Learning and Instruction 32 (2014) 73e80

Contents lists available at ScienceDirect

Learning and Instruction

journal homepage: www.elsevier.com/locate/learninstruc

Construction play and cognitive skills associated with the

development of mathematical abilities in 7-year-old children
Swiya Nath*, Dénes Szücs*
Centre for Neuroscience in Education, Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge, CB2 3EB, UK

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

Article history: Construction play is thought to develop logico-mathematical skills, however the underlying mechanisms
Received 16 May 2013 have not been defined. In order to fill this gap, this study looks at the relationship between Lego con-
Received in revised form struction ability, cognitive abilities and mathematical performance in 7-year-old, Year 2 primary school
17 January 2014
children (N ¼ 66). While studies have focused on the relationship between mathematics performance
Accepted 19 January 2014
and verbal memory, there are limited studies focussing on visuospatial memory. We tested both vi-
Available online 11 February 2014
suospatial and verbal working memory and short term memory, as well as non-verbal intelligence.
Mathematical performance was measured through the WIAT-II numerical operations, and the word
Keywords:
Mathematics
reading subtest was used as a control variable. We used a Lego construction task paradigm based on four
Construction play task variables found to systematically increase construction task difficulty. The results suggest that Lego
Visuospatial memory construction ability is positively related to mathematics performance, and visuospatial memory fully
Lego mediates this relationship. Future work of an intervention study using Lego construction training to
develop visuospatial memory, which in turn may improve mathematics performance, is suggested.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction 2. Construction play, spatial ability, and mathematics

Children learn through play. Construction play is defined by
Piaget as activities producing symbolic products, and is thought to
develop logico-mathematical knowledge (Wolfgang & Phelps,
1983). However, the mechanisms of how construction play can
develop logico-mathematical skills are not defined. Baddeley’s
(1986, 2000) influential memory model differentiated between
verbal and visuospatial memory systems. Ever since, there has been
high interest in analysing the relationship between memory sys-
tems and learning. Previous studies on construction play and
mathematics have focused on the relationship with verbal memory
(Richardson & Richardson, 2011), but there are no studies on con-
struction play and visuospatial memory, which may be more rele-
vant. In fact, there is a general paucity of research on visuospatial
memory, in comparison to the research on verbal memory
(Raghubar, Barnes, & Hecht, 2010). In this study we analyse the
relationship between construction play and mathematics perfor-
mance and test whether visuospatial memory is a mechanism
mediating this relationship.
* Corresponding authors.
E-mail addresses: sn413@cam.ac.uk (S. Nath), ds377@cam.ac.uk (D. Szücs).
performance
Studies analysing the relationship between construction play
and cognitive abilities have focused on spatial ability (Caldera et al.,
1999; Connor & Serbin, 1977; Robert & Heroux, 2004). Spatial ability
has been further divided into static (fixed objects) or dynamic
(movement of objects), and intrinsic (comparison within an object)
or extrinsic (comparison between objects) (Uttal et al., 2013).
Studies have found correlations between construction play and
spatial abilities. Connor and Serbin (1977) first asked undergradu-
ate students to categorize toys as masculine or feminine, and then
analysed play preferences of children according to the categoriza-
tion. They found boys playing with masculine activities (blocks, and
large motor toys) were correlated with higher performance of
spatial ability as measured by the Wechsler Preschool and Primary
Scale of Intelligence Block Design and Preschool Embedded
Figures Test (EFT). However for girls, they did not find significant
relationships between cognitive development and play preferences
for either masculine or feminine activities (Connor & Serbin, 1977).
In a study with older children (9-year-old), past experience in
playing with Lego did not correlate with any of the spatial mea-
sures. There was a negative relationship between the mental
rotation ability, and the number of errors and the time taken to

