Content uploaded by Aina Puce
Author content
All content in this area was uploaded by Aina Puce
Content may be subject to copyright.
Activation of Human Prefrontal Cortex
during Spatial and Nonspatial Working
Memory Tasks Measured by Functional
MRI
Gregory McCarthy,
123
Aina Puce,
12
R. Todd Constable,
4
John H. Krystal,
5
John C. Gore
4
and Patricia Goldman-Rakic
6
l
Neuropsychology Laboratory, VA Medical Center, West Haven,
CT 06516, and Departments of
2
Surgery (Neurosurgery),
3
Neurology,
4
Diagnostic Radiology,
5
Psychiatry, and
^eurobiology, Yale University School of Medicine, New
Haven, CT 06510, USA
Separate working memory domains for spatial location, and for
objects, faces, and patterns, have been identified in the prefrontal
cortex (PFC) of nonhuman primates. We have used functional
magnetic resonance imaging to examine whether spatial and
nonspatial visual working memory processes are similarly
dissociable in human PFC. Subjects performed tasks which required
them to remember either the location or shape of successive visual
stimuli.
We found that the mnemonic component of the working
memory tasks affected the hemispheric pattern of PFC activation.
The spatial (LOCATION) working memory task preferentially
activated the middle frontal gyrus (MFG) in the right hemisphere,
while the nonspatial (SHAPE) working memory task activated the
MFG in both hemispheres. Furthermore, the area of activation in
the left hemisphere extended into the inferior frontal gyrus for the
nonspatial SHAPE task. A perceptual target (DOT) detection task
also activated the MFG bilaterally, but at a level approximately half
that of the working memory tasks. The activation in the MFG
occurred within 3-6 s of task onset and declined following task
offset. Time-course analysis revealed a different pattern for the
cingulate gyrus, in which activation occurred upon task completion.
Cingulate activation was greatest following the
SHAPE
task and was
greater in the left hemisphere. The present results support the
prominent role of the PFC and, specifically, the MFG in working
memory, and indicate that the mnemonic content of the task affects
the relative weighting of hemispheric activation.
Introduction
Working memory is a system for the transient storage and
processing of information (Baddeley, 1992). The functional
anatomy of working memory has been investigated in
physiological studies in nonhuman primates, and such studies
have demonstrated that the dorsolateral prefrontal cortex (dPFQ
is involved in short-term mnemonic coding (Fuster, 1973;
Funahashi et aL, 1989). Goldman-Rakic (1987) proposed an
architecture for working memory in which different subregions
of the dPFC maintain on-line representations of different
informational domains (such as shape and object features) in
memory through control over reciprocally interconnected brain
regions or networks. This model of compartmentalized working
memory systems has been supported by recent physiological
data acquired from monkeys engaged in delayed response tasks
requiring memory for location (Funahashi et
aL,
1989), features
of objects, and faces (Wilson et aL, 1993). Neurons in the PFC
that are responsive during the performance of nonspatial
working memory are located more ventrally than those
responsive during the performance of spatial working memory
tasks (Wilson etaL, 1993).
The neural basis for working memory in humans has been
investigated by neuroimaging studies using both positron
emission tomography (PET) and functional magnetic resonance
imaging (fMRI). A recent review of this literature (McCarthy,
1995) has identified consistent activation of the
PFC
among tasks
in which short-term memory storage has been manipulated,
although there have been inconsistencies with regard to the
specific prefrontal regions activated and their hemispheric
lateralization. For example, Petrides et
al.
(1993a) used PET to
demonstrate activation of the right middle frontal gyrus in a task
requiring memory for spatial patterns, and Cohen et
aL
(1994)
used fMRI to demonstrate bilateral activation of dPFC in a
letter-monitoring task. We have previously used fMRI to
demonstrate activation of the right middle frontal gyrus in a task
requiring short-term storage of spatial locations (McCarthy et al.,
1994),
and Smith et
aL
(1995) have also reported activation of
the right
MFG
in a task which engaged memory for the location
of visual forms. However, working memory for stimulus location
has not invariably led to MFG activation (cf. Jonides et
al.,
1993;
Courtney etaL,
1995;
Smith et al., 1995). In addition, the pattern
of hemispheric activation of the PFC, and the extent of ventral
activation has varied among studies (e.g., Jonides et aL, 1993;
Paulesu et
al.,
1993; Petrides et aL, 1993b; Smith et al., 1995,
1996).
The degree to which differential activation of subregions of
frontal cortex reflects the memory domain manipulated has
important implications for the neural organization of cognitive
processing in humans (Goldman-Rakic, 1987) and for the
structure of working memory (Baddeley, 1992). Here we have
employed fMRI to investigate activation of the PFC in subjects
performing tasks requiring either the transient storage of the
location or
the
shape of visual stimuli, as exemplars of spatial and
nonspatial working memory, respectively. The pattern of
activation, hemispheric symmetry, and temporal course of
activation within subregions of the PFC was investigated.
Materials and Methods
Subjects
Ten neurologically normal, right-handed subjects (aged 23-42 years;
seven males, three females) participated in the primary study and
performed both spatial and nonspatial working memory tasks. All
subjects had prior experience in fMRI studies. The experimental protocol
was approved by the Human Investigation Committee of Yale University
School of Medicine.
At the conclusion of the study, four additional right-handed subjects
(aged 25-34
years;
one male, three females) participated in a second study
of either spatial or nonspatial working memory to test the stability of the
group activation time-course (described below) and the influence of the
image acquisition sequence upon activation. One subject was tested in
both studies.
Working Memory Tasks
Visual stimuli were delivered under computer control to an active matrix
LCD panel whose images were back-projected onto a translucent
Plexiglas screen mounted on the patient gurney of a whole-body MRI
system. The subject viewed stimuli on the screen through a prism mirror
mounted in the head coil. All stimuli were colored white and were
presented against a dark background with a central fixation cross. All
Cerebral Cortex July/Aug 1996;6:6OO-6ll; 1047-3211/96/*4.00
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
behavioral responses were made with a fiber-optic response box on
which subjects depressed a button with the index finger of their right
hand.
Each experimental run lasted 96 s, during which functional images
were acquired every 1.5 s (see below). Each run consisted of
a
pre-task
baseline interval (27 s duration), a task interval in which subjects
performed a working memory or control task (27 or 28.5 s), and a
post-task interval (40.5 or 42 s). Subjects were instructed to maintain
fixation and minimize blinking during the run.
The experimental design is illustrated in Figure 1. There were three
experimental tasks: LOCATION, SHAPE, and
DOT.
The LOCATION and
SHAPE tasks were memory-guided in that subjects were obliged to
remember the location or shape, respectively, of each stimulus to
perform the tasks accurately. In contrast, the DOT task was a perceptual
target detection and vigilance task that did not require memory of
previous stimuli. All three tasks, however, required subjects to fixate, to
attend to visual stimuli, and to make two or three button presses per run.
