.54
ⴱ
0.76
ⴱⴱ
—
Delay distraction 0.57
ⴱⴱ
0.70
ⴱⴱ
0.43
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
963
DISSOCIATING DISTRACTOR-FILTERING

As shown by Table 3, all predictors together significantly pre-
dicted performance during the short delay condition (adjusted
R
2
⫽0.290, F
3,3243
⫽442.079, p⬍.001). Over and above the
no-distractor condition, an additional 13% of the variability (p⬍
.001) in WMC was explained by adding both encoding and delay
distractor performance to the model in which no distraction per-
formance was used to predict WM performance in the short delay
condition. Both performance in encoding and delay distraction con-
ditions significantly and uniquely predicted WM performance in the
short delay condition, while controlling for performance in the other
two conditions (encoding distraction: standardized ␤⫽0.263, p⬍
.001; delay distraction: standardized ␤⫽0.230, p⬍.001), in keeping
with separate mechanisms for distractor-filtering at encoding and
delay making a unique contribution to WMC. The positive associa-
tions indicate that greater distractor filtering performance is associated
with greater WMC. This time, a significant correlation was seen
between performance on encoding and delay distraction conditions
(partial correlation controlling for performance on the no distraction
condition, r⫽.315, p⬍.001), indicating that although they make a
unique contribution to WMC, their performance was not completely
unrelated in the context of the smartphone study. Possible explana-
tions for this are considered in the discussion.
Similarly, when we used performance on the short delay con-
dition as the baseline condition (as a predictor variable, and per-
Table 3
Results of the Hierarchical Regression in Which Model 1 Predicts WMC Using Performance
From the No Distraction Condition, and Model 2 Predicts WMC Using Performance From No
Distraction Condition Together With Both the Encoding Distraction and Delay
Distraction Conditions
Study Model Predictor R
2
change Standardized
beta p
value Partial
correlation
Laboratory study 1 No distraction 0.20 0.45 0.04
2 No distraction 0.28 ⫺0.45 0.21 ⫺0.30
Encoding distraction 0.62 0.03 0.48
ⴱ
Delay distraction 0.62 0.04 0.51
ⴱ
Smartphone study 1 No distraction 0.16 0.40 0.00
2 No distraction 0.13 0.18 0.00 0.18
ⴱⴱⴱ
Encoding distraction 0.26 0.00 0.25
ⴱⴱⴱ
Delay distraction 0.23 0.00 0.23
ⴱⴱⴱ
Note. The results of partial correlations are also shown for each of the predictors in Model 2, showing the
extent to which they correlate with WMC, whilst controlling for the other two predictor variables.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
ⴱⴱⴱ
p⬍.001.
Figure 2. Positive associations between WMC and the residual after accounting for no distraction performance
in a) the encoding distraction condition and b) the delay distraction condition for the laboratory study, and
between the short delay condition and the residual after accounting for no distraction performance in c) the
encoding distraction condition and d) the delay distraction condition for the smartphone study. The positive
associations indicate that greater distractor filtering performance is associated with greater WMC.
964 MCNAB AND DOLAN

formance on the WM no distraction condition as our measure of
WMC (the dependent variable), both encoding and delay distrac-
tion performance significantly and uniquely predicted WMC (ad-
justed R
2
⫽0.294, F
3,3243
⫽451.38, p⬍.001, for encoding
distraction performance standardized ␤⫽0.285, p⬍.001; for
delay distraction performance standardized ␤⫽0.214, p⬍.001).
Again there was a significant correlation between performance on
the encoding and delay distraction conditions (partial correlation
controlling for performance in the short delay condition, r⫽.316,
p⬍.001).
Because the maximum possible score for each condition was 10
(as a result of the game timing-out), to ensure that our results were
not a consequence of ceiling effects we repeated the analysis only
considering data from participants who scored less than 10 in the
no distraction and the short delay conditions (n⫽986). Regression
analysis again showed a unique contribution from both encoding
and delay distraction to performance in the short delay condition
(adjusted R
2
⫽0.234, F
3,982
⫽101.26, p⬍.001, encoding
distraction: standardized ␤⫽0.242, p⬍.000; delay distraction:
standardized ␤⫽0.233, p⬍.000). Using a partial correlation
(controlling for performance on the no distraction condition) we
again found a significant correlation between performance on
encoding and delay distraction conditions (r⫽.324, p⬍.001).