0959-4752/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.learninstruc.2014.01.006
74 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80
construct the Lego model (Brosnan, 1998). Similarly, in another
study, although preference for construction play was not signifi-
cantly correlated with any of the spatial tests, accuracy on the
structured block play was correlated with Block Design and
Copying Blocks, but not with the Children’s EFT (Caldera et al.,
1999). In 9 and 12-year-old children, construction play was signif-
icantly related to Water Level tasks and Block Design but not EFT
(Robert & Heroux, 2004). It is important to note that two of the
studies (Brosnan, 1998; Caldera et al., 1999) correlating construc-
tion ability and spatial ability found that accuracy, rather than
preference for construction play, was positively correlated with
spatial ability. The other studies only looked at preference of con-
struction play, but did not have a measure to test accuracy in
construction ability.
There are a few studies analysing the relationship between
construction ability and mathematics. In adolescents, a study ana-
lysing the relationship between block building and mathematical
performance found that structural balance of a block building ac-
tivity was correlated to mathematics performance (Casey, Pezaris, &
Bassi, 2012). In younger children, a study found that building a
model according to instructions was correlated to early maths
ability in 3-year-olds (Verdine et al., 2013). Another study on the
adaptiveness and complexity of construction play and mathematics
performance was conducted by observing and recording construc-
tion play with manipulatives (blocks, Lego, carpentry) in a child care
program with preschool children 3e4 years of age (Stannard,
Wolfgang, Jones, & Phelps, 2001). They found that the correlation
between the children’s construction play and mathematics perfor-
mance was not significant at younger years, but the correlation was
significant in grade 7 (12-year-old) and beyond. The researchers
suggest this as evidence that construction play in preschool can have
long term effects on logico-mathematical knowledge (Stannard
et al., 2001). In a subsequent study, researchers looked at the rela-
tionship between block construction and reading and mathematical
abilities, using Tests of Early Mathematics and Reading Abilities in
preschool children (Hanline, Milton, & Phelps, 2010). Children
played once a week for 90 min and their block constructions were
scored according to a 19-point scale based on the complexity and
symbolic representational properties of the constructions. They did
not find a relationship between block construction and mathematics
performance, but with reading performance at 8 years of age
(Hanline et al., 2010). These findings may be a result of the study
focussing on symbolic play, defined by Piaget as transforming ob-
jects into make-believe play through motor or verbal actions
(Wolfgang & Phelps, 1983), more than construction play. Both the
studies mentioned found a relationship between construction
ability and mathematics performance (Stannard et al., 2001) or
reading performance (Hanline et al., 2010) several years after con-
struction ability was tested. It could be that construction ability
develops more complex academic skills that are not tested till later
years. Nevertheless, with the time lapse between the observations
of construction ability and the measures of academic performance, it
is difficult to assess whether the findings relate specifically to con-
struction ability or as a result of general cognitive development,
which was not tested. Another study with children (10-year-old)
found that spatial ability is uniquely correlated to scientific
reasoning, when taking into account verbal and non-verbal intelli-
gence (Mayer, Sodian, Koerber, & Schwippert, 2014). This suggests
that spatial ability may be a unique underlying mechanism ac-
counting for individual differences in academic performance.
The only study on construction play and mathematics that ac-
counts for cognitive skills is by Richardson and Richardson (2011)
who tested the relationship between Lego construction ability,
spatial ability, verbal memory and mathematics performance in 3
groups of children (7e8-year-old, 10e11-year-old, and 13e14-year-
old). Lego construction ability was measured using the Lego con-
struction paradigm (Richardson, Jones, & Torrance, 2004) in which
eighteen different single coloured Lego construction tasks of pro-
gressive difficultly were used (see Methodology section). In-
structions were provided on a one page isometric view of the
models, that children had to create with Lego blocks, while their
construction time was recorded. They found significant correlations
between Lego construction time and mathematical performance as
well as spatial ability, but not with verbal memory. They found that
in older children (10e11-year-old, and 13e14-year-old) the rela-
tionship between construction time and mathematical perfor-
mance was mediated by spatial ability. Richardson and Richardson
(2011) was the only study to date to analyse construction play
together with mathematical performance and verbal memory.
However, a control task such as linguistic performance, was not
included to test whether the observed relationship with mathe-
matics performance was domain general, or specific to Lego con-
struction ability and mathematics development. They tested Digit
Span, a measure of verbal memory, however visuospatial memory
may be more relevant to construction tasks. In fact, no studies have
so far measured visuospatial working memory in relation to con-
struction tasks.
Miyake et al. analysed the relationship between spatial abilities,
visuospatial memory, and executive functions. They found that
although the results show a higher unique correlation between
spatial abilities and executive functions, visuospatial memory
accounted for some of the variance found in spatial abilities
(Miyake, Friedman, Rettinger, Shah, & Hegarty, 2001). Taking this as
theoretical support, we focused our study on the relationship be-
tween visuospatial memory and construction ability.
2.1. Verbal and visuospatial memory models, and mathematics
The Baddeley (1986, 2000) memory model describes verbal and
visuospatial memory as slave systems, with the central executive as
an overarching umbrella term. Verbal memory has been further
subdivided into verbal short term memory (STM), defined as simple
storage, while verbal working memory (WM) is thought to involve
STM as well as controlled attention, which is a domain of executive
functions (Miyake et al., 2001). Researchers suggest that the most
parsimonious model is the one in which STM is domain specific to
verbal or visuospatial domains, while WM is domain general
(Alloway, Gathercole, & Pickering, 2006). In contrast, others suggest
that visuospatial memory should be divided along characteristics
different than that of verbal memory. For example, some suggest
that a visual cache stores information about form and colour,
whereas the inner scribe would store information about movement
sequences (Logie & Pearson, 1997). Yet, other models divide visuo-
spatial memory into visual or spatial domains (Mammarella et al.,
2006; Passolunghi & Mammarella, 2012; Pickering, Gathercole,
Hall, & Lloyd, 2001). Visual tasks require recognition of form,
shape and colour of objects, whereas spatial tasks require the
recognition of location, position and configuration of objects, while
processing them simultaneously or sequentially (Mammarella et al.,
2006).
Visuospatial working memory has been correlated with math-
ematics performance in several studies (Ashkenazi, Rosenberg-Lee,
Metcalfe, Swigart, & Menon, 2013; Holmes & Adams, 2006;
McKenzie, Bull, & Gray, 2003; Meyer, Salimpoor, Wu, Geary, &
Menon, 2010; Passolunghi & Mammarella, 2012). One study ana-
lysed specific visual and spatial memory impairments and mathe-
matics ability (Passolunghi & Mammarella, 2012). The tasks
consisted of visual and spatial tasks at both simple and complex
levels, and intrusion errors (test of inhibition) were also measured.
Children with mathematics learning disability were selectively
 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80
poorer at spatial tasks, with more intrusion errors, but were not
