NeuroImage 55 (2011) 773–787

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

Complex span tasks and hippocampal recruitment during working memory

Carlos Cesar Faraco a,⁎, Nash Unsworth b, Jason Langley c, Doug Terry b, Kaiming Li d, Degang Zhang d,
Tianming Liu d, L. Stephen Miller a,b
a
Biomedical Health Sciences Institute, Division of Neuroscience, University of Georgia, Psychology Building, Athens, GA., 30602, USA
b
Department of Psychology, University of Georgia, Psychology Building, Athens, GA., 30602, Athens, GA., USA
c
Department of Physics and Astronomy, University of Georgia, Physics Building, Athens, GA., 30602, USA
d
Department of Computer Science, University of Georgia, 415 Boyd Graduate Studies Research Center, Athens, GA., 30602, USA

a r t i c l e i n f o a b s t r a c t

Article history: The working memory (WM) system is vital to performing everyday functions that require attentive, non-
Received 7 May 2010 automatic processing of information. However, its interaction with long term memory (LTM) is highly
Revised 6 December 2010 debated. Here, we used fMRI to examine whether a popular complex WM span task, thought to force the
Accepted 9 December 2010
displacement of to-be-remembered items in the focus of attention to LTM, recruited medial temporal regions
Available online 21 December 2010
typically associated with LTM functioning to a greater extent and in a different manner than traditional
Keywords:
neuroimaging WM tasks during WM encoding and maintenance. fMRI scans were acquired while participants
Working memory performed the operation span (OSPAN) task and an arithmetic task. Results indicated that performance of
Hippocampus both tasks resulted in significant activation in regions typically associated with WM function. More
Complex span importantly, significant bilateral activation was observed in the hippocampus, suggesting it is recruited
fMRI during WM encoding and maintenance. Right posterior hippocampus activation was greater during OSPAN
Executive control than arithmetic. Persitimulus graphs indicate a possible specialization of function for bilateral posterior
hippocampus and greater involvement of the left for WM performance. Recall time-course activity within this
region hints at LTM involvement during complex span.
© 2010 Elsevier Inc. All rights reserved.

Introduction is at some point stored in a location, LTM, from where it can later be
retrieved by another system, short term memory (STM). The informa-
Working memory (WM) is thought of as a system in which tion is then manipulated, updated, and maintained in accordance with
information currently in the focus of attention can be maintained and the aim of the present goal state. Of importance is also the idea that the
manipulated. It is also seen as a gateway through which sensory information held in STM does not have to be retrieved from LTM, but
information can enter into long term memory (LTM) or through maybe newly acquired information that has been linked with other
which information can be recruited from LTM into the focus of information in LTM. Linking, or relational encoding, is necessary in order
attention (Atkinson and Shiffrin, 1968; Baddeley and Hitch, 1974; to attach meaning to the newly acquired information.
Cowan, 1988; Engle et al., 1999a; Unsworth and Engle, 2007b). A Many of the current discussions on WM have emphasized the
properly functioning WM system enables an individual to keep concept of capacity limits. To describe this concept, Cowan's (1988,
attention on a desired goal while preventing other environmental or 1999, 2005) embedded processes model examines three states of
cognitive stimuli from interfering with the completion of the desired memory: the information residing in LTM, recently perceived or
goal. Furthermore, WM is crucial when attempting to override our accessed information that is in an easily accessible (activated) state in
automatic responses through a set of cognitively salient processes LTM, and a sub-portion of that information which we are consciously
(Unsworth and Engle, 2007b), thus making it critical for the aware of, known as the focus of attention. WM capacity differences are
performance of a variety of everyday tasks. believed to arise from the ability to keep the focus of attention on the
Atkinson and Shiffrin (1968) and Baddeley and Hitch (1974) task at hand while suppressing interference from environmental stimuli
espoused the idea that the process of WM is that by which information or irrelevant cognitions caused by the activation of other memory traces
in LTM. Much in the same way, Unsworth and Engle (2006, 2007b)
suggest that differences in WM capacity arise from an individual's ability
to actively maintain information in primary memory (i.e., the focus of
⁎ Corresponding author. Fax: +1 706 542 5285. attention) while also performing a controlled search of the information
E-mail addresses: cfaraco@uga.edu (C.C. Faraco), nunswor@uga.edu (N. Unsworth),
impulse@physast.uga.edu (J. Langley), dpterry@uga.edu (D. Terry),
residing in secondary memory (i.e., LTM).
likaiming@gmail.com (K. Li), lczhdgm@gmail.com (D. Zhang), tliu@cs.uga.edu (T. Liu), Several tasks have traditionally been used to assess WM capacity.
lsmiller@uga.edu (L.S. Miller). For example, the typical digit span task assesses capacity by

1053-8119/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2010.12.033
774 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

determining the maximum length of numbers that an individual can has been shown to occur specifically during the retrieval stage of the
serially recall. More involved tests, such as the digits backwards and recent negative condition (D'Esposito et al., 1999) and is linked to the
letter-number sequencing tasks, assess capacity while also requiring resolution of interference. A more recent review by Blumenfeld and
the ability of mental double-tracking, meaning that memorizing and Ranganath (2007) has further indicated the VLPFC's role in the
reversing/ordering operations must be performed simultaneously resolution of interference by noting that it is consistently recruited
(Lezak et al., 2004, pp. 359–363). Other tasks like Daneman and when controlled selection of items is required. The DLPFC's roles in
Carpenter's (1980) reading span task and Turner and Engle's (1989) executive processing are further confirmed by demonstrating it is
operation span (OSPAN) task are complex working memory span highly recruited for the organizational processing of information.
(CWMS) tasks that require the participant to engage in a processing Blumenfeld and Ranganath (2007) have also summarized the roles of
activity that is irrelevant to the to-be-remembered information. They VLPFC and DLPFC in LTM, suggesting VLPFC supports the formation of
involve encoding, maintenance, storage, and processing of various LTMs through the process of controlled item selection, while the
types of information. Proper performance on CWMS tasks requires a DLPFC aids in building associations between items in LTM and those in
high degree of executive attentional-control (Conway et al., 2003; the focus of attention. Another key region in the frontal lobes, the
Engle et al., 1999a; Kane et al., 2007a) to switch between tasks and anterior cingulate cortex (ACC), is also believed to be necessary for
maintain attention on the task at hand. Of critical importance, the proper WM function and is thought to be involved in conflict
irrelevant task is often thought to force the to-be-remembered monitoring and error detection (Bernstein et al., 1995; Botvinick et
information to be temporarily displaced from the focus. If the to-be- al., 2001; MacDonald et al., 2000). Both of these are attentional control
remembered information is properly encoded, it may be placed into processes and as such the ACC is believed to be critical to cognitive
and retrieved from LTM as required (Kane et al., 2007a). The control (Smith and Jonides, 1999; Osaka et al., 2003). Furthermore,
displacement of task relevant information from the focus occurs Kaneda and Osaka (2008) suggest that the ACC may play a greater role
because the irrelevant task usually requires controlled, effortful in executive functioning than the DLPFC.
processing (Conway and Engle, 1996; Engle et al., 1999b); it is of The parietal lobes are thought to function as associative centers
sufficiently high cognitive load that it may occupy the whole of the and be involved in higher level cognitive processes. They are also
focus of attention, thereby displacing any information which exceeds believed to be crucial to WM processes and serve as storage regions
the individual's immediate WM capacity (Bunting, 2006; McCabe, (Hamidi et al., 2008; Postle, 2006; Postle and D'Esposito, 1999; Rowe
2008). It is this type of complex processing, and the resultant et al., 2000; Srimal and Curtis, 2008). More specifically, the superior
interactions of items coming into and going out of the focus, that make parietal lobule (SPL) and the precuneus (Brodmann area 7) are
CWMS tasks invaluable for the cognitive study and neuroimaging of believed to be crucial in maintaining and organizing items held in the
WM. WM store (Wager and Smith, 2003; Wendelken et al., 2008), while
the supramarginal gyrus (part of the inferior parietal lobule) is
Neuroanatomical regions traditionally associated with WM tasks thought to retrieve the temporal ordering of items that have been
displaced from the focus of attention through serial scanning (Öztekin
Functional neuroimaging experiments of WM have typically used et al., 2008). Parietal cortex is also thought to select the appropriate
tasks such as the Sternberg (1966) or the n-back (Gevins et al., 1990); response for a specific stimulus, known as stimulus-response
we will refer to these types of tasks as traditional neuroimaging WM mapping (Corbetta and Shulman, 2002; Miller, 2000; Miller and
(TNWM) tasks. During the Sternberg task subjects are presented with Cohen, 2001).
a set of stimuli and are asked if the target stimulus matches any of the The medial temporal lobes (MTL) have traditionally been asso-
stimuli presented in the previous set. The n-back task is more complex ciated with the encoding and maintenance of declarative LTMs. For
in that there is a continuous presentation of stimuli; on target trials example, Scoville and Milner's (1957) classic study demonstrated that
subjects are asked either if the target stimulus matches a stimulus bilateral lesions to the hippocampal formation produced an extremely
presented n trials back (usually 1 to 3) or to identify how many trials detrimental impact on the retention of new memories. More recently,
back the target stimulus was shown. Generally speaking, most neuroimaging studies have challenged this limited conception of MTL
neuroimaging investigations of WM have associated a core of regions regions by demonstrating hippocampal recruitment during various
with WM functioning. types of WM tasks, and specifically, during the maintenance phases of
The prefrontal cortex (PFC) is believed to be integral to WM and some of these tasks. Öztekin et al. (2009) found the hippocampus was
executive control (D'Esposito et al., 2000; Owen et al., 2005; Wager active during item recognition trials of a serial position task and that
and Smith, 2003). Sub-sections of the PFC, such as the ventrolateral activity increased for earlier items rather than the last item on a
prefrontal cortex (VLPFC) and dorsolateral prefrontal cortex (DLPFC), judgment of recency task. Using neurosurgically implanted electro-
have been said to be involved in object and spatial domain specific encephalograph (EEG) electrodes, Axmacher et al. (2007) detected
processing (Courtney et al., 1998; Smith and Jonides, 1999), significant load dependent negative DC potential shifts and increases
respectively. However, an extensive review of the neuroimaging in synchronized gamma band activity in the rhinal cortex during the
literature by Wager and Smith (2003) indicated that PFC sub-regions maintenance of multiple items during a visual Sternberg task. The
were not so much domain specific as they were process specific. The negative DC shift likely indicating membrane potential depolarization
DLPFC being involved in executive processes, such as attentional and increased firing and/or synaptic activation of rhinal cortex
control or the monitoring of complex information (Cabeza and neurons, while synchronized gamma band activity further indicated
Nyberg, 2000), while the VLPFC is involved in storage-related recruitment of the rhinal cortex. Van Vugt et al. (2010) furthered these
processes such as the maintenance of spatial information (Toepper findings by demonstrating a local load dependent gamma oscillatory
et al., 2010) or the rehearsal of verbal information (Cabeza and power increase in the hippocampus during Sternberg task mainte-
Nyberg, 2000). Bor et al. (2006) showed activation of the VLPFC nance. Additionally, this increase was greater for non-verbal items
during a task where spatial information had to be kept online without (faces) than for verbal items (letters).
aid of a spatial strategy; when spatial strategies for remembering the
target stimuli were given, activation was only exhibited in the DLPFC. What makes CWMS tasks valuable neuroimaging tools?
Further involvement of the VLPFC in storage processes has been
demonstrated in proactive interference tasks where the left inferior Even though item recognition tasks such as the n-back and the
frontal gyrus has shown significant activation during a recent negative Sternberg have proven to be in valuable in dissociating many of the
condition (Jonides et al., 2000; Jonides and Nee, 2006). This activity brain regions involved in WM functioning, there are reasons to
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 775