Subjects were practiced on all tasks prior to entering the magnet. The
requirement for central eye fixation was stressed during these practice
runs and throughout the imaging sessions.
In the LOCATION task, a single square appeared in a spatial location
randomly chosen from among
20
possible
locations.
These locations were
chosen to avoid obvious nameable patterns such as a clockface. After a
1.25s exposure, the square disappeared for
0.25
s and was then replaced
by another square at another randomly chosen location. The subject was
instructed to respond with
a
button press whenever
a
square appeared in
a location previously occupied during that run (a 'target'). A run
consisted of 18 or 19 stimuli which included two or three targets. The
exact number of targets and their position within the run were randomly
determined. After the last stimulus was presented, the number of targets
presented and the number of correct responses and reaction time were
displayed at the center of the screen and remained visible during the
post-task interval.
The SHAPE and DOT tasks used identical timing to the LOCATION
task, and presented the subject with the same numbers of stimuli and
targets. In the SHAPE task, a single irregular shape was randomly chosen
from among 20 possible shapes and appeared at the center of the screen.
After 1.25 s, the shape disappeared and was replaced by another
randomly chosen shape 0.25 s later. The subject was instructed to
respond with a button press whenever a shape appeared which had
previously appeared during that run. The shapes were the same as those
used by McCarthy et aL (1994) and were chosen not to
resemble
obvious
objects such as circles or squares. The shapes were sized to fit within the
bounding limits of the squares used in the LOCATION task.
SHAPE LOCATION DOT
Figure 1. Schematic illustration of the
SHAPE,
LOCATION,
and DOT tasks (see text for
details).
Each
run consisted of 18 or 19 stimuli presented at
a
rate of
one
every 1.5
s
and
containing two or three randomly distributed targets. The stimuli in the
SHAPE
and DOT
tasks were presented in the center of the
screen,
while the stimuli in the
LOCATION
task
zy&zxeA at randomly chosen locations.
For
convenience,
target stimuli for
each
task are
shown as the fourth stimulus in each sequence.
In the
DOT
task, a single square appeared at the center of the screen.
After 1.25 s, the square disappeared for
0.25
s
and then reappeared, in the
manner of the SHAPE task. During the appearance of two or three
randomly chosen squares, a single pixel brightened for
a
100
ms
interval.
The location of the pixel within the square, and the timing of its
appearance were randomly determined. Subjects were instructed to
press a button whenever they detected such a target event. Each square
was identical in size to those used in the LOCATION task.
Each imaging session consisted of 12 experimental runs in which
each of the three tasks was replicated four
times.
The experimental runs
were blocked, but the task order was balanced across subjects. For the
four subjects of the second study, only a single working memory task
(SHAPE or LOCATION) was used and was replicated 16 times in a single
imaging session.
MRI Studies
A 1.5 T MRI scanner (General Electric Signa, Milwaukee WD with a
quadrature head coil and echoplanar capability (Instascan, ANMR
Systems Inc., Wilmington MA) was used. The subject's head was
immobilized using a vacuum cushion and a Velcro forehead strap.
Anatomical sagittal localizer scans were acquired (Ti-weighted:
TR
=
500,
T
E
-
11,
NEX
=
1,
FOV =
24 cm, slice thickness
=
5 mm, skip
=
2.5 mm,
imaging matrix
256 » 192)
to identify the anterior
(AC)
and posterior (PC)
commissures. Four Ti-weighted coronal scans
(TR
=
500,
TE
=
11,
NEX =
2,
FOV =
40 cm, skip
=
0 mm, slice thickness
=
7 mm, imaging matrix 256 *
192) were then acquired centered 4 cm anterior to the AC (measured
along the AC-PC line) to encompass the same prefrontal cortical regions
previously shown active in working memory tasks (McCarthy et ai,
1994) and to be used for co-registration with the functional images.
Coronal magnetic resonance angiography (MRA) images were also
obtained to identify the major venous vasculature in this region
(TR
=45,
T
E
= 7.7, a = 40°, NEX = 1, FOV = 24 cm, flow compensation, slice
thickness 2 mm, imaging matrix 256 * 128).
Functional images were acquired using a gradient-echo echoplanar
image acquisition sequence (T
R
=
1500, T
E
=
45,
a
=
60°,
NEX =
1,
FOV =
40
x
20 cm, slice thickness
= 7
mm, skip factor
=
0, imaging matrix 256 *
128).
Images were acquired for each of the four anatomical coronal
images described above. Each run consisted of the acquisition of 64
images
for each of these four
slices.
Individual
images
were acquired in 67
ms and the interval between successive acquisitions of the same
anatomical slice was 1.5 s (yielding
a
total imaging time of 96
s).
Each run
was preceded by four radiofrequency excitations to achieve steady-state
transverse magnetization. In each 64-image acquisition, the task
commenced at image 18.
In addition to the gradient-echo echoplanar imaging described above,
the four subjects of the secondary study were also studied using
spin-echo
(TR
=
1500, T
E
=
120 or 100,
NEX =
1, FOV
=
40 x
20
cm, slice
thickness
=
7 mm, skip factor
=
0, imaging matrix 256
x
128) echoplanar
image acquisition. Eight experimental runs were performed with each
imaging sequence.
Analysis of MRI Data
The images comprising each experimental run were examined for head
movement and acquisition artifacts by plotting the center of
mass
of each
image in each run and by animating the image time series. Individual
images that showed obvious acquisition artifacts were eliminated from
further
analysis.
Two primary methods of
analysis
were used to emphasize individual
and group effects. Data from individuals were analyzed by computing
{-tests
for each voxel's signal intensity comparing pre-task and task
intervals. The Mests were computed separately for each of the four
replications of each task. Images 3-20 comprised the pre-task baseline
analysis period, while images 23-43 comprised the task analysis interval.
These analysis periods were offset from the start and end of the task
period to accommodate the -6 s rise time and somewhat slower fall time
of the task-related
MR
signal changes
as
observed previously (McCarthy et
aL,
1994). The resulting Mmages of the first and second, and third and
fourth task replicates were then averaged. These two average f-images
were then subjected to a logical 'and' procedure
('split
Mest') such that
only voxels which exceeded
a
t-value of 1.5 (or were less than
a
t-value
of
-1.5) in both /images were retained (Schneider et
aL,
1993; Puce et al.
Cerebral Cortex July/Aug 1996,
V
6 N 4 601
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
Figure 2. Activated voxels (color overlay) from the LOCATION (ajb), SHAPE (c,rf), and DOT [ef) tasks in an individual subject identified by the split-f procedure described in Materials
and Methods. Each pixel indicates
a
significant increase in activation (t > 2) during the task interval relative to the pre-task
baseline.
