Allo Way 2006

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Child Development, November/December 2006, Volume 77, Number 6, Pages 1698 – 1716

Verbal and Visuospatial Short-Term and Working Memory in Children:


Are They Separable?
Tracy Packiam Alloway and Susan J. Pickering
Susan Elizabeth Gathercole University of Bristol
University of Durham

This study explored the structure of verbal and visuospatial short-term and working memory in children be-
tween ages 4 and 11 years. Multiple tasks measuring 4 different memory components were used to capture the
cognitive processes underlying working memory. Confirmatory factor analyses indicated that the processing
component of working memory tasks was supported by a common resource pool, while storage aspects depend
on domain-specific verbal and visuospatial resources. This model is largely stable across this developmental
period, although some evidence exists that the links between the domain-specific visuospatial construct and the
domain-general processing construct were higher in the 4- to- 6-year age group. The data also suggest that all
working memory components are in place by 4 years of age.

The term working memory refers to the capacity to explore, in children, whether verbal and visuospatial
store and manipulate information over brief periods memory are supported by a common resource pool
of time (Baddeley & Hitch, 1974; Just & Carpenter, or whether they are maintained by separable cogni-
1992). Extensive research over the past three decades tive resources. An issue of particular interest is
has established that working memory is not a single whether the structure of working memory under-
store, but a memory system comprised of separable goes changes across the childhood years.
interacting components. Functioning in concert, One account of domain-general working memory
these components provide a kind of flexible mental capacity is provided by the Baddeley and Hitch
workspace that can be used to maintain and trans- (1974) model. The central executive is the component
form information in the course of demanding cog- responsible for controlling resources and monitoring
nitive activities, and that acts as a temporary bridge information processing across informational do-
between externally and internally generated mental mains (e.g., Baddeley, Emslie, Kolodny, & Duncan,
representations. There are a range of alternative 1998; Baddeley & Hitch, 1974). This system is also
theoretical accounts of working memory capacity. responsible for a range of regulatory functions, in-
Some researchers suggest that working memory cluding the retrieval of information from long-term
consists of a domain-general aspect that coordinates memory and attentional control. In this model,
information in two independent domain-specific storage of information is mediated by two domain-
storage components for verbal and visuospatial specific slave systems: the phonological loop, which
codes (Baddeley, 1986; see Engle, Kane, & Tuholski, provides temporary storage of verbal information,
1999a, for a review). Others argue instead that and the visuospatial sketchpad, specialized for the
working memory resources are separated into verbal maintenance and manipulation of visual and spatial
and visuospatial constructs (Shah & Miyake, 1996). representations (see Baddeley & Logie, 1999, for a
These studies have largely been restricted to adult review). A fourth component of this model has re-
populations. The aim of the present study was to cently been added, the episodic buffer, responsible
for binding information across informational do-
mains and memory subsystems into integrated
This research was supported by a research grant awarded by the chunks (Baddeley, 2000). This model of working
Economic and Social Research Council of Great Britain to Tracy memory has been supported by evidence from
Packiam Alloway, Susan Elizabeth Gathercole, and Susan J. Pick-
studies of children (e.g., Alloway, Gathercole, Willis,
ering. The authors thank the schools, parents, and children who
consented to participate in this study. The authors also thank Chris & Adams, 2004), adult participants, neuropsycho-
Jarrold and two anonymous reviewers for their helpful comments logical patients (see Baddeley, 1996; Jonides, Lacey, &
on an earlier draft of this manuscript.
Correspondence concerning this article should be addressed to
Dr. Tracy Packiam Alloway, School of Education, University of
Durham, Leazes Road, Durham DH1 1TA, U.K. Electronic mail r 2006 by the Society for Research in Child Development, Inc.
may be sent to t.p.alloway@durham.ac.uk. All rights reserved. 0009-3920/2006/7706-0014
Verbal and Visuospatial Short-Term and Working Memory in Children 1699

Nee, 2005, for reviews), and neuroimaging investi- Research on adult participants and on older children
gations (Vallar & Papagno, 2003). supports this distinction (Friedman & Miyake, 2000;
Domain-general accounts of working memory Jarvis & Gathercole, 2003; Miyake, Friedman, Ret-
capacity have also been advanced by other theorists. tinger, Shah, & Hegarty, 2001).
Engle and colleagues (Engle, Tuholski, Laughlin, & Working memory capacity is measured by com-
Conway, 1999b; see also Kane et al., 2004) have plex span tasks that require simultaneous short-term
suggested that working memory capacity is limited storage of information while processing additional,
by controlled attention, the ability to allocate atten- and sometimes unrelated, information. According to
tional resources despite distraction or interference. domain-general accounts of working memory, the
Engle et al. (1999b) used a latent-variable analysis in processing aspect of the task is controlled by a cen-
order to extract the common variance shared by tralized component (i.e., the central executive or
verbal short-term memory tasks such as digit span, controlled attention), while the short-term storage
and verbal working memory tasks such as reading aspect is supported by a domain-specific component
span. They argued that the residual variance in (i.e., verbal or visuospatial store). From a domain-
verbal working memory reflected controlled pro- specific perspective, performance in complex tasks is
cessing, which was uniquely linked with general a function of efficiency in either verbal or visuospa-
fluid intelligence. The commonality between this tial abilities. Consequently, performance in verbal
model and the Baddeley and Hitch (1974) working working memory tasks would not predict spatial
memory model (see also Baddeley, 1986) is that there abilities, nor would spatial working memory mea-
is a central component that functions as a coordin- sures be highly associated with verbal skills. Evidence
ator between ongoing processing of information for this position has been provided by Shah and
(controlled attention and the central executive, in the Miyake (1996; see also Friedman & Miyake, 2000).
respective models) and storage of information in the The principal aim of the present study was to in-
subsystems. Both models also incorporate domain- vestigate the detailed structure of working memory
specific storage components: one for verbal infor- in childhood. A key issue addressed in the present
mation and another for visuospatial information. study is whether working memory during the
However, they can be distinguished in terms of the childhood years is best characterized by a model
relationship between these storage components and incorporating domain-general resources supple-
of high-level cognitive abilities such as reasoning mented by domain-specific storage such as the
and comprehension. While in the Engle et al. (1999b) working memory model of Baddeley (2000), or by a
model only the controlled attentional aspect of model in which working memory resources are
working memory capacity is predictive of learning, fractionated across the verbal and visuospatial do-
in the Baddeley and Hitch (1974) model links have mains (e.g., Shah & Miyake, 1996). In addition to
been established between both the central executive research on adult populations (e.g., Kane et al., 2004),
and the specialized storage systems and academic developmental research has supported the distinc-
attainment (e.g., Pickering & Gathercole, 2004). This tion between verbal and visuospatial short-term
theoretical distinction is addressed in a recent study storage in children as young as 5 years of age. For
by Bayliss, Jarrold, Gunn, and Baddeley (2003), who example, Pickering, Gathercole, and Peaker (1998)
found that the ability to coordinate the processing found that for short-term serial recall, scores on
and storage aspects of working memory tasks con- verbal and spatial tasks were dissociable in 5- and 8-
tributes to the prediction of reading and mathematic year-olds, although error patterns were similar. This
ability. The authors suggested that these findings indicates that while different memory systems are
provide support for a multicomponent model of involved in verbal and spatial short-term memory,
working memory that includes a domain-general there may be a common mechanism underlying the
processing component and domain-specific storage reconstruction of serial order that contributes to
components. performance in both domains (see also Chuah &
An alternative account of working memory cap- Maybery, 1999).
acity is provided by Miyake and colleagues, who In a broader individual differences analysis of
have proposed that working memory capacity is both short-term storage tasks and working memory
supported by two separate pools of domain-specific tasks in children, Bayliss et al.’s (2003; see also Bay-
resources for verbal and visuospatial information liss, Jarrold, Baddeley, Gunn, & Leigh, 2005) findings
(Shah & Miyake, 1996; see also Daneman & Tardiff, favored both a domain-general processing compon-
1987). Each domain is independently capable of ent and domain-specific storage components in 7- to
manipulating and keeping information active. 9-year-olds. Gathercole, Pickering, Ambridge, and
1700 Alloway, Gathercole, and Pickering

