Individual differences in
Running head: WORKING MEMORY AND CLASSROOM BEHAVIOR
Individual differences in processing speed mediate a relationship between working
memory and children’s classroom behavior
Christopher Jarrold,, Naomi Mackett, & Debbora Hall
University of Bristol
Address correspondence to:
Chris Jarrold
School of Experimental Psychology
University of Bristol
12a Priory Road
Bristol
BS8 1TU
UK
Electronic mail: C.Jarrold@bristol.ac.uk
Telephone: +44 (0)117 928 8450
Facsimile: +44 (0)117 928 8588
1
Individual differences in
2
Individual differences in processing speed mediate a relationship between working
memory and children’s classroom behavior
Previous studies have shown an association between children’s working memory
performance and teacher ratings of classroom inattention, leading to the suggestion that
children who appear inattentive may in fact suffer from reduced working memory
capacity. However, working memory performance is determined by a range of factors
and in this study we examine the relationships between teacher ratings of classroom
behavior and the various constraints on working memory performance in a representative
sample of 6- to 8-year-olds in mainstream education. Analysis of individual differences
confirmed that working memory scores could be decomposed into the following
components: storage capacity, processing efficiency, and the residual variance that results
from combining storage and processing operations. However, only processing efficiency
was reliably related to teacher ratings of individuals’ ability to concentrate and learn in
the classroom, suggesting that individual differences in basic speed of processing, rather
than in memory capacity, drive this relationship.
Keywords: working memory, speed of processing, inattention, classroom behavior
Body text: 4984 words
Individual differences in
3
Individual differences in processing speed mediate a relationship between working
memory and children’s classroom behavior
Working memory is the ability to hold in mind information in the face of distraction
in order to engage in goal-directed behavior (Kane, Bleckley, Conway, & Engle, 2001).
Current theoretical models therefore emphasise the need for individuals to employ some
form of executive control (Baddeley, 1986) or controlled attention (Cowan et al., 2005;
Engle, Tuholski, Laughlin, & Conway, 1999; Kane et al., 2001) to keep representations
active in working memory. Indeed, Engle et al. (1999) measured adults’ working
memory and short-term memory capacities, with the latter being defined in terms of
participants’ ability to remember items in correct serial order in the absence of any
distraction. They found that working memory performance was related to short-term
memory capacity, reflecting a common need for storage of to-be-remembered
information. However, their working memory measures captured additional variance,
that was also related to fluid intelligence, and which they ascribed to executive control
abilities (see also Kane et al., 2004). Subsequent work has shown that working memory
performance may in fact depend on at least three component abilities – short-term storage
capacity, the ability to carry out the distracting ‘processing’ that is necessarily embedded
in a working memory task, and the ability to combine these two demands (Bayliss,
Jarrold, Gunn, and Baddeley, 2003). Taken together, these findings suggest that
combining storage and processing operations in a working memory paradigm recruits
additional, and potentially executive, resources over and beyond those involved in the
storage and processing components themselves.
Individual differences in
4
These theoretical analyses are consistent with evidence that measures of adults’
working memory are stronger predictors of higher-level abilities such as reading,
mathematics, and indices of intelligence than are measures of short-term memory (e.g.,
Oberauer, Schulze, Wilhelm, & Süß, 2005). Importantly, this greater predictive power of
working memory measures has also been observed in children (e.g., Bayliss, Jarrold,
Baddeley, Gunn, & Leigh, 2005; Bayliss, Jarrold, Gunn, & Baddeley, 2003; Hitch,
Towse, & Hutton, 2001). However, in addition to these relationships with measures of
academic achievement, researchers and educational practitioners are increasingly
suggesting that more general aspects of classroom behavior might depend on working
memory capacity. For example, Gathercole, Lamont, and Alloway (2006) studied the
classroom behavior of three boys who had previously been identified as having poor
working memory. They found that these individuals had difficulty in following complex
instructions, arguably because of the need to simultaneously hold in mind information
from the start of a complex sentence while processing the remainder of it (see also
Gathercole, Durling, Evans, Jeffcock, & Stone, 2008). They suggested that apparent
problems of inattention in such individuals might be better understood as working
memory difficulties; individuals who struggle to hold in mind classroom instructions in
the face of other distractions are likely to forget what has been asked of them, fail to stay
‘on-task’, and appear distractible.
In a series of subsequent studies, Gathercole and colleagues (Alloway, Gathercole,
Kirkwood, & Elliot, 2009; Gathercole, Alloway et al., 2008; Gathercole, Durling et al.,
2008) examined teacher ratings of classroom behavior in samples of children who had
previously been identified as showing particularly poor working memory performance
Individual differences in
5
using the Conners’ Teaching Rating Scale - Revised, Short Form (Conners, 2001). In
each study individuals with poor working memory function were particularly impaired on
the cognitive problems/inattention subscale of this version of the Conners’ form, relative
to comparison groups without working memory difficulties, supporting the view that poor
working memory performance is associated with apparent attentional problems in a
classroom setting. This work is of considerable importance because it suggests that
professionals risk incorrectly ascribing fundamental problems of attention to children that
are instead largely mediated by working memory difficulties (see also Lui & Tannock,
2007; Rogers, Hwang, Toplak, Weiss, & Tannock, 2011).
