Brain Injury

ISSN: 0269-9052 (Print) 1362-301X (Online) Journal homepage: www.tandfonline.com/journals/ibij20

Rehabilitation of verbal working memory after left

hemisphere stroke

C. Vallat, P. Azouvi, H. Hardisson, R. Meffert, C. Tessier & P. Pradat-Diehl

To cite this article: C. Vallat, P. Azouvi, H. Hardisson, R. Meffert, C. Tessier & P. Pradat-Diehl
(2005) Rehabilitation of verbal working memory after left hemisphere stroke, Brain Injury,
19:13, 1157-1164, DOI: 10.1080/02699050500110595

To link to this article: https://doi.org/10.1080/02699050500110595

Published online: 22 Nov 2010.

Submit your article to this journal

Article views: 1066

View related articles

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ibij20
Brain Injury, December 2005; 19(13): 1157–1164

CASE STUDY

Rehabilitation of verbal working memory after left hemisphere stroke

C. VALLAT1, P. AZOUVI2, H. HARDISSON3, R. MEFFERT3,

C. TESSIER3, & P. PRADAT-DIEHL3
1
Antenne UEROS-UGECAN, Hôpital Raymond Poincare´, Garches, France, 2Service de Re´e´ducation Neurologique
Hôpital Raymond Poincare´, Universite´ de Versaille Saint Quentin, Garches, France, and 3Service de Me´decine Physique
et Re´adaptation, Hoˆpital de la Salpe´trie`re, Paris, France

(Received 20 July 2004; accepted 14 February 2005)

Abstract
Primary objective: To assess a rehabilitation programme of working memory.
Research design: Single case multiple-baseline-across-behaviour design in a patient with chronic stroke suffering from a
specific impairment of working memory.
Methods: Rehabilitation included training of storage and processing components of verbal working memory. Outcome was
assessed with specific working memory tests, non-specific tasks requiring working memory, non-target tasks not requiring
working memory and questionnaires addressing daily-life functioning. This assessment was performed twice before the
experimental therapy and once after.
Results: Cognitive measures were stable before therapy. A statistically significant improvement was found for target measures
and for daily-life skills. In contrast, no change was found on non-target tasks.
Conclusion: Improvement was not apparently related to spontaneous recovery, since performance was stable before therapy.
This study suggests that specific cognitive training may improve working memory in patients with brain injury.

Keywords: Working memory, rehabilitation, stroke

Introduction The phonological loop is responsible for the proces-

Working memory is used as a system for temporary
maintenance and manipulation of information.
Baddeley and Hitch [1] and Baddeley [2] described
a working memory model divided into three compo-
nents. The central executive is an attentional control
system, while the phonological loop and the
visuo-spatial sketchpad are two modality-specific
slave systems. The central executive supervises and
co-ordinates the regulation of information flow and
controls strategy switching. The central executive is
more involved in processing than in temporary
storage [3] and has a limited capacity. Furthermore,
it also serves as an interface between the two slave
systems. This component is assumed to be similar
to the supervisor attentional system (SAS) of
Norman and Shallice’s [4] attentional model.
sing and maintenance of verbal information.
This system comprises two components: the phono-
logical store in which representations of verbal
material are held and a sub-vocal rehearsal mechan-
ism which serves to refresh the contents of the
phonological store. The visuo-spatial sketchpad is
devoted to the processing of visual and spatial
information. Although this component has received
less attention, it is currently assumed to com-
prise two components: a ‘passive’ visual store and
an active mechanism to refresh the contents of the
visuo-spatial store.
Working memory is involved in many cognitive
processes. The central executive is assumed to be
involved in higher order mental functions, such as
logical reasoning, mental calculation, text compre-
hension [5] and in the ability to understand and