Wearing (2004; see also Gathercole & Pickering, 2000) assessments in English, mathematics, and science.
extended these findings and established that there is Multiple tasks measuring four different memory
a processing component and domain-specific storage components were taken in order to capture the pro-
systems in place from as young as 6 years of age. cesses underpinning working memory. The mea-
This tripartite structure of working memory was sures were taken from the Automated Working
consistent across the childhood period up to 15 Memory Assessment (AWMA; Alloway, Gathercole,
years. While this study confirms the relative inde- & Pickering, 2004). The AWMA is a computerized
pendence of verbal and visuospatial short-term tool for assessing short-term and working memory in
memory from an executive component associated children aged 4 – 11 years. One benefit of the AWMA
with verbal working memory tasks, there are some is that it is designed to provide a practical and con-
limitations. In particular, there was no separate as- venient way for nonexpert assessors such as teachers
sessment of visuospatial working memory, because to screen their pupils for significant working mem-
until recently the processing-plus-storage span tasks ory problems, with a user-friendly interface. The
used to assess working memory capacities appro- automated presentation and scoring of tasks provide
priate for use with children, such as listening span, a consistency in presentation of stimuli across
counting span, and reading span, were exclusively participants, thus reducing experimenter error.
verbal in nature (Case, Kurland, & Goldberg, 1982; The tasks incorporated into the computerized
Daneman & Carpenter, 1980). battery were selected on the basis of providing reli-
There is considerable merit in including visuo- able and valid assessments of verbal and visuospa-
spatial working memory measures in a younger tial short-term and working memory in the relevant
population. For example, in older children (11- and research literature. In line with this substantial body
14-year-olds), Jarvis and Gathercole (2003) found a of prior evidence, verbal and visuospatial working
dissociation between verbal and visuospatial work- memory were measured using tasks involving sim-
ing memory systems in this age group. In adult ultaneous storage and processing of information,
studies, Kane et al. (2004) also found that verbal and whereas tasks involving only the storage of infor-
visuospatial working memory capacity was separ- mation were used to measure verbal and visuospa-
able. It is important to determine the extent to which tial short-term memory. The inclusion of visuospatial
these findings will generalize to younger children in working memory tasks in the present study allows
order to fully understand the cognitive mechanisms us to extend findings in previous studies (e.g.,
underpinning working memory. In the present Gathercole et al., 2004) to provide a comprehensive
study, the inclusion of suitable verbal and visuo- investigation of whether verbal and visuospatial
spatial short-term and working memory tasks, as memory skills develop independently across the
well as a large sample of children aged between childhood years.
4 and 11 years, allow us to track any developmental The multiple assessments of each memory com-
changes in the structure of memory. ponent taken for a large sample of children in this
A second aim was to investigate whether visuo- study provided the opportunity to use the method of
spatial short-term memory tasks draw more on ex- confirmatory factor analysis to test alternative mod-
ecutive resources than verbal short-term memory els of the underlying structure of working memory.
ones (e.g., Miyake et al., 2001; Shah & Miyake, 1996). Model 1 is a two-factor domain-specific model, with
These data from adult samples indicate that there latent constructs for all verbal and visuospatial
may be a common visuospatial construct that in- memory tasks. Model 2 has a two-factor structure
cludes both short-term and working memory tasks. corresponding to working memory and short-term
However, Kane et al. (2004) did not find any support memory, and is a fully domain-general model. The
for this position, and suggested instead that there are third model comprises four factors, with separate
distinct components associated with visuospatial domain-specific working memory and short-term
short-term and working memory tasks in adults (see memory constructs (e.g., Friedman & Miyake, 2000;
also Bayliss et al., 2003). Again, this is an area of Miyake et al., 2001). The fourth model contained
debate where it would be useful to identify the re- three factors: a single domain-general working
lationship between visuospatial memory skills and memory factor, and two separable storage factors for
executive resources in young children before we are verbal and visuospatial short-term memory; this
able to reach any clear conclusions about specific corresponds to the Baddeley and Hitch (1974)
theoretical frameworks of working memory. working memory model and the model advanced by
Participants were recruited from schools repre- Engle et al. (1999b; see also Conway, Cowan, Bunt-
senting a broad range of performance in national ing, Therriault, & Minkoff, 2002). Given the close
Verbal and Visuospatial Short-Term and Working Memory in Children 1701

links between working memory and academic at- less, only 7% had no academic or vocational quali-
tainment, it is important to establish whether it is fications), and the average amount of time the chil-
domain-general executive resources or domain-spe- dren in the present study attended preschool or
cific skills represented by verbal and visuospatial nursery was 2 years (77% attended for a length of
working memory tasks that contribute to learning. time or less).
One important issue addressed by the study is
whether the theoretical structure of working mem-
Procedure
ory is consistent across childhood, or is subject to
changes as a function of developmental growth. One Each child was tested individually in a quiet area
possibility is that younger children draw more on of the school for a single session lasting up to 40 min.
executive resources even to perform short-term All 12 memory measures were taken from the
memory tasks. This could be due to the fact that the AWMA (Alloway et al., 2004). Six of these measures
brain areas related to higher level cognition are still involved storage-plus-processing components and
developing, and while older children will have ac- are referred to as working memory tasks. Of these,
quired greater cognitive resources, younger children 3 measures tapped verbal ability and 3 tapped visuo-
will not have the benefit of these additional support spatial ability. The remaining 6 measures involved
systems (Nelson, 1995, 2000; see also Fuster, 2000). storage-only components and are referred to as
Another possibility is that the working memory short-term memory tasks; 3 tasks were verbal in
structure remains consistent across development, in nature and 3 were visuospatial.
line with previous findings from Gathercole et al. All tests were initially piloted on groups of chil-
(2004) using a more restricted set of measures. In the dren aged 4 – 5 and 9 – 10 years. As a result of this
present study, the fit of each theoretical model to the pilot, the tests were refined to ensure that both the
data was tested separately in each age group, in or- practice and the experimental trials were age ap-
der to identify potential changes in short-term and propriate. Extensive practice trials that included
working memory between 4 and 11 years. visual illustrations of the working memory tasks
were included, especially for the younger children.
In particular, it should be noted that the processing
Method
and storage components of all working memory
tasks were presented separated with multiple prac-
Participants
tice trials before these two elements of the task were
A total of 709 children from primary schools in the combined in practice.
northeast of England participated in the study. Par- Test reliability of the AWMA was assessed in a
ental consent was obtained for each child partici- subset of children (n 5 105) from the standardization
pating in the study. Participating schools were study randomly selected across schools. The age of
selected on the basis of the national average of per- children ranged from 59 to 140 months (mean 5 99
formance on Key Stage 2 assessments (national as- months, SD 5 24). A period of 4 weeks separated the
sessments in reading, writing, and mathematics) in two successive testing administrations. Test – retest
the year 2002, and represented a range of low reliabilities are reported with the description of each
(o65%), middle (75%), and high (485%) test results. test. A mixed analysis of variance on recall per-
The age of children ranged from 54 to 141 months formance at testing times 1 and 2 indicated no sig-
(mean 5 97 months, SD 5 24). The majority of fam- nificant difference as a function of age.
ilies served by the schools are of European heritage, The tests were administered in a fixed sequence
with roughly 15% from immigrant families repre- designed to vary task demands as widely as possible
senting countries such as Pakistan, Bangladesh, across successive tests and to reduce fatigue. All tests
China, Africa, and the Caribbean. were presented on a laptop computer with the screen
Information was provided by each child’s princi- resolution set to 600  480 pixels. The tests were
pal caregiver about maternal educational level (i.e., designed using Borland’s C11 Builder 5 (2004). For
GCSEs, A levels, vocational training, or higher edu- the spoken presentation of stimuli, audio files were
cation) and the age at which the mother left school. recorded using a minidisk player and then edited on
According to these socioeconomic indices, the aver- the GoldWave program (2004). All picture files were
age age at which the mother left school was 17 years created in Microsoft Powerpoint using the standard
(68% of the mothers had a level of education up to shape graphics.
this point or less), the average qualification was the In all of the following tests, the instructions were
final year of secondary school (54% achieved this or presented as a sound file while the computer screen
1702 Alloway, Gathercole, and Pickering

was blank. Practice trials followed the instructions. circles in an array and then recall the tallies of circles
The test trials were presented as a series of blocks; in the arrays that were presented. The test trial begins
each block consists of six trials. The experimenter with one visual array, and increases by an additional
recorded the child’s response using the right arrow visual array in each block, until the child is unable to
key on the keyboard ( ! ) for a correct response and correctly recall four trials. Each visual array stayed
the left arrow key on the keyboard ( ) for an in- on the computer screen until the child indicated that
correct response. The computer program automat- she or he had completed counting all the circles. If the
ically credits a correct trial with a score of 1. child made an error in counting the circles and re-
According to the ‘‘move on’’ rule, if a child responds called this incorrect sum, they were not penalized. In
correctly to the first four trials within a block of trials, the backwards digit recall task, the child is required to
the program automatically proceeds to the next recall a sequence of spoken digits in the reverse or-
block and gives credit for trials that were not ad- der. Test trials begin with two numbers, and increase
ministered. However, if three or more errors are by one number in each block, until the child is unable
made within a block of trials, the program stops the to recall four correct trials at a particular block. The
test and automatically returns to the main menu. The number of correct trials was scored for each child. For
score for that test reflects the number of correct re- children aged 4.5 and 11.5 years, test – retest reliabil-
sponses up to the point at which the test was ended. ity is .81, .79, and .64 for listening recall, counting
Performance on the processing component of the recall, and backward digit recall, respectively.
working memory tasks was monitored; however, It is worth noting that the present study included
errors were not included in the recall score. forward digit recall as a measure of verbal short-term
memory and backward digit recall as a measure of
verbal working memory. This decision was based on
Tests
findings that in forward digit recall, the processing
Verbal short-term memory. Three measures were load is minimal as children immediately recall num-
administered (modified from the Working Memory ber sequences. In contrast, in the backward digit recall
Test Battery for Children, WMTB – C; Pickering & task there is an added requirement to recall the digits
Gathercole, 2001). In the digit recall task, the child in reverse sequence that imposes a substantial pro-
hears a sequence of digits and has to recall each se- cessing load on the child as illustrated by the finding
quence in the correct order. In the word recall task, the that forward digit spans are higher than backward
child hears a sequence of words and has to recall digit spans (Isaacs & Vargha Khadem, 1989; see also
each sequence in the correct order. In the nonword Vandierendonck, Kemps, Fastame, & Szmalec, 2004).
recall task, the child hears a sequence of nonwords Correspondingly, short-term memory skills (such as
and has to recall each sequence in the correct order. forward digit recall) are much more weakly associ-
For children aged 4.5 and 11.5 years, test – retest re- ated with general academic and cognitive perform-
liability is .84, .76, and .64 for digit recall, word recall, ance than working memory skills as measured by
and nonword recall, respectively. backward digit recall (e.g., Daneman & Merikle, 1996;
Verbal working memory. The following three mea- see also Gathercole & Alloway, 2006, for a review).
sures were administered (also modified from the Visuospatial short-term memory. Three measures
WMTB – C; Pickering & Gathercole, 2001). In order to were administered. In the dot matrix task, the child is
ensure that the task demands were not too difficult shown the position of a red dot in a series of 4  4
for the younger children (see Gathercole et al., 2004), matrices and has to recall this position by tapping
the storage and processing components of the tasks the squares on the computer screen. The position of
were presented independently and then combined. each dot in the matrix is held on the computer for 2 s.
Practice trials could also be repeated to ensure com- The sequences were random with no location being
prehension of the task. In the listening recall task, the highlighted more than once within a trial. The re-
child is presented with a series of spoken sentences, maining two visuospatial short-term memory
has to verify the sentence by stating ‘‘true’’ or ‘‘false,’’ measures were modified for use in the present study
and recalls the final word for each sentence in se- from the WMTB – C (Pickering & Gathercole, 2001).
quence. Test trials begin with one sentence, and In the mazes memory task, the child is shown a maze
continue with additional sentences in each block with a red path drawn through it for 3 s. She or he
until the child is unable to recall three correct trials at then has to trace in the same path on a blank maze
a block. In the counting recall task, the child is pre- presented on the computer screen. In the block recall
sented with a visual array of red circles and blue task, the child views a video of a series of blocks
triangles. She or he is required to count the number of being tapped, and reproduces the sequence in the
Verbal and Visuospatial Short-Term and Working Memory in Children 1703

