NeuroImage 54 (2011) 550–559

Contents lists available at ScienceDirect

NeuroImage
j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y n i m g

Domain-general mechanisms of complex working memory span
Jason M. Chein a,⁎, Adam B. Moore b, Andrew R.A. Conway b
a
b

Temple University, Department of Psychology, 6th Fl. Weiss Hall, Philadelphia PA, 19122, USA
Princeton University, Department of Psychology, Green Hall, Princeton, NJ, 08544, USA

a r t i c l e

i n f o

Article history:
Received 9 September 2009
Revised 3 June 2010
Accepted 22 July 2010
Available online 4 August 2010

a b s t r a c t
A new fMRI complex working memory span paradigm was used to identify brain regions making domaingeneral contributions to working memory task performance. For both verbal and spatial versions of the task,
complex working memory span performance increased the activity in lateral prefrontal, anterior cingulate,
and parietal cortices during the Encoding, Maintenance, and Coordination phase of task performance.
Meanwhile, overlapping activity in anterior prefrontal and medial temporal lobe regions was associated with
both verbal and spatial recall from working memory. These findings help to adjudicate several contested
issues regarding the executive mechanisms of working memory, the separability of short-term and working
memory in the verbal and spatial domains, and the relative contribution of short-term and long-term
memory mechanisms to working memory capacity. The study also provides a vital bridge between
psychometric and neuroimaging approaches to working memory, and constrains our understanding of how
working memory may contribute to the broader landscape of cognitive performance.
© 2010 Elsevier Inc. All rights reserved.

Introduction
Given its importance in so many mental aptitudes – reasoning,
language comprehension, spatial visualization, problem solving, etc. –
it is unsurprising that there exists a large neuroimaging literature
focused on the nature of working memory (WM) and its neural
underpinnings. By many accounts, this work has greatly informed
current thinking regarding the structure and function of WM, and its
position in the general architecture of human cognition (Kane and
Engle, 2002; Owen et al., 2005; Postle, 2006; Smith, 2000). However,
certain factors may limit the relevance of this large body of work to an
understanding of WM's role in complex cognition. In the current
paper we present a new fMRI WM paradigm designed to address
these limitations. This paradigm replicates standard behavioral and
imaging results, and yields novel fMRI findings regarding the domaingeneral processes involved in WM.
The neuroimaging and behavioral literatures on WM are disjointed
for several reasons. First, prior neuroimaging studies have often
utilized tasks that require only the passive storage of information in
short-term memory (STM), and not the coordination of storage with
additional processing demands. Yet it is the ability to maintain and

Abbreviations: WM, working memory; STM, short-term memory; CWMS, complex
working memory span; FMRI, functional magnetic resonance imaging; PFC, prefrontal
cortex; ACC, anterior cingulate cortex; PPC, posterior parietal cortex; MTL, medial
temporal lobe.
⁎ Corresponding author. Temple University, Department of Psychology, 6th Fl. Weiss
Hall, 1701 N. 13th Street, Philadelphia, PA, 19122, USA. Fax: +1 215 204 5539.
E-mail address: jchein@temple.edu (J.M. Chein).
1053-8119/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2010.07.067

manipulate information in coordination with ongoing processing that
is the very hallmark of WM (Baddeley and Hitch, 1974; Miller et al.,
1960) and that distinguishes the complex working memory span
(CWMS) tasks that yield the strongest correlations with complex
cognition from other measures of short-term memory (Engle et al.,
1999). Second, although a handful of prior neuroimaging studies has
employed CWMS tasks (Bunge et al., 2000; Kondo et al., 2004; Osaka
et al., 2003, 2004; Smith et al., 2001), all were conducted in only the
verbal domain, thus leaving it indeterminate whether the observed
brain activity reflected domain-specific or domain-general processing. It is important to make this distinction because psychometric
data (based on individual differences in WM capacity) indicate that
domain-general mechanisms of WM are most predictive of complex
cognition (Kane et al., 2004). Third, many prior neuroimaging studies
of WM have relied on item-recognition tasks, such as the Sternberg
memory search task (Jonides et al., 1998) and the n-back task (Braver
et al., 1997; Owen et al., 2005). Although a relatively small number of
studies have required recall rather than recognition (Gilbert and Fiez,
2004; Osaka et al., 2004), very few studies have also included an
analysis of neural activity during the recall period (cf. Chein and Fiez,
2001). This emphasis on recognition tasks, and on trial periods that
precede retrieval, is unfortunate in light of recent work indicating that
an important mechanism underlying individual differences in WM
capacity is the ability to select and use retrieval cues to optimally focus
memory search during recall (Unsworth and Engle, 2007a,b). The
requirement for recall also more realistically reflects the complex
cognitive behavior exhibited in such tasks as reading, reasoning, and
problem solving, which all require the rapid recall of recently acquired
knowledge. The present study reflects an effort to bridge the divide

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

between the neuroimaging and psychometric/behavioral literatures
by addressing these earlier limitations.
The measurement of WM capacity
Variation in WM capacity across individuals correlates strongly
with performance on a range of complex cognitive tasks (e.g., verbal
SAT, Ravens Matrices, etc.). Through nearly 30 years of careful
investigation, we and others (Bayliss et al., 2005; Conway et al.,
2005; Engle et al., 1999; Miyake, 2001) have concluded that the most
valid and reliable measures of WM capacity are CWMS tasks, such as
reading span (Daneman and Carpenter, 1980) and operation span
(Turner and Engle, 1989). In CWMS tasks, subjects remember a short
stimulus list for later recall, and must simultaneously engage in a
secondary “processing” task. CWMS tasks differ from traditional STM
tasks, such as digit span and word span, in that they require the
subject to shift attention away from each successive to-be-remembered stimulus in order to perform some other task. These tasks are
thought to be a realistic reflection of WM in everyday cognition
because they require maintenance of information in the face of, or in
the service of, concurrent processing; the very definition of WM
proposed by Miller et al. (1960) and Baddeley and Hitch (1974).
Consistent with this view, CWMS tasks account for a very large
proportion of the variation in individuals' fluid intelligence and
performance across a very wide range of complex cognitive tasks (for
a meta-analysis, see Kane et al., 2005). Although other WM tasks
involving the manipulation and updating of stored information have
been used to explore the neural correlates of WM in prior research
(e.g., N-back, tasks that require mental reordering of stimuli, etc.), few
of these alternative WM assessments have been tested as predictors of
individual differences in higher cognitive ability, and when tested,
these tasks generally explain substantially less variance than do
CWMS tasks (Kane et al., 2007). CWMS tasks are thus highly reliable
and valid measures of WM capacity, ecologically valid, predictive of
many other cognitive abilities, and consistent with the original
definition of WM.
The demands of CWMS task performance can be decomposed into
two phases: the first involving the period in which successive stimuli
are encoded and maintained in coordination with the requirements of
the secondary processing task — we refer to this as the Encoding,
Maintenance, and Coordination, or EMC, phase; the second involving
the period in which to-be-remembered items are retrieved and
reported, typically through recall — we refer to this as the Recall
phase.
Neuroimaging of CWMS
There have been only five published papers involving imaging of
CWMS. The small size of this literature is surprising given the vast
number of papers that have used these types of tasks in the
psychometric literature to establish the relationship between WM
and complex cognition. One of the few labs to administer CWMS in an
imaging environment (Smith et al., 2001) noted this juxtaposition:
“…by and large the WM tasks used in neuroimaging are not the
kinds of tasks that correlate with higher-level cognition. Thus,
much of what we know about the neural bases of WM may not
bear on the way in which WM is used in higher-level cognition”
(p. 2095).
Neuroimaging studies using CWMS tasks, which do correlate with
higher-level cognition, may thus be especially important in connecting the behavioral and neuroscientific findings on WM.
Unfortunately, it is somewhat difficult to summarize even the
small body of prior CWMS imaging work due to fundamental
differences in methodology. To begin, prior studies involving imaging

