ORIGINAL ARTICLE

Mirror therapy for phantom limb pain: Brain changes and the
role of body representation
J. Foell1,2*, R. Bekrater-Bodmann1*, M. Diers1, H. Flor1
1 Department of Cognitive and Clinical Neuroscience, Central Institute of Mental Health, Medical Faculty Mannheim, Heidelberg University,
Mannheim, Germany
2 Department of Psychology, Florida State University, Tallahassee, USA

Correspondence
Herta Flor
E-mail: herta.or@zi-mannheim.de
Funding sources
This work was supported by the PHANTOM
MIND project (Phantom phenomena: a
window to the mind and the brain), which
receives research funding from the European
Communitys Seventh Framework Programme
(FP7/20072013)/ERC Grant Agreement No.
230249, and by a subproject of the
Collaborative Research Project Bionic-Hand
funded by the Bundesministerium fr Bildung
und Forschung (V4UKF02). This manuscript
reects only the authors views, and the
Community is not liable for any use that may
be made of the information contained
therein. The authors have no other relevant
afliations or nancial involvement with any
organization or entity with a nancial interest
in or nancial conict with the subject matter
or materials discussed in the manuscript
apart from those disclosed. No writing
assistance was utilized in the production of
this manuscript.
Conicts of interest
None declared.

Abstract
Background: Phantom limb pain (PLP) is a common consequence of
amputation and is difficult to treat. Mirror therapy (MT), a procedure
utilizing the visual recreation of movement of a lost limb by moving the
intact limb in front of a mirror, has been shown to be effective in reducing
PLP. However, the neural correlates of this effect are not known.
Methods: We investigated the effects of daily mirror training over 4
weeks in 13 chronic PLP patients after unilateral arm amputation. Eleven
participants performed hand and lip movements during a functional
magnetic resonance imaging (fMRI) measurement before and after MT.
The location of neural activity in primary somatosensory cortex during
these tasks was used to assess brain changes related to treatment.
Results: The treatment caused a significant reduction of PLP (average
decrease of 27%). Treatment effects were predicted by a telescopic
distortion of the phantom, with those patients who experienced a
telescope profiting less from treatment. fMRI data analyses revealed a
relationship between change in pain after MT and a reversal of
dysfunctional cortical reorganization in primary somatosensory cortex.
Pain reduction after mirror training was also related to a decrease of
activity in the inferior parietal cortex (IPC).
Conclusions: Experienced body appearance seems to be an important
predictor of mirror treatment effectiveness. Maladaptive changes in
cortical organization are reversed during mirror treatment, which also
alters activity in the IPC, a region involved in painful perceptions and in
the perceived relatedness to an observed limb.

*These authors contributed equally to this

work.
Accepted for publication
6 November 2013
doi:10.1002/j.1532-2149.2013.00433.x

1. Introduction
After the amputation of a limb, most patients report
awareness of a phantom (Giummarra et al., 2007), i.e.,
the continuing perception of the missing limb, with or

without proportional distortions. Additionally,

6090% (Jensen et al., 1983; Hanley et al., 2009) of
amputees report phantom limb pain (PLP), a sensation
of pain located in the amputated limb, which has a
high rate of chronicity and is difficult to treat (Weeks

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC. Eur J Pain 18 (2014) 729739 729
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium,
provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

Mirror therapy for phantom limb pain

Whats already known about this topic?

It is known that mirror therapy provides relief for
some patients with chronic phantom limb pain.
We also know that the intensity of this pain is
correlated with the extent of cortical reorganization in primary somatosensory cortex.
What does this study add?
This study shows that the pain relief induced by
mirror therapy is accompanied by a reversion of
cortical reorganization, and that the treatment
effect is dependent on properties of the phantom
limb.

et al., 2010). Central changes have been proposed to

contribute to PLP (Flor et al., 2006). Reorganization of
somatosensory (cf. Flor et al., 1995; Birbaumer et al.,
1997) and motor cortex (cf. Karl et al., 2001;
Koppelstaetter et al., 2007) e.g., the invasion of areas
neighbouring the representation of the amputated
limb into the cortical representation zone was shown
to be related to PLP intensity, suggesting that it may be
related to this change after amputation (Flor et al.,
2006).
Since there is a strong influence of vision on the
perception (Hunter et al., 2003) and movement
(Brodie et al., 2003) of a phantom, the use of experimental set-ups inducing the illusion of two intact
limbs might influence these central alterations
(Ramachandran and Rogers-Ramachandran, 1996).
When persons with PLP are asked to place their
healthy limb in front of a mirror such that its reflection
visually replaces the missing limb, this can result in
pain alleviation in some patients. Chan et al. (2007)
published a randomized placebo-controlled study
showing significant PLP decrease with recently amputated leg amputees after 4 weeks of training. However,
in subacute pain spontaneous recovery can occur
(Schley et al., 2008), requiring a study of these effects
in chronic patients.
The mechanisms behind the mode of action of
mirror therapy (MT) are not clear. One possible
mechanism is the representational restitution of the
missing limb in the brain (Foell et al., 2011) by the
convergence of concurrent visual and proprioceptive
input. Additionally, previous evidence indicates that
an external object can be more easily integrated into
ones own body representation if there is a high degree
of congruence between sensory modalities (Tsakiris,
2010). For example, deviations between the seen and
felt position of an artificial limb impede the sensation

J. Foell et al.

of ownership (Lloyd, 2007). These results might be of

significance in MT, as perceived distortions of the
phantom might complicate the integration of visual
and proprioceptive input into a coherent percept and
reduce the effects of mirror treatment. This might be
one factor why about 40% of the PLP patients do not
benefit from MT (Weeks et al., 2010). The identification of predictors that influence the effectiveness of
MT would help to understand the mechanism of this
treatment and could improve its clinical use.
This study was designed to (1) evaluate the effects of
MT on pain in chronic PLP patients, (2) compare the
brain changes related to treatment effects before and
after therapy and (3) identify predictors of treatment
success. Our hypotheses were (1) a subset of patients
will benefit from MT, and (2) these patients will
display an individual reduction of dysfunctional cortical shift, which will be stronger in patients with large
treatment benefit.

