Content uploaded by Torkel Klingberg
Author content
All content in this area was uploaded by Torkel Klingberg
Content may be subject to copyright.
Training of Working Memory in Children
With ADHD
Torkel Klingberg, Hans Forssberg, and Helena Westerberg
Department of Neuropediatrics, Karolinska Institute, Stockholm, Sweden
ABSTRACT
Working memory (WM) capacity is the ability to retain and manipulate information during a short period of
time. This ability underlies complex reasoning and has generally been regarded as a ®xed trait of the
individual. Children with attention de®cit hyperactivity disorder (ADHD) represent one group of subjects
with a WM de®cit, attributed to an impairment of the frontal lobe. In the present study, we used a new
training paradigm with intensive and adaptive training of WM tasks and evaluated the effect of training with
a double blind, placebo controlled design. Training signi®cantly enhanced performance on the trained WM
tasks. More importantly, the training signi®cantly improved performance on a nontrained visuo-spatial WM
task and on Raven's Progressive Matrices, which is a nonverbal complex reasoning task. In addition, motor
activity ± as measured by the number of head movements during a computerized test ± was signi®cantly
reduced in the treatment group. A second experiment showed that similar training-induced improvements on
cognitive tasks are also possible in young adults without ADHD. These results demonstrate that performance
on WM tasks can be signi®cantly improved by training, and that the training effect also generalizes to
nontrained tasks requiring WM. Training improved performance on tasks related to prefrontal functioning
and had also a signi®cant effect on motor activity in children with ADHD. The results thus suggest that WM
training potentially could be of clinical use for ameliorating the symptoms in ADHD.
The ability to retain and manipulate informa-
tion in WM depends on the prefrontal cortex
(Fuster, 1989; Goldman-Rakic, 1987) and under-
lies several cognitive abilities, including logical
reasoning and problem-solving (Engle, Kane, &
Tuholski, 1999; Hulme & Roodenrys, 1995;
Klingberg, 2000). Furthermore, WM capacity
has been regarded as a permanent trait of the
individual, closely related to g, a proposed mea-
sure of general cognitive ability (Engle, Kane, &
Tuholski, 1999; Kyllonen & Christal, 1990).
Attention de®cit hyperactivity disorder (ADHD)
is characterized by inattention, impulsivity, and
hyperactivity (American Psychiatric Associa-
tion, 1994). Among the cognitive de®cits in this
disorder, WM impairment is of central impor-
tance (Barkley, 1997; Kuntsi, Oosterlaan, & Ste-
venson, 2001; Mariani & Barkley, 1997; Rapport,
Chung, Shore, Denney, & Isaacs, 2000; Wester-
berg, Hirvikoski, Forssberg, & Klingberg, sub-
mitted) and has been suggested to be associated
with an impaired function of the frontal lobe
(Rubia et al., 1999; Schweitzer et al., 2000;
Zametkin et al., 1990).
In the present study we investigated whether
WM capacity could be improved by training.
Furthermore, if impairment of WM is a core
de®cit in ADHD, this would imply that improve-
ment of WM would decrease the symptoms in
ADHD. Previous attempts to improve WM by
training have only achieved moderate success. In
experiments where subjects perform repeated
Address correspondence to: Torkel Klingberg, Department of Neuropediatrics, Karolinska Institute, Astrid
Lindgrens Barnsjukhus Q2:07, 171 76 Stockholm, Sweden. Tel.: 46-8-5177-7357. Fax: 46-8-5177-7349.
E-mail: torkel.klingberg@neuro.ki.se
Accepted for publication: March 8, 2002.
Journal of Clinical and Experimental Neuropsychology 1380-3395/02/2406-781$16.00
2002, Vol. 24, No. 6, pp. 781±791 # Swets & Zeitlinger
WM trials, without adapting the dif®culty level,
this typically only leads to faster reaction times,
but no increase in WM capacity has been report-
ed (Kristofferson, 1972; Phillips & Nettelbeck,
1984). Some success has been achieved in teach-
ing rehearsal strategies to children with learning
disabilities (Brown, Campione, Bray, & Wilcox,
1973; Butter®eld, Wambold, & Belmont, 1973;
Hulme, 1992). There are also case studies of
subjects who have learned strategies to retain a
large number of digits (Ericsson, Chase, & Faloon,
1980). However, in these studies the strategies
were not useful for improving on other WM or
reasoning tasks. Therefore, such training does not
increase general WM capacity.