551

of CWMS have used somewhat varied measures of verbal CWMS (i.e.,
reading span, listening span, and operation span). Another problem is
that the key fMRI contrasts have differed considerably across studies.
For example, one approach (which we favor) is to contrast a CWMS
condition (combined storage and processing) first with a processingonly condition, then separately with a storage-only condition, and to
then take the conjunction of the two contrasts (Bunge et al., 2000).
Another approach (Kondo et al., 2004; Osaka et al., 2003; Osaka et al.,
2004) is to simply contrast the CWMS task with processing-only, but
this necessarily confounds activity specific to the CWMS task with
activity due to the storage component. Yet another approach (Smith
et al., 2001) involves a contrast of the WM span condition with an
average of the processing and storage components, but this approach
may be flawed when regions active in storage- or processing-only do
not reach threshold in averaging.
Another difference across experiments has been the response
requirements, with different studies requiring serial recall (Osaka
et al., 2003), cued serial recall (Bunge et al., 2000), order recognition
(Smith et al., 2001), and forced-choice recognition (Kondo et al., 2004;
Osaka et al., 2004). Also, the studies that did require recall did not
analyze the data from the recall period.
Despite these differences in methodology, some consistencies
have emerged. One universal finding is increased activity in lateral
prefrontal cortex (PFC) in the CWMS condition relative to other
conditions. Although some experiments have also implicated the
anterior cingulate cortex (ACC) in CWMS (Kondo et al., 2004; Osaka
et al., 2003, 2004), others have not (Bunge et al., 2000; Smith et al.,
2001). Likewise, some experiments have found activity in posterior
parietal cortex (PPC) in CWMS (Kondo et al., 2004; Osaka et al., 2004;
Smith et al., 2001), while others have not (Bunge et al., 2000; Osaka
et al., 2003). Despite the use of different tasks (listening span, reading
span, and operation span), different response requirements (e.g.,
recall, cued recall, and recognition), different contrasts, and possible
statistical power limitations, this set of findings does yield the basic
prediction that the EMC phase of CWMS conditions should recruit
lateral PFC, and possibly ACC and PPC areas.
Domain-general vs. domain-specific contributions of WM
The brain regions identified in neuroimaging studies using CWMS –
lateral PFC, ACC, and PPC – are widely discussed as constituents of a
domain-general cognitive control system that guides the selection and
processing of mental representations and thus scaffolds controlled
cognitive performance (Chein and Schneider, 2005; Duncan and
Owen, 2000; Miller and Cohen, 2001; Schneider and Chein, 2003).
Domain-general, as used in the neuroimaging and psychometric
literatures, refers simply to involvement in tasks with varying material
types. Despite the use of exclusively verbal tasks, activity in these
regions during CWMS task performance has been interpreted as
reflecting the engagement of domain-general, or executive, WM
mechanisms. This interpretation is broadly consistent with psychometric data showing that shared variance across verbal and spatial
measures of CWMS is highly predictive of individual differences in
various measures of reasoning and fluid intelligence (Conway et al.,
2003).
However, two contested issues in the psychometric literature are
relevant to this interpretation, and to the present study. First,
although much prior work (Conway et al., 2002; Daneman and
Carpenter, 1980; Engle et al., 1999; Kane et al., 2004) establishes that
verbal CWMS tasks more strongly predict cognitive ability than do
verbal STM tasks, this distinction is less well established in the spatial
domain. While some researchers have found spatial CWMS tasks to be
more predictive than spatial STM tasks (Kane et al., 2004), others have
not (Miyake et al., 2001). Second, the presumed domain-generality of
“executive” WM resources has also been brought into question by
conflicting psychometric findings. One hypothesis is that there is

552

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

indeed a unitary, domain-general component of WM, explaining why
individual differences in spatial WM are strongly correlated with
individual differences in verbal WM (Conway and Engle, 1996; Kane
et al., 2004; Turner and Engle, 1989). By this account, one would
expect that verbal and spatial CWMS tasks should engage overlapping
(domain-general) neural resources. An alternative hypothesis is that
there are separate executive WM resources for different types of
information (MacDonald and Christiansen, 2002; Shah and Miyake,
1996). According to this latter view, which rejects the basic notion of
domain-generality, one would expect distinct cortical regions to be
activated by verbal and spatial versions of a CWMS task.

dependent search process elicited during recall relies on coordinated
activity between PFC and MTL areas. This speculation is again
consistent with computational models that explain both neural and
behavioral findings by simulating the interaction between PFC and
MTL in a variety of memory tasks (Polyn et al., 2009). These claims
lead us to the previously untested prediction that the hippocampus
and neighboring medial temporal cortices should be more strongly
recruited during the recall period of a CWMS task than during the
recall period of an STM span (storage-only) task (so long as the
number of to-be-remembered items does not exceed the limited
capacity of attention).

Short-term and long-term memory mechanisms underlying WM
capacity

Novel CWMS paradigm for fMRI

A further development in our current view of WM regards the
relative contribution of short-term and long-term memory mechanisms to CWMS. While the terms “working memory” and “short-term
memory” have often been used interchangeably, many recent studies
indicate that WM and long-term memory tasks (e.g., episodic
retrieval) tap into common neural mechanisms (Cabeza et al., 2002;
Nee and Jonides, 2008; Ranganath et al., 2004; Talmi et al., 2005).
These neuroimaging findings have motivated several memory
theorists to revisit their assumptions regarding the relationships
between short-term, long-term, and working memory (Baddeley,
2000; Cowan, 2008; Healey and Miyake, 2009; Unsworth and Engle,
2007a). Of direct relevance to the present work, Unsworth and Engle
(2007a,b) have argued that there are two dissociable domain-general
mechanisms that influence WM capacity; (1) a dynamic attention
component that maintains a small amount of information in a
transiently more accessible state (i.e., in STM), and (2) a search
component, responsible for guiding a search through information that
has been lost from the focus of attention and thus displaced into
secondary (long-term) memory. According to this perspective, as a
subject performs a CWMS task, the dynamic attention component is
necessary during the EMC phase, to coordinate the processing and
storage demands of the task and to maintain the to-be-remembered
items in an active and accessible state. The search component is
necessary during the Recall phase to support recovery of to-beremembered items that have been lost from the focus of attention.
While Unsworth and Engle do not provide a neural model, the
dynamic attentional mechanisms described in their account are
consistent with neuroanatomically motivated computational models
of WM implicating the lateral PFC and ACC as regions involved in the
active maintenance, updating, and monitoring of information in WM
(Miller and Cohen, 2001; O'Reilly and Frank, 2006). The engagement
of this active maintenance system is thus consistent with the reported
increases in PFC and ACC activities in association with CWMS
performance.
Unsworth and Engle further speculate that the medial temporal
lobes (MTL) should be important for CWMS performance. Historically,
the MTL has been associated specifically with long-term memory, in
large part because damage to the MTL is associated with profound
long-term memory deficits but typically leaves short-term memory
intact (Squire and Schacter, 2002). Most neuroimaging studies of
short-term and working memory have, consistent with this interpretation of the neuropsychological findings, failed to elicit activation of
the MTL (Chein et al., 2002; Owen et al., 2005; Wager and Smith,
2003). There is evidence, however, that MTL structures may play a
role in supporting WM encoding and maintenance when the
memoranda are unfamiliar, complex, or require relational processing
(Cabeza et al., 2002; Olson et al., 2006; Ranganath and D'Esposito,
2001). Unsworth and Engle's view further implicates the MTL in WM
recall, even with simple and familiar stimuli, when the task demands
exceed the limitations of active maintenance processes; as is the case
for CWMS tasks. Specifically, Unsworth and Engle argue that the cue-