explore the use of CWMS tasks in neuroimaging settings. For example, recruitment occurs during the encoding and maintenance phase of
the n-back has been shown to account for variability in general fluid the task.
intelligence (Gf), but it has done so only under a 3-back condition, and We hypothesized that 1) the OSPAN encoding and maintenance
this variance in Gf is separate than that accounted for by WM span phase and Arithmetic task would yield activation in regions
(Kane et al., 2007b). WM span, as measured by a CWMS task though, commonly associated with WM and the resolution of interference
has been shown to account for up to half the variability in Gf (Conway during on-going retrieval, such as VLPFC, DLPFC, ACC, SPL, and inferior
et al., 2003; Kane et al., 2005). A CWMS task like the OSPAN has been parietal lobule (IPL), 2) activations in these regions would be greater
shown to possess high levels of reliability and internal consistency as during the OSPAN since CWMS tasks should require greater executive
compared to other measures of WM capacity (Klein and Fiss, 1999). control than a typical neuroimaging WM task, 3) due to the nature of
An automated version of the OSPAN has also demonstrated high levels the OSPAN, activation would also be evidenced in areas typically
of reliability and internal consistency, and shown high levels of associated with LTM binding and retrieval, specifically the hippocam-
correlation with other measures of WM (Unsworth et al., 2005). pus, and 4) the nature of CWMS tasks would be sufficiently different
Moreover, more recent work has demonstrated correlations between from TNWM tasks, resulting in unique patterns of activity within LTM
complex WM span tasks and traditional measures of LTM (Unsworth associated regions such as the hippocampus. This would be the first
et al., 2009; Unsworth, 2010) at sub-span levels, unlike TNWM tasks. instance where such activity would be demonstrated for a CWMS task
Such correlations make the case for the use of CWMS tasks in during maintenance and encoding. We also aimed to explore what
neuroimaging WM research, especially when attempting to decipher pattern of brain activity during a CWMS task is correlated with correct
the possible interplay of WM and LTM. and incorrect recall. In other words, we explored what patterns of
Unfortunately, CWMS tasks have seen little use in neuroimaging brain activation are associated with WM capacity and proper storage
studies. In one of the few, Kondo et al. (2004) found the OSPAN and retrieval. Edin et al. (2009) indicates we may find correct recall is
elicited activation in regions usually activated during the n-back (e.g., associated with heightened DLPFC activity which modulates parietal
PFC, ACC, and SPL), while the high-span group also exhibited activation.
significant activation in the inferior temporal cortex. Kondo et al.
(2004) was limited, however, in that they mainly restricted their Materials and methods
analysis to the functional connectivity differences of the cingulo-
frontal network between high-span and low-span individuals. Participants
Recently Chein et al. (2010) examined domain general mechanisms
during encoding and maintenance, and examined MTL activity during Twenty-five young adults from the University of Georgia were
recall for verbal and spatial complex span tasks. They found activity in recruited for this study through the university's research participant
areas typically associated with WM during encoding and mainte- pool and through word of mouth; 17 females and 8 males, average
nance, and found the posterior hippocampus and immediately age = 24.8 ± 2.8. Exclusion criteria included self-report of previous
inferior portion of the parahippocampal gyrus were involved in the head injury, history of loss of consciousness, current drug abuse,
recall portion of the task. However, they did not specifically examine evidence of neurodegenerative processes, and an estimated below
MTL recruitment during encoding and maintenance. Therefore, a average IQ. Participants could also not have a history of, present
more in depth neuroimaging exploration of CWMS tasks and the roles clinical signs of, or currently be under treatment for, any major
of MTL ROIs during encoding and maintenance is warranted. psychiatric symptoms or disorders. The exception to the latter
exclusion criteria was a past history of depression, since a significant
portion of the population may have at one point presented with
The current study clinically diagnosable symptoms (Pratt and Brody, 2008). Incompat-
ibility with the MRI environment (e.g., metallic implants, pacemakers,
In this study we aimed to elucidate whether significant differences stents, etc.) was assessed through a standardized screening form and
exist between the neural resources required for the performance of participants were excluded given any signs of incompatibility.
CWMS and TNWM tasks. More precisely, we wanted to determine if the
encoding and maintenance phase of a CWMS task results in significantly Measures
greater recruitment of areas typically associated with LTM functioning
than might occur during a TNWM task. As previously stated, Axmacher et Participants were made aware of the exclusion criteria before
al. (2007) and Van Vugt et al. (2010) demonstrated hippocampal activity participating in the study. Upon meeting with the investigator,
during a Sternberg task maintenance, a TNWM task. To examine this, we participants were fully screened. Screening included completing the
compared functional magnetic resonance (fMRI) activity observed MRI screening form, answering questions from the psychotic
during the OSPAN task (complex span; letter span + equation verifica- symptoms screening portion of the Structured Clinical Interview for
tion) with that of an arithmetic task (traditional type of neuroimaging DSM-IV (SCID-I; First et al., 1997), and being asked the exclusion
WM task; equation verification). criteria questions described earlier. Additionally, female participants
The goal of the OSPAN task is to recall the to-be-remembered were asked to take a pregnancy test; even though the MRI
items (letters) in serial order. During the OSPAN, equation verification environment has been shown to have no adverse side-effects, this
is presented as the irrelevant, cognitively demanding task. As a result, was taken as a precautionary measure. If screening was successful,
participants should often not have sufficient cognitive resources participants were given a brief IQ estimate, the Wechsler Test of Adult
available to rehearse the to-be-remembered letters while performing Reading (WTAR; Wechsler, 2001) to rule out below average IQ.
equation verification. In other words, the equation verification should
occupy the focus of attention causing a displacement of the to-be Task and stimuli
remembered letters from the focus to LTM. If the to-be-remembered
items have been displaced from the focus and properly stored in LTM, Participants performed the OSPAN task in a similar fashion to
they can later be retrieved from LTM as needed. By contrasting the Kondo et al. (2004). The full presentation of the task lasted 6 m and
OSPAN and Arithmetic conditions we hoped to control for the 45 s, with fixed alternating conditions of OSPAN, Arithmetic, and
common activation patterns resulting from equation verification in Baseline; there were 3 OSPAN, 3 Arithmetic, and 6 Baseline epochs,
order to demonstrate that the OSPAN task forces recruitment of each lasting 30 s. The OSPAN epochs were always followed by 15 s
regions associated with LTM binding and retrieval, and that Response epochs (Fig. 1). Each run was preceded by a set of visual
776 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

instructions and contained a total of 15 epochs; 3 OSPAN + 3 Procedure

Response + 3 Arithmetic + 6 Baseline. The OSPAN, Arithmetic, and
Baseline conditions were structured so that participants received Participants initially practiced the task by viewing it on a computer
similar amounts of visual input and gave the same amount of motor monitor and tapping their finger to the appropriate response as they
output (Fig. 1). would with the response pads in the MRI unit. On average,
During the Baseline condition participants were presented with participants required 3 min before the investigator acknowledged
arrows pointing either left or right (4 s) and responded by indicating they were performing the task appropriately. After practice, partici-
which direction the arrow was pointing with the appropriate button pants were placed in the MRI scanner. During the structural scan
press. Presentation of arrows alternated with the presentation of participants performed a practice run of the task in order to further
asterisks (2 s). During the Arithmetic condition participants were become acquainted with the task and scanner environment. Partici-
presented with an equation (4 s) consisting of two operations, pants then performed 2 runs while fMRI data was acquired. The task
multiplication or division and addition or subtraction. Their task was designed in E-Prime, version 1.2 (Psychology Software Tools,
was to judge whether the equation was correct or incorrect and to 2006), stimuli were presented through MRI compatible goggles
respond by pressing the appropriate button on the response pad. In (Resonance Technology Inc.), participants responded through Cedrus
between the presentation of the equations they were shown an LUMINA MRI compatible response pads by using their index and
asterisk (2 s). During the OSPAN task participants also had to judge middle fingers, and responses and reaction times (RTs) were recorded
equations (4 s), but instead of asterisks they were presented with by E-Prime. Behavioral data was acquired during all 3 (1 practice, 2
letters (2 s) which they were instructed to remember in serial order. experimental) runs in the scanner.
Within all these block types, the sequences repeated five times. After
each OSPAN block, there was a response block in which participants MRI acquisition
identified which letters were previously presented. They were shown
5 separate arrays, each consisting of 4 letters, for 3 s each. They were 3D structural scans were acquired using a fast spoiled gradient
to identify the letters presented with the appropriate button presses. recalled echo (FSPGR) protocol; TE = min full, TR = 7.5 ms, flip
For the first array the participant identified which of the letters was angle = 20°, 154 axial slices, slice thickness = 1.2 mm, and
first presented, for the second array they identified which was FOV = 256 × 256 mm. These images covered from the top of the
presented second, and so on. For any of the epochs, if the participants head to the brainstem and acquisition took approximately 6 m and
responded to a prompt after the allotted time, the response was 20 s. Functional scans were acquired using a T2*-weighted single shot
considered incorrect. echo planar imaging (EPI) sequence and were aligned to the

Condition O R B A B

Duration (s) 30 15 30 30 30

OSPAN
4s 6s

2s
(5 x 3) + 7 = 23
12 s
C

(8 x 5) - 3 = 37
18 s
Y

(24 ÷ 2) - 1 = 11
Response 3s 24 s
3s
Q
6s

Y F Q C (15 ÷ 5) - 8 = -5 30 s
9s
F
Q Y L G
12 s
(23 x 2) + 7 = 53
C F L Q 15 s
L
L R F C