Correspondence: Claire Vallat, Antenne UEROS-UGECAN, Service de Rééducation Neurologique, Hôpital Raymond Poincaré, 92380 Garches, France.
Tel: þ 331 47 10 76 47. Fax: þ33 1 47 10 70 73. E-mail: claire.vallat@rpc.ap.hp.fr
ISSN 0269–9052 print/ISSN 1362–301X online # 2005 Taylor & Francis
DOI: 10.1080/02699050500110595
1158 C. Vallat et al.
manipulate both spoken and written language. It
is also implicated in control-type functions and
mainly in non-routine processes [6]. The main func-
tions of the phonological loop are to provide support
for the acquisition of new vocabulary, long sentences
comprehension [7] and repetition in order to main-
tain and refresh information in a short-term store.
The function of the visuo-spatial sketchpad is less
well known.
The more frequently reported causes of working
memory deficits in adults are traumatic brain injury
(TBI) and Alzheimer’s disease, both associated
with central executive impairments [8, 9]. A few
cases of phonological loop impairments have been
reported after left hemisphere stroke [7, 10].
However, the incidence of working memory deficits
after stroke is probably under-estimated and has
not been systematically studied. Furthermore, resi-
dual impairments frequently reported by recovered
aphasic patients could be related to a working
memory deficit: difficulty to hold a conversation for
more than a few minutes, particularly with a high
speech flow, to understand complex sentences or
long written texts, to participate in a conversation
between a number of people or to do two things
simultaneously (talking while walking for example).
However, there has been, to date, very little
research on rehabilitation of working memory.
Cicerone [11] has recently reported successful work-
ing memory rehabilitation in a study of five patients
with TBI. In addition, a few rehabilitation studies
of attention have included training tasks addressing
components of working memory, mainly the central
executive. For example, Sturm et al. [12] proposed
a rehabilitation programme focused on specific
attentional modules, including divided attention,
which is currently thought to rely on the central
executive of working memory. However, it is difficult
to disentangle in these studies the effect on atten-
tional processes and on working memory itself.
Considering the dramatic consequences of
working memory deficits on higher level cognitive
processes and its impact on daily-life skills, a rehabili-
tation programme of working memory was developed.
This programme was designed within Baddeley’s
[2, 6] theoretical framework and mainly addressed
the central executive and the phonological loop. A
single case design was used, with multiple baselines
across behaviours, including an assessment of the
generalization to daily life activities. The objective of
the present study was to assess both the efficacy and
the specificity of this working memory therapy in a
patient with a chronic stroke suffering from a selective
impairment of the central executive and the
phonological loop.
Case report
CPP is a 53 year-old right-handed, high school
graduate male. He was a computer scientist and
managing director of a computer business. In
November 1997, he suffered a left parietal lobe
infarction, responsible for an isolated conduction
aphasia, without any other neurological deficit.
Conventional speech therapy was undertaken and
his language abilities progressively and dramatically
improved. However, CPP still complained of diffi-
culties in everyday life, suggesting an impairment of
working memory. He reported a difficulty to hold a
conversation involving several persons or to take
notes during professional meetings. It was also
difficult for him to do shopping because he tended
to forget some of the items he was supposed to buy
and to make mistakes in evaluating the total cost of
his purchases. These difficulties contrasted with his
apparently recovered language functions and his
ability to solve complex cognitive tasks, such as
difficult arithmetic problems. In March and April
1998, 5 months post-onset, a first neuropsychologi-
cal assessment was carried out (which will be
presented in detail later) which showed a working
memory deficit, mainly involving the verbal sub-
system and the central executive. He then continued
to receive conventional speech therapy for a further
9 months. In January 1999, 14 months post-onset,
CPP intended to return to work, but still had the
same complaints and a second neuropsychological
assessment (that will be presented in detail later) was
carried out, which showed that working memory
impairments had not improved. This deficit was
quite isolated, without any accompanying major
deficit of other cognitive functions, including lan-
guage, long-term memory and attention. Indeed,
CPP obtained a normal performance on a widely
used French aphasia battery [13], he performed at
ceiling on a naming test (score 80/80 on the test de
dénomination orale 80 [14]) and his oral compre-
hension, as assessed with the Token test (score:
36/36) was normal as well. He also performed within
normal limits on two tests of verbal long-
term memory, the paired associates sub-test of the
Wechsler Memory Scale-revised (WMS-R)
[15] (score: 23/24) and the Rey Auditory Verbal
Learning Test [16] (score, free recall: 54/60).
Attention was assessed with a computerized atten-
tional assessment battery [17], using the alertness
and selective attention sub-tests, which were both
performed within normal limits (standard notes for
reaction time, respectively, 61 and 51). The only test
that was found to be slightly impaired was visual
long-term memory, as assessed with Rey complex
figure recall [18] (score: 20/36). It was then decided

