Active maintenance in prefrontal area 46 creates distractor-resistant memory.

Abstract

How does the brain maintain information in working memory while challenged by incoming distractions? Using functional magnetic resonance imaging (fMRI), we measured human brain activity during the memory delay of a spatial working memory task with distraction. We found that, in the prefrontal cortex (PFC), the magnitude of activity sustained throughout the memory delay was significantly higher on correct trials than it was on error trials. By contrast, the magnitude of sustained activity in posterior areas did not differ between correct and error trials. The correlation of activity between posterior areas was, however, associated with correct memory performance after distraction. On the basis of these findings, we propose that memory representations gain resistance against distraction during a period of active maintenance within working memory. This may be mediated by interactions between prefrontal and posterior areas.

Full text

Page 1

Patients with prefrontal lesions show particularly impaired per-
formance on working memory tasks that include distractors1.
This is consistent with the view that a function of the PFC is to
protect memories from distraction2. In monkeys, sustained neur-
al activity during the memory delay of a working memory task
is taken as the neural correlate of actively maintained informa-
tion, and such activity is seen in the PFC3–6as well as in the pos-
terior association areas5,7,8. When distractors intervene during
the memory delay, the PFC sustains stimulus-selective activity2,
whereas the posterior association areas do not2,7,8.
These observations indicate that the PFC protects memory
against distraction, but how? One possibility is that a local neur-
al circuit within the PFC itself maintains information against dis-
traction. Modeling studies show that this can be achieved by
recurrent interaction of excitatory and inhibitory circuits9. Such
a system would provide distraction resistance within the main-
tenance component of working memory. Here, we propose
another mechanism—interaction between the PFC and posteri-
or association areas—that creates resistance against distraction.
According to our model, the PFC plays an executive role in trans-
forming the maintained information into a representation that
can survive distraction.
Normal human subjects performed a spatial working memo-
ry task in which they viewed, and tried to memorize, a sequence of
five spatial positions indicated by red squares on a screen (Fig. 1).
After a memory delay (during which no stimuli were presented),
the subjects performed a distractor task in which they viewed,
and tried to remember, a sequence of five blue dots. The distrac-
tor task required subjects to shift spatial attention to the posi-
tions of distractors, which made it impossible to mentally rehearse
the first sequence10,11. After making a recognition judgment
Active maintenance in prefrontal
area 46 creates distractor-resistant
memory
Katsuyuki Sakai1, James B. Rowe1and Richard E. Passingham1,2
1Wellcome Department of Cognitive Neurology, Institute of Neurology, 12 Queeen Square, London WC1N 3BG, UK
2Department of Experimental Psychology, University of Oxford, South Parks Road, Oxford OX1 3UD,UK
Correspondence should be addressed to K.S. (ksakai@fil.ion.ucl.ac.uk)
Published online: 15 April 2002, DOI: 10.1038/nn846
How does the brain maintain information in working memory while challenged by incoming distrac-
tions? Using functional magnetic resonance imaging (fMRI), we measured human brain activity dur-
ing the memory delay of a spatial working memory task with distraction. We found that, in the
prefrontal cortex (PFC), the magnitude of activity sustained throughout the memory delay was
significantly higher on correct trials than it was on error trials. By contrast, the magnitude of
sustained activity in posterior areas did not differ between correct and error trials. The correlation of
activity between posterior areas was, however, associated with correct memory performance after
distraction. On the basis of these findings, we propose that memory representations gain resistance
against distraction during a period of active maintenance within working memory. This may be
mediated by interactions between prefrontal and posterior areas.
about the blue dots, participants were tested on their memory
for the spatial order of the red squares. These ‘distractor-plus’
trials were intermixed with ‘distractor-minus’ trials, in which the
memory delay was immediately followed by a memory test for
the spatial sequence of red squares. In control trials, there was
no spatial order to remember (see Methods). We used event-relat-
ed fMRI to identify activity sustained during the memory delay.
Long and variable lengths for the memory-delay period (8–16 s)
allowed us to distinguish sustained activity from the phasic activ-
ity time-locked to viewing stimuli (red squares), viewing dis-
tractors (blue dots) and performing a memory test. We compared
fMRI data collected during the distractor-plus and distractor-
minus conditions with data collected during control conditions.
RESULTS
Behavioral results
Accuracy on the distractor task was 98% for both memory and
control conditions. On the memory test for the spatial order of
the red squares, the accuracy was 68% in the distractor-plus con-
dition (with 22% ‘forgotten’ responses, see Methods) and 97%
in the distractor-minus condition. Accuracy was not correlated
with the length of the memory delay (P > 0.1).
Participants were instructed not to move their eyes during the
memory delay (a rehearsal strategy that can aid in remembering
the locations). The extent and tracking distance of subjects’ eye
movements were minimal as compared to the extent of the array
of the five red squares (24° × 18° of visual angle), and did not dif-
fer significantly between the distractor-plus, distractor-minus and
control conditions (P >0.1; mean extent of eye movements, 0.58°,
0.52°and 0.61°of visual angle; mean tracking distance, 1.74°, 1.97°
and 1.55°of visual angle, respectively). During the memory delay,
articles
nature neuroscience • volume 5 no 5 • may 2002479
© 2002 Nature Publishing Group http://neurosci.nature.com