Discussion
In line with previous work in which encoding distractor-filtering
has been identified as a potential basis for WMC (Vogel et al.,
2005), in two studies we found that participants with greater WMC
were less affected by distraction presented during encoding, indi-
cating more effective distractor-filtering. We extend this work by
comparing performance associated with encoding and delay dis-
traction, while controlling for performance in the absence of dis-
traction, in the same individuals. We used one standard laboratory
experiment with data from 21 participants, and a large-scale study
with data collected from 3,247 participants using smartphones. For
both studies multiple regression analysis revealed that both encod-
ing and delay distractor-filtering significantly predicted WMC,
each accounting for unique variance in WMC, highlighting a
distinction between encoding and delay distractor filtering.
The unique contributions we observe from encoding/delay dis-
tractor filtering in predicting WMC provide tentative evidence that
encoding and delay distractor-filtering make separate contributions
to WMC, and that separate mechanisms for encoding and delay
distractor-filtering might contribute to WMC. The fact that we see
such associations when controlling for WM performance in the
absence of distraction argues against the alternative account in
which those with high WMC have extra WM resources available
to successfully encode the distractors. Such extra capacity would
only be available if baseline performance was at ceiling. Because
no participant reached the maximum of K⫽5 in the WMC
condition of the laboratory study, and we have replicated the
results of the smartphone study while excluding participants who
reached the maximum level (level 10) in the no distraction or short
delay conditions, our results are not explicable by mere ceiling
effects. Furthermore, in the case of encoding distractor filtering,
preparatory brain activity associated with distractor filtering, be-
fore the onset of target stimuli has been shown to predict WMC
and the unnecessary storage of distractors (McNab & Klingberg,
2008) indicating that, at least for encoding distraction, preparation
to filter distractors makes a contribution to subsequent WM per-
formance.
The laboratory experiment and the smartphone study differed in
a number of ways besides the size of the sample. For example, in
the laboratory experiment we used the Kvalue as a measure of
WM performance, whereas for the smartphone study, estimates of
WM performance were obtained by increasing WM load and
identifying the load at which the participant failed. The Kvalue
can be obtained by averaging over many trials and takes into
account guessing, so may be a more precise measure, whereas
increasing load to find the participant’s limit is more suited to a
game format, and may therefore be more motivating. As antici-
pated, the different measures of WMC yielded very different
values of WMC in the two studies (see Table 1). Many other
differences between the two studies may have contributed to this
difference. For example, stimuli were presented on smaller screens
in the smartphone study, in a rectangular grid (perhaps enabling
easier labeling of different grid positions, and easier chunking
compared to the circular grid used in the laboratory study) and
participants were encouraged to play competitively. Furthermore,
we were unable to control the environment in which the smart-
phone game was played. Regardless of these differences, we find
a unique contribution from both encoding and delay distraction
cost to WMC with both approaches. The different encoding and
delay distractor durations used in the laboratory study (1 s and 2 s,
respectively) and the smartphone study (1 s and 1 s, respectively)
indicate that differences in distractor duration do not affect the
separate contribution of encoding and delay distractor filtering to
WMC that we observe in both studies.
The positive correlation between encoding and delay distraction
observed in the smartphone study suggests that although we find
evidence for unique contributions from encoding and delay dis-
traction, shared variability may also exist. A common mechanism
may also contribute differently to conditions of encoding and delay
distraction, for example being sustained for one but not the other.
Another difference between the laboratory and the smartphone
study was a significant positive correlation between encoding and
delay distraction cost in the latter but not the former, when con-
trolling for performance in the absence of overt distractors. Al-
though this does not affect our main result (the significant and
unique contribution made by encoding and delay distraction to
WMC in the two studies), the difference is intriguing and indicates
that encoding and delay distraction performance is not entirely
unrelated under certain conditions. One possibility is that a differ-
ence in WM load may have affected their association. In the
laboratory study WM load was held constant at load 3, whereas
WM load in the smartphone study increased in line with perfor-
mance, so that participants were exposed to higher load trials.
Another possibility is that in more distracting environments the
ability to ignore encoding and delay distraction becomes more
closely associated. Whether there is a closer association between
encoding and delay distractor filtering at higher WM loads or in
more distracting environments, and the mechanisms responsible
for this, are questions for future research.
Separate mechanisms for distractor-filtering at encoding and
during the delay could resolve an apparent inconsistency high-
lighted by Cools et al. (2010). They observed impaired backward
span performance in patients with PD off medication, which is
965
DISSOCIATING DISTRACTOR-FILTERING

presumably a result of deficient striatal dopamine, yet enhanced
distractor-filtering. Similarly, Mehta et al. (2004) observed a task
set shifting impairment but an improvement in resistance to dis-
traction after sulpiride, a D2 receptor antagonist that specifically
affects the striatum (Mehta et al., 1999;Mehta et al., 2003). Cools
et al. (2010) acknowledge that distractor filtering has also been
linked to the functions of the basal ganglia (McNab & Klingberg,
2008;Gruber et al., 2006), revealing an apparent inconsistency.