to start an experimental therapy, with special
emphasis on working memory.
Rehabilitation method
A specific cognitive retraining of the central execu-
tive and the phonological loop was used. Eight
different tasks were used, each one arranged in a
difficulty hierarchy order organized following two
criteria, capacity and level of processing. Capacity
was related to the amount of information to be stored
(e.g. number of words, syllables or phonemes). The
level of cognitive processing was related to stimulus
complexity (e.g. frequency of usage, concreteness).
Stimuli were presented auditorily, only once, and at
a regular speed (1 item per second) and responses
were given orally.
CPP received three 1-hour training sessions per
week for 6 months. In each session each one of
the eight tasks was trained in the same order. The
difficulty level was progressively adapted to CPP’s
performance, starting at a 3-item level (i.e. CPP’s
n-1 backward digit span). Each level was trained
until the patient succeeded in 90% of the trials.
There were eight types of tasks. For the first five
tasks described below there were six levels of diffi-
culty depending on the number of syllables (one,
two or three syllables) and the degree of imagery
(concrete or abstract). In each one of these six
levels, words were presented in increasing order of
number of letters and decreasing order of frequency.
Before starting the therapy, the feasibility of these
tasks was assessed in 10 control subjects, matched
for age, sex and educational level to CPP. All tasks
were completed at ceiling by control subjects.
Following is a description of the eight tasks:
. Reconstitution of words from oral spelling: The
patient had to reconstruct and pronounce a word
that was spelled by the therapist.
. Reconstitution of words from oral spelling with a letter
omitted: This is the same task as the previous one,
but with a letter omitted, replaced by a ‘bip’. For
example, at level 3 with a two syllable concrete
word: the item spelt is p-a.‘bip’-i-e-r- and the
subject had to find: ‘papier’.
. Oral spelling.
. Odd or even number of letters in a word: The task was
to find whether the number of letters in a word
heard was odd or even. For example, at level 5
with a three syllable concrete word: [kanape],
‘canapé’ is heard and the answer must be ‘even’.
. Reconstitution of words from syllables: The therapist
presented cluttered syllable words that the subject
had to reorder and pronounce. There were three,
four or five syllables per word. For example,
Rehabilitation of verbal working memory 1159
at level 1 with a three syllable concrete word:
therapist spelled [ta/pã/lõ], ‘ta-pan-lon’ and the
subject have to answer [pãtalõ], ‘pantalon’.
. Alphabetic way: The therapist presented a letter
and a calculation (addition or subtraction) and the
patient had to find the target letter following the
increasing or decreasing alphabetical order (for
example, B þ 6 ¼ H; E
2 ¼ C). There were four
levels of difficulty, depending on the numbers to
be added or subtracted (either short numbers,
below 5 or long numbers, from 6–9).
. Word sorting in alphabetic order: The patient was
required to sort a series of concrete words into
alphabetic order. There were nine levels of
difficulty, depending on the number of words
(three, four or five) and of syllables (one, two or
three). For example, at level 1 with three one-
syllable-words: [s"l/vwal/mul] ‘sel,voile,moule’
was heard and the answer was: [mul/s"l/vwal].
. Acronyms: The task was to find the word made
from the initial phonemes of a list of words. Five
levels of difficulty depending on the number of
words (two to five) and the number of syllables
(one or two) were designed. For example, at level 1
with two words of one syllable: [bal/wa], ‘bal, oie’
was heard and the answer was [bwa], ‘bois’.
Experimental design and outcome measures
A single-case methodology with a multiple-baseline-
across-behaviour design was used [19]. As previously
stated, the therapy was started 14 months post-onset.
At that time, spontaneous recovery was unlikely.
Moreover, the working memory deficit had not
improved during the previous 9 months, despite
conventional speech therapy. The outcome measures
were similar to those used on baseline but, to limit
retest effects, parallel versions of tests were used. In
addition, to control for a non-specific effect of
training, non-target measures were also taken.
These measures were assumed not to require work-
ing memory and were not specifically trained.
Hence, they were not expected to improve after
therapy. The outcome measures that will be pre-
sented hereafter were obtained just after the end
of the therapy. A 3-month follow-up was initially
planned, but could not be done, because CPP had
returned to work at that time.
Three sets of tasks were used: (i) cognitive tasks
specifically designed to assess the different compo-
nents of working memory; (ii) related verbal complex
tasks, i.e. tasks that were not specifically designed to
assess working memory, but that nevertheless
included a high working memory load (text compre-
hension and arithmetic problem solving); and
1160 C. Vallat et al.
Table I. Digit spans. Correct responses (%) (out of five trials) at each span length for controls (n ¼ 10) and for patient CPP. For control
subjects, results are the mean (SD).
2 3
Forward digit span
Controls 100 (0.0) 100 (0.0)
Baseline 1 100 80
Baseline 2 100 100
Post-treatment 100 100
Backward digit span
Controls 100 (0.0) 100 (0.0)
Baseline 1 100 100
Baseline 2 100 100
Post-treatment 100 100
(iii) ecological questionnaires, designed to assess
the consequences of working memory deficit in
everyday life. For all cognitive tasks, three parallel
versions were designed in order to minimize re-test
effects. For specific working memory tasks, CPP’s
performance was compared to a group of 10 matched
control subjects, of similar age and education
background (mean age ¼ 48.2, SD ¼ 6.2, range:
40–56, all subjects having 12 years or more of
education).
Statistical analysis used corrected chi-square tests
on the total number of correct responses. For each
measure, two separate tests were performed, one to
compare performance on the two baseline sessions
and one to compare performance on the second
baseline and the post-therapy assessment.
Results
As previously stated, two baseline assessments of
working memory were obtained, respectively, 5 and
14 months post-onset, to control for re-test effects
and/or spontaneous recovery. Outcome assess-
ment was completed after the end of the 6-month
therapy.
Cognitive tasks specifically designed to assess
the different components of working memory
Digit and visuo-spatial spans. Both forward and
backward digit and visuo-spatial spans were assessed.
Five trials were given for each span length, starting
with two items. The test ended when the patient
failed the five consecutive trials for a given length.
The measures were the percentage of correct
responses out of five trials for each span length.
Visuo-spatial spans were within the normal range
(maximal level achieved: 5 for both forward and
backward visuo-spatial spans) and were not signifi-
cantly modified post-therapy. In contrast, CPP’s
forward and backward digit spans were both
Span length
4 5 6 7
100 (0.0) 96.6 (7.0) 86.6 (17.2) 64.9 (31.8)
80 40 60 0
80 60 60 0
100 100 100 80
94.9 (8.1) 53.3 (39.9) 38.3 (46.5) 9.9 (21.1)
40 0 0 0
60 0 0 0
80 80 60 0
dramatically reduced, in comparison to matched
controls (Table I). This impairment was particularly
marked for backward digit span. Chi-square statistics
were used to compare the percentage of correct
responses between the first and the second baseline
assessments and between the second baseline and
post-therapy. Forward and backward digit spans
were stable before the therapy (corrected
2
(df ¼ 1) ¼ 0.07 and 0.07, respectively, both p > 0.1).
Forward span improved significantly after training
(corrected
2 ¼ 7.1, p < 0.01), while backward span
improvement fell just short of significance (corrected