Page 2

saccadic eye movements occurred in only 2.2% and 2.5% of the
distractor-plus and -minus trials, respectively, and in 3.0%of con-
trol trials. Upon presentation of distractors, eye movements nat-
urally followed the order of the distractor locations.
Sustained activity on distractor-plus trials
We found sustained activity in two prefrontal regions during the
memory delay of correct trials in the distractor-plus condition
(P <0.05 corrected, compared to control condition). One region
was the mid-part of the middle frontal gyrus (area 46 [ref. 12];
Fig.2a), and the other was the posterior part of the superior frontal
sulcus just anterior to the precentral sulcus (area 8; Fig. 2b). There
was also sustained activity in the posterior part of the intraparietal
sulcus (IPS, Fig. 2cleft) as well as in other areas (Table 1).
Time course of sustained activity
On correct trials, the activity in area 46, area 8 and the IPS was sus-
tained throughout the memory delay, and expanded according to
the length of memory delay (Fig. 2, center panels). This finding is
consistent with previous unit recording studies in monkeys3,4,13.
On error trials, however, the activity in area 46 was not sustained
during the memory delay (Fig. 2a, right). The activity on error tri-
als was not significantly different from that on control trials
(P > 0.1) and was significantly less than that on correct trials
(P < 0.05). This finding is in accord with previous unit recording
studies4,6,14, although there was no distraction in those studies. By
contrast, the activity in area 8 and the IPS was sustained during the
memory delay even on error trials (Fig. 2b and c, right) and did
not differ significantly between correct and error trials (P >0.1).
Differential activity for correct and error trials
We investigated the times at which activity on correct and error
trials diverged. Trials with different lengths of the memory delay
were collapsed together and the time-series of BOLD signals
(see Methods) was realigned at the onset (x = 0 on the left pan-
els in Fig. 3) and offset (x = 0 on the right panels in Fig. 3) of the
memory delay. For area 46, the activity on correct trials (blue)
differed significantly from that on error trials (red) during the
period between 6 s after onset and 6 s after offset of the memory
delay (P < 0.05, Fig. 3a, see also Fig. 2a). Given the delayed tim-
ing of the BOLD signal compared with electrical activity, this
period functionally corresponds to the memory delay. On error
trials, area 46 showed a biphasic response: the first response
peaked at 4 s after the onset of the memory delay, that is, 8.75 s
after the presentation of the first stimulus, whereas the second
response peaked at 8 s after the offset of memory delay, that is,
8 s after the presentation of the first distractor (Fig. 3a, red). At
these peaks, the activity in area 46 did not differ significantly
between the correct and error trials (P > 0.1). In contrast, the
activity in area 8 and the IPS did not differ between the correct
and error trials at any time throughout the task epoch (P > 0.1,
Fig. 3b and c). A three-way ANOVA with factors of area, accu-
racy and time-bin confirmed the existence of a significant dif-
ference between the pattern of sustained activation in area 46
and that in area 8 and the IPS (P < 0.05).
Tight signal coupling on correct trials
Prefrontal areas 46 and 8 and the IPS are heavily interconnect-
ed15,16. To see how they interact, we first examined the correla-
tion (by calculating the correlation coefficient, r) of activity
between area 8 and the IPS (Fig. 4a). This correlation was sig-
nificantly higher for correct trials (blue) than for error trials (red).
Data were pooled from all the subjects (correct, r = 0.766,
n = 312; error, r = 0.562, n = 156; P < 0.01). The slope and inter-
cept of the regression lines did not differ: IPS = 0.690 × area 8 +
0.305 for correct trials and IPS = 0.672 × area 8 + 0.316 for error
trials. We further calculated the correlation coefficient for each
subject. Paired t-tests across subjects showed that the correlation
was significantly higher on correct trials than on error trials
(P < 0.01), indicating that it applies to the general population.
The tighter signal coupling between area 8 and the IPS on cor-
rect trials may be due to common inputs from area 46. The cor-
relation of activity between area 46, area 8 and the IPS supports
the idea. We found that when the signal in area 46 was high, the
signals in area 8 and IPS were tightly clustered (Fig. 4b), and this
was true even when data for correct trials (blue in Fig. 4b) were
considered alone. We divided the data for area 46 into two
halves—higher and lower activity. The correlation of signals in
area 8 and the IPS was significantly higher (P < 0.01) for the high-
er-activity half in area 46 (r = 0.909, n = 156) than for the lower-
Fig. 1. Schematic diagram of a distractor-plus trial. Subjects remembered
a sequence of five spatial positions that were indicated by red squares
(reproduced here as orange, but red in the scanner). After an unfilled
memory delay of 8–16 s (varied in steps of 2 s), they performed a distrac-
tor task in which they remembered positions of five blue dots and then
judged whether the position of a blue asterisk matched one of the posi-
tions of the blue dots. Immediately after the distractor judgment, subjects
saw an array of five boxes and an arrow and were asked to judge whether
the arrow replicated the spatial order of the red squares. Bottom, the tim-
ing of the task epochs relative to the onset of memory delay at time ‘0’.
480
nature neuroscience • volume 5 no 5 • may 2002
articles
Table 1. Sustained activation during the memory delay of
‘distractor-plus’ trials.
AreaSideCoordinates
y
46
54
t- value
xz
22
22
62
48
64
36
40
36
58
Prefrontal area 46
Prefrontal area 46
Prefrontal area 8
IPS/SPL
IPS/SPL
IPS
IPS
Premotor cortex
SMA
Inferior temporal cortex
Cerebellum
Right
Left
Right
Right
Left
Right
Left
Right
Right
Right
Left
406.49
6.38
8.25
13.5
9.66
9.83
11.23
9.88
14.67
9.63
9.56
–44
32
18
–18
34
–52
52
6
–62
–62
–50
–40
16
44
62 –44
–62
–22
–32 –24
IPS, cortex in the intraparietal sulcus; SPL, superior parietal lobule; SMA, sup-
plementary motor area.
© 2002 Nature Publishing Group http://neurosci.nature.com