For
these
pixels,
activation decreased by at least
50%
by the conclusion of the post-task interval. Activation data are overlaid over their corresponding anatomical Ti-weighted images. The two adjacent slices shown were those
which contributed to the spatially normalized group average shown
in
Figure 3. In this and all subsequent
figures,
the right hemisphere is shown on the left side of the images.
1995).
Finally, the time-course of activation for each 'significant' voxel
was evaluated to determine whether its activation was due to low-
frequency physiological noise or other sources of drift. Our experience
has shown that drift can falsely inflate the number of significant /-values
obtained when comparing long duration baseline and task intervals
(especially when the comparison intervals occur consecutively in fixed
order). Voxels for which mean intensity change during the task interval
did not fall by at least 50% by the last 15 s of the post-task interval were
eliminated from further consideration. This imposition of an expected
activation time-course has been used previously in fMRI studies (e.g.,
Bandettiniefai, 1993; McCarthy
etal,
1993).
Group effects were evaluated by spatially normalizing the raw images
from each subject and averaging them into a group mean series for each
task (McCarthy et al, 1994). A representative subject was selected whose
coronal images were most similar to those of the atlas of Talairach and
Toumoux (1988). Because of variability in head size across subjects, only
two contiguous images (from among the four acquired) could be
identified
as
anatomical matches in all 10 subjects. Two blinded operators
independently superimposed each subject's images upon the corres-
ponding images from the representative subject using translation and
rotation, and stretching (independently in two dimensions). Particular
care was used to align the midline, and the superior and inferior frontal
sulci. As the frontal sinuses caused some signal loss at the inferior surface
of these slices, the region of the gyrus rectus and the orbital frontal gyms
was less well aligned. The two operators jointly reviewed all
superimpositions and resolved differences in alignment. The spatially
normalized group images were evaluated by comparing pre-task and task
analysis intervals by f-test with the further requirement of
a
50%
drop in
signal intensity by the end of the post-task interval as described above.
As a further test, the group mean time-series for each task were
interrogated by drawing anatomical regions of interest (ROI) over the
superior frontal gyri (SFG), middle frontal gyri (MFG), inferior frontal gyri
(IFG),
cingulate gyri (Cing), and white matter control regions, and
integrating the
MR
signal intensity. The mean signal intensity and standard
deviation were calculated for the pre-task analysis interval across the
three tasks for each anatomical
ROI.
The integrated intensity for each
ROI
in each image in the series was then converted to a z-score deviation from
this common pre-task value so that the task effects would be in
comparable units expressing statistical distance in standard deviations
from common baseline 'noise'.
The mean activation time-courses produced by the group analysis
were then used in a second study to test further the reliability of the
findings, hi this analysis, the group mean activation time-course obtained
from the anatomical region of interest analysis described above was
convolved with the time-series for each voxel from the data acquired in a
series of four new subjects. This technique identified voxels which had a
similar time-course to the group activation time-course.
Results
Performance Data
The mean reaction time (RT) and standard error for the three
tasks were as follows: LOCATION, 673
±
38 ms; SHAPE, 709
±
15
ms;
and DOT, 452 ± 13 ms. An analysis of variance (ANOVA)
showed a significant main effect of task iF
=
32.7, P
<
0.01) for
RT.
Post-hoc analysis using Tukey's HSD revealed that both
working memory tasks had significantly longer RTs than the
602
Spam!
and Nonspatial Working Memory • McCarthy et
aL
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
DOT
task
(P
<
0.05), but the working memory tasks did not differ
significantly from each other. The percentage of correct
responses was 56 ±
4%
for the LOCATION task, 74 ±
8%
for the
SHAPE task, and 78 ±
13%
for the DOT task. While there was a
trend for lower percentage correct responses for the LOCATION
task, the differences in accuracy between the tasks were not
statistically significant
(JF=
2.8, P
=
0.08).
t-Test Analyses of/MRI Images
Individual Subject Images
Across subjects, the working memory tasks most consistently
activated the MFG where clusters of significantly activated
voxels were observed in 9 of the 10 subjects. The activated
regions typically included the inferior frontal sulcus which
separates the MFG from the IFG. Representative data for an
individual subject are presented in Figure 2a-f in which the
significantly activated voxels (as described in Materials and
Methods) are represented as -color overlays displayed upon
anatomically consecutive Ti-weighted images. The two coronal
slices shown for each task are those which contributed to the
spatially normalized group mean images. Figure 2a,b represents
activations associated with the LOCATION task for the anterior
and posterior slice, respectively. Figure 2c,d represents
activation associated with the SHAPE task for the same slices;
while Figure 2e/ represents activation associated with the DOT
task. In the LOCATION task, a cluster of activated voxels was
obtained in the right MFG in the anterior slice (Fig. 2a) and on
the border of the right MFG and IFG in the posterior slice (Fig.
2b).
No activation exceeded the t criterion of 2.0 in the same
region of the left
MFG,
although a small patch of activated voxels
occurred more inferiorly in the left IFG (Fig. 26). The SHAPE
task produced larger clusters of activated voxels in the right
MFG
and IFG bilaterally, most notably in the posterior slice (Fig. 2d).
Few activated voxels were obtained in the
DOT
task compared to
the working memory tasks.
A
few scattered voxels were seen in
the right SFG in both the anterior
(Fig. 2e)
and posterior
(Fig.
2/)
slices,
and in more inferior aspects of the right hemisphere.
Activated voxels were evident at the base of the brain between
the eyes in both the LOCATION and SHAPE tasks. This region,
which occurred at the most inferior edge of the gradient-echo
images, showed an inconsistent pattern across subjects. The
proximity of this region to the eyes and to susceptibility artifacts
caused by the sinuses suggested that these effects, when
observed, were artifactual.
Group Mean Images
The ttest analysis performed on the group mean images
confirmed the observations in individual subjects. The
significantly activated voxels for the LOCATION task are
displayed in Figure 3a,b as a red overlay upon the group average
anatomical images. The results for the SHAPE and DOT tasks are
similarly presented in Figure 3c,d and Figure 3e/, respectively.
Across subjects, the
LOCATION
task was associated with clusters
of activated voxels in the right
MFG
of both slices with minimal
activations of the left MFG of either slice. The SHAPE task also
produced clusters of activated voxels in the right MFG of both
slices.
The distributions of activated voxels were similar to those
obtained for the LOCATION task. Unlike the LOCATION task,
however, SHAPE was associated with a cluster of activated
voxels in the left MFG extending inferiorly into the IFG (Fig.
3c,d).