significantly different on visual tasks from typical populations
(Passolunghi & Mammarella, 2012). Likewise, children with devel-
opmental dyscalculia (diagnosed as children with poor mathe-
matical performance, with normal IQ and reading performance)
were found to have impairments in visuospatial WM, STM and
inhibition ability compared to typically developing populations
(Szucs, Devine, Soltesz, Nobes, & Gabriel, 2013). This highlights the
relationship between inhibition, visuospatial processing and
mathematics performance, in isolation from other memory mea-
sures and the executive functions of attention, and task switching
(Szucs, Devine, Soltesz, Nobes, & Gabriel, 2014).
Studies have compared verbal memory and visuospatial mem-
ory in relations to mathematics performance. In a study of children
(8e9-year-old), only visuospatial working memory, and not verbal
working memory, remained significantly correlated with mathe-
matics performance when age-related variance was controlled
(Holmes & Adams, 2006). Visuospatial working memory measures
predicted unique variance in younger (8-year-old) children’s
mathematics performance, but only predicted unique variance in
older children (9-year-old) for difficult mathematics questions. The
researchers suggest that younger children have a higher de-
pendency on visuospatial working memory than older children,
who revert back to visuospatial working memory for more difficult
tasks (Holmes & Adams, 2006). In another study, younger children
(6-year-old) performed significantly poorer on mathematics tasks
during the visuospatial interference, but not the phonological
interference. Older children’s (8-year-old) performance was
significantly disrupted during the phonological interference, as
well as the visuospatial interference, but not to the same extent as
the younger children (McKenzie et al., 2003). This suggests that
younger children are more reliant on visuospatial working memory
for their mathematics performance than older children. On the
other hand, when verbal STM, WM, and visuospatial memory were
compared to WIAT-II numerical operations and mathematical
reasoning scores, younger children (7e8-year-old) showed a cor-
relation between verbal STM scores and mathematical reasoning,
while older children (8e9-year-old) showed a correlation between
visuospatial memory and numerical operations (Meyer et al., 2010).
They conclude that younger children are more reliant on verbal
processing and then shift to visuospatial processing. It could also be
that WIAT-II mathematical reasoning is based on word problems
which may be more salient to younger children, than older children
who are more proficient readers. The authors (Meyer et al., 2010)
suggest that verbal processing may play a larger role in initial
learning and performance, while visuospatial processing may
facilitate development of mathematics at later stages. In all the
studies comparing and contrasting verbal and visuospatial mem-
ory, the unique role of visuospatial memory in the development of
mathematics has not been disputed. Taking this as evidence, we
tested whether visuospatial memory could be the underlying
mechanism in the relationship between construction ability and
mathematics performance.
2.2. Overview of current study
Construction play is thought to develop logico-mathematical
skills, however the underlying mechanisms have not been
defined. Studies have found a relationship between performance on
construction play and spatial abilities (Caldera et al., 1999; Connor &
Serbin, 1977; Robert & Heroux, 2004) as well as mathematics per-
formance (Richardson & Richardson, 2011; Stannard et al., 2001).
Although there are studies that have factored in cognitive abilities of
IQ (Stannard et al., 2001), and verbal memory (Richardson &
Richardson, 2011), there is no study that compares construction
75
ability and visuospatial memory. Given the visuospatial processing
characteristics inherent in construction ability, it is more relevant to
test visuospatial memory than verbal memory as previously done
(Richardson & Richardson, 2011). Visuospatial memory has been
correlated with spatial abilities (Miyake et al., 2001) as well as
mathematics performance, particularly in younger children (6e8-
year-old) (Holmes & Adams, 2006; McKenzie et al., 2003). With
this rationale, we hypothesized that visuospatial memory as the
mechanism underlying the relationship between construction play
and mathematics performance. Here we operationally tested con-
struction ability in a Lego construction task paradigm (Richardson
et al., 2004). For the first time in a construction task setting we
tested several measures of verbal STM and WM, as well as visuo-
spatial STM and visuospatial WM. We used standardized measures
of WIAT-II mathematics and reading performance, as well as stan-
dardized measures of memory (AWMA). Reading performance was
measured to test whether construction ability is domain specific to
mathematics, or a domain general cognitive ability. We also tested
general non-verbal intelligence by using Raven’s Coloured Pro-
gressive Matrices (CPM). We recruited children 7 years of age. We
hypothesized that Lego construction ability (LegoCA) would be
correlated with mathematics performance (Hypothesis 1), with
visuospatial WM and STM mediating this relationship (Hypothesis
2), and tested this using mediational analysis. We did not expect to
find a relationship between LegoCA and verbal WM and STM (Hy-
pothesis 3).
3. Methodology
3.1. Participants
There were 66 participants in total (M ¼ 7.4 years, SD ¼ 3.7
months, 30 females). Ethical approval was provided by the Cam-
bridge Psychology Research Ethics Committee. After the study,
children were given gift vouchers and sweets to thank them for
participating in our study.
3.2. Tests
The Wechsler Individual Achievement Test e Second UK Edition
(WIAT-II) numerical operations and reading subtests were used as
they are specific to the UK population. The Raven’s Coloured Pro-
gressive Matrices (CPM) was used as it is designed for children, and
the scores can be standardized. The CPM tests non-verbal abstract
reasoning.
3.3. Memory tasks
The Automated Working Memory Assessment-1 (AWMA-1) was
used since it consists tests of verbal STM, verbal WM, visuospatial
STM and visuospatial WM (corresponding to the phonological loop,
visuospatial sketch pad and central executive components of Bad-
deley and Hitch’s 1974 model of working memory), which are
factors of particular interest to this study (Alloway, 2007). Scores
were standardized across individuals (M ¼ 100, SD ¼ 15). The test
reliability of the AWMA-1 was between a ¼ 0.69 and a ¼ 0.90 when
tested across 128 individuals ranging in age from 4-year-old to 23-
year-old (Alloway, 2007).
The AWMA-1 consisted of four pre-recorded tasks. In Digit
Recall (verbal STM) participants had to recall numbers presented
aurally. In Listening Recall (verbal WM) participants had to first
process whether the sentence was true or false, and then recall the
last word of the sentences in the order they were presented. In Dot
Matrix (visuospatial STM) participants had to recall the position of
a dot in a matrix in the order that it was presented. This task is akin
76 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80
to the simple spatial-sequential task as there is a recall of the order
(Passolunghi & Mammarella, 2012). In Odd One Out (visuospatial
WM) participants had to identify the odd shape, and then recall the
position of the odd shape in the order they were presented. This
task is akin to the complex spatial-sequential task as it requires
processing of the stimuli followed by recall of the order
(Passolunghi & Mammarella, 2012). In all the AWMA-I tests, four
correct answers led to the next level, while failing three consecu-
tive answers self-terminated the subtest.
3.4. Construction ability tasks
To measure Lego construction ability (LegoCA), participants
were given Lego pieces and given instructions on building the
model. The instructions were displayed on one page, and consisted
of exploded isometric models. The highest level of the Lego model
constructed accurately was used as the participants’ construction
ability score in the data analysis. The instructions for the Lego
construction tasks were provided by Miles Richardson (see Fig. 1 for
