Content uploaded by Torkel Klingberg
Author content
All content in this area was uploaded by Torkel Klingberg
Content may be subject to copyright.
Prefrontal cortex and basal ganglia control access
to working memory
Fiona McNab & Torkel Klingberg
Our capacity to store information in working memory might be determined by the degree to which only relevant information is
remembered. The question remains as to how this selection of relevant items to be remembered is accomplished. Here we show
that activity in the prefrontal cortex and basal ganglia preceded the filtering of irrelevant information and that activity, particularly
in the globus pallidus, predicted the extent to which only relevant information is stored. The preceding frontal and basal ganglia
activity were also associated with inter-individual differences in working memory capacity. These findings reveal a mechanism by
which frontal and basal ganglia activity exerts attentional control over access to working memory storage in the parietal cortex in
humans, and makes an important contribution to inter-individual differences in working memory capacity.
Working memory capacity is an important factor for a wide range of
cognitive abilities, including general fluid intelligence
1,2
. Recent studies
of humans using functional magnetic resonance imaging (fMRI) and
electroencephalography have identified a region in the parietal lobe
where brain activity reflects the amount of stored visuo-spatial infor-
mation
3,4
. Furthermore, subsequent studies have shown that when a
working memory trial contains both relevant and irrelevant informa-
tion, storage-related parietal activity for distractors is negatively corre-
lated with working memory capacity, so that individuals with high
working memory capacity are less likely to store irrelevant distractors,
which would unnecessarily consume capacity
5
. This suggests that the
extent to which only relevant information is stored is related to, and
may form a basis for working memory capacity
5
.
The neural basis for the control of access to working memory
storage, the possible neural determinant of working memory capacity,
is still unknown. It has been suggested that such control may stem from
a bias signal from the prefrontal cortex
5
, and recordings from the lateral
prefrontal cortex of monkeys indicate that this region is involved in the
selection of behaviorally relevant information
6
.However,theregions
involved in such top-down control, and the relationship between their
activity and working memory capacity, have not been investigated.
To address this, we conducted an fMRI study that was designed to
identify activity associated with preparation to filter out irrelevant items
that were presented during encoding in a visual-spatial working
memory task. Task instructions were given before the presentation of
the memory stimuli, a method that has been used previously to identify
the neural correlates of various task sets (for example, see ref. 7), and
that, in this study, enabled us to isolate top-down control processes
from processes related to the encoding of stimuli into working memory.
In each trial the task instruction took the form of a geometric shape
(a square or a triangle) that indicated whether yellow circles should act
as distractors to be ignored (the ‘distraction’ condition) or target
stimuli to be remembered (the ‘no distraction’ condition) in the
subsequent working memory task (Fig. 1). In the distraction task,
subjects needed to remember three red circles (targets) and ignore two
yellow circles (distractors). In the no distraction task, the number of
targets were either three (all red) or five (3 red circles and 2 yellow).
Activity that was associated with preparation to filter out irrelevant
items, before the processing of the memory stimuli, was identified by
contrasting the instruction periods of the distraction task condition
and the no distraction task condition.
RESULTS
Task difficulty
The inclusion of distractors increased task difficulty, as seen by the
accuracy in trials of three target circles with and without distraction
(accuracy was 80% ± 14% and 85% ± 11%, respectively, mean ± s.d.;
paired t-test, t ¼ –2.3; P ¼ 0.015, n ¼ 24). However, as the no
distraction condition sometimes included trials with three and some-
times five targets, there was no difference in accuracy, on average,
between the distraction and the no distraction conditions (80% ± 14%
and 78% ± 10%, respectively, n ¼ 24). Therefore, the task instruction
did not predict differences in task difficulty.
Filtering set activity
‘Filtering set activity’ was defined as the difference in brain activity
between the instruction periods of the distraction trials and the no
distraction trials. Such activity was observed in three regions: bilaterally
in the posterior part of the middle frontal gyrus (in and anterior to the
precentral sulcus) and in left basal ganglia (with one local maxima in
the putamen and one in the global pallidus) (P o 0.05, corrected for
multiple comparisons). We determined the time course of activity at
Received 14 August; accepted 6 November; published online 9 December 2007; doi:10.1038/nn2024
Developmental Cognitive Neuroscience, Stockholm Brain Institute, Karolinska Institutet, MR Centrum, N8:00, 17176 Stockholm, Sweden. Correspondence should be
addressed to T.K. (Torkel.Klingberg@ki.se).