The DOT task produced fewer activated voxels in this
analysis. One cluster was evident in the right MFG of the
Table
1
Mean activation levels
for
each task and anatomical location
for the
region
of
interest analysis
calculated
for the group-averaged,
spatially normalized images
LOCATION SHAPE
DOT
Right
SFG
MFG
IFG
Cing
Left
SFG
MFG
IFG
Cing
Both
SFG
MFG
IFG
Cing
1.40
4.44
1.53
0.76
-1.10
2.65
-0.46
-0.15
0.19
3.55
0.54
0.31
1.73
4.88
1.06
1.72
0.85
5.33
2036
1.26
1.29
5.11
1.71
1.49
3.42
2.57
-1.17
1.24
-0.23
2.12
-2.31
0.50
1.60
2.35
-1.74
0.87
Activation is expressed
in
z-scores
(standard deviations) comparing
the
task interval to
the
common pre-task
baseline. SFG,
superior frontal
gyrus;
MFG.
middle frontal
gyrus;
IFG.
inferior
frontal
gyrus; Cing,
cingulate
gyrus.
posterior slice
(Fig.
3/) and a cluster of activation was evident in
the inferior aspect of both slices
(Fig.
3ef).
Region of Interest Analysis
The anatomical extent of activation across subjects was more
systematically evaluated in the ROI analyses performed on the
spatially normalized data. Figure 4 presents the spatially
normalized anterior
(Fig.
4a) and posterior
(Fig.
4b) group mean
anatomical images and the anatomical ROIs used to interrogate
the mean image series in each task (because of
signal
loss caused
by the frontal sinuses, ROIs for the gyrus rectus and the orbital
frontal gyri were not drawn). Table 1 presents the mean
activation in the task analysis interval (i.e., images 23-43) for
each ROI. The overall distribution of activation summed across
working memory tasks is presented in Figure 5. Mean activation
across hemispheres exceeded ± 2 jr-scores [or standard
deviations (SDs)] above the pre-task interval only in the
MFG.
Time-Course
of Activation for
MFG
The Mest analyses for the group-averaged data
(Fig.
3) suggested
a hemispheric difference between the activation patterns of the
LOCATION and SHAPE tasks. This was investigated further in
the anatomical ROI analysis (see
Fig.
4 for regions interrogated).
Since the ROIs integrate activity over a large number of voxels,
they should be maximally sensitive to small but consistent
activations which might produce few significant rvalues in the
presence of noise. The time-courses of activation obtained for
the right (red ROI, Fig. 4a,b) and left MFG are shown in Figure
6a,b,
respectively. The activation time-courses were combined
for both the anterior and posterior slices for clarity since the
signal changes associated with each task were similar in both
slices.
For the right
MFG (Fig.
6a), the activation time courses for
all three tasks rose above 2.0 SDs above the pre-task baseline
within 3 s and reached a plateau by -6 s. This level of activation
was maintained until 6 s after the task interval, at which time the
activation slowly declined. The mean activation over the task
analysis interval was similar for both working memory tasks (see
Table 1) but only reached slightly more than half of
this
level for
the DOT task. In contrast to the robust activation time course for
the MFG, the ROIs drawn for the adjacent white matter (blue
Cerebral Cortex July/Aug 1996,
V
6 N 4 603
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
Figure 3. Red pixels indicate significant increases
in
activation ((> 8] during the task interval relative to the pre-task
baseline
for the group-averaged
data.
For
these
pixels,
activation
decreased by at least
50%
by the conclusion of the post-task
interval.
Results for the anterior
and
posterior slices are shown for the LOCATION (ah). SHAPE (cd). and
DOT
[efl tasks.
Activations are shown superimposed upon spatially normalized, group-averaged anatomical images. The approximate Talairach coordinates for the centers of the activations were
LOCATION (x = 37, y = 40^ = 29). SHAPE (x =
31.
K
=
40.
z = 40 for right hemisphere, x = -37, y =
40.
z = 22 for left hemisphere). Note that the/-coordinate was limited
by the slice selection.
ROI, Fig. 4a,b) did not show any consistent task-related activa-
tion (Fig. 7a,fe).
Hemispheric Differences in
MFG
Activation.
The time-courses of activation in the left MFG (Fig. 6b) showed
a
similar onset latency
as
the right
MFG.
The SHAPE task produced
a slightly greater mean level of activation for the left MFG (5.33
SDs) than it did for the right MFG. The DOT task produced a
mean activation of
2.12
SDs—also similar to its value for the right
MFG. However, the LOCATION task produced a mean activation
of only 2.65 SDs in the left MFG—a value half that produced by
SHAPE and similar to that produced by DOT. The differential
604 Spatial and Nonspatial Working Memory • McCarthy ct al.
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
10
5
-5
4
9
14 19 24 29 34 39 44 49 54 59 64
IMAGE
10
5
•
-5
llllilllllllllllllillllll
4
9
14 19 24 29 34 39 44 49 54 59 64
IMAGE
Figure 6. Group mean time-courses of activation for the right (a) and left (6)
MFG
averaged over the anterior and posterior slices (LOCATION = heavy solid line, SHAPE = light solid
line,
DOT = dotted line). In this and Rgures 6 and 7, the image number appears on thex-axis and image intensity (expressed as a^-score relative to the pre-task baseline) is shown
on the y-axis.
The
vertical lines demarcate task onset
and
offset.
10
5
-5
4
9
14 19 24 29 34 39 44 49 54 59 64
IMAGE
10
5
-5
4
9
14 19 24 29 34 39 44 49 54 59 64
IMAGE
Figure 7. Group mean time-courses of activation for the right (a) and left (6) white matter averaged over the anterior and posterior slices (LOCATION
=
heavy solid line, SHAPE
:
light solid line, DOT = dotted line).
Hgure 8. Activated voxels are shown overlaid upon Ti-weighted anatomical images for an individual subject performing the
SHAPE
task using a gradient-echo imaging sequence,
(a) Convolution analysis showing clusters of activation in the left
and
right middle frontal
gyri.
(6) Split-f map (f > 2) showing two clusters of activation in the left middle frontal gyrus.
Color
scale:
yellow (f = 2) to red (f = 5).
in the immediate post-task interval, with somewhat greater
activation following the cessation of the SHAPE task (Fig. 10).
This difference was pronounced in the left cingulate where the
activation produced by the SHAPE task persisted longer than for
the other two tasks. This more persistent activation for SHAPE
was also apparent in the left MFG (Fig. 6b).
606 Spatial and Nonspatial Working Memory • McCarthy
a
aL
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
Figure 4. Spatially normalized group mean images for the anterior (a) and posterior (A) slices. Regions of interest used in the group analyses are indicated in color for the right
hemisphere. Similar ROIs (not shown) were analyzed for the left hemisphere. ROIs encompassing the superior
frontal
gyri are shown in green; middle frontal gyri in
red;
inferior frontal
gyri in yellow; cingulate gyri in pink; and white matter in blue.
activation produced in the left MFG for the SHAPE compared to
LOCATION working memory tasks was consistent with the Mest
analysis shown in Figure 3.
