Content uploaded by Jim D Richards
Author content
All content in this area was uploaded by Jim D Richards
Content may be subject to copyright.
Clinical Rehabilitation 2003; 1 7: 264–272
© Arnold 2003 10.1191/0269215503cr607oa
Address for correspondence: VM Pomeroy, Professor of
Rehabilitation for Older People, Centre for Rehabili-
tation and Ageing, St George’s Hospital Medical School,
Cranmer Terrace, London SW17 0RE, UK. e-mail:
v.pomeroy@sghms.ac.uk
Agreement between physiotherapists on quality
of movement rated via videotape
VM Pomeroy, A Pramanik, L Sykes The Stroke Association’s Therapy Research Unit, University of Manchester
J Richards University of Salford, and E Hill The Stroke Association’s Therapy Research Unit, University of Manchester,
Salford, UK
Received 24th February 2001; returned for revisions 3rd August 2001; revised manuscript accepted 9th November
2001.
Background: Although achieving quality of movement after stroke is an
important aim of physiotherapy it is rarely measured objectively or described
explicitly.
Objective: To test whether physiotherapists agree on a composite measure
of quality of movement.
Setting: A movement analysis laboratory
Subjects: Ten stroke patients and 10 healthy age-matched volunteers.
Design: Prospective correlational.
Procedure: All subjects were videolmed performing three trials of six
standardized functional tasks. Two videotapes were made, each with a
different randomized order of appearance of the trials. Ten senior
physiotherapists independently rated the videotapes twice using a 100-mm
visual analogue scale.
Analysis: Analysis of variance models were tted to transformed data.
Estimates of components of variance were calculated and presented as a
percentage of the total variance for differences, within subjects (intra-subject),
between raters (inter-rater) and within raters (intra-rater). An acceptable
percentage was set at less than 10%.
Results: The percentage of intra-subject variance ranged from 1% (pick up
box and walking) to 9% (step on block). The percentage of inter-rater variance
ranged from 18% (pick up pencil) to 38% (sit to stand). The percentage of
intra-rater variance was less than 1% for all tasks.
Conclusions: Although physiotherapists disagreed with each other on quality
of movement they were more consistent in their own scoring.
Agreement on measuring quality of movement
265
Introduction
Achieving quality of movement during functional
activity post stroke is an important aim of phys-
ical therapy intervention in the UK
1
but it is not
measured objectively in routine clinical practice.
Although physiotherapists appear to have an
implicit knowledge of the meaning of quality of
movement, explicit description appears to be dif-
cult other than by equating it to normal move-
ment, which has been described differently by
different authors.
1–3
Whilst sophisticated move-
ment analysis can be used to identify parameters
of normal movement and is sometimes used in
research, this equipment is expensive, requires
considerable expertise for its operation and the
volume of data produced can often be over-
whelming. Given these difculties, it is not sur-
prising that research into the effects of
physiotherapy treatment in stroke rehabilitation
has not fully addressed quality of movement.
1
Progress in stroke rehabilitation research
requires the outcomes of physical therapy inter-
vention to be evaluated
4
with appropriate mea-
sures of performance.
5
As quality of movement
is an important aim of physical therapy, progress
therefore requires that appropriate measures of
quality of movement be developed for use in clin-
ical practice and research.
Previous research into the development of
measures of quality of movement have been
focused on children with cerebral palsy where a
need for measurement has also been identifed.
6
One approach involved obtaining clinical con-
sensus on observable attributes of quality of
movement whilst children were performing
specic motor tasks.
7,8
The motor tasks were
scored on ability to complete and an additional
dimension was added by scoring the quality of
the performance of movement.
9,10
Attributes
identied included: range of movement, target
accuracy, uency, co-ordination, alignment, dis-
sociated movement, stability and weight
shift.
8,10
Although this is clinically intuitive it is a
complex assessment and appears to arbitrarily
apply attributes of quality to scale items. It was
interesting that the originators of this measure
reported that raters did not nd the measure
easy to use, e.g., rating three specic quality
attributes at the same time as functional ability
during one motor task was difcult.
11
Given that (a) dening different quality attrib-
utes for different functional tasks has resulted in
a complex measure, and (b) that in conventional
clinical practice physiotherapists appear to make
a composite judgement of quality of movement
using many attributes of quality,
3,4,7,8,10–13
a way
forward is to use a measure that does not spec-
ify exactly what attribute of quality is under test.