correct order by tapping on a picture of the blocks. memory tasks were defined as scores more than 3.5
The blocks were tapped at a rate of one block per standard deviations above or below the mean. Ten
second. For children aged 4.5 and 11.5 years, test – values out of the 8,496 in the data set met this cri-
retest reliability is .83, .81, and .83 for dot matrix, terion and were replaced with values corresponding
mazes memory, and block recall, respectively. to  3.5 standard deviations as appropriate. One
Visuospatial working memory. Three measures to multivariate outlier with a Mahanalobis d2 score
assess visuospatial working memory that are suit- (po.001) was eliminated. The final data set for sub-
able for school-aged children were developed. As sequent analyses consisted of 708 children.
with the verbal working memory tasks, the storage
and processing components of the tasks were pre-
Descriptive Statistics
sented independently and then combined. In the first
task, the odd-one-out task (adapted from Russell, The mean raw scores for each measure as a func-
Jarrold, & Henry, 1996), the child views three shapes, tion of single year bands and gender are provided in
each in a box presented in a row, and identifies the Table 1. In all cases, improvements in performance
odd-one-out shape. At the end of each trial, the child were observed across the age bands. A series of
recalls the location of each odd-one-out shape, in the multivariate analyses of variance (MANOVAs) were
correct order, by tapping the correct box on the performed on each of the three measures associated
screen. Each array is presented on the computer with the different working memory components as a
screen for 2 s. Test – retest reliability is .81 for children function of age in years (4 – 11 years) and gender. The
aged 4.5 and 11.5 years. MANOVA performed on the three verbal short-term
In the Mr. X task (adapted from Hamilton, Coates, memory tasks yielded a highly significant effect of
& Heffernan, 2003), fictitious cartoon figures known age, F(2, 702) 5 52.44, po.01, but no significant effect
as Mr. X were designed to be unfamiliar yet likable to of gender, F(1, 702) 5 1.96, p4.05, and no significant
children. The child is presented with a picture of two interaction between age and gender, F(2, 702) 5 1.73,
Mr. X figures. The child identifies whether the Mr. X p4.05. The same pattern of significance was ob-
with the blue hat is holding the ball in the same hand served in the MANOVAs performed on the tasks
as the Mr. X with the yellow hat. The Mr. X with the for verbal working memoryFage: F(2, 702) 5 75.05,
blue hat may also be rotated. At the end of each trial, po.01; gender: F(2, 702) 5 1.26, p4.05; interaction:
the child has to recall the location of each ball in Mr. F(2, 702) 5 2.52, p4.05, visuospatial short-term
X’s hand in sequence, by pointing to a picture with memoryFage: F(2, 702) 5 78.52, po.01; gender:
eight compass points. Both the Mr. X figures and the F(2, 702) 5 1.95, p4.05; interaction: F(2, 702) 5 1.27,
compass points stayed on the computer screen until p4.05, and visuospatial working memoryFage:
the child provided a response. Test – retest reliability F(2, 702) 5 75.91, po.01; gender: F(2, 702) 5 1.64,
is .77 for children aged 4.5 and 11.5 years. p4.05; interaction: F(2, 702) 5 1.22, p4.05. The age
In the spatial span task (adapted from Jarvis & effects across all measures reflect the increasing
Gathercole, 2003), the child views a picture of two memory capacity as children get older.
arbitrary shapes, where the shape on the right has a This pattern of performance growth as a function
red dot on it. The child identifies whether the shape of age is demonstrated in Figure 1, which plots the z
on the right is the same as or opposite to the shape on scores for each year group from 4 to 11 years; these
the left. The shape with the red dot may also be ro- scores were calculated on the basis of the entire
tated. At the end of each trial, the child has to recall sample of children. All 12 tests indicate broadly
the location of each red dot on the shape in sequence, similar developmental functions, with performance
by pointing to a picture with three compass points. increasing across each year group. The main excep-
Both the shapes and the compass points stayed on tion to this pattern can be observed with respect to
the computer screen until the child provided a re- the verbal short-term memory tasks, where per-
sponse. Test – retest reliability is .82 for children aged formance leveled off between 10 and 11 years.
4.5 and 11.5 years. One of the principal aims of the present study was
to investigate the structural organization of working
memory across the range of ages of the children
Results
participating in this study. For this purpose, the
children were grouped into three age bands con-
Data Screening
sisting of more than 200 children each to provide
The data were screened for university and multi- satisfactory and comparable statistical stability.
variate outliers. Univariate outliers on each of the 12 Means for each age band are summarized in Table 2.
1704 Alloway, Gathercole, and Pickering

Spatial
Correlational Analyses

6.51
6.42
9.09
9.54

11.82
Visuospatial working memory

10.37

13.30
13.74
15.21
15.06
17.37
17.48
18.53
17.56
span
Correlations among all variables were conducted
on the full age range (rather than age bands) using
the raw task scores. Zero-order correlations are dis-
Mr. X

10.82

12.16
14.12
13.46
4.31
4.16
6.07
5.42
7.24
7.80
9.54
9.09
11.00

11.08
played in the lower triangle in Table 3. The inter-
correlations between measures purportedly tapping
the different working memory components were
Odd-one

10.95

13.68
13.84
15.43
17.17
17.02
18.29
18.35
19.24
19.15
8.57
8.82

11.13
11.86 substantial in magnitude, with rs ranging from .56 to
-out

.73 for the verbal short-term memory tasks, .69 to .74


for the verbal working memory tasks, .68 to .75 for
the visuospatial short-term memory tasks, and .64
Block
recall

11.79
12.00

14.28
15.00
15.84
16.39
17.39
18.08
20.00
19.16
19.74
21.48
22.02
21.92
to .73 for the visuospatial working memory tasks
Visuospatial short-term

(po.001 probability level in each case). As none of


the zero-order correlations were higher than .80,
memory

multicollinearity was not a problem in this data set


memory

Mazes

13.21
13.54
14.17
18.02
16.12
18.95
18.87
21.02
18.51
5.65
6.03
8.74
8.81
11.98

(Kline, 1998). It is, however, worth noting that these


coefficients are inflated by the large age variation in
this group. In order to adjust for this, a partial cor-
matrix

relation analysis with age in months partialed out


12.82
12.87
14.37
15.25
16.53
17.18
17.98
18.91
20.32
20.08
21.84
22.54
23.37
22.36
Dot

was calculated. These are shown in the upper tri-


angle of Table 3. The intercorrelations between
working memory measures remained moderate to
digit recall
Backward

high even after age is partialed out (rs ranging from


4.82
4.68
6.79
7.38
8.82
9.36
9.62

11.38
11.98
10.28

12.74
12.83
14.12
13.84

.35 to .62). The within-construct coefficients were


Verbal working memory

higher than between-construct coefficients, suggest-


ing good internal validity of the measures purport-
edly tapping four subcomponents of working
Counting
recall

memory.
8.98
9.00
12.07
12.73
13.22
15.48
15.72
17.21

18.69

20.86
22.37
21.56
19.11

20.11

Zero-order correlations for children in the age


band 4 – 6 years are displayed in the lower triangle of
Table 4, while zero-order correlations for children in
Listening

the age band 7 – 8 years are displayed in the upper


recall

4.43
4.37
6.26
6.67
7.67
8.54
8.74
9.81

11.87
10.74
10.98

12.05
13.08
12.23

triangle of Table 4. Zero-order correlations for chil-


dren in the age band 9 – 11 years are displayed in
Table 5. The intercorrelations between measures
purportedly tapping the different memory compon-
Nonword
recall

8.69
8.53
9.00
9.48
8.73

11.28
10.96
10.92

12.30
12.45
13.42
13.84
13.57
13.28
Verbal short-term memory

ents were robust, and in most cases there were


higher within-construct correlation coefficients than
between-construct coefficients. This was true for all
three age bands, and reflects the pattern of relation-
Word
recall

12.10
12.47
13.37
14.19
14.43
15.66
16.13
17.17
17.70
18.08
19.58
19.27
19.51
19.72

ships for the full data set as shown in Table 3.