To address the aforementioned limitations, and to support
investigation of domain-general and long-term memory contributions to CWMS performance, we developed a novel CWMS paradigm
adapted for use with fMRI. Beyond its incorporation of both verbal and
spatial CWMS conditions and a recall procedure, our implementation
involves two further important departures from prior neuroimaging
studies. First, our paradigm used a fully crossed design which included
both same-domain CWMS conditions (e.g., verbal storage combined
with verbal processing) as well as cross-domain CWMS conditions
(e.g., verbal storage combined with spatial processing). A few
behavioral studies have manipulated and crossed both the storage
and processing domains of WM span tasks (Bayliss et al., 2003; Shah
and Miyake, 1996). The key behavioral result in these studies, which
we used as a benchmark to validate our paradigm, was that
interference effects were stronger in the same-domain tasks than in
the cross-domain tasks (i.e., recall accuracy was lowest when the
storage and processing were conducted in the same-domain).
Second, our CWMS conditions included an important change from
prior implementations of the pacing of the processing component. A
complication with prior neuroimaging studies of CWMS is that they
have always presented the processing component for a fixed amount
of time between each to-be-remembered item. For example, a single
sentence in reading span was presented for 4 s (Osaka et al., 2004).
The problem inherent with this approach is that subjects who process
quickly have excess time to devote to the storage component of the
task. Such fixed-pacing has been shown to undermine the strong
correlation between CWMS performance and complex cognition in
psychometric studies (Conway et al., 2005; Friedman and Miyake,
2004), likely because it encourages the engagement of domainspecific maintenance strategies (e.g., subvocal rehearsal). We
addressed this dilemma by requiring subjects to make as many
rapid forced-choice decisions as they could in each 4 s epoch of the
processing task component. Designed this way, the processing
components were still of fixed overall duration (4 s), which
accommodates fMRI design considerations, but a subject's attention
was continuously occupied for the entire duration of the processing
component. Barrouillet, Camos, and colleagues have recently shown
that CWMS tasks with this type of repetitive processing component
are as predictive of cognitive ability as are traditional CWMS tasks
(Lepine et al., 2005). By deploying a novel CWMS task tailored to the
pragmatic constraints of fMRI experimentation, with both verbal and
spatial materials and a recall procedure, the present study extends a
small but promising foundation of prior neuroimaging work on the
neural mechanisms underlying CWMS task performance.
Materials and methods
Subjects
Twelve right-handed, native English speaking, subjects (age range
19–24 years, 7 females) selected from the Princeton University

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

553

community took part in the fMRI experiment after providing written
consent.
Behavioral procedures
Design
Two independent variables were manipulated, both within
subjects: (1) processing component (spatial, verbal, and control),
and (2) storage component (spatial, verbal, and control). This resulted
in 8 different task conditions, with control/control being an empty
cell in the design. The order of presentation of task conditions was
pseudo-randomized across subjects.
Processing components
There were two processing tasks, lexical decision (LEX) and
symmetry decision (SYM). Decisions in each task were indicated by
button presses on an MRI compatible response unit held in the
subject's left hand. In lexical decision, a string of letters was presented
and the subject was asked to indicate whether the string formed a
word or not by pressing the designated button. A pool of 200 letter
strings, half of which were words, was sampled without replacement
throughout the task. In the symmetry decision condition, cells of an
8 × 8 matrix were colored either black or white and the subject had to
indicate whether or not the matrix was symmetrical about the vertical
axis by pressing the designated response button. A pool of 200
matrices, half of which were symmetrical, was sampled without
replacement throughout the task.
Storage components
There were two storage conditions, letters (LET) and locations
(LOC), in which subjects attempted to maintain a sequence of 4 items
for later recall. In the letter condition, a lowercase letter was
presented in 24 point, white Geneva font, shown in the center of a
black screen. The pool of sampled letters consisted of 16 consonants
(b, c, d, f, g, j, k, l, m, n, p, s, t, v, x, and z). In the locations condition, one
cell in a 4 × 4 matrix was highlighted in red. A given letter or location
was never repeated within a given trial.
Trial conditions
There were 8 trial conditions involving various combinations of
the processing and storage components (Fig. 1a). Four conditions
were dual-task CWMS conditions (LEX–LET, LEX–LOC, SYM–LET, and
SYM–LOC), and the other four were baseline conditions, involving
either the processing component with no storage requirement (LEX,
SYM) or the storage component with no processing requirement (LET,
LOC).
Run and trial sequences
The entire session consisted of 8 “runs”, with each run involving
8 trials. Each trial lasted 52 s, and included (Fig. 1b): an instruction
period (2 s), four repetitions of the processing (4 s) and storage (1 s)
elements (4 × (4 + 1) = 20 s total), a recall period (8 s), feedback
(2 s), and finally, a resting baseline period (20 s). Subjects recalled the
to-be-remembered stimuli for each trial by selecting the appropriate
cells in a 4 × 4 matrix using an MRI compatible mouse held in the right
hand (resting on a lap desk positioned on the subject's lower
abdomen). For LOC items, the locations in the 4 × 4 recall grid
corresponded to the to-be-remembered locations. For LET items, the
16 letters used as possible stimuli throughout the experiment were
re-presented in the 4 × 4 matrix (one letter per cell). The locations of
the letters within the matrix shifted from trial to trial so that subjects
could not adopt a spatial strategy to aid verbal recall. In both storage
tasks, the subject had to click on the appropriate items in correct serial
order.

Fig. 1. (a) Schematic representation of the 8 trial types included in the fMRI design, and
(b) timeline of trial events and the corresponding GLM model regressors.

FMRI procedures
FMRI data acquisition
Scanning was conducted on a 3-Tesla head-only Siemens Allegra
magnet equipped with a standard transmit/receive head coil. Subjects
lay supine, and stimuli were projected onto a visual display positioned
inside the magnet's bore (viewed through a mirror placed above the
subjects' eyes).
Prior to functional scanning, a high-resolution 3D structural
volume (Siemens MPRAGE, 1 mm cubic voxels) was collected from
each subject. Subsequent functional series were acquired using a T2*weighted echo-planar imaging (EPI) sequence (TE = 30 ms, TR = 2 s,
FOV = 200 mm, slice-thickness = 4.0 mm, flip angle = 90, in-plane
resolution = 3.125 mm). Functional MRI data were collected in
8 separate runs corresponding to sequential runs of the cognitive
paradigm.
FMRI data analysis
Data analysis was conducted off-line using AFNI (Cox, 1996). A
series of preprocessing steps was employed to correct for artifacts and
individual subject differences. Slice scan time correction using sinc
interpolation was first applied to compensate for variation in
acquisition timing. Then, images were adjusted for subject motion
through a six-parameter rigid-body automated registration algorithm.
Next, temporal high-pass filtering (.01 Hz cutoff) and 3D isotropic
Gaussian smoothing (8 mm FWHM) were applied. In order to obtain
group composite results, structural images from each subject were
transformed into the space of the MNI sterotaxic atlas (Mazziotta
et al., 2001), and the resulting transformation was applied to each
subject's functional data.
Statistical analysis of the functional data employed least-squares
estimation based on the general linear model (GLM), allowing fMRI
BOLD signal changes occurring during particular temporal stages of

