Developmental Cognitive Neuroscience 6 (2013) 162–175

Contents lists available at ScienceDirect

Developmental Cognitive Neuroscience

journal homepage: http://www.elsevier.com/locate/dcn

Fractionating the neural correlates of individual working

memory components underlying arithmetic problem solving
skills in children
Arron W.S. Metcalfe a,∗ , Sarit Ashkenazi a,b , Miriam Rosenberg-Lee a ,
Vinod Menon a,c,d,e,∗
a
Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA 94305, USA
b
School of Education, The Hebrew University of Jerusalem – Mt. Scopus, Jerusalem 91905, Israel
c
Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, CA 94305, USA
d
Program in Neuroscience, Stanford University School of Medicine, Stanford, CA 94305, USA
e
Symbolic Systems Program, Stanford University School of Medicine, Stanford, CA 94305, USA

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

Article history: Baddeley and Hitch’s multi-component working memory (WM) model has played an endur-
Received 3 May 2013 ing and influential role in our understanding of cognitive abilities. Very little is known,
Received in revised form
however, about the neural basis of this multi-component WM model and the differential
29 September 2013
role each component plays in mediating arithmetic problem solving abilities in children.
Accepted 1 October 2013
Here, we investigate the neural basis of the central executive (CE), phonological (PL) and
visuo-spatial (VS) components of WM during a demanding mental arithmetic task in 7–9
Keywords:
Working memory year old children (N = 74). The VS component was the strongest predictor of math ability
Arithmetic cognition in children and was associated with increased arithmetic complexity-related responses in
Development left dorsolateral and right ventrolateral prefrontal cortices as well as bilateral intra-parietal
Central executive sulcus and supramarginal gyrus in posterior parietal cortex. Critically, VS, CE and PL abili-
Visuo-spatial ties were associated with largely distinct patterns of brain response. Overlap between VS
fMRI and CE components was observed in left supramarginal gyrus and no overlap was observed
Individual differences between VS and PL components. Our findings point to a central role of visuo-spatial WM
during arithmetic problem-solving in young grade-school children and highlight the use-
fulness of the multi-component Baddeley and Hitch WM model in fractionating the neural
correlates of arithmetic problem solving during development.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction in particular (Geary et al., 2004; Imbo and Vandierendonck,

2008; Raghubar et al., 2010). A long line of research in
Working memory (WM) is now increasingly consid- adults has established that complex arithmetic problem
ered to be an important factor in the development of solving tasks require active maintenance and manipulation
mathematical cognition in general (LeFevre et al., 2005; of task relevant visuo-spatial and phonological information
Raghubar et al., 2010) and arithmetic problem solving skills in WM (Hitch, 1978; LeFevre et al., 2005). In contrast, how
different components of WM contribute to the develop-
ment of arithmetic skills is poorly understood. Baddeley
∗ Corresponding authors at: Department of Psychiatry and Behavioral and Hitch’s multi-component WM model offers a powerful
Sciences, Stanford University School of Medicine, 401 Quarry Road, CA approach for studying this question. This model has played
94305, USA. Tel.: +1 650 736 0128; fax: +1 650 736 7200.
a prominent role in our understanding of the fundamen-
E-mail addresses: [email protected], [email protected]
(A.W.S. Metcalfe), [email protected] (V. Menon). tal constituents of general cognitive abilities (Baddeley,

1878-9293/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.dcn.2013.10.001
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 163

1986, 1996, 2003, 2012; Baddeley and Hitch, 1974). Only during a visuo-spatial WM task. While these studies have
a few behavioral studies have, however, used such models provided some evidence for the differential influence of
to examine how different components of WM contribute WM components in numerical problem solving abilities
to arithmetic problem solving skills in children (LeFevre in children, their precise neural representations remain
et al., 2005; Raghubar et al., 2010). Emerging behavioral unknown. Critically, no previous brain imaging studies
data suggest that individual WM components make unique have simultaneously examined the role of the three WM
contributions to the development of arithmetic problem components in children’s problem solving abilities.
solving skills (De Smedt et al., 2009; Meyer et al., 2010; Here we use an individual differences approach to frac-
Simmons et al., 2012). Whether individual WM compo- tionate the neural correlates of individual WM components
nents rely on different brain systems during arithmetic underlying arithmetic problem solving skills in children.
problem solving is currently not known. Our central goal is to test the hypothesis that the CE, VS
Baddeley and Hitch’s multicomponent WM model and PL play distinct roles in problem solving. To the extent
includes a central executive (CE) responsible for high that these core components of WM engage different brain
level control, monitoring and task switching, along with areas our findings would provide novel support for theories
subordinate phonological (PL) and visuo-spatial (VS) com- which posit that CE, VS and PL encapsulate distinct cogni-
ponents for short term storage and maintenance of verbal tive processes. The period encompassing ages 7–9 years is a
and visuo-spatial information (Baddeley, 2012). Behav- time of significant developmental change in the acquisition
ioral studies have suggested that each WM component of single digit addition skills as evidenced by shifts from
plays a different role in arithmetic problem solving. The unsophisticated to mature strategy use (Siegler et al., 1995;
CE is required for the complex arithmetic procedures of Siegler and Shrager, 1984) and by dynamic shifts in the
carrying and borrowing (Imbo et al., 2007). The phonolog- role of different WM components (De Smedt et al., 2009;
ical loop is used for active maintenance of intermediate Meyer et al., 2010), making it an ideal period for investigat-
results (Trbovich and LeFevre, 2003). And finally, the visuo- ing this question. Based on previous behavioral studies in
spatial sketchpad is involved in the solution of multi children, we predicted that VS and CE would emerge as the
digit operations (Heathcote, 1994), single digit subtraction strongest behavioral predictors of basic arithmetic skills.
problems (Lee and Kang, 2002), and has been connected Studies in patients with brain lesions have emphasized
to the representation of quantities in the format of an dissociations linking the CE to the lateral prefrontal cortex
internal mental number line representation (Rotzer et al., (Baddeley et al., 1997), the PL to the inferior frontal cor-
2009). In children specifically, CE predicts performance on tex and supramarginal gyrus (Vallar and Papagno, 2002),
single-digit addition tasks in 5–8 year old children (De and the VS to ventral occipito-temporal and dorsal poste-
Smedt et al., 2009; Meyer et al., 2010; Simmons et al., rior parietal cortex (Della Sala and Logie, 2002). A recent
2012). Geary et al. (2012) found that starting in grade 1, meta-analysis of verbal and visual WM tasks has pointed
higher CE scores predicted faster transitions from counting to a common fronto-parietal network active across differ-
(e.g., 6 + 5 = 6 + 1 + 1 + 1 +1 + 1 = 11) to a more sophisticated ent domains, with differential activation of Broca’s area for
decomposition strategy (e.g., 6 + 5 = 5 + 5 + 1 = 10 + 1 = 11). verbal tasks, premotor cortex for object and location tasks,
Simmons et al. (2012) found a trend for PL ability predict- posterior lateral PFC for maintenance of items in WM and
ing multiplication performance in 7–8 year olds and other anterior lateral PFC for maintaining task set (Rottschy et al.,
studies have demonstrated a link between PL ability and 2012).
math performance tests including word problems in grade Extrapolating from these WM studies, we hypothesized
2 (De Smedt et al., 2009; Meyer et al., 2010). VS ability that the CE, PL and VS components would engage distinct
has been found to predict magnitude judgment and num- brain systems during arithmetic problem solving. In par-
ber writing skills in 5–8 year old children (Simmons et al., allel with the hypothesized behavioral results, we further
2012), and arithmetic performance in 6–8 year old children predicted that VS and CE components would strongly mod-
(De Smedt et al., 2009; Meyer et al., 2010). ulate brain activity in distinct fronto-parietal regions, while
In contrast to behavioral studies, brain imaging stud- PL effects would be generally weak.
ies of arithmetic cognition in children have not directly
examined the role of individual WM components. For the 2. Methods
most part, the role of WM has been surmised based on
greater prefrontal cortex engagement in children (Cantlon 2.1. Participants
et al., 2009; Grabner et al., 2009; Ischebeck et al., 2007;
Rivera et al., 2005). Two recent brain imaging studies Seventy-four participants (40 female, 34 male) were
have indirectly addressed the link between WM abili- recruited from a wide range of schools in the San Francisco
ties and numerical problem solving skills. Dumontheil and Bay Area using mailings to schools and postings at libraries
Klingberg (2012) found that activity in the intra-parietal and community groups. Our theoretical focus on vari-
sulcus during a visuo-spatial WM task predicted arith- ability in working memory recruitment during complex
metic performance two years later in a sample of 6–16 arithmetic task performance dictated a tightly constrained
year old children and adolescents. Rotzer et al. (2009) focus on 7–9 year old participants (M = 7.8 years, SD = 0.7).
found that compared to typically developing controls, chil- All participants were right-handed (Oldfield, 1971) and
dren with low math ability had lower VS abilities as reported no history of psychiatric illness or medication
well as lower activity levels in the right inferior frontal use. All participants completed the Wechsler Abbreviated
gyrus, right anterior intra-parietal sulcus, and right insula Scale of Intelligence (Wechsler, 1999) and met the inclusion
164 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

