www.elsevier.

com/locate/ynimg
NeuroImage 22 (2004) 562 – 573

Functional dissociations within the inferior parietal cortex in verbal

working memory
Susan M. Ravizza, a,b,* Mauricio R. Delgado, b Jason M. Chein, a,c
James T. Becker, c,d,e and Julie A. Fiez a,c,d
a
Department of Psychology, University of Pittsburgh, Pittsburgh, PA 15260, USA
b
Department of Psychology, New York University, New York, NY 10003, USA
c
Center for the Neural Basis of Cognition, Pittsburgh, PA 15260, USA
d
Department of Neuroscience, University of Pittsburgh, Pittsburgh, PA 15260, USA
e
Department of Psychiatry, University of Pittsburgh, Pittsburgh, PA 15260, USA

Received 10 July 2003; revised 19 December 2003; accepted 28 January 2004

Available online 20 April 2004

Neuroimaging studies of working memory have revealed two sites in According to one influential model of working memory (Bad-
the left supramarginal gyrus that may support the short-term storage deley and Hitch, 1974), the ability to remember verbal information
of phonological information. Activation in the left dorsal aspect of the for a brief time is contingent upon a system dedicated to the storage
inferior parietal cortex (DIPC) has been observed in contrasts of of phonological material. Verbal items (e.g., digits, letters, or
working memory load, whereas activation in the ventral aspect of the words) are maintained in this store and are periodically refreshed
inferior parietal cortex (VIPC) has been found primarily in contrast of
by articulatory rehearsal to prevent their rapid decay. Auditory
information type (verbal vs. nonverbal). Our goal was to determine
whether these two areas are functionally distinct or if instead they are information is claimed to have obligatory access to this dedicated
part of a homogeneous region with large variations in the focus of peak linguistic short-term store (STS), although visually presented items
activity. Toward this end, we used fMRI to assess the neural response can also be maintained through this route if they are phonologically
in two working memory tasks (N-back and item recognition) in which recoded. However, the STS is claimed to be a separate system from
we also manipulated memory load and the type of information to be that used for phonological perception (Baddeley et al., 1998;
recalled (verbal vs. nonverbal). We found both DIPC and VIPC Martin and Breedin, 1992).
activation in the same group of subjects and further demonstrated that The primary evidence for a distinct phonological store and
they have differential sensitivity to our experimental factors. Only the perceptual system includes several accounts of patients with
DIPC showed robust load effects, whereas only the VIPC showed reduced spans for verbal information without a concomitant deficit
reliable effects of information type. These results help to account for
in the identification or discrimination of phonemes (Vallar and
the differences observed in between-subject comparisons, and they
indicate that the two regions are functionally dissociable. In contrast to Baddelely, 1984; Warrington and Shallice, 1969). Research with
the DIPC, activity of the VIPC was also recruited in the fixation and these patients has also suggested that the STS is specialized for
low-load conditions, a surprising result that has not been fully explored linguistic material as these patients have no difficulty with visuo-
in prior studies. Despite their distinctive patterns of performance, spatial working memory tasks (Shallice and Warrington, 1974;
neither of these regions displayed a pattern of activity that entirely Vallar and Papagno, 1995). Their pattern of deficit, then, suggests a
corresponds to common assumptions of a dedicated phonological neural area that is dedicated to the storage of verbal material but is
short-term store (STS). Instead, we hypothesize that the DIPC may not involved in the initial perception of the items to be remem-
support domain-general executive processes, while the VIPC may bered. The left temporoparietal junction has been suggested as the
support phonological encoding – recoding processes central to a variety anatomic substrate of such a store, as it is the most commonly
of language tasks.
reported lesion site for patients with a selective deficit of verbal
D 2004 Elsevier Inc. All rights reserved.
working memory (Shallice and Vallar, 1990).
Keywords: Inferior parietal cortex; Verbal working memory; Phonologic
Imaging studies of verbal working memory have attempted to
short-term store; N-back; Item recognition localize the site of the phonological store more precisely (e.g.,
Paulesu et al., 1993; Smith et al., 1996). However, two distinct
regions of the supramarginal gyrus are claimed to function as the
phonological STS (Becker et al., 1999)—a region of the inferior
* Corresponding author. Room 640, Learning and Research Develop-
parietal cortex with a more dorsal focus [dorsal aspect of the
ment Center, University of Pittsburgh, 3939 O’Hara Street, Pittsburgh, PA inferior parietal cortex (DIPC)] and a region with a more ventral
15260. Fax: +1-412-624-9149. focus [ventral aspect of the inferior parietal cortex (VIPC)] near the
E-mail address: [email protected] (S.M. Ravizza). site associated with short-term memory deficits in patients. While
Available online on ScienceDirect (www.sciencedirect.com.) these regions are often considered to function homogenously

1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2004.01.039
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 563