However, while undoubtedly plausible there are two reasons why this suggestion
may be premature at this stage. First, the individuals with low working memory
performance assessed in these studies also tended to have low IQ, raising the possibility
that rated problems of inattention in these groups were driven by a more general factor
rather than by working memory difficulties specifically. Second, as outlined at the
outset, working memory performance is multiply determined. Consequently, impaired
working memory performance might reflect diminished executive control abilities, but it
might equally result from impaired short-term memory performance, or a reduction in the
efficiency with which the processing component of a working memory task is performed.
One aim of the current work, therefore, was to attempt to replicate the finding of an
association between working memory task performance and teacher ratings of classroom
behavior in a sample where individuals would be expected to be performing in the typical
IQ range (cf. Lui & Tannock, 2007). The second aim was to better understand the nature
of any relationship that might be observed between working memory performance and
Individual differences in
6
teacher ratings of behavior. Specifically, in addition to measuring working memory
using standard ‘complex span’ measures of working memory that combine processing
and storage demands (Conway et al., 2005) we also took independent measures of the
storage and the processing components of these complex span tasks (cf. Bayliss et al.,
2003) in order to isolate the key factor underpinning any relation between working
memory performance and teacher ratings of classroom behavior.
The current study did not manipulate the type of processing involved in our
complex span tasks because previous work had indicated that individual differences in
processing speed were domain-general (Bayliss et al., 2003). Rather, two complex span
tasks were employed that both involved verbal processing, with one requiring verbal
storage and the other requiring visuo-spatial storage. In addition to measuring
performance on these two complex span tasks, participants’ verbal and visuo-spatial
short-term memory performance was assessed using ‘simple span’ tasks that exactly
matched the storage requirements of the complex span tasks but without any concurrent
processing. Similarly, individuals’ processing speed was measured using the same
processing task as employed in the complex span tasks, but in the absence of any storage
load. Finally, teachers rated classroom behavior using a recent version of the Conners’
scale. As a result, this collection of measures allowed us to examine the relationship
between working memory and ratings of classroom behavior, and then to break down this
relationship in terms of the component processes that constrain an individual’s working
memory performance.
Method
Individual differences in
7
Participants
Participants were 47 children, who represented all individuals from four school
classes for whom parental consent was obtained. Two of these classes were for children
in UK Year 2 (US Grade 1), and were situated in an infant school, the other two classes
were for children in UK Year 3 (US Grade 2) and were in the linked junior school on the
same, shared geographical site. These schools were chosen because they showed close to
national average levels of attainment on ‘Key Stage 2’ assessments of reading and
mathematics for children aged 11 years. In addition, the percentage of children in the
infant and junior school recorded as eligible for receiving free school meals in the last
available national census (January 2010) was 3.9 and 15.7 respectively (national average
for this age range = 18.5%). Twenty-one children (10 boys) were in Year 2 and 26 (17
boys) were in Year 3. The age of the sample ranged between 6 years 10 months and 8
years 3 months, with a mean age of 7 years 6 months (SD = 4 months).
Procedure
Participants were tested two different complex span tasks (verbal and visuospatial), two different simple span tasks (digit and Corsi), and a measure of processing
speed that was conducted twice. These tasks were presented in two sessions, each of
around 30 minutes in length. In the first session individuals received processing speed
assessment 1, verbal complex span, and Corsi span, in that order. In the second session
they were given the visuo-spatial complex span, digit span, and processing speed
assessment 2, in that order. In addition, each participant’s classroom behavior was rated
by a teacher who had taught that individual for the past 3 months or more, using the
Conners’ 3 Teachers’ Short Form (Conners, 2008).
Individual differences in
8
Complex span tasks. The two complex span tasks required concurrent storage and
processing, and were formed by crossing two types of storage (verbal or visuo-spatial)
with a verbal processing component. On any trial participants were presented with a
series of storage items, with a 3 s processing window following the presentation of each
storage item. The processing task involved making a phonological discrimination on a
series of nonwords that were presented during the processing window. Specifically,
participants had to press one key if the nonword began with a ‘k’ sound and another key
if it did not. Nonwords were selected from a pre-recorded set of 84 one-syllable
nonwords that were recorded in a female voice, and which lasted 500 ms each. Half of
the nonwords began with a ‘k’ sound. When a participant made a key press response to a
nonword presented in any given processing window, a further nonword was presented
following a gap of 250ms. In this way, sufficient nonwords were presented within a
given processing window to fill its 3 s length. At this point the next storage item was
presented, or, if the end of the trial had been reached, recall was signalled by the onset of
a recall screen.
In the verbal complex span task, storage items were numbers drawn from the set 1
to 9, which were visually presented individually in the centre of the screen for 1 s in 120
point Arial font. In the visuo-spatial complex span task, storage items were selected from
a 3 x 3 matrix of 9 squares (each approximately 2.5 cm x 2.5 cm). This matrix was
displayed on the screen for 1 s during each storage item presentation phase, with one of
the squares highlighted in red. Recall from a trial in the verbal complex span task
involved the participant saying the list of numbers that had been presented, with
instructions that recall should be in correct serial order. Participants recalled the items
Individual differences in
9
from trials in the visuo-spatial complex span task by touching on the appropriate squares
of a blank matrix shown on the computer screen, again under serial order recall
instructions.