Page 3

activity half in area 46 (r = 0.638, n = 156). This finding supports
the idea of a common input from area 46 to area 8 and IPS.
To what extent did the activation in area 46 predict the
memory performance? We plotted the accuracy of memory
performance as a function of the BOLD signal changes in area
46 (Fig. 5). For distractor-plus trials (filled circles), the accuracy
of performance fell to chance when activity in area 46 was low,
and improved with an increase in activity.
Sustained activity on distractor-minus trials
We also examined sustained activity in the distractor-minus con-
dition. For all brain regions, the activity in the distractor-minus
condition did not differ significantly from that in the distractor-
plus conditions when correct and error trials were taken together
(P > 0.1). This was expected, because the subjects were unable to
predict whether or not there would be distractors. The correlation
of the BOLD signals between area 8 and the IPS did not differ
between the distractor-minus and distractor-plus conditions when
correct and error trials were taken together (Fig. 4c, r = 0.684,
n = 492 for distractor-minus condition and r = 0.677, n = 468 for
distractor-plus condition, P > 0.1). Moreover, the distribution of
Fig. 2. Sustained activation during the memory
delay. (a) Prefrontal area 46 (x, y, z coordinates:
40, 46, 22). (b) Prefrontal area 8 (32, 6, 62).
(c) Intraparietal sulcus region (IPS; 18, –62, 48).
The left side of the images (or top in b) corre-
sponds to the left side of the brain. Data for
correct and error trials are shown separately in
the center and right panels, respectively (a–c).
The adjusted BOLD signal data from the activa-
tion peaks (blue cross-hairs) were temporally
realigned at the onset of the memory delay and
are shown as relative increase of BOLD signals
(y-axis) over time (x-axis) for each length of
memory delay (z-axis). Time ‘0’ on the x-axis
corresponds to the onset of memory delay.
the signals in area 46 on correct trials in the
distractor-minus condition (blue in
Fig. 4d) was similar to that on all the trials
in the distractor-plus condition (blue and
red in Fig. 4b). Our subjects were almost
always successful at recall in the distractor-
minus condition (accuracy 97%) and this
high accuracy was maintained regardless
of the amount of activity in area 46
(Fig. 5). Note again that, even in the dis-
tractor-minus condition, higher activation in area 46 was associat-
ed with tighter coupling of area 8 and the IPS (Fig. 4d). Thus
inter-regional interactions during the memory delay predicted suc-
cessful memory recall with distraction, but had no correlation with
performance on trials without distraction.
DISCUSSION
Our results support the view that the PFC is centrally involved
in protecting working memory representations against distrac-
tion1,2,17. In the condition with distractors, we found significant
activation in prefrontal area 46 during the memory delay on cor-
rect trials but not on error trials. Notably, it was the sustained
prefrontal activity before, not after2, presentation of the distrac-
tors that corresponded with resistance to distraction. Whereas
activity after distraction is taken to reflect on-line maintenance
of information that survives distraction, activity before distraction
may reflect executive processes that transform the maintained
information into distraction-resistant representations.
Prefrontal mechanisms to protect memory against distractors
could operate at various stages during memory tasks. Upon pre-
sentation of distractors, the PFC might be involved in filtering
articles
nature neuroscience • volume 5 no 5 • may 2002481
a
b
c
Fig. 3. Time course of activation for correct and error trials. (a) In prefrontal area 46, correct and error trials have divergent time courses of activation.
(b, c) In prefrontal area 8 and the intraparietal sulcus region (IPS), activation is similar for correct and error trials. The time series of the adjusted BOLD sig-
nal data were collapsed across different lengths of memory delay and were realigned at the onset (time ‘0’ on the left panel) and offset (time ‘0’ on the right
panel) of memory delay. For each 2-s bin, the mean and 95% confidence interval of the signals are shown separately for correct (blue) and error (orange)
trials. The red and blue shaded horizontal bars at the bottom of each graph indicate the time epoch for presentation of stimuli and distractors (see Fig. 1).
abc
© 2002 Nature Publishing Group http://neurosci.nature.com