(Bunge et al., 2000; Cabeza et al., 2002; Clark et al., 2000; Henson DIPC and VIPC activity using simple load or information-type
et al., 2000; Tsukiura et al., 2001), one way to reconcile the contrasts. Toward this end, we examined parietal activity by
paradoxical neuroimaging and neuropsychological findings is to observing load effects in the verbal condition alone, in line with
propose that these regions make different contributions to verbal previous verbal working memory studies (Cohen et al., 1997;
working memory tasks. D’Esposito et al., 1998; Jonides et al., 1997). Moreover, to
For instance, most studies reporting DIPC activity have con- replicate previous information type effects (Paulesu et al., 1993,
trasted memory load (e.g., delay interval, number of items), 1996; Salmon et al., 1996), we assessed whether parietal activity
whereas studies reporting VIPC activity have contrasted verbal was greater in the verbal than the nonverbal, high-load condition.
and nonverbal working memory conditions. These findings suggest In the next section, we determined whether either parietal region
that the DIPC is recruited in memory-intensive conditions when displayed the interaction effect that we argue is predicted for a
attentional demands are high. On the other hand, the VIPC appears phonological store by the Baddeley model. Finally, we examined
to be selectively active for verbal material regardless of memory parietal regions that exhibited a load effect in the nonverbal
load. To test these accounts of DIPC and VIPC activation, load and memory task to assess whether these areas were selective for
information type (verbal – nonverbal) were manipulated in a single verbal items.
scanning session; this allowed us to determine whether both DIPC
and VIPC foci of activation could be found within the same group
of subjects and, if so, whether these regions showed functionally Methods
distinct patterns of activation. Based upon prior observations of
inferior parietal activity during verbal working memory tasks, we Participants
hypothesize that the DIPC and VIPC are functionally dissociable,
with the DIPC demonstrating sensitivity to load but not informa- Twenty-one right-handed adults (10 male, 11 female) with ages
tion type and the VIPC showing sensitivity to information type but ranging from 18 to 37 were paid US$46 to participate in this
not load. experiment (1.5 T: n = 10; 3 T: n = 11). All provided informed
More specifically, performance on two working memory tasks consent following procedures approved by the Institutional Review
that have been previously associated with parietal activation were Board at the University of Pittsburgh. One participant’s N-back
compared. The N-back task was employed, because it has evoked data had to be excluded due to poor task compliance.
robust activity of the DIPC in previous studies (Cohen et al., 1997;
D’Esposito et al., 1998; Jonides et al., 1997). Given that the Stimuli
majority of the working memory studies reporting VIPC activity
have used an item-recognition paradigm, participants were tested The verbal stimuli consisted 18 English letters (B, C, D, F, G,
on this task as well. Moreover, both load (high or low) and H, J, K, M, N, P, Q, R, S, T, V, X, and Z). A set of 18 Korean letters
information type (verbal or nonverbal) were manipulated in these that looked least like English letters (a, chiuch, hiuh, i, iung,
working memory tasks to assess inferior parietal activity across the khukh, kiyek, mium, niun, o, ou, phuph, piup, sios, tikut, xpiup,
entire set of conditions. For the verbal versus nonverbal contrast, xsios, and ye) comprised the nonverbal set (see Fig. 1). All of the
English and Korean letters were used as stimuli, since prior studies participants confirmed that they could not read Korean.
have shown greater VIPC activation for English versus Korean
letters (Paulesu et al., 1993, 1996; Salmon et al., 1996). Procedure
If the DIPC and VIPC display the response we have predicted
to manipulations of load and information type, their suitability as The experiment was identical regardless of the scanner used to
the locus of the STS is in question. Based upon Baddeley’s model collect images. Versions of the N-back and item-recognition tasks
of verbal working memory, the phonological store might be were created to manipulate both load and information type. Thus,
expected to show both effects of load (since more information for each task, there were four conditions—high-load verbal, low-
would need to be encoded and maintained in the store as the load is load verbal, high-, and low-load nonverbal (see Figs. 1 and 2). For
increased) and effects of item type (since the store is thought to be the N-back task, items were presented every 3 s. In the low-load
specifically dedicated to the maintenance of verbal information). version of the N-back (0-back) task, participants were asked to
More precisely, a within-subjects design should reveal an interac- press a button with their right index finger if a specific target
tion between load and information type in a region that functions as appeared. The target was the letter ‘‘X’’ in verbal blocks and was
a dedicated phonological short-term store, with robust activation the Korean letter ‘‘sios’’ in the nonverbal conditions. If any other
for the verbal high-load condition and low activation for all other item besides the target appeared on the screen, participants were
conditions. Thus, if our hypothesis is upheld, it would suggest that asked to press a button with their right middle finger. In the high-
neither the DIPC nor the VIPC exhibits a pattern of activity load version of the N-back task (3-back), participants determined
consistent with current conceptions of phonological storage and whether an item was the same as one at three trials back. If the item
that alternative interpretations of both DIPC and VIPC function was the same, participants pressed the button under their right
should be considered. index finger. Participants pressed the button under their right
middle finger if the item was different than the one presented three
trials back. In the verbal 3-back task, participants were encouraged
Experiment to rehearse the letters presented in the last three trials while
continuously updating their list as each new letter appeared. Items
To increase the reliability of our results, we tested two separate were visible for 500 ms and were followed by a fixation cross that
groups of participants using either a 1.5- or a 3-Tesla (T) magnet on appeared for 2500 ms. Twelve items were presented in each block
identical tasks. Our first aim was to replicate previous reports of of trials, so that each block lasted 36 s. The probability of an item
564 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573

Fig. 1. Examples of the 0- and 3-back tasks for both the English and Korean conditions in the N-back paradigm.

being a target was 33%, whereas new distracters and repeated items on the list or a button with their middle finger if it did not.
distracters appeared 47% and 20% of the time, respectively. Five lists were presented in a block of trials, and each block lasted
The item-recognition task began with the serial presentation of 56 s.
six items followed by a short delay and then the presentation of a Each participant in the scanner performed five runs of both the
probe item (see Fig. 2). Each item was visible for 500 ms, after N-back and item-recognition tasks. A run was composed of five
which a fixation cross would appear for 500 ms. After the set of blocks—each combination of the English or Korean and high- or
six items had been presented, a short delay of 2000 ms occurred low-load conditions and one fixation control block. For the fixation
and was succeeded by the appearance of a probe item for another block, participants were instructed to fixate on the cross presented
2000 ms. In the low-load condition (detect), participants deter- in the center of the screen. The probability of a run beginning with
mined whether a specific target had been in the list of six items. the English or Korean conditions was 50%. Load was counter-
The probe was the same as those used in the 0-back task described balanced across participants so that half began each run with the
previously (i.e., ‘‘X’’ or ‘‘sios’’). Conversely, participants were low-load conditions and half began with the high-load conditions.
unaware of what the probe item would be in the high-load version The fixation block was always the third block in the run. Thus, a
of the task (rehearse) and were encouraged to rehearse the set of run for the N-back task might be (1) Korean – 0-back, (2) Korean –
six letters in the English condition. The probability that the 3-back, (3) Fixation, (4) English – 0-back, or (5) English – 3-back.
subsequent probe matched one of the six items in the list was All participants received one training block on both working
50%. Items that matched the probes could be presented in any memory tasks before being scanned.
portion of the list in the rehearse condition, while in the detect
condition, they were weighted to occur more frequently in the final Functional MRI protocol
three items to increase the likelihood that participants would pay
attention to the entire set of items. Participants pressed a button Ten participants performed the experiment in a 1.5-T GE Signa
with their index finger if the probe item matched one of the six scanner. The scanning session began by obtaining 30 T1-weighted
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 565

Fig. 2. Examples of the detect and rehearse tasks for both the English and Korean conditions in the item-recognition paradigm.