Here, we investigated whether WM capacity
could be improved by a new type of computer-
ized, cognitive training and whether this would
reduce motor activity in children with ADHD. We
adopted two key features of a training regime
previously used to enhance sensory discrimina-
tion and induce cortical plasticity in sensory and
motor cortices (Buonomano & Merzenich, 1998;
Tallal et al., 1996): (1) training was performed
close to the capacity of the individual by using an
adaptive staircase method that adjusted dif®culty
on a trial-by-trial basis; and (2) training was
performed at least 20 min per day, 4±6 days a
week, for at least 5 weeks.
We designed a computer program based upon
these principles in which subjects practiced WM
tasks. Fourteen children with ADHD undertook
training of a visuo-spatial WM task, a visuo-
spatial version of backwards digit-span, and a
spatial-verbal WM task. Visual and verbal feed-
back was implemented in the computer program
to increase compliance during the training. We
also designed a ``placebo'' or ``low-dose'' compu-
ter-program, which was similar to the treatment-
program, but did not include the two key features:
thus dif®culty level was not interactively adjusted,
and daily training amounted to less than 10 min
per day. The study was designed as a double-blind
study where children, parents, and the psychol-
ogist administrating pre- and posttraining tests
were blinded to which version of the computer
program the children had practiced and to
the difference in expected effect of the two
versions.
Subjects performed a battery of cognitive
tasks before and after training to evaluate WM
capacity and prefrontal functioning. These tasks
included Raven's Coloured Progressive Matrices
(Raven, 1995), which is a series of complex
reasoning tasks thought to measure prefrontal
functioning and general cognitive ability. Perfor-
mance on this task has been estimated to have
about r
2
:9 correlation with g (Engle et al.,
1999). Impulsivity was estimated using the Stroop
task, on which children with ADHD are known to
have impaired ability to perform. The Stroop task
also relies on activation of the prefrontal cortex
(Bench et al., 1993; Pardo, Pardo, Janer, &
Raichle, 1990). To get a measure of motor activity
level before and after training we used an infrared
camera that detected movements of a marker
placed on the child's head while the child
performed a 15-min continuous performance task
on a computer. Motor activity as measured by this
method has previously been shown to correlate
with behavioral ratings of hyperactivity in
children with ADHD (Teicher, Ito, Glod, &
Barber, 1996) and to be sensitive to measure the
effects of stimulant treatment (Teicher et al.,
2000). In a subsequent experiment, we tested
whether young adults without ADHD or WM
de®cits would be able to increase WM capacity by
training.
EXPERIMENT 1
METHODS
Subjects
Children participating in the study were between 7 and
15 years of age and diagnosed with ADHD by a
pediatrician according to the guidelines of DSM-IV.
The treatment group included 1 girl and 6 boys (mean
age 11.0, SD 2.0), and the control group includ-
ed 2 girls and 5 boys (mean age 11.4, SD 3.0).
Three subjects in the treatment group and 2 subjects in
the control group were on medication. Movement
analysis could not be performed in 2 subjects in
the treatment group due to technical failures. There
were no signi®cant differences in age between the
groups, or any differences in pretest scores for either of
the cognitive task or for head movements. The study
was conducted at the Astrid Lindgren Children's
Hospital, which is part of the Karolinska Hospital
782 TORKEL KLINGBERG ET AL.
and approved by the local ethics committee at the
Karolinska Hospital.
Tests for Pre- and Posttraining Evaluation
All subjects were tested on ®ve cognitive tasks:
(1) Trained version of the visuo-spatial WM task:
Circles were presented one at a time in a four-by-four
grid. After a delay the subjects indicated the positions
of the circles. The number of circles in the sequence
was successively increased until the subject missed two
trials in a row. The score was the maximum number of
circles remembered, with half a point given for one
correct trial out of two. (2) Span board: Ten blocks were
arranged in an irregular pattern in front of the subject.
The testing psychologist pointed to a sequence of
blocks and the subject then pointed to the same blocks
in the same order (forward version) or in the reverse
order (backward version). Scoring was performed as
for the trained version of the visuo-spatial WM task.
(3) Stroop task: Words describing colors were printed
with ink in a color that was incongruent with the word,
that is, ``green'' printed in yellow ink. The subjects
were asked to name the color of the ink for each word.