However, both Mehta et al. (2004) and Cools et al. (2010) pre-
sented distractors during the delay period, whereas the study which
linked distractor-filtering to the striatum presented distractors only
during encoding (McNab & Klingberg, 2008). If encoding, but not
delay distractor filtering, involves the striatum this apparent in-
consistency would be resolved.
A model for selection in WM has been proposed based on
disinhibitory gating in the motor domain, putatively involving the
basal ganglia selectively initiating the storage of new memories by
“unlocking the gate” to memory maintenance in the frontal cortex
(Frank et al., 2001). In this way frontal memory representations
can be rapidly updated depending on task-relevance. One possible
account for our findings is that a striatal gating mechanism is
specific to encoding, but not delay, distractor-filtering. In the case
of delay distractor filtering, the gate allowing access to WM has
presumably already been “locked,” and an alternative mechanism
is then required to preserve a frontal memory representation in the
face of distraction. This may take the form of an increase in the
power of alpha oscillations which serve to close “the doors of
perception,” allowing effective delay distractor filtering (Bonne-
fond & Jensen, 2012). It also seems likely that delay distractor-
filtering involves the frontal cortex but not the striatum as during
delay distraction, Cools et al. (2007) observed that activity in
lateral frontal cortex, but not the striatum, was potentiated by the
dopamine agonist Bromocriptine. Single cell recordings have
shown that some PFC neurons create spatially tuned delay activity
as a result of delay distraction (Artchakov et al., 2009). Similarly,
middle frontal cortex/dorsolateral prefrontal cortex (DLPFC) and
inferior frontal gyrus (IFG) have been specifically implicated in
delay distractor-filtering (Dolcos, Miller, Kragel, Jha & McCarthy,
2007;Sakai, Rowe & Passingham, 2002) with DLPFC linked to
the enhancement of relevant information in the presence of dis-
traction (Feredoes, Heinen, Weiskopf, Ruff & Driver, 2011) and
IFG linked to the inhibition of interference (Dolcos et al., 2007;
Aron, Robbins & Poldrack, 2004). Furthermore, Minamoto et al.
(2010) observed that participants with high compared with low
WMC showed greater activity in left middle frontal gyrus during
delay distraction, greater accuracy on delay distraction trials, and
less activity in sensory cortices (fusiform gyrus) during distractor
presentation. We propose that the evidence suggests top-down
modulation from the left middle frontal gyrus provides more
effective delay distractor-filtering in participants with high WMC,
and is a possible neural basis for the delay distraction-filtering
mechanism we identify. A frontal delay distractor-filtering mech-
anism would also fit with the data of Cools et al. (2010) and Mehta
et al. (2004) if compensatory upregulation of frontal dopamine
underlies their findings of improved delay distractor-filtering in
PD and with sulpiride, respectively.
As well as the basal ganglia, middle frontal cortex is implicated
in encoding distraction filtering (McNab & Klingberg, 2008).
Toepper et al. (2010) observed ventrolateral and dorsolateral PFC
activity during a Corsi Block Tapping task, and greater DLPFC
activity in a version where distractors were added “during encod-
ing.” Based upon this finding it was suggested that DLPFC is
involved with higher-level executive processes including both
inhibition of spatial distraction (Toepper et al., 2010). However, in
this study spatial locations to be remembered were presented
sequentially, so that a distractor presented during the encoding of
the second stimulus in the sequence, for example, would have also
acted as a delay distractor for the first stimulus in the sequence.
Therefore, it is difficult to attribute enhanced DLPFC activity they
observe specifically to the effects of encoding distraction. Further
work is needed to establish whether encoding and delay distractor-
filtering involve different frontal mechanisms, and to understand
the basis for the dissociation we observe between encoding and
delay distractor-filtering.
It is clear that there are a number of differences between
encoding and delay distractor conditions. For example, encoding
distractors may group with the targets given their common onsets,
whereas the delay distractors have a sudden onset after presenta-
tion of the targets, which may capture attention. There may also be
differences in arousal during the different stages of the WM task.
Further work is needed to manipulate and control such factors to
decipher their relative contribution to WMC and the neural mech-
anisms responsible for overcoming their distracting effects.