Each task began with 4 trials at list length 2. If the participant correctly recalled all
of the storage items in correct serial order on at least one of these trials, they then moved
on to 4 trials at list length 3, if not, the task ended at that point. The same progression
rule was operated up to a list length of 6, giving a total possible maximum of 20 trials.
Performance was coded using a partial credit score (see Conway et al., 2005) in which the
proportion of items on each trial recalled in correct serial position was totalled across all
trials (maximum score of 20).
Simple span tasks. The two simple span tasks were explicitly designed in order to
measure the short-term memory demands inherent in the two complex span tasks. A digit
span task involved presenting the same storage items as in the verbal complex span task,
but with no intervening processing. Similarly a Corsi span task (Milner, 1971) presented
series of squares in the same 3 x 3 matrix as used in the visuo-spatial complex span task,
but without any interleaved processing demands.
As in the complex span tasks, storage items were presented for 1 s each, and there
were 4 trials at each list length. List lengths increased from 2 to a maximum of 6,
dependent on the participant recalling all of the items in correct serial order on at least
one trial of a given list length. Indeed, each trial in a simple span task used exactly the
same storage items, presented in the same order, as a trial in the corresponding complex
span task. However, the ordering of these yoked trials was different in the simple and
Individual differences in
10
complex span tasks. Performance was again coded in terms of partial credit scores
(maximum 20 for each task).
Processing speed. Speed of processing was assessed using exactly the same
processing task as was embedded in the two complex span tasks. The same set of
nonword stimuli as used in the processing component of the complex span tasks was used
as the stimulus set here, but in this case processing judgements were made in the absence
of any storage load. Participants completed two processing speed assessments, one at the
start and one at the end of the test battery. Each consisted of a 30 s period of processing
judgements. The dependent measure from each processing assessment was the
participant’s median reaction time for correct responses only. Each assessment was
preceded by 5 s of practice judgements.
Conners’ 3 Teachers Short Form. Each participant’s class teacher completed the
Conners’ 3 Teachers’ Short Form, rating that child’s behavior in the past 3 months. This
form consists of 39 statements assessing five sub-components of classroom behavior:
inattention, hyperactivity/impulsivity, learning problems/executive functioning,
defiance/aggression, and peer relations. Higher scores on each subscale reflect a greater
prevalence of problematic behaviors.
Results
The experimental variables derived from the memory task, and the scores from the
five subscales of the Conners’ form, were examined for univariate and multivariate
outliers. Four outlying scores were detected representing atypically high values on the
following measures: processing RT 1, hyperactivity/impulsivity, defiance/aggression,
peer relations. In each case the value in question was reduced to the next nearest value
Individual differences in
11
for that variable. No significant multivariate outliers were detected on the basis of
Mahalanobis distance.
Table 1 presents descriptive statistics for the experimental measures and the
subscales of the Conners’ form. As noted above, the dependent measure extracted from
the two processing tasks was median RT for correct responses, average accuracies for
assessment 1 and 2 were 81.0% (SD = 20.4%) and 81.3% (SD = 21.0%) respectively.
Table 1 also provides reliability estimates. Reliability estimates for the experimental
measures were good, with the exception of processing RT 1. Reliability of the Conners’
subscales was acceptable with the exception of the defiance/aggression measure. Item
analysis showed that the reliability of this variable would be improved by the omission of
one question. Doing this raised the alpha value to .592, and all subsequent analyses were
therefore based on a total score that excluded this item.
All of the experimental measures showed acceptable levels of skewness and
kurtosis apart from the processing RT 1 measure. Given this non-normality of the
processing RT 1 variable, its low reliability (see Table 1 and above), and the fact that it
did not correlate to any meaningful degree with the processing RT 2 score, r = .209, p =
.158, processing RT 1 values were not analysed further.2 Instead, processing RT 2
values were used as the sole index of processing speed in the individual differences
analyses reported below. All of the Conners’ scale scores were significantly skewed and
a square root transform was therefore applied to each total score prior to subsequent
analyses.
Table 2 presents the correlation matrices for the inter-relationships between the
Conners’ subscales, both with and without age partialled. This table shows moderate to
Individual differences in
12
high correlations between the first three Conners’ subscales. In contrast, the peer
relations subscale correlated significantly only with the defiance/aggression subscale.
Given this pattern of inter-relationships, and to reduce the number of subsequent
analyses, two separate composite scores were calculated from the Conners’ scale. The
first was an ‘individual behavior’ composite that was calculated by averaging normalized
scores on the inattention, hyperactivity/impulsivity, and learning problems/executive
functioning subscales. The second was a ‘social behavior’ composite that averaged
normalized scores on the defiance/aggression and peer relations subscales.