(4) Raven's Colored Progressive Matrices: A series of
nonverbal reasoning tasks described in detail elsewhere
(Raven, 1995). This test is often used for estimating
general nonverbal mental ability. (5) Choice reaction
time task. A computer was used for displaying visual
stimuli. A separate serial-response box (Psychology
Software Tools, Pittsburg) was used for collecting
responses. Yellow circles would appear in one out of
two possible locations on the screen, to the left or to the
right. The task was to press a button when a yellow
circle appeared. The yellow circle was preceded by
a warning cue consisting of a grey circle that appeared
1±4 s before the yellow circle. Each subject was ®rst
tested on a simple-reaction time version with only one
possible stimulus location (®rst circles appearing on
the left location, with responses made by the left
hand, then repeated on the right hand side). Then
subjects were tested on a choice reaction time task
with warning cues appearing simultaneously to the
right and left, and yellow circles appearing randomly
either in the left or the right location. The dependent
measures were (1) reaction time for one-choice trials;
(2) increase in reaction time for two-choice as com-
pared with one-choice trails; and (3) variance of reac-
tion times.
The measurement of head movements has been
described in previous publications (Teicher et al., 1996).
An infrared motion analysis system (OPTAx Systems,
Burlington, MA) recorded the movements of a small
re¯ective marker attached to the back of the head of
the child. A movement was de®ned to begin when
the marker moved 1.0 mm or more from its most
recent resting location. The number of movements was
recorded during a 15-min period when the child was
performing a version of a continuous performance task.
In this task subjects were asked to respond to a target
and withhold response to nontargets, with no require-
ment of holding any information in WM. Stimuli
were presented every 2.0 s, and 50% of stimuli were
targets.
Computerized Training Program
Four subtests were presented during each training
session: (1) A visuo-spatial WM task where circles
were presented one at a time in a four-by-four grid
(same task as the one used for testing). (2) Backwards
digit-span. A keyboard with numbers was shown and
digits read aloud. The subject then marked the digits,
but in the reverse order. (3) Letter-span task. Letters
were read aloud one at a time. The subject had to
remember the identity and order of the letters. A row of
lamps was then visible and a ¯ashing lamp cued the
subject as to which letter should be reported back,
for example, if lamp no. 3 was lit, the subject should
report the third letter that they previously heard.
(4) Choice reaction time task. This task was not a
WM task but a mixture of a reaction-time task and a go/
no-go task, and was included based on the evidence
from our laboratory that children with ADHD are im-
paired on similar tasks. Two grey circles were presented
on the screen (horizontally oriented in the two-choice
condition). Subjects were told to press a spatially
congruent key when one of the circles became green,
and to withhold responding when one of the circles
became red.
For all WM tasks, dif®culty was adjusted by chang-
ing the number of stimuli to be remembered. Subjects
completed 30 trials on each WM task every day, and
the daily training time was approximately 25 min.
Time between test and retest was 5±6 weeks during
which the subjects trained 24.3 (2.2 SEM) days. For
two types of tasks: the visuo-spatial WM task and the
digit-span task, subjects proceeded to train a second
and more demanding version after 10±18 days of
training on the ®rst version. In this second version a
visual distraction occurred during the delay in the WM
tasks.
The placebo/low-dose version included 10 trails per
task, with 2 stimuli to remember in the visuo-spatial
and digit span task, and 3 stimuli in the letter-span task.
The placebo version was intended to control for the
effect of taking the evaluation test repeatedly, for the
spontaneous improvement that could occur over the
training period of 5±6 weeks, and for possible bias for
the psychologist administrating the tasks as well as for
effects of expectancy from the children. The placebo-
version did not, however, control for the total amount of
time in front of a computer.
TRAINING OF WORKING MEMORY 783
RESULTS
Test-retest changes in the group of subjects under-
taking the treatment program (treatment group)
was compared to the test-retest changes in the
group of subjects using the placebo program
(control group). This comparison demonstrated
a signi®cant treatment effect for the practiced
visuo-spatial WM task (Table 1) as well as for
the Span board task, a nonpracticed visuo-spatial
WM task (Fig. 1a; Table 1). In the Span board
task, subjects remembered the position of small
cubes arranged in a pseudo- random order. While
the task tested visuo-spatial WM, it differed in
several ways from the practiced visuo-spatial
WM task: it was not computerized, but involved
three-dimensional stimuli, had a different spatial
arrangement of stimuli, and involved different
modes of stimulus presentation and response.
Substantial improvements were evident for all
children on the Span board task (Fig. 1a), and
group differences were signi®cant (P < .0001;
performance on the backward and forward ver-
sions were averaged). Signi®cant improvement on
Raven's Progressive Matrices was also evident
(Table 1). Again, all children in the treatment
group improved (Fig. 1b). Signi®cant group dif-
ference was found also for the Stroop task
(Table 1). However, only weak and inconsistent
effects were seen in a choice reaction time task
(Table 1).