Page 4

Fig. 4. Correlation of activation. (a, b) On the distractor-
plus trials, higher correlation was associated with correct
performance. (c, d) On the distractor-minus trials, correct
performance was achieved regardless of correlation. The
data for correct (blue circle) and error (orange cross) trials
are shown separately. For visualization purposes, the number
of data points for correct trials is matched to that for error
trials in the distractor-plus condition (n = 152 for each).
482
nature neuroscience • volume 5 no 5 • may 2002
out irrelevant information18,19. At retrieval of mem-
ory items, the PFC might be involved in selecting rel-
evant information to overcome interference effects
from the preceding trials20,21. In the present study,
we have shown that the PFC starts to operate before
presentation of the distractors.
As participants knew that there might be distrac-
tors after the memory delay, they may have been
preparing to switch tasks22–24; it has been shown pre-
viously that the PFC is involved task switching25,26.
In this case, one would expect to see a rise in PFC
activity at the time subjects would begin to anticipate
the new task. Our data show, however, that activity
in prefrontal area 46 remained high immediately after
the presentation of the spatial stimuli. We think it
unlikely that subjects started to prepare for the task
switch this early, as they knew it would be at least 8 s before onset
of the distractor task.
It is also unlikely that the activity in area 46 reflected simple
maintenance of information because subjects were successful in the
distractor-minus condition irrespective of the level of activation in
this area. Sustained activity in area 46 may not be necessary for sim-
ple maintenance in the absence distraction. This view is consistent
with previous imaging results that did not show sustained activa-
tion in area 46 when subjects simply maintained spatial informa-
tion without distraction27,28. By contrast, studies in monkeys have
shown that sustained activity in area 46 is associated with correct
performance even when there is no explicit distraction4,6,14. It is
possible that monkeys are more prone to distraction even without
distractors presented as such. Monkeys with prefrontal lesions per-
formed at 97% accuracy on a delayed response task in complete
darkness, whereas they performed at chance when the task was
administered in light conditions29. Furthermore, the number of
stimulus locations used in monkey studies is relatively small (two
to eight) and many trials are given in one session; thus there could
be high interference in memory across trials.
On error trials, prefrontal area 46 showed a normal phasic
response to the stimuli, but did not show sustained activity during
the memory delay. The phasic response to the stimuli is associated
with registration of the spatial information13. As subjects were
almost always successful on distractor-minus trials, it is clear that
they successfully registered or encoded the spatial order. In this
respect, our finding differs from previous studies30,31in which pha-
sic prefrontal activity at incidental encoding was associated with
subsequent performance on a recognition test given 30 minutes
later. Our study has shown that sustained prefrontal activation dur-
ing the memory delay of intentional remembering was associated
with success on a recall test given after 4.5 s of a distractor task. Both
studies suggest that PFC activity enhances the strength of memory.
In addition to area 46, we also found sustained activation in
the more posterior prefrontal area 8 and the IPS. These areas have
been consistently found to show sustained activity during a mem-
ory delay in imaging as well as in unit recording studies5,27,28,32.
Sustained activity in these areas may reflect simple maintenance
of information in working memory, without additional executive
processing. Unlike area 46, the activity in area 8 and the IPS did
not differ between correct and error trials in the distractor-plus
condition. Both prefrontal area 8 and the IPS have been shown to
be involved in covert shifts of attention33–35. Therefore, this find-
ing weakens the possibility that the subjects made errors simply
because they did not pay attention to the remembered positions.
We found that the correlation of the sustained activity between
area 8 and the IPS was associated with resistance against distrac-
tion. The degree of coupling between these areas was associated
with correct memory retrieval after distraction. This indicates that
the correlation of neural activity can be controlled independent-
ly of neural firing rates36. Increased correlation is thought to
strengthen information representations36,37. In the present study,
the tight coupling of area 8 and the IPS may reflect the robust-
ness of a memory representation in the face of distraction. Our
results also point to the possibility that area 46 controls the cou-
pling between area 8 and the IPS. Although activity in area 46 was
articles
Fig. 5. Increased accuracy of memory performance as a function of activity in pre-
frontal area 46 for distractor-plus trials (filled circles). On distractor-minus trials
(open circles), memory performance was maintained at a high level regardless of the
magnitude of prefrontal activation. Note that 50% accuracy is chance performance.
a
b
c
d
© 2002 Nature Publishing Group http://neurosci.nature.com