A set of hierarchical regressions then examined the extent to which the constituent
parts of working memory (storage capacity, processing efficiency, and any residual
variance) related to these individual behavior (see Table 3) and social behavior (see Table
4) Conners’ composites. In each case one set of analyses examined these relationships
for verbal working memory performance and another examined them for visuo-spatial
working memory performance. Age was entered on the first step of each regression to
control for the effects of any general age-related improvements. Either processing RT 2
was entered on step 2 with the appropriate storage measure (digit span or Corsi span) then
entered on step 3, or the appropriate storage measure was entered on Step 2 with
processing RT 2 entered on step 3. On step 4 the corresponding complex span measure
(verbal or visuo-spatial) was entered. This final step was included to examine the extent
to which residual variation in working memory performance, over and beyond the
component storage and processing contributions, contributed to the explanation of
Conners’ scores.
Individual differences in
13
Table 3 shows that the only significant predictor of the individual behavior
Conners’ composite was processing RT 2, which explained a significant amount of
variance even when entered on the 3rd step of each regression. Indeed, subsequent
analysis confirmed that processing RT 2 explained a significant proportion of the
variance in individual behavior even when entered after all of age, digit span, and verbal
complex span, ∆r2 = .126, F(1, 42) = 7.775, p = .008, or when entered after all of age,
Corsi span, and visuo-spatial complex span, ∆r2 = .128, F(1, 42) = 7.329 p = .010.
Inspection of the regression coefficients confirmed that those individuals who took longer
to make their responses in the second processing assessment had higher Conners’ scores.
In contrast, Table 4 indicates that the social behaviour Conners’ composite was only
predicted significantly by Corsi span, even when this measure was entered on the 3rd step
of the regression. In this case, the regression coefficients indicated that this association
ran in a counter-intuitive direction, with those individuals with higher Corsi spans having
higher scores on the social behaviour composite. Corsi span was not a significant
predictor of social behavior when entered after all of age, processing RT 2, and visuospatial complex span, ∆r2 = .041, F(1, 42) = 2.064, p = .158, because of the variance it
shared with visuo-spatial complex span.
A subsequent analysis examined whether the observed relationship between
processing RT2 and the individual behaviour Conners’ composite was seen equally for
the three subscales that made up this aggregate measure. This looked at the contribution
of processing RT2 to the prediction of i) inattention, ii) hyperactivity/impulsivity, and iii)
learning problems/executive functioning using regressions in which age and either digit
or Corsi span had already been entered (cf. Table 4). These regressions showed that
Individual differences in
14
processing RT2 predicted significant additional variance in inattention when either digit
span, ∆r2 = .148, F(1, 43) = 8.381, p = .006, or Corsi span, ∆r2 = .155, F(1, 43) = 8.539,
p = .006, were controlled for. Similarly, smaller but still significant proportions of
additional variance in hyperactivity/impulsivity were accounted for by processing RT2,
∆r2 = .098, F(1, 43) = 6.232, p = .016 controlling for digit span, ∆r2 = .103, F(1, 43) =
6.374, p = .015 controlling for Corsi span. The additional contribution of processing RT2
to the prediction of learning problems/executive functioning was also close to significant,
regardless of whether digit span, ∆r2 = .073, F(1, 43) = 3.656, p = .063, or Corsi span, ∆r2
= .084, F(1, 43) = 3.974, p = .053, were controlled for.
A final analysis explored whether the residual variance in complex span
performance, once the component variance in processing speed and storage ability was
accounted for, was inter-related across the two working memory tasks. For each complex
span task, residual variance was determined by partialling out the appropriate simple span
measure and the second assessment of processing speed. These two residual scores
correlated reliably with one another even with age partialled, r = .346, p = .018.
Discussion
This study had two related aims. The first was to attempt to replicate previous
findings of an association between experimental assessments of working memory and
teacher ratings of classroom behavior in a relatively representative sample of typically
developing children. The second was to partition working memory performance into its
component parts in order to determine the cause of any observed relationship between the
experimental measures and the behavior ratings. To that end we employed a previously
validated methodology that involved presenting independent measures of processing
Individual differences in
15
efficiency and short-term storage capacity in addition to complex span working memory
tasks that combined these processing and storage demands.
A first point to note is that the data suggest that the two measures of processing
efficiency were not indexing a common construct. The processing task employed here
required participants to make speeded decisions on the start sound in each of a series of
nonwords. One would certainly expect such a task to be constrained by general speed of
processing, but it is also possible that it was limited by individuals’ phonological
awareness, at least to begin with. Other data indicate that children of this age find initial
phoneme deletion challenging (Muter, Hulme, Snowling, & Taylor, 1997) but improve
substantially with training (Content, Morais, Alegria, & Bertleson, 1982). These points
are consistent with the view that the processing RT 2 measure is a purer index of basic
speed of processing than is the processing RT 1 score, but nevertheless the use of a single
measure of processing speed is a potential limitation of the current work.
Table 1 shows that both complex span tasks were noticeably harder than their
corresponding simple span equivalents, indicating that the processing demands of these
tasks did impact on memory storage. In addition, an important finding was that the two
‘residual variances’, extracted from the complex span tasks by partialling out processing
speed and the appropriate measure of storage capacity, were themselves significantly
related. This replicates previous work (see Jarrold & Bayliss, 2007) and suggests that
this residual variance is measuring a domain-general and potentially executive
requirement common to the two working memory measures. It also highlights the
importance of determining which if any of the three potential components of working
Individual differences in
16
memory performance – domain-specific storage capacity, domain-general processing
efficiency, or domain-general residual variance – relates to classroom behavior.