The number of head movements was signi®-
cantly reduced in the treatment group compared
to the control group (Table 1, Fig. 1c). Again, this
effect was evident in all subjects in the treatment
group (Fig. 1c). The number of head movements
during retest in the control group was about 6%
higher than during the ®rst testing. This is consist-
ent with previous data on test-retest changes after
administration of pharmacological placebo, where
an increase of about 8% was found on the second
testing (Teicher et al., 2000; Teicher, personal
communication). The reduction of head move-
ments in the treatment group was 74% (SEM 7).
In comparison, a probe dose of methylphenidate
(approximately 0.4 mg/kg) reduced the number
of head movements by 62% (Teicher et al., 2000).
Test-retest and the treatment-control com-
parisons were signi®cant for four tests: trained
WM, Span board, Raven's Progressive Matrices,
Stroop accuracy, and number of head movements
(Table 1). Correlation analysis on the test-retest
differences showed that improvement in the two
WM tasks were signi®cantly correlated, as was
the correlation between improvement on Raven's
Progressive Matrices and the trained WM task
(Table 2). The reduction in head movements was
highly correlated with improvements on both the
trained WM task (r
2
:73) and Raven's Pro-
gressive Matrices (r
2
:74).
EXPERIMENT 2
METHODS
Subjects
Four male, healthy volunteer subjects (AP, DE, DH, IK),
aged 23, 29, 20 and 22 years, who were all university
students, participated in the training. The subjects had
no history of psychiatric or neurological disease. The
local ethics committee at the Karolinska Hospital
approved the study.
Tests for Pre- and Posttraining Evaluation
Before and after training, all subjects were tested on
similar tasks as described in Experiment 1: (1) Trained
version of the visuo-spatial WM task; (2) Span board
task; (3) Stroop task; (4) Raven's Advanced Progressive
Matrices; and (5) Choice reaction time task. Posttrain-
ing results on the choice reaction time task was lost
for two of the subjects. In contrast to Experiment 1,
the Advanced Progressive Matrices were given
(Raven, 1990). Eighteen problems (even numbered
problems) were given before testing, and a new set of
18 problems (odd numbered problems) was given
afterwards. No measurement of head movement was
made.
Subjects undertook 5 weeks of training on the same
computerized training program as described in Experi-
ment 1, with an average of 26 days of training.
RESULTS
During training, performance improved gradually
on all trained tasks, with an increased amount of
information kept in WM and decreased reaction-
times (Fig. 2). On the cognitive testing performed
before and after training, improvement was evi-
dent for all subjects and all tasks (Table 3). Adult
784 TORKEL KLINGBERG ET AL.
Table 1. Task Performance in Children With ADHD in the Treatment Group, Before and After Training.
Control Treatment Test-retest in Group difference
b
Before
Mean (SEM)
After
Mean (SEM)
Before
Mean (SEM)
After
Mean (SEM)
treatment group
a
Trained visuo-spatial WM 5.0 (0.22) 4.79 (0.21) 4.71 (0.21) 6.43 (0.41) P .0007 P .0006
Span board
c
4.54 (0.46) 4.93 (0.28) 4.36 (0.12) 6.32 (0.25) P .0001 P .001
Stroop task
Accuracy (max 60) 56.3 (0.8) 55.1 (2.6) 55.4 (1.2) 59.4 (0.3) P .03 P .02
Time for completion (s) 80.3 (6.7) 86.3 (15.1) 101 (7.9) 90.9 (7.2) P .17 P .12
Raven's progressive matrices 28.7 (0.8) 29.3 (1.0) 26.4 (1.2) 33.3 (1.6) P .001 P .001
Choice reaction time task
RT Latency (ms) 314 (18) 342 (30) 282 (22) 296 (31) P .20 P .27
Two ± One choice (ms) 91 (19) 79 (17) 146 (24) 71 (14) P .07 P .05
RT standard deviation 128 (31) 117 (20) 106 (24) 92 (17) P .18 P .49
Number of head movements 1496 (579) 1881 (616) 1001 (269) 315 (148) P .002 P .00008
Note.
a
Test-retest differences (before and after training) in the treatment group (one-tailed t-test).
b
Intra-individual test-retest differences were compared between the treatment group and the placebo group (one-tailed t-test).
c
Mean improvement for both backward and forward version.