criterion of full-scale IQ above 80. Informed consent was and (4) Passive fixation. For Complex addition, participants
obtained from the legal guardian of the child, and all study were presented with an equation involving two addends
protocols were approved by the Stanford University Insti- and asked to indicate via button box, if the proposed answer
tutional Review Board. was correct or incorrect (e.g., “3 + 4 = 8 ). One operand
ranged from 2 to 9, the other from 2 to 5 (tie problems,
2.2. Cognitive assessments such as “5 + 5 = 10 , were excluded), and half of the tri-
als were presented as correct equations. Incorrect answers
2.2.1. Mathematical abilities deviated by ±1 or ±2 from the correct sum (Ashcraft and
Numerical Operation and Mathematical Reasoning sub- Battaglia, 1978). Control addition was based on the same
tests of the WIAT-II (Wechsler, 2001) were used to assess task model as Complex however, one addend was always
domain-wide mathematical proficiency. Numerical Oper- 1 (e.g., 3 + 1 = 4). Our use of this task was based on pilot
ations is a pencil-and-paper test that measures number studies, which suggested that children are consistently
writing and identification, rote counting, number produc- faster on these problems compared to the Complex addi-
tion, and simple addition, subtraction, multiplication, and tion problems (Ashkenazi et al., 2012; Cho et al., 2012;
division calculations. Age-appropriate examples include Rosenberg-Lee et al., 2011). Furthermore, children show
horizontally presented problems such as 4 − 2 = ? and less strategy variability for “N + 1 problems, thus serving
vertically presented problems such as 37 + 54 = ?. Math- as ideal control problems for our study (Baroody, 1985;
ematical Reasoning is a verbal problem-solving test that Barrouillet and Lépine, 2005; Siegler, 1987). In the num-
measures counting, geometric shape identification, and ber identification task, arithmetic symbols were replaced
single and multi-step word problem-solving involving by irrelevant keyboard characters (e.g., “4 o 5 @ 7”) and
time, money, and measurement with both verbal and visual participants were asked to assess if “5” was among the
prompts. This test also includes problem solving with presented digits. Finally, in the Passive fixation task, the
whole numbers, fractions and decimals, graph interpreta- symbol “*” appeared at the center of the screen and par-
tion, mathematical pattern identification, and solution of ticipants were asked to focus their attention on it. To aid
statistics and probability problems. Age selected examples children’s performance, specific task instructions appeared
include presentation of a drawing of a dime coupled with below each problem. During the Complex and Control
verbal presentation of the question: “How many pennies addition tasks, the word “Solve” appeared below the prob-
does it take to equal the value of one dime?”. A probability lem. In the number identification task, the word “Find”
problem asks: “If you flipped a coin ten times, how many appeared on the screen, and during the passive fixation
times would the coin be most likely to land on heads?”. trials, the word “Look” appeared on the screen. Stimuli
were presented in a block fMRI design in order to opti-
2.2.2. Working memory mize signal detection (Friston et al., 1999). In each task,
Three subtests of the Working Memory Test Battery stimuli were displayed for 5 s with an inter-trial interval
for Children (WMTB-C; Pickering and Gathercole, 2001) of 500 ms. There were 18 trials of each task condition, bro-
were used to assess central executive, phonological, and ken up into 4 blocks of 4 or 5 trials, thus each block lasted
visuo-spatial ability. Central executive ability was assessed either 22 or 27.5 s. The order of the blocks was random-
using Counting Recall wherein participants sequentially ized across participants with the following constraints:
counted aloud arrays of 4–7 dots on cards of progressively in every set of 4 blocks, all of the conditions were pre-
increasing sequence length, and then recited the totals sented and the Complex and Control addition task blocks
for each card in the order that the cards appeared in the were always separated by either a Find or a Passive fixa-
sequence. For each test, the score was determined by the tion block; all other orders of addition and non-addition
total number of trials successfully completed before the task conditions were equally likely. The total length of the
participant failed to produce the correct sequence three experimental run was 6 min and 36 s. A verification, rather
times at a given span level. Each span level consisted than verbal production, format was used in the scanner
of 6 possible trials. Phonological capacity was assessed because overt verbal responses can result in significant
using Digit Recall wherein participants were presented head movement in children resulting in unusable fMRI
with incrementally larger digit strings and then asked to data.
immediately reproduce the string. Visuo-spatial ability was
assessed using a variant of the Corsi block tapping task
called Block Recall wherein the assessor tapped out pro- 2.3.2. Stimulus presentation
gressively longer sequences of block patterns in a staggered The task was programed using E-Prime (Psychology
three-dimensional 9 block array; participants were asked Software Tools, Inc., Pittsburgh, PA) on a PC running
to immediately reproduce the sequence. For both tasks, Windows XP. The onsets of the fMRI scan and experimen-
the inter stimulus interval was approximately 1 second in tal task were synchronized using a TTL pulse delivered
duration. to the scanner timing microprocessor board from a
serial response box connected to the computer. Stimuli
2.3. Brain imaging were presented visually at the center of a screen using
a custom-built magnet compatible projection system.
2.3.1. Experimental procedures The temporal precision of stimulus presentation and
The fMRI experiment consisted of blocks of: (1) Complex response onset detection was accurate to approximately
addition, (2) Control addition, (3) Number identification ±1 ms.
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 165