554

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

each trial to be assessed. The full model included separate covariates
associated with the 20 s EMC period, and with the 8 s Recall period, for
each of the eight trial conditions (Fig. 1b).
As detailed in the Results section, several statistical contrasts based
on this full model were conducted. Each contrast was conducted via a
whole-brain, voxel-wise, random-effects test based on the coefficients
obtained from the GLM model. To reduce noise variance, individual
subject runs producing outlier (N2 standard deviations) coefficient
estimates were censored from further analysis. EMC regions were
identified by using a conjunction approach in which the relevant
CWMS condition (EMC regressor) was contrasted with its processingonly counterpart, then with its storage-only counterpart, and then a
conjunction of the two independent contrasts was formed. Recall
regions were identified by contrasting the relevant CWMS condition
to its storage-only counterpart, using the recall period regressors. For
each individual contrast, a voxel-wise False Discovery Rate (FDR)
threshold of q b .05 was used, followed by the removal of small
clusters (b135 mm3). A gray matter mask, which excluded voxels
from outside of the brain, voxels in large white matter regions, and
voxels from inside ventricles, was also applied.
Results
Behavioral results
Analysis of performance focused on recall accuracy (Fig. 2, see online
supplementary material for complementary results obtained from
analysis of accuracy and reaction time data associated with processing
judgments). All null hypothesis significance tests were non-directional
with alpha= .05. Effect sizes for ANOVA results were calculated using
the formula for partial eta-squared, η2, and for pairwise comparisons
using Cohen's d. The numbers of stimuli correctly recalled per trial were
averaged within condition (max. of 4, min. of 0). Recall accuracy means
were then submitted to a 2 × 3 repeated measures ANOVA with 2 levels
of storage (LET and LOC) and 3 levels of processing (LEX, SYM, and
control). This analysis revealed a marginally significant main effect of
storage condition, F(1,11) = 4.44, p = .06, η2 = .29, a significant main
effect of processing condition, F(2,22) = 25.26, p b .05, η2 = .70, and a
significant interaction, F(2,22) = 15.51, p b .05, η2 = .59. The slight main
effect of storage reflects the fact that average recall was slightly better in
the dual-task LET conditions than in the dual-task LOC conditions.
However, there were no differences between the verbal and spatial
storage-only tasks, suggesting that they were equally demanding. The
main effect of processing condition is unsurprising and reveals the fact
that recall was better on trials with no processing decisions than with
either SYM or LEX decisions (pb .05 for both pairwise comparisons to
control; the pairwise comparison between SYM and LEX was not

Fig. 2. Average number of letters or location recalled in the same-domain CWMS (LEX–
LET and SYM–LOC), cross-domain CWMS (SYM–LET and LEX–LOC) and storage-only
(LET and LOC) conditions.

significant, p N .05). The key predicted behavioral result is the significant
interaction, which demonstrates greater interference when the processing and storage domains matched than when they were different. In
order to test this prediction more directly, we conducted a planned
contrast comparing average interference in same-domain conditions to
average interference in cross-domain conditions. This contrast was
significant, t(11) = 4.08, p b .05, d = 1.18, with greater same-domain
interference (M= .72) than cross-domain (M= .30).
The behavioral results replicate previous experiments on crossdomain WM span, in which subjects experienced greater interference
when processing and storage domains matched than when they were
different. These results are reassuring because they suggest that our
CWMS paradigm, developed to be compatible with the fMRI
environment, is valid and produces the same signature behavioral
results as traditional CWMS tasks. The behavioral effects also support
more detailed interpretation of brain activation patterns, as we
discuss below.
Imaging results
Separate analyses focused on two segments of task performance:
1) the EMC period, during which participants alternated between the
processing and storage task components, and 2) the Recall period,
during which participants attempted to serially recall the earlier
presented storage items.
Encoding, Maintenance, and Coordination in WM
We identified EMC regions by using a conjunction approach in
which the relevant CWMS condition (verbal or spatial) was
contrasted with its processing-only counterpart, then with its
storage-only counterpart, and then a conjunction of the two
independent contrasts was formed (Bunge et al., 2000). This
conjunction can be described as [(CWMS–P) ∩ (CWMS–S)] and is
consistent with the conjunction methodology suggested by Nichols et
al. (2005). Each contrast was conducted via a whole-brain, voxelwise, random-effects test based on the coefficients obtained from each
subject for the EMC covariates in the GLM model. Regions showing
significantly increased activity for the same-domain verbal CWMS
condition (LET–LEX) relative to both verbal processing alone (LEX)
and verbal storage alone (LET) are shown in Fig. 3a (left), and partially
detailed in Table 1 (see Table 2, included in the online supplement, for
a listing of domain-specific verbal CWMS sites).
As can be seen, the verbal CWMS condition was associated with
increased activation in the bilateral PFC (BA 9/46), bilateral PPC (BA 7),
and ACC (BA 32), among other regions. This pattern of activation
essentially replicates the findings from prior neuroimaging studies of
CWMS, all having used a same-domain verbal CWMS task (Bunge et al.,
2000; Kondo et al., 2004; Osaka et al., 2003, 2004; Smith et al., 2001).
Results from the contrast of the analogous spatial CWMS condition
(LOC–SYM) relative to its counterpart storage-only (LOC) and
processing-only (SYM) conditions are also shown in Fig. 3a (right)
and Table 1. While this conjunction identified some brain regions not
detected in the verbal contrast (see Table 2, included in online
supplement, for a detailing of domain-specific spatial CWMS sites),
the overall pattern of activity was highly consistent with verbal
CWMS; with bilateral PFC, bilateral PPC, and ACC regions again
exhibiting increased engagement in the spatial CWMS task.
A central aim in analyzing the EMC period was to evaluate the
domain-general nature of PFC, PPC, and ACC contributions to CWMS.
In the independent analyses of the verbal and spatial CWMS
conditions described above, each of these regions was found to
exhibit significantly increased activity (see Table 1). While the exact
clusters of activation differed slightly across the verbal and spatial
CWMS tasks, in every case the peak of activation was within 15 mm
(Euclidean distance) of the analogous activation peak, and the
identified clusters were mostly overlapping across domains (i.e.,

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

555

recruitment in both the verbal and spatial CWMS conditions during
the EMC period.
To further elucidate the involvement of these brain areas, we
examined average estimated percent signal change during the EMC
period for the same-domain CWMS (LET–LEX and LOC–SYM), crossdomain CWMS (LET–SYM and LOC–LEX), processing-only (LEX and
SYM), and storage-only (LET and LOC) conditions; shown in Fig. 3b.
These activation patterns show that the PFC and ACC are most
strongly engaged for same-domain CWMS conditions, but also exhibit
significant recruitment in cross-domain CWMS, regardless of whether
the to-be-remembered materials are verbal (letters) or spatial
(locations). Overall, the general pattern of activity observed in PFC
and ACC regions is entirely consistent with a domain-general
interpretation of their involvement in CWMS task performance.
Activity in the PPC partially paralleled that observed in anterior
regions, by again revealing increased engagement for same-domain
CWMS relative to the corresponding processing- and storage-only
controls. However, the right PPC was substantially more activated in
the cross-domain SYM–LET condition (relative to LEX–LET), and
strong recruitment of bilateral PPC regions was also apparent for the
spatial control conditions (LOC and SYM). Moreover, the crossed LEX–
LOC condition did not increase activity beyond the level observed for
location memory alone (LOC). Thus, although the PPC shows
significant involvement in CWMS for both the verbal and spatial
domains, the interpretation of PPC activity must also account for
additional sensitivity to spatial processing demands, as evidenced by
the contribution that this region makes to single-task spatial storage
(LOC), and non-mnemonic visuo-spatial processing (SYM).