TRAINING OF WORKING MEMORY 785
Fig. 1. Test-retest differences in the treatment and placebo group. Bars show mean values for each group. Mar-
kers represent individuals in each group. The group differences were signi®cant for each of the three tasks
(Table 1). (a) Span board task, a nontrained visuo-spatial WM task; (b) Raven's Progressive Matrices; (c)
number of head movements.
786 TORKEL KLINGBERG ET AL.
test-retest values were compared to those of the
control group in Experiment 1. Alternatively, the
test-retest scores for the adults could have been
compared with an adult control group. However,
because it is the change between test and retest
that is compared, the differences in pretest scores
are subtracted away. The test-retest improvement
of the control group in Experiment 1 is also
consistent with previous reports of test-retest
improvements in adults on spatial WM (Lowe &
Rabbitt, 1998) and the Stroop task (Salinsky,
Storzbach, Dodrill, & Binder, 2001). Compared
to the placebo-group in Experiment 1, the impro-
vement was signi®cant for the trained visuo-
spatial WM task, Span board, Stroop time and
Raven's Progressive Matrices (P < .05). However,
all subjects were correct on the highest level
included on the visuo-spatial WM tasks (level 9)
and 2 achieved highest score on the Span board
(level 8, both on the forward and the backward
version), and 2 subjects achieved maximum score
on the Stroop task (100). The differences between
the trained group and the control group in Experi-
ment 1 was thus probably underestimated due to
ceiling effects in the adult group. On Raven's
Progressive Matrices no subject achieved max-
imum score and the improvement was 24.5%,
which is similar to the 26.1% improvement for
Table 2. Correlations (r
2
) Between Test-Retest Differences in the Treatment Group.
Trained WM Corsi block RPM Head movements
Span board .85**
RPM 0.74* 0.55
Head movements 0.73 0.56 0.74
Stroop accuracy 0.42 0.59 ÿ0.22 0.20
Note. RPM Raven's Progressive Matrices.
*P < 0:05; **P < 0:01.
Table 3. Task Performance in Four Healty Adults, Before and After Training.
Before
Mean (SEM)
After
Mean (SEM)
Improvement
by treatment
a
Group difference
b
Trained visuo-spatial WM 7.12 (0.62) 9.0 (0.00) P .03 P .005
Span board
c
5.6 (0.39) 7.25 (0.48) P .02 P .02
Forwards (items) 5.62 (0.71) 7.0 (0.81) P .02 P .02
Backwards (items) 5.62 (0.24) 7.50 (0.35) P .01 P .06
Stroop task
Accuracy (max 100) 98.25 (0.85) 99.5 (0.29) P .10 P .24
Time for completion (s) 109.7 (7.4) 86.7 (4.1) P .01 P .02
Raven's progressive matrices
(max 18) 12.25 (0.25) 15.25 (0.85) P .01 P .04
Choice reaction time task
RT Latency (ms) 248 (16) 220 (5) P .12 P .06
Two ± One choice (ms) 60 (25) 44 (3) P .30 P .43
RT standard deviation 54 (17) 32 (10) P .10 P .42
Note.
a
Paired t-tests of test-retest differences (before and after training) in the treatment group (one-tailed t-test).
b
Intra-individual test-retest differences were compared between the treatment group and the placebo group in
Experiment 1 (one-tailed t-test).
c
Mean improvement for both backward and forward version.
TRAINING OF WORKING MEMORY 787
children in the treatment group in Experiment 1.
Raven's Progressive Matrices is known as a
highly reliable test where test-retest improvement
is typically 0±8% (Raven, 1990). With test-retest
intervals as short as 2 weeks it has been shown to
be 9% (Costenbader & Ngari, 2001). With longer
retest intervals the improvement is less, and in
Experiment 1 the test-retest improvement after 5
weeks was 2%.
DISCUSSION
The present study showed that intensive and adap-
tive, computerized WM training gradually increa-
sed the amount of information that the subjects
could keep in WM (Tables 1 and 3, Figs. 1 and 2).
The improved performance occurred over weeks
of training, and is in this respect similar to the
slow acquisition of a perceptual skill or a motor
skill (Karni et al., 1995; Recanzone, Schreiner, &
Merzenich, 1993; Tallal et al., 1996). Further-
more, the improvement from training was evident
both for a group of children with ADHD (Experi-
ment 1), as well as for adult subjects without
ADHD (Experiment 2). This shows that an initial
de®cit in WM was not necessary for improvement
to occur.