2.3.3. fMRI data acquisition Table 1

Raw and standardized cognitive assessment scores.
Images were acquired on a 3T GE Signa scanner (Gen-
eral Electric, Milwaukee, WI) using a custom-built head coil Measure Raw Standardized
at Stanford University’s Richard M. Lucas Center for Imag-
M SD M SD
ing. Head movement was minimized during the scan by
Full-scale IQ – – 111 12.5
cushions placed around the participant’s head. A total of
WIAT-II
29 axial slices (4.0 mm thickness, 0.5 mm skip) parallel to Numerical Operations 17 4.3 105 15.3
the AC-PC line and covering the whole brain were imaged Mathematical Reasoning 35 6.0 110 13.8
with a temporal resolution of 2 s using a T2* weighted gra- WMTB-C
dient echo spiral in-out pulse sequence (Glover and Lai, Central executive 18 5.1 84 18.7
Phonological 28 3.7 106 15.5
1998) with the following parameters: TR = 2 s, TE = 30 ms,
Visuo-spatial 24 2.7 95 10.8
flip angle = 80◦ , 1 interleave. The field of view was 20 cm,
Note: WIAT-II = Wechsler Individual Achievement Test-Second Edition;
and the matrix size was 64 × 64, providing an in-plane spa-
WMTB-C = Working Memory Test Battery for Children.
tial resolution of 3.125 mm. To reduce blurring and signal
loss from field inhomogeneity, an automated high-order
shimming method based on spiral acquisitions was used shared variance among the three WM measures the con-
before acquiring functional MRI scans (Kim et al., 2002). trast images for Complex minus Control were examined
using three separate ANCOVA models including one for
2.3.4. fMRI preprocessing CE, PL and VS. Single sample t-tests were used to iden-
fMRI data were analyzed using SPM8 (http://www.fil. tify significant clusters of activity in separate whole brain
ion.ucl.ac.uk/spm/). The first 5 volumes were not analyzed analyses; in order to fully characterize WM-related effects,
to allow for T1 equilibration. A linear shim correction results from the regression analyses were not masked by
was applied separately for each slice during reconstruc- task-related activation. For all analyses, significant activa-
tion (Glover and Lai, 1998). Images were realigned to tion clusters were determined using a height threshold
the first scan to correct for motion and slice acquisition of p < 0.01, with cluster correction for multiple compar-
timing. Following procedures used in AFNI Despike (Cox, isons within a gray matter mask at p < 0.05 (k = 100 voxels)
1996), deviant volumes resulting from spikes in movement determined using Monte Carlo simulations implemented in
greater than 0.5 voxels (1.56 mm) (translational displace- MatLab using methods similar to AFNI’s AlphaSim program
ment computed as the square root of sum of the x, y, z (Forman et al., 1995; Ward, 2000).
plane) or spikes in the global signal greater than 5% were
then interpolated using the two adjacent scans. No partici- 3. Results
pant had greater than 15% of volumes interpolated. Images
were spatially normalized to standard MNI space using the 3.1. Cognitive assessments
echo-planar imaging template provided with SPM8, resam-
pled every 2 mm using trilinear sinc interpolation, and Raw and standardized scores for WIAT-II mathematical
smoothed with a 6 mm full-width half-maximum Gaussian subtests and the WMTB-C measures of working mem-
kernel to decrease spatial noise prior to statistical anal- ory components are presented in Table 1. Averages for all
ysis. Mean average motion over x, y, and z planes for all scores were in the normal range, with the exception of cen-
subjects was 1.0 mm (SD = 0.8) and unidirectional motion tral executive (CE). Due to our tight age range (7–9) and lack
was: x (M = 0.5 mm, SD = 0.5 mm), y (M = 0.9, SD = 0.9), and of correlations between age and the raw working memory
z (M = 1.5, SD = 1.0). measures (all p > .1), all subsequent analyses were per-
formed using these values rather than standardized scores.
2.3.5. Individual subject and group analyses We analyzed zero-order correlations between multiple
Task-related brain activation was identified using the component WM scores and assessments of math ability to
general linear model implemented in SPM8. In the indi- avoid elimination of shared variance consistent with the
vidual subject analyses, interpolated volumes flagged at main fMRI correlational analyses. Correlations between the
the preprocessing stage were de-weighted. For the math- behavioral measures are presented in Table 2. Each of the
ematical cognition task, brain activity related to each three WM measures was significantly correlated with all of
task condition was modeled using boxcar functions cor- the standardized measures of math achievement, with the
responding to the block length and convolved with a exception of PL ability and Numerical Operations.
canonical hemodynamic response function and a temporal
dispersion derivative to account for voxel-wise latency dif- 3.2. Brain imaging
ferences in hemodynamic response. Low-frequency drifts
at each voxel were removed using a high-pass filter 3.2.1. Behavioral performance
(0.5 cycle/min). Serial correlations were accounted for by In-scanner behavioral performance was analyzed using
modeling the fMRI time series as a first-degree autore- participant’s median reaction time (RT) and accuracy.
gressive process. Voxel-wise t-statistics maps contrasting Overall, average correct RT was 2572 ms (SD = 552) and
Complex and Control addition problems were generated average accuracy was 83% (SD = 12). Paired samples t-
for each participant. For group analyses, contrast images tests indicated that performance on Complex problems
corresponding to the Complex minus Control addition tasks was slower (2843 ms, SD = 621) and less accurate (77%,
were analyzed using a random effects analysis. To preserve SD = 16) than compared to Control problems (2366 ms,
166 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

Table 2
Correlation between cognitive measures.

Variables Mathematical Reasoning Central executive Phonological Visuo-spatial

Numerical Operations .61*** .26* .08 .44***

Mathematical Reasoning .37*** .26* .25*
Central executive .39*** .14
Phonological −.02

df = 72.
*
p < .05.
***
p ≤ .001.

SD = 554; 88%, SD = 12; t (73) = 8.3, SE = 57, p < .001 and t results are presented below in order of their rank-order for
(73) = -5.9, SE = 1.7, p < .001, respectively). Following the prediction of math achievement: VS, CE, and lastly, PL.
analytical model of our core imaging analyses, Complex
minus Control accuracy and median RT were analyzed as 3.2.3.1. VS. The VS component of WM was associated with
zero-order correlations with each WM component. In con- task-related activation in left supramarginal gyrus, left pos-
trast to correlations to the more behaviorally sensitive terior intra-parietal sulcus, right angular gyrus, left middle
measures included in the assessments of math ability, all frontal gyrus (dorsolateral prefrontal cortex), left infe-
scanner task behavior correlations were non-significant, rior frontal gyrus (ventrolateral prefrontal cortex), and the
p > .5. medial aspects of the superior frontal gyrus (Table 4 and
Fig. 2). No brain regions showed negative correlations with
VS ability.
3.2.2. Arithmetic task activation
We first examined overall levels of activation inde-
3.2.3.2. CE. The CE component of WM was associated with
pendent of WM. Overall arithmetic complexity related
task-related activation in bilateral superior parietal lob-
activation (Complex minus Control) was detected in a dis-
ule, bilateral intra-parietal sulcus, bilateral occipital cortex
tributed set of brain regions including bilateral lingual
and fusiform gyrus, left supramarginal gyrus, right angular
gyrus, bilateral fusiform gyrus, bilateral insula, left intra-
gyrus, right middle frontal gyrus (dorsolateral prefrontal
parietal sulcus and left pre-motor cortex (Table 3 and
cortex), and right hippocampus (Table 5 and Fig. 3). No
Fig. 1). Task-related deactivation was observed in bilateral
brain regions showed negative correlations with CE ability.
paracingulate cortex, anterior cingulate cortex and right
frontal pole.
3.2.3.3. PL. The PL component of WM was associated with
task-related activation in left anterior intra-parietal sulcus
3.2.3. Relation between arithmetic task activation and and lingual gyrus (Table 6 and Fig. 4). PL ability was neg-
individual WM scores atively correlated with task-related activation in the left
Regression analysis was used to investigate the relation frontal pole.
between arithmetic complexity related brain activation
and CE, PL, and VS ability scores. We performed three 3.2.4. Functional dissociations and overlap between WM
separate regression analyses, one for each subcomponent, components
in order to preserve the shared variance between WM Fig. 5 shows brain areas that demonstrated distinct
measures. This allowed us to detect functional dissocia- and overlapping arithmetic task-related activation associ-
tions as well as overlap across the three WM measures. ated with each WM component. Only two circumscribed
This approach is key to our research questions assessing regions of overlap were observed. First, overlap between
the uniqueness and overlap of WM components during VS and CE related activation was observed in the left supra-
arithmetic verification problem solving. The neuroimaging marginal gyrus. Second, overlap between CE and PL related

Table 3
Peak activation for whole brain analysis of Complex task performance.

