BRAIN INJURY, VOL.

18,

NO.

4 (APRIL 2004), 391401

Assessing stroke patients prospective memory

using virtual reality
B . M . B R O O K S y, F . D . R O S E y,
J . P O T T E R z, S . J A Y A W A R D E N A z a n d
A. MORLINGz
y School of Psychology, University of East London, London, UK
z Stroke Rehabilitation Unit, Kent & Canterbury Hospital, Canterbury, UK
(Received 11 February 2003; revised 12 August 2003; accepted 15 August 2003)
There is a dearth of empirical evidence about prospective memory (remembering to perform actions
in the future) in stroke patients. A probable reason for this is that it is difficult to perform a realistic
and controlled assessment of prospective memory ability in a rehabilitation setting. Virtual reality may
provide a solution to this difficulty by allowing prospective memory to be tested in a simulation of a
real-life situation whilst retaining a laboratory level of scientific control. This exploratory study assessed
the performance of stroke patients and age-matched control participants on event-, time- and activitybased prospective memory retrieval tasks in a personal computer-based virtual environment. Stroke
patients were severely impaired at the event- and activity-based tasks compared with age-matched
controls, but only marginally impaired at the time-based task. The additional knowledge gained from
this form of assessment could direct rehabilitation more effectively towards specific impairments of
individual patients.

Introduction
One of the most disabling forms of memory impairment is the inability to remember to perform actions in the future (prospective memory failure) [13]. Impaired
prospective memory is more likely than any other form of memory impairment to
interfere with independent living as sufferers may forget to switch off the stove, to
light the gas, or to take medication. A realistic assessment of a patients prospective
memory ability should therefore be a major focus of any cognitive rehabilitation
programme.
Unfortunately, it is not currently possible to perform a comprehensive assessment of prospective memory ability in a rehabilitation setting because no standardized test is yet available. The most relevant test is the Rivermead Behavioural
Memory Test (RBMT) [4], which was developed as a method of identifying
everyday memory problems. However, only two, or possibly three, items in the
RBMT relate to prospective memory, an insufficient number on which to base a
realistic assessment of prospective memory ability. A comprehensive prospective

Correspondence to: Dr B. M. Brooks, School of Psychology, University of East London, London

E15 4LZ, UK. e-mail: b.m.brooks@uel.ac.uk
Brain Injury ISSN 02699052 print/ISSN 1362301X online # 2004 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/02699050310001619855

392

B. M. Brooks et al.

memory assessment should ideally include a number of measures of each of the

different types of prospective memory retrieval occasions that have been proposed
in the literature. For example, Kvavilashvili and Ellis [5] suggest that prospective
memory retrieval occasions can be classified as event-, time- or activity-based. In
event-based tasks, an action is required to be performed in response to a cue (e.g.
relaying a message when you see a friend); in time-based tasks, an action is required
at a previously specified future time (e.g. keeping an appointment at 3 p.m.); and
in activity-based tasks, an action is required before or after performing an activity
(e.g. switching off the stove after cooking). Based on this classification, a comprehensive prospective memory assessment should include a number of instances of
each of these three retrieval tasks.
The probable reason for the paucity of prospective memory assessment material
is that it is difficult to devise a comprehensive assessment that does not require testing in real-life situations with all the logistical problems that entails. Such an assessment would be impossible in a rehabilitation unit but can be devised using virtual
reality (VR).
VR is an interactive computer technology that can create the illusion of being
in an artificial world. It provides a detailed three-dimensional computer-generated
representation of the world that can be manipulated, and within which one can
move around. A form of VR may be presented on a computer monitor and can be
explored and manipulated using a joystick or other control device. Virtual environments using this technology can be designed to represent many real-life
situations and are programmed to record accurate measurements of performance
within them. They can combine aspects of the realism of naturalistic scenarios with
the control of laboratory-based studies.
There is considerable potential for using these virtual environments in the
assessment and rehabilitation of people with brain damage which is only just beginning to be realized (see [6], for a review). For example, they have been used in
the assessment and rehabilitation of people with traumatic brain injury [7, 8], executive function disorders [9], contralateral neglect [10], and temporal lobectomy [11].
Pertinent to the proposed investigation, they have been used to study memory
processes in vascular brain injury patients [12], and in memory rehabilitation of
a vascular brain injury patient with amnesia [13]. A virtual environment using this
technology would appear to be an ideal medium for assessing prospective memory
ability.
It is assumed that many stroke patients have problems with their prospective
memory, but there is a lack of empirical evidence to support this assumption [14].
No known study has compared the performance of stroke patients and age-matched
control participants to investigate whether stroke patients show significant deficits
in performing different types of prospective memory task. The following study
used a virtual environment of a four-room bungalow to assess the ability of stroke
patients and age-matched control participants to perform event-, time- and activitybased prospective memory tasks whilst engaged in a furniture removal task. The
event-based task was to remember to put Fragile notices on five glass items
before moving them; the time-based task was to allow the removal men access at
exactly 5-min intervals; the activity-based task was to keep the kitchen door closed
to stop the cat escaping.
Participants were also tested on two real-life prospective measures in order that
performance in the VR-based prospective memory test could be compared with