Fig. 3. (a) Regions exhibiting significantly increased activity during the Encoding,
Maintenance, and Coordination (EMC) phase for same-domain verbal and spatial
CWMS conditions, relative to the corresponding storage-only and processing-only
controls, and (b) the patterns of signal change observed in each domain-general
(overlapping) region across the 8 trial types.

included common voxels). Beyond these a priori regions, only the
medial frontal gyrus (medial BA 6), left precentral gyrus (lateral BA 6)
and left anterior insula (BA 13) exhibited common and significant

Controlled retrieval from LTM
As discussed above, recent studies indicate that CWMS performance may be mediated not only by dynamic attentional processes,
but also by controlled retrieval mechanisms used to recall information
that has been displaced from active memory, but that can be
recovered from LTM. To identify candidate regions involved in this
putative controlled retrieval process, we contrasted recall period
activity for the verbal and spatial CWMS conditions with their
storage-only controls (LET–LEX vs. LET, LOC–SYM vs. LOC). Since
subjects had nothing to recall in the processing-only conditions (LEX,
SYM), these conditions were not included in the analysis of recall
period activity.
Unsworth and Engle (2007b) predicted that recall in CWMS tasks
might involve interactions between PFC and MTL mechanisms. These
regions could thus be considered a priori regions of interest in our
analysis of recall activity. Once again, the primary focus of analysis
was to identify regions making a similar contribution to recall for both
verbal and spatial domains (since the controlled retrieval process
under investigation is presumably used to support recall regardless of
content). Accordingly, active clusters were first identified by independent contrasts conducted within each domain, and then common
sites of recall-related activity were identified. Regions exhibiting
activity consistent with a domain-general contribution to recall (see
Table 1) included bilateral PFC areas (BA 46/10, anterior and inferior
to the loci implicated in active maintenance), the left posterior MTL,
and few other regions. An analogous right lateralized MTL region
displayed significant activity increases for verbal CWMS, but fell just
below the statistical threshold for spatial CWMS (for a full list of
domain-specific activations associated with the recall period, see
Table 2 in the online supplement).
To further investigate MTL involvement in CWMS recall, and to
determine the consistency of its recruitment across subjects, we
inspected the position and pattern of MTL activity on a subject-bysubject basis. While this approach revealed some inter-individual
variation in the position of activation, almost every individual (10 out
of 12) exhibited significant MTL recruitment during CWMS recall, and
the most common localization was in the posterior hippocampus and

556

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

Table 1
Peak coordinates and descriptive statistics for regions exhibiting domain-general engagement during the Encoding, Maintenance, and Coordination (EMC) and Recall phases of
CWMS task performance.
EMC Regions

L mid. frontal (L PFC)
R mid. frontal (R PFC)
Ant. cingulate (ACC)
L sup. parietal (L PPC)
R sup. parietal (R PPC)
L med. frontal
L precentral
L ant. insula
Recall regions

Verbal conjunction

Spatial conjunction

BA

x

y

z

mm3

9/46
9
32
7
7
6
6
13

−37
37
−4
−30
20
−2
−31
−31

30
41
11
−60
−71
2
−10
19

38
36
44
53
48
52
56
9

594
1350
1620
5535
2565
3429
837
540

Z-score

Z-score

x

y

z

mm3

Z-score

Z-score

LOC–SYM vs. LOC

LOC–SYM vs. SYM

25
38
13
−68
−71
2
−14
17

33
38
39
43
40
51
57
11

216
864
864
5697
2862
1728
702
135

2.62
2.85
3.09
3.61
3.84
3.73
3.70
3.55

2.95
3.09
3.09
3.02
3.73
2.62
2.97
2.42

13.9
4.1
5.4
14.6
10.0
1.4
4.1
3.0

Z-score

Δ

LET–LEX vs. LET

LET–LEX vs. LEX

3.29
2.51
3.09
3.67
2.44
3.71
3.29
3.35

2.65
2.11
2.70
2.88
3.09
3.27
2.62
3.29

−49
35
−4
−23
26
−1
−31
−32

Z-score

x

Verbal recall
BA

x

Spatial recall
y

z

mm3

y

z

mm3

LET–LEX vs. LET
L mid./inf. frontal (PFC)
R mid./inf. frontal (PFC)
L hipp./parahipp. (MTL)
L ant. insula
Thalamus
Midbrain/cerebellum

46
46
13

−41
47
−29
−29
2
2

Δ

41
44
−41
20
−20
−37

9
3
3
−1
3
−31

162
135
1539
216
297
567

3.15
2.90
3.89
3.26
3.7
2.98

LOC–SYM vs. LOC
−42
44
−26
−36
2
−2

45
47
−43
20
−17
−34

13
13
3
−4
3
−34

2349
3780
189
1863
2295
999

3.90
3.93
3.02
3.89
3.22
3.31

5.7
10.9
3.6
7.6
3.0
4.2

Notes: BA = Brodmann Area; x, y, z = MNI coordinates; Δ = Euclidean distance between verbal CWMS and spatial CWMS peaks.

immediately inferior portion of the parahippocampal gyrus. Left MTL
activation sites from illustrative individual subjects are shown in
Fig. 4, together with an averaged time-course of activity in this recallrelated MTL site.

Fig. 4. (a) Significant regions of activity in the left posterior MTL during the recall period
for verbal and spatial CWMS conditions relative to their corresponding storage-only
controls, for 3 illustrative subjects, and (b) The average activation profile for recallrelated activity in the left MTL for CWMS and storage-only (STM) conditions.

Discussion
Working memory is a central construct in cognitive neuroscience
because it plays a critical role in such a wide range of cognitive
behaviors, and because it is dependent upon the function of the
prefrontal cortex (PFC), an area of the brain that has traditionally held
a prominent position in neuroscientific accounts of why humans and
non-human primates exhibit superior cognitive abilities relative to
other species (Fuster, 1995; Goldman Rakic, 1998). The findings from
the current experiment (1) replicate and extend previous neuroimaging results on verbal CWMS, (2) generalize these earlier findings into
the spatial domain — thus demonstrating that the implicated brain
regions support domain-general mechanisms of WM, and (3) provide
a novel discovery with respect to the role of the MTL in CWMS — thus
informing recent theoretical claims regarding the involvement of both
short-term and long-term memory mechanisms in mediating the
relationship between WM and other measures of cognitive function.
Taken together, the findings provide a number of important insights
regarding the nature of WM and the neural mechanisms underlying
the predictive validity of CWMS tasks (i.e., why CWMS tasks are so
successful in predicting individual differences across a wide range of
cognitive tasks).
The finding of increased activity in the PFC and ACC during verbal
CWMS replicates previous imaging studies (Kondo et al., 2004; Osaka
et al., 2003, 2004). As well, we found significantly greater behavioral
interference for same-domain than for cross-domain CWMS conditions, providing a replication of prior behavioral studies (Bayliss et al.,
2003; Shah and Miyake, 1996). These replications give assurance that
our novel CWMS paradigm affords a valid measure of WM that shares
its essential features with the CWMS tasks used more broadly in the
psychometric literature.
The inclusion of cross-domain CWMS conditions is also novel
among neuroimaging studies, and affords unique insights into the
specific mechanisms that link CWMS tasks and measures of higher
cognitive function. In PFC (bilaterally) and ACC regions, activity for
both same-domain verbal CWMS (LEX–LET) and cross-domain verbal
CWMS (SYM–LET) conditions was significantly greater than in either
control condition. Without data from the cross-domain condition (as
was the case for all prior neuroimaging studies of CWMS), activity in
these regions could be interpreted as a neural response to conflict
produced when to-be-remembered information must compete for