Region Cluster size (voxels) Peak T-score Peak MNI coordinates (mm)

X Y Z

Complex > Control

B lingual gyrus 6188 5.57 16 −84 6
R fusiform gyrus 38 −68 −12
R insula cortex 610 4.36 38 22 0
L pre-motor cortex 409 4.13 −54 −6 48
L insula cortex 279 4.00 −28 22 2
L supramarginal gyrus/intra-parietal sulcus 209 3.30 −44 −44 46
Control > Complex
B paracingulate/anterior cingulate cortex 224 3.12 −4 48 −4
R frontal pole/paracingulate cortex 157 2.93 2 56 18

Note: p < .01, non-parametric FWE p < .05. B = bilateral. R = right; L = left.
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 167

Fig. 1. Arithmetic complexity-related brain activation. Significant group-level activations for the contrast of Complex–Control were detected in bilateral
anterior insula, right inferior frontal gyrus, left premotor cortex, right posterior hippocampus, bilateral lingual and fusiform gyri, and left intra-parietal
sulcus. Deactivation was detected only in the ventromedial prefrontal cortex. Complex addition problems were of the form x + y = z with one operand
from 2 to 9 and the other from 2 to 5; the Control problems had the same format except that x or y was set to 1. PMC, premotor cortex; Ins, insula; IPS,
intra-parietal sulcus; LG, lingual gyrus; FG, fusiform gyrus; IFG, inferior frontal gyrus; SFG, superior frontal gyrus; CG, cingulate gyrus; LG, lingual gyrus;
Hipp, hippocampus (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Table 4
Brain areas associated with visuo-spatial component of working memory.

Region Cluster size (voxels) Peak T-score Peak MNI coordinates (mm)

X Y Z

L supramarginal/angular gyrus 707 3.79 −52 −46 32

L inferior frontal gyrus 150 3.67 40 42 −14
R superior frontal gyrus 103 3.29 8 2 64
R intra-parietal sulcus (posterior) 156 3.20 38 −50 32
L middle frontal gyrus 134 2.97 −42 34 32

Note: p < .01, non-parametric FWE p < .05. R = right. L = left. No significant deactivation was observed.

Table 5
Brain areas associated with the central executive component of working memory.

Region Cluster size (voxels) Peak T-score Peak MNI coordinates (mm)

X Y Z

L superior parietal lobule 10,651 4.69 −28 −48 52

R superior parietal lobule 4.49 26 −50 68
L cuneus/superior occipital cortex 4.26 −20 −82 28
R occipital cortex 4.07 16 −88 14
R fusiform gyrus 4.04 26 −64 −20
R cuneus 3.39 30 −62 38
R angular gyrus 2.98 58 −46 24
R middle frontal gyrus 259 3.74 42 22 46
R hippocampus 183 3.27 30 −18 −10
R pre-motor cortex 115 3.23 22 −18 60
L supramarginal gyrus 160 3.18 −54 −50 34
R brain stem 182 3.12 6 −22 −6
R fusiform gyrus 116 3.00 40 −48 −26

Note: p < .01, non-parametric FWE p < .05. R = right. L = left. No significant deactivation was observed. Sub-peaks for the large posterior cluster (first in table)
were chosen at distance 32 mm apart.
168 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

Fig. 2. Brain areas associated with visuo-spatial component of working memory. Individual differences in visuo-spatial ability emerged as the strongest
and most consistent predictor of behavior and brain response during arithmetic problem solving. Increased arithmetic complexity-related brain responses
were located in left dorsolateral and right ventrolateral prefrontal cortex, as well as bilateral posterior parietal cortex including intra-parietal sulcus. MFG,
middle frontal gyrus; IPS, intra-parietal sulcus; SMG, supramarginal gyrus; SFG, superior frontal gyrus; IFG, inferior frontal gyrus. Bottom panel: regions
of interest depicted as orange open circles (radius: 6 mm). Scatter plots are based on functional clusters identified using whole-brain regression analysis,
and are provided for the purpose of visualization. L, left; R, right. Behavioral Performance: for comparison, plots of cognitive assessments of math and
visuo-spatial ability. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 169

Fig. 3. Brain areas associated with the central executive component of working memory. Central executive ability was correlated with distributed brain
areas including dorsolateral prefrontal cortex, posterior parietal cortex, and hippocampus during arithmetic problem solving. SPL, superior parietal lobule;
SMG, supramarginal gyrus; IPS, intra-parietal sulcus; MFG, middle frontal gyrus; AG, angular gyrus; FG, fusiform gyrus; OC, occipital cortex. Bottom panel:
regions of interest depicted as orange open circles (radius: 6 mm). Scatter plots are based on functional clusters identified using whole-brain regression
analysis, and are provided for the purpose of visualization. L, left; R, right. Behavioral Performance: for comparison, plots of cognitive assessments of math
and visuo-spatial ability. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
170 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

Table 6
Brain areas associated with the phonological component of working memory.

Region Cluster size (voxels) Peak T-score Peak MNI coordinates (mm)

X Y Z

Positive correlations
L intra-parietal sulcus 329 4.28 −24 −46 54
L lingual gyrus 129 3.50 −8 −86 −2
L superior occipital cortex 269 3.46 −22 −74 32
Negative correlations
L frontal pole 151 3.53 −20 56 2

Note: p < .01, non-parametric FWE p < .05. L = left.