Assessing stroke patients prospective memory using virtual reality

393

real-life prospective ability. In one real-life task, they loaned the experimenter
a personal item, e.g. a watch, and they were required to remember to ask for its
return at the end of the study. In the other real-life task, they were instructed to
remember to ask for a written explanation of the study. Participants perceptions of
their own prospective memory ability were assessed using a 10-item questionnaire
about their real-life prospective memory performance.
As a measure of retrospective memory, participants were given a free recall test,
so that the effects of stroke on retrospective and prospective memory could be
compared. Premorbid intelligence of patients and controls was assessed using the
National Adult Reading Test (NART) [15]. Fluid intelligence was assessed using
the Block Design and Matrix Reasoning subtests from the Wechsler Abbreviated
Scale of Intelligence (WASI) [16].
Methods
Participants
Twenty-three male and 19 female stroke patients, 21 with right hemisphere impairments, 20 with left hemisphere impairments, and one with a bilateral impairment, from the Stroke Rehabilitation Unit at the Kent and Canterbury Hospital,
Canterbury, participated in the study. The stroke patients ranged from 1 week to
2 months post-stroke. The criteria for the participation of the stroke patients were
that their physiotherapist assessed them as physically able to perform the tasks and
the psychologist running the study assessed them as cognitively able to perform the
tasks. Eleven male and 18 female age-matched adults (mainly relatives of the stroke
patients) also participated in the study as control participants.
Equipment and materials
The virtual environment was a four-room bungalow. It was constructed using
Superscape VRT software, run on a desktop computer, explored using a joystick,
and manipulated using a mouse. Photographs of the virtual bungalow are shown
in figure 1. In the bungalow were 30 items with to go labels attached to them. Five
of these items had glass components (computer, wine rack, television, grandfather
clock and microwave). In the hall was a digital clock with a red button to its left,
whose significance and use is explained below.

Figure 1. Virtual bungalow.

B. M. Brooks et al.

394

A set of 12 photographs of common objects, e.g. a pen, a bus, was used in the
free recall test. A 10-item questionnaire about typical prospective memory tasks,
e.g. How often do you set off to do something then cant remember what?,
How often do you forget to take medication?, each item followed by seven rating
boxes ranging from Never to Always, was used in a subjective prospective
memory assessment.
Procedure
Participants were first asked their age and the number of years they had spent in
full-time education. They were also asked if they would lend the experimenter
a personal item, e.g. a watch, which was put out of sight. They were instructed that
they should remember to ask for the item to be returned and where the item was
hidden at the end of the study.
They were then given a free recall test in which they were shown 12 photographs of common objects, one photograph at a time for 3 s each. As soon as they
had seen all 12 photographs, they were asked to recall aloud any objects they could
remember and were given as long as they required for this task.
In the prospective memory task, participants were first informed that the owner
of the virtual bungalow was moving to a larger house with a hall and seven rooms
(lounge, dining room, nursery, kitchen, study, music room and bedroom) and that
he had put to go labels on the furniture and objects to be moved. Their task was to
organize these items for the removal men by asking the experimenter to move any
items that they considered should go in each of the rooms of the new house in turn
beginning with the new hall. (They were given a list of the rooms in the new house
to refer to during this task.)
They were also given the following three instructions concerning activities that
they were required to remember to perform during the removal task:
Ask the experimenter to put Fragile notices on the items with glass components (computer,
wine rack, television, grandfather clock and microwave) before you move them.
Allow the removal men access to the bungalow by asking the experimenter to click on the red
button beside the clock in the hall at exactly five-minute intervals. You may ask the experimenter to check the time on the clock in the hall whenever you like, but you should not
use your watch to check the time.
Ask the experimenter to close the door every time you leave the kitchen to keep the cat in.

Before beginning the task, participants were informed that they should remember to
ask the experimenter for a written explanation of the study when they had finished
performing the task.
The experimenter first entered the bungalow using a joystick. At the clock in
the hall, each participant was asked to recall the three additional activities they were
required to perform during the removal task (the prospective memory tasks). When
the experimenter was confident that each participant could recall the three tasks,
and the time on the clock in the hall registered an exact multiple of five, e.g. 10.05,
12.30, the furniture removal task began. The task continued until all the items
with to go labels had been moved. Immediately they had completed the task,
participants were asked to recall the three prospective memory tasks.
After the prospective memory task, participants completed a 10-item prospective
memory questionnaire. If participants were unable to complete questionnaires

Assessing stroke patients prospective memory using virtual reality

395

themselves, the experimenter completed them in accordance with their responses.

At the end of the study, if participants had failed to ask for the return of their
belongings or for the written explanation of the study, these items were given to
them as they left the study room.
Forty-one of 50 participants who recalled all three prospective memory tasks
at the end of the test, were assessed on the NART and the two performance subtests from the WASI. In order not to tire the participants unduly, these assessments
were completed on a different occasion. Unfortunately, some participants had
already left the unit before they were assessed on these measures, and six patients
and three controls, either could not be traced, or declined to participate further.

Results
According to Kvavilashvili and Ellis [5], there is a retrospective memory component
of remembering to perform actions in the future, that of actually remembering
what the action is. Seventeen stroke patients and four control participants failed
to recall the three prospective memory tasks correctly when they were questioned
immediately the removal task had finished. 2-tests revealed a significant difference between patients and controls recall of the three prospective memory
tasks (2 6.24, p 0.01). Since participants who failed to recall what actions
they were required to perform would obviously forget when to perform them,
these 21 participants were not included in the following analyses.
In all the statistical analyses, the probability level was set at 0.05. There were
no significant differences between the remaining 25 stroke patients and 25 controls
in terms of age (patients mean 71.8 years, SD 9.40, controls mean 68.44 years,
SD 7.11, t(48) 1.43, p 0.16), or years in full-time education (patients mean
10.76 years, SD 2.05, controls mean 11.04 years, SD 2.01, t(48) 0.49,
p 0.63). There was also no significant difference in patients and controls performance in the NART (patients mean 32.68, SD 10.75, controls mean
36.5, SD 6.54, t(39) 1.39, p 0.17).
VR-based prospective memory tasks
Each VR-based prospective memory task produced three-to-six retrieval occasions.
Figure 2 shows the mean probability of stroke patients and controls correctly
performing the three VR-based prospective memory tasks. It indicates that the
performance of the controls was higher than that of the patients in all three prospective memory tasks. However, there appears to be less difference between patients
and controls in the time-based task than in the activity- and event-based tasks.
Planned t-tests confirmed this view of the data as there was only a marginally
significant difference between patients and controls in the probability of correctly
performing the time-based prospective memory task (patients mean 0.39, SD
0.35, controls mean 0.58, SD 0.32, t(48) 2.02, p 0.05), whereas there
were highly significant differences between the probability of patients and controls
correctly performing the activity-based task (patients mean 0.27, SD 0.31,
controls mean 0.61, SD 0.33, t(48) 3.83, p < 0.001), and the event-based
task (patients mean 0.31, SD 0.40, controls mean 0.66, SD 0.40, t(48)
3.05, p 0.004).

B. M. Brooks et al.

396
0.7
0.
0.6
0.
0.5
0.
0.4
0.
0.3
0.
0.
0.2
0.1
0.
0
Time-based

Activity-based

Patients
Figure 2.

Event-based

Controls

Mean probability of correct performance in the three VR-based prospective memory tasks.

Real-life prospective memory measures

Nineteen of the 25 patients and 20 of the 25 controls remembered to ask for
the return of their belongings at the end of the study. Everyone who remembered
to ask for its return also remembered where the belonging was hidden. Conversely,
only four patients and 11 controls remembered to ask for a written explanation of
the study. 2-tests revealed that there was no significant difference between patients
and controls remembering to ask for their belongings (2 < 0.001, p 1), but there
was a significant difference between patients and controls remembering to ask for
a written explanation of the study (2 4.89, p 0.03).
Prospective memory questionnaire
Results of the prospective memory questionnaire indicated that patients rated their
own prospective memory abilities as highly as controls (patients mean 55.52,
controls mean 54.50). An independent t-test confirmed that there was no statistical difference between the patients and controls ratings (t(47) 0.58, p 0.57).
Interestingly, there was no significant correlation between participants overall performance in the VR prospective memory test and their scores in the prospective
memory questionnaire (Pearsons correlation 0.12, p 0.41).
A comparison of performance across the three different
prospective memory measures
Figure 3 highlights the differences between patients and controls overall performance in the VR-based prospective memory test (collapsed across the three
VR-based prospective memory tasks), their performance in the real prospective
memory tasks and their subjective ratings in the prospective memory questionnaire.
As can be seen, patients were impaired at the VR-based prospective memory test

Assessing stroke patients prospective memory using virtual reality

397

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
VR Test

Real 1

Patients

Real 2

Quest

Controls

Figure 3. Performance probabilities in the different prospective memory measures. VR Test, overall performance
in the VR-based prospective memory test; Real 1, remembering to request the return of belonging; Real 2,
remembering to request a written explanation; Quest, performance in the subjective prospective memory
questionnaire.

and one real-life prospective memory task, requesting a written explanation of

the study. However, they were not impaired at the other real-life task, requesting
the return of their belonging. Neither did they subjectively rate their prospective
memory as being worse than the controls.
Free recall (ability to recall events in the past)
Mean performance of patients in the free recall test was 6.4 compared with the
mean performance of controls, which was 7.48. An independent t-test found a
significant difference between patients and controls in the free recall test
(t(48) 2.2, p 0.03).
Figure 4 compares overall performance probabilities in the VR-based prospective memory test with performance probabilities in the free recall test. It appears
that patients were more impaired at the prospective memory tasks than the free
recall test whereas the performance of the controls was similar in the two tests.
Stroke patients prospective memory would therefore appear to be more impaired
than their retrospective memory, as measured by a free recall test. This interpretation of the data was confirmed by a 2
2 analysis of variance (ANOVA) with
one between-participants factor, Impairment (patients versus controls), and one
within-participants factor, Test (prospective memory test versus free recall test),
in which there was a significant interaction between Impairment and Test (F(48)
7.02, p 0.01).
Fluid intelligence tests
There was a significant difference between patients and controls in the Block
Design performance subtest from the WASI (patients mean 46.77, SD 12.66,
controls mean 54.45, SD 10.02, t(39) 2.23, p 0.03), and a marginally

B. M. Brooks et al.

398
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
VR-based PM

Patients

Free Recall

Controls

Figure 4. Performance probabilities in the VR-based prospective memory test and the free recall test.

significant difference in the Matrix Reasoning subtest (patients mean 48.00,

SD 11.18, controls mean 54.90, SD 11.60, t(39) 1.94, p 0.06). When
these scores were added and converted into performance IQ scores, there was
a significant difference between patients and controls (patients mean 96.33,
SD 18.17, controls mean 108, SD 15.23, t(39) 2.22, p 0.03). In addition,
performance IQ scores correlated with overall prospective memory performance
(Pearsons correlation 0.56, p < 0.001).
Discussion
In this study, a virtual environment was used to compare realistically the performance of stroke patients and age-matched control participants on event-, activityand time-based prospective memory retrieval tasks whilst they were performing a
furniture removal task. Also included in the study were two measures of real-life
prospective memory, a questionnaire relating to participants subjective ratings of
their own prospective memory abilities, a free recall test, and two fluid intelligence
tests.
An important finding that emerged from the study was that 17 of the 42 stroke
patients and four of the 29 control participants were unable to recall all three of
the prospective memory task instructions immediately they had finished the furniture removal task, even though they had been able to recall them immediately
before beginning the task. If a stroke patient cannot remember what is required,
it is highly unlikely that he/she will remember when to do it. This is obviously a very
real problem and should be borne in mind by therapists during rehabilitation.
The main result from this study was that the remaining stroke patients were
impaired at all three prospective memory tasks, but less impaired at the time-based
task. This result is counter to Maylors [17] prediction regarding the effects of age
on prospective memory ability, that age-related impairments should be more
apparent in time-based tasks than in event-based tasks (which contain environmental support in the form of some external cue event) (p. 178). This prediction is
based on Craiks [18] theory of age-related decrements in memory which proposes

Assessing stroke patients prospective memory using virtual reality

399

that self-initiated processing becomes more difficult for the elderly. If self-initiated
retrieval is not more difficult for stroke patients, their inability to remember to
perform tasks in the future must be attributable to a different impairment, perhaps
an inability to multitask. Alternatively, it is possible that the time-based task
(instructing the experimenter to press a button by the clock in the hall at exactly
5-min intervals) did not represent a realistic time-based task because the interval
between each retrieval occasion was too short and stroke patients continually
kept the task in consciousness. However, it is unlikely that the control participants
adopted this behaviour as their performance in the time-based task was marginally
worse than their performance in the two other tasks (figure 2). Further research
using different time intervals in the time-based task is required to investigate this
effect.
The finding that stroke patients were not impaired compared with controls
at one of the real-life prospective memory measures from the RBMT, requesting
the return of their property, but they were impaired at the VR-based prospective
memory tasks, indicates that the VR-based assessment is more sensitive to prospective memory impairments than the RBMT measure. This is not surprising, however, as the RBMT measure is a single-item test, whereas the VR-based tasks are
based on three-to-six retrieval occasions. Comparison of the high level of performance in this real-life task with the low level of performance in the other real life
task, asking for a written explanation of the study, and the finding that stroke
patients were not impaired at asking for the return of their property task but
were impaired at asking for a written explanation of the study, indicates that
motivation is important in prospective memory performance, especially for stroke
patients. (Participants were probably more motivated to ask for the return of their
property than for the written explanation.) If rehabilitation strategies can encourage
stroke patients to be more motivated to remember, it follows that their prospective
memory abilities would improve.
Another aim of rehabilitation should be to explain to stroke patients that they
might be likely to experience difficulties in remembering to perform tasks in the
future, particularly as the results of the questionnaire which assessed patients and
controls subjective knowledge of their own prospective memory abilities indicated that stroke patients were not aware how impaired they were at prospective
memory tasks. It is possible, however, that the stroke patients may have been
referring back to their prestroke behaviour when completing the questionnaire,
as some of the questions referred to situations which they would not encounter
in a stroke rehabilitation ward.
It would appear from the significant interaction between Impairment (patients
versus controls) and Test (prospective memory versus free recall) that prospective
memory ability is more affected by stroke than retrospective memory ability. However, this conclusion should be viewed with caution as the stroke patients included
in this analysis were those who could recall all three prospective memory tasks at
the end of the study, and were therefore those whose retrospective memory ability
was less affected by stroke.
Conclusions
In the present exploratory study, VR proved a useful medium to test stroke patients
prospective memory ability. The results indicated that stroke patients have particular

B. M. Brooks et al.

400

problems with both the content and retrieval of prospective memory tasks, i.e. what
to remember and when to remember it. However, when they were very motivated,
they overcame their impairments. Rehabilitation therapists should also note that
stroke patients may not be aware that they are susceptible to prospective memory
problems.
Future studies may benefit from using this VR-based testing procedure in conjunction with functional magnetic resonance imaging to investigate whether there is
any relationship between the area of the brain affected by stroke and prospective
memory ability. If such a relationship were found, future stroke rehabilitation could
be directed more effectively towards specific impairments of individual patients.
Acknowledgements
The authors express their appreciation to Dr Colin Cryer, Centre for Health
Services Studies, University of Kent, for valuable statistical help. They also sincerely
thank all participants for taking the time to perform the study.
References
1. COCKBURN, J.: Task interruption in prospective memory: a frontal lobe function? Cortex, 31:
8797, 1995.
2. KINSELLA, G., MURTAGH, D., LANDRY, A., HOMFRAY, K., HAMMOND, M., OBEIRNE, L., DWYER,
L., LAMONT, M. and PONSFORD, J.: Everyday memory following traumatic brain injury. Brain
Injury, 10: 499507, 1996.
3. TITOV, N. and KNIGHT, R. G.: A procedure for testing prospective remembering in persons
with neurological impairments. Brain Injury, 14: 877886, 2000.
4. WILSON, B. A., COCKBURN, J. and BADDELEY, A.: The Rivermead Behavioral Memory Test (Bury
St Edmunds: Thames Valley Test Co.), 1985.
5. KVAVILASHVILI, L. and ELLIS, J.: Varieties of intention: Some distinctions and classifications. In:
M. Brandimonte, G. O. Einstein and M. A. McDaniel (editors) Prospective Memory: Theory and
Applications (Hillsdale: Lawrence Erlbaum), pp. 2351, 1996.
6. ROSE, F. D. and FOREMAN, N.: Virtual Reality. Psychologist, 12: 550554, 1999.
7. CHRISTIANSEN, C., ABREU, B, OTTENBACHER, K., HUFFMAN, K., MASEL, B. and CULPEPPER, R.:
Task performance in virtual environments used for cognitive rehabilitation after traumatic brain
injury. Archives of Physical Medicine and Rehabilitation, 79: 888892, 1998.
8. DAVIES, R. C., JOHANSSON, G., BOSCHIAN, K., LINDEN, A., MINOR, U. and SONESSON, B.:
A practical example of using virtual reality in the assessment of brain injury. In: Proceedings of
the 2nd European Conference on Disability, Virtual Reality and Associated Technologies, Sweden, 1998,
pp. 6168.
9. MENDOZZI, L., MOTTA, A., BARBIERI, E., ALPINI, D. and PUGNETTI, L.: The application of virtual
reality to document coping deficits after a stroke. Report of a case study. Cyberpsychology and
Behavior, 1: 7991, 1998.
10. WANN, J. P., RUSHTON, S. K., SMYTH, M. and JONES, D.: Virtual environments in the rehabilitation of disorders of attention and movement. In: G. Riva (editor) Virtual Reality in Neuro-psychophysiology: Cognitive, Clinical and Methodological Issues in Assessment and Rehabilitation (Amsterdam:
IOS Press), pp. 157164, 1997.
11. MORRIS, R. G., PARSLOW, D. and RECCE, M. D.: Using immersive virtual reality to test allocentric spatial memory impairment following unilateral temporal lobectomy. In: Proceedings of the
3rd International Conference on Disability, Virtual Reality and Associated Technologies, Sardinia,
pp. 189196, 2000.
12. ROSE, F. D., BROOKS, B. M., ATTREE, E. A., PARSLOW, D. M., LEADBETTER, A. G., MCNEIL,
J. E., JAYAWARDENA, S., GREENWOOD, R. and POTTER, J.: A preliminary investigation into the use
of virtual environments in memory retraining after vascular brain injury: Indications for future
strategy? Disability and Rehabilitation, 21: 548554, 1999.

Assessing stroke patients prospective memory using virtual reality

401

13. BROOKS, B. M., NCNEIL, J. E., ROSE, F. D., GREENWOOD, R. J., ATTREE, E. A. and LEADBETTER,
A. G.: Route learning in a case of amnesia: a preliminary investigation into the efficacy of training
in a virtual environment. Neuropsychological Rehabilitation, 9: 6376, 1999.
14. TITOV, N. and KNIGHT, R. G.: A procedure for testing prospective remembering in persons
with neurological impairments. Brain Injury, 14: 877886, 2000.
15. NELSON, H. E.: The National Adult Reading Test (NART) (Windsor: NFER), 1982.
16. WECHSLER, D.: The Wechsler Abbreviated Scale of Intelligence (San Antonio: Psychological Corporation), 1999.
17. MAYLOR, E. A.: Does prospective memory decline with age? In: M. Brandimonte, G. O. Einstein
and M. A. McDaniel (editors) Prospective Memory: Theory and Applications (Hillsdale: Lawrence
Erlbaum), pp. 173197, 1996.
18. CRAIK, F. I. M.: A functional account of age differences in memory. In: F. Klix and H. Hagendorf
(editors) Human Memory and Cognitive Capabilities, Mechanisms and Performances (Amsterdam:
Elsevier), pp. 409422, 1986.