L C Y Q

Fig. 1. Stimulus presentation. Top: Block design for stimuli; this sequence was repeated 3 times per run. O: OSPAN, R: word recognition, B: baseline, and A: Arithmetic condition.
Bottom: Progression of OSPAN block and the proceeding Response block where participants identify the letters presented during the OSPAN block in serial order. For the Arithmetic
blocks, participants were presented with equations, as in the OSPAN, but asterisks were presented instead of letters. During Baseline participants were presented with either left or
right pointing arrows instead of equations, and asterisks instead of letters.
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 777

intercommisural line (AC–PC line); TE = 25 ms, TR = 1500 ms, 90° RF 2002). Images were smoothed using a 6.875 mm isotropic FWHM
pulse, 30 interleaved slices, acquisition matrix = 64 × 64, spa- Gaussian smoothing kernel, twice the voxel dimensions in the x and y
cing = 0 mm, slice thickness = 4 mm, FOV = 240 × 240 mm, and planes. A high-pass temporal filter, calculated at 135 s (OSPAN +
ASSET factor = 2. Functional images covered the entire cortical surface Response + Baseline + Arithmetic + Baseline times), was applied. The
and a portion of the cerebellum. Each run consisted of 270 volumes data were prewhitened to remove inter-voxel auto-correlation. Head
and 10 dummy samples were discarded during the initial acquisition. motion parameters estimated from MCFLIRT were added as confound/
regressor variables to the design. A standard hemodynamic response
Data analysis function was convolved with each run's time course. All results were
warped to the 91x109x91mm MNI standard brain using a 12-degree
All data were processed using the FMRIB Software Library (FSL; affine transform.
Smith et al., 2004; Woolrich et al., 2009). Before MRI data sets were We contrasted the OSPAN and Arithmetic blocks with Baseline,
analyzed using the FMRI Expert Analysis Tool (FEAT), they were and the OSPAN to the Arithmetic blocks. Comparing the task blocks to
converted from their native GE DICOM format to NIFTI format using Baseline allowed us to determine the regions recruited for each task
the dcm2nii conversion tool (Rorden, 2007). Each participant's fMRI and also gave us a qualitative indication of how similar or distinct
data was motion corrected using the Motion Correction FMRIB Linear these regions were. The OSPAN N Arithmetic contrast yielded regions
Registration Tool (MCFLIRT; Jenkinson et al., 2002). The images were likely associated with the high degree of attentional control required
then slice time corrected to account for the interleaved acquisition during CWMS tasks and the storage and retrieval of information to
and then brain extracted using the Brain Extraction Tool (BET; Smith, and from LTM. Between-subjects voxel-wise analyses were performed

Fig. 2. Depiction of linear trend which created a bias in the hippocampal signals. The top two averaged time series depict the linear trend that occurred for the first 53 volumes of
acquisition within the hippocampus; this included the initial OSPAN, baseline, and Arithmetic blocks. The lower figure depicts the time-series for the left DLPFC. This region was
randomly chosen from our list of ROIs to determine whether this trend was global or localized to the hippocampi. The linear least squares lines of best fit are plotted on the graphs to
further emphasize the linear trend within the hippocampus for the first 53 volumes.
778 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

using a mixed-effects model. Whole brain group results were segmentation. A mask was then made for each subject's hippocampi
thresholded using FSL's cluster thresholding, Z N 4.0 and p = 0.05. from the respective segmentation results. Hippocampus analysis in
Even at this high threshold, results for the task minus baseline FEAT followed the standard procedure outlined above, except that at the
contrasts were still extremely robust. We increased the Z threshold first level each subject's brain was only registered to their respective 3D
for these contrasts to 4.5 in order to parcellate some of the clusters anatomical scans. The affine transformations from FIRST, optimized for
into more meaningful regions. Increasing the threshold did not cause subcortical alignment to the standard MNI brain, were then convolved
any activation areas of interest to become non-significant. with the EPI to 3D transformations in order optimize the EPI to MNI
For hippocampal ROI analysis, FMRIB's Integrated Registration Tool registration for hippocampal alignment between subjects. A smaller
(FIRST; Patenaude et al., 2007) was implemented for subcortical 5 mm isotropic FWHM Gaussian smoothing kernel was applied to the

Fig. 3. Percent signal change values for ROIs. Graphs represent percent signal changes for activated ROIs with the S.E.M. OSPAN and Arithmetic conditions are presented with bilateral
measures where appropriate.
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 779

hippocampal EPI data due to the restricted area of analysis. Group-wise motor output that occurs during the OSPAN and Arithmetic blocks.
ROI hippocampal analysis used a group modal mask of the hippocampi Since participants performed well on both tasks, volumes spanning
dilated by one voxel in every direction. incorrect equation judgment or letters incorrectly recalled during the
Persitimulus plots for the hippocampi were calculated by taking Response block were not excluded from the analysis. Activation sites
the mean values of the hippocampi for each time point across for the OSPAN and Arithmetic blocks were similar with more robust
participants, using the modal mask mentioned above, and averaging activation occurring during the OSPAN. Activity in frontal regions was
these values across blocks and then runs. Percent intensity change detected in the DLPFC, frontal orbital cortex, frontal operculum,
(PIC) plots were calculated by taking the PIC values for each subjects' middle frontal gyrus, precentral gyrus, superior frontal gyrus,
runs and averaging them across time, resulting in a scalar value for supplementary motor area (SMA), and VLPFC. In parietal regions,
each subject. These values were then averaged across subjects. this included the precuneus, the IPL (supramarginal and angular
Additionally, PIC and persitimulus plots were calculated from voxels gyrus), and the SPL. Activation was also evidenced in the ACC,
falling within the top 10% intensity range within the ROI. Baseline paracingulate gyrus, posterior cingulate gyrus, and the anterior
values were calculated by averaging the last two volumes of each insular cortex. Percent signal changes for the main ROIs are displayed
baseline block. Mean hippocampal PIC values used baseline values in Fig. 3. Activity detected in occipital regions included the lingual
from the entire hippocampus, while the top 10% used the baseline gyrus, occipital fusiform gyrus, and the occipital pole. In general, both
values from the top 10% voxels. Task and baseline values were tasks elicited bilateral activation in areas typically associated with
calculated separately for each hemisphere. WM functioning.
Notably, plots of the entire mean and top 10% time-series depicted Of greater interest, was that the whole brain group analysis gave
linear increases in voxel intensity through approximately the first 53 indication that OSPAN recruited some hippocampal areas by demon-
volumes of each run. This trend greatly affected the resulting statistics strating activation in regions bordering, or encompassing parts of,
and persistimulus plots because intensity values for the initial OSPAN bilateral hippocampus (Table 1). The Arithmetic task also appeared to
blocks start below baseline values and increase linearly, through the recruit some hippocampal regions as indicated by an area of activation
initial baseline, until the third or fourth volume of the initial bordering around the left hippocampus. ROI voxel-wise analysis of the
Arithmetic blocks (Fig. 2). A time series plot was also calculated for hippocampi (task N baseline; Z N 2.3, p =0.05) resulted in strong
another region, the left DLPFC, to examine whether the linear trend activation encompassing the dentate gyrus, posterior portions of the
was global or localized to the hippocampi. This trend was not subiculum, and posterior/mid-posterior cornu ammonis regions of the
observed for the DLPFC (Fig. 2). Thus we present hippocampus results hippocampi for both tasks as compared to baseline (Fig. 4). The ROI
calculated from 200 volumes of data per run, or 80 volumes of data for results further demonstrate that MTL regions are recruited during
both the OSPAN and Arithmetic conditions. These results excluded the CWMS tasks, but also demonstrate that they are recruited at least
first 70 volumes, or the first OSPAN (TRs 1–30), baseline (TRs 31–50), during more traditional types of neuroimaging WM tasks such as
and Arithmetic (TRs 51–70) blocks. The hippocampal ROI voxel-wise arithmetic.
group analysis was thresholded at Z N 2.3 and p = 0.05.
An uncorrected, voxel threshold of p = 0.005, whole brain group
analysis was performed on the OSPAN N Arithmetic contrast using the Table 1
MTL regions of activation.
average number of correct letter identification responses (demeaned)
as a regressor. This was done to preliminarily identify regions whose Region Coordinates in mm, MNI Z-
activity during the encoding and maintenance might be significantly score
x y z
correlated with correct and incorrect letter identification during the
Whole brain analysis
recall period. Given the drawbacks associated with uncorrected voxel- OSPAN N Baseline
wise analysis, a cluster threshold of 20 voxels was chosen in order to L thalamus − 10 − 20 16 6.66
extract clusters with a higher likelihood of truly significant activation. L caudate − 14 2 16 5.90
Peak activated voxels from the identified clusters were dilated by an L hippocampus/lateral GENICULATE − 24 − 32 −2 5.10
R thalamus 18 − 14 8 4.73
8 mm sphere to use as masks for a more stringent cluster thresholded
R hippocampus 30 − 30 −4 4.73
voxel-wise ROI analysis, Z N 2.3 and p b 0.05, correlating these regions
with correct and incorrect letter identification responses, respectively. Arithmetic N Baseline
L hippocampus/lateral geniculate − 24 − 28 −4 5.3
Results
ROI analysis
OSPAN N Baseline
Behavioral results R hippocampus, DG/CA 28 − 30 −6 6.58
R hippocampus, DG/CA 26 − 34 0 5.88
All participants completed both functional runs; answering the R hippocampus, CA/DG 36 − 26 − 10 5.37
L hippocampus, DG/CA − 24 − 34 −2 7.46
equations during the OSPAN and Arithmetic conditions at 87% (M = 13,
L hippocampus, DG − 26 − 26 − 10 7.06
SD= 2.35) and 85% (M = 12.8, SD= 2.93) accuracy, respectively, and
recalling the letters in serial order during the OSPAN at 87% (M = 13.04, Arithmetic N Baseline
SD= 3.43) accuracy. Average RTs for the equations were 2371.29 ms R hippocampus, DG/CA 28 − 28 −8 6.04
(SD= 169.31) for the OSPAN blocks and 2458.11 ms (SD= 175.39) for R hippocampus, DG/CA 26 − 34 0 5.02
R Hippocampus, CA/DG 36 − 26 − 12 4.63
the Arithmetic blocks. Paired sample t-tests revealed a significant effect
R hippocampus, CA 18 − 40 2 2.5
for equation verification RT, t(24) = 4.15, p b 0.001, with RTs during the L hippocampus, DG/CA − 26 − 30 −8 8.25
OSPAN blocks occurring faster; no significant effect was found for L hippocampus, DG −22 − 32 −6 7.13
number of correct equations t(24) = 0.661, p b 0.515. These RT results L hippocampus, DG/CA − 24 − 34 −2 6.68
are in accordance with those in Kondo et al. (2004).
OSPAN N Arithmetic
R hippocampus, CA/DG 36 − 26 − 12 4.3
fMRI results
Regions of MTL activity for the Task N Baseline and OSPAN N Arithmetic contrasts. The
top half presents activity detected for the whole brain contrast, while the bottom half
The OSPAN and Arithmetic blocks were contrasted against presents the hippocampus ROI analysis results. Whole brain thresholded at Z N 4.5,
Baseline (Z N 4.5, p = 00.05), controlling for the visual input and p b 0.05; ROI thresholded at Z N 2.3, p b 0.05.
780 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