In the current study classroom behavior was measured by the Conners’ 3 Teachers’
Short Form, which has five potentially distinct subscales. However, the inter-relation of
these subscales (see Table 2) suggested that they might be sensibly grouped into two
clusters with the defiance/aggression and peer relations subscales separating to some
extent from the other three subscales. This division is obviously at odds with the fivefactor structure that underpins the scale and that emerged from a substantially larger
normative sample (Conners, 2008). It is also not consistent with evidence of a distinction
between inattentive and hyperactive/impulsive symptoms in attention deficit
hyperactivity disorder (ADHD). However, it is in line with hierarchical models of
ADHD that include a degree of commonality between these symptoms (see Toplak et al.,
2012), and there is also some intuitive sense in the current pattern of clustering. The
defiance/aggression and peer relations subscales clearly index problematic social
behaviors, while the remaining three subscales tap problematic behaviors that are
presumably more often reflective of cognitive difficulties experienced by the individual.
Further analysis showed that the social Conners’ composite was only related
significantly to visuo-spatial short-term memory as measured by the Corsi span task.
However, this association ran in an unexpected direction with individuals with relatively
higher Corsi span performance showing more problematic social behaviors. There is no
obvious theoretical explanation for this finding, and it might simply reflect an artefactual
chance association. Alternatively, a post-hoc explanation follows from the fact that
previous experimental and factor analytic studies have sometimes shown a closer relation
Individual differences in
17
between Corsi spans and complex span tasks than between digit spans and complex span
measures of working memory (e.g., Alloway, Gathercole, & Pickering, 2006; Ang & Lee,
2008). In the current study, Corsi span did not correlate with the social Conners’
composite if entered into the regression after visuo-spatial complex span, and it may
therefore be the case that Corsi span partly captures individual differences in executive
control abilities. Furthermore, there is some support for a positive relationship between
executive function and sensation-seeking in adolescence (Romer et al., 2011) and the
correlation between Corsi span and social behavior seen here may perhaps reflect those
with greater executive abilities seeking stimulation through negative interactions with
their peers, or frustration among these individuals due to a lack of sufficient stimulation
leading to problematic social activities. However, these suggestions are clearly
speculative, and are weakened by the fact that residual variance in the verbal complex
span task did not correlate with the social behavior Conners’ composite. The finding of a
positive association between Corsi span and problematic social behavior should,
therefore, be interpreted with caution and subjected to replication in future work.
In contrast the clear relationship with the individual behavior Conners’ composite
and individuals’ processing speed (see Table 3) is much more easily explained. Here,
individuals with lower processing speeds were rated as showing more problems of
inattention, hyperactivity/impulsivity, and learning/executive functioning. Very similar
findings have been reported in recent work on preterm individuals by Mulder and
colleagues. Mulder, Pitchford, and Marlow (2011a) examined teacher and parent ratings
of impulsivity and inattention in very preterm children and age-matched full term
controls. Participants’ speed of processing and working memory were also assessed.
Individual differences in
18
Mulder et al. found group differences in parental ratings of inattention, with greater levels
of inattention in the preterm group. However, these group differences were reduced to
non-significant levels when verbal processing speed was accounted for, with no
additional variance being explained by the subsequent addition of working memory into
the model. In other words, for this dependent measure, processing speed and not working
memory mediated the relationship between prematurity and problematic behavior.3 In
related work, Willcutt, Pennington, Olson, Chhabildas, and Hulslander, (2005) found that
while individuals with ADHD were impaired, relative to typically developing comparison
individuals, on composite measures of both verbal working memory and processing
speed, this group effect was numerically stronger for the processing speed factor. In that
study, processing speed was more strongly related to level of participants’ inattention
symptoms than their degree of hyperactivity-impulsivity (see also Lui & Tannock, 2007;
Martinussen & Tannock, 2006), though both symptom measures were significantly
related to the processing speed factor score. A similar finding emerged in the current
data; post-hoc analysis showed that variation on all three of the individual behaviour
Conners’ subscales was linked to differences in processing speed, but the relationship
was strongest for the inattention subscale. It may therefore be the case that any reduction
in speed of processing has a relatively general effect on individuals’ behaviour, but that
this is most easily observed in terms of problems of attention.
It should be noted that individual differences in IQ were not assessed in the current
study. However, there is considerable evidence for a close association between IQ and
speed of processing (e.g., Anderson, 1992; Fry & Hale, 2000), and it is quite possible that
the variance in processing speed seen here, which drove the relationship with classroom
Individual differences in
19
behavior, overlapped with variation in IQ in this sample (see also Mulder et al. 2011a, b).
Clearly, this suggestion is also consistent with the view (see above) that previous
evidence of a link between working memory performance and classroom behavior in
individuals with indentified working memory problems is a corollary of the low IQ of
these individuals.
Certainly, the current data are not consistent with the view that the relationship
between problematic classroom behavior and working memory performance is mediated
by the ability to maintain information, either in short-term memory or via the imposition
of executive controlled attention. There was no significant relationship between the
individual behavior Conners’ composite and either digit or Corsi span, measures that tap
the ability to store in correct serial order verbal or visuo-spatial information respectively.