2 ¼ 3.3, p ¼ 0.06).
Assessment of the phonological loop. This was only
performed on the second baseline assessment. As
previously stated (see Introduction), the phonologi-
cal loop includes two sub-components, the phono-
logical store and the sub-vocal rehearsal process. The
phonological store was tested using the phonological
similarity effect [2]. According to Baddeley, short-
term recall of phonologically close letters (such as t,
p, b, v), is lower than phonologically dissimilar letters
(such as r, m, l, s). This effect is usually attributed to
the phonological store. Again, this was assessed with
five trials at each span length. As shown in Table II,
CPP’s letter span was impaired, as compared to
controls, for dissimilar letters, but lay within normal
limits for phonologically close letters. He did not
show any phonological similarity effect. Indeed,
while, as expected, control subjects performed
better for phonologically dissimilar letters, CPP
presented the reverse pattern of performance
(Table II). This suggested an impairment of the
phonological store [6]. Post-therapy assessment
showed a slight non-significant improvement of
both phonologically close and dissimilar letter
spans (corrected
2 ¼ 0.9 and 0.05, respec-
tively, both p > 0.1). CPP continued to show a
better performance for phonologically close letters
(Table II).
 Rehabilitation of verbal working memory 1161

Table II. Letter spans. Correct responses (%) (out of five trials) at each span length for controls (n ¼ 10) and for patient CPP. For control
subjects, results are the mean (SD).