Mean Test Scores as a Function of Age and Sex

recall

21.29
20.24
23.33
23.98
23.88
25.20
25.05
26.02
26.89
27.06
28.45
28.73
29.84
28.44

Confirmatory Factor Analyses


Digit

The degree to which the data fitted alternative


models of short-term and working memory was
49
38
43
48
51
56
61
53
47
49
38
63
51
61
N

tested formally using confirmatory factor analysis


(Bentler, 2001; Bentler & Wu, 1995). This method
Sex

provides a means of testing the adequacy of com-


M

M
F

peting theoretical accounts of the relationships be-


tween measures, with each model specified in terms
Table 1

(years)

of paths between observed variables and latent


Age

10.5
4.5

5.5

6.5

7.5

8.5

9.5

constructs, and between constructs. A commonly


Verbal and Visuospatial Short-Term and Working Memory in Children 1705

2.00
Verbal short-term memory Visuo-spatial STM
2.00

1.50 1.50

1.00 1.00
Mean z-score

Mean z-score
.50 .50

0.00 0.00

− .50 − .50
Digit recall Dot matrix

− 1.00 Word recall −1.00 Mazes memory

Nonword recall Block recall


− 1.50 −1.50
4.5 5.5 6.5 7.5 8.5 9.5 10.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5

2.00
Visuo-spatial working memory
Verbal working memory
2.00

1.50
1.50

1.00
1.00
Mean z-score
Mean z-score

.50
.50

0.00
0.00

− .50 − .50
Listening recall Odd-one-out

− 1.00 Counting recall −1.00 Mr X

Backward digit recall Spatial span


− 1.50 −1.50
4.5 5.5 6.5 7.5 8.5 9.5 10.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5

Figure 1. Mean z scores as a function of age for each of the 12 measures, grouped by task type.

used index of goodness of fit for each model is the good fit. Further assessment of the extent to which
chi-square statistic, which compares the degree to the specified model approximates to the true model
which the predicted covariances in the model differ is the root mean square error of approximation
from the observed covariances. A good fit is deter- (RMSEA). An RMSEA value of .08 or lower is ac-
mined by small and nonsignificant chi-square val- ceptable, and a value below .05 indicates a good fit
ues. Because this statistic is sensitive to variances in (see McDonald & Ho, 2002).
sample sizes, with very large samples as in the In the series of models tested, paths between la-
present study even the best-fitting models frequently tent constructs were left free to covary (represented
yield significant chi-square values (Kline, 1998). diagrammatically as bidirectional links) in the ab-
Model adequacy was therefore evaluated using ad- sence of justifiable assumptions concerning direction
ditional global fit indices that are more sensitive to of causality. Such models are known as measurement
model specification than to sample size (Jaccard & models. In each case, the level of significance of the
Wan, 1996; Kline, 1998). Fit indices such as the path weights between each observed variable and its
comparative fit index (CFI; Bentler, 1990), the bollen associated factor, and the correlations between all
fit index (IFI; Bollen, 1989), and the normed fit index pairs of factors, was set at an a level of .05. Statistical
(NFI; Bentler & Bonett, 1980) provide a further comparison of models was achieved by performing
measure of fit computed by comparing the hy- chi-square difference tests (McDonald & Ho, 2002).
pothesized model against a null model in which the The statistics and fit indices generated by each of the
relations between the latent variables are not speci- measurement models are summarized in Table 6. In
fied and consequently are set at 0. Fit indices with order to identify the best-fitting model, a series of
values equal to or higher than .90 demonstrate a chi-square difference tests were performed and are
1706 Alloway, Gathercole, and Pickering

Table 2
Descriptive Statistics for All Test Scores as a Function of Age Band

4 – 6 years 7 – 8 years 9 – 11 years All: 4 – 11 years


(n 5 285) (n 5 210) (n 5 213) (n 5 708)

Measure M SD M SD M SD M SD

Verbal short-term memory


Digit recall 23.14 4.18 26.18 3.63 28.86 4.85 25.76 4.87
Word recall 13.81 3.32 17.20 3.25 19.51 3.67 16.53 4.17
Nonword recall 9.30 3.09 11.68 3.22 13.54 3.30 11.28 3.65
Verbal working memory
Listening recall 6.46 3.17 9.98 2.91 12.31 3.03 9.27 3.93
Counting recall 12.12 4.27 17.55 4.70 21.29 4.59 16.49 5.93
Backward digit recall 7.14 3.55 10.73 3.74 13.41 3.85 10.09 4.54
Visuospatial short-term memory
Dot matrix 14.99 3.64 19.23 3.47 22.56 3.70 18.53 4.81
Mazes memory 9.32 5.69 15.30 5.23 19.30 4.69 14.10 6.74
Block recall 14.37 3.88 18.56 3.57 21.42 3.96 17.74 4.83
Visuospatial working memory
Odd-one-out 10.99 3.69 15.73 4.61 18.78 4.51 14.74 5.36
Mr. X 5.95 3.25 10.05 4.03 12.81 3.99 9.23 4.71
Spatial span 9.13 4.30 14.25 4.46 17.73 4.03 13.24 5.60

reported following the presentation of the different structs. The first factor was associated with all six
model solutions. verbal memory measures, while Factor 2 was asso-
First, models were tested on the full data set for all ciated with the six visuospatial ones. The model is
ages. The input to each model was the partial cor- summarized in Figure 2, and the fit statistics are
relation matrix, with age partialed out. Once the shown in Table 4. This two-factor model does not
best-fitting model was identified, the fit of this model provide a good fit to the data; the chi-square value is
was also tested for each individual age band. In these highly significant and the fit indices are less than .90.
analyses also, the partial correlation matrix, con- In Model 2, the first factor was associated with all
trolling for age, was used as the input. six working memory measures and the second factor
Model 1 is a two-factor model that distinguishes was associated with the six short-term memory ones.
only between verbal and visuospatial memory con- The model is summarized in Figure 3, and the fit

Table 3
Correlations Between All Memory Scores; Partial Correlations (Controlling for Age in Months) in Upper Triangle

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Age (months) F
2. Digit recall .54 F .47 .40 .41 .36 .45 .34 .26 .28 .30 .19 .30
3. Word recall .60 .64 F .62 .31 .35 .34 .28 .32 .25 .30 .17 .23
4. Nonword recall .49 .56 .73 F .30 .32 .31 .27 .35 .26 .30 .20 .24
5. Listening recall .69 .63 .60 .53 F .49 .48 .35 .34 .33 .41 .37 .45
6. Counting recall .71 .60 .63 .55 .74 F .43 .42 .43 .43 .53 .41 .47
7. Backward digit recall .65 .64 .60 .53 .71 .69 F .35 .30 .34 .42 .35 .32
8. Dot matrix .70 .58 .58 .51 .67 .71 .65 F .42 .54 .37 .31 .36
9. Mazes memory .68 .53 .60 .56 .65 .70 .61 .70 F .41 .38 .31 .36
10. Block recall .67 .54 .55 .50 .64 .70 .63 .75 .68 F .39 .33 .40
11. Odd-one-out .66 .55 .58 .52 .68 .75 .67 .67 .66 .66 F .35 .49
12. Mr. X .66 .48 .50 .46 .66 .69 .63 .63 .62 .63 .64 F .51
13. Spatial span .69 .56 .55 .49 .71 .73 .62 .67 .66 .68 .73 .73 F
Verbal and Visuospatial Short-Term and Working Memory in Children 1707

Table 4
Correlations Between all Memory Scores for Age Band 4 – 6 Years in Lower Triangle (n 5 285) and for Age Band 7 – 8 Years in Upper Triangle (n 5 210)

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Age (months) F .23 .17 .19 .29 .30 .25 .25 .28 .25 .27 .23 .25
2. Digit recall .39 F .44 .42 .48 .38 .50 .29 .27 .37 .32 .26 .23
3. Word recall .36 .54 F .60 .39 .33 .37 .26 .28 .28 .29 .24 .17
4. Nonword recall .16 .38 .55 F .42 .33 .39 .30 .38 .41 .35 .28 .19
5. Listening recall .51 .48 .37 .25 F .53 .54 .40 .39 .45 .51 .43 .29
6. Counting recall .54 .50 .48 .34 .61 F .42 .43 .44 .46 .59 .46 .30
7. Backward digit recall .52 .48 .44 .29 .58 .61 F .43 .33 .36 .44 .45 .25
8. Dot matrix .49 .46 .37 .21 .50 .60 .49 F .46 .45 .42 .32 .25
9. Mazes memory .52 .39 .40 .33 .48 .59 .48 .53 F .49 .44 .36 .28
10. Block recall .47 .35 .33 .19 .44 .57 .50 .66 .50 F .47 .42 .48
11. Odd-one-out .51 .46 .45 .34 .54 .65 .57 .48 .51 .47 F .50 .59
12. Mr. X .45 .30 .23 .22 .52 .55 .40 .49 .47 .48 .47 F .57
13. Spatial span .45 .40 .30 .22 .54 .57 .41 .44 .44 .45 .53 .57 F