example). Only one colour was used for each Lego model. Accuracy
and construction time was recorded, and a picture of the con-
structed model was taken.
The Lego construction task paradigm was formulated by
Richardson et al. (2004). They found that the four construction task
variables of 1) symmetrical planes (number of orientations
required to find correct placement), 2) novel assemblies (number of
unique assemblies), 3) components (number of pieces to create
structure), and 4) selections (total number of components to select
required components from) were systematically correlated with
the time taken to construct the structure. Through regressional
analysis, the researchers derived a mathematical formula to model
task difficulty:
Task difficulty ¼ 10½ð0:020 componentsÞþð0:117 symmetrical planesÞþð0:047 novel assembliesÞþð0:028 selectionsÞþ1:464
The task is more difficult with a higher number of novel
assemblies, components and selections. The task is easier with a
Fig. 1. Exploded isometric model instructions for Lego construction (Richardson et al.,
2004).
higher number of symmetrical planes since there are fewer re-
strictions on the orientation of the component.
The Lego construction task paradigm has been validated in
adults and children as young as 7 years of age (a ¼ 0.83) through
replication studies (Richardson, Jones, Croker, & Brown, 2011). The
construction tasks were made easier than the tasks used by
Richardson and Richardson (2011) by not including components
other than those required to construct the structure. For example, if
a construction task required 4 Lego components, then only 4 Lego
components were provided, so that children did not have to choose
the correct 4 Lego components out of a selection of 6 Lego com-
ponents. Using the mathematical model of Richardson and
Richardson (2011) the 7 Lego construction tasks were presented
in order of task difficulty (see Table 1 for example).
The testing session took approximately an hour and a half. All
tasks were tested in individual sessions. Children were given breaks
between tasks. The tasks were administered in randomised order.
3.5. Statistical analysis
The relationship between performance on LegoCA, WM tasks,
STM tasks, CPM, and WIAT-II (Reading and Numerical Operations),
were tested using Spearman’s ranked correlation, since the Lego
task was ranked by level. A mediation analysis was used to test if
cognitive abilities mediated the relationship between the perfor-
mance on LegoCA and WIAT-II numerical operations. Mediation
analysis followed the recommendations of Baron and Kenny (1986)
and Preacher and Hayes (2008) of four steps in the procedure to test
for mediation: (1) testing if the relationship between the inde-
pendent variable and the dependent variable is significant, (2)
testing if the relationship between the independent variable and
the mediator variable is significant, (3) testing if the relationship
between the mediator variable and the dependent variable while
controlling for the independent variable is significant, and (4) if the
relationship between the independent variable and the dependent
variable becomes non-significant after running the analysis, there
is a full mediation (Baron & Kenny, 1986; Preacher & Hayes, 2008).
The mediation analysis had LegoCA as the independent variable,
CPM visuospatial WM and STM as the mediator variables, and
WIAT-II numerical operations as the dependent variable. The data
was standardized, and so the mediational analysis reports stan-
dardized beta values. Since the raw order data consists variables
that were not normally distributed, bootstrapping was used to
assess the significance of the mediation analysis results as recom-
mended by Preacher and Hayes (2008). For this purpose we used
the SPSS script provided by Hayes (2013). For bootstrapping anal-
ysis the data was resampled 5000 times. The outcome of the test
can be considered significant if the 95% confidence interval of
computed statistics does not include zero. Only the bootstrapping
results of tests which fully satisfied the requirements of media-
tional analysis are presented. All statistical testing was done using
SPSS Statistics Version 20.
4. Results
None of the children could complete the last Lego construction
task of the highest task difficulty level, and so there were no ceiling
effects. The last Lego construction task was removed from analysis.
Maximum level of Lego construction task was used in the data
 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80
Table 1
Lego models in order of task difficulty (Richardson et al., 2004).
Model Components Symmetrical planes
4 1
5 1.2
7 2
5 1.8
6 1.8
9 2
8 1
analysis. The recall and processing scores of the Odd-One-Out test
were strongly correlated r ¼ 0.94, p < .001. Hence, in order to avoid
multicollinearity problems they were combined into an average
and named Visuospatial WM (VSWM) score. Likewise, the recall
and processing scores of the Listening Recall test were strongly
correlated r ¼ 0.94, p < .001, they were combined into an average
and named Verbal WM score.
Descriptive statistics are provided in Table 2. Spearman’s rank
correlations between measures are shown in Table 3. WIAT-II nu-
merical operations were significantly correlated with visuospatial
STM and WM, CPM, verbal WM and WIAT-II word reading, but not
significantly correlated to verbal STM. LegoCA was significantly
correlated with WIAT numerical operations, visuospatial WM and
STM, and CPM. The correlation between performance on LegoCA
and verbal memory scores, and reading scores were not significant.
A mediation analysis was used to test for group mediation, with
LegoCA as the independent variable, CPM, visuospatial WM and
STM as the mediator variables, and WIAT-II numerical operations as
the dependent variable. The mediation models are shown in Fig. 2.
Fig. 2B presents the results of the mediation analysis of visuo-
spatial memory and CPM. LegoCA was significantly related to the
dependent variable WIAT-II numerical operations, thus satisfying
the first step of the mediation analysis procedure. LegoCA was
significantly related to all three mediators, thus satisfying the
second step of the mediation analysis procedure. The mediators
were not significantly related to WIAT-II numerical operations
when LegoCA was controlled, and so the model does not satisfy the
third step of the mediation analysis procedure. These results
77
Novel assemblies Selections Task difficulty
1 4 38.55
1 5 40.79
2 7 45.71
4 5 48.02
4 6 53.20
2 9 57.02
5 8 92.47
suggest that there was no group mediation effect in the relation-
ship between LegoCA and WIAT-II numerical operations. CPM, Vi-
suospatial WM and STM, were then analysed individually to test for
unique mediation effects.
Fig. 2C presents the results of the mediation analysis used to test
whether CPM mediated the relationship between LegoCA and
WIAT-II numerical operations. LegoCA was significantly related to
the dependent variable WIAT-II numerical operations, thus satis-
fying the first step of the mediation analysis procedure. LegoCA was
significantly related the mediator variable CPM, thus satisfying the
second step of the mediation analysis procedure. The mediator
CPM was not significantly related to WIAT-II numerical operations
when LegoCA was controlled, and so the model does not satisfy the
third step of the mediation analysis procedure. These results sug-
gest that CPM does not mediate the relationship between LegoCA
and WIAT-II numerical operations.
Fig. 2D presents the results of the mediation analysis used to test
whether VSWM mediated the relationship between LegoCA and
WIAT-II numerical operations. LegoCA was significantly related to
the dependent variable WIAT-II numerical operations, thus satis-
fying the first step of the mediation analysis procedure. LegoCA was
significantly related the mediator variable VSWM, thus satisfying
the second step of the mediation analysis procedure. The mediator
VSWM was significantly related to WIAT-II numerical operations
when LegoCA was controlled, satisfying the third step of the
mediation analysis procedure. The relationship between LegoCA
and WIAT-II numerical operations becomes non-significant when
VSWM is controlled, satisfying the fourth step of the mediation
78 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80