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

representation with the products of the secondary processing task. For
instance, the phonological representations of to-be-remembered
letters (words in some prior studies) may compete directly with the
phonological representations that underlie lexical decisions (verbal
representations of sentences or equations in prior studies), thus
producing a high conflict situation that elicits an increase in the
recruitment of cognitive control areas (Botvinick et al., 2001).
However, the observation that these regions are also significantly
recruited when storage and processing involve highly disparate
representations (as in the case of verbal storage with visuo-spatial
processing) suggests that direct representational conflict may not be
the principle source of PFC or ACC activity. Instead, we interpret
increased PFC and ACC activities during CWMS conditions as
indicating the engagement of attention control and selection
mechanisms that directly support active maintenance in WM.
Moreover, while both verbal CWMS conditions produced significant increases in activity for these regions, stronger magnitude
increases were consistently observed for the same-domain, relative to
the cross-domain, condition.1 This pattern provides still further
traction in explaining PFC and ACC contributions to CWMS. Specifically, the pattern shows that the degree of activation in these regions
is dependent on the magnitude of behavioral interference produced
by the secondary task, not simply by the demand for dual-tasking or
the coordination of task-switching (between storage and processing
tasks). These findings inform an ongoing debate regarding the basis
for the predictive validity of CWMS tasks. While we and others hold
that CWMS tasks are predictive because they place strong demands on
the domain-general attention control mechanisms that maintain
information in an accessible state (Hudjetz and Oberauer, 2007; Kane
et al., 2001), an alternative theory is that CWMS tasks are predictive
because they index the ability to execute and coordinate taskswitching operations (Towse et al., 2000). The finding that activity in
these regions is contingent on the degree of interference produced by
the secondary task, despite similar demands for task-coordination,
seems to favor the controlled attention view over the taskcoordination view.
A somewhat different pattern of activation was observed in the
PPC, where the cross-domain SYM–LET task condition produced
stronger bilateral PPC activity than did the same-domain LEX–LET
condition (though again, both were significantly active above the
storage-only and processing-only control conditions). This pattern of
PPC engagement can be more fully understood in light of the findings
from spatial CWMS conditions, which we consider next.
One of the central goals of the present study was to extend
previous results by testing whether the brain regions implicated in
verbal CWMS performance are also engaged by spatial CWMS tasks.
By examining verbal and spatial CWMS activity using a common
methodology, we hoped to demonstrate that regions previously
implicated in only verbal CWMS performance in fact mediate domaingeneral WM processes. Consistent with this aim, we found that during
the active maintenance phase, PFC, ACC, and PPC regions again
exhibited increased activity when same-domain spatial CWMS (SYM–
LOC) was contrasted with its counterpart storage-only (LOC) and
processing-only (SYM) control conditions. While there was some
variation in the precise set of voxels activated in the analogous verbal
and spatial CWMS contrasts, the verbal and spatial CWMS clusters
were in very close proximity to one another, included overlapping
voxels, and evinced nearly identical patterns of activity across the
conditions belonging to each domain. We believe that this overall
1
Though not discussed in the Results section, we note that stronger PFC and ACC
activations in the same-domain conditions (relative to the cross-domain conditions)
were observed even when the clusters were defined by a conjunction contrast of the
cross-domain CWMS conditions with their counterpart storage-only and processingonly conditions. In general, the cross-domain clusters identified in these regions
comprised a sub-set of the voxels revealed in association with the same-domain
CWMS contrasts.

557

pattern of results strongly supports a domain-general interpretation
of activity in these regions, and converges with psychometric data
indicating that verbal and spatial WM tasks tap into shared underlying
cognitive processes (Kane et al., 2004).
These results are generally inconsistent with the idea that WM
supports cognition through primarily domain-specific mechanisms
(MacDonald and Christiansen, 2002), or the idea that there are
separate pools of executive resources for verbal and spatial WM (Shah
and Miyake, 1996). Admittedly, one could still defend these latter
domain-specific WM positions by looking beyond apparent commonalities and instead emphasizing observed differences between the
verbal and spatial CWMS contrasts (e.g., the right PFC is more strongly
active in spatial contrasts, the obtained clusters have slightly disparate
peak locations and contain non-overlapping voxels, there are other
regions that appear active in the contrast for only one domain).
However, differences of this type would be expected even when
attempting a replication of the identical task condition, due to the
inherent noisiness of fMRI (Manoach et al., 2001; Noll et al., 1997),
and taking this position necessitates that one ignore the overwhelmingly parallel findings from the verbal and spatial domain in a priori
regions.
The data from the spatial CWMS conditions may also help
adjudicate a further contested issue arising from conflicting psychometric studies. While there is general agreement that verbal CWMS
tasks are typically more predictive of complex cognition than are
traditionally used verbal STM tasks (e.g., digit span), there is less
consensus regarding the dissociability of spatial CWMS from spatial
STM span measures. In showing that spatial CWMS is distinguished
neurally from spatial STM in much the same fashion as verbal CWMS
is distinguished from verbal STM (i.e., stronger PFC and ACC activity in
the contrast of CWMS to STM for both domains), the current findings
seem to support models that dichotomize CWMS and STM in both the
verbal and spatial domains (Kane et al., 2004). However, it is
interesting and potentially informative that in at least one region
implicated in domain-general WM processing, the PPC, spatial STM
(LOC) and processing-only (SYM) conditions evoked activity
increases that were similar (if not greater) in magnitude to those
observed for the same-domain verbal CWMS condition (LEX–LET) and
the cross-domain spatial CWMS condition (LEX–LOC). By contrast,
verbal STM (LET) and processing-only (LEX) conditions evinced very
little response from the PPC. These findings suggest the possibility
that common recruitment of PPC mechanisms in both spatial STM and
CWMS contexts may explain a higher degree of shared variance
between these spatial conditions than is observed between the
analogous verbal conditions.
Perhaps the most novel aspect of the current findings pertains to
the observation of greater activity in the MTL in CWMS conditions
relative to storage-only conditions during recall. This result was
explicitly predicted by Unsworth and Engle (2007a) but had not been
tested. Bilateral PFC activations (anterior to those observed for the
active maintenance period) were similarly found to exhibit domaingeneral involvement in CWMS recall. These are important findings in
that they suggest a key distinction between CWMS tasks and other
WM tasks that have previously been used in imaging studies (and
which typically do not show MTL activity). The findings are consistent
with the proposal of a controlled search mechanism that is supported
by interactions between lateral PFC and MTL processes, and are
compatible with lesion and neurophysiology studies indicating that
these same regions are critically involved in reactivating information
from long-term memory (Ranganath and Rainer, 2003). Although
information can be lost from its activated state over the course of
CWMS performance, the surviving cues in primary memory may be
sufficient to guide search through recent LTM representations,
allowing for recovery of to-be-remembered traces. This guided search
process appears to be orchestrated by domain-general MTL and
anterior PFC mechanisms.