activation was observed in left anterior intra-parietal sul- PL were generally weak both at the behavioral and neu-
cus. No overlap was observed between brain responses ral levels. Notably, no brain regions showed overlap across
associated with VS and PL ability. all three components. Interestingly, all pairwise overlap
was localized to left posterior parietal cortex. The effects
of VS and CE overlapped in the left supramarginal gyrus,
4. Discussion whereas the CE and PL showed overlap in left anterior
intra-parietal sulcus. Collectively these results are the first
Behavioral models of cognitive skill development have demonstration of distinct patterns of brain activity related
emphasized the separate roles of different WM compo- to multi-component WM in children as well as the first con-
nents. Here we examined whether CE, VS and PL, the key vincing demonstration of tripartite WM dissociation over
components of WM, map onto distinct patterns of brain multiple functional systems underlying arithmetic cogni-
activity during arithmetic verification problem solving. We tion in children. Below, we discuss the implication of our
are aware of no prior studies examining individual differ- results for a number of extant hypotheses concerning the
ences in multiple components of WM ability associated role of multiple WM components in children’s arithmetic
with problem solving in adults or children, thus, a pri- cognition including the fundamental role of VS ability in
ori, both positive and negative correlations might have numerical problem solving.
been possible: positive correlations could reflect appropri-
ate engagement of task relevant processes and negative
correlations could represent inefficient processing. It is 4.1. Differential role of visuo-spatial working memory
therefore noteworthy that, with the exception of the left underlying problem solving in children
frontal pole involvement in PL, no brain areas in our study
showed negative correlations with the CE and VS. This find- Behaviorally, VS ability emerged as the strongest predic-
ing emphasizes the facilitating effects of increased WM tor of math skills. Greater VS ability was associated with
capacity via increased cortical engagement during early better performance on Numerical Operations and Math
arithmetic development. Reasoning, two standardized measures of math ability from
Our study revealed a striking functional dissociation the WIAT-II. The largest correlation obtained between any
in arithmetic-related brain activity associated with each math and WM measure was between VS ability and Numer-
individual WM component. The VS, CE and PL compo- ical Operations ability (r = .44), a standardized measure
nents of WM were associated with distinct patterns of of arithmetic proficiency based on numerical calculations
task-related brain activation in prefrontal, posterior pari- and fact retrieval without any linguistic demands. For
etal and inferior temporal cortices. Notably, the VS and CE Mathematical Reasoning, which involves word-based math
were associated with distinct prefrontal-parietal circuits problem solving, VS ability was the second ranking predic-
known to be involved in WM. In contrast, the effects of tor compared to CE.

Fig. 4. Brain areas associated with the phonological component of working memory. Negative correlations were observed in the frontal pole (FP). Positive
correlates were detected in left anterior intra-parietal sulcus (IPS) and lingual gyrus (LG). Regions of interest depicted as open orange circles (radius: 6 mm).
Scatter plots are based on functional clusters identified using whole-brain regression analysis, and are provided for the purpose of visualization.
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 171

Fig. 5. Functional dissociations and overlap between brain areas associated with each of the three components of working memory. Overlap between
central executive (CE) and visuo-spatial (VS) components was observed only in left supramarginal gyrus (SMG), overlap between CE and phonological (PL)
components was observed only in the left intra-parietal sulcus (IPS) and no overlap was observed between VS and PL components. Negative correlation
between activity and PL ability is not depicted (see Fig. 4). No overlap for VS and PL (magenta) was observed. Bottom panel: coronal slices depict regions of
interest selected as overlap in correlations of activity and individual working memory components. Scatter plots are based on functional clusters identified
using whole-brain regression analysis, and are provided for the purpose of visualization. L, left.

VS abilities were also strongly correlated with brain Evidence for direct links between visuo-spatial WM and
responses during arithmetic problem solving. Individual arithmetic is also beginning to accumulate; Dumontheil
differences in VS ability were associated with increased and Klingberg (2012) found that activity level in the intra-
task-related brain responses in left dorsolateral and right parietal sulcus of the posterior parietal cortex during a
ventrolateral prefrontal cortex, as well as bilateral intra- visuo-spatial WM task predicted arithmetic performance
parietal sulcus and supramarginal gyrus in the posterior two years later. Rotzer et al. (2009) found that compared
parietal cortex. These responses were largely specific to VS to controls, children with low math ability had lower visuo-
ability, as overlap between the VS and CE components was spatial WM performance and lower levels of activity in
observed only in a circumscribed cluster of voxels in left the right inferior frontal gyrus, right intra-parietal sulcus,
supramarginal gyrus; no overlap was observed between VS and right insula during VS WM task performance. Taken
and PL abilities in any brain region. together, these findings provide important new insights
Multiple lines of evidence in children have pointed into the role of the VS component of WM in numerical
to overlapping patterns of activation of fronto-parietal problem solving in young children and they provide further
regions in VS WM and arithmetic problem solving evidence for direct links between the two domains. Our
(Kaufmann et al., 2011; Klingberg, 2006; Klingberg et al., results complement these findings by demonstrating that
2002; Kwon et al., 2002; Rivera et al., 2005; Rotzer et al., children with greater visuo-spatial WM capacity engage
2009). A core theme in these studies is that fronto-parietal frontal and parietal regions to a greater extent than chil-
circuits are involved in both VS WM and arithmetic tasks. dren with lower visuo-spatial WM capacity.
172 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

4.2. VS, CE and PL components of WM are associated 2012; Logie, 2011) in the domain of children’s arithmetic
with distinct brain responses cognition and highlight largely distinct neural processes
correlated with each component. Additionally, our study
Our findings demonstrate that CE, VS and PL are associ- also highlights a specific role of VS ability both at the
ated with distinct patterns of brain response, and provide behavioral and neural levels. This prominent role of VS abil-
new evidence to substantiate the Baddeley model of work- ity in both behavioral and brain measures of performance
ing memory in a developmental context (Baddeley, 2012). is consistent with models of arithmetic that emphasize
We observed a striking functional dissociation between spatial number line representation of quantity (Ashkenazi
brain regions associated with each individual WM compo- and Henik, 2010; Ashkenazi et al., 2013; Dehaene et al.,
nent. Importantly, the VS and PL components showed no 2003; Rotzer et al., 2009). The link between quantity and
overlap and the CE was the only component that showed space requires domain-general VS processes (de Hevia and
overlap with each of the other two components; this fea- Spelke, 2010; Dehaene et al., 2003); one central example for
ture of the data also supports the Baddeley and Hitch model the support systems that link number and space is visuo-
as PL and VS are characterized as distinct subordinate sys- spatial WM (De Smedt et al., 2009; Geary and Hoard, 2001;
tems under the control of the CE. Finally, it is noteworthy Holmes and Adams, 2006; Meyer et al., 2010; Rasmussen
that overlap between the CE and the two other compo- and Bisanz, 2005). During the initial stages of arithmetic
nents were observed only in the posterior parietal cortex, knowledge acquisition children use these mental number
suggesting that this region is an important locus for the line representations along with visuo-spatial WM to help
integration of information in WM during arithmetic prob- solve problems and build domain-specific representations
lem solving (Ansari, 2008; Dehaene et al., 2003; Kawashima (de Hevia and Spelke, 2010; Rasmussen and Bisanz, 2005).
et al., 2004; Kucian et al., 2008; Menon et al., 2000; Rivera In addition to the interesting findings for the role of VS
et al., 2005; Rosenberg-Lee et al., 2011). Although both the abilities in complex problem solving, the current paradigm
VS and CE ability were each associated with activation in afforded simultaneous examination of the role of CE pro-
multiple prefrontal and parietal areas, there was no overlap cesses during this period of arithmetic skill development.
between these components except in the ventral aspects Increased fronto-parietal recruitment helps explain why
of the left supramarginal gyrus. Notably, core prefrontal better CE ability leads to better long term retrieval of facts
regions associated with the VS and CE showed no over- (Bailey et al., 2012; Geary et al., 2012) despite the prox-
lap; this dissociation of WM-related functional activity is imal importance of VS ability and dovetails with other
consistent with cognitive conceptualizations of fraction- behavioral data showing that more successful attempts
ated WM (Baddeley, 2012) and suggest that behavioral at early executive intensive non-retrieval strategies leads
models separating the contribution of these three WM to higher correct retrieval later in development (Ashcraft,
components are fundamentally related to neural function 1982; Groen and Parkman, 1972; Siegler et al., 1995; Siegler
during an arithmetic task. The greater involvement of dis- and Shrager, 1984). Overlap in VS and CE correlations in
tinct prefrontal and parietal cortices in children with higher the left supramarginal gyrus represents another intriguing
CE ability is consistent with behavioral and neuroimaging aspect of our findings. Although multiple parietal regions
studies which have suggested that skill acquisition in the were engaged during our task, only the supramarginal
developmental window studied here is characterized by gyrus showed overlap between the CE and VS components.
greater reliance on control mechanisms that facilitate con- This ventral subdivision of the posterior parietal cortex is
trolled retrieval and the inhibition of irrelevant math facts typically engaged by cognitive tasks that require manipula-
or operations (Chen and Siegler, 2000; Cho et al., 2012; tion of the contents of WM (Kwon et al., 2002). In contrast,
Geary, 2004). Indeed, Bailey et al. (2012) recently showed the dorsal aspects of the parietal cortex, including the
that CE capacity in first graders (age 6) was predictive of IPS, have generally been implicated in visual short term
ability to retrieve facts from memory in later grades. In con- memory (Todd and Marois, 2005), spatial attention (Knops
trast to the VS and CE components, PL was not associated et al., 2009; Silk et al., 2010) and semantic representations
with increases in brain responses in any prefrontal cortex of quantity (Cohen Kadosh et al., 2008; Hubbard et al.,
regions. PL ability was associated with greater responses in 2005). Behavioral research has suggested that the visuo-
circumscribed regions of the superior parietal lobule that spatial store is supported by the more general attentional
overlapped with brain areas that were correlated with the resources of the CE component of WM (Alloway et al., 2006;
CE. Our findings suggest a pattern of largely distinct brain Thompson et al., 2006). Our study provides partial, but not
responses associated with each WM component. full, support for this view and identifies the left supra-
marginal gyrus as a locus of this interaction. It is, however,
4.3. Multi-component WM and problem solving skills in important to note that this was the only brain area where
children the VS and CE showed prominent overlap, suggesting that
the two components largely operate independently.
Although the role of multi-component WM in children’s Previous studies of numerical cognition have reported
arithmetic problem solving has been examined in several an anterior to posterior shift in activation with develop-
previous behavioral studies, no consensus has yet emerged ment (Kawashima et al., 2004; Kucian et al., 2008; Rivera
for the differential roles of importance of the three compo- et al., 2005). Within the posterior parietal cortex, at least
nents (Geary, 2004; Raghubar et al., 2010). Findings from two studies of arithmetic have reported greater activity
our study support the fundamental explanatory properties in older subjects in the left supramarginal gyrus (Rivera
of Baddeley and Hitch’s tripartite model of WM (Baddeley, et al., 2005; Rosenberg-Lee et al., 2011). These results
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 173