2. Materials and methods

2.1 Patients
The sample consisted of 13 patients with major unilateral
upper limb amputation [four women, mean age 50.6, standard deviation (SD) = 15.8 years, range: 2674 years] who
experienced PLP regularly at least once a week with an
average intensity of at least 20 on a visual analogue scale
(VAS; ranging from 0 to 100) and who had been amputated
for more than 2 years to rule out acute PLP (average time
since amputation 21.3, SD = 12.7 years, range: 649 years).
All subjects reported feeling a phantom limb. Supporting
Information Table S1 provides information on all participants, including cause of amputation, prosthesis use, pain
medication, PLP frequency and intensity, and the experience
of telescoping. All participants gave written informed
consent prior to taking part in the study, and the Ethics
Committee of the Medical Faculty Mannheim of Heidelberg
University approved the protocol, which adhered to the Declaration of Helsinki. Two patients were excluded from the
functional magnetic resonance imaging (fMRI) measurements or analysis (due to a tattoo in one case and an incidental finding in anatomic images in another); thus, 11
patients were included in the brain imaging part of the study
(four women, mean age 49.3 years, SD = 15.3, range: 2668
years; average time between amputation and measurement
21.3 years, SD = 13.8, range: 649 years), whereas all 13
patients participated in the mirror treatment. PLP intensity as
assessed by the German version of the West Haven-Yale
Multidimensional Pain Inventory (MPI; Kerns et al., 1985;
Flor et al., 1990; modified to separately assess phantom and
residual limb pain) averaged 2.23 (SD = 0.72) for the entire
group and 2.12 (SD = 0.70) for the fMRI subgroup. Twelve
patients reported that they experienced non-painful
phantom phenomena, such as a feeling of pressure or slight

730 Eur J Pain 18 (2014) 729739 2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC

J. Foell et al.

tingling, at least several times per week. The mean intensity

for these sensations was 48.38 (SD = 26.58, range: 092) on
VAS from 0 = no sensation to 100 = very intense sensation
(fMRI subgroup: mean 46.27, SD = 26.09, range: 092).
Patients experiencing a telescope (n = 8) reported a mean
intensity of 80.0 (SD = 26.4, range: 23100) for this sensation, measured using VAS, ranging from 0 = no telescope to
100 = very intense (fMRI subgroup: mean 89.50, SD = 9.01,
range: 77100).
Over the course of the study, one patient (patient 1) used
a steady amount of pain medication (a non-steroidal antiinflammatory drug) everyday for 8 weeks of the study. Three
patients (3, 7 and 13) used analgesics occasionally, i.e., on 5
or fewer days spread over the entire 8 weeks of the study
(patient 3: an opioid, patients 7 and 13: a non-steroidal
anti-inflammatory drug).

2.2 Design
Each patient gave daily reports of pain (using VAS) for a total
of 8 weeks, commencing 2 weeks prior to therapy (prephase) to assess the baseline level of PLP, followed by a
4-week training phase with daily mirror exercises and concluding with 2 weeks without any treatment (but with daily
pain ratings; post-phase) to evaluate possible long-term
effects of the therapy. On the day before the first training
unit as well as after the last one, the patients brain activation
was assessed: Patients performed (1) mirrored hand movements and (2) a lip-pursing task during fMRI measurement
(cf. Lotze et al., 2001; MacIver et al., 2008). During the same
visit, patients were asked about the frequency, intensity and
relieving factors for current and past PLP as well as current
and past non-painful phantom sensations (including telescoping) using a structured interview (Winter et al., 2001).
For everyday of the study (pre-phase, training phase and
post-phase), patients were asked about any pain medication
they were using.

2.3 Mirror training

Patients received specific verbal and written instructions for
4-week mirror training during a one-on-one appointment
prior to the training phase. Patients were instructed to try to
consciously relate the movement observed in the mirror to
their phantom at any point during the training and to keep
their attention focused on the task. Each instruction was
explained verbally, demonstrated by a therapist and performed by the patient in front of her or him. Patients then
trained on their own everyday for a period of 4 weeks. Five
different movement tasks were devised to improve patient
compliance and to reduce the risk of straining. The tasks
included (1) opening and closing of fingers: repeated converging of the fingertips, starting with a loosely opened
hand, palm towards the mirror, but without any tactile
contact among the fingers or between fingertips and palm;
(2) stretching of fingers, with palm towards the mirror; (3)
turning the hand, so that the palm alternately faced upwards

Mirror therapy for phantom limb pain

and downwards; (4) sequential converging of fingertips and

thumb, palm towards the mirror, without actual contact
between the fingertips; and (5) tracing figures with the index
finger in the manner of a concert conductor.
All these tasks were designed for maximum visibility of
the movement while excluding any kind of tactile feedback.
Patients were instructed to perform each task for 3 min
(total: 15 min daily) and to record the actual amount of time
they needed to perform each of the tasks on each day.
Patients were advised to keep the frequency of the hand
movements constant (at approximately 0.5 Hz), but this was
not enforced; instead, patients were free to use the frequency
that best enabled them to relate the observed movement to
the movement of their phantom limb. They were instructed
to abort any movement task if it intensified their pain level
and to document if this happened. After explanation and
demonstration of all tasks, the patients practiced the tasks
themselves and received a detailed written explanation of
the movement tasks including colour photographs. No
patient reported a lack of understanding of or an inability to
perform the mirrored movement task.
Patients received a large transportable mirror (30 50 cm
or 12 20 in.) and were required to give a daily report on
the intensity of PLP on that day in general (on VAS with the
endpoints no pain and very intense pain, ranging from 0
to 100) and the occurrence and intensity of PLP during the
training (on a scale from 0 for none at all to 6 for most
intense pain ever experienced, taken from the MPI; Flor
et al., 1990). They were also asked daily about the degree to
which they were able to relate the movement seen in the
mirror to their phantom limb, on a scale from 0 (not at all)
to 6 (as vivid as a real perceptual experience), and whether
they felt movement in the phantom limb (on the same scale
as the item before). Patients were contacted by phone at least
once a week to check for questions or problems, and were
instructed to contact the experimenters if any problems
should arise.
The entire procedure of the study, except for the time
spent in the MRI scanner and the explanation of the fMRI
measurement procedure, was identical for the patients with
and without fMRI data.