images for anatomical localization (0.9375  0.9375  3.8 mm). Images were first corrected for motion using a six-parameter
Participants then performed the experiment while T2*-weighted rigid-body automated image registration (AIR) algorithm and then
gradient echo, 1-shot spiral scans were acquired parallel to the detrended with a linear regression to remove scanner drift. We
AC – PC line. Twenty-six 3.8-mm, contiguous oblique-axial slices chose one participant’s structural scans from each scanner group to
were obtained every 2 s with each voxel having an in-plane serve as a common reference brain that was then used to coregister
resolution of 3.75  3.75 mm. (TR = 2 s; TE = 35 ms; flip angle all participants’ structural and functional images within that group.
= 70j). Functional images were acquired in the same plane as the These reference brains were also transformed into Talairach space
anatomical scans, but coverage was limited to the top 26 slices— using the AFNI program (Cox, 1996). To minimize differences in
extending from superior portions of the cortex to more superior intensity between participants, functional images were normalized
regions of the cerebellum. This protocol permitted us to acquire 90 by scaling each image to a global mean intensity. These images
and 140 volumes for each of the five conditions in the N-back task were then smoothed using an 8-mm full-width, half-maximum
and item-recognition tasks, respectively. Gaussian kernel to reduce anatomical differences between subjects.
Eleven participants performed the experiment in a 3-T GE
Signa scanner. The scanning session began by obtaining 36 T1- Functional MRI data analysis
weighted images for anatomical localization (0.78125  0.78125
 3.2 mm). Acquisition was identical to that reported for the 1.5-T All ANOVAs were computed using the Neuroimaging Soft-
magnet (1-shot, spiral scans; see above), except that thirty-six 3.2- ware package (NIS 3.5;, http://kraepelin.wpic.pitt.edu/nis/) devel-
mm slices were obtained with each voxel having a resolution of oped at the University of Pittsburgh and Princeton University. The
3.125  3.125 mm. (TR = 2 s; TE = 18 ms; flip angle = 80j). images acquired for each participant in each condition were first
566 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573

averaged and then subjected to the ANOVA program that gener- ROI analyses
ated spatial F-maps. By averaging images across participants, we To determine the full pattern of load and information type
were able to determine reliable areas of activity within the group effects, we then selected a subset of the general areas identified by
while discounting individual differences in functional anatomy. these voxelwise ANOVAs for further analysis. The selected areas
Given that we were using coordinates averaged across multiple consisted of our a priori regions of interest in parietal cortex (DIPC
studies to define the VIPC and DIPC, it was more appropriate to and VIPC) and a set of comparative areas previously associated
analyze the group data to compare our results with previous with the specific rehearsal of verbal information [left Broca’s area
studies. For these tests, subject was designated as a random factor. (BA), the SMA, and the cerebellum] and general executive
Moreover, ROIs had to survive a threshold of P < 0.005 and processing (left DLPFC and right DIPC).
consist at least four contiguous pixels to be considered a source of Our goal was to determine the typical pattern of activity across
significant activation. Main and interaction effects were assessed conditions, regardless of task, statistical contrast, magnet (1.5 or 3
using a critical F value of 13.61 [ F(1,9) = 13.61, P < 0.005]. If T), or precise location of the peak of activation. For each general
post hoc tests were necessary to determine significance between area of interest, there were potentially eight contributory locations
multiple variables, a t test was performed using the mean intensity within a given region found in our first analysis step: the specific
values for each participant in a given condition for a particular regions of significant activation found in our contrasts of load in
ROI. the N-back task, information type in the N-back task, load in the
It is important to note that our threshold in the voxelwise test item-recognition task, and information type in the item-recognition
results in a mapwise false-positive rate that exceeds 0.05 across the task for participants scanned in the 1.5- and 3-T scanners. In cases
entire set of brain voxels. However, the primary aim of these tests where multiple regions of activation were found in a single area for
was to serve as a method of localization for our a priori regions of a given contrast and group (e.g., the two left DIPC regions in the
interest so that we could explore the pattern of parietal, frontal, and contrast of load in the item-recognition task for participants in the
cerebellar activity across the full set of conditions. By using 3-T group), we selected the region that was closest to the center of
functionally defined regions rather than predefined ROIs based the search criteria for our two parietal regions or (for areas outside
on previous literature, we ensure that the maximally affected of the parietal lobe) the region that was most consistent with the
voxels lie within a given area. Moreover, as we predicted that peak coordinates obtained across other contrasts in our data. First,
some regions would not be modulated by certain manipulations we averaged the data produced by each group in each region,
(e.g., DIPC should not show an effect of information type), we felt combining the data from the first subject scanned in the 1.5- and 3-
that a more liberal criterion would actually serve as a more T magnets, the data from the second subject scanned in the 1.5-
conservative test of our predictions. and 3-T magnets, and so forth. After averaging across magnets, we
averaged the signal values obtained from locations within each
Voxelwise tests region that was identified by the load and information contrasts for
To compare our results to previous verbal working memory the N-back task, and we averaged the signal values obtained in the
studies, we began by performing the analyses typically used in load and information-type contrasts for the item-recognition task.
these studies. Thus, we performed separate load and information Finally, the averaged values for each task were averaged together
type voxelwise ANOVAs for the N-back and item-recognition tasks to obtain one set of values in each condition. The resulting set of
to see if we could replicate load effects in the DIPC and informa- signal values was then analyzed using an ANOVA with load and
tion type effects in the VIPC. information type as factors, for each of our six general areas of
A region was considered part of the DIPC if its peak was interest (left VIPC, left DIPC, right DIPC, left DLPFC, SMA, left
centered in the x and y dimension F1.5 cm (approximately 2 SD) Broca’s area, and left and right cerebellum).
away from mean coordinates (x = 34, y = 51) derived from 23
studies of working memory (Awh et al., 1996; Barch et al., 1997;
Braver et al., 1997; Bunge et al., 2000; Callicott et al., 1999; Chein Results and discussion
and Fiez, 2001; Clark et al., 2000; Cohen et al., 1997; Coull et al.,
1996; Davachi et al., 2001; D’Esposito et al., 1998; de Zubicaray et Given that the procedure was identical for both groups (1.5 and
al., 1998; Henson et al., 2000; Jonides et al., 1997, 2000; Marsh- 3 T), they are discussed in the same section, although the results
uetz et al., 2000; Petrides et al., 1993; Reuter-Lorenz et al., 2000; are reported separately for each magnet in the tables.
Rypma et al., 1999; Salmon et al., 1996; Schumacher et al., 1996;
Smith et al., 1996; Tsukiura et al., 2001). Moreover, the peak had Behavioral data
to fall within the range z = 32 to 52 mm reported in previous
studies. To be considered part of the VIPC, the peak activity had to Participants’ accuracy in each task was analyzed using a 2  2
be within F1.5 cm of the x and y dimensions (x = 52 mm, y =  2  2 (task  load  information type  magnet) repeated-
27 mm) obtained from averaging coordinates across studies measures ANOVA. Main effects for load [ F(1,18) = 195.23, P <
reporting VIPC activity (Becker et al., 1996; Paulesu et al., 0.001] and information type [ F(1,18) = 23.68, P < 0.001] were
1993, 1996; Salmon et al., 1996) and also fall within the range obtained, as well as the interaction effect of these two variables
of z = 10 to 30 mm. Although inferior parietal cortex ends at [ F(1,18) = 15.19, P < 0.001] (see Fig. 3). An analysis of simple
approximately z = 14 mm, we considered that the VIPC and effects confirmed that the difference between the English and
portions of the superior temporal gyrus may not be functionally Korean high-load conditions was significant [t(19) = 4.56, P <
distinct given that short-term memory patients have lesions that 0.001], whereas the difference between the low-load conditions
encompass the temporoparietal junction. Furthermore, this range was not ( P > 0.1). Furthermore, participants were slightly less
more closely approximates the height criteria used to define areas accurate in the item-recognition task in general than the N-back
of the DIPC. task (92% vs. 93%), although the effect was not significant
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 567