Even though our initial OSPAN N Baseline results were encourag-

ing, our main goal was to demonstrate neural activation supporting
the idea that during encoding and maintenance CWMS tasks recruit
areas traditionally associated with LTM (i.e., the hippocampus) to a
greater extent than TNWM tasks. The whole brain, OSPAN N Arith-
metic, group activation map (Z N 4.0, p = 0.05; Fig. 5) revealed robust,
bilateral differences in a variety of regions. Areas of increased
activation for the OSPAN N Arithmetic contrast are listed in Table 3.
Results demonstrated that the areas typically associated with WM
tasks are much more active during complex than TNWM tasks. For
example, in the frontal lobes greater activation was observed in the
superior and middle frontal gyri, the paracingulate gyri, ACC, IFG, pars
opercularis, DLPFC, precentral gyri. In the parietal cortex, greater
activation was observed in the SPL, IPL, and precuneus. The
supplementary motor cortex and the precentral gyrus exhibited the
greatest differences, as the largest, most robust activation clusters
encompassed these regions. Differences were also observed in regions
less commonly reported in past studies, such as the right insula and
left posterior cingulate cortex.
Fig. 4. Activation map for hippocampal ROI analysis. The top two figures represent The hippocampus ROI voxel-wise analysis (Z N 2.3, p = 0.05)
activation for OSPAN (left) and Arithmetic (right) as compared to Baseline, cluster revealed a cluster of 110 voxels with significantly greater activation
thresholded at Z N 2.3, p b 0.05. Activation covers a portion of the subiculum and part of in the posterior right hippocampus for OSPAN N Arithmetic (Fig. 4,
the cornu ammonis regions. B) OSPAN N Arithmetic activation demonstrating signifi-
Table 1). The Arithmetic N OSPAN contrast did not yield any
cantly greater right posterior hippocampus activation, cluster thresholded at Z N 2.3,
p b 0.05. significant hippocampal differences. Mean and upper 10% values

Fig. 5. Activation map for the OSPAN N Arithmetic contrast. Cluster thresholded, Z N 4.5, p b 0.05. The contrast image shows greater activation in cortical regions typically associated
with working memory, this is likely due to the higher level of executive attentional control required during the OSAPN task. Images are in radiological convention, from inferior to
superior.
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 781

Fig. 6. Peristimulus graphs and percent signal change graphs for the hippocampi. Persitimulus graphs are presented for the upper 10% and mean values of both hippocampi. Values
were taken from an a-priori modal mask of the hippocampus that was dilated by one voxel in every direction. Differences between values in each of the graphs are significant across
time. Peak and mean persitimulus graphs are drawn with a line at the 20 TR mark signaling the end of the OSPAN encoding and maintenance phase, and the end of the Arithmetic
task. Following the OSPAN encoding and maintenance phase is the recall portion of the task. The initial peak in the mean graphs likely signals the retrieval of task relevant
instructional sets, while a similar peak is evidenced after termination of the blocks, likely indicating retrieval of the instructional set for the next task. The mean peristimulus graph
shows a deactivation of overall hippocampus function. This is in contrast to the upper 10% graphs which show above baseline levels of activity. Voxels falling within the upper 10%
were in the posterior hippocampus across all subjects. Overall, these graphs indicate a possible specialization of function within the posterior hippocampus for working memory
encoding and maintenance. Additionally, the increased activation in the posterior left hippocampus during OSPAN retrieval hints that items may have been displaced to LTM during
complex span encoding and maintenance.

for the hippocampi (Fig. 6) were significantly greater for OSPAN diate memory functioning; 2) complex arithmetic (semantic fact-
across time between conditions; mean: left, t(39) = 2.573, p b 0.014 retrieval) and CWMS tasks do involve access to LTM through the
and right, t(39) = 3.594, p b 0.001; upper 10%: left, t(39) = 3.796, hippocampus, but for different reasons; and 3) anterior hippocam-
p b 0.001 and right, t(39) = 3.124, p b 0.003. Values were also pal regions may specifically be part of the default/resting state
significantly different across hemispheres within conditions, with network.
mean values greater on the right and upper 10% greater on the left; Group analyses using the average number of correct serially
mean: OSPAN, t(39) = 6.330, p b 0.000 and Arithmetic, t(39) = identified letters per participant as a regressor, thresholded at an
3.995, p b 0.000; upper 10%: OSPAN, t(39) = 3.068, p b 0.004 and uncorrected p = 0.005 with a cluster size threshold ≥ 20, yielded
Arithmetic, t(39) = 2.534, p b 0.015 (refer to Table 2 for a summary regions that could be positively and negatively correlated with the
of these results). These results support our hypothesis and suggest a correct and incorrect recall (Table 4). Regions positively correlated
few possibilities in regards to specialization of function within the with letter identification included left paracingulate/ACC, left medial
hippocampus: 1) posterior hippocampus may be critical to imme- temporal gyrus, right frontal pole, and pars opercularis of the right
782 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

Table 2
Paired differences.

Paired%-sc Mean SD S.E.M. 95% C.I. t df Sig.

t-test % Diff. (2-tailed)
Lower Upper
differences for
the hippocampus

Differences in mean hippocampal signal change

Left hippocampus 0.0197 0.0484 0.0077 0.0421 0.0352 2.573 39 0.014
(Arithmetic–OSPAN)
Right hippocampus 0.0257 0.0453 0.0072 0.0113 0.0402 3.594 39 0.001
(Arithmetic–OSPAN)
Arithmetic (right–left) 0.0151 0.0238 0.0038 0.0227 0.0074 3.995 39 0.000
OSPAN (right–left) 0.0211 0.0211 0.0033 0.0278 0.0143 6.330 39 0.000

Differences in upper 10% hippocampal signal change

Left hippocampus 0.0176 0.0293 0.0082 0.0082 0.0269 3.796 39 0.001
(Arithmetic–OSPAN)
Right hippocampus 0.0156 0.0315 0.0050 0.0055 0.0256 3.124 39 0.003
(Arithmetic–OSPAN)
Arithmetic (left–right) 0.0246 0.0246 0.0039 0.0020 0.0177 2.534 39 0.015
OSPAN (left–right) 0.0244 0.0244 0.0039 0.0040 0.0197 3.068 39 0.004

Paired t-test differences for percent signal values from the right and left hippocampi under the OSPAN and Arithmetic conditions. All t-test were significantly different across
comparisons. Values were tested across time and the means of each individuals mean time points were supplied in the analysis.

inferior frontal gyrus. Regions negatively correlated with letter ing parts of the post-central and supramarginal gyri was found to be
identification included left superior parietal lobule, right lateral significant (Table 4). The apparent ceiling effect with regards to
occipital/cuneal cortex, left parietal operculum cortex, right post- number of correct responses (87%) may partly explain why few
central gyrus, and left angular gyrus. Masks from these regions were cortical regions were found to be positively and negatively correlated
then used for ROI analysis of positive and negative correlation with with the encoding, maintenance, and possible displacement of items
correct recall using a more stringent cluster threshold (Z N 2.3; to WM under the ROI analysis.
p = 0.05). Of the positively correlated regions used as masks, a cluster
encompassing the anterior cingulate and the frontal pole was found to
be significant. For negatively correlated regions, a cluster encompass- Table 4
Regions correlated with correct letter serial recall.

Cluster Region Brodmann Coordinates Z-

Table 3 size area in MNI, mm score
Regions of significant differences; OSPAN N Arithmetic.
x y z
Cluster Region Brodmann Coordinates in Z- Whole brain
area mm, MNI score Positively correlated
x y z 49 L paracingulate/anterior 32 −2 54 6 3.16
cingulate cortex
1 R and L supplementary motor 6 0 0 62 7.62 45 L middle temporal gyrus − 52 − 46 −8 4.28
cortex 38 R occipital pole 17 12 − 92 −6 4.21
L precentral gyrus 3/4 − 50 −6 42 6.41 27 R DLPFC 9 2 60 12 4.16
L superior/middle frontal gyrus 6 − 24 −4 52 6.26 24 R inferior frontal gyrus, 44 60 14 18 4.09
R and L paracingulate gyrus 32 0 14 46 5.95 pars opercularis
R anterior cingulate gyrus 24 8 12 46 5.71
L inferior frontal gyrus, pars 44 − 54 12 22 5.59 Negatively correlated
opercularis 102 L superior parietal lobule 5 − 22 − 44 60 3.21
2 R precuneus 7 10 − 66 38 6.24 94 R precentral gyrus 6 20 − 18 62 4.18
R angular 39 44 − 48 24 5.13 84 R lateral occipital cortex/cuneal 7/19 14 − 82 42 4.18
R superior parietal lobule/ 7/40 38 − 48 42 4.80 cortex
supramarginal 64 L parietal operculum cortex 40 − 54 − 38 22 4.19
3 L precuneus 7 −8 − 72 38 5.45 58 R post-central gyrus 3 34 − 34 66 3.11
L supramarginal 40/39 − 44 − 40 34 5.41 50 L angular gyrus 39 − 40 − 60 18 4.27
L superior parietal lobule 7 − 26 − 56 46 5.2 31 L superior parietal lobule/lateral 7 34 − 58 62 4.05
4 L DLPFC–middle frontal gyrus 9 − 38 34 26 5.38 occipital cortex
5 R DLPFC–middle frontal gyrus 9 34 38 38 6.24 25 L superior frontal gyrus, 6 − 16 −6 74 3.05
6 R cerebellum, anterior 26 − 64 − 28 4.96 premotor cortex
7 L lateral occipital cortex, inferior 18 − 28 − 88 4 5.47
8 R precentral gyrus 44/6 54 6 12 5.52 ROI analysis
9 R lateral occipital cortex, 18 28 − 84 2 5.45 Positively correlated
superior 112 Bilateral paracingulate 10 0 54 6 3.45
10 R insula/frontal operculum 13 38 18 6 5.45 R paracingulate 10 −2 50 4 2.27
11 R lateral occipital cortex, 19 30 − 72 28 4.98 L frontal pole 9 2 58 10 2.92
superior
12 R lateral occipital cortex, inferior 18 38 − 82 − 8 5.02 Negatively correlated
13 L thalamus − 16 −4 0 4.25 106 R post–central gyrus/spl 5/7 32 − 38 70 2.30
14 R white matter, adjacent to 24 − 28 30 4.53
supramarginal Regions from the OSPAN N Arithmetic contrast positively and negatively correlated with
15 L posterior cingulate cortex 23 − 4 − 28 26 4.34 correct letter responses. Top half is from the whole brain, uncorrected voxel threshold
analysis, p b 0.005, cluster threshold ≥ 20. The bottom half is the ROI analysis performed
OSPAN N Arithmetic. Z N 4.0, p = 0.05. Here we present the results of the whole brain using the regions detected in the whole brain analysis; cluster thresholded at Z N 2.3;
voxel-wise analysis for the OSPAN N Arithmetic contrast. p b 0.05.
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 783