Increased performance was seen for both train-
ed and nontrained visuo-spatial WM tasks, show-
ing that the training effect generalized to other
settings (Tables 1 and 3). Only inconsistent effects
where seen for a choice reaction time task
(Table 1), which is in agreement with a previous
observation that cognitive tasks are more acces-
sible for improvement from training than are
attentional tasks (Sohlberg, McLaughlin, Pavese,
Heidrich, & Posner, 2000). A signi®cant training
effect was also seen for the reasoning task
(Tables 1 and 3, Fig. 1b). Although the improve-
ment on the Span board task was evidence some
transfer of training, it is dif®cult to evaluate the
commonalities of the two tasks, and to exclude
that the subjects were trained towards improve-
ment on the Span board task. However, the
improvement on the reasoning task is a clear
Fig. 2. Improvement during training of the visuo-spatial WM task in four adult subjects. Each graph shows data
from a single individual.
788 TORKEL KLINGBERG ET AL.
evidence of that the training effect generalized to
nonpracticed tasks, since the training did not
include any problem solving or reasoning exer-
cises at all. This is true also for the enhanced
performance on the Stroop task. The improvement
in reasoning ability is likely due to the fact that
complex reasoning depends on WM, or more
precisely, that the trained WM tasks and the
reasoning task rely on the same cortical areas. A
common factor underlying the WM tasks, Raven's
Progressive Matrices and the Stroop task is that
they all depend on the prefrontal cortex (Bench
et al., 1993; Duncan & Owen, 2000; Klingberg,
Forssberg, & Westerberg, 2002; Pardo et al.,
1990; Prabhakaran, Smith, Desmond, Glover, &
Gabrieli, 1997). The association between the
reasoning task and the WM tasks is further subs-
tantiated by the signi®cant correlation between
improvement on the visuo-spatial WM task and
improvement on Raven's Progressive Matrices
(Table 2).
The second main ®nding was that training on
WM tasks signi®cantly reduced the number of
head movements (Table 1, Fig. 1c), with improve-
ments on WM being correlated with the reduction
in movements (Table 2). The measurement of
head movements was an attempt to get an object-
ive measurement of the children's motor activity.
This measure of motor activity has previously
been shown to correlate with behavioral ratings
(Teicher et al., 1996). An obvious drawback of the
test is that it only measures behavior during a
short period of time, and in a laboratory situation.
It would therefore ideally be complemented with
additional measurement of hyperactivity. Never-
theless, the decrease in head movements gives a
clue to the question of how the cognitive de®cits,
the impulsivity and the motor symptoms in ADHD
are related to each other, and whether some symp-
toms are more fundamental than others (Barkley,
1997). The present results, where increased WM
capacity resulted in decreased movements, suggest
that there could be a causal relationship between
cognitive functioning and motor behavior.
The fact that WM capacity including reasoning
ability, could be substantially affected by training
deserves some comment since this seems to
challenge the notion that WM capacity re¯ects a
®xed property of the individual (Engle et al.,
1999; Kyllonen & Christal, 1990). The results
also seem to contradict the notion that tasks such
as Raven's Progressive Matrices measure a ®xed
cognitive ability (Jensen, 1998). However, only
one reasoning test was given, and to test the hypo-
thesis about how experience affects cognitive
ability, a wider battery of tasks would be needed.
Future research will have to investigate to what
extent the training also effects everyday life for
children with ADHD, and document the dur-
ability of the training effects. However, here we
could show that training signi®cantly improved
performance on tasks related to cognitive func-
tion, inhibition (as measured by the Stroop task),
as well as motor activity, which suggests that WM
training could be of interest for future clinical use
in children with ADHD.
ACKNOWLEDGMENTS
We thank Jonas Beckeman and David Skoglund for
programming, graphical design and helpful discussions
on task design. We thank Russell Poldrack for valuable
comments on this manuscript, and Elisabeth Fernell,
Kirsten Holmberg, Maria Silverberg, and Christer
Hurve for patient referrals, and Viveka Johansson for
assistance during testing. This work was funded by
Jeansson Stiftelse, The Swedish Medical Research
Foundation, Hja
È
rnfonden, Svenska Dyslexifo
È
reningen,
Sven Jerrings Stiftelse, Frimurarna Barnahuset, and
Sa
È
llskapet Barnava
Ê
rd.
REFERENCES
American Psychiatric Association. (1994). Diagnostic
and statistical manual of mental disorders (4th ed.).
Washington, DC: Author.