2.4 MRI scanning session

The MRI measurements took place at the beginning and at
the end of the 4-week training period. During the first
appointment, patients were informed about the purpose of
the study and the exact movement tasks to perform in the
MRI scanner. Before the measurement, patients were trained
to execute the requested hand movements with the intact
hand without producing muscle activity in the residual arm
(using an electromyography feedback procedure prior to the
first measurement; see Lotze et al., 2001). Lip and hand
movements were demonstrated to the patients and practiced
by the patients before the measurement. During the fMRI
scan, the patients gaze was redirected using a mirror
attached to the MRI head coil. This way, they could easily

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC Eur J Pain 18 (2014) 729739

731

Mirror therapy for phantom limb pain

view their intact hand lying on the abdomen, as well as its

reflection, which was produced using a mirror placed on the
patients body. For the hand movement condition, patients
were instructed to close their hand to a fist (yet without any
actual touching of the fingertips to the palm or between the
fingers) and to open it again whenever they heard a sound
signal. This signal was presented with a frequency of 0.5 Hz
during activity phases (on-block; duration: six scans or
19.8 s), whereas no sound was given during the resting
phase (off-block; duration: six scans or 19.8 s), for which
patients were instructed to lie still, watching their hand and
the mirror image. For the on-block, they were explicitly
instructed to imagine their phantom hand moving in accordance with the reflected intact hand. For the lip-pursing
condition, also assessed with an fMRI measurement, patients
were asked to lie still with their eyes closed and to purse their
lips whenever they heard the sound signal (which was again
presented with a frequency of 0.5 Hz during activity phases).

2.5 Image acquisition and analysis

The fMRI scans were conducted with a 3 Tesla Siemens Trio
MR (Siemens AG, Erlangen, Germany) scanner using echoplanar imaging (matrix 64 64, TE 45 ms, TR 3.3 s) and 40
slices of 2.3 mm thickness, tilted in accordance with the
transverse plane adjusted to include all frontal, central, parietal and occipital cortical areas as well as upper parts of the
temporal cortex and the cerebellum. Seventy-eight wholebrain scans including six blocks of hand or lip movements
with six scans each paired with seven blocks of six scans of
rest were gathered per condition, and the first three volumes
were excluded from the analysis to allow for signal stability
following onset transients. For anatomical reference, a
T1-weighted anatomical data set (magnetization-prepared
rapid acquisition with gradient echo; slice thickness 1.1 mm,
TR 2300 ms, TE 2.98 ms, flip angle 9) was obtained.
fMRI data were evaluated using SPM8 (Wellcome Institute of Imaging Neuroscience, London, UK) implemented in
Matlab 7.1 (MathWorks Inc., Natick, MA, USA). The echoplanar images of each subject were coregistered to the individual anatomical data sets after the anterior commissure
had been manually defined as the reference point. Further
pre-processing included motion and slicetime correction,
normalization to the standard space using a template [Montreal Neurological Institute (MNI)] and smoothing with a
Gaussian kernel of 8 mm3 (full width at half maximum) to
decrease spatial noise. For group-level analyses, the images
acquired from patients with a right-hand amputation were
flipped (cf. Lotze et al., 2001; MacIver et al., 2008; Diers et al.,
2010) so that the hemisphere unaffected by (i.e., ipsilateral
to) the amputation is always shown on the left. For the
individual analyses of cortical reorganization (lip movement
data), this flip was not performed. The hemispheres of the
patients brains that were ipsilateral to the amputation
served as control for the measured changes in cortical activity since no pain-specific alterations were expected on the
intact hemisphere as a consequence of the study. Whole-

J. Foell et al.

brain activation during lip-pursing and hand movement

tasks, both before and after treatment, was observed, and
differences in activation during hand movement, before and
after treatment, in areas related to the processing of bodyrelated information and pain [primary and secondary somatosensory cortex (S1 and S2), primary motor cortex (M1),
anterior cingulate cortex, insula, parietal cortex] were correlated with the benefit from therapy.

2.6 Statistical analysis of clinical and

behavioural data
The average daily pain ratings were aggregated into weekly
averages, and the mean ratings before training were then
compared with those after training using a one-tailed pairedsample t-test. As a measure of effect size, Cohens d was
calculated using the means and SDs of the pain ratings before
and after treatment. Connections between non-painful
phantom phenomena (i.e., telescoping and the feeling of
relatedness between phantom hand and mirrored hand, time
since amputation) and PLP as well as individual treatment
effect were calculated using Spearmans rank correlations.
Missing values caused by a patient skipping some daily
ratings were replaced by the averaged value of the day before
and the day after.
Ratings about the relatedness between the mirror image
and the phantom hand were gathered after the performance
of each of the five tasks. For the depiction of results, these
values were summed up over 3 days each. The averaged
value for the first 3 days of the training was taken as an
indicator for the individual trait of the patient. A 3-day
period was used in order to ensure that the rating was not
influenced by either an initial first-day surprise effect or a
training effect caused by the mirror treatment. Scores were
averaged over the five tasks.

2.7 Brain changes related to treatment

Several measures were used to assess treatment effects on
neuronal activity. For both hand and lip stimulation, brain
activity during the task was observed before and after treatment on a whole-brain level and when anatomical regions of
interest (ROIs) were used. For the identification of the lip
and hand areas in S1 and M1, the functional ROIs defined by
MacIver et al. (2008) were used (MNI coordinates: lip S1:
58, 18, 24; lip M1: 52, 8, 36; hand S1: 34, 30, 58;
hand M1: 34, 34, 52). Differences in activation related to
MT were assessed using family-wise error correction. In
addition, a multiple regression was performed, using pain
reduction after treatment as a covariate in order to reveal
differences in activation related to treatment effects.
For the investigation of a shift in cortical lip representation, the analysis of the fMRI data of the lip-pursing task was
used to determine the location of activity peaks in S1 and M1
using pre- and post-central ROIs (as defined by the Wake
Forest University PickAtlas toolbox version 2.4 for SPM8;
Maldjian et al., 2003), both on the hemisphere affected by