Although the overall level of activation was smaller, the
SHAPE task also produced a stronger activation (Table 1) of the
left inferior frontal gyrus (2.36 SDs) than of the right (1.06 SDs).
In contrast, no activation of the left inferior frontal gyrus was
observed for the LOCATION task. The more inferior extent of
activation for the SHAPE task in the left hemisphere was seen in
the group Mest analysis (Fig. 3d) and in individual analyses (Fig.
Id).
Activation produced by the
DOT
task showed
a
hemispheric
asymmetry in the SFG. The right side showed a mean activation
of 3-42 SDs while the left showed no activation.
Application of Group Activation
Time-Course
to New Subjects
The group mean activation time-courses from the MFG (Fig.
6a,b) for the LOCATION and SHAPE tasks were averaged,
low-pass filtered to remove noise, and then used in
a
convolution
analysis of the data obtained for four additional subjects. This
analysis was performed to test the reliability of the ROI
time-course analyses obtained in the group analysis. Figures 8
and 9 present the results for subject AP who performed eight
runs of the SHAPE task using gradient-echo echoplanar
acquisition. Figure 8a shows the result of the convolution
analysis. Several clusters of activated voxels appear in the left
MFG, and two smaller clusters appear in the right MFG. A few
additional clusters appear near the bottom of the image. Figure
8b shows the result of the Mest analysis performed on these
same data. Two clusters of activated voxels were obtained in the
left MFG, and one small cluster was obtained in the right MFG.
The activated regions identified by the Mest overlapped with
those identified by the convolution analysis.
The results for this individual showed a predominantly left
MFG activation for the SHAPE task which extended more
inferiorly in the left hemisphere, and thus replicated two of the
main findings from the primary
study.
Subject
AP was
not run on
the spatial working memory task in this new study, but had
previously demonstrated a strong right dominant activation for
spatial working memory tested previously by McCarthy et at
(1994-see their fig. 4).
The time-courses of activation (Fig. 9) were determined for
SFG
MFG Cing
Figure 5. Distribution of activation summed across working memory tasks and
hemispheres
in
the spatially normalized
images,
z-scores represent integrated activation
for each anatomical ROI expressed relative to a common pre-task baseline. SFG,
superior frontal gyrus; MFG, middle frontal gyrus; IFG, inferior frontal gyrus; Cing,
cingulate gyrus.
the activated voxels from the left
MFG
from the convolution (Fig.
8a) and Mest analysis
(Fig.
8b), respectively.
A
mean activation of
1.43 and 1.9% was calculated for each method for the
gradient-echo images (solid lines). These same voxels were
interrogated in the spin-echo image series and showed mean
activations of
0.60
and 0.77%, respectively (broken
lines).
These
values were 42 and 41% of the measures obtained for
gradient-echo imaging.
The convolution and Mest analyses revealed similar clusters of
activation in the MFG with gradient-echo imaging in all four
subjects, and similar activation time-courses were obtained for
both gradient and spin-echo images in three of four subjects. Of
the three subjects with concordant activation time-courses for
gradient and spin-echo imaging, the mean signal changes
obtained for spin-echo were 37% (range 35-42%) of those
obtained for gradient-echo.
Time
Course
of Activation for Cingulate Cortex
The ROI analysis for the cingulate gyri (pink ROI, Fig. 4a,b)
showed a distinct pattern in which little or no activation
occurred during the task, but all three tasks produced activation
Cerebral
Cortex July/Aug
1996, V6N4 605
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
%
SIGNAL
CHANGE
-J.U
2.5
2.0
1.5
1.0
•
0.5
0.0
-0.5
-1.0
V •: ^V7
•.•
-M+t
H
t + l-H > +
'*"'•
*'»-•*
"•/•/*
"2S8S8PISS3SS88S
% SIGNAL
CHANGE
HI >H-t
H-fKW
t-H M f M4-4 I I H IH H-
Figure
9.
Time-courses
of
activation
for
ROIs
in the
left middle frontal gyrus from Rgure
8. (a)
Convolution ROI from Rgure 8a.
(A)
Split Mest ROI from Rgure 8b. Solid line
gradient-echo sequence; dotted line
=
spin-echo sequence.
10
5
-
+H+
4 9 14 19 24 29 34 39 44 49 54 59 64
IMAGE
10
5-
4 9 14 19 24 29 34 39 44 49 54 59 64
IMAGE
Rgure 10. Group mean time-courses of activation for the right (a) and left \b) cingulate gyri averaged over the anterior and posterior slices (LOCATION
=
heavy solid line, SHAPE
=
light solid line, DOT
=
dotted line).
Cerebral CortexJuly/Aug 1996, V6N4
607
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
Discussion
The present study demonstrated that the middle frontal gyri
were consistently activated during the performance of working
memory tasks, and that the hemispheric pattern and extent of
activation was dependent upon the working memory domain.
Activation commenced by 3 s after task onset and declined at a
slower
rate,
beginning -6 s after task completion. This pattern of
activation was not observed in white matter, and similar levels
and/or patterns of activation were not observed in the superior
and inferior frontal gyri, or in the cingulate gyri. The pattern of
activation in the MFG was reliable in that the group activation
time course of this region successfully identified activation in
the MFG in data collected subsequently in four additional
subjects.
Differential Hemispheric Activity Related to Location
and Shape
The present findings replicate our previous study performed at
2.1 T where we found that spatial working memory prefer-
entially activated the right MFG (McCarthy et al., 1994). The
spatial memory (LOCATION) task produced a right MFG
activation which was 168% of that produced in the left MFG in
the group-averaged data. The nonspatial memory (SHAPE) task
produced nearly equal activation of the right MFG as the spatial
task, but even greater activation was produced in the left MFG
where the average level of activation was more than twice the
level measured for the spatial task. Thus while the greatest levels
of activation for the working memory tasks were consistently
found in the
MFG,
the activation showed a different hemispheric
pattern as a function of the memory task performed (see Smith
et
al.,
1996, fora related discussion).
Activation of prefrontal cortex during working memory tasks
has been reported in several PET and fMRI studies
(e.g.,
Jonides
et al., 1993; Petrides et al., 1993a,b; Cohen et al., 1994; Swartz
et al, 1994, 1995; Smith et al, 1995, 1996; see review by
McCarthy, 1995). This region has also been activated by tasks
which, though not presented in the context of working memory,
nevertheless required the transient storage of sensory
information (e.g., Corbetta et al., 1991; O'Sullivan et
al.
1994).