Similarly, there was no suggestion of an association between this Conners’ composite and
the residual variance in complex span associated with the need to combine storage and
processing operations in a working memory task. The predictive power of this residual
variance was examined in our analyses by entering the complex span measures in the
final steps of our regressions; because the independent assessments of storage capacity
and processing efficiency had already been entered into the model by this point, these
aspects of complex span performance had already been controlled for. In each case this
remaining variance did not add significantly to the prediction of the individual behaviour
Conners’ composite. This residual variance is often seen to reflect the executive control
required to maintain information in memory in the face of distracting processing, and
may well relate to individuals’ ability to inhibit this distraction (Kane et al., 2001).
Direct measures of inhibitory control were not included in this study, but future work
Individual differences in
20
would benefit from including a direct assessment of individuals’ ability to inhibit
distraction and to exert executive control in order to fully test the possible association
between these constructs and classroom behaviour.
In summary, the current data replicate previous evidence of a relationship between
teacher ratings of inattentive / hyperactive / distracted behavior in the classroom and
working memory measures. They therefore support the view that apparent problems of
individual behavior in the classroom might be secondary to more fundamental difficulties
that are picked up by working memory tasks. However, they go further than previous
work in suggesting that this relationship might be mediated primarily by individual
differences in processing efficiency rather than by variance in memory capacity. One
might well expect individuals with a reduced speed of processing to struggle to process
complex classroom instructions, not because they have difficulty in remembering these
instructions, but rather because they struggle to encode and construct the correct semantic
meaning of the words they are hearing. If this is correct, then the appropriate intervention
might not be to reduce the length of complex classroom instructions, but instead would be
to reduce the pace with which these instructions are presented.
Individual differences in
21
References
Alloway, T. P., Gathercole, S. E., Kirkwood, H., & Elliott, J. (2009). The cognitive and
behavioural characteristics of children with low working memory. Child
Development, 80, 606-621. doi: 10.1111/j.1467-8624.2009.01282.x
Alloway, T. P., Gathercole, S. E., & Pickering, S. J. (2006). Verbal and visuospatial
short-term and working memory in children: Are they separable? Child
Development, 77, 1698-1716. doi: 10.1111/j.1467-8624.2006.00968.x
Anderson, M. (1992). Intelligence and Development: A Cognitive Theory. Oxford:
Blackwell.
Ang, S. Y., & Lee, K. (2008). Central executive involvement in children's spatial
memory. Memory, 16, 918-933. doi: 10.1080/09658210802365347
Baddeley, A. D. (1986). Working Memory. Oxford: Oxford University Press.
Bayliss, D. M., Jarrold, C., Baddeley, A. D., Gunn, D. M., & Leigh, E. (2005). Mapping
the developmental constraints on working memory span performance.
Developmental Psychology, 41, 579-597. doi: 10.1037/0012-1649.41.4.579
Bayliss, D. M., Jarrold, C., Gunn, D. M., & Baddeley, A. D. (2003). The complexities of
complex span: explaining individual differences in working memory in children
and adults. Journal of Experimental Psychology: General, 132, 71-92. doi:
10.1037/0096-3445.132.1.71
Conners, C. K. (2001). Conners Rating Scale—Revised. North Tunawanda, NY:
Multihealth Systems.
Conners, C. K. (2008). Conners 3. North Tunawanda, NY: Multihealth Systems.
Content, A., Morais, J., Alegria, J., & Bertleson, P. (1982). Accelerating the development
Individual differences in
22
of phonemic segmentation skills in kindergarteners. Cahiers de Psychologie
Cognitive, 2, 259-269.
Conway, A. R. A., Kane, M. J., Bunting, M. F., Hambrick, D. Z., Wilhelm, O., & Engle,
R. W. (2005). Working memory span tasks: A methodological review and user's
guide. Psychonomic Bulletin and Review, 12, 769-786. doi: 10.3758/BF03196772
Cowan, N., Elliot, E. M., Saults, J. S., Morey, C., Mattox, S., Hismjatullina, A., &
Conway, A. R. A. (2005). On the capacity of attention: Its estimation and its role
in working memory and cognitive abilities. Cognitive Psychology, 51, 42-100.
doi: 10.1016/j.cogpsych.2004.12.001
Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999). Working
memory, short-term memory, and general fluid intelligence: A latent-variable
approach. Journal of Experimental Psychology: General, 128, 309-311. doi:
10.1037/0096-3445.128.3.309
Fry, A. F., & Hale, S. (2000). Relationships among processing speed, working memory,
and fluid intelligence in children. Biological Psychology, 54, 1-34. doi:
10.1016/S0301-0511(00)00051-X
Gathercole, S. E., Alloway, T. P., Kirkwood, H. J., Elliott, J. G., Holmes, J., & Hilton, K.
A. (2008). Attentional and executive function behaviours in children with poor
working memory. Learning and Individual Differences, 18, 214-223. doi:
10.1016/j.lindif.2007.10.003
Gathercole, S. E., Durling, M., Evans, S., Jeffcock, E., & Stone, S. (2008). Working
memory abilities and children’s performance in laboratory analogues of classroom
activities. Applied Cognitive Psychology, 22, 1019-1037. doi: 10.1002/acp.1407
Individual differences in
23
Gathercole, S. E., Lamont, E., & Alloway, T. P. (2006). Working memory in the
classroom. In S. Pickering (Ed.), Working memory and education (pp. 219-240).