If reliability is found to be acceptable then an
objective measure could probably be introduced
into clinical and research practice. If reliability is
found to be unacceptable then the case for devel-
oping what will probably be more complex mea-
sures will be made more strongly.
This study considered three areas of reliability:
1) Whether subjects exhibit the same amount of
quality when a movement is repeated con-
secutively, i.e., consecutive trials of the same
movement (intra-subject).
2) Whether different raters give the same rating
of quality as each other for the same move-
ment (inter-rater).
3) Whether individual raters give the same rat-
ing of quality of exactly the same movement
when they view it on different occasions
(intra-rater).
Methods
Ten stroke patients were recruited to this study.
They:
were aged between 65 and 74 years;
had sustained a stroke between 6 and 12
months previously;
were able to achieve all the movement tasks
required (details below);
had no other musculoskeletal or neurological
pathology.
In addition, ten healthy, age-matched volunteers
were also recruited to ensure that the ‘high qual-
ity’ area of the VAS would probably be included
in the scores given by raters.
Local ethical approval was given for this study
and all subjects gave written informed consent.
266
VM Pomeroy
et al.
Format of measure
The two formats considered were a Likert
scale and a visual analogue scale (VAS). The
clinical physiotherapists we consulted felt that a
VAS would be easier to use although we were
aware that Likert scales have been shown to pro-
duce less variability in scoring than VAS
14
and
therefore might produce higher reliability for a
measure of quality of movement. However when
used to measure phenomena with less subjective
attributes, a VAS has acceptable reliability.
15,16
We therefore proceeded with a VAS. The format
used was a 100-mm vertical VAS with a bottom
anchor point of ‘essentially no quality’ and a top
anchor point of ‘essentially total quality’. Explo-
ration of the potential clinical usefulness of this
tool suggested that physiotherapists felt that a
vertical scale made more clinical sense than a
horizontal one and that it allowed them to use
their existing practice to judge quality of move-
ment as the ‘worse’ quality of movement was at
the bottom and the ‘best’ at the top. The VAS
therefore has face validity.
Rating of videotapes
Raters in this study were 10 senior physiother-
apists with at least two years experience in stroke
rehabilitation. They rated the quality of move-
ment for each trial using a 100-mm vertical VAS
with the lowest anchor point ‘essentially no qual-
ity’ score 0, and the highest anchor point ‘essen-
tially total quality’ score 100. Raters used a
ballpoint pen to mark the VAS with a cross (x).
If the centre of the cross was ‘off-line’ a hori-
zontal line was projected from the centre of the
cross to the VAS line. Specic criteria for the
assessment of quality were not provided and
raters were asked to use their expert clinical
judgement to judge the level of quality for each
trial. After rating videotape 1 each rater returned
the tape and rating sheets to the research team
without keeping any records of their ratings. Four
to six weeks later they were sent videotape 2 and
repeated the procedure.
Analysis
The analysis was undertaken using data from
all 20 subjects. The calculation of limits of agree-
ment was contraindicated due to the large num-
ber of comparisons that would have to be made.
Functional tasks
The functional tasks used in this study were
adapted from the gross function and arm sections
of the Rivermead Motor Assessment
17
with stan-
dardized starting positions
18
and chair and table
adjusted correctly for the height of each partici-
pant
19
:
1) Sit to stand
2) Walk 10 metres
3) Step affected foot onto a block (dominant
lower limb for healthy volunteers)
4) Extension of affected elbow, wrist and ngers
to point at a target (dominant upper limb for
healthy volunteers)
5) Pick up a box from a table using two hands
6) Pick up a pencil from a table using affected
hand (dominant hand for healthy volunteers).
Full details of the functional tasks are given in
the Appendix. These tasks were chosen to give a
selection of functional activities, involving both
the upper and lower limbs.
Procedure
Each subject was videolmed performing three
consecutive trials of each functional task during
a one-hour period in the gait laboratory. Each
trial was conducted with the subject waiting for
a ash of light that indicated that they should
start the task. Lower limb tasks were lmed in
the sagittal plane and for the upper limb tasks the
two video cameras were placed diagonally in
front of the subject so that the whole upper limb
was in view throughout each of the three tasks.