However, reports have differed with regard to the hemispheric
distribution of effects, and in the relative extent of dorsal and
ventral prefrontal activation.
Activation of the right PFC has been observed during the
performance of spatial working memory tasks (Jonides et al.,
1993;
McCarthy et al, 1994; Smith et al, 1995), and in memory
tasks using spatially oriented stimuli (Petrides et al, 1993a). Of
these studies, Petrides et al (1993a) and McCarthy et al (1994)
showed bilateral but right dominant activation of the dPFC,
while Jonides et
al.
(1993) found activation in right inferior
PFC
(area
47).
Smith et al (1995) replicated the findings of Jonides et
al (1993) with regard to the right inferior PFC, but also found
activation of dPFC area 46 in one variant of their task. More
recently, this latter group (Smith etal, 1996) reported activation
of dPFC area 46 when the spatial working memory task was
made more continuous. Activation was bilateral but was greater
in the right hemisphere. These results and the present study
support the conclusion that working memory tasks primarily
involving spatial memory preferentially activate the right PFC.
However, others have not found this relationship. Sweeney et al
(1996) found bilateral activation of dPFC area 46 using an
oculomotor spatial delayed-response task similar to that
employed in the nonhuman primate studies. Courtney et al
(1996) found predominantly right dPFC activation for a face
working memory task, but not for a location working memory
task where more posterior activation in premotor cortex was
found when comparing their location and face tasks. Thus, at
present, it is not possible to incorporate all studies of spatial
working memory within a common framework.
In contrast to the spatial working memory tasks reviewed
above, nonspatial tasks have shown a different hemispheric
pattern of activation. Cohen etal. (1994) used
a
letter monitoring
task to demonstrate bilateral activation of
dPFC
and Swartz et al
(1994,
1995) showed bilateral activation of dorsal
PFC
(including
dPFC and dorsal pole) in a visual memory task similar to the
SHAPE task used
here.
Smith etal. (1995) found activation of left
premotor cortex in an object working memory task.
The degree to which bilateral activation may represent the
addition of verbal mediation to the subject's task strategy
remains undetermined. Petrides et al (1993b) found bilateral
activation of dPFC using a verbal working memory task, and
Smith et al. (1996) showed predominantly left prefrontal
activation in their verbal working memory study. In the present
study we attempted to limit verbal mediation by using shapes
without obvious verbal labels. Nevertheless, some subjects
reported using verbal labels as a mnemonic aid. However, the
argument that differential verbal processing in working memory
tasks determines completely the hemispheric pattern of dPFC
activation
is
challenged by those
PET
studies of working memory
which explicitly required phonological processing and in which
differential activation of Broca's area rather than dPFC was
obtained (Zatorre etal, 1992; Paulesu etal, 1993).
It is noteworthy that activation of the PFC has also been
obtained in long-term recognition memory tasks (e.g., Grasby et
al. 1993; Moscovitch et al 1995; Shallice et al, 1994). For
example, Moscovitch et al. (1995) have reported increased
activation of the right PFC (including areas 44, 45, and 46) for
both spatial and object recognition
tasks.
It has been argued that
the PFC is critical for episodic memory and, furthermore, that
there is hemispheric specialization with regard to episodic
memory function (Shallice et al, 1994; Tulving et al, 1994).
Both Shallice et al (1994) and Tulving et al. (1994) have
proposed that the right PFC is important for episodic retrieval
while the left PFC is important for encoding into episodic
memory. In the model of Tulving etal (1994), the left PFC is also
involved in retrieving information from semantic memory. These
studies are not necessarily in conflict with the present study in
that it is likely that working memory plays a role in encoding to
and retrieval from episodic memory. However, the present
results and prior literature (e.g., Smith et al 1995) indicate that
the spatial or object domain tested by the working memory task
is important in determining the hemispheric pattern of frontal
lobe activation.
Comparison of Perceptual Control and Working
Memory Tasks
Although not clearly evident in the Mest images
(Fig.
3eJ), the
ROI time-course analyses showed that the non-memory-guided
DOT task also produced bilateral activation of the
MFG,
although
at approximately half the level of the SHAPE task. This is
consistent with our previous study in which non-memory color
detection and dot detection tasks both produced activation of
the right MFG, but at a level less than that obtained for a spatial
working memory task using identical stimuli (McCarthy et al,
1994).
The present study revealed bilateral MFG activation for
the DOT control task, although the activation of the right MFG
was somewhat greater than the left.
608 Spatial and Nonspatial Working Memory • McCarthy ct aL
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
We note that the level of activation produced by the
LOCATION task in the left MFG was similar to that produced by
the DOT task. Pardo et al (1991) have reported a PET study in
which increases in blood flow in the PFC occurred during
performance of vigilance tasks requiring sustained attention,
such as the DOT task used here. If the level of activation
produced by the DOT task represents processes related to
attention, motor preparation, and/or other common task
requirements, and if the difference between the DOT and
working memory tasks represents additional processes related to
the maintenance of memory for past
items,
then the hemispheric
dissociation between the spatial and nonspatial working
memory tasks becomes even more apparent. A comparison of
the activation time-courses (see
Fig.
6a,b) shows that subtraction
of the DOT task would leave residual activation of the left MFG
only for the nonspatial SHAPE working memory task. However,
both spatial and nonspatial working memory tasks would show
residual activation of the right MFG.
The monitoring of visually presented stimuli might be
expected to activate the
MFG
in humans as studies in nonhuman
primates have shown that neurons in the PFC (particularly
dorsolateral areas 8 and 46) respond to attended visuospatial
stimuli—whether or not they are to be recalled (Funahashi et al,
1991;
Wilson et al, 1993). Furthermore, quantitative 2-deoxy-
glucose metabolic studies in monkeys performing cognitive
tasks have consistently shown that the difference between
control and experimental tasks is one of degree rather than of
anatomical location (Friedman and Goldman-Rakic, 1988).
Therefore, the present findings are entirely in line with
expectations based upon cortical physiology.
Task Difficulty
Other task differences or systematic artifacts may, however,
underlie the difference in activation between the DOT and
working memory tasks. One concern is task difficulty. The
percentage of correct responses made during performance of
the DOT task was not significantly different from either working
memory task, but the RT for DOT detection was significantly
faster. This is not surprising as one would expect that scanning
a
memory buffer to identify targets (as in the working memory
tasks) would take longer than detecting a brightened pixel. Thus,
the operation of the very process of interest (memory) is the
likely basis for the RT difference. While perceptual tasks can be
made arbitrarily difficult to equalize RTs, the difficulty level in
each task would be based upon different processes. An
argument that greater difficulty would lead to greater activation
due to a non-specific process such as arousal is countered by the
results of the LOCATION task, which, though showing a trend
for a greater error rate, nevertheless produced less overall
activation than
SHAPE.