data from both magnets to separate repeated-measures ANOVAs

using mean intensity values from the English 3- and 0-back
conditions. As expected, in the 1.5-T group, a number of regions
were affected by load (see Table 1), including areas generally
associated with central executive processes (e.g., dorsolateral
prefrontal cortex) and articulatory rehearsal (e.g., the supplemen-
tary motor area, the left premotor cortex, Broca’s area, and the
cerebellum). Note that the DIPC, but not the VIPC, was signifi-
cantly active in this load contrast—a result in line with previous N-
back experiments. Moreover, these results were replicated by the
participants scanned in the 3-T magnet. The DIPC was active in
both studies regardless of the magnet used, whereas VIPC activity
Fig. 3. Accuracy in the N-back and item-recognition tasks for Groups 1 and 2. was not detected by a load contrast for either group.
In contrast, VIPC activity is usually reported when verbal and
nonverbal high-load conditions are compared. When we compared
[ F(1,18) = 3.16, P = 0.093]. Neither a main effect of magnet nor an the English and Korean 3-back tasks in repeated-measures
interaction of magnet with any other variable was obtained. ANOVAs, we found both DIPC and VIPC activities for partic-
The fact that participants were more accurate in the low-load ipants scanned with the 1.5-T magnet, but only VIPC activity for
conditions confirmed that our load manipulation was effective in those scanned with the 3-T magnet (see Table 2). Thus, the VIPC,
increasing task difficulty. Moreover, the difference in accuracy but not the DIPC, displayed robust effects of information type.
between the English and Korean conditions indicates that partic-
ipants were using a less effective strategy to perform the task when Imaging data—item recognition
having to remember Korean letters. Although it is possible to
verbally encode Korean letters (e.g., ‘‘ring,’’ ‘‘lambda’’), the ease To assess load effects for this task, we compared scans acquired
of encoding English letters provided an advantage for the verbal in the English rehearse (high-load) condition to those in the English
items that were especially apparent in the high-load conditions. detect condition (low-load). Neither parietal region was active in
this contrast, and concerned that our low-load condition placed an
Imaging data—N-back observable demand on memory, we additionally compared the
English rehearse task to the fixation control condition. For both
Studies reporting DIPC activity have typically compared verbal experiments, DIPC was recruited more heavily in the high-load
high- and low-load conditions. Thus, we subjected the imaging condition than in the fixation condition. In contrast, the VIPC was

Table 1
Neural regions showing a consistent effect of load in the English conditions (P < 0.005)

Except for the inferior parietal cortex, only regions that were active in two of four experiments are listed.
568 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573

Table 2
Neural regions in Experiments 1 and 2 showing an effect of information type in the high-load conditions (P < 0.005)

Except for the inferior parietal cortex, only regions that were active in two of four experiments are listed.

not apparent using a load contrast. Thus, the DIPC displayed load articulatory loop (e.g., Broca’s area, SMA, and bilateral cerebel-
effects in both the N-back and item-recognition paradigms, whereas lum). Moreover, we wanted to assess the pattern of activity of the
the VIPC was not affected by load in either task. right homologue of the left DIPC. Activity of the right DIPC was
Information type effects (verbal vs. nonverbal) in the item- observed in this and other verbal working memory experiments
recognition task were examined by a repeated-measures ANOVA (Cohen et al., 1997; Henson et al., 2000; Jonides et al., 1997;
using scans acquired in the English and Korean rehearse condi- Tsukiura et al., 2001) and may be part of a general attentional
tions. For those scanned with the 1.5-T magnet, we were unable to network (Chein et al., in press; Corbetta et al., 2000; Wojciulik and
find inferior parietal involvement or indeed any neural activity that Kanwisher, 1999). Activity of an additional executive region, the
was greater in the English condition in our voxelwise analyses. For left DLPFC, was also observed across conditions. Like the left
the 3-T group, a region of the VIPC/superior temporal gyrus DIPC, the left DLPFC is often reported in both verbal and spatial
displayed greater activity in the English than in the Korean working memory studies but has been claimed to be more
conditions. Thus, in both memory paradigms, the VIPC was important for remembering verbal items (Reuter-Lorenz et al.,
engaged by verbal stimuli, whereas the DIPC displayed less robust 2000; Smith et al., 1996; although see D’Esposito et al., 1998).
effects of the type of information to be remembered. Regions associated with articulatory rehearsal displayed a
The null result of information type for images acquired in the significant interaction effect [Broca’s area: F(1,9) = 9.06, P <
1.5-T magnet is somewhat surprising given that our version of the 0.05; SMA: F(1,9) = 30.44, P < 0.001; right cerebellar cortex:
item-recognition task was almost identical to those used in previ- F(1,9) = 9.98, P < 0.05; left cerebellar cortex: F(1,9) = 8.02, P <
ous studies of working memory for English and Korean letters. 0.05] (see Fig. 4 for a subset of these areas). All of these regions
One important difference between the studies reporting VIPC displayed the same pattern of results; that is, they were recruited
involvement and the experiment reported here is the methodology more heavily in the English high-load condition than all three other
used to image neural activity. Our experiment implemented fMRI conditions (all Ps < 0.05). Thus, neural markers of articulatory
to explore verbal working memory effects, while previous studies rehearsal performed in a way consistent with Baddeley’s model.
finding VIPC activity have all used PET (Becker et al., 1996; The VIPC also displayed a significant interaction effect [ F(1,9)
Paulesu et al., 1993, 1996; Salmon et al., 1996). PET imaging = 26.15, P < 0.001] but did not behave as would be predicted of a
differs from fMRI in ways that may produce differences in phonological short-term store. Although activity was greater in the
sensitivity to neural activity, especially in high-order linguistic verbal high-load condition than the nonverbal [t(9) = 8.42, P <
tasks (Veltman et al., 2000), and so a small difference between the 0.001], this area showed no effect of load in the verbal conditions.
English and Korean conditions may be less detectable using fMRI A neural substrate of the short-term store should show a load effect
than PET. Although the timing of acquisition was identical for both given that the area would be periodically refreshed by the rehearsal
scanners, boosting the signal strength may have allowed us to system in the high-load but not the low-load condition. Moreover,
detect responses with the 3-T magnet in the item-recognition task, this area was just as active in conditions that place little demand on
which were below our statistical threshold with the 1.5-T magnet. working memory (i.e., the fixation and Korean low-load condi-
tions) as in the memory-intensive verbal condition. In contrast to
Interaction effects across tasks rehearsal areas, the VIPC did not behave in a way that corresponds
to current notions of the phonological short-term store.
A simple prediction of the Baddeley model is that regions The DIPC fares no better in this regard. This area did not display
selectively involved in either articulatory rehearsal or phonological an interaction effect of load and information type [ F(1,9) = 3.08, P =
storage should be affected by the interaction of load and type of 0.113] and exhibited only a marginal effect of information type
information, with robust activation observed only in the verbal [ F(1,9) = 4.92, P = 0.054]. However, a main effect of load was
high-load condition. Specifically, markers of storage and rehearsal obtained [ F(1,9) = 44.32, P < 0.001]. The DIPC was primarily
should display greater activity in the English high-load conditions recruited when load was heavy regardless of the linguistic nature of
(i.e., 3-back, rehearse) than in all the other conditions. Of interest is the items to be remembered. In fact, this region tended to be more or
the pattern of activity of six regions previously associated with equally active in the Korean high-load than the English high-load
either phonological storage (i.e., left DIPC, left VIPC) or the condition in the item-recognition task [1.5 T: English (Eng) =
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 569