NATURE NEUROSCIENCE VOLUME 11
[
NUMBER 1
[
JANUARY 2008 103
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/natureneuroscience
the maxima for each region, averaged across sessions and subjects
(Fig. 2). Activation peaked at 6 s after the onset of the instruction,
which corresponds to the delay of the hemodynamic response. These
results indicate that bilateral middle frontal gyrus and left basal ganglia
are involved in the preparation to select the information that is to be
stored in working memory. We next investigated the relationship
between this preparatory activity and both working memory capacity
and the storage of distractors.
Working memory capacity
We conducted a behavioral experiment outside of the scanner to
establish each participant’s working memory capacity. The procedure
was the same as that used for the no distraction condition of the
scanning task, except that trials always began with the presentation of a
diamond and the grid contained three, four, five or six red circles.
Participants were required to remember the positions of the circles and
to indicate, via a button press, whether the probe position corre-
sponded to one of these target positions (a yes or no response).
Working memory capacity was estimated with the K-value, estimating
how much information can be stored in working memory, using a
standard formula
5,8
. For both the frontal and basal ganglia regions that
showed filtering set activity (Fig. 2) we extracted the mean relative
signal change from the distraction versus no distraction contrast,
during the instruction period, for each participant, and correlated
these values with working memory capacity (the average from the two
prefrontal regions was used). There was a moderate, but significant,
positive correlation between the mean filtering set activity in
the prefrontal cortex and working memory capacity (r ¼ 0.35,
P ¼ 0.045; Fig. 3a) and between the mean filtering set activity in
the basal ganglia and working memory capacity (r ¼ 0.35,
P ¼ 0.042; Fig. 3b).
The activity (mean beta values for the regressor) associated with the
instruction for the no distraction condition did not correlate with
working memory capacity (prefrontal cortex: r ¼ 0.20, P ¼ 0.168; basal
ganglia: r ¼ –0.06, P ¼ 0.391). Therefore, it was specifically in the
contrast between the distraction and no distraction conditions
that prefrontal and basal ganglia activity correlated with working
memory capacity.
The basal ganglia cluster contained local maxima located in the
putamen and globus pallidus, respectively. Because these are function-
ally different parts of the basal ganglia circuit, separate correlation
analyses were carried out on the values of mean relative signal change
extracted from the maxima in the putamen and the maxima in the
globus pallidus, revealing that, although activity in the putamen voxel
did not significantly correlate with working memory capacity (r ¼ 0.27,
P ¼ 0.097; Fig. 3c), the activity in the globus pallidus voxel did correlate
with working memory capacity (r ¼ 0.57, P ¼ 0.001; Fig. 3d). There-
fore, in line with our hypothesis, preparatory activity in the frontal
regions and left basal ganglia (in particular, the globus pallidus) was
significantly correlated with working memory capacity. We next
investigated the relationship between preparatory filtering set activity
and brain activity that is related to the storage of distractors.
Unnecessary storage activity
As previously mentioned, the extent to which irrelevant distractors are
unnecessarily stored is reflected in event-related potentials that are
recorded over load-sensitive lateral occipital and parietal lobes
5
.To
identify such activity, we first located a load-sensitive parietal region by
contrasting the activity associated with the encoding and storage of five
circles (load 5) with that of three circles (load 3) in the no distraction
task condition, considering the period between the onset of the circles
and the onset of the probe stimulus. The maximum parietal difference
was seen in the right posterior parietal cortex, which may correspond to
the load-sensitive parietal region identified by previous studies
3–5
.We
identified the medial/lateral (x), anterior/posterior (y) and dorsal/
Distraction task
3 or 4 s
Task
instruction
Memory
stimuli
Time
Probe
2, 3 or 4 s
3, 4 or 5 s
2 s
1 s
No distraction task
Figure 1 The distraction condition (one third of trials) and the no distraction
condition (one third of trials) included in the scanning task (see the
manuscript text). The remaining third of trials involved a no memory
condition, indicated by a diamond, which followed the same format, but
required participants to make a color judgment. The results of this task
condition are not reported here.