suggest that in math cognition, this area may be a key locus separable? Child Development 77, 1698–1716, http://dx.doi.org/
10.1111/j. 1467-8624.2006.00968.x.
for the intersection of development and individual differ-
Ansari, D., 2008. Effects of development and enculturation on number
ences in working memory, but further research is needed representation in the brain. Nature Reviews Neuroscience 9, 278–291.
to fully elucidate this relationship. In the present study Ashcraft, M.H., 1982. The development of mental arithmetic: a chrono-
we used a verification task which did not require a ver- metric approach. Developmental Review 2, 213–236.
Ashcraft, M.H., Battaglia, J., 1978. Cognitive arithmetic: evidence for
bal response, whether a verbal production task involves retrieval and decision processes in mental addition. Journal of Exper-
greater engagement of the PL remains to be investigated. imental Psychology: Human Learning and Memory 4, 527.
Finally, the current data suggest future possibilities for Ashkenazi, S., Henik, A., 2010. A disassociation between physi-
cal and mental number bisection in developmental dyscalcu-
some integration between developmental models of arith- lia. Neuropsychologia 48, 2861–2868, http://dx.doi.org/10.1016/
metic cognition that emphasize the importance of central j.neuropsychologia.2010.05.028.
executive processes and those that emphasize spatial- Ashkenazi, S., Rosenberg-Lee, M., Metcalfe, A.W.S., Swigart, A.G., Menon,
V., 2013. Visual-spatial working memory is an important source of
number related hypotheses by demonstrating connections domain-general vulnerability in the development of arithmetic cog-
to both, generalized over separate cortical areas during this nition. Neuropsychologia 51, 2305–2317.
important window in the development of cognitive arith- Ashkenazi, S., Rosenberg-Lee, M., Tenison, C., Menon, V., 2012. Weak task-
related modulation and stimulus representations during arithmetic
metic skill.
problem solving in children with developmental dyscalculia. Devel-
opmental Cognitive Neuroscience 2, 152–166.
5. Conclusions Baddeley, A.D., 1986. Working Memory. Clarendon Press/Oxford Univer-
sity Press, New York, NY.
Baddeley, A.D., 1996. Exploring the central executive. Quarterly Journal of
We examined Baddeley and Hitch’s multi-component Experimental Psychology 49A, 5–28.
WM in the context of arithmetic problem solving abilities Baddeley, A.D., 2003. Working memory: looking back and looking forward.
in children. We used a combination of neuropsychologi- Nature Reviews Neuroscience 4, 829–839.
Baddeley, A.D., 2012. Working memory: theories, models, and contro-
cal assessments of math ability and working memory, and versies. Annual Review of Psychology 63, 1–29, http://dx.doi.org/
functional neuroimaging, within the context of a single 10.1146/annurev-psych-120710-100422.
well-controlled cognitive task involving arithmetic prob- Baddeley, A.D., DellaSala, S., Papagno, C., Spinnler, H., 1997. Dual-task per-
formance in dysexecutive and nondysexecutive patients with a frontal
lem solving. We found a striking concordance between lesion. Neuropsychology 11, 187–194.
cognitively derived fractionation of WM and arithmetic Baddeley, A.D., Hitch, G.J., 1974. Working memory. In: Bower, G.H. (Ed.),
complexity-related functional brain activity. This sug- The Psychology of Learning and Motivation: Advances in Research and
Theory. Academic, New York, pp. 47–89.
gests that multi-component WM as a model of arithmetic Bailey, D.H., Littlefield, A., Geary, D.C., 2012. The codevelopment of
development explains fundamental relationships that are skill at and preference for use of retrieval-based processes for solv-
manifest at the level of neural activity. Our findings point ing addition problems: individual and sex differences from first to
sixth grades. Journal of Experimental Child Psychology 113, 78–92,
to a central role for VS abilities during ages 7–9, an impor-
http://dx.doi.org/10.1016/j.jecp.2012.04.014.
tant stage in the development of math abilities in children. Baroody, A.J., 1985. Mastery of basic number combinations: internaliza-
Finally, it is noteworthy that the VS and PL components tion of relationships or facts. Journal for Research in Mathematics
Education 16, 83–98.
showed no overlap in brain activation and the CE was the
Barrouillet, P., Lépine, R., 2005. Working memory and children’s use of
only component that showed overlap with each of the other retrieval to solve addition problems. Journal of Experimental Child
two components, consistent with the Baddeley and Hitch Psychology 91, 183–204.
model which places PL and VS as distinct subordinate sys- Cantlon, J.F., Libertus, M.E., Pinel, P., Dehaene, S., Brannon, E.M., Pelphrey,
K.A., 2009. The neural development of an abstract concept of number.
tems under the control of the CE. Overlap between the Journal of Cognitive Neuroscience 21, 2217–2229.
CE and the two other WM components was observed in Chen, Z., Siegler, R.S., 2000. Intellectual development in childhood. In:
the posterior parietal cortex, suggesting that this region Sternberg, R.J. (Ed.), Handbook of Intelligence. Cambridge University
Press, Cambridge, pp. 92–116.
is a locus for integration of information in WM during Cho, S., Metcalfe, A.W.S., Young, C.B., Ryali, S., Geary, D.C., Menon, V., 2012.
arithmetic problem solving in children. Further studies are Hippocampal-prefrontal engagement and dynamic causal interac-
needed to examine how brain systems associated with each tions in the maturation of children’s fact retrieval. Journal of Cognitive
Neuroscience 24, 1849–1866.
WM component interact dynamically to support arithmetic Cohen Kadosh, R., Lammertyn, J., Izard, V., 2008. Are numbers special? An
problem solving and skill development in young children. overview of chronometric, neuroimaging, developmental and com-
parative studies of magnitude representation. Progress in Neurobio-
logy 84, 132–147, http://dx.doi.org/10.1016/j.pneurobio.2007.11.001.
Conflicts of interest
Cox, R.W., 1996. AFNI: software for analysis and visualization of func-
tional magnetic resonance neuroimages. Computers and Biomedical
The authors declare they have no conflicts of interest. Research 29, 162–173, http://dx.doi.org/10.1006/cbmr.1996.0014.
de Hevia, M.D., Spelke, E.S., 2010. Number–space mapping in human
infants. Psychological Science 21, 653–660, http://dx.doi.org/10.1177/
Acknowledgments 0956797610366091.
De Smedt, B., Janssen, R., Bouwens, K., Verschaffel, L., Boets, B., Ghesquiere,
This research was supported by an NSERC postdoctoral P., 2009. Working memory and individual differences in mathemat-
ics achievement: a longitudinal study from first grade to second
fellowship (387787-2010) to AWSM and by grants from grade. Journal of Experimental Child Psychology 103, 186–201,
NIH (HD047520, HD059205, HD057610) and NSF (DRL- http://dx.doi.org/10.1016/j.jecp.2009.01.004.
0750340) to VM. Dehaene, S., Piazza, M., Pinel, P., Cohen, L., 2003. Three parietal cir-
cuits for number processing. Cognitive Neuropsychology 20, 487–506,
http://dx.doi.org/10.1080/02643290244000239.
References Della Sala, S., Logie, R.H., 2002. Neuropsychological impairments of visual
and spatial working memory. In: Baddeley, A.D., Wilson, B., Kopelman,
Alloway, T.P., Gathercole, S.E., Pickering, S.J., 2006. Verbal and vis- M. (Eds.), Handbook of Memory Disorders. John Wiley, Chichester, pp.
uospatial short-term and working memory in children: are they 271–292.
174 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175