The evidence we provide, suggesting separate mechanisms for
encoding and delay distractor-filtering, and their separate contri-
butions to WMC, could have implications for understanding the
basis of WMC impairments, for example in schizophrenia and
healthy aging. In these groups distractor-filtering has tended to be
studied with distractors presented during a delay, or between the
sequential presentation of the stimuli that are to be remembered,
preventing a clear distinction between encoding and delay
distractor-filtering (e.g., Anticevic, Repovs, Krystal & Barch,
2012;Cameron et al., 2002;Gazzaley, Cooney, Rissman &
D’Esposito, 2005). Furthermore, cognitive training can lead to
improvements in WMC (Klingberg, Forssberg & Westerberg,
2002;Dahlin et al., 2008) with good evidence to suggest that
distractor-filtering can be trained (Melara, Rao & Tong, 2002). By
identifying the extent to which encoding or delay distractor-
filtering is impaired in a certain individual, it may be possible to
tailor the cognitive training to the individual and increase improve-
ments.
In conclusion, with two separate studies (one a standard labo-
ratory experiment with data from 21 participants, and the other a
large-scale study with data collected from 3,247 participants using
smartphones), we provide evidence for a dissociation between
encoding and delay distractor-filtering, with both separately con-
tributing toward WMC. These therefore represent two potential
bases for WMC.
References
Anticevic, A., Repovs, G., Krystal, J. H., & Barch, D. M. (2012). A broken
filter: Prefrontal functional connectivity abnormalities in schizophrenia
during working memory interference. Schizophrenia Research.doi:
10.1016/j.schres.2012.07.007
Aron, A. R., Robbins, T. W., & Poldrack, R. A. (2004). Inhibition and the
right inferior frontal cortex. Trends in Cognitive Sciences, 8, 170–177.
doi:10.1016/j.tics.2004.02.010
966 MCNAB AND DOLAN

Artchakov, D., Tikhonravov, D., Ma, Y., Neuvonen, T., Linnankoski, I., &
Carlson, S. (2009). Distractor impair and create working memory-related
neuronal activity in the prefrontal cortex. Cerebral Cortex, 19, 2680–
2689. doi:10.1093/cercor/bhp037
Baddeley, A. (2003). Working memory: Looking back and looking for-
ward. Nature Reviews Neuroscience, 4, 829. doi:10.1038/nrn1201
Bonnefond, M., & Jensen, O. (2012). Alpha oscillations serve to protect
working memory maintenance against anticipated distracters. Current
Biology, 22, 1969–1974. doi:10.1016/j.cub.2012.08.029
Bopp, K. L., & Verhaeghen, P. (2005). Aging and verbal memory span: A
meta-analysis. Journals of Gerontology. Series B, Psychological Sci-
ences and Social Sciences, 60, 223. doi:10.1093/geronb/60.5.P223
Cameron, A. M., Oram, J., Geffen, G. M., Kavanagh, D. J., McGrath, J. J.,
& Geffin, L. B. (2002). Working memory correlates of three symptom
clusters in schizophrenia. Psychiatry Research, 110, 49–61. doi:
10.1016/S0165-1781(02)00036-7
Castellanos, F. X., & Tannock, R. (2002). Neuroscience of attention-
deficit/hyperactivity disorder: The search for endophenotypes. Nature
Reviews Neuroscience, 3, 617–628.
Cools, R., Miyakawa, A., Sheridan, M., & D’Esposito, M. (2010). En-
hanced frontal function in Parkinson’s disease. Brain, 133, 225–233.
doi:10.1093/brain/awp301
Cools, R., Sheridan, M., Jacobs, E., & D’Esposito, M. (2007). Impulsive
personality predicts dopamine-dependent changes in fronto-striatal ac-
tivity during component processes of working memory. The Journal of
Neuroscience, 27, 5506–5514. doi:10.1523/JNEUROSCI.0601-07.2007
Cowan, N. (2001). The magical number 4 in short-term memory: A
reconsideration of mental storage capacity. Behavioral and Brain Sci-
ences, 24, 87–114. doi:10.1017/S0140525X01003922
Dahlin, E., Neely, A. S., Larsson, A., Bäckman, L., & Nyberg, L. (2008).
Transfer of learning after updating training mediated by the striatum.
Science, 320, 1510–1512. doi:10.1126/science.1155466
Dolcos, F., Miller, B., Kragel, P., Jha, A., & McCarthy, G. (2007).
Regional brain differences in the effect of distraction during the delay
interval of a working memory task. Brain Research, 1152, 171–181.
doi:10.1016/j.brainres.2007.03.059
Feredoes, E., Heinen, K., Weiskopf, N., Ruff, C., & Driver, J. (2011).