0.08
0.04
0.04
0.06
0.02
0.00
–0.02
–0.04
0.04
0.02
0.00
–0.02
–0.04
0.00
–0.04
–0.08
4.2
8.4
12.6
16.8
21.0
25.2
29.4
4.2
8.4
12.6
16.8
21.0
25.2
29.4
4.2
8.4
12.6
16.8
21.0
25.2
29.4
Bilateral middle frontal gyrus
Time (s) Time (s) Time (s)
Relative signal change
Relative signal change
Relative signal change
Left basal ganglia
ab c de
Figure 2 Preparatory filtering set activity. The results from the distraction versus no distraction contrast are shown for the task instruction period (P o 0.05,
corrected for multiple comparisons). (a–c) Significant task-dependent differences were observed in bilateral middle frontal gyri (a, maximum at MNI
coordinates in mm (x, y, z): –40, –12, 50 and 48, –10, 44). The time series of relative signal change are shown for signals at the peak of task-dependent
differences in each cluster (b, left middle frontal gyrus; c, right middle frontal gyrus) for both distraction (shown in black) and no distraction (shown in gray)
task conditions. (d,e) Significant task-dependent differences were also observed in left basal ganglia (d, maximum at –18, 6, –6), and the time series of
relative signal change is also shown for the signal at the peak of task-dependent differences in this cluster (e). The error bars indicate s.e.m. and the shaded
section represents the times at which the memory stimuli were presented (which varied between 3, 4 and 5 s after the onset of the instruction cue).
104 VOLUME 11
[
NUMBER 1
[
JANUARY 2008 NATURE NEUROSCIENCE
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/natureneuroscience
ventral (z) extremes of this activity, and con-
verted these to Talairach coordinates. The
cluster extent was x ¼ 24/59, y ¼ –50/–77
and z ¼ 36/51, which corresponds closely with
the previously identified load-sensitive parie-
tal region
3
of x ¼ 17/29, y ¼ –52/–69 and z ¼
38/54 (personal communication, J.J. Todd &
R. Marois, Vanderbilt Vision Research Center,
Vanderbilt University).
If distractors are filtered effectively, the
response to the distraction condition in the
load-sensitive parietal area should be similar
to that observed for load 3 of the no distrac-
tion condition (as both included three target
stimuli). Conversely, if filtering is not effec-
tive, there should be greater activity in this region, reflecting the greater
working memory load that is associated with the additional storage of
distractors. Therefore, in the load-sensitive parietal cluster, we extracted
values of mean relative signal change from the contrast between the
distraction condition and load 3 trials of the no distraction condition
for each participant, considering the period between the onset of the
memory stimuli and the onset of the probe (corresponding to encoding
and storage), as this should reflect the extent to which irrelevant
distractors were unnecessarily stored.
We then correlated this parietal ‘unnecessary storage activity’ with
the preparatory filtering set activity for the regions in which significant
task-dependent differences had been observed (the prefrontal regions
and the globus pallidus) during the preceding instruction period.
A significant negative correlation was seen for the globus pallidus
(r ¼ –0.50, P ¼ 0.005; Fig. 4), but not for the prefrontal regions
(r ¼ –0.06, P ¼ 0.387), indicating that enhanced activity in the globus
pallidus region was associated with fewer distractors being unnecessa-
rily stored. Furthermore, the unnecessary storage activity was also
negatively correlated with working memory capacity (r ¼ –0.43, P ¼
0.016), which is consistent with the hypothesis that unnecessary storage
accounts for the correlation between filtering set activity and working
memory capacity. Filtering set activity in the globus pallidus was also
significantly negatively correlated with the difference in accuracy
between the distraction condition (three target circles and two
distractors) and the no-distraction load-3 condition (three target
circles) (r ¼ –0.40, P ¼ 0.028) during scanning, indicating that greater
filtering set activity was linked to a reduced loss in accuracy associated
with distracter presentation.
DISCUSSION
The present study identified the basal ganglia as being responsible for
allowing only relevant information into working memory. Consistent
with the theory that an individual’s working memory capacity is
determined by their ability to selectively filter irrelevant distractors
5
,
prefrontal and basal ganglia activity was a significant predictor of
working memory capacity (measured in the absence of overt distrac-
tors), and basal ganglia activity significantly negatively correlated with
parietal load effects that reflected the unnecessary storage of distractors.
The present results therefore reveal a specific neural mechanism by
which an individual’s ability to exert control over the encoding of new
information is linked to their working memory capacity
9–11
,measured
in the absence of overt distraction.