Span length

2 3 4 5 6 7 8

Phonologically close letters

Controls 100 (0.0) 100 (0.0) 100 (0.0) 85.0 (26.8) 55.0 (38.7)
Baseline 2 100 100 100 60 40 0
Post-treatment 100 100 100 100 60 40 0
Phonologically dissimilar letters
Controls 100 (0.0) 100 (0.0) 100 (0.0) 95.0 (10.5) 77.5 (32.2) 42.5 (44.2) 20.0 (34.9)
Baseline 2 100 100 100 40 20 0
Post-treatment 100 100 100 60 40 0

Table III. Word spans. Correct responses (%) (out of five trials) at each span length for controls (n ¼ 10) and for patient CPP. For control
subjects, results are the mean (SD).

Span length

2 3 4 5 6 7 8

Short words
Controls 100 (0.0) 100 (0.0) 100 (0.0) 77.5 (14.2) 35.0 (33.7) 12.5 (21.2) 0.0 (0.0)
Baseline 2 100 100 100 100 40 0
Post-treatment 100 100 100 100 100 0
Long words
Controls 100 (0.0) 100 (0.0) 67.5 (26.5) 10.0 (17.5) 0.0 (0.0)
Baseline 2 100 100 0 0
Post-treatment 100 100 80 0

The sub-vocal rehearsal process was assessed Table IV. Brown–Peterson task. Correct responses (%) at three
using the word-length effect, with monosyllabic and recall delays for controls (n ¼ 10) and for patient CPP. For control
quadri-syllabic words. As can be seen in Table III, subjects, results are the mean (SD).
CPPs short-word span was normal, but his long- Recall delay
word span was impaired comparatively to controls.
5 seconds 10 seconds 20 seconds
However, he showed a preserved word-length
effect, as indicated by a higher span for short than Articulatory suppression
for long words (higher level achieved, respectively, Controls 100 (0.0) 100 (0.0) 100 (0.0)
6 and 3), similarly to controls, suggesting that the Baseline 1 57 66 16
Baseline 2 83 50 33
sub-vocal rehearsal process was preserved. A slight Post-treatment 66 66 66
post-therapy improvement occurred for both short Digit addition
and long word spans, but that again did not reach Controls 92.0 (13.9) 86.0 (16.5) 76.0 (27.9)
statistical significance (corrected
2 ¼ 0.4 and 0.9, Baseline 1 33 0 33
both p > 0.1) (Table III). Baseline 2 16 33 0
Post-treatment 66 83 100