Thus while differences in task difficulty
may underlie some activation differences, activation levels did
not correlate with the performance measures obtained in the
working memory tasks.
Differential Dorsal and Ventral Prefrontal Activation
While the present study strongly indicates differential activation
of the PFC for spatial and nonspatial working memory tasks as a
function of hemisphere, the question of differential dorsal and
ventral distribution of activation within a hemisphere for the
spatial and nonspatial working memory tasks is less clear. The
largest and most consistent activations associated with the
working memory tasks occurred in the MFG, which
cytoarchitectonic studies in human cortex have shown
correspond primarily to area 46 (Rajkowska and Goldman-Rakic,
1995).
This region is thus analogous to the principal sulcus of
monkey shown to be active in working memory
tasks
(Funahashi
et al, 1989; Goldman-Rakic, 1987). In the present study,
activation was also seen in the left
IFG
for the nonspatial SHAPE
working memory task, while the spatial LOCATION task showed
no consistent activation of this same region. The degree of
activation was small relative to the MFG and requires further
study to assess its reliability. Nevertheless, it is consistent with
prior monkey studies (Wilson et al, 1993) which showed that
nonspatial memory
tasks
involve more ventral prefrontal regions
than spatial memory tasks. It is also consistent with prior
neuroimaging working memory studies in humans: Cohen et al
(1994),
for example, reported activation extending into more
ventral prefrontal regions in their letter monitoring
task.
It is also
consistent with Courtney et al (1996) who used face stimuli in a
nonspatial working memory task and found activation extending
into the inferior PFC, albeit predominantly in the right
hemisphere.
Activation of the Cingulate Gyrus
An unexpected finding in the present study was the pattern of
activation in the cingulate gyrus. While little or no activation
occurred during task performance, significant activation
occurred for all three tasks upon task cessation. The reason for
this nonspecific post-task activation is not clear, but the
temporal dissociation of activity between the cingulate and
prefrontal region is striking. PET studies have reported
co-activation of the anterior cingulate and dorsolateral PFC in
tasks with working memory components (e.g., Petrides et al,
1993a,b), and we have observed similar co-activation, though
not reliably across replications (McCarthy et al, 1994). Pardo et
al (1990) have implicated the anterior cingulate in selecting
processing systems necessary for task execution. Simple
sustained attention, however, did not activate this region (Pardo
et al, 1991). Interpretation of the cingulate activation observed
in the present experiment is difficult because it occurred at a
point when the tasks ended and when performance feedback
was provided. This suggests that the post-response activation
may be related to performance feedback or, alternatively, to the
disengagement of sustained attention. It is unlikely to reflect
response selection, the mnemonic content of the working
memory task, or attention per se, which would have been
maximal during the tasks themselves. Of possible relevance are
recent single-unit recordings in the posterior cingulate cortex of
rhesus monkeys which have revealed a population of neurons
that consistently respond only after the delay of delayed-response
trials,
independent of the direction of the response (Carlson et
al, 1994). This neuronal temporal profile is distinctly different
from that recorded in prefrontal neurons, just as the fMRI
activation of the cingulate cortex differs from that of the
prefrontal cortex. Both findings open new questions on the
contribution of the cingulate cortex to cognitive functions.
Gradient-Echo versus
Spin-Echo
Imaging Sequences
The degree to which fMRI can distinguish activity in closely
adjacent anatomical regions that are nonetheless functionally
distinct (as determined by other methods such as single unit
recording) remains to be determined. One issue that has been
raised is the degree to which fMRI performed at 1.5 T using
gradient-echo acquisition may partly reflect increased signals in
large venules and veins which may be distant to the activated
neurons (e.g., Constable et al, 1994; Lai etal, 1993). Spin-echo
Cerebral CortexJuly/Aug 1996, V6N 4 609
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
acquisition is less sensitive to magnetic susceptibility effects,
especially those arising from large vessels. While activation
effects in spin-echo images emphasize smaller vessels relative to
gradient-echo images, the overall sensitivity is reduced, making
functional imaging with spin-echo difficult at 1.5 T. In the
present study, we identified voxels activated by the working
memory tasks using gradient-echo sequences, but then
interrogated these same voxels in spin-echo sequences. Using
this procedure, activation effects were observed in the
spin-echo images with signal changes 37% of those obtained in
the gradient-echo images. This value is within the range
expected from theoretical calculations of the relative signal
changes in microvascular structures (Kennan et aL, 1994).
Moreover, the time-courses of activation for the gradient-echo
and spin-echo sequences were similar. One might expect
a
delay
in the gradient-echo time-course if the activity primarily
reflected flow downstream of an activated region of
cortex.
It is
therefore unlikely that the activation effects found in the MFG
are greatly displaced from the true locus of activation.
Nevertheless, until the precise anatomical size and proximity of
spatial and nonspatial visual processing centers in human PFC
are established, the resolution of the imaging method remains an
issue.
Notes
This work was supported by the Department of Veterans Affairs, NIMH
Grants MH-44866 and MH-05286, and the McDonell-Pew Program in
Cognitive Neuroscience. We thank Marie Luby, Francis Favorini, and Dr
Anthony Adrignolo for assistance in data analysis.
Address correspondence to Dr Gregory McCarthy, Neuropsychology
Laboratory/116B1, VA Medical Center, West Haven, CT 06516, USA.
References
Baddeley A (1992) Working memory. Science 255:556-559.
Bandettini PA, Jesmanowica A, Wong EC, Hyde JS (1993) Processing
strategies for time-course data sets in functional MR1 of the human
brain. Magn Reson Med 30:161-173.
Carlson S, Mikami A, Goldman-Rakic, PS (1993) Omnidirectional delay
activity in the monkey posterior cingulate cortex during the
performance of an oculomotor delayed response task. Soc Neurosci
Abstr 19:800.
Cohen
JD,
Forman SD, Braver
TS,
Casey BJ, Servan-Schreiber D, Noll DC
(1994) Activation of prefrontal cortex in a non-spatial working
memory task with functional
MRI.
Hum Brain Map 1:293-304.
Constable RT, Kennan RP, Puce A, McCarthy G, Gore JC (1994)
Functional NMR imaging using fast spin echo at 1.5T. Magn Reson
Med 31:1-6.
Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE (1991)
Selective and divided attention during visual discriminations of shape,
color, and speed: functional anatomy by positron emission
tomography. J Neurosci 11:2383-2402.
Courtney SM, Ungerleider
LG,
Keel K, Haxby
JV
(1996) Object and spatial
visual working memory activate separate neural systems in human
cortex. Cereb Cortex 6:39-49.
Friedman HR, Goldman-Rakic PS (1988) Activation of the hippocampus
by working memory: a 2-deoxyglucose study of behaving rhesus
monkeys.