732 Eur J Pain 18 (2014) 729739 2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC

J. Foell et al.

the amputation and on the opposing hemisphere. The shift

in cortical organization caused by the amputation was
defined as the Euclidean distance between the peak activity
locations under the assumption that both locations lie in the
same hemisphere; this was carried out by multiplying the
x-coordinate of the left activity peak with 1, thereby effectively flipping the coordinates of the left side in MNI space
onto the right side (cf. MacIver et al., 2008). This procedure
allowed us to estimate the position of the mouth representation under healthy conditions and to calculate the distance
between this position and the affected position (see Elbert
et al., 1994; Karl et al., 2001). The calculation resulted in a
measure of the shift that had occurred on the contralateral
side of the cortex relative to the ipsilateral side represented as
the distance in millimetres, where the distance was calculated using the Euclidean distance formula. The individual
amount of this cortical shift was correlated with the individual intensity of pain using Spearmans rank correlation.
The degree of cortical shift as described above immediately
before the 4-week training was compared with the same
value immediately after the training as a measure for the
degree of change in cortical organization, which might have
occurred as a result of MT. Cortical reorganization during the
training phase was defined as the Euclidean distance between
left and right activity peaks (as described above) before training minus that same distance after training. This results in the
amount of relative change in location during the training
period given in millimetres and creates a measure of activity
shift on the affected side back towards a healthy state. Spearmans rank correlation was used to investigate the connection
between this measure and pain ratings.
In addition, we calculated the cortical shift from the first to
the second measurement in the S1 and M1 regions of the
hemisphere ipsilateral to the amputation. This shift was then
checked for correlations with benefit from treatment. Since
we hypothesized a specific effect of MT only on the affected
hemisphere, we expected no significant correlation for the
hemisphere ipsilateral to amputation site.
In order to assess the contribution of the hand movement
(cf. Makin et al., 2013), we also defined functional hand ROIs
using group means of M1 and S1 during hand movement
(contralateral to amputation site) and defined 5-mm spherical ROIs around these coordinates. We used these ROIs to
extract the percentage signal change in these areas (using the
REX ROI extraction toolbox version 2.1 for SPM8;
Whitfield-Gabrieli, 2009). We correlated the intensity of activation at the first measurement with pain levels before
therapy, the activation at the second measurement with pain
levels after therapy and the difference in activation intensities with benefit from treatment.

3. Results
3.1 Mirror training performance
The patients did not report any problems in performing the mirrored movements or with the duration of

Mirror therapy for phantom limb pain

the training per day, and no patient dropped out of

treatment. Three participants skipped individual training days because of time constraints or illness (patients
7 and 13 missed 1 day, patient 12 missed 2 nonconsecutive days). Patients consistently performed all
five movement tasks for 3 min each with only minor
deviations in training duration (patients 7, 11 and 12,
range of duration 24 min per task).

3.2 Pain
The patients reported stable pain ratings 2 weeks
before the training, with a decline after the first week
of daily training. The average pain rating in the week
after the training was significantly lower than that
from the week before start of the training (week 1:
M = 28.21, SD = 11.52; week 2: M = 28.26,
SD = 16.27; week 7: M = 20.60, SD = 12.80; week 8:
M = 23.44, SD = 13.04; week 2 vs. week 7: t12 = 1.78,
p = 0.05). Cohens d was 0.52, indicating a mediumsized effect of treatment on pain scores. Individual
pain ratings before and after the mirror treatment are
given in Supporting Information Table S2 and illustrated in Fig. 1A; Fig. 1B shows the average pain
ratings before, during and after the mirror treatment.

3.2.1 fMRI data: Group-level analysis

Lip movement task: At both time points, patients
showed significant bilateral activation during lip
pursing in S1, M1 and insular cortex. Significant brain
activations both before and after treatment are
described in Supporting Information Table S3.
Mirrored hand movement task: The mirrored hand
movements caused significant bilateral activation in
somatosensory and motor regions before and after
treatment, as well as in the insular cortex and, at the
first time point, in the inferior parietal cortex (IPC)
and thalamus (Supporting Information Table S4).
Both before and after treatment, activation in S1
cortex was less intense on the hemisphere affected by
the amputation. A paired-sample t-test revealed no
significant pre-post change in the mirror task in S1 or
M1. The multiple regression analysis showed a significant connection between the individual amount of
pain reduction during the treatment and a decrease of
activity in the IPC on the hemisphere affected by the
amputation over the course of therapy (p = 0.001,
uncorrected, peak activity: T = 7.31, Z = 4.08, MNI
coordinates: x = 52, y = 38, z = 46; see Supporting
Information Fig. S1).

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC Eur J Pain 18 (2014) 729739

733

Mirror therapy for phantom limb pain

J. Foell et al.

Pain rating (VAS, 0100)

Pre-training
Post-training

Subject #

B
Training phase

Post-phase

Pain rating (VAS, 0100)

Pre-phase

(Cohens d) of treatment showed no significant result

for M1 (r = 0.15, p = 0.34), whereas the amount of
reduction in the cortical shift of S1 was found to be
significantly positively correlated with the reduction
in PLP measured as the difference between the preand post-phases (r = 0.75, p < 0.01). Fig. 3 shows a
scatter plot of this relationship.
The cortical shift in the control regions, i.e., S1 and
M1 in the hemisphere ipsilateral to the amputation,
was not significantly correlated with treatment benefit
(S1: r10 = 0.18, p = 0.59; M1: r10 = 0.16, p = 0.64).
The intensity of activation in the phantom hand
region during mirrored hand movements before the
training was not significantly correlated with
pre-training pain (S1: r10 = 0.24, p = 0.48; M1:
r10 = 0.37, p = 0.26), the same activation after the

All patients (N = 11)

Pre
Post

Week

Figure 1 (A) Individual pre- and post-training pain ratings, ordered from
most to least benet. Triangles denote individual patients without a telescopic distortion of the phantom. (B) Daily pain ratings averaged over
weeks and over all participants. Error bars depict standard error.