558

J.M. Chein et al. / NeuroImage 54 (2011) 550–559

Minimal engagement of the MTL in the storage-only (STM)
conditions is addressed within this account by assuming that active
maintenance processes (e.g., the focus of attention) offer sufficient
capacity to maintain the integrity of the four presented items in an
accessible form, thus obviating the need to conduct controlled search
of LTM during recall. According to this explanation, one might predict
that if the number of to-be-remembered items exceeds the capacity of
the focus of attention (i.e., too many items are presented and the
capacity of active maintenance is exceeded), or if the task design
somehow discourages the utilization of active maintenance processes
(e.g., speeded presentation, use of materials that are too complex or
novel to actively maintain), then the MTL–PFC retrieval path would be
necessary to recover items regardless of whether the task involves
storage and processing (CWMS) or storage-only (STM). Indeed,
consistent with these predictions, others have reported retrievalbased MTL involvement in short-term storage-only tasks that have
these relevant characteristics, including the recall of primacy items
from a long-list (12-item) recognition task (Talmi et al., 2005), the
recognition of earlier serial items (the first 2 of 3) in a rapidly
presented item-recognition test (Nee and Jonides, 2008), and the
short-term recognition of complex visual scenes (Schon et al., 2009).
These results, together with the present findings, thus lend support to
the view that WM often involves controlled retrieval from long-term
memory, as portayed in Unsworth and Engle's (2007a,b) framework.
In conclusion, neuroimaging research has greatly informed current
theoretical accounts of WM and its place in cognition. However,
several factors have limited the relevance of neuroimaging work for
the question of how WM contributes to complex cognition. Here we
present a new experimental procedure that addresses several of these
limitations, replicates previous behavioral and imaging results, and
provides novel data on the involvement of the PFC, ACC, and PPC in
both verbal and spatial WM, and on the role of the MTL in WM. Future
work will be needed to explore the relative importance of different
brain regions for the performance of various types of WM tasks, and to
explore how activity in these brain regions gives rise to individual
differences in cognitive ability.
The following are the supplementary materials related to this
article. Supplementary Material.
Table 2. Peak coordinates for regions exhibiting domain-specific
engagement during the Encoding, Maintenance, and Coordination
(EMC) and Recall periods of CWMS task performance.
Supplementary materials related to this article can be found online
at doi:10.1016/j.neuroimage.2010.07.067.
Acknowledgments
The authors would like to thank Michael Kane, Jonathan Cohen,
and Leigh Nystrom for their general advice, and for the helpful
comments on the manuscript.
References
Baddeley, A.D., 2000. The episodic buffer: a new component of working memory?
Trends Cogn. Sci. 4, 417–423.
Baddeley, A.D., Hitch, G., 1974. Working memory. In: Bower, G.H. (Ed.), The Psychology
of Learning and Motivation. Academic Press, New York, pp. 47–89.
Bayliss, D.M., Jarrold, C., Baddeley, A.D., Gunn, D.M., 2005. The relationship between
short-term memory and working memory: complex span made simple? Memory
13, 414–421.
Bayliss, D.M., Jarrold, C., Gunn, D.M., Baddeley, A.D., 2003. The complexities of complex
span: explaining individual differences in working memory in children and adults.
J. Exp. Psychol. Gen. 132, 71–92.
Botvinick, M.M., Braver, T.S., Barch, D.M., Carter, C.S., Cohen, J.D., 2001. Conflict
monitoring and cognitive control. Psychol. Rev. 108, 624–652.
Braver, T.S., Cohen, J.D., Nystrom, L.E., Jonides, J., Smith, E.E., Noll, D.C., 1997. A
parametric study of prefrontal cortex involvement in human working memory.
Neuroimage 5, 49–62.
Bunge, S.A., Klingberg, T., Jacobsen, R.B., Gabrieli, J.D., 2000. A resource model of the
neural basis of executive working memory. Proc. Natl Acad. Sci. USA 97,
3573–3578.

Cabeza, R., Dolcos, F., Graham, R., Nyberg, L., 2002. Similarities and differences in the neural
correlates of episodic memory retrieval and working memory. Neuroimage 16, 317–330.
Chein, J.M., Fiez, J.A., 2001. Dissociation of verbal working memory system components
using a delayed serial recall task. Cereb. Cortex 11, 1003–1014.
Chein, J.M., Fissell, K., Jacobs, S., Fiez, J.A., 2002. Functional heterogeneity within Broca's
area during verbal working memory. Physiol. Behav. 77, 635–639.
Chein, J.M., Schneider, W., 2005. Neuroimaging studies of practice-related change: fMRI
and meta-analytic evidence of a domain-general control network for learning.
Brain Res. Cogn. Brain Res. 25, 607–623.
Conway, A.R., Engle, R.W., 1996. Individual differences in working memory capacity:
more evidence for a general capacity theory. Memory 4, 577–590.
Conway, A.R., Kane, M.J., Bunting, M.F., Hambrick, D.Z., Wilhelm, O., Engle, R.W., 2005.
Working memory span tasks: a methodological review and user's guide. Psychon.
Bull. Rev. 12, 769–786.
Conway, A.R., Kane, M.J., Engle, R.W., 2003. Working memory capacity and its relation
to general intelligence. Trends Cogn. Sci. 7, 547–552.
Conway, A.R.A., Cowan, N., Bunting, M.F., Therriault, D.J., 2002. A latent variable analysis
of working memory capacity, short-term memory. Intelligence 30, 163–184.
Cowan, N., 2008. What are the differences between long-term, short-term, and working
memory? Prog. Brain Res. 169, 323–338.
Cox, R.W., 1996. AFNI: software for analysis and visualization of functional magnetic
resonance neuroimages. Comput. Biomed. Res. 29, 162–173.
Daneman, M., Carpenter, P.A., 1980. Individual differences in working memory and
reading. J. Verbal Learn. Verbal Behav. 19, 450–466.
Duncan, J., Owen, A.M., 2000. Common regions of the human frontal lobe recruited by
diverse cognitive demands. Trends Neurosci. 23, 475–483.
Engle, R.W., Tuholski, S.W., Laughlin, J.E., Conway, A.R.A., 1999. Working memory,
short-term memory, and general fluid intelligence: a latent-variable approach. J.
Exp. Psychol. Gen. 128, 309–331.
Friedman, N.P., Miyake, A., 2004. The reading span test and its predictive power for
reading comprehension ability. J. Mem. Lang. 51, 136–158.
Fuster, J.M., 1995. Frontal cortex and the cognitive support of behavior. In: McGaugh, J.
L., Bermudez Rattoni, F., et al. (Eds.), Plasticity in the Central Nervous System:
Learning and memory, pp. 149–160.
Gilbert, A.M., Fiez, J.A., 2004. Integrating rewards and cognition in the frontal cortex.
Cogn. Affect. Behav. Neurosci. 4, 540–552.
Goldman Rakic, P.S., 1998. The prefrontal landscape: implications of functional
architecture for understanding human mentation and the central executive. In:
Roberts, A.C., Robbins, T.W. (Eds.), The Prefrontal Cortex: Executive and Cognitive
Functions. Oxford University Press, New York, NY, pp. 87–102.
Healey, M.K., Miyake, A., 2009. The role of attention during retrieval in workingmemory span: a dual-task study. Q. J. Exp. Psychol. (Colchester) 62, 733–745.
Hudjetz, A., Oberauer, K., 2007. The effects of processing time and processing rate on
forgetting in working memory: testing four models of the complex span paradigm.
Mem. Cogn. 35, 1675–1684.
Jonides, J., Smith, E.E., Marshuetz, C., Koeppe, R.A., Reuter-Lorenz, P.A., 1998. Inhibition
in verbal working memory revealed by brain activation. Proc. Natl Acad. Sci. USA
95, 8410–8413.
Kane, M.J., Bleckley, M.K., Conway, A.R.A., Engle, R.W., 2001. A controlled-attention
view of working-memory capacity. J. Exp. Psychol. Gen. 130, 169–183.
Kane, M.J., Conway, A.R.A., Miura, T.K., Colflesh, G.J.H., 2007. Working memory,
attention control, and the n-back task: a question of construct validity. J. Exp.
Psychol. Learn. Mem. Cogn. 33, 615–622.
Kane, M.J., Engle, R.W., 2002. The role of prefrontal cortex in working-memory capacity,
executive attention, and general fluid intelligence: an individual-differences
perspective. Psychon. Bull. Rev. 9, 637–671.
Kane, M.J., Hambrick, D.Z., Conway, A.R.A., 2005. Working memory capacity and fluid
intelligence are strongly related constructs: comment on Ackerman, Beier, and
Boyle (2005). Psychol. Bull. 131, 66–71.
Kane, M.J., Hambrick, D.Z., Tuholski, S.W., Wilhelm, O., Payne, T.W., Engle, R.W., 2004.
The generality of working memory capacity: a latent-variable approach to verbal
and visuospatial memory span and reasoning. J. Exp. Psychol. Gen. 133, 189–217.
Kondo, H., Morishita, M., Osaka, N., Osaka, M., Fukuyama, H., Shibasaki, H., 2004.
Functional roles of the cingulo-frontal network in performance on working
memory. Neuroimage 21, 2–14.
Lepine, R., Barrouillet, P., Camos, V., 2005. What makes working memory spans so
predictive of high-level cognition? Psychon. Bull. Rev. 12, 165–170.
MacDonald, M.C., Christiansen, M.H., 2002. Reassessing working memory: comment on
Just and Carpenter (1992) and Waters and Caplan (1996). Psychol. Rev. 109, 35–54
discussion 55–74.
Manoach, D.S., Halpern, E.F., Kramer, T.S., Chang, Y., Goff, D.C., Rauch, S.L., Kennedy, D.
N., Gollub, R.L., 2001. Test–retest reliability of a functional MRI working memory
paradigm in normal and schizophrenic subjects. Am. J. Psychiatry 158, 955–958.
Mazziotta, J., Toga, A., Evans, A., Fox, P., Lancaster, J., Zilles, K., Woods, R., Paus, T.,
Simpson, G., Pike, B., Holmes, C., Collins, L., Thompson, P., MacDonald, D., Iacoboni,
M., Schormann, T., Amunts, K., Palomero-Gallagher, N., Geyer, S., Parsons, L., Narr,
K., Kabani, N., Le Goualher, G., Boomsma, D., Cannon, T., Kawashima, R., Mazoyer, B.,
2001. A probabilistic atlas and reference system for the human brain: International
Consortium for Brain Mapping (ICBM). Philos. Trans. R. Soc. Lond. B Biol. Sci. 356,
1293–1322.
Miller, E.K., Cohen, J.D., 2001. An integrative theory of prefrontal cortex function. Annu.
Rev. Neurosci. 24, 167–202.
Miller, G.A., Galanter, E., Pribram, K.H., 1960. Plans and the Structure of Behavior. Holt,
Rinehart & Winston, New York.
Miyake, A., 2001. Individual differences in working memory: introduction to the special
section. J. Exp. Psychol. Gen. 130, 163–168.