Note. For ages 4 – 6 years, all coefficients 4.18 are significant at the .001 level; for ages 7 – 8 years, all coefficients 4.23 are significant at the
.001 level.

statistics are summarized in Table 4. This model also correlation between Factors 2 (verbal working
did not provide a satisfactory fit of the data: the chi- memory) and 4 (visuospatial working memory) was
square value (po.001) is highly significant, all fit very high (.91), with 83% of their variance shared. In
indices are less than .90, and the RMSEA value ex- contrast, Factors 1 (verbal short-term memory) and 3
ceeds .08. (visuospatial short-term memory) only shared 33%
Model 3a consists of four domain-specific con- of their variance. This suggests that there is a greater
structs: verbal short-term memory (Factor 1), verbal degree of domain specificity in the short-term
working memory (Factor 2), visuospatial short-term memory tasks than the working memory ones. In
memory (Factor 3), and visuospatial working mem- order to confirm statistically that verbal and visuo-
ory (Factor 4). This model provided a good fit of the spatial working memory were more strongly corre-
data (see Figure 4). The chi-square value is lower lated than were verbal and visuospatial short-term
than the other models (although still highly signifi- memory, the path between the working memory
cant, po.001), all fit indices are above .90, and the constructs was fixed to be equal to the path between
RMSEA 5 .07. It should, however, be noted that the the short-term memory constructs. The fit indices for

Table 5
Correlations Between All Memory Scores for Age Band 9 – 11 Years

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Age (months) F
2. Digit recall .14 F
3. Word recall .08 .51 F
4. Nonword recall .03 .48 .73 F
5. Listening recall .20 .46 .36 .35 F
6. Counting recall .24 .39 .38 .37 .52 F
7. Backward digit recall .21 .52 .38 .37 .51 .47 F
8. Dot matrix .11 .39 .34 .39 .37 .43 .35 F
9. Mazes memory .14 .32 .45 .44 .35 .46 .31 .47 F
10. Block recall .17 .33 .29 .30 .34 .47 .38 .61 .42 F
11. Odd-one-out .16 .32 .31 .29 .38 .52 .44 .42 .41 .42 F
12. Mr. X .27 .23 .21 .19 .39 .43 .42 .32 .33 .31 .29 F
13. Spatial span .12 .35 .32 .31 .51 .53 .37 .44 .45 .48 .51 .53 F

Note. All coefficients 4.20 are significant at the .001 level.


1708 Alloway, Gathercole, and Pickering

Table 6 with Model 3a resulted in a significant decrease in


Goodness-of-Fit Statistics for the Different Measurement Models for the model fit (Dw2 5 54.85, df 5 1, po.001). This in-
Each Age Band dicates that working memory capacity is a more
domain-general construct than short-term memory.
Age
Model band w2 df p CFI IFI NFI RMSEA
The next issue we investigated was whether vis-
uospatial memory tasks require greater involvement
1 All 450.55 53 o.001 .86 .86 .85 .103 of the central executive component than do verbal
4–6 522.59 53 o.001 .82 .82 .80 .112 memory tasks. In Model 3a, the correlation between
7–8 426.33 53 o.001 .87 .87 .86 .100 visuospatial short-term and working memory was
9 – 11 683.15 53 o.001 .82 .82 .81 .130 only slightly higher than that between verbal short-
2 All 441.96 53 o.001 .86 .87 .85 .102 term and working memory (.78 compared with .68).
4–6 510.45 53 o.001 .82 .82 .80 .110
In Model 3c, the path between visuospatial short-
7–8 452.31 53 o.001 .87 .87 .85 .103
term and working memory constructs was fixed to
9 – 11 666.66 53 o.001 .82 .82 .81 .128
3a All 192.24 48 o.001 .95 .95 .94 .065 be equal to the path between the verbal short-term
4–6 273.89 48 o.001 .91 .91 .90 .082 and working memory constructs, and the fit values
7–8 215.45 48 o.001 .94 .94 .93 .070 were good. The fit indices for this Model 3c, are also
9 – 11 337.78 48 o.001 .92 .92 .90 .092 included in Table 4. A chi-square difference test
3b All 247.09 49 o.001 .93 .93 .92 .076 comparing Model 3c with Model 3a indicated that
4–6 351.44 49 o.001 .88 .88 .87 .090 the model fit was not significantly reduced
7–8 247.99 49 o.001 .93 .93 .92 .076 (Dw2 5 4.50, df 5 1, p4.01). This suggests that visuo-
9 – 11 383.76 49 o.001 .90 .90 .89 .098 spatial short-term memory tasks do not tap executive
3c All 196.74 49 o.001 .95 .95 .93 .065 resources any more than verbal short-term memory
4–6 276.04 49 o.001 .91 .91 .89 .080
ones.
7–8 216.29 49 o.001 .94 .94 .93 .069
9 – 11 348.00 49 o.001 .91 .91 .90 .093
While Model 3a seems to provide an adequate fit
4a All 226.52 51 o.001 .94 .94 .92 .070
to the data, there are concerns of multicollinearity
4–6 288.03 51 o.001 .91 .91 .89 .081 due to the high correlation between Factors 2 (verbal
7–8 292.14 51 o.001 .92 .92 .90 .082 working memory) and 4 (visuospatial working
9 – 11 382.22 51 o.001 .90 .90 .89 .096 memory). In order to address this problem, we tested
4b All 85.37 45 o.001 .98 .98 .96 .062 a three-factor model consisting of the following: a
4–6 297.24 45 o.001 .90 .90 .89 .089 domain-general construct incorporating both the
7–8 193.00 45 o.001 .95 .95 .94 .068 verbal and visuospatial storage-plus-processing
9 – 11 326.25 45 o.001 .92 .92 .91 .094 tasks, a verbal short-term memory construct, and a
visuospatial short-term memory construct (see Fig-
ure 5). This model is consistent with the tripartite
this Model 3b, are included in Table 4. A chi-square working memory model of Baddeley and Hitch
difference test comparing the fit of this model (3b) (1974), as well as that of Engle et al. (1999b). The

Dot matrix
Digit recall

Mazes memory
Word recall

Block recall
Nonword recall .85
Verbal Visuo-spatial
memory memory
Listening recall Odd-one-out

Counting recall
Mr X

Backward digit Spatial span


recall

Figure 2. Path model for the two-factor model based on the distinction between verbal and visuospatial memory skills (Model 1). For
clarity, only path coefficients between latent variables for the full data set are shown.
Verbal and Visuospatial Short-Term and Working Memory in Children 1709

Listening recall Digit recall

Counting recall Word recall

Backward digit Nonword recall


recall .85
Working Short-term
memory memory
Odd-one-out Dot matrix

Mr X
Mazes memory

Spatial span Block recall

Figure 3. Path model for the two-factor model based on a domain-specific view of short-term and working memory (Model 2). For clarity,
only path coefficients between latent variables for the full data set are shown.

model is summarized in Figure 5, and the model between both the processing and storage aspects of
statistics are shown in Table 4. This model (4a) pro- the verbal and visuospatial working memory tasks
vided a better fit to the data than either Models 1 or was represented by one factor. The chi-square value
2, with a lower chi-square value (although still was considerably lower and all fit indices were
highly significant, po.001). All fit indices were .94 or above .90, and the RMSEA 5 .06. In addition to
above, and the RMSEA 5 .05, indicating a satisfac- providing a good account of the data, the correl-
tory fit. ations between latent constructs were also satisfac-
In order to apply a current theoretical model of tory. The fit indices for this model are included in
working memory based on adult data to the present Table 4 (Model 4b). A chi-square difference test
data from children (e.g., Conway et al., 2002), a final comparing the fit of this model with Models 3a
model (4b) was tested. In this, the variance shared by (Dw2 5 141.15, df 5 3, po.001) and 4a (Dw2 5 106.87,
the processing aspect of verbal and visuospatial df 5 6, po.001) confirmed that this model provides
working memory tasks was represented by a com- the best account of these data. This indicates that
mon factor, while the storage element involved in while verbal and visuospatial working memory
these tasks loaded on both this factor as well as do- tasks share the involvement of executive functioning
main-specific constructs for the verbal and visuo- (Baddeley, 2000) or controlled attention (Engle et al.,
spatial short-term memory tasks. This model 1999b), the storage requirements of these tasks are
contrasts with Model 4a, where shared variance separable. The model solution for the full data set is

Digit recall
Listening recall
.68
Counting recall Verbal working Verbal short- Word recall
memory term memory

Backward digit .80 .50 Nonword recall


recall
.91
.57
Odd-one-out Dot matrix

Mr X Visuo-spatial Visuo-spatial Mazes memory


WM memory STM
.78
Spatial span Block recall

Figure 4. Path model based on distinct verbal and visuospatial working memory components and domain-specific storage components
(Model 3a). For clarity, only path coefficients between latent variables for the full data set are shown.
1710 Alloway, Gathercole, and Pickering

Listening recall Counting recall Odd-one-out Mr X

Backward digit Spatial span


recall

Working
memory

.81
Digit recall .61
Dot matrix

Word recall Verbal short- Visuo-spatial Mazes memory


term memory short-term
memory
Nonword recall .57
Block recall

Listening recall Odd-one-out

Backward Working Mr X
digit recall memory
Spatial span
Counting recall

.50 .80
Dot matrix
Digit recall

Verbal STM Visuo-spatial Block Recall


Word recall
STM
.58
Mazes
Nonword recall Memory

All ages
Figure 5. Path model for the three-factor model based on the tripartite model of working memory, with a distinction between verbal short-
term and working memory, and a separate factor for visuospatial memory (Model 4a). For clarity, only path coefficients between latent
variables for the full data set are shown.