London: Elsevier Academic Press.
Hitch, G. J., Towse, J. N., & Hutton, U. (2001). What limits children's working memory
span? Theoretical accounts and applications for scholastic development. Journal
of Experimental Psychology: General, 130, 184-198. doi: 10.1037/00963445.130.2.184
Jarrold, C. & Bayliss, D. M. (2007). Variation in working memory due to typical and
atypical development. In A. R. A. Conway, C. Jarrold, M. J. Kane, A. Miyake, &
J. N. Towse (Eds.), Variation in working memory (pp. 134-161). New York:
Oxford University Press.
Kane, M. J., Bleckley, M. K., Conway, A. R. A., & Engle, R. W. (2001). A controlledattention view of working-memory capacity. Journal of Experimental
Psychology: General, 130, 169-183. doi: 10.1037/0096-3445.130.2.169
Kane, M. J., Hambrick, D. Z., Tuholski, S. W., Wilhelm, O., Payne, T. W., & Engle, R.
W. (2004). The generality of working memory capacity: A latent-variable
approach to verbal and visuospatial memory span and reasoning. Journal of
Experimental Psychology: General, 133, 189-217. doi: 10.1037/00963445.133.2.189
Lui, M., & Tannock, R. (2007). Working memory and inattentive behaviour in a
community sample of children. Behavioral and Brain Functions, 3, 12. doi:
10.1186/1744-9081-3-12
Martinussen, R., & Tannock, R. (2006). Working memory impairments in children with
Individual differences in
24
attention-deficit hyperactivity disorder with and without comorbid language
learning disorders. Journal of Clinical and Experimental Neuropsychology, 28,
1073-1094. doi: 10.1080/13803390500205700
Milner, B. (1971). Interhemispheric differences in the localisation of psychological
processes in man. British Medical Bulletin, 27, 272-277.
Mulder, H., Pitchford, N. J., & Marlow, N. (2011a). Inattentive behaviour is associated
with poor working memory and slow processing speed in very pre-term children
in middle childhood. British Journal of Educational Psychology, 81, 147-160.
doi: 10.1348/000709910X505527
Mulder, H., Pitchford, N. J., & Marlow, N. (2011b). Processing speed mediates executive
function difficulties in very preterm children in middle childhood. Journal of the
International Neuropsychological Society, 17, 445-454. doi:
10.1017/S1355617711000373
Muter, V., Hulme, C., Snowling, M., & Taylor, S. (1997). Segmentation, not rhyming,
predicts early progress in learning to read. Journal of Experimental Child
Psychology, 65, 370-396. doi: 10.1006/jecp.1996.2365
Oberauer, K., Schulze, R., Wilhelm, O., & Süß, H.-M. (2005). Working memory and
intelligence – their correlation and their relation: comment on Ackerman, Beier,
and Boyle (2005). Psychological Bulletin, 131, 61-65. doi: 10.1037/00332909.131.1.61
Romer, D., Betancourt, L. M., Brodsky, N. L., Giannetta, J. M., Yang, W., & Hurt, H.
(2011). Does adolescent risk taking imply weak executive function? A
prospective study of relations between working memory performance,
Individual differences in
25
impulsivity, and risk taking in early adolescence. Developmental Science, 14,
1119-1133. doi: 10.1111/j.1467-7687.2011.01061.x.
Rogers, M., Hwang, H., Toplak, M., Weiss, M., & Tannock, R. (2011). Inattention,
working memory, and academic achievement in adolescents referred for attention
deficit/hyperactivity disorder (ADHD). Child Neuropsychology, 17, 444-458. doi:
10.1080/09297049.2010.544648
Toplak, M. E., Sorge, G. B., Flora, D. B., Chen, W., Banaschewski, T., Buitelaar, J., . . .
Faraone, S. V. (2012). The hierarchical factor model of ADHD: invariant across
age and national groupings? Journal of Child Psychology and Psychiatry, 53,
292-303. doi: 10.1111/j.1469-7610.2011.02500.x
Willcutt, E. G., Pennington, B. F., Olson, R. K., Chhabildas, N., & Hulslander, J. (2005).
Neuropsychological analyses of comorbidity between reading disability and
attention deficit hyperactivity disorder: In search of the common deficit.
Developmental Neuropsychology, 27, 35-78. doi: 10.1207/s15326942dn2701_3
Individual differences in
26
Author Note
Preparation of this paper was supported by a grant from the Economic and Social
Research Council of the UK: RES-062-23-2467. We are grateful to the pupils and staff
of Ashley Down Infants and Juniors Schools, Bristol, for their cooperation with this
work.
Individual differences in
27
Footnotes
1 Cronbach’s alpha values for the memory partial credit scores were calculated by
deriving separate partial credit scores based on performance on the first, second, third,
and fourth trial at each span level. However, it should be noted that these values are not
completely independent of each other because of the stopping rule employed on these
tasks. Consequently this approach risks over-estimating task reliability.