Dumontheil, I., Klingberg, T., 2012. Brain activity during a visuo- Knops, A., Thirion, B., Hubbard, E.M., Michel, V., Dehaene, S., 2009. Recruit-
spatial working memory task predicts arithmetical performance ment of an area involved in eye movements during mental arithmetic.
2 years later. Cerebral Cortex 22, 1078–1085, http://dx.doi.org/ Science 324, 1583–1585, http://dx.doi.org/10.1126/science.1171599.
10.1093/cercor/bhr175. Kucian, K., von Aster, M., Loenneker, T., Dietrich, T., Martin, E., 2008. Devel-
Forman, S.D., Cohen, J.D., Fitzgerald, M., Eddy, W.F., Mintun, M.A., Noll, opment of neural networks for exact and approximate calculation: a
D.C., 1995. Improved assessment of significant activation in functional FMRI study. Developmental neuropsychology 33, 447–473.
magnetic resonance imaging (fMRI): use of a cluster-size threshold. Kwon, H., Reiss, A.L., Menon, V., 2002. Neural basis of protracted devel-
Magnetic Resonance in Medicine 33, 636–647. opmental changes in visuo-spatial working memory. Proceedings of
Friston, K.J., Zarahn, E., Josephs, O., Henson, R., Dale, A.M., 1999. Stochastic the National Academy of Sciences of the United States of America 99,
designs in event-related fMRI. Neuroimage 10, 607–619. 13336–13341, http://dx.doi.org/10.1073/pnas.162486399.
Geary, D.C., 2004. Mathematics and learning disabilities. Journal of Learn- Lee, K.M., Kang, S.Y., 2002. Arithmetic operation and working memory:
ing Disabilities 37, 4–15. differential suppression in dual tasks. Cognition 83, B63–B68.
Geary, D.C., Hoard, M.K., 2001. Numerical and arithmetical deficits in LeFevre, J.A., DeStefano, D., Coleman, B., Shanahan, T., 2005. Mathematical
learning-disabled children: relation to dyscalculia and dyslexia. Apha- cognition and working memory. In: Campbell, J.I.D. (Ed.), Handbook
siology 15, 635–647, http://dx.doi.org/10.1080/02687040143000113. of Mathematical Cognition. Psychology Press, New York, pp. 361–395.
Geary, D.C., Hoard, M.K., Byrd-Craven, J., DeSoto, M.C., 2004. Strategy Logie, R.H., 2011. The functional organization and capacity limits of work-
choices in simple and complex addition: contributions of working ing memory. Current Directions in Psychological Science 20, 240–245,
memory and counting knowledge for children with mathematical http://dx.doi.org/10.1177/0963721411415340.
disability. Journal of Experimental Child Psychology 88, 121–151, Menon, V., Rivera, S.M., White, C.D., Eliez, S., Glover, G.H., Reiss,
http://dx.doi.org/10.1016/j.jecp.2004.03.002. A.L., 2000. Functional optimization of arithmetic processing
Geary, D.C., Hoard, M.K., Nugent, L., 2012. Independent contributions in perfect performers. Cognitive Brain Research 9, 343–345,
of the central executive, intelligence, and in-class attentive behav- http://dx.doi.org/10.1016/S0926-6410(00)00010-0.
ior to developmental change in the strategies used to solve addition Meyer, M.L., Salimpoor, V.N., Wu, S.S., Geary, D.C., Menon, V., 2010.
problems. Journal of Experimental Child Psychology 113, 49–65, Differential contribution of specific working memory components
http://dx.doi.org/10.1016/j.jecp.2012.03.003. to mathematics achievement in 2nd and 3rd graders. Learn-
Glover, G.H., Lai, S., 1998. Self-navigated spiral fMRI: interleaved versus ing and Individual Differences 20, 101–109, http://dx.doi.org/
single-shot. Magnetic Resonance in Medicine 39, 361–368. 10.1016/j.lindif.2009.08.004.
Grabner, R.H., Ansari, D., Koschutnig, K., Reishofer, G., Ebner, F., Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edin-
Neuper, C., 2009. To retrieve or to calculate? Left angular burgh inventory. Neuropsychologia 9, 97–113.
gyrus mediates the retrieval of arithmetic facts during problem Pickering, S., Gathercole, S., 2001. Working Memory Test Battery for Chil-
solving. Neuropsychologia 47, 604–608, http://dx.doi.org/10.1016/ dren. The Pschological Corporation, London.
j.neuropsychologia.2008.10.013. Raghubar, K.P., Barnes, M.A., Hecht, S.A., 2010. Working memory and
Groen, G.J., Parkman, J.M., 1972. A chronometric analysis of simple addi- mathematics: a review of developmental, individual difference,
tion. Psychological Review 79, 329. and cognitive approaches. Learning and Individual Differences 20,
Heathcote, D., 1994. The role of visuo-spatial working memory in the 110–122, http://dx.doi.org/10.1016/j.lindif.2009.10.005.
mental addition of multi-digit addends. Cahiers de Psychologie Cog- Rasmussen, C., Bisanz, J., 2005. Representation and working memory
nitive/Current Psychology of Cognition 13, 207–245. in early arithmetic. Journal of Experimental Child Psychology 91,
Hitch, G.J., 1978. Role of short-term working memory in mental 137–157, http://dx.doi.org/10.1016/j.jecp.2005.01.004.
arithmetic. Cognitive Psychology 10, 302–323, http://dx.doi.org/ Rivera, S.M., Reiss, A.L., Eckert, M.A., Menon, V., 2005. Developmental
10.1016/0010-0285(78)90002-6. changes in mental arithmetic: evidence for increased functional spe-
Holmes, J., Adams, J.W., 2006. Working memory and children’s cialization in the left inferior parietal cortex. Cerebral Cortex 15,
mathematical skills: implications for mathematical development 1779–1790, http://dx.doi.org/10.1093/cercor/bhi055.
and mathematics curricula. Educational Psychology 26, 339–366, Rosenberg-Lee, M., Barth, M., Menon, V., 2011. What difference does a