3.2.2 fMRI data: Individual analysis

The initial measurement of cortical activation during
the lip-pursing task showed an average dysfunctional
cortical shift of 15.4 mm (SD = 9.6 mm) in S1 as well
as an average shift of 16.6 mm (SD = 6.7 mm) in M1.
The comparison of the cortical shift before and after
training revealed an average reduction of this shift of
2.9 mm (SD = 11.4 mm) for S1 and 1.5 mm
(SD = 10.5 mm) for M1. This difference is not statistically significant on a group level (Z = 0.889,
p = 0.37). Fig. 2 displays the difference in distribution
of cortical activity for all patients during the lippursing task in S1 on the hemisphere that is affected
by the amputation before and after treatment
(Fig. 2A), as well as individual data for the amount of
dysfunctional shift in S1 before and after treatment
(Fig. 2B).
The correlation between the individual amount of
reduction in the shift and the individual effect size

B
Pre-training
Post-training

Figure 2 (A) Neuronal activity in somatosensory cortex during lippursing task before (red) and after (blue) mirror therapy. (B) Amount of
dysfunctional shift for individual patients before (red) and after (blue)
training, ordered from most to least benet.

734 Eur J Pain 18 (2014) 729739 2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC

J. Foell et al.

training was not significantly correlated with

post-training pain (S1: r10 = 0.40, p = 0.22; M1:
r10 = 0.46, p = 0.16), and the difference between the
activations before and after was not significantly correlated with treatment benefit (S1: r10 = 0.24,
p = 0.47; M1: r10 = 0.17, p = 0.62).
3.2.3 Predictors of treatment effects
There was no significant correlation between benefit
from treatment and time since amputation (r = 0.27,
p = 0.42). Eight of the 13 patients reported a telescopic
distortion of the phantom limb (see Supporting Information Table S1). The perceived intensity of this telescope was negatively correlated with the treatment
effect (r = 0.94, p < 0.01). Apart from the intensity of
the telescope, the mere presence of this phenomenon
had an influence on pain alleviation: Five patients
without a telescopic distortion reported an average of
51.23% decrease in PLP (pre-phase: M = 32.01,
SD = 14.34 vs. post-phase: M = 15.61, SD = 6.64;
t4 = 2.20, p < 0.05 one tailed; Cohens d = 1.64),
whereas the training had almost no effect (changes in
PLP ratings <1%) on the averaged PLP ratings of
patients with a telescope (pre-phase: M = 25.88,
SD = 12.60 vs. post-phase: M = 26.03, SD = 14.09,
t7 = 0.07, p = 0.47 one tailed; Cohens d = 0.01). Consequently, both groups differed significantly in their
reduction in pain intensity after training (t11 = 2.14,
p < 0.05 one tailed; Cohens d = 1.62). This difference
(shown in Fig. 4) indicates that the average alleviation

in pain over all patients was almost exclusively caused

by five participants without a telescope.
The feeling of being able to relate the observed
movement to the phantom limb did not significantly
differ between the five training tasks, with mean
values ranging from 1.74 to 1.95 (on a scale from 0 for
not at all to 6 for as vivid as a real perceptual experience), while SDs ranged from 1.61 to 1.77. Individual participants differed in the mean value of this
item, with a range across nearly the entire scale (0.00
5.91). Overall, the average intensity of this sensation
rose over the course of the training from 1.40
(SD = 1.62) on the first day of the training to 2.15
(SD = 1.89) on the last one; a difference bordering on
statistical significance (t12 = 1.76; p = 0.05 one tailed)
indicating a medium-sized effect (Cohens d = 0.43).
The initial starting point of the patients for this item
(measured as its average for the first 3 days of training)
varied widely, with a range from 0.00 to 5.80. There
was a strong positive correlation between individual
values of this 3-day average and the amount of pain
relief reported by the patients as a result of the training
(r = 0.873, p < 0.001).
At the beginning of training, those patients without
a telescopic distortion started out with a slightly higher
intensity (not statistically significant) of the sensation
of relatedness as described above, after which they
reported no significant change from the beginning to
the end of training (first 3 days: M = 2.29, SD = 2.16;
last 3 days: M = 2.36, SD = 2.38; t4 = 0.22, p = 0.42
one tailed; Cohens d = 0.03). Patients with a telescope, however, displayed a gradual increase in the

Pre-phase

Treatment phase

Post-phase

Pain intensity (VAS 0100)

Figure 3 Correlation between cortical reorganization in S1 (positive

values indicate shift in expected direction, i.e., towards healthy location)
and effect size of treatment. Empty circles indicate patients with telescoping, and lled circles patients without telescoping.

Mirror therapy for phantom limb pain

Week

Figure 4 Difference in pain development of patients with and without

telescopic distortion of the phantom limb. Error bars indicate standard
error.

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC Eur J Pain 18 (2014) 729739

735

Mirror therapy for phantom limb pain

sensation, ranging from 0.73 (SD = 0.26) on the first 3

days of training to 1.67 (SD = 0.59) on the last 3 days.
The change in group average described above was
almost entirely caused by patients with a telescope:
For this group, there was a significant difference
between the first and the last 3 days of training
(t7 = 2.21, p < 0.05 one tailed; Cohens d = 2.06).

4. Discussion
One aim was to determine whether MT is effective in
chronic PLP patients. Overall, treatment led to a pain
reduction of 27%, a medium-sized effect. This extends
earlier findings in acute amputees (Chan et al., 2007)
to chronic PLP. However, our results also suggest that
only some patients profit from MT, and that individual
differences determine treatment effectiveness.
Treatment effects were connected to cortical reorganization: As PLP decreased, the representation in the
somatosensory cortices of both hemispheres became
more similar, partially recreating their presumed
normal state. For motor cortex, no connection was
found. MacIver et al. (2008) investigated the effects of
mental imagery on PLP and found similar changes,
including more concise activation after treatment.
Makin et al. (2013) recently hypothesized that preserved hand function rather than dysfunctional reorganization (Flor et al., 1995) in S1 is positively
associated with PLP, suggesting that an effective PLP
treatment should be accompanied by reduced activity
in the hand representation in S1 rather than a relocation of cortical lip representation. However, we found
no significant alterations in S1 activity representing the
hand after training, and no significant correlation
between activation changes in the hand region and the
treatment effect. Instead, we found the already
reported shift in cortical lip localization. Makin et al.
had patients perform lip movements comparable with
the ones in this study and found no connection
between PLP and neural activation. Their finding of
continuing activation in the region previously representing the now-amputated hand seems to contradict
the model of maladaptive cortical reorganization.
However, a combination of conceptual and methodological factors suggests that these viewpoints are
complementary rather than contradictory (Flor et al.,
2013). Makin et al. (2013) used the intensity of activation, whereas cortical reorganization mainly relates to
its location. Furthermore, the direction of this shift can
vary between persons, diminishing the reliability of an
averaged approach: In their analysis, activation during
movement in PLP patients was weaker and less spread
out, consistent with averaged heterogeneous activa-