For pragmatic reasons, i.e., having to readjust the
position and height of furniture, the tasks were
performed in the following order: walking, sit-to-
stand, step onto block, point at a target, pick up
a box, and pick up a pencil.
The videotaped trials were transferred in a ran-
dom order to videotape 1. Viewing order was
reordered by trial rather than subject. The same
videotaped trials were then transferred in a dif-
ferent random order to videotape 2 so that the
trials were shown in different sequences on the
two videotapes. No information was given to the
raters about which subjects had suffered a stroke.
Agreement on measuring quality of movement
267
To examine inter-rater alone, 45 comparisons
would have to be made. The conventional form
of analysis for examination of reliability is to cal-
culate intraclass correlation coefcients (ICC)
from an analysis of variance model (ANOVA
20
)
and to calculate the limits of agreement to indi-
cate the region within which 95% of individual
subject differences can be expected to fall.
21
However, the conventional ANOVA model was
unsuitable for this study which examined three
aspects of reliability: intra-subject, intra-rater,
and inter-rater. We therefore tted an ANOVA
model which accounted for the three aspects of
reliability. An ANOVA model was tted to a log
transformation of the data with rater effects,
rater replication effects, subject effects, subject
replication nested within the subject and their
two-way interactions. Rater, subject and rater
replication were treated as random effects. Esti-
mates of components of variance were calculated
and presented as a percentage of the total vari-
ance for differences, within subjects (intra-sub-
ject), between raters (inter-rater) and within
raters (intra-rater). An acceptable percentage
was set at less than 10%.
22
Summaries were produced to allow the three
areas of reliability to be examined on the scale
which the clinicians used to rate the subjects. To
examine whether each subject performed the
movement to the same standard on the three tri-
als (intra-subject reliability), the range of ratings
given to a subject over the three trials was cal-
culated for each rater (highest and lowest score
given by the same rater and over time). To exam-
ine how much raters agreed with each other
(inter-rater reliability) the interquartile range
(IQR) of ratings given to each subject on each
trial by the raters was calculated. To examine
whether raters gave the same rating for both
viewings (intra-rater reliability) the difference
between the two scores was calculated. Box and
Whisker plots of all of these aspects of reliability
were produced for each of the functional
tasks.
23,24
Results
Only the Box and Whisker plots for the results
of the pick up box task are presented and dis-
cussed, but similar results were found for the
other tasks, the condensed results are shown in
Table 1.
The ranges of scores given to each subject over
the three trials by each rater on the pick up box
task are shown in Figure 1 (only the rst viewing
of each rater is shown, the second viewing gave
similar results). The range of scores over trial
varied from 0 to 75 points on the VAS scale.
Some raters do appear to be less consistent
between trials than others (e.g., rater 8) but even
the smaller ranges are in the region of 20 points.
The IQR of scores given by raters for the same
trial on each subject performing the pick up box
task are shown in Figure 2. The IQRs varied
between 2 and 62, with the median IQR being 10.
IQRs do not seem to change too much between
trials but there is perhaps evidence that raters
disagree more on the second viewing than the
rst.
Table 1 The extent of disagreement of scoring in intra-subject, inter-rater and intra-rater reliability (mm on 10-cm VAS)
Area of reliability
Range of three trials IQR of raters (inter-rater) Actual difference between
(intra-subject) 1st and 2nd viewing
(intra-rater)
Median (min, max) Median (min, max) Median (min, max)
Pick up box 7 (0, 75) 10 (2, 63) 4.5 (0, 68)
Pick up pencil 9 (0, 65) 13 (1, 43) 5 (0, 61)
Walking 9 (0, 50) 12 (1, 44) 5 (0, 58)
Sit to stand 9 (0, 65) 15 (2, 45) 4 (0, 47)
Step on block 9 (0, 65) 15 (2, 45) 5 (0, 54)
Point at target 9 (0, 80) 12 (2, 50) 5 (0, 73)
268
VM Pomeroy
et al.
Figure 1 Intra-subject reliability for pick up box task for all 20 subjects.
Figure 2 Inter-rater reliability for pick up box task.
Agreement on measuring quality of movement
269
zero. The proportion of variability due to raters
for each of the tasks is shown in Table 2. The
variability due to difference between trials (intra-
subject) was very small for each of the tasks,
<10%. For each task the proportion of variabil-
ity due to raters (inter-rater) accounts for more
than 10% of the total variability. The proportion
of variability due to differences between viewings
(intra-rater) is very small, less than 1% for all
tasks.