Assessment of the central executive system. The central

executive was tested using the Brown Peterson
paradigm [20]. Subjects were asked to recall con-
sonant trigrams after three delays (5, 10 or
20 seconds) with or without an interfering task
(motor task, articulatory suppression and digit addi-
tion). While CPPs performance without interference
or with a motor task was at ceiling, it dropped
dramatically, as compared to controls, with more
demanding interfering tasks (articulation or digit
addition) and at all recall delays (5 to 20 seconds)
(Table IV). The scores were not modified from the
first to the second baseline sessions. In contrast, after
rehabilitation, performance improved dramatically,
particularly under the most demanding condition,
with digit addition (Table IV). This effect was
statistically highly significant (corrected
2 ¼ 13.4,
p < 0.001). In this most difficult condition, CPPs
post-therapy performance was close to ceiling and did
not significantly differ from that of controls. In
contrast, although there was also a mild improv-
ement under the easier condition (articulatory
1162 C. Vallat et al.
suppression), this effect was not statistically signi-
ficant (corrected
2 ¼ 0.1, p > 0.1).
In summary, these results suggested that CPP
suffered from an impairment of working memory,
mainly involving the central executive and the
phonological store, that was stable before treatment
and that significantly improved after rehabilitation.
Related complex tasks
Text oral comprehension was assessed with 10
sentences of different length and syntactic complex-
ity, designed following the methodology of a widely
used French aphasia battery [13]. Performance
was slightly impaired (correct responses ¼ 80%).
Arithmetic problem solving was assessed with 15
problems of increasing complexity, close to those
used in the arithmetic sub-test of the Wechsler Adult
Intelligence Scale–revised [21]. Performance was
severely impaired (correct responses on the two
successive baselines ¼ 20% and 33%). These two
tasks were performed without difficulty (100%
correct responses) by matched control subjects.
Post-treatment, there was a trend for a mild
improvement on both tasks (correct responses,
respectively, 100% and 66%) that did not reach
statistical significance, presumably due to a ceiling
effect for text comprehension.
Ecological questionnaire
Two questionnaires were used to assess the con-
sequences of working memory impairments on daily-
life activities. A verbal communication questionnaire
(L’Echelle de communication verbale de Bordeaux:
ECVB) [22] was used, because CPPs difficulties
were mainly related to the verbal components of
working memory. This questionnaire includes 38
questions addressing eight domains of verbal com-
munication, such as shopping, social interaction or
phone conversation. For each item, a 4-level scale
was used to assess the frequency of occurrence of the
trouble (0 ¼ always, 3 ¼ never). For clarity of pre-
sentation, scores in each domain were recoded to
obtain a maximum score of 100, higher scores
indicating better functioning. Difficulties were
found in all daily-life situations addressed by the
questionnaire, mainly in conversation, phone use
Table V. Verbal communication questionnaire. Mean score (maximum score ¼ 100, higher scores indicating better functioning).
Intention
Total expression Conversation
Baseline 69 89 54
Post-treatment 83 100 63
and shopping including money use (Table V),
contrasting with CPPs good performance on tradi-
tional aphasia testing.
In addition, a specific questionnaire was designed,
to assess working memory functioning in various
daily-life situations. This questionnaire included 16
questions addressing three different domains:
storage/processing aspects of working memory
(eight questions, e.g. ‘do you have difficulty in recal-
ling a series of digits, such as a telephone number?’
or ‘do you have difficulty in understanding what
you read?’); attentional aspects of working memory
(six questions, e.g. ‘do you have difficulty in doing
something with background noise?’ or ‘does fatigue
affect your concentration?’); executive aspects of
working memory (two questions, such as ‘do you
have difficulty in doing two things simultaneously?’).
A dichotomous scale was used (0 ¼ difficulty present,
1 ¼ no difficulty). A total score (maximum ¼ 16) and
three sub-scores (storage/processing; attention;
executive) were computed. Again, for clarity of
presentation, all scores were recoded to obtain a
maximum score of 100, higher scores indicating
better functioning. As shown in Table VI, CPP
reported difficulties for the three sub-scales.
An improvement was found on both question-
naires (verbal communication questionnaire, total
score:
2 ¼ 6.2; p ¼ 0.01; working memory question-
naire:
2 ¼ 4.800; p < 0.05) (Tables V and VI). CPP’s
complaints in everyday life were significantly reduced
after treatment. In addition, after rehabilitation,
CPP returned to full-time work at his previous
level. He was contacted by phone 3 months post-
return to work. Unfortunately, he was too busy to
complete a follow-up neuropsychological assess-
ment. Nevertheless, he was working full-time and
did not further mention any significant difficulty
either at home or in his job.
Table VI. Working memory questionnaire. Mean score for each
category (maximum score ¼ 100, higher scores indicating better
functioning).
Total Storage Attention Executive
Baseline 43.7 50 33.3 50
Post-treatment 81.2 100 50 100
Phone Social
conversation Shopping interaction Reading Writing Calculation
57 67 81 92 83 0
81 83 95 100 100 0