J. Foell et al.

tion. In addition, input from the periphery was shown

to increase cortical reorganization especially when it is
random (Spitzer, 1997; Mackert et al., 2003). Overall,
these issues lead us to assume that methodological
differences kept the effects of cortical reorganization
from being found by that group. Their further finding
relating PLP to disrupted connectivity and grey matter
volume is highly interesting and promotes further discussion about the mechanisms behind the disorder.
In accordance with previous work (Weeks et al.,
2010; Darnall and Li, 2012), not all patients reported
pain alleviation after treatment. Our data suggest that
this is connected to the ability to relate the mirrored
movement to their phantom: Although the patients,
on average, reported an increase in the intensity of
this sensation during training, they started out very
differently. This suggests that the ability to relate the
mirror image to the phantom is a characteristic that
patients bring into the therapy. The concept of the
body matrix (Moseley et al., 2012) proposes that
activity in specific cortical regions (including S1) is
related to the ability to subjectively incorporate a limb
not belonging to ones body. This embodiment has
been found to vary between individuals while being
stable over time (Bekrater-Bodmann et al., 2012). It
has also been argued that integration of a foreign body
part, rather than movement of the phantom, is necessary for pain relief (Schmalzl et al., 2011). Here, the
relatedness towards the mirrored hand at the beginning of treatment was predictive of pain relief, also
suggesting that the incorporation of the mirrored
hand, i.e., the feeling of observing ones own hand, is
necessary for treatment benefit.
Farrer et al. (2003) investigated agency (the feeling
that leads us to attribute an action to ourselves) over a
virtual hand and examined neural correlates. When
the feeling of relatedness was reduced by distorting the
image, activation in IPC was present. The intensity of
this activation increased along as reported relatedness
decreased. Interrupting agency elicits activation in the
same area (Yomogida et al., 2010). Our results support
these findings, since we observed activation in this
region to be higher before treatment, when relatedness
to the mirror image was lowest. We found this change
in activation to be directly associated with pain relief.
The same area has been connected to pain processing
(Dunckley et al., 2005; Cauda et al., 2009; Uematsu
et al., 2011) and may be part of a neuronal network
involved in the perpetuation of neuropathic pain.
These results identify IPC as being connected both to
disruption of agency over a body part and painful
somatic sensations and emphasize its role regarding the
resiliency of PLP and treatment effectiveness. A reduc-

736 Eur J Pain 18 (2014) 729739 2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC

J. Foell et al.

tion of IPC activation by providing the brain with

agency over the mirrored hand may thus be an alternative or additional neuronal mechanism of MT.
However, these mechanisms require that patients
accept the mirrored hand as their own. We found a
large variation in this acceptance, and our results
suggest a possible reason: Patients with a telescope had
virtually no benefit from treatment, whereas those
without a telescope reported an average pain alleviation of over 50%, and all three patients with a major
treatment effect (i.e., pain reduction of more than a
third from baseline) belonged to this group. Also, telescope vividness was negatively correlated with treatment benefit. The possible interaction of telescoping
and MT has not yet been studied, but it seems plausible that a distortion or displacement of the phantom
limb on a proprioceptive level causes a contradiction if
the patient is confronted with an intact visual representation of that limb. We recently found that incongruence between a mirrored and an actual arm can
lead to an illusory third arm (Foell et al., 2013), suggesting that the incongruent mirror image is not perceived as a representation of the hidden arm. In
amputees, the difference between a deformed
phantom and the non-deformed mirror image of an
intact arm may likewise lead to non-integration, with
the mirror image perceived as an additional entity,
unrelated to the phantom. This non-integration might
impede treatment benefit. Telescoping might thus be a
predictor for the effectiveness of MT.
It might be possible to circumvent this influence of
a telescope by introducing a training system that takes
the distortion into account. One solution is a virtual
reality (VR) set-up capable of fitting the virtual image
to the distorted phantom. Since we have shown that
the ability to include the mirrored hand increases
during training in patients with a telescope, we
assume that this ability can be improved. Recently, our
group introduced and tested a VR-based controllable
hand for use inside an MRI scanner, which induces a
feeling of ownership (Bach et al., 2012; Trojan et al., in
press). The role of telescoping, however, needs to be
further examined. It should be noted that dividing our
sample according to telescoping creates very small
groups and has happened post hoc. Further investigations need to replicate these findings.
This study has some limitations. No control group
without treatment or with an alternate treatment was
used. Thus, placebo effects or spontaneous changes of
PLP cannot be excluded (although the latter are
unlikely given the chronic condition of the patients).
However, our results on reorganizational alterations in
S1 and the correlation between pain relief and IPC