It has previously been suggested that individual differences in the
efficiency with which items are filtered from working memory may
stem from a bias signal emanating from the prefrontal cortex
5
. Here we
show prefrontal activity that meets this criterion. The activity is
associated with the preparation to filter items from working memory,
consistent with a role for the prefrontal cortex as a control region
12,13
.
The results also suggest that such a process is carried out in concert with
the basal ganglia, presumably according to one of the previously
described fronto-striatal loops
14
.
The basal ganglia are activated during planning and set-shifting
15–17
,
and have been shown to be important in the pathophysiology of several
diseases affecting sensory gating
18
. The globus pallidus is the output
module of the basal ganglia and contains motor, limbic and associative
regions, of which the latter is crucial for spatial attention
19
. Although
it has been acknowledged that the basal ganglia is involved in
working memory
20,21
, such an involvement is not well understood.
However, there is evidence for an involvement of the globus pallidus
during working memory–guided movement sequencing
22
,and
electrophysiological studies in primates have indicated that globus
pallidus activation is modulated by memory requirements during
motor sequencing
23
.
Furthermore, the basal ganglia have a high density of dopamine
receptors, which are central to working memory
24
. Using the idea that
dopamine can carry out a gating function by transiently strengthening
the efficiency of inputs to the frontal cortex, and by extending models
of disinhibitory gating in the motor domain, an interaction between
the frontal cortex and the basal ganglia has been modeled
25
.Inthis
model, the basal ganglia contribute a selective gating mechanism that
disinhibits thalamocortical loops and the influence of incoming stimuli
1.2
1.5
2.0
0.8
0.4
–0.4
0.0
1.0
0.0
0123456
1.0
0.5
–0.5
0.0
0.8
0.4
0.0
012
Relative signal change
Relative signal change
Relative signal change
Relative signal change
3
r = 0.35 r = 0.36 r = 0.27 r = 0.57
456
Working memory capacity
0123456
Working memory capacity Working memory capacity
0123456
Working memory capacit
y
Middle frontal gyrus Left basal ganglia
Not significant
Putamen Global pallidus
ab c d
Figure 3 Correlations between working memory capacity and preparatory filtering set activity.
(a–d) Correlations between working memory capacity and the mean relative signal change from the
distraction versus no distraction contrast in the instruction period from frontal clusters (a, P ¼ 0.045)
and the basal ganglia (b, P ¼ 0.042). Correlations are also shown for the maxima of the basal ganglia
cluster in the putamen (c, –18, 6, –6; P ¼ 0.097) and the globus pallidus (d,–12,–2,–8;P ¼ 0.001).
4
3
2
1
0
–1
–2
–3
–4
–0.4 0.40 0.8
Globus pallidus filtering set activity
(relative si
g
nal chan
g
e)
Parietal unnecessary storage
activity (relative signal change)
r = –0.50
ab
Figure 4 Correlations between preparatory filtering set activity and
unnecessary storage activity. (a) The right parietal region in which the load
effect was observed from the contrast between load 5 and load 3 in the no
distraction condition (maximum at 48, –66, 48). (b) The negative correlation
between the relative signal change extracted from the globus pallidus voxel in
the distraction versus no distraction contrast during the instruction period
and the mean relative signal change from the distraction versus no-distraction
load-3 contrast between onset of the circles and onset of the probe stimulus.
NATURE NEUROSCIENCE VOLUME 11
[
NUMBER 1
[
JANUARY 2008 105
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/natureneuroscience
on the working memory system is regulated. Similarly, dopamine in the
basal ganglia has been modeled as gating stabilization against distrac-
tion by enhancing select memories
26
.
The present findings also address the question of the neural basis for
the connection between attention and working memory. It is known
that attention and working memory are closely connected (as well as
being partly overlapping concepts), and working memory capacity
correlates with the efficiency of controlled attention
9–11,27–29
. Such
findings have led to the suggestion that attention can serve as a
gatekeeper for working memory, by biasing the encoding of informa-
tion toward items that are most relevant
10
. The present results provide
a neural basis for such a gatekeeping function.
In conclusion, we have shown that activity in the prefrontal cortex
and basal ganglia precedes the filtering of irrelevant items during the
encoding of working memory. This preparatory activity predicts the
extent to which only relevant information is stored, as reflected by
parietal storage–related activity, and predicts inter-individual differ-
ences in working memory capacity. This activity therefore reveals a
specific mechanism that could contribute toward an individual’s
working memory capacity.