Non-target measures
Non-target measures included verbal fluency and
long-term memory tests. Two conditions were used
to assess verbal fluency [23]. For each one, the
patient was asked to spell out as many words as
possible within 2 minutes, first animals then words
starting with a ‘p’. Scores were, respectively, 24
and 16. Visual long-term memory was assessed
with Complex Figure recall [18, 24] (score: 20/36).
Verbal long-term memory was assessed with the Rey
Auditory Verbal Learning Test [16], including 12
words repeated five times for a maximum score of 60
(free recall score 54/60). None of these non-target
measures changed after rehabilitation.
Discussion
The present study suggested an effectiveness of a
specific therapy focusing on verbal and attentional
aspects of working memory and its generalization to
daily-life activities in a patient with a chronic stroke
suffering from a selective impairment of the central
executive and the phonological loop. CPP presented,
1 year after a left hemisphere stroke, an isolated
impairment of working memory. This impairment
mainly involved verbal aspects of working memory,
while the visuo-spatial sketchpad seemed to be
preserved. The patient’s spontaneous complaints
mainly concerned complex resource-demanding
verbal tasks. Neuropsychological testing showed
reduced forward and backward digit spans, an
impairment of the phonological store (as shown by
a lack of the phonological similarity effect) and a
poor performance in the Brown–Peterson task with
demanding interference tasks. Visuo-spatial spans
were within normal limits. These results suggested
an impairment of the phonological store and of the
central executive system. This deficit had serious
consequences on his everyday life skills. On formal
testing, deficits were found on complex tasks requir-
ing working memory, such as text comprehension
and arithmetic problem-solving. When interviewed
with standardized ecological questionnaires, CPP
reported difficulties in a wide range of everyday
life situations, such as shopping, conversations or
phone use.
This deficit did not improve despite 1 year of
conventional speech therapy. Indeed, two baseline
measures of working memory were taken 5 and 14
months post-stroke onset and did not show any
improvement. At that time, when spontaneous
recovery seemed very unlikely, a specific rehabilita-
tion programme was undertaken. The training tasks
were arranged in a progressive hierarchical order of
difficulty, regarding both storage and processing
aspects of working memory.
Rehabilitation of verbal working memory 1163
Post-therapy assessment showed a significant
improvement of the main target measures, particu-
larly forward and backward digit spans, which were
close to the tasks used during training, but also of
the Brown–Peterson task, using a procedure that
was not trained during therapy. There was also a
trend for an improvement of other related but
untrained tasks (arithmetic problem-solving). More
importantly, an improvement was also found in
everyday life situations requiring working memory,
as rated by the ecological questionnaires. In addition,
after this experimental rehabilitation, CPP returned
to full-time work at his previous level. By contrast,
there was no improvement of non-target measures
that were not trained and were not expected to
improve, such as visuo-spatial spans (visuo-spatial
working memory was not directly trained) and
other tasks addressing different cognitive domains,
such as long-term memory or verbal fluency.
Hence, CPP’s improvement could not readily be
attributed to spontaneous recovery, non-specific
cognitive stimulation or to re-test effects. Indeed,
the deficits had been stable for many months before
the experimental rehabilitation started, despite
conventional speech therapy and, moreover, the
improvement was limited to the domain that was
specifically addressed by the treatment. In addition,
to control for any re-test effect, parallel forms of
tests were used in the different testing sessions and
the outcome measures were always different from
the tasks used for the therapy. These results suggest
that the training resulted in a specific improvement
of verbal and attentional aspects of working
memory, with a generalization to untrained tasks
and a transfer to everyday life skills.
To the authors’ knowledge, there has been only
little research on the rehabilitation of working
memory. Cicerone [11] reported a remediation study
of the central executive after mild TBI. Four patients
were included and the outcome measures were the
Paced Auditory Serial Addition Test (PASAT) and
an attentional questionnaire. The present study has
several similarities with Cicerone’s study, the same
theoretical model was used and the training
method was also derived from previous studies on
rehabilitation of attention. However, the training
tasks that were used by Cicerone [11] were different
and consisted of n-back, random generation and
dual tasks, with special emphasis on allocation of
attentional resources and management of informa-
tion processing speed, similar to time pressure
management [25]. By contrast, the training strategy
was based on a direct restoration strategy of impaired
working memory.
There has been more research on the rehabilitation