Mirror therapy for phantom limb pain

activity suggest a mechanism of MT that goes beyond

a placebo effect. Also, it is unclear what might be an
appropriate placebo control: It is interesting to note
that mental visualization, which was found to be an
effective form of treatment by MacIver et al. (2008)
and Mercier and Sirigu (2009), was used as a control
by Chan et al. (2007). Moreover, in a previous study
on sensory discrimination training, an attention
placebo condition was ineffective (Flor et al., 2001).
For these reasons, even though the effect that we
found for cortical reorganization was in line with prior
findings in terms of its effect size, directionality and
location on the cortex, and even though its theoretical
mechanism was corroborated by the findings regarding parietal activation and telescoping, the influence
of a possible placebo effect cannot be excluded. We
assumed that the hemisphere ipsilateral to the amputation would be unimpaired and did not undergo
changes due to cortical reorganization. Subjects thus
acted as their own controls. This use of the unimpaired
hemisphere as a control is a limitation, since it may be
possible that the treatment causes changes on both
hemispheres. However, we found no significant correlation between the cortical shift in the unimpaired
hemisphere and treatment benefit, suggesting that
any changes in this hemisphere did not systematically
influence our baseline. The sample sizes for behavioural and fMRI measurements that were used were
relatively small. However, we found significant effects
for both, with a medium overall effect size for treatment effectiveness. A study using a larger sample size
might detect additional differences in response to MT.
This study shows that mirror treatment for PLP has
a measurable effect on pain ratings of chronic PLP
patients, and that this benefit is correlated with
changes in neural activation, particularly with a
reduction of dysfunctional cortical reorganization in
S1. Furthermore, we found that treatment effectiveness depended on an ability to relate the mirror image
to ones phantom. This finding is supported by a
decrease in cortical activity in IPC, an area connected
to a feeling of agency and pain generation. A telescopic
distortion of the phantom seems to be a determining
factor for treatment benefit.

Author contributions
All authors of this paper were involved in the conception and
design of this study as well as the analysis and interpretation
of data. Further, all authors revised the manuscript for
important intellectual content and gave final approval of the
version to be published. Results were discussed among all
authors, and all authors commented on the manuscript.

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC Eur J Pain 18 (2014) 729739

737

Mirror therapy for phantom limb pain

References
Bach, F., Cakmak, H., Maass, H., Bekrater-Bodmann, R., Foell, J., Diers,
M., Trojan, J., Fuchs, X., Flor, H. (2012). Illusory hand ownership
induced by an MRI compatible immersive virtual reality device. Biomed
Tech (Berl) 57, 718720.
Bekrater-Bodmann, R., Foell, J., Diers, M., Flor, H. (2012). The perceptual
and neuronal stability of the rubber hand illusion across contexts and
over time. Brain Res 1452, 130139.
Birbaumer, N., Lutzenberger, W., Montoya, P., Larbig, W., Unertl, K.,
Topfner, S., Grodd, W., Taub, E., Flor, H. (1997). Effects of regional
anesthesia on phantom limb pain are mirrored in changes in cortical
reorganization. J Neurosci 17, 55035508.
Brodie, E.E., Whyte, A., Waller, B. (2003). Increased motor control of a
phantom leg in humans results from the visual feedback of a virtual leg.
Neurosci Lett 341, 167169.
Cauda, F., DAgata, F., Sacco, K., Duca, S., Cocito, D., Paolasso, I., Isoardo,
G., Geminiani, G. (2009). Altered resting state attentional networks in
diabetic neuropathic pain. J Neurol Neurosurg Psychiatry 81, 806811.
Chan, B.L., Witt, R., Charrow, A.P., Magee, A., Howard, R., Pasquina, P.F.,
Heilman, K.M., Tsao, J.W. (2007). Mirror therapy for phantom limb
pain. N Engl J Med 357, 22062207.
Darnall, B.D., Li, H. (2012). Home-based self-delivered mirror therapy for
phantom pain: A pilot-study. J Rehabil Med 44, 254260.
Diers, M., Christmann, C., Koeppe, C., Ruf, M., Flor, H. (2010). Mirrored,
imagined and executed movements differentially activate sensorimotor
cortex in amputees with and without phantom limb pain. Pain 149,
296304.
Dunckley, P., Wise, R.G., Aziz, Q., Painter, D., Brooks, J., Tracey, I.,
Chang, L. (2005). Cortical processing of visceral and somatic stimulation: Differentiating pain intensity from unpleasantness. Neuroscience
133, 533542.
Elbert, T., Flor, H., Birbaumer, N., Knecht, S., Hampson, S., Larbig, W.,
Taub, E. (1994). Extensive reorganization of the somatosensory cortex
in adult humans after nervous system injury. Neuroreport 5, 25932597.
Farrer, C., Franck, N., Georgieff, N., Frith, C.D., Decety, J., Jeannerod, M.
(2003). Modulating the experience of agency: A positron emission
tomography study. Neuroimage 18, 324333.
Flor, H., Denke, C., Schaefer, M., Grusser, S. (2001). Effect of sensory
discrimination training on cortical reorganisation and phantom limb
pain. Lancet 357, 17631764.
Flor, H., Diers, M., Andoh, J. (2013). The neural basis of phantom limb
pain. Trends Cogn Sci 17, 307308.
Flor, H., Elbert, T., Knecht, S., Wienbruch, C., Pantev, C., Birbaumer, N.,
Larbig, W., Taub, E. (1995). Phantom-limb pain as a perceptual correlate of cortical reorganization following arm amputation. Nature 375,
482484.
Flor, H., Nikolajsen, L., Staehelin Jensen, T. (2006). Phantom limb pain: A
case of maladaptive CNS plasticity? Nat Rev Neurosci 7, 873881.
Flor, H., Rudy, T.E., Birbaumer, N., Streit, B., Schugens, M.M. (1990).
[The applicability of the West Haven-Yale multidimensional pain inventory in German-speaking countries. Data on the reliability and validity
of the MPI-D.]. Schmerz 4, 8287.
Foell, J., Bekrater-Bodmann, R., Flor, H., Cole, J. (2011). Phantom limb
pain after lower limb trauma: Origins and treatments. Int J Low Extrem
Wounds 10, 224235.
Foell, J., Bekrater-Bodmann, R., McCabe, C., Flor, H. (2013). Sensorimotor incongruence and body perception: An experimental investigation.
Front Hum Neurosci 7, 310. doi: 10.3389/fnhum.2013.00310.
Giummarra, M.J., Gibson, S.J., Georgiou-Karistianis, N., Bradshaw, J.L.
(2007). Central mechanisms in phantom limb perception: The past,
present and future. Brain Res Rev 54, 219232.
Hanley, M.A., Ehde, D.M., Jensen, M., Czerniecki, J., Smith, D.G.,
Robinson, L.R. (2009). Chronic pain associated with upper-limb loss.
Am J Phys Med Rehabil 88, 742751.
Hunter, J.P., Katz, J., Davis, K.D. (2003). The effect of tactile and visual
sensory inputs on phantom limb awareness. Brain 126, 579589.
Jensen, T.S., Krebs, B., Nielsen, J., Rasmussen, P. (1983). Phantom limb,
phantom pain and stump pain in amputees during the first 6 months
following limb amputation. Pain 17, 243256.