Fig. 4. Overall pattern of activity across the entire set of conditions for the VIPC, Broca’s area, the right cerebellum, the left and right DIPC, and left DLPFC.

2189.97, Korean (Kor) = 2190.32; 3 T: Eng = 2224.31, Kor = the criteria we established for the simple load and information-type
2224.42]. Moreover, the DIPC displayed the same pattern of activity contrasts (see Voxelwise tests). For both tasks and magnets, the left
across conditions as the two attentional – executive regions—the DIPC showed greater activity in the high-load nonverbal condition
right DIPC and left DLPFC. Neither of these latter regions showed a than in the low-load nonverbal condition (1.5-T N-back = 35,
significant interaction effect [rDIPC: F(1,9) = 3.88, P = 0.08; 50, 43/ 23, 68, 36; 1.5-T item recognition = 27, 62, 45/
lDLPFC: F(1,9) = 0.02, P = 0.889], but both were recruited when 44, 40, 43/ 32, 47, 35; 3-T N-back = 48, 44, 43; 3-T
load was high regardless of information type [rDIPC: F(1,9) = 44.98, item recognition = 36, 46, 38/ 26, 59, 34). In contrast,
P < 0.001; lDLPFC: F(1,9) = 47.7, P < 0.001]. activity of the VIPC did not emerge when memory load was high
To assess whether the VIPC and DIPC were functionally in the nonverbal conditions.
distinct, we directly compared signal levels from these regions in
a 2 (area)  2 (load)  2 (information type) ANOVA. The three-
way interaction was significant [ F(1,9) = 17.24, P < 0.005]. General discussion
Paired-sample t tests using the difference scores demonstrated that
(a) there was a greater difference between the English high- and Previous attempts at locating the neural instantiation of the
low-load condition in the DIPC than the VIPC [t(9) = 2.4, P < phonological short-term store have assessed the effects of mem-
0.05], (b) the difference between the Korean high- than low-load ory load or the type of stimuli to be encoded on brain activity. A
conditions was positive for the DIPC and negative for the VIPC
[t(9) = 7.96, P < 0.001], (c) there was less of a difference between
the English and Korean high-load conditions in the DIPC than the
VIPC [t(9) = 4.1, P < 0.005], and (d) there was less of a difference
between the English and Korean low-load conditions in the DIPC
than VIPC [t(9) = 2.77, P < 0.05] (see Fig. 5).
The same analysis was performed on the signal values obtained
for the left DIPC in comparison to the right DIPC. None of the
interactions of load and information type with area were significant
( Ps > 0.1). Thus, the left DIPC was not functionally dissociable
from the right DIPC.

Nonverbal working memory

Our results suggest that the DIPC is active when memory load
is high regardless of information type. To provide further evidence
for this claim, we performed voxelwise ANOVAs comparing the Fig. 5. Difference in mean signal values in simple load and information type
imaging data from the Korean high- and low-load conditions using contrasts for the VIPC and DIPC.
570 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573