The difference between the two viewings by
each rater for the pick up box task is shown in
Figure 3 (only the plots for the rst trial for each
subject are shown, results were similar for the
other trials). The differences ranged from 0 to 68.
Some raters appeared to be less consistent than
others, e.g., rater 10, but many of the raters do
not differ by more than 10 points for the major-
ity of the trials.
Some of the estimates of variance were calcu-
lated to be negative values, these were set to
Table 2 Components of variance as a percentage of total variance
Component of Component of Component of Variance due
variance due to variance due to variance due to to remaining
differences differences differences components (%)
between trials between raters between viewings
(intra-subject) (%) (inter-rater) (%) (intra-rater) (%)
Pick up box 1 25 <1 73
Pick up pencil 2 18 <1 81
Walking 1 20 <1 79
Sit to stand 6 38 <1 55
Step on block 9 31 <1 59
Point at target 5 28 <1 66
Figure 3 Intra-rater reliability for pick up box task.
270
VM Pomeroy
et al.
iotherapists.
3,4,7,8,10–13
In addition, minimal differ-
ences between inter- and intra-rater reliability as
assessed by ICC coefcients (0.92 and 0.96
respectively) were found when quality attributes
were specied for the Gross Motor Performance
Measure (GMPM) and rater training was pro-
vided.
27
It might be possible to improve inter-rater reli-
ability if all physiotherapists could be trained to
assess the same quality attributes, but the rst
requirement would be the identication and def-
inition of the quality attributes for each func-
tional task. Obtaining such consensus for children
with cerebral palsy (GMPM) resulted in different
combinations of quality attributes being assigned
to each task
6–8
and reports of difculty in using
the GMPM.
27
It is reasonable to expect that
obtaining consensus on quality attributes for
stroke patients could also result in similar com-
plexity and difculties for clinical use although a
more complex measure might be acceptable for
research use. However, as these ndings indicate
that conventional clinical practice of judging
quality of movement may be unreliable, the case
for developing more complex clinical measures
has been strengthened. Consensus building might
therefore be an option to begin to develop qual-
ity of movement measurement.
Another explanation of the higher intra-rater
reliability compared with inter-rater reliability
could be that physiotherapists are measuring the
same quality attributes but that they have a dif-
ferent interpretation of the level of quality.
Training might improve agreement but this
would require the denition of levels of quality
and at least denition of the reference points:
‘essentially total quality’ and ‘essentially no qual-
ity’.
A starting point for development might be the
physiotherapy belief that quality of movement is
closely related to, if not dependent on normal
movement,
28
but this statement immediately
leads to the question ‘what is normal movement?’
The answer would be fairly simple if normal
movement always consisted of a xed sequence
of standardized components but considerable
variation has been found between individuals in
the combinations of component parts of move-
ment patterns for the performance of functional
tasks.
29–33
However, under stringent conditions
Discussion
Disagreement between raters was found for this
measure of quality of movement for all six of the
functional tasks to which it was applied but intra-
rater reliability was found to be acceptable for all
tasks. These ndings suggest that this quality of
movement measure could be used in research
studies if (a) it was restricted to one rater only,
(b) pilot work was undertaken to identify the
inter-rater reliabilities which could then be incor-
porated into a stratied design in future studies
or (c) replicates of independent observations
were to be obtained on each subject and then
averaged.
25
Use of the current tool in clinical
practice is not recommended.
Some of the rater disagreement may have
resulted from variability in scoring observed pre-
viously with the VAS,
14
particularly when used
with a vertical orientation.
26
Likert scales have
been shown to produce less variability in scor-
ing
14
and might have produced higher reliability
in this context. However, as (a) reliability is
acceptable when less subjective phenomena are
measured with a VAS
15,16
and (b) this study
found acceptable overall intra-rater reliability it
seems reasonable to propose that a considerable
proportion of the variability between raters was
due to different perceptions of quality of move-
ment between physiotherapists rather than just to
the properties of the VAS. Support for this inter-
pretation is given by the clinical literature and
previous studies which suggest that there may be
many quality attributes which are used by phys-
Clinical messages
These ndings suggest that although phys-
iotherapists disagree with each other on
quality of movement they are more consis-
tent in their own scoring.