Discussion present task demands and must retrieve the appropriate response-set
from LTM. Both tasks also require sub-vocal rehearsal; one of the
In this study, we used fMRI in conjunction with a CWMS task, the largest clusters of activation during both tasks encompassed the SMA
OSPAN, and a task resembling TNWM tasks, equation verification. Our and precentral gyrus, areas believed to be responsible for verbal
aims were to 1) examine how brain activity differs between CWMS production and which may also contribute to executive functioning
and TNWM tasks; 2); determine whether the hippocampus, typically (Koelsch et al., 2009; Ridderinkhof et al., 2004).
associated with the retrieval and formation of LTMs, is significantly Parietal regions are reported to be active during a number of
more active during the encoding and maintenance phase of a CWMS different cognitive processes. A review of the precuneus by Cavanna and
task than during a TNWM task; and 3) explore how activity during Trimble (2006) indicated that this region is involved in a diverse array of
OSPAN encoding and maintenance may be correlated with later highly integrated functions, consistent with its role as an associative
correct and incorrect recall. region and its high level of cortico-cortical connectivity. Wager and
LTM access during CWMS tasks is supported by the embedded Smith (2003) indicate that BA7 is the most significantly activated region
processes (Cowan, 1988, 1999) and active maintenance (Unsworth during the executive processes of updating, ordering, and manipulation.
and Engle, 2007b) models of WM and the limited capacity of the focus During equation verification, which involved two arithmetic calcula-
of attention (Cowan, 2001). For example, during the OSPAN task tions in our paradigm, these processes were necessary as participants
participants may be forced to access LTM in order to store and retrieve decided how to approach the equations, manipulated the calculations
to-be-remembered letters (task relevant), as they are also asked to within the equations, and updated portions of the equation with the
perform equation verification (task irrelevant) while trying to serially proper solutions. Fehr et al. (2007) have also reported precuneus
encode the to-be-remembered letters. That is to say, equation activation during both simple and complex arithmetic tasks, further
verification is a complex process which, when combined with the supporting the idea that the precuneus is highly responsible for
increasing list length, will likely occupy the limited capacity of the updating and manipulating information.
focus of attention. As such, the to-be-remembered letter sequence IPL activation during these two tasks likely relates to a few
may be displaced from the focus of attention and stored in LTM. functions specifically associated with the IPL. Firstly, the supramar-
Consequently, this memory trace likely resides in an activated state ginal gyrus contributes to reading regardless of task demands
(above threshold) and, depending on the individual's level of overall (Stoeckel et al., 2009). Arithmetic computation is a verbally based
attentional control, can be retrieved back into the focus as needed skill and should therefore have elicited activation in regions
(Oberauer, 2002). associated with reading since the equations were visually presented.
Recently, neuroimaging studies have begun to show hippocampal Secondly, activity in the IPL is thought to increase as responses are
activity during TNWM tasks. Typically, this activity has been shown to made under uncertainty (Vickery and Jiang, 2009), possibly indicating
increase along with load. Unsworth and Engle (2007a) indicate that it may additionally contribute to the error monitoring/checking roles
simple WM span tasks, such as TNWM tasks, may access LTM as long typically assigned to the ACC and other frontal regions. During both
as a supra-span load is presented; a supra-span load would cause tasks, participants had a limited amount of time, 4 s, to respond to a
some items to be displaced from the focus, possibly into LTM. Even complex arithmetic equation. Therefore, it is likely that under various
though there is strong evidence supporting the latter, we have chosen instances they were unsure of their responses by the time they were
to examine the possible access of LTM associated regions during a required to respond, at times resulting in incorrect responses. Lastly,
CWMS task because these tasks require an added degree of executive the IPL, specifically the angular gyrus, is recruited during arithmetic
control not required for the performance of simple WM span tasks, fact retrieval (Dehaene and Cohen, 1997; Wood et al., 2008). Another
they have been shown to exhibit a higher level of ecological validity possibility for IPL activation is that it forms part of the structural core
than typical WM tasks, and because the neuroimaging literature on of the human brain (Hagmann et al., 2008). Therefore it likely serves a
popular CWMS tasks is scarce and deserves further exploration. facilitative, associative role during complex cognitive operations such
as the ones performed in these tasks.
Shared cortical resources between CWMS and TNWM tasks in this study
Differences in cortical activation between CWMS and TNWM tasks
Cortical areas of activation during the OSPAN and Arithmetic
conditions were similar (Fig. 3), with the OSPAN exhibiting more We contrasted the OSPAN and Arithmetic blocks to examine how
robust activation in regions common to both tasks. This similar encoding, maintenance, and the possible storage and retrieval of
pattern would be expected given both tasks require the verification of information to and from LTM manifest neurally during a CWMS task
a complex equation, a cognitively demanding task; it also consistent as compared to a TNWM task. The contrast revealed that, as
with the dual task literature which states that dual tasks will have expected, differences in neural recruitment between CWMS and
similar but more intense and dispersed regions of activity (Adcock et equation verification were significant (Fig. 5). However, both tasks
al., 2000; Bunge et al., 2000). Cortical regions seen to be active in both appear to recruit the same, or similar, cortical regions since
conditions and commonly associated with WM are DLPFC (BA 9 and differences were mainly evident in regions common to both tasks.
46), inferior frontal gyrus (IFG; BA 44), middle frontal gyrus (MFG; BA This supports the idea that WM processes, regardless of their
6 and 9), precuneus cortex (BA 7), SPL (BA 7), IPL (BA 39 and 40), and complexity, stem from a common network (e.g., Anurova et al.,
ACC (BA 24/32). Additionally, one of the largest clusters of activation 2003; Cabeza and Nyberg, 2000; Linden, 2007). It is also possible
for both tasks encompassed the SMA and precentral gyrus. that some of the regions within this network may assume further
DLPFC has typically been associated with overall executive roles, e.g., aiding in the storage and retrieval of items to and from
functioning (Wager and Smith, 2003), but more recent evidence LTM as task complexity and interference increase.
suggests that it is specifically involved in focusing attention on task During the OSPAN participants require a higher degree of
relevant info in LTM (Abe et al., 2007). It is also reported to be executive control in order to switch between serial encoding and
involved in information source-monitoring (Wood et al., 2008). Kong the additional processing task of equation verification, while also
et al. (2005) indicated that complex arithmetic procedures are keeping track of an increasing number of task relevant items in WM.
supported by bilateral MFG and ACC activation. It has also been Participants also assumed WM loads for the to-be-remembered items
suggested that the ACC may play a greater role in executive that surpassed the average limit of WM capacity, currently thought to
functioning than the DLPFC (Kaneda and Osaka, 2008). Common to average around four items (Cowan, 2001). Under these conditions, it
both tasks in our paradigm is that participants must be aware of is evident that participants would benefit from increased WM
784 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

capacity. The DLPFC is thought to boost visuospatial WM capacity pal formation activity is negatively correlated to that of DLPFC as WM
through top-down excitation of intraparietal sulcal circuits (Edin et load increases and therefore conclude that hippocampal formation
al., 2009). This function of the DLPFC is thought not to be limited to activity is not necessary for WM. Additionally, they show that
visuospatial WM and likely occurs for a variety of cognitive functions. hippocampal formation activity deficiency in WM performance in
DLPFC activity has also been shown to be predictive of LTM formation schizophrenia patients may in part be due to functional decoupling
through the binding of related items residing in WM (Blumenfeld and between the hippocampus and DLFPC as WM load increases. Critical
Ranganth, 2006). Bunge et al. (2001) found that DLPFC activity in the to the interpretation of these results is the fact that a 2-back n-back
MFG, among other regions typically associated with WM, was task was used rather than a CWMS task. Even though the n-back task
significantly correlated with resolving task interference. More recent is designed as a cognitively demanding task requiring high levels of
studies (e.g., Blumenfeld and Ranganath, 2007) have demonstrated attentional control and the updating of mental representations, it may
that even more critical to the resolution of interference is the VLPFC. more accurately represent an individual's immediate memory
VLPFC is thought to aid in the controlled selection of items, and capacity rather than their WM capacity (Conway et al., 2005).
similarly, is thought to aid the formation of LTMs through controlled For our study, hippocampal voxel-wise ROI analysis indicated
selection. These functions are critical during the OSPAN as the posterior bilateral hippocampal activity during the OSPAN. We also
formation of proper LTMs becomes more crucial as set sizes increase. found similar bilateral hippocampal activity during the Arithmetic
In the previous section we outlined the possible role of parietal task. These findings indicate that Arithmetic and CWMS tasks both
regions during both the OSPAN and Arithmetic tasks. We indicated recruit the hippocampus for WM performance and at first glance
that parietal regions are mainly involved in updating and manipulat- suggest that hippocampus may be recruited for immediate memory
ing the information currently in the focus of attention. A review by functioning. However, studies of hippocampectomized patients have
Koenigs et al. (2009) indicated that the SPL in particular may be either shown no deficits, or deficits only at supra-span levels, on
responsible for these operations. The increased involvement of the simple WM span tasks (e.g., Milner, 1972; Owen et al., 1996).
SPL during CWMS provides further evidence for this role of SPL, as Therefore, it is possible that some arithmetic operations recruit
participants must update and possibly manipulate the retrieved to- additional processes not required during sub-span simple WM span or
be-remembered letter sets as more letters are presented. Successful TNWM tasks. There are indications that during more complex
updating will rely on the accuracy of the order in which the letters are arithmetic problems, such as the large number multiplication or
kept in WM and retrieved from LTM. If participants make a division problems found in our study, back up strategies are used to
discrepancy judgment and believe the ordering of the retrieved retrieve solutions to problems that are similar to the ones in question
items is incorrect, the SPL may be recruited to rearrange the set based (Jost et al., 2009; Smith-Chant and LeFevre, 2003). For example, 7 × 9
on an alternative representation. can be solved by retrieving the solution to 7 × 10 and subtracting 7. For
other problems, various back-up strategies may be used to retrieve or
Hippocampal recruitment in CWMS and TNWM tasks calculate information, and are cycled through while trying to identify
the correct answer. Furthermore, even though it has traditionally
We were particularly interested in the involvement of the been argued that the hippocampus is not necessary for semantic-fact
hippocampus during CWMS, since the hippocampus has been retrieval from LTM, such as that required during arithmetic, recent
traditionally associated with the formation and retrieval of long studies suggest its involvement (e.g., Burianova et al., 2010; Hoscheidt
term memories (e.g., Davachi et al., 2006; Eichenbaum et al., 2007). et al., 2010; Ryan et al., 2008; Whatmough and Chertkow, 2007).
Recent WM neuroimaging studies haves indicated the involvement of Consequently, hippocampal activity during complex arithmetic
the hippocampus during WM encoding, maintenance and, retrieval operations may signal a role for the hippocampus as part of a
(e.g., Axmacher et al., 2009a, 2009b; Fletcher et al., 2003; Öztekin et temporary WM storage buffer for items retrieved from LTM not
al., 2009; Schon et al., 2009). Using intracranial EEG, Axmacher et al. currently in the focus of attention. Axmacher et al. (2009a, 2009b)
(2007) demonstrated reduced hippocampus/MTL gamma power recently arrived at a similar conclusion and this idea is akin to Cowan's
during maintenance of a single item, but increased power during embedded processes model where items retrieved from LTM not
maintenance of multiple items; fMRI data concurred with their currently in the focus reside in an activated, easily accessible state
intracranial EEG recordings. Later, Axmacher et al. (2009a, 2009b) within LTM.
also found MTL activation under low load visual WM conditions. Support for the hypothesis that the hippocampus is involved in the
However, one must take into account that even under “low load” maintenance of items during a CWMS task to a greater extent than
conditions, complex visual stimuli, such as the faces used in the latter during a TNWM task came from our OSPAN N Arithmetic contrast. This
study, likely incur a high WM/attentional load due to the numerous yielded a cluster of significantly greater activation in the right
features present in such stimuli. Öztekin et al., 2009 used an item posterior hippocampus (Fig. 4). Mean peristimulus plots (Fig. 6) for
recognition and judgment of recency task to show that the the hippocampal ROIs indicate an initial bilateral peak after task
hippocampus, along with DLPFC and IFG, collectively support WM onsets, possibly indicating a retrieval of task specific instructional sets,
retrieval. Additionally, their data support the distinction between and a decrease in activity below baseline for the remainder of the
items maintained within and outside of the focus of attention, and tasks (30 s; 20 TRs). Persitimulus plots for voxels with intensity values
argue that the mechanisms responsible for these distinctions are in the upper 10% show above baseline activity levels for bilateral
closely intertwined. posterior hippocampi during OSPAN and Arithmetic (Fig. 6), with
On the contrary, Zarahn et al. (2005) have argued that hippocam- significantly greater activation in left posterior hippocampus versus
pus activity is independent of WM load, or functioning, for familiar right (Table 2).
stimuli. They have argued this point based on a hippocampal activity These results indicate that the left hippocampus plays at least an
pattern that follows an inverted-U in relation to WM load. However, equivalent or greater role than the right during WM encoding,
an inverted-U shape is often shown for different regions of the brain in maintenance, and updating. The difference between peak activity
relation to WM load and is considered a standard aspect of proper occurring in posterior hippocampi and mean activity for the whole
WM functioning. For example, the inverted-U characteristic of brain hippocampi suggests a specialization of WM functioning within the
function in relation to WM load is thought to be shifted in posterior hippocampus. The overall decrease in mean bilateral
schizophrenia and is believed to explain their relatively poorer hippocampal activation below baseline for both tasks may suggests
performance on measures of WM (e.g., Callicott et al., 2003). In a possible resting state functions for anterior regions of the hippocam-
different vain, Meyer-Lindenberg et al. (2005) argue that hippocam- pus. Recent studies of the brain's resting state networks have
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 785