METHODS
Participants. Twenty-five healthy participants (13 females, ages 19–33,
right handed) gave informed consent to participate in the study, which
was approved by the local ethics committee of the Karolinska Hospital
(Forskningetikprovning).
Tasks and stimuli. In the tasks carried out during scanning, the assignment of
geometric shapes to the distraction and no distraction task conditions
was counterbalanced across participants so that comparisons between task
instructions would not be confounded by differences in visual stimuli. The
mean magnitude of the cosine of the angle between regressors (which is 1
for collinear regressors and 0 for orthogonal regressors) indicated that the
activity associated with the presentation of the task instruction and
that associated with the presentation of the stimulus array was suffi-
ciently separated (0.046 in the distraction condition and 0.040 in the
no distraction condition).
There were 16 positions in the grid. In the behavioral experiment, the grid
subtended a visual angle of 201 both horizontally and vertically, and the
minimum difference between circles was 41, from center to center. In the fMRI
experiment, the grid subtended a visual angle of 111 both horizontally and
vertically, and the minimum difference between circles was 21, from center to
center. The instruction cues were presented for 3, 4 or 5 s. The stimulus array
was shown for 1 s and was followed by a delay of 2, 3 or 4 s, and then the probe
stimulus was displayed for 2 s.
In the behavioral experiment used to obtain a measure of working memory
capacity for each participant, conducted at least 1 week before scanning, all
parameters were the same as in the scanning procedure, but the experiment
only involved the no distraction task. The geometric shape was always a
diamond and target stimuli were three, four, five or six red circles (ten trials of
each, yellow circles were never shown). On presentation of the probe stimulus,
participants were required to make a button press with the index or middle
finger of their right hand, depending on whether a circle had appeared at the
location indicated (which was the case for half of the trials). When the probe
stimulus was not in a target position, it was in a position adjacent to one of the
target positions. The required response (yes or no) and the different durations
of presentation of the diamond and both fixation crosses were distributed
evenly across trials of each array size.
Visual working memory capacity was computed with the standard formu-
la
5,8
K ¼ S (H F), where K is the working memory capacity, S is the array
size, H is the observed hit rate and F is the false alarm rate. This formula uses
the false alarm rate to correct for guessing and assumes that if K items can be
held in working memory, from an array of S items, the probed item would have
been one of those held in memory on K/S of trials, so that performance will be
correct on K/S of the trials. For each participant, we computed the K value for
each of the four array sizes and used the mean K of array sizes 5 and 6 as our
measure of working memory capacity.
In the no distraction condition, the grid contained three red circles for half
of the trials and three red circles with two yellow circles for the other half. The
required response and durations of presentation of the instruction and both
fixation crosses were distributed evenly across task conditions and the two load
conditions. For each session, the stimulus configurations were generated
pseudo-randomly, with the criteria that a maximum of two target items could
be presented in adjacent locations and that one of the two yellow circles was
always in a location adjacent to a red circle. The stimulus configurations were
assigned to the different task conditions pseudo-randomly, but with the same
assignment for each participant. In 60% of trials that occurred in the
distraction condition, the probe appeared in a position that had been occupied
by a distracter. Trials were presented pseudo-randomly (with the trial types also
randomized) in an event-related design.
Before going into the scanner room, each participant completed one practice
session of the scanning task (30 trials, ten of each condition). In the scanner, 22
of the 25 participants completed four sessions of 30 trials (ten trials of each
condition), with the order of sessions being counterbalanced across partici-
pants. Button presses were recorded for all but one of the participants. Twenty-
two participants completed all four sessions, two completed three sessions and
one completed two sessions.
MRI acquisition. Images were acquired using a 1.5-T GE Signa scanner.
T2*-weighted, gradient echo echo-planar images were acquired with a repeti-
tion time of 2.1 s, an echo time of 40 ms, a flip angle of 761, 22 axial slices, 5-
mm slice thickness, 220-mm FOV and a 64 64 grid. Each session lasted
7 min and involved the acquisition of 195 volumes. T1-weighted spoiled
gradient images (FOV 240 mm) were acquired in the same position as the
functional images.