region that was more active when memory load was high or Dorsal inferior parietal cortex
when linguistic stimuli were used was considered a potential
candidate for such a store. Studies comparing verbal and non- Across magnets and tasks, the DIPC displayed robust load
verbal memory pointed to the VIPC as the site of the phonolog- effects. However, we predicted that a region acting as a phono-
ical STS, whereas others that manipulated memory load indicated logical short-term store should be more active in verbal than
involvement of the DIPC. In studies where load and information nonverbal memory conditions. The DIPC (1) did not show greater
type were manipulated in the same set of subjects, direct activity for verbal items, (2) was recruited more heavily in the
comparisons of load but not information type were performed nonverbal than the verbal high-load condition in the item-recog-
(D’Esposito et al., 1998; Reuter-Lorenz et al., 2000; Smith et al., nition task, (3) was not functionally dissociable from the right
1996). Thus, confusion arose regarding the locus of the phono- DIPC, and (4) displayed load effects in the nonverbal task. These
logical STS, because load and information type were almost results suggest that the DIPC may be acting as part of a frontal –
never manipulated or contrasted in the same design. It was parietal executive system (Corbetta et al., 2000; Posner and
unclear whether these anatomically nonoverlapping regions were Dehaene, 1994). In our experiment, DIPC activity was modulated
functionally distinct or whether the peak of parietal activity by load and information type similarly to two other regions, the left
varied widely across studies. By manipulating both these varia- DLPFC and right DIPC, posited to be part of a general attentional
bles in the same design, we were able to demonstrate that these network (Smith et al., 1996; Wojciulik and Kanwisher, 1999).
regions of the inferior parietal cortex behave quite differently in Indeed, some researchers have speculated that the inferior parietal
response to variations of load or the type of the items to be cortex serves to focus attention on items in working memory rather
remembered. Moreover, the areas responded in ways that we than serving as a phonological store per se (Chein et al., in press;
predicted; the DIPC was more sensitive to load manipulations, Cowan, 1999). As part of a frontal – parietal executive system, it is
whereas the VIPC showed a preference for the type of material to not surprising that the DIPC is involved in working memory tasks
be remembered. employing a variety of verbal, spatial, and visual stimuli. This
Based on our experiment, we can also explain why inferior region may be important for retaining temporal order information
parietal locations have varied across previous working memory (Marshuetz et al., 2000), attentionally reactivating sources of
studies. When low-load conditions are subtracted from high-load information in neural regions (Corbetta et al., 2002), rapid switch-
conditions, DIPC, but not VIPC, activity is manifested. This is in ing of attention (La Bar et al., 1999), or preparing for a given task
accordance with reports of DIPC activity in a wide variety of (Sohn et al., 2000)—all domain-general functions that may be
working memory tasks that implement a control condition that tapped in verbal working memory tasks.
places little or no demand on working memory. Indeed, the only
study to manipulate and directly contrast both load and information Ventral inferior parietal cortex
type in the same design found greater activity in the left DIPC in
the object working memory condition than in the verbal working On the surface, the VIPC would appear to be the most likely
memory condition (Nystrom et al., 2000). candidate for the phonological short-term store. Activation of this
We also were able to demonstrate that VIPC activity is area has been reported in working memory tasks manipulating
contingent upon the type of information to be remembered—a information type (Paulesu et al., 1993, 1996; Salmon et al., 1996)
result consistent with earlier reports of VIPC activity in studies and in passive listening paradigms (Fiez et al., 1996; Petersen et al.,
that used a nonverbal condition with a similar load as a control. 1998). Moreover, patients suffering from a selective short-term
Although DIPC activity tended to be higher for English letters in memory deficit have damage at or near this area (Shallice and Vallar,
the high-load condition in the N-back task, the effect was 1990; Vallar and Papagno, 1995). Accordingly, the VIPC was more
reversed in the item-recognition task—a result that may explain active in the English high-load condition compared to the Korean
why this area has not been found in item-recognition studies high-load condition in three of the four information-type contrasts.
comparing verbal and nonverbal stimuli (Paulesu et al., 1993, Despite fitting so many of the criteria of a short-term store, this
1996; Salmon et al., 1996). Martin et al. (in press) have reported region displayed two inconsistent findings. First, there was no
similar results in their study of phonological and semantic verbal difference between the verbal high- and low-load conditions.
working memory. They took the parietal ROIs reported in Although inconsistent with notions of a dedicated phonological
Becker et al. (1999) and examined how they were affected by store, this results in line with evidence of VIPC involvement in
load and task type (i.e., semantic or phonological). They found more basic speech processing such as phonological discrimination
that the DIPC was recruited more heavily when load was high and identification tasks (Caplan et al., 1995), reading (Paulesu et
regardless of the type of task being performed, whereas the al., 1996; Rumsey et al., 1997), the mediation between auditory
VIPC was unaffected by memory demand. Instead, VIPC activ- and articulatory representations (Hickok and Poeppel, 2000), and
ity was affected by whether the task was phonological or auditory imagery (McGuire et al., 1996; Shergill et al., 2001).
semantic; activity of this region was greater when the task Thus, the VIPC may be more sensitive to the amount of phono-
was phonological. logical encoding or recoding that occurs in a given condition,
This experiment strongly supports our hypothesis that DIPC rather than how much needs to be recalled.
and VIPC are functionally dissociable. Our use of a complete The other result that conflicts with its potential role in phono-
experimental design allowed us to replicate prior observations of logical storage concerns the greater involvement of the VIPC in
load effects in DIPC and information type in VIPC, but it also conditions where memory load was low or nonexistent (i.e., 0-
revealed unexpected complexities in the activation patterns. As back, detect, fixation). A study by Greicius et al. (2003) reported a
discussed below, neither region behaves in a manner entirely similar result; that is, they found the VIPC to be more active in a
consistent with current ideas of how verbal items are maintained spatial 0- than 2-back condition. It may be that activity in this area