J. Foell et al.

Karl, A., Birbaumer, N., Lutzenberger, W., Cohen, L.G., Flor, H. (2001).
Reorganization of motor and somatosensory cortex in upper extremity
amputees with phantom limb pain. J Neurosci 21, 36093618.
Kerns, R.D., Turk, D.C., Rudy, T.E. (1985). The West Haven-Yale Multidimensional Pain Inventory (WHYMPI). Pain 23, 345356.
Koppelstaetter, F., Siedentopf, C.M., Rhomberg, P., Lechner-Steinleitner,
S., Mottaghy, F.M., Eisner, W., Golaszewski, S.M. (2007). [Functional
magnetic resonance imaging before motor cortex stimulation for
phantom limb pain]. Nervenarzt 78, 14351439.
Lloyd, D.M. (2007). Spatial limits on referred touch to an alien limb may
reflect boundaries of visuo-tactile peripersonal space surrounding the
hand. Brain Cogn 64, 104109.
Lotze, M., Flor, H., Grodd, W., Larbig, W., Birbaumer, N. (2001). Phantom
movements and pain. An fMRI study in upper limb amputees. Brain
124, 22682277.
MacIver, K., Lloyd, D.M., Kelly, S., Roberts, N., Nurmikko, T. (2008).
Phantom limb pain, cortical reorganization and the therapeutic effect of
mental imagery. Brain 131, 21812191.
Mackert, B.M., Sappok, T., Grsser, S., Flor, H., Curio, G. (2003). The
eloquence of silent cortex: Analysis of afferent input to deafferented
cortex in arm amputees. Neuroreport 14, 409412.
Makin, T.R., Scholz, J., Filippini, N., Henderson Slater, D., Tracey, I.,
Johansen-Berg, H. (2013). Phantom pain is associated with preserved
structure and function in the former hand area. Nat Commun 4, 1570.
doi: 10.1038/ncomms2571.
Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H. (2003). An automated method for neuroanatomic and cytoarchitectonic atlas-based
interrogation of fMRI data sets. Neuroimage 19, 12331239.
Mercier, C., Sirigu, A. (2009). Training with virtual visual feedback to
alleviate phantom limb pain. Neurorehabil Neural Repair 23, 587594.
Moseley, G.L., Gallace, A., Spence, C. (2012). Bodily illusions in health
and disease: Physiological and clinical perspectives and the concept of a
cortical body matrix. Neurosci Biobehav Rev 36, 3446.
Ramachandran, V.S., Rogers-Ramachandran, D.C. (1996). Synaesthesia in
phantom limbs induced with mirrors. Proc Biol Sci 263, 377386.
Schley, M.T., Wilms, P., Toepfner, S., Schaller, H.P., Schmelz, M., Konrad,
C.J., Birbaumer, N. (2008). Painful and nonpainful phantom and stump
sensations in acute traumatic amputees. J Trauma 65, 858864.
Schmalzl, L., Thomke, E., Ragn, C., Nilseryd, M., Stockselius, A.,
Ehrsson, H.H. (2011). Pulling telescoped phantoms out of the stump:
Manipulating the perceived position of phantom limbs using a full-body
illusion. Front Hum Neurosci 5, 121.
Spitzer, M. (1997). Noise-driven neuroplasticity in self-organizing feature
maps: A neurocomputational model of phantom limbs. MD Comput 14,
192199.
Trojan, J., Diers, M., Fuchs, X., Bach, F., Bekrater-Bodmann, R., Foell, J.,
Kamping, S., Rance, M., Maass, H., Flor, H. (in press). An augmented
reality home-training system based on the mirror and imagery treatment approach. Behav Res Methods.
Tsakiris, M. (2010). My body in the brain: A neurocognitive model of
body-ownership. Neuropsychologia 48, 703712.
Uematsu, H., Shibata, M., Miyauchi, S., Mashimo, T. (2011). Brain
imaging of mechanically induced muscle versus cutaneous pain. Neurosci Res 70, 7884.
Weeks, S.R., Anderson-Barnes, V.C., Tsao, J.W. (2010). Phantom limb
pain: Theories and therapies. Neurologist 16, 277286.
Whitfield-Gabrieli, S. (2009). Region of Interest Extraction (REX) Toolbox
(Boston, MA: MIT Department of Brain and Cognitive Sciences).
Winter, C., Fritsche, K., Karl, A., Huse, E., Labig, W., Grusser, S.M., Flor,
H. (2001). [The phantom and stump phenomena interview (PSPI)].
Schmerz 15, 172178.
Yomogida, Y., Sugiura, M., Sassa, Y., Wakusawa, K., Sekiguchi, A.,
Fukushima, A., Takeuchi, H., Horie, K., Sato, S., Kawashima, R. (2010).
The neural basis of agency: An fMRI study. Neuroimage 50, 198207.

Supporting Information
Additional Supporting Information may be found in the
online version of this article at the publishers web-site:

738 Eur J Pain 18 (2014) 729739 2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC

J. Foell et al.

Figure S1. Differences in cortical activity before and after

treatment phase in correlation with the individual change in
pain ratings from pre- to post-therapy phase (p = 0.001
uncorrected, voxel threshold: 10 voxels). MNI coordinates of
peak activity: x = 52, y = 38, z = 46; T = 7.31; Z = 4.08;
cluster size: 29 voxels.

Mirror therapy for phantom limb pain

Table S1.
Table S2.
functional
Table S3.
Table S4.

Details of participants.
Individual scores for phantom limb pain and dyscortical shift before and after treatment.
Activation during lip movement.
Activation during hand movement.

2013 The Authors. European Journal of Pain published by John Wiley & Sons Ltd on behalf of European Pain Federation - EFIC Eur J Pain 18 (2014) 729739

739