http://dx.doi.org/10.1080/01443410500341056. year of schooling make? Maturation of brain response and connectiv-
Hubbard, E.M., Piazza, M., Pinel, P., Dehaene, S., 2005. Interactions between ity between 2nd and 3rd grades during arithmetic problem solving.
number and space in parietal cortex. Nature Reviews Neuroscience 6, Neuroimage 57, 796–808.
435–448, http://dx.doi.org/10.1038/Nrn1684. Rottschy, C., Langner, R., Dogan, I., Reetz, K., Laird, A.R., Schulz, J.B., Fox,
Imbo, I., Vandierendonck, A., 2008. Effects of problem size, oper- P.T., Eickhoff, S.B., 2012. Modelling neural correlates of working mem-
ation, and working-memory span on simple-arithmetic strate- ory: a coordinate-based meta-analysis. Neuroimage 60, 830–846,
gies: differences between children and adults? Psychological http://dx.doi.org/10.1016/j.neuroimage.2011.11.050.
Research-Psychologische Forschung 72, 331–346, http://dx.doi.org/ Rotzer, S., Loenneker, T., Kucian, K., Martin, E., Klaver, P., von
10.1007/s00426-007-0112-8. Aster, M., 2009. Dysfunctional neural network of spatial
Imbo, I., Vandierendonck, A., Vergauwe, E., 2007. The role working memory contributes to developmental dyscalculia.
of working memory in carrying and borrowing. Psycho- Neuropsychologia 47, 2859–2865, http://dx.doi.org/10.1016/
logical Research-Psychologische Forschung 71, 467–483, j.neuropsychologia.2009.06.009.
http://dx.doi.org/10.1007/s00426-006-0044-8. Siegler, R.S., 1987. The perils of averaging data over strategies: an example
Ischebeck, A., Zamarian, L., Egger, K., Schocke, M., Delazer, M., 2007. Imag- from children’s addition. Journal of Experimental Psychology: General
ing early practice effects in arithmetic. Neuroimage 36, 993–1003. 116, 250.
Kaufmann, L., Wood, G., Rubinsten, O., Henik, A., 2011. Meta- Siegler, R.S., Shipley, C., Simon, T.J., Halford, G.S., 1995. Variation, selection,
analyses of developmental fMRI studies investigating typical and cognitive change. In: Simon, T., Halford, G. (Eds.), Developing Cog-
and atypical trajectories of number processing and cal- nitive Competence: New Approaches to Process Modeling. Erlbaum,
culation. Developmental Neuropsychology 36, 763–787, Hillsdale, NJ, pp. 31–76.
http://dx.doi.org/10.1080/87565641.2010.549884. Siegler, R.S., Shrager, J., 1984. Strategy choice in addition and subtraction:
Kawashima, R., Taira, M., Okita, K., Inoue, K., Tajima, N., Yoshida, H., how do children know what to do? In: Sophian, C. (Ed.), Origins of
Sasaki, T., Sugiura, M., Watanabe, J., Fukuda, H., 2004. A func- cognitive skills. Erlbaum, Hillsdale, NJ, pp. 229–293.
tional MRI study of simple arithmetic – a comparison between Silk, T.J., Bellgrove, M.A., Wrafter, P., Mattingley, J.B., Cunnington, R., 2010.
children and adults. Cognitive Brain Research 18, 227–233, Spatial working memory and spatial attention rely on common neu-
http://dx.doi.org/10.1016/j.cogbrainres.2003.10.009. ral processes in the intraparietal sulcus. Neuroimage 53, 718–724,
Kim, D.H., Adalsteinsson, E., Glover, G.H., Spielman, D.M., 2002. Regular- http://dx.doi.org/10.1016/j.neuroimage.2010.06.068.
ized higher-order in vivo shimming. Magnetic Resonance in Medicine Simmons, F.R., Willis, C., Adams, A.M., 2012. Different components of
48, 715–722. working memory have different relationships with different mathe-
Klingberg, T., 2006. Development of a superior frontal- matical skills. Journal of Experimental Child Psychology 111, 139–155,
intraparietal network for visuo-spatial working memory. http://dx.doi.org/10.1016/j.jecp.2011.08.011.
Neuropsychologia 44, 2171–2177, http://dx.doi.org/10.1016/ Thompson, J.M., Hamilton, C.J., Gray, J.M., Quinn, J.G., Mackin, P.,
j.neuropsychologia.2005.11.019. Young, A.H., Ferrier, I.N., 2006. Executive and visuospatial sketch-
Klingberg, T., Forssberg, H., Westerberg, H., 2002. Increased brain pad resources in euthymic bipolar disorder: implications for
activity in frontal and parietal cortex underlies the devel- visuospatial working memory architecture. Memory 14, 437–451,
opment of visuospatial working memory capacity during http://dx.doi.org/10.1080/09658210500464293.
childhood. Journal of Cognitive Neuroscience 14, 1–10, Todd, J.J., Marois, R., 2005. Posterior parietal cortex activity predicts
http://dx.doi.org/10.1162/089892902317205276. individual differences in visual short-term memory capacity.
 A.W.S. Metcalfe et al. / Developmental Cognitive Neuroscience 6 (2013) 162–175 175

Cognitive Affective & Behavioral Neuroscience 5, 144–155, Ward, B.D., 2000. Simultaneous inference for fMRI data. AFNI 3dDecon-
http://dx.doi.org/10.3758/Cabn.5.2.144. volve Documentation. Medical College of Wisconsin.
Trbovich, P., LeFevre, J.-A., 2003. Phonological and visual working Wechsler, D., 1999. Wechsler Abbreviated Scale of Intelligence. Psycho-
memory in mental addition. Memory & Cognition 31, 738–745, logical Corporation, San Antonio, TX.
http://dx.doi.org/10.3758/bf03196112. Wechsler, D., 2001. The Wechsler Individual Achievement Test — Second
Vallar, G., Papagno, C., 2002. Neuropsychological impairments of verbal Edition (WIAT-II). The Psychological Corporation.
short-term memory. In: Baddeley, A.D., Kopelman, M.D., Wilson, B.A.
(Eds.), Handbook of Memory Disorders. , second ed. Wiley, Chichester,
pp. 249–270.