Page 5

not significantly correlated with activity in either area 8 or the IPS
alone (r = 0.281 and 0.279, respectively), it was associated with
tighter coupling between area 8 and IPS. This higher-order cor-
relation might indicate a modulatory role of area 46 rather than a
transmissive one17. Because our data are correlational, only an
intervention study could settle the direction of the causality.
What processes could underlie the memory maintenance seen
here? One possibility is that during the memory delay, participants
engaged in a process of active rehearsal, covertly shifting spatial
attention to the remembered target positions11. This would be
equivalent to repeatedly selecting representations for the appropri-
ate spatial locations. Selection may be achieved by a top-down sig-
nal from prefrontal area 46 to posterior association areas where the
memory representations are stored17,27,38,39. Repeated inputs from
area 46 may cause reverberation of activity within area 8 and the
IPS, thereby enhancing the memory representation during dis-
traction40. Alternatively, representations of spatial order may have
been re-organized or elaborated during the memory delay: the
sharpness of the spatial tuning of prefrontal neurons increases dur-
ing the memory delay6. It has also been shown that prefrontal
area 46 is involved in the elaboration of maintained information,
which subsequently leads to better memory performance41.
Our findings support a marked difference between the nature
of the sustained activity in prefrontal area 46 and that of other areas.
We suggest the term ‘active maintenance’17to distinguish activity
in area 46 from activity in other areas that may be associated with
‘simple maintenance’. We propose that active maintenance makes
demands on executive processes, consistent with claims that pre-
frontal area 46 has an executive role in memory42–44. We further
propose that the mechanism by which these executive processes are
carried out involves higher-order interactions between prefrontal
and posterior association areas45. In the face of distraction, these
interactions are essential for correct memory performance.
METHODS
Subjects. Fourteen normal right-handed volunteers (age range 20–40 years,
seven males, seven females) gave written informed consent to participate in
the study. The study was approved by the joint ethics committee of the Insti-
tute of Neurology and University College London Hospital, London, UK.
Behavioral task. Participants were asked to remember the spatial posi-
tions of a sequence of five red squares, which were presented on a screen
for 750 ms each with an inter-stimulus interval of 250 ms. The array of the
five squares subtended a visual angle of 24°×18°(Fig. 1). This was followed
by an unfilled memory delay, lasting 8–16 s (varied in steps of 2 s), during
which subjects were instructed not to rehearse the remembered locations
by moving their eyes. On ‘distractor-plus’ trials, a spatial distractor task fol-
lowed the delay: five blue dots were presented successively for 500 ms each
with an interstimulus interval of 250 ms. Then, after 500 ms, participants
were shown an asterisk for 500 ms and made a button press with the right
index or middle finger to report whether or not a blue circle had appeared
at the position of the asterisk. Subjects were explicitly instructed to perform
the distractor task as accurately as possible.
One second after the distractor task, participants were tested on their
memory for the spatial order of the red squares. We presented an array of
five boxes that matched the positions of the five red squares, with an
arrow pointing from one box to another (Fig. 1). Participants respond-
ed with a button press (again using the right index or middle finger) to
report whether or not the arrow indicated the correct spatial order of the
red squares. They pressed a button with the right ring finger if they had
forgotten the order. The distractor task was demanding enough to prevent
people from mentally rehearsing the spatial order of red squares. This
was confirmed by a pilot experiment in which three subjects were asked
to continue the rehearsal of red squares while performing the distractor
task. With an accuracy of 95% on the memory task, performance on the
distractor task deteriorated nearly to chance (56%).
Intermixed with these distractor-plus trials were memory trials without
distractors (‘distractor-minus’ trials). On these trials, the unfilled memory
delay was followed immediately by a memory test for the spatial order of
red squares. The conditions were otherwise identical to those in ‘distrac-