in memory. is inhibited by the central executive in cases where people are
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 571

purposefully trying to maintain items in working memory. In the Conclusions

verbal conditions, once the items are encoded and placed in short-
term memory, it may be sensible for the store to block further input Clearly, dorsal and ventral regions of the supramarginal gyrus
from more basic processing areas. A certain amount of noise may respond in a distinct and dissociable manner during the performance
be introduced into the phonological buffer if phonological pro- of working memory tasks. Indeed, their performance was quite
cessors are allowed unrestricted input into the short-term store. If disparate across a number of conditions. Whereas the VIPC was
this were true, one would have to assume that the VIPC is responsive to the type of information to be maintained but not to
suppressed regardless of the number of items being maintained. memory load, the DIPC displayed the inverse pattern. Furthermore,
In the nonverbal high-load condition, activity of nonessential areas VIPC activity was active in conditions with low memory demand,
may also be occurring due to its general difficulty, and in fact, whereas the DIPC was more active when memory load was high.
suppression of this area may be easier than in the verbal conditions, We have proposed several ideas to explain the pattern of activity
because there is no need to use the verbal system at all. observed in these two parietal regions. The VIPC may be involved in
Although these explanations of VIPC activity need to be tested, more basic speech processes that are suppressed during working
the evidence is not supportive of its role as a phonological short- memory tasks. Rather than being selective for verbal material, the
term store. This is puzzling given that some have suggested that DIPC may contribute to executive processes such as retaining
this area is damaged in patients exhibiting a verbal working temporal order (Marshuetz et al., 2000), attentionally reactivating
memory deficit without a concomitant impairment in phonological sources of information in neural regions (Corbetta et al., 2002), or
processing (Shallice and Vallar, 1990; Vallar and Papagno, 1995). task preparation (Sohn et al., 2000). However, neither area is
Evidence for this claim, however, is based on the occurrence of affected by load and information type effects in a way predicted
common lesion sites reported across studies of short-term memory by a region serving as a phonological short-term store.
patients; thus, the types of tests used to assess spatial working
memory and phonological processing deficits differ widely across
these studies. Only one study has examined lesion sites within a Acknowledgment
large group of patients with verbal short-term memory deficits
(Bartha and Benke, 2003). This study finds that the common area This research was supported by NIH grant RO1 MH59256.
of damage for conduction aphasics with low verbal spans but
normal auditory processing and spatial spans was in an inferior
temporal region (BA 37). Although this region of the temporal lobe References
is not commonly reported in imaging studies of verbal working
memory, it is possible that activity of a phonological store occurs Awh, E., Jonides, J., Smith, E.E., Schumacher, E.H., Koeppe, R.A., Katz,
so fleetingly that it goes undetected by current imaging techniques. S., 1996. Dissociation of storage and rehearsal in verbal working mem-
ory: evidence from positron emission tomography. Psychol. Sci. 7 (1),
Rehearsal-related regions 25 – 31.
Baddeley, A., Hitch, G., 1974. Working memory. In: Bower, G.H. (Ed.),
The Psychology of Learning and Motivation. Academic Press, San
While neither parietal region displayed the predicted interaction
Diego, pp. 47 – 90.
effect of load and information type, we also examined the
Baddeley, A., Gathercole, S., Papagno, C., 1998. The phonological loop as
activation in areas commonly associated with articulatory rehearsal a language learning device. Psychol. Rev. 105 (1), 158 – 173.
in verbal working memory: Broca’s area, the SMA, and bilateral Barch, D., Braver, T.S., Nystrom, L.E., Forman, S.D., Noll, D.C., Cohen,
cerebellar cortex. All these regions behaved in a way consistent J.D., 1997. Dissociating working memory from task difficulty in human
with their presumed role in verbal working memory. Specifically, prefrontal cortex. Neuropsychologia 35 (10), 1373 – 1380.
they were recruited most heavily in the verbal, high-load condi- Bartha, L., Benke, T., 2003. Acute conduction aphasia: an analysis of 20
tion. One explanation for these results is that, as suggested by cases. Brain Lang. 85 (1), 93 – 108.
Baddeley and Hitch (1974), the rehearsal system refreshes verbal Becker, J.T., Mintun, M.A., Aleva, K., Wiseman, M.B., Nichols, T.,
items in a short-term store, but that activity of the store is so DeKosky, S.T., 1996. Compensatory reallocation of brain resources
supporting verbal episodic memory in Alzheimer’s disease. Neurology
transient that it is not detected by current imaging techniques.
46 (3), 692 – 700.
Alternatively, our findings may indicate that verbal working
Becker, J.T., MacAndrew, D.K., Fiez, J.A., 1999. A comment on the func-
memory can be carried out solely by articulatory mechanisms tional localization of the phonological storage subsystem of working
(Gruber, 2001). Additional evidence for this claim is provided memory. Brain Cogn. 41 (1), 27 – 38.
from a study by Chein and Fiez (2001) who reported sustained Braver, T.S., Cohen, J.D., Nystrom, L.E., Jonides, J., Smith, E.E., Noll,
activation of Broca’s area, but not parietal cortex, during the D.C., 1997. A parametric study of prefrontal cortex involvement in
maintenance interval of a delayed serial recall task. Moreover, it human working memory. NeuroImage 5, 49 – 62.
has been suggested that Broca’s area supports phonological storage Bunge, S.A., Klingberg, T., Jacobsen, R.B., Gabrieli, J.D., 2000. A re-
as well as rehearsal. Sakai et al. (2002) asked participants to source model of the neural basis of executive working memory. Proc.
perform a distractor task during the memory delay of an item- Natl. Acad. Sci. 97 (7), 3573 – 3578.
Cabeza, R., Dolcos, F., Graham, R., Nybert, L., 2002. Similarities and
recognition task. After the distractor interval, Broca’s area and the
differences in the neural correlates of episodic memory retrieval and
parahippocampus were active at retrieval, but the parietal cortex
working memory. NeuroImage 16, 317 – 330.
was not, prompting Sakai et al. (2002) to suggest that Broca’s area Callicott, J.H., Mattay, V.S., Bertolino, A., Finn, K., Coppola, R., Frank,
may serve to store information that is then reactivated by the J.A., et al., 1999. Physiological characteristics of capacity constraints in
parahippocampus. Thus, evidence is accumulating for the notion working memory as revealed by functional MRI. Cereb. Cortex 9 (1),
that items can be stored and maintained by traditional rehearsal 20 – 26.
areas without recourse to the parietal cortex. Caplan, D., Gow, D., Makris, N., 1995. Analysis of lesions by MRI in
572 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573