Causal evidence for frontal involvement in memory target maintenance
by posterior brain areas during distractor interference of visual working
memory. PNAS Proceedings of the National Academy of Sciences of the
United States of America, 108, 17510–17515. doi:10.1073/pnas
.1106439108
Frank, M. J., Loughry, B., & O’Reilly, R. C. (2001). Interactions between
frontal cortex and basal ganglia in working memory: A computational
model. Cognitive, Affective & Behavioural Neuroscience, 1, 137–160.
doi:10.3758/CABN.1.2.137
Gazzaley, A., Cooney, J. W., Rissman, J., & D’Esposito, M. (2005).
Top-down suppression deficit underlies working memory impairment in
normal aging. Nature Neuroscience, 8, 1298–1300. doi:10.1038/nn1543
Goldman-Rakic, P. S. (1994). Working memory dysfunction in schizo-
phrenia. Journal of Neuropsychiatry and Clinical Neurosciences, 6, 348.
Gruber, A. J., Dayan, P., Gutkin, B. S., & Solla, S. A. (2006). Dopamine
modulation in the basal ganglia locks the gate to working memory.
Journal of Computational Neuroscience, 20, 153–166. doi:10.1007/
s10827-005-5705-x
Ihssen, N., Linden, D. E. J., & Shapiro, K. L. (2010). Improving visual
short-term memory by sequencing the stimulus array. Psychonomic
Bulletin & Review, 17, 680–686. doi:10.3758/PBR.17.5.680
Klingberg, T., Forssberg, H., & Westerberg, H. (2002). Training of work-
ing memory in children with ADHD. Journal of Clinical and Experi-
mental Neuropsychology, 24, 781–791. doi:10.1076/jcen.24.6.781.8395
McNab, F., & Klingberg, T. (2008). Prefrontal cortex and basal ganglia
control access to working memory. Nature Neuroscience, 11, 103–107.
doi:10.1038/nn2024
McNab, F., Varrone, A., Farde, L., Jucaite, A., Bystritsky, P., Forssberg,
H., & Klingberg, T. (2009). Changes in cortical dopamine D1 receptor
binding associated with cognitive training. Science, 323, 800–802.
doi:10.1126/science.1166102
Mehta, M. A., Manes, F. F., Magnolfi, G., Schakian, B. J., & Robbins,
T. W. (2004). Impaired set-shifting and dissociable effects on tests of
spatial working memory following the dopamine D2 receptor antagonist
sulpiride in human volunteers. Psychopharmacology, 176, 331–342.
doi:10.1007/s00213-004-1899-2
Mehta, M. A., McGowan, S., Lawrence, A. D., Aitkin, M. R. F., Mont-
gomery, A. J., & Grasby, P. R. (2003). Systemic sulpiride modulates
striatal blood flow: Relationships to spatial working memory and plan-
ning. NeuroImage, 20, 1982–1994. doi:10.1016/j.neuroimage.2003.08
.007
Mehta, M. A., Sahakian, B. J., McKenna, P. J., & Robbins, T. W. (1999).
Systemic sulpiride in young adult volunteers simulates the profile of
cognitive deficits in Parkinson’s disease. Psychopharmacology, 146,
162–174. doi:10.1007/s002130051102
Melara, R. D., Rao, A., & Tong, Y. (2002). The duality of selection:
Excitatory and inhibitory processes in auditory selective attention. Jour-
nal of Experimental Psychology: Human Perception and Performance,
28, 279–306. doi:10.1037/0096-1523.28.2.279
Minamoto, T., Osaka, M., & Osaka, N. (2010). Individual differences in
working memory capacity and distractor processing: Possible contribu-
tion of top-down inhibitory control. Brain Research, 1335, 63–73.
doi:10.1016/j.brainres.2010.03.088
Sakai, K., Rowe, J. B., & Passingham, R. E. (2002). Active maintenance in
prefrontal area 46 creates distractor-resistant memory. Nature Neurosci-
ence, 5, 479–484.
Sü␤, H-M., Oberauer, K., Wittmann, W. W., Wilhelm, O., & Schulze, R.
(2002). Working-memory capacity explains reasoning ability—and a
little bit more. Intelligence, 30, 261–288. doi:10.1016/S0160-
2896(01)00100-3
Toepper, M., Gebhardt, H., Beblo, T., Thomas, C., Driessen, M., Bischoff,
M....Sammer, G. (2010). Functional correlates of distractor suppres-
sion during spatial working memory encoding. Neuroscience, 165,
1244–1253. doi:10.1016/j.neuroscience.2009.11.019
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. doi:10.1038/nature04171
Received July 15, 2013
Revision received January 13, 2014
Accepted January 15, 2014 䡲
967
DISSOCIATING DISTRACTOR-FILTERING