tor-plus’ trials. We also presented control trials in which five red squares
appeared in the same position at the center of the screen. After a delay, sub-
jects were shown an array of five squares, one at the center and the others at
the four corners of the screen. They pressed a button with their index finger
if the square at the center was indicated by an arrow and with the middle
finger if a square at the corner was indicated by an arrow. In half of the con-
trol trials, a distractor task came after the delay, as in ‘distractor-plus’ trials.
After participants had practiced the task for 15 min, imaging began.
Eighteen distractor-plus trials, 18 distractor-minus trials and 36 control
trials (18 with distractors and 18 without distractors) were randomly
presented with inter-trial intervals of 8–16 s, which were varied in steps
of 2 s. For each trial we used different arrays of five boxes. Throughout the
task, the subjects’ eye movements were recorded using an infrared eye-
tracking system (Model 504LRO, Applied Science Laboratories, Bedford,
Massachusetts) that detected eye movements larger than 0.25° of visual
angle at a sampling rate of 240 Hz. We measured the extent of eye move-
ments during the memory delay by calculating the diameter of the min-
imum circle that covered all eye-traces. We also measured the eye-tracking
distance, or the total distance of eye movements, and the frequency of
saccadic eye-movements (velocity > 30°/s)46during the memory delay.
fMRI imaging. Imaging was performed using a 2-Tesla scanner (Siemens
Vision, Erlagen, Germany). The functional images sensitive to blood oxy-
genation level–dependent (BOLD) contrasts were acquired by T2*-weight-
ed echo planar imaging (TR = 4.5 s, TE = 40 ms, 540 sequential whole
brain volume acquisitions, 64 × 64 × 48 voxels at 3 mm isotropic resolu-
tion). The onset of each task trial relative to the preceding image acquisition
was jittered in steps of 0.75 s within 1 TR (4.5 s). High-resolution struc-
tural T1-weighted MPRAGE images (TR = 9.5 s, TE = 4 ms, TI = 600 ms,
voxel size 1 ×1 ×1.5 mm, 108 axial slices) were also acquired on all subjects.
Data analysis. We used SPM99 software (http://www.fil.ion.ucl.ac.uk/
spm) for image processing and analysis. The first five volumes of images
were discarded to allow for T1 equilibration. The remaining 535 image
volumes were realigned to the first image, sinc-interpolated over time to
correct for phase advance during volume acquisition, and normalized to
the Montreal Neurological Institute (Montreal, Canada) reference brain.
The data were spatially smoothed with a Gaussian kernel (10 mm, full-
width at half-maximum). Statistical parametric maps of t-statistics were
calculated for condition-specific effects within a general linear model.
Sustained activity was modeled as an epoch time-locked to the start of
the memory delay, with duration matched to the length of the memory
delay. We created separate models for correct and error trials for each of
the distractor-plus, distractor-minus and control conditions. Error tri-
als were those for which subjects gave an incorrect response or reported
that they could not remember. The trials in which subjects made errors
in the distractor task were also modeled separately. The model included
separate covariates for transient activation in response to the presenta-
tion of each stimulus, distractor and memory test. All events were con-
volved with a canonical hemodynamic response function. The data were
high-pass filtered with a frequency cutoff at 100 s.
We performed a random effects analysis. Images of parameter esti-
mates for the contrast of interest were created for each subject (first-level
analysis), and were then entered into a second-level analysis using a one-
sample t-test across the subjects. To equate the weighting of each subject
contributing to the second-level analysis, we selected 12 subjects who
made at least 10 correct responses and at least 5 error responses in the
distractor-plus condition. For each subject, 10 correct trials and 5 error
trials were randomly chosen for the statistical comparisons; thus, approx-
imately 120-s memory-delay epochs for correct trials and 60-s memo-
ry-delay epochs for error trials were chosen for each subject.
Identification of activation foci.First, we identified areas that showed sus-
tained activity during the memory delay. We made comparisons between
the memory delay of the correct trials in the distractor-plus condition and
the delay of the control condition (P <0.05 corrected; t11> 8.95). Because
articles
nature neuroscience • volume 5 no 5 • may 2002483
© 2002 Nature Publishing Group http://neurosci.nature.com