stroke patients with acoustic – phonetic processing deficits. Neurology Marshuetz, C., Smith, E.E., Jonides, J., DeGutis, J., Chenevert, T.L.,
45 (2), 293 – 298. 2000. Order information in working memory: fMRI evidence for pa-
Chein, J.M., Fiez, J.A., 2001. Dissociation of verbal working memory rietal and prefrontal mechanisms. J. Cogn. Neurosci. 12 (Suppl. 2),
system components using a delayed serial recall task. Cereb. Cortex 130 – 144.
11 (11), 1003 – 1014. Martin, R.C., Breedin, S.D., 1992. Dissociations between speech percep-
Chein, J.M., Ravizza, S.M., Fiez, J.A., 2003. Using neuroimaging to eval- tion and phonological short-term memory deficits. Cogn. Neuropsy-
uate models of working memory and their implications for language chol. 9 (6), 509 – 534.
processing. J. Neurolinguist. 16 (4 – 5), 315 – 339. Martin, R.C., Wu, D., Freeman, M., Jackson, E.F., Lesch, M., 2003. An
Clark, C.R., Egan, G.F., McFarlane, A.C., Morris, P., Weber, D., Sonkkilla, event-related fMRI investigation of phonological versus semantic short-
C., et al., 2000. Updating working memory for words: a PET activation term memory. J. Neurolinguist. 16 (4 – 5), 341 – 360.
study. Hum. Brain Mapp. 9 (1), 42 – 54. McGuire, P.K., Silbersweig, D.A., Murray, R.M., Frackowiak, R.S.J.,
Cohen, J.D., Perlstein, W.M., Braver, T.S., Nystrom, L.E., Noll, D.C., Frith, C.D., 1996. Functional neuroanatomy of inner speech and
Jonides, J., et al., 1997. Temporal dynamics of brain activation during auditory verbal imagery. Psychol. Med. 26 (1), 29 – 40.
a working memory task. Nature 386, 604 – 608. Nystrom, L.E., Braver, T.S., Sabb, F.W., Delgado, M.R., Noll, D.C.,
Corbetta, M., Kincade, J.M., Ollinger, J.M., McAvoy, M.P., Shulman, G.L., Cohen, J.D., 2000. Working memory for letters, shapes and locations:
2000. Voluntary orienting is dissociated from target detection in human fMRI evidence against stimulus-based regional organization in human
posterior parietal cortex. Nat. Neurosci. 3 (3), 292 – 297. prefrontal cortex. NeuroImage 11, 424 – 446.
Corbetta, M., Kincade, J.M., Shulman, G.L., 2002. Neural systems for Paulesu, E., Frith, C.D., Frackowiak, R.S., 1993. The neural correlates
visual orienting and their relationships to spatial working memory. of the verbal component of working memory. Nature 362 (6418),
J. Cogn. Neurosci. 14 (3), 508 – 523. 342 – 345.
Coull, J.T., Frith, C.D., Frackowiak, R.S., Grasby, P.M., 1996. A fronto- Paulesu, E., Frith, U., Snowling, M., Gallagher, A., Morton, J., Fracko-
parietal network for rapid visual information processing: a PET study of wiak, R.S.J., et al., 1996. Is developmental dyslexia a disconnection
sustained attention and working memory. Neuropsychologia 34 (11), syndrome? Evidence from PET scanning. Brain 119 (1), 143 – 157.
1085 – 1095. Petersen, S.E., Fox, P.T., Mintun, M.A., Raichle, M.E., 1998. Positron
Cowan, N., 1999. An embedded-processes model of working memory. In: emission tomographic studies of the cortical anatomy of single-word
Miyake, A., Shah, P. (Eds.), Models of Working Memory: Mechanisms processing. Nature 331, 585 – 589.
of Active Maintenance and Executive Control. Cambridge Univ. Press, Petrides, M., Alivisatos, B., Meyer, E., Evans, A.C., 1993. Functional
New York, NY, pp. 62 – 101. activation of the human frontal cortex during the performance of
Cox, R.W., 1996. AFNI: software for analysis and visualization of func- verbal working memory tasks. Proc Natl. Acad. Sci. U. S. A. 90,
tional magnetic resonance neuroimages. Comput. Biomed. Res. 29, 878 – 882.
162 – 173. Posner, M.I., Dehaene, S., 1994. Attentional networks. Trends Neurosci. 17
Davachi, L., Maril, A., Wagner, A.D., Mail, E., 2001. When keeping in (2), 75 – 79.
mind supports later bringing to mind: neural markers of phonological Reuter-Lorenz, P.A., Jonides, J., Smith, E.E., Hartley, A., Miller, A.,
rehearsal predict subsequent remembering. J. Cogn. Neurosci. 13 (8), Marshuetz, C., et al., 2000. Age differences in the frontal lateralization
1059 – 1070. of verbal and spatial working memory revealed by PET. J. Cogn. Neu-
D’Esposito, M., Aguirre, G.K., Zarahn, E., Ballard, D., Shin, R.K., Lease, rosci. 12 (1), 174 – 187.
J., 1998. Functional MRI studies of spatial and nonspatial working Rumsey, J.M., Nace, K., Donohue, B., Wise, D., Ma, M.J., Andreason, P.,
memory. Cogn. Brain Res. 7 (1), 1 – 13. 1997. A positron emission tomographic study of impaired word recog-
de Zubicaray, G.I., Williams, S.C., Wilson, S.J., Rose, S.E., Brammer, M.J., nition and phonological processing in dyslexic men. Arch. Neurol. 54
Bullmore, E.T., et al., 1998. Prefrontal cortex involvement in selective (5), 562 – 573.
letter generation: a functional magnetic resonance imaging study. Cor- Rypma, B., Prabhakaran, V., Desmond, J., Glover, G., Gabrieli, J., 1999.
tex 34 (3), 389 – 401. Load dependent roles of frontal brain regions in the maintenance of
Fiez, J.A., Raichle, M.E., Balota, D.A., Tallal, P., Petersen, S.E., 1996. PET working memory. NeuroImage 9, 216 – 226.
activation of posterior temporal regions during auditory word presenta- Sakai, K., Rowe, J.B., Passingham, R.E., 2002. Parahippocampal reactiva-
tion and verb generation. Cereb. Cortex 6, 1 – 10. tion signal at retrieval after interruption of rehearsal. J. Neurosci. 22
Greicius, M.D., Krasnow, B., Reiss, A.L., Menon, V., 2003. Functional (15), 6315 – 6320.
connectivity in the resting brain: a network analysis of the default mode Salmon, E., Van der Linden, M., Collette, F., Delfiore, P., Maquet, P.,
hypothesis. Proc. Natl. Acad. Sci. U. S. A. 100 (1), 253 – 258. Degueldre, C., et al., 1996. Regional brain activity during working
Gruber, O., 2001. Effects of domain-specific interference on brain activa- memory tasks. Brain 119, 1617 – 1628.
tion associated with verbal working memory task performance. Cereb. Schumacher, E.H., Lauber, E., Awh, E., Jonides, J., Smith, E.E., Koeppe,
Cortex 11, 1047 – 1055. R.A., 1996. PET evidence for an amodal verbal working memory sys-
Henson, R.N.A., Burgess, N., Frith, C.D., 2000. Recoding, storage, re- tem. NeuroImage 3, 79 – 88.
hearsal and grouping in verbal short-term memory: an fMRI study. Shallice, T., Vallar, G., 1990. The impairment of auditory – verbal short-
Neuropsychologia 38 (4), 426 – 440. term storage. In: Vallar, G., Shallice, T. (Eds.), Neuropsychological
Hickok, G., Poeppel, D., 2000. Towards a functional neuroanatomy of Impairments of Short-Term Memory. Cambridge Univ. Press, New
speech perception. Trends Cogn. Sci. 4 (4), 131 – 138. York, pp. 11 – 53.
Jonides, J., Schumacher, E.H., Smith, E.E., Lauber, E., Awh, E., Minosh- Shallice, T., Warrington, E., 1974. The dissociation between short term
ima, S., et al., 1997. Verbal working memory load affects regional brain retention of meaningful sounds and verbal material. Neuropsychologia
activation as measured by PET. J. Cogn. Neurosci. 9 (4), 462 – 475. 12, 553 – 555.
Jonides, J., Marshuetz, C., Smith, E.E., Reuter-Lorenz, P.A., Koeppe, R.A., Shergill, S.S., Bullmore, E.T., Brammer, M.J., Williams, S.C.R., Mur-
2000. Age differences in behavior and PET activation reveal differences ray, R.M., McGuire, P.K., 2001. A functional study of auditory
in interference resolution in verbal working memory. J. Cogn. Neurosci. verbal imagery. Psychol. Med. 31 (2), 241 – 253.
12 (1), 188 – 196. Smith, E.E., Jonides, J., Koeppe, R.A., 1996. Dissociating verbal and
La Bar, K.S., Gitelman, D.R., Parrish, T.B., Mesulam, M.-M., 1999. spatial working memory using PET. Cereb. Cortex 6 (1), 11 – 20.
Neuroanatomic overlap of working memory and spatial attention net- Sohn, M.-H., Ursu, S., Anderson, J.R., Stenger, V.A., Carter, C.S., 2000.
works: a functional MRI comparison within subjects. NeuroImage 10, The role of prefrontal cortex and posterior parietal cortex in task switch-
695 – 704. ing. Proc. Natl. Acad. Sci. 97 (24), 13448 – 13453.
 S.M. Ravizza et al. / NeuroImage 22 (2004) 562–573 573

Tsukiura, T., Fujii, T., Takahashi, T., Xiao, R., Inase, M., Iijima, T., et al., F.N. (Eds.), Handbook of Memory Disorders. Wiley, New York,
2001. Neuroanatomical discrimination between manipulating and main- pp. 135 – 165.
taining processes involved in verbal working memory: a functional MRI Veltman, D.J., Friston, K.J., Sanders, G., Price, C.J., 2000. Regionally
study. Cogn. Brain Res. 11 (1), 13 – 21. specific sensitive differences in fMRI and PET: where do they come
Vallar, G., Baddelely, A., 1984. Fractionation of working memory: neuro- from? NeuroImage 11, 575 – 588.
psychological evidence for a phonological short-term store. J. Verbal Warrington, E., Shallice, T., 1969. The selective impairment of auditory
Learn. Verbal Behav. 23 (2), 151 – 161. verbal short-term memory. Brain 92 (4), 885 – 896.
Vallar, G., Papagno, C., 1995. Neuropsychological impairments of Wojciulik, E., Kanwisher, N., 1999. The generality of parietal involvement
short-term memory. In: Baddelely, A., Willson, B.A., Watts, in visual attention. Neuron 23, 747 – 764.