J.M. Chein et al. / NeuroImage 54 (2011) 550–559
Miyake, A., Friedman, N.P., Rettinger, D.A., Shah, P., Hegarty, M., 2001. How are
visuospatial working memory, executive functioning, and spatial abilities related?
A latent-variable analysis. J. Exp. Psychol. Gen. 130, 621–640.
Nee, D.E., Jonides, J., 2008. Neural correlates of access to short-term memory. Proc. Natl
Acad. Sci. USA 105, 14228–14233.
Nichols, T., Brett, M., Andersson, J., Wager, T., Poline, J.-P., 2005. Valid conjunction
inference with the minimal statistic. Neuroimage 25, 653–660.
Noll, D.C., Genovese, C.R., Nystrom, L.E., Vazquez, A.L., Forman, S.D., Eddy, W.F.,
Cohen, J.D., 1997. Estimating test–retest reliability in functional MR imaging. II:
Application to motor and cognitive activation studies. Magn. Reson. Med. 38,
508–517.
O'Reilly, R.C., Frank, M.J., 2006. Making working memory work: a computational
model of learning in the prefrontal cortex and basal ganglia. Neural Comput. 18,
283–328.
Olson, I.R., Page, K., Moore, K.S., Chatterjee, A., Verfaellie, M., 2006. Working memory for
conjunctions relies on the medial temporal lobe. J. Neurosci. 26, 4596–4601.
Osaka, M., Osaka, N., Kondo, H., Morishita, M., Fukuyama, H., Aso, T., Shibasaki, H., 2003.
The neural basis of individual differences in working memory capacity: an fMRI
study. Neuroimage 18, 789–797.
Osaka, N., Osaka, M., Kondo, H., Morishita, M., Fukuyama, H., Shibasaki, H., 2004. The
neural basis of executive function in working memory: an fMRI study based on
individual differences. Neuroimage 21, 623–631.
Owen, A.M., McMillan, K.M., Laird, A.R., Bullmore, E., 2005. N-back working memory
paradigm: a meta-analysis of normative functional neuroimaging studies. Hum.
Brain Mapp. 25, 46–59.
Polyn, S.M., Norman, K.A., Kahana, M.J., 2009. A context maintenance and retrieval
model of organizational processes in free recall. Psychol. Rev. 116, 129–156.
Postle, B.R., 2006. Working memory as an emergent property of the mind and brain.
Neuroscience 139, 23–38.
Ranganath, C., Cohen, M.X., Dam, C., D'Esposito, M., 2004. Inferior temporal, prefrontal,
and hippocampal contributions to visual working memory maintenance and
associative memory retrieval. J. Neurosci. 24, 3917–3925.

559

Ranganath, C., D'Esposito, M., 2001. Medial temporal lobe activity associated with
active maintenance of novel information. Neuron 31, 865–873.
Ranganath, C., Rainer, G., 2003. Neural mechanisms for detecting and remembering
novel events. Nat. Rev. Neurosci. 4, 193–202.
Schneider, W., Chein, J.M., 2003. Controlled and automatic processing: behavior, theory,
and biological mechanisms. Cogn. Sci. 27, 525–559.
Schon, K., Quiroz, Y.T., Hasselmo, M.E., Stern, C.E., 2009. Greater working memory load
results in greater medial temporal activity at retrieval. Cereb. Cortex.
Shah, P., Miyake, A., 1996. The separability of working memory resources for spatial
thinking and language processing: an individual differences approach. J. Exp.
Psychol. Gen. 125, 4–27.
Smith, E.E., 2000. Neural bases of human working memory. Curr. Dir. Psychol. Sci. 9,
45–49.
Smith, E.E., Geva, A., Jonides, J., Miller, A., Reuter-Lorenz, P., Koeppe, R.A., 2001. The
neural basis of task-switching in working memory: effects of performance and
aging. Proc. Natl Acad. Sci. USA 98, 2095–2100.
Squire, L.R., Schacter, D.L. (Eds.), 2002. Neuropsychology of Memory, 3rd ed. Guilford
Press, New York, NY, US.
Talmi, D., Grady, C.L., Goshen-Gottstein, Y., Moscovitch, M., 2005. Neuroimaging the serial
position curve. A test of single-store versus dual-store models. Psychol. Sci. 16, 716–723.
Towse, J.N., Hitch, G.J., Hutton, U., 2000. On the interpretation of working memory span
in adults. Mem. Cogn. 28, 341–348.
Turner, M.L., Engle, R.W., 1989. Is working memory capacity task dependent? J. Mem.
Lang. 28, 127–154.
Unsworth, N., Engle, R.W., 2007a. The nature of individual differences in working
memory capacity: active maintenance in primary memory and controlled search
from secondary memory. Psychol. Rev. 114, 104–132.
Unsworth, N., Engle, R.W., 2007b. On the division of short-term and working memory:
an examination of simple and complex span and their relation to higher order
abilities. Psychol. Bull. 133, 1038–1066.
Wager, T.D., Smith, E.E., 2003. Neuroimaging studies of working memory: a metaanalysis. Cogn. Affect. Behav. Neurosci. 3, 255–274.