Table 2 Lego construction ability was significantly related to, visuospa-

Descriptive statistics.
Task Mean
WIAT-II numerical operations 111.65
Lego max level (LegoCA) 4.12
Visuospatial working memory (VSWM) 114.20
Dot matrix (Visuospatial STM) 103.74
Verbal working memory (Verbal WM) 111.70
Digit recall (Verbal STM) 108.82
Raven’s coloured progressive matrices (CPM) 104.77
WIAT-II word reading 112.97
analysis procedure. These results suggest full mediation by VSWM
between LegoCA and WIAT-II numerical operations. Since not all
variables were normally distributed, bootstrapping was used. The
confidence interval (0.040e0.282) at 95% suggests that these
findings are significant.
Fig. 2E presents the results of the mediation analysis used to test
whether visuospatial STM mediated the relationship between
LegoCA and WIAT-II numerical operations. LegoCA was significantly
related to the dependent variable WIAT-II numerical operations,
thus satisfying the first step of the mediation analysis procedure.
LegoCA was significantly related the mediator variable visuospatial
STM, thus satisfying the second step of the mediation analysis pro-
cedure. The mediator visuospatial STM was significantly related to
WIAT-II numerical operations when LegoCA was controlled, satis-
fying the third step of the mediation analysis procedure. The rela-
tionship between LegoCA and WIAT-II numerical operations
becomes non-significant when visuospatial STM is controlled,
satisfying the fourth step of the mediation analysis procedure. These
results suggest full mediation by visuospatial STM between LegoCA
and WIAT-II numerical operations. Since not all variables were nor-
mally distributed, bootstrapping was used. The confidence interval
(0.025e0.235) at 95% suggests that these findings are significant.
5. Discussion
The aim of this study was to assess the relationship between
Lego construction ability and mathematics performance, while
testing visuospatial memory as a mechanism mediating this rela-
tionship. We hypothesized that Lego construction ability would be
correlated with mathematics performance (Hypothesis 1), with
visuospatial WM and STM mediating this relationship (Hypothesis
2), and tested it using mediational analysis. We did not expect to
find a relationship between Lego construction ability and verbal
WM and STM (Hypothesis 3).
tial memory, CPM and mathematics performance, but not to verbal
SD memory and reading performance. These findings support our
18.48
Hypothesis 1 and 3. This suggests that Lego construction ability is
1.22 uniquely correlated with mathematics performance, and not driven
11.85 by general cognitive ability or intelligence. A mediation analysis
15.85 was run to account for cognitive abilities in the relationship be-
13.54
tween Lego construction ability and mathematics performance. The
11.58
14.56 model inclusive of both CPM and visuospatial memory did not find
11.51 mediation effects. The mediators were then tested individually. We
found that the relationship between Lego construction ability and
mathematics performance is uniquely mediated by both visuo-
spatial WM and STM, but not CPM. These findings support our
Hypothesis 2. This suggests that visuospatial short-term and
working memory is the mechanism mediating the relationship
between Lego construction ability and mathematics performance.
Since the CPM was not a mediating factor, the relationship between
Lego construction ability and mathematics is not explained as a
general non-verbal intelligence, but a relationship specifically
mediated by visuospatial memory.
There seems to be a disassociation between the relationship of
Lego construction ability, visuospatial memory and CPM to math-
ematics performance and between the relationship of verbal
memory and word reading abilities to mathematical performance.
Lego construction ability is only correlated to the visuospatial
memory and CPM, but not to verbal memory or word reading
ability. Several studies analysing the relationship between both
verbal and visuospatial memory, mathematics and reading, have
found that visuospatial memory is correlated only to mathematics
(Kyttälä, 2008; Passolunghi, Mammarella, & Altoè, 2008; Szucs
et al., 2013, 2014). Our data finds the same. Although mathe-
matics and reading performance are correlated, visuospatial mea-
sures are only correlated to mathematics performance and not
reading performance. This suggests that Lego construction ability is
correlated with a unique visuospatial aspect of mathematics that is
not transferable to word reading and verbal memory.
Mathematics performance was correlated with word reading,
CPM, visuospatial STM and WM, and verbal WM. Mathematics
performance was not significantly correlated with verbal STM. A
study comparing children in grade 2 (7e8-year-old) and grade 3
children (8e9-year-old) (Meyer et al., 2010), found mathematical
reasoning was correlated with verbal WM, and not verbal STM in
younger children, but not in older children. The same study, found
WIAT-II numerical operations correlated with visuospatial memory
only in older children (Meyer et al., 2010). In our study, we found
WIAT-II numerical operations correlated with visuospatial memory

Table 3
Spearman’s rho correlation matrix for all variables.