Barkley, R.A. (1997). Behavioral inhibition, sustained
attention, and executive functions: Constructing a
unifying theory of ADHD. Psychological Bulletin,
121, 65±94.
Bench, C.J., Frith, C.D., Grasby, P.M., Friston, K.J.,
Paulesu, E., Frackowiak, R.S., & Dolan, R.J. (1993).
Investigations of the functional anatomy of atten-
tion using the Stroop test. Neuropsychologia, 31,
907±922.
Brown, A.L., Campione, J.C., Bray, N.W., & Wilcox,
B.L. (1973). Keeping track of changing variables:
Effects of rehearsal training and rehearsal preven-
tion in normal and retarded adolescents. Journal of
Experimental Psychology, 101, 123±131.
TRAINING OF WORKING MEMORY 789
Buonomano, D.V., & Merzenich, M.M. (1998). Corti-
cal plasticity: From synapses to maps. Annual
Reviews of Neuroscience, 21, 149±186.
Butter®eld, E.C., Wambold, C., & Belmont, J.M.
(1973). On the theory and practice of improving
short-term memory. American Journal of Mental
De®ciency, 77, 654±669.
Costenbader, V., & Ngari, S.M. (2001). A Kenya
Standardization of the Raven's Coloured Progres-
sive Matrices. School Psychology International, 22,
258±268.
Duncan, J., & Owen, A.M. (2000). Common regions of
the human frontal lobe recruited by diverse
cognitive demands [In Process Citation]. Trends
Neuroscience, 23, 475±483.
Engle, W.R., Kane, J.M., & Tuholski, S.W. (1999).
Individual differences in working memory capa-
city and what they tell us about controlled atten-
tion, general ¯uid intelligence, and functions of
the prefrontal cortex. In A. Myake & P. Shah
(Eds.), Models of working memory (pp. 102±134).
Cambridge: Cambridge University Press.
Ericsson, K.A., Chase, W.G., & Faloon, S. (1980).
Acquisition of a memory skill. Science, 208, 1181±
1182.
Fuster, J.M. (1989). The prefrontal cortex. New York:
Raven Press.
Goldman-Rakic, P.S. (1987). Circuitry of primate pre-
frontal cortex and regulation of behavior by repre-
sentational memory. In V.B. Mountcastle (Ed.),
Handbook of physiology (pp. 373±417). Bethesda,
MD: American Physiological Society.
Hulme, C. (1992). Working memory and severe learning
dif®culties. Hove: Lawrence Erlbaum Associates.
Hulme, C., & Roodenrys, S. (1995). Practitioners
review: Verbal working memory development and
its disorders. Journal of Child Psychology and
Psychiatry, 36, 373±398.
Jensen, A.R. (1998). The g factor: The science of
mental ability. Westport, Conneticut: Praeger.
Karni, A., Meyer, G., Jezzard, P., Adams, M.M., Turner,
R., & Ungerleider, L.G. (1995). Functional MRI
evidence for adult motor cortex plasticity during
motor skill learning. Nature, 377, 155±158.
Klingberg, T., Forssberg, H., & Westerberg, H. (2002).
Increased brain activity in frontal and parietal cortex
underlies the development of visuo-spatial working
memory capacity during childhood. Journal of
Cognitive Neuroscience, 14, 1±10.
Klingberg, T. (2000). Limitations in information pro-
cessing in the human brain: Neuroimaging of dual
task performance and working memory tasks.
Progress in Brain Research, 126, 95±102.
Kristofferson, M.W. (1972). Effects of practice on
character-classi®cation performance. Canadian
Journal of Psychology, 26, 54±60.
Kuntsi, J., Oosterlaan, J., & Stevenson, J. (2001).
Psychological mechanisms in hyperactivity: I. Res-
ponse inhibition de®cit, working memory impair-
ment, delay aversion, or something else? Journal of
Child Psychology and Psychiatry, 42, 199±210.
Kyllonen, P.C., & Christal, R.E. (1990). Reasoning
ability is (little more than) working-memory capa-
city?! Intelligence, 14, 389±433.
Lowe, C., & Rabbitt, P. (1998). Test/re-test reliability
of the CANTAB and ISPOCD neuropsycholo-
gical batteries: Theoretical and practical issues.
Cambridge Neuropsychological Test Automated
Battery. International Study of Post-Operative Cogni-
tive Dysfunction. Neuropsychologia, 36, 915±923.