of attention, which is a concept closely related to
working memory. However, in most of these studies,
1164 C. Vallat et al.
attention was trained as a whole, without clearly deli-
neating effects on the different attentional domains.
In a recent review, Park and Ingles [26] concluded
that there is only limited evidence of the effectiveness
of such rehabilitation programmes. Sturm et al. [12]
recently emphasized that different attentional deficits
should be treated with specific training methods,
focusing on the particular impairment exhibited by
an individual patient. Accordingly, in the present
study it was also found that improvement was limited
to the sub-components of working memory which
had been specifically trained (verbal and executive),
while visual working memory, which was not addres-
sed by the treatment, remained unchanged. From a
theoretical point a view, such results also support
the modular approaches of working memory and
attention.
In conclusion, the present study suggested an
effectiveness of a specific therapy focusing on verbal
and central executive aspects of working memory
and its generalization to daily-life activities. Such
therapy can easily be used in a conventional rehabili-
tation setting. In the present case, a long-lasting
rehabilitation programme was given. The effective-
ness of a shorter training duration is currently
under study. In addition, further study could address
the specificity of therapy on different components of
working memory.
References
1. Baddeley A, Hitch G. Working memory. In: Bower GA, editor.
Recent advances in learning and motivation. New York:
Academic Press; 1974. pp 47–90.
2. Baddeley AD. Working memory. New York: Oxford University
Press; 1986.
3. Baddeley A, Logie RH. Models of working memory:
Mechanisms of active maintenance and executive control. In:
Miyake A, Shah P, editors. Working memory: The multiple-
component model. Cambridge, UK: Cambridge University
Pres; 1999. pp 28–61.
4. Norman DA, Shallice T. Attention to action. Willed and
automatic control of behavior. University of California, San
Diego, CHIP report N 99; 1980.
5. Kintsch W, Van Dijk T. Strategies of discours comprehension.
New York: Academic Press; 1983.
6. Baddeley A. Recent developments in working memory.
Current Opinion in Neurobiology 1998;8:234–238.
7. Vallar G, Baddeley A. Phonological short-term store and
sentence processing. Cognitive Neuropsychology 1987;4:
479–491.
8. Azouvi P, Couillet J, Leclercq M, Martin Y, Asloun S,
Rousseaux M. Divided attention and mental effort after
severe traumatic brain injury. Neuropsychologia 2004;42:
1260–1268.
9. Baddeley A, Della Sala S, Spinnler H. The two-component
hypothesis of memory deficit in Alzheimer’s disease. Journal
of Clinical and Experimental Neuropsychology 1991;13:
372–380.
10. Waters G, Caplan D, Hidebrandt N. On the structure of
verbal short-term memory and its functional role, in sentence
comprehension: Evidence from neuropsychology. Cognitive
Neuropsychology 1991;8:81–126.
11. Cicerone K. Remediation of ‘working attention’ in mild
traumatic brain injury. Brain Injury 2002;16:185–195.
12. Sturm W, Willmes K, Orgass B, et al. Do specific attention
deficits need specific training? Neuropsychological
Rehabilitation 1997;7: 81–103.
13. Ducarne B. Examen de l’aphasie. Paris: Edition du Centre de
Psychologie Appliquée; 1989.
14. Deloche G, Metz-Lutz MN, Kremin H, Hannequin D,
Ferrand I, Perrier D, Dordain M, Quint S, Cardebat D,
Lota AM, van der Linden M, Larroque C, Bunel G,
Pichard B, Naud E. Test de dénomination orale de 80
images: Do 80. Paris: ECPA; 1997.
15. Wechsler D. Wechsler memory scale-revised manual.
San Antonio: The Psychological Corporation; 1987.
16. Rey A. Examen clinique en psychologie. Paris: Presse
Universitaire de France; 1964.
17. Zimmermann P, Fimm B. Batterie de tests neuropsychologi-
ques pour l’évaluation des troubles de l’attention. Freiburg:
Vera Fimm, Psychologische Testsysteme; 1992.
18. Osterrieth PA. Le test de copie d’une figure complexe.
Archives de Psychologie 1944;30:206–356.
19. Wilson B. Single-case experimental designs in neuropsycho-
logical rehabilitation. Journal of Clinical and Experimental
Neuropsychology 1987;9:527–544.
20. Van der Linden M, Coyette F, Seron X. Selective impairment
of the ‘central executive’ component of working memory:
A single case study. Cognitive Neuropsychology 1992;9:
301–326.
21. Wechsler D. Wais-R manual. New York: The Psychological
Corporation; 1981.
22. Darrigrand B, Mazaux J. Echelle de communication verbale
de bordeaux. Isbergues: Ortho Edition; 2000.
23. Cardebat D, Doyon B, Puel M, Goulet P, Joanette Y.
Evocation lexicale formelle et sémantique chez des sujets
normaux. Performances et dynamiques de production en
fonction du sexe, de l’âge et du niveau d’étude. Acta
Neurologica Belgica 1990;90:207–217.
24. Taylor E. Psychological appraisal of children with cerebral
deficits. Cambridge: Harvard University Press; 1959.
25. Fassoti L, Kovacs F, Eling P, Brouwer W. Time pressure
management as a compensatory strategy training after
closed head injury. Neuropsychological Rehabilitation 2000;
10:47–65.
26. Park NW, Ingles JL. Effectiveness of attention rehabilita-
tion after an acquired brain injury: A meta-analysis.
Neuropsychology 2001;15:199–210.