Numerical LegoCA VSWM Visuospatial vWM Verbal STM CPM

operations STM

LegoCA Correlation .298*

p .014
VSWM Correlation .357** .377**
p .003 .002
Visuospatial STM Correlation .373** .311* .516**
p .002 .011 .000
vWM Correlation .250* .066 .295* .203
p .043 .597 .016 .102
Verbal STM Correlation .034 .132 .042 .128 .276*
p .786 .292 .740 .304 .025
CPM Correlation .256* .265* .424** .431** .148 .064
p .038 .031 .000 .000 .235 .608
Word reading Correlation .498** .066 .026 .108 .409** .222 .229
p .000 .619 .847 .416 .001 .092 .081

Note: * ¼ p < .05; ** ¼ p < .01.

 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80 79

A) Mediator
a Variable b

Independent Variable Dependent Variable

c (c1)
Note: * = p <.05; ** = p <.01, n.s = not significant

B) CPM
.291* .061n.s
.372 ** VSWM .213n.s

Visuospatial STM
.312* .223n.s
Lego Construction Ability WIAT-II Numerical
.275* (.109n.s) Operations

C) .291* CPM .189n.s D) .372** VSWM .326*

LegoCA Math LegoCA Math

.275* (.220.s) .275* (.154n.s)
E) Visuospatial
.312* STM .333**

LegoCA Math
.275* (.172n.s)

Fig. 2. Mediation analysis, (A) The mediation analysis model used in all analyses, with the paths labelled (a), (b), (c) and (c1). Path (a) represents the standardized beta coefficient of
the direct relationship between the independent variable and the mediator variable(s). Path (b) represents the standardized beta coefficient of the indirect relationship between the
mediator variable(s) and the dependent variable, while controlling for the independent variable. Path (c) represents the standardized beta coefficient of the direct relationship
between the independent variable and the dependent variable. The notation (c1) represents the standardized beta coefficient of the indirect relationship between the independent
variable and the dependent variable, while controlling for the mediator variables. (B) Mediation analysis of visuospatial memory and CPM, (C) Mediation analysis of CPM, (D)
Mediation analysis of VSWM, (E) Mediation analysis of Visuospatial STM.