implicated the hippocampus as part of some of these networks, focus of attention and would help to reduce the information load
specifically the default mode network (Buckner et al., 2008; parietal regions may have to engage or manipulate, thereby reducing
Fransson and Marrelec, 2008; Frings et al., 2009). It is thought that parietal activation.
the hippocampi and surrounding medial temporal structures play a A similar view of frontal and parietal interaction was also recently
key role in providing associational and relational information from espoused by Edin et al. (2009). Through a computational model, verified
memory to aid in mentalizing (Addis and Schacter, 2008; Buckner by analysis of actual fMRI data, they demonstrated that DLPFC boosts the
et al., 2008.) visuospatial WM capacity of parietal regions through increased
Possible direct evidence for the idea that the hippocampus may be functional coupling between the regions. The middle and superior
associated with LTM access during WM encoding and maintenance is frontal gyri, which are thought to play a regulatory role in relation to the
seen in persitimulus plots of the left hippocampus during OSPAN DLPFC, also evidenced activation consistent with a boosting function.
retrieval (time points 21–30). Mean hippocampus signal maintains an Taken together, our findings and those of Edin et al. (2009) support the
intensity similar to the signal during maintenance and encoding, embedded processes and active maintenance models of WM. They
while voxels in the upper 10% range decrease for the right indicate that frontal and cingulate regions play a role in attentional
hippocampus, and initially decrease then increase for the left regulation/control, or enhancing WM capacity in parietal regions, while
hippocampus. Therefore, it is plausible that the developing signal parietal regions play a role in storing and manipulating the contents in
within the hippocampus indicates increasing access to LTM for items the focus of attention. Therefore, an individual's WM capacity can be
displaced from the focus. Alternatively, this may indicate that the said to be limited by the degree of interaction, or coupling, between
hippocampus is cycling through items in immediate memory, but the frontal/cingulate and parietal regions.
intact immediate memory performance in hippocampectomized To further constrain our exploratory analysis we performed a
patients suggests otherwise. cluster-thresholded ROI analysis on these regions. This analysis
In general, these results provide support for the hippocampus' role in yielded a bilateral cluster overlapping part of the ACC and the frontal
WM maintenance during a CWMS task above and beyond that in TNWM pole was positively correlated with correct recall, while a cluster
tasks. Given the coupling between hippocampal activity and LTM overlapping the right post central gyrus and SPL was negatively
functioning, our findings are also suggestive of LTM functioning during correlated with correct recall. Such findings are very promising
WM maintenance of a complex WM span task. Pattern and intensity because they suggest that an interaction between two of the regions
differences between the hippocampus as a whole and posterior regions believed to be most responsible for attentional control (ACC) and
suggest a specialization of function, such that posterior regions are updating, ordering, and manipulation processes (SPL) may be most
directly involved in WM functioning. Lastly, the hippocampus may act as linked to the appropriate maintenance of information resulting in
part of a temporary storage buffer by helping to hold information correct recall. The promising nature of our exploratory results warrant
outside of the focus of attention in an activated state. further investigation into the nature of successful encoding and
maintenance during CWMS paradigms.
Further support for the active maintenance and embedded processes
models of WM
Conclusion
The fronto-parietal network is often mentioned in the context of
Our fMRI results demonstrated that as expected, the OSPAN
WM (e.g., Bledowski et al., 2009; Champod and Petrides, 2007; Colom
yielded greater activation than Arithmetic in regions typically
et al., 2007; Vincent et al., 2008). Regressing the number of correct
associated with WM. Of greater consequence, the OSPAN recruited
letter responses on the OSPAN N Arithmetic contrast gave us a further
bilateral posterior hippocampi to a greater extent than Arithmetic
indication of the roles that frontal and parietal regions, along with the
while overall hippocampal activation decreased below baseline levels
ACC, may play during WM encoding and storage of to-be-remem-
of activity during performance of both tasks. This suggests posterior
bered items and indicated support for the embedded processes
hippocampal specialization during WM performance and further
(Cowan, 1999, 2005) and active maintenance models (Unsworth and
supplements the evidence presenting the hippocampal formation as
Engle, 2007b) of WM. Together, these models propose that frontal and
part of the brain's default mode network. A rise in intensity signals
parietal regions are highly interconnected during WM processes and
from the left hippocampus during retrieval hint at the idea that LTM
the effectiveness of their interactions dictates WM capacity limits.
access occurs during encoding and maintenance of items presented in
Specifically, they suggest that the PFC is involved in monitoring and
CWMS tasks. Such evidence calls for more in depth examination of
directing the attentional resources required during WM functioning
CWMS tasks through neuroimaging, as they may provide crucial
and that IFG functioning in particular is critical to the retrieval of
insight into the role of the hippocampus during WM functioning. High
information in the face of interference (Jonides and Nee, 2006).
resolution neuroimaging studies have already alluded to the idea of
Information storage and processing tasks, however, are not relegated
encoding and retrieval specificity between the dentate gyrus, cornu
to PFC regions, but rather information currently in the focus of
ammonis regions, and the subiculum (Eldridge et al., 2005; Preston
attention is stored and manipulated in parietal regions. Additionally, it
et al., 2010; Zeineh et al., 2003). Lastly, the linear increase in activity
is suggested that the ACC acts a secondary monitor by biasing the
during the initial blocks of each run also warrants further exploration.
degree of executive control exerted by the IFG.
We are not aware of any cognitive or physiological findings that may
Our preliminary, exploratory results indicated that activity in
explain these trends, but can only speculate that they may be due to
frontal/cingulate regions, including the DLPFC, IFG (pars opercularis),
some form of habituation or task detection function exhibited by the
and the ACC, during WM encoding and maintenance, tended to
hippocampi.
positively correlate with number of correct letter identification
responses, and that areas mainly in the parietal lobe, including the
SPL, IPL, cuneal cortex, parietal operculum, and post-central gyrus, Acknowledgments
tended to negatively correlate with correct responses. If frontal and
cingulate regions are involved in regulating the influence of We would like to thank the University of Georgia's BioImaging
interfering information it is probable they are working at a heightened Research Center for providing the necessary imaging acquisition hours.
level in individuals who are successfully filtering this information We would also like to thank Kim Mason and Devin Smith for their
compared to those who are not. An effective, dual fronto-cingulate assistance in acquiring MRI and behavioral data, and Brett Clementz,
filter would then limit the amount of information entering into the Jennifer McDowell, Chris Rorden, Dean Sabatinelli, and Nathan
786 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787