Data analysis. Preprocessing and statistical analysis were carried out with
SPM5 (Welcome Department of Cognitive Neurology, http://www.fil.ion.ucl.
ac.uk/spm/software/spm5). Preprocessing included slice-time correction,
motion correction, normalization to the template EPI (interpolating to
2-mm cubic voxels) and spatial smoothing with an 8-mm Gaussian kernel.
The models used a canonical hemodynamic response and its temporal
derivative; however, to plot the time course of the preparation activity, we
estimated the model again omitting the temporal derivative, and a finite
impulse response (FIR) approach was used. From this model, maxima were
located at Montreal Neurological Institute (MNI) coordinates –42, –10, 52, 48,
–8, 42, and –16, –4, –6, and these voxels were used to plot time courses (shown
in Fig. 2). The first model, used to identify filtering set activity, included
separate regressors for each of the instruction conditions, a regressor for the
presentation of the memory stimuli (duration 1 s, with a covariate of the
number of circles presented), a regressor for storage (beginning at the
presentation of the memory stimuli and ending at the onset of the probe,
with a covariate for the number of circles to remember) and a regressor for the
probe stimulus. A second model was used to investigate the unnecessary storage
activity. In this model there were separate regressors for each of the instruction
conditions and regressors for storage in the distraction condition, the no-
distraction load-3 condition, the no-distraction load-5 condition, the no
memory condition, and a regressor for the probe stimulus. In both cases, only
trials that received a correct response were included in the model. Correlation
analyses between fMRI data and behavioral outcomes (working memory
capacity and filtering ability), as well as between preparatory activity and
unnecessary storage activity, were carried out using SPSS for Windows (Rel.
11.5.0, SPSS) and the accompanying P values were determined by one-tailed
analysis with the hypothesis that the preparatory filtering ability would
determine working memory capacity and filtering ability.
Comparisons of interest were implemented as linear contrasts. The analysis
was carried out individually, and contrast images for each subject were used in
a second-level analysis, treating subjects as a random effect. For analysis of the
instruction period, the statistical map was thresholded with a false discovery
rate of P o 0.05, and differences were considered to be significant if they
fulfilled the criteria of an extent threshold of 150 voxels and a corrected cluster
level requirement of P o 0.05. To identify the parietal load-sensitive region, we
106 VOLUME 11
[
NUMBER 1
[
JANUARY 2008 NATURE NEUROSCIENCE
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/natureneuroscience
used a threshold of P o 0.01, uncorrected for multiple comparisons. This less
stringent threshold was used because we had an a priori hypothesis for a load
effect in this region, and the purpose of this analysis was only to identify the
load-sensitive region for further analysis of unnecessary storage.
ACKNOWLEDGMENTS
The authors thank G. Leroux, P. Fransson, F. Edin, A. Compte and
A.-C. Ingridsson for their help. This work was supported by the Foundation for
Strategic Research and the Knut and Alice Wallenberg Foundation.
AUTHOR CONTRIBUTIONS
F.M. and T.K. designed the tasks and wrote the manuscript together. F.M.
conducted the experiments and analyzed the data.
Published online at http://www.nature.com/natureneuroscience
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Conway, A.R.A., Kane, M.J. & Engle, R.W. Working memory capacity and its relation to
general intelligence. Trends Cogn. Sci. 7, 547–552 (2003).
2. Su
¨
, H.-M., Oberauer, K., Wittmann, W.W., Wilhelm, O. & Schulze, R. Working-memory
capacity explains reasoning ability—and a little bit more. Intelligence 30, 261–288
(2002).
3. Todd, J.J. & Marois, R. Capacity limit of visual short-term memory in human posterior
parietal cortex. Nature 428, 751–754 (2004).
4. Vogel, E.K. & Machizawa, M.G. Neural activity predicts individual differences in visual
working memory capacity. Nature 428, 748–751 (2004).
5. Vogel, E.K., McCollough, A.W. & Machizawa, M.G. Neural measures reveal individual
differences in controlling access to working memory. Nature 438, 500–503 (2005).
6. Rainer, G., Asaad, W.F. & Miller, E.K. Selective representation of relevant information by
neurons in the primate prefrontal cortex. Nature 393, 577–579 (1998).
7. Sakai, K. & Passingham, R.E. Prefrontal set activity predicts rule-specific neural
processing during subsequent cognitive performance. J. Neurosci. 26, 1211–1218
(2006).