measures in 7-year-old children. These conflicting results question
the suggestion that younger children rely on verbal processing and
then switch to visuospatial processing, with mathematical devel-
opment. The authors (Meyer et al., 2010) suggest that verbal pro-
cessing facilitates initial learning, whereas visuospatial processing
is involved with long term mathematical development.
Going further, since WM is thought to include STM and
controlled attention, future work can test executive functions to
theoretically isolate the visuospatial aspects and differentiate it
from generalized executive functions. This has been previously done
by testing inhibition, attention and task switching (Szucs et al.,
2014), intrusion tasks (Passolunghi & Mammarella, 2012), and a
latent variable approach comparing STM, WM and fluid intelligence
(Engle, Tuholski, Laughlin, & Conway, 1999). Such a study would
allow us to further differentiate whether Lego construction ability is
uniquely correlated to visuospatial memory, or generalized execu-
tive functions.
Intervention studies have found training with manipulatives
improves performance in visuospatial memory. A study compared
musical training and mental abacus training on verbal WM and STM,
and spatial simultaneous and spatial sequential at simple and
complex levels (Lee, Lu, & Ko, 2007). Children (12-year-old) with
mental abacus training showed improved performance on spatial
tasks compared to the control group, while children with musical
training showed improved performance on both verbal and spatial
tasks (Lee et al., 2007). Another intervention study with children
(9e10-year-old), has shown that it is possible to improve visuo-
spatial memory through training of spatial sequential tasks (Caviola,
Mammarella, Cornoldi, & Lucangeli, 2009). The intervention was
presented as a game where a protagonist carried out several spatial
sequential tasks. Teachers discussed strategies of coding and
analysing the stimuli, creating chunks of visuospatial stimuli and
using mental images, followed by a discussion on recognizing the
best strategy to use. The intervention group showed marked
improvement on spatial sequential tasks, while spatial simulta-
neous tasks and verbal memory remained comparable to the control
group (Caviola et al., 2009). The Lego construction task paradigm
and isometric model instructions require the usage of similar stra-
tegies of coding and analysing the components to build Lego con-
struction models with increasing task difficulty, and may be a
valuable intervention tool to fine tune visuospatial memory.
Visuospatial memory is correlated with mathematics perfor-
mance, particularly in younger children 6-year-old (Holmes &
Adams, 2006) to 8-year-old (McKenzie et al., 2003). Children
begin to rely on verbal working memory and revert back to vi-
suospatial memory for more difficult mathematics questions
(Holmes & Adams, 2006). A reason for this development may be the
educational system that enforces verbal memory above and beyond
visuospatial memory. There is a paucity of research on visuospatial
memory in comparison with verbal memory. Studies do show a
unique correlation between visuospatial memory and mathe-
matics, and that children with mathematics learning disabilities
and developmental dyscalculia have visuospatial memory impair-
ments (Passolunghi & Mammarella, 2012; Szucs et al., 2014). Pre-
vious research has shown that visuospatial memory can be
specifically trained (Caviola et al., 2009; Lee et al., 2007). These
findings highlight educational implications for using intervention
tools for children with visuospatial memory impairments, which
are correlated with mathematics performance. The results from
this study suggest that the relationship between mathematics
performance and Lego construction ability is uniquely mediated by
visuospatial memory. The benefit of using the Lego construction
80 S. Nath, D. Szücs / Learning and Instruction 32 (2014) 73e80
paradigm (Richardson et al., 2004) as an intervention tool is that
the task difficulty can be systematically increased in increments.
This raises the possibility that Lego construction tasks may be used
to facilitate visuospatial memory, and thus the development of
mathematical performance. Future work with an intervention
study will be required to test this hypothesis.
In conclusion we found that Lego construction play is strongly
related to mathematical performance in primary school children
and this relationship is mediated by visuospatial memory. Our re-
sults raise the possibility that construction play may be effectively
used to train and strengthen mathematical skill in young children
and future studies can investigate using construction play for the
remediation of specific mathematical learning difficulties, like
developmental dyscalculia.
Acknowledgements
This research was supported by UK Medical Research Council
grant (G90951) to Denes Szucs. Sincere gratitude to Miles Richard-
son for providing the Lego construction task paradigm instructions.
The data was collected with the help of Amy Devine, Alison Nobes,
Dr. Florence Gabriel, Jacqueline Chung, Jacqueline Ding and Eleanor
Smith.
References
Alloway, T. P. (2007). Automated working memory assessment (AWMA) (1st ed.).
Pearson Assessment.
Alloway, T. P., Gathercole, S. E., & Pickering, S. J. (2006). Verbal and visuospatial
short-term and working memory in children: are they separable? Child Devel-
opment, 77(6), 1698e1716. http://dx.doi.org/10.1111/j.1467-8624.2006.00968.x.
Ashkenazi, S., Rosenberg-Lee, M., Metcalfe, A. W., Swigart, A. G., & Menon, V. (2013).
Visuo-spatial working memory is an important source of domain-general
vulnerability in the development of arithmetic cognition. Neuropsychologia,
51(11), 2305e2317. http://dx.doi.org/10.1016/j.neuropsychologia.2013.06.031.
Baddeley, A. D. (1986). Working memory. Oxford: Oxford University Press.
Baddeley, A. D. (2000). The episodic buffer: a new component of working memory?
Trends in Cognitive Sciences, 4(11), 417e422. http://dx.doi.org/10.1016/S1364-
6613(00)01538-2.
Baron, R. M., & Kenny, D. A. (1986). The moderatoremediator variable distinction in
social psychological research: conceptual, strategic, and statistical consider-
ations. Journal of Personality and Social Psychology, 51(6), 1173e1182. http://
dx.doi.org/10.1037/0022-3514.51.6.1173.
Brosnan, M. J. (1998). Spatial ability in children’s play with Lego blocks. Perceptual
and Motor Skills, 87(1), 19e28. http://dx.doi.org/10.2466/pms.1998.87.1.19.
Caldera, Y. M., McDonald-Culp, A., O’Brien, M., Truglio, R., Alvarez, M., &
Huston, A. C. (1999). Children’s play preferences, construction play with blocks,
and visualespatial skills: are they related? International Journal of Behavioral
Development, 23, 855e872. http://dx.doi.org/10.1080/016502599383577.
Casey, B. M., Pezaris, E. E., & Bassi, J. (2012). Adolescent boys’ and girls’ block con-
structions differ in structural balance: a block-building characteristic related to
math achievement. Learning and Individual Differences, 22(1), 25e36. http://