Yansak for their excellent advice throughout the writing of this Courtney, S.M., Petit, L., Haxby, J.V., Ungerleider, L.G., 1998. The role of prefrontal cortex
in working memory: examining the contents of consciousness. Philos. Trans. R. Soc.
manuscript. Lond. B Biol. Sci. 353, 1819–1828.
Cowan, N., 1988. Evolving conceptions of memory storage, selective attention, and their
mutual constraints within the human information-processing system. Psychol.
References Bull. 104, 163–191.
Cowan, N., 1999. An embedded-processes model of working memory. In: Miyake, A.,
Abe, M., Hanakawa, T., Takayama, Y., Kuroki, C., Ogawa, S., Fukuyama, H., 2007. Shah, P. (Eds.), Models of Working Memory. Cambridge University Press,
Functional coupling of human prefrontal and premotor areas during cognitive Cambridge, pp. 62–101.
manipulation. J. Neurosci. 27, 3429–3438. Cowan, N., 2001. The magical number 4 in short-term memory: a reconsideration of
Adcock, R.A., Constable, R.T., Gore, J.C., Goldman-Rakic, P.S., 2000. Functional mental storage capacity. Behav. Brain Sci. 24, 87–185.
neuroanatomy of executive processes involved in dual-task performance. Proc. Cowan, N., 2005. Working Memory Capacity. Psychology Press, New York.
Natl Acad. Sci. USA 97, 3567–3572. Daneman, M., Carpenter, P.A., 1980. Individual-differences in working memory and
Addis, D.R., Schacter, D.L., 2008. Constructive episodic simulation: temporal distance reading. J. Verbal Learn. Verbal Behav. 19, 450–466.
and detail of past and future events modulate hippocampal engagement. Davachi, L., Maril, A., Wagner, A.D., 2006. When keeping in mind supports later bringing
Hippocampus 18, 227–237. to mind: neural markers of phonological rehearsal predict subsequent remember-
Anurova, I., Artchakov, D., Korvenoja, A., Ilmoniemi, R., Aronen, H.J., Carlson, S., 2003. ing. J. Cogn. Neurosci. 13, 1059–1070.
Differences between auditory evoked responses recorded during spatial and Dehaene, S., Cohen, L., 1997. Cerebral pathways for calculation: Double dissociation
nonspatial working memory tasks. Neuroimage 20, 1181–1192. between rote verbal and quantitative knowledge of arithmetic. Cortex 33, 219–250.
Atkinson, R.C., Shiffrin, R.M., 1968. Human memory: a proposed system and its control D'Esposito, M., Postle, B.R., Jonides, J., Smith, E., 1999. The neural substrate and temporal
processes. In: Spence, K.W., Spence, J.T. (Eds.), The Psychology of Learning and dynamics of interference effects in working memory as revealed by event-related
Motivation (Volume 2). Academic Press, New York. functional MRI. Proc. Natl Acad. Sci. USA 96, 7514–7519.
Axmacher, N., Mormann, F., Fernández, G., Cohen, M., Elger, C., Fell, J., 2007. Sustained D'Esposito, M., Postle, B.R., Rypma, B., 2000. Prefrontal cortical contributions to working
neural activity patterns during working memory in the human medial temporal memory: evidence from event-related fMRI studies. Exp. Brain Res. 133, 3–11.
lobe. J. Neurosci. 27, 7807–7816. Edin, F., Klingberg, T., Johansson, P., McNab, F., Tegner, J., Compte, A., 2009. Mechanism
Axmacher, N., Elger, C.E., Fell, J., 2009a. Working memory related hippocampal for top-down control of working memory capacity. Proc. Natl Acad. Sci. USA 106,
deactivation interferes with long-term memory formation. J. Neurosci. 28, 6802–6807.
1052–1060. Eichenbaum, H., Yonelinas, A.P., Raganath, C., 2007. The medial temporal lobe and
Axmacher, N., Haupt, S., Cohen, M.X., Elger, C.E., Fell, J., 2009b. Interference of recognition memory. Annu. Rev. Neurosci. 30, 123–152.
working memory load with long-term memory formation. Eur. J. Neurosci. 29, Eldridge, L.L., Engel, S.A., Zeineh, M.M., Bookheimer, S.Y., Knowlton, B.J., 2005. A
1501–1513. dissociation of encoding and retrieval processes in the human hippocampus.
Baddeley, A.D., Hitch, G.J., 1974. Working memory. In: Bower, G.A. (Ed.), Recent Advances J. Neurosci. 25, 3280–3286.
in Learning and Motivation (Vol. 8). Academic Press, New York, pp. 47–89. Engle, R.W., Kane, M.J., Tuholski, S.W., 1999a. Individual differences in working memory
Bernstein, P.S., Scheffers, M.K., Coles, M.G., 1995. “Where did I go wrong?” A capacity and what they tell us about controlled attention, general fluid intelligence,
psychophysiological analysis of error detection. J. Exp. Psychol. Hum. Percept. and functions of the prefrontal cortex. In: Miyake, A., Shah, P. (Eds.), Models of
Perform. 21, 1312–1322. Working Memory. Cambridge University Press, Cambridge, pp. 103–134.
Bledowski, C., Rahm, B., Rowe, J., 2009. What “works” in working memory? Separate Engle, R.W., Tuholski, S.W., Laughlin, J.E., Conway, A.R., 1999b. Working memory, short-
systems for selection and updating of critical information. J. Neurosci. 29, term memory, and general fluid intelligence: a latent-variable approach. J. Exp.
13735–13741. Psychol. Gen. 128, 309–331.
Blumenfeld, R.S., Ranganath, C., 2007. Prefrontal cortex and long-term memory Fehr, T., Code, C., Herrmann, M., 2007. Common brain regions underlying different arithmetic
encoding: an integrative review of findings from neuropsychology and neuroima- operations as revealed by conjunct fMRI-BOLD activation. Brain Res. 1172, 93–102.
ging. Neuroscientist 213, 280–291. First, M.B., Spitzer, R.L., Gibbon, M., Williams, J.B., 1997. Sctructured Clinical Interview
Blumenfeld, R.S., Ranganth, C., 2006. Dorsolateral prefrontal cortex promotes long-term for DSM-IV Axis I Disorders (SCID-I), Clinical Version. American Psychiatric
memory formation through its role in working memory organization. J. Neurosci. Publishing, Inc., Arlington, VA.
26, 916–925. Fletcher, P.C., Stephenson, C.M., Carpenter, T.A., Donovan, T., Bullmore, E., 2003.
Bor, D., Duncan, J., Lee, A.C., Parr, A., Owen, A.M., 2006. Frontal lobe involvement in Regional brain activations predicting subsequent memory success: an event-
spatial span: converging studies of normal and impaired function. Neuropsycho- related fmri study of the influence of encoding tasks. Cortex 39, 1009–1026.
logia 44, 229–237. Fransson, P., Marrelec, G., 2008. The Precuneus/posterior cingulate cortex plays a
Botvinick, M.M., Braver, T.S., Barch, D.M., Carter, C.S., Cohen, J.D., 2001. Conflict pivotal role in the default mode network: Evidence from a partial correlation
monitoring and cognitive control. Psychol. Rev. 108, 624–652. network analysis. Neuroimage 42, 1178–1184.
Buckner, R.L., Andrews-Hanna, J.R., Schacter, D., 2008. The brain's default network. Frings, L., Schulze-Bonhage, A., Spreer, J., Wagner, K., 2009. Remote effects of
Anatomy, function, and relevance to disease. Ann. NY Acad. Sci. 1124, 1–38. hippocampal damage on default network connectivity in the human brain. J.
Bunge, S.A., Klingberg, T., Jacobsen, R.B., Gabrieli, J.D., 2000. A resource model of the Neurol. 2556, 2021–2029.
neural basis of executive working memory. Proc. Natl Acad. Sci. USA 97, Gevins, A.S., Bressler, S.L., Cutillo, B.A., Illes, J., Miller, J.C., Stern, J., Rex, H.R., 1990. Effects
3573–3578. of prolonged mental work on functional brain topography. Electroencephalogr.
Bunge, S.A., Ochsner, K.N., Desmond, J.E., Glover, G.H., Gabrieli, J.D., 2001. Prefrontal Clin. Neurophysiol. 76, 339–350.
regions involved in keeping information in and out of mind. Brain 124, 2074–2086. Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C.J., Wedeen, V.J., Sporns, O.,
Bunting, M., 2006. The role of processing difficulty in the predictive utility of working 2008. Mapping the structural core of human cerebral cortex. PLoS Biol. 6, e159.
memory span. Psychon. B Rev. 13, 998–1004. Hamidi, M., Tononi, G., Postle, B.R., 2008. Evaluating frontal and parietal contributions
Burianova, H., McIntosh, A.R., Grady, C.L., 2010. A common functional brain network for to spatial working memory with repetitive transcranial magnetic stimulation. Brain
autobiographical, episodic, and semantic memory retrieval. Neuroimage 49, Res. 1230, 202–210.
865–874. Hoscheidt, S.M., Nadel, L., Payne, J., Ryan, L., 2010. Hippocampal activation during
Cabeza, R., Nyberg, L., 2000. Imaging cognition II: an empirical review of 275 PET and retrieval of spatial context from episodic and semantic memory. Behav. Brain Res.
fMRI studies. J. Cogn. Neurosci. 12, 1–47. 212, 121–132.
Callicott, J.H., Mattay, V.S., Verchinski, B.A., Marenco, S., Egan, M.F., Weinberger, D.R., Jenkinson, M., Bannister, P., Brady, M., Smith, S., 2002. Improved optimization for the
2003. Complexity of prefrontal cortical dysfunction in schizophrenia: more than up robust and accurate linear registration and motion correction of brain images.
or down. Am. J. Psychiatry 160, 2209–2215. Neuroimage 17, 825–841.
Cavanna, A.E., Trimble, M.R., 2006. The precuneus: a review of its functional anatomy Jonides, J., Nee, D.E., 2006. Brain mechanisms of proactive interference in working
and behavioural correlates. Brain 129, 564–583. memory. Neuroscience 139, 181–193.
Champod, A.S., Petrides, M., 2007. Dissociable roles of the posterior parietal and the Jonides, J., Marshuetz, C., Smith, E.E., Reuter-Lorenz, P.A., Koeppe, R.A., Hartley, A., 2000.
prefrontal cortex in manipulation and monitoring processes. Proc. Natl Acad. Sci. Age differences in behavior and PET activation reveal differences in interference
USA 104, 14837–14842. resolution in verbal working memory. J. Cogn. Neurosci. 12, 188–196.
Chein, J.M., Moore, A.B., Conway, A.R., 2010. Domain-general mechanisms of complex Jost, K., Khader, P., Burke, M., Bien, S., Rösler, F., 2009. Dissociating the solution process of
working memory span. Neuroimage 54, 550–559. small, large, and zero, multiplications by means of fMRI. Neuroimage 46, 308–318.
Colom, R., Jung, R.E., Haier, R.E., 2007. General intelligence and memory span: Kane, M.J., Hambrick, D.Z., Conway, A.R.A., 2005. Working memory capacity and fluid
evidence for a common neuroanatomic framework. Cogn. Neuropsychol. 24, intelligence are strongly related constructs: comment on Ackerman, Beier, and
867–878. Boyle (2005). Psychol. Bull. 131, 66.
Conway, A.R., Engle, R.W., 1996. Individual differences in working memory capacity: Kane, M.J., Conway, A.R., Miura, T.K., Colflesh, G.J., 2007a. Working memory, attention
more evidence for a general capacity theory. Memory 4, 577–590. control, and the N-back task: a question of construct validity. J. Exp. Psychol. Learn.
Conway, A.R., Kane, M.J., Engle, R.W., 2003. Working memory capacity and its relation Mem. Cogn. 33, 615–622.
to general intelligence. Trends Cogn. Sci. 7, 547–552. Kane, M.J., Conway, A.R., Hambrick, D.Z., Engle, R., 2007b. Variation in working memory
Conway, A.R., Kane, M.J., Bunting, M.F., Hambrick, D.Z., Wilhelm, O., Engle, R.W., 2005. capacity as variation in executive attention and control. In: Conway, A.R.A., Jarrold,
Working memory span tasks: a methodological review and user's guide. Psychon. C., Kane, M.J., Miyake, A., Towse, J. (Eds.), Variation in Working Memory. Oxford
Bull. Rev. 12, 769–786. University Press, New York, pp. 21–48.
Corbetta, M., Shulman, G.L., 2002. Control of goal-directed and stimulus-driven Kaneda, M., Osaka, N., 2008. Role of anterior cingulate cortex during semantic coding in
attention in the brain. Nat. Rev. Neurosci. 3, 201–215. verbal working memory. Neurosci. Lett. 436, 57–61.
 C.C. Faraco et al. / NeuroImage 55 (2011) 773–787 787