8. Cowan, N. The magical number 4 in short-term memory: a reconsideration of mental
storage capacity. Behav. Brain Sci. 24, 87–185 (2001).
9. Conway, A.R. & Engle, R.W. Individual differences in working memory capacity: more
evidence for a general capacity theory. Memory 4, 577–590 (1996).
10. Awh, E., Vogel, E.K. & Oh, S.H. Interactions between attention and working memory.
Neuroscience 139, 201–208 (2006).
11. Awh, E. & Jonides, J. Overlapping mechanisms of attention and spatial working memory.
Trends Cogn. Sci. 5, 119–126 (2001).
12. Miller, E.K. & Cohen, J.D. An integrative theory of prefrontal cortex function. Annu. Rev.
Neurosci. 24, 167–202 (2001).
13. Sakai, K. & Passingham, R.E. Prefrontal interactions reflect future task operations.
Nat. Neurosci. 6, 75–81 (2003).
14. Alexander, G.E., DeLong, M.R. & Strick, P.L. Parallel organization of functionally
segregated circuits linking basal ganglia and cortex. Annu. Rev. Neurosci. 9,
357–381 (1986).
15. Monchi, O., Petrides, M., Strafella, A.P., Worsley, K.J. & Doyon, J. Functional role of the
basal ganglia in the planning and execution of actions. Ann. Neurol. 59, 257–264
(2006).
16. Dubois, B. & Pillon, B. Cognitive deficits in Parkinson’s disease. J. Neurol. 244,2–8
(1997).
17. Taylor, A.E. & Saint-Cyr, J.A. The neuropsychology of Parkinson’s disease. Brain Cogn.
28, 281–296 (1995).
18. Schneider, J.S. Basal ganglia role in behavior: importance of sensory gating and its
relevance to psychiatry. Biol. Psychiatry 19, 1693–1710 (1984).
19. Grabli, D. et al. Behavioural disorders induced by external globus pallidus dysfunction in
primates: I. Behavioural study. Brain 127, 2039–2054 (2004).
20. Lewis, S.J.G., Dove, A., Robbins, T.W., Barker, R.A. & Owen, A.M. Striatal contributions
to working memory: a functional magnetic resonance imaging study in humans. Eur . J.
Neurosci. 19, 755–760 (2004).
21. Postle, B.R. & D’Esposito, M. Dissociation of human caudate nucleus activity in spatial
and nonspatial working memory: an event-related fMRI study. Brain Res. Cogn. Brain
Res. 8, 107–115 (1999).
22. Menon, V., Anagnoson, R.T., Glover, G.H. & Pfefferbaum, A. Basal ganglia involvement
in memory-guided movement sequencing. Neuroreport 11, 3641–3645 (2000).
23. Mushiake, H. & Strick, P.L. Pallidal neuron activity during sequential arm movements.
J. Neurophysiol. 74, 2754–2758 (1995).
24. Goldman-Rakic, P.S. Regional and cellular fractionation of working memory. Proc. Natl.
Acad. Sci. USA 93, 13473–13480 (1996).
25. Frank, M.J., Loughry, B. & O’Reilly, R.C. Interactions between frontal cortex and basal
ganglia in working memory: a computational model. Cogn. Affect. Behav. Neurosci. 1,
137–160 (2001).
26. Gruber, A.J., Dayan, P., Gutkin, B.S. & Solla, S.A. Dopamine modulation in the basal
ganglia locks the gate to working memory. J. Comput. Neurosci. 20, 153–166
(2006).
27. Kane, M.J., Bleckley, M.K., Conway, A.R.A. & Engle, R.W. A controlled-attention view of
working-memory capacity: Individual differences in memory span and the control of
visual orienting. J. Exp. Psychol. Gen. 130, 169–183 (2001).
28. Kane, M.J. & Engle, R.W. Working-memory capacity and the control of attention: the
contributions of goal neglect, response competition and task set to Stroop interference.
J. Exp. Psychol. Gen. 132, 47–70 (2003).
29. Bleckley, M.K., Durso, F.T., Crutchfield, J.M., Engle, R.W. & Khanna, M.M. Individual
differences in working memory capacity predict visual attention allocation. Psychon.
Bull. Rev. 10, 884–889 (2003).
NATURE NEUROSCIENCE VOLUME 11
[
NUMBER 1
[
JANUARY 2008 107
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/natureneuroscience