J
Neurosci 8:4693-4706.
Funahashi S, Bruce CJ, Goldman-Rakic PS (1989) Mnemonic coding of
visual space in the monkeys dorsolateral prefrontal cortex. J
Neurophysiol 61:1-19.
Funahashi S, Bruce CJ, Goldman-Rakic PS (1991) Visuospatial coding in
primate prefrontal neurons revealed by oculomotor paradigms. J
Neurophysiol 63:814-831.
Fuster J (1973) Unit activity in prefrontal cortex during delayed-response
performance: neuronal correlates of transient memory. J
Neurophysiol 36:61-78.
Goldman-Rakic PS (1987) Circuitry of primate prefrontal cortex and
regulation of behavior by representational memory. In: Handbook of
physiology, the nervous system, higher functions of the brain, Section
I, Vol. V, Part 1 (Plum F, ed), pp. 373-417. Bethesda, MD: American
Physiological Society.
Grasby PM, Frith CD, Friston KJ, Bench C, Frackowiak RSJ, Dolan RJ
(1993) Functional mapping of brain areas implicated in
auditory-verbal memory function. Brain 116:1-20.
Jonides J, Smith
EE,
Koeppe RA, Awh E, Minoshima S, Mintun MA (1993)
Spatial working memory in humans as revealed by PET. Nature
363:623-625.
Kennan RP, Zhong
J,
Gore
JC
(1994) Intravascular susceptibility contrast
mechanisms in tissues. Magn Reson Med 31:9-21.
Lai S, Hopkins AL, Haacke EM, Ii D, Wasserman BA, Buckley P, Friedman
L,
Meltzer H, Hedera P, Friedland R (1993) Identification of vascular
structures as a major source of signal contrast in high resolution 2D
and 3D functional activation imaging of the motor cortex at 1.5 T:
preliminary results. Magn Reson Med 30:380-386.
McCarthy G (1995) Functional neuroimaging of memory. Neuroscientist
1:155-163.
McCarthy G, Blamire AM, Rothman DL, Gruetter R, Shulman RG (1993)
Echo-planar magnetic resonance imaging studies of frontal cortex
activation during word generation in humans. Proc Natl Acad Sci USA
90:4952-4956.
McCarthy G, Blamire AM, Puce A, Nobre AC, Bloch G, Hyder F,
Goldman-Rakic P, Shulman
R
(1994) Functional MRI studies of frontal
cortex activation during a spatial working memory task in humans.
Proc Natl Acad Sci USA 91:8690-8694.
Moscovitch M, Kapur S, Kohler S, Houle S (1995) Distinct neural
correlates of visual long-term memory for spatial location and object
identity: a positron emission tomography study in humans. Proc Natl
Acad Sci USA 92:3721-3725.
O'Sullivan BT, Roland PE, Kawashima R (1994) A PET study of
somatosensory discrimination in man. Microgeometry versus
macrogeometry. Eur J Neurosci 6:137-148.
Pardo
JV,
Fox PT, Raichle ME (1991) Localization of a human system for
sustained attention by positron emission tomography. Nature
349:61-64.
Pardo
JV,
Pardo PJ, Janer KW, Raichle ME (1990) The anterior cingulate
cortex mediates processing selection in the Stroop attentional conflict
paradigm. Proc Natl Acad Sci USA 87:256-259.
Paulesu E, Frith CD, Frackowiak RSJ (1993) The neural correlates of the
verbal component of working memory. Nature 362:342-345.
Petrides
M,
Alivisatos
B,
Evans
AC,
Meyer
E
(1993a) Dissociation of human
mid-dorsolateral from posterior dorsolateral frontal cortex in memory
processing. Proc Natl Acad Sci USA 90:873-877.
Petrides M, Alivisatos B, Meyer
E,
Evans
AC
(1993b) Functional activation
of the human frontal cortex during the performance of verbal
working memory tasks. Proc Natl Acad Sci USA 90:878-882.
Puce A, Constable RT, Luby ML, McCarthy G, Nobre AC, Spencer DD,
Gore
JC,
Allison T (1995) Functional magnetic resonance imaging of
sensory and motor cortex: comparison with electrophysiological
localization. J Neurosurg 83:262-270.
Rajkowska
G
and Goldman-Rakic PS (1995) Cytoarchitectonic definition
of prefrontal areas in the normal human cortex: II. Variability in
locations of areas 9 and 46 and relationship to the Talairach coordinate
system. Cereb Cortex 5:323-338.
Schneider
W,
Casey
BJ,
Noll D (1993) Functional
MRI
mapping of stimulus
rate effects across visual processing stages. Hum Brain Map
1:117-133.
Shallice
T,
Fletcher
P,
Frith CD, Grasby
P,
Frackowiak
RSJ,
Dolan
RJ
(1994)
Brain regions associated with acquisition and retrieval of verbal
episodic memory. Nature 368:633-635.
Smith
EE,
Jonides J, Koeppe RA, Awh E, Schumacher EH, Minoshima S
(1995) Spatial versus object working memory: PET investigations. J
Cog Neurosci 7:337-356.
Smith EE, Jonides J, Koeppe RA (1996) Dissociating verbal and spatial
working memory using
PET.
Cereb Cortex 6:11-20.
Swartz
BE,
Halgren
E,
Fuster J, Mandelkern
M
(1994) An
18
FDG-PET study
of cortical activation during a short-term visual memory task in
humans. NeuroReport 5:925-928.
Swartz BE, Halgren E, Fuster J, Simpkins F, Gee M, Mandelkern M (1995)
Cortical metabolic activation in humans during a visual memory task.
Cereb Cortex 3:205-214.
Sweeney
JA,
Mintun MA, Kwee BS, Wiseman MB, Brown DL, Rosenberg
DR, Carl
JR
(1996)
A
positron emission tomography study of voluntary
saccadic eye movements and spatial working memory.
J
Neurophysiol
75:454-468.
610 Spatial and Nonspatial Working Memory • McCarthy et
aL
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the human Wilson FAW, O'Scalaidhe SP, Goldman-Rakic PS (1993) Dissociation of
brain. New
York:
Thieme. object and spatial processing domains in primate prefrontal cortex.
Tulving E, Kapur S, Craik FIM, Moscovitch M, Houle S (1994) Science 260:1955-1958.
Hemispheric encoding/retrieval asymmetry in episodic memory: ZatorreRJ, Evans
AC,
Meyer, E, Gjedde
A
(1992) LateraJization of phonetic
positron emission tomography findings. Proc Natl Acad Sci USA and pitch discrimination in speech processing. Science 256:846-849.
91:2016-2020.
Cerebral CortexJuly/Aug 1996, V6N4 611
by guest on July 13, 2011cercor.oxfordjournals.orgDownloaded from