2 Analyses that combine processing RTs 1 and 2 into a single measure lead to
substantially weaker effects than those reported below; details available on request from
the first author.
3 Having said this, a somewhat different pattern emerged from the corresponding analysis
of teacher ratings of inattention in Mulder et al. (2011a). Here, the introduction of verbal
processing speed on step 2 of the model again reduced the previously significant group
effect to a non-significant level. However, an index of working memory added further
significant variance to the model when entered on step 3. Mulder et al. suggest that
working memory may be more of a determinant of ability to stay on task in the classroom
than at home. While this may be true, it is worth noting that in a parallel report of
additional analyses from essentially the same dataset, Mulder, Pitchford, and Marlow
(2011b) found that a significant group difference in working memory in these samples
was itself mediated by differences in processing speed.
Individual differences in
28
Table 1
Descriptive statistics for the experimental measures in the study (processing RTs in ms,
all span scores are partial credit scores), plus Conners’ subscale raw scores (I =
Inattention, HI = Hyperactivity/impulsivity, LE = Learning problems/executive
functioning, DA = Defiance/aggression, PR = Peer relations). Reliability estimates are
split half correlations with Spearman-Brown correction for the two processing time
measures, and Cronbach’s alpha values for all other variables.1
M
SD
Range
Skewness
Kurtosis
Reliability
Processing RT 1
1642
454
993 - 3068
1.66
3.10
.469
Processing RT 2
1323
253
914 - 2038
0.18
0.04
.864
Digit span
15.21
2.29
8.53 - 18.50
-0.71
0.10
.838
Corsi span
13.46
3.49
2.67 – 18.02
-0.91
0.41
.927
Verbal complex
7.57
4.22
1.50 - 17.20
0.53
-0.72
.915
Visual complex
7.83
4.02
0.50 – 17.33
0.14
-0.47
.922
I
2.74
3.84
0 - 14
1.60
1.87
.943
HI
2.83
3.16
0 - 12
1.29
1.12
.891
LE
3.34
3.39
0 - 13
1.24
0.89
.832
DA
0.62
1.10
0-3
1.56
0.84
.418
PR
0.98
1.80
0-7
2.04
3.83
.866
Individual differences in
29
Table 2
Correlations between the five Conners’ subscales. Zero-order coefficients above the
leading diagonal, coefficients for partial correlations controlling for age below the
leading diagonal (LE = Learning problems/executive functioning) (* p < .05, ** p < .01)
Measure
1
2
3
4
5
1. Inattention
-
.679**
.733**
.477**
.346*
2. Hyperactivity/impulsivity
.655**
-
.328*
.435**
.174
3. LE
.746**
.345*
-
.353*
.255
4. Defiance/aggression
.471*
.441*
.351*
-
.464**
5. Peer relations
.319
.113
.252
.457*
-
Individual differences in
Table 3
Hierarchical regressions predicting variation in the ‘individual behaviour’ composite of the Conners’ ratings
Storage on step 2 and processing on step 3
Processing on step 2 and storage on step 3
Step
IV
∆r2
F
df
p
Step
IV
∆r2
F
df
p
1
Age
.080
3.890
1, 45
.055
1
Age
.080
3.890
1, 45
.055
2
Digit span
.055
2.772
1, 44
.103
2
RT 2
.169
9.927
1, 44
.003
3
RT 2
.157
9.510
1, 43
.004
3
Digit span
.042
2.543
1, 43
.118
4
Verbal complex
.028
1.715
1, 42
.197
4
Verbal complex
.028
1.715
1, 42
.197
1
Age
.080
3.890
1, 45
.055
1
Age
.080
3.890
1, 45
.055
2
Corsi span
.002
0.076
1, 44
.784
2
RT2
.169
9.927
1, 44
.003
3
RT 2
.168
9.612
1, 43
.003
3
Corsi Span
< .001
0.002
1, 43
.967
4
Visual complex
.017
0.957
1, 42
.333
4
Visual complex
.017
0.957
1, 42
.333
2
Individual differences in
Table 4
Hierarchical regressions predicting variation in the ‘social behaviour’ composite of the Conners’ ratings
Storage on step 2 and processing on step 3
Processing on step 2 and storage on step 3
Step
IV
∆r2
F
df
p
Step
IV
∆r2
F
df
p
1
Age
.022
1.014
1, 45
.319
1
Age
.022
1.014
1, 45
.319
2
Digit span
.014
0.651
1, 44
.424
2
RT 2
.031
1.462
1, 44
.233
3
RT 2
.035
1.606
1, 43
.212
3
Digit span
.018
0.811
1, 43
.373
4
Verbal complex
.045
2.150
1, 42
.150
4
Verbal complex
.045
2.150
1, 42
.150
1
Age
.022
1.014
1, 45
.319
1
Age
.022
1.014
1, 45
.319
2
Corsi span
.104
5.243
1, 44
.027
2
RT 2
.031
1.462
1, 44
.233
3
RT 2
.023
1.151
1, 43
.289
3
Corsi span
.095
4.823
1, 43
.034
4
Visual complex
.009
0.424
1, 42
.518
4
Visual complex
.009
0.424
1, 42
.518
3