Mariani, M.A., & Barkley, R.A. (1997). Neuropsy-
chological and academic functioning in preschool
boys with attention de®cit hyperactivity disorder.
Developmental Psychology, 13, 111±129.
Pardo, J.V., Pardo, P.J., Janer, K.W., & Raichle, M.E.
(1990). The anterior cingulate cortex mediates
processing selection in the Stroop attentional con-
¯ict paradigm. Proceedings of the National Acad-
emy of Sciences of the United States of America, 87,
256±259.
Phillips, C.J., & Nettelbeck, T. (1984). Effects of
practice on recognition memory of mildly mentally
retarded adults. American Journal of Mental
De®ciency, 88, 678±687.
Prabhakaran, V., Smith, J.A., Desmond, J.E., Glover,
G.H., & Gabrieli, J.D. (1997). Neural substrates of
¯uid reasoning: An fMRI study of neocortical
activation during performance of the Raven's
Progressive Matrices Test. Cognitive Psychology,
33, 43±63.
Rapport, M.D., Chung, K.M., Shore, G., Denney, C.B., &
Isaacs, P. (2000). Upgrading the science and technol-
ogy of assessment and diagnosis: Laboratory and
clinic-based assessment of children with ADHD.
Journal of Clinical Child Psychology, 29, 555±568.
Raven, J.C. (1990). Advanced Progressive Matrices.
Set II. Oxford, United Kingdom: Oxford Psychol-
ogists Press.
Raven, J.C. (1995). Coloured Progressive Matrices.
Oxford, United Kingdom: Oxford Psychologists
Press.
Recanzone, G.H., Schreiner, C.E., & Merzenich, M.M.
(1993). Plasticity in the frequency representation of
primary auditory cortex following discrimination
training in adult owl monkeys. Journal of Neu-
roscience, 13, 87±103.
Rubia, K., Overmeyer, S., Taylor, E., Brammer, M.,
Williams, S.C., Simmons, A., & Bullmore, E.T.
(1999). Hypofrontality in attention de®cit hyper-
activity disorder during higher-order motor control:
A study with functional MRI. American Journal of
Psychiatry, 156, 891±896.
790 TORKEL KLINGBERG ET AL.
Salinsky, M.C., Storzbach, D., Dodrill, C.B., & Binder,
L.M. (2001). Test-retest bias, reliability, and regre-
ssion equations for neuropsychological measures
repeated over a 12±16-week period. Journal of
the International Neuropsychology Society, 7, 597±
605.
Schweitzer, J.B., Faber, T.L., Grafton, S.T., Tune,
L.E., Hoffman, J.M., & Kilts, C.D. (2000). Altera-
tions in the functional anatomy of working
memory in adult attention de®cit hyperactivity
disorder. American Journal of Psychiatry, 157,
278±280.
Sohlberg, M.M., McLaughlin, K.A., Pavese, A.,
Heidrich, A., & Posner, M.I. (2000). Evaluation of
attention process training and brain injury education
in persons with acquired brain injury. Journal of
Clinical and Experimental Neuropsychology, 22,
656±676.
Tallal, P., Miller, S.L., Bedi, G., Byma, G., Wang, X.,
Nagarajan, S.S., Schreiner, C., Jenkins, W.M., &
Merzenich, M.M. (1996). Language comprehension
in language-learning impaired children improved
with acoustically modi®ed speech. Science, 271,
81±84.
Teicher, M.H., Anderson, C.M., Polcari, A., Glod, C.A.,
Maas, L.C., & Renshaw, P.F. (2000). Functional
de®cits in basal ganglia of children with attention-
de®cit/hyperactivity disorder shown with functional
magnetic resonance imaging relaxometry. Nature
Medicine, 6, 470±473.
Teicher, M.H., Ito, Y., Glod, C.A., & Barber, N.I.
(1996). Objective measurement of hyperactivity
and attentional problems in ADHD. Journal of
American Academy for Child and Adolescent Psy-
chiatry, 35, 334±342.
Westerberg, H., Hirvikoski, T., Forssberg, H., & Kling-
berg, T. (submitted). Visuo-spatial working mem-
ory: A sensitive measurement of cognitive de®cits in
ADHD.
Zametkin, A.J., Nordahl, T.E., Gross, M., King, A.C.,
Semple, W.E., Rumsey, J., Hamburger, S., & Cohen,
R.M. (1990). Cerebral glucose metabolism in adults
with hyperactivity of childhood onset. New England
Journal of Medicine, 323, 1361±1366.
TRAINING OF WORKING MEMORY 791