Klein, K., Fiss, W.H., 1999. The reliability and stability of the Turner and Engle working Ryan, L., Cox, C., Hayes, S.M., Nadel, L., 2008. Hippocampal activation during episodic
memory task. Behav. Res. Methods Instrum. Comput. 31, 429–432. and semantic memory retrieval: comparing category production and category cued
Koelsch, S., Schulze, K., Sammler, D., Fritz, T., Müller, K., Gruber, O., 2009. Functional recall. Neuropsychologia 46, 2109–2121.
architecture of verbal and tonal working memory: an fMRI study. Hum. Brain Schon, K., Quiroz, Y.T., Hasselmo, M.E., Stern, C.E., 2009. Greater working memory load
Mapp. 30, 859–873. results in greater medail temporal activity at retrieval. Cereb. Cortex 19, 2561–2571.
Koenigs, M., Barbey, A.K., Postle, B.R., Grafman, J., 2009. Superior parietal cortex is Scoville, W.B., Milner, B., 1957. Loss of recent memory after bilateral hippocampal
critical for the manipulation of information in working memory. J. Neurosci. 29, lesions. J. Neurol. Neurosurg. Psychiatr. 20, 11–21.
14980–14986. Smith, S.M., 2002. Fast robust automated brain extraction. Hum. Brain Mapp. 17,
Kondo, H., Morishita, M., Osaka, N., Osaka, M., Fukuyama, H., Shibasaki, H., 2004. 143–155.
Functional roles of the cingulo-frontal network in performance on working Smith, E.E., Jonides, J., 1999. Storage and executive processes in the frontal lobes.
memory. Neuroimage 21, 2–14. Science (New York, N.Y.) 283, 1657–1661.
Kong, J., Wang, C., Kwong, K., Vangel, M., Chua, E., Gollub, R., 2005. The neural substrate Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E., Johansen-
of arithmetic operations and procedure complexity. Cogn. Brain Res. 22, 397–405. Berg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R., Saunders, J.,
Lezak, M.D., Howieson, D.B., Loring, D.W., 2004. Neurospychological Assessment, Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M., 2004. Advances in
Fourth Edition. Oxford Univeristy Press, New York, pp. 359–363. functional and structural MR image analysis and implementation as FSL. Neuro-
Linden, D.E., 2007. The working memory networks of the human brain. Neuroscientist image 23 (Suppl 1), S208–S219.
13, 257–267. Smith-Chant, B.L., LeFevre, J., 2003. Doing as they are told and telling it like it is: self-
MacDonald, A.W., Cohen, J.D., Stenger, V.A., Carter, C.S., 2000. Dissociating the role of reports in mental arithmetic. Mem. Cogn. 31, 516–528.
the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Srimal, R., Curtis, C.E., 2008. Persistent neural activity during the maintenance of spatial
Science 288, 1835–1838. position in working memory. Neuroimage 39, 455–468.
McCabe, D.P., 2008. The role of covert retrieval in working memory span tasks: Sternberg, S., 1966. High speed scanning in human memory. Science 153, 652–654.
evidence from delayed recall tests. J. Mem. Lang. 58, 480–494. Stoeckel, C., Gough, P.M., Watkins, K.E., Devlin, J.T., 2009. Supramarginal gyrus
Meyer-Lindenberg, A.S., Olsen, R.K., Kohn, P.D., Brown, T., Egan, M.F., Weinberger, D.R., involvement in visual word recognition. Cortex 45, 1091–1096.
Berman, K.F., 2005. Regionally specific disturbance of dorsolateral prefrontal- Toepper, M., Gebhardt, H., Beblo, T., Thomas, C., Driessen, M., Bischoff, M., Blecker, C.R.,
hippocampal functional connectivity in schizophrenia. Arch. Gen. Psychiatry 62, Vaitl, D., Sammer, G., 2010. Functional correlates of distractor suppression during
379–386. spatial working memory encoding. Neuroscience 165, 1244–1253.
Miller, E.K., 2000. The neural basis of top-down control of visual attention in the Turner, M.L., Engle, R.W., 1989. Is working memory capacity task dependent. J. Mem.
prefrontal cortex. In: Monsell, S., Driver, J. (Eds.), Attention and Performance. Lang. 28, 127–154.
Massachusetts Institute of Technology Press, Cambridge, MA, pp. 511–534. Unsworth, N., 2010. On the division of working memory and long-term memory and
Miller, E.K., Cohen, J.D., 2001. An integrative theory of prefrontal cortex function. Annu. their relation to intelligence: a latent variable approach. Acta Psychol. 134, 16–28.
Rev. Neurosci. 24, 167–202. Unsworth, N., Engle, R.W., 2006. Simple and complex memory spans and their relation
Milner, B., 1972. Disorders of learning and memory after temporal lobe lesions in man. to fluid abilities: evidence from list-length effects. J. Mem. Lang. 54, 68–80.
Clin. Neurosurg. 19, 421–446. Unsworth, N., Engle, R.W., 2007a. On the division of short-term and working memory:
Oberauer, K., 2002. Access to information in working memory: exploring the focus of an examination of simple and complex span and their relation to higher order
attention. J. Exp. Psychol. 28, 411–421. abilities. Psychol. Bull. 133, 1038–1066.
Osaka, M., Osaka, N., Kondo, H., Morisihita, M., Fukiyama, H., Aso, T., Shibasaki, H., 2003. Unsworth, N., Engle, R.W., 2007b. The nature of individual differences in working
The neural basis of individual differences in working memory capacity: An fMRI memory capacity: active maintenance in primary memory and controlled search
study. NeuroImage 18, 789–797. from secondary memory. Psychol. Rev. 114, 104–132.
Owen, A.M., Morris, R.G., Sahakian, B.J., Polkey, C.E., Robbins, T.W., 1996. Double Unsworth, N., Heitz, R.R., Schrock, J.C., Engle, R.W., 2005. An automated version of the
dissociations of memory and executive functions in working memory tasks operation span task. Behav. Res. Methods 37, 498–505.
following frontal lobe excisions, temporal lobe excisions or amygdalo-hippocam- Unsworth, N., Redick, T.S., Heitz, R.P., Broadway, J.M., Engle, R.W., 2009. Complex
pectomy in man. Brain 119, 1597–1615. working memory span tasks and higher-order cognition: a latent-variable analysis
Owen, A.M., McMillan, K.M., Laird, A.R., Bullmore, E., 2005. N-back working memory of the relationship between processing and storage. Memory 17, 635–654.
paradigm: a meta-analysis of normative functional neuroimaging studies. Hum. Van Vugt, M.K., Schulze-Bonhage, A., Litt, B., Brandt, A., Kahana, 2010. Hippocampal
Brain Mapp. 25, 46–59. gamma oscillations increase with memory load. J. Neurosci. 30, 2694–2699.
Öztekin, I., McElree, B., Staresina, B.P., Davachi, L., 2008. Working memory retrieval: Vickery, T.J., Jiang, Y.V., 2009. Inferior parietal lobule supports decision making under
contributions of the left prefrontal cortex, the left posterior parietal cortex, and the uncertainty in humans. Cereb. Cortex 19, 916–925.
hippocampus. J. Cogn. Neurosci. 21, 581–593. Vincent, J.L., Kahn, I., Snyder, A.Z., Raichle, M.E., Buckner, R.L., 2008. J. Neurophysiol. 100,
Öztekin, I., McElree, B., Staresina, B.P., Davachi, L., 2009. Working memory retrieval: 3328–3342.
contributions of the left prefrontal cortex, the left posterior parietal cortex, and the Wager, T.D., Smith, E.E., 2003. Neuroimaging studies of working memory: a meta-
hippocampus. J. Cogn. Neurosci. 21, 581–593. analysis. Cogn. Affect. Behav. Neurosci. 3, 255–274.
Patenaude, B., Smith, S., Kennedy, D., Jenkinson, M., 2007. FIRST — FMRIB's integrated Wechsler, D., 2001. Wechsler Test of Adult Reading. The Psychological Corporation, San
registration and segmentation tool. Human Brain Mapping Conference. Antonio, TX.
Postle, B.R., 2006. Working memory as an emergent property of the mind and brain. Wendelken, C., Bunge, S.A., Carter, C.S., 2008. Maintaining structured information: an
Neuroscience 139, 23–38. investigation into functions of parietal and lateral prefrontal cortices. Neuropsy-
Postle, B.R., D'Esposito, M., 1999. “What”–Then–“Where” in visual working memory: an chologia 46, 665–678.
event-related fMRI study. J. Cogn. Neurosci. 11, 585–597. Whatmough, C., Chertkow, H., 2007. rCBF of the hippocampal complex covaries with
Pratt, L.A., Brody, D.J., 2008. Depression in the United States Household Population, superior semantic memory retrieval. Behav. Brain Res. 181, 262–269.
2005–2006. U.S. Department of Health and Human Services. DHHS Publication Wood, G., Nuerk, H.-C., Moeller, K., Geppert, B., Schnitker, R., Weber, J., Wilmes, K., 2008.
(PHS) 2008–1209, Hyattsville, MD. All for one but not one for all: how multiple number representations are recruited
Preston, A.R., Bornstein, A.M., Hutchinson, B., Gaare, M.E., Glover, G.H., Wagner, A.D., in one numerical task. Brain Res. 1187, 154–166.
2010. High-resolution fmri of content-sensitive subsequent memory responses in Woolrich, W.M., Jbabdi, S., Patenaude, B., Chappell, M., Makni, S., Behrens, T., Beckmann,
human medial temporal lobs. J. Cogn. Neurosci. 22, 156–173. C., Jenkinson, M., Smith, S.M., 2009. Bayesian analysis of neuroimaging data in FSL.
PST, 2006. E-Prime (Version 1.2.1.845). Psychology Software Tools, Pittsburgh. Neuroimage 45, S173–S186.
Ridderinkhof, K.R., Ullsperger, M., Crone, E.A., Nieuwenhuis, S., 2004. The role of the Zarahn, E., Rakitin, B., Abela, D., Flynn, J., Stern, Y., 2005. Positive evidence against
medial frontal cortex in cognitive control. Science 306, 443–447. human hippocampal involvement in working memory maintenance of familiar
Rorden, C. 2007. DCM2NII (Version October 7) [Computer software]. stimuli. Cereb. Cortex 15, 303–316.
Rowe, J.B., Toni, I., Josephs, O., Frackowiak, R.S., Passingham, R.E., 2000. The prefrontal Zeineh, M.M., Engel, S.A., Thompson, P.M., Bookheimer, S.Y., 2003. Dynamics of the
cortex: response selection or maintenance within working memory? Science 288, hippocampus during encoding and retrieval of face–name pairs. Science 299,
1656–1660. 577–580.