summarized in Figure 6. For reasons of parsimony for Models 3a and 3c. Across the three age bands,
and theoretical coherence (given the multicollinear- Model 4b provided a very good fit with the data.
ity problem of Model 3a), this model is accepted as Two chi-square difference test comparing the fit of
providing a suitable fit to the data. this model with Models 3a and 4a confirm that this
A series of confirmatory factor analyses was con- three-factor order model provides the best account of
ducted to test the fit of each of these different these data. The model comparison statistics are de-
measurement models across the age range repre- scribed in Table 7. The path coefficients between each
sented in the present study. The partial correlation of the latent constructs in Model 4b across each of
matrix was used as the input for three different age three age bands are summarized in Figure 6.
bands: 4 – 6, 7 – 8, and 9 – 11 years. The fit indices for It is worth noting that in Model 4b, the path be-
each of these models are provided in Table 4. Of tween the working memory construct and visuo-
specific interest is whether the components of Model spatial short-term memory was very high for the
4b are equivalent across the three age bands. The 4- to 6-year-olds (.97). It is possible that these children
findings are similar to that of the model for the full were reliant solely on executive resources to perform
data set. Models 1 and 2 did not provide a good fit the visuospatial tasks. In order to test this, we fixed the
with the data, and multicollinearity was a problem correlations between working memory and visuospatial
Verbal and Visuospatial Short-Term and Working Memory in Children 1711

Listening recall Odd-one-out

Backward Working Mr X
digit recall memory
Spatial span
Counting recall
.32
.97
Dot matrix
Digit recall

Verbal STM Visuo-spatial Block Recall


Word recall
STM
.55
Mazes
Nonword recall Memory

4 to 6 years

Listening recall Odd-one-out

Backward Working Mr X
digit recall memory
Spatial span
Counting recall

.51 .82
Dot matrix
Digit recall

Verbal STM Visuo-spatial Block Recall


Word recall
STM
.63
Mazes
Nonword recall Memory

7 to 8 years

Listening recall Odd-one-out

Backward Working Mr X
digit recall memory
Spatial span
Counting recall

.43 .71
Dot matrix
Digit recall

Verbal STM Visuo-spatial Block Recall


Word recall
STM
.62
Mazes
Nonword recall Memory

9 to 11 years
Figure 6. Four path models for each age band (4 – 6, 7 – 8, 9 – 11 years, and the full data set) for the three-factor model based on a common
processing factor for verbal and visuospatial working memory tasks and domain-specific factors for verbal and visuospatial storage
(Model 4b). All factor loadings are significant at the po.05 level.

short-term memory at 1.0 to indicate a perfect cor- children (7 – 8, 9 – 11 years, and the full data set),
relation. The data indicate no significant decrease in there was a significant decrease in the model fit. This
the fit of the model when the correlation was set to finding is important as it indicates that for the vis-
1.0 for the 4- to 6-year-olds. However, for the older uospatial short-term memory tasks, the younger
1712 Alloway, Gathercole, and Pickering

Table 7 performance in verbal short-term memory tasks


Model Comparison Statistics for Each Age Band Between Models 4b and leveled off sooner than in other tasks. No sex dif-
3a, and 4b and 4a ferences were found in any of the tasks. The corre-
lation analyses indicate stronger relationships
Model 3a Model 4a
between measures purported to tap the same
Age band Dw2 df p Dw2 df p underlying memory component than with measures
of other memory components. As careful consider-
All 141.15 3 o.001 106.87 6 o.001
ation was given to design of the memory tasks, the
4 – 6 years  23.35 3 o.001  9.21 6 ns
theoretical structure appears to reflect underlying
7 – 8 years 22.45 3 o.001 99.14 6 o.001
9 – 11 years 11.53 3 o.01 55.97 6 o.001
differences in the cognitive processes associated with
these different measures rather than task-specific
Note. ns 5 not significant. Dw2 5 w2 difference. similarities and differences. While there were dif-
ferences in surface features across the tasks, an im-
portant underlying aspect was that all tasks
children base their performance on resources asso- provided a measure of the capacity of memory, ra-
ciated with the central executive (or controlled at- ther than a reflection of knowledge or information
tention) to a larger extent than older children do. acquired by the child. For example, in the verbal
short-term memory tasks, both word and number
tasks were used in order to avoid performance re-
Discussion
flecting knowledge-based skills in only literacy or
The purpose of the present study was to investigate numeracy. Also, consistent with research on working
whether the structure of working memory is con- memory, the verbal and visuospatial short-term
sistent across the childhood years. The first issue memory tasks involved only storage of information,
addressed in the study was whether working mem- while the verbal and visuospatial working memory
ory capacity in childhood could be viewed as a do- measures involved simultaneous processing and
main-general construct that coordinates separate storage of information. For example, in the listening
codes for verbal and visuospatial storage or whether recall task, the child judged the veracity of the sen-
it is a domain-specific construct with distinct re- tence and then recalled the final word; in the spatial
sources for verbal and visuospatial information. The span task, the child determined whether one shape
second issue was whether visuospatial short-term was a mirror image of the other and then recalled the
memory tasks draw more on executive resources location of a red dot. The findings from the present
than verbal short-term memory ones. A key feature study support the idea that any differential patterns
of the present study is that it is one of the few studies of performance reflect the underlying cognitive
to examine these theoretical issues of the structure mechanisms responsible for performance rather than
and interrelationships of working memory com- the nature of the tasks.
ponents in childhood using multiple assessments, In order to address the issue of domain generality
including tasks for visuospatial working memory, in versus specificity of working memory capacity, sev-
a large sample with a wide age range. It extends eral key theoretical models were compared using
previous developmental research such as that of confirmatory factor analysis. A three-factor model,
Gathercole et al. (2004; see also Bayliss et al. 2003, with related but separable constructs representing
2005) where the theoretical structure of working measures of verbal and visuospatial storage and a
memory from 6 years old was examined. In the third factor representing the shared variance be-
present study, the inclusion of children as young as 4 tween the verbal and visuospatial working memory
years and visuospatial working memory measures tasks (Model 4b), provided the best account of the
such as the odd-one-out, Mr. X, and spatial span interrelationships between measures of short-term
provides a better understanding of the theoretical and working memory across the different age bands.
structure of memory in childhood and how it de- Thus, the interrelationships between measures of
velops in this crucial period. working memory and short-term memory appeared
The developmental functions from 4 to 11 years to be stable across this developmental period, and
for each aspect of memory tapped by the AWMA also suggested that this underlying cognitive struc-
(verbal short-term and working memory and vis- ture for memory is in place as young as 4 years.
uospatial short-term and working memory) were Although there are domain-specific aspects related
comparable, showing steady improvements in ac- to short-term memory capacity associated with
curacy across age groups. One exception was that verbal and visuospatial codes, working memory
Verbal and Visuospatial Short-Term and Working Memory in Children 1713

capacity is primarily driven by a domain-general populations (e.g., Gathercole et al., 2004) and in adult
mechanism. This is in line with the view that the samples (e.g., Kane et al., 2004; Park et al., 2002) that
processing aspect of the tasks shares the involvement verbal short-term memory tasks consist of a storage-
of executive functioning (Baddeley, 2000) or con- only component, whereas working memory mea-
trolled attention (Engle et al., 1999a). It is worth sures require executive resources for the processing
noting that there is some evidence of development aspect of the task. This pattern is consistent across all
change; in particular, links between the domain- three age groups, indicating no developmental
general processing construct and the domain-spe- change in the relationship between verbal short-term
cific visuospatial construct were higher in the 4 – 6 and working memory.
age band compared with the other two age groups, a With respect to the visuospatial domain, the re-
point that we will return to later. sults indicate differences in the developmental pat-
An alternative model consistent with the view of terns. Specifically, the data reveal that the link
domain-specific constructs for verbal and visuospa- between the domain-specific visuospatial construct
tial memory measures (e.g., Friedman & Miyake, and the domain-general processing construct was
2000; Miyake et al., 2001) was also tested (Model 3a). higher in the 4 – 6 age band compared with the other
However, in this model the correlation between two age groups. Indeed, when the correlation be-
verbal and visuospatial working memory was very tween these two variables was fixed to represent a
high and they shared a large amount of variance perfect association, there was no significant decrease
(83%). This suggests that verbal and visuospatial in the fit of this model for the youngest age group.
working memory capture more common underlying However, this was not the case for the older age
cognitive skills than verbal and visuospatial short- groups, confirming that young children draw more
term memory tasks. In adult studies reporting dis- on executive resources (or controlled attention) than
tinguishable verbal and spatial working memory older children when performing the visuospatial
constructs, Kane et al. (2004) point out that these short-term memory tasks (e.g., Cowan et al., 2005).
constructs also share 70 – 90% of their variance (e.g., One explanation for this finding can be drawn
Kyllonen, 1993; Oberauer, Sü, Schulze, Wilhelm, & from the idea of developmental fractionation (Hitch,
Wittmann, 2000). It is then perhaps unsurprising that 1990), that is cognitive mechanisms can develop at
a similar pattern was observed in our developmental different rates. Using tasks of visuospatial short-term
population. This proposal fits in with the view of the memory, Pickering, Gathercole, Hall, and Lloyd
central executive, which is responsible for control- (2001) found evidence of a developmental fraction-
ling resources and monitoring information process- ation, which they proposed was the result of whether
ing (Baddeley, 2000). It is also compatible with the the tasks had a static or dynamic presentation for-
view that controlled attention works to keep task- mat. There is growing evidence to suggest that ex-
relevant information active in working memory (e.g., ecutive functions support the more dynamic aspects
Engle et al., 1999a). of visuospatial tasks, including tracking visual se-
A strength of the present study was the hetero- quences and transforming visuospatial images (e.g.,
geneity of the population sampled, which was Duff & Logie, 1999; Logie, 1995). In the present study,
drawn from a full range of demographic back- the visuospatial short-term memory tasks included
grounds and academic attainment. One explanation dynamic formats, such as the perceptuo-motor
for the finding of domain specificity in adult samples tracking of dots and block in the dot matrix and
could be the result of a homogeneous population, a block recall tasks, respectively. Thus, the high asso-
point made by Shah and Miyake (1996) to explain the ciation between visuospatial short-term memory
domain specificity in their data. Kane et al. (2004) tasks and working memory for younger children in
also suggested that when participants are pooled the present study fits in well with the idea that dy-
from selective universities, the range of cognitive namic formats of visuospatial tasks involve execu-
abilities is restricted, and it is therefore more useful tive functions as well. This is an important finding as
to recruit from a broad participant pool in order to very few developmental studies have explored this
gain a more general understanding about working relationship between visuospatial short-term and
memory skills. working memory across a range of ages, including
The second research question addressed the issue children as young as 4 years old (however, for re-
of whether visuospatial short-term memory tasks search on older children, see Bayliss et al., 2003;
draw more on executive resources than verbal ones. Jarvis & Gathercole, 2003; Pickering et al., 1998).
Looking at the verbal domain first, the findings are A related issue regarding the theoretical structure
consistent with previous research in developmental of visuospatial memory concerns the extent to which
1714 Alloway, Gathercole, and Pickering