At present the variation is too great for this
measure to be used in clinical practice.
Further development work is required to
examine the possibility that obtaining con-
sensus on quality attributes and/or level of
quality enables acceptable inter-rater relia-
bility.
Agreement on measuring quality of movement
271
11 Bobath B. Adult hemiplegia. Evaluation and
treatment. Oxford: Butterworth Heinemann, 1993:
34–42.
12 Mawson S. TELER normal movement indicators.
Shefeld: TELER, 1995.
13 Weisglas-Kuperus N, Naerts W, Fetter WPF et al. .
Minor neurological dysfunction and quality of
movement in relation to neonatal cerebral damage
and subsequent development. Dev Med Child Neurol
1994; 36: 727–35.
14 Guyatt GH, Townsend M, Berman LB, Keller JL. A
comparison of Likert and visual analogue scales for
measuring change in function. J Chron Dis 1987; 12:
1129–33.
15 Tiplady B, Jackson SHD, Maskrey VM, Swift CG.
Validity and sensitivity of visual analogue scales in
young and older healthy subjects. Age Ageing 1998;
27: 63–66.
16 Maxwell C. Sensitivity and accuracy of the visual
analogue scale: a psycho-physical classroom
experiment. Br J Clin Pharmacol 1978; 6: 15–24.
17 Lincoln N, Leadbitter D. Assessment of motor
function in stroke patients. Physiotherapy 1979; 65:
48–51.
18 Colson JHC, Collison FW. Starting positions. In:
Progressive exercise therapy in rehabilitation and
physical education. Bristol: Wright PSG, 1983:
261–72.
19 Mandell AC. The correct height of school furniture.
Physiotherapy 1984; 70: 48–53.
20 Eliaswziw M, Young SL, Woodbury MG, Fryday-
Field K. Statistical methodology for the concurrent
assessment of interrater and intrarater reliability:
using goniometric measurements as an example.
Phys Ther 1994; 74: 777–88.
21 Bland JM, Altman DG. Statistical methods for
assessing agreement between two methods of clinical
measurement. Lancet 1986; 1: 307–10.
22 Fleiss JL, Slakter MJ, Fischman SL, Park MH,
Chilton NW. Inter-examiner reliability in caries
trials. J Dent Res 1979; 58: 604–609.
23 Altman DG. Practical statistics for medical research,
London: Chapman and Hall, 1997: 33.
24 Stata Corp. Stata statistical Software, release 6.0.
College Stata, TX: Stata Corporation, 1999.
25 Fleiss JL. The design and analysis of clinical
experiments. New York: John Wiley and Sons, 1986.
26 Swiwatanakul K, Kelvie W, Lasagna L, Calimlim JF,
Weis OF, Mehta G. Studies with different types of
visual analogue scales for measurement of pain. Clin
Pharmacol Ther 1983; 34: 234–39.
27 Gowland C, Boyce WF, Wright V et al. Reliability of
the Gross Motor Performance Measure. Phys Ther
1995; 7: 597–602.
28 Partridge CJ. Physiotherapy approaches to the
treatment of neurological conditions – an historical
perspective. In: Neurological physiotherapy: a
problem solving approach. Edwards S ed. New
such as the standardized starting positions and
procedures used in this study, components of
gross motor patterns have been shown to be con-
sistent.
34–39
Focus groups could therefore identify
and dene the attributes and levels of quality of
movement for each functional task to develop the
measure.
Acknowledgements
We gratefully acknowledge the nancial sup-
port provided by the Stroke Association and
REMEDI. We thank all the physiotherapists who
acted as raters in this study and the volunteers
who were subjects in this study. We also thank
Dr EB Faragher for his statistical advice and sup-
port.
References
1 Ashburn A, Partridge C, DeSouza L. Physiotherapy
in the rehabilitation of stroke: a review. Clin Rehabil
1993; 7: 337–45.
2 Partridge CJ, Morris LW. Recovery from physical
disability after stroke: proles for different levels of
starting severity. Clin Rehabil 1993; 7: 210–17.
3 Edwards S. An analysis of normal movement as the
basis for the development of treatment techniques.
In: Edwards S ed. Neurological physiotherapy. A
problem solving approach. New York: Churchill
Livingstone, 1996: 15–40.