Page 6

484
nature neuroscience • volume 5 no 5 • may 2002
of our prior hypothesis regarding the role of the PFC, we also corrected for
a reduced search volume within the lateral PFC (t11> 5.50). Then we used
this activation map as an inclusive mask. Within this mask, we made com-
parisons between correct and error trials in the distractor-plus condition,
and also between the error trials in this condition and all trials in the con-
trol condition. The contrasts were thresholded at P <0.05, corrected for a 10-
mm-radius spherical search volume centered at the peaks of activation foci
shown in Table 1 (t11> 3.55). We also compared the memory delay of cor-
rect trials in the distractor-minus condition and the memory delay of all
the trials in the distractor-plus condition (P <0.05 corrected).
Time course of activation. The time course of the BOLD signals was
realigned at the onset and offset of the memory delay and was re-sam-
pled in 2-s time bins. The signals within each bin were then averaged
across the trials for the 12 subjects (120 correct trials and 60 error tri-
als). The signals for the correct and error trials were compared for each
bin using an unpaired t-test.
Correlation of activation.The signals during the period of sustained acti-
vation, that is, between 6 s after the onset and 6 s after the offset of the
memory delay, were collected from the 10 correct and 5 error trials for
each of the 12 subjects. We obtained 312 and 156 data points for correct
and error trials, respectively, in the distractor-plus condition and 492
data points for correct trials in the distractor-minus condition. We cal-
culated Pearson’s product-moment correlation coefficients (r) to esti-
mate the linear correlation of the signals. The difference in the strength of
correlation between the correct and error trials was tested by transforming
the correlation coefficients into z-scores (Fisher’s z-transformation) and
comparing them using a χ2-test.
Acknowlegments
We are grateful to C. Frith, M. Rugg and R. Frackowiak for comments. This
study was supported by the Wellcome Trust. K.S. was supported by the Human
Frontier Science Program.
Competing interests statement
The authors declare that they have no competing financial interests.
RECEIVED 2 JANUARY; ACCEPTED 4 MARCH 2002
1. D’Esposito, M. & Postle, B. R. The dependence of span and delayed-response
performance on prefrontal cortex. Neuropsychologia 37, 1303–1315 (1999).
2. Miller, E. K., Erickson, C. A. & Desimone, R. Neural mechanisms of visual
working memory in prefrontal cortex of the macaque. J. Neurosci. 16,
5154–5167 (1996).
3. Fuster, J. M. Unit activity in prefrontal cortex during delayed-response
performance: neuronal correlates of transient memory. J. Neurophysiol. 36,
61–78 (1973).
4. Funahashi, S., Bruce, C. J. & Goldman-Rakic, P. S. Mnemonic coding of
visual space in the monkey’s dorsolateral prefrontal cortex. J. Neurophysiol.
61, 331–349 (1989).
5. Chafee, M. V. & Goldman-Rakic, P. S. Matching patterns of activity in
primate prefrontal area 8a and parietal area 7ip neurons during a spatial
working memory task. J. Neurophysiol. 79, 2919–2940 (1998).
6. Sawaguchi, T. & Yamane, I. Properties of delay-period neuronal activity in the
monkey dorsolateral prefrontal cortex during a spatial delayed matching-to-
sample task. J. Neurophysiol. 82, 2070–2080 (1999).
7. Miller, E. K., Li, L. & Desimone, R. Activity of neurons in anterior inferior
temporal cortex during a short-term memory task. J. Neurophysiol. 13,
1460–1478 (1993).
8. Constantinidis, C. & Steinmetz, M. A. Neuronal activity in posterior parietal
area 7a during the delay periods of a spatial memory task. J. Neurophysiol. 76,
1352–1355 (1996).
9. Compte, A., Brunel, N., Goldman-Rakic, P.S. & Wang, X. J. Synaptic
mechanisms and network dynamics underlying spatial working memory in a
cortical network model. Cereb. Cortex 10, 910–923 (2000).
10. Smyth, M. M. Interference with rehearsal in spatial working memory in the
absence of eye movements. Q. J. Exp. Psychol. A 49, 940–949 (1996).
11. Awh, E. & Jonides, J. Overlapping mechanisms of attention and spatial
working memory. Trends Cogn. Sci. 5, 119–126 (2001).
12. Rajkowska, G. & Goldman-Rakic, P. S. Cytoarchitectonic definition of
prefrontal areas in the normal human cortex: variability in locations of areas
9 and 46 and relationship to the Talairach coordinate system. Cereb. Cortex 5,
323–337 (1995).
13. Kojima, S. & Goldman-Rakic, P. S. Delay-related activity of prefrontal
neurons in rhesus monkeys performing delayed response. Brain Res. 248,
43–49 (1982).
14. Funahashi, S., Inoue, M. & Kubota, K. Delay-period activity in the primate
prefrontal cortex encoding multiple spatial positions and their order of
presentation. Behav. Brain Res. 84, 203–223 (1997).
15. Petrides, M. & Pandya, D. N. Projections to the frontal cortex from the
posterior parietal region in the rhesus monkey. J. Comp. Neurol.228, 105–116
(1984).
16. Cavada, C. & Goldman-Rakic, P. S. Posterior parietal cortex in rhesus
monkey: II. Evidence for segregated corticocortical networks linking sensory