it is visual or spatial. Neuropsychological evidence data. However, it does provide some initial support
provides evidence for an impaired spatial but pre- for development of strategies as a function of age.
served visual memory (Hanley, Young, & Pearson, The findings have valuable implications for
1991), and vice versa (Farah, Hammond, Levine, & learning. There is now extensive evidence of a causal
Calvanio, 1988). Data on selective interference of link between impairments of verbal working memory
memory for spatial information with concurrent and learning difficulties (Alloway, Gathercole,
spatial movements (Baddeley & Lieberman, 1980), Adams, & Willis, 2005; Gathercole & Alloway, 2004;
and of visual memory with interference by visual Swanson & Siegel, 2001). In a recent observational
items (Quinn & McConnell, 1996) also support this study, we found that children with such impairments
position. Other studies have found differential levels often fail in classroom learning activities that place
of performance on a block sequence and visual pat- heavy demands on working memory (Gathercole,
tern memory tasks, which has been interpreted as Lamont, & Alloway, 2006). On this basis, we have
supporting a fractionation of visual and spatial suggested that excessive working memory load dis-
working memory (e.g., Logie & Pearson, 1997; rupts the incremental process of acquiring skill and
Pickering et al., 2001). It is important to note, how- knowledge, and so underlies the learning difficulties
ever, that in the present study there is no support for associated with working memory impairments
findings of dissociable visual and spatial aspects of (Gathercole, 2004). In atypical populations, there is
the visuospatial short-term memory tasks. Perform- a specificity of memory impairments. For example, a
ance on the dot matrix task, which is similar to the recent study has found that children with specific
visual patterns test, and the block recall task is not language impairment perform poorly on all verbal
markedly different (see Tables 1 and 2). Moreover, memory measures but within the normal range on
the age-related improvements are consistent for both visuospatial memory ones, while children with de-
visual and spatial short-term memory tasks (see velopmental coordination disorder perform below
Figure 1). Indeed, Gathercole et al. (2004) found age-expected levels in both verbal and visuospatial
similar developmental trajectories for these tasks as memory measures (Alloway & Archibald, 2005). The
well (see also Alloway, 2006). There are two possible development of a standardized tool such as the
explanations for why other studies have found a AWMA provides an important opportunity for the
dissociation between visual and spatial short-term theoretical exploration of the specificity of links be-
memory. First, Gathercole (1998) suggested that in tween verbal and visuospatial memory and scholastic
practice it is difficult to utilize a task that taps purely attainment in an atypically developing population.
visual or spatial components of memory. Second, an In conclusion, the findings from the present study
alternative explanation for findings of dissociable indicate that the theoretical structure of working
visual and spatial memory performance could be memory capacity is consistent with the view that
due to the static versus dynamic presentation for- there are domain-specific components for storage,
mats of the tasks rather than a visual/spatial dis- and a domain-general component for processing in-
tinction (Pickering et al., 2001). Similarly, in the formation. This structural organization of memory
present study, the dynamic presentation of all the can be reliably assessed from as young as 4 years and
visuospatial tasks may have resulted in the undif- remains constant across the childhood years.
ferentiated performance for the traditionally visual
(dot matrix) and spatial (block recall) tasks.
References
A further interesting point concerns the relation-
ship between verbal and visuospatial short-term Alloway, T. P. (2006). Working memory and children with
memory. The strength of association between these developmental coordination disorders. In T. P. Alloway
two variables increases slightly from the 4- to 6-age & S. E. Gathercole (Eds.), Working memory and neurode-
group to the older age bands (7 – 11 years). This is in velopmental conditions (pp. 161 – 187). New York: Psych-
ology Press.
line with the view that younger children rely more
Alloway, T. P., & Archibald, L. M. (2006). Working memory
on visual codes initially, but as they get older they
and learning in children with developmental coordination
use strategies such as rehearsal to recode visual disorder and specific language impairment. Manuscript
material using verbal labels (e.g., Hitch, Halliday, submitted for publication.
Schaafstal, & Schraagen, 1988). This is not inconsis- Alloway, T. P., Gathercole, S. E., Adams, A. M., & Willis, C.
tent with the view that these two types of storage (2005). Working memory and other cognitive skills as
tasks are domain specific as the data from the pres- predictors of progress towards early learning goals at
ent study indicate that a common short-term mem- school entry. British Journal of Developmental Psychology,
ory construct (Model 2) was not a suitable fit to the 23, 417 – 426.
Verbal and Visuospatial Short-Term and Working Memory in Children 1715