dx.doi.org/10.1016/j.lindif.2011.11.008.
Caviola, S., Mammarella, I. C., Cornoldi, C., & Lucangeli, D. (2009). A metacognitive
visuospatial working memory training for children. International Electronic
Journal of Elementary Education, 2(1), 122e136.
Connor, J. M., & Serbin, L. A. (1977). Behaviorally based masculine-and feminine-
activity-preference scales for preschoolers: correlates with other classroom
behaviors and cognitive tests. Child Development, 48(4), 1411e1416. http://
dx.doi.org/10.2307/1128500.
Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. (1999). Working
memory, short-term memory, and general fluid intelligence: a latent-variable
approach. Journal of Experimental Psychology: General, 128(3), 309e331. http://
dx.doi.org/10.1037/0096-3445.128.3.309.
Hanline, M. F., Milton, S., & Phelps, P. C. (2010). The relationship between preschool
block play and reading and maths abilities in early elementary school: a longi-
tudinal study of children with and without disabilities. Early Child Development
and Care, 180(8), 1005e1017. http://dx.doi.org/10.1080/03004430802671171.
Hayes, A. F. (2013). My macros and code for SPSS and SAS. Retrieved July 26, 2013,
from http://www.afhayes.com/spss-sas-and-mplus-macros-and-code.html.
Holmes, J., & Adams, J. W. (2006). Working memory and children’s mathematical
skills: implications for mathematical development and mathematics
curricula. Educational Psychology, 26(3), 339e366. http://dx.doi.org/10.1080/
01443410500341056.
Kyttälä, M. (2008). Visuospatial working memory in adolescents with poor per-
formance in mathematics: variation depending on reading skills. Educational
Psychology, 28(3), 273e289. http://dx.doi.org/10.1080/01443410701532305.
Lee, Y. S., Lu, M. J., & Ko, H. P. (2007). Effects of skill training on working memory
capacity. Learning and Instruction, 17(3), 336e344. http://dx.doi.org/10.1016/
j.learninstruc.2007.02.010.
Logie, R. H., & Pearson, D. G. (1997). The inner eye and the inner scribe of
visuo-spatial working memory: evidence from developmental fractionation.
European Journal of Cognitive Psychology, 9(3), 241e257. http://dx.doi.org/
10.1080/713752559.
Mammarella, I. C., Cornoldi, C., Pazzaglia, F., Toso, C., Grimoldi, M., & Vio, C. (2006). Evi-
dence for a double dissociation between spatial-simultaneous and spatial-sequential
working memory in visuospatial (nonverbal) learning disabled children. Brain and
Cognition, 62(1), 58e67. http://dx.doi.org/10.1016/j.bandc.2006.03.007.
Mayer, D., Sodian, B., Koerber, S., & Schwippert, K. (2014). Scientific reasoning
in elementary school children: assessment and relations with cognitive
abilities. Learning and Instruction, 29, 43e55. http://dx.doi.org/10.1016/
j.learninstruc.2013.07.005.
McKenzie, B., Bull, R., & Gray, C. (2003). The effects of phonological and visual-
spatial interference on children’s arithmetical performance. Educational and
Child Psychology, 20(3), 93e108.
Meyer, M. L., Salimpoor, V. N., Wu, S. S., Geary, D. C., & Menon, V. (2010). Differential
contribution of specific working memory components to mathematics
achievement in 2nd and 3rd graders. Learning and Individual Differences, 20(2),
101e109. http://dx.doi.org/10.1016/j.lindif.2009.08.004.
Miyake, A., Friedman, N. P., Rettinger, D. A., Shah, P., & Hegarty, M. (2001). How are
visuospatial working memory, executive functioning, and spatial abilities
related? A latent-variable analysis. Journal of Experimental Psychology: General,
130(4), 621e640. http://dx.doi.org/10.1037/0096-3445.130.4.621.
Passolunghi, M. C., & Mammarella, I. C. (2012). Selective spatial working memory
impairment in a group of children with mathematics learning disabilities and
poor problem-solving skills. Journal of Learning Disabilities, 45(4), 341e350.
http://dx.doi.org/10.1177/0022219411400746.
Passolunghi, M. C., Mammarella, I. C., & Altoè, G. (2008). Cognitive abilities as
precursors of the early acquisition of mathematical skills during first through
second grades. Developmental Neuropsychology, 33(3), 229e250. http://
dx.doi.org/10.1080/87565640801982320.
Pickering, S. J., Gathercole, S. E., Hall, M., & Lloyd, S. A. (2001). Development of
memory for pattern and path: further evidence for the fractionation of visuo-
spatial memory. The Quarterly Journal of Experimental Psychology: Section A,
54(2), 397e420. http://dx.doi.org/10.1080/713755973.
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for
assessing and comparing indirect effects in multiple mediator models. Behavior
Research Methods, 40(3), 879e891. http://dx.doi.org/10.3758/BRM.40.3.879.
Raghubar, K. P., Barnes, M. A., & Hecht, S. A. (2010). Working memory and mathe-
matics: a review of developmental, individual difference, and cognitive ap-
proaches. Learning and Individual Differences, 20(2), 110e122. http://dx.doi.org/
10.1016/j.lindif.2009.10.005.
Richardson, M., Jones, G., Croker, S., & Brown, S. (2011). Identifying the task char-
acteristics that predict children’s construction task performance. Applied
Cognitive Psychology, 25(3), 377e385. http://dx.doi.org/10.1002/acp.1702.
Richardson, M., Jones, G., & Torrance, M. (2004). Identifying the task variables that
influence perceived object assembly complexity. Ergonomics, 47(9), 945e964.
http://dx.doi.org/10.1080/00140130410001686339.
Richardson, M., & Richardson, C. (2011). Building blocks of mathematics: Under-
standing children’s construction play as a basis for educational and cognitive
assessment (Unpublished manuscript).
Robert, M., & Heroux, G. (2004). Visuo-spatial play experience: forerunner of visuo-
spatial achievement in preadolescent and adolescent boys and girls? Infant and
Child Development, 13(1), 49e78. http://dx.doi.org/10.1002/icd.336.
Stannard, L., Wolfgang, C. H., Jones, I., & Phelps, P. (2001). A longitudinal study of the
predictive relations among construction play and mathematical achievement.
Early Child Development and Care, 167(1), 115e125. http://dx.doi.org/10.1080/
0300443011670110.
Szucs, D., Devine, A., Soltesz, F., Nobes, A., & Gabriel, F. (2013). Developmental
dyscalculia is related to visuo-spatial memory and inhibition impairment.
Cortex, 49(10), 2674e2688. http://dx.doi.org/10.1016/j.cortex.2013.06.007.
Szucs, D., Devine, A., Soltesz, F., Nobes, A., & Gabriel, F. (2014). Cognitive compo-
nents of a mathematical processing network in 9-years-old children. Develop-
mental Science (in press).
Uttal, D. H., Meadow, N. G., Tipton, E., Hand, L. L., Alden, A. R., Warren, C., et al.
(2013). The malleability of spatial skills: a meta-analysis of training
studies. Psychological Bulletin, 139(2), 352e402. http://dx.doi.org/10.1037/
a0028446.
Verdine, B. N., Golinkoff, R. M., Hirsh-Pasek, K., Newcombe, N. S., Filipowicz, A. T., &
Chang, A. (2013). Deconstructing building blocks: Preschoolers’ spatial assem-
bly performance relates to early mathematical skills. Child development, 1e15.
http://dx.doi.org/10.1111/cdev.12165.
Wolfgang, C. H., & Phelps, P. (1983). Preschool play materials preference inventory.
Early Child Development and Care, 12(2), 127e141. http://dx.doi.org/10.1080/
0300443830120204.