and limbic areas with the frontal lobe. J. Comp. Neurol. 287, 422–445 (1989).
17. Miller, E. K. & Cohen, J. D. An integrative theory of prefrontal cortex
function. Annu. Rev. Neurosci. 24, 167–202 (2001).
18. Chao, L. L. & Knight, R. T. Human prefrontal lesions increase distractibility
to irrelevant sensory inputs. Neuroreport 6, 1605–1610 (1995).
19. Guillery, R. W., Feig, S. L., & Lozsadi, D. A. Paying attention to the thalamic
reticular nucleus. Trends Neurosci. 21, 28–32 (1998).
20. D’Esposito, M., Postle, B. R., Jonides, J. & Smith, E. E. The neural substrate
and temporal dynamic of interference effects in working memory as revealed
by event-related functional MRI. Proc. Natl. Acad. Sci. USA 96, 7514–7519
(1999).
21. Bunge, S. A., Ochsner, K. N., Desmond, J. E., Glover, G. H. & Gabrieli, J. D. E.
Prefrontal regions involved in keeping information in and out of mind. Brain
124, 2074–2086 (2001).
22. Rogers, R. D. & Monsell, S. Costs of a predictable switch between simple
cognitive tasks. J. Exp. Psychol. Gen. 124, 207–231 (1995).
23. Gopher, D., Armony, L. & Greenshpan, Y. Switching tasks and attention
policies. J. Exp. Psychol. Gen. 129, 308–339 (2000).
24. Rubinstein, J. S., Meyer, D. E. & Evans, J. E. Executive control of cognitive
processes in task switching. J. Exp. Psychol. Hum. Percept. Perform. 27,
763–797 (2001).
25. D’Esposito, M. et al. The neural basis of the central executive system of
working memory. Nature 378, 279–281 (1995).
26. Sohn, M-H., Ursu, S., Anderson, J. R., Stenger, V. A. & Carter, C. S. The role of
prefrontal cortex and posterior parietal cortex in task switching. Proc. Natl.
Acad. Sci. USA 97, 13448–13453 (2000).
27. Rowe, J. B., Toni, I., Josephs, O., Frackowiak, R. S. J. & Passingham, R. E. The
prefrontal cortex: response selection or maintenance within working
memory? Science 288, 1656–1660 (2000).
28. Rowe, J. B. & Passingham, R. E. Working memory for location and time:
activity in prefrontal area 46 relates to selection rather than maintenance in
memory. Neuroimage 14, 77–86 (2001).
29. Malmo, R. B. Interference factors in delayed response in monkeys after
removal of frontal lobes. J. Neurophysiol. 5, 295–308 (1942).
30. Brewer, J. B., Zhao, Z., Desmond, J. E., Glover, G. H. & Gabrieli, J. D. E.
Making memories: brain activity that predicts how well visual experience will
be remembered. Science 281, 1185–1187 (1998).
31. Wagner, A. D. et al. Building memories: remembering and forgetting of
verbal experiences as predicted by brain activity. Science 281, 1188–1191
(1998).
32. Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G. & Haxby, J. V. An
area specialized for spatial working memory in human frontal cortex. Science
279, 1347–1351 (1998).
33. Rosen, A. C. et al. Neural basis of endogenous and exogenous spatial
orienting: a functional MRI study. J. Cognit. Neurosci. 11, 135–152 (1999).
34. Corbetta, M., Kincade, J. M., Ollinger, J. M., McAvoy, M. P. & Shulman, G. L.
Voluntary orienting is dissociated from target detection in human posterior
parietal cortex. Nat. Neurosci. 3, 292–297 (2000).
35. Perry, R. J. & Zeki, S. The neurology of saccades and covert shifts in spatial
attention. An event-related fMRI study. Brain 123, 2273–2288 (2000).
36. Salinas, E. & Sejnowski, T. Correlated neuronal activity and the flow of neural
information. Nat. Rev. Neurosci. 2, 539–550 (2001).
37. Varela, F., Lachaux, J. -P., Rodriguez, E. & Martinerie, J. The brainweb: phase
synchronization and large-scale integration. Nat. Rev. Neurosci. 2, 229–239
(2001).
38. Frith, C. in Control of Cognitive Processes: Attention and Performance XVIII(eds.
Monsell, S. & Driver, J.) 549–565 (MIT Press, Cambridge, Massachusetts, 2000).
39. Miyashita, Y. & Hayashi, T. Neural representation of visual objects: encoding
and top-down activation. Curr. Opin. Neurobiol. 10, 187–194 (2000).
40. Wang, X. J. Synaptic reverberation underlying mnemonic persistent activity.
Trends Neurosci. 24, 455–463 (2001).
41. Wagner, A. D., Maril, A., Bjork, R. A. & Schacter, D. L. Prefrontal
contributions to executive control: fMRI evidence for functional distinctions
within lateral prefrontal cortex. Neuroimage 14, 1337–1347 (2001).
42. Petrides, M. in The Prefrontal Cortex (eds. Roberts, A. C., Robins T. W. &
Weiskrantz, L.) 103–116 (Oxford Univ. Press, Oxford, UK, 1998).
43. Owen, A. M. et al. Redefining the functional organization of working
memory processes within human lateral prefrontal cortex. Eur. J. Neurosci.
11, 567–574 (1999).
44. D’Esposito, M., Postle, B. R., Ballard, D. & Lease, J. Maintenance versus
manipulation of information held in working memory: an event-related
fMRI study. Brain Cogn. 41, 66–86 (1999).
45. Engel, A. K., Fries, P. & Singer, W. Dynamic predictions: oscillations and
synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716 (2001).
46. Kato, M. et al. Eye movements in monkeys with local dopamine depletion in
the caudate nucleus. I. Deficits in spontaneous saccades. J. Neurosci. 15,
912–927 (1995).
articles
© 2002 Nature Publishing Group http://neurosci.nature.com