Alloway, T. P., Gathercole, S. E., & Pickering, S. J. (2004). opmental change? Journal of Experimental Child Psych-
The automated working memory assessment. Test battery ology, 73, 7 – 44.
available from authors. Conway, A. R. A., Cowan, N., Bunting, M. F., Therriault, D.
Alloway, T. P., Gathercole, S. E., Willis, C., & Adams, A. M. J., & Minkoff, S. R. B. (2002). A latent variable analysis of
(2004). A structural analysis of working memory and working memory capacity, short-term memory capacity,
related cognitive skills in early childhood. Journal of processing speed, and general fluid intelligence. Intelli-
Experimental Child Psychology, 87, 85 – 106. gence, 30, 163 – 183.
Baddeley, A. D. (1986). Working memory. New York: Oxford Cowan, N., Elliott, E. M., Saults, J. S., Morey, C. C., Mattox, S.,
University Press. Hismjatullina, A., et al. (2005). On the capacity of atten-
Baddeley, A. D. (1996). Exploring the central execu- tion: Its estimation and its role in working memory and
tive. Quarterly Journal of Experimental Psychology, 49A, cognitive aptitudes. Cognitive Psychology, 51, 42 – 100.
5 – 28. Daneman, M., & Carpenter, P. A. (1980). Individual dif-
Baddeley, A. D. (2000). The episodic buffer: A new com- ferences in working memory and reading. Journal of
ponent of working memory? Trends in Cognitive Sciences, Verbal learning and Verbal Behaviour, 19, 450 – 466.
4, 417 – 422. Daneman, M., & Merikle, P. M. (1996). Working memory
Baddeley, A. D., Emslie, H., Kolodny, J., & Duncan, J. and language comprehension: A meta-analysis. Psycho-
(1998). Random generation and the executive control of nomic Bulletin and Review, 3, 422 – 433.
working memory. Quarterly Journal of Experimental Daneman, M., & Tardiff, T. (1987). Working memory and
Psychology, 51A, 819 – 852. reading skill reexamined. In M. Coltheart (Ed.), Attention
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. and performance XII: The psychology of reading (pp. 491 –
Bower (Ed.), The psychology of learning and motivation 508). Hillsdale, NJ: Erlbaum.
(Vol. 8, pp. 47 – 90). New York: Academic Press. Duff, S. C., & Logie, R. H. (1999). Storage and processing in
Baddeley, A. D., & Lieberman, K. (1980). Spatial working visuo-spatial working memory. Scandinavian Journal of
memory. In R. S. Nickerson (Ed.), Attention and per- Psychology, 40, 251 – 259.
formance VIII (pp. 521 – 539). Hillsdale, NJ: Lawrence Engle, R. W., Kane, M. J., & Tuholski, S. W. (1999a). Indi-
Erlbaum Associates Inc. vidual differences in working memory capacity and
Baddeley, A. D., & Logie, R. H. (1999). The multiple-com- what they tell us about controlled attention, general
ponent model. In A. Miyake & P. Shah (Eds.), Models of fluid intelligence, and functions of the prefrontal cortex.
working memory: Mechanisms of active maintenance and In A. Miyake & P. Shah (Eds.), Models of working memory:
executive control (pp. 28 – 61). New York: Cambridge Mechanisms of active maintenance and executive control (pp.
University Press. 102 – 134). New York: Cambridge University Press.
Bayliss, D. M., Jarrold, C., Baddeley, A. D., Gunn, D. M., & Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A.
Leigh, E. (2005). Mapping the developmental constraints R. A. (1999b). Working memory, short-term memory, and
on working memory span performance. Developmental general fluid intelligence: A latent variable approach.
Psychology, 41, 579 – 597. Journal of Experimental Psychology: General, 125, 309 – 331.
Bayliss, D. M., Jarrold, C., Gunn, D. M., & Baddeley, A. D. Farah, M. J., Hammond, K. M., Levine, D. N., & Calvanio,
(2003). The complexities of complex span: Explaining R. (1988). Visual and spatial mental imagery: Dissocia
individual differences in working memory in children ble systems of representation. Cognitive Psychology, 20,
and adults. Journal of Experimental Psychology: General, 439 – 462.
132, 71 – 92. Friedman, N. P., & Miyake, A. (2000). Differential roles for
Bentler, P. M. (1990). Comparative fit indexes in structural visuospatial and verbal working memory in situation
models. Psychological Bulletin, 107, 238 – 246. model construction. Journal of Experimental Psychology:
Bentler, P. M. (2001). EQS Structural Equation Modelling General, 129, 61 – 83.
Software (Version 6.0) [Computer software]. USA: mvsoft. Fuster, J. M. (2000). Executive frontal functions. Experi-
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and mental Brain Research, 133, 66 – 70.
goodness of fit in the analysis of covariance structures. Gathercole, S. E. (1998). The development of memory.
Psychological Bulletin, 88, 588 – 606. Journal of Child Psychology and Psychiatry, 39, 3 – 27.
Bentler, P. M., & Wu, E. J. C. (1995). EQS for Macintosh Gathercole, S. E. (2004). Working memory and learning
User’s Guide. Encino, CA: Multivariate Software Inc. during the school years. Proceedings of the British Acad-
Bollen, K. A. (1989). Structural equations with latent variables. emy, 125, 365 – 380.
New York: John Wiley & Sons Inc. Gathercole, S.E, & Alloway, T. P. (2004). Working memory
Borland’s C11 Builder 5. (2004). Borland International. and classroom learning. Dyslexia Review, 15, 4 – 9.
Case, R., Kurland, D. M., & Goldberg, J. (1982). Working Gathercole, S.E, & Alloway, T. P. (2006). Working memory
memory capacity as long-term activation: An individual deficits in neurodevelopmental disorders. Journal of
differences approach. Journal of Experimental Psychology: Child Psychology and Psychiatry, 47, 4 – 15.
Learning, Memory and Cognition, 19, 1101 – 1114. Gathercole, S. E., Lamont, E., & Alloway, T. P. (2006). Working
Chuah, Y. M. L., & Maybery, M. T. (1999). Verbal and memory in the classroom. In S. Pickering (Ed.), Working
spatial short-term memory: Common sources of devel- memory and education (pp. 219 – 240). Elsevier Press.
1716 Alloway, Gathercole, and Pickering

Gathercole, S. E., & Pickering, S. J. (2000). Assessment of Miyake, A., Friedman, N., Rettinger, D. A., Shah, P., &
working memory in six- and seven-year-old children. Hegarty, M. (2001). How are visuospatial working mem-
Journal of Educational Psychology, 92, 377 – 390. ory, executive functioning, and spatial abilities related?
Gathercole, S. E., Pickering, S. J., Ambridge, B., & Wearing, A latent-variable analysis. Journal of Experimental Psych-
H. (2004). The structure of working memory from 4 to 15 ology: General, 130, 621 – 640.
years of age. Developmental Psychology, 40, 177 – 190. Nelson, C. A. (1995). The ontogeny of human memory: A
GoldWave Inc. (2004). GoldWave Digital Audio Editor Soft- cognitive neuroscience perspective. Developmental
ware, Version 5. St. Johns: GoldWave Inc. Psychology, 31, 723 – 738.
Hamilton, C. J., Coates, R. O., & Heffernan, T. (2003). What Nelson, C. A. (2000). Neural plasticity and human devel-
develops in visuo-spatial working memory develop- opment: The role of early experience in sculpting
ment? European Journal of Cognitive Psychology, 15, 43 – 69. memory systems. Developmental Science, 3, 115 – 136.
Hanley, J. R., Young, A. W., & Pearson, N. A. (1991). Im- Oberauer, K., Sü, H.-M., Schulze, R., Wilhelm, O., &
pairment of the visuo-spatial sketchpad. Quarterly Jour- Wittmann, W. W. (2000). Working memory capacityF
nal of Experimental Psychology, 43A, 101 – 125. facets of a cognitive ability construct. Personality and
Hitch, G. J. (1990). Developmental fractionation of working Individual Differences, 29, 1017 – 1045.
memory. In G. Vallar & T. Shallice (Eds.), Neuropsycho- Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N.,
logical impairment of short-term memory (pp. 221 – 246). Smith, A. D., & Smith, P. K. (2002). Models of visuo-
Cambridge, UK: Cambridge University Press. spatial and verbal memory across the adult life span.
Hitch, G. J., Halliday, S., Schaafstal, A. M., & Schraagen, J. Psychology and Aging, 17, 299 – 320.
M. C. (1988). Visual working memory in young children. Pickering, S. J., & Gathercole, S. E. (2001). Working memory
Memory and Cognition, 16, 120 – 132. test battery for children. London: Psychological Corpor-
Isaacs, E. B., & Vargha Khadem, F. (1989). Differential ation Europe.
course of development of spatial and verbal memory Pickering, S. J., & Gathercole, S. E. (2004). Distinctive
span: A normative study. British Journal of Developmental working memory profiles in children with special edu-
Psychology, 7, 377 – 380. cational needs. Educational Psychology, 24, 393 – 408.
Jaccard, J., & Wan, C. K. (1996). LISREL approaches to Pickering, S. J., Gathercole, S. E., Hall, M., & Lloyd, S. A.
interaction effects in multiple regression. Thousand Oaks, (2001). Development of memory for pattern and path:
CA: Sage Publications. Further evidence for the fractionation of visuo-spatial
Jarvis, H. L., & Gathercole, S. E. (2003). Verbal and non- memory. Quarterly Journal of Experimental Psychology,
verbal working memory and achievements on national 54A, 397 – 420.
curriculum tests at 11 and 14 years of age. Educational Pickering, S. J., Gathercole, S. E., & Peaker, S. M. (1998).
and Child Psychology, 20, 123 – 140. Verbal and visuo-spatial short-term memory in children:
Jonides, J., Lacey, S. C., & Nee, D. E. (2005). Processes of Evidence for common and distinct mechanisms. Memory
working memory in mind and brain. Current Directions and Cognition, 26, 1117 – 1130.
in Psychological Science, 14, 2 – 5. Quinn, J. G., & McConnell, J. (1996). Irrelevant pictures in
Just, M. A., & Carpenter, P. A. (1992). A capacity theory of visual working memory. Quarterly Journal of Experimental
comprehension: Individual differences in working Psychology, 49A, 200 – 215.
memory. Psychological Review, 99, 122 – 149. Russell, J., Jarrold, C., & Henry, L. (1996). Working memory
Kane, M. J., Hambrick, D. Z., Tuholski, S. W., Wilhelm, O., in children with autism and with moderate learning
Payne, T. W., & Engle, R. W. (2004). The generality of difficulties. Journal of Child Psychology and Psychiatry, 37,
working-memory capacity: A latent-variable approach 673 – 686.
to verbal and visuo-spatial memory span and reasoning. Shah, P., & Miyake, A. (1996). The separability of working
Journal of Experimental Psychology: General, 133, 189 – 217. memory resources for spatial thinking and language
Kline, R. B. (1998). Principles and practice of structural equa- processing: An individual differences approach. Journal
tion modelling. New York: Guilford. of Experimental Psychology: General, 125, 4 – 27.
Kyllonen, P. C. (1993). Aptitude testing inspired by infor- Swanson, H. L., & Siegel, L. (2001). Learning disabilities as
mation processing: A test of the four-sources model. a working memory deficit. Issues in Education. Contri-
Journal of General Psychology, 120, 375 – 405. butions from Educational Psychology, 7, 1 – 48.
Logie, R. H. (1995). Visuo-spatial working memory. Hove, UK: Vallar, G., & Papagno, C. (2003). Neuropsychological im-
Erlbaum. pairments of short-term memory. In A. D. Baddeley, M.
Logie, R. H., & Pearson, D. G. (1997). The inner eye and the D. Kopelman, & B. A. Wilson (Eds.), Handbook of memory
inner scribe of visuo-spatial working memory: Evidence disorders (pp. 249 – 270). Chichester, UK: John Wiley &
from developmental fractionation. European Journal of Sons.
Cognitive Psychology, 9, 241 – 257. Vandierendonck, A., Kemps, E., Fastame, M. C., & Szm-
McDonald, R. P., & Ho, R. M. (2002). Principles and prac- alec, A. (2004). Working memory components involved
tice in reporting structural equation analyses. Psycho- in the Corsi block test. British Journal of Psychology, 95,
logical Methods, 7, 64 – 82. 57 – 79.

You might also like