4 Partridge C. Evaluation of physiotherapy for people
with stroke. London: King’s Fund Centre, 1994.
5 Scholz JP. Dynamic pattern theory – some
implications for therapeutics. Phys Ther 1990; 70:
827–43.
6 Boyce WF, Gowland C, Rosenbaum PL et al.
Measuring quality of movement in cerebral palsy:
a review of instruments. Phys Ther 1991; 71:
813–19.
7 Boyce WF, Gowland C, Hardy S et al. Development
of a quality of movement measure for children with
cerebral palsy. Phys Ther 1991; 71: 820–32.
8 Boyce W, Gowland C, Russell D et al. Consensus
methodology in the development and content
validation of a gross motor performance measure.
Physiother Can 1993; 45: 94–100.
9 Boyce WF, Gowland C, Rosenbaum PL et al. The
gross motor performance measure: validity and
responsiveness of a measure of quality of movement.
Phys Ther 1995; 75: 603–13.
10 Johnson LM, Randall MJ, Reddihough DS et al.
Development of a clinical assessment of quality of
movement for unilateral upper-limb function. Dev
Med Child Neurol 1994; 36: 965–73.
272
VM Pomeroy
et al.
performance: a kinematic perspective. Am J Occup
Ther 1994; 48: 679–87.
35 Hirokawa S. Normal gait characteristics under
temporal and distance constraints. J Biomed Eng
1989; 11: 449–56.
36 Gabell A, Nayak USL. The effect of age on
variability in gait. J Gerontol 1984; 39: 662–66.
37 Winter-DA. Foot trajectory in human gait: a precise
and multifactorial motor control task. Phys Ther
1992; 72: 45–56.
38 Ikeda ER, Schenkman ML, Riley PO, Hodge WA.
Inuence of age on dynamics of rising from a chair.
Phys Ther 1991; 71: 473–81.
39 Doorennbosch CAM, Harlaar J, Roebroeck ME,
Lankhorst GJ. Two strategies of transferring from
sit-to-stand; the activation of monoarticular and
biarticular muscles. J Biomech 1994; 27: 1299–307.
York: Churchill Livingstone, 1996: 3–14.
29 VanSant AF. Rising from a supine position to erect
stance. Description of adult movement and a
developmental hypothesis. Phys Ther 1988; 68:
185–92.
30 Richter RR, VanSant AF, Newton RA. Description
of adult rolling movements and hypothesis of
developmental sequences. Phys Ther 1989; 69: 63–71.
31 McCoy JO, VanSant AF. Movement patterns of
adolescants rising from a bed. Phys Ther 1993; 73:
182–93.
32 VanSant AF. Life-span development in functional
tasks. Phys Ther 1990; 70: 788–98.
33 Woollacott MH, Shumway-Cook A. Changes in
postural control across the life span – a systems
approach. Phys Ther 1990; 70: 799–807.
34 Wu C, Trombly CA, Lin K. The relationship
between occupational form and occupational
Appendix – Movement tasks adapted from Rivermead Motor Assessment
17
1) Start in sitting, sitting to standing, stand for 15 seconds (gross function section, item 3).
2) Start in standing, walk 10 metres indoors with no stand-by help, calliper, splint or walking aid (gross
function section, item 8).
3) Start in standing, step affected leg (dominant leg for healthy volunteers) onto a 17-cm-high block and
hold the position for 5 seconds (adapted from gross function section item 7).
4) Start in sitting with affected glenohumeral joint (dominant for healthy volunteers) in 90° exion, elbow
and mid-pronated forearm resting on a high table with ngers touching opposite shoulder, move forearm
until elbow is fully extended and tips of ngers are extended in line with target place directly in front of
glenohumeral joint (adapted from arm section, item 3).
5) Start in sitting with hands and wrists resting on table and elbows in 90° exion directly under
glenohumeral joints, reach forward, pick up empty cardboard box with both hands until box is level
with glenohumeral joints and hold this position for 3 seconds – box on table so that the subject has to
extend elbows fully to reach it, (adapted from arm section, item 5).
6) Start in sitting with hand resting on table with elbows in 90° exion directly under glenohumeral joints,
stretch affected arm forward (dominant for healthy volunteers), pick up pencil from table, lift from table,
lower to table and release – pencil on table so that the subject has to extend elbow fully to reach it
